file
stringlengths 27
28
| text
stringlengths 791
1.41M
|
|---|---|
PMC008xxxxxx/PMC8983004.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
8710219
1493
AIDS
AIDS
AIDS (London, England)
0269-9370
1473-5571
35197437
8983004
10.1097/QAD.0000000000003169
NIHMS1770030
Article
HIV & SARS-CoV-2 biochemical interactions may not explain clinical outcomes among adults hospitalized with COVID-19 co-infected with HIV
Response to “Antagonism between hydrogen bonding and secondary chemical bonding to calcium in viruses”
DURSTENFELD Matthew S. MD 12
HSUE Priscilla Y. MD 12
1 Department of Medicine, University of California, San Francisco, CA, USA
2 Division of Cardiology, UCSF at Zuckerberg San Francisco General, San Francisco, CA, USA
Address for Correspondence & Reprints: Matthew S. Durstenfeld, MD, Tel: +1 628 206 5461; Fax: +1 628 206 5447, [email protected] twitter: @durstenfeld; Division of Cardiology, UCSF at Zuckerberg San Francisco General Hospital, 1001 Potrero Avenue, 5G4, San Francisco, CA 94110, USA
10 1 2022
15 3 2022
15 3 2023
36 4 616617
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
pmcWe appreciate the interest by Huang et al [1] in our paper on clinical outcomes among people living with HIV hospitalized with COVID-19. We hypothesized that people living with HIV hospitalized with COVID-19 would be at increased risk of mortality and major adverse cardiac events compared to those without HIV, but we did not find evidence of significantly increased risk [2]. From a biochemistry perspective, Huang et al propose that HIV and SARS-CoV-2 do not show mutual interaction due to counteraction of hydrogen bonding and secondary chemical bonding to calcium. They suggest this accounts for the lack of an effect of HIV-1 infection on COVID-19 clinical outcomes among those with SARS-CoV-2 coinfection in our study.
Our study did not investigate whether there are direct viral interactions between HIV and SARS-CoV-2. Nonetheless Huang et al’s proposed mechanism is unlikely to explain our findings for the following reasons: (1) HIV targets CD4+ T cells[3, 4], whereas SARS-CoV-2 primarily infects respiratory epithelial cells via the ACE-2 receptor [5, 6]. Respiratory tract tissues most susceptible to SARS-CoV-2 [7] do not overlap with HIV reservoirs [8]. Simply put, SARS-CoV-2 and HIV are not in the same place at the same time. (2) Most individuals with HIV included in our cohort are likely to be treated with antiretroviral therapy and the majority are likely to be virally suppressed (per CDC surveillance data in the US: 96% are on antiretroviral therapy [9], and 85% are virally suppressed [10]). We were not able to assess the proportion in our study treated with antiretroviral therapy or viral suppression, but others have found increased risk among those with lower CD4 counts[11–13]. In other words, most people with HIV in our study likely do not have high levels of circulating HIV especially compared to the levels of SARS-CoV-2 during acute infection.
If there are differences among PLWH in post-acute sequelae of COVID-19 (PASC) or “Long COVID,” then chronic inflammation, immune activation, autoantibodies, or microvascular dysfunction are plausible mechanisms that could explain such differences. Viral RNA persistence may drive these pathologic processes [14], although HIV co-infection has not been shown to result in differences in RNA persistence. Combining insights from clinical studies with basic science may catalyze advances in understanding and treating these two viral infections, especially when they infect the same host.
Conflicts of Interest and Sources of Funding:
MSD has no disclosures. PYH has received honoraria from Gilead and Merck, research grant from Novartis, unrelated to this work. Dr Durstenfeld is now supported by NIH/NHLBI 5K12 HL143961. Dr Hsue is supported by NIH/NIAID 2K24AI112393-06.
References
1. Huang J , Liao L , Wang Y , Liu Q . Antagonism 1 between hydrogen bonding and Secondary chemical bonding to calcium in viruses. AIDS 2022.
2. Durstenfeld MS , Sun K , Ma Y , Rodriguez F , Secemsky EA , Parikh RV , Association of human immunodeficiency virus infection with outcomes among adults hospitalized with COVID-19. AIDS 2022.
3. Barré-Sinoussi F , Chermann JC , Rey F , Nugeyre MT , Chamaret S , Gruest J , Isolation of a T-Lymphotropic Retrovirus from a Patient at Risk for Acquired Immune Deficiency Syndrome (AIDS). Science 1983; 220 (4599 ):868–871.6189183
4. Nishimura Y , Brown CR , Mattapallil JJ , Igarashi T , Buckler-White A , Lafont BAP , Resting naïve CD4+ T cells are massively infected and eliminated by X4-tropic simian–human immunodeficiency viruses in macaques. Proceedings of the National Academy of Sciences 2005; 102 (22 ):8000–8005.
5. Shang J , Ye G , Shi K , Wan Y , Luo C , Aihara H , Structural basis of receptor recognition by SARS-CoV-2. Nature 2020; 581 (7807 ):221–224.32225175
6. Lan J , Ge J , Yu J , Shan S , Zhou H , Fan S , Structure of the SARS-CoV-2 spike receptor-binding domain bound to the ACE2 receptor. Nature 2020; 581 (7807 ):215–220.32225176
7. Jackson CB , Farzan M , Chen B , Choe H . Mechanisms of SARS-CoV-2 entry into cells. Nature Reviews Molecular Cell Biology 2022; 23 (1 ):3–20.34611326
8. Churchill MJ , Deeks SG , Margolis DM , Siliciano RF , Swanstrom R . HIV reservoirs: what, where and how to target them. Nature Reviews Microbiology 2016; 14 (1 ):55–60.26616417
9. Monitoring selected national HIV prevention and care objectives by using HIV surveillance data—United States and 6 dependent areas, 2018. In: HIV Surveillance Supplemental Report: Centers for Disease Control and Prevention. ; 2020.
10. Behavioral and clinical characteristics of persons with diagnosed HIV infection— Medical Monitoring Project, United States, 2018 Cycle (June 2018–May 2019). . In: HIV Surveillance Special Report: Centers for Disease Control and Prevention. ; 2020.
11. Tesoriero JM , Swain C-AE , Pierce JL , Zamboni L , Wu M , Holtgrave DR , COVID-19 Outcomes Among Persons Living With or Without Diagnosed HIV Infection in New York State. JAMA Network Open 2021; 4 (2 ):e2037069–e2037069.33533933
12. Hoffmann C , Casado JL , Harter G , Vizcarra P , Moreno A , Cattaneo D , Immune deficiency is a risk factor for severe COVID-19 in people living with HIV. HIV Med 2021; 22 (5 ):372–378.33368966
13. Dandachi D , Geiger G , Montgomery MW , Karmen-Tuohy S , Golzy M , Antar AAR , Characteristics, Comorbidities, and Outcomes in a Multicenter Registry of Patients with HIV and Coronavirus Disease-19. Clin Infect Dis 2020.
14. Daniel C , Sydney S , Sabrina R , Alison G , Joon-Yong C , Manmeet S , Nature Portfolio 2021.
|
PMC008xxxxxx/PMC8983005.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
101723284
47427
Biomed J Sci Tech Res
Biomed J Sci Tech Res
Biomedical journal of scientific & technical research
2574-1241
35392255
8983005
10.26717/BJSTR.2022.41.006668
NIHMS1790042
Article
Value of BI-RADS 3 Audits
Roychowdhury Prithwijit BS 1
Vijayaraghavan Gopal R. MD, MPH 12
Roubil John MD 1
Williams Imani M. MS 1
Siddiqui Efaza MBBS 2
Vedantham Srinivasan PhD 3
1. Department of Medicine, University of Massachusetts Medical School, 55 Lake Ave N, Worcester, MA 01655
2. Department of Radiology, UMass Memorial Healthcare, University of Massachusetts Medical School, 55 Lake Ave N, Worcester, MA 01655
3. Department of Medical Imaging, University of Arizona, Tucson, AZ 85724
Author Contributions:
PR: Study design, acquisition of data, data analysis, manuscript preparation
JR: Acquisition of data, manuscript preparation
IW: Acquisition of data, manuscript preparation
ES: Acquisition of data, data analysis, manuscript preparation
GV: Study design, acquisition of data, data analysis, manuscript preparation
SV: Study design, data analysis, manuscript preparation
Corresponding Author: Srinivasan Vedantham, PhD, Professor, Department of Medical Imaging, 1501 N Campbell Ave., PO Box 245067, University of Arizona, Tucson, AZ 85724, [email protected], Tel: 520-626-6641
24 3 2022
2 2022
14 2 2022
14 2 2023
41 5 3308633092
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Objectives:
BI-RADS 3 is an established assessment category in which the probability of malignancy is equal to or less than 2%. However, monitoring adherence to imaging criteria can be challenging and there are few established benchmarks for auditing BI-RADS 3 assignments. In this study, we explore some parameters that could serve as useful tools for quality control and clinical practice management.
Materials and Methods:
This retrospective study covered a 4-year period (Jan 2014-Dec 2017) and included all women over 40 years who were recalled from a screening exam and had an initial assignment of BI-RADS 3 (probably benign) category after diagnostic workup. A follow-up period of 2 years following the assignment of BI-RADS 3 was used for quantitative quality control metrics.
Results:
Among 135,765 screening exams, 13,453 were recalled and 1,037 BI-RADS 3 cases met inclusion criteria. The follow-up rate at 24 months was 86.7%. The upgrade rate was 7.4% (77/1,037) [CI: 5.9–9.2%] and the PPV3 was 33.8% (26/77) [CI: 23.4–45.5%]. The cancer yield was 2.51% (26/1,037) [CI: 1.64–3.65%] and did not differ (p=0. 243) from the 2% probability of malignancy. The initial BI-RADS3 per screening exam and per recall from screening were 0.76% (1,037/135,765) [CI: 0.72–0.81%] and 7.7% (1,037/13,453) [CI: 7.26–8.17%], respectively.
Conclusion:
Regular audit of BIRADS 3 metrics has the potential to provide additional insights for clinical practice management. Data from varied clinical settings with input from an expert committee could help establish benchmarks for these metrics.
radiology
mammography
breast cancer
screening
BI-RADS criteria
BI-RADS 3
pmcIntroduction
Screening mammography is a vital element of breast cancer detection that has helped to reduce disease mortality.1–4 With the current screening strategy, yearly cancer detection rate in the US is approximately five per 1000 screens and fewer than 2% of screens prove suspicious and require biopsy.5–7 In an effort to improve specificity, decrease cost, and reduce harm the American College of Radiology (ACR) established the Breast Imaging Reporting and Data System (BI-RADS) category 3 - probably benign designation to be used for short-term surveillance instead of immediate biopsy.8–10
The morphological criteria for BI-RADS 3 include a solitary circumscribed mass with a solid ultrasound (US) correlate, focal asymmetry without an US correlate, and grouped, round calcifications.8,9,11 Typically the designation of BI-RADS 3 is made after an initial diagnostic work-up and should not be assigned on a screening mammogram. The assignment of BI-RADS 3 activates a short-term (6-, 12-, and 24-months) follow-up protocol which has been demonstrated to reduce false-positive findings at biopsy, while also retaining a high sensitivity for early-stage breast cancer.9
The designation of BI-RADS 3 is meant to indicate that a finding has a 2% or less risk of malignancy8 and a recent retrospective report of 45,202 BI-RADS 3 cases from the National Mammography Database suggests that this expectation is concordant with reality.12 However, institution-level evidence still suggest that in practice 0.9 – 7.9% of BI-RADS 3 lesions are upgraded to BI-RADS 4 and sent for biopsy.9,13–15 Additionally, as the BI-RADS 3 designation is afforded some flexibility there is an appreciable amount of interobserver variability within each modality.16–18 As a result, monitoring adherence to imaging criteria can be challenging and there are relatively few established benchmarks for auditing BI-RADS 3 assignment. Herein, we share BI-RADS 3 audit results from our own institution over a four-year period and propose discrete auditing criteria that may help to establish performance benchmarks. We introduce the following metrics while on surveillance and which may serve as useful benchmarks: (i) percentage of initial BI-RADS 3 to total screens, (ii) percentage of initial BI-RADS 3 to screen-recalled cases (BI-RADS 0), (iii) BI-RADS 3 upgrade rates within 24 months, (iv) positive predictive value (PPV3) of lesions biopsied within 24 months, (v) distribution of imaging morphology assigned a BI-RADS 3 category, and (vi) cancer yield.
Materials and Methods
Our institute is a large tertiary academic medical center (a NAPBC accredited and a breast imaging center of excellence by the ACR) in the north east United States with an effective catchment area of nearly 1 million individuals. This retrospective study was approved by the institutional review board (IRB) and is compliant with the Health Insurance Portability and Accountability Act. Information regarding the annual number of screening mammograms and the specific number of BI-RADS 0, and BI-RADS 3 cases were obtained from the radiology information system (RIS). All relevant BI-RADS 3 medical record numbers (MRNs) were identified with the assistance of the institute’s translational science core. All cases were reviewed in the electronic medical records at our institution. All data was extracted and compiled in RedCap19 by study personnel. Efforts were taken to standardize the data extraction process and to minimize inter-observer variability. A sample of ten records was collaboratively reviewed by all study personnel to standardize the data extraction and compiling of records from radiologist’s interpretation. Subsequently, the data were extracted from the remaining charts independently by four study personnel.
Subjects
The study included all women over 40 years of age recalled (BI-RADS 0) from screening and assigned BI-RADS 3 at a follow-up diagnostic evaluation from January 2014 through December 2017 at our institution. Our inclusion criteria were women who were assigned BI-RADS 0 on initial screening exam, and, assigned BI-RADS 3 from a diagnostic follow-up exam performed within 90 days of the screening exam, and, had at least one follow-up visit in the subsequent 24-month period. Exclusion criteria were women under 40 years of age at the date of their initial screening exam, or, BI-RADS 3 assessment following diagnostic assessment in a symptomatic patient, or, the follow-up diagnostic evaluation from a screening mammogram exceeded the 90-day time limit, or, did not have one or more evaluations in the 2-year follow-up period.
The study was limited to mammographic and ultrasound evaluations only. All of the digital mammograms were performed at our multiple clinical sites on Hologic (Bedford, MA) Selenia® or Selenia® Dimensions™ units. Both full-field digital mammograms (2D) and digital breast tomosynthesis (DBT) techniques20 are employed at the time of the screening examinations. There are no clearly defined criteria with regards to who is offered a 2D mammogram and who is offered a DBT study. All breast ultrasounds were performed on a Phillips (Bothell, WA) iU-22 unit by a dedicated breast sonographer, and when necessary, the radiologist will also personally scan the patient. At our institute BI-RADS 3 cases are evaluated at 6 months (ipsilateral breast), 12 months (bilateral) and 24 months (bilateral). At each time point, supplemental ultrasound as indicated was also performed.
The data abstracted from the chart included the patient age at time of BI-RADS 3 designation as well as if the preceding BI-RADS 0 mammogram was their baseline. We also recorded whether the BI-RADS 3 designation was made via diagnostic mammogram, or ultrasound, or both. The radiologist who assigned the BI-RADS 3 designation, the breast density category (A-D), the quadrant-based location, and the morphology of the BI-RADS 3 finding from mammography and ultrasound were recorded. The presence of follow-up imaging at 6, 12, 24 months was recorded and was used to calculate the follow-up rate. If a patient was deemed to be loss-to-follow up at 24 months, the last known finding was recorded. If a biopsy was completed, the duration (months) after BI-RADS 3 assignment, modality used image guidance, and the histopathologic findings from the biopsied specimen were all captured.
Statistical Methods
The quantitative measures in this study are all reported as proportions/percentages. The Clopper-Pearson exact 95% confidence interval was computed. One sample tests of proportions were used to determine if the quantitative metrics differed from values reported in literature. All tests were two-tailed. Effects associated with p<0.05 were considered statistically significant. All analyses were conducted using statistical software (SAS version 9.4, SAS Institute, Inc., Cary, NC).
Results
Demographics
A total of 135,765 screening exams were performed during the four-year period from which 13,453 were recalled (Figure 1). A total of 1,360 women were assigned BI-RADS 3 of which 1,037 women met the study eligibility criteria during the four-year period. There were 24 unique radiologists who assigned BI-RADS 3 category during the study period. Eight out of the 24 radiologists were fellowship-trained in breast imaging and each of these eight radiologists assigned 50 or more BI-RADS 3 studies and accounted for 93% (n=969) of all included BI-RADS 3 cases. The mean age at time of initial BI-RADS 3 assignment was 56.6 ± 11.1 years with range of 40–94 years (Table 1). For 165 (15.9%) women, the BI-RADS 0 mammogram that preceded their BI-RADS 3 assignment was the patient’s first mammogram. In terms of breast density, nearly half (49.6%, n=514) of all of the breasts studied were category B, followed by 37.1% (n=385) in category C, 8.29% (n=86) in category A, and 4.82% (n=50) in category D.
BI-RADS 3 Features: Morphology, Laterality and Location
Nearly all (95.9%, n=994) of the BI-RADS 3 cases were assigned BI-RADS 3 on either mammogram/DBT alone, or mammogram/DBT with ultrasound. The remainder (3.95%, n=41) of cases were assigned BI-RADS 3 on ultrasound (Table 2). The imaging morphology breakdown of the 1037 cases were asymmetry/architectural distortion (n=512, 49%), grouped calcifications (n=398, 38%), and non-calcified circumscribed mass (n=90, 9%). The remaining 37 BI-RADS 3 cases (4%) were called at the discretion of the radiologist and the electronic records did not document the classic descriptors for a BI-RADS 3 assessment. The assignment of BI-RADS 3 lesions was relatively even with 49.8% (n=516) in the left breast, 44.6% (n=462) in the right breast, and 5.70% (n=59) of cases bilaterally. The upper outer quadrant had the greatest number of lesions in both the right (n=232, 38.0%) and the left (n=195, 35.3%) breasts, followed by the subareolar/central region in the right (n=140, 22.9%) and left (n=115, 20.8%) breasts.
Follow-up of BI-RADS 3 lesions
The follow-up rate at 6 months was 97.1% (1,007/1,037) and decreased progressively to 95.8% (979/1,022) at 12 months and 86.6% (876/1,011) at 24 months (Table 2). The denominator is adjusted for lesion downgrade due to benign pathology from biopsy at prior follow-up. Among the 1,037 BI-RADS 3 patients, 7.4% (n=77) of all the cases underwent biopsy, of which n=23, n=40 and n=14 cases were biopsied at 6 months, 12 months and 18-24 months, respectively. A majority of the biopsies (n=47, 61%) of the biopsies were performed under ultrasound guidance and the remainder (n=30, 39%) using stereotactic mammography. The distribution of biopsies at different follow-up periods was as follows: 23/77 (30%) at 6 months, 40/77 (52%) at 12 months, and 14/77 (18%) were performed between 18-24 months.
Quantitative benchmarks
The quantitative benchmarks suggested for routine clinical practice management are summarized in Table 3. The percentage of initial BI-RADS 3 to total screens was 0.76% (1,037/135,765) and the percentage of initial BI-RADS 3 to screen-recalled cases (BI-RADS 0) was 7.7% (1,037/13,453). Within the 24-month follow-up period, the BI-RADS 3 upgrade rate was 7.4% (77/1,037). Among the 77 lesions biopsied within 24 months following BI-RADS 3 assignment, there were 26 malignancies, resulting in positive predictive value (PPV3) of 33.8% (26/77). Among the 26 cancers, 62% (n=16) were biopsied under ultrasound guidance, while 38% (n=10) were biopsied under stereotactic mammography. The cancer yield within the 24-month follow-up period was 2.51% (26/1,037). Among these 26 cancers, 30.8% (8/26) were detected at 6 months, 57.7% (15/26) at 12 months and 11.5% (3/26) at 18-24 months. The most frequently identified cancer type was ductal carcinoma in situ (DCIS) with 46% (12/26) of the cases. This was followed by invasive ductal carcinoma (IDC) at 42% (n=11) and invasive lobular carcinoma (ILC) at 12% (n=3).
Discussion
The purpose of introducing the BI-RADS 3 categorization in the BI-RADS Atlas8 was to reduce the harms of screening by decreasing the number of false positives biopsies, reducing the cost of health care and yet maintaining sensitivity for early detection of breast cancers. Although the BI-RADS atlas specifies the probability of cancer in this subset as 2% or less, there has been no established routine audit in recent times for various clinical practice settings17,21. We therefore conducted a retrospective review of our own data as a quality assurance project to better guide clinical practice management. In our study over a 4-year period of 1,037 BI-RADS 3 cases following an inconclusive (BI-RADS 0) screening mammogram, the cancer yield was 2.5% (n=26) during the 2-year surveillance period. The observed cancer yield was not statistically different (p=0.243) from the 2% probability of malignancy as described in the BI-RADS atlas. Our cancer yield did not significantly differ with the 1.86% cancer yield reported by Berg et al12 (p=0.123), but was significantly higher than the 1.47% reported by Michaels et al21 (p=0.006), the 1.02% reported by Lehman et al22 (p<0.001), and the 0.8% reported by Baum et al23 (p<0.001).
Among the 26 cancers detected within the 2-year follow-up period, 8/26 (30.8%) were detected within the first 6 months and supports the value of the short-term (6 months) follow-up. The ratio in our series was different from Berg et al12, where 58% cancers were identified at 6 months (p=0.005). During the first 12 months of follow-up, 23/26 (89%) cancers were detected and is comparable to the 73% reported by Chung et al24 (p=0.076). In keeping with multiple prior studies11,12,21 most of our cancers were DCIS 12/26 (46%). There were 11/26 (42%) invasive ductal carcinomas and 3/26 (12%) invasive lobular carcinomas in our series. The invasive cancers were early-stage cancers.
In our study, during the 2-year surveillance, 77/1,037 (7.4%) cases were upgraded to BIRADS 4/5 and were biopsied. This rate was higher than the 5.9% reported by Michaels et al21 (p=0.037) and 0.88% reported by Vizcaino et al15 (p<0.001). The positive predictive value (PPV3) in our series was 26/77 (34%), which is larger than the 16.6% in Berg et al12 (p<0.001) and comparable to the 25% in Michaels et al21 (p=0.076).
In our study, the proportion of BI-RADS 3 to the number of recalls (BI-RADS 0) was 10.1% (1,360/13,453) among all women and 7.7% (1,037/13,453) among study eligible women. In our literature search on PubMed, we could not identify any publication that reported on the use of this metric. We suggest including this metric as part of routine audits for clinical practice management. To establish a benchmark across different practice settings, there is need for sharing recent data from varied clinical settings (academic and private, dedicated and non-dedicated breast imaging practices). The above referred indices could serve as a useful benchmark of a practice’s quality assurance.
Age, ethnicity, lack of transport, education, and cost of care all result in disparities and barriers that contribute to a poor follow-up. Poor compliance to follow-up would directly impact the cancer yield in BIRADS-3 cases. While the literature12,21,23,24 describes loss to follow-up as a major concern, in our series the follow-up rates were good with 97% at 6 months, 94% at 12 months and 84% at 24 months. In Michaels et al21 the compliance for follow-up progressively declined from 83% at 6 months to 54% at 24 months. In Baum et al23, the studied cohort only had a 71% compliance with follow-up.
The current edition of BI-RADS atlas clearly discourages assignment of BI-RADS 3 from a screening examination without a complete diagnostic workup. However, prior literature did not make that clear distinction12. The BI-RADS atlas clearly outlines the morphology criteria for assignment of BI-RADS 3 under mammogram, ultrasound and MRI; however, it also mentions that the radiologist’s experience and discretion could determine the assignment. The distribution of the different morphologies contributing to a BIRADS-3 assignment in our study was asymmetry/focal asymmetry/architectural distortion was 49% (512/1,037), microcalcifications 38% (398/1,037), non-calcified circumscribed mass on mammogram or ultrasound or both was 9% (90/1,037) and 4% (37/1,037) of the assignments were at the discretion of the interpreting radiologist without one of the above descriptors in the report. In most studies13,14,15,21 calcifications accounted for greater than 50% of the BI-RADS 3 assignments, except in Varas et al14, where calcifications accounted only for 19% of the BI-RADS 3 assignment. Institutional policies, reader variability and access to care may be contributing to these differences. Also, radiologist’s experience and fellowship-training may influence interpretation.18 Dedicated fellowship-trained breast imagers and general radiologists performing breast imaging are known to differ in their evaluation and assessment of breast lesions.17,18 Literature also mentions of varying cancer yields depending on whether dedicated breast imagers or general radiologists interpret breast exams.12,18,21 The majority of our BIRADS 3 cases at our facility were reviewed by dedicated fellowship-trained breast imagers.
Another factor contributing to variability that has been recently reported is the patient’s age with cancer yield exceeding 2% for women older than 60 years of age.25 Also, after the introduction of DBT, there is literature indicating better visualization of architectural distortion, some of which lack an ultrasound correlate.26 During the early stages of DBT adoption in clinical practice, there was lack of a DBT-guided biopsy device and hence consensus among the radiologists on the management of these lesions. Further, there is also variability among radiologists16 in terms of lesion descriptors that could contribute to variability in assigning BI-RADS 3 category. Ambinder et al18, refers to the decreasing incidence of BI-RADS 3 post-DBT implementation. All of these factors contribute to inter-reader and inter-facility variability and have resulted in wide variability across practices in the assignment of BIRADS 3 as a percentage of the total screens. We feel that larger data set from across the country may help us define some benchmarks necessitating practices to review their policies should there be large variances from established benchmarks.
Limitations
Our study had limitations. The study was retrospective in nature. Only mammographic and ultrasound features were considered. Prior to mid-2016, when we acquired the capability to perform tomosynthesis guided biopsies, architectural distortion without an ultrasound correlate were assigned BIRADS 3 at our institute. On review of our records, architectural distortion and asymmetry, though distinct morphologies, were sometimes used interchangeably in the report. Hence, we merged the two categories for analysis rather than attempt to distinguish them. We did not specifically account for downgrades to BIRADS 1 and 2 during follow-up, which is likely a very small proportion, since a majority of our breast imagers continue to follow up cases assigned a BIRADS-3 for the entire 24-month surveillance.
Conclusions
Audit of BIRADS 3 metrics has the potential to provide additional insights for clinical practice management. Many of the criteria referred to in this paper (cancer yield, BI-RADS 3 as a percentage of screens, as a percentage of BI-RADS 0, distribution of the morphology of BI-RADS3 assignments, upgrade rates, positive biopsy rates) may serve a useful role in monitoring clinical practice and for establishing the optimal range for the appropriate use of the BI-RADS 3 category. Larger data sets from varied clinical settings, with inputs from an expert committee could help establish benchmarks for these metrics.
Acknowledgements
The authors thank Yurima Guilarte-Walker and the Data Lake team as well as Pratik Patel from the RedCap team for their assistance with data collection and data management. This work was supported in part by the National Cancer Institute (NCI) of the National Institutes of Health (NIH) grants R01CA195512 and R01CA199044. The contents are solely the responsibility of the authors and do not necessarily represent the official views of the NCI or the NIH.
Figure 1: Flowchart describing the assignment and follow-up of probably benign findings and the associated quantitative metrics for clinical practice management.
Table 1: Patient demographics, prior mammograms, and breast density of BIRADS 3 patients (n =1,037)
Characteristic Value
Age (Mean ± SD) 56.6 ± 11.1
Range 40-94
Baseline Mammogram* 165 (15.9%)
Breast Density Category
Category A – Almost Entirely Fatty 86 (8.29%)
Category B – Scattered Areas of fibroglandular tissue 514 (49.6%)
Category C – Heterogeneously Dense 385 (37.1%)
Category D – Extremely Dense 50 (4.82%)
* Mammogram that was assigned BI-RADS 0 was first-ever mammogram
Table 2: Imaging characteristics including modality that resulted in BIRADS 3, lesion location and lesion morphology, and follow-up.
Characteristic Value (%)
Modality
Mammogram & Ultrasound 590/1,037 (56.9%)
Mammogram alone 404/1,037 (39.0%)
Ultrasound alone 41/1,037 (3.95%)
Laterality
Left 462/1,037 (44.6%)
Right 516/1,037 (49.8%)
Both 59/1,037 (5.7%)
Morphology
Asymmetry/ Architectural Distortion 512/1,037 (49%)
Grouped calcifications 398/1,037 (38%)
Noncalcified circumscribed mass 90/1,037 (9%)
Other 37/1,037 (4%)
Follow-up rates *
At 6 months 1,007/1,037 (97.1%)
At 12 months 979/1,022 (95.8%)
At 18-24 months 876/1,011 (86.7%)
* Denominator adjusted for benign pathology from biopsy at prior follow-up.
Table 3: Quantitative benchmarks for clinical practice management
Characteristic Value (%) [95% CI]
Percentage of initial BI-RADS 3 to total screens 1,037/135,765 (0.76%) [ 0.72% - 0.81%]
Percentage of initial BI-RADS 3 to screen-recalled (BI-RADS 0) cases 1,037/13,453 (7.71%) [7.26% - 8.17%]
Upgrade rate within 24 months of BI-RADS 3 77/1,037 (7.43%) [5.90% - 9.19%]
PPV3 of biopsies within 24 months of BI-RADS 3 26/77 (33.77%) [23.38% - 45.45%]
Cancer yield within 24 months of BI-RADS 3 26/1,037 (2.51%) [1.64% - 3.65%]
Cancer type
Ductal carcinoma in situ (DCIS) 12/26 (46.2%) [26.6% - 66.6%]
Invasive ductal carcinoma (IDC) 11/26 (42.3%) [23.4% - 63.1%]
Invasive lobular carcinoma (ILC) 3/26 (11.5%) [2.5% - 30.2%]
Key Messages
Audit of BIRADS 3 metrics can provide additional insights for clinical practice management.
BIRADS 3 metrics to monitor could include cancer yield, BI-RADS 3 as a percentage of screens, BI-RADS 3 as a percentage of BI-RADS 0, distribution of the morphology of BI-RADS 3 assignments, upgrade rates, and positive biopsy rates may serve a useful role in quality evaluation and establishing the optimal range for the appropriate use of the BI-RADS 3 category.
Larger data sets from varied clinical settings, with inputs from an expert committee could help establish benchmarks for these metrics.
Disclosures: GRV – Research collaboration with DeepHealth, Inc.; SV - NIH funded research collaboration with GE Global Research; SV - NIH funded research collaboration with Koning Corporation.
Data Access and Integrity: The authors declare that they had full access to all of the data in this study and the authors take complete responsibility for the integrity of the data and the accuracy of the data analysis.
References
1. Niell BL , Freer PE , Weinfurtner RJ , Arleo EK , Drukteinis JS . Screening for Breast Cancer. Radiol Clin North Am. 2017;55 (6 ):1145–1162. doi:10.1016/j.rcl.2017.06.004 28991557
2. Nelson HD , Fu R , Cantor A , Pappas M , Daeges M , Humphrey L . Effectiveness of Breast Cancer Screening: Systematic Review and Meta-analysis to Update the 2009 U.S. Preventive Services Task Force Recommendation. Ann Intern Med. 2016;164 (4 ):244–255. doi:10.7326/M15-0969 26756588
3. Løberg M , Lousdal ML , Bretthauer M , Kalager M . Benefits and harms of mammography screening. Breast Cancer Res. 2015;17 :63. doi:10.1186/s13058-015-0525-z 25928287
4. DeSantis CE , Ma J , Gaudet MM , Newman LA , Miller KD , Goding Sauer A , Jemal A , Siegel RL . Breast cancer statistics, 2019. CA Cancer J Clin. 2019 Nov;69 (6 ):438–451. doi: 10.3322/caac.21583.31577379
5. Lehman CD , Arao RF , Sprague BL , Lee JM , Buist DS , Kerlikowske K , Henderson LM , Onega T , Tosteson AN , Rauscher GH , Miglioretti DL . National Performance Benchmarks for Modern Screening Digital Mammography: Update from the Breast Cancer Surveillance Consortium. Radiology. 2017 Apr;283 (1 ):49–58. doi: 10.1148/radiol.2016161174.27918707
6. Rosenberg RD , Yankaskas BC , Abraham LA , Sickles EA , Lehman CD , Geller BM , Carney PA , Kerlikowske K , Buist DS , Weaver DL , Barlow WE , Ballard-Barbash R . Performance benchmarks for screening mammography. Radiology. 2006 Oct;241 (1 ):55–66. doi: 10.1148/radiol.2411051504.16990671
7. Lee CS , Bhargavan-Chatfield M , Burnside ES , Nagy P , Sickles EA . The National Mammography Database: Preliminary Data. AJR Am J Roentgenol. 2016;206 (4 ):883–890. doi:10.2214/AJR.15.14312 26866649
8. D’Orsi CJ ACR BI-RADS® Atlas, Breast Imaging Reporting and Data System. Published online 2013.
9. Sickles EA . Periodic mammographic follow-up of probably benign lesions: results in 3,184 consecutive cases. Radiology. 1991;179 (2 ):463–468. doi:10.1148/radiology.179.2.2014293 2014293
10. Liberman L , Menell JH . Breast imaging reporting and data system (BI-RADS). Radiol Clin North Am. 2002;40 (3 ):409–430, v. doi:10.1016/s0033-8389(01)00017-3 12117184
11. Sickles EA . Breast calcifications: mammographic evaluation. Radiology. 1986;160 (2 ):289–293. doi:10.1148/radiology.160.2.3726103 3726103
12. Berg WA , Berg JM , Sickles EA , Burnside ES , Zuley ML , Rosenberg RD , Lee CS . Cancer Yield and Patterns of Follow-up for BI-RADS Category 3 after Screening Mammography Recall in the National Mammography Database. Radiology. 2020 Jul;296 (1 ):32–41. doi: 10.1148/radiol.2020192641.32427557
13. Helvie MA , Pennes DR , Rebner M , Adler DD . Mammographic follow-up of low-suspicion lesions: compliance rate and diagnostic yield. Radiology. 1991;178 (1 ):155–158. doi:10.1148/radiology.178.1.1984295 1984295
14. Varas X , Leborgne F , Leborgne JH . Nonpalpable, probably benign lesions: role of follow-up mammography. Radiology. 1992;184 (2 ):409–414. doi:10.1148/radiology.184.2.1620838 1620838
15. Vizcaíno I , Gadea L , Andreo L , Salas D , Ruiz-Perales F , Cuevas D , Herranz C , Bueno F ; Sceening Program Working Group. Short-term follow-up results in 795 nonpalpable probably benign lesions detected at screening mammography. Radiology. 2001 May;219 (2 ):475–83. doi: 10.1148/radiology.219.2.r01ma11475.11323475
16. Grimm LJ , Anderson AL , Baker JA , Johnson KS , Walsh R , Yoon SC , Ghate SV . Interobserver Variability Between Breast Imagers Using the Fifth Edition of the BI-RADS MRI Lexicon. AJR Am J Roentgenol. 2015 May;204 (5 ):1120–4. doi: 10.2214/AJR.14.13047.25905951
17. Michaels AY , Chung CSW , Frost EP , Birdwell RL , Giess CS . Interobserver variability in upgraded and non-upgraded BI-RADS 3 lesions. Clin Radiol. 2017;72 (8 ):694.e1–694.e6. doi:10.1016/j.crad.2017.03.005
18. Ambinder EB , Mullen LA , Falomo E , Myers K , Hung J , Lee B , Harvey SC . Variability in Individual Radiologist BI-RADS 3 Usage at a Large Academic Center: What’s the Cause and What Should We Do About It? Acad Radiol. 2019 Jul;26 (7 ):915–922. doi: 10.1016/j.acra.2018.09.002.30268720
19. Harris PA , Taylor R , Thielke R , Payne J , Gonzalez N , Conde JG . Research electronic data capture (REDCap)--a metadata-driven methodology and workflow process for providing translational research informatics support. Journal of biomedical informatics. 2009;42 (2 ):377–381. doi:10.1016/j.jbi.2008.08.010 18929686
20. Vedantham S , Karellas A , Vijayaraghavan GR , Kopans DB . Digital Breast Tomosynthesis: State of the Art. Radiology 277 (3 ):663–84. doi: 10.1148/radiol.2015141303.
21. Michaels AY , Chung CSW , Birdwell RL , Frost EP , Giess CS . Imaging and histopathological features of BI-RADS 3 lesions upgraded during imaging surveillance. Breast J 2017; 1 :10–16.
22. Lehman CD , Arao RF , Sprague BL , Lee JM , Buist DS , Kerlikowske K , Henderson LM , Onega T , Tosteson AN , Rauscher GH , Miglioretti DL . National Performance Benchmarks for Modern Screening Digital Mammography: Update from the Breast Cancer Surveillance Consortium. Radiology 283 (1 ):49–58. doi: 10.1148/radiol.2016161174.
23. Baum JK , Hanna LG , Acharyya S , Conant EF , Bassett LW , Pisano ED . Use of BI-RADS 3- Probably Benign category in the American College of Radiology Network Digital Mammographic Imaging Screening trial. Radiology 2011: 260 (1 ):61–67.21502382
24. Chung CSW , Giess CS , Gombos EC , Frost EP , Yeh ED , Raza S , Birdwell RL . Patient compliance and diagnostic yield of 18 month follow up of surveillance in probably benign mammographic lesions. AJR 2014; 202 :922–927.24660725
25. Lee CS , Berg JM , Berg WA . Cancer Yield Exceeds 2% for BI-RADS 3 Probably Benign Findings in Women Older Than 60 Years in the National Mammography Database. Radiology. 2021 Mar 30 :204031. doi: 10.1148/radiol.2021204031.
26. Vijayaraghavan GR , Newburg A , Vedantham S . Positive Predictive Value of Tomosynthesis-guided Biopsies of Architectural Distortions Seen on Digital Breast Tomosynthesis and without an Ultrasound Correlate. J Clin Imaging Sci. 2019 Nov 18;9 :53. doi: 10.25259/JCIS_134_2019.31819830
|
PMC008xxxxxx/PMC8983006.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
9918281681406676
50993
Glob Implement Res Appl
Glob Implement Res Appl
Global implementation research and applications
2662-9275
35392361
8983006
10.1007/s43477-021-00033-0
NIHMS1792060
Article
Evaluation of organizational capacity in the implementation of a church-based cancer education program
Knott Cheryl L. PhD http://orcid.org/0000-0002-2261-7875
1*
Miech Edward J. EdD 2
Slade Jimmie MA 3
Woodard Nathaniel MPH 1
Robinson-Shaneman Barbara-Jean 4
Huq Maisha MPH 1
1 University of Maryland School of Public Health, Department of Behavioral and Community Health, 1234 School of Public Health Building, College Park, MD, 20742, USA.
2 Center for Health Services Research, Regenstrief Institute, Indianapolis, IN, USA.
3 Community Ministry of Prince George’s County, PO Box 250, Upper Marlboro, MD 20773, USA.
4 Department of Pharmaceutical Health Services Research, School of Pharmacy, University of Maryland, Baltimore, MD, USA.
* Corresponding author: Cheryl L. Knott, PhD, University of Maryland School of Public Health, 1234W School of Public Health Building, College Park, MD 20742. Phone: 301-405-6659; Fax: 301-314-9167; [email protected]; Twitter: ChampUMD; Tumblr: champlabumd
25 3 2022
3 2022
07 1 2022
01 3 2023
2 1 2233
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Implementation evaluations have increasingly taken into account how features of local context help determine implementation outcomes. The purpose of this study was to determine which contextual features of organizational capacity led directly to the RE-AIM Framework implementation outcomes of intervention reach and number of days taken to implement, in an implementation trial of a series of cancer education workshops conducted across 13 African American churches in Maryland. We used a configurational approach with Coincidence Analysis to identify specific features of organizational capacity that uniquely distinguished churches with implementation success from those that were less successful. Aspects of organizational capacity (e.g., congregation size, staffing/volunteers, health ministry experience) were drawn from an existing measure of church organizational capacity for health promotion. Solution pathways leading to higher intervention reach included: having a health ministry in place for 1–4 years; or having fewer than 100 members; or mid-size churches that had conducted health promotion activities in 1–4 different topics in the past 2 years. Solution pathways to implementing the intervention in fewer number of days included: having conducted 1–2 health promotion activities in the past 2 years; or having 1–5 part-time staff and a pastor without additional outside employment; or churches with a doctorally prepared pastor and a weekly attendance of 101–249 members. Study findings can inform future theory, research, and practice in implementation of evidence-based health promotion interventions delivered in faith-based and other limited-resource community settings. Findings support the important role of organizational capacity in implementation outcomes in these settings.
implementation
organizational capacity
configurational analysis
coincidence analysis
cancer
churches
pmcImplementation evaluations have increasingly considered how context helps determine implementation outcomes (Durlak & DuPre, 2008; Gingiss et al., 2006; Kelly et al., 2004; Norton, 2012; Riley et al., 2003; Woltmann et al., 2008). Organizational factors have been explicitly incorporated into multiple implementation science theories and/or models, including the Consolidated Framework for Implementation Research (CFIR) (Damschroder et al., 2009), which features the role of inner setting factors such as structure, communication, and readiness. Similarly, the Practice, Robust Implementation and Sustainability Model (PRISM) (Feldstein & Glasgow, 2008) highlights the importance of infrastructure and organizational characteristics, which along with other factors, impact implementation outcomes. These theories and others (Aarons et al., 2012; Meyers et al., 2012) have guided a rich wave of recent studies that examine the role of context in implementation outcomes.
One particular aspect of context of special interest to implementation science is organizational capacity. Though sometimes conflated with organizational readiness, organizational capacity is a discrete construct (Rabin & Brownson, 2012; Weiner et al., 2008). Organizational capacity is defined as the observable, structural aspects of an organization, assessed at the organizational level (Rabin & Brownson, 2012; Tagai et al., 2018), and can include things such as financial and material resources. Organizational capacity has been shown to be relevant to a number of implementation outcomes. Findings from 81 studies included in a meta-analysis indicate that organizational capacity (e.g., staffing and funding) has critical implications for the process of implementation (Durlak & DuPre, 2008). Financial provisions and leadership support were positively associated with intervention adoption in healthcare settings (Norton, 2012). Organizational capacity (e.g., staffing and funds) has been linked to increased implementation fidelity in school-based tobacco prevention programs (Gingiss et al., 2006). Staff turnover has been linked to decreased program implementation (i.e., fidelity and penetration) in mental healthcare settings (Woltmann et al., 2008). Research suggests adequate staffing positively affects the number of heart health promotion strategies implemented in public health agencies (Riley et al., 2003), as well as number of program sessions delivered and resulting changes in dietary behavior in health education programs (Kelly et al., 2004).
The role of organizational capacity is likely to be particularly salient when interventions are implemented in limited resource settings, such as community settings outside the healthcare system. These are settings that regularly engage in health promotion activities, but the primary mission, and corresponding resource allocation, is something other than health promotion. A notable example here is the health promotion work done in the context of churches, which have served key community support and resource functions in African American communities in America since the time of slavery (Lincoln & Mamiya, 1990). Allen and colleagues (Allen et al., 2015) recognized the importance of church organizational capacity when they designed and implemented an organization-level intervention to increase capacity of Catholic churches serving Latinos to provide evidence-based cancer control interventions. The team concluded that many Catholic parishes have the capacity to implement evidence-based cancer control interventions if they are adapted to fit the setting (Allen et al., 2016). Guided by CFIR, a recent study examined predictors of 12-month implementation of the Faith, Activity, and Nutrition intervention in United Methodist churches (Wilcox et al., 2021). The authors reported that intervention coordinator ratings of inner setting factors (e.g., communication, culture, organizational rewards) were predictive of their implementing the intervention’s core components.
The Present Study
Both the recent growing attention to the role of organizational factors in implementation science research and the observation that African American churches vary greatly in their organizational capacity (Tagai et al., 2018), lead to the question of what are the organizational determinants of successful implementation of evidence-based interventions delivered in these settings. The purpose of the present study was to determine which individual and combinations of organizational capacity indicators were co-present with key implementation outcomes, identified based on the RE-AIM Framework. These outcomes included intervention reach, defined as the proportion of eligible individuals participating in the intervention (Glasgow et al., 1999) and the number of days taken to implement a series of three cancer educational workshops, in an implementation trial conducted in African American churches in central Maryland. Because churches vary so greatly in their organizational capacities and often serve as health promotion settings even in the face of limited resources, it is important to identify the organizational conditions that can facilitate successful intervention implementation.
We used configurational analysis to identify determinants of implementation success. Configurational analysis is a mathematical, cross-case approach that operates within a regularity framework (Baumgartner & Falk, in press) and applies Boolean algebra and formal logic to identify a “minimal theory,” i.e., the key difference-making conditions that uniquely distinguish one group from another. Configurational analysis, also known as Configurational Comparative Methods, incorporates both Qualitative Comparative Analysis (QCA) and more recently Coincidence Analysis (CNA), the approach used here. Configurational analysis searches for necessary and sufficient conditions for an outcome to appear and is well-suited to identify both Boolean conjunctivity (when the joint appearance of several conditions yields an outcome) and disjunctivity (when multiple pathways lead to the same outcome) (Furnari et al., 2020; Palinkas et al., 2019; Ragin, 2014). Because configurational analysis draws on Boolean algebra (as opposed to linear algebra and probability), it does not require large sample sizes and can be applied with small-n studies. Unlike traditional variable-oriented methods, configurational analysis retains a persistent link to individual cases, applying formal logic to develop models identifying the specific bundles of conditions that distinguish cases with an outcome of interest from those without. Configurational analysis in general–and Coincidence Analysis in particular–has appeared across a wide variety of health-related implementation contexts in the published literature since 2020 (Cohen et al., 2021; Coury et al., 2021; Hickman et al., 2020; Miech et al., 2021; Petrik et al., 2020; Whitaker et al., 2020; Yakovchenko et al., 2020).
Methods
Project HEAL Intervention
The intervention platform for this secondary analysis is “Project HEAL” (Health through Early Awareness and Learning) (Holt et al., 2018; Holt et al., 2014; Santos et al., 2017; Santos et al., 2014; Scheirer et al., 2017). The Project HEAL intervention trains lay members of African American churches as Community Health Advisors, who then conduct a series of three cancer educational workshops in their churches on breast, prostate, and colorectal cancer, with an emphasis on early detection. Details on the intervention are described elsewhere (Holt et al., 2014). The Project HEAL intervention was implemented in 14 African American churches in Prince George’s County, Maryland. The Community Health Advisors complete a 13-module training and pass a cancer knowledge examination with a score of 85% or greater before they become certified to implement the workshop series. The cancer educational workshops are designed to both reinforce and build on each other and are intended to be delivered within a 60-day period of time (workshop 1, workshop 2 at 30 days later, workshop 3 at 30 days later), which makes the number of days to complete the workshop series a primary implementation outcome. The Community Health Advisors use a PowerPoint slide deck with direct spiritual references to present the workshop material, and they distribute project-specific print educational booklets to attendees. The Project HEAL workshops have been shown to result in significant baseline to 24-month increases in breast, prostate, and colorectal cancer knowledge, as well as significant increases in reports of fecal occult blood test and colonoscopy, mammography maintenance among women, and digital rectal exams in men (Holt et al., 2018).
Measures
Organizational Capacity
Indicators of church organizational capacity for health promotion were utilized from the Faith-Based Organization Capacity Inventory (FBO-CI) (Tagai et al., 2018). The FBO-CI was developed to assess organizational capacity specific to faith-based organizations and was grounded on theoretical work by Greenhalgh and colleagues (Greenhalgh et al., 2004). The instrument contains three subscales, including staffing and space (e.g., number of full- and part-time staff), health promotion experience (e.g., presence of health ministry; health promotion activities conducted in the past 2 years), and external collaboration (e.g., collaborations with outside organizations; participation in research). Internal reliability of the subscales ranged from 0.67 to 0.83 (Tagai et al., 2018). Select individual capacity indicators (items) were coded as multi-value ordinal factors with up to 4 possible levels for analysis based on their response scales and distributions (see Table 1).
Implementation Outcomes
Two primary implementation outcomes from the Project HEAL intervention were reach and time to intervention completion. These were selected because of their alignment with the RE-AIM (Reach, Efficacy, Adoption, Implementation, and Maintenance) Framework (Glasgow et al., 1999), which was used to guide the evaluation, and there was considerable between-church variability in the data, suggesting both uneven and in some cases sub-optimal implementation with opportunity for future growth. Reach was calculated as the proportion of church members who enrolled in the intervention divided by an estimate of the total number of eligible members. Reach was dichotomized into low and high on the basis of a median split (10%−29% = low; 30%−78% = high). Time to completion was defined as the number of days that the church took to implement all three cancer educational workshops in the series and was also dichotomized into low and high based on a median split and on the recommended implementation time frame from the study protocol (28–60 days = more timely completion; 61–84 days = less timely completion).
Factor Selection
The original dataset had over 50 factors based on the FBO-CI indicators described above. To reduce this data, we implemented a configurational approach to factor selection described in detail elsewhere (Cohen et al., 2021; Coury et al., 2021; Hickman et al., 2020; Petrik et al., 2020; Yakovchenko et al., 2020). Briefly, we began by using the “minimally sufficient conditions” (msc) function within the R package “cna” to comprehensively scan the full original dataset in order to identify configurations of conditions with strong connections to the outcomes (e.g., reach and time to completion). This process exhaustively considers all one-, two- and three-condition configurations instantiated in the dataset, assesses each configuration against a prespecified consistency threshold, retains all configurations that satisfy this criterion, and then generates a “condition table” to list and organize the Boolean output. In a condition table, rows contain all configurations of conditions that meet a specified consistency level while column variables include outcome, conditions, consistency, and coverage. We started the msc routine by specifying a consistency threshold of 100%; if no configurations met this threshold, we iteratively lowered the specified consistency level by 5 points (e.g., from 100% to 95%, etc.) and repeated the process to generate a new condition table. We continued lowering the consistency threshold until there were at least two potential configurations of church-level conditions that met the specified consistency level. Using this approach, we inductively analyzed the entire dataset and used the condition table output to identify a subset of candidate factors for model development in the next step of configurational analysis.
Model Development
We next developed models for reach and time to workshop completion by iteratively using model-building functions within the “cna” software package in R. We assessed models based on their overall consistency and coverage as well as potential model ambiguity (when more than one model satisfies the specified consistency and coverage thresholds and explains the outcome of interest). We selected a final model based on the same criteria of overall consistency and coverage, with no model ambiguity. The Coincidence Analysis package (“cna”) in R (Ambühl et al., 2021), R (version 3.5.0), R Studio (version 1.1.383) and Microsoft Excel were used to support the analyses. One church did not have a recorded reach outcome and was thus dropped from the analysis, resulting in an analytic sample of 13 churches.
Results
Table 1 provides the organizational capacity indicators, their codings, and the sample distribution for each indicator. Over half of the churches in the sample were in the medium size range of 101–499 members on their rosters, most of the pastors did not have outside employment in addition to their role in the church, and both full- and part-time staff numbers were modest in number while the number of volunteers was considerably greater. Nearly three-fourths of the churches had an existing health ministry.
With regard to the outcome variables, reach varied from 10 to 78%, with a median of 29% and a mean of 33% (SD=0.23) of eligible participants enrolled in the intervention. Time to complete the 3-workshop series varied from 28 to 84 days, with a median of 59.5 days. These data reflected considerable between-church heterogeneity in the implementation outcomes.
Configurational Analyses
Table 2 shows the configurational model for the higher value of the reach outcome, with 30–78% of eligible members participating in the intervention. This model had three distinct solution pathways composed of just three factors. Churches that had greater intervention reach were those that had: 1) 1–4 years of health ministry experience; OR 2) fewer than 100 members; OR 3) 101–499 members AND had conducted health promotion activities in 1–4 different health topics in the previous 2 years. Configurational models have two main parameters of fit: consistency and coverage. Consistency measures how reliably a model yields an outcome, and is calculated by “the number of cases identified by a model that also have the outcome of interest” divided by “the total number of cases identified by the model.” This solution for higher reach had 100% consistency (6/6), as there were six cases identified by the model that also had the outcome (n=6) divided by six cases total that were identified by the model (n=6). Coverage measures the explanatory breadth of a model, and is calculated by “the number of cases identified by a model that also have the outcome of interest” divided by “the total number of cases that have the outcome of interest.” This solution had 100% coverage (6/6), as there were six cases identified by the model that also had the outcome (n=6) divided by six cases total that had the outcome of interest (n=6).
Table 3 shows the configurational model for the lower value of the reach outcome, with 10–29% of eligible members participating in the intervention. This model had three distinct solution pathways composed of just three factors. Churches that had lower intervention reach were those that had: 1) 500 or more members; OR 2) a medium membership size (101–499) AND had conducted health promotion activities in 9 or more different health topics in the previous 2 years; OR 3) a moderate (5–10 years) history of health ministry experience AND had conducted health promotion activities in 5–8 different health topics in the previous 2 years. As before, each solution pathway was sufficient by itself for the outcome of interest. This configurational model for lower reach likewise had both 100% consistency (6/6) and 100% coverage (6/6).
Table 4 shows the configurational model for more timely completion of the 3-workshop series (28–60 days). This model had three distinct solution pathways composed of just five factors. Churches that took fewer days to complete the 3-workshop series were those that had: 1) 1–2 health activities in the previous 2 years; OR 2) 1–5 part-time staff AND the pastor did not have outside employment in addition to the church; OR 3) pastors who had a doctoral degree AND mid-sized (101–249) weekly attenders. This model similarly had both 100% consistency (6/6) and 100% coverage (6/6).
Table 5 shows the configurational model for less timely completion of the 3-workshop series (61–84 days). This model also had three distinct solution pathways composed of just five factors. Churches that took more days to complete the 3-workshop series were those that had: 1) pastors with outside employment in addition to the church; OR 2) conducted 3–4 health promotion activities in the previous 2 years AND had 6 or more part-time staff; OR 3) had 51 or more volunteers AND 250 or more members in weekly attendance. This configurational model also had both 100% consistency (6/6) and 100% coverage (6/6).
Discussion
The current study identified features of organizational capacity that can help explain key implementation outcomes in a trial that evaluated implementation of an evidence-based cancer control intervention delivered by volunteers in African American churches. Though organizational factors are clearly important to implementation success, to date there has been limited ability to examine these factors using traditional variable-oriented methods in small-n studies. Overall, there were several church organizational capacity factors consistently related to the implementation outcomes of intervention reach and the number of days that it took the churches to implement the full series of three Project HEAL cancer educational workshops.
Reach
Reach is a key outcome in implementation research and the first element of the RE-AIM Framework (Glasgow et al., 1999). Reach reflects the penetration of an intervention to impact its intended audience. Though in many cases, as in the current one, it is an estimate due to an uncertain denominator (Santos et al., 2017), data on reach is informative as an implementation outcome. Churches with better reach were those that had some, but not many, years of health ministry history. This reflects the idea that there may not be a linear, but rather a curvilinear, relationship between history of health promotion and better implementation outcomes. It is possible that in churches that had some experience, members viewed the Project HEAL intervention as novel or were curious or otherwise had an appetite for the information and were thus more inclined to engage than churches in which health ministry activities were commonplace. In the mid-size churches, reach was greater if they also had focused on just a few health topics in the past two years. Again, this may reflect a relative novelty of the Project HEAL workshops.
It is notable that reach was higher in the smaller congregations. This finding reflects the idea that bigger does not automatically equate to better when it comes to church size and the ability of its programming to serve the membership. The smaller churches may be more socially cohesive, with a fewer number of competing ministries and with most members attending a single event rather than having to choose between multiple concurrent offerings on the church calendar. It may have been easier for the Community Health Advisors to get the word out about the workshops in the smaller congregations due to the nature of the social network. In small congregations leaders other than the pastor (e.g., deacon/deaconess; first lady [pastors’s wife]) often can suggest the inclusion of an announcement in the church bulletin or make an announcement to reach the entire congregation (Slade, 2021). Nevertheless, the finding can be considered good news because most congregations in the US have fewer than 100 members (Chaves & Eagle, 2015), so even small organizations have great potential to reach people with health promotion.
Other churches had lower reach, with the intervention serving fewer than one in three members. This reflects a considerable missed opportunity to benefit members with cancer education. Community Health Advisors at times expressed disappointment and frustration that more people did not come to hear their presentations. Community Health Advisors, after a disappointing initial workshop turnout, often consulted with the study team on active recruitment techniques in an effort to increase participation. As previously discussed, reach was lower in the larger congregations. Mid-size churches that had offered their members nine or more different types of health topics in the past two years had poor reach. Again, this may have resulted in over-saturation of health promotion among the members, making them less inclined to engage, or the Community Health Advisors may have been burnt out and less effective in their recruitment efforts. Finally, churches with 5–10 years of formal health ministry history and that had conducted health promotion activities on an average of 5–8 different health topics in the prior two years had lower reach. This again may be due to over-saturation or competing activities on busy church calendars. Of note, the three factors in the model for higher values of reach (years of having a health ministry in place; membership size; or number of health promotion topics in the past 2 years) were exactly the same factors in the model for lower values of reach. This provides compelling evidence of their difference-making roles, as different values of these three factors corresponded directly to higher and lower values of reach; when these three factors varied, so did the outcome.
Time to Implement the Workshop Series
The Project HEAL intervention was designed with the intent of multiple cancer educational workshops rather than a one-time event and included both: 1) repeated exposure to key cancer risk and early detection information to account for the fact that not every enrollee will attend all workshops; and 2) unique cancer educational information that built upon the previous workshops, to give those who attended multiple workshops new and valuable information each time. Therefore, the timing of the workshops was deliberate, in an attempt to avoid too rapid of a succession that would limit feasibility of attendance yet maintain enough momentum to keep enrollees interested and to facilitate a layered dialogue with the Community Health Advisors based on the material from the previous workshop. Optimally, the three workshops were designed to be delivered approximately monthly, so that all three could be completed in around a 60-day period.
There was variability in the workshop delivery time, with some churches delivering in quicker succession and others taking more time for implementation. It was initially surprising to us that churches that had conducted fewer health promotion activities in the previous two years had more timely implementation. Upon further reflection, however, it is possible that these churches had fewer competing events as opposed to those that were saturated with activities. Furthermore, volunteers assigned by the pastor to implement the intervention may have had fewer commitments, particularly in the area of health activities, and thus were able to move at a quicker pace. Finally, if health promotion were a novel activity for the volunteer interventionists, perhaps enthusiasm translated to a brisker pace.
The combination of a handful of part-time staff and having a pastor who did not have additional employment outside the church likewise linked directly to quicker implementation. This is likely due to the pastor’s ability to focus his or her attention more exclusively on church activities. Finally, the churches with more educated pastors and a mid-size active membership also had more timely implementation. Future research is needed to qualitatively examine how pastor education relates to their support of health ministry activities. It is possible that even if a pastor is highly educated, if he/she heads a small congregation there may be limited resources for implementation, and in larger congregations, there may be many competing activities as well as additional organizational layers (e.g., board/elder approvals, church politics) that can slow implementation. In addition, larger congregations often implement formal procedures to communicate with congregants, which can make it difficult to inform the full congregation about planned health ministry activities (Slade, 2021). Future research should examine ways to optimize reach in large congregations to increase impact of health promotion activities and reduce the resource inefficiency resulting from poor reach.
When considering churches with less timely implementation, similar organizational capacity factors came into play. The two churches where the pastor had outside employment were also those that took more time to implement the workshop series. This stands to reason, as in many churches the pastor is the gatekeeper of decision-making and has multiple employment roles; this can limit the pastor’s availability and slow the approval of workshop scheduling. In addition, in some churches the pastor preferred to attend the workshops to show support, which may have resulted in a lengthier scheduling timeline due to their outside employment obligations.
The combination of a moderate number of health promotion activities in the past two years and a greater number of part-time staff was also related to less timely implementation. Though previous experience and potentially available staff would appear to facilitate implementation, these churches may have also had busier church calendars with many competing activities, resulting in workshop scheduling delays. Collecting data on the church calendar (e.g., number and frequency of events and the scheduling and prioritization process) may have provided helpful insights to understand this finding, as busy church calendars tended to be a key barrier to getting the workshops scheduled (Slade, 2021).
Finally, churches that had both a large number of volunteers and mid-size active congregation size tended to take longer to implement the workshops. Given that the Project HEAL intervention is delivered by volunteers, the number of available volunteers and their portfolio of church commitments comes into play in implementation. This is reinforced by the finding that the number of staff alone did not appear to play a role in implementation timelines. While it may seem counterintuitive that having a large number of volunteers is associated with slower implementation in the mid-size churches, it may be that these congregations offer so many activities that a large volunteer base is needed and could still be insufficient to cover the workload. More nuanced insights are warranted regarding the role of volunteers (e.g., typical number and intensity of activities) and how pastors counted the number of volunteers (e.g., activity threshold for a member being counted as a volunteer). Of note, four of the five factors in the model for higher values of “days to workshop completion” (e.g., number of health activities in the past 2 years, pastor outside employment, number of part-time staff, member weekly attendance) likewise appeared in the model for lower values of “days to completion.” The appearance of these four factors in both models underscores their difference-making roles, as different values of these four factors corresponded directly to high and low values of the outcome; when these four factors varied, so did the outcome.
Organizational Capacity: Is More Better?
A somewhat unexpected finding in our results was that the middle value of an organizational capacity indicator (e.g., member size; number of health activities or topic areas conducted in the previous 2 years), rather than the highest value, was related to better performance on an implementation outcome. An example of this is the reach model, where reach was lower in the larger congregations. Though our previous work evaluated reach and indicated that it was modest overall (Santos et al., 2017), statistical power with church as the unit of analysis limited our previous ability to evaluate reach by church characteristics such as membership size. We previously reported that church member turnout at the workshops was better when the pastor also attended (Williams et al., 2018). Church size and pastor support may work together to influence intervention reach. For example, it may be more difficult for pastors of larger congregations to show their support for church initiatives such as Project HEAL, due to the complexity of the communication structure in these organizations and the high likelihood of concurrent activities.
In the aforementioned analysis of the Faith, Activity, and Nutrition intervention in United Methodist churches (Wilcox et al., 2021), while implementer ratings of inner setting characteristics were associated with implementation outcomes, pastor ratings of these characteristics were not. The study reported that implementation of the healthy eating intervention was greater in churches with fewer than 500 members, a finding consistent with the current finding of better implementation in smaller churches. While the Faith, Activity, and Nutrition study reported that inner setting ratings (culture, readiness) and implementation process (e.g., opinion leader/champion engagement) were most closely linked with implementation of core components, the current study focused more on the structural aspects of capacity (Tagai et al., 2018).
Strengths/Limitations
Study strengths include use of configurational analysis, a cross-case approach that can be applied in studies of differing sample sizes, including small-n studies. Though it would be optimal to have had more intervention sites, this was not feasible given the intensity of the investment required to enroll and then maintain the relationship with each church. In addition, the community-engaged nature of this study overall, from the intervention (Holt et al., 2014) to the instrumentation (Tagai et al., 2018) to the evaluation (Holt et al., 2018; Santos et al., 2017), makes for an authentic and relevant contribution in the area of community-based implementation science. In terms of limitations, there are other indicators across the implementation continuum (e.g., fidelity, sustainment) beyond the scope of the current analysis that should be considered for future study. Factors beyond those included in our original dataset may have played a role in implementation outcomes. Capacity indicators may have served as proxies for other factors not included in the evaluation. In addition, while dichotomizing the outcome variables was necessary for the analysis, doing so can obscure the distribution of the data particularly when scores fall near the median. In the current case, there were a number of churches characterized as having more timely implementation, with number of days to implement the workshop series near sixty. Finally, as previously mentioned, additional qualitative research into the roles of pastors and Community Health Advisors could help explain the configurational findings though within-case and cross-case qualitative analysis that allowed for further examination of these conditions in greater depth and context.
Practice Implications
The current findings can be used to inform future church-based health promotion activities. While the social networks and less complex organizational structure of small churches appears to be conducive to health promotion interventions reaching most members, there are things that larger congregations can do to maximize the benefit of health programming for their members. This may include bolstering communication about health promotion activities both centrally and within all church ministries (e.g., health, outreach, education, family life, women’s, men’s, seniors). This niche saturation approach could help get the word out so that all can benefit. Intentional coordination with the church calendar is advised to avoid situations where multiple events are offered at the same time. While some health promotion activities are one-time events (e.g., health fairs, single educational sessions), those that involve information delivered over multiple sessions should be carefully considered in the context of the recent history of health promotion activities in the church. If there is a precedent for health promotion in the church but not a preponderance of recent health activities, the timing may be right for a more intensive program. In any case, when it comes to health promotion activities offered in the church setting, planning, preparation, and prioritization bode well for success.
Conclusions
This is one of the first studies to apply configurational analysis to evaluate implementation outcomes related to health promotion activities outside of a traditional healthcare setting. A small set of organizational capacity features, primarily membership size and previous health promotion experience, proved to be consistent difference-makers vis-a-vis intervention reach and implementation timeline. Our configurational findings indicate great potential for reaching community members through implementation of evidence-based cancer education in places of worship, particularly in small church congregations, along with identifying the specific aspects of organizational capacity that consistently distinguished churches achieving higher levels of implementation success. Future efforts should focus on scale-up of these activities for population-level impact on cancer disparities.
Acknowledgements:
The authors acknowledge Ms. Deborah Bors, who assisted with manuscript preparation.
Funding:
This work was supported by the National Cancer Institute under Grant R01CA147313 and by funds through the Maryland Department of Health’s Cigarette Restitution Fund Program.
Availability of data and materials:
Study materials are available upon request from the corresponding author. The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request upon execution of a data use agreement.
Table 1 Indicators of Church Organizational Capacity and Sample Distribution in N=13 Churches
FBO-CI Construct Item Levels Coding N (%)
Staffing/space Membership size 3 0–100 4 (30.8)
101–499 7 (53.8)
500+ 2 (15.4)
Weekly attendance 3 0–100 4 (30.8)
101–249 6 (46.2)
250+ 3 (23.1)
Pastor outside employment 2 no 11 (84.6)
yes 2 (15.4)
Pastor education 3 other 3 (23.1)
masters 5 (38.5)
doctorate 4 (30.8)
Building ownership 2 no 3 (23.1)
yes 9 (69.2)
Number full-time staff 3 0 3 (23.1)
1 5 (38.5)
2+ 5 (38.5)
Number part-time staff 3 0 4 (30.8)
1–5 4 (30.8)
6+ 5 (38.5)
Number volunteers 3 0 0 (0.0)
1–25 2 (15.4)
26–50 4 (30.8)
51+ 7 (53.8)
Health Promotion Experience Current health ministry 2 no 4 (30.8)
yes 9 (69.2)
Number health areas past 2 years 4 0 0 (0.0)
1–4 5 (38.5)
5–8 3 (23.1)
9+ 5 (38.5)
Number health activities past 2 years 4 0 0 (0.0)
1–2 3 (23.1)
3–4 5 (38.5)
5+ 5 (38.5)
Years of health ministry existence 4 0 4 (30.8)
1–4 3 (23.1)
5–10 4 (30.8)
11+ 2 (15.4)
Note. FBO-CI = Faith-Based Organization Capacity Inventory. Numbers might not sum due to missing data. External collaboration subscale items were not significant in any analysis and are not shown.
Table 2 Configurational Model with Three Solution Pathways for Higher Value of Reach Outcome in N=13 Churches
Outcome: Reach # Years health ministry history Number of members Number of members # Health areas in past 2 years
High 5–10 <100 <100 9+
High 1–4 101–499 101–499 1–4
High 0 <100 <100 1–4
High 0 101–499 101–499 1–4
High 1–4 <100 <100 5–8
High 1–4 <100 <100 9+
Low 0 101–499 101–499 9+
Low 5–10 101–499 101–499 5–8
Low 5–10 500+ 500+ 5–8
Low 11+ 101–499 101–499 9+
Low 5–10 101–499 101–499 9+
Low 0 500+ 500+ 1–4
Note. Reach = Percent of eligible members participating in the intervention: 10%−29% = Low; 30%−78% = High (based on median split). Shading is used for clarity in reporting the configurational analysis.
Table 3 Configurational Model with Three Solution Pathways for Lower Value of Reach Outcome in N=13 Churches
Outcome: Reach Number of members Number of members # Health areas in past 2 years # Years health ministry history # Health areas in past 2 years
Low 101–499 101–499 9+ 0 9+
Low 101–499 101–499 5–8 5–10 5–8
Low 500+ 500+ 5–8 5–10 5–8
Low 101–499 101–499 9+ 11+ 9+
Low 101–499 101–499 9+ 5–10 9+
Low 500+ 500+ 1–4 0 1–4
High <100 <100 9+ 5–10 9+
High 101–499 101–499 1–4 1–4 1–4
High <100 <100 1–4 0 1–4
High 101–499 101–499 1–4 0 1–4
High <100 <100 5–8 1–4 5–8
High <100 <100 9+ 1–4 9+
Note. Reach = Percent of eligible members participating in the intervention: 10%−29% = Low; 30%−78% = High (based on median split). Shading is used for clarity in reporting the configurational analysis.
Table 4 Configurational Model with Three Solution Pathways for More Timely Completion (MTC) of 3-Workshop Series in N=13 Churches
Outcome: # Days to complete 3 workshops # Health activities in past 2 years # Part-time staff Pastor outside employment Pastor education Weekly member attendance
MTC 5+ 0 No Doctorate 101–249
MTC 5+ 1–5 No Doctorate <100
MTC 3–4 0 No Doctorate 101–249
MTC 1–2 6+ No Masters 101–249
MTC 5+ 1–5 No Masters 250+
MTC 1–2 1–5 No Masters <100
MTC 1–2 6+ No Other 101–249
LTC 3–4 6+ No Masters 101–249
LTC 5+ 0 No Doctorate 250+
LTC 3–4 6+ No Masters 250+
LTC 3–4 0 Yes Other <100
LTC 5+ 1–5 Yes Other <100
Note. Number of days to complete workshop series: 28–60 days = More Timely Completion (MTC); 61–84 days = Less Timely Completion (LTC).
Shading is used for clarity in reporting the configurational analysis.
Table 5 Configurational Model with Three Solution Pathways for Less Timely Completion (LTC) of the 3-Workshop Series in N=13 Churches
Outcome: # Days to complete 3 workshops Pastor outside employment # Health activities in past 2 years # Part-time staff # Volunteers Weekly member attendance
LTC No 3–4 6+ 51+ 101–249
LTC No 5+ 0 51+ 250+
LTC No 3–4 6+ 51+ 101–249
LTC No 3–4 6+ 51+ 250+
LTC Yes 3–4 0 26–50 <100
LTC Yes 5+ 1–5 26–50 <100
MTC No 5+ 0 51+ 101–249
MTC No 5+ 1–5 1–25 <100
MTC No 3–4 0 26–50 101–249
MTC No 1–2 6+ 51+ 101–249
MTC No 5+ 1–5 26–50 250+
MTC No 1–2 1–5 1–25 <100
MTC No 1–2 6+ 51+ 101–249
Note. Number of days to complete workshop series: 28–60 days = More Timely Completion (MTC); 61–84 days = Less Timely Completion (LTC).
Shading is used for clarity in reporting the configurational analysis.
Ethics approval and consent to participate: All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. All study methods were approved by the University of Maryland Institutional Review Board (#10-0691), and all participants provided written informed consent to participate.
Consent for publication: All authors have provided consent for publication.
Competing and financial interests: The authors declare that they have no competing or financial interests.
Clinical trial registration: The study was pre-registered at clinicaltrials.gov. Identifier: NCT02076958. Registered 3/4/2014. https://clinicaltrials.gov/ct2/show/NCT02076958
References
Aarons GA , Green AE , Palinkas LA , Self-Brown S , Whitaker DJ , Lutzker JR , Silovsky JF , Hecht DB , & Chaffin MJ (2012). Dynamic adaptation process to implement an evidence-based child maltreatment intervention. Implementation Science, 7 , 32. 10.1186/1748-5908-7-32 22512914
Allen JD , Torres MI , Tom LS , Leyva B , Galeas AV , & Ospino H (2016). Dissemination of evidence-based cancer control interventions among Catholic faith-based organizations: results from the CRUZA randomized trial. Implementation Science, 11 , 74. 10.1186/s13012-016-0430-6 27193768
Allen JD , Torres MI , Tom LS , Rustan S , Leyva B , Negron R , Linnan LA , Jandorf L , & Ospino H (2015). Enhancing organizational capacity to provide cancer control programs among Latino churches: design and baseline findings of the CRUZA Study. BMC Health Services Research, 15 , 147. 10.1186/s12913-015-0735-1 25889628
Ambühl M , Baumgartner M , Epple R , Parkkinen V-P , & Thiem A (2021). Cna: causal modeling with coincidence analysis. https://cran.microsoft.com/snapshot/2021-10-10/web/packages/cna/cna.pdf
Baumgartner M , & Falk C (in press). Boolean difference-making: a modern regularity theory of causation. The British Journal for the Philosophy of Science. 10.1093/bjps/axz047
Chaves M , & Eagle A (2015). Religious congregations in 21st century America: National Congregations Study. https://sites.duke.edu/ncsweb/files/2019/02/NCSIII_report_final.pdf
Cohen DJ , Sweeney SM , Miller WL , Hall JD , Miech EJ , Springer RJ , Balasubramanian BA , Damschroder L , & Marino M (2021). Improving smoking and blood pressure outcomes: The interplay between operational changes and local context. Annals of Family Medicine, 19 (3 ), 240–248. 10.1370/afm.2668 34180844
Coury J , Miech EJ , Styer P , Petrik AF , Coates KE , Green BB , Baldwin LM , Shapiro JA , & Coronado GD (2021). What’s the “secret sauce”? How implementation variation affects the success of colorectal cancer screening outreach. Implementation Science Communications, 2 , 5. 10.1186/s43058-020-00104-7 33431063
Damschroder LJ , Aron DC , Keith RE , Kirsh SR , Alexander JA , & Lowery JC (2009). Fostering implementation of health services research findings into practice: A consolidated framework for advancing implementation science. Implementation Science, 4 , 50. 10.1186/1748-5908-4-50 19664226
Durlak JA , & DuPre EP (2008). Implementation matters: A review of research on the influence of implementation on program outcomes and the factors affecting implementation. American Journal of Community Psychology, 41 (3–4 ), 327–350. 10.1007/s10464-008-9165-0 18322790
Feldstein AC , & Glasgow RE (2008). A practical, robust implementation and sustainability model (PRISM) for integrating research findings into practice. The Joint Commission Journal on Quality and Patient Safety, 34 (4 ), 228–243. 10.1016/s1553-7250(08)34030-6 18468362
Furnari S , Crilly D , Misangyi VF , Greckhamer T , Fiss PC , & Aguilera R (2020). Capturing causal complexity: Heuristics for configurational theorizing. Academy of Management Review, 46 (4 ), 778–799. 10.5465/amr.2019.0298
Gingiss PM , Roberts-Gray C , & Boerm M (2006). Bridge-It: A system for predicting implementation fidelity for school-based tobacco prevention programs. Prevention Science, 7 (2 ), 197–207. 10.1007/s11121-006-0038-1 16791522
Glasgow RE , Vogt TM , & Boles SM (1999). Evaluating the public health impact of health promotion interventions: The RE-AIM framework. American Journal of Public Health, 89 (9 ), 1322–1327. 10.2105/ajph.89.9.1322 10474547
Greenhalgh T , Robert G , Macfarlane F , Bate P , & Kyriakidou O (2004). Diffusion of innovations in service organizations: Systematic review and recommendations. The Milbank Quarterly, 82 (4 ), 581–629. 10.1111/j.0887-378X.2004.00325.x 15595944
Hickman SE , Miech EJ , Stump TE , Fowler NR , & Unroe KT (2020). Identifying the implementation conditions associated with positive outcomes in a successful nursing facility demonstration project. Gerontologist, 60 (8 ), 1566–1574. 10.1093/geront/gnaa041 32440672
Holt CL , Tagai EK , Santos SLZ , Scheirer MA , Bowie J , Haider M , & Slade JL (2018). Web-based versus in-person methods for training lay community health advisors to implement health promotion workshops: Participant outcomes from a cluster-randomized trial. Translational Behavioral Medicine, 9 (4 ), 573–582. 10.1093/tbm/iby065
Holt CL , Tagai EK , Scheirer MA , Santos SL , Bowie J , Haider M , Slade JL , Wang MQ , & Whitehead T (2014). Translating evidence-based interventions for implementation: Experiences from Project HEAL in African American churches. Implementation Science, 9 , 66. 10.1186/1748-5908-9-66 24885069
Kelly CM , Baker EA , Williams D , Nanney MS , & Haire-Joshu D (2004). Organizational capacity’s effects on the delivery and outcomes of health education programs. Journal of Public Health Management and Practice, 10 (2 ), 164–170. 10.1097/00124784-200403000-00011 14967984
Lincoln CE , & Mamiya LH (1990). The black church in the African American experience. Duke University Press.
Meyers DC , Durlak JA , & Wandersman A (2012). The quality implementation framework: a synthesis of critical steps in the implementation process. American Journal of Community Psychology, 50 (3–4 ), 462–480. 10.1007/s10464-012-9522-x 22644083
Miech EJ , Freitag MB , Evans RR , Burns JA , Wiitala WL , Annis A , Raffa SD , Spohr SA , & Damschroder LJ (2021). Facility-level conditions leading to higher reach: a configurational analysis of national VA weight management programming. BMC Health Services Research, 21 , 797. 10.1186/s12913-021-06774-w 34380495
Norton WE (2012). An exploratory study to examine intentions to adopt an evidence-based HIV linkage-to-care intervention among state health department AIDS directors in the United States. Implementation Science, 7 , 27. 10.1186/1748-5908-7-27 22471965
Palinkas LA , Mendon SJ , & Hamilton AB (2019). Innovations in mixed methods evaluations. Annual Review of Public Health, 40 , 423–442. 10.1146/annurev-publhealth-040218-044215
Petrik AF , Green B , Schneider J , Miech EJ , Coury J , Retecki S , & Coronado GD (2020). Factors influencing implementation of a colorectal cancer screening improvement program in community health centers: An applied use of configurational comparative methods. Journal of General Internal Medicine, 35 (Suppl 2 ), 815–822. 10.1007/s11606-020-06186-2
Rabin BA , & Brownson RC (2012). Developing the terminology for dissemination and implementation research. In Brownson RC , Colditz GA , & Proctor EK (Eds.), Dissemination and implementation research in health: Translating science to practice (pp. 23–54). Oxford University Press. 10.1093/acprof:oso/9780199751877.003.0002
Ragin CC (2014). The comparative method: Moving beyond qualitative and quantitative strategies. University of California Press.
Riley BL , Taylor SM , & Elliott SJ (2003). Organizational capacity and implementation change: a comparative case study of heart health promotion in Ontario public health agencies. Health Education Research, 18 (6 ), 754–769. 10.1093/her/cyf051 14654507
Santos SLZ , Tagai EK , Scheirer MA , Bowie J , Haider M , Slade J , Wang MQ , & Holt CL (2017). Adoption, reach, and implementation of a cancer education intervention in African American churches. Implementation Science, 12 , 36. 10.1186/s13012-017-0566-z 28292299
Santos SLZ , Tagai EK , Wang MQ , Scheirer MA , Slade JL , & Holt CL (2014). Feasibility of a web-based training system for peer community health advisors in cancer early detection among African Americans. American Journal of Public Health, 104 (12 ), 2282–2289. 10.2105/ajph.2014.302237 25320894
Scheirer MA , Santos SL , Tagai EK , Bowie J , Slade J , Carter R , & Holt CL (2017). Dimensions of sustainability for a health communication intervention in African American churches: A multi-methods study. Implementation Science, 12 , 43. 10.1186/s13012-017-0576-x 28351405
Slade J (2021, August 25). Personal Communication: Church communication to C. L. Knott.
Tagai EK , Holt CL , Scheirer MA , Santos SLZ , Haider M , Bowie J , Slade J , Whitehead TL , Atkinson M , & Wang MQ (2018). Assessing capacity of faith-based organizations for health promotion activities. Health Promotion Practice, 19 (5 ), 414–723. 10.1177/1524839917737510
Weiner BJ , Amick H , & Lee S-YD (2008). Conceptualization and measurement of organizational readiness for change: a review of the literature in health services research and other fields. Medical Care Research and Review, 65 (4 ), 379–436. 10.1177/1077558708317802 18511812
Whitaker RG , Sperber N , Baumgartner M , Thiem A , Cragun D , Damschroder L , Miech EJ , Slade A , & Birken S (2020). Coincidence analysis: a new method for causal inference in implementation science. Implementation Science, 15 , 108. 10.1186/s13012-020-01070-3 33308250
Wilcox S , Jake-Schoffman DE , Saunders RP , Kinnard D , Kaczynski AT , Hutto B , & James KL (2021). Predictors of implementation in the Faith, Activity, and Nutrition dissemination and implementation study: application of the Consolidated Framework for Implementation Research (CFIR) in a statewide initiative. Translational Behavioral Medicine, 11 (2 ), 419–429. 10.1093/tbm/ibaa025 32221601
Williams RM , Tagai EK , Santos SLZ , Slade JL , Carter RL , & Holt CL (2018). The role of leadership support in a church-based cancer education implementation study. Journal of Religion and Health, 57 (1 ), 146–156. 10.1007/s10943-017-0427-8 28597333
Woltmann EM , Whitley R , McHugo GJ , Brunette M , Torrey WC , Coots L , Lynde D , & Drake RE (2008). The role of staff turnover in the implementation of evidence-based practices in mental health care. Psychiatric Services, 59 (7 ), 732–737. 10.1176/ps.2008.59.7.732 18586989
Yakovchenko V , Miech EJ , Chinman MJ , Chartier M , Gonzalez R , Kirchner JE , Morgan TR , Park A , Powell BJ , Proctor EK , Ross D , Waltz TJ , & Rogal SS (2020). Strategy configurations directly linked to higher hepatitis c virus treatment starts: an applied use of configurational comparative methods. Medical Care, 58 (5 ), e31–e38. 10.1097/mlr.0000000000001319 32187105
|
PMC008xxxxxx/PMC8983010.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
0372477
965
Behav Res Ther
Behav Res Ther
Behaviour research and therapy
0005-7967
1873-622X
35248749
8983010
10.1016/j.brat.2022.104065
NIHMS1785658
Article
Lessons Learned from Designing an Asynchronous Remote Community Approach for Behavioral Activation Intervention for Teens
Jenness Jessica L. 1
Bhattacharya Arpita 2
Kientz Julie A. 3
Munson Sean A. 3
Nagar Ria 4
1 Department of Psychiatry and Behavioral Sciences, University of Washington, Seattle, WA
2 Department of Informatics, University of California Irvine, Irvine, CA
3 Department of Human Centered Design and Engineering, University of Washington, Seattle, WA
4 School of Psychology, University of Glasgow, Glasgow, Scotland
Correspondence should be addressed to Jessica Jenness, Ph.D., Department of Psychiatry and Behavioral Sciences, University of Washington; Child Health Institute, 6200 NE 74th St. Building 29, Suite 110; Seattle, WA 98115. [email protected]. Phone: 1-206-616-7967
26 3 2022
4 2022
15 2 2022
01 4 2023
151 104065104065
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Adolescent depression is common; however, over 60% of depressed adolescents do not receive mental health care. Digitally-delivered evidence-based psychosocial interventions (EBPIs) may provide an opportunity to improve access and engagement in mental health care. We present a case study that reviews lessons learned from using the Discover - Design - Build - Test (DDBT) model to create, develop, and evaluate a high-fidelity prototype of an app to deliver an EBPI for depression, behavioral activation (BA), on an Asynchronous Remote Communities (ARC) platform (referred to as ActivaTeen). We review work at each stage of the DDBT framework, including initial formative work, iterative design and development work, and an initial feasibility study. We engaged teens with depression, mental health clinicians, and expert evaluators through the process. We found that the DDBT model supported the research team in understanding the requirements for our prototype system, ActivaTeen, and conceiving of and developing specific ideas for implementation. Our work contributes a case study of how the DDBT framework can be applied to adapting an EBPI to a new, scalable and digital format. We provide lessons learned from engaging teens and clinicians with an asynchronous approach to EBPIs and human centered design considerations for teen mental health.
Digital Mental Health
Discover-Design-Build-Test
Asynchronous Remote Communities
Adolescent Depression
Depression Treatment
Behavioral Activation
pmcIntroduction
Approximately 9–11% of U.S. adolescents are diagnosed with depression each year (Mojtabai et al., 2016; SAMHSA, 2016). Adolescent depression increases risk for lifelong negative outcomes such as psychosocial disability and suicide (Clayborne et al., 2019). Despite the rise of depression in adolescence, only 1% of all U.S. youth receive outpatient care for depression each year, with low engagement among those treated (Olfson et al., 2003, 2009). Furthermore, uptake of evidence-based psychosocial interventions (EBPI) into usual care has been slow given training and implementation challenges (Fixsen et al., 2009; Williams & Beidas, 2019). It is critical to investigate innovations capitalizing on the near ubiquitous use of technology platforms among adolescents for improving the usability of and engagement in EBPIs. Here we present a case study reviewing lessons learned from application of the Discover, Design, Build, Test (DDBT) implementation model (Lyon et al., 2019) to understand how a specific EBPI, Behavioral Activation (BA) (McCauley, Schloredt, et al., 2016), may be improved by delivery on an asynchronous online platform that provides technological and peer support.
Technology-Mediated Interventions to Address Adolescent Depression
Traditional EBPIs are greatly limited due to implementation cost (Eiraldi et al., 2015), time burden, and lack of administrative support (Langley et al., 2010). Digital mental health innovations have the potential to transform EBPIs in ways that increase accessibility, engagement, and scalability, especially for traditionally underserved individuals (Comer, 2015). Previous digital adolescent depression treatments have included a range of options including multi-modal (e.g., videos, online workbooks and modules, animations) self- and therapist-guided computer-based cognitive behavioral therapy (CBT), video game based EBPI learning, and self-guided smartphone applications or text-based therapy (see Grist et al., 2019 and Hollis et al., 2017 for reviews). In a systematic reviews of digital health interventions, computer-based CBT was found effective for adolescents with anxiety and depression (Grist et al., 2019; Hollis et al., 2017). Pilot randomized control trials have demonstrated acceptability and feasibility of websites and mobile applications in delivering BA modules to adolescents with depression (Davidson et al., 2014; Rohani et al., 2020). While self-administered digital approaches, such as smartphone apps and online modules, may improve certain barriers related to access, maintaining user engagement over time continues to be a significant limitation (Fritz et al., 2014; Tatara et al., 2013). Similar to in-person EBPI delivery approaches, digital interventions have also been critiqued for their poor, unresponsive design, reducing their potential for real-world impact (Grist et al., 2019). Further, only 10% of CBT/BA mobile applications use core EBPI components, and many do not address safety and privacy needs crucial to safe platform engagement (Huguet et al., 2016). Therefore, our work aimed to address the need for creative use and design of technology to engage teens in EPBIs for mental health management.
Asynchronous Remote Communities (ARC) are a promising technology-based approach that capitalizes on the improved reach and usability of technology while providing support, social interactions, and peer-driven motivation, features that have increased engagement in digital health platforms (Epstein et al., 2016). Specifically, ARC platforms- like Slack- use private online spaces to both collect data and deliver weekly tasks to groups of target users or other stakeholders through automated chatbots and human moderators. Participants in ARCs may respond with multimedia options (e.g., de-identified photos, videos) and collaborate among themselves to generate ideas, problem-solve challenges, and provide supportive coaching. ARC platforms were originally used as a data collection tool in qualitative research allowing for asynchronous discussions by target users (e.g., clinicians, patients) on design and user needs for integrating technology into healthcare. Recently, work from our group has evaluated whether ARC may be a feasible delivery tool for modified EBPIs (Bhattacharya et al., 2021).
Specifically, the same ARC attributes that make them valuable for research–flexibility, transportability, and scalability–may also support their use to improve EBPI usability, reach, and engagement for adolescents who have limited access to mental health facilities. For example, ARC platforms could either supplement or reduce the need for synchronous therapy interactions through technology-based symptom and behavioral data collection and homework completion. Alternatively, ARC could serve as a fully self-directed tool by delivering EBPI modules within a community of peer learners. In order to determine the feasibility of ARC delivered EBPIs, we adapted the ARC method to (1) deliver early prototypes based on the weekly evidence-based strategies of behavioral activation, which is a straightforward and adaptable intervention for depression management (Tindall et al., 2017) and (2) facilitate remote participation and feedback from teenagers and clinicians in the design and testing process.
BA is based on a functional analytic model of depression that highlights the transactional associations among environmental stress, behavior, and mood (McCauley, Schloredt, et al., 2016). BA addresses depression through a number of behavioral strategies (e.g., mood-activity logging, goal-setting, problem-solving barriers) with the goal of (1) increasing the experience of positive reinforcement (rewarding experiences) to help improve mood and (2) decreasing avoidance of reinforcing activities (McCauley et al., 2016). Importantly, BA is easily tailored to fit each individual’s values and goals based on their context. BA has been successfully delivered across many platforms including self-directed online courses and clinician-directed approaches in primary care, telephone, and brief-group therapy (Dimidjian, 2011).
Our previous work examining teen stress management using ARC indicated that teens wanted digital tools to 1) support reflection/logging of factors that may impact mood management; 2) social/peer support; and 3) planning support for goal completion (Bhattacharya et al., 2019). Indeed, features such as mood and activity logging and visualizations, as well as text reminders for skill use, was feasible, acceptable, and improved platform engagement (Davidson et al., 2014; Rohani et al., 2020). While previous digital mental health tools have examined each of these aspects independently or in limited combinations (Grist et al., 2019; Hugeuet et al., 2016), maintaining user engagement (e.g., time spent using components of the digital tool, frequency of skill use/treatment recommendations in situ) over time continues to be a significant limitation (Grist et al., 2019). The unique aspect of using the ARC approach for mental health interventions is leveraging the combination of social connection and interactive support tools like the ability to integrate new and interactive media such as chatbots, photos, and videos on portable platforms across multiple devices. Specifically, the core components of ARC platforms that may improve patient engagement in BA and, ultimately, mental health and functional outcomes are: (1) between-session therapy homework and goal completion support through interactive chatbots, reminders, and access to online worksheets; (2) ecological momentary assessment (EMA) logging and visualizations of symptoms and behavior that can been viewed and shared by both patient and clinician; (3) asynchronous therapist coaching through direct messaging; and (4) scaffolded peer communities that collectively respond to moderated engagement prompts to share homework goals and challenges, provide asynchronous peer coaching, and general peer support. Therefore, we sought to extend our previous work by developing an app to deliver BA on an ARC platform (referred to as ActivaTeen).
Human-Centered Design for Mental Health and the Discover Design Build Test Model
Human-centered design (HCD) describes methods for learning about people and contexts to develop engaging, usable, and effective products and services (Holtzblatt et al., 2005; Holtzblatt & Beyer, 1997; Norman & Draper, 1987; Zomerdijk & Voss, 2010), as well as design patterns that can be translated from one system or intervention to another (Dearden & Finlay, 2006; Fincher et al., 2003). Intervention developers, implementers, and designers have increasingly applied HCD methods to support development and refinement of a range of mental health services and systems (e.g., Alexopoulos et al., 2015; Graham et al., 2019; Lyon, Dopp, et al., 2020; Lyon & Bruns, 2019; Mohr et al., 2013, 2017; Murnane et al., 2018; Rohani et al., 2020; J. Suh et al., 2020). For example, 1) the HCD technique of journey mapping–understanding a user’s current or intended path through use of a service and/or products–has been used to support staff and clinician understanding of patient experiences (Percival & McGregor, 2016); 2) Lyon et al. (2021) apply the HCD technique of cognitive walkthroughs–having domain experts go through prototype system according to scripted scenarios to identify barriers to effective use–to identify implementation barriers that could be addressed before implementation (Lyon et al., 2021). HCD methods aim to improve feasibility, acceptability, and effectiveness by shifting intervention design and implementation away from top-down processes to methods that engage the patients and clinicians who will be delivering them. These methods aim to tailor interventions to the people and contexts in which they will be used and might be led by design researchers or practitioners or may be even more participatory (e.g., Björling & Rose, 2019; Wadley et al., 2013).
The NIMH-funded ALACRITY center at the University of Washington funded the work discussed in this paper and seeks to support and further develop the use of HCD for mental health interventions and implementations. This center adapted the HCD process and aspects of implementation science into a Discover - Design - Build - Test (DDBT) model (Lyon et al., 2019) that scaffolds the application of HCD to mental health services (Figure 1). While the overall model is linear, results in any phase could indicate a need to return to an earlier phase.
Using DDBT in Designing ActivaTeen
We organized our research agenda, and the remainder of this paper, according to the DDBT model. We review lessons learned from the Discover and Design phases, as well as Building and Testing of preliminary interactive prototypes of ActivaTeen to assess the feasibility and acceptability of our design concepts. In the Discover phase, we invited teens with depression and low mood and mental health clinicians to begin using an ARC platform (Slack) and gave them design activities to understand how components of BA (McCauley et al., 2016) and ARC could be used with these stakeholder groups. In the Design and Build phases, we designed and built an ActivaTeen prototype that included a chatbot, digital homework completion, and ecological momentary assessment (EMA) logging and visualizations of mood and behavior and conducted expert review and usability evaluation to iterate on the designs to assure they were suitable for testing. Finally, in the Test phase, we evaluated our prototypes of ActivaTeen with a new set of teens experiencing depression or low mood and mental health clinicians to understand whether the overall approach was acceptable and feasible and to assess a more holistic evaluation of the user experience. We leave developing and testing a scalable, robust ActivaTeen platform for our future work.
This paper presents a case study reviewing our experience and lessons learned from applying the DDBT model’s methods to designing a technology-based mental health intervention (Figure 1). We summarize our process of using the DDBT framework in practice by summarizing the methods and results of our work within each phase of the DDBT model (see Table 1 for a broad overview). We end by reflecting on our findings related to using ARC as a new platform for mental health intervention delivery and how DDBT allowed us to identify and shape the intervention’s possibilities in collaboration with teens diagnosed with depression and their clinicians.
Discover
In the Discover phase, we sought to understand the needs of our target users, worked closely within our interdisciplinary team of psychologists and HCD experts, and involved target users in obtaining feedback on the early design of our ActivaTeen activities and prototype. This section reviews our Discover phase work as relevant to the scope of the paper in explaining the aspects that informed the DDBT process. In-depth details on methods and results are available in our prior publication (Bhattacharya et al., 2021).
Method
We enrolled target users, including teens experiencing depression and adolescent mental health clinicians, in an ARC based research study in order to understand their needs with depression management and obtain their feedback on low fidelity prototypes1 for ActivaTeen (Bhattacharya et al., 2020). Specifically, we conducted two separate ARC studies on Slack across 10 weeks with ten mental health clinicians (31–50 years of age), including therapists, primary care, and school counselors) who work with teens with depression, and eight teen participants (between 15 and 19 years of age2) who were currently experiencing mild-to-moderate symptoms of depression based on the PHQ-8 (Richardson et al., 2010; Wu et al., 2019). Please see Bhattacharya et al. 2021 Table 1 and Table 2 for further demographic details. We gave them weekly prompts sent to a Slack channel to evaluate target user feedback on several topics including (1) current technology use in therapy, (2) preferences on how to use ARC with therapy, and (3) feedback on prototypes for adapting paper-based BA content and homework to a digital format for delivery on an ARC platform (see Bhattacharya et al., 2021 for examples of prompts). Users were directed to respond to the prompt in the channel, but also had the option to direct message study team members if they had questions or thoughts they wanted to share privately. The team created prompts prior to study start but adapted the planned prompt each week based on target user responses from the previous week. Specifically, our group moderator monitored and summarized all participant responses as they came in, and our research team met weekly to discuss the prompts and target user feedback to inform adaptions to the upcoming week’s prompt. In developing and adapting these prompts, we drew on our team’s clinician investigator’s experience and knowledge of areas for potential improvements in the study design, activity design, and prototype design.
BA content and homework prototypes included mood and activity logging, chatbots for planning goals to improve mood and overcome barriers, including avoidance. We created low-fidelity prototypes to show how some BA activities might be adapted to an interactive asynchronous remote platform and showed them to our clinician and teen participants to obtain feedback. Prototypes included low-fidelity drawings or online survey-based mock-ups and screenshots where the research team could easily incorporate feedback and iterate on the design before investing resources into building the tool. Teen and clinicians did not interact with each other in the Discover phase. Teen participants interacted with each other by commenting on other teen’s posts and utilizing emoji reactions to each other’s posts. There were also structured activities that involved moderated direct message discussions between teens (see Bhattacharya et al. 2021 for more details on weekly activities). After the online activities ended, 9 clinicians and 5 teens interviewed with us about their experience in the study and completed exit surveys (Bhattacharya et al. 2021). Participants were compensated $5 per week (teens), $10 per week (clinicians), and $20 for completing the exit interviews.
Data Analysis
Our study data included Slack transcripts of discussions among all participants including polls and DMs, interview transcripts, and survey responses. Researchers inductively analyzed qualitative data (participants’ weekly responses on Slack and the exit interview transcripts). Three researchers initially read and discussed four interviews with the lead coder then generated a preliminary codebook based on that discussion. Three researchers then used this codebook to code the initial four interviews, discussed, and revised the codebook. After discussion, the researchers divided the remaining interviews and data from weekly activities to code independently. Because the weekly activities sometimes reflected different themes than the interviews, the coders periodically discussed additional themes and updated the codebook, recoding the data as-needed. Because the researchers started by coding the exit interviews–which spanned experiences in the entire study–changes from the initial codebook were minimal. We quantitatively evaluated target user feedback using the Acceptability, Intervention Appropriateness, and Feasibility of Intervention Measure (Weiner et al., 2017) and six constructs of the User Burden Scale, (Suh et al., 2016) including physical, mental & emotional, time & social, financial, difficulty of use, and privacy.
Summary of Results
Across both qualitative and quantitative measures, we found that ARC is feasible for engaging teens and clinicians in the early phases of the design process and obtaining their feedback on incorporating technical functionalities and BA concepts into the prototype. The key themes (see Table 1 for more details) we found from the Discover phase included: (1) clinicians and teens perceived benefits in using the ARC platform to support logging mood and activities and engaging with BA content and (2) both teens and clinician participants wanted interactive online support as a supplement (versus replacement) to in-person therapy, (3) clinicians highlighted concerns about managing boundaries around expectations of constant asynchronous access and crisis support, and (4) both teens and clinicians raised the importance of privacy and data security (Bhattacharya et al., 2021). To summarize the design insights gained from this study, teens did not want a chatbot to replace or emulate a human but envisioned its function as an interactive tool for self-reflection and completion of therapy goals.
Design and Build
The Design and Build phases of the DDBT model are typically intertwined in an iterative process until usability scores and/or expert consensus indicate that the prototype is suitable for the testing phase. In our process, we applied findings from our Discover phase to design and develop the ActivaTeen prototype (Table 1), evaluated its usability with HCD and clinical psychology experts, and iteratively refined aspects of our prototype before testing it with target users in the next phase.
Methods
Based on our results in the Discover phase that highlighted the themes of design needs in a prototype to supplement face-to-face therapy between teens and clinicians, we further made refinements to the ARC method and identified the technical and usability requirements of translating BA components into an ARC format.
Video for Psychoeducation on BA:
Clinicians and teens highlighted a need for interactive training and education components for individuals new to BA (Bhattacharya, 2020). To introduce BA to teens, we created an animated video (Nagar & Kemp, 2020) that follows the style of showing real time whiteboard drawings. The content for this video was prepared by storyboarding iteratively and was led by our teen volunteer. We obtained feedback from two clinicians who are experts in BA and the visual design and video-making was outsourced to a freelance Undergraduate designer.
ActivaTeen:
Based on feedback from teens in the Discover phase, we designed an application on the Slack platform called ActivaTeen in the format of an interactive smart diary. We did not aim to simulate human conversation through ActivaTeen but intended for teens to reflect through interactive prompts sent by the bot. This expectation was set for the teens at the start of the study. We incorporated four main modules of BA into ActivaTeen: (1) logging prompts and data visualizations for activity and mood tracking, (2) interactive, conversational style chatbot module to prompt teens to reflect on upward and downward spirals in mood and behavior, (3) interactive module for planning SMART (Specific, Measureable, Appealing, Realistic, Time-bound) goal and mini-stems, and (4) interactive, conversational style chatbot module to reflect on and problem-solve barriers to SMART goals.
We built the backend of the application (i.e., the parts of the system that retrieve, store, and manipulate data, but with which that the user does ot directly interact) using the Slack API, Javascript, and Node.js, building the modular functions for each BA module on the open source botkit library. Our approach of using an existing platform allowed us prototype key features of ActivaTeen while relying on the platform’s implementation of necessary-but-not-novel features, such as user authentication and profiles and discussion channels, a form of piggyback prototyping (Grevet & Gilbert, 2015). We parsed and cleaned the JSON data using Python scripts and prepared the visualizations summarizing teens’ logging data using Tableau.
Preliminary Testing with Design Students:
In February 2020, we invited one Undergraduate and three Masters’ students in human centered design & engineering as users with design expertise to conduct preliminary user testing and cognitive walkthroughs of the ActivaTeen Slack app for 7 weeks (Doherty et al., 2010). Each week, we conducted an in-person meeting and asked designers to try out the respective activity on Slack for 20 minutes while two researchers observed them. During each session, two researchers provided the prompts to the users, asking them to respond on Slack while observing them. After all students completed responding, researchers asked them to elaborate on what went well, what was challenging and why and the group discussed and brainstormed how to resolve the challenges within the limits of technical feasibility. Changes were addressed based on what the group considered important, and we iterated on the prototype weekly to get feedback on the changes till everyone was satisfied with it. Design students were also asked to log their mood and activity for 3 weeks while obtaining intermediate feedback on design changes each week. Key changes were made to the interface, including adding a freeform text box for users to provide context for logging activities and mood, support for understanding psychoeducational content, and visual representations of mood logging data. We then conducted a focus group for 40 minutes to discuss the designers’ experience, challenges, and changes they want in the design. We also asked our volunteer teen and two additional teens and a clinician from prior studies to remotely test and provide feedback on the four ActivaTeen modules via a survey or email. As is standard practice when involving expert evaluators in the design and build process, we did not obtain any demographic information on expert evaluators (Doherty et al., 2010).
Summary of Results
The final design and functionality of the modules are explained below. Table 2 shows the alignment between ActivaTeen modules and the BA protocol content.
The mood-activity logging module prompts teens to log an activity from the past 3–6 hours, the type of activity, how they were feeling during the activity, and intensity of the feeling (Figure 2a). The nine types of activities included everyday activities, hobby, physical activity, relaxation, school, socializing, work, and other. After week 6 when they planned a SMART goal, a category of “SMART goal” was added. Options for feelings were based on the circumplex model of emotions (Russell, 1980) which included positive (content, relaxed, focused, happy, excited), neutral, and negative (angry, overwhelmed, depressed, anxious, tired, sad, bored) valence. Teens could log the intensity of their feeling ranging from 1: least intense to 10: most intense. Based on feedback from expert designers, we added a text box for teens to freely journal any additional context that may be relevant to the activity or how they were feeling.
The reflecting on upward and downward mood spirals module prompts teens to think about a time in the week when they noticed their mood was “down” and a time it was “up”. In a conversational style, the chatbot app prompts the teens to respond to: what happened, how they felt about it, and what they did as a result of feeling that way.
The planning SMART goals module presents teens with a digital form to support setting a SMART goal and mini-steps. To support teens with generating ideas for goals, we also created examples of SMART goals and mini-steps and added suggestion cards for 6 types of activities: hobby, everyday activities or chores, socializing, relaxation, do less of something, and physical activity (Figure 2b).
The overcoming barriers module uses an interactive chatbot to ask teens what barrier(s) they encountered in completing their mini-steps (Figure 2c), educate teens about the internal and external types of barriers with examples, and ask them to plan how they would overcome these barriers. Suggestion cards were designed for both internal and external barriers with an intent to help teens brainstorm ideas and adapt it to what was feasible in their context. The teens could then plan their SMART goal by getting redirected to module 3 with revised mini-steps for overcoming barriers.
The key outcome of this stage was the implemented and refined prototype for the Test phase. Through this development and refinement, we also identified ideas for improvement and new features. Changes included having a space for free reflection (context in the EMA logs), framing of the interactive app as a smart diary instead of a chat bot, and revising the framing of the weekly activities and ActivaTeen prompts. We also identified questions that participants would like to answer based on their logs, which helped us develop visualizations to support reflection. We stopped making changes to the prototype when expert designers, the design team, and teen volunteers agreed that all identified concerns that could interfere with an informative Test phase had been addressed. This qualitative approach and feedback were valuable for rapidly iterating and testing during the early stages of prototyping to prioritize changes before testing it with teen and clinician participants. These reflections supported the team in making decisions about proceeding to the next phase.
Test
Once our Design and Build phase prototype met the requirements identified in the Discover phase, we moved onto the Test phase to assess the feasibility and acceptability of the prototype with teens and clinicians. In this phase, we sought to inform future planned clinical trials of the proposed system.
Methods
We started the Test phase of the study during the onset of the COVID-19 pandemic in the US. To recruit teens between February and March 2020, we advertised our study in online groups, sent messages and flyers to clinicians and a mailing list of parents with teenagers. We directed interested teen participants to fill out a screener survey with contact information and enrolled teens experiencing mild to moderately severe depression symptoms based on the PHQ-8 questionnaire (Richardson et al., 2010; Wu et al., 2019). If the teen was experiencing moderately severe depression, we also required that they had a current therapist. We recruited clinicians through Seattle Children’s Hospital Mood and Anxiety Disorders outpatient clinic (the child psychiatry clinic connected to the University of Washington) via emailed flyers. Participants were compensated with $10 on completing each week’s activity and $20 for completing the exit interviews. All study activities were conducted between May and August 2020.
Eleven teens (Table 3) consented and joined Slack. We present the study recruitment and retention flow in Figure 3 and summarize the demographic information of Teens and Clinicians in Tables 3 and 4, respectively. Clinicians and teens dropped out of the study due to personal life challenges, transitions, and difficulties with keeping up (which may have also been aggravated due to the pandemic). Please see Figure 3 and Table 5 for a count of participants who participated each week.
We conducted the online activities for 8 weeks (Table 5). We designed each week’s activity to guide teens through BA content and corresponding supports in our ARC, and to learn about their experiences with those supports. Similar to our Discover phase process, we discussed progress and preliminary data each week to finalize the activity for next week for clinicians and teens. The content for each module is explained in Appendix A.
Online group activities:
In Weeks 1 and 2, clinicians and teens were enrolled in the same Slack group. They were provided a short tutorial on how to use Slack features, asked to introduce themselves with ice-breaker questions, and asked for feedback on the BA psychoeducation video. The clinician and teen groups were separated in week 3 as we wanted the teens to be able to reflect on their own accord without the influence of clinicians.
Both teens and clinicians were asked to track their mood and activities in week 3 and reflect on their experience in week 4 in their respective groups. The teens additionally tracked their activities and mood for 5 weeks (weeks 3, 4, 6, 7, and 8). We asked teens to log mood and activity on ActivaTeen with 112 reminders over 5 weeks (35 days) of the study at 3–6-hour intervals between 9 am and 9 pm. We changed the timing and frequency of the reminders based on the preliminary analysis of the times at which the participants logged and the feedback from participants. Participants were asked to let us know any time if they wanted us to change the frequency of or stop the reminders.
Each ActivaTeen module was interleaved with one week of reflection and feedback on the module. Teens completed the ActivaTeen modules on upward /downward spiral in week 3, SMART goal planning module in week 6, and the overcoming barriers module in week 8. The clinicians also piloted these activities and provided feedback. In weeks 4 and 8, we created a direct message (DM) group in which a teen, clinician, and one researcher present as a moderator interacted with each other in mock therapy scenarios. For example, the clinicians were asked to initiate the conversation and provided example prompts to help teens reflect on their logging data in week 4 and review SMART goals, brainstorm how to overcome barriers, and review data in week 6. Across all weeks, teen participants interacted with each other by commenting on other teen’s posts and utilizing emoji reactions to each other’s posts. The creation of a clinician facing ActivaTeen “dashboard” to provide data visualizations and summary of patient homework completion is a next step for our work, so summaries and visualizations were created by research staff and sent to clinicians via direct message on the ActivaTeen platform.
Exit Interviews and Surveys:
Nine interviews (7 teens, 2 clinicians) were conducted on Zoom and participants were given the choice to turn off the video. One teen (T31) chose to respond via text chat only with her audio and video turned off as the interviewer asked questions verbally while being available on video. The ninth interview was conducted over the phone with a clinician. Interviews lasted between 30–60 minutes. All interviews were transcribed automatically via Zoom, and we read through and edited the transcript for errors while watching each video.
For each interview, we prepared a slide deck with screenshots of the four main modules of ActivaTeen, summarized graphs for each teen participant, and SMART goals (as available). The graphs included representations of valence of emotions across categories of activities, time, frequency of logging, and a summary table of specific activities across time and valence (Appendix B). For the clinician interviews, we used graphs from two teens as examples. During the interview, we shared our screen co-viewing the deck of slides with the participants with screenshots of ActivaTeen user interface (UI) to ask them for feedback on the design. Seven participants (5 teens, 2 clinicians) also completed a post-study PHQ-8 survey (Richardson et al., 2010; Wu et al., 2019).
Ethical Considerations:
All phases of this study are a part of a larger project for which we obtained emergency contact information from all participants and parental permission was not obtained formally. During recruitment in our prior work (Bhattacharya et al. 2019), we found that some interested teens were unable to participate as their parental contact was not responsive or teens were unwilling to provide contact information of the parent. To reduce such barriers for teens in participating, we consulted with our university’s Institutional Review Board (IRB) and obtained a complete waiver of parental permission. In the Test phase, the majority of the teens were recruited through parents. All participants chose anonymous pseudonyms, and no identifiable information was shared on the groups or direct messages. Teens were informed during the consent process that clinicians will be on the Slack group and will be able to see their mood and activity logging data (with their pseudonym). Given the joint teen and clinician involvement, researchers made sure to clarify to participants that they were enrolling in a feasibility, acceptability, and usability study and not a treatment study. Adversity mitigation protocols were still relevant and used from prior studies and emergency contact information was obtained for all participants. Three researchers moderated the group and responded to participants within 24 hours. There we no adverse or crisis events and no need to contact emergency contacts throughout all phases of our project. This study was approved as minimal risk by the University’s Human Subject Division.
Data Analysis
The coders first read and inductively coded a part of the data (four teen-interview transcripts) line by line and wrote their codes in a cumulative document. The coders then met for weekly discussions to prepare a codebook. We distributed the weekly Slack data and interview transcripts between three coders and the remainder of the data was coded using the final codebook (Appendix C). All coders read and shared memos with each other and met for discussions to resolve discrepancies in coding. We conducted affinity modelling (Holtzblatt et al., 2004), an open-ended process by which researchers iteratively discuss and cluster data, to identify emergent themes. Three main themes emerged from our analysis. The research team further iterated on these themes through discussion and while writing. We summarize the results focusing on the insights we gained for iterating on our design below and a detailed empirical analysis can be found in (Bhattacharya, 2020, Chapter 5).
To provide descriptive characterization of our sample, we calculated average scores of PHQ-8 pre- and post-study. We calculated an average score for each item of the Acceptability, Intervention Appropriateness, and Feasibility of Intervention Measure for the clinician group and teen group, respectively. For the User Burden Scale (Suh et al., 2016), we computed average scores of teens and clinician groups separately across each of the 6 constructs – physical, mental & emotional, time & social, financial, difficulty of use, and privacy. In the IUS, items are rated on a Likert scale from 0 (strongly disagree) to 4 (strongly agree), with half of the items reverse-scored (Lyon, Pullman, et al., 2020). A total score is normally calculated by multiplying the sum of these scores by 2.78 (range: 0–100). Drawing parallels from scoring the system usability scale (SUS), above a 68 would be considered above average and anything below 68 is below average (Lewis, 2018).
Summary of Results
ActivaTeen was found to be feasible, usable, and acceptable across qualitative and quantitative measures. Specifically, three main themes emerged from our qualitative analysis of the weekly feedback and interview data based on our goals. Based on our empirical analysis, participants reported that ActivaTeen supported teens in moving from avoidance to action by (1) providing structure for data collection and reflection on the self-tracked mood-behavior data to better set and follow through meaningful goals. Specifically, data collected from BA homework modules and ActivaTeen logging and visualizations of mood and behavior that was viewed and shared by both patient and clinician assisted in supporting use and engagement of BA content; (2) scaffolding interactions with clinicians asynchronously through direct messaging; and (3) scaffolding peer communities to share insights, goals, and challenges and general peer support (Bhattacharya, 2020).
Due to our sample size, quantitative data were collected with the aim of describing our sample and providing pilot feasibility, usability, and acceptability data. All teens engaged in the mood-activity logging module with the total number of logs ranging from 3 to 125 out of 112 logging prompts (M= 52.13, SD= 45.53) and number of days logged ranging from 3 to 37 (M= 12.71, SD=10.34) across 35 days of logging (two teens continued to log after reminders were stopped). Five teens and two clinicians responded to the end of study survey with the usability metrics. On average, teen participants reported depression scores in the moderate range pre-study (M= 14.44, SD= 3.94) and the mild range post-study (M= 9.60, SD= 6.27). Overall, participants indicated low burden and high acceptability/feasibility for the online intervention. On the User Burden Scale (1: not at all burdensome, 4: very burdensome) (Suh et al., 2016), participants scored an average of 0.6 (clinicians) and 0.6 (teens) on difficulty of using Slack and 0–0.2 on all other types of burden. Target users reported high acceptability, intervention appropriateness, and feasibility of intervention (1: Strongly Disagree, 5: Strongly agree) (Weiner et al., 2017) with average scores between 3.4 and 3.5 for teens and 3.6 and 4 for clinicians. The mean IUS score was > 68 (above average; Lewis, 2018; Lyon, Pullmann, et al., 2020).
Discussion
In our application of DDBT, we combined the clinical and human centered design expertise of our research team with teens’ and clinicians’ experiences to understand the design needs for the successful use of ARC supported BA (Discover phase). This also facilitated generation of ideas on what additional resources could support teens and clinicians in their use of BA (a combination of Discover and Design) and iterative building and testing of the advanced prototype of our ARC app, ActivaTeen. Through this work, we reflected on design considerations gained from our work, applications of DDBT to innovate on mental health care approaches, and future directions.
Reflection on Design Considerations
Our work led to several general design considerations that may also apply to the design of other digitally-delivered EBPIs. Teens and clinicians reported that ActivaTeen’s design supported teens in reducing avoidance by providing structure through logging and adjusting their perception of control by planning SMART goals and breaking down goals into smaller tasks. Using reminders and planning tools, technology can more directly intervene in the avoidant thought process in-situ with psychoeducation and reminders for action supporting teens in breaking down activities into manageable steps. Further, technologies can be used to help teens keep track of progress with their mini-steps and provide intermediate rewards for small achievements.
Understanding one’s behavior and mood patterns through reflection is crucial for teens who are undergoing psychosocial transitions and developing routines and habits that may last a lifetime (Galván, 2017; McCauley, Gudmundsen, et al., 2016). The levels of reflections demonstrated by teens in this study ranged from teens merely describing what they notice in the data, supporting inquiry to elaborate more on why they observe certain patterns in their routines and making connections, and ultimately taking transformative action to change their activities and plan goals (Baumer, 2015; Fleck & Fitzpatrick, 2010; See Appendix C). Asking participants early on to reflect on why they are logging and providing intermediate summaries to support potential outcomes would be helpful to revisit intermittently for digitally delivered EBPIs.
As we developed ActivaTeen, we needed to make design changes to meet the temporal and structural expectations of teens such as personalizing the timing and frequency of logging reminders based on the participants’ routines and the times they feel most comfortable to log. Rohani et al. (2019) recommended providing options for not only in-the-moment logging but designing to support participants in retroactively logging multiple activities. To enable accurate interpretations and account for recall bias, these logs need to be differentiated by providing visual indicators for what data was logged in-the-moment versus data that was logged later. Teens also reported using non-digital tools such as sticky notes or retaining the information without reminders. It is important to identify and be inclusive of what meaningful engagement means for the teen even if it does not always involve interacting with the technology and what engagement would mean for a therapist-client interaction for a particular evidence-based practice. For example, for ActivaTeen we conceptualize meaningful engagement as (1) compliance or adherence, in whether participants act on what the treatment recommends and (2) a proxy for whether participants can receive support for the intervention in digital form when they need it in the absence of a therapist.
Consistent with research showing that the inclusion of even minimal therapist contact improves outcomes of digitally delivered treatments (Grist et al., 2019), both teens and clinicians wanted to preserve the human connection with a clinician, online or in-person. Though teens preferred to have asynchronous direct messaging in addition to reviewing the data during therapy, clinicians spoke about the burden of their caseload which did not provide them with additional time outside of weekly synchronous therapy sessions. Both teens and clinicians found value in reflecting on the data summaries and graphs and discussing it together. Clinicians wanted to use the data summary and visualizations to identify activities and contexts that triggered negative emotions and the intensity of the emotion, identify positively reinforcing activities, and plan future goals and recommendations with teen clients during synchronous online or in-person therapy sessions. Research suggests that visualizations summarizing logged data based on personally meaningful criteria supports meaningful reflection (Epstein et al., 2016). Relatedly, our findings show that problem-solving and goal-planning discussions between patients and therapists were supported by data visualizations displaying patterns and triggers associated with positive and negative changes in mood. Data visualization tools may also provide agency to the teens by allowing them to share what they think is valuable to their treatment with their clinicians. Using the app as a central repository of psychoeducational information with examples and suggestion cards can support teens in further reflection on reviewing and generating ideas for SMART goals.
Our work also highlights the inclusion of a peer community component to digitally-supported EBPIs to potentially improve engagement in and effectiveness of care. Built-in peer support may help maintain user engagement by increasing feelings of social belongingness and decreasing stigma around mental health concerns. Social belonging has been identified as an important component in effective mental health outcomes (Allen et al., 2018; Newman et al., 2007; Nuttman-Shwartz, 2019) and is inversely correlated with behavior problems in adolescents (Newman et al., 2007), provides protection against stress for adolescents (Nuttman-Shwartz, 2019), and may be an essential component to improve feelings of isolation (Allen et al., 2018). The integrated peer support network also works to combat mental health stigma, a common barrier for adolescents seeking mental health treatment (Chandra & Minkovitz, 2006; Gulliver et al., 2010). Indeed, reductions in mental health stigma have been observed for individuals interacting with other treatment seeking peers (Pinfold et al., 2005; Wade et al., 2011).
Reflection on Using the DDBT Framework
As we consider the next steps in our research, we anticipate questions that might support further development of the DDBT model and provide guidance to other research teams applying DDBT in their own work. First, while DDBT emphasizes the iteration between the Design and Build phase, the design process is overall iterative. For example, in our work we used Slack because it was a flexible, user-friendly platform for exploring different designs and connecting with participants as we conducted our initial Discover, Design, Build, and Test phases. As we move toward implementing tools that providers can adopt in their day-to-day work, HIPAA regulations and various organizational factors mean that Microsoft Teams will be the ARC platform better suited for our intended use context. This work necessitates backing up and revisiting certain aspects of the design: not every design choice that worked well on Slack is appropriate to just copy to Microsoft Teams, and so we must redesign some features.
In our Discover and Test phases, the ARC was a both platform for collecting input from participants and for rapidly prototyping design ideas, leading to conflation between aspects of the research study and aspects of the intervention. When researchers integrate using ARC as both a research tool to gather information and as an intervention delivery method for BA, it may be most usable for participants, but it also may confuse and conflate our qualitative research with the intervention. This is a tradeoff we and others have noted in previous uses of ARC in research (Bhattacharya et al., 2019; Maestre et al., 2018). While we deemed it most beneficial to combine the research platform with the prototyping platform at this stage of the research, future work oriented toward evaluating ActivaTeen’s efficacy will need to test the prototype on its own, without conflating research components.
Finally, the DDBT framework felt familiar to the members of the research team trained in human centered design, but teams can borrow from and apply other frameworks and methods from human centered design and other fields. Recent work has described the potential of HCD for mental and behavioral health (e.g., Graham et al., 2019; Lyon, Brewer, et al., 2020; Lyon, Dopp, et al., 2020) as well as the alignment between HCD and implementation science (e.g., Dopp et al., 2019, 2020). Others may see similarities to the plan-do-study-act (PSDA) cycles common in quality improvement work (Langley et al., 2009), though DDBT focuses more on initial discovery and more open-ended design phases in pursuit of new solutions, while PSDA is focused more on continuous improvement in the field. Each framework will have different emphases and may also support buy-in from different stakeholders based on their prior experiences with that framework.
Limitations and Future Work
Teens with depression are difficult to reach for participation in research. During the Test phase, the onset of the pandemic impacted and delayed recruitment as we could no longer reach teens through in-person measures, participants’ schedule and activities changed (e.g., no school, lack of or no access to peers, migration to online activities, and cancellation of camps or part time work), and the coping strategies that teens could feasibly use were limited due to social distancing. Due to recruitment challenges related to the pilot nature of our project and COVID-19 pandemic, our sample was skewed towards people who could be reached through researchers’ contacts, mailing lists, clinics, and snowball sampling and were willing to participate during the COVID-19 pandemic. Our sample is not racially or ethnically diverse, which limits the generalizability of our findings. We could not further evaluate the usability with real-world clinician-client dyads and could only ask participants to speculate how this type of guided ARC system would work with their real-life therapists or patients. Recruiting real world client-therapist pairs and a diverse and representative group of participants is the next step for our work in order to provide more helpful insights into understanding how a system like this would feasibly fit into the clinicians’ workflow, influence the therapeutic relationship, and influence cultural barriers to care.
While both teens and clinicians highlighted that they preferred ARC as a supplement versus replacement to therapy during our Discover phase, this phase was conducted prior to the COVID-19 pandemic, and it is possible that these views may have changed. Indeed, the pandemic was a catalyst for mental health care systems to migrate to online telehealth using teleconferencing tools (e.g., Torous et al., 2020; Torous & Wykes, 2020). Specifically, as social distancing requirements due to the pandemic continued for over one year in the United States at the time of writing, we speculate that compared to our data, mental health clinicians and teens may have increased acceptance over the potential use of online technologies for therapy and may have devised creative strategies for compartmentalizing their online time, incorporating contextual information from clients’ homes, and using available mental health apps and digital tools. Clinicians might want to continue using these digital tools when they return to in-person therapy creating more opportunities for investigating hybrid in-person and ARC formats of therapy.
Additional options to integrate ARC into clinical work include using ARC as a platform for the client-clinician dyad alone and/or creating a peer group with multiple patients at a clinic who all join the same group (either for patients attending group therapy or seeing individual clinicians at the same clinic). Understanding the feasibility of this in-context would require more intensive recruitment efforts, using a HIPAA compliant system such as Microsoft Teams, and connecting with administrative and technological stakeholders at clinics. Though technology has the potential to provide increased access, we acknowledge that ARC approaches exclude teens who do not have regular and consistent access to digital tools and internet connectivity. Indeed, a nationally representative sample of teens shows that, while the vast majority of teens own a smartphone in both low- and high-income households, there continues to be disparities in smartphone ownership (i.e., 74% vs. 89%, respectively) (Rideout & Robb, 2019). Therefore, it is important for efforts to improve access to care also consider approaches that do not solely rely on technology.
In our future work, we aim to we aim to evaluate mental health and usability outcomes in a larger scale RCT with and without the guided ARC enhancement in conjunction with synchronous BA therapy. As part of this work, we plan to further collaborate with clinical stakeholders to examine whether and how the ARC approach increases the feasibility and effectiveness of hybrid synchronous and asynchronous care. It is integral for researchers to work with organizational decision-makers and stakeholders in adolescent clinical settings and industry partners to create innovative solutions and work towards real-world deployment to improve scalability and accessibility of evidence-based mental health care for teens. Given the varied components of ARC platforms that may improve engagement in skill use, clinical outcomes, and functioning, we also plan to utilize platform engagement metrics (e.g., time spent with chatbot teaching modules, frequency of EMA logging, frequency of peer and patient-therapist direct messaging) to explore which aspects of ActivaTeen are most associated with improvements in symptoms and outcomes.
Conclusion
Using the DDBT framework, we designed, developed, and studied the feasibility of using integrated peer, clinician, and interactive ActivaTeen app support with teens experiencing depression. Our work demonstrated that the DDBT model was a useful frame in which to understand target user needs for adapting BA to an ARC format, build an advanced prototype and to empirically understand how engagement with this prototype varied across teens and mental health clinicians. Our work provides a platform for future research to investigate the use of ARC supported BA, and other EBPIs, in real-world clinical contexts with the ultimate aim of improving access to, engagement in, and scalability of mental health care for teens.
Supplementary Material
1
2
3
Acknowledgements:
We thank all participants. We are grateful for constructive discussions with colleagues at the University of Washington and for feedback provided by Drs. Schueller and Cohen. This work is funded by the University of Washington ALACRITY Center under NIMH Award #1P50MH115837. This project was supported by the National Center for Advancing Translational Sciences of the National Institutes of Health under Award Number UL1 TR002319 and the National Institute of Mental Health (K23MH112872). The content is solely the responsibility of the authors and does not necessarily represent the official views of funding agencies.
Figure 1: Summary of the overall DDBT process and the methods used in our study using DDBT approach for designing for mental health interventions.
Figure 2: Four modules of BA supported by ActivaTeen: (a) logging activity: teens log feeling, intensity, and journaling context, (b) planning SMART goals and mini-steps, and (c) overcoming barriers to SMART goals (author name as user for anonymity)
Fig. 3. Study recruitment and retention flowchart.
Table 1: Summary of design of our prototype through the different phases of DDBT based and takeaways from the empirical work with teens, clinicians, and experts which informed the next phase
Module Discover Phase Design & Build Phase Test Phase
BA Psychoeducational Video Prototype: NIMH BA video and examples of animated educational videos Prototype: Storyboard with hand drawn panels outlining the video script and visuals reviewed by experts turned into animated video Prototype: Animated BA video showing the journey of a teen experiencing depression and BA treatment
Takeaways:
- Make the video interactive
-Use tangible examples for teens
-Use animations Takeaways:
- Minor edits by experts (child clinical psychologists) to content and story of storyboard and video to be aligned with BA principles Takeaways:
- Add content explaining the upcoming modules as well to what to expect in the coming weeks
Module 1: Mood and Activity Logging Prototype: Mock-up of mood and activity logging in survey format Prototype: ActivaTeen programmed to send an automatic prompt to log activity, emotion and intensity every 3 hours between 9 am-9 pm for 3 weeks. The app provided a summary for each log. Expert designers were asked to reflect on their summaries and what they would like to learn from these data. Prototype: ActivaTeen sends an automatic prompt to log activity, emotion, intensity, and journaling context every 3–6 hours between 9 am-9 pm. Visualization of data provided to teen and clinician. Direct message between teen and clinicians two weeks after logging to discuss data.
Takeaways:
-Add reminders to the tracking prompts
-Ensure in-person review with a clinician
-Allow for data review to check on progress
-Consider number of notifications to avoid teens ignoring prompts Takeaways:
-Need to provide more context for elaborating on emotion through a free text journaling box
-Frame the prompt to allow for tracking of multiple activities
- increase the interval between prompts (max 3 daily)
-Consider randomly sending prompts through the day to get better sampling of activity/mood Takeaways:
- Account for difficulty in identifying negative emotions by setting expectations for logging a range of emotions and not just positive emotions
- Show real-time visualizations for reflection to facilitate what they understand from these data and to motivate logging
- Reduce guilt and overwhelming backlog when wanting to return to logging after not logging
- Provide space for both logging in the moment and logging later
- Continue to allow for clinician-teen reflection on data
Module 2: Upward and Downward Spiral Prototype: Survey format with prompts on reflecting on upward and downward spirals in mood and action Prototype: ActivaTeen programmed with Interactive chatbot with prompts Prototype: Interactive chatbot with prompts
Takeaways:
-Teens noted survey was clear, simple and liked that it supported reflection on situation-mood-activity spirals Takeaways:
- Add a prompt to think about what they would want to do next after experiencing an emotion Takeaways:
-No changes: Teens and clinicians continued to note the prompts supported reflection
Module 3: plan SMART goals Prototype: Survey format for individually planning a SMART goal, mini steps, and setting reminders. Illustrated screenshot mock-up of chatbot format. Evaluated direct messaging format in which participants pair up with a peer and a researcher moderator. Participants were asked to share SMART goals and provide feedback on each other’s goals. Prototype: ActivatTeen programmed with Interactive chatbot that allowed teens to generate a SMART goal and ministeps. Included suggestion cards to help generate SMART goals. Prototype: Added visual example of SMART goal that defined SMART acronym. Added prompt to direct message with clinicians the week after planning their SMART goal and ministeps
Takeaways: -preference for interactive format of chatbot vs survey
- Consideration of what to do when direct messaging falls through if one of the teens in the pairing does not respond Takeaways:
- Include a visual explanation of what a SMART is that includes an example
-Include more support when choosing and setting SMART goals Takeaways:
- Clinicians said quality of SMART goals would be improved by synchronous support to help teens plan SMART goals ahead of time instead of after
-Most teens found it helpful to plan and break down the goals to reduce avoidance
-Teen did not find Slack reminder tool to be useful
- Some teens used sticky notes to remind themselves of mini-steps others did not
- Mini-steps should include some tracking of progress when each step was completed and enable setting bulk reminders for regular activities
- All teens found messaging with clinicians to be helpful
Module 4 Overcoming barriers to SMART goals Prototypes: Survey format for assessing barriers. Illustrated mockup of chatbot format with prompts to overcome barriers. Evaluated direct messaging with paired peer from week 8 to allow for follow up and sharing barriers Prototype: ActivaTeen programmed with interactive chatbot to help identify internal and external barriers, suggest strategies to overcome barriers, and re-plan SMART goal to account for overcoming barriers Prototype: No changes from Design/Build phase
Takeaways: -Preference for interactive format of chatbot vs survey
- Moderator and researchers provided direct message support if teen’s peer partner did not engage Takeaways:
-No changes suggested by expert evaluators Takeaways:
- Some teens found it difficult to identify barriers
- Certain external barriers outside of teens’ control were unpredictable to plan for and chatbot did not help (e.g., waiting to hear back from a friend or medication taking effect)
Table 2. Overview of the alignment between the BA protocol content (McCauley et al., 2016) and ActivaTeen app delivered via an ARC platform
BA Protocol Content ActivaTeen Support via ARC
Session Modules and Content Overarching features: Direct message therapist, prompts to share success and challenges to peer community
1–2 Intro to BA: BA Model of Depression, Activity/Symptom Monitoring Peer community introductions, BA Psychoed video, daily EMA mood-activity tracking
3–4 Activation/Goal Directed Behavior: Situation-Activity-Mood Connection, Regulating Mood with Activity, Linking Behavior to Short- & Long-term Outcomes, Making the Most of a Good Feeling ActivaTeen homework support for tracking upward and downward mood spirals; mood-activity EMA tracking
5–8 Core Skill Building: Problem Solving, Goal Setting, Identifying Internal & External Barriers, Overcoming Avoidance, Use Alternative Coping ActivaTeen Goal setting, problem-solving, and overcoming barriers chatbot support; BA goal and mini-step tracking; EMA tracking
9–11 Practice/Application: Review and Integrate Content, Practice Skills ActivaTeen Goal setting, problem-solving, and overcoming barriers chatbot support; goal, mini-step, and mood EMA tracking
12 Relapse Prevention & Termination: Strategies for Preventing Relapse, Identify Strengths & Challenges for Moving Forward Continued access to ActivaTeen chatbots, therapist messaging, and peer community
Table 3: Summary of teen participants’ demographic information in study 3 (n=9)
Age 13–19 years (mean= 16 years)
Gender Female (n=4), Male (n=4), Non-binary (n=1),
Education level Some high school (n=5), Some college (n=1), No response (n=3)
Race White (n=5), More than one race (unspecified) (n=1), no response (n=3)
Therapy experience Currently seeing therapist (n=5), past therapy (n=2), no previous therapy (n=5)
Pre-study PHQ M= 14.44, SD= 3.94, range= 6–19
Table 4: Summary of clinician participants’ demographic information in study 3 (n=2)
Age 29–41 years
Gender Female (n=3)
Number of years of clinical experience treating teens 5–12 years
Table 5: Number of activities completed by teens in each week in Study 3 (total number of teens who joined the Slack group n=11, number of teens who continued after week 1 n=9)
Activity type in Test Phase Number of teens who completed
Week 1: Introductions and Ice Breakers 10
Week 2: Review of BA psychoeducation video & Begin Mood-Activity Logging 6
Week 3: Upward/Downward Spiral Chatbot 6
Week 4: Clinician-Teen DM to review logging 9
Week 5: Reflection on Clinician-Teen DM 6
Week 6: SMART Goal Planning 7
Week 7: Sharing on group about SMART goal 5
Week 7: Peer feedback on SMART goals 0
Week 8: Clinician-Teen DM SMART goal barriers 7
Week 8: Barriers to SMART goals 6
Mood and activity Logging 9
Interviews 7
End of study survey 5
* Note: DM= Direct message
Highlights
Digital platforms may improve teen access to and engagement in mental health care.
This is a case study of the Discover-Design-Build-Test (DDBT) implementation model.
We applied DDBT to adapt a treatment to an Asynchronous Remote Communities platform.
The DDBT model supported our understanding of prototype requirements.
Future work will test a robust digital platform to improve access and engagement in care.
Conflict of Interest: The authors declare no conflict of interest.
Author Credit Statement:
Jessica L. Jenness: Conceptualization; methodology; resources; writing- original draft, review & editing; visualization, supervision; project administration; funding acquisition
Arpita Bhattacharya: Conceptualization; methodology; resources; writing- original draft, review & editing; visualization; data curation; investigation; formal analysis; project administration
Julie A. Kientz: Conceptualization; methodology; resources; writing- original draft, review & editing; supervision; project administration; funding acquisition
Sean A. Munson: Conceptualization; methodology; resources; writing- original draft, review & editing; visualization, supervision; project administration; funding acquisition
Ria Nagar: data curation; investigation; formal analysis; project administration
1 In prototyping, fidelity refers to the level of a prototype’s detailing and finish, with respect to both functionality and look and feel. Low fidelity prototypes facilitate understanding how a system might work and initial evaluations before investing in development of high-fidelity prototypes.
2 Allowable age range of 13- to 19-years old was the same across all phases of our work. The presented age ranges reflect the ages of enrolled participants.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
References
Agapie E , Avrahami D , & Marlow J (2016). Staying the Course: System-Driven Lapse Management for Supporting Behavior Change. Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems, 1072–1083. 10.1145/2858036.2858142
Alexopoulos GS , Raue PJ , Kiosses DN , Seirup JK , Banerjee S , & Arean PA (2015). Comparing engage with PST in late-life major depression: A preliminary report. The American Journal of Geriatric Psychiatry, 23 (5 ), 506–513.25081818
Allen K , Kern ML , Vella-Brodrick D , Hattie J , & Waters L (2018). What schools need to know about fostering school belonging: A meta-analysis. Educational Psychology Review, 30 (1 ), 1–34.
Baumer EP (2015). Reflective informatics: Conceptual dimensions for designing technologies of reflection. Proceedings of the 33rd Annual ACM Conference on Human Factors in Computing Systems, 585–594.
Bhattacharya A (2020). Designing Guided Asynchronous Remote Communities to Support Teen Mental Health [Unpublished Doctoral Dissertation].
Bhattacharya A , Nagar R , Jenness JL , Munson S , & Kientz JA (2021). Designing Asynchronous Remote Interventions to Support Teens with Depression Management using Behavioral Activation. JMIR Formative Research, 5 (7 ), e20969.34255665
Bhattacharya A , Liang C , Zeng EY , Shukla K , Wong MER , Munson SA , & Kientz JA (2019). Engaging Teenagers in Asynchronous Online Groups to Design for Stress Management. Proceedings of the 18th ACM International Conference on Interaction Design and Children, 26–37. 10.1145/3311927.3323140
Björling EA , & Rose E (2019). Participatory research principles in human-centered design: Engaging teens in the co-design of a social robot. Multimodal Technologies and Interaction, 3 (1 ), 8.
Chandra A , & Minkovitz CS (2006). Stigma starts early: Gender differences in teen willingness to use mental health services. Journal of Adolescent Health, 38 (6 ), 754–e1.
Clayborne ZM , Varin M , & Colman I (2019). Systematic review and meta-analysis: Adolescent depression and long-term psychosocial outcomes. Journal of the American Academy of Child & Adolescent Psychiatry, 58 (1 ), 72–79.30577941
Comer JS (2015). Introduction to the special series: Applying new technologies to extend the scope and accessibility of mental health care. Cognitive and Behavioral Practice, 22 (3 ), 253–257.
Davidson TM , Yuen EK , Felton JW , McCauley J , Gros KS , & Ruggiero KJ (2014). Feasibility assessment of a brief, web-based behavioral activation intervention for adolescents with depressed mood. The International Journal of Psychiatry in Medicine, 48 (1 ), 69–82.25354927
Dearden A , & Finlay J (2006). Pattern languages in HCI: A critical review. Human–Computer Interaction, 21 (1 ), 49–102.
Dimidjian S , Barrera M Jr , Martell C , Muñoz RF , Lewinsohn PM . The origins and current status of behavioral activation treatments for depression. Annual review of clinical psychology. 2011;7 :1–38.
Dopp AR , Parisi KE , Munson SA , & Lyon AR (2019). Integrating implementation and user-centred design strategies to enhance the impact of health services: Protocol from a concept mapping study. Health Research Policy and Systems, 17 (1 ), 1–11.30621706
Dopp AR , Parisi KE , Munson SA , & Lyon AR (2020). Aligning implementation and user-centered design strategies to enhance the impact of health services: Results from a concept mapping study. Implementation Science Communications, 1 (1 ), 1–13.32885176
Doherty G , Coyle D , & Matthews M (2010). Design and evaluation guidelines for mental health technologies. Interacting with computers, 22 (4 ), 243–252.
Eikey EV , & Reddy MC (2017). “It’s Definitely Been a Journey”: A Qualitative Study on How Women with Eating Disorders Use Weight Loss Apps. Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems, 642–654. 10.1145/3025453.3025591
Eiraldi R , Wolk CB , Locke J , & Beidas R (2015). Clearing hurdles: The challenges of implementation of mental health evidence-based practices in under-resourced schools. Advances in School Mental Health Promotion, 8 (3 ), 124–140.26336512
Epstein DA , Caraway M , Johnston C , Ping A , Fogarty J , & Munson SA (2016). Beyond abandonment to next steps: Understanding and designing for life after personal informatics tool use. Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems, 1109–1113.
Fincher S , Finlay J , Greene S , Jones L , Matchen P , Thomas J , & Molina PJ (2003). Perspectives on HCI patterns: Concepts and tools. CHI’03 Extended Abstracts on Human Factors in Computing Systems, 1044–1045.
Fixsen DL , Blase KA , Naoom SF , Van Dyke M , & Wallace F (2009). Implementation: The missing link between research and practice. NIRN Implementation Brief, 1 , 218–227.
Fleck R , & Fitzpatrick G (2010). Reflecting on reflection: Framing a design landscape. Proceedings of the 22nd Conference of the Computer-Human Interaction Special Interest Group of Australia on Computer-Human Interaction, 216–223.
Fritz T , Huang EM , Murphy GC , & Zimmermann T (2014). Persuasive technology in the real world: A study of long-term use of activity sensing devices for fitness. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, 487–496.
Galván A (2017). The neuroscience of adolescence. Cambridge University Press.
Graham AK , Wildes JE , Reddy M , Munson SA , Barr Taylor C , & Mohr DC (2019). User-centered design for technology-enabled services for eating disorders. International Journal of Eating Disorders, 52 (10 ), 1095–1107.
Grevet C , & Gilbert E (2015). Piggyback prototyping: Using existing, large-scale social computing systems to prototype new ones. Proceedings of the 33rd Annual ACM Conference on Human Factors in Computing Systems, 4047–4056.
Grist R , Croker A , Denne M , & Stallard P (2019). Technology delivered interventions for depression and anxiety in children and adolescents: A systematic review and meta-analysis. Clinical Child and Family Psychology Review, 22 (2 ), 147–171.30229343
Gulliver A , Griffiths KM , & Christensen H (2010). Perceived barriers and facilitators to mental health help-seeking in young people: A systematic review. BMC Psychiatry, 10 (1 ), 1–9.20055988
Hetrick SE , Robinson J , Burge E , Blandon R , Mobilio B , Rice SM , Simmons MB , Alvarez-Jimenez M , Goodrich S , & Davey CG (2018). Youth codesign of a mobile phone app to facilitate self-monitoring and management of mood symptoms in young people with major depression, suicidal ideation, and self-harm. JMIR Mental Health, 5 (1 ), e9.29362208
Hollis C , Falconer CJ , Martin JL , Whittington C , Stockton S , Glazebrook C , & Davies EB (2017). Annual Research Review: Digital health interventions for children and young people with mental health problems–a systematic and meta-review. Journal of Child Psychology and Psychiatry, 58 (4 ), 474–503.27943285
Holtzblatt K , & Beyer H (1997). Contextual design: Defining customer-centered systems. Elsevier.
Holtzblatt K , Wendell JB , & Wood S (2005). Rapid contextual design: A how-to guide to key techniques for user-centered design (Vol. 2005 ).
Huguet A , Rao S , McGrath PJ , A systematic review of cognitive behavioral therapy and behavioral activation apps for depression. PloS one. 2016;11 (5 ):e0154248.27135410
Kelley C , Lee B , & Wilcox L (2017). Self-Tracking for Mental Wellness: Understanding Expert Perspectives and Student Experiences. Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems, 629–641. 10.1145/3025453.3025750
Langley AK , Nadeem E , Kataoka SH , Stein BD , & Jaycox LH (2010). Evidence-based mental health programs in schools: Barriers and facilitators of successful implementation. School Mental Health, 2 (3 ), 105–113.20694034
Langley GJ , Moen RD , Nolan KM , Nolan TW , Norman CL , & Provost LP (2009). The improvement guide: A practical approach to enhancing organizational performance. John Wiley & Sons.
Lewis JR (2018). The system usability scale: Past, present, and future. International Journal of Human–Computer Interaction, 34 (7 ), 577–590.
Lyon AR , Brewer SK , & Areán PA (2020). Leveraging human-centered design to implement modern psychological science: Return on an early investment. American Psychologist, 75 (8 ), 1067.
Lyon AR , & Bruns EJ (2019). User-centered redesign of evidence-based psychosocial interventions to enhance implementation—Hospitable soil or better seeds? JAMA Psychiatry, 76 (1 ), 3–4.30427985
Lyon AR , Coifman J , Cook H , McRee E , Liu FF , Ludwig K , … & McCauley E (2021). The Cognitive Walkthrough for Implementation Strategies (CWIS): a pragmatic method for assessing implementation strategy usability. Implementation Science Communications, 2 (1 ), 1–16.33413699
Lyon AR , Dopp AR , Brewer SK , Kientz JA , & Munson SA (2020). Designing the future of children’s mental health services. Administration and Policy in Mental Health and Mental Health Services Research, 47 (5 ), 735–751.32253634
Lyon AR , Munson SA , Renn BN , Atkins DC , Pullmann MD , Friedman E , & Areán PA (2019). Use of human-centered design to improve implementation of evidence-based psychotherapies in low-resource communities: Protocol for studies applying a framework to assess usability. JMIR Research Protocols, 8 (10 ), e14990.31599736
Lyon AR , Pullmann MD , Jacobson J , Osterhage K , Al Achkar M , Renn BN , Munson SA , & Areán PA (2020). Assessing the Usability of Complex Psychosocial Interventions: The Intervention Usability Scale.
MacLeod H , Jelen B , Prabhakar A , Oehlberg L , Siek K , & Connelly K (2016). Asynchronous remote communities (arc) for researching distributed populations. 10th EAI International Conference on Pervasive Computing Technologies for Healthcare.
Maestre JF , MacLeod H , Connelly CL , Dunbar JC , Beck J , Siek KA , & Shih PC (2018). Defining Through Expansion: Conducting Asynchronous Remote Communities (ARC) Research with Stigmatized Groups. Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems, 1–13. 10.1145/3173574.3174131
McCauley E , Gudmundsen G , Schloredt K , Martell C , Rhew I , Hubley S , & Dimidjian S (2016). The adolescent behavioral activation program: Adapting behavioral activation as a treatment for depression in adolescence. Journal of Clinical Child & Adolescent Psychology, 45 (3 ), 291–304.25602170
McCauley E , Schloredt KA , Gudmundsen GR , Martell CR , & Dimidjian S (2016). Behavioral activation with adolescents: A clinician’s guide. Guilford Publications.
Mohr DC , Burns MN , Schueller SM , Clarke G , & Klinkman M (2013). Behavioral intervention technologies: Evidence review and recommendations for future research in mental health. General Hospital Psychiatry, 35 (4 ), 332–338.23664503
Mohr DC , Weingardt KR , Reddy M , & Schueller SM (2017). Three problems with current digital mental health research… And three things we can do about them. Psychiatric Services, 68 (5 ), 427–429.28412890
Mojtabai R , Olfson M , & Han B (2016). National trends in the prevalence and treatment of depression in adolescents and young adults. Pediatrics, 138 (6 ).
Munson SA (2017). Rethinking assumptions in the design of health and wellness tracking tools. Interactions, 25 (1 ), 62–65.
Murnane EL , Walker TG , Tench B , Voida S , & Snyder J (2018). Personal informatics in interpersonal contexts: Towards the design of technology that supports the social ecologies of long-term mental health management. Proceedings of the ACM on Human-Computer Interaction, 2 (CSCW ), 1–27.
Nagar R , & Kemp N (2020). BA Video. https://youtu.be/Le10_EyamDw
Newman BM , Lohman BJ , & Newman PR (2007). Peer group membership and a sense of belonging: Their relationship to adolescent behavior problems. Adolescence, 42 (166 ).
Norman D , & Draper S (1987). User centered system design: New perspectives on human-computer interaction. Journal Educational Computing Research, 3 (1 ), 129–134.
Nuttman-Shwartz O (2019). The moderating role of resilience resources and sense of belonging to the school among children and adolescents in continuous traumatic stress situations. The Journal of Early Adolescence, 39 (9 ), 1261–1285.
Olfson M , Gameroff MJ , Marcus SC , & Waslick BD (2003). Outpatient treatment of child and adolescent depression in the United States. Archives of General Psychiatry, 60 (12 ), 1236–1242.14662556
Olfson M , Mojtabai R , Sampson NA , Hwang I , Druss B , Wang PS , Wells KB , Pincus HA , & Kessler RC (2009). Dropout from outpatient mental health care in the United States. Psychiatric Services, 60 (7 ), 898–907.19564219
Percival J , & McGregor C (2016). An evaluation of understandability of patient journey models in mental health. JMIR human factors, 3 (2 ), e20.27471006
Pinfold V , Thornicroft G , Huxley P , & Farmer P (2005). Active ingredients in anti-stigma programmes in mental health. International Review of Psychiatry, 17 (2 ), 123–131.16194782
Richardson LP , McCauley E , Grossman DC , McCarty CA , Richards J , Russo JE , Rockhill C , & Katon W (2010). Evaluation of the Patient Health Questionnaire-9 Item for detecting major depression among adolescents. Pediatrics, 126 (6 ), 1117–1123. 10.1542/peds.2010-0852 21041282
Rideout V , & Robb M (2019). The Common Sense Census: Media Use by Tweens and Teens (2019). Common Sense Media. https://www.commonsensemedia.org/research/the-commonsense-census-media-use-by-tweens-and-teens-2019
Rohani DA , Quemada Lopategui A , Tuxen N , Faurholt-Jepsen M , Kessing LV , & Bardram JE (2020). MUBS: A personalized recommender system for behavioral activation in mental health. Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems, 1–13.
Rohani DA , Tuxen N , Lopategui AQ , Faurholt-Jepsen M , Kessing LV , & Bardram JE (2019). Personalizing mental health: A feasibility study of a mobile behavioral activation tool for depressed patients. Proceedings of the 13th EAI International Conference on Pervasive Computing Technologies for Healthcare, 282–291. 10.1145/3329189.3329214
Russell JA (1980). A circumplex model of affect. Journal of Personality and Social Psychology, 39 (6 ), 1161.
Substance Abuse and Mental Health Services Administration (SAMHSA). (2016). 2015 National survey on drug use and health. https://www.samhsa.gov/data
Suh H , Shahriaree N , Hekler EB , & Kientz JA (2016). Developing and Validating the User Burden Scale: A Tool for Assessing User Burden in Computing Systems. Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems, 3988–3999. 10.1145/2858036.2858448
Suh J , Williams S , Fann JR , Fogarty J , Bauer AM , & Hsieh G (2020). Parallel Journeys of Patients with Cancer and Depression: Challenges and Opportunities for Technology-Enabled Collaborative Care. Proceedings of the ACM on Human-Computer Interaction, 4 (CSCW1 ), 1–36.
Tatara N , \AArsand E , Skrøvseth SO , & Hartvigsen G (2013). Long-term engagement with a mobile self-management system for people with type 2 diabetes. JMIR MHealth and UHealth, 1 (1 ), e1.25100649
Thrul J , & Ramo DE (2017). Cessation strategies young adult smokers use after participating in a facebook intervention. Substance Use & Misuse, 52 (2 ), 259–264.27759475
Tindall L , Mikocka-Walus A , McMillan D , Wright B , Hewitt C , & Gascoyne S (2017). Is behavioural activation effective in the treatment of depression in young people? A systematic review and meta-analysis. Psychology and Psychotherapy: Theory, Research and Practice, 90 (4 ), 770–796. 10.1111/papt.12121
Torous J , Myrick KJ , Rauseo-Ricupero N , & Firth J (2020). Digital mental health and COVID-19: Using technology today to accelerate the curve on access and quality tomorrow. JMIR Mental Health, 7 (3 ), e18848.32213476
Torous J , & Wykes T (2020). Opportunities from the coronavirus disease 2019 pandemic for transforming psychiatric care with telehealth. JAMA Psychiatry, 77 (12 ), 1205–1206.32391857
Wade NG , Post BC , Cornish MA , Vogel DL , & Tucker JR (2011). Predictors of the change in self-stigma following a single session of group counseling. Journal of Counseling Psychology, 58 (2 ), 170.21319898
Wadley G , Lederman R , Gleeson J , & Alvarez-Jimenez M (2013). Participatory design of an online therapy for youth mental health. Proceedings of the 25th Australian Computer-Human Interaction Conference: Augmentation, Application, Innovation, Collaboration, 517–526.
Weiner BJ , Lewis CC , Stanick C , Powell BJ , Dorsey CN , Clary AS , Boynton MH , & Halko H (2017). Psychometric assessment of three newly developed implementation outcome measures. Implementation Science, 12 (1 ), 108.28851459
Williams NJ , & Beidas RS (2019). Annual research review: The state of implementation science in child psychology and psychiatry: A review and suggestions to advance the field. Journal of Child Psychology and Psychiatry, 60 (4 ), 430–450.30144077
Wu Y , Levis B , Riehm KE , Saadat N , Levis AW , Azar M , Rice DB , Boruff J , Cuijpers P , Gilbody S , & others. (2019). Equivalency of the diagnostic accuracy of the PHQ-8 and PHQ-9: A systematic review and individual participant data meta-analysis. Psychological Medicine.
Zomerdijk LG , & Voss CA (2010). Service design for experience-centric services. Journal of Service Research, 13 (1 ), 67–82.
|
PMC008xxxxxx/PMC8983012.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
9202995
36185
J Offender Rehabil
J Offender Rehabil
Journal of offender rehabilitation
1050-9674
1540-8558
35386231
8983012
10.1080/10509674.2022.2045528
NIHMS1792050
Article
A psychometric reevaluation of the TCU criminal thinking scales (CTS)
Sease Thomas B.
Joe George
Pankow Jennifer
Lehman Wayne E. K.
Knight Kevin
Institute of Behavioral Research, Texas Christian University, Fort Worth, Texas, USA
✉ CONTACT Thomas B. Sease [email protected] Institute of Behavioral Research, Texas Christian University, 3034 Sandage Avenue, Fort Worth, TX 76109, USA.
25 3 2022
2022
15 3 2022
05 4 2022
61 3 135147
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
In the United States, approximately 9 million people cycle in and out of jail and more than 600,000 people are released from prison each year. Unfortunately, the reentry process includes several barriers people must overcome (e.g., criminal thinking) to achieve adequate psychosocial functioning. As such, valid and reliable assessments that allow correctional staff to monitor clients’ progress in treatment and test program effectiveness are paramount to reducing this major public safety concern. The TCU Criminal Thinking Scales (CTS) are a widely used assessment of criminal thinking in correctional settings. This study reevaluated the psychometric properties of the TCU CTS using Item Response Theory. Results showed the TCU CTS had good internal reliability and each scale loaded onto one factor. Item level analysis revealed most items adequately fit the model, generally measuring moderate levels of criminal thinking. Furthermore, several TCU CTS scales were negatively correlated with motivation for treatment and psychosocial functioning.
assessment
criminal thinking
evaluation
criminal justice system
measurement
pmcIntroduction
In the United States, approximately 9 million people cycle in and out of jail and more than 600,000 people are released from prison each year (U.S. Department of Health & Human Services, 2021). Unfortunately, the reentry process includes contextual barriers justice-involved people must overcome, such as finding employment, housing, and transportation (see Bushway et al., 2007 for a full review). Exacerbating these concerns are maladaptive cognitive patterns or negative attitudes that interfere with psychosocial functioning. Together, difficulties associated with the reentry process place this population at an elevated risk for subsequent involvement in the criminal justice system. A longitudinal study conducted by the U.S. Department of Justice found that more than three-fourths (83%) of people released from state prisons recidivated within nine years of release (Alper et al., 2018). To address this public safety challenge, considerable effort has been devoted to providing justice-involved people with services that reduce their likelihood of criminal activity post-release.
The Risk-Need-Responsivity (RNR; Andrews & Bonta, 2010) model of risk assessment and treatment needs posits that services with the greatest potential to reduce recidivism are those that (1) match the intensity of care with the level of individual need, (2) target the client’s unique needs as the mechanism of change, and (3) provide support that maximizes a person’s likelihood of benefiting from the intervention. A fundamental component of this approach is mitigating criminogenic factors that contribute to a person’s risk for recidivism. This is performed under the assumption that attenuating maladaptive social, psychological, and personal patterns of behavior will reduce a person’s probability of reoffending. Noteworthy criminogenic factors amenable through intervention are antisocial personality patterns (e.g., impulsivity, aggressiveness, irritability) and pro-criminal attitudes (Bonta & Andrews, 2007). Antisocial personality patterns have been connected to criminal behavior in prospective research, and pro-criminal attitudes (e.g., Personal Irresponsibility, Justification) have been correlated with substance use among youth in the juvenile justice system (Dembo et al., 2007; Fridell et al., 2008). Similarly, decreases in criminal thinking have been associated with fewer disciplinary infractions and less prison misconduct in justice-involved populations (Folk et al., 2016; Walters, 2017).
Meta-analytic strategies have consistently demonstrated a connection between criminal thinking and criminal behavior (Walters, 2002, 2012). For example, criminal thinking predicted recidivism even while controlling for criminal history (Walters, 2012). Relatedly, criminal thinking was positively correlated with justice-involved individuals’ history of lifetime arrests and negatively correlated with recovery attitudes (Bartholomew et al., 2018). The literature has identified several status factors linked to criminal thinking, such as more education, a longer sentence length, and more time served (Mandracchia & Morgan, 2010). Justice-involved males tend to report higher levels of criminal thinking when compared to their female counterparts (Taxman et al., 2011). Studies examining the proximal outcomes of criminal thinking, in combination with those investigating personal factors related to criminal thinking, provide critically important information regarding intervention opportunities that have the potential to improve longer-term public safety and health outcomes.
An easy to administer and cost-effective measure of criminal thinking in justice-involved populations is the 36-item Criminal Thinking Scales (CTS; Knight et al., 2006). The TCU CTS was initially developed for the evaluation of cognitive-oriented substance use treatment programs. The TCU CTS includes the following six scales: Entitlement (EN), Justification (JU), Criminal Rationalization (CN) Personal Irresponsibility (PI), Power Orientation (PO) and Coldheartedness (CH). The administration takes less than 15 minutes, and the scoring of the scales is achieved by computing the average of items responses within each scale, minimizing the challenges often associated with the administration and scoring of screening instruments used with justice-involved populations. In this way, the TCU CTS offers correctional staff an easy, cost-effective, and easily interpretable tool that can be used to monitor client progress in treatment.
The psychometrics of the TCU CTS have been established using Classical Test Theory (e.g., Knight et al., 2006). In the original paper, all scales of the TCU CTS were found to have acceptable internal reliability scores (α = .68–.81) and scale structures. Subsequent investigations have supported the validity of TCU CTS, with criminal thinking being correlated with less engagement in substance use treatment, worse client functioning, and decreased adherence to treatment (Best et al., 2009; Dembo et al., 2007; Mitchell et al., 2013; Sana & Batool, 2017). Personal Irresponsibility and Justification were positively associated with crack cocaine use severity, and Personal Irresponsibility was associated with more cocaine use dependence and weekly drug use (Packer et al., 2009). Criminal thinking is not only associated with increased substance use severity but also decreased psychological well-being (Best et al., 2009; Butt et al., 2019).
Although the TCU CTS scale is a valid and reliable assessment of criminal thinking, advancements in scale development make it possible for researchers to assess the psychometric properties of a scale using item-level analyses (i.e., Item Response Theory; IRT); these contrast measurement-level analyses used by Classical Test Theory. This process allows for the assessment of each item’s validity within a construct and the removal of ill-fitting items without diminishing the overall validity of the assessment. These procedures provide researchers additional information about the scale, which can be used to refine assessments with greater detail. To this end, the present study focuses on a psychometric reevaluation of the TCU CTS using IRT.
Current study
This study examined the psychometric properties of each TCU CTS scale using IRT procedures. This was achieved using exploratory and confirmatory factor analysis, followed by an examination of internal reliability scores, item fit, and difficulty scores. Furthermore, this study evaluated the validity of the TCU CTS by examining its association with motivation for treatment (e.g., desire for help, problem recognition, treatment readiness) and psychosocial functioning (e.g., depression, anxiety, self-esteem). The authors expected that criminal thinking scores to be negatively associated with treatment motivation and self-esteem while being positively associated with measures of depression and anxiety.
Method
Participants and procedure
This study was a data reanalysis using secondary data collected from eight prison-based substance use disorder treatment programs that were part of the TCU Disease Risk Reduction (DRR) Project (see Lehman et al., 2015). Five male-only and three female-only units from two different states were included in the study. Treatment programs were modified therapeutic community programs wherein clients were required to receive a minimum of 20 h of programing each week. All new intakes completed a battery of forms as part of routine clinical practice which included the TCU Global Risk assessment Adult version (TCU A-RSKForm), the TCU Drug Screen II, TCU Client Evaluation of Self in Treatment (CEST) Forms, and the TCU Criminal Thinking Scales (CTS). All data were de-identified and analyzes were performed on responses provided at time-point one, making this study cross-sectional. All procedures associated with this study were approved by the authors’ Institutional Review Board.
The resultant sample consisted of 8,351 justice-involved males (67.1%) and females (32.9%) ranging in age from 17 to 71 (M = 34.69, SD = 9.78). Half of the sample reported their race as White (50.7%), followed by Black/African American (25.6%) or Other (23.7%). The majority of respondents identified as non-Hispanic (81.2%) and only a quarter (24.1%) were married. According to the TCU Drug Screen II (Knight et al., 2002), about two-thirds (63.1%) of the sample met the criteria for a severe substance use problem.
Measures
Criminal thinking
The 36-item TCU CTS was used to assess criminal thinking in the following six dimensions: Personal Irresponsibility, Coldheartedness, Criminal Rationalization, Power Orientation, Entitlement, and Justification. Using a 5-point Likert scale (1 = Strongly Disagree, 5 = Strongly Agree), participants rated items such as, “When people tell you what you do, you become aggressive,” “You are not to blame for everything you have done,” and “You feel that you are above the law” for Power orientation, Personal Irresponsibly, and Entitlement, respectively. In previous work, all scales on the TCU CTS have demonstrated strong psychometric properties (Knight et al., 2006).
Motivation for change
The TCU MOTForm (Simpson & Joe, 1993) assessed desire for help, problem recognition, and treatment readiness. Using a 5-point Likert scale (1 = Strongly Disagree, 5 = Strongly Agree), participants were asked to report on items about their motivation for substance use treatment. Sample items include, “Your drug use is a problem for you,” “You need help dealing with your drug use,” and “You want to be in drug treatment” for problem recognition, desire for help, and treatment readiness, respectively. Extant literature indicates these scales are valid and reliable in justice-involved populations (Pankow et al., 2012).
Psychological functioning
To measure psychological functioning, the self-esteem, depression, and anxiety scales on the TCU PSYForm were used. Respondents are asked to rate how much they agreed or disagreed with each item using a 5-point Likert scale (1 = Strongly Disagree, 5 = Strongly Agree). Example items include, “You have a lot to be proud of,” “You feel hopeless about the future,” and “You feel tense or keyed-up” for self-esteem, depression, and anxiety, respectively. The TCU PSYForm has demonstrated strong psychometric properties in justice-involved populations (Pankow et al., 2012).
Analytic plan
The reanalysis of the TCU CTS was performed using the concepts from IRT (see De Ayala, 2013 for a comprehensive review of IRT). First, descriptive statistics (M, SD) were computed for each TCU CTS scale (see Table 1). Next, each CTS scale was examined to ensure the assumptions of IRT were met (e.g., unidimensionality, monotonicity). The assumption of local independence was not tested because it was assumed that the correlation of the residuals between any two items would be uncorrelated after the trait being measured by the items was partitioned out. Finally, one random subsample of 400 participants was obtained from the full dataset to test for construct validity. All analyses listed herein were completed using Winsteps Version 4.4.6, SAS Version 9.4, and SPSS Version 25.0.
Item response theory (IRT)
Exploratory and confirmatory factor analysis was used to test the latent structure of each scale and evaluate the assumption of unidimensionality. Factors with an eigen value greater than one were retained (Kaiser, 1960) and internal reliability scores were computed. Confirmatory factor analysis was used to verify the structure of each CTS scale and test each model using indicators of model fit (e.g., χ2, GFI, SRMR, RMSEA). Then, item fit was evaluated using the point-measure correlations, mean-squared fit statistics (MNSQ), and difficulty scores provided in Winsteps. Items with a point-measure correlation greater than 0.40 were considered to have an acceptable fit (Kean et al., 2018). Likewise, items with a MNSQ infit and outfit score between 0.60 and 1.4 considered well-fitting (Kean et al., 2018). The former provides an estimate of model fit that is sensitive to unusual scores toward the middle the sampling distribution whereas the latter is sensitive to unusual scores at either tail of the distribution. Finally, items were evaluated using the difficulty score for each item within their respective scale. Difficulty scores represent the average ability levels (i.e., degree of criminal thinking) captured by a particular item. Higher difficulty scores suggest that people endorsing these items are more likely to have higher “criminal thinking” whereas people only endorsing items with low difficulty are more likely to have lower “criminal thinking.”
Validity
To evaluate the construct validity of the TCU CTS, correlation analysis determined the associations among the TCU CTS scales, motivation for treatment, and psychosocial functioning.
Results
Dimensionality
Using the full sample (N = 8,351), exploratory factor analysis with a varimax rotation tested the latent structure of each scale on the TCU CTS. Results showed each of the scales loaded best onto one factor, all achieving eigenvalues greater than 2 and explaining a large portion of the observed variance in their respective latent construct (≥39.49%; see Table 2 for factor loading). All scales showed acceptable internal reliability, with Cronbach alphas of 0.83, 0.76, 0.82, 0.64, 0.75, and 0.68 for EN, JU, PO, CH, CN, and PI, respectively.
Confirmatory factor analysis tested the model fit of each individual TCU CTS scale when being forced onto a single factor. Results showed that the minimum fit χ2 test was significant (ps < .001), which was expected considering the large sample size. Follow-up tests using other estimation methods of model fit (e.g., GFI, SRMR, RMSEA) showed that each TCU CTS scale achieved acceptable model fit when examined as a single factor (see Table 2).
Item fit
Results from the point-measure correlations showed that all items were positively correlated with each other, exceeding the 0.40 criteria for determining acceptable item fit. In addition, MNSQ infit and outfit statistics showed that all items except Item 15 adequately fit the model for their respective scale (see Table 3). The MNSQ outfit for Item 15 was 1.59. When examining difficulty values, the item with the highest difficulty score was Item 3 (i.e., “The real reason you are locked-up is because of your race”) and the item with the lowest difficulty score was Item 15 (i.e., “You like to be in control”). In general, results showed that the TCU CTS scales are primarily composed of items with moderate difficulty.
Validity
Using a random sample of 400 respondents, correlation analysis tested the validity of the TCU CTS. As shown in Table 4, most TCU CTS scales were associated in a positive direction. The exception to this trend was the CH scale, which was not associated with the PO, JU, or CN scales.
Most TCU CTS scales were negatively correlated with desire for help, problem recognition, and treatment readiness. That is, higher TCU CTS scores were related to lower motivation for treatment. All scales except CH were negatively associated with self-esteem and multiple scales were positively associated with anxiety and depression. These results indicate that CTS scores were related to lower psychological well-being in this sample of justice involved individuals.
Discussion
This study reevaluated the psychometric properties of the TCU CTS. Exploratory factor analysis supported a one-factor solution for each TCU CTS scale and confirmatory factor analysis suggested the proposed one-factor model for each scale appropriately fit the data. Reliability coefficients ranged from 0.64 to 0.83, indicating these scales can be used as reliable measures of criminal thinking. Next, items were evaluated within their respective scale using indicators of item fit. Results showed most items achieved an acceptable item fit score, as indicated by point-measure correlations, MNSQ infit scores, and MNSQ outfit scores. In evaluating difficulty scores, TCU CTS scales most commonly contained a high number of items with moderate difficulty. This trend across scales implies that the items on the TCU CTS are most adept in capturing information among people with moderate levels of criminal thinking.
This study also revealed that measurements of criminal thinking were negatively associated with desire for help, problem recognition, and treatment readiness—key indicators of motivation for treatment. Motivation for changing substance use behavior has been linked to improved treatment satisfaction and client retention (Joe et al., 1998; Sia et al., 2000). As such, criminal thinking could be a barrier for clients entering substance use treatment, potentially predicting clients’ responsiveness to treatment. The TCU CTS scales were also correlated with variables measuring psychosocial functioning (i.e., depression, anxiety, self-esteem). These findings converge with existing literature reporting a negative association between criminal thinking and psychological well-being (Best et al., 2009; Butt et al., 2019). Together, criminogenic cognitions are associated with decreased motivation for treatment and psychosocial functioning. These findings support the need for evidence-based services that ameliorate criminal thinking patterns.
Findings provide further psychometric support for the use of the TCU CTS as a valid and reliable assessment of criminogenic cognitions. In accordance with the TCU treatment model (Simpson, 2000), criminal thinking measures would ideally be used in two ways. First, for people entering substance use treatment, assessments of criminal thinking can be utilized in conjunction with other validated screeners to examine the client’s problem severity. As demonstrated here, people with high levels of criminal thinking may concurrently experience low motivation for change and psychosocial functioning. This information could be used to make treatment decisions intended to match the level of care with the amount of need in the individual. Second, the TCU CTS scales should be used to monitor client progress in treatment as a pre-post assessment survey. Applications of the TCU treatment model have demonstrated early recovery is associated with treatment adherence, which in turn improves treatment outcomes (Simpson & Joe, 2004). Early recovery in treatment could be measured using the TCU CTS wherein positive changes in criminal thinking would be expected to be linked to improved client outcomes.
While these findings support the utility of the TCU CTS, ensuing investigations could focus on improving these measures. Most items on the TCU CTS had moderate difficulty scores. Future iterations of the scale might consider including items intended capture low and high ability levels. This could include adding items that have a high probability of being selected by most respondents as well as items with a low probability of being highly endorsed. Future studies may also consider exploring the feasibility of creating a TCU CTS that has the capacity to be scored as a composite scale. While the individual scales demonstrated strong psychometric properties in this study, a composite scale could be of practical importance; providing correctional staff and treatment providers a single score for risk classification. Finally, the TCU CTS would benefit from a more comprehensive investigation of its validity. This study showed the TCU CTS scales were more highly correlated with psychosocial functioning variables (e.g., self-esteem, depression, anxiety) than motivation for change. As a result, the TCU CTS could be measuring aspects of criminal thinking that are more relevant to a person’s overall well-being than potential responsiveness to interventions for substance use. Succeeding investigations are needed to explore this possibility.
There are several limitations of the present study. This study focused on reevaluating the psychometrics of the TCU CTS using a sample of justice-involved individuals incarcerated at eight different correctional facilities. These data do not provide information about the correlates of criminal thinking for people involved in the criminal justice system in other ways (e.g., probation, parole). Such information is warranted considering one investigation found criminal thinking to be correlated with recidivism for people in jail, but not for people on probation (Caudy et al., 2015). Data used for this study was cross-sectional, measured at prison intake. Thus, associations between variables do not imply causal relationships. Lastly, this study only investigated the associations between the TCU CTS scales, motivation for change, and psychosocial functioning. This approach fails to account for potential mediators or moderators that may be influencing these relationships. Studies should continue to investigate mediators and moderators of criminal thinking and its distal outcomes to help elucidate how and when criminal thinking is most predictive of negative psychosocial outcomes.
In summation, a primary aim of the criminal justice system is to provide services that reduce clients’ risk for future criminal activity, involvement in the criminal justice system, and recidivism. Equally, these services aim to provide people involved in the justice system the skills needed to succeed as a member of society following reentry. Criminal thinking is a dynamic cognitive process malleable through an intervention that, if accurately measured, could serve as a crucial component of an individual’s treatment plan.
Table 1. Descriptive statistics for criminal thinking scores.
Mean SD 33rd percentile 67th percentile
Cold Heartedness 2.16 0.60 2.00 2.40
Personal Irresponsibility 2.03 0.61 1.83 2.17
Power Orientation 2.32 0.71 2.00 2.57
Justification 2.01 0.63 1.83 2.17
Criminal Rationalization 2.84 0.77 2.50 3.17
Entitlement 1.73 0.56 1.33 2.00
Note. Scores for each scale range from 1 (Strongly Disagree) to 5 (Strongly Agree). SD: standard deviation.
Table 2. Model of fit indices for each CTS subscale.
Scale χ 2 SRMR RMSEA GFI
Cold Heartedness <.001 0.02 0.05 0.99
Personal Irresponsibility <.001 0.02 0.03 0.99
Power Orientation <.001 0.03 0.07 0.98
Justification <.001 0.03 0.08 0.98
Criminal Rationalization <.001 0.01 0.02 0.99
Entitlement <.001 0.03 0.08 0.97
Note. SRMR: Standardized Root Mean Square Residual; RMSEA: Root Mean Square Error of Approximation; GFI: Goodness of Fit Index.
Table 3. Item fit statistics.
Item # Infit MNSQ Infit ZSTD Outfit MNSQ Outfit ZSTD PTMZ R corr. Difficulty scores
CH
1 1.08 4.35 1.08 4.05 0.62 −0.04
6 0.87 −7.97 0.87 −7.49 0.65 −0.25
12 1.07 4.38 1.17 9.61 0.62 −0.69
17 1.21 9.48 1.16 7.70 0.54 0.74
27 0.83 −9.40 0.81 −9.90 0.61 0.23
PI
2 1.11 6.68 1.22 9.90 0.59 −0.59
3 1.08 3.91 0.93 −3.05 0.48 1.28
21 1.06 3.39 1.19 9.87 0.55 −0.26
29 0.99 −0.74 1.08 4.35 0.60 −0.43
31 0.79 −9.90 0.79 −9.90 0.61 0.09
36 1.04 1.92 1.05 2.56 0.59 −0.09
PO
4 0.94 −3.63 0.96 −2.22 0.66 0.29
10 0.85 −8.98 .84 −9.35 0.67 0.44
13 1.09 5.42 1.17 9.14 0.62 −0.06
14 0.89 −7.02 0.86 −8.06 0.71 0.16
15 1.37 9.90 1.59 9.90 0.61 −1.23
20 0.82 −9.90 0.82 −9.90 0.69 0.29
28 1.08 4.46 1.08 4.49 0.66 0.10
JU
7 1.03 1.74 1.17 9.51 0.62 −0.50
11 1.03 1.67 1.17 9.53 0.65 −0.57
16 1.05 2.75 1.06 2.97 0.63 0.02
25 1.13 5.74 1.03 1.55 0.57 0.76
26 0.94 −2.92 0.93 −3.55 0.66 0.10
35 0.93 −3.79 0.89 −5.71 0.64 0.19
CN
5 1.12 7.87 1.12 7.54 0.63 0.16
8 0.84 −9.90 0.85 −9.90 0.70 −0.54
18 1.40 9.90 1.44 9.90 0.54 0.17
19 0.77 −9.90 0.82 −9.90 0.67 −0.34
30 1.02 1.50 0.98 −1.29 0.63 0.57
34 0.86 −9.90 0.90 −6.96 0.67 −0.02
EN
9 0.99 −0.26 1.09 3.74 0.69 −0.28
22 1.07 2.79 0.98 −1.02 0.69 0.55
23 1.05 1.83 0.92 −3.17 0.71 0.37
24 1.13 5.32 1.16 6.42 0.70 −0.27
32 1.07 3.18 1.27 9.90 0.68 −0.65
33 0.79 −8.99 0.70 −9.90 0.74 0.28
Note. MNSQ: mean-square; ZSTD: standardized weighted (infit) and unweighted (outfit) mean-squared fit statistics; PTMZ: point-measure correlation.
Table 4. Bivariate correlations.
CH PI PO JU CN EN AX SE DP DH TR
CH
PI .17*
PO .02 .49**
JU .05 .64** .69**
CN .01 .51** .44** .42**
EN .21** .70** .59** .73** .39**
AX −.12* .28** .43** .37** 0.31** .28**
SE .04 −.18** −.32** −.29** −.18** −.16** −49**
DP −.03 .35** .45** .39** .32** .29** .71** −.59**
DH −.23** −.26** −.02 .01 −.10* −.16** .20** −.28** .27**
TR −.21** −.31** −.11* −.04 −.22** −.17** .04 −.13** −.02 .72**
PR −.17* −.17* .05 .12* −.05 −.04 .25** −.32** .22** .84** .68**
Note. CH: Cold Heartedness; PI: Personal Irresponsibility; PO: Power Orientation; JU: Justification; CN: Criminal Rationalization; EN: Entitlement; AX: Anxiety; SE: Self-Esteem; DP: Depression; DH: Desire for Help; TR: Treatment Readiness; PR: Problem Recognition.
** p < .01,
* p < .05.
References
Alper M , Durose MR , & Markman J (2018). 2018 update on prisoner recidivism: A 9-year follow-up period (2005–2014). US Department of Justice, Office of Justice Programs, Bureau of Justice Statistics. https://babsim.com/wp-content/uploads/2019/04/2018-update-on-prisoner-recidivism.pdf
Andrews DA , & Bonta J (2010). Rehabilitating criminal justice policy and practice. Psychology, Public Policy, and Law, 16 (1 ), 39–55. 10.1037/a0018362
Bartholomew NR , Morgan RD , Mitchell SM , & Van Horn SA (2018). Criminal thinking, psychiatric symptoms, and recovery attitudes among community mental health patients: An examination of program placement. Criminal Justice and Behavior, 45 (2 ), 195–213. 10.1177/0093854817734007
Best D , Day E , Campbell A , Flynn PM , & Simpson DD (2009). Relationship between drug treatment engagement and criminal thinking style among drug-using offenders. European Addiction Research, 15 (2 ), 71–77. 10.1159/000189785 19142006
Bonta J , & Andrews DA (2007). Risk-need-responsivity model for offender assessment and rehabilitation. Rehabilitation, 6 (1 ), 1–22. https://www.securitepublique.gc.ca/cnt/rsrcs/pblctns/rsk-nd-rspnsvty/rsk-nd-rspnsvty-eng.pdf
Bushway SD , Stoll MA , & Weiman D (Eds.). (2007). Barriers to reentry? The labor market for released prisoners in post-industrial America. Russell Sage Foundation. https://books.google.com/books?hl=en&lr=&id=YOeFAwAAQBAJ&oi=fnd&pg=PR7&dq=Barriers+to+Reentry%3F+The+Labor+Market+for+Released+Prisoners+in+Post-Industrial+America&ots=SsnPF_jFTq&sig=4fVz_6DZ9L0OraS0ldxDf_y0mO0#v=onepage&q=Barriers%20to%20Reentry%3F%20The%20Labor%20Market%20for%20Released%20Prisoners%20in%20Post-Industrial%20America&f=false
Butt A , Abdi SK , Hamid A , Dogar FA , & Fatima J (2019). Criminal thinking, moral disengagement and psychological wellbeing in prisoners. Journal of Fatima Jinnah Medical University, 13 (2 ), 51–54. https://www.jfjmu.com/index.php/ojs/article/view/590
Caudy MS , Folk JB , Stuewig JB , Wooditch A , Martinez A , Maass S , Tangney JP , & Taxman FS (2015). Does substance misuse moderate the relationship between criminal thinking and recidivism? Journal of Criminal Justice, 43 (1 ), 12–19. 10.1016/j.jcrimjus.2014.11.001 25598559
De Ayala RJ (2013). The theory and practice of item response theory. Guilford Publications.
Dembo R , Turner CW , & Jainchill N (2007). An assessment of criminal thinking among incarcerated youths in three states. Criminal Justice and Behavior, 34 (9 ), 1157–1167. 10.1177/0093854807304348
Folk JB , Disabato DJ , Daylor JM , Tangney JP , Barboza S , Wilson JS , Bonieskie L , & Holwager J (2016). Effectiveness of a self-administered intervention for criminal thinking: Taking a chance on change. Psychological Services, 13 (3 ), 272–282. 10.1037/ser0000079 27243111
Fridell M , Hesse M , Jaeger MM , & Kühlhorn E (2008). Antisocial personality disorder as a predictor of criminal behavior in a longitudinal study of a cohort of abusers of several classes of drugs: Relation to type of substance and type of crime. Addictive Behaviors, 33 (6 ), 799–811. 10.1016/j.addbeh.2008.01.001 18258375
Joe GW , Simpson DD , & Broome KM (1998). Effects of readiness for drug abuse treatment on client retention and assessment of process. Addiction, 93 (8 ), 1177–1190. 10.1080/09652149835008 9813899
Kaiser HF (1960). The application of electronic computers to factor analysis. Educational and Psychological Measurement, 20 (1 ), 141–151. 10.1177/001316446002000116
Kean J , Bisson EF , Brodke DS , Biber J , & Gross PH (2018). An introduction to item response theory and Rasch analysis: Application using the eating assessment tool (EAT-10). Brain Impairment, 19 (1 ), 91–102. 10.1017/BrImp.2017.13 29606914
Knight K , Garner BR , Simpson DD , Morey JT , & Flynn PM (2006). An assessment for criminal thinking. Crime & Delinquency, 52 (1 ), 159–177. 10.1177/0011128705281749
Knight K , Simpson DD , & Morey JT (2002). An evaluation of the TCU drug screen. National Institute of Justice, Office of Justice Programs. U.S. Department of Justice. https://www.ojp.gov/pdffiles1/nij/grants/196682.pdf
Lehman WE , Rowan GA , Greener JM , Joe GW , Yang Y , & Knight K (2015). Evaluation of WaySafe: A disease-risk reduction curriculum for substance-abusing offenders. Journal of Substance Abuse Treatment, 58 , 25–32. 10.1016/j.jsat.2015.05.007 26059002
Mandracchia JT , & Morgan RD (2010). The relationship between status variables and criminal thinking in an offender population. Psychological Services, 7 (1 ), 27–33. 10.1037/a0016194
Mitchell D , Tafrate RC , Hogan T , & Olver ME (2013). An exploration of the association between criminal thinking and community program attrition. Journal of Criminal Justice, 41 (2 ), 81–89. 10.1016/j.jcrimjus.2012.09.003
Packer G , Best D , Day E , & Wood K (2009). Criminal thinking and self-control among drug users in court mandated treatment. Criminology & Criminal Justice, 9 (1 ), 93–110. 10.1177/1748895808099182
Pankow J , Simpson DD , Joe GW , Rowan-Szal GA , Knight K , & Meason P (2012). Examining concurrent validity and predictive utility for the Addiction Severity Index and Texas Christian University (TCU) short forms. Journal of Offender Rehabilitation, 51 (1–2 ), 78–95. 10.1080/10509674.2012.633021 23087588
Sana F , & Batool I (2017). Development and validation of an indigenous criminal thinking scale. Pakistan Journal of Psychological Research, 32 (1 ), 117–139. http://www.pjprnip.edu.pk/index.php/pjpr/article/view/605/523
Sia TL , Dansereau DF , & Czuchry ML (2000). Treatment readiness training and probationers’ evaluation of substance abuse treatment in a criminal justice setting. Journal of Substance Abuse Treatment, 19 (4 ), 459–467. 10.1016/S0740-5472(00)00139-2 11166510
Simpson DD , (2000). Research summary: Focus on treatment process and outcomes. TCU Model of treatment process and outcomes Institute of Behavioral Research, Texas Christian University. https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.80.3414&rep=rep1&type=pdf
Simpson DD , & Joe GW (1993). Motivation as a predictor of early dropout from drug abuse treatment. Psychotherapy: Theory, Research, Practice, Training, 30 (2 ), 357–368. 10.1037/0033-3204.30.2.357
Simpson DD , & Joe GW (2004). A longitudinal evaluation of treatment engagement and recovery stages. Journal of Substance Abuse Treatment, 27 (2 ), 89–97. 10.1016/j.jsat.2004.03.001 15450643
Taxman FS , Rhodes AG , & Dumenci L (2011). Construct and predictive validity of criminal thinking scales. Criminal Justice and Behavior, 38 (2 ), 174–187. 10.1177/0093854810389550
U.S. Department of Health & Human Services. (2021). Incarceration & reentry. Office of the Assistant Secretary for Planning and Evaluation. https://aspe.hhs.gov/incarceration-reentry
Walters GD (2002). The Psychological Inventory of Criminal Thinking Styles (PICTS): A review and meta-analysis. Assessment, 9 (3 ), 278–291. 10.1177/1073191102009003007 12216785
Walters GD (2012). Criminal thinking and recidivism: Meta-analytic evidence on the predictive and incremental validity of the Psychological Inventory of Criminal Thinking Styles (PICTS). Aggression and Violent Behavior, 17 (3 ), 272–278. 10.1016/j.avb.2012.02.010
Walters GD (2017). Effect of a brief cognitive behavioural intervention on criminal thinking and prison misconduct in male inmates: Variable-oriented and person-oriented analyses. Criminal Behaviour and Mental Health, 27 (5 ), 457–469. 10.1002/cbm.2028 28589632
|
PMC008xxxxxx/PMC8983016.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
101686921
45447
J Biosci Med (Irvine)
J Biosci Med (Irvine)
Journal of biosciences and medicines
2327-5081
2327-509X
35386489
8983016
10.4236/jbm.2022.103019
NIHMS1792265
Article
Evaluation of Insulin Infusion Rates for the Treatment of Diabetic Ketoacidosis in the Emergency Department
Bass Megan E. 1
Paavola Nicole 2
Kiser Tyree H. 1
Wright Garth 1
Jacknin Gabrielle 2*
1 Department of Clinical Pharmacy, Skaggs School of Pharmacy and Pharmaceutical Sciences, University of Colorado, Aurora, CO, USA
2 Department of Pharmacy, UCHealth, University of Colorado Hospital, Aurora, CO, USA
* [email protected]
26 3 2022
3 2022
17 3 2022
05 4 2022
10 3 203211
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Introduction:
There is minimal literature to support the appropriate dosing for the initiation of IV regular insulin therapy in DKA patients. A 0.1 unit/kg bolus followed by 0.1 units/kg/hour or 0.14 units/kg/hour is commonly utilized and recommended in guidelines.
Objective:
We sought to assess clinical and safety outcomes associated with various insulin infusion starting doses in patients diagnosed with DKA in the emergency department in an effort to help guide prescribing.
Methods:
A retrospective cohort study was conducted within an academic emergency department and included patients who received continuous infusion regular insulin with an ICD-10 code for DKA between January 2016 and January 2019. A predictive regression model was applied to test if predefined lab values influenced the starting insulin infusion rates. Clinical and safety outcomes were evaluated by starting insulin infusion rate. Data was analyzed based on starting insulin infusion rates.
Results:
347 patients met inclusion criteria with 92 (26.5%) patients receiving <0.07 units/kg/hr, 123 (35.4%) patients receiving 0.07 to 0.099 units/kg/hr, 123 (35.4%) patients receiving 0.10 to 0.139 units/kg/hr, and 9 (2.6%) patients receiving ≥0.14 units/kg/hr. After adjusting for baseline labs, glucose was the only significant predictor of the initial infusion rate (p < 0.001). For every 100 mg/dL increase in the baseline glucose value, the initial infusion rate increased by 0.005 units/kg/hr. There was no difference between insulin starting infusion rates and length of stay, rates of hypoglycemia, hypokalemia, or dysrhythmias.
Conclusion:
Glucose levels significantly influenced the insulin starting infusion rate, with no identified differences in adverse effects or clinical outcomes.
Diabetic Ketoacidosis
Regular Insulin
Intravenous Insulin
Dose
Infusion Rates
pmc1. Introduction
Positive therapeutic responses have previously been established with the use of intravenous (IV) insulin therapy in adult patients with diabetic ketoacidosis (DKA) [1] [2] [3]. There is minimal literature to support the appropriate dosing for the initiation of IV insulin therapy in these patients. The role of insulin in the inhibition of ongoing lipolysis, suppression of further glucose production from the liver, and providing insulin for adequate glucose uptake into cells is widely understood and accepted. What is yet to be established is the impact of the starting insulin infusion rate that achieves all of the above metabolic functions while also down trending the glucose in a manner that is conducive to optimal time to gap closure and mitigation of risk associated with the infusion. Safety concerns, specifically cerebral edema and the rate of serum glucose reduction, have been identified and are associated with bolus doses of insulin prior to starting the infusion [4] [5]. A previous small prospective cohort study concluded that a bolus dose of insulin is not required if an adequate dose of regular insulin, defined as 0.14 units/kg/hour, is administered as a continuous infusion. Utilizing this insulin dosage strategy resulted in no difference in time to pH, bicarbonate, and glucose control. However, patients who received a starting dose of 0.07 units/kg/hour with no bolus were identified as not adequate to get the same response to insulin therapy [2]. Similarly, Goyal et al. also concluded that a bolus insulin dose was not associated with significant benefit in adult patients with DKA, and showed increased rates of hypoglycemic episodes [6]. While the data to support no bolus initiation in these patients has become a more common practice, the appropriate starting infusion rate without a bolus remains variable and controversial even in various national guidelines [4] [5] [7] [8] [9] [10].
Our large academic medical center emergency department has a provider pathway that reflects an initial starting infusion rate of IV regular insulin at 0.14 units/kg/hour with no associated bolus. Despite this dose embedded in the pathway, the dose that is chosen to start patients’ on varies tremendously based on patient specific factors as well as pharmacist and provider specific practices. Some patient specific factors commonly mentioned as being evaluated include baseline lab values to help classify DKA severity (including pH, bicarbonate, anion gap, and glucose), lab values associated with safety (potassium), diabetes mellitus type (type I, II, or other) and known insulin history, including being insulin naive or insulin resistant. There is a wide range of providers and pharmacists that work within the emergency department which naturally lends itself to variability in these practices.
We aimed to evaluate patients presenting to a large, academic emergency department for treatment of DKA receiving IV regular insulin via continuous infusion. The goal of this study was to assess descriptive characteristics, clinical outcomes, and safety of various starting insulin infusion rates (units/kg/hour) for the treatment of DKA within the emergency department. A predictive regression model was applied to test if predefined lab values influenced the starting insulin infusion rates.
2. Methods
A retrospective cohort study utilizing electronic medical records was conducted from January 2016 through January 2019 within a large, academic medical center’s emergency department. Our research was approved by the Colorado Multiple Institutional Review Board (COMIRB #19–2060). Health Data Compass was utilized to pull medication, lab, and encounter data from the electronic health record for each patient meeting defined inclusion criteria. Health Data Compass is a local, enterprise health data warehouse that integrates patient clinical data from the electronic health record [11]. Inclusion criteria were defined as patients 1) between 18 and 89 years old; 2) administration of continuous infusion regular insulin with or without a bolus within the UCH emergency department; and 3) an ICD-10 diagnosis code for DKA during the same encounter (Table 1).
Data was analyzed based on starting insulin infusion rates < 0.07 units/kg/hr, 0.07 – 0.099 units/kg/hr, 0.1 – 0.139 units/kg/hr, and ≥0.14 units/kg/hr. DKA severity was defined as mild, moderate, or severe to help categorize the population in the study (Table 2). A predictive regression model was applied to test if initial lab values influenced the starting insulin infusion rates. The regression model accounted for baseline lab values of glucose, potassium, anion gap, pH, and bicarbonate. These covariates were analyzed for multicollinearity. Other analyzed outcomes included duration of insulin infusion and mean ICU and hospital length of stay. In order to assess metabolic outcomes, time to goal values were assessed including time to blood glucose ≤ 250 mg/dL, pH ≥ 7.3, bicarbonate ≥ 15 mmol/L, and anion gap ≤ 12 mmol/L. Hypoglycemia, dysrhythmias, and hypokalemia were defined by documented ICD-10 codes within 0 to 48 hours of starting insulin therapy.
Descriptive analyses were completed for baseline population characteristics. The Chi-squared test or Fisher’s Exact Test were used to analyze categorical variables and student’s t-test or the Kruskal Wallis test for continuous data. The analysis for this project was generated using SAS software, Version 9.4 of the SAS System for Windows. Copyright © 2016 SAS Institute Inc. SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc., Cary, NC, USA.
3. Results
A total of 347 patients met inclusion criteria with 92 (26.5%) patients in the <0.07 units/kg/hour cohort, 123 (35.4%) patients in the 0.07 to 0.099 units/kg/hour cohort, 123 (35.4%) patients in the 0.10 to 0.139 units/kg/hour cohort, and 9 (2.6%) patients in the ≥0.14 units/kg/hour cohort. Patient demographics are shown in Table 3. Of note, patients in the <0.07 units/kg/hour cohort had a statistically higher mean weight compared to other starting infusion rate cohorts and the mean infusion rate in units/kg/hour was statistically different between cohorts as expected. Although mean infusion duration was not statistically different between cohorts, there were 42 patients with insulin infusion discontinuation in 8 hours or less, with a mean pH of 7.15. Table 4 shows a significant difference was found in the distribution of patients by starting insulin infusion rate and DKA severity (p < 0.001).
After adjusting for other baseline lab values, glucose was the only significant predictor of the initial insulin infusion rate (p < 0.05). For every 100 mg/dL increase in the baseline glucose value, the initial infusion rate increased by 0.005 units/kg/hr. A higher anion gap was correlated with a higher initial infusion rate and a higher bicarbonate and pH were correlated with a lower initial infusion rate, though not statistically significant. No multicollinearity was found among covariates. For all values associated with the linear regression, see Table 5. Time to goal glucose less than or equal to 250 mg/dL was statistically different between starting infusion rate cohorts (p = 0.006), with no significant difference identified for time to bicarbonate, pH, or anion gap goals (Figure 1). There was no significant difference in ICU length of stay or total hospital length of stay between starting infusion rate cohorts (p = 0.095 and 0.06 respectively, Table 6). Safety outcomes of dysrhythmia, hypoglycemia, and hypokalemia were also not different between starting infusion rates (Table 6).
4. Discussion
Insulin administration, in addition to fluid and electrolyte management, continue to be the mainstay of treatment for DKA, and emergency departments have continued to develop protocols in line with national guidelines [4] [12]. Despite the large number of annual emergency department visits for DKA management, there is minimal literature to support the wide range of dosing strategies used to initiate continuous infusion regular insulin therapy. While some protocols call for a standard starting infusion rate for all patients regardless of individual patient characteristics, the current process followed within our site facilitated the ability to study ordering trends occurring with interdisciplinary collaboration between the emergency department providers and pharmacists to determine starting insulin infusion rates [13]. To our knowledge, this type of model has never been assessed in this setting for DKA management.
This retrospective cohort study showed similar trends to previous literature, with no statistically significant difference between clinical or safety outcomes associated with starting IV insulin infusion rate [1] [3] [8] [9] [10]. Although time to glucose goal less than or equal to 250 mg/dL was statistically different, this is not unexpected as the baseline glucose value was statistically different between groups. The linear regression model supports that baseline glucose level is the only significant predictor for insulin starting infusion rate in our interdisciplinary collaboration model of ordering. Only nine patients over several years were initiated on the starting insulin infusion rate of 0.14 units/kg/hour. This indicates that a change to our emergency department DKA pathway is indicated to better represent the majority of insulin infusion ordering practices. This seems particularly relevant as there are minimal safety or efficacy data available for this cohort of higher starting insulin infusion rates. Another interesting finding from this study was the lack of significant difference in mean insulin infusion duration between starting infusion rate cohorts. Although our current protocol does not support early administration of long acting insulin to help facilitate transition off of infusion to subcutaneous therapy, this is a future consideration for our pathway and population based on newer literature showing potential benefit, particularly in mild-moderate DKA patients [10] [14] [15] [16]. As the starting insulin infusion rate does not appear to impact duration of infusion, it would be interesting to assess the addition of long acting insulin to this outcome in a future pathway. Lastly, the 42 patients who had their insulin infusions discontinued within 8 hours had a clear pH association of ≥7.15 with this early discontinuation. This finding has directly led to changes in our practice for the management of patients with mild to moderate DKA, partially defined as having a pH ≥ 7.15, being treated in our observation unit utilizing a subcutaneous insulin regimen.
5. Limitations
This study had several limitations. First, it was a single site, retrospective analysis with data obtained from a large database warehouse, and as a result could be impacted by missing data and selection bias. Additionally, with only a small cohort of patients in the initial infusion rate cohort of ≥0.14 units/kg/hour, our results and data analysis for this cohort are unable to provide any sound clinical conclusions about this cohort. Lastly, we calculated time to goal lab values based on previous studies and clinically relevant endpoints. We analyzed our endpoints based on the initial and final lab values from the patients’ stay. While these are important to being able to assess disease severity and time to resolution, the progression of disease and clinical implications of the infusion rates may benefit with analysis of the changes between these values as well.
6. Conclusion
There is a wide range of starting insulin infusion rates utilized in patients with DKA. Glucose levels significantly influenced the insulin starting infusion rate, with no identified differences in adverse effects or clinical outcomes between starting infusion rate cohorts.
Funding
This study was supported by the Colorado Clinical and Translational Sciences Institute (CCTSI). The CCTSI is supported in part by Colorado CTSA Grant UL1TR002535 from NCATS/NIH.
Figure 1. Median time to lab outcome in minutes based on starting insulin drip rate.
Table 1. Inclusion diagnoses by defined ICD-10 code.
Code Description
E13.11 Other specified diabetes mellitus with ketoacidosis with coma
E10.10 Type I (juvenile type) diabetes mellitus with ketoacidosis, uncontrolled
E13.10 Type II or unspecified type diabetes mellitus with ketoacidosis, uncontrolled
E08.1 Diabetes mellitus due to underlying condition with ketoacidosis
E08.11 Diabetes mellitus due to underlying condition with ketoacidosis with coma
E10.11 Type 1 diabetes mellitus with ketoacidosis with coma
Table 2. Diabetic ketoacidosis severity definitions (baseline).
DKA Category Blood Glucose (mg/dL) pH Bicarbonate
Mild >250 7.25 – 7.30 15 – 18
Moderate >250 7.00 – 7.24 10 – (<15)
Severe >250 <7.00 <10
Table 3. Demographics.
Characteristic <0.07 units/kg/hr (n = 92) 0.07 – 0.099 units/kg/hr (n = 123) 0.10 – 0.139 units/kg/hr (n = 123) ≥0.14 units/kg/hr (n = 9) P-Value
Sex, N (%)
Male 51 (55.4%) 70 (56.9%) 65 (53.3%) 5 (55.6%) 0.864
Female 41 (44.6%) 53 (43.1%) 57 (46.7%) 4 (44.4%)
Mean Weight, kg 90.7 ± 31.9 75.4 ± 20 68.4 ± 16.7 69.1 ± 9.2 <0.001
Mean infusion rate, units/kg/hr 0.04 ± 0.03 0.09 ± 0.02 0.11 ± 0.02 0.13 ± 0.03 <0.001
Mean infusion duration, hours 38.2 ± 43.2 48.7 ± 56.5 48.9 ± 46 41.8 ± 32.2 0.153
Table 4. Diabetic ketoacidosis baseline severity by starting insulin drip rate.
Starting Insulin Drip Rate by DKA Category Mild DKA (n = 102) Moderate DKA (n = 121) Severe DKA (n = 115) P-value
Number of patients, n (%)
<0.07 units/kg/hr 38 (45.2%) 34 (40.5%) 12 (14.3%)
0.07 – 0.099 units/kg/hr 39 (31.7%) 41 (33.3%) 43 (40%) <0.001
0.10 – 0.139 units/kg/hr 23 (18.9%) 42 (34.4%) 57 (46.7%)
≥0.14 units/kg/hr 2 (22.2%) 4 (44.5%) 3 (33.3%)
Table 5. Linear regression model for baseline labs and starting insulin infusion rate.
Variable Parameter Estimate Standard Error P-value
Intercept 0.28660 0.16317 0.08
Anion Gap 0.00014 0.00050 0.77
Glucose 0.00005 0.00001 <0.0001
Bicarbonate −0.00096 0.00062 0.12
Potassium −0.00033 0.00260 0.90
pH −0.03031 0.02295 0.19
Table 6. Clinical and safety outcomes by starting insulin infusion rate.
Characteristic <0.07 units/kg/hr (n = 92) 0.07 – 0.099 units/kg/hr (n = 123) 0.10 – 0.139 units/kg/hr (n = 123) ≥0.14 units/kg/hr (n = 9) P-Value
Total hospital Length of Stay, Days 1.18 ± 0.38 1.09 ± 0.29 1.07 ± 0.26 1 ± 0 0.06
ICU Length of Stay, Minutes 202.9 ± 489.3 347.7 ± 591.8 393.4 ± 669.5 415.9 ± 420.5 0.095
Dysrhythmia, n (%) 12 (14%) 22 (17.9%) 20 (16.3%) 3 (33.3%) 0.42
Hypoglycemia, n (%) 5 (5.4%) 10 (8.1%) 18 (14.6%) 1 (11.1%) 0.13
Hypokalemia, n (%) 30 (32.6%) 25 (20.3%) 28 (22.8%) 0 (0%)
Conflicts of Interest
The authors report no conflicts of interest related to this work.
References
[1] Fisher JN , Shahshahani MN and Kitabchi AE (1977) Diabetic Ketoacidosis: Low-Dose Insulin Therapy by Various Routes. New England Journal of Medicine, 297 , 238–241. 10.1056/NEJM197708042970502
[2] Kitabchi AE , Murphy MB , Spencer J , Matteri R and Karas J (2008) Is a Priming Dose of Insulin Necessary in a Low-Dose Insulin Protocol for the Treatment of Diabetic Ketoacidosis? Diabetes Care, 31 , 2081–2085. 10.2337/dc08-0509 18694978
[3] Kitabchi AE , Umpierrez GE , Fisher JN , Murphy MB and Stentz FB (2008) Thirty Years of Personal Experience in Hyperglycemic Crises: Diabetic Ketoacidosis and Hyperglycemic Hyperosmolar State. Journal of Clinical Endocrinology & Metabolism, 93 , 1541–1552. 10.1210/jc.2007-2577 18270259
[4] Kitabchi AE , Umpierrez GE , Miles JM and Fisher JN (2009) Hyperglycemic Crises in Adult Patients with Diabetes. Diabetes Care, 32 , 1335–1343. 10.2337/dc09-9032 19564476
[5] Kitabchi AE , Umpierrez GE , Murphy MB , Barrett EJ , Kreisberg RA , Malone JI and Wall BM (2004) American Diabetes A. Hyperglycemic Crises in Diabetes. Diabetes Care, 27 , S94–S102. 10.2337/diacare.27.2007.S94 14693938
[6] Goyal N , Miller JB , Sankey SS and Mossallam U (2010) Utility of Initial Bolus Insulin in the Treatment of Diabetic Ketoacidosis. Journal of Emergency Medicine, 38 , 422–427. 10.1016/j.jemermed.2007.11.033
[7] Kitabchi AE , Umpierrez GE , Murphy MB , Barrett EJ , Kreisberg RA , Malone JI and Wall BM (2001) Management of Hyperglycemic Crises in Patients with Diabetes. Diabetes Care, 24 , 131–153. 10.2337/diacare.24.1.131 11194218
[8] Dhatariya KK , Nunney I , Higgins K , Sampson MJ and Iceton G (2016) National Survey of the Management of Diabetic Ketoacidosis (DKA) in the UK in 2014. Diabetic Medicine, 33 , 252–260. 10.1111/dme.12875 26286235
[9] Dhatariya KK and Vellanki P (2017) Treatment of Diabetic Ketoacidosis (DKA)/Hyperglycemic Hyperosmolar State (HHS): Novel Advances in the Management of Hyperglycemic Crises (UK versus USA). Current Diabetes Reports, 17 , Article No. 33. 10.1007/s11892-017-0857-4 28364357
[10] Doshi P , Potter AJ , De Los Santos D , Banuelos R , Darger BF and Chatham-pally Y (2015) Prospective Randomized Trial of Insulin Glargine in Acute Management of Diabetic Ketoacidosis in the Emergency Department: A Pilot Study. Academic Emergency Medicine, 22 , 657–662. 10.1111/acem.12673 26013711
[11] Bass ME , Kiser TH , Page RL 2nd , McIlvennan CK , Allen LA , Wright G and Shakowski C (2021) Comparative Effectiveness of Direct Oral Anticoagulants and Warfarin for the Treatment of Left Ventricular Thrombus. Journal of Thrombosis and Thrombolysis, 52 , 517–522. 10.1007/s11239-020-02371-6 33420896
[12] Dhatariya K (2021) The Management of Diabetic Ketoacidosis in Adults. Joint British Diabetes Societies for Inpatient Care. ABCD (Diabetes Care) Ltd., Miria House, Knowle, Solihull, West Midlands.
[13] Gosmanov AR , Gosmanova EO and Kitabchi AE (2000) Hyperglycemic Crises: Diabetic Ketoacidosis and Hyperglycemic Hyperosmolar State. In: Feingold KR , Anawalt B , Boyce A , Chrousos G , de Herder WW , Dhatariya K , , Eds., Endotext, MDText.com, South Dartmouth (MA).
[14] Hsia E , Seggelke S , Gibbs J , Hawkins RM , Cohlmia E , Rasouli N , Wang C , Kam I and Draznin B (2012) Subcutaneous Administration of Glargine to Diabetic Patients Receiving Insulin Infusion Prevents Rebound Hyperglycemia. The Journal of Clinical Endocrinology & Metabolism, 97 , 3132–3137. 10.1210/jc.2012-1244 22685233
[15] Houshyar J , Bahrami A and Aliasgarzadeh A (2015) Effectiveness of Insulin Glargine on Recovery of Patients with Diabetic Ketoacidosis: A Randomized Controlled Trial. Journal of Clinical and Diagnostic Research, 9 , OC01–OC05. 10.7860/JCDR/2015/12005.5883
[16] Rappaport SH , Endicott JA , Gilbert MP , Farkas JD , Clouser RD and McMillian WD (2019) A Retrospective Study of Early vs Delayed Home Dose Basal Insulin in the Acute Management of Diabetic Ketoacidosis. Journal of the Endocrine Society, 3 , 1079–1086. 10.1210/js.2018-00400 31069278
|
PMC008xxxxxx/PMC8983017.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
101523164
37432
Arch Appl Mech
Arch Appl Mech
Archive of applied mechanics = Ingenieur-Archiv
0939-1533
1432-0681
35386426
8983017
10.1007/s00419-021-01985-3
NIHMS1788626
Article
A multiphasic model for determination of water and solute transport across the arterial wall: effects of elastic fiber defects
Guang Young Department of Biomedical Engineering, Washington University, St. Louis, MO, USA
Cocciolone Austin J. Department of Biomedical Engineering, Washington University, St. Louis, MO, USA
Crandall Christie L. Department of Mechanical Engineering and Materials Science, Washington University, St. Louis, MO, USA
Johnston Benjamin B. Department of Biomedical Engineering, Washington University, St. Louis, MO, USA
Setton Lori A. Department of Biomedical Engineering, Washington University, St. Louis, MO, USA
Wagenseil Jessica E. http://orcid.org/0000-0001-7972-448X
Department of Mechanical Engineering and Materials Science, Washington University, St. Louis, MO, USA
[email protected]
14 3 2022
2 2022
03 6 2021
05 4 2022
92 2 447459
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Transport of solute across the arterial wall is a process driven by both convection and diffusion. In disease, the elastic fibers in the arterial wall are disrupted and lead to altered fluid and mass transport kinetics. A computational mixture model was used to numerically match previously published data of fluid and solute permeation experiments in groups of mouse arteries with genetic (knockout of fibulin-5) or chemical (treatment with elastase) disruption of elastic fibers. A biphasic model of fluid permeation indicated the governing property to be the hydraulic permeability, which was estimated to be 1.52×10–9, 1.01×10–8, and 1.07×10–8 mm4/μN.s for control, knockout, and elastase groups, respectively. A multiphasic model incorporating solute transport was used to estimate effective diffusivities that were dependent on molecular weight, consistent with expected transport behaviors in multiphasic biological tissues. The effective diffusivity for the 4 kDA FITC-dextran solute, but not the 70 or 150 kDa FITC-dextran solutes, was dependent on elastic fiber structure, with increasing values from control to knockout to elastase groups, suggesting that elastic fiber disruption affects transport of lower molecular weight solutes. The model used here sets the groundwork for future work investigating transport through the arterial wall.
Biphasic
Febio
Permeation
Dextran
Mouse model
Elastin
Elastase
Finite element
pmc1 Introduction
The elastic fiber network is a major extracellular matrix component of the large conduit arteries. Composed of cross-linked elastic fibers made up of elastin and microfibrils, the elastic fiber network is organized in concentric layers called elastic laminae [1]. The internal elastic lamina separates the single layer of endothelial cells at the arterial lumen from the smooth muscle cells in the medial layer. Additional elastic laminae alternate with layers of smooth muscle cells, associated collagen fibers, and interconnecting elastic fibers throughout the medial layer of the arterial wall [2].
Defects in the elastic fiber network are common to many cardiovascular diseases but are a direct factor in the onset, progression, and rupture of aneurysms. Aneurysms are defined as a permanent localized dilation of the vascular diameter at least 50% greater than the normal size and may dissect or rupture leading to severe cardiovascular complications or death [3]. Elastic fiber defects in aneurysmal disease can be caused by genetic mutations that affect elastic fiber assembly [4] or by increased activity of proteases that degrade elastic fibers due to inflammation associated with risk factors such as smoking, alcohol consumption, aging, obesity, and hypertension [5]. Elastic fiber defects manifest as an absence or fragmentation of the elastic laminae, allowing arterial dilation and subsequent wall remodeling.
Pharmacokinetics, or the transport of solutes across the arterial wall associated with progression or treatment of disease, may be altered due to severe tissue remodeling in aneurysms. Previous studies illustrate that both convection and diffusion play a role in governing solute transport across the arterial wall in a manner that is dependent on solute size, porosity, and deformation of the wall, and microstructural components including smooth muscle cells and extracellular matrix components, such as the elastic laminae [6–9]. In work from Tarbell and collaborators, a “mechano-hydraulic model” of the arterial wall as a deformable and porous construct was developed by accounting for the physical presence and organization of smooth muscle cells, elastic fibers, collagen, and proteoglycan constituents [10–14]. This model enabled the prediction that solute, i.e., albumin, transport is dominated by convection rather than diffusion, supporting experimental studies that find the same for medium to high molecular weight solutes [15–17]. This body of work was important for demonstrating that extracellular matrix components can control the mechanics of water and solute transport in arteries, for representing the strain-dependent porosity of arteries, and for incorporating a coupling of fluid to solid phases in these structures. Still, this approach relies upon a priori assumptions of microstructural composition and morphometric information that may be lacking for many arteries, including diseased arteries of interest here.
Continuum models of fluid–solid interactions have been developed to represent mechanical contributions of deforming extracellular matrices in soft tissues, as well as strain-dependent porosities, hydraulic permeability, and solute diffusivities that may govern the movement of fluid and solute through tissue [18–20]. Here we develop a model of the arterial wall as a continuum mixture of fluid, solid and solute phases to determine intrinsic coefficients of hydraulic permeability and effective diffusivity from experimental data on arterial conductance in denuded (endothelial cell layer removed) mouse arteries [21]. The use of a multiphasic model here enables the study of an aqueous fluid phase as distinct from that of the solute phase, to reveal how fluid movement can impact solute transport in a porous and permeable solid. Using an axisymmetric finite element model of the arterial wall in FEBio [22], we numerically match model predictions for solute flux to measured volume and solute flux for arteries from wild-type control (CTL) and fibulin-5 knockout (KO) mice as a means to determine a role for elastic fiber fragmentation due to genetic mutations on arterial transport. We additionally determine properties of arteries following luminal elastase treatment (ELA) to independently assess a role for elastic fiber disruption due to protease digestion on arterial mechanics [21].
We use our multiphasic mixture model to address the hypothesis that genetic or proteolytic disruption of the elastic fibers increases intrinsic coefficients of hydraulic permeability and effective diffusivity in the arterial wall. These intrinsic coefficients are critical for understanding the transport of solutes that may play a role in the progression or treatment of arterial diseases, such as aneurysms, that are associated with elastic fiber defects.
2 Methods
2.1 Experimental data
Experiments were performed to evaluate perfusion of fluid and solute in mouse carotid arteries as described by Cocciolone and co-workers [21]. In brief, wild-type control (CTL), fibulin-5 knockout (Fbln5−/−) (KO) [22], or elastase-treated (ELA) denuded mouse carotid arteries were mounted on a pressure myograph in phosphate-buffered saline (N = 5–10/group). KO arteries have fragmented elastic fibers due to the loss of fibulin-5, a protein necessary for elastic fiber assembly [23]. ELA arteries have disrupted elastic fibers due to a brief treatment with elastase, a protease that degrades elastic fibers [24]. The carotids were pressurized to 100 mmHg (13.3 kPa), and volumetric flow rate, Q, through the arterial wall was calculated from the steady-state displacement of a bubble in the myograph inlet tubing (Fig. 1). Solute flux of the arterial tissue was determined by adding 2.5 mg/mL of 4, 75, or 150 kDa fluorescein isothiocyanate (FITC)-dextran (Sigma-Aldrich; #46,944, #46,945, and #46,946, respectively) to the arterial lumen (Clum) and measuring the concentration of FITC-dextran in the myograph fluid bath (Cbath) over time (Fig. 1). These data were used by Cocciolone and co-workers to calculate a transmural hydraulic conductance for both fluid and FITC-dextrans [21]; here, they were used to estimate intrinsic hydraulic permeability of the porous and permeable arterial wall, and effective diffusivities governing diffusion of the FITC-dextran solutes, as described below.
2.2 Biphasic computational model prediction of hydraulic flux
A biphasic model of the arterial wall as a porous, fluid-saturated mixture was constructed in FEBio [22] to simulate the flow of fluid under an intraluminal pressure gradient. The arterial wall was modeled as an axisymmetric cylinder with an inner radius Ri = 230μm, outer radius Ro = 269μm, and a length of 5.42 mm, taken from the average dimensions of mouse carotid arteries in the pressurized, axially stretched state [21]. The assumption of axisymmetry with spatial gradients only in the radial direction enabled modeling of a segment of the arterial wall cross section (3° arc) [25] meshed as shown in Fig. 2a using eight-node hexahedral elements refined closer to the inner radius [22]. In preliminary studies, we confirmed that this model of an axisymmetric radial arc gave rise to equivalent model predictions as for the full 360° cross-sectional area, with reduced computational time.
The solid phase for the wall was represented as an isotropic, neo-Hookean solid with uniform modulus (E) and Poisson’s ratio (ν) and with a strain-dependent Holmes–Mow hydraulic permeability k(J) [26] as, (1) k(J)=k0(J−ϕs1−ϕs)αe12M(J2−1),
where k0 is the reference state hydraulic permeability, J is the Jacobian of the deformation, ϕs is the solid volume fraction, α is the power-law exponent, and M is a coefficient defining the nonlinearity of the strain-dependent effect. In model development, the Poisson’s ratio (ν) was assumed to be 0.4 [27] and the modulus (E), hydraulic permeability in the reference state (k0), and solid volume fraction (ϕs) were varied to test for theoretical solution sensitivity to parameter choice. In the absence of experimental information indicating a nonlinear strain dependence of the hydraulic permeability, the values for α and M were assumed to be unity (Fig. 2b).
The test bath was modeled as a biphasic material and thus was prescribed solid material properties, even though the solid volume fraction (ϕs) is zero to model the fluid phase. As the fluid test bath was fully interfaced with the biphasic model of the arterial wall, parameters for the solid phase of the fluid bath were selected and chosen to represent an incompressible material including a Poisson’s ratio (ν) of 0.49, modulus (E) of 100,000 kPa (much greater than that of the arterial wall), and a constant isotropic hydraulic permeability (ko) of 1×10−9mm4/μN.s (similar to the arterial wall).
Boundary conditions were set to 100 mmHg (13.3 kPa) effective fluid pressure at the inner wall and 0 mmHg on the outer wall, with displacements constrained in the θ- and z-directions. Plots of finite element model (FEM)-predicted steady-state fluid flux were obtained for this range of parameters to reveal a role for E, k0, and ϕs of the arterial wall in contributing to predictions of fluid flux (w) as defined by, (2) w=−k•(∇p),
where k = k(J)I and ∇p is the gradient of the fluid pressure.
In comparison with experimental data, values for the solid volume fraction (ϕs) and the modulus (E) for the arterial wall were selected from the parametric sensitivity study in model development, and FEM predictions of fluid flux at the inner arterial wall (w) were numerically compared to the experimentally determined values for fluid flux (Jv) [21] as defined by, (3) |w|=Jv=QAA,
where Q is the volumetric flow rate and AA is the surface area of the inner arterial wall measured in the experimental studies, to determine the intrinsic permeability in the reference state (k0) for each arterial tissue type (CTL, KO, and ELA) (Fig. 2b).
2.3 Multiphasic computational model of solute flux
The transmural solute flux experiment was also modeled in FEBio [22] as described here. Values for the outer arterial radii were obtained from experimentally measured values for each tested artery [21], with an assumed arterial wall thickness of 40 μm to construct a representative axisymmetric mesh of the arterial wall as a porous and permeable, neo-Hookean solid for the FEM. Values for the modulus (E) and solid volume fraction (ϕs) of the arterial wall were chosen from the results of the parametric sensitivity to hydraulic conductance as described in the prior section with values of E = 400 Pa and ϕs = 0.6 for all arteries (Fig. 2b). Poisson’s ratio (ν) was assumed to be 0.4 [27] for all arteries. The hydraulic permeability in the reference state (k0) for each tissue type was estimated from the corresponding data for the hydraulic conductance experiment in Cocciolone et al. [21], as described in the preceding section (Fig. 2b).
In the model of the solute flux experiment, w is the volumetric flux of solvent relative to the solid and j is the molar flux of solute relative to the solid. In general, w and j are given by [28], (4) w=−k⋅(∇p+RTdDfree⋅∇c)
(5) j=d⋅(−ϕw∇c+cDfreew)
where k = k(J) I and k(j) is defined in Eq. 1, ∇p is the gradient of the effective pressure (p), (6) p=p0+RTcΦ
po is the hydrostatic fluid pressure, R is the ideal gas constant and is equal to 8.314 J/mol K, T is the absolute temperature and is equal to 310 K, c is the solute concentration, Φ is the osmotic coefficient and is assumed to be unity, d = Def f I, ∇c is the gradient of the solute concentration, ϕw is the solvent volume fraction, and Dfree is the diffusivity in free solution.
Balance of momentum for the mixture was enforced subject to boundary conditions corresponding to the solute transport experiment. Hydrostatic fluid pressure, po, between the bath and the lumen was 100 mmHg (13.3 kPa). The solute concentration at the inner wall corresponding to the lumen (Clum) was held constant for each solute (4, 70, 150 kDa FITC-dextran). The free diffusivity (Dfree) (Fig. 2b) was estimated for each solute using the approximate Stokes radii for each molecule provided by the manufacturer. FITC-dextran particles were imaged with transmission electron microscopy, as described below, to confirm their approximate size and shape.
The 10-ml test bath used for collection of diffusing solute was modeled as a multiphasic material with solid volume fraction ϕs = 0, representative of a fluid phase at the outer boundary of the arterial wall. The diffusivity for each solute in the bath was modeled with “well-mixed conditions” (i.e., assumed diffusivity 10,000 × higher than free diffusivity) [29, 30]. For these conditions, an effective diffusivity (Deff) for each solute within the arterial wall was obtained by minimizing the difference between the calculated solute flux at the outer wall to that measured experimentally in the test bath with a nonlinear Levenberg–Marquardt optimization algorithm available in FEBio [22]. Following the optimization to determine the effective diffusivity (Deff), the fluid flow (from Eq. 4) and solute flux (Eq. 5) were predicted to estimate the contribution of fluid movement to the total solute flux, as measured.
2.4 Transmission electron microscopy
FITC-dextran particles of each molecular weight were suspended at low density (0.02 mg/ml) in phosphate-buffered saline. Glow-discharged carbon-coated 200 mesh copper grids were placed on a 10 μl drop of the FITC-dextran solution and incubated for 1 min at room temperature. Post-incubation, the grids were washed serially with 5 ddH2O drops and stained with 0.75% uranyl formate for 2 min. Excess uranyl formate was blotted off using filter paper then the grids were air dried. Grids were imaged on a JEOL 1400 transmission electron microscope equipped with an AMT CCD camera operating at 120 kV at 80,000 × nominal magnification resulting in a magnified pixel size of 0.23 nm. Representative images of clearly delineated particles for each molecular weight were captured.
2.5 Statistical analysis
Values for the effective diffusivity (Deff) for each of the three FITC-dextran molecules in the arterial wall were determined by nonlinear optimization to experimental solute flux data for CTL, KO, and ELA carotid arteries on a specimen-specific basis. Tests of normality were performed and differences between tissue types and among solutes were analyzed with a two-way ANOVA followed by Tukey’s post hoc test at a significance level of p < 0.05.
3 Results
Transmission electron microscopy images of the FITC-dextran particles show that they are approximately circular with a radius comparable to the Stokes radii reported by the manufacturer (1.4, 6, and 8.5 nm, respectively for 4, 70, and 150 kDa FITC-dextran) (Fig. 3). Detailed measurements were not taken as size and shape are likely affected by the processing required for imaging. However, the images confirm that Dfree calculations based on a spherical particle with the reported Stokes radii are a reasonable assumption.
FEM predictions of fluid flux for a range of parametric values for solid volume fraction, modulus, and hydraulic permeability of the arterial wall are shown in Fig. 4. Fluid flux through the model arterial wall was determined to be insensitive to solid volume fraction (ϕs) for values in the range of 0.6−0.8, contributing less than 4% variation (Fig. 4a) and covering the range of solid volume fractions measured for undeformed and deformed arterial tissue [31]. Thus, we assumed a value of ϕs = 0.6 in further models for estimation of hydraulic permeability. Similarly, the predicted fluid flux was relatively insensitive to values for moduli (E) in the range of 126 kPa–1 MPa, contributing less than 8% variation (Fig. 4b). As the modulus of the mouse arterial wall has been measured to be within this range [32, 33], we chose a value of E = 400 kPa without introducing large variations in predictions of fluid flux, as shown by the parametric study results.
With fixed values for E and ϕs, fluid flux is linearly dependent on hydraulic permeability (Fig. 4). FEM predictions of fluid flux over time were matched to experimental data [21] to obtain values for hydraulic permeability (ko) of 1.52×10−9, 1.01×10−8, and 1.07×10−8 mm4/μN.s for CTL, KO and ELA groups, respectively. These values were then used as model inputs for the solute flux predictions (Fig. 2b).
An effective diffusivity (Deff) for each solute within the arterial wall was obtained by nonlinear optimization of the model predicted solute flux (Eq. 5) to the bath concentration (Cbath) over time measured experimentally for individual artery specimens. Representative FEBio predictions and experimental measurements are shown in Fig. 5. Results of the nonlinear optimization yielded an effective diffusivity (Deff) that showed a dependence on solute molecular weight (Fig. 6a). Deff significantly varied among arterial tissue types for the low molecular weight solute (4 kDa FITC-dextran) (p < 0.05), with the ELA group having the highest Deff. No differences in Deff were observed across tissue types for either the 70 kDa or 150 kDa FITC-dextran; however, there was a trend of increasing Deff with elastic fiber fragmentation for the 70 kDa FITC-dextran that was consistent with that for the 4 kDa FITC-dextran. Further, the highest molecular weight solute (150 kDa FITC-dextran) had the lowest Deff in all tissue groups.
Simulated volumetric fluid flow through the arterial wall in the solute flux experiment was different among the three groups of arterial tissue types (Fig. 6b). The CTL arteries were predicted to have the lowest fluid flow for all solutes while arteries in the ELA group have the highest fluid flow. The fluid flow behaviors within each arterial tissue type correspond to the hydraulic permeabilities as determined from the hydraulic conductance experiments, vary according to the geometry of each individual arterial specimen, and reflect the presence of a hydraulic pressure gradient across the arterial wall. The simulated solute flux similarly varies across arterial tissue types and according to individual specimen geometry (Fig. 6c) and is much higher for the 4 kDa compared to the 70 and 150 kDa FITC-dextrans, reflecting the differences in effective diffusivities. Statistical analyses were not performed on the simulated fluid flow and solute flux as they are derived values from previously determined parameters.
4 Discussion
Previous investigations have shown the importance of considering the arterial wall as a deformable and porous construct to estimate fluid and solute flux [6–9]. While prior models represented the microstructural elements of smooth muscle cells, elastic fibers, collagen, and proteoglycan constituents as regulators of arterial wall transport [10–14], we adopted a continuum model here that obviates the need to independently estimate microstructural parameters of apparent pore size or biochemical composition. Our multiphasic model is capable of revealing how the aqueous fluid phase can move relative to the solute (also fluid) phase. In this manner, we can understand the contributions of a hydraulic pressure relative to an osmotic pressure in driving transmural solute transport and the impact that solute size, or arterial wall microstructure, might have on those contributions. Our results show that the mobile aqueous phase dominates fluid and solute transport here, particularly for small solutes, as a result of the applied hydraulic pressure gradient that dominates over gradients in chemical potential. A parametric study revealed that the model predictions for fluid flux in this geometry were relatively insensitive to the arterial wall modulus (< 8% variation between E = 125 kPa–1 MPa) and solid volume fraction (< 4% variation between ϕs = 0.6–0.8) within a range of physiologic values, and that the movement of fluid could be governed by a single term corresponding to the undeformed, or reference state hydraulic permeability (k0).
Estimates for k0 given here are the first available for the arterial wall using a biphasic continuum model and show evidence of differences with genetic mutation or proteolytic cleavage of arterial wall constituents, namely fragmentation of elastic laminae as observed in aneurysmal disease. We estimate k0 values on the order of 10−9 mm4/μN.s for control arteries and 10−8 mm4/μN.s for arteries with genetic or proteolytic fragmentation of the elastic laminae (Fig. 2b). Our estimates are of the same order of magnitude as those for bovine cartilage (10−9 mm4/μN.s) [34] and porcine skin (10−8 mm4/μN.s) [35], and lower than those obtained for mouse skin and tendon (10−7 mm4/μN.s) [36]. Additional studies are needed to confirm our values of ko for arterial tissue and investigate nonlinear strain dependence.
An advantage of the continuum mixture approach is the ability to model additional phases representing solvent and solute flux through porous and permeable solid materials. Following the estimation of parameters governing fluid flow in the hydraulic conductance experiment, we incorporated a solvent phase representing FITC-dextran within the fluid to estimate solute flux into the test bath under a concentration gradient. Thus, we were able to estimate an effective diffusivity (Deff) within the arterial wall tissue that showed a dependence upon molecular weight (Fig. 6a), as expected [9, 37]. Our calculated Deff ranges from 14 to 30% (CTL), 10 to 36% (KO), and 41 to 72% (ELA) of the Dfree value for each size FITC-dextran molecule based on a Stokes radius calculation, indicating altered diffusion within the arterial wall tissue and a further dependence on elastic fiber structure. Our calculated Deff for CTL arterial tissue is 5–10 times higher than Deff values for similar sized dextran or albumin molecules within the cell cytoplasm [37, 38], suggesting that the FITC-dextran molecules are moving around (not through) the cells within the arterial wall. Our Deff values are similar to those measured by Hwang and Edelman [9] for transmural transport of albumin across the bovine arterial wall, however their Deff values for planar transport in rectangular tissue samples are 1–2 orders of magnitude higher than albumin diffusion in free solution. We assumed transmural (radial) transport only in our current model. Anisotropy of the multiphasic mechanical behavior should be investigated in future work.
In addition to predicting molecular size-dependent effective diffusivities, the model predicted that Deff was sensitive to elastic fiber structure in the arterial wall for the 4 kDa solute, but not for the 70 or 150 kDa solutes (Fig. 6a). There was a trend toward dependence on elastic fiber structure for the larger molecules in some cases. For example, Deff for low, medium and high molecular weight FITC-dextran was highest in the ELA arteries, as compared to KO and CTL groups. The results indicate that elastic fiber structure affects solute transport in a size-dependent manner. Size-dependent molecular sieving that depends on extracellular matrix content has been well described in cartilage tissue [39]. Studies of solute diffusion under a concentration gradient only, without the imposition of a hydraulic pressure gradient across the wall, would be useful to confirm whether Deff is differentially affected by elastic fiber organization in the arterial wall.
An important limitation of the experimental study, as well as the FEM prediction, is the absence of the intact endothelial cell layer that is a key regulator of fluid and solute transport in healthy arterial walls [40]. However, local loss of or increased permeability of the endothelial cell layer is a hallmark of many cardiovascular diseases, including aneurysms. An advantage of the approach used here is the ability to incorporate layers corresponding to the endothelial cells and concentric layers of smooth muscle cells and elastic laminae that could be used to model different disease processes in future work. As data required to independently determine the full set of physical properties were not available even for the simple denuded homogeneous arterial wall used here, we chose to focus on assessing the hydraulic permeability and effective diffusivity as a first step toward this goal. The continuum model allows separation of diffusivity and fluid flow, which is important for understanding different contributions of diffusive and advective transport. Our modeling results suggest that convection plays a dominant role in this experimental configuration (Figs. 6b and c), consistent with prior studies in large arteries with and without an intact endothelial cell layer [15–17].
Our continuum model begins with a volume-averaged approach that assumes a homogeneous and continuous fluid phase throughout the arterial wall that is not impacted by local variations in wall architecture at the nanometer level. These features could be incorporated with representations of solid phase microstructure in future work with evidence that fluid–solid interactions at the nanometer-level impact fluid transport beyond the porosity and permeability. Many different features contribute to the microstructure of the solid phase (e.g., elastic fibers and collagen) and also the composition of the aqueous fluid phase (i.e., charged macromolecules). Additional features of solute or matrix charge that could affect binding of the molecules during transport, or cellular uptake, could also be included in future models. There are numerous examples of elastic laminae binding with macromolecules in the arterial wall [41–43]. Hydrophobic molecules, such as paclitaxel and sirolimus, routinely used to target the arterial wall for therapeutic purposes [44] can locally partition resulting in wall concentrations higher than applied levels [45]. Degradation of elastic laminae with aneurysmal disease may affect molecular binding or partitioning within the arterial wall.
Despite the benefits of the continuum model approach, the predictions rely upon assumed parameters that have not been independently measured for mouse arteries, and that would need to be experimentally determined for the CTL, KO and ELA groups. However, the parameters (namely the solid volume fraction and Young’s modulus of the arterial wall) used in the current study are within physiologic ranges and our sensitivity study showed that variations over a broader range had negligible effects on fluid flux. Additional studies could be undertaken to determine nonlinear strain-dependent material properties, solid volume fraction, and their changes with experimental treatment. Further studies with more complex, structurally based constitutive models may reveal additional insight into how fluid–solid interactions impact transport across the arterial wall. The modeling developed here lays out an approach that can be used to simulate transport under different boundary conditions or microstructure to reveal how genetic mutations in or proteolytic degradation of elastic fibers affect specific features of arterial mechanics relevant for progression or treatment of human disease.
Funding
This work was supported by grants from the American Heart Association 19TPA-34910047 (JEW) and the National Institute of Arthritis and Musculoskeletal and Skin Diseases of the NIH under award number R01AR070975 (LAS). The FEBio software suite has been developed with partial support from the NIH. Transmission electron microscopy was performed by the Washington University Center for Cellular Imaging supported by Washington University School of Medicine, the Children’s Discovery Institute of Washington University, St. Louis Children’s Hospital (CDI-CORE-2015-505 and CDI-CORE-2019-813), and the Foundation for Barnes-Jewish Hospital (3770 and 4642).
Availability of data
All data are included in the published manuscript or references.
Fig. 1 Schematic (a) and photographs (b) of the transport experiments (not to scale). An artery was mounted on a pressure myograph in a bath of phosphate-buffered saline. The myograph inlet was connected to a hydrostatic pressure column (100 mmHg/13.3 kPa) with a reservoir of phosphate-buffered saline that was either solute free or contained 2.5 mg/mL of 4, 70, or 150 kDa FITC-dextran. The direction of flow is from the column toward the arterial lumen and then across the wall of the artery, as the myograph outlet is clamped. Displacement of an air bubble in the inlet tubing was used to determine volumetric fluid flow through the arterial wall and calculate hydraulic conductance. Total solute flux through the arterial wall was measured by recording the starting concentration (Clum) of FITC-dextran within the arterial lumen and the change in FITC-dextran concentration (Cbath) in the external bath over time. Ri and Ro are the inner and outer radii of the artery, respectively, while Rb is the equivalent radius of the 10-mL volume bath. These measurements were used to model the experimental setup in FEBio
Fig. 2 Schematic of the model geometry and variables. a The arterial wall is modeled as an axisymmetric wedge of 3° (yellow) with a wall thickness defined by the inner (Ri) and outer (Ro) radii. In addition, the model geometry requires specification of an equivalent radius for the 10-mL fluid bath, Rb. b The table provides the multiphasic model parameters for the arterial wall.
Fig. 3 Transmission electron micrographs of individual FITC-dextran particles of 4 (a), 70 (b), and 150 (c) kDa molecular weight. The particles are approximately circular with sizes near the reported Stokes radii from the manufacturer. Scale bar = 10 nm
Fig. 4 Parametric study for determination of biphasic model parameters. a Fluid flux plotted against hydraulic permeability for differing values of solid volume fraction and assumed modulus (E = 251.2 kPa). b Fluid flux plotted against hydraulic permeability for differing modulus values and an assumed solid volume fraction (ϕs = 0.6). Hydraulic permeability is the main determinant of fluid flux and fluid flux is largely independent of solid volume fraction and modulus within ranges considered physiologic for the mouse carotid artery
Fig. 5 Representative experimental data (symbols) and model predictions (lines) for the solute transport experiments for individual artery specimens. Model predictions were obtained by nonlinear optimization of the effective diffusivity (Deff) to minimize differences between the experimental and calculated solute flux (Eq. 5). Concentration in the external bath (Cbath) over time was recorded experimentally and predicted computationally for 4 (a), 70 (b), and 150 (c) kDa FITC-dextran after addition to the arterial lumen in wild-type (CTL), Fbln5−/− (KO), and elastase-treated (ELA) mouse carotid arteries
Fig. 6 Effective diffusivity values and simulated fluid flow and solute flux for the multiphasic model. a Effective diffusivity (Deff) values were determined from the model fits to solute flux data on a specimen-specific basis. * = p < 0.05 among all tissue types for 4 kDa FITC-dextran. # = p < 0.05 for 4 kDa compared to 70 and 150 kDa FITC-dextran for KO tissue. & = p < 0.05 for 4 kDa compared to 70 and 150 kDa FITC-dextran for ELA tissue. Significance was determined by two-way ANOVA followed by Tukey’s post hoc test. b Corresponding predictions of relative fluid flow and c solute flux shown on a specimen-specific basis. Fluid flow depends on the imposed pressure gradient, arterial geometry, and hydraulic permeability, and so varies among the CTL, KO and ELA conditions as expected. Solute flux similarly varies among CTL, KO and ELA conditions but is also found to be much higher for the transport of the 4 kDa FITC-dextran due to the increased effective diffusivity values. Significance was not determined for data shown in panels B and C because the values are derived from previously determined values. N = 5–10/group
Code availability The model was developed in FEBio, an open-source software. Code for the model is available on request from the corresponding author.
Conflict of interest None.
References
1. Cocciolone AJ , Hawes JZ , Staiculescu MC , Johnson EO , Murshed M , Wagenseil JE : Elastin, arterial mechanics, and cardiovascular disease. Am J Physiol Heart Circ Physiol (2018). 10.1152/ajpheart.00087.2018
2. O’Connell MK , Murthy S , Phan S , Xu C , Buchanan J , Spilker R , Dalman RL , Zarins CK , Denk W , Taylor CA : The three-dimensional micro- and nanostructure of the aortic medial lamellar unit measured using 3D confocal and electron microscopy imaging. Matrix Biol 27 (3 ), 171–181 (2008). 10.1016/j.matbio.2007.10.008 18248974
3. Hiratzka LF , Bakris GL , Beckman JA , Bersin RM , Carr VF , Casey DE Jr. , Eagle KA , Hermann LK , Isselbacher EM , Kazerooni EA , Kouchoukos NT , Lytle BW , Milewicz DM , Reich DL , Sen S , Shinn JA , Svensson LG , Williams DM , American College of Cardiology Foundation/American Heart Association Task Force on Practice G, American Association for Thoracic S, American College of R, American Stroke A, Society of Cardiovascular A, Society for Cardiovascular A, Interventions, Society of Interventional R, Society of Thoracic S, Society for Vascular M: 2010 ACCF/AHA/AATS/ACR/ASA/SCA/SCAI/SIR/STS/SVM guidelines for the diagnosis and management of patients with thoracic aortic disease: a report of the american college of cardiology foundation/american heart association task force on practice guidelines, American Association for Thoracic Surgery, American College of Radiology, American Stroke Association, Society of Cardiovascular Anesthesiologists, Society for Cardiovascular Angiography and Interventions, Society of Interventional Radiology, Society of Thoracic Surgeons, and Society for Vascular Medicine. Circulation 121 (13 ), e266–369 (2010). 10.1161/CIR.0b013e3181d4739e 20233780
4. Faggion Vinholo T , Brownstein AJ , Ziganshin BA , Zafar MA , Kuivaniemi H , Body SC , Bale AE , Elefteriades JA : Genes associated with thoracic aortic aneurysm and dissection: 2019 update and clinical implications. Aorta (Stamford) 7 (4 ), 99–107 (2019). 10.1055/s-0039-3400233 31842235
5. Jones GT , Tromp G , Kuivaniemi H , Gretarsdottir S , Baas AF , Giusti B , Strauss E , Van’t Hof FN , Webb TR , Erdman R , Ritchie MD , Elmore JR , Verma A , Pendergrass S , Kullo IJ , Ye Z , Peissig PL , Gottesman O , Verma SS , Malinowski J , Rasmussen-Torvik LJ , Borthwick KM , Smelser DT , Crosslin DR , de Andrade M , Ryer EJ , McCarty CA , Bottinger EP , Pacheco JA , Crawford DC , Carrell DS , Gerhard GS , Franklin DP , Carey DJ , Phillips VL , Williams MJ , Wei W , Blair R , Hill AA , Vasudevan TM , Lewis DR , Thomson IA , Krysa J , Hill GB ,Roake J ,Merriman TR ,Oszkinis G ,Galora S ,Saracini C ,Abbate R ,Pulli R ,Pratesi C ,Saratzis A ,Verissimo AR , Bumpstead S , Badger SA , Clough RE , Cockerill G , Hafez H , Scott DJ , Futers TS , Romaine SP , Bridge K , Griffin KJ , Bailey MA , Smith A , Thompson MM , van Bockxmeer FM , Matthiasson SE , Thorleifsson G , Thorsteinsdottir U , Blankensteijn JD , Teijink JA , Wijmenga C , de Graaf J , Kiemeney LA , Lindholt JS , Hughes A ,Bradley DT ,Stirrups K ,Golledge J ,Norman PE ,Powell JT ,Humphries SE ,Hamby SE ,Goodall AH ,Nelson CP , Sakalihasan N , Courtois A , Ferrell RE , Eriksson P , Folkersen L , Franco-Cereceda A , Eicher JD , Johnson AD , Betsholtz C , Ruusalepp A , Franzen O , Schadt EE , Bjorkegren JL , Lipovich L , Drolet AM , Verhoeven EL , Zeebregts CJ , Geelkerken RH , van Sambeek MR , van Sterkenburg SM , de Vries JP , Stefansson K , Thompson JR , de Bakker PI , Deloukas P , Sayers RD , Harrison SC , van Rij AM , Samani NJ , Bown MJ : Meta-analysis of genome-wide association studies for abdominal aortic aneurysm identifies four new disease-specific risk loci. Circ Res 120 (2 ), 341–353 (2017). 10.1161/CIRCRESAHA.116.308765 27899403
6. Caro CG , Lever MJ : Factors influencing arterial wall mass transport. Biorheology 21 (1–2 ), 197–205 (1984)6380612
7. Tedgui A ,Lever MJ :Filtration through damaged and undamaged rabbit thoracic aorta. Am J Physiol 247 (5Pt2 ), H784–791 (1984). 10.1152/ajpheart.1984.247.5.H784 6496759
8. Baldwin AL , Wilson LM , Simon BR : Effect of pressure on aortic hydraulic conductance. Arterioscler Thromb 12 (2 ), 163–171 (1992)1543691
9. Hwang CW , Edelman ER : Arterial ultrastructure influences transport of locally delivered drugs. Circ Res 90 (7 ), 826–832 (2002)11964377
10. Huang ZJ , Tarbell JM : Numerical simulation of mass transfer in porous media of blood vessel walls. Am J Physiol 273 (1 Pt 2 ), H464–477 (1997). 10.1152/ajpheart.1997.273.1.H464 9249521
11. Kim WS , Tarbell JM : Macromolecular transport through the deformable porous media of an artery wall. J Biomech Eng 116 (2 ), 156–163 (1994). 10.1115/1.2895714 8078322
12. Kim WS , Tarbell JM : Prediction of macromolecular transport through the deformable porous media of an artery wall by pore theory. Korean J Chem Eng 13 (5 ), 457–465 (1996). 10.1007/Bf02705994
13. Tarbell JM : Mass transport in arteries and the localization of atherosclerosis. Annu Rev Biomed Eng 5 , 79–118 (2003). 10.1146/annurev.bioeng.5.040202.121529 12651738
14. Tada S , Tarbell JM : Internal elastic lamina affects the distribution of macromolecules in the arterial wall: a computational study. Am J Physiol Heart Circ Physiol 287 (2 ), H905–913 (2004). 10.1152/ajpheart.00647.2003 15016628
15. Ramirez CA , Colton CK , Smith KA , Stemerman MB , Lees RS : Transport of 125I-albumin across normal and deendothelialized rabbit thoracic aorta in vivo. Arteriosclerosis 4 (3 ), 283–291 (1984)6712542
16. Tedgui A , Lever MJ : The interaction of convection and diffusion in the transport of 131I-albumin within the media of the rabbit thoracic aorta. Circ Res 57 (6 ), 856–863 (1985). 10.1161/01.res.57.6.856 4064259
17. Lever MJ , Jay MT : Convective and diffusive transport of plasma proteins across the walls of large blood vessels. Front Med Biol Eng 5 (1 ), 45–50 (1993)8323882
18. Ateshian GA , Maas S , Weiss JA : Solute transport across a contact interface in deformable porous media. J. Biomech 45 (6 ), 1023–1027 (2012). 10.1016/j.jbiomech.2012.01.003 22281406
19. Ateshian GA , Maas S , Weiss JA (2010) Finite element algorithm for frictionless contact of porous permeable media under finite deformation and sliding. J. Biomech. Eng. Trans. Asme 132 (6 ). doi: 10.1115/1.4001034
20. Lai WM , Hou JS , Mow VC : A triphasic theory for the swelling and deformation behaviors of articular cartilage. J Biomech Eng 113 (3 ), 245–258 (1991). 10.1115/1.2894880 1921350
21. Cocciolone AJ , Johnson E , Shao JY , Wagenseil JE : Elastic fiber fragmentation increases transmural hydraulic conductance and solute transport in mouse arteries. J Biomech Eng (2018). 10.1115/1.4042173
22. Maas SA , Ellis BJ , Ateshian GA , Weiss JA (2012) FEBio: Finite elements for biomechanics. J. Biomech. Eng. Trans. Asme 134 (1 ). doi: 10.1115/1.4005694
23. Budatha M , Roshanravan S , Zheng Q , Weislander C , Chapman SL , Davis EC , Starcher B , Word RA , Yanagisawa H : Extracellular matrix proteases contribute to progression of pelvic organ prolapse in mice and humans. J Clin Invest 121 (5 ), 2048–2059 (2011). 10.1172/JCI45636 21519142
24. Wagenseil JE , Nerurkar NL , Knutsen RH , Okamoto RJ , Li DY , Mecham RP : Effects of elastin haploinsufficiency on the mechanical behavior of mouse arteries. Am J Physiol Heart Circ Physiol 289 (3 ), H1209–1217 (2005). 10.1152/ajpheart.00046.2005 15863465
25. Maas S , Ateshian GA , Weiss JA , Rawlins D (2019) Section 2.9: Tutorial 9: axisymmetric analysis of biphasic indentation. In: Preview Finite Element Preprocessing, User’s Manual Version 2.1. https://help.febio.org/PreView/PreView_2_1/index.html,
26. Holmes MH , Mow VC : The nonlinear characteristics of soft gels and hydrated connective tissues in ultrafiltration. J Biomech 23 (11 ), 1145–1156 (1990). 10.1016/0021-9290(90)90007-p 2277049
27. Ebrahimi AP : Mechanical properties of normal and diseased cerebrovascular system. J Vasc Interv Neurol 2 (2 ), 155–162 (2009)22518247
28. Maas S , Rawlins D , Weiss JA , Ateshian GA (2020) FEBio Theory Manual Version 3.0
29. Guang Y , McGrath TM , Klug N , Nims R , Shih CC , Bayguinov PO , Guilak F , Pham CTN , Fitzpatrick JAJ , Setton LA : Combined experimental approach and finite element modeling of small molecule transport through joint synovium to measure effective diffusivity. J Biomech Eng (2019). 10.1115/1.4044892
30. Nims RJ , Cigan AD , Albro MB , Vunjak-Novakovic G , Hung CT , Ateshian GA : Matrix production in large engineered cartilage constructs is enhanced by nutrient channels and excess media supply. Tissue Eng. Part C Methods 21 (7 ), 747–757 (2015). 10.1089/ten.TEC.2014.0451 25526931
31. Chooi KY , Comerford A , Sherwin SJ , Weinberg PD (2016) Intimal and medial contributions to the hydraulic resistance of the arterial wall at different pressures: a combined computational and experimental study. J R Soc Interface 13 (119 ). doi:10.1098/rsif.2016.0234
32. Ferruzzi J , Bersi MR , Uman S , Yanagisawa H , Humphrey JD (2015) Decreased elastic energy storage, not increased material stiffness, characterizes central artery dysfunction in fibulin-5 deficiency independent of sex. J Biomech Eng 137 (3 ). doi:10.1115/1.4029431
33. Pavey SN , Cocciolone AJ , Marty AG , Ismail HN , Hawes JZ , Wagenseil JE : Pentagalloyl glucose (PGG) partially prevents arterial mechanical changes due to elastin degradation. Exp. Mech (2020). 10.1007/s11340-020-00625-1
34. Mow VC , Holmes MH , Lai WM : Fluid transport and mechanical properties of articular cartilage: a review. J Biomech 17 (5 ), 377–394 (1984). 10.1016/0021-9290(84)90031-9 6376512
35. Oomens CWJ , Vancampen DH , Grootenboer HJ : A mixture approach to the mechanics of skin. J. Biomech 20 (9 ), 877–885 (1987). 10.1016/0021-9290(87)90147-3 3680313
36. Oftadeh R , Connizzo BK , Nia HT , Ortiz C , Grodzinsky AJ : Biological connective tissues exhibit viscoelastic and poroelastic behavior at different frequency regimes: application to tendon and skin biophysics. Acta Biomater 70 , 249–259 (2018). 10.1016/j.actbio.2018.01.041 29425716
37. Peters R : Nucleo-cytoplasmic flux and intracellular mobility in single hepatocytes measured by fluorescence microphotolysis. EMBO J 3 (8 ), 1831–1836 (1984)6207019
38. Wojcieszyn JW , Schlegel RA , Wu ES , Jacobson KA : Diffusion of injected macromolecules within the cytoplasm of living cells. Proc Natl Acad Sci U S A 78 (7 ), 4407–4410 (1981). 10.1073/pnas.78.7.4407 6945591
39. Maroudas A : Transport of solutes through cartilage—permeability to large molecules. J Anat 122 , 335–347 (1976)1002608
40. Penn MS , Saidel GM , Chisolm GM : Relative significance of endothelium and internal elastic lamina in regulating the entry of macromolecules into arteries in vivo. Circ Res 74 (1 ), 74–82 (1994)8261597
41. Penn MS , Koelle MR , Schwartz SM , Chisolm GM : Visualization and quantification of transmural concentration profiles of macromolecules across the arterial wall. Circ Res 67 (1 ), 11–22 (1990). 10.1161/01.res.67.1.11 2364485
42. Winlove CP , Parker KH , Ewins AR : Reversible and irreversible interactions between elastin and plasma lipoproteins. Biochem. Biophys. Acta 838 (3 ), 374–380 (1985). 10.1016/0304-4165(85)90237-5 3970975
43. Chappell DC , Parker KH , Winlove CP : Mechanisms of interaction of albumin with arterial elastin. Connect Tissue Res 30 (2 ), 157–163 (1993). 10.3109/03008209309041329 8149747
44. Lee DH , de la Torre Hernandez JM : The Newest Generation of Drug-eluting Stents and Beyond. Eur Cardiol 13 (1 ), 54–59 (2018). 10.15420/ecr.2018:8:2 30310472
45. Creel CJ , Lovich MA , Edelman ER : Arterial paclitaxel distribution and deposition. Circ Res 86 (8 ), 879–884 (2000). 10.1161/01.res.86.8.879 10785510
|
PMC008xxxxxx/PMC8983018.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
101264945
34537
Curr Opin HIV AIDS
Curr Opin HIV AIDS
Current opinion in HIV and AIDS
1746-630X
1746-6318
35225247
8983018
10.1097/COH.0000000000000716
NIHMS1771153
Article
Critical Social and Behavioral Sciences Perspectives on Ending the HIV Epidemic
Auerbach Judith D. a
Dubé Karine b
a Division of Prevention Sciences, Center for AIDS Prevention Studies, School of Medicine, University of California, San Francisco, San Francisco, California, USA
b Gillings School of Global Public Health, University of North Carolina, Chapel Hill, Chapel Hill, North Carolina, USA
Correspondence to Judith D. Auerbach, PhD., University of California, San Francisco, San Francisco, CA 94143, USA. [email protected]
15 1 2022
01 3 2022
01 3 2023
17 2 3739
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
pmcIntroduction
Recent development of highly efficacious long-acting, injectable drugs to prevent and treat HIV have excited the world. At the same time, the reticence of a significant proportion of the eligible population to use these technologies and the existence of major faults in the systems through which these technologies are distributed to those who do want them give us all pause. The tension between the possibility of new discoveries and the reality of their uptake and use (or lack thereof) at its core reflects the tension between “efficacy” and “effectiveness.” While a technology may prove highly efficacious for individuals in a controlled research environment, when it hits the “real world,” its true effectiveness in a population is contingent on physical, psychological, and social (including political, economic, and cultural) and behavioral factors that influence its adoption and impact [1]. These factors play out differentially across societies, although there are some common features.
HIV-associated Health and Social Inequalities
In all societies, HIV is not distributed equally among social groups. HIV-associated inequalities (or “disparities,” as they are popularly called) have received a great deal of attention as of late, including in major policy agendas, such as the United States’ Ending the HIV Epidemic and National HIV/AIDS Strategy, but there has been very little focus on the social and behavioral sciences related to them. Beyond naming and documenting HIV-associated inequalities (e.g., by race, ethnicity, sex, gender, age, geography, class, etc.), it is imperative to better define, operationalize, measure, and analyze how these inequalities concretely affect HIV outcomes in order to redress them. Inequalities are not simply ideas and beliefs manifest in humans’ behavior to each other, they are structured and institutionalized in ways that make them core determinants of HIV, although, as Bowleg et al [2] argue in the case of structural racism in the United States, this has not been sufficiently recognized.
Additionally, inequalities are often overlapping and intersecting and are at the root of syndemics—“disease concentrations of two or more epidemics and disease interactions that share social, ecological, and structural factors that exacerbate their exposure and intensification [3].” The review of recent syndemics literature by Logie, et al. [3] reveals how health conditions interact with environmental factors to foment and/or exacerbate HIV and other epidemics to produce differential and unequal outcomes for different people. Importantly, their review also shows that protective social conditions or practices, such as social support and community resources, can mitigate the effects of these interactions and operate as “counter-syndemics.”
Addressing intersecting social inequalities is core to realizing the “promise” (i.e., effectiveness) of new biomedical HIV prevention and treatment technologies. As Philbin and Perez-Brumer [4] detail, this includes attention not only to how new efficacious technologies are implemented, but also to how they are researched in the first place. The criteria for participating in clinical trials, along with the need to locate these trials in settings with high HIV burden, means that some populations are excluded from research and that some settings may not have the resources to implement the trial findings once interventions are proven efficacious.
The Dynamic Nature of Interpersonal and Social Processes
Social factors are not really determinants, in that they are dynamic and not fixed, nor are they linear. Rather, whether related to beliefs (e.g., stigma, religion), behaviors and practices (e.g., intimate relationships, drug use, sex), or structures (e.g., housing, politics, economy, law) social factors are processes that are shaped and reshaped by the actions of individuals and collectivities. Moreover, they are not only or always barriers—forces that operate to impede people’s ability to take up and use HIV prevention, treatment, and care strategies—but they often can be facilitators. As noted by Logie et al. [3] and Ruiz et al. [5], social capital, social support systems, and social networks can be harnessed, along with psychological strengths, to improve the health and well-being of people living and aging with HIV—to move from surviving to thriving.
Multiple Levels of Analysis and Engagement
The complexity of the HIV epidemic calls for theoretical and applied research that addresses all levels of social organization – individual, relational, societal. Gamarel at al. [6] review social and behavioral interventions that focus on individuals’ cognitive, affective, and behavioral processes, relational factors in dyads, and structural factors operating within organizations that are important for optimizing risk reduction and uptake and use of HIV prevention and treatment technologies. Such interventions are critical to addressing HIV-related inequities but will only be effective at a population if they are sustained.
At the societal level, policy interventions play a key role in how HIV epidemics unfold and affect individuals and communities. As Hoppe et al. [7] detail, laws and policies that criminalize HIV infection and transmission have had disparate impact on various populations in North America as they have been unevenly applied to people based on race, gender, and sexuality. Moreover, most of these laws have yet to be reformed or “modernized” in light of recent scientific developments in HIV prevention and treatment – particularly the efficacy of pre-exposure prophylaxis in preventing acquisition and the efficacy of highly active antiretroviral therapy in preventing transmission. New HIV surveillance technologies based on phylogenetic science also must be monitored for their potential impact on criminalization statutes and how they are applied.
Multiple Disciplines and Methods
Addressing all social levels in which HIV operates requires engagement with and among multiple scientific disciplines and affected communities. As the review by Pyra et al. [8] makes clear, there is a critical intersection between social and behavioral sciences and implementation science that can strengthen progress in moving from efficacy to effectiveness in the HIV response. That intersection, or “transdisciplinary” approach, would involve incorporating social and behavioral theories (e.g., on power and inequality, structural racism, intersectionality, etc.) and methods (e.g., psychometric tools, ethnography) into implementation theories and frameworks to improve understanding of mechanisms of change in implementation goals, such as equitable delivery of interventions.
Budhwani et al. [9] describe how implementation of HIV prevention and treatment strategies also can benefit from better adaptation and integration of new digital health technologies for information dissemination and intervention delivery. Such technologies, when tailored appropriately with community involvement, have the potential to protect confidentiality and reduce stigma which otherwise hamper many people’s engagement with HIV services.
The Centrality of Communities
The importance of community engagement in the HIV response cannot be underestimated. This means much more than inviting individuals to sit on community advisory boards to provide input into other people’s research ideas. Rather it means understanding whether and how HIV is present in the everyday lives of people, and how that understanding can best be integrated into scientific and programmatic responses within and in collaboration with communities [10]. The best ways to ascertain the dynamics and nuances of HIV prevention, treatment, and care, in the lives of people is to center community perspectives. As Arnold et al. [11] detail, this can be done with ethnographic and other qualitative methods, such as in-depth interviews, focus groups, participant observation, and innovative methods, such as photovoice and “go-along” observations—alone or in combination with quantitative methods. Centering community in the HIV response is essential for moving from efficacy to effectiveness in equitable and sustainable ways.
Conclusion
The papers in this volume review recent literature that aims to unpack and address key social and behavioral factors relevant to the effectiveness of HIV prevention and treatment strategies and the possibilities of “ending the HIV epidemic.” They reflect “critical” social and behavioral sciences perspectives on HIV and AIDS in both sense of the word – by providing insights that are critical to a complete understanding of the global HIV epidemic, and by providing critiques of the ways in which those insights have been marginalized in medicine, public health, and research to the detriment to the HIV response. Our goal in this volume is not to counterpoise different disciplines and perspectives, but rather to bring the social and behavioral into conversation with the biomedical as we aim toward a more truly integrated science (and practice) approach to HIV, which we know is necessary for ending the global epidemic.
Acknowledgements
We would like to thank all lead and co-authors of papers in this section of Current Opinion in HIV and AIDS and Steve Deeks and Giuseppe Pantaleo for inviting us to organize a volume dedicated to social and behavioral sciences.
Financial support and sponsorship
J.D.A.’s work was supported in part by National Institutes of Health/National Institute of Mental Health 3P30MH062246-20S2.
Conflicts of interest
None.
REFERENCES
1. Kippax S , Stephenson N . Beyond the Distinction between Biomedical and Social Dimensions of HIV Prevention through the Lens of a Social Public Health. Am J Public Heal. 2012;102 : 789–99.
2. Bowleg L , Malekzadeh A , Boone C , Mbaba M . Ending the HIV Epidemic for All, Not Just Some: Structural Racism as a Fundamental but Overlooked Social-Structural Determinant of the U.S. HIV Epidemic. Curr Opin HIV AIDS. 2022.
3. Logie C , Coelho M , Kohrt B , Tsai A , Mendenhall E . Exploring the State of Syndemics and HIV Research. Curr Opin HIV AIDS. 2022.
4. Philbin M , Perez-Brumer A . Promise, Perils and Cautious Optimism: The Next Frontier in Long-Acting Modalities for the Prevention and Teatment of HIV and AIDS. Curr Opin HIV AIDS. 2022.
5. Ruiz E , Greene K , Galea J , Brown B . From Surviving to Thriving: The Current Status of the Behavioral, Social, and Psychological Issues of Aging with HIV. Curr Opin HIV AIDS. 2022.
6. Gamarel K , King W , Operatio D . Behavioral and Social Interventions to Promote Optimal HIV Prevention and Care Continua Outcomes in the United States. Curr Opin HIV AIDS. 2022.
7. Hoppe T , McClelland A , Pass K . Beyond Criminalization: Reconsidering HIV Criminalization in an Era of Reform. Curr Opin HIV AIDS. 2022.
8. Pyra M , Motley D , Bouris A . Moving Towards Equity: Fostering Transdisciplinary Research between the Social Behavioral Sciences and Implementation Science to End the HIV Epidemic. Curr Opin HIV AIDS. 2022.
9. Budhwani H , Kiszla B , Hightow-Weidman L . Adapting Digital Health Interventions for the Evolving HIV Landscape: Examples to Support Prevention and Treatment Research. Curr Opin HIV AIDS. 2022.
10. MacQueen KM , Auerbach JD (Guest Editors). Science, Theory, and Practice of Engaged Research: Good Participatory Practices and Beyond. J Int AIDS Soc. 2018; 21 (S7 ):e25179.30334608
11. Arnold E , Campbell C , Koester K . The Innovative Use of Qualitative and Mixed Methods Research to Advance Improvements along the HIV Prevention and Care Continuum. Curr Opin HIV AIDS. 2022.
|
PMC008xxxxxx/PMC8983019.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
8301365
4429
J Am Coll Cardiol
J Am Coll Cardiol
Journal of the American College of Cardiology
0735-1097
1558-3597
35144750
8983019
10.1016/j.jacc.2021.11.048
NIHMS1767191
Article
Immune Checkpoint Therapies and Atherosclerosis: Mechanisms and Clinical Implications: JACC State-of-the-Art Review
Vuong Jacqueline T. MD 1
Stein-Merlob Ashley F. MD 2
Nayeri Arash MD 2
Sallam Tamer MD PhD 2
Neilan Tomas G. MD MPH 3
Yang Eric H. MD 24
1. Department of Medicine, Ronald Reagan UCLA Medical Center, Los Angeles, California 90095
2. Division of Cardiology, Department of Medicine, Ronald Reagan UCLA Medical Center, Los Angeles, CA 90095
3. Division of Cardiology, Department of Medicine, Massachusetts General Hospital, Boston, MA 02114
4. UCLA Cardio-Oncology Program, Division of Cardiology, Department of Medicine, University of California at Los Angeles, Los Angeles, CA
Corresponding Author: Eric H. Yang, MD, UCLA Cardiovascular Center, 100 Medical Plaza, Suite 630, Los Angeles, CA 90095, Work Phone: (310)825-9011, Fax #: (310)825-9012, Twitter: @datsunian
6 1 2022
15 2 2022
15 2 2023
79 6 577593
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Immune checkpoint inhibitor therapy has revolutionized the treatment of advanced malignancies in recent years. Numerous reports have detailed the myriad of possible adverse inflammatory effects of immune checkpoint therapies, including within the cardiovascular system. However, these reports have been largely limited to myocarditis. The critical role of inflammation and adaptive immunity in atherosclerosis has been well characterized in preclinical studies, and several emerging clinical studies indicate a potential role of immune checkpoint targeting therapies in the development and exacerbation of atherosclerosis. In this review, we provide an overview of the role of T-cell immunity in atherogenesis and describe the molecular effects and clinical associations of both approved and investigational immune checkpoint therapy on atherosclerosis. We also highlight the role of cholesterol metabolism in oncogenesis and discuss the implications of these associations on future treatment and monitoring of atherosclerotic cardiovascular disease in the oncologic population receiving immune checkpoint therapy.
Condensed Abstract
In the last decade, immune checkpoint inhibitors have been revolutionary in the field of oncology and has become the standard of care in numerous advanced malignancies. As T cell immunity plays an integral role in atherogenesis through promotion of inflammation, preclinical studies and recently emerging clinical studies indicate a role of immune checkpoint inhibitors in the development of atherosclerosis. This review provides a comprehensive analysis of the molecular effects and clinical associations between immune checkpoint altering therapy and atherosclerotic cardiovascular disease (ASCVD) risk, and identifies important considerations to the surveillance and treatment of atherosclerosis in patients receiving immune checkpoint therapy.
Inflammation
Immunology
Cardiovascular Disease
Cardio-Oncology
Atherosclerosis
pmcIntroduction: Emergence of Immune Checkpoint Inhibitors
Modern day immunotherapy originates from the cancer immunosurveillance hypothesis, which states that immune cells are responsible for the surveillance and elimination of nascent transformed cells in host tissues(1). Enhanced appreciation for the ability for tumor cells to evade immune surveillance (2) galvanized efforts to develop therapies restoring the host immune system’s antineoplastic defenses. The culmination of these efforts led to the development of immune checkpoint inhibitors (ICIs) that restore the host T cell immune response against cancer cells. The development of ICIs in the last decade has rapidly revolutionized cancer therapy, with applications of ICIs targeting CTLA-4 and PD-1/PDL-1 pathways in cancers including melanoma, non-small cell lung cancers (NSCLC), and urothelial carcinomas(3-5).
When ICIs were being approved, it was anticipated that diffuse leveraging of the immune system would lead to off-tumor adverse events, which occur in up to 70% of patients and may affect any organ system, but are typically easily managed. Initial reports detailing the potential cardiovascular effects of ICIs focused principally on the uncommon, but highly morbid, occurrence of myocarditis, mediated by direct T cell infiltration of the myocardium expressing PD-L1(6). More recently, the potential for ICIs to affect the cardiovascular system beyond myocarditis has been reported. Specifically, preclinical evidence supports that treatment with ICIs can theoretically promote the development and acceleration of atherosclerosis, and recent emerging clinical evidence suggests an increase in atherosclerotic-related cardiovascular events in patients receiving ICI therapy and a possible connection between ICIs and increased atherosclerotic cardiovascular disease (ASCVD) risk (7).
The role of inflammation in atherosclerosis has been well established, with the CANTOS, LoDoCo, and COLCOT trials demonstrating that targeting immune responses can improve clinical cardiovascular outcomes(8-10). These studies showed that targeting innate immunity, such as inhibition of IL-1β via canakinumab and neutrophil function via colchicine, may mitigate the progression of ASCVD. As adaptive immunity via T cell activation has also been shown to play a crucial role in the development and progression of atherosclerosis, ICIs may have important implications on atherosclerotic disease (11). This review aims to describe the role of T-cell mediated immunity in atherogenesis, the molecular implications of various pathways of immune checkpoint alteration in atherosclerosis, the current clinical associations between treatment with ICIs and atherosclerosis, and potential treatment avenues.
Overview of T Cell Mediated Immunity and Key Immune Checkpoint Pathways
1. Overview of T Cell Mediated Immunity
Tumor associated antigens are recognized and phagocytosed by antigen presenting cells (APCs) such as dendritic cells and macrophages. These antigens are presented via major histocompatibility complex (MHC) molecules on the surface of APCs, which interact with T cell receptors (TCRs) and serve as the primary stimulatory signal leading to the intracellular cascade that activates naïve CD4+ and CD8+ T cells. However, full T cell activation is dependent on the presence of co-stimulatory signals involving the binding of CD28 on T cell surfaces to B7-1 (CD80) or B7-2 (CD86) on APCs (Figure 1A).
Subsequently, the cytokine composition of the surrounding environment determine the fate of T cell differentiation into various subclasses, including CD8+ cells into cytotoxic T cells, and CD4+ T cells into helper T cells (Th1, Th2 , and Th17) and regulatory T cells (Treg) (Figure 1A). Cytotoxic T cells secrete cytotoxins and promote apoptosis of its target cells. Of the CD4+ T cell derivatives, Th1 promotes cell-based immunity by activating macrophages and cytotoxic T cells, while Th2 cells promote antibody-mediated immunity and the recruitment of eosinophils. Th17 cells assist with extracellular pathogen clearance at mucosal surfaces and promote antibody production (12). Treg cells promote peripheral tolerance by secreting inhibitory cytokines, promoting cytolysis, and suppressing the activation, proliferation and cytokine production of CD4+ and CD8+ T cells (13).
2. T Cell Anergy and Co-Inhibitory Signaling
Several mechanisms exist to prevent immune overactivation and promote self-tolerance. Activation of T cells requires binding of a second co-stimulatory signal with stimulatory checkpoint molecules such as CD28 to B7-1/B7-2, without which T cells would remain anergic. Another class of immune checkpoint molecules induce co-inhibitory signals, including cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) and the programmed cell death protein −1 (PD-1) receptor.
CTLA-4 is an analog similar to CD28 that is constitutively expressed on Treg cell surfaces and upregulated on other T cell classes after initial activation. It exerts inhibitory effects in the early phases of T cell activation within lymphoid tissues by directly competing with CD28 to bind to B7-1 and B7-2 ligands on APCs with higher affinity. Its binding launches a signaling cascade preventing TCR signal transduction (14). Later studies showed a signaling-independent mechanism of inhibition via activation of trans-endocytosis that removes B7 ligands from the surfaces of APCs (15). CTLA-4 expression has been demonstrated in T cells of atherosclerotic plaque (16), and CTLA-4 inhibition has been linked to multiorgan lymphocyte infiltration, including the heart, leading to ICI fulminant myocarditis (17).
The PD-1 receptor is present on activated T cells and exerts an inhibitory effect via binding its two ligands PD-L1 (CD274) and PD-L2 (CD273) in peripheral tissues. While PD-L2 is present mostly on macrophages and dendritic cells, PD-L1 is expressed in hematopoietic cells and tissue cells in various organs, including in cardiomyocytes, vascular endothelial cells, and leukocytes within atherosclerotic plaque (18-20). The binding of PD-1 to PD-L1 or PD-L2 leads activation of the Akt pathway that decreases the production of inflammatory cytokines and cell survival proteins (21). Disruption in the PD-1/PD-L1 pathway in PD-1 knockout mice led to rapid onset autoimmune mediated dilated cardiomyopathy with diffuse IgG deposition, and is a potential mechanism of cardiotoxicity in ICI induced myocarditis (19).
3. Immune Checkpoint Alterations in Cancer and the Advent of ICIs
In the tumor microenvironment, tumor cells promote recruitment and production of CTLA-4 expressing Treg cells via TGF-β secretion. (22). In addition, prolonged activation of T cells in the cancer immune response leads to exhaustion, where T cells upregulate inhibitory molecules including PD-1 and CTLA-4 to limit their own proliferative potential (23) (Figure 1B). The surrounding tissue hypoxia promotes an increase in PD-L1 expression in various tumors (18).
Targeted ICIs utilize the altered oncologic expression of immune checkpoints to augment the host immune response, beginning with the approval of the CTLA-4 checkpoint inhibitor ipilimumab (Yervoy) for use in metastatic melanoma in 2011 (4). Shortly after came approval for PD-1 inhibitors pembrolizumab (Keytruda) and nivolumab (Opdivo) (3, 5) (Figure 1B). Since their initial FDA approval, the indications for ICIs have expanded rapidly to include NSCLCs, non-Hodgkin’s lymphoma, urothelial carcinoma, colorectal cancers, and more. Several investigational therapies targeting other immune checkpoints are in various stages of development (24). The pathways utilized in anti-CTLA-4 and PD-1/PD-L1 inhibitor therapies as well as those of investigational immune checkpoint agents have been implicated in atherogenesis (Central Illustration, Table 1).
Role of Immune Cells in Atherosclerosis and Effects of Current Immune Checkpoint Inhibitors
1. Overview of Atherogenesis
The pathophysiology of atherosclerosis begins with damaged endothelial cell expression of cellular adhesion molecules that attract monocytes to enter the subendothelium, where they transform into macrophages that are polarized into subtypes depending on their microenvironment. The classically activated pro-inflammatory phenotype is promoted by exposure to free fatty acids, oxidized lipids, and factors such as IFN-γ whereas the alternatively activated anti-inflammatory phenotype is promoted by factors such as IL-4, IL-13, and IL-10. Pro-inflammatory macrophages predominate in early atherosclerotic plaque and are responsible for the accumulation of intracellular lipids, formation of foam cells, and secretion proinflammatory cytokines such as IL-1β and IL-6 (25) (Figure 2). Anti-inflammatory macrophages promote collagen formation and effective lipid clearance, and are associated with atherosclerosis regression (26). Necrotic core formation is the hallmark of chronic inflammation in atherosclerosis, and is rich in lipids and cellular debris as a result of foam cell and vascular smooth muscle cell (VSMC) apoptosis and necrosis (27).
T cell activation is also integral to atherogenesis. During early atherosclerosis, APCs present atheroma-related antigens to naïve T cells in lymphoid tissues, leading to T cell migration towards the plaque (28). The initial recruitment of T cells is nonspecific, and they undergo further selective activation and clonal expansion into atherogenic T cell subtypes (28, 29). Atheroma-related T cell antigens have been difficult to identify, but include oxidized LDL (oxLDL) particles, heat shock proteins, and Apolipoprotein B (30). OxLDL are abundant in atheromatous plaque, and serves as an antigen triggering T cell autoimmune response that leads to inflammation and macrophage activation via IFN-γ secretion (31). CD8+ cytotoxic T cells, Th1 cells, Th2 cells, Th17 cells, and Treg cells have all been identified in atherosclerotic plaques (30).
2. Role of Differentiated T cells in Atherogenesis
Th1 cells are the predominant CD4+ T cell present in plaques (32) and the most directly associated with atherogenesis due to their production of inflammatory cytokines, including IFN-γ and TNF-α (33, 34) (Figure 2). The production of IFN-γ enhances recruitment of macrophages and T cells and promotes macrophage polarization, cytokine secretion and foam cell formation (35). IFN-γ is also thought to inhibit vascular smooth muscle cell proliferation, thereby contributing to decreased plaque stability (36). The atherogenic role of IFN-γ is most clearly demonstrated in a study where mice with dysfunctional IFN-γ receptors developed smaller and more phenotypically stable atheromas compared to controls (37). TNF-α promotes atherosclerosis through leukocyte recruitment, inflammatory cytokine production, and promotion of endothelial cell damage and oxidative stress (38). TNF-α deficient mice have been shown to exhibit smaller plaque lesions (33), and the presence of TNF-α is associated with increased lesion necrosis and more advanced plaque progression in mice (39). The inhibition of Th1 differentiation in mice is also atheroprotective and reduces the amount of IFN-γ detected in plaques (40).
Treg cells’ atheroprotective role has been well studied, and their effects are primarily mediated through secretion of TGF-β and IL-10 (Figure 2). TGF-β inhibits recruitment and activation of T-cells and macrophages, and increases plaque stability by promoting VSMC proliferation (41). Mice with defective TGF-β receptors have larger atherosclerotic lesions with increased T cell and macrophage composition, increased IFN-γ expression, and more vulnerable plaque phenotype (42). IL-10 reduces Th1 differentiation and prevents the recruitment and cytokine secretion of T cells and macrophages (43). A study with IL-10 deficient mice demonstrated increased susceptibility to atherosclerosis, and higher T cell infiltration and IFN-γ expression compared to controls (44). In addition, Treg secretion of IL-10 promotes the transformation of macrophages from the pro-inflammatory to the anti-inflammatory phenotype, potentiating its atheroprotective effect (45). Ait-Oufella et al. demonstrated that deficiency in Treg was associated with increased plaque size and more advanced plaque phenotype in mice, and that subsequent co-transference of Treg reduced inflammatory cell infiltration and plaque size (46). In addition, the number of Treg cells was inversely correlated with plaque vulnerability in human carotid arteries (47).
The roles of the remaining T cell subtypes in atherosclerosis are less well defined. Th2 cells oppose the production of IFN-γ, which suggested a potentially atheroprotective role. This was further supported as deficiency in IL-5, one of the primary Th2 secreted cytokines, was shown to accelerate atherosclerosis development in mice (48, 49). However, the deletion of another Th2 cytokine, IL-4, has been shown to decrease plaque lesion size in mice (50). Thus, the role of Th2 cells in atherogenesis remains unclear.
Th17 cells have received considerable attention in atherosclerosis in recent years, though its definitive role is still controversial. A study analyzing human coronary atherosclerotic plaques demonstrated that IL-17, the principal cytokine released by Th17, worked synergistically with IFN-γ to promote inflammation by increasing secretion of IL-6 (51). However, cytokine secretion of Th17 cells relies on environmental context, and subsets secrete IL-17 in conjunction with the anti-inflammatory IL-10, making the role of Th17 cells in atherosclerosis difficult to define (52). Several studies have indicated an atherogenic role to IL-17,(53) while others suggest an atheroprotective role and promotion of plaque stability via Type I collagen production (54, 55). Despite these inconclusive findings, the ratio between Th17 and Treg cells have been implicated in atherosclerosis progression, with increased Th17 and decreased Treg levels observed in patients with coronary atherosclerosis. Recent data suggests that inhibition of CD69, a key molecule expressed on early T lymphocytes regulating T cell differentiation, leads to elevated Th17/Treg ratios and exacerbation of atherosclerosis in mice (56). Indeed, the delicate balance between Th17 and Treg cells has important implications on autoimmune conditions and off-target adverse events related to ICI therapy, including myocarditis (57, 58).
Despite CD8+ T cells comprising about 50% of lymphocytes in atherosclerotic lesions (59), fewer studies have addressed the role of CD8+ cells. Cytotoxic CD8+ T cells have been shown to promote atherosclerosis in mouse models through IFN-γ secretion (60) and necrotic core formation by perforin and granzyme B-mediated apoptosis of macrophages (61). However, another mouse model suggest an atheroprotective role by promoting cytolysis of APCs (62).
Immune Checkpoint Therapies and Atherosclerosis
1. B7-1/B7-2 and CD28/CTLA-4 Pathway
The immunoregulatory effects of CTLA-4 and B7-1/B7-2 binding suggests an atheroprotective role for this pathway. The use of abatacept, a synthetic analog of CTLA-4, prevented CD4+ cell activation and reduced atherosclerosis development in murine femoral arteries by 78%, whereas the administration of CTLA-4 blocking antibodies increased atherosclerotic lesion sizes (63). Similarly, mice with elevated homocysteine levels had larger atherosclerotic plaque sizes associated with decreased membrane expression of CTLA-4, and pretreatment of these mice with abatacept ameliorated plaque development with reduced IFN-γ and IL-2 production and decreased macrophage content (64). Matsumoto et al. demonstrated that transgenic mice with constitutive CTLA-4 overexpression exhibited a significant reduction in atherosclerotic lesion size at the aortic root (16). In this study, CTLA-4 overexpression appeared to reduce atherosclerosis by decreasing plaque inflammation, as evidenced by a 38% decrease in macrophage accumulation and 42% decrease in CD4+ T cell infiltration, and downregulation of T cell proliferative capacity and proinflammatory cytokine production (16).
Modeling the role of ICIs in atherosclerosis, Poels et al. evaluated the role of antibody mediated CTLA-4 inhibition on atherogenesis. They demonstrated a two-fold increase in the size of atherosclerotic lesions in mice treated with CTLA-inhibiting antibodies, which was primarily mediated by a transition to an activated T cell profile without significant alterations in the macrophage inflammatory profile. CTLA-4 inhibition was also associated with plaque progression to more advanced phenotypes, with decreased collagen content and increased intimal thickening and necrotic core areas (65).
2. PD-1 and PD-L1/PD-L2 Pathway
PD-1 has been shown to suppress Th1 cytokine production and promote the development of Treg cells, suggesting a potentially atheroprotective role for this pathway (66). In hyperlipidemic mice, both PD-L1/PD-L2 deficiency and PD-1 receptor inhibition have been associated with increased atherosclerotic lesion size, increased plaque T cell activation , and enhanced TNF-α secretion (67, 68) . In contrast to CTLA-4 inhibition, PD-1 inhibition also exhibited increased lesion macrophage content, and enhanced the cytotoxicity of lesion CD8+ T cells (68). PD-1 deficiency has been shown to activate both CD4+ T cells and regulatory T cells, but the net effect was still exacerbated atherosclerotic lesion growth and increased plaque T cell infiltration (69).
In humans, several studies have observed decreased expression of PD-1 or its ligands in patients with coronary artery disease (CAD) and acute coronary syndrome (ACS), suggesting its protective role in atherogenesis and progression to advanced plaque phenotype (70, 71). At baseline, human atherosclerotic plaque T cells express high levels of PD-1 as a marker of T cell exhaustion in chronic inflammation (72). Thus, the delicate co-existence of PD-1 expressing T cells with activated T-cells within human plaques raises concern that PD-1 inhibition would lead to exacerbation of atherosclerosis.
Investigational Immune Checkpoint Targets and Atherosclerosis
1. CD40-CD40L Pathway
CD40 is a T cell costimulatory molecule and a member of the TNF receptor family that is present on APCs as well as non-immune cells such as endothelial cells, VSMCs, and platelets (73). Its classic ligand, CD40L, is expressed on activated CD4+ T cells. The binding of CD40 to CD40L promotes B cell proliferation and improves dendritic cell antigen presentation and T cell activation (74). CD40/CD40L ligation leads to recruitment of TNF receptor associated factors (TRAFs), that lead to downstream increases in atherogenic cytokines including IL-1β, IL-6, TNF-α and IFN-γ, which enhances the activities of macrophages (73).
Activation of the CD40/CD40L pathway has been shown to enhance antitumor immune response and promote tumor cell apoptosis and has been explored in cancer immunotherapy(75). Several ongoing Phase I and II clinical trials are evaluating the effect of anti-CD40 agonist antibodies, such as elicrelumab and dacetuzumab in various hematologic and solid tumor malignancies (76).
Several studies have explored role of the CD40/CD40L pathway in atherosclerosis. CD40L ligation to CD40 promotes thrombosis due to its activating effects on endothelial cells and platelets and its promotion of tissue factor production(77, 78). Endothelial cells exposed to CD40L exhibit increased expression of adhesion molecules such as VCAM-1 and P-selectin that enhances leukocyte recruitment, a key process in atherosclerosis initiation (79). CD40L has also been shown to promote foam cell formation in atherosclerotic plaques (80). Bruemmer et al. observed that in human iliac arteries, the expression of CD40 increased with atherosclerotic stage, and the largest increases occurred in the early stages of atherosclerosis (81). In later stages of atherosclerosis, the CD40/CD40L pathway stimulation increased secretion of matrix-degrading metalloproteinases that increase plaque rupture vulnerability (78).
Several preclinical studies have explored the inhibition of CD40/CD40L in reducing atherosclerosis. Blockade of CD40L either via antibodies or knock out models resulted in decreased leukocyte composition in the atherosclerotic plaque and phenotypic evidence of increased plaque stability (82-85). However, CD40L blockade was inconsistent in achieving plaque size reduction, where studies by Bavendiek et al. and Mach et al. observed reduced plaque size (82, 85), while two studies from Lutgens et al. did not (83, 84). In addition, CD40L inhibitory antibodies has been shown to increase the risk of thromboembolic events via platelet activation, limiting its systemic use in atherosclerosis treatment (86). Recent data from Lacy et al. demonstrated differing roles for CD40L depletion on atherosclerosis depending on cell source, with T cell CD40L inhibition leading to reduced Th1 polarization and IFN-γ secretion, whereas platelet-CD40L inhibition led to reduced atherothrombosis (87). These data hold promise for the development of cell-specific CD40L inhibition in the treatment of atherosclerosis.
With further research, CD40L/CD40 blockade may be pursued as a treatment for atherosclerosis. Conversely, with CD40/CD40L agonists being studied in clinical trials for cancer immunotherapy, their possible implications on cardiovascular disease must be carefully considered.
2. GITR-GITRL Pathway
Similar to CD40, glucocorticoid-induced TNF receptor-related protein (GITR) is a T cell costimulatory molecule present on T and NK cells and binds to its ligand, GITRL, expressed on APCs and endothelial cells (88). GITR-GITRL signaling promotes macrophage activity, increases CD8+ T cell cytotoxicity, and enhances effector T cell activation by increasing IFN-γ and IL-2 secretion (89, 90). Simultaneously, Treg cells experience a decrease in immunosuppressive effect downstream of GITR-GITRL interaction (89). The efficacy of tumor inhibition by GITR agonists was directly correlated with intratumor CD4+ and CD8+ levels, suggesting its therapeutic potential in NSCLC, renal cell carcinoma, and melanoma (91). Several phase I clinical trials have shown limited therapeutic response as monotherapy, but possible synergistic activity in combination with anti-PD-1 agents (92).
Macrophage and effector T cell activation and Treg suppression by the GITR/GITRL pathway suggests a proatherogenic role. GITR+ macrophages, T cells, and endothelial cells have been identified within atherosclerotic plaques (93, 94). Moreover, Shami et al. demonstrated increased GITR expression in carotid artery plaques in patients with symptomatic cerebrovascular disease, which correlated with increased plaque vulnerability (94). Kim et al. showed that GITR agonist antibodies promoted macrophage production of pro-atherogenic cytokines, plaque-destabilizing metalloproteinases, and intracellular cell adhesion molecule 1 (ICAM-1) (95). Conversely, GITR deficiency has been associated with decreased plaque size and increased plaque stability in mice (94). Thus, GITR agonism may be beneficial in combination with other ICIs such as PD-1 in cancer immunotherapy, but may worsen atherosclerosis, though studies of the direct effect of GITR agonistic monoclonal antibody on atherosclerosis are needed.
3. Other Costimulatory Pathways: CD27/CD70 and ICOS
Agonism of several other co-stimulatory pathways are being explored as immunotherapy. CD27 is a transmembrane glycoprotein that is part of the TNF superfamily that binds to CD70 (also known as CD27L) expressed on activated T cells. Its agonism enhances T cell expansion and promotion of memory T cells, with a predilection toward IFN-γ production (96). It has gained interest as a target for immunotherapy, with ongoing phase I/II trials for anti-CD27 agonists such as varlilumab (97). With its promotion of IFN-γ production, initial studies postulated that CD27/CD70 agonism would worsen atherogenesis. Surprisingly, chronic inflammation from CD70 overexpression was atheroprotective, as the CD27/CD70 pathway has shown important roles in macrophage efferocytosis, oxLDL clearing, and promotion of atheroma Treg survival (98, 99).
ICOS, or inducible costimulator, is a protein involved in CD4+ and CD8+ differentiation, survival and proliferation, promoting activation of cytotoxic T cells, enhancing effector T cell production of IL-4, IL-5, IL-10 and TNF-α, as well as promoting Treg activity (100). Given its prominent role in the survival of both pro-tumor and anti-tumor T cells, both ICOS agonists and antagonists are in early phase cancer immunotherapy trials (100). In preclinical studies, ICOS inhibition was associated with aggravated atherosclerosis via a reduced Treg population, as well as enhanced IFN-γ production and diminished IL-10 production (101, 102). Thus, contrary to CD40 and GITR, further study of CD27/CD70 and ICOS agonism may one day prove beneficial in atherosclerosis.
4. Investigational Co-Inhibitory Molecules: TIM-3 and LAG-3
TIM-3 is a negative regulatory immune checkpoint present on effector T cells, Treg cells, and dendritic cells. It binds to four separate ligands, most notably to galectin-9 that triggers Th1 and CD8+ T cell death (103, 104). Ongoing clinical trials are studying anti-TIM-3 antibodies such as cobolimab both as monotherapy or in combination with anti-PD-1 inhibitors(24).
Several preclinical studies have suggested that TIM-3 is a negative regulator of atherosclerosis. Foks et al. found that TIM-3 expression was increased in mice fed with atherogenic diet, and that TIM-3 antibody blockade was associated with larger atherosclerotic plaque size, increased lesion macrophage content, and decreased lesion Treg cells (105). In atherosclerotic lesions, CD8+ T cells exhibits co-expression of PD-1 and TIM-3, and inhibiting both immune checkpoints was associated with increase TNF-α and IFN-γ and decrease in IL-10 and IL-4) when compared to singular PD-1 or TIM-3 blockade (106). Thus, the development of anti-TIM-3 checkpoint inhibitors pose theoretical risks of exacerbating atherosclerosis.
LAG-3 is a cell surface protein analogous to CD4 present on T cells and dendritic cells that binds MHC class II molecules and serves as an inhibitory signal. It directly competes with TCR and CD4 binding to MHC class II, suppresses T cell expansion and promotes CD4+ T cell differentiation into Treg (107). LAG-3 targeting ICIs, such as relatlimab, are being investigated in Phase II and III clinical trials (108). One observational study of the Multiethnic Study of Atherosclerosis (MESA) cohort demonstrated that patients with CAD had higher LAG-3 levels, and LAG-3 was a significant coronary artery disease risk predictor (109). Despite this positive correlation, no causal relationship between LAG-3 and atherosclerosis has been established. Preclinical studies are needed to evaluate whether increased LAG-3 expression is a contributory or compensatory mechanism in atherosclerosis.
5. CD47-SIRPα Pathway
In addition to T cell mediated immune checkpoint inhibitors, novel classes of immune check therapies being explored include the macrophage-mediated immune checkpoint CD47. CD47 is an immunoglobulin-like molecule that binds to signal regulatory protein alpha (SIRPα) and impairs phagocytosis. In areas with high rates of apoptosis and cell turnover, such as within tumors and atherosclerotic necrotic cores, effective clearance of apoptotic cellular debris helps prevent further inflammatory response (110). This process, known as “efferocytosis” refers to programmed cell removal by which macrophages detect cell surface markers that signal phagocytosis, collectively termed “eat me” signals (111). By contrast, cells may express markers that impair phagocytosis, termed “don’t eat me” signals, such as CD47.
Kojima et al. demonstrated an upregulation of CD47 in both murine and human atherosclerotic plaque, particularly in the necrotic core (112). Treatment of atherosclerotic models with CD47 inhibitory antibodies markedly reduced atherosclerosis by restoring efferocytosis, evidenced by the enhanced removal of diseased vascular smooth muscles and macrophages in-vivo (112, 113). In addition, CD47 inhibition was shown to downregulate genes implicated in macrophage response to IL-1 and IFN-γ, leading to significant reduction in atherosclerotic inflammation in PET/CT imaging of mouse models (113). CD47 has also been suggested to reduce the ability of macrophages to remove opsonized targets, including the removal of opsonized clonal smooth muscle cells thought to give rise to the majority of cells in atherosclerotic plaques (114).
In recent years, CD47 inhibitor therapies have been used in clinical trials aimed at increasing tumor cell recognition and phagocytosis by macrophages. Magrolimab, the first-in-class anti-CD47 antibody, showed promising results in Phase 1B trials of relapsed and refractory non-Hodgkin’s lymphoma (115). It recently gained breakthrough therapy designation by the FDA, and ongoing trials are evaluating its efficacy in various hematologic and solid tumor malignancies. Interestingly, a small retrospective analysis on the non-Hodgkin’s lymphoma trial participants demonstrated a reduction of FDG uptake in the carotid arteries after 9 weeks of Magrolimab treatment, implying that CD47 inhibition reduces vascular inflammation (116). Thus, CD47 inhibition may be a shared pathway against both oncogenesis and atherogenesis.
Macrophage efferocytosis has also been shown to be influenced by Treg cell activity. Proto et al. showed that Treg cell expansion enhanced efferocytosis via expression of IL-13, which in turn increased IL-10 production in macrophages and led to apoptotic cell engulfment and clearance (117). Since existing PD-1 and CTLA-4 targeted ICI therapies have been suggested to suppress Treg activity, decreased efferocytosis may be another avenue by which other classes of immune checkpoint therapies may exacerbate atherosclerosis. However, extensive research is needed to substantiate this potential link.
Cholesterol Metabolism and T Cell Mediated Immunity in Cancer
Clinically, obesity and metabolic syndrome have been clearly linked with increased cancer risk, mediated by many shared risk factors (118). One shared link is energy metabolism, where alterations in lipid metabolism can shape the immune system’s response to tumor activity. Cholesterol metabolism is essential for cancer progression, including the formation of cellular membranes during rapid proliferation and in tumor migration and invasion (119).
The features of metabolic syndrome, including hypertriglyceridemia, hyperglycemia and hypercholesterolemia, promotes a chronic inflammatory state that paradoxically blocks physiologic immune function (120). Chronic inflammation and hypercholesterolemia promote an increase in the production of myeloid-derived suppressor cells, or MDSCs, through a process termed “emergency myelopoiesis” (121). MDSCs inhibit T cell functions through alterations in TCR receptors that impair downstream signaling, secretion of TGF-β, IL-10 and cytokines that decrease effector T cell function, and overexpression of PD-L1 (122). In addition, Ma et al. utilized an Apoe−/− mouse model traditionally used to study atherosclerosis to demonstrate negative regulatory effect of hypercholesterolemia on IL-9 levels, which impair CD8+ T cell differentiation and antitumor response (123). The enriched cholesterol content of the tumor microenvironment induces T cell exhaustion and the expression of immune checkpoints such as PD-1, TIM-3 and LAG-3 on CD8+ T cells (124).
Given the intricate relationship of hypercholesterolemia and repressed T cell function, it is anticipated that cholesterol levels modulate patient response to immunotherapy. Perrone et al. found that among patients receiving ICIs, baseline hypercholesterolemia was associated with improved overall survival rates (125). Other studies have established improved prognosis in obese patients treated with ICIs (126, 127). Thus it can be postulated that cancer patients with higher cholesterol levels may have worse T cell dysfunction by mechanisms described above, and subsequently exhibit a more pronounced response to the restoration of T cell function by immune checkpoint inhibition. However, these data are limited to date, and whether hypercholesterolemia is a causative factor in improved ICI response or simply a biomarker of chronic inflammation rendering patients more susceptible to immune restoration has yet to be established.
Cholesteryl esters are a storage form of cholesterol that are observed in increased numbers in the tumor microenvironment. One of the key enzymes in this process is ACAT1 (Acetyl-coenzyme A acetyltransferase), which promotes cholesterol esterification into the storage form and facilitates cholesterol transport in blood (128), has been studied in antitumor therapy and atherosclerosis. ACAT1 inhibition has been shown to decrease cancer cell migration and tumor progression in breast cancer and glioblastomas (129, 130). In the early stages of atherosclerosis, ACAT1 expression is increased in macrophages to enhance their ability to store free cholesterol, and the use of ACAT inhibitors had been shown to promote cell death (131). The ACAT1 inhibitor avasimibe has been studied in animals and humans with hyperlipidemia, however their effects on plasma cholesterol and atherosclerotic plaque size were inconsistent (132-134). Yang et al applied ACAT inhibition to cancer immunotherapy, and demonstrated that the addition of avasimibe to anti-PD-1 therapy was more effective in the treatment of melanoma in mice, with demonstrated enhancement in CD8+ T cell antitumor activity (135). Together, these findings imply an important relationship between hypercholesterolemia, a major driver of atherosclerosis, and tumor progression.
Clinical Implications of ICI Therapy on ASCVD
Recent clinical data have emerged to suggest an association between ICI use and accelerated atherosclerosis and atherosclerotic cardiovascular events. Initially, a case series of 11 patients suggested that PD-1 blockade may actually improve complicated plaque burden, as PD-1 blockade was associated with regression of atherosclerotic plaque in 3 patients (27%), no change in 7 (64%) and an increase in 1 (9%). However, this study was limited by a short CT scan follow up period and imaging protocols that were not intended for vascular imaging (136). Since, two case reports have linked PD-L1 and PD-1 inhibitors with rapid interval progression of CAD on left heart catheterization and fatal ACS in metastatic lung and giant bone cell cancer, respectively (137, 138).
Several single center studies have documented the incidence of ASCVD in ICI therapy. In a retrospective analysis of ICI clinical trials where PD-1 and PD-L1 inhibitors including nivolumab, pembrolizumab, atezolizumab, avelumab, and durvalumab were tested in patients with NSCLC, Hu et al. reported a 1% incidence of myocardial infarction and 2% incidence of stroke (139). In a single center registry by Oren et al. of 3,326 patients with any malignancy on ICI therapy including atezolizumab, avelumab, ipilimumab, nivolumab and pembrolizumab, there was a 7% incidence for both myocardial infarction and stroke within a 16 month period (140).
In the largest single center study to date, Drobni et al. compared 2842 patients on any ICIs (the majority of which being PD-1 inhibitors) with controls matched for age, cancer type and cardiovascular history on their risk of ASCVD related events over a 2 year follow up period. ICI use was associated with >4-fold greater risk for composite cardiovascular events (HR 4.7, [95% CI 3.5-6.2, p<0.001), >7 fold greater risk of myocardial infarction (HR 7.2 [95% CI 4.5-11.5, p<0.001], 3-fold greater risk of coronary revascularization (HR 3.3, [95% CI 2.0-5.5], p<0.001) and 4-fold greater risk of stroke (HR 4.6 [95% CI 2.9-7.2], p<0.001) (7). Small scale human imaging and histologic studies have attempted to substantiate the possibility that ICI therapy increases atheromatous inflammation and atherosclerosis. In Calabretta et al., PET scans from twenty melanoma patients treated with ICIs (80% PD-1 inhibitors, 5% CTLA-4 inhibitors and 15% combination) showed significant elevation in FDG uptake of all aortic segments about 6 months post-treatment, particularly in noncalcified and mildly calcified segments (p<0.001) (141). However, in a study by Poels et al., ten melanoma patients treated with pembrolizumab or combination nivolumab/ipilimumab showed no absolute change in overall vascular FDG-positive inflammation after 6 weeks.(142) These discrepant results may be due to small sample sizes and varying follow up periods. In a sub-study by Drobni et al., 40 melanoma patients were evaluated via CT at three different time points, and ICI treatment was shown to accelerate the rate of atherosclerotic plaque progression 3-fold (from 2.1%/yr pre-ICI to 6.7%/yr after ICI, p=0.02). In an autopsy study of 11 patients on ICI therapy, although the absolute number of T cells was unchanged, there was an increased ratio of CD3+ cells to CD68+ cells leading to a shift from macrophage- to lymphocyte-predominant inflammation (143). Future large-scale, long term imaging and histologic studies will be needed to further clarify the role of ICI on atheromatous plaque progression and immune response.
Role of Pharmacotherapy in Cardiovascular Risk Reduction in Patients Receiving ICIs
Growing clinical data signaling increased atherosclerosis in ICI therapy invite consideration of pharmacotherapies to reduce cardiovascular events in this patient population. In addition to lipid lowering, statins have been associated with plaque stabilization, endothelial dysfunction reversal, and inflammation reduction. How statins reduce inflammation are not fully understood, but may involve inhibition of beta-2 integrin leukocyte function antigen-1 (LFA-1), an adhesion molecule with a role in T cell activation (144). In Drobni et al. the markedly increased rate of aortic plaque progression associated with ICI therapy was attenuated by the use of statins, though no direct comparisons of cardiovascular events based on statin use were performed (7). PCSK9 inhibitors are an increasingly popular class of monoclonal antibodies that reduce serum LDL and atherosclerotic events in higher risk patients (145). In contrast to statins, there is much less known about potential anti-inflammatory effects of PCSK9 inhibitors.
The potential for cholesterol levels to modulate ICI response raises important considerations for the effects of these cholesterol lowering agents on ICI efficacy. Despite the correlation between hypercholesterolemia and improved outcomes of ICI therapy, statins and PCSK9 inhibitors have shown preliminary evidence of synergistic benefit when paired with ICI therapy independent of their cholesterol lowering effects. In particular, statins have been shown to inhibit protein prenylation leading to enhanced antigen presentation that may be synergistic with immunotherapy (146). Several clinical studies in patients with advanced NSCLC and malignant pleural mesothelioma treated with ICIs demonstrated that statins were associated with increased response rate, improved time to treatment failure, progression free and overall survival (147, 148). Liu et al. showed that proprotein convertase substilisin/kexin type 9 (PCSK9) inhibition with evolocumab synergized with anti-PD-1 therapy to suppress tumor growth by increasing the expression of MHC I proteins and enhancing lymphocyte proliferation into the tumor (149). These findings highlight the complex interplay between cholesterol metabolism and the immune system that requires further research, but demonstrate that statins and PCSK9 inhibitors have the potential to both enhance ICI efficacy and treat ICI related atherosclerosis. However, Drobni et al. recently demonstrated increased risk of myopathy in patients treated concurrently with statins and ICIs (150). Thus, further study is needed to confirm the safety of existing therapies and identify novel therapeutics in the treatment of ASCVD in the context of ICI use.
Conclusions
Numerous studies have demonstrated the role of various immune checkpoint pathways in atherogenesis, and emerging clinical studies have begun describing an association between ICI use and increased risk of ASCVD. Taken together, these reports indicate that the cardiovascular effects of immune checkpoint therapies extend beyond the rare incidences of myocarditis. As ASCVD is one of the most prominent causes of morbidity and mortality worldwide, if further clinical studies confirm this association, the risk of major adverse cardiovascular events in patients receiving ICIs must be carefully considered.
The current level evidence linking ICI therapy to atherogenesis and atherosclerotic events is not without its limitations. Preclinical studies thus far have not clearly delineated how immune checkpoint alteration will impact each stage of atherosclerosis. It has become increasingly recognized that the microenvironmental context of cell death and apoptosis is important in determining whether the net effect is atherogenic or atheroprotective (111). Thus, the effect of ICIs on atherosclerosis may prove to be more heterogeneous and evolve as clinical disease progresses. Careful mechanistic understanding of the effect of ICI at each stage of atherosclerosis is needed to determine the timing and need of interventions.
On a clinical level, the linkage between ICI use and atherosclerosis has only been suggested by smaller observational studies, and extensive further evidence with larger sample sizes and longer study duration are needed to confirm this association and estimate the rate of adverse cardiovascular events in this population. If future studies continue to demonstrate significant ASCVD burden in patients receiving ICIs, effort must be taken to identify atherosclerotic development and control reversible atherosclerosis risk factors. In this effort, integration of baseline and routine atherosclerotic imaging in these patients may help to quantify atherosclerotic burden.
Patients who develop ASCVD from ICI therapy represent a unique patient population that warrant our investigation of potential pharmacotherapeutic options. The role of statins, PCSK9 inhibitors and other pharmacotherapies should be explored in animal models, and randomized controlled trials are needed to prove their effectiveness and safety in this population.
Sources of Funding:
Dr. Stein-Merlob and Dr. Nayeri are supported by the National Institutes of Health Cardiovascular Scientist Training Program (grant T32HL007895). Dr. Sallam is supported by the AHA Transformational Project Award and the National Institutes of Health grant HL149766. Dr. Neilan is supported by a gift from A. Curt Greer and Pamela Kohlberg, and grants from the National Institutes of Health/National Heart, Lung, and Blood Institute grants R01HL130539, R01HL137562, K24HL150238.
Disclosures:
Dr. Neilan has been a consultant to and received fees from Amgen, H3-Biomedicine, and AbbVie outside of the current work. Dr. Neilan also reports consultant fees from Bristol Myers Squibb for a Scientific Advisory Board focused on myocarditis related to immune checkpoint inhibitors and grant funding from AstraZeneca on atherosclerosis with immune checkpoint inhibitors. Dr. Yang is supported by a grant from CSL Behring. The remaining authors have no disclosures.
Nonstandard Abbreviations
ASCVD Atherosclerotic cardiovascular disease
CANTOS Canakinumab Anti-Inflammatory Thrombosis Outcomes Study
COLCOT Colchicine Cardiovascular Outcomes Trial
CTLA-4 Cytotoxic T-lymphocyte associated protein 4
ICI Immune checkpoint inhibitor
IFN-γ Interferon gamma
LAG-3 Lymphocyte-activation gene 3
LoDoCo Low-Dose Colchicine
PD-1 Programmed cell death protein 1
PD-L Programmed cell death ligand
Figure 1 T Cell Activation and Effect of Immune Checkpoints.
A. T Cell Activation. Antigen presenting cells (APCs) present antigens via major histocompatibility complex (MHC) molecules that bind to T cell receptors (TCR) on naïve CD4+ and CD8+ T cells. T cell activation requires the binding of costimulatory molecules B7.1(CD80) or B7.2(CD86) on APCs to CD28. Naïve CD8+ T cells are activated into cytotoxic T cells that secrete perforins and granzymes, leading to apoptotic cascades in target cells. The presence of particular cytokines and growth factors in the microenvironment determine naïve CD4+ T cell differentiation.
B. Immune Alterations in Cancer and Effect of Immune Checkpoint Inhibitors. Prolonged inflammation in cancer leads to T cell exhaustion that promotes recruitment of Treg cells and increased expression of cytotoxic T-lymphocyte associated protein 4 (CTLA-4) and programmed cell death protein 1 receptor (PD-1) on T cells. Programmed cell death protein 1 receptor ligand (PD-L1) on tumor cells binds to PD-1 to decrease the production of inflammatory cytokines. Immune checkpoint inhibitors enhance tumor cell killing by targeting PD-1, PD-L1 and CTLA-4 to restore T cell activation and inflammatory cytokine production. Created with BioRender.
Figure 2 Atherogenesis and Effects of T Cell Activation.
Damaged endothelial cells express platelet endothelial cell adhesion molecules (PECAM), intracellular adhesion molecules (ICAM), vascular cell adhesion protein (VCAM) and selectins that recruit monocytes to the subendothelium. Monocytes are activated into macrophages that consume oxidized low density lipoproteins (ox-LDL) and produce inflammatory cytokines, leading to foam cell and necrotic core formation. T cells are recruited to the subendothelium and activated by APCs. Activation of T helper 1 (Th1) cells leads to tumor necrosis factor alpha (TNF-α) secretion that promotes monocyte recruitment and endothelial damage, and interferon gamma (IFN-γ) secretion that promotes macrophage activation. T cell activation decreases the function of regulatory T cells (Treg) that normally decreases T cell and monocyte recruitment via transforming growth factor-beta (TGF-β) secretion and reduces Th1 differentiation via interleukin-10 (IL-10) secretion. T cell activation is enhanced by immune checkpoint altering antibodies. Created with BioRender.
Central Illustration: Summary of Immune Checkpoint Alterations and their Effects on Atherogenesis.
CTLA-4 and PD-1/PDL-1 blockade are well-studied examples of co-inhibitory molecule blockade that worsen atherosclerosis. CD40/CD40L agonism and GITR agonism are examples of co-stimulatory molecule agonism that worsen atherosclerosis. CD-47 blockade restores efferocytosis and may be atheroprotective. APC indicates antigen presenting cell; MHC, major histocompatibility complex, TCR, T cell receptor, CTLA4, cytotoxic T-lymphocyte associated protein 4; PD-1, Programmed cell death protein 1 receptor; PD-L1, Programmed cell death protein 1 receptor ligand; SIRP-α, signal-regulatory protein alpha; CD, cluster of differentiation; GITR, glucocorticoid-induced tumor necrosis factor family-related protein; GITRL, glucocorticoid-induced tumor necrosis factor family-related protein ligand; Treg, regulatory T cell; VCAM, vascular cell adhesion molecule; and MMP, matrix metalloproteinase. Created with BioRender.
Table 1: Summary of Immune Checkpoint Pathway Alterations and Effects on Cancer Therapy and ASCVD
Immune
Checkpoint
Class Immune
Checkpoint
Target Effect on Cancer Therapy† Effect on ASCVD† Example Therapies
Co-Inhibitory Signal Blockade CTLA-4 Restores T cell activation and enhances cancer cell destruction (3, 24) -Increases atherosclerotic lesion size (63-65)
-Progression to advanced phenotype (65)
-Decreases collagen content (65) Ipilimumab (4)
Tremelimumab (151)*
Zalifrelimab (AGEN1884)* (152)
PD-1 and PD-L1 -Increases atherosclerotic lesion size (67-69)
-Enhances TNF-α secretion (67, 68)
-Enhances lesion helper T cell, cytotoxic T cell and macrophage activation (67-69) Anti-PD-1: nivolumab (5), pembrolizumab (3), cemiplimab (153), spartalizumab* (154), camrelixumab* (154), tislelizumab* (154)
Anti- PD-L1: atezolizumab (155), avelumab (156), durvalumab (157), cosebelimab*(154),sugemalimab*(154), CX-072*(154)
TIM-3* -Increases atherosclerotic lesion size (105)
-Increases lesion macrophage content
-Decreases lesion Treg content
-Enhances TNF-α and IFN-γ secretion in combination with anti-PD-1 in human cells (106) Cobolimab* (24) BMS-986258*(158), MBG453* (158), TSR-022*(158), Sym023*(158), BGB-A425* (158)
LAG-3* -Effect unknown, a marker positively correlated with ASCVD in a human observational study (109) Relatlimab*, Eftilagimod alpha*, Tebotelimab*, Favezelimab*, LAG525*, REGN 3767* (159)
Co-Stimulatory Signal Agonism GITR* -Enhances macrophage activity (90)
-Promotes CD4+ and CD8+ T cell activation (89, 90)
-Decreases immunosuppressive ability of Treg (89) -Promotes macrophage secretion of atherogenic cytokines (94, 95)
-Increases MMP production (95)
-Enhances plaque leukocyte recruitment via ICAM-1 (95)
-GITR inhibition decreased plaque size and increased plaque stability (94) AMG228*, MEDI-1873*, BMS-986156* (160)
CD40 and CD40L* -Enhances antigen presentation by APCs (74)
-Promotes B cell proliferation and antibody production (74)
-Promotes tumor cell apoptosis (75) -Increases secretion of IL-1β, IL-6, TNF-α, IFN-γ (73)
-Endothelial cell and smooth muscle cell activation in human cells (78)
-Increases leukocyte recruitment by expression of VCAM-1 and P-selectin (77)
-Increases MMP production (77)
-Promotes foam cell production (77)
Inhibition of CD40L:
-increased plaque stability (82-85)
- inconsistently decreased plaque size (82, 85) vs no change (83, 84)
-reduced Th1 polarization and IFN-γ secretion (87)
-reduced atherothrombosis (87) Selicrelumab* (161), dazetuzumab* (162), CP-870893*(162) , JNJ-107*(162), APX005M (162)*
CD27 and CD70* -Enhanced T cell expansion and survival (96)
-Enhanced memory T cell production (96)
-Polarization toward IFN-γ producing effector T cells (96) -decreased lesion size (98, 99)
-enhanced macrophage efferocytosis (98)
-enhanced oxLDL clearing (98)
-promotion of Treg survival (98, 99) Anti-CD27:
Varlilumab*, AMG-172*
Anti-CD70:
Cusatuzumab*, BMS-936561* (163)
ICOS and ICOSL* -Promotion of cytotoxic T cells (100)
-Enhancing effector T cell function (100)
-Promotion of Treg activity (100) Inhibition of ICOS:
- increased atherosclerotic plaque size (101, 102)
-reduced atherosclerotic Treg population (102)
-increased IFN-γ secretion (101)
-decreased IL-10 secretion (101) Anti- ICOS Agonists:
Feladilimab*,vopratelimab (100)
Anti- ICOS Antagonists:
MEDI-570*, KY1044 *(100)
“Don’t eat me” Blockade CD47* -Blockade of interaction with SIRP-α (111)
-Enhances phagocytosis of apoptotic cells and debris (111)
-Reduces inflammation (111) -Reduces atherosclerotic lesion size and inflammation (112, 113)
-Decreases macrophage response to IL-1β and IFN-γ (113)
-Improves clearance of VSMCs (114) Magrolimab* (115), SRF231*(164), AO-176*(164), CC-90002* (164)
* denotes investigational therapies.
† denotes findings of preclinical/animal studies unless otherwise specified. AGEN indicates Agenus; AMG, Amgen; AO, Arch Oncology; APX, Apexigen; APCs, antigen presenting cells; ASCVD, atherosclerotic cardiovascular disease; BGB-A, BeiGene; BMS, Bristol Myers Squibb; CC, Celgene; CD, cluster of differentiation; CTLA-4, cytotoxic T-lymphocyte associated protein; CX, CytomX; GITR, glucocorticoid-induced tumor necrosis factor receptor-related protein; ICAM-1, intracellular adhesion molecule 1; IFN-γ, interferon gamma; IL, interleukin; JNJ, Johnson & Johnson; KY, Kymab; LAG-3, lymphocyte-activation gene 3; MMP, matrix metalloproteinase; PD-1, programmed cell death protein 1; PD-L1, programmed cell death ligand 1; REGN, Regeneron Pharmaceuticals; SIRP-α, signal regulatory protein alpha; SRF, Surface Oncology; Sym, Symphogen; TIM-3, T cell immunoglobulin and mucin-domain-containing-3; TNF-α, tumor necrosis factor alpha; Treg, regulatory T cell; VCAM-1, vascular cell adhesion molecule 1; and VSMCs, vascular smooth muscle cells.
Highlights
Immune checkpoint therapy may increase atherosclerotic cardiovascular disease (ASCVD).
Cholesterol metabolism modulates the response to immune checkpoint therapies and may be a common target for treatment of both cancer and atherosclerosis.
Further research should define cardiovascular endpoints for patients receiving immune checkpoint therapies and standardize treatment and surveillance strategies.
References
1. Dunn GP , Bruce AT , Ikeda H , Old LJ , Shreiber RD . Cancer Immunoediting: From Surveillance to Escape. Cancer Immunother. Immune Suppr. Tumor Growth Second Ed. 2013;3 :85–99.
2. Spranger S , Gajewski TF . Impact of oncogenic pathways on evasion of antitumour immune responses. Nat. Rev. Cancer 2018;18 :139–147. Available at: 10.1038/nrc.2017.117.29326431
3. Balar AV , Castellano D , O’Donnell PH , First-line pembrolizumab in cisplatin-ineligible patients with locally advanced and unresectable or metastatic urothelial cancer (KEYNOTE-052): a multicentre, single-arm, phase 2 study. Lancet Oncol. 2017;18 :1483–1492. Available at: 10.1016/S1470-2045(17)30616-2.28967485
4. Hodi SF , O’Day SJ , McDermott DF , Improved Survival with Ipilimumab in Patients with Metastatic Melanoma. N. Engl. J. Med 2010;363 :711–723.20525992
5. Rizvi NA , Mazières J , Planchard D , Activity and safety of nivolumab, an anti-PD-1 immune checkpoint inhibitor, for patients with advanced, refractory squamous non-small-cell lung cancer (CheckMate 063): A phase 2, single-arm trial. Lancet Oncol. 2015;16 :257–265.25704439
6. Palaskas N , Lopez-Mattei J , Durand JB , Iliescu C , Deswal A . Immune Checkpoint Inhibitor Myocarditis: Pathophysiological Characteristics, Diagnosis, and Treatment. J. Am. Heart Assoc 2020;9 :1–12.
7. Drobni ZD , Alvi RM , Taron J , Association between Immune Checkpoint Inhibitors with Cardiovascular Events and Atherosclerotic Plaque. Circulation 2020:2299–2311.33003973
8. Ridker PM , Everett BM , Thuren T , Antiinflammatory Therapy with Canakinumab for Atherosclerotic Disease. N. Engl. J. Med 2017;377 :1119–1131.28845751
9. Nidorf SM , Eikelboom JW , Budgeon CA , Thompson PL . Low-dose colchicine for secondary prevention of cardiovascular disease. J. Am. Coll. Cardiol 2013;61 :404–410.23265346
10. Tardif J-C , Kouz S , Waters DD , Efficacy and Safety of Low-Dose Colchicine after Myocardial Infarction. N. Engl. J. Med 2019;381 :2497–2505.31733140
11. Wolf D , Ley K . Immunity and Inflammation in Atherosclerosis. Circ. Res 2019;124 :315–327.30653442
12. Miossec P , Korn T , Kuchroo VK . Interleukin-17 and Type 17 Helper T Cells. N. Engl. J. Med 2009;361 :888–898.19710487
13. Vignali DAA , Collison LW , Workman CJ . How regulatory T cells work. Nat. Rev. Immunol 2008;8 :523–532.18566595
14. Pardoll DM . The blockade of immune checkpoints in cancer immunotherapy. Nat. Rev. Cancer 2012;12 :252–264.22437870
15. Qureshi OS , Zheng Y , Nakamura K , UKPMC Funders Group cell extrinsic function of CTLA-4. Science (80-. ). 2011;332 :600–603.
16. Matsumoto T , Sasaki N , Yamashita T , Overexpression of cytotoxic T-lymphocyte-associated antigen-4 prevents atherosclerosis in mice. Arterioscler. Thromb. Vasc. Biol 2016;36 :1141–1151.27055906
17. Brunner-Weinzierl MC , Rudd CE . CTLA-4 and PD-1 control of T-cell motility and migration: Implications for tumor immunotherapy. Front. Immunol 2018;9 :1–8.29403488
18. Sun C , Mezzadra R , Schumacher TN . Regulation and Function of the PD-L1 Checkpoint. Immunity 2018;48 :434–452. Available at: 10.1016/j.immuni.2018.03.014.29562194
19. Nishimura H , Okazaki T , Tanaka Y , Autoimmune dilated cardiomyopathy in PD-1 receptor-deficient mice. Science (80-. ). 2001;291 :319–322.
20. Tajiri K , Ieda M . Cardiac Complications in Immune Checkpoint Inhibition Therapy. Front. Cardiovasc. Med 2019;6 :1–7.30740396
21. Kim ES , Kim JE , Patel MA , Mangraviti A , Ruzevick J , Lim M . Immune Checkpoint Modulators: An Emerging Antiglioma Armamentarium. J. Immunol. Res 2016;2016 .
22. Zou W Regulatory T cells, tumour immunity and immunotherapy. Nat. Rev. Immunol 2006;6 :295–307.16557261
23. Baitsch L , Baumgaertner P , Devêvre E , Exhaustion of tumor-specific CD8+ T cells in metastases from melanoma patients. J. Clin. Invest 2011;121 :2350–2360.21555851
24. Qin S , Xu L , Yi M , Yu S , Wu K , Luo S . Novel immune checkpoint targets: Moving beyond PD-1 and CTLA-4. Mol. Cancer 2019;18 :1–14.30609930
25. Leitinger N , Schulman IG . Phenotypic polarization of macrophages in atherosclerosis. Arterioscler. Thromb. Vasc. Biol 2013;33 :1120–1126.23640492
26. Barrett TJ . Macrophages in Atherosclerosis Regression. Arterioscler. Thromb. Vasc. Biol 2020;40 :20–33.31722535
27. Bentzon JF , Otsuka F , Virmani R , Falk E . Mechanisms of plaque formation and rupture. Circ. Res 2014;114 :1852–1866.24902970
28. Robertson AKL , Hansson GK . T cells in atherogenesis: For better or for worse? Arterioscler. Thromb. Vasc. Biol 2006;26 :2421–2432.16973967
29. Paulsson G , Zhou X , Törnquist E , Hansson GK . Oligoclonal T cell expansions in atherosclerotic lesions of apolipoprotein E-deficient mice. Arterioscler. Thromb. Vasc. Biol 2000;20 :10–17.10634795
30. Saigusa R , Winkels H , Ley K . T cell subsets and functions in atherosclerosis. Nat. Rev. Cardiol 2020;17 :387–401. Available at: 10.1038/s41569-020-0352-5.32203286
31. Stemme S , Faber B , Holm J , Wiklund O , Witztum JL , Hansson GK . T lymphocytes from human atherosclerotic plaques recognize oxidized low density lipoprotein. Proc. Natl. Acad. Sci. U. S. A 1995;92 :3893–3897.7732003
32. Frostegård J , Ulfgren AK , Nyberg P , Cytokine expression in advanced human atherosclerotic plaques: Dominance of pro-inflammatory (Th1) and macrophage-stimulating cytokines. Atherosclerosis 1999;145 :33–43.10428293
33. Brånén L , Hovgaard L , Nitulescu M , Bengtsson E , Nilsson J , Jovinge S . Inhibition of tumor necrosis factor-α reduces atherosclerosis in apolipoprotein E knockout mice. Arterioscler. Thromb. Vasc. Biol 2004;24 :2137–2142.15345516
34. Rocha VZ , Folco EJ , Sukhova G , Interferon-γ, a Th1 cytokine, regulates fat inflammation: A role for adaptive immunity in obesity. Circ. Res 2008;103 :467–476.18658050
35. Gotsman I , Lichtman AH . Targeting interferon-γ to treat atherosclerosis. Circ. Res 2007;101 :333–334.17702979
36. Amento EP , Ehsani N , Palmer H , Libby P . Cytokines and growth factors positively and negatively regulate interstitial collagen gene expression in human vascular smooth muscle cells. Arterioscler. Thromb. Vasc. Biol 1991;11 :1223–1230.
37. Gupta S , Pablo AM , Jiang XC , Wang N , Tall AR , Schindler C . IFN-γ, potentiates atherosclerosis in ApoE knock-out mice. J. Clin. Invest 1997;99 :2752–2761.9169506
38. McKellar GE , McCarey DW , Sattar N , McInnes IB . Role for TNF in atherosclerosis? Lessons from autoimmune disease. Nat. Rev. Cardiol 2009;6 :410–417. Available at: 10.1038/nrcardio.2009.57.19421244
39. Boesten LSM , Zadelaar ASM , Van Nieuwkoop A , Tumor necrosis factor-α promotes atherosclerotic lesion progression in APOE*3-leiden transgenic mice. Cardiovasc. Res 2005;66 :179–185.15769461
40. Laurat E , Poirier B , Tupin E , In vivo downregulation of T helper cell 1 immune responses reduces atherogenesis in apolipoprotein E-knockout mice. Circulation 2001;104 :197–202.11447086
41. Grainger DJ . Transforming Growth Factor β and Atherosclerosis: So Far, So Good for the Protective Cytokine Hypothesis. Arterioscler. Thromb. Vasc. Biol 2004;24 :399–404.14699019
42. Robertson AKL , Rudling M , Zhou X , Gorelik L , Flavell RA , Hansson GK . Disruption of TGF-β signaling in T cells accelerates atherosclerosis. J. Clin. Invest 2003;112 :1342–1350.14568988
43. Ou H xiao , Guo B bing , Liu Q , Regulatory T cells as a new therapeutic target for atherosclerosis. Acta Pharmacol. Sin 2018;39 :1249–1258.29323337
44. Mallat Z , Besnard S , Duriez M , Protective role of interleukin-10 in atherosclerosis. Circ. Res 1999;85 .
45. Lin J , Li M , Wang Z , He S , Xuming M , Li D . The role of CD4+CD25+ regulatory T cells in macrophage-derived foam-cell formation. J. Lipid Res 2010;51 :1208–1217.20007839
46. Ait-Oufella H , Salomon BL , Potteaux S , Natural regulatory T cells control the development of atherosclerosis in mice. Nat. Med 2006;12 :178–180.16462800
47. Dietel B , Cicha I , Voskens CJ , Verhoeven E , Achenbach S , Garlichs CD . Decreased numbers of regulatory T cells are associated with human atherosclerotic lesion vulnerability and inversely correlate with infiltrated mature dendritic cells. Atherosclerosis 2013;230 :92–99. Available at: 10.1016/j.atherosclerosis.2013.06.014.23958259
48. Binder CJ , Hartvigsen K , Chang MK , IL-5 links adaptive and natural immunity specific for epitopes of oxidized LDL and protects from atherosclerosis. J. Clin. Invest 2004;114 :427–437.15286809
49. Ketelhuth DFJ , Hansson GK . Adaptive Response of T and B Cells in Atherosclerosis. Circ. Res 2016;118 :668–678.26892965
50. King VL , Szilvassy SJ , Daugherty A . Interleukin-4 deficiency decreases atherosclerotic lesion formation in a site-specific manner in female LDL receptor−/− mice. Arterioscler. Thromb. Vasc. Biol 2002;22 :456–461.11884290
51. Eid RE , Rao DA , Zhou J , Interleukin-17 and interferon-γ Are produced concomitantly by human coronary artery-infiltrating T cells and act synergistically on vascular smooth muscle cells. Circulation 2009;119 :1424–1432.19255340
52. Taleb S , Tedgui A , Mallat Z . IL-17 and Th17 cells in atherosclerosis: Subtle and contextual roles. Arterioscler. Thromb. Vasc. Biol 2015;35 :258–264.25234818
53. Smith E , Prasad KMR , Butcher M , Blockade of interleukin-17A results in reduced atherosclerosis in apolipoprotein E-deficient mice. Circulation 2010;121 :1746–1755.20368519
54. Gisterå A , Robertson AKL , Andersson J , Transforming growth factor-β signaling in T cells promotes stabilization of atherosclerotic plaques through an interleukin-17-dependent pathway. Sci. Transl. Med 2013;5 :18–23.
55. Danzaki K , Matsui Y , Ikesue M , Interleukin-17A deficiency accelerates unstable atherosclerotic plaque formation in apolipoprotein e-deficient mice. Arterioscler. Thromb. Vasc. Biol 2012;32 :273–280.22116098
56. Tsilingiri K , De La Fuente H , Relaño M , Oxidized Low-Density Lipoprotein Receptor in Lymphocytes Prevents Atherosclerosis and Predicts Subclinical Disease. Circulation 2019;139 :243–255.30586697
57. Knochelmann HM , Dwyer CJ , Bailey SR , When worlds collide: Th17 and Treg cells in cancer and autoimmunity. Cell. Mol. Immunol 2018;15 :458–469. Available at: 10.1038/s41423-018-0004-4.29563615
58. Rikhi R , Karnuta J , Hussain M , Immune Checkpoint Inhibitors Mediated Lymphocytic and Giant Cell Myocarditis: Uncovering Etiological Mechanisms. Front. Cardiovasc. Med 2021;8 :1–9.
59. Gewaltig J , Kummer M , Koella C , Cathomas G , Biedermann BC . Requirements for CD8 T-cell migration into the human arterial wall. Hum. Pathol 2008;39 :1756–1762. Available at: 10.1016/j.humpath.2008.04.018.18706675
60. Kolbus D , Ramos OH , Berg KE , CD8+T cell activation predominate early immune responses to hypercholesterolemia in Apoe−/−mice. BMC Immunol. 2010;11 :58. Available at: http://www.biomedcentral.com/1471-2172/11/58. 21126329
61. Kyaw T , Winship A , Tay C , Cytotoxic and proinflammatory CD8+ T lymphocytes promote development of vulnerable atherosclerotic plaques in ApoE-deficient mice. Circulation 2013;127 :1028–1039.23395974
62. Chyu KY , Zhao X , Dimayuga PC , CD8 + T cells mediate the athero-protective effect of immunization with an ApoB-100 peptide. PLoS One 2012;7 .
63. Ewing MM , Karper JC , Abdul S , T-cell co-stimulation by CD28-CD80/86 and its negative regulator CTLA-4 strongly influence accelerated atherosclerosis development. Int. J. Cardiol 2013;168 :1965–1974. Available at: 10.1016/j.ijcard.2012.12.085.23351788
64. Ma K , Lv S , Liu B , CTLA4-IgG ameliorates homocysteine-accelerated atherosclerosis by inhibiting T-cell overactivation in apoE−/− mice. Cardiovasc. Res 2013;97 :349–359.23118130
65. Poels K , van Leent MMT , Reiche ME , Antibody-Mediated Inhibition of CTLA4 Aggravates Atherosclerotic Plaque Inflammation and Progression in Hyperlipidemic Mice. Cells 2020;9 .
66. Latchman Y , Wood CR , Chernova T , PD-L2 is a second ligand for PD-1 and inhibits T cell activation. Nat. Immunol 2001;2 :261–268.11224527
67. Gotsman I , Grabie N , Dacosta R , Sukhova G , Sharpe A , Lichtman AH . Proatherogenic immune responses are regulated by the PD-1/PD-L pathway in mice. J. Clin. Invest 2007;117 :2974–2982.17853943
68. Bu DX , Tarrio M , Maganto-Garcia E , Impairment of the programmed cell death-1 pathway increases atherosclerotic lesion development and inflammation. Arterioscler. Thromb. Vasc. Biol 2011;31 :1100–1107.21393583
69. Cochain C , Chaudhari SM , Koch M , Wiendl H , Eckstein HH , Zernecke A . Programmed cell death-1 deficiency exacerbates T cell activation and atherogenesis despite expansion of regulatory T cells in atherosclerosis-prone mice. PLoS One 2014;9 :1–10.
70. Lee J , Zhuang Y , Wei X , Contributions of PD-1/PD-L1 pathway to interactions of myeloid DCs with T cells in atherosclerosis. J. Mol. Cell. Cardiol 2009;46 :169–176. Available at: 10.1016/j.yjmcc.2008.10.028.19056397
71. Li S-H , Chen W-J , Yan M , Shu Y-W , Liao Y-H . Expression of coinhibitory PD-L1 on CD4+CD25+FOXP3+ regulatory T cells is elevated in patients with acute coronary syndrome. Coron. Artery Dis 2015;26 . Available at: https://journals.lww.com/coronary-artery/Fulltext/2015/11000/Expression_of_coinhibitory_PD_L1_on.10.aspx.
72. Fernandez DM , Rahman AH , Fernandez N , Single-cell immune landscape of human atherosclerotic plaques. Nat Med 2019;25 :1576–1588.31591603
73. Michel NA , Zirlik A , Wolf D . CD40L and Its Receptors in Atherothrombosis—An Update. Front. Cardiovasc. Med 2017;4 .
74. Li DK , Wang W . Characteristics and clinical trial results of agonistic anti-CD40 antibodies in the treatment of malignancies (Review). Oncol. Lett 2020;20 .
75. Eliopoulos AG , Young LS . The role of the CD40 pathway in the pathogenesis and treatment of cancer. Curr. Opin. Pharmacol 2004;4 :360–367.15251129
76. Richards DM , Sefrin JP , Gieffers C , Hill O , Merz C . Concepts for agonistic targeting of CD40 in immuno-oncology. Hum. Vaccines Immunother 2020;16 :377–387. Available at: 10.1080/21645515.2019.1653744.
77. Lievens D , Eijgelaar WJ , Biessen EAL , Daemen MJAP , Lutgens E . The multi-functionality of CD40L and its receptor CD40 in atherosclerosis. Thromb. Haemost 2009;102 :206–214.19652870
78. Mach F , Schoneck U , Bonnefoy J-Y , Pober JS , Libby P . Activation of Monocyte/Macrophage Functions Related to Acute Atheroma Complication by Ligation of CD40. Circulation 1997;96 :396–399.9244201
79. Kotowicz K , Dixon GLJ , Klein NJ , Peters MJ , Callard RE . Biological function of CD40 on human endothelial cells: Costimulation with CD40 ligand and interleukin-4 selectively induces expression of vascular cell adhesion molecule-1 and P-selectin resulting in preferential adhesion of lymphocytes. Immunology 2000;100 :441–448.10929070
80. Yuan M , Fu H , Ren L , Wang H , Guo W . Soluble CD40 ligand promotes macrophage foam cell formation in the etiology of atherosclerosis. Cardiol. 2015;131 :1–12.
81. Bruemmer D , Riggers U , Holzmeister J , Expression of CD40 in vascular smooth muscle cells and macrophages is associated with early development of human atherosclerotic lesions. Am. J. Cardiol 2001;87 :21–27.11137828
82. Mach F , Schönbeck U , Sukhova GK , Atkinson E , Libby P . Reduction of atherosclerosis in mice by inhibition of CD40 signalling. Nature 1998;394 :200–203.9671306
83. Lutgens E , Cleutjens KBJM , Heeneman S , Koteliansky VE , Burkly LC , Daemen MJAP . Both early and delayed anti-CD40L antibody treatment induces a stable plaque phenotype. Proc. Natl. Acad. Sci. U. S. A 2000;97 :7464–7469.10861013
84. Lutgens E , Gorelik L , Daemen MJAP , Requirement for CD154 in the progression of atherosclerosis. Nat. Med 1999;5 :1313–1316.10546000
85. Bavendiek U , Zirlik A , LaClair S , MacFarlane L , Libby P , Schönbeck U . Atherogenesis in mice does not require CD40 ligand from bone marrow-derived cells. Arterioscler. Thromb. Vasc. Biol 2005;25 :1244–1249.15746436
86. Robles-Carrillo L , Meyer T , Hatfield M , Anti-CD40L Immune Complexes Potently Activate Platelets In Vitro and Cause Thrombosis in FCGR2A Transgenic Mice. J. Immunol 2010;185 :1577–1583.20585032
87. Lacy M , Bürger C , Shami A , Cell-specific and divergent roles of the CD40L-CD40 axis in atherosclerotic vascular disease. Nat. Commun 2021;12 :1–12. Available at: 10.1038/s41467-021-23909-z.33397941
88. Nocentini G , Riccardi C . Therapeutic Targets of the TNF Super Family. 2009.
89. Mahne AE , Mauze S , Joyce-Shaikh B , Dual roles for regulatory T-cell depletion and costimulatory signaling in agonistic GITR targeting for tumor immunotherapy. Cancer Res. 2017;77 :1108–1118.28122327
90. Gerdes N , Zirlik A . Co-stimulatory molecules in and beyond co-stimulation-tipping the balance in atherosclerosis? Thromb. Haemost 2011;106 :804–813.21979444
91. Vence L , Bucktrout SL , Curbelo IF , Characterization and comparison of GITR expression in solid tumors. Clin. Cancer Res 2019;25 :6501–6510.31358539
92. Zappasodi R , Sirard C , Li Y , Rational design of anti-GITR-based combination immunotherapy. Nat. Med 2019;25 :759–766. Available at: 10.1038/s41591-019-0420-8.31036879
93. Kim WJ , Kang YJ , Suk K , Park JE , Kwon BS , Lee WH . Comparative analysis of the expression patterns of various TNFSF/TNFRSF in atherosclerotic plaques. Immunol. Invest 2008;37 :359–373.18569075
94. Shami A , Atzler D , Bosmans LA , Glucocorticoid-induced tumour necrosis factor receptor family-related protein (GITR) drives atherosclerosis in mice and is associated with an unstable plaque phenotype and cerebrovascular events in humans. Eur. Heart J 2020;41 :2938–2948.32728688
95. Kim WJ , Bae EM , Kang YJ , Glucocorticoid-induced tumour necrosis factor receptor family related protein (GITR) mediates inflammatory activation of macrophages that can destabilize atherosclerotic plaques. Immunology 2006;119 :421–429.17067317
96. Van De Ven K , Borst J . Targeting the T-cell co-stimulatory CD27/CD70 pathway in cancer immunotherapy: Rationale and potential. Immunotherapy 2015;7 :655–667.26098609
97. Ansell SM , Flinn I , Taylor MH , Safety and activity of varlilumab, a novel and first-in-class agonist anti-CD27 antibody, for hematologic malignancies. Blood Adv. 2020;4 :1917–1926.32380537
98. Winkels H , Meiler S , Smeets E , CD70 limits atherosclerosis and promotes macrophage function. Thromb. Haemost 2017;117 :164–175.27786334
99. Winkels H , Meiler S , Lievens D , CD27 co-stimulation increases the abundance of regulatory T cells and reduces atherosclerosis in hyperlipidaemic mice. Eur. Heart J 2017;38 :3590–3599.29045618
100. Solinas C , Gu-Trantien C , Willard-Gallo K . The rationale behind targeting the ICOS-ICOS ligand costimulatory pathway in cancer immunotherapy. ESMO Open 2020;5 :e000544. Available at: 10.1136/esmoopen-2019-000544.32516116
101. Afek A , Harats D , Roth A , Keren G , George J . A functional role for inducible costimulator (ICOS) in atherosclerosis. Atherosclerosis 2005;183 :57–63.15941568
102. Gotsman I , Grabie N , Gupta R , Impaired regulatory T-cell response and enhanced atherosclerosis in the absence of inducible costimulatory molecule. Circulation 2006;114 :2047–2055.17060381
103. Kang CW , Dutta A , Chang LY , Apoptosis of tumor infiltrating effector TIM-3+CD8+ T cells in colon cancer. Sci. Rep 2015;5 :1–12.
104. Chiba S , Baghdadi M , Akiba H , Tumor-infiltrating DCs suppress nucleic acid-mediated innate immune responses through interactions between the receptor TIM-3 and the alarmin HMGB1. Nat. Immunol 2012;13 :832–842.22842346
105. Foks AC , Ran IA , Wasserman L , T-cell immunoglobulin and mucin domain 3 acts as a negative regulator of atherosclerosis. Arterioscler. Thromb. Vasc. Biol 2013;33 :2558–2565.23990206
106. Qiu MK , Wang SC , Dai YX , Wang SQ , Ou JM , Quan ZW . PD-1 and Tim-3 pathways regulate CD8+ T cells function in atherosclerosis. PLoS One 2015;10 :1–11.
107. Long L , Zhang X , Chen F , The promising immune checkpoint LAG-3: From tumor microenvironment to cancer immunotherapy. Genes and Cancer 2018;9 :176–189.30603054
108. Andrews LP , Marciscano AE , Drake CG , Vignali DAA . LAG3 (CD223) as a Cancer Immunotherapy Target. Immunol. Rev 2017;276 :80–96.28258692
109. Golden D , Kolmakova A , Sura S , Lymphocyte activation gene 3 and coronary artery disease. JCI Insight 2016;1 :1–15.
110. Fadok VA , Bratton DL , Konowal A , Freed PW , Westcott JY , Henson PM . Macrophages that have ingested apoptotic cells in vitro inhibit proinflammatory cytokine production through autocrine/paracrine mechanisms involving TGF-β, PGE2, and PAF. J. Clin. Invest 1998;101 :890–898.9466984
111. Rayner KJ . Cell Death in the Vessel Wall: The Good, the Bad, the Ugly. Arterioscler. Thromb. Vasc. Biol 2017;37 :e75–e81.28637702
112. Kojima Y , Volkmer JP , McKenna K , CD47-blocking antibodies restore phagocytosis and prevent atherosclerosis. Nature 2016;536 :86–90.27437576
113. Flores AM , Hosseini-Nassab N , Jarr KU , Pro-efferocytic nanoparticles are specifically taken up by lesional macrophages and prevent atherosclerosis. Nat. Nanotechnol 2020;15 :154–161. Available at: 10.1038/s41565-019-0619-3.31988506
114. Wang Y , Nanda V , Direnzo D , Clonally expanding smooth muscle cells promote atherosclerosis by escaping efferocytosis and activating the complement cascade. Proc. Natl. Acad. Sci. U. S. A 2020;117 :15818–15826.32541024
115. Advani R , Flinn I , Popplewell L , CD47 Blockade by Hu5F9-G4 and Rituximab in Non-Hodgkin’s Lymphoma. N. Engl. J. Med 2018;379 :1711–1721.30380386
116. Jarr K-U , Nakamoto R , Doan BH , Effect of CD47 Blockade on Vascular Inflammation. N. Engl. J. Med 2021;384 :382–383.33503349
117. Proto JD , Doran AC , Gusarova G , Regulatory T Cells Promote Macrophage Efferocytosis during Inflammation Resolution. Immunity 2018;49 :666–677.e6. Available at: 10.1016/j.immuni.2018.07.015.30291029
118. Lauby-Secretan B , Ph D, Scoccianti C , Ph D, Loomis D , Ph D. Body Fatness and Cancer — Viewpoint of the IARC Working Group. N. Engl. J. Med 2016;375 :794–798.27557308
119. Huang B , Song B liang , Xu C . Cholesterol metabolism in cancer: mechanisms and therapeutic opportunities. Nat. Metab 2020;2 :132–141. Available at: 10.1038/s42255-020-0174-0.32694690
120. Bersanelli M , Cortellini A , Buti S . The interplay between cholesterol (and other metabolic conditions) and immune-checkpoint immunotherapy: shifting the concept from the “inflamed tumor” to the “inflamed patient.” Hum. Vaccines Immunother 2021;00 :1–5. Available at: 10.1080/21645515.2020.1852872.
121. Porta C , Marino A , Consonni FM , Metabolic influence on the differentiation of suppressive myeloid cells in cancer. Carcinogenesis 2018;39 :1095–1104.29982315
122. Safari E , Ghorghanlu S , Ahmadi-khiavi H , Mehranfar S , Rezaei R , Motallebnezhad M . Myeloid-derived suppressor cells and tumor: Current knowledge and future perspectives. J. Cell. Physiol 2019;234 :9966–9981.30537008
123. Ma X , Bi E , Huang C , Cholesterol negatively regulates IL-9-producing CD8+ T cell differentiation and antitumor activity. J. Exp. Med 2018;215 :1555–1569.29743292
124. Ma X , Bi E , Lu Y , Cholesterol Induces CD8+ T Cell Exhaustion in the Tumor Microenvironment. Cell Metab. 2019;30 :143–156.e5. Available at: 10.1016/j.cmet.2019.04.002.31031094
125. Perrone F , Minari R , Bersanelli M , The Prognostic Role of High Blood Cholesterol in Advanced Cancer Patients Treated with Immune Checkpoint Inhibitors. J. Immunother 2020;43 :196–203.32404654
126. Young AC , Quach HT , Song H , Impact of body composition on outcomes from anti-PD1 +/−anti-CTLA-4 treatment in melanoma. J. Immunother. Cancer 2020;8 :1–8.
127. Cortellini A , Bersanelli M , Buti S , A multicenter study of body mass index in cancer patients treated with anti-PD-1/PD-L1 immune checkpoint inhibitors: When overweight becomes favorable. J. Immunother. Cancer 2019;7 :1–11.30612589
128. Leon C , Hill JS , Wasan KM . Potential role of Acyl-coenzyme A:cholesterol transferase (ACAT) inhibitors as hypolipidemic and antiatherosclerosis drugs. Pharm. Res 2005;22 :1578–1588.16180116
129. Antalis CJ , Uchida A , Buhman KK , Siddiqui RA . Migration of MDA-MB-231 breast cancer cells depends on the availability of exogenous lipids and cholesterol esterification. Clin. Exp. Metastasis 2011;28 :733–741.21744083
130. Geng F , Cheng X , Wu X , Inhibition of SOAT1 suppresses glioblastoma growth via blocking SREBP-1-mediated lipogenesis. Clin. Cancer Res 2016;22 :5337–5348.27281560
131. Tabas I , Marathe S , Keesler GA , Beatini N , Shiratori Y . Evidence that the initial upregulation of phosphatidylcholine biosynthesis in free cholesterol-loaded macrophages is an adaptive response that prevents cholesterol-induced cellular necrosis. Proposed role of an eventual failure of this response in foam c. J. Biol. Chem 1996;271 :22773–22781. Available at: 10.1074/jbc.271.37.22773.8798453
132. Tardif JC , Grégoire J , L’Allier PL , Effects of the acyl coenzyme A:cholesterol acyltransferase inhibitor avasimibe on human atherosclerotic lesions. Circulation 2004;110 :3372–3377.15533865
133. Bocan TMA , Krause BR , Rosebury WS , The combined effect of inhibiting both ACAT and HMG-CoA reductase may directly induce atherosclerotic lesion regression. Atherosclerosis 2001;157 :97–105.11427208
134. InsullJr W , Koren MJ , Davignon J , Efficacy and short-term safety of a new ACAT inhibitor, avasimibe, on lipids, lipoproteins and apolipoproteins, in patientswith combined hyperlipidemia. Atherosclerosis 2001;157 :137–144.11427213
135. Yang W , Bai Y , Xiong Y , Potentiating the antitumour response of CD8+ T cells by modulating cholesterol metabolism. Nature 2016;531 :651–655. Available at: 10.1038/nature17412.26982734
136. Gelsomino F , Fiorentino M , Zompatori M , Programmed death-1 inhibition and atherosclerosis: Can nivolumab vanish complicated atheromatous plaques? Ann. Oncol 2018;29 :284–285. Available at: 10.1093/annonc/mdx718.29106493
137. Cautela J , Rouby F , Salem J-E , Acute Coronary Syndrome With Immune Checkpoint Inhibitors: A Proof-of-Concept Case and Pharmacovigilance Analysis of a Life-Threatening Adverse Event. Can. J. Cardiol 2020;36 :476–481. Available at: 10.1016/j.cjca.2019.11.035.32144037
138. Kwan JM , Cheng R , Feldman LE . Hepatotoxicity and Recurrent NSTEMI While on Pembrolizumab for Metastatic Giant Cell Bone Tumor. Am. J. Med. Sci 2019;357 :343–347. Available at: http://www.sciencedirect.com/science/article/pii/S0002962918304476. 30638772
139. Hu Y-B , Zhang Q , Li H-J , Evaluation of rare but severe immune related adverse effects in PD-1 and PD-L1 inhibitors in non-small cell lung cancer: a meta-analysis. Transl. Lung Cancer Res 2017;6 :S8–S20.29299404
140. Oren O , Yang EH , Molina JR , Bailey KR , Blumenthal RS , Kopecky SL . Cardiovascular Health and Outcomes in Cancer Patients Receiving Immune Checkpoint Inhibitors. Am. J. Cardiol 2020;125 :1920–1926. Available at: 10.1016/j.amjcard.2020.02.016.32312493
141. Calabretta R , Hoeller C , Pichler V , Immune Checkpoint Inhibitor Therapy Induces Inflammatory Activity in Large Arteries. Circulation 2020:1–3.
142. Poels K , van Leent MMT , Boutros C , Immune Checkpoint Inhibitor Therapy Aggravates T Cell–Driven Plaque Inflammation in Atherosclerosis. JACC CardioOncology 2020.
143. Newman JL , Stone JR . Immune checkpoint inhibition alters the inflammatory cell composition of human coronary artery atherosclerosis. Cardiovasc. Pathol 2019;43 :107148. Available at: http://www.sciencedirect.com/science/article/pii/S1054880719302339. 31518915
144. Weitz-Schmidt G , Welzenbach K , Brinkmann V , Statins selectively inhibit leukocyte function antigen-1 by binding to a novel regulatory integrin site. Nat. Med 2001;7 :687–692.11385505
145. Sabatine MS , Giugliano RP , Keech AC , Evolocumab and Clinical Outcomes in Patients with Cardiovascular Disease. N. Engl. J. Med 2017;376 :1713–1722.28304224
146. Bird L Statins as adjuvants. Nat. Rev. Immunol 2018;18 :669.30297717
147. Omori M , Okuma Y , Hakozaki T , Hosomi Y . Statins improve survival in patients previously treated with nivolumab for advanced non-small cell lung cancer: An observational study. Mol. Clin. Oncol 2018:137–143.
148. Cantini L , Pecci F , Hurkmans DP , High-intensity statins are associated with improved clinical activity of PD-1 inhibitors in malignant pleural mesothelioma and advanced non-small cell lung cancer patients. Eur. J. Cancer 2021;144 :41–48. Available at: 10.1016/j.ejca.2020.10.031.33326868
149. Liu X , Bao X , Hu M , Inhibition of PCSK9 potentiates immune checkpoint therapy for cancer. Nature 2020;588 :693–698. Available at: 10.1038/s41586-020-2911-7.33177715
150. Drobni ZD , Murphy SP , Alvi RM , Association between incidental statin use and skeletal myopathies in patients treated with immune checkpoint inhibitors. Immunother. Adv 2021;1 :1–14.
151. Ribas A , Kefford R , Marshall MA , Phase III randomized clinical trial comparing tremelimumab with standard-of-care chemotherapy in patients with advanced melanoma. J. Clin. Oncol 2013;31 :616–622.23295794
152. Coward J , Lemech C , Meniawy T , Phase I/II study of CTLA-4 inhibitor AGEN1884 + PD-1 Inhibitor AGEN2034 in patients with advanced/refractory solid tumors, with expansion into 2L cervical cancer and solid tumors. Ann. Oncol 2018;29 :viii417.
153. Migden MR , Rischin D , Schmults CD , PD-1 Blockade with Cemiplimab in Advanced Cutaneous Squamous-Cell Carcinoma. N. Engl. J. Med 2018:341–351.29863979
154. Guo L , Wei R , Lin Y , Kwok HF . Clinical and Recent Patents Applications of PD-1/PD-L1 Targeting Immunotherapy in Cancer Treatment—Current Progress, Strategy, and Future Perspective. Front. Immunol 2020;11 :1–19.32038653
155. Herbst RS , Giaccone G , de Marinis F , Atezolizumab for First-Line Treatment of PD-L1–Selected Patients with NSCLC. N. Engl. J. Med 2020;383 :1328–1339.32997907
156. Powles T , Park SH , Voog E , Avelumab Maintenance Therapy for Advanced or Metastatic Urothelial Carcinoma. N. Engl. J. Med 2020;383 :1218–1230.32945632
157. Antonia SJ , Villegas A , Daniel D , Overall Survival with Durvalumab after Chemoradiotherapy in Stage III NSCLC. N. Engl. J. Med 2018;379 :2342–2350.30280658
158. Acharya N , Acharya N , Sabatos-Peyton C , Anderson AC , Anderson AC . Tim-3 finds its place in the cancer immunotherapy landscape. J. Immunother. Cancer 2020;8 :1–11.
159. Lecocq Q , Keyaerts M , Devoogdt N , Breckpot K . The next-generation immune checkpoint lag-3 and its therapeutic potential in oncology: Third time’s a charm. Int. J. Mol. Sci 2021;22 :1–17.
160. Buzzatti G , Dellepiane C , Del Mastro L . New emerging targets in cancer immunotherapy: The role of GITR. ESMO Open 2020;4 .
161. Byrne KT , Betts CB , Mick R , Neoadjuvant Selicrelumab, an Agonist CD40 Antibody, Induces Changes in the Tumor Microenvironment in Patients with Resectable Pancreatic Cancer. Clin. Cancer Res 2021;27 :4574–4586.34112709
162. Marin-Acevedo JA , Kimbrough EMO , Manochakian R , Zhao Y , Lou Y . Immunotherapies targeting stimulatory pathways and beyond. J. Hematol. Oncol 2021;14 :1–31. Available at: 10.1186/s13045-021-01085-3.33402199
163. Starzer AM , Berghoff AS . New emerging targets in cancer immunotherapy: CD27 (TNFRSF7). ESMO Open 2020;4 .
164. Zhang W , Huang Q , Xiao W , Advances in Anti-Tumor Treatments Targeting the CD47/SIRPα Axis. Front. Immunol 2020;11 :1–15.32038653
|
PMC008xxxxxx/PMC8983022.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
0121103
280
Adv Exp Med Biol
Adv Exp Med Biol
Advances in experimental medicine and biology
0065-2598
2214-8019
26427485
8983022
10.1007/978-3-319-17121-0_100
NIHMS1783661
Article
Live-Cell Imaging of Phagosome Motility in Primary Mouse RPE Cells
Hazim Roni
Jiang Mei
Esteve-Rudd Julian
Diemer Tanja
Lopes Vanda S.
Williams David S.
UCLA School of Medicine, Jules Stein Eye Institute, 100 Stein Plaza, Los Angeles, CA 90095, USA
Roni Hazim, Mei Jiang authors, Contributed equally
[email protected]
24 3 2022
2016
05 4 2022
854 751755
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
The retinal pigment epithelium (RPE) is a post-mitotic epithelial monolayer situated between the light-sensitive photoreceptors and the choriocapillaris. Given its vital functions for healthy vision, the RPE is a primary target for insults that result in blinding diseases, including age-related macular degeneration (AMD). One such function is the phagocytosis and digestion of shed photoreceptor outer segments. In the present study, we examined the process of trafficking of outer segment disk membranes in live cultures of primary mouse RPE, using high speed spinning disk confocal microscopy. This approach has enabled us to track phagosomes, and determine parameters of their motility, which are important for their efficient degradation.
Live-cell imaging
Retinal pigment epithelium
Intracellular trafficking
Photoreceptor outer segment
Phagocytosis
pmc100.1 Introduction
The retinal pigment epithelium (RPE) is a post-mitotic epithelial monolayer of cuboidal cells situated between the light-sensitive photoreceptors and the choriocapillaris (Bok 1993). The RPE performs numerous functions vital to the health of photoreceptors and thus to healthy vision. These functions include recycling of retinoids during the visual cycle, transport of nutrients from the blood to the photoreceptors, and secretion of growth factors, such as vascular endothelial growth factor (VEGF) and pigment epithelial-derived factor (PEDF) (Strauss 2005). One of the most critical functions performed by the RPE is the phagocytosis of photoreceptor outer segment (POS) tips (Young and Bok 1969), an event that occurs on a daily cycle (LaVail 1976).
The RPE is a professional phagocyte, internalizing and degrading approximately 10 % of each photoreceptor outer segment on a daily basis. Phagosomes containing POS membranes move from the apical region of the RPE towards the basal region (Herman and Steinberg 1982; Gibbs et al. 2003), fusing with degradative organelles such as endosomes and lysosomes along the way (Wavre-Shapton et al. 2014; Bosch et al. 1993). By-products that are not completely degraded tend to form constituents of aggregates, such as lipofuscin or sub-RPE deposits, common features associated with macular degeneration (Brunk and Terman 2002). Given the movement of phagosomes from the apical region, their motility is closely related with their degradation. In an early study, it was shown that colchicine, which disrupts microtubules, inhibited the translocation of phagosomes from the apical region (Herman and Steinberg 1982). More recently, the importance of actin-based motility was demonstrated in mice lacking MYO7A, an unconventional myosin. In those mice, phagosomes were retained longer in the apical region of the RPE, and were degraded more slowly (Gibbs et al. 2003). In the present report, we describe the use of live-cell imaging, using spinning disk confocal microscopy, to study the intracellular trafficking of POS-containing phagosomes within primary mouse RPE cells.
100.2. Isolation and Culture of Primary Mouse RPE
Primary mouse RPE were isolated as previously described (Gibbs et al. 2003). Intact eyes were enucleated from P10-P15 mice and washed 3–4 times by inversion with growth medium (Dulbecco’s modified Eagle’s medium (DMEM) with 4.5 g/L glucose, L-glutamine, and sodium pyruvate). The eyes were then incubated in a 2 % dispase solution for 45 min at 37° C. Following removal of the enzyme solution, the eyes were washed 3 times with growth medium containing 10 % fetal bovine serum (FBS) and 20 mM HEPES. The eyes were dissected into eyecups by making an incision along the ora serrata to remove the cornea, iris, lens, and ciliary body. Eyecups were then incubated in growth medium for 20 min at 37° C, as this facilitates the separation of the RPE from the retina and Bruch’s membrane. Sheets of RPE were gently scraped from Bruch’s membrane and collected in growth medium with 10 % FBS. The sheets were then washed 3 times with growth medium and twice with calcium- and magnesium-free Hank’s Balanced Salt Solution (HBSS). The cells were then briefly and gently triturated and plated on Lab-Tek chambered coverglass. Live-cell imaging experiments were carried out on 3–7 day old cultures.
100.3. Isolation and Labeling of Mouse POSs
Mouse POSs were isolated as previously described (Gibbs et al. 2003). Mouse retinas were collected under dim red light and homogenized in Ringer’s solution (130 mM NaCl, 3.6 mM KCl, 2.4 mM MgCl2, 1.2 mM CaCl2, 10 mM HEPES, and 0.02 mM EDTA). The homogenate was cleared by centrifugation for 30 s at 100 g, and then the supernatant was layered on top of a discontinuous Optiprep 8 %−10 %−15 % step gradient in Ringer’s solution and spun at 12,000 g for 20 min at 4° C. POSs were collected at the 10 %/15 % interface and diluted 3 times with Ringer’s solution. POSs were then pelleted by spinning the solution at 10,300 g for 10 min at 4° C. The POSs were then labeled by incubation with 0.1 mg Texas Red-X, succinimidyl ester or 5 % (v/v) Alexa Fluor 488 carboxylic acid, succinimidyl ester, mixed isomers in 1 mL 0.1 M NaHCO3, pH 8.3 for 1 h at 4° C. POSs were then washed with Ringer’s solution, resuspended in RPE growth medium, and counted using a haemocytometer to determine the yield.
100.4. Live Imaging Using Spinning Disk Confocal Microscopy
Figure 100.1a depicts a schematic diagram of the protocol used for live-cell imaging. We used C57BL/6J mice for both the RPE cells and the POSs. Cultured RPE cells were incubated with 1–5 × 106 fluorescently-labeled POSs in growth medium with 10 mM HEPES for 20 min at 37° C, washed extensively with growth medium, and then immediately imaged for a maximum of 1 h, using an Ultraview Spinning Disk Confocal Microscope system with a Zeiss Axiovert photomicroscope, including an environment chamber. Movies were acquired at 3 frames per second with the Volocity software (PerkinElmer), using a 63x oil immersion objective and a Hamamatsu EM-CCD camera (see supplementary video). Trajectories of phagosomes were analyzed using the Volocity software (Fig. 100.1b).
Not all phagosomes were moving at a given time, however, the paths of those that were moving typically followed relatively straight lines, with back and forth movements along these lines. This motility is consistent with movements along microtubules, as cargos of plus- and minus-end directed microtubule motors. The paths can be analyzed to assess a variety of phagosome motility parameters. Speed and distance traveled represent two basic parameters. From analysis of the paths of phagosomes that traveled at least 3 μm in a 24-second interval, we found a mean speed of 1.2 ± 0.1 μm/s and a mean total distance traveled of 11.3 ± 1.9 μm, during the 24-sec interval. This speed is typical of transport by microtubule motors (Okada et al. 1995). Transport along actin filaments by myosins is typically many fold slower (Boal 2012), suggesting that the observed motility was dominated by the microtubule motors, kinesin and dynein.
100.5. Conclusions
Fluorescently-labeled POS phagosomes can be monitored in live RPE cells, using spinning disk confocal microscopy. Their motility can be determined by tracking their trajectories, thus providing a sensitive, real-time measurement of a critical parameter of RPE health—one, which we are finding in other studies, feeds directly into the efficiency of phagosome degradation, and the propensity for the accumulation of debris and consequent activation of downstream events, such as inflammation and oxidative stress.
Supplementary Material
Supplementary Video
Acknowledgements
We thank Barry Burgess for technical assistance. This study was supported by NIH R01 grant EY 07042 and P30 grant EY 00331.
Fig. 100.1 Method for monitoring phagocytosis of photoreceptor outer segments in live RPE. a Scheme of phagocytosis assay for live cultures of RPE, b Magnified view of a fluorescently-labeled phagosome and its trajectory inside a live RPE cell
The online version of this chapter 10.1007/978-3-319-17121-0_100 contains supplementary video material, which can be downloaded from: http://extra.springer.com.
References
Boal D (2012) Mechanics of the cell. In: Boal D (eds) Dynamic filaments Cambridge University, Cambridge
Bok D (1993) The retinal pigment epithelium: a versatile partner in vision. J Cell Sci 17 :189–195
Bosch E , Horwitz J , Bok D (1993) Phagocytosis of outer segments by retinal pigment epithelium: phagosome-lysosome interaction. J Histochem Cytochem 41 :253–263 8419462
Brunk UT , Terman A (2002) Lipofuscins: mechanisms of age-related accumulation and influence on cell function. Free Rad Biol Med 33 :611–619 12208347
Gibbs D , Kitamoto J , Williams DS (2003) Abnormal phagocytosis by retinal pigmented epithelium that lacks myosin VIIa, the usher syndrome 1B protein. Proc Natl Acad Sci U S A 100 :6481–6486 12743369
Herman KG , Steinberg RH (1982) Phagosome movement and the diurnal pattern of phagocytosis in the tapetal retinal pigment epithelium of the opossum. Invest Ophthalmol Vis Sci 23 : 277–290 7107156
LaVail MM (1976) Rod outer segment disk shedding in rat retina: relationship to cyclic lighting. Science 194 :1071–1074 982063
Okada Y , Yamazaki H , Sekine-Aizawa Y (1995) The neuron-specific kinesin superfamily protein KIF1A is a unique monomeric motor for anterograde axonal transport of synaptic vesicle precursors. Cell 81 :769–780 7539720
Strauss O (2005) The retinal pigment epithelium in visual function. Physiol Rev 85 :845–881 15987797
Wavre-Shapton ST , Meschede IP , Seabra MC (2014) Phagosome maturation during endosomes interaction revealed by partial rhodopsin processing in retinal pigment epithelium. J Cell Sci 127 :3852–3861 25074813
Young RW , Bok D (1969) Participation of the retinal pigment epithelium in the rod outer segment renewal process. J Cell Biol 42 :392–403 5792328
|
PMC008xxxxxx/PMC8983023.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
101232704
33038
J Chem Theory Comput
J Chem Theory Comput
Journal of chemical theory and computation
1549-9618
1549-9626
34844409
8983023
10.1021/acs.jctc.1c00770
NIHMS1784425
Article
Gaussian-Accelerated Molecular Dynamics with the Weighted Ensemble Method: A Hybrid Method Improves Thermodynamic and Kinetic Sampling
Ahn Surl-Hee http://orcid.org/0000-0002-3422-805X
§Department of Chemistry, University of California San Diego, La Jolla 92093 California, United States
Ojha Anupam A. http://orcid.org/0000-0001-6588-3092
§Department of Chemistry, University of California San Diego, La Jolla 92093 California, United States
Amaro Rommie E. http://orcid.org/0000-0002-9275-9553
Department of Chemistry, University of California San Diego, La Jolla 92093 California, United States
McCammon J. Andrew http://orcid.org/0000-0003-3065-1456
Department of Chemistry, University of California San Diego, La Jolla 92093 California, United States; Department of Pharmacology, University of California San Diego, La Jolla 92093 California, United States
§ Author Contributions
S.-H.A. and A.A.O. have contributed equally.
Corresponding Author: Surl-Hee Ahn – Department of Chemistry, University of California San Diego, La Jolla 92093 California, United States; s3ahn@ ucsd.edu
12 3 2022
14 12 2021
30 11 2021
05 4 2022
17 12 79387951
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Gaussian-accelerated molecular dynamics (GaMD) is a well-established enhanced sampling method for molecular dynamics simulations that effectively samples the potential energy landscape of the system by adding a boost potential, which smoothens the surface and lowers the energy barriers between states. GaMD is unable to give time-dependent properties such as kinetics directly. On the other hand, the weighted ensemble (WE) method can efficiently sample transitions between states with its many weighted trajectories, which directly yield rates and pathways. However, convergence to equilibrium conditions remains a challenge for the WE method. Hence, we have developed a hybrid method that combines the two methods, wherein GaMD is first used to sample the potential energy landscape of the system and WE is subsequently used to further sample the potential energy landscape and kinetic properties of interest. We show that the hybrid method can sample both thermodynamic and kinetic properties more accurately and quickly compared to using either method alone.
Graphical Abstract
pmc1. INTRODUCTION
Molecular dynamics (MD) simulations are becoming quintessential tools in many fields, including biology,1–3 chemistry,4–7 materials science,8,9 and chemical and biological engineering.10,11 An increasing number of researchers have used MD simulations to uncover mechanisms of their biological system of interest in atomistic detail. Applications of MD simulations range from studying protein folding3,6,12–14 and protein–protein or protein–ligand interactions15,16 to computer-aided drug design (virtual screening and ligand docking).4,5,7 However, MD simulations are not without their challenges. MD simulations have to be run using femtosecond time steps due to being limited by the fastest motions in the system. In contrast, biological processes of interest are on the order of microseconds or longer. Additionally, systems often get “stuck” in metastable states and do not change conformations for an extended period. Hence, MD simulations can be computationally costly when attempting to observe rare events, which is often the case of interest.
Fortunately, researchers have developed several “enhanced sampling methods” to overcome this timescale gap between MD simulations and biological processes. Many enhanced sampling methods work by adding a biasing potential to force the system away from metastable states. These include but are not limited to Gaussian-accelerated molecular dynamics (GaMD),17–20 metadynamics,21–25 umbrella sampling,26–29 and adaptive biasing force (ABF).30–34 Among these, GaMD has the advantage of not requiring any collective variables (CVs) to steer the simulation. Rather, it allows unrestrained sampling of configuration space. Another similar class of methods changes the system’s temperature instead to sample states difficult to reach at room temperature, including replica exchange molecular dynamics (REMD) or parallel tempering.35–39 Although both these classes of methods are effective at obtaining thermodynamic properties like the free-energy landscape of the system, they alter the actual kinetics of the system, preventing them from directly getting kinetic properties from the system. Note that there are methods to derive kinetic properties like rate constants from simulations that used these methods. Still, they need to be obtained either by using Kramers’ rate theory in the high friction or “overdamping” regime,40 constructing a master equation,41,42 or assuming a low residence time in the transition states,43 which are either approximations or are limiting conditions for these methods to be used for all cases. Moreover, since the real kinetics of the system is altered, continuous pathways cannot be obtained from these methods.
As a result, several path-sampling methods focus on sampling kinetic properties, such as rate constants from the reactant state to the product state, including milestoning,44–48 forward flux sampling,49–52 transition interface sampling,53–56 and others. These methods divide the path space from the reactant state to the product state into many interfaces and run many short simulations to efficiently obtain rate constants and free-energy landscape of the path space. However, since these methods primarily focus on sampling the path of interest, the rest of the free-energy landscape is not extensively sampled. If a more comprehensive picture of the entire configuration space is needed along with its thermodynamic and kinetic properties, then the weighted ensemble (WE) method57–67 can be used on the system of interest. WE decomposes the configuration space into small volume elements called “bins” with appropriate CVs and run many short simulations with probabilities or “weights” for τ amount of time (“resampling time”) for many iterations to obtain good statistics. WE has proven to be useful in uncovering insights into the mechanisms of important biophysical systems.68–76 However, WE requires a sufficient sampling of the configuration space by having many initial configurations with close to steady-state probabilities to get accurate results quickly.
As a result, we have developed a hybrid enhanced sampling method that combines GaMD and WE called GaMD–WE. There also exist hybrid methods that combine REMD and GaMD77,78 and well-tempered metadynamics and GaMD,79 but both aim to improve sampling of thermodynamic properties. In contrast, GaMD–WE aims to enhance sampling of both thermodynamic and kinetic properties. There is also another hybrid method called Markovian WE milestoning (M-WEM)80 that combines milestoning and WE and several reweighting methods for WE81 that also aim to accelerate sampling of both thermodynamic and kinetic properties. However, M-WEM cannot generate continuous pathways since WE is used to accelerate convergence of milestones in milestoning. The reweighting methods are iterative methods that can be used additionally for any WE simulation, including GaMD–WE simulations. In GaMD–WE, GaMD is initially run to sample the free-energy landscape efficiently with its harmonic boost potentials. Then, after reweighting is performed to recover the original free-energy landscape, WE is run with many initial configurations produced from the GaMD run. This way, the two methods complement each other and reduce each other’s limitations. This paper will introduce both methods, the hybrid method, and the results that demonstrate the hybrid method’s power to obtain thermodynamic and kinetic properties more accurately and more quickly than one method by itself.
2. METHODS
2.1. GaMD.
GaMD is an enhanced sampling method for MD simulations that can efficiently sample thermodynamic proper ties such as the free-energy landscape of the system. When the system potential V(r), where r denotes the position vector of an N-atom system, is lower than the threshold energy E, GaMD fills the energy wells by adding a harmonic boost potential ΔV(r), that is (1) ΔVr=12kE−Vr2
where k denotes the harmonic force constant. If V(r) ≥ E, then no boost potential is added. There are several criteria that the boost potential ΔV(r) needs to satisfy for GaMD to work, and readers can refer to the original GaMD paper17 for specific details about the boost potential ΔV(r) and the harmonic force constant k.
If the anharmonicity of the harmonic boost potential ΔV(r) is small, then ΔV(r) follows a near Gaussian distribution and the cumulant expansion to the second order can be used to approximate the exponential average term ⟨eβΔV(r)⟩, where β denotes the thermodynamic beta or 1/kBT. This exponential average term ⟨eβΔV(r)⟩ is needed to reweight and recover the original free-energy landscape from GaMD. Readers can refer to the original GaMD paper17 for specific details about energetic reweighting with cumulant expansion to the second order.
A significant advantage of GaMD is that CVs, which describe the state of a molecular system, are not needed. Identifying the appropriate CVs for a particular system is still an active area of research and can be difficult for new or unfamiliar systems.82–84 In contrast, metadynamics requires CVs to be chosen a priori, which similarly fills energy wells with repulsive Gaussian potentials. However, metadynamics does not suffer from having unconverged high-energy regions like GaMD since metadynamics recovers the original free-energy landscape as the opposite sum of all Gaussians. Nonetheless, GaMD is one of the few enhanced sampling methods that does not require tuning of many parameters and can be easily applied to various systems. Additionally, GaMD is fully implemented in Amber85 (starting from Amber16) and NAMD86 (starting from 2.13),18 which makes it easier for users to use the method.
2.2. WE Method.
The WE method is another enhanced sampling method for MD simulations that runs many short simulations instead of one long simulation to sample thermodynamic and kinetic properties efficiently. These short simulations or “walkers” carry probabilities or “weights” that evolve throughout the simulation via “resampling,” a statistical procedure to maintain a number of these short simulations at visited regions of the configuration space. More details can be found in the original WE paper,57 in a review article,58 and in the Weighted Ensemble Simulation Toolkit with Parallelization and Analysis (WESTPA) papers,87,88 but the general scheme is as follows.
The following parameters are chosen a priori for the WE simulation: CVs, resampling time τ for walkers, partitioning of bins (small volume elements of the configuration space), and target number of walkers per bin nw. If there is only one initial state, then there will be nw walkers, each with a weight of 1/nw. Otherwise, there will be multiple nw walkers, each with an appropriate weight that sums up to 1 for the entire system.
Walkers are run for τ amount of time and binned to appropriate bins depending on their CV values.
Walkers go through “resampling,” that is, merged or replicated in a statistically correct way so that the target number of walkers per bin nw is maintained for each bin. Each walker ends up with a weight between Pi/nw and 2Pi/nw where Pi denotes the sum of the weights in bin i.
Steps 2 and 3 are repeated until the desired convergence is reached.
With resampling, the walkers are maintained in each visited bin regardless of its energy barrier height. Computational cost is also curtailed since walkers are merged in oversampled, low-energy regions and replicated in rare, high-energy regions. Since no statistical bias is added to the system, one can directly obtain both thermodynamic and kinetic properties of the system from the evolution of walkers’ weights in each bin.
Although several parameters need to be chosen a priori as stated in step 1, the resampling time τ can be selected without having to worry about fulfilling the Markovian property, a requirement that other enhanced sampling methods have such as milestoning. The resampling time τ should be chosen to be short enough so that WE does not inadvertently miss transitions.61,76,89 However, since many bins have to reach convergence to extract correct thermodynamic and kinetic properties of the system, WE can be computationally costly if the initial states are not close to the steady state.
In our study, equilibrium WE was used to sample the thermodynamic and kinetic properties of the system since we were interested in obtaining rate constants between more than two states. The rate constants were obtained by defining states post simulation. If one is interested in obtaining kinetic properties between two states, however, then steady-state WE can be used, which “recycles” or feeds back walkers to the initial state once they reach the target state. This way, WE focuses its computational effort in sampling the transition of interest in nonequilibrium steady-state conditions.
2.3. GaMD–WE Method.
The hybrid GaMD–WE method aims to combine strengths and mitigate weaknesses of both methods. By initially running GaMD to sample the free-energy landscape of the system, one can obtain a well-sampled initial state distribution for WE. Then with WE, one can get a more refined free-energy landscape closer to steady state and sample kinetic properties such as rate constants from one state to another state. We show that the hybrid method is significantly more effective than running a conventional WE to sample thermodynamic and kinetic properties within the same amount of simulation time in the subsequent Results section.
We have developed a GaMD–WE package for users to run GaMD and prepare initial states for WE, specifically for the WESTPA.87,88 The current package is not fully integrated with WESTPA, that is, the user needs to use the GaMD–WE package for the GaMD portion and run a WESTPA simulation separately for the WE portion using initial states from the GaMD–WE package, but we plan to make it fully integrated in the future. The GaMD–WE package is fully customizable, that is, the desired force fields, water models, and others can be added, and it follows a series of scripts as the following.
System is prepared for simulation after the Protein Data Bank (PDB) structure is downloaded from the PDB server. Appropriate force field parameters are added followed by system solvation.
Solvated system is then minimized, heated, and equilibrated using the OpenMM90 simulation engine. Directories are created for the subsequent GaMD simulations.
Six GaMD simulations are run using the Amber85 simulation engine for varying degrees of potential boosts, that is, lower bound dihedral potential boost, upper bound dihedral potential boost, lower bound total potential boost, upper bound total potential boost, lower bound dual (dihedral + total) potential boost, and upper bound dual potential boost, which are all of the possible combinations for potential boosts, for the desired amount of simulation time.
Simulation data are then extracted from the output log files for each of the six GaMD simulations. Reweighting is then performed with several reweighting methods, that is, cumulant expansion to the first order, second order, and third order, to recover the original free-energy landscape. Reweighting is done for bins with more than 10 frames. Then, with the desired target number of walkers per bin nw, initial structures for WE are saved with appropriate weights from the reweighted probabilities. If there are more frames in the chosen bin than the target number of walkers per bin, then the frames are chosen at random.
Initial structures for WE are minimized in two steps (Step 1: Heavy atoms Cα, N, C, and O of the protein are minimized. Step 2: Entire system including the solvent is minimized) so that none of them “crash” during WE. User can set the minimization steps.
WE simulation directory is created with proper initial structures. Number of initial structures can be compared among the six simulation outputs (see the Supporting Information). GaMD simulation that yielded the largest number of initial structures can be subsequently used for WE simulation since that would indicate the greatest coverage of the free-energy landscape. All of the initial structures from GaMD are used for WE. Hence, the current GaMD–WE implementation is suitable for equilibrium WE simulations that are focused on sampling the entire free-energy landscape and rate constants for regions defined post simulation.
In the subsequent Results section, we show that GaMD has greater coverage of the free-energy landscape and closer to steady-state probabilities compared to WE within the same simulation time. Even if WE has a similar amount of coverage of the free-energy landscape compared to GaMD for some systems, GaMD has an advantage over WE with reweighting since appropriate probabilities or weights can be recovered from the added biasing potentials. In contrast, since WE does not add any statistical bias to the system, the system needs to evolve naturally or reach convergence to appropriate probabilities or weights, which in most cases would take longer than adding a biasing potential to the system.
3. RESULTS
We have tested our hybrid method on two systems: alanine dipeptide in explicit solvent and chignolin in implicit solvent. We show how GaMD–WE outperforms either method in obtaining thermodynamic and kinetic properties. To further illustrate that GaMD surpasses WE in getting the free-energy landscape, we show the free-energy landscapes of bovine pancreatic trypsin inhibitor (BPTI) in the explicit solvent obtained from the two methods. The three systems are illustrated in Figure 1a–c. Amber ff14SB force field parameters91 were used for all three systems. Simulations were run under the canonical ensemble with the temperature T set to 300 K using the Langevin thermostat with friction coefficient γ = 1.0 ps−1 for all three systems.
Three simulations were run for each system (alanine dipeptide and chignolin) using GaMD–WE and WE, and the average of the three is shown as the final result. The simulations were run until the rate constants had leveled off and remained consistent. Error bars for WE and GaMD–WE rate constants represent 95% confidence intervals (i.e., 1.96×σ3where σ denotes standard deviation). Each point in the WE and GaMD–WE rate constant graphs was calculated cumulatively in 50 iteration blocks.
3.1. Alanine Dipeptide.
Alanine dipeptide is a 22-atom system that is commonly used as a test system for new methods. Initial structure was obtained from https://markovmodel.github.io/mdshare/ALA2/. TIP3P water model92 was used to solvate the system explicitly. For GaMD–WE, GaMD was run for 50 ns and WE was run for 11.95 μs, so that the total simulation time amounted to 12 μs. The total simulation time for conventional WE was also 12 μs (three independent 4 μs runs). Resampling time τ for WE was set to 10 ps, equal to GaMD’s sampling frequency. The target number of walkers per bin nw was set to 4. CVs were set to be the dihedral angles, ϕ and ψ. Bins were evenly spaced in intervals of 10° for both ϕ and ψ (ranging from −180 to 180°).
First, we show that GaMD covers the free-energy landscape more than brute force simulation and WE within the same simulation time. Figure 2a–c shows the average free-energy landscape (−kBT ln P, where P denotes probability) of alanine dipeptide after 50 ns of brute force simulation, GaMD, and WE, respectively. The lowest energy state was set to be zero for each of the free-energy landscapes. In particular, the GaMD run with the upper bound of the dihedral boost potential yielded the maximum number of initial structures on average as compared to the other five GaMD potential settings (see the Supporting Information). Cumulant expansion to the second order was used for GaMD reweighting. As seen in Figure 2, GaMD covered most of the metastable regions that includes PII, the rest of the β-sheet region, and the right-handed α-helix region αR on the left side compared to brute force simulation and WE within 50 ns of the simulation time. This shows that it can be more beneficial to use GaMD instead of brute force simulation or WE to sample the free-energy landscape, even for this simple system. Figure 2d,e shows that the free-energy landscapes are comparable after 12 μs of brute force simulation and WE, respectively.
Second, we show that GaMD–WE can converge to the correct rate constants faster and more accurately than WE. In particular, rate constants between the three regions of interest shown in Figure 2e were measured over simulation time. αR region was defined to be −120° ≤ ϕ ≤ 0°, −100° ≤ ψ ≤ 50°, PII region was defined to be −120° ≤ ϕ ≤ 0°, 100° ≤ ψ ≤ 180°, and αL region was defined to be 0° ≤ ϕ ≤ 120°, −50° ≤ ψ ≤ 100°. Initial structures for the three GaMD–WE runs were from one of the three GaMD runs with an upper bound of dihedral boost potential that yielded the largest number of initial structures. WE and brute force simulations used the same initial structure as the GaMD runs. In addition, since GaMD covered a wider free-energy landscape within 50 ns, another set of GaMD–WE simulations was run with equal weights. Although reweighting would give more accurate weights for each region, we wanted to investigate whether there will be any improvements in obtaining kinetics solely from covering more of the free-energy landscape with GaMD versus WE. Figure 3 shows the evolution of rate constants over aggregate simulation time, and Table 1 summarizes the final rate constants for brute force, WE, and GaMD–WE simulations after 12 μs of simulation time. Reference brute force simulation values were obtained from averaging all first passage times from three independent 4 μs runs and performing Bayesian bootstrapping for 95% confidence intervals.88 The first 200 ns of simulation time was cut off in the rate constant calculation to eliminate the initial structure bias for brute force, WE, and GaMD–WE simulations.
Figure 3a,b shows that the convergence is comparable between WE and GaMD–WE for the rate constants between the two central metastable states αR and PII since both methods covered both regions well. However, GaMD–WE under-estimated both rate constants (see Table 1), which WE obtained more accurately. This might be due to GaMD–WE having lower weights than actual for the initial αR and PII structures, which would slow down the rate of convergence for GaMD–WE. The GaMD–WE simulations with equal weights, on the other hand, had larger error bars than regular GaMD–WE and performed similarly to GaMD–WE.
For the rate constants that involved the higher-energy region αL, however, GaMD–WE performed better than WE. Figure 3c,e highlights that GaMD–WE having the rate constants go from αL to either primary metastable state αR or PII converge faster with smaller error bars (see Table 1) compared to WE. GaMD–WE simulations with equal weights, on the other hand, did not perform better than either WE or GaMD–WE. As for the reverse rate constants that go from either primary metastable state αR or PII to αL, GaMD–WE and WE have comparable performances, with GaMD–WE slightly underestimating both rate constants as seen from Figure 3d,f and Table 1. GaMD–WE simulations with equal weights performed marginally better than GaMD–WE by underestimating the rate constants lesser. This might be due to having higher weights for the regions of interest than GaMD–WE.
These results indicate that GaMD–WE can obtain kinetics involving higher-energy regions like αL faster and more accurately than WE alone. In contrast, WE performs as well as GaMD–WE in getting kinetics involving central metastable states, which both methods can sample sufficiently well. In addition, GaMD–WE needs reweighting to have more accurate weights and have an advantage over WE, that is, GaMD–WE does not necessarily have an advantage over WE from solely covering more of the free-energy landscape within the same simulation time. However, this might not be the case if GaMD had covered a region with a higher energy barrier that is difficult for conventional WE to sample quickly, so this hypothesis needs to be tested on more complex systems in the future.
3.2. Chignolin.
To investigate whether GaMD–WE will be significantly more effective than WE for more complex systems, we tested the two methods on chignolin, a 138-atom system with 10 residues (PDB: 1UAO). The modified generalized Born implicit model with the model II radii93 was used to solvate the system implicitly. For GaMD–WE, GaMD was run for 500 ns, and WE was run for 39.5 μs, so that the total simulation time amounted to 40 μs. The total simulation time for conventional WE was also 40 μs (five independent 8 μs runs). Resampling time τ for WE was set to 20 ps, equal to GaMD’s sampling frequency and same as in ref 88, which had a chignolin example. Target number of walkers per bin nw was set to 4. CVs were set to be the mass-weighted root-mean-square-deviation (RMSD) of Cα atoms from the initial folded state (PDB: 1UAO) and the mass-weighted radius of gyration (Rg) of Cα atoms. Bins were evenly spaced in intervals of 0.2 Å for both RMSD and Rg (ranging from 0 to 8 Å).
Figure 4a–c shows the average free-energy landscape (−kBT ln P, where P denotes probability) of chignolin after 500 ns of brute force simulation (three out of five simulations), GaMD, and WE, respectively. The lowest energy state was set to be zero for each free-energy landscape. In particular, the GaMD run with the upper bound of the dihedral boost potential yielded the maximum number of initial structures on average as compared to the other five GaMD potential settings (see Supporting Information). Cumulant expansion to the second order was used for GaMD reweighting. Although GaMD and WE have comparable free-energy landscape coverage as seen in Figure 4b,c, GaMD has probabilities closer to actual values due to reweighting as seen in Figure 4d,e (or 4f), which show that the free-energy landscapes are comparable after 40 μs of brute force simulation and WE, respectively. In chignolin’s case, however, brute force simulation had the most coverage of the free-energy landscape out of the three cases. It is also a common practice to run a short brute force simulation to sample the free-energy landscape and run WE using the sampled states. Hence, using either brute force simulation or GaMD would have been more beneficial than WE to initially sample the free-energy landscape in chignolin’s case.
We also show that GaMD–WE can converge to the correct rate constants faster and more accurately than WE, more notably than in the alanine dipeptide case. Specifically, the rate constants between the four regions of interest shown in Figure 4 e,f were measured over the simulation time. The folded region was defined to be 0.0 Å ≤ RMSD ≤ 1.0 Å, the unfolded region was defined to be 4.0 Å ≤ RMSD, the intermediate I region was defined to be 3.5 Å ≤ RMSD ≤ 4.5 Å, 4.5 Å ≤ Rg ≤ 6.5 Å, and the intermediate II region was defined to be 6.0 Å ≤ RMSD ≤ 7.0 Å, 7.0 Å ≤ Rg ≤ 8.0 Å. Initial structures for the three GaMD–WE runs were from one of the three GaMD runs with an upper bound of the dihedral boost potential that yielded the largest number of initial structures. WE and brute force simulations used the same initial structure as the GaMD runs. Figures 5 and 6 show the evolution of rate constants over aggregate simulation time, and Table 2 summarizes the final rate constants for brute force, WE, and GaMD–WE simulations after 40 μs of simulation time. The reference brute force simulation values were obtained from averaging all first passage times from five independent 8 μs runs and performing Bayesian bootstrapping for 95% confidence intervals.88 The first 2 μs of simulation time was cut off in the rate constant calculation to eliminate the initial structure bias for brute force, WE, and GaMD–WE simulations.
Figure 5a,b shows that GaMD–WE is faster than WE at converging to the reference rate constants between the folded and unfolded regions. GaMD–WE is significantly better at obtaining the rate constants from unfolded to folded since GaMD–WE had closer to actual probabilities for the unfolded region with reweighting. Without reweighting, WE takes a significantly longer time to converge to the reference rate constants. Figure 6a–d also shows similar results that highlight GaMD–WE converging faster to the reference rate constants, especially for rate constants that go from either intermediate I or II region to the folded region. Performance for GaMD–WE is only slightly better than WE in obtaining the rate constants between the intermediate I region and the intermediate II region; However, as seen in Figure 6e,f, the two regions are close to each other, making sampling between these two regions and converging to the actual rate constants easier than the other cases. Finally, Table 2 indicates that the only rate constant that WE had within the reference confidence interval was for the folded → intermediate II rate constant. In contrast, the GaMD–WE had all of the rate constants fall within the confidence intervals except for the intermediate I → intermediate II rate constant. Error bars for GaMD−WE were lower than WE for all rate constants except for the folded → unfolded, unfolded → folded, and folded → intermediate I rate constants. These results indicate that GaMD–WE’s performance in obtaining kinetics is significantly better than conventional WE for more complex systems than alanine dipeptide. Nonetheless, as previously mentioned, the brute force simulation had more coverage of the free-energy landscape than GaMD and had close to true probabilities. Hence, it is plausible that the common practice of running WE, that is, run a short brute force simulation to initially sample the free-energy landscape before running WE, would have been as good or better than GaMD–WE in obtaining the rate constants in chignolin’s case.
3.3. BPTI.
As a final test, we ran a brute force simulation, GaMD, and WE of BPTI, an 892-atom system with 58 residues (PDB: 5PTI). This was to test whether GaMD will be more effective than brute force simulation and WE at covering the free-energy landscape of a bigger protein system than alanine dipeptide or chignolin. TIP4P-Ew water model94 was used to solvate the system explicitly. The total simulation time for all three simulations was 500 ns. Resampling time τ for WE was set to 40 ps since BPTI is a bigger system than chignolin, and the resampling time needs to be long enough to cross over to the next bin. Simulation points were sampled every 2 ps to match GaMD’s sampling frequency. Target number of walkers per bins nw was set to 4. CVs were set to be the dihedral angles χ1 − C14 and χ1 − C38 associated with the disulfide bond formed between cysteine 14 and cysteine 38.95,96 Bins were evenly spaced in intervals of 10° for both (ranging from −180 to 180°).
Figure 7a–c shows the free-energy landscape (−kBT ln P, where P denotes probability) of BPTI after one 500 ns run of brute force simulation, GaMD, and WE, respectively. The lowest energy state was set to be zero for each of the free-energy landscapes. In particular, the GaMD run with the upper bound of the dual boost potential yielded the maximum number of initial structures on average as compared to the other five GaMD potential settings (see Supporting Information). Maclaurin expansion to the 10th order was used for GaMD reweighting since using cumulant expansion to the second order is limited for small proteins with 40 residues or less.17 Figure 7b shows a free-energy landscape similar to the one obtained from accelerated molecular dynamics (aMD), which was also obtained using a dual boost potential and Maclaurin expansion to the 10th order for reweighting.96 Slight differences between the two free-energy landscapes may be from using different force field parameters: aMD used the modified Amber ff99SB-ILDN force field,97,98 which removed modifications to leucine, aspartic acid, and asparagine to mimic the Anton simulation of BPTI,99 and GaMD used the Amber ff14SB force field.91
Metastable states of interest, including the major state M and two minor or excited states mC14 and mC38,100 are marked in Figure 7a. M region was defined to be −120° < χ1 − C14 < 0°, 0° < χ1 − C38 < 120°, 0° < χ3 < 180°, mC14 was defined to be 0° < χ1 – C14 < 120°, 0° < χ1 – C38 < 120°, − 180° < χ3 < 0°, and mC38 was defined to be −120° < χ1 − C14 < 0°, −120° < χ1 − C38 < 0°, −180° < χ3 < 0° as in the BPTI aMD refs 95 and 96 Dihedral angle χ3 is associated with the disulfide bond formed between cysteine 14 and cysteine 38. It is clear that GaMD is more effective at exploring other metastable states present in BPTI including mC14 as compared to brute force simulation or WE. However, note that the other five GaMD runs yielded similar results as brute force simulation or WE (see the Supporting Information), and only the upper bound of the dual boost potential was able to explore the various metastable states. In addition, WE is better than the brute force simulation at sampling metastable states. However, even after extending the WE simulation for 2 μs (total simulation time: 2.5 μs), WE is still not able to sample mC14 as seen in Figure 7d. These results are similar to those in ref 96, which highlighted the fact that a 500 ns aMD simulation of BPTI was able to sample important metastable states and as much as a 1 ms brute force simulation of BPTI from Anton. This highlights the power of GaMD being able to sample more of the configuration space than WE. GaMD can sample orthogonal modes to the chosen CVs χ1 − C14 and χ1 − C38 since it is a CV-free enhanced sampling method. In contrast, WE mainly samples along the chosen CVs and can encounter difficulties in the sampling regions when there are orthogonal modes present to the chosen CVs. Although rate constants between these metastable states were not measured, it is expected that GaMD–WE will sample them significantly faster than conventional WE.
4. DISCUSSION
Three examples mentioned in the previous sections highlight how GaMD–WE can be more powerful than either GaMD or WE by itself. On the other hand, this hybrid method also reveals both methods’ advantages and disadvantages. GaMD is more effective at sampling the configuration space than WE by being a CV-free method. By adding boost potentials to “fill” the energy wells in a CV-free manner, GaMD can sample the configuration space more evenly across different modes in the system. On the other hand, WE is mainly limited to efficiently sampling along the chosen CVs. This is not problematic if the chosen CVs sufficiently describe the dynamics of the system, but in most cases, it is difficult to know the best CVs a priori. In such cases, WE (and other enhanced sampling methods that need CVs a priori) could miss sampling orthogonal modes and be slow at sampling the configuration space. For alanine dipeptide and chignolin, the chosen CVs have been commonly used in existing literature.17,59,88 They are small enough systems for brute force simulation, GaMD, and WE to sufficiently sample well. However, for BPTI, the dihedral angles χ1 − C14 and χ1 − C38 are not commonly used CVs and have shown to be insufficient for WE to sample as effectively as GaMD. BPTI is also a much bigger system than alanine dipeptide or chignolin. In this case, principal component analysis (PCA) vectors have been commonly used instead as CVs for BPTI.96,99,101
However, GaMD’s conventional reweighting method, which is used in GaMD–WE, is reliable for small systems up to 100 residues since the energetic noise becomes too high for accurate reweighting for larger systems.17 A longer simulation or many simulation frames for each conformation would be necessary for larger systems to get good statistics and low error for reweighting. Hence, reweighting has been done for each structural cluster with many simulation frames, instead of individual frames, for larger systems like G-protein-coupled receptors protein complexes.102 We plan to implement this for larger systems in our next installment of GaMD–WE.
Although WE could use PCA vectors as CVs in theory, prior brute force simulation data will be needed to accurately calculate PCA vectors, time-structure-based independent component analysis vectors, and other dimensionality reduction vectors to use them as CVs and describe the system.103,104 Nonetheless, WE can have nondifferentiable CVs such as the number of hydrogen bonds, which can be helpful for many systems. In contrast, other methods such as metadynamics and ABF need differentiable CVs. Moreover, WE does not add any statistical bias to the system and is exact regardless of the parameters,105 so it can reliably obtain the actual kinetics of the system. Since a long simulation is typically needed to get reasonable estimates of the kinetics, researchers have recently developed methods for WE to estimate the actual kinetics faster.60,106,107 If GaMD–WE is combined with these current methods, more improvements will be seen in obtaining thermodynamic and kinetic properties.
5. CONCLUSIONS
We have combined two well-established enhanced sampling methods, GaMD and WE, into a hybrid method GaMD–WE to create a more powerful enhanced sampling method for MD simulations. GaMD is used to sample the free-energy landscape initially, and WE is used to further sample the free-energy landscape and ascertain rate constants between two states of interest in the system of interest. We have shown how the hybrid method performs better than conventional WE in sampling thermodynamic and kinetic properties for two systems, and its performance significantly improves as the system size grows. We have also noted that running GaMD initially before WE will be beneficial for BPTI due to its greater coverage of the free-energy landscape. For future directions, we plan to fully integrate the hybrid method with a WE simulation toolkit such as WESTPA and possibly combine it with other WE enhancing algorithms to create state-of-the-art enhanced sampling methods. The GaMD–WE package is available at https://github.com/anandojha/gamd_we, and the documentation is available at https://gamd-we.readthedocs.io/en/latest/.
Supplementary Material
Supplementary information
2
ACKNOWLEDGMENTS
S.H.A. acknowledges support from NIH GM31749 and the University of California San Diego. A.A.O. acknowledges support from the fellowship from the Molecular Sciences Software Institute (MolSSI) under NSF grant OAC-1547580. R.E.A. acknowledges support from NSF XSEDE CHE060063. The authors thank Lillian Chong, Daniel Zuckerman, Jeremy Copperman, and Yinglong Miao for giving helpful suggestions and feedback and Saumya Thakur for help with the supplementary cover art. All simulations were done using the Triton Shared Computing Cluster (TSCC), San Diego Supercomputing Center (SDSC).
Figure 1. Representative pictures of the systems tested: (a) alanine dipeptide, (b) chignolin, and (c) BPTI.
Figure 2. Average free-energy landscape (−kBT ln P, where P denotes probability) of alanine dipeptide after (a) 50 ns of brute force simulations, (b) 50 ns of GaMD (with the upper bound of the dihedral boost potential), and (c) 50 ns of WE. (d,e) show the average free-energy landscape obtained after 12 μs of brute force simulation and WE (after cutting out the first 200 ns of simulation time to eliminate the initial structure bias), respectively. (e) shows the regions of interest (αR, αL, and PII) marked. The lowest energy state was set to be zero for each free-energy landscape.
Figure 3. Evolution of rate constants over aggregate simulation time for WE (in red), GaMD–WE (in blue), and GaMD–WE with equal weights (in magenta). The reference brute force values are in black. (a,b) show the rate constants between the two major metastable states αR and PII. (c,d) show the rate constants between αR and αL, a higher-energy region, and (e,f) show the rate constants between PII and αL.
Figure 4. Average free-energy landscape (−kBT ln P, where P denotes probability) of chignolin after (a) 500 ns of brute force simulation, (b) 500 ns of GaMD (with the upper bound of the dihedral boost potential), and (c) 500 ns of WE (d,e) (or (f)) show the average free-energy landscape obtained after 40 μs of brute force simulation and WE (after cutting out the first 2 μs of simulation time to eliminate the initial structure bias), respectively. (e,f) show the regions of interest marked. The lowest energy state was set to be zero for each free-energy landscape.
Figure 5. Evolution of rate constants over aggregate simulation time for WE (in red) and GaMD–WE (in blue). The reference brute force values are in black. (a,b) show the rate constants between the folded region and the unfolded region.
Figure 6. Evolution of rate constants over aggregate simulation time for WE (in red) and GaMD–WE (in blue). The reference brute force values are in black. (a,b) show the rate constants between the folded region and the intermediate I region. (c,d) show the rate constants between the folded region and the intermediate II region. (e,f) show the rate constants between the intermediate I region and the intermediate II region.
Figure 7. Free-energy landscape (−kBT ln P, where P denotes probability) of BPTI after (a) one 500 ns brute force simulation run, (b) one 500 ns run of GaMD (with the upper bound of dual boost potential) with metastable states of interest M, mC14, and mC38 marked, (c) one 500 ns run of WE, and (d) extending WE run for 2 μs (total simulation time: 2.5 μs), respectively. The lowest energy state was set to be zero for each of the free-energy landscapes.
Table 1. Alanine Dipeptide Rate Constants (in ns−1) after 12 μs of Simulation Timea
brute force WE GaMD–WE GaMD–WE with equal weights
αR → PII 6.81, [6.71, 6.90] 6.79 ± 0.052 6.54 ± 0.088 6.47 ± 0.12
PII → αR 2.97, [2.92, 3.02] 2.96 ± 0.055 2.88 ± 0.035 2.88 ± 0.040
αL → αR 0.30, [0.25, 0.36] 0.38 ± 0.25 0.34 ± 0.087 0.23 ± 0.14
αR → αL 0.0099, [0.0083, 0.012] 0.0083 ± 0.0011 0.0081 ± 0.0021 0.0092 ± 0.0017
αL → PII 0.33, [0.27, 0.40] 0.32 ± 0.020 0.33 ± 0.012 0.31 ± 0.019
PII → αL 0.0099, [0.0083, 0.012] 0.0085 ± 0.0014 0.0082 ± 0.0020 0.0090 ± 0.0018
a In the brute force simulation column, the first value indicates the average rate constant value, and the second value indicates the 95% confidence interval calculated from Bayesian bootstrapping. For WE and GaMD–WE, the error bars represent 95% confidence intervals calculated from the standard deviation of three independent runs.
Table 2. Chignolin Rate Constants after 40 μs of Simulation Timea
brute force [ns−1] WE [ns−1] GaMD–WE [ns−1]
folded → unfolded 1.25, [1.17, 1.33] 1.16 ± 0.056 1.26 ± 0.067
unfolded → folded 0.026, [0.024, 0.028] 0.037 ± 0.0049 0.027 ± 0.0049
folded → intermediate I 1.41, [1.32, 1.50] 1.26 ± 0.046 1.46 ± 0.089
intermediate I → folded 0.028, [0.026, 0.031] 0.041 ± 0.0056 0.029 ± 0.0045
folded → intermediate II 0.42, [0.40, 0.45] 0.44 ± 0.040 0.40 ± 0.020
intermediate II → folded 0.020, [0.019, 0.022] 0.024 ± 0.0039 0.021 ± 0.0019
intermediate I → intermediate II 1.15, [1.13, 1.17] 1.19 ± 0.037 1.20 ± 0.024
intermediate II → intermediate I 4.52, [4.46, 4.58] 4.67 ± 0.19 4.52 ± 0.17
a In the brute force simulation column, the first value indicates the average rate constant value, and the second value indicates the 95% confidence interval calculated from Bayesian bootstrapping. For WE and GaMD–WE, the error bars represent 95% confidence intervals calculated from the standard deviation of three independent runs.
Complete contact information is available at: https://pubs.acs.org/10.1021/acs.jctc.1c00770
The authors declare no competing financial interest.
REFERENCES
(1) Karplus M ; McCammon JA Molecular dynamics simulations of biomolecules. Nat. Struct. Biol 2002, 9 , 646–652.12198485
(2) Hollingsworth SA ; Dror RO Molecular dynamics simulation for all. Neuron 2018, 99 , 1129–1143.30236283
(3) Karplus M ; Kuriyan J Molecular dynamics and protein function. Proc. Natl. Acad. Sci. U.S.A 2005, 102 , 6679–6685.15870208
(4) Hospital A ; Goñi JR ; Orozco M ; Gelpí JL Molecular dynamics simulations: advances and applications. Adv. Appl. Bioinf. Chem 2015, 8 , 37.
(5) Durrant JD ; McCammon JA Molecular dynamics simulations and drug discovery. BMC Biol 2011, 9 , 71.22035460
(6) Dror RO ; Dirks RM ; Grossman JP ; Xu H ; Shaw DE Biomolecular simulation: a computational microscope for molecular biology. Annu. Rev. Biophys 2012, 41 , 429–452.22577825
(7) De Vivo M ; Masetti M ; Bottegoni G ; Cavalli A Role of molecular dynamics and related methods in drug discovery. J. Med. Chem 2016, 59 , 4035–4061.26807648
(8) Steinhauser M ; Hiermaier S A review of computational methods in materials science: examples from shock-wave and polymer physics. Int. J. Mol. Sci 2009, 10 , 5135–5216.20054467
(9) Reed EJ ; Soljačić M ; Gee R ; Joannopoulos JD Molecular dynamics simulations of coherent optical photon emission from shock waves in crystals. Phys. Rev. B: Condens. Matter Mater. Phys 2007, 75 , 174302.
(10) Maginn EJ ; Elliott JR Historical perspective and current outlook for molecular dynamics as a chemical engineering tool. Ind. Eng. Chem. Res 2010, 49 , 3059–3078.
(11) Xu J ; Li X ; Hou C ; Wang L ; Zhou G ; Ge W ; Li J Engineering molecular dynamics simulation in chemical engineering. Chem. Eng. Sci 2015, 121 , 200–216.
(12) Chodera JD ; Swope WC ; Pitera JW ; Dill KA Long-time protein folding dynamics from short-time molecular dynamics simulations. Multiscale Model. Simul 2006, 5 , 1214–1226.
(13) Levitt M Protein folding by restrained energy minimization and molecular dynamics. J. Mol. Biol 1983, 170 , 723–764.6195346
(14) Lindorff-Larsen K ; Piana S ; Dror RO ; Shaw DE How fast-folding proteins fold. Science 2011, 334 , 517–520.22034434
(15) Pan AC ; Jacobson D ; Yatsenko K ; Sritharan D ; Weinreich TM ; Shaw DE Atomic-level characterization of protein–protein association. Proc. Natl. Acad. Sci. U.S.A 2019, 116 , 4244–4249.30760596
(16) Zhu J ; Lv Y ; Han X ; Xu D ; Han W Understanding the differences of the ligand binding/unbinding pathways between phosphorylated and non-phosphorylated ARH1 using molecular dynamics simulations. Sci. Rep 2017, 7 , 12439.28963484
(17) Miao Y ; Feher VA ; McCammon JA Gaussian accelerated molecular dynamics: Unconstrained enhanced sampling and free energy calculation. J. Chem. Theory Comput 2015, 11 , 3584–3595.26300708
(18) Pang YT ; Miao Y ; Wang Y ; McCammon JA Gaussian accelerated molecular dynamics in NAMD. J. Chem. Theory Comput 2017, 13 , 9–19.28034310
(19) Miao Y ; McCammon JA Annual Reports in Computational Chemistry; Elsevier, 2017; Vol. 13 ; pp 231–278.29720925
(20) Wang J ; Arantes PR ; Bhattarai A ; Hsu RV ; Pawnikar S ; Huang Y.-m. M. ; Palermo G ; Miao Y Gaussian Accelerated Molecular Dynamics: Principles and Applications. Wiley Interdiscip. Rev.: Comput. Mol. Sci 2021, 11 , No. e1521.34899998
(21) Barducci A ; Bonomi M ; Parrinello M Metadynamics. Wiley Interdiscip. Rev.: Comput. Mol. Sci 2011, 1 , 826–843.
(22) Laio A ; Rodriguez-Fortea A ; Gervasio FL ; Ceccarelli M ; Parrinello M Assessing the accuracy of metadynamics. J. Phys. Chem. B 2005, 109 , 6714–6721.16851755
(23) Bussi G ; Laio A ; Parrinello M Equilibrium free energies from nonequilibrium metadynamics. Phys. Rev. Lett 2006, 96 , 090601.16606249
(24) Barducci A ; Bussi G ; Parrinello M Well-tempered metadynamics: a smoothly converging and tunable free-energy method. Phys. Rev. Lett 2008, 100 , 020603.18232845
(25) Laio A ; Gervasio FL Metadynamics: a method to simulate rare events and reconstruct the free energy in biophysics, chemistry and material science. Rep. Prog. Phys 2008, 71 , 126601.
(26) Torrie GM ; Valleau JP Nonphysical sampling distributions in Monte Carlo free-energy estimation: Umbrella sampling. J. Comput. Phys 1977, 23 , 187–199.
(27) Kästner J Umbrella sampling. Wiley Interdiscip. Rev.: Comput. Mol. Sci 2011, 1 , 932–942.
(28) Virnau P ; Müller M Calculation of free energy through successive umbrella sampling. J. Chem. Phys 2004, 120 , 10925–10930.15268122
(29) Warmflash A ; Bhimalapuram P ; Dinner AR Umbrella sampling for nonequilibrium processes. J. Chem. Phys 2007, 127 , 154112.17949137
(30) Darve E ; Rodríguez-Gómez D ; Pohorille A Adaptive biasing force method for scalar and vector free energy calculations. J. Chem. Phys 2008, 128 , 144120.18412436
(31) Comer J ; Gumbart JC ; Hénin J ; Leliev̀re T ; Pohorille A ; Chipot C The adaptive biasing force method: Everything you always wanted to know but were afraid to ask. J. Phys. Chem. B 2015, 119 , 1129–1151.25247823
(32) Darve E ; Pohorille A Calculating free energies using average force. J. Chem. Phys 2001, 115 , 9169–9183.
(33) Chipot C ; Pohorille A Springer Series in Chemical Physics; Springer, 2007; Vol. 86 ; pp 159–184.
(34) Leliev̀re T ; Rousset M ; Stoltz G Long-time convergence of an adaptive biasing force method. Nonlinearity 2008, 21 , 1155.
(35) Sugita Y ; Okamoto Y Replica-exchange molecular dynamics method for protein folding. Chem. Phys. Lett 1999, 314 , 141–151.
(36) Sindhikara D ; Meng Y ; Roitberg AE Exchange frequency in replica exchange molecular dynamics. J. Chem. Phys 2008, 128 , 024103.18205439
(37) Zhang W ; Wu C ; Duan Y Convergence of replica exchange molecular dynamics. J. Chem. Phys 2005, 123 , 154105.16252940
(38) Sindhikara DJ ; Emerson DJ ; Roitberg AE Exchange often and properly in replica exchange molecular dynamics. J. Chem. Theory Comput 2010, 6 , 2804–2808.26616081
(39) Rosta E ; Hummer G Error and efficiency of replica exchange molecular dynamics simulations. J. Chem. Phys 2009, 131 , 165102.19894977
(40) Miao Y Acceleration of biomolecular kinetics in Gaussian accelerated molecular dynamics. J. Chem. Phys 2018, 149 , 072308.30134710
(41) Stelzl LS ; Hummer G Kinetics from replica exchange molecular dynamics simulations. J. Chem. Theory Comput 2017, 13 , 3927–3935.28657736
(42) Buchete N-V ; Hummer G Peptide folding kinetics from replica exchange molecular dynamics. Phys. Rev. E: Stat., Nonlinear, Soft Matter Phys 2008, 77 , 030902.
(43) Tiwary P ; Parrinello M From metadynamics to dynamics. Phys. Rev. Lett 2013, 111 , 230602.24476246
(44) Faradjian AK ; Elber R Computing time scales from reaction coordinates by milestoning. J. Chem. Phys 2004, 120 , 10880–10889.15268118
(45) Vanden-Eijnden E ; Venturoli M ; Ciccotti G ; Elber R On the assumptions underlying milestoning. J. Chem. Phys 2008, 129 , 174102.19045328
(46) Bello-Rivas JM ; Elber R Exact milestoning. J. Chem. Phys 2015, 142 , 094102.25747056
(47) Májek P ; Elber R Milestoning without a reaction coordinate. J. Chem. Theory Comput 2010, 6 , 1805–1817.20596240
(48) Vanden-Eijnden E ; Venturoli M Markovian milestoning with Voronoi tessellations. J. Chem. Phys 2009, 130 , 194101.19466815
(49) Allen RJ ; Valeriani C ; Rein ten Wolde P Forward flux sampling for rare event simulations. J. Condens. Matter Phys 2009, 21 , 463102.
(50) Allen RJ ; Frenkel D ; Ten Wolde PR Forward flux sampling-type schemes for simulating rare events: Efficiency analysis. J. Chem. Phys 2006, 124 , 194111.16729807
(51) Valeriani C ; Allen RJ ; Morelli MJ ; Frenkel D ; Rein ten Wolde P Computing stationary distributions in equilibrium and nonequilibrium systems with forward flux sampling. J. Chem. Phys 2007, 127 , 114109.17887830
(52) Hussain S ; Haji-Akbari A Studying rare events using forward-flux sampling: Recent breakthroughs and future outlook. J. Chem. Phys 2020, 152 , 060901.32061206
(53) Van Erp TS ; Bolhuis PG Elaborating transition interface sampling methods. J. Comput. Phys 2005, 205 , 157–181.
(54) Moroni D ; van Erp TS ; Bolhuis PG Investigating rare events by transition interface sampling. Phys. A 2004, 340 , 395–401.
(55) Borrero EE ; Weinwurm M ; Dellago C Optimizing transition interface sampling simulations. J. Chem. Phys 2011, 134 , 244118.21721623
(56) Du W-N ; Marino KA ; Bolhuis PG Multiple state transition interface sampling of alanine dipeptide in explicit solvent. J. Chem. Phys 2011, 135 , 145102.22010733
(57) Huber GA ; Kim S Weighted-ensemble Brownian dynamics simulations for protein association reactions. Biophys. J 1996, 70 , 97–110.8770190
(58) Zuckerman DM ; Chong LT Weighted ensemble simulation: review of methodology, applications, and software. Annu. Rev. Biophys 2017, 46 , 43–57.28301772
(59) Zhang BW ; Jasnow D ; Zuckerman DM The “weighted ensemble” path sampling method is statistically exact for a broad class of stochastic processes and binning procedures. J. Chem. Phys 2010, 132 , 054107.20136305
(60) Copperman J ; Zuckerman DM Accelerated Estimation of Long-Timescale Kinetics from Weighted Ensemble Simulation via Non-Markovian “Microbin” Analysis. J. Chem. Theory Comput 2020, 16 , 6763–6775.32990438
(61) Bhatt D ; Zhang BW ; Zuckerman DM Steady-state simulations using weighted ensemble path sampling. J. Chem. Phys 2010, 133 , 014110.20614962
(62) Dickson A ; Brooks CL III WExplore: hierarchical exploration of high-dimensional spaces using the weighted ensemble algorithm. J. Phys. Chem. B 2014, 118 , 3532–3542.24490961
(63) Donyapour N ; Roussey NM ; Dickson A REVO: Resampling of ensembles by variation optimization. J. Chem. Phys 2019, 150 , 244112.31255090
(64) Ahn S-H ; Grate JW ; Darve EF Efficiently sampling conformations and pathways using the concurrent adaptive sampling (CAS) algorithm. J. Chem. Phys 2017, 147 , 074115.28830168
(65) Abdul-Wahid B ; Feng H ; Rajan D ; Costaouec R ; Darve E ; Thain D ; Izaguirre JA AWE-WQ: fast-forwarding molecular dynamics using the accelerated weighted ensemble. J. Chem. Inf. Model 2014, 54 , 3033–3043.25207854
(66) Yu L ; Rajan D ; Feng H ; Darve E ; Thain D ; Izaguirre JA , Folding Proteins at 500 ns/hour with Work Queue. 2012 IEEE 8th International Conference on E-Science, 2012; pp 1–8.
(67) Costaouec R ; Feng H ; Izaguirre J ; Darve E Analysis of the Accelerated Weighted Ensemble Methodology; Conference Publications, 2013; p 171.
(68) Adelman JL ; Grabe M Simulating current–voltage relationships for a narrow ion channel using the weighted ensemble method. J. Chem. Theory Comput 2015, 11 , 1907–1918.26392816
(69) Adhikari U ; Mostofian B ; Copperman J ; Subramanian SR ; Petersen AA ; Zuckerman DM Computational estimation of microsecond to second atomistic folding times. J. Am. Chem. Soc 2019, 141 , 6519–6526.30892023
(70) Saglam AS ; Chong LT Protein–protein binding pathways and calculations of rate constants using fully-continuous, explicit-solvent simulations. Chem. Sci 2019, 10 , 2360–2372.30881664
(71) Dixon T ; Lotz SD ; Dickson A Predicting ligand binding affinity using on-and off-rates for the SAMPL6 SAMPLing challenge. J. Comput-Aided Mol. Des 2018, 32 , 1001–1012.30141102
(72) Lotz SD ; Dickson A Unbiased molecular dynamics of 11 min timescale drug unbinding reveals transition state stabilizing inter-actions. J. Am. Chem. Soc 2018, 140 , 618–628.29303257
(73) Sztain T ; Ahn S-H ; Bogetti AT ; Casalino L ; Goldsmith JA ; Seitz E ; McCool RS ; Kearns FL ; Acosta-Reyes F ; Maji S ; Mashayekhi G ; McCammon JA ; Ourmazd A ; Frank J ; McLellan JS ; Chong LT ; Amaro RE A glycan gate controls opening of the SARS-CoV-2 spike protein. Nat. Chem 2021, 13 , 963–968.34413500
(74) Ahn S-H ; Grate JW ; Darve EF Investigating the role of non-covalent interactions in conformation and assembly of triazine-based sequence-defined polymers. J. Chem. Phys 2018, 149 , 072330.30134719
(75) Ahn S-H ; Grate JW Foldamer Architectures of Triazine-Based Sequence-Defined Polymers Investigated with Molecular Dynamics Simulations and Enhanced Sampling Methods. J. Phys. Chem. B 2019, 123 , 9364–9377.31603686
(76) Ahn S-H ; Jagger BR ; Amaro RE Ranking of ligand binding kinetics using a weighted ensemble approach and comparison with a multiscale milestoning approach. J. Chem. Inf. Model 2020, 60 , 5340–5352.32315175
(77) Huang Y.-m. M. ; McCammon JA ; Miao Y Replica exchange Gaussian accelerated molecular dynamics: Improved enhanced sampling and free energy calculation. J. Chem. Theory Comput 2018, 14 , 1853–1864.29489349
(78) Oshima H ; Re S ; Sugita Y Replica-exchange umbrella sampling combined with gaussian accelerated molecular dynamics for free-energy calculation of biomolecules. J. Chem. Theory Comput 2019, 15 , 5199–5208.31539245
(79) Chen H ; Fu H ; Chipot C ; Shao X ; Cai W Overcoming Free-Energy Barriers with a Seamless Combination of a Biasing Force and a Collective Variable-Independent Boost Potential. J. Chem. Theory Comput 2021, 17 , 3886.34106706
(80) Ray D ; Stone SE ; Andricioaei I Markovian Weighted Ensemble Milestoning (M-WEM): Long-time Kinetics from Short Trajectories. 2021, bioRxiv:2021.06.26.450057.
(81) Russo JD ; Copperman J ; Zuckerman DM Iterative trajectory reweighting for estimation of equilibrium and non-equilibrium observables. 2020, arXiv preprint arXiv:2006.09451
(82) Trapl D ; Horvacanin I ; Mareska V ; Ozcelik F ; Unal G ; Spiwok V Anncolvar: approximation of complex collective variables by artificial neural networks for analysis and biasing of molecular simulations. Front. Mol. Biosci 2019, 6 , 25.31058167
(83) Fiorin G ; Klein ML ; Hénin J Using collective variables to drive molecular dynamics simulations. Mol. Phys 2013, 111 , 3345–3362.
(84) Sittel F ; Stock G Perspective: Identification of collective variables and metastable states of protein dynamics. J. Chem. Phys 2018, 149 , 150901.30342445
(85) Case DA ; Belfon K ; Ben-Shalom I ; Brozell SR ; Cerutti D ; Cheatham T ; Cruzeiro VWD ; Darden T ; Duke RE ; Giambasu G ; Amber 2020, 2020.
(86) Phillips JC ; Hardy DJ ; Maia JDC ; Stone JE ; Ribeiro JV ; Bernardi RC ; Buch R ; Fiorin G ; Hénin J ; Jiang W ; Scalable molecular dynamics on CPU and GPU architectures with NAMD. J. Chem. Phys 2020, 153 , 044130.32752662
(87) Zwier MC ; Adelman JL ; Kaus JW ; Pratt AJ ; Wong KF ; Rego NB ; Suárez E ; Lettieri S ; Wang DW ; Grabe M ; WESTPA: An interoperable, highly scalable software package for weighted ensemble simulation and analysis. J. Chem. Theory Comput 2015, 11 , 800–809.26392815
(88) Bogetti AT ; Mostofian B ; Dickson A ; Pratt A ; Saglam AS ; Harrison PO ; Adelman JL ; Dudek M ; Torrillo PA ; DeGrave AJ ; A suite of tutorials for the WESTPA rare-events sampling software [Article v1. 0]. Living J. Comp. Mol. Sci 2019, 1 , 10607.
(89) Pratt A ; Suárez E ; Zuckerman DM ; Chong LT Extensive Evaluation of Weighted Ensemble Strategies for Calculating Rate Constants and Binding Affinities of Molecular Association/Dissociation Processes. 2019, bioRxiv:671172
(90) Eastman P ; Swails J ; Chodera JD ; McGibbon RT ; Zhao Y ; Beauchamp KA ; Wang L-P ; Simmonett AC ; Harrigan MP ; Stern CD ; OpenMM 7: Rapid development of high performance algorithms for molecular dynamics. PLoS Comput. Biol 2017, 13 , No. e1005659.28746339
(91) Maier JA ; Martinez C ; Kasavajhala K ; Wickstrom L ; Hauser KE ; Simmerling C ff14SB: improving the accuracy of protein side chain and backbone parameters from ff99SB. J. Chem. Theory Comput 2015, 11 , 3696–3713.26574453
(92) MacKerell AD Jr ; Bashford D ; Bellott M ; Dunbrack RL Jr ; Evanseck JD ; Field MJ ; Fischer S ; Gao J ; Guo H ; Ha S ; All-atom empirical potential for molecular modeling and dynamics studies of proteins. J. Phys. Chem. B 1998, 102 , 3586–3616.24889800
(93) Onufriev A ; Bashford D ; Case DA Modification of the generalized Born model suitable for macromolecules. J. Phys. Chem. B 2000, 104 , 3712–3720.
(94) Horn HW ; Swope WC ; Pitera JW ; Madura JD ; Dick TJ ; Hura GL ; Head-Gordon T Development of an improved four-site water model for biomolecular simulations: TIP4P-Ew. J. Chem. Phys. 2004, 120 , 9665–9678.15267980
(95) Xue Y ; Ward JM ; Yuwen T ; Podkorytov IS ; Skrynnikov NR Microsecond time-scale conformational exchange in proteins: using long molecular dynamics trajectory to simulate NMR relaxation dispersion data. J. Am. Chem. Soc 2012, 134 , 2555–2562.22206299
(96) Pierce LCT ; Salomon-Ferrer R ; Augusto F. de Oliveira C ; McCammon JA ; Walker RC Routine access to millisecond time scale events with accelerated molecular dynamics. J. Chem. Theory Comput 2012, 8 , 2997–3002.22984356
(97) Hornak V ; Abel R ; Okur A ; Strockbine B ; Roitberg A ; Simmerling C Comparison of multiple Amber force fields and development of improved protein backbone parameters. Proteins: Struct., Funct., Bioinf 2006, 65 , 712–725.
(98) Lindorff-Larsen K ; Piana S ; Palmo K ; Maragakis P ; Klepeis JL ; Dror RO ; Shaw DE Improved side-chain torsion potentials for the Amber ff99SB protein force field. Proteins: Struct., Funct., Bioinf 2010, 78 , 1950–1958.
(99) Shaw DE ; Maragakis P ; Lindorff-Larsen K ; Piana S ; Dror RO ; Eastwood MP ; Bank JA ; Jumper JM ; Salmon JK ; Shan Y ; Atomic-level characterization of the structural dynamics of proteins. Science 2010, 330 , 341–346.20947758
(100) Grey MJ ; Wang C ; Palmer AG Disulfide bond isomerization in basic pancreatic trypsin inhibitor: multisite chemical exchange quantified by CPMG relaxation dispersion and chemical shift modeling. J. Am. Chem. Soc 2003, 125 , 14324–14335.14624581
(101) Ernst M ; Sittel F ; Stock G Contact-and distance-based principal component analysis of protein dynamics. J. Chem. Phys 2015, 143 , 244114.26723658
(102) Wang J ; Miao Y Mechanistic insights into specific G protein interactions with adenosine receptors. J. Phys. Chem. B 2019, 123 , 6462–6473.31283874
(103) Pérez-Hernández G ; Paul F ; Giorgino T ; De Fabritiis G ; Noé F Identification of slow molecular order parameters for Markov model construction. J. Chem. Phys 2013, 139 , 015102.23822324
(104) Schwantes CR ; Pande VS Improvements in Markov state model construction reveal many non-native interactions in the folding of NTL9. J. Chem. Theory Comput 2013, 9 , 2000–2009.23750122
(105) Aristoff D ; Zuckerman DM Optimizing weighted ensemble sampling of steady states. Multiscale Model. Simul 2020, 18 , 646–673.34421402
(106) DeGrave AJ ; Bogetti AT ; Chong LT The RED scheme: Rate-constant estimation from pre-steady state weighted ensemble simulations. J. Chem. Phys 2021, 154 , 114111.33752378
(107) Torrillo PA ; Bogetti AT ; Chong LT A minimal, adaptive binning scheme for weighted ensemble simulations. J. Phys. Chem. A 2021, 125 , 1642–1649.33577732
|
PMC008xxxxxx/PMC8983024.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
101313252
34584
J Vis Exp
J Vis Exp
Journal of visualized experiments : JoVE
1940-087X
34958089
8983024
10.3791/63291
NIHMS1783777
Article
Visualizing Cytoskeleton-Dependent Trafficking of Lipid-Containing Organelles in Drosophila Embryos
Kilwein Marcus D.
Welte Michael A.
Department of Biology, University of Rochester, RC Box 270211, Rochester, NY 14627, USA
Corresponding Author: Michael A. Welte ([email protected])
9 3 2022
13 12 2021
13 12 2021
13 12 2022
178 10.3791/63291This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Early Drosophila embryos are large cells containing a vast array of conventional and embryo-specific organelles. During the first three hours of embryogenesis, these organelles undergo dramatic movements powered by actin-based cytoplasmic streaming and motor-driven trafficking along microtubules. The development of a multitude of small, organelle-specific fluorescent probes (FPs) makes it possible to visualize a wide range of different lipid-containing structures in any genotype, allowing live imaging without requiring a genetically encoded fluorophore. This protocol shows how to inject vital dyes and molecular probes into Drosophila embryos to monitor the trafficking of specific organelles by live imaging. This approach is demonstrated by labeling lipid droplets (LDs) and following their bulk movement by particle image velocimetry (PIV). This protocol provides a strategy amenable to the study of other organelles, including lysosomes, mitochondria, yolk vesicles, and the ER, and for tracking the motion of individual LDs along microtubules. Using commercially available dyes brings the benefits of separation into the violet/blue and far-red regions of the spectrum. By multiplex co-labeling of organelles and/or cytoskeletal elements via microinjection, all the genetic resources in Drosophila are available for trafficking studies without the need to introduce fluorescently tagged proteins. Unlike genetically encoded fluorophores, which have low quantum yields and bleach easily, many of the available dyes allow for rapid and simultaneous capture of several channels with high photon yields.
SUMMARY:
In the early Drosophila embryo, many organelles are motile. In principle, they can be imaged live via specific fluorescent probes, but the eggshell prevents direct application to the embryo. This protocol describes how to introduce such probes via microinjection, and then analyze bulk organelle motion via particle image velocimetry.
pmcINTRODUCTION:
Vital dyes and molecular probes are powerful tools to image specific cellular structures and organelles live. In the Drosophila embryo, many different organelles display cytoskeleton-driven localization1–4, but the application of these small molecules is challenging because the eggshell is impermeable to many of them. This protocol describes a method to use fluorescent probes (FPs) in live embryos via microinjection in order to detect large-scale trafficking of organelles. The procedure covers preparation of the injection solution, egg collection and preparation of embryos, microinjection, imaging, and image analysis.
Dramatic spatial rearrangements of organelles are common in many animal oocytes, eggs, and embryos, in part because of the large size of these cells. In the Drosophila embryo, for example, lipid droplets (LDs) and yolk vesicles move toward the embryo center just before cellularization5. This motion depends on microtubules and leaves a ~40 μm region all around the periphery of the embryo depleted of the two organelles. During earlier cleavage stages, many organelles are transported by cytoplasmic flows that are driven by actin-myosin-based contractions at the embryo surface6. Although the embryos of many species exhibit similar rearrangements, the Drosophila embryo is particularly suited for following these processes by imaging because it develops externally at standard laboratory ‘room temperature’, is relatively transparent, small enough to fit on most microscope setups, and can be manipulated using powerful genetic tools.
For some organelles, fluorescently tagged proteins are available that specifically label these structures. For example, LSD-2 (also known as dPLIN2) is a protein that in embryos specifically targets LDs7. Fly lines are available that carry either inducible transgenes encoding a fusion between Green Fluorescent Protein (GFP) and LSD-28,9 or a gene trap in which yellow fluorescent protein (YFP) is inserted into the coding region of the endogenous LSD-2 gene10. However, this approach has limitations, including that these fusion proteins have low quantum yields and tend to bleach easily. In addition, labeling multiple different structures simultaneously can be challenging: for many organelles, only one type of fluorescent tag (often GFP or mCherry) is currently available, so imaging two organelles at the same time may require new transgenes or insertions; also, even if compatible tags are available, introducing them into a single strain can require time-consuming crosses. It also makes using the many powerful genetic resources less convenient, e.g., two organelle markers, a Gal4 driver, and an inducible RNAi construct all have to be present in the same mother.
In principle, these limitations can be overcome with the use of FPs, including vital dyes (e.g., LysoTracker to mark lysosomes), molecular probes (e.g., SiR-tubulin to label microtubules), and fluorescently labeled biological molecules (e.g., C12 BODIPY to probe fatty acid metabolism11). From use in cultured cells, they are typically well-validated as powerful tools for probing cellular biology. FPs are versatile, have superior photo properties, and are compatible with fluorescent proteins. Multiple dyes can be mixed and applied simultaneously, often with the benefits of separation into the violet/blue and far-red areas of the spectrum and small Stokes shifts, preventing channel bleed through. Small Stokes shifts allow for simultaneous capture of multiple imaging channels, enabling the tracking of several organelles at once. Finally, they can equally be applied to labeling organelles in the embryos of other Drosophila species or even other insects where no fluorescently tagged proteins may be available.
However, most of these FPs cannot traverse the elaborate eggshell of the Drosophila embryo. It consists of five layers: three outer chorionic layers (chorion) that prevent mechanical damage plus a waxy layer surrounding the vitelline membrane that creates a chemical barrier12. For simplicity, the combination of the waxy layer and the vitelline membrane will be referred to as the “vitelline membrane” below. To bypass the eggshell, this protocol adapts an established embryo microinjection approach to introduce FPs into the Drosophila embryo. The protocol describes how to monitor the cytoplasmic flow of LDs in cleavage stage embryos. It includes the preparation of injection needles and egg collection cages, the process of egg collection, and mechanical removal of the chorion. It goes over how to microinject and image the embryos and how to analyze the bulk flow of LDs using particle image velocimetry (PIV, adapted from6). It provides advice on troubleshooting to ensure embryo survival and create the best system for the image. Also discussed is how the protocol can be modified to simultaneously image LDs and microtubules or to apply it to the study of other organelles, including lysosomes, mitochondria, yolk vesicles, and the endoplasmic reticulum (ER).
PROTOCOL:
1. Prepare necessary materials
NOTE: These preparations are best done days or weeks ahead of time.
1.1. Prepare injection needles.
NOTE: Needles can be stored indefinitely in a covered container. Needles must be fine enough to deliver ~700 fL while being strong enough to pierce the vitelline membrane.
1.1.1. Place capillary into the capillary holder on the needle puller. Secure the capillary with the wingnuts. Align the center of the capillary with the heating element so that two symmetric needles are generated.
1.1.2. Choose appropriate settings on the needle puller according to the manufacturer’s instructions.
1.1.3. Perform quality control using a dissecting scope.
NOTE: Needle tips should be as fine as possible. Cracked, jagged, or large-bore needle tips are discarded. Initial quality control can be done using a dissecting scope.
1.1.4. Prepare batches of 10 needles (5 capillaries worth) at a time.
1.1.5. Place a deformable putty, which has been rolled up into a cylinder, at the center of the bottom of a container with a lid. Carefully press the needles into the putty such that it holds them horizontally with the needle tips safely suspended in the air to prevent damage. Cover with the lid, and store until use.
1.2. Prepare egg collection plates. Store at 4 °C.
NOTE: Egg collection plates are small agar plates supplemented with fruit juice to encourage egg-laying. How to prepare them is described in many protocols, including13.
1.3. Prepare heptane glue.
NOTE: This glue is extracted from double-sided tape and is used to securely attach the embryos to a coverslip. It stops the embryo from floating out of focus during injection and imaging. The glue can be prepared days or weeks ahead of time.
1.3.1. Ball up double-sided tape to a size bigger than a golf ball and place it at the bottom of a ~50 mL glass container with a tightly sealing lid. Pack tape tightly so enough adhesive will be present.
1.3.2. In a fume hood, fill the glass container with heptane to cover all the tape. Keep the container tightly closed when not in use.
CAUTION: Heptane is flammable as a liquid and vapor. It can cause eye, skin, and respiratory tract irritation. Breathing vapors may cause drowsiness, dizziness, and lung damage. Keep heptane away from sources of ignition and store it in a well-ventilated area meant for flammables and away from incompatible substances.
1.3.3. Let the heptane glue sit overnight or longer. For best results, place the container on a shaker or another agitator overnight to help with dissolving the adhesive.
NOTE: The tape itself will remain, but the adhesive will be dissolved in the heptane. In step 5.1.2, the glue is applied to a coverslip; heptane evaporates, leaving the glue behind.
1.3.4. Test the heptane glue preparation by placing a drop of it on a glass slide and making sure a sticky residue remains once the heptane has evaporated. If adhesive residue is not visible afterward, add more tape to the glass container and repeat the previous step.
NOTE: Once prepared, the glue can be used for months or even years.
1.4. Prepare a 1 mg/mL (3.8 mM) BODIPY493/503 stock solution by diluting commercially obtained preparation in pure anhydrous DMSO. Keep it covered to protect it from light and water. Store indefinitely at −20 °C or short term at 4 °C.
2. Prepare egg collection cages
NOTE: Do this at least a day (preferably 2 or 3 days) before the planned injection(s).
2.1. Prepare yeast paste.
NOTE: Yeast paste supplements protein and other vital nutrients not provided in the apple juice plates. It promotes egg production and egg-laying.
2.1.1. Add 2–10 g of dry baker’s yeast, and then 1 mL of tap water to a small beaker. Mix using a spatula.
2.1.2. Keep adding tap water in 1 mL increments until the desired, tooth-paste-like consistency is reached.
2.1.3. Cover the mixture and store it at 4 °C.
2.2. Set up fly cages for egg collection.
NOTE: Male and female flies of the desired genotype are required: 10–50 females less than 2 weeks old and a similar number of males. A number of different options for fly cages are available, including homemade options14,13. It is important that the size of the apple juice plates is matched to the size of the fly cages.
2.2.1. Place a small smear of yeast paste on an apple juice plate. Keep it covered and let it come to room temperature because flies will not lay eggs on the plate if it is too cold.
2.2.2. Transfer flies to an embryo collection cage, seal with the yeasted apple juice plate, and secure the plate to the embryo collection cage.
2.2.3. Allow newly transferred flies to acclimate to the cage for 1–2 days, replacing the plate with a fresh one daily. If all the yeast has been eaten by the next day, increase the amount of yeast paste for future collections.
NOTE: This is an important step to increase egg yield as well-fed females lay more eggs.
3. On the day of injection, prepare the injection solution and load it into the needle.
3.1. Use the stock solution of BODIPY 493/503 (3.8 mM in DMSO) as the injection solution.
3.2. Load a single needle with 1 μL of the injection solution using needle loading tips.
NOTE: Strive to get the liquid all the way to the tip without trapping air bubbles. Be ready to quickly replace the loaded needle in case of accidental breakage.
3.2.1. Attach a loading tip to a micropipette and draw 1 μL of the injection solution into the tip. Using gloves, hold the needle in one hand with its tip pointed away to minimize the risk of breakage.
3.2.2. Carefully insert the loading tip into the needle and push it close to the needle tip. Dispense the liquid near the needle tip. After removing the pipette, hold the needle vertically with its tip down until the liquid flows to the tip.
3.3. Store the loaded needles in a separate container with putty as above (see step 1.1.5).
NOTE: Needles should be prepared in advance (up to several hours) of injection but should not be used after more than a day.
3.4. Keep needles out of ambient light to prevent bleaching of the dyes. Cover the storage container with aluminum foil without damaging the needle tips. Make sure all light is obscured.
4. Collecting embryos for injection
NOTE: The timing of collection depends on what stage of embryos the injection needs to be performed at. With the timing scheme below, the embryos at the time of injection will be 0–90 min old, which corresponds to cleavage stages15.
4.1. On the day of injection, prepare apple juice plates with yeast, as in step 2.2.1. If N rounds of injections are planned, prepare N + 2 plates. Keep these plates at room temperature.
NOTE: In addition to the N plates to collect embryos for injection, one plate is needed as a pre-collection plate and another one to feed the flies in the cage once collections are completed.
4.2. Replace the plate on the cage with a fresh yeasted plate (pre-collection plate) and leave it on the cage for 1–2 h.
4.3. Replace the plate on the cage with a fresh plate (collection plate). Discard the pre-collection plate, as it will contain embryos older than desired. Then, allow the flies to lay eggs for 1.5 h.
Note: As well-fed flies typically lay their eggs shortly after the eggs have been fertilized, a 1.5 h collection time assures that at the end of the collection, most of the embryos on the plate are 0–90 min old, i.e., are in cleavage stages. Female flies that have not been fed fresh yeast since the previous day will tend to retain fertilized egg for some indeterminate amount of time before laying. Hence, the pre-collection plate may have embryos that were fertilized before the start of collection and thus are older than 60 min, sometimes much older.
4.4. Replace the plate on the cage with a fresh yeasted plate. Cover the collection plate so that stray flies do not lay eggs on them.
5. Prepare embryos for microinjection
5.1. Assemble the materials needed.
5.1.1. Attach a piece of double-sided tape to a glass slide. Avoid touching the tape with fingers as that reduces its stickiness.
NOTE: This slide will be used for removing the chorion and not for imaging.
5.1.2. Using a transfer pipette or a micropipettor (p200 or p1000), place a small drop (200 μL or less) of heptane glue roughly in the center of a rectangular coverslip (60 × 25 mm). Allow the heptane to evaporate (this takes less than a min).
NOTE: This coverslip will be used to mount the embryos for injection and imaging. The dimensions are chosen to fit into an adjustable metal holder on the confocal microscope.
5.1.3. Assemble a desiccation chamber, a sealed chamber (e.g., a Tupperware sandwich box) containing desiccation beads. Only use desiccation beads that have not hydrated.
NOTE: To ensure the embryos take up the injection solution, the embryo must be slightly desiccated to reduce internal pressure. This is done by placing the coverslip with the dechorionated, air-exposed embryos into the desiccation chamber.
5.2. Mechanically remove the chorion.
5.2.1. Cover the appropriately aged embryo plate with a thin layer of Halocarbon Oil 27 to turn the eggshell transparent.
NOTE: Embryos become translucent within tens of seconds15. If this step does not occur efficiently, the apple juice plates may be too wet and need to be dried off before use.
5.2.2. View the plate under a dissection scope with transillumination (i.e., the light going through the plate into the eyepieces) to confirm the stage of the embryos.
5.2.3. Select embryos in cleavage stages.
NOTE: Cleavage stage embryos are entirely opaque; the subsequent blastoderm stages can be recognized by a band of transparent cytoplasm all around the opaque center15. A good introduction on how to recognize various embryonic stages is available on the website (given in reference16) maintained by the Society of Developmental Biology.
5.2.4. Using fine tweezers, grab an embryo of the desired stage by its dorsal appendages and transfer it onto the prepared glass slide covered with a piece of double-sided tape. Place the embryo on the tape. Minimize the transfer of oil.
5.2.5. As gently as possible, roll the embryo across the surface of the tape by gently nudging the embryo with the side of the tip of the tweezers. Do not poke the embryo directly with the sharp tweezer tips.
NOTE: If the force applied is too low, the embryo will not roll. If it is too high, it will burst. Finding the appropriate intermediate force requires experience and, thus, this step should be practiced beforehand.
5.2.6. Continue rolling the embryo until the chorion transiently adheres to the tape and cracks.
5.2.7. Once the chorion cracks, keep rolling to separate the embryo (still inside the vitelline membrane) from the chorion as the chorion remains stuck to the tape.
5.2.8. Confirm that the chorion is completely removed by observing the loss of the dorsal appendages from the embryo.
5.2.9. Roll the embryo back onto the chorion, which is less adhesive than the tape. Gently rub the embryo with the tweezers until it attaches to the tweezers for transfer.
5.2.10. Transfer the embryo to the coverslip with the heptane glue and bring it in contact with the glue, which typically detaches it from the tweezers.
5.2.11. Adjust the orientation of the embryo on the coverslip.
5.2.11.1. Embed the lateral surface of the embryo into the glue.
NOTE: The surface of the embryo embedded into the glue is the one that will be imaged. Embedding the lateral surface is the simplest, but this can be adjusted depending on personal preference or the biological question to be answered.
5.2.11.2. Orient the long axis of the embryo perpendicular to the long axis of the coverslip.
5.2.11.3. If the embryo does not lay in the preferred orientation, clean the tweezers to remove any oil that would detach the embryo from the glue. Then, gently attempt to roll the embryo into position.
5.2.12. Prepare the desired number of embryos for injection in the manner described above.
NOTE: As time passes after removal of the chorion, the ambient air will begin to desiccate the embryos and thus the effect of step 5.3 will be uneven for the embryos on the coverslip, which were dechorionated at different times. Typically, 1–3 embryos are the most manageable.
5.3. Desiccate embryos.
5.3.1. Place the coverslip with embryos into the prepared desiccation chamber and seal it.
5.3.2. Allow the embryos to desiccate for 5–12 min.
NOTE: The timing of this step is dependent on the ambient temperature and humidity and thus needs to be determined empirically.
5.3.3. Remove the coverslip from the chamber and place a drop of Halocarbon Oil 700 onto it, fully covering the embryos, to prevent further desiccation.
5.3.4. Inspect the embryos on a dissecting scope to judge proper desiccation.
5.3.5. If the embryos are just slightly shriveled, but not deflated, proceed to microinjection (step 6).
5.3.6. If the embryos are not sufficiently or overly desiccated, return to step 5.2 and prepare a new batch of embryos as the Halocarbon Oil 700 cannot be removed. If the embryos were overly desiccated, shorten desiccation time for the next batch of embryos by 3 min; if they were under-desiccated, increase desiccation time by 3 min.
6. Microinject embryos
NOTE: Ensure the microinjection setup includes an inverted microscope, a micromanipulator to hold and position the injection needle, and a commercial microinjector to deliver controlled volumes.
6.1. Power on the microinjector and input the preferred settings.
NOTE: For this protocol, it is recommended to use linear motion output and the fast setting, but many others will work. The operator should determine personal preference.
6.2. Load the needle into the micromanipulator. To prevent the needle from being damaged while loading the coverslip containing the embryos; move it out of the way into a safe position.
6.3. Place the coverslip onto the stage, with embryos on top, toward the needle. Then, carefully move the needle back into position for injection, with its tip still well above the stage.
6.4. Put the 4x objective into the light path. Using the focus knob on the microscope, bring the embryos into focus.
NOTE: This will be the focal plane used for injection and will be adjusted only minimally going forward.
6.5. Using the stage controls, move the embryos horizontally into the center of the field of view.
6.5.1. Pan away from the embryos, keeping them at the edge of the field of view, in preparation for lowering the needle. Stay in the same focal plane to ensure the needle is lowered to the correct position.
6.6. Lower the needle tip into the correct focal plane and ensure that it is visible in the field of view together with the embryos.
6.6.1. Using the micromanipulator controls and looking from above (not yet through the eyepieces), slowly drop the needle tip into the oil, aiming for the area where the objective points to. As the oil is quite viscous, go slowly to avoid damaging the needle.
6.6.2. Once the needle is visible through the eyepieces, continue using the micromanipulator controls until the needle tip is in focus and at the center of the field of view.
6.6.3. By moving the stage, bring the embryos back to the center of the field of view. Use the stage control, micromanipulator controls, and focus knob to fine-tune the position of the embryo and needle tip relative to each other while both are in focus. Aim to perform injections as close to the coverslip as possible.
6.7. Perform needle quality control.
6.7.1. To ensure that the needle is functional, dispense some of the injection solution into the Halocarbon Oil 700 surrounding the embryo, visible as a bubble in the oil.
6.7.2. If nothing is flowing out of the needle, gradually increase pressure at the injector. If this does not work, get a new needle.
6.8. Inject the embryo.
NOTE: Acceptable injection volumes range from ~0.06–1 pL. The lower limit is set by what can be visualized entering the embryo. The upper limit is set by the trauma the injection exerts on the embryo.
6.8.1. Inject along the lateral edges of the embryo as this is the least invasive.
6.8.2. Using the stage controls, move the embryo toward the needle tip until the latter gently punctures the embryo and enters it.
6.8.3. At the same time, start the flow of injection solution and watch for the appearance of a transient clear spot at the site of the needle tip, which indicates a successful transfer of liquid into the embryo.
6.8.4. Monitor the embryo. If the embryo resists the needle and ruptures (cytoplasm squirts out) upon entry, return to step 5 and increase the desiccation time. If the embryo is flat against the coverslip and appears floppy during the injection, return to step 5, and shorten the desiccation time.
6.8.5. If no dye entered the embryo, confirm that dye can still freely flow from the needle as in step 6.7. If the dye does not flow, replace the needle. If the dye does flow, inject a new embryo, and begin releasing the dye just before the needle punctures the embryo.
NOTE: Repeated injections of the same embryo are not recommended as it ruptures the previous wound/injection site.
6.9. Repeat until all the embryos are injected.
7. Image embryos
NOTE: This protocol utilizes a laser scanning confocal microscope.
7.1. Place coverslip into the metal holder on the confocal microscope so that the embryos are imaged directly from below through the coverslip; above the coverslip, there is only oil and no other barrier.
7.2. As a quality control step, image first using the epifluorescence function on the confocal microscope, with a 40x objective. Make sure the fluorophore is visible in the embryo before proceeding.
7.3. If the desired plane of imaging is different from the site of injection, allow enough time for the dye to diffuse to the target area.
NOTE: For BOPIPY 493/503, this time typically ranges from 30–60 min. As diffusion time depends greatly on the dye, other dyes may require optimizing the site of injection and waiting time.
7.4. Use different objectives for imaging at different scales. Use a 40x objective to image all of the embryo and a 63x objective to image subsections for smaller scales.
7.5. For live imaging during the syncytial stages before cellularization, use the following conditions to start with: image size of 512 × 512 pixels, line average of 3, frame rate of 0.1 frames per second (i.e., 1 frame acquired every 10 s). Adjust conditions according to the capabilities of the microscope employed.
NOTE: These conditions allow acquiring ~500+ images from BOPIPY 493/503 and capturing most of the motion.
8. Analyze LD flow by first using FIJI to prepare a time series of images, and then python for PIV analysis
8.1. Optimize image acquisition.
8.1.1. Perform injections of the desired dye at the appropriate stage. Repeat this step until day-to-day variation is negligible.
8.1.2. Optimize image acquisition settings including magnification, zoom, resolution, and frame rate.
NOTE: There is a tradeoff between high-quality images (resolution + line averaging) and acquisition speed (frame rate).
8.2. Prepare data in FIJI.
8.2.1. Acquire a high-quality time series to be analyzed.
8.2.2. Open one frame of the time series to generate a mask with FIJI. Duplicate the image. Convert the Type of the image to an 8 bit.
8.2.3. Define the boundary of the embryo using the Polygon or Freehand Selection Tool. Use Clear Outside to set pixel values outside the selection to 0. Clear and then Invert within the selection to set the value of the pixels within the selection to the maximum value (255).
8.2.4. Unselect, then use the Histogram command to ensure that only pixel values of 0 and 255 are present.
8.2.5. Save this new image mask for the specific time series.
8.2.6. Open the time series of interest. Choose which ten frames best capture the time of interest, for example, nuclear cycle 9 at the onset of the cortical contraction. Duplicate the ten frames of interest, forming a substack.
8.2.7. Use the Stack to Images function to generate 10 images to analyze with PIV.
8.2.8. Save the individual files in a manner that preserves their order (SeriesX_1, SeriesX_2, SeriesX_3 …).
8.3. Use the prepared mask and individual frames for PIV analysis, employing the provided sample python script (Supplemental data) or a custom script generated by the user.
NOTE: The provided script is based on the python application of OpenPIV17. It outputs distance in pixels which can be converted using the pixel width and frame rate to generate speeds or velocities.
8.4. Generate replicates by completing the previous steps for additional embryos and compiling the output values.
REPRESENTATIVE RESULTS:
Following injection, the dye will be localized only at the site where the needle tip was inserted. The dye will then diffuse away from the injection site depending on its diffusive characteristics. Figure 1 shows injection of BODIPY 493/503, soon after injection (panel A) and 24 min later (panel B). After 24 min, the dye has made it to roughly the midpoint of the embryo’s long axis.
Analyzing organelle motility can be achieved through dye injections and time-lapse imaging. In Figure 2, an embryo was co-injected with BODIPY 493/503 (Figure 2A) and LysoTracker Red (Figure 2B) and imaged using laser excitation at 488 and 596 nm, respectively. This embryo was then time-lapse imaged (one frame every 30 s for 30 min, 5 min analyzed). The time series was then run through PIV analysis, the output of which is shown via streamlines in Figure 2A,B. Note that the streamlines do not represent the trajectory of individual particles, but the cytoplasmic flow is inferred from analyzing all the particles in that region of the cytoplasm. Through labeling of two independent cellular structures (LDs and acidic organelles), the PIV analysis finds similar flows, with both labels converging on the central region of the embryo where the cytoplasm is flowing into the embryo’s interior6.
Currently available FPs allow the labeling of many other organelles and cellular structures. Figure 3 shows the labeling of the ER via ER tracker Green. ER tracker provides a nice resolution of the nuclear envelope, allowing visualization of major cell cycle stages. ER tracker Green is imaged using 488 nm excitation.
Labeling of the mitochondria is tricky, as most dyes tested seem to be trapped in the first mitochondrion they enter. On the other hand, no sign of dye toxicity was detected, making it possible to follow labeled mitochondria through cellularization and the ectodermal nuclear cycle 15. Figure 4 shows an ectodermal cell several hours post-injection with Mitoview 633 (excitation wavelength 633 nm).
BODIPY, Lysotracker, and LipidSpot are robust and can be used to acquire ~500+ images at 512 × 512, line average 3, frame rates from 1/s to 0.1/s. ER tracker Green, SIR-tubulin, and the mitochondrial dyes mentioned are less robust and yield ~50–200 images under the same conditions.
Starting at blastoderm stages, LDs move bidirectionally along microtubules, powered by the opposite-polarity motors kinesin-1 and cytoplasmic dynein5. This motion can be visualized by co-labeling LDs and microtubules via injecting both BODIPY 493/503 and SiR-Tubulin (Figure 5). As LDs frequently reverse their direction of movement (as they switch between kinesin-1 and cytoplasmic dynein), higher frame rates during acquisition better capture critical details of LD motility.
Live imaging of autofluorescent yolk vesicles is possible without any form of dye injection (Figure 6). However, the autofluorescence is dim, and the excitation laser is phototoxic. Thus, live imaging of autofluorescent yolk has a poor signal-to-noise ratio relative to dye injection.
DISCUSSION:
The Drosophila embryo is a powerful and convenient model to study fundamental questions in cellular and organismal biology. Its relative simplicity, powerful genetics, and small size make it an excellent system for imaging both cellular processes and development. Here, a standard microinjection protocol is adapted to enable FP usage in embryos. This approach allows for fluorescent imaging of specific cellular structures without the need for genetically encoded fluorophores, opening many genetic backgrounds to imaging. Combining multiple dyes plus strategically chosen fluorescently tagged proteins can open multichannel live imaging spanning the whole spectrum of visible light.
Critical steps in the protocol:
This protocol uses BODIPY 493/503 to label LDs. This approach can easily be adapted to mark other cellular structures. For subsequent image analysis, one of the most important factors is the signal-to-noise ratio, i.e., the brightness of the dye compared to the background signal. Lysosomes have been successfully imaged (LysoTracker Red, 1 mM), as well as mitochondria (Mitoview 633, 200 μM), the ER (ER tracker Green, 10 μM), and microtubules (SiR tubulin, 200 nM in DMSO), as shown in Figures 2–5. In addition, yolk vesicles are autofluorescent and give off blue light upon UV excitation (image using 405 nm as the excitation wavelength (Figure 6)). For other dyes, aim for the dye concentration to be 100–1,000x of what would be needed for staining cultured cells live; this is similar to the concentration of a stock solution that would be diluted into cell culture media. As this protocol calls for an injection of 100 fL and the Drosophila embryo is roughly 9 nL in volume18, these dye concentrations will average out to an internal embryonic concentration of under 1/100th of what is present in the cell culture media. Temporarily, the local concentration will be higher at the site of injection, which is most relevant for FPs that do not diffuse well (i.e., mitochondrial dyes and SIR-tubulin). For these FPs, start at the recommended high concentrations; if unexpected death is observed, successively dilute two-fold until an acceptable compromise between survival and signal strength is reached.
When co-injecting multiple dyes, both dyes should either be in the same solvent, or both the solvents and dyes should be compatible with the mixture (alcohol concentrations exceeding ~10% are not recommended).
The quality of the needles is essential for the success of this procedure, as the tip needs to be as fine as possible. Otherwise, damage from the injection wound can compromise the subsequent development of the embryo. As commercial needle pullers differ, it is important to follow the suggestions of the manufacturer and try out multiple pulling parameters until the desired shape is achieved. It is critical to perform the quality control step 1.1.3 as working with a cracked, jagged, or large bore needle tip will make the successful injection more difficult or even impossible.
The embryo needs to be partially desiccated so that additional volumes of liquid can be added during the injection. If the embryo is under-desiccated, the needle will not enter easily, and cytoplasm will squirt out as the needle penetrates or as the solution is injected. If the embryo is over-desiccated, it will look deflated and will not develop properly. The exact drying time depends on local conditions, e.g., air humidity, and can change from day to day. It has to be determined empirically for each session.
The ability of the embryo to survive after microinjection depends critically on the quality of the needle, proper desiccation, and limiting injection volume to less than 1 pL (ideally 100 fL). As long as these parameters are optimized, no significant toxicity is apparent when the described dyes are injected at the recommended concentrations. If embryos survive the desiccation and injection steps, they typically develop successfully well into germ-band extension, an exception being the microtubule and ER probes which caused cellularization defects at high levels (100 fL injection of the stock concentration of each). Testing found no obvious developmental defects when DMSO, water, and mixtures of the two were injected at the recommended volumes, with an embryo survival rate through the germ-band extension of ~75% or more. Injection volumes more than over 1 pL caused defects and embryos injected with ~4 pL volumes developed for less than 1 h. Therefore, injection volumes need to be kept low, which means that dye concentrations have to be high.
Generally, injections along the lateral edge of the embryo are recommended as those result in the least damage. However, the injection site may need to be adjusted depending on the diffusive properties of the FPs employed. BODIPY 493/503 and LysoTracker diffuse faster across the entire embryo than Lipid Spot 610 (another dye to mark LDs), while Sir-Tubulin and Mito View 633 never diffuse fully across the embryo (imaging as late as 7 h post-injection). Thus, injection in or near the site of interest may be necessary. When injecting in the anterior or posterior regions, a particularly fine needle is recommended.
Image acquisition relies on confocal microscopy to optically section and resolve small organelles and all cytoskeletal components. Techniques requiring analysis of many images (e.g., STORM or PALM) will not work because the embryonic contents are in motion and the fluorophores are not optimized for photoswitching. Epifluorescence microscopy lacks the lateral and axial resolution to make out most organelles and smaller cellular structures. For these reasons, it is strongly recommended to use a confocal microscope or employ light sheet technology.
The reproducibility of the image analysis relies greatly on consistent imaging data. For the greatest chance of success, optimization of the injection technique and image acquisition is required. Establishing and practicing a technique where the dye(s) of interest, site of injection, age of the embryo, injection volume, and acquisition setup are all consistent will generate the most robust data for image analysis.
Modifications and troubleshooting of the method:
This protocol demonstrates a method for analyzing the bulk flow of LDs in cleavage stage embryos using particle image velocimetry. The same approach can be used for other organelles, other developmental stages, and other analysis methods. For example, Figure 2 shows analysis of LDs and acidic organelles flowing in the syncytial stages of embryogenesis, visualized by co-injecting BODIPY 493/503 and LysoTracker Red. Further successful imaging of LD motion in embryos up to 7 h post-fertilization has been achieved; these embryos do retain the injection wound but are able to develop for several hours.
Data gathered using this protocol has been used for particle image velocimetry, but many other analysis techniques are available. For example, particle tracking programs like those found in ImageJ, Imaris, or manual tracking can be used to obtain velocities and directionalities of moving structures. Note that most such tracking software are built to work with data from planar cell culture systems and do not always adapt well to 3D structures like the Drosophila embryo. Further, for the generation of the best quality particle tracking data, multiple Z planes would need to be imaged; this should be feasible if image stack acquisition times are under ~2 s. This benchmark should be reachable on spinning disc confocal, lattice light sheet, and recent laser scanning confocal systems. However, the feasibility of particle tracking for abundant organelles such as LDs, mitochondria, and lysosomes is low as the amount of positive signal in a field of view is too high for the current tracking methods. Tracking of less abundant structures like nuclei or yolk vesicles may be possible. PIV for flow analysis works well for LDs and acidic organelles in cleavage stages because both organelles move freely. Organelles like nuclei, ER, and mitochondria are tethered to other cellular structures and thus do not move freely and are not suited to software analysis that assumes free motion. The investigator should pick the techniques best suited to the organelle of interest.
During the syncytial and cellular blastoderm stages, LDs (as well as some other organelles) move along radially oriented microtubules5. It is therefore possible to find cross-sectional views (like in Figure 5) where single microtubules are in focus for long distances, allowing particle tracking in 2D. Since these optical planes are deep within the embryo, overall signal strength is diminished, and the signal-to-noise ratio is reduced.
For tracking analysis, imaging as fast as possible can reveal crucial details of the motion and thus of the motile machinery. For example, lipid-droplet motion is a mixture of two motile states, slow-short motion (~200 nm/s; average travel distance ~100 nm) and fast-long motion (~450 nm/s; average travel distance ~1,000 nm)19; thus, if images are taken every second or even less frequently, the slow-short state becomes undetectable. However, frequent imaging also induces fluorophore bleaching and phototoxicity. Imaging conditions, therefore, have to be adjusted depending on the exact question to be addressed.
Limitations of the method:
Depending on the dye desired, the method can be limited by the compatibility between the dye solubility and the toxicity of the injection solution. Alcohols like isopropanol and ethanol are difficult to handle within a needle due to their lower viscosity and appear to damage cellular components and kill the embryo.
The method is also not well suited for visualizing the earliest steps in embryogenesis because it takes 30+ min to prepare the embryos for injection. At room temperature, the initial cell cycles of the embryo are just ~10 min long each; so, even if one were to pick a newly fertilized egg in step 5.2.3, the first few cell cycles would already be completed by the time the embryo is ready for imaging.
Illumination with light in the UV/blue range is considerably more phototoxic than for longer wavelengths. Under such conditions (e.g., to follow autofluorescent yolk vesicles; Figure 6), one has to limit imaging time (leading to shorter time series) or use lower laser power (resulting in a reduced signal-to-noise ratio).
After cellularization, dyes injected in a specific location tend to diffuse poorly, as they must traverse many cell membranes. This limits the region of observation in later developmental stages.
The significance of the method with respect to existing/alternative methods:
The motion of LDs and other lipid-containing organelles in early embryos can be visualized with genetically encoded fluorophores, label-free techniques, and by the introduction of FPs. The latter can be achieved by permeabilization of the vitelline membrane12 or the microinjection approach discussed here.
Genetically encoded fluorophores are versatile markers whose levels are typically highly reproducible from embryo to embryo. However, they have lower quantum yields and bleach more easily than FPs. Typically, they are only available in one or two tagged version(s) (e.g., GFP or mCherry), limiting the choice of which structures can be imaged simultaneously. FPs, on the other hand, often exist in a large variety; for example, various lipid-droplet specific dyes are available with emission spectra from Autodot in the UV/blue spectrum to Lipidtox and LipidSpot 610 in the far-red spectrum. FPs can also be directly applied to any strain of interest, and thus do not require strain construction to, for example, introduce the desired organelle marker into a mutant strain of interest. This advantage is particularly pronounced when multiple structures are to be labeled simultaneously; instead of time-consuming crosses spanning multiple generations, this can be achieved in a single day by mixing the relevant dyes and introducing them at the same time. Finally, if cellular processes are to be probed with pharmacological inhibition, drugs and dyes can be introduced together.
Label-free methods are very powerful approaches for detecting specific cellular structures. For example, LDs can be specifically detected in early embryos by third-harmonic generation microscopy20 or by femtosecond Stimulated Raman Loss microscopy21. Like FPs, these approaches can be applied in any genetic background, and because they do not cause bleaching, they potentially allow for faster image acquisition. However, they are typically limited to specific organelles and thus do not by themselves support multiplex imaging; they also require specialized microscopes.
There are two general strategies for introducing small molecules into embryos. One is the microinjection approach employed here; the other is chemical (terpene) treatment to permeabilize the vitelline membrane. The latter approach12 is less involved than microinjection, but also more variable from embryo to embryo. In addition, after permeabilization, the protection provided by the vitelline membrane is compromised and the embryo proper is accessible to the external medium, making it more challenging to keep it alive. Microinjection is much less likely to derail embryonic development than permeabilization. However, permeabilization is recommended if many embryos need to be monitored simultaneously, e.g., for drug screening purposes. To follow the movement of cellular structures and obtain reproducible image series suitable for image analysis, microinjection is the method of choice.
Importance and potential applications of the method in specific research areas:
The Drosophila embryo is an important model system for studying many cell-biological and developmental processes1,5,6. Tagging organelles with fluorescent proteins has made major contributions to the understanding of how the early embryo develops, how various organelles traffic, and how such trafficking is modulated developmentally and genetically. However, their propensity to bleach and the challenges of generating strains in which multiple organelles are labeled with different colors limit the application of this approach. The use of FPs introduced by microinjection solves many of these challenges and can even be combined with fluorescently tagged proteins. This technique allows for the imaging of multiple organelles, cell structures, and cytoskeletal components in any genetic background. As a result, several genotypes can be compared via live imaging, making it possible to determine the effect of mutations on the trafficking of multiple organelles.
This protocol demonstrates the FP injection approach for embryos of Drosophila melanogaster, but in principle, this approach applies to any insect eggs for which microinjection techniques have been established, including other species of Drosophila, crickets22, and aphids23.
Supplementary Material
Supplemental data - python script
ACKNOWLEDGMENTS:
We thank Pakinee Phromsiri, Brian Jencik, Jinghong (James) Tang, and Roger White for their comments on the manuscript. We thank Patrick Oakes, Stefano Di Talia, and Victoria Deneke for sharing their expertise on how to perform PIV analysis. This work was supported by National Institutes of Health grants F31 HD100127 (to M. D. K.) and R01 GM102155 (to M. A. W).
Figure 1: Diffusion of BODIPY 493/503 through the embryo.
The dye was injected along the lateral edge toward the anterior end (top right) and diffuses from this injection site into the embryo, labeling LDs. (A) The dye has diffused through portions of the embryo adjacent to the injection site. (B) Roughly 24 min/2 nuclear cycles later, the dye has diffused past the midpoint of the embryo. Scale bar: 100 μm. A 1024 × 1024 frame (line average 4) was acquired every 30 s.
Figure 2: Particle image velocimetry (PIV) for LDs and acidic organelles.
An embryo was injected with both BODIPY 493/503 and LysoTracker Red. (A) BODIPY channel. (B) LysoTracker Red channel. (A’,B’) Streamline diagrams generated by PIV analysis of the two channels generated from 10 sequential frames, including those shown in A and B. A’ corresponds to the flow of LDs, and B’ corresponds to the flow of acidic organelles. Note that both A’ and B’ show a left-of-center confluence where embryonic contents are flowing out of the plane of view, into the center of the embryo. Also, note that BODIPY has diffused more than LysoTracker as different dyes have different diffusive properties. Scale bar: 100 μm. A 1024 × 1024 frame (line average 4) was acquired every 30 s.
Figure 3: ER tracker labels syncytial nuclear divisions.
A syncytial blastoderm embryo was microinjected with ER tracker and a portion of its surface was imaged over time. (A) Spindle assembly during a nuclear division. (B) Abscission of the nuclear envelope during the same division. (C) Subsequent interphase. (D) The onset of the next division is indicated by centrosome appearance (occurrence of circular ER-free regions, marked by arrowheads). Note the gradual dye bleaching. Excitation wavelength: 488 nm. Scale bar: 5 μm. A: initial frame, B: 3 min elapsed, C: 10 min elapsed, D: 13 min elapsed. A 1024 × 1024 frame (line average 4) was acquired every 30 s.
Figure 4: Mitoview 633 labeling of mitochondria.
An embryo was injected with Mitoview 633 during the syncytial blastoderm stage and imaged 4 h later, after cellularization. The image shows a neuroectodermal cell of an embryo in germ-band extension. Scale bar: 5 μm. A 1024 × 1024 frame (line average 4) was acquired every 30 s.
Figure 5: Co-labeling of LDs and microtubules.
A cellularizing embryo was injected with a mixture of BODIPY 493/503 (yellow A,B,C) and SiR Tubulin (magenta A’,B’,C’). A”, B”, C” show the merged channels. Panels A, A’, and A” show the initial frame, panels B, B’ and B” show the frame after 5 s, and panels C, C’ and C’’ show the frame after 10 s. Scale bar: 5 μm. A 512 × 512 frame (line average 3) was acquired every 2.5 s.
Figure 6: Imaging yolk vesicle autofluorescence during syncytial cleavage stages. (A) At the start of the acquisition. (B) After 8 min. (C) After 16 min. Excitation wavelength: 405 nm. Low excitation intensity was used to keep the embryo alive. Scale bar: 100 μm. A 1024 × 1024 frame (line average 4) was acquired every 30 s.
A complete version of this article that includes the video component is available at http://dx.doi.org/10.3791/63291.
DISCLOSURES:
The authors have no conflicts of interest to disclose.
REFERENCES:
1. Chowdhary S , Tomer D , Dubal D , Sambre D , Rikhy R Analysis of mitochondrial organization and function in the Drosophila blastoderm embryo. Scientific Reports. 7 (1 ), 5502 (2017).28710464
2. Mavor LM Rab8 directs furrow ingression and membrane addition during epithelial formation in Drosophila melanogaster. Development. 143 (5 ), 892–903 (2016).26839362
3. Riggs B Actin cytoskeleton remodeling during early Drosophila furrow formation requires recycling endosomal components Nuclear-fallout and Rab11. Journal of Cell Biology. 163 (1 ), 143–154 (2003).
4. Welte MA , Gross SP , Postner M , Block SM , Wieschaus EF Developmental regulation of vesicle transport in Drosophila embryos: forces and kinetics. Cell. 92 (4 ), 547–557 (1998).9491895
5. Welte MA As the fat flies: The dynamic lipid droplets of Drosophila embryos. Biochimica et Biophysica Acta. 1851 (9 ), 1156–1185 (2015).25882628
6. Deneke VE Self-organized nuclear positioning synchronizes the cell cycle in Drosophila embryos. Cell. 177 (4 ), 925–941 e917 (2019).30982601
7. Welte MA Regulation of lipid-droplet transport by the perilipin homolog LSD2. Current Biology. 15 (14 ), 1266–1275 (2005).16051169
8. Bailey AP Antioxidant role for lipid droplets in a stem cell niche of Drosophila. Cell. 163 (2 ), 340–353 (2015).26451484
9. Stephenson RA Sequestration to lipid droplets promotes histone availability by preventing turnover of excess histones. Development. 148 (15 ) (2021).
10. Lowe N Analysis of the expression patterns, subcellular localisations and interaction partners of Drosophila proteins using a pigP protein trap library. Development. 141 (20 ), 3994–4005 (2014).25294943
11. Rambold AS , Cohen S , Lippincott-Schwartz J Fatty acid trafficking in starved cells: regulation by lipid droplet lipolysis, autophagy, and mitochondrial fusion dynamics. Developmental Cell. 32 (6 ), 678–692 (2015).25752962
12. Rand MD , Kearney AL , Dao J , Clason T Permeabilization of Drosophila embryos for introduction of small molecules. Insect Biochemistry and Molecular Biology. 40 (11 ), 792–804 (2010).20727969
13. Tran SL , Welte MA In-vivo centrifugation of Drosophila embryos. Journal of Visualized Experiments: JoVE. (40 ) (2010).
14. Rothwell WF , Sullivan W Drosophila embryo collection. CSH Protocols. 2007 pdb prot4825 (2007).
15. Wieschaus E , Nüsslein-Volhard C in Drosophila: A Practical Approach (ed Roberts DB ) Oxford University Press. 179–214 (1998).
16. Brody TB The Interactive Fly: Stages of Development and Mitotic Domains, <https://www.sdbonline.org/sites/fly/aimain/2stages.htm> (1999).
17. Liberzon A OpenPIV/openpiv-python: OpenPIV - Python (v0.22.2) with a new extended search PIV grid option.
18. Markow TA , Beall S , Matzkin LM Egg size, embryonic development time and ovoviviparity in Drosophila species. Journal of Evolutionary Biology. 22 (2 ), 430–434 (2009).19032497
19. Gross SP , Welte MA , Block SM , Wieschaus EF Dynein-mediated cargo transport in vivo. A switch controls travel distance. Journal of Cell Biology. 148 (5 ), 945–956 (2000).
20. Debarre D Imaging lipid bodies in cells and tissues using third-harmonic generation microscopy. Nature Methods. 3 (1 ), 47–53 (2006).16369553
21. Dou W , Zhang D , Jung Y , Cheng JX , Umulis DM Label-free imaging of lipid-droplet intracellular motion in early Drosophila embryos using femtosecond-stimulated Raman loss microscopy. Biophysical Journal. 102 (7 ), 1666–1675 (2012).22500767
22. Barry SK Injecting Gryllus bimaculatus Eggs. Journal of Visualized Experiments: JoVE. (150 ) (2019).
23. Le Trionnaire G An integrated protocol for targeted mutagenesis with CRISPR-Cas9 system in the pea aphid. Insect Biochemistry and Molecular Biology. 110 , 34–44 (2019).31015023
|
PMC008xxxxxx/PMC8983026.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
9215515
20498
Neuroimage
Neuroimage
NeuroImage
1053-8119
1095-9572
34933123
8983026
10.1016/j.neuroimage.2021.118790
NIHMS1779554
Article
Deep Learning for Alzheimer's Disease: Mapping Large-scale Histological Tau Protein for Neuroimaging Biomarker Validation
Ushizima Daniela 123
Chen Yuheng 4
Alegro Maryana 124
Ovando Dulce 4
Eser Rana 4
Lee WingHung 4
Poon Kinson 4
Shankar Anubhav 4
Kantamneni Namrata 4
Satrawada Shruti 4
Amaro Edson Junior 5
Heinsen Helmut 56
Tosun Duygu 78
Grinberg Lea T. 1459
1 Bakar Institute for Computational Health Sciences, University of California San Francisco, CA, USA
2 Berkeley Institute for Data Science, University of California Berkeley, CA, USA
3 Lawrence Berkeley National Laboratory, Berkeley, CA, USA
4 Department of Neurology, University of California San Francisco, San Francisco, CA, USA
5 University of Sao Paulo Medical School, Sao Paulo, Brazil
6 Julius-Maximilians University Würzburg, Würzburg, Germany
7 Department of Radiology, University of California San Francisco, San Francisco, CA, USA
8 Veterans Affairs San Francisco, CA, USA
9 Department of Pathology, University of California San Francisco, San Francisco, CA, USA
Correspondence: [email protected], 675 Nelson Rising Lane. PO Box 1207, San Francisco, CA, 94158, USA
18 2 2022
3 2022
20 12 2021
01 9 2022
248 118790118790
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Abnormal tau inclusions are hallmarks of Alzheimer's disease and predictors of clinical decline. Several tau PET tracers are available for neurodegenerative disease research, opening avenues for molecular diagnosis in vivo. However, few have been approved for clinical use. Understanding the neurobiological basis of PET signal validation remains problematic because it requires a large-scale, voxel-to-voxel correlation between PET and (immune)histological signals. Large dimensionality of whole human brains, tissue deformation impacting co-registration, and computing requirements to process terabytes of information preclude proper validation. We developed a computational pipeline to identify and segment particles of interest in billion-pixel digital pathology images to generate quantitative, 3D density maps. The proposed convolutional neural network for immunohistochemistry samples, IHCNet, is at the pipeline's core. We have successfully processed and immunostained over 500 slides from two whole human brains with three phospho-tau antibodies (AT100, AT8, and MC1), spanning several terabytes of images. Our artificial neural network estimated tau inclusion from brain images, which performs with ROC AUC of 0.87, 0.85, and 0.91 for AT100, AT8, and MC1, respectively. Introspection studies further assessed the ability of our trained model to learn tau-related features. We present an end-to-end pipeline to create terabytes-large 3D tau inclusion density maps co-registered to MRI as a means to facilitate validation of PET tracers.
Machine learning
deep learning
convolutional neural networks
Alzheimer's disease
histopathology
digital pathology
big data
imaging
pmcIntroduction
An estimated 1 out of 9 people over the age of 65 has Alzheimer's (AD) (2021), and the cost of dementia surpassed the cost of heart conditions and cancer (Kelley et al., 2015). Neurodegenerative diseases feature abnormal protein deposits and neuronal loss that progressively overtake an expanding landscape of brain areas in stereotypical patterns, which provides the neuropathological basis of clinical staging systems (Seeley, 2017). AD shows a stereotypical spread of phospho-tau and Aβ deposits (Montine et al., 2012), which could only be detected in postmortem examination until recently.
Positron emission tomography (PET) radioligands allow spatially-resolved imaging of cerebral metabolism, neuroreceptors, and, more recently, pathological protein deposits (Dani et al., 2016; Tiepolt et al., 2019; Wang and Edison, 2019). PET-based labeling of tau and Aβ hold the promise to facilitate understanding of neurodegenerative diseases by enabling (differential) diagnosis, monitoring progression starting from preclinical stages, and evaluating therapeutic efficacy of new drugs in living patients (Rabinovici et al., 2019; Tiepolt et al., 2019).
From the early days of radiolabeling, rigorous criteria based on pharmacokinetics, binding properties in vitro, and match between autoradiography results with target protein burden measured by immunohistochemistry determine efficiency and suitability for novel tracers (Klunk, 2018).
However, the F.D.A requires postmortem correlation studies for testing clinical utility before approving PET tracers for clinical use. In vivo studies with humans testing "suitable" tau and Aβ tracer candidates soon started to show unexpected results. For instance, Aβ PET imaging shows white matter retention. Several phospho-tau tracers show PET signal in basal ganglia, areas that lack respective protein deposits in aging and most neurodegenerative diseases. Also, in vivo studies show signal retention in tauopathies to which the candidate tau tracer has a low affinity in vitro (Leuzy et al., 2019; Soleimani-Meigooni et al., 2020). Besides showing off-target binding, tau and Aβ tracers seem to lack sensibility to detect protein deposits at earlier disease stages (Leuzy et al., 2019; Salloway et al., 2017; Soleimani-Meigooni et al., 2020). Because of technical limitations, most postmortem correlation studies either compare PET signal with semi-quantitative pathological stagings, such as Braak stage, CERAD score, and Thal phases of β-amyloid deposition or compare regional PET signal with the density of targeted protein of approximately similar spatial areas instead of performing voxel-to-voxel comparison (Curtis et al., 2015; Leuzy et al., 2019; Pontecorvo et al., 2020; Sabri et al., 2015; Salloway et al., 2017). As a result, many candidate tracers have been stuck in experimental stages for years as their sensibility, accuracy, and specificity rates, the nature of the off-target binding, the influence of aging, and comorbid pathologies in their binding properties remain poorly understood (Barrio, 2018; Lemoine et al., 2018; Marquie et al., 2017; Smith et al., 2017). Lack of reliable validated means to directly assign molecular signals detected by PET imaging to the corresponding neural microstructures they originate from, persist as a pivotal barrier to widely incorporating PET tracers into clinical pipelines for neurodegenerative diseases.
We developed an end-to-end solution for performing large-scale, voxel-to-voxel correlations between PET and high-resolution histological signals using open-source resources and magnetic resonance imaging (MRI) as the common registration space. Our semi-automated computational pipeline introduces a wide range of computer vision techniques, modern deep learning (DL) algorithms, and high-performance computing (HPC) capabilities to process images of large-scale histological datasets to generate quantitative 3D maps of abnormal protein deposits of whole human brains. We provide a design rationale, a detailed description of the protocol, and results of validation and quality control steps supporting the robustness of our computational methods. The pipeline tacks two main limitations precluding histology to neuroimaging voxel-to-voxel correlations: inherited non-linear tissue deformation during processing that challenges co-registration and immense computational and wet lab capacity required to generate quantitative 3D protein maps of the whole human brain. To illustrate the utility of our pipeline for calculating anatomical priors to validate molecular neuroimaging methods accurately, we applied it to two postmortem human brains, harboring moderate and severe Alzheimer's disease (AD) pathology. We immunostained over 500 whole mount human brain slides and created 3D histology maps co-registered to the corresponding MRI volume of three different forms of abnormal tau protein, spanning several Gigabytes of data. Quality control steps showed excellent agreement between DL-based segmentation of tau inclusions against manual segmentation results based on the area under the receiver operating characteristic (ROC) curve and Dice coefficient metrics. This pipeline represents a way forward to decrease the time between PET tracer development and clinical approval in neurodegenerative diseases and possibly, other conditions.
METHODS
UCSF IRB approved this study under #13-12079. Algorithms used in this study can be found in (https://github.com/grinberglab/high-res-3D-tau). Figure 1 illustrates the workflow steps.
Specimen Procurement, Histological Processing and brain slabbing
We tested the entire pipeline using two whole human brains. The first specimen, Case 1, belonged to a 88 years-old cognitively normal donor with moderate AD pathology (AD neuropathologic change A2B2C1 score (Montine et al., 2012)). The second specimen, Case 2, belonged to a 76 years-old donor diagnosed with dementia and severe AD pathology (AD neuropathologic changes A3B3C3 score). Case 1 underwent postmortem MPRAGE MRI acquisition, while Case 2 underwent postmortem SPGE acquisition within 10 hours of death (Fig. 1a). To minimize tissue deformation, upon autopsy, we stored the specimens upside down, hanging by the circle of Willis for three days. After that, we mounted them in plastic skulls, 3D-printed from the patient's computerized tomography (CT) images. Fixation continued for 18 days inside a bucket filled with buffered 4% paraformaldehyde (Alegro et al., 2016). Next, the specimens were embedded in celloidin. The brains were sectioned in serial sets of coronal slides, each set containing four 160 μm-thick sections. The section thickness was selected to balance the risk of tissue tear and the efficiency of antibody penetration. Representative 2 x 1” celloidin samples were cut for paraffin embedding for neuropathological diagnosis (Alegro et al., 2016). During sectioning, digital photographs were acquired directly from the blockface following each stroke using a high-definition, computer-controlled DSLR camera (EOS 5D Mark II, Canon, Tokyo, Japan) mounted on a copy stand arm (Kaiser Fototechnik, Germany) (Fig. 1b). During processing, brain tissue shrank about 30 to 40%. Thus, each coronal PET scan voxel would contain at least two complete histological sets, opening the opportunity to probe multiple antibodies per voxel. Details of tissue processing celloidin embedding and cutting have been described by us elsewhere (Alegro et al., 2016; Alegro et al., 2017; Theofilas et al., 2014).
CT and MRI acquisition
Postmortem structural T1-weighted MR imaging of Case 1 was performed on a 3T Siemens Skyra MRI system with a transmit and 32-channel receive coil using a 3D MPRAGE T1-weighted sequence with the following parameters: TR/TE/TI = 2300/2.98/900ms, 176 sagittal slices, within plane FOV = 256×240mm2, voxel size = 1×1×1mm3, flip angle = 9°, bandwidth = 240Hz/pix.
Postmortem structural T1-weighted MR imaging of Case 2 was performed on a GE Discovery 3T MR750 system with a transmit and 32-channel receive coil using a 3D SPGR T1-weighted sequence with the following parameters: TI = 400ms, 200 sagittal slices, within plane FOV = 256×256mm2, voxel size = 1×1×1mm3, flip angle = 11°, bandwidth = 31.25Hz/pix.
CT was acquired in life with a slice thickness of 1.25 mm, reconstruction diameter of 300 mm, and image matrix of 512x512 and only used as a model for the 3D printed skull (Fig. 1b)
Immunohistochemical labeling of abnormal tau protein inclusions
We created 3D histological maps for two widely used antibodies against phospho-tau (AT100 (pThr212+Ser214, Thermo), and AT8 (pSer202+Thr205, Thermo)) and one antibody that detects tau conformational changes (MC1, gift of Peter Davies). As a rule, 1st sections from each set were immunostained with AT8, 2nd sections from even-numbered sets were immunostained with AT100, and 3rd sections from even-numbered sets were immunostained with MC1. Free-floating immunohistochemistry reactions were done in batches of 50 (Fig. 1d). Each batch contained positive and negative controls for comparison. The controls were serial sections from a single subject not belonging to the study. Also, we excluded batches in which positive control showed low-quality staining, and in those, the staining process was repeated on the 2nd set of contiguous tissue sections. We performed single-label immunohistochemistry to avoid signal mixing and mounted each histological section on 6" x 4" glass slides.
Whole Slide Imaging
We built a cost-effective whole slide scanner (Fig. 1e, Fig. 2) to accommodate our histological slides (4" x 6"), which do not fit into regular microscope stages or cannot be fully imaged due to short stage travel range. The hardware comprises a high-precision, 6" travel range, industrial XY stage (Griffin Motion), an Olympus manual focusing box, color CCD camera (Qimaging Micro publisher 6), and 5.5X machine vision objective (Navitar Zoom 6000) mounted directly on the camera. Illumination is performed by a lightbox with diffuser, mounted on top of the XY stage. Sections were loaded directly to a 3D printed slide mount fixed on top of the lightbox. At 6.75x magnification, the objective field of view was 3.28 x 2.6 mm. The scanner was controlled by software developed in-house using Macro Manager 2.042, which has a user interface that allows defining the region-of-interest (ROI), performs white balance, and select lens magnification parameters. Macro Manager 2.0 computes the image tiles coordinates necessary to cover the selected ROI and synchronizes the XY stage movements with image capture. TeraStitcher (Bria and Iannello, 2012), which is capable of working with several Gigabytes of data while maintaining a small memory footprint were used to stitch full resolution tiles (1.22μm/ pixel resolution) and create histological images of each slide (Fig. 3). A 10% resolution version of each slide image was also created during the stitching process for visual quality control and aiding with histology pre-processing and registration steps (Fig. 1g). Our scanner software can be downloaded at (https://github.com/grinberglab/high-res-3D-tau).
Creation of imaging datasets for IHCNet training and validation
We generated 1024x1024 pixels (1.25 x 1.25 mm) patches from randomly selected gray matter locations throughout the full-resolution brain image datasets to create training, testing, and validation datasets. Before patch extraction, we applied manually segmented white matter masks to the full-resolution brain image datasets. This step aimed to generate training and validation datasets enriched for gray matter patches (because tau pathology is predominantly located in gray matter in AD). We created 100 patches per antibody (AT100, AT8, and MC1) per case, totaling six datasets and 600 patches. The patch extraction routine was written in Python and completely automated, running on UCSFs' Wynton cluster (https://wynton.ucsf.edu), exploring computational parallelism while extracting patches, i.e., the different images are split into patches simultaneously (Fig. 1f). Each patch was manually masked for background and tau inclusion with the help of Fiji’s Trainable Weka Segmentation plugin (www.cs.waikato.ac.nz/ml/weka) (Fig. 3d-e) (Schindelin et al., 2012). Here, the user manually selected sample pixels belonging to tau and background classes. These pixels were used to compute Gaussian filters, Hessian, membrane projections, mean, maximum, anisotropic diffusion, Lipschitz, Gabor, Laplacian, entropy, Sobel, a difference of Gaussians, variance, minimum, median, bilateral filter, Kuwahara, derivatives, structure, and neighbor values. A linear SVM classifier (LibLinear) was then used to generate an initial tau segmentation (Fan et al., 2008). First, a user retrained and refined the initial segmentation using an image editor (Gimp) (https://www.gimp.org). Next, all final masks went through quality control by a highly experienced pathologist (LTG). Labeling took approximately 2 hours per patch (total 1200 h or 150 days of specialized work).
Finally, patches for each antibody (AT8, AT100, and MC1) were combined on bigger datasets of 100 or 200 patches each and were randomly split into 80% of patches for training, 10% for testing, and the remaining for validation.
Histology Images Pipeline
Pre-processing blockface images
Blockface images had their background segmented using a semi-automated graph-based algorithm (Fig. 1c). Briefly, the user selects brain and background sample pixels using a graphical user interface (GUI). Images are then converted to LAB color space (L* for perceptual lightness, and a* and b* for the four unique colors of human vision: red, green, blue, and yellow), and the algorithm computes mean color difference maps (ΔE). ΔE is defined as the distance in LAB space: ΔE=Ldiff2+Adiff2+Bdiff2
With Ldiff, Adiff, Bdiff being the difference values computed as: Ldiff=Li−μ(Lr)Adiff=Ai−μ(Ar)Bdiff=Bi−μ(Br)
Where Li, Ai and Bi are image LAB channels, Lr, Ar, Br are LAB channels of reference pixels selected by the user, and μ(.) is the mean. Pixels in ΔE, whose color is similar to the reference values, appear dark (smaller distance) while cell pixels are brighter (larger distance). We computed brain and background ΔE maps using the manually selected pixels as reference values and performed a global histogram threshold using the Otsu's method to obtain binary masks. We, in turn, combine both masks to obtain brain segmentation. It is expected that several undesired objects linger after the initial segmentation. The segmentation is further refined using a graph-based method to remove the undesired objects. In this method, image objects and their relationship are modeled as a weighted graph, where connected structures are considered the vertices. Edge weights are computed using a similarity function computed from color and distance values. The graph is partitioned using NCuts (Jianbo and Malik, 2000), leaving just the brain area. Commonly, the camera or brain must be repositioned several times during sectioning to adjust for changes in block size, causing the blockface images to be misaligned in relation to each other. We used Matlab's registration GUI to select landmarks for computing affine registrations (2018). Finally, the aligned blockface images were stacked together to form the blockface 3D volume using Insight Segmentation and Registration Toolkit (ITK) (Avants et al., 2014). This 3D volume was used as an intermediate space for creating the 3D tau-protein maps (Fig. 1h).
Low-resolution histology background segmentation
The 10% resolution histological image datasets were converted to LAB color space (Fig. 1g). In that space, background pixels are consistently darker than brain pixels, and segmentation is performed through histogram thresholding using the triangle algorithm (Zack et al., 1977). The resulting binary masks are used to erase all background pixels. Moreover, brain masks are combined with white matter masks to create gray matter masks that guided the entire segmentation process.
2D registration to blockface image
After background segmentation, the 10% resolution histological images are aligned to their respective blockface images using a combination of manual and automatic registration (Fig. 1g). Due to an excessive number of artifacts caused by histological processing, we initialized the registration manually using MIPAV spline-based registration (McAuliffe et al., 2001). Here, the user manually selects landmarks on both the histology and blockface images. MIPAV then generates a warped image and a registration warp file. After the initial registration, the image went through a diffeomorphic registration using the 2D SyN algorithm based on the large diffeomorphic deformation model metric mapping (LDDMM) method (Beg et al., 2005).
Image and Mask Tiling
Each full-resolution histological slide image was first tiled to reduce memory footprint during image segmentation, with tile size corresponding to approximately 5mm2 of tissue (Fig. 1f). Tile coordinates and dimensions are saved as XML metadata files. The histological images’ respective 10% resolution gray matter masks (whole image minus white matter masks) are rescaled to match their full-resolution dimension and tiles. Finally, histological tiles were masked using their respective gray matter tiles, leaving only the ROIs for image segmentation. Tiles with less than 5% of tissue pixels were ignored during segmentation to reduce the overall computational time. Image and mask tiling routines were developed in Python (Van Rossum and Drake Jr, 1995) and ran on UCSFs' Wynton cluster exploring computational parallelism, having one pipeline for each histological image.
Deep learning-based segmentation of particles of interest (Fig. 1f, Fig. 4)
We designed IHCNet, a UNet (Ronneberger et al., 2015) based neural network capable of working with color information and outputting tau presence confidence maps that were later thresholded to create binary maps.
Model development and training
Our model has an 204x204x3 pixels input layer to accommodate RGB images – we chose this size to match 1mm2 of tissue in our whole slide imaging scanner resolution. Figure 4 shows the IHCNet architecture together with each layer tensor size. The image is pushed through three contractions blocks, the bottleneck, and upsampled by three expansion blocks in our model.
Each contraction block comprises two convolution layers that use 3x3 kernels, stride of 1 and ReLu activation, followed by a 2x2 max pooling layer and a 0.1 rate dropout. The bottleneck comprises two convolutional layers that use 1x1 kernels, stride of 1, and ReLu activation. Expansion blocks are composed of a 2x2 upsampling layer followed by a 0.1 rate dropout and two convolutional layers that use 3x3 kernels, stride of 1 and ReLu, except for the last expansion block that uses 3x3 upsampling. The last layer reshapes the data to a 20000x2 tensor and softmax activation.
Training is performed using standard backpropagation, with binary cross-entropy as the loss function and Adam algorithm for optimization. The learning rate is estimated using the cyclical learning rate method (Smith, 2015). The network was developed in Keras (https://keras.io) on top of TensorFlow (https://www.tensorflow.org). We trained the network using mini-batches of 32 images during 100 epochs or until the loss curve plateaued. We also used massive data augmentation, performing real-time random rotations, shear, horizontal and vertical flips. All training and inference were performed on an NVIDIA Titan V GPU with 12 GB of RAM. IHCNet outputs confidence maps of the existence of tau inclusions, which are thresholded to generate binary masks. The threshold value was set heuristically to 0.5 for AT100, AT8, and MC1 datasets, and pixel with confidence higher than this value was considered tau (Fig. 3d-e). IHCNet training took up to 2 days per antibody using a workstation with 2 GPUs. The networks were trained until we observed a plateau in the accuracy and loss function graphs.
Heatmap computation
The binary tiles were transferred back to the cluster for computing heatmaps (Fig. 1f). We computed the mean amount of tau on each tile, indicated by the mean number of pixels belonging to foreground, inside a 1 μm2 of tissue, an 8x8 pixels block. The heatmaps were generated as tiles having the same dimension of binary tiles where each 1 μm2 block is filed out with the mean number of pixels belonging to tau. Tiles are stitched together to create a heatmap having the same resolution as the original histological image and are later resized a 10% resolution.
Heatmap normalization
Each of the six heatmap datasets (AT100, AT8, and MC1 for each brain) was normalized to mitigate batch staining inhomogeneity (Fig. 1f). A trained neuropathologist (LTG) selected strong signal ROIs from slices whose staining is deemed optimal and weak signal ROIs from slices whose staining is suboptimal for each dataset. Outliers were removed from each ROI by thresholding the values between the first and third quartiles, and mean strong and weak signals were calculated. We then computed an increase factor as the absolute difference value between both means. All heatmaps from slices deemed suboptimal during quality control were adjusted by summing their non-zero values with this increase factor.
Heatmap to blockface alignment and stacking
The 2D registration maps computed during the 2D registration to blockface image step were applied to the 10% resolution normalized heatmaps, yielding heatmap registered to their respective blockface images. These images are then stacked together for a 3D volume (Fig. 1f).
3D Histological Reconstruction and Visualization and MRI-Histology registration
The 3D registration maps generated during the Histology to MRI 3D registration were inverted and used to warp the MRI to blockface space, allowing direct comparison between MRI signal and tau density. Freeview (Fischl, 2012) was used for visualizing 2D slices. Amira (ThermoFisher Scientific) was used for 3D visualization.
The accuracy of registration approaches for medical images relies on the presence and detection of homologous features in both target image (i.e., 3D blockface reconstruction) and the spatially warped image (i.e., structural T1-weighted MRI). To satisfy this prerequisite, we used the Advanced Normalization Tools (ANTs) (Avants et al., 2009), the N4 non-parametric non-uniform intensity normalization bias correction function (Tustison and Avants, 2013; Tustison et al., 2010) followed by skull-stripping (Fig. 1h). Briefly, structural T1-weighted MRIs were warped into the 3D image space of the corresponding blockface reconstructions using ANTs (Avants et al., 2008)). First, a linear rigid transformation was performed. Then a diffeomorphic transformation using the Symmetric Normalization (SyN) transformation model was performed. SyN uses a gradient-based iterative convergence using diffeomorphisms to converge on an optimal solution based on a similarity metric (e.g., cross-correlation) (Klein et al., 2009). To validate the registration, we calculated the Dice coefficient of ventricles (Dice, 1945).
RESULTS
Pipeline for creating High-Resolution 3D mapping of immunohistochemical findings in whole human brains
We engineered a comprehensive pipeline suitable for rendering quantitative 3D mapping of particles of interest (lesions, inclusions) detected by immunohistochemistry in whole postmortem human brains that can be co-registered with other histology- and neuroimaging-based 3D maps of the same brain. We broke down our pipeline into a series of histological and computation modules (Fig.1) that can be executed independently for scalability. The complete pipeline relies on three pillars: whole-brain histological processing and large slide immunostaining (Alegro et al., 2016; Alho et al., 2017), 3D and 2D registration algorithms (Alegro et al., 2016), and novel DL algorithms to segment and quantify particles of interest in terabytes-large imaging datasets.
Preliminary protocols for histological processing and registration have been published previously (Alegro et al., 2016; Alho et al., 2017; Theofilas et al., 2014). Here, we focused on pipeline development and quality control steps to enable scaling-up whole-brain slice immunostaining (including the hardware we assembled to accommodate oversized biological specimens), developing and validating DL algorithms for segmenting particles of interest, and integration of all steps. We processed two whole human brains in 1608 coronal, whole-brain sections, of which 524 underwent immunohistochemistry. Then, we used our computational pipeline to scan the histological sections at microscopic resolution, segment tau inclusions in each voxel, create 3D inclusion maps and register those maps to the corresponding MRI volume.
Briefly, the first module comprises acquiring structural MRI to create an MRI dataset (Neuroimaging Acquisition Module- Fig. 1a). We obtained postmortem MRI sequences in-cranio right before procurement (Alegro et al., 2016). Upon procurement, the brains were processed, embedded as a whole, and coronally cut (Alegro et al., 2016; Heinsen et al., 2000) (Histological Processing Module- Fig. 1b). Brain slicing generates one physical and one image datasets: (a) serial histological sections (about 800 per case, coronal axis), which are the input to the Immunostaining Module (Fig. 1d), and (b) blockface image dataset comprising digital images of each histological section generated by photographing the celloidin block before each microtome stroke (Theofilas et al., 2014). The blockface image dataset proceeds to the Blockface Imaging Processing Module (Fig. 1c), which delivers a digital blockface volume, an intermediate space for registering 2D histological maps, and the MRI volume to enable neuroimaging and histology voxel-to-voxel comparisons (Alegro et al., 2016). Further details about the Neuroimaging Acquisition, Histological Processing, and Blockface Imaging Processing modules are found in Methods and our previous publications (Alegro et al., 2016; Alho et al., 2017; Grinberg et al., 2009; Theofilas et al., 2014). Hereon, we will focus on newly developed Modules (blue/gray shaded in Fig. 1) and their integration into the pipeline.
Different types of abnormal tau protein accumulate along AD progression. Because autoradiography studies are only partially translatable to an in vivo condition (Okamura et al., 2018), it is often unclear which abnormal tau species correlates better to a tau PET signal. Therefore, a robust validation pipeline should enable probing different histological markers within the same PET voxel. In the Immunostaining Module (Fig. 1d), we used free-floating immunohistochemistry to independently detect three types of abnormal tau (detected by AT8, AT100, and MC1 antibodies) per case (6 whole-brain maps in total).
Immunostained slides proceed to the Scanning and Stitching module (Fig. 1e). We designed a custom-built whole-slide scanner (Fig. 2) to digitalize large and thick tissue sections at the microscopic resolution. Each histological section generated about 1500 image tiles (20GBs of data) at a resolution of 1.22μm/pixel (Fig. 3). We scanned a total of 524 immunostained sections at one plane of focus. It took approximately 90 min to scan and two days to stitch each section. Thus we used an HPC system to perform stitching, exploring computational parallelism for each section since the tiles of individual sections can be stitched concurrently and independently from other sections. The stitching process generated two new imaging datasets: a high-resolution digital section dataset (Fig. 1f) with images of approximately 80000x50000 pixels per tissue section (another 20GBs of data per slide) and a low-resolution digital section dataset (Fig. 1g) in which the original digital sections were downsampled to a 10% resolution. Low-resolution images facilitate all computing steps that did not require microscopic resolution, including visual inspection of the final stitched images, masking background, or any brain area of interest. Low-resolution images also facilitated registering images of immunostained slides to the original blockface coordinates. We developed the 2D Histology Registration Module (Fig. 1g) to register each low-resolution histological digital image to the corresponding blockface image. First, we applied background masks to the low-resolution digital section dataset. Then, we used a combination of open-source registration algorithms to align these masked images to their corresponding pre-processed blockface images and create warping maps for each histological section. The warping results were visually inspected, and imperfections were corrected manually—most of the imperfections were located around the limbic structures because the region has delicate sulci and crests, making it more prone to tears and folding during processing.
Segmenting and quantifying of tau inclusions were performed in the high-resolution digital section dataset using the Inclusion Segmentation and Mapping Module (Fig. 1f). This module contains IHCNet for segmenting particles of interest in a setting of low signal-to-noise ratio. IHCNet is the core innovation of our pipeline, with higher segmentation accuracy and robustness to histological artifacts than thresholding algorithms. The architecture, rationale, and validation steps for IHCNet are described in Methods. We used the warping maps generated in the 2D Histology Registration Module to register the high-resolution heatmap of each histological section, created via IHCNet, to their corresponding blockface image. The 2D registered heatmaps were stalked and normalized to create 3D heatmaps of each tau type. The final 3D tau heatmaps were stored in standard nifti medical imaging file format
Finally, we used the 3D Histology Reconstruction Module (Fig. 1h) to register the histology to MRI 3D maps. Briefly, structural T1-weighted MRIs were warped into the 3D image space of the corresponding blockface reconstructions (Avants et al., 2008; Avants et al., 2014). The diffeomorphic transform matrices were then applied to all the tau inclusion heatmaps for mapping into the structural MR imaging space and resolution. The final product of our test was six 3D tau heatmaps registered to the corresponding MRI showing the distribution of tau detected by AT8, AT100, and MCI antibodies, from two different whole human brain samples (Fig. 5, Supplementary Videos 1-6).
IHCNet, a convolutional neural network (CNN) for semantic segmentation
Architecture and rationale
We assumed it would be possible to apply thresholding algorithms in our first attempts to segment tau deposits in histological images, assuming that the immunostaining label is brown (DAB −3,3′-Diaminobenzidine staining), and the background is transparent. However, the thresholding strategy failed despite several optimizing steps to ensure a high signal-to-noise ratio on immunohistochemistry. Conventional 5-um thick histological sections have a transparent background. However, we had to cut our brain blocks at a 160 μm minimum thickness to minimize tissue tearing. The background of those 32x thicker-than-conventional sections is opaque, especially in the white matter area because of the myelin. Second, hard-to-control factors such as submicrometric thickness variations and regional differences in hydration level influence the staining hue in large sections, making thresholding algorithms useless for the detection task. To address these challenges, we designed IHCNet, a CNN. IHCNet uses a UNet based architecture (Fig. 4), trained to compute a confidence map of the 2D distribution of particles of interest (i.e., tau inclusion here) instead of a single class confidence value. Furthermore, we extended IHCNet architecture to leverage color information and optimally work with microscopic imaging resolution (1.22 μm/pixel).
IHCNet has an input layer measuring 204×204×3 pixels corresponding to 1mm2 of tissue to accommodate RGB images. IHCNet features a validation hold-out scheme, where training and testing sets were used for model training, and a validation set was used for independent network performance statistics.
Validation of the network training
To validate the IHCNet model, we computed the ROC and precision-recall curves over the testing and validation sets for each antibody. For ROC, the AT8 model achieved an area under the curve (AUC) of 0.85 for both testing and validation datasets, while the AT100 model achieved an AUC of 0.87 on testing and 0.89 on the validation dataset. MC1 model achieved the best ROC results with an AUC of 0.91 on the testing and 0.88 on the validation dataset (Fig. 6). From thresholded masks, we calculated precision and recall values. We obtained precision values of 0.87, 0.87, 0.90 respectively for AT8, AT100, and MC1 in the testing datasets and 0.92, 0.92, 0.91 for AT8, AT100, and MC1, respectively, in the validation datasets. As for the recall, the thresholded masks obtained 0.16, 0.19, 0.12 respectively for AT8, AT100, and MC1 in the testing datasets and 0.18, 0.13, and 0.08 for AT8, AT100, and MC1, respectively, in the validation datasets.
In terms of precision-recall, the AT8 model achieved an AUC of 0.68 for the testing and 0.72 for validation; AT100 had a 0.66 for testing and 0.68 for validation, and MC1 had a 0.63 for testing and 0.63 for validation.
Neuronal network tau segmentation of the full resolution dataset
CNN-based segmentation is computationally intense. Thus, we first masked regions of interest to decrease computing time. Specifically, we masked out background and white matter tissue pixels to focus on the regions with tau deposition. Masks were created using the low-resolution digital section dataset and subsequently upsampled and applied to the high-resolution histological images. Moreover, to facilitate computation, we divided the high-resolution images into tiles. It took about four days for IHCNet to segment the largest dataset (AT8, case 2,160 tiles). The network output probability maps were thresholded to create binary masks. The exact cut-off values for thresholding were applied to all datasets. Figure 6 shows an example of the histology tiles, the probability maps created by the CNN, and the thresholded binary masks (top rows).
Transfer learning and additional validation experiments
Considering it is unfeasible to generate training data for each new antibody and case, we tested how our CNN behaves with different training datasets size and composition. We used the following training datasets for these experiments: #1: all labeled patches from cases 1 and 2, divided per antibody (AT8, AT100, and MC1), totalizing three databases with 200 patches each. #2 all labeled patches divided per antibody and case, totalizing six databases of 100 patches each. In summary, we created nine training models. The rationale for cross-testing is that morphology of tau inclusions detected by different tau antibodies is similar. Thus, in theory, networks trained for one tau antibody could segment histological images from a different tau antibody.
First, using these nine training modules, we tested how a CNN trained with labeled patches from a given tau antibody from a specific case(s) performs in segmenting images labeled with the same tau antibody in another case. For instance, the CNN trained using AT8 patches from case 1 was used to segment AT8 signal on case 2 and rendered an AUC of 0.7855 in the testing set and 0.7847 in the validation set. Inversely, when the network trained on AT8 patches from case 2 was applied to segment AT8 signal on case 1, we obtained an AUC of 0.8642 in the testing set and 0.7544 in the validation set (Fig. 7).
Second, we used labeled patches of a given tau antibody to train a CNN and then applied the CNN to another tau antibody validation dataset (cross-testing). When a network trained on AT100 patches was applied to segment AT8 datasets, we achieved an AUC of 0.8459 in the AT8 testing dataset segmentation and 0.8219 in the AT8 validation dataset. Inversely, a network trained on AT8 patches when applied to segment AT100 datasets resulted in an AUC of 0.8677 in the AT100 testing dataset and 0.8501 in the AT100 validation dataset (Fig. 8).
The results show that the difference between models is negligible. The curves for the models trained with case 1 AT8-only and AT100-only images are slightly inferior to the dataset.
Network introspection experiments show the neural network correctly learns tau inclusions
Standard precision-recall and ROC curve evaluation cannot show whether the network effectively learns image features of particles of interest (tau inclusions) or is simply overfitting patterns. Thus, we used neural network inspection techniques to evaluate the network ability to learn. We first used gradient-guided class activation maps (Grad-CAM) (Selvaraju et al., 2017) on randomly selected images to evaluate the most critical features to drive the network to a particular decision. Then, we added perturbations for checking whether the network was segmenting tau inclusions solely based on spatial localization or using relevant features. Standard Grad-CAM works by computing the gradient between a user-defined class (artificial) neuron in the last network layer and an intermediary target layer and then multiplying its mean value with the target class activations. This technique was designed to work with networks with a fully connected or a global pooling layer (Cireşan et al., 2011), responsible for mapping the information spread across the convolutional layers to a single neuron at the last layer. As IHCNet outputs 2D probability maps of inclusions and lacks the property of fragmented mapping information to a single neuron, we adapted Grad-CAM by selecting multiple output neurons inside a tau inclusion, on randomly selected test images, i.e., neurons that should be activated for the class “tau,” with the help of a binary mask, and computing the mean of the Grad-CAM maps generated for each selected neuron. As seen in Figure 9, a visual inspection of the output CAM maps shows a good agreement between the features that drove network decision and tau signal. The network correctly responded to perturbations. We repeated the same experiment to interrogate if the network could correctly discriminate the background. We used a mask to select background pixels. The resulting Grad-CAM shows good localization of the background. Finally, we interrogated whether the network was learning to detect tau based on pixel spatial localization or using other relevant information. We created a perturbed image by partially covering a tangle with a background patch and repeated the background Grad-CAM experiment. The network correctly recognized the patch as background, suggesting that other relevant features inform the network.
Registration validation
We quantitatively evaluated registration quality by computing the Dice coefficient (Dice, 1945) between stacks of masks manually labeled on MRI and histological slices. The Dice coefficient measures how well two sets of images co-localize.
Histology to blockface 2D registration evaluation:
we manually labeled the lateral ventricles on histological images containing ventricles. All histology ventricle labels were registered to the blockface using the pre-computed registration maps. We computed the 2D Dice coefficient for all registered histology/block label pairs. The same procedure was performed for our two cases. Mean Dice coefficient values for case 1 were 0.9082±0.053 (AT8), 0.9112±0.0405 (AT100) and 0.88±0.0531 (MC1). Mean Dice coefficient values for case 2 were 0.8185±0.1070 (AT8), 0.8385±0.1444 (AT100) and 0.8605±0.1823 (MC1).
MRI to blockface 3D registration evaluation:
here, we used the lateral ventricle masks. The Dice coefficient values were then computed between each blockface ventricle label created to evaluate the 2D registration and its respective MRI ventricle label (after registration). The same procedure was performed for the two cases. We obtained a mean Dice coefficient of 0.9047 (±0.032) and 0.8570 (±0.095) for case 1 and case 2, respectively.
Discussion
We developed a scalable pipeline for creating MR-registered 3D maps of particles-of-interest detected with immunohistochemistry in oversized (whole human brain) histological slides as a means to enable histology to neuroimaging voxel-to-voxel comparisons. This pipeline comprises histopathological and computational modules and incorporates DL algorithms and HPC capabilities into a previously developed framework to facilitate high precision 3D histology to neuroimaging registration. The pipeline is generalizable to a broad array of studies focusing on validating any neuroimaging modality that can be registered to MR space, such as PET, against stained-based histological data.
Molecular imaging is the most promising technique for quantifying proteins spatially-resolved in the living brain. Because neurodegenerative diseases spread stereotypically (Seeley, 2017), molecular imaging promises to render reliable preclinical diagnosis, monitor disease progression, and measure the efficacy of experimental treatments. Attempts to validate Aβ and tau tracers by comparing the histological properties of postmortem brains to the PET signal obtained closer to death proved valuable to facilitate clinical approval (Curtis et al., 2015; Leuzy et al., 2019; Pontecorvo et al., 2020; Sabri et al., 2015; Salloway et al., 2017; Soleimani-Meigooni et al., 2020). Still, methodological issues limiting large-scale voxel-to-voxel comparisons precluded these studies from answering several critical questions about the neurobiological basis of the observed PET signal, including the nature of off-target signal or each kind of posttranslational modification of the target protein better correlate with PET signal intensity. Thus, the diagnostic scope of clinically approved tracers remains limited, and the time between tracer development and clinical approval is still very long.
Immunohistochemistry and autoradiography are methods of choice to detect the anatomical distribution of a protein of interest in situ. The former is more broadly used because it takes less time, requires less equipment, avoids radioactivity, and is less prone to artifacts. Voxel-to-voxel comparison of PET signal to immunohistochemistry results can inform the neurobiological basis of PET signal. However, previous studies were limited by difficulties in locating precisely in PET imaging the areas probed with histology or performing histology at a large scale.
The resolution of PET scans is low. Anatomical parcellation, which is particularly relevant In neurodegenerative diseases because distribution and the regional load of abnormal protein deposition have implications for differential diagnosis and staging disease severity, is better achieved by co-registering PET to MRI space. Thus, co-registering histological data, the gold standard for detecting abnormal protein deposition, to MR space is crucial to facilitate voxel-to-voxel validation of PET signal. We and others have recently offered solutions to enable structural histology to MR co-registration of human whole-brain volumes (Alegro et al., 2016; Alkemade et al., 2020). Such solutions use modified brain tissue processing and computational algorithms to minimize and correct the significant non-linear tissue deformation inherited from tissue processing. Still, these solutions only use tissue stained with simple dyes, such as Nissl. Immunohistochemistry adds several processing steps, resulting in additional tissue deformation. Also, immunohistochemistry is optimized for thin tissues (3 to 10 um) to ensure good antibody penetration and signal-to-noise imaging results. Optimal histology to neuroimaging registration requires whole-brain histological 3D reconstructions. However, because brain tissue is only semi-rigid, cutting larger histological blocks thin results in tears and folds. Thus, the larger is the tissue area, the thicker it needs to be cut. Free-floating techniques facilitate antibody penetration on thicker histological slides, but the signal-to-noise ratio is always poor since thicker slides have many cell layers overlapped.
For incorporating immunohistochemistry 3D maps into our original pipeline (Alegro et al., 2016), we had to overcome significant challenges, including (1) how to digitalize images of hundreds of supersized and thick histological slides (that do not fit regular histological scanners, require stronger light sources, are prone to imaging inhomogeneity, and generate gigabytes of data each) and (2) how to perform large-scale image segmentation on several hundreds of terabytes of inhomogeneous images with accuracy and precision.
We addressed the first challenge by engineering a cost-effective whole slide imaging system with an extensive travel range capable of imaging the entire area of our slides. Options of open-hardware microscopy equipment expanded recently (Campbell et al., 2014; Nunez et al., 2017; Phillips et al., 2020; Rosenegger et al., 2014) with projects relying on a wide variety of materials, from inexpensive 3D printed parts to high-end optical kits. However, most designs accommodate small samples only. Hardware robustness and availability of customizable open-source software were imperative for our needs. Our configuration prioritizes cost-benefit to minimize production costs. Machine vision objectives offer a reasonable field of view while incorporating all the optical elements necessary for microscopic imaging in one set of lenses attached to the camera, thus improving scanning time. Our scanner took 50 to 90 minutes to image one plan of focus for each large section, plus loading and focusing time. The scanner never failed during the over 1000h of scanning time.
The second challenge was the most complex to solve. Microscopy lenses have a limited field of view. Thus large surfaces need to be imaged in several parts image inhomogeneity. Also, in histological thicker sections, several layers of cells overlap, resulting in an opaque background. These factors contributed to the failure of thresholding, or machine-learning-based segmentation algorithms developed to thin histological slides (Koga et al., 2021; Signaevsky et al., 2019) to segment tau inclusions in our preparations. Therefore, we took the challenge to create a CNN to segment tau signal with precision. We elected a DL algorithm for its ability to automatically discover unknown patterns that best characterize a raw data set by capturing semantic meaning (LeCun et al., 2015). DL models data in a bottom-up approach, using small low-level features such as edges, first in lower network layers, and increasing abstraction and complexity in the following layers. This approach makes DL more robust to identify immunostaining and tissue artifacts, signal inhomogeneity, and false positives.
IHCNet enabled segmenting morphologically diverse tau inclusions from blurred images. Moreover, its computational modules facilitate processing the large volume of data simultaneously by harnessing the power of HPC to run the different parts of the pipeline, using a paradigm known as embarrassingly parallel workload, where hundreds of copies of the same pipeline are run at the same time.
To test the robustness of IHCNet, we conducted thorough validation experiments. On ROC AUC, our classification results (mean 0.88 on the testing set and 0.87 on the validation set) were just below the reported values for cell culture (0.95) (Al-Kofahi et al., 2018). Noteworthy, cell culture images have a clearer background, fewer artifacts, and much smaller surfaces than our histological images. When compared to digital pathology methods for astrocyte detection (Suleymanova et al., 2018), which is a more challenging segmentation problem since digital pathology images usually have a cluttered background, our pipeline also yielded mean precision (0.88 for testing and 0.92 for validation) above the reported value of 0.86, although much worse recall performance (0.16 for testing, 0.13 validation) when compared to their reported value of 0.78. The same pattern of similar precision but worse recall rates were seen when comparing our results to other studies focusing on segmenting or classifying tau lesions (Koga et al., 2021; Signaevsky et al., 2019). This worse recall performance was expected because we dealt with the blurriness and hue variations inherited from thick and large histological slides.
In contrast, other studies using DL algorithms to segment tau from histological sections ~30x thinner and with higher spatial resolution (1.22 μm/pixel here vs. 0.5 μm/pixel), much smaller sampling areas (standard 2×1” vs. 4×6”). Our pipeline can accommodate increasing scanning spatial resolution (by replacing the lens) and improving image blurriness by compacting images acquired at different depths of focus to improve recall values. The choice of compromising spatial resolution and image sharpness vs. increasing computational capability needs in an already computing-demanding pipeline depends on the aim of each project. The other pipelines focused on classifying tau lesions to facilitate neuropathological staging and diagnosis, thus requiring better spatial resolution. The cost-benefit of increasing scanning resolution is limited for our goals because, at the current level of PET resolution, each PET voxel represents a summary signal from thousands of histology voxels, and histology-based maps need to be downsampled for MRI/PET co-registration. Although it is beyond the scope of this manuscript, assessing the distribution of signals from all these histology voxels relative to a single PET voxel could provide further insights into PET imaging mechanisms. Therefore, we opted to employ a tissue processing pipeline optimized for minimizing tissue deformation and enable staining of larger samples. A more recent investigation focused on segmenting tau at 50 μm thick immunostained slides from medial temporal lobes blocks that underwent MR ex-cranio (Yushkevich et al., 2021). This study aligned with ours and used similar elements of previous versions of our histology-to-imaging registration pipeline 20, reported AUC values similar to ours. However, precision and recall values are missing, precluding further comparisons. Here, we focused on 3D histological maps for PET tracer validation.
The capacity of transfer learning is another IHCNet strength as it allows us to scale up our pipeline. CNN training is time and resource-consuming because it requires a significant amount of labeled data for performing effectively. Cross-training results indicate that mixing images from different cases – as in the full model training where images from the two cases were combined, not only does not interfere with training, but it is beneficial. However, while the ROC AUC values do not show a considerable difference between models, the precision-recall curves show decreased model accuracy in some cases and should be computed before using transfer learning. Also, it may be necessary to add extra labels for applying IHCNet to other tauopathies beyond AD.
Although we and others have previously developed and successfully tested algorithms for registering images of whole brain-histological sections dyed with Nissl staining on blockface images (Alegro et al., 2016; Alkemade et al., 2020), co-registering the digital images from immunostained slices to the blockface images represented an additional challenge in our pipeline. Immunohistochemistry processing requires harsher chemicals than Nissl staining, causing more tissue deformation and tears. These artifacts were prominent on the severely affected case 2, in which the brain tissue was extremely atrophic and friable. We had to add a manual affine registration step during the 2D histology to blockface alignment. Although this solution is robust to resolve histology artifacts, it impairs the scalability of our pipeline.
In conclusion, we present an unprecedented solution to improve PET tracer validation against histopathological changes measured on postmortem tissue. A crucial component of this pipeline is IHCNet, a CNN capable of processing large image datasets to locate and segment particles of interest. IHCNet has robust transfer learning capability. Our pipeline innovates for its ability to process massive datasets and perform well in blurred images with broad hue variations and obtain excellent co-registration results. Our 3D mapping at microscopic resolution coupled with our previously developed 3D registration algorithms for combining histological and imaging volumes can potentially open avenues for thorough and systematic validation of new neuroimaging tracers and expediting their approval for clinical use.
Supplementary Material
Svideo1
Svideo 2
Svideo 3
Svideo 4
Svideo6
Svideo 5
ACKNOWLEDGEMENTS and FUNDING
We thank the brain donors for contributing tissue to this study. This work was supported by AVID Radiopharmaceuticals, NIH R01AG056573, R01AG056573, U54 NS100717 and BrightFocus Foundation. MA and DU were partially funded by the Berkeley Institute of Data Science and the Office of Science of the U.S. Department of Energy under Contract No. DE-AC02-05CH11231.
Abbreviations
AD Alzheimer's disease
AUC area under the curve
CNN convolutional neural network
DL deep learning
Grad-CAM gradient-guided class activation maps
HPC high-performance computing
MRI magnetic resonance imaging
PET positron emission tomography
ROC receiver operating characteristic
Figure 1: Step-by-step pipeline for high-resolution 3D mapping of immunohistochemical findings in whole human brains. Blue shading represents unpublished steps, including IHCNET (darker blue shading).
Fig. 2- Scheme of the Slide Scanner set-up built for digitalizing and stitching 6 x 4” histological slides at 1.22μm/pixel resolution
Figure 3: a) example of a stitched image (Case 1- AT8); b) zoom in of the area delimitated by the black square in 3a (scalebar is 10 μm) to illustrate tau inclusions our system is capable of detecting (green arrow: neurofibrillary tangle, red arrow: neuropil threads, and blue arrow: neuritic component of a plaque; c) example of training tile initially segmented using Weka (scalebar is 100 μm); d) result fo the tile segmentation after manual quality control (red: neurofibrillary tangle, green: threads and plaques, purple: background)
Figure 4: IHCNet architecture. The arrows indicate skip connections that are followed by cropped when the right-hand side layer is smaller than the left side layer
Figure 5: AT8, AT100 and MC1 quantitative heatmaps registered to blockface (1st column), MRI (2nd column) of Cases 1 and 2. The 3rd column depict the respective 3D renderings of each tau map. Heatmaps were normalized to the same scale to facilitate comparisions. As expected from neuropathological studies, AT8 detected higher amounts of tau than AT100 and MC1. Videos of the full 3D tau rendering are found in Supplementary Material.
Figure 6: patches of digitalized images of an AD case (case #1) immunostained for AT8. Upper panel (a/c/e) - patch #1 and lower panel (b/d/f) – patch #2. a/b) ROIs of AT8 antibody-stained brain tissue with high amounts (a) and low amounts (b) of tau inclusions; c/d) their respective probability maps of tau signals generated by IHCNet. Green color symbolizes tau signal. The brighter the signal, higher the tau probability is and e/f) binary segmentation after thresholding c /d at 0.5. Graphs show ROC (g) and precision-recall curves (h) for models trained with all full datasets from all antibodies. Stars on the precision-recall graphs show where the 0.5 threshold is localized for each stain.
Figure 7: Comparison of ROC (left) and precision-recall (right) curves for models trained with complete datasets (images of both cases/antibody) vs. models trained with a dataset comprised of images from one brain only/antibody. Top row: AT100; Middle row: AT8 and Botton row: MC1. The solid orange and solid purple lines show the results for the testing and validation using the full datasets. Dashed red and blue lines show testing and validation results for the model trained with images from case 2 alone and dashed green and brown lines show testing and validation results for the model trained with images from case 1 alone. Stars on the precision-recall graphs show where the 0.5 threshold is localized
Figure 8: Comparison of ROC and precision-recall curves for cross-trained models, where we performed training using images from one antibody and tested the model on a dataset of a different antibody. Top row: dashed red and blue lines show testing and validation results for a model trained using AT100 data for segmenting AT8 images. For comparison, the orange and purple lines show the test and validation results for the model trained with AT8 data. Middle row: model trained on AT8 data to segment AT100 images. Bottom row: model trained on AT100 images to segment MC1 images. Stars on the precision-recall graphs show where the 0.5 threshold is localized.
Fig 9: Network interpretability test using Grad-Cam and perturbation techniques. a. Example of original images with neuritic plaques (i.e. a1/f1 (lower magnification) and a2/f2 (higher magnification) and neurofibrillary tangles (i.e. a3/f3). Mask used for foreground pixels selection from (a), indicated by red arrow. c. Mask used to locate background pixels selection from (a), indicated by blue arrow. d. Grad-CAM result of original image using pixels from mask in (b) as references. e. Grad-CAM result of original image using pixels from mask in (c) as references. f. Perturbed image, where a background image patch (indicated by the yellow arrow) was artificially placed to partially cover the tau inclusion. g. Neural network model segmentation of the original image (area with in the green shapes). h. Neural network model segmentation of the perturbed image (area with in the green shapes). i. Grad-CAM result of perturbed image using pixels from mask in (b) as references. j. Grad-CAM result of perturbed image using pixels from mask in (c) as references. In Grad-CAM results (d, e, i, j), brighter/warmer heat-map values indicating stronger influence in the network decision.
COMPETING INTERESTS
The authors declare no competing interests.
ETHICS DECLARATIONS
This project was approved the UCSF IRB (protocol 13-12079).
REFERENCES
2018. MATLAB The MathWorks Inc, Natick, Massachusetts.
2021. 2021 Alzheimer's disease facts and figures. Alzheimers Dement 17 , 327–406.33756057
Al-Kofahi Y , Zaltsman A , Graves R , Marshall W , Rusu M , 2018. A deep learning-based algorithm for 2-D cell segmentation in microscopy images. BMC Bioinformatics 19 , 365.30285608
Alegro M , Júnior EA , Loring B , Heinsen H , Alho EJL , Zöllei L , Ushizima D , Grinberg LT , 2016. Multimodal Whole Brain Registration: MRI and High Resolution Histology. 2016 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 634–642.
Alegro M , Theofilas P , Nguy A , Castruita PA , Seeley W , Heinsen H , Ushizima DM , Grinberg LT , 2017. Automating cell detection and classification in human brain fluorescent microscopy images using dictionary learning and sparse coding. Journal of Neuroscience Methods 282 , 20–33.28267565
Alho A , Hamani C , Alho EJL , da Silva RE , Santos GAB , Neves RC , Carreira LL , Araujo CMM , Magalhaes G , Coelho DB , Alegro MC , Martin MGM , Grinberg LT , Pasqualucci CA , Heinsen H , Fonoff ET , Amaro E Jr. , 2017. Magnetic resonance diffusion tensor imaging for the pedunculopontine nucleus: proof of concept and histological correlation. Brain Struct Funct 222 , 2547–2558.28283747
Alkemade A , Pine K , Kirilina E , Keuken MC , Mulder MJ , Balesar R , Groot JM , Bleys R , Trampel R , Weiskopf N , Herrler A , Moller HE , Bazin PL , Forstmann BU , 2020. 7 Tesla MRI Followed by Histological 3D Reconstructions in Whole-Brain Specimens. Front Neuroanat 14 , 536838.33117133
Avants BB , Epstein CL , Grossman M , Gee JC , 2008. Symmetric diffeomorphic image registration with cross-correlation: evaluating automated labeling of elderly and neurodegenerative brain. Medical Image Analysis 12 , 26–41.17659998
Avants BB , Tustison N , Song G , 2009. Advanced normalization tools (ANTS). Insight 2 .
Avants BB , Tustison NJ , Stauffer M , Song G , Wu B , Gee JC , 2014. The Insight ToolKit image registration framework. Front Neuroinform 8 , 44.24817849
Barrio JR , 2018. The Irony of PET Tau Probe Specificity. J Nucl Med 59 , 115–116.29123015
Beg MF , Miller MI , Trouvé A , Younes L , 2005. Computing Large Deformation Metric Mappings via Geodesic Flows of Diffeomorphisms. International Journal of Computer Vision 61 , 139–157.
Bria A , Iannello G , 2012. TeraStitcher - a tool for fast automatic 3D-stitching of teravoxel-sized microscopy images. BMC Bioinformatics 13 , 316.23181553
Campbell RA , Eifert RW , Turner GC , 2014. OpenStage: a low-cost motorized microscope stage with sub-micron positioning accuracy. PLoS One 9 , e88977.24586468
Cireşan DC , Meier U , Masci J , Gambardella LM , Schmidhuber J , 2011. Flexible, High Performance Convolutional Neural Networks for Image Classification. Proceedings of the Twenty-Second International Joint Conference on Artificial Intelligence 2 , 1237–1242.
Curtis C , Gamez JE , Singh U , Sadowsky CH , Villena T , Sabbagh MN , Beach TG , Duara R , Fleisher AS , Frey KA , Walker Z , Hunjan A , Holmes C , Escovar YM , Vera CX , Agronin ME , Ross J , Bozoki A , Akinola M , Shi J , Vandenberghe R , Ikonomovic MD , Sherwin PF , Grachev ID , Farrar G , Smith AP , Buckley CJ , McLain R , Salloway S , 2015. Phase 3 trial of flutemetamol labeled with radioactive fluorine 18 imaging and neuritic plaque density. JAMA Neurol 72 , 287–294.25622185
Dani M , Brooks DJ , Edison P , 2016. Tau imaging in neurodegenerative diseases. Eur J Nucl Med Mol Imaging 43 , 1139–1150.26572762
Dice LR , 1945. Measures of the Amount of Ecologic Association Between Species. Ecology 26 , 297–302.
Fan RE , Chang KW , Hsieh CJ , Wang XR , Lin CJ , 2008. LIBLINEAR: A Library for Large Linear Classification J. Mach. Learn. Res 9 , 1871–1874.
Fischl B , 2012. FreeSurfer. Neuroimage 62 , 774–781.22248573
Grinberg LT , Amaro Junior E , da Silva AV , da Silva RE , Sato JR , dos Santos DD , de Paula Pacheco S , de Lucena Ferretti RE , Paraizo Leite RE , Pasqualucci CA , Teipel SJ , Flatz WH , Brazilian Aging Brain Study, G., Heinsen H , 2009. Improved detection of incipient vascular changes by a biotechnological platform combining post mortem MRI in situ with neuropathology. J Neurol Sci 283 , 2–8.19289241
Heinsen H , Arzberger T , Schmitz C , 2000. Celloidin mounting (embedding without infiltration) - a new, simple and reliable method for producing serial sections of high thickness through complete human brains and its application to stereological and immunohistochemical investigations. J Chem Neuroanat 20 , 49–59.11074343
Jianbo S , Malik J , 2000. Normalized cuts and image segmentation. IEEE Transactions on Pattern Analysis and Machine Intelligence 22 , 888–905.
Kelley AS , McGarry K , Gorges R , Skinner JS , 2015. The burden of health care costs for patients with dementia in the last 5 years of life. Ann Intern Med 163 , 729–736.26502320
Klein A , Andersson J , Ardekani BA , Ashburner J , Avants B , Chiang MC , Christensen GE , Collins DL , Gee J , Hellier P , Song JH , Jenkinson M , Lepage C , Rueckert D , Thompson P , Vercauteren T , Woods RP , Mann JJ , Parsey RV , 2009. Evaluation of 14 nonlinear deformation algorithms applied to human brain MRI registration. Neuroimage 46 , 786–802.19195496
Klunk WE , 2018. Molecular imaging: What is right and what is an illusion? Alzheimers Dement (Amst) 10 , 217–220.29780866
Koga S , Ghayal NB , Dickson DW , 2021. Deep Learning-Based Image Classification in Differentiating Tufted Astrocytes, Astrocytic Plaques, and Neuritic Plaques. J Neuropathol Exp Neurol 80 , 306–312.33570124
LeCun Y , Bengio Y , Hinton G , 2015. Deep learning. Nature 521 , 436–444.26017442
Lemoine L , Leuzy A , Chiotis K , Rodriguez-Vieitez E , Nordberg A , 2018. Tau positron emission tomography imaging in tauopathies: The added hurdle of off-target binding. Alzheimers Dement (Amst) 10 , 232–236.29780868
Leuzy A , Chiotis K , Lemoine L , Gillberg PG , Almkvist O , Rodriguez-Vieitez E , Nordberg A , 2019. Tau PET imaging in neurodegenerative tauopathies-still a challenge. Mol Psychiatry 24 , 1112–1134.30635637
Marquie M , Normandin MD , Meltzer AC , Siao Tick Chong M , Andrea NV , Anton-Fernandez A , Klunk WE , Mathis CA , Ikonomovic MD , Debnath M , Bien EA , Vanderburg CR , Costantino I , Makaretz S , DeVos SL , Oakley DH , Gomperts SN , Growdon JH , Domoto-Reilly K , Lucente D , Dickerson BC , Frosch MP , Hyman BT , Johnson KA , Gomez-Isla T , 2017. Pathological correlations of [F-18]-AV-1451 imaging in non-alzheimer tauopathies. Ann Neurol 81 , 117–128.27997036
McAuliffe MJ , Lalonde FM , McGarry D , Gandler W , Csaky K , Trus BL , 2001. Medical Image Processing, Analysis and Visualization in clinical research. Proceedings 14th IEEE Symposium on Computer-Based Medical Systems. CBMS 2001 , pp. 381–386.
Montine TJ , Phelps CH , Beach TG , Bigio EH , Cairns NJ , Dickson DW , Duyckaerts C , Frosch MP , Masliah E , Mirra SS , Nelson PT , Schneider JA , Thal DR , Trojanowski JQ , Vinters HV , Hyman BT , 2012. National Institute on Aging-Alzheimer’s Association guidelines for the neuropathologic assessment of Alzheimer’s disease: a practical approach. Acta Neuropathologica 123 , 1–11.22101365
Nunez I , Matute T , Herrera R , Keymer J , Marzullo T , Rudge T , Federici F , 2017. Low cost and open source multi-fluorescence imaging system for teaching and research in biology and bioengineering. PLoS One 12 , e0187163.29140977
Okamura N , Harada R , Ishiki A , Kikuchi A , Nakamura T , Kudo Y , 2018. The development and validation of tau PET tracers: current status and future directions. Clinical and Translational Imaging 6 , 305–316.30148121
Phillips ZF , Dean S , Recht B , Waller L , 2020. High-throughput fluorescence microscopy using multi-frame motion deblurring. Biomedical Optics Express 11 , 281–300.32010517
Pontecorvo MJ , Keene CD , Beach TG , Montine TJ , Arora AK , Devous MD Sr. , Navitsky M , Kennedy I , Joshi AD , Lu M , Serrano GE , Sue LI , Intorcia AJ , Rose SE , Wilson A , Hellstern L , Coleman N , Flitter M , Aldea P , Fleisher AS , Mintun MA , Siderowf A , 2020. Comparison of regional flortaucipir PET with quantitative tau immunohistochemistry in three subjects with Alzheimer's disease pathology: a clinicopathological study. EJNMMI Res 10 , 65.32542468
Rabinovici GD , Gatsonis C , Apgar C , Chaudhary K , Gareen I , Hanna L , Hendrix J , Hillner BE , Olson C , Lesman-Segev OH , Romanoff J , Siegel BA , Whitmer RA , Carrillo MC , 2019. Association of Amyloid Positron Emission Tomography With Subsequent Change in Clinical Management Among Medicare Beneficiaries With Mild Cognitive Impairment or Dementia. JAMA 321 , 1286–1294.30938796
Ronneberger O , Fischer P , Brox T , 2015. U-Net: Convolutional Networks for Biomedical Image Segmentation. In: Navab N , Hornegger J , Wells WM , Frangi AF (Eds.), Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015. Springer International Publishing, Cham, pp. 234–241.
Rosenegger DG , Tran CH , LeDue J , Zhou N , Gordon GR , 2014. A high performance, cost-effective, open-source microscope for scanning two-photon microscopy that is modular and readily adaptable. PLoS One 9 , e110475.25333934
Sabri O , Sabbagh MN , Seibyl J , Barthel H , Akatsu H , Ouchi Y , Senda K , Murayama S , Ishii K , Takao M , Beach TG , Rowe CC , Leverenz JB , Ghetti B , Ironside JW , Catafau AM , Stephens AW , Mueller A , Koglin N , Hoffmann A , Roth K , Reininger C , Schulz-Schaeffer WJ , Florbetaben Phase 3 Study, G., 2015. Florbetaben PET imaging to detect amyloid beta plaques in Alzheimer's disease: phase 3 study. Alzheimers Dement 11 , 964–974.25824567
Salloway S , Gamez JE , Singh U , Sadowsky CH , Villena T , Sabbagh MN , Beach TG , Duara R , Fleisher AS , Frey KA , Walker Z , Hunjan A , Escovar YM , Agronin ME , Ross J , Bozoki A , Akinola M , Shi J , Vandenberghe R , Ikonomovic MD , Sherwin PF , Farrar G , Smith APL , Buckley CJ , Thal DR , Zanette M , Curtis C , 2017. Performance of [(18)F]flutemetamol amyloid imaging against the neuritic plaque component of CERAD and the current (2012) NIA-AA recommendations for the neuropathologic diagnosis of Alzheimer's disease. Alzheimers Dement (Amst) 9 , 25–34.28795133
Schindelin J , Arganda-Carreras I , Frise E , Kaynig V , Longair M , Pietzsch T , Preibisch S , Rueden C , Saalfeld S , Schmid B , Tinevez JY , White DJ , Hartenstein V , Eliceiri K , Tomancak P , Cardona A , 2012. Fiji: an open-source platform for biological-image analysis. Nat Methods 9 , 676–682.22743772
Seeley WW , 2017. Mapping Neurodegenerative Disease Onset and Progression. Cold Spring Harb Perspect Biol 9 .
Selvaraju RR , Cogswell M , Das A , Vedantam R , Parikh D , Batra D , 2017. Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Localization. 2017 IEEE International Conference on Computer Vision (ICCV), pp. 618–626.
Signaevsky M , Prastawa M , Farrell K , Tabish N , Baldwin E , Han N , Iida MA , Koll J , Bryce C , Purohit D , Haroutunian V , McKee AC , Stein TD , White CL 3rd , Walker J , Richardson TE , Hanson R , Donovan MJ , Cordon-Cardo C , Zeineh J , Fernandez G , Crary JF , 2019. Artificial intelligence in neuropathology: deep learning-based assessment of tauopathy. Lab Invest 99 , 1019–1029.30770886
Smith LN , 2015. cyclical learning Rates for Training Neural Networks. ArXiv150601186 Cs.
Smith R , Schain M , Nilsson C , Strandberg O , Olsson T , Hagerstrom D , Jogi J , Borroni E , Scholl M , Honer M , Hansson O , 2017. Increased basal ganglia binding of (18) F-AV-1451 in patients with progressive supranuclear palsy. Mov Disord 32 , 108–114.27709757
Soleimani-Meigooni DN , Iaccarino L , La Joie R , Baker S , Bourakova V , Boxer AL , Edwards L , Eser R , Gorno-Tempini ML , Jagust WJ , Janabi M , Kramer JH , Lesman-Segev OH , Mellinger T , Miller BL , Pham J , Rosen HJ , Spina S , Seeley WW , Strom A , Grinberg LT , Rabinovici GD , 2020. 18F-flortaucipir PET to autopsy comparisons in Alzheimer's disease and other neurodegenerative diseases. Brain 143 , 3477–3494.33141172
Suleymanova I , Balassa T , Tripathi S , Molnar C , Saarma M , Sidorova Y , Horvath P , 2018. A deep convolutional neural network approach for astrocyte detection. Sci Rep 8 , 12878.30150631
Theofilas P , Polichiso L , Wang X , Lima LC , Alho AT , Leite RE , Suemoto CK , Pasqualucci CA , Jacob-Filho W , Heinsen H , Brazilian Aging Brain Study, G., Grinberg LT , 2014. A novel approach for integrative studies on neurodegenerative diseases in human brains. J Neurosci Methods 226 , 171–183.24503023
Tiepolt S , Patt M , Aghakhanyan G , Meyer PM , Hesse S , Barthel H , Sabri O , 2019. Current radiotracers to image neurodegenerative diseases. EJNMMI Radiopharm Chem 4 , 17.31659510
Tustison NJ , Avants BB , 2013. Explicit B-spline regularization in diffeomorphic image registration. Front Neuroinform 7 , 39.24409140
Tustison NJ , Avants BB , Cook PA , Zheng Y , Egan A , Yushkevich PA , Gee JC , 2010. N4ITK: improved N3 bias correction. IEEE Trans Med Imaging 29 , 1310–1320.20378467
Van Rossum G , Drake FL Jr , 1995. Python tutorial. . Centrum voor Wiskunde en Informatica, Amsterdam, The Netherlands.
Wang YT , Edison P , 2019. Tau Imaging in Neurodegenerative Diseases Using Positron Emission Tomography. Curr Neurol Neurosci Rep 19 , 45.31172290
Yushkevich PA , Lopez MM , Martin M , Ittyerah R , Lim S , Ravikumar S , Bedard ML , Pickup S , Liu W , Wang J , Hung LY , Lasserve J , Vergnet N , Xie L , Dong M , Cui S , McCollum L , Robinson JL , Schuck T , de Flores R , Grossman M , Tisdall MD , Prabhakaran K , Mizsei G , Das SR , Artacho-Perula E , Jimenez M , Rabal M , Romero FJM , Sanchez SC , Gonzalez JCD , de la Rosa-Prieto C , Parada MC , Lee EB , Trojanowski JQ , Ohm DT , Wisse LEM , Wolk DA , Irwin DJ , Insausti R , 2021. Three-dimensional mapping of neurofibrillary tangle burden in the human medial temporal lobe. Brain.
Zack GW , Rogers WE , Latt SA , 1977. Automatic measurement of sister chromatid exchange frequency. J Histochem Cytochem 25 , 741–753.70454
|
PMC008xxxxxx/PMC8983028.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
9208117
21801
Platelets
Platelets
Platelets
0953-7104
1369-1635
30422039
8983028
10.1080/09537104.2018.1542125
NIHMS1778778
Article
Chronic liver disease, thrombocytopenia and procedural bleeding risk; are novel thrombopoietin mimetics the solution?
Olson Sven R. http://orcid.org/0000-0002-2717-4620
1
Koprowski Steven 1
Hum Justine 2
McCarty Owen J.T. 13
DeLoughery Thomas G. 1
Shatzel Joseph J. 13
1 Division of Hematology and Medical Oncology, Oregon Health and Science University, Knight Cancer Institute, Portland, OR, USA
2 Division of Gastroenterology and Hepatology, Oregon Health & Science University, Portland, OR, USA
3 Department of Biomedical Engineering, Oregon Health & Science University, Portland, OR, USA
Correspondence: Sven R. Olson, Division of Hematology and Medical Oncology, Oregon Health & Science University, 3181 SW Sam Jackson Park Rd, Portland, OR 97239. [email protected]
13 2 2022
2019
13 11 2018
05 4 2022
30 6 796798
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Chronic liver disease (CLD) alters normal hemostatic and thrombotic systems via multiple mechanisms including reduced platelet function and number, leading to challenging peri-operative planning. Hepatic thrombopoietin (TPO) synthesis is reduced in CLD, leading to several recent randomized, placebo-controlled trials examining the utility of TPO-mimetics to increase platelet counts prior to surgery. While these trials do suggest that TPO-mimetics are efficacious at increasing platelet counts in patients with CLD and have led to several recent drug approvals in this space by the U.S. Food & Drug Administration, it remains unclear whether these results translate to the relevant clinical endpoint of reduced perioperative bleeding rate and severity. In this article, we review several recently-published, phase 3 trials on the TPO-mimetics eltrombopag, avatrombopag and lusutrombopag, and discuss the clinical significance of their results.
Bleeding
Platelets
thrombopoietin
pmcChronic liver disease (CLD) is associated with complex changes in the body’s normal hemostatic and thrombotic systems including altered synthesis of pro- and anticoagulant proteins, accelerated fibrinolysis, and reduced platelet function and number [1]. The magnitude of these changes increases with worsening hepatic fibrosis, resulting in unpredictable bleeding and thrombotic risks and challenging surgical planning. Thrombocytopenia increases risk for procedural bleeding in the general population, with a goal platelet count of >50,000/mL recommended by societal guidelines prior to major surgeries [2]; similar consensus does not exist for platelet goals in CLD. Platelet transfusions can be complicated by transfusion reactions, infections, and platelet alloimmunization, all of which can increase health care costs. Based on the pathophysiology of CLD-associated thrombocytopenia, which includes reduced hepatic synthesis of thrombopoietin (TPO), multiple recent randomized controlled trials (RCT) have explored the pre-operative administration of oral thrombopoietin mimetics to increase platelet counts in CLD prior to elective surgery. These trials include a study of the TPO-mimetic Eltrombopag (Promacta®, Novartis) in ELEVATE [3], and four studies of the newer second-generation Avatrombopag (Doptelet®, Dova Pharmaceuticals) in ADAPT-1, ADAPT-2 [4], and Lusutrombopag (Mulpleta®, Shionogi Inc.), in the L-PLUS 1, and L-PLUS 2 trials.
Eltrombopag was approved by the U.S. Food and Drug Administration in 2008 for second-line treatment of immune thrombocytopenic purpura, as well as thrombocytopenia due to hepatitis C infection to facilitate interferon-based therapy. Given these promising data, the phase III ELEVATE trial examined Eltrombopag vs. placebo for treatment of thrombocytopenia in CLD with platelets <50,000/mL prior to elective procedures [3]. The primary endpoint was avoidance of platelet transfusion before or 7 days after procedures, with a secondary endpoint consisting of number of patients experiencing WHO grade ≥ 2 bleeding. The majority of patients underwent “low risk” procedures including paracentesis and endoscopy ± biopsies. The high rates of portal vein thrombosis (PVT) (6 Eltrombopag arm, 1 placebo arm) lead to early trial termination; however, the Eltrombopag arm did have a statistically significantly smaller proportion of patients requiring platelet transfusions (19% vs. 72% p < 0.0001), the number of platelets transfused per episode was lower, and bleeding rates were equivalent (Table 1). A common criticism of this study included the lack of screening ultrasound at enrollment to evaluate for pre-existing occult PVT, which are not uncommon in CLD. Given the safety concerns with Eltrombopag, further large-scale studies in this setting were abandoned until recently, with the development of several newer TPO mimetics.
Four recent, phase III RCTs evaluating Avatrombopag and Lusutrombopag in CLD patients have generated renewed interest in the use of TPO mimetic prior to elective procedures. The ADAPT-1 and ADAPT-2 trials examined Avatrombopag vs. placebo in a total 435 patients, with the primary endpoint of reducing platelet transfusions or rescue procedures for bleeding up to 7 days post-operatively [4]. Inclusion criteria were similar to ELEVATE, and the majority of procedures performed were similarly low risk. Cohorts with low (<40,000/mL) and high (40–50,000/mL) baseline platelet counts were studied; in both trials and both cohorts, Avatrombopag statistically significantly reduced the number of platelet transfusions required (Table 1), with an effect magnitude similar to ELEVATE. Thrombosis occurred less frequently (placebo 2, Avatrombopag 1). WHO Grade ≥ 2 bleeding rates were similar between experimental and placebo arms in both low and high baseline platelet cohorts (3.8% vs. 3.3%, 2.6% vs. 4.6%, respectively). The amount of platelets transfused for each episode and frequency of more aggressive hemostatic measures (coagulation factors, vitamin K, anti-fibrinolytics, surgical procedures) were not reported.
Finally, the phase III L-PLUS 1 and L-PLUS 2 trials, results of which were presented at the American Association of the Study of Liver Diseases 2017 annual meeting, compared the novel TPO-mimetic Lusutrombopag against placebo in 311 patients for the same indication as the ELEVATE and ADAPT trials, with a primary endpoint of proportion of patients needing pre-operative (L-PLUS 1) or either pre- or post-operative (L-PLUS 2) platelet transfusions. These trials also found statistically significant reductions in platelet transfusion requirements. Full details of these trials, including specifics of bleeding outcomes, are not currently published. Based on the ADAPT and L-PLUS trials, both Avatrombopag and Lusutrombopag received FDA approval in May and July 2018, respectively.
Collectively, these trials emphasize several important takeaways: TPO-mimetics are effective at improving platelet counts in CLD and are generally well-tolerated. The increased signal for PVT seen with Eltrombopag was not seen in trials of newer TPO-mimetics, possibly due to drug differences, but more likely due to the ADAPT and L-PLUS trials’ screening and exclusion of patients with previous venous thromboembolic disease, and their avoidance of inducing platelet levels >100,000/mL.
While clearly effective at reducing the need for perioperative platelet transfusion, several questions still remain surrounding the clinical utility of TPO mimetics in this setting. Most critical is whether these drugs actually reduce the risk of peri-procedural bleeding in thrombocytopenic patients with CLD; unfortunately, the design of these trials leaves this question somewhat unanswered as bleeding rates were not the main study endpoints or were reported with limited detail. Integral to this issue is the knowledge that CLD affects many coagulation parameters, such that a single variable such as platelet count may be insufficient to predict bleeding risk. No other hemostatic assays were reported in the current trials, namely prothrombin time, activated partial thromboplastin time, fibrinogen or global point of care assays such as thromboelastography (TEG). In particular, abundant literature has been published on the ability of TEG to accurately predict bleeding risk in trauma and liver transplant settings, with emerging literature on its use in CLD [5]. As the ADAPT authors themselves point out, the optimal safe platelet level in CLD prior to procedures is also unclear. Indeed, recent data from Basili et al. suggests that platelet levels correlate poorly with bleeding in CLD [6]. These studies also excluded patients with more severe liver dysfunction (Childs-Pugh C or MELD > 24), a population more likely to experience procedural bleeding. Finally, the financial burden of these drugs relative to the cost of platelet transfusions must be considered. Avatrombopag is listed at $1,080 USD per 20 mg tablet, with a typical course consisting of 10 tablets over 5 days, amounting to $10,800 per treatment course. In contrast, a recent analysis estimated the cost of single apheresis unit of platelets (including administration and facility costs) to be anywhere from $3,723 to $4,436 USD [7], though it must be noted that this study was performed via financial support from Dova Pharmaceuticals, the makers of Avatrombopag. These cost comparisons should be viewed in the context of the high proportion of patients in the experimental arms of these studies who ended up requiring platelet transfusions despite use of a TPO-mimetic (anywhere from 12% to 35%). Only ELEVATE reported the average number of platelet units transfused per episode, with a difference of 1 unit between the Eltrombopag and placebo groups (median 3 vs. 4) [3]. Another limitation is a lack of reporting of transfusion-related complications, including transfusion reactions, infections, and platelet alloimmunization which are not well described in patients with CLD, and if significant, may favor the use of TPO-mimetics.
Future studies would benefit from more precise evaluation of bleeding rates within cohorts, the predictive value of other markers of hemostasis, better descriptions of transfusion complications, and cost effectiveness analysis. Nevertheless, there are multiple populations within CLD potentially worthy of future studies with TPO-mimetics, including those with chronic, severe thrombocytopenia ± recurrent bleeding events, or those who are already platelet-refractory. TPO mimetics are clearly effective at increasing platelet counts in CLD, and the findings from recent large-scale prospective trials are compelling in their exploration of an alternative to platelet transfusions in patients with CLD undergoing invasive procedures. Many patients will justifiably wish to avoid the need for platelet transfusions in favor of taking an oral medication prior to surgeries, which could potentially decrease costs and complication rates.
Funding
This work was supported by grants from the National Institutes of Health (R01HL101972, R01GM116184).
Table I. Descriptions of RCTs examining TPO-mimetics for patients with CLD undergoing procedures.
Trial ELEVATE ADAPT-1 ADAPT-2 L-PLUS 1 L-PLUS 2
Drug Eltrombopag Avatrombopag Avatrombopag Lusutrombopag Lusutrombopag
Published N. Eng. J Med 2012 J. Gastroenterol 2018 J. Gastroenterol 2018 Abstract only Abstract only
Trial phase -Phase 3 -Phase 3 -Phase 3 -Phase 3 -Phase 3
-Randomized -Randomized -Randomized -Randomized -Randomized
-Double-blinded -Double-blinded -Double-blinded -Double-blinded -Double-blinded
-Placebo-controlled -Placebo-controlled -Placebo-controlled -Placebo-controlled -Placebo-controlled
Inclusion criteria -CLD, Child Pugh 5–12, MELD ≤ 24 -CLD, MELD ≤ 24 -CLD, MELD ≤ 24 -Chronic liver disease (CLD) -CLD, Child-Pugh A or B
-Plt count <50 × 109/L -Plt count <50 × 109/L -Plt count <50 × 109/L -Plt count <50 × 109/L -Plt count <50 × 109/L
-Planned elective procedure -Planned procedure (low, moderate, high bleeding risk) -Planned procedure (low, moderate, high bleeding risk) -Planned invasive procedure -Planned non-emergency invasive procedure
Primary endpoint -No plt transfusion before, during, up to 7 days post-op -No plt transfusion or rescue procedures for bleeding up to 7 days post-op -No plt transfusion or rescue procedures for bleeding up to 7 days post-op -No preoperative plt transfusion -No preoperative plt transfusion
-No rescue therapy for bleeding up to 7 days post-op
Secondary endpoint(s) -Proportion of plts with bleeding WHO 2 or greater, before, during, up to 7 after proc -Proportion of responders (Plt count > 50 × 109/L by day of surgery) -Proportion of responders (Plt count > 50 × 109/L by day of surgery) -Proportion of responders (Plt count ≥ 50 × 109/L or increase by > 20 × 109/L) -Proportion of responders (Plt count ≥ 50 × 109/L or increase by > 20 × 109/L)
-Number of plt transfusions before, during, and up to 30 days post-op -Duration of plt increase -Duration of plt increase
-Plt transfusion or rescue therapy through 35 days post-op
Total number of patients 292 231 204 96 215
Number with TPO 145 149 (low and high plts) 128 (low and high plts) 48 108
Number with placebo 147 82 76 48 107
Number not reaching primary endpoint TPO 41 -Low plts (31/90) -Low plts (22/70) 10 38
-High plts (7/59) -High plts (7/58)
Percent of patients not reaching primary endpoint TPO 28% -Low plts (34.4%) -Low plts (31.4%) 20.80% 35.2%
-High plts (11.9%) -High plts (12.1%)
Number not reaching primary endpoint Placebo 119 -Low plts (37/48) -Low plts (28/43) 42 76
-High plts (21/34) -High plts (22/33)
Percent of patients not reaching primary endpoint Placebo 81% -Low plts (77.1%) -Low plts (65.1%) 87.50% 71%
-High plts (61.8%) -High plts (66.67%)
Total bleeding related adverse events TPO 17% 3.80% 2.60% 14.60% 2.8%
Total bleeding related adverse events Placebo 23% 3.30% 4.60% 27.10% 5.6%
Major bleeding TPO - - - - 0%
Major bleeding placebo - - - - 1.9% (required rescue therapy for bleeding)
Thrombosis TPO 6 portal vein thrombus 1 1 1 portal vein thrombus 1 portal vein thrombus
Thrombosis placebo 2 portal vein thrombus 0 2 1 portal vein thrombus 2 portal vein thrombus
Mean baseline plt count - -Low - 31 × 109/L TPO+ placebo -Low - 33 × 109/L TPO+ placebo −40.4 × 109/L
-High - 45 × 109/L placebo, 44 × 109/L TPO -High - 45 × 109/L placebo, 44 × 109/L TPO
Procedure breakdown Per bleeding risk:
-Category 1 (56% placebo, 62% TPO)
-Category 2 (14% placebo, 16% TPO)
-Category 3 (14% placebo, 10% TPO
-Category 4 (3% placebo, 3% TPO) Separated in to bleeding risk:
-Low (59.7% placebo, 61.4% TPO)
-Moderate (20.1% placebo, 15.5% TPO)
-High (20.1% placebo, 23.1% TPO) -Percutaneous liver ablation (42.7%)
-Transcatheter arterial chemoembolization (25.0%)
-Endoscopic variceal ligation (14.6%)
-High (20.1% placebo, 23.1% TPO) -Endoscopic variceal ligation (28%)
-Endoscopy (27%)
-Dental extraction (11%)
-TACE (%)
Disclosure statement
The authors state no conflict of interest.
References
1. Tripodi A , Mannucci PM . The coagulopathy of chronic liver disease. New England J Med 2011;365 (2 ):147–156. doi:10.1056/NEJMra1011170 21751907
2. Kaufman RM , Djulbegovic B , Gernsheimer T , Kleinman S , Tinmouth AT , Capocelli KE , Cipolle MD , Cohn CS , Fung MK , Grossman BJ , Platelet transfusion: a clinical practice guideline from the aabb. Ann Intern Med 2015;162 (3 ):205–213. doi:10.7326/M14-1589 25383671
3. Afdhal NH , Giannini EG , Tayyab G , Mohsin A , Lee J-W , Andriulli A , Jeffers L , McHutchison J , Chen P-J , Han K-H , Eltrombopag before procedures in patients with cirrhosis and thrombocytopenia. N Engl J Med 2012;367 (8 ):716–724. doi:10.1056/NEJMoa1110709 22913681
4. Terrault N , Chen Y-C , Izumi N , Kayali Z , Mitrut P , Tak WY , Allen LF , Hassanein T . Avatrombopag before procedures reduces need for platelet transfusion in patients with chronic liver disease and thrombocytopenia. Gastroenterology 2018;155 (3 ):705–718. doi:10.1053/j.gastro.2018.05.025 29778606
5. Stravitz RT . Potential applications of thromboelastography in patients with acute and chronic liver disease. Gastroenterol Hepatol (N Y) 2012;8 (8 ):513–520.23293564
6. Basili S , Raparelli V , Napoleone L , Talerico G , Corazza GR , Perticone F , Sacerdoti D , Andriulli A , Licata A , Pietrangelo A , Platelet count does not predict bleeding in cirrhotic patients: results from the PRO-LIVER study. Am J Gastroenterol 2018;113 (3 ):368–375. doi:10.1038/ajg.2017.457 29257146
7. Barnett CL , Mladsi D , Vredenburg M , Aggarwal K . Cost estimate of platelet transfusion in the United States for patients with chronic liver disease and associated thrombocytopenia undergoing elective procedures. J Med Econ 2018;21 (8 ):827–834. doi:10.1080/13696998.2018.1490301 29912593
|
PMC008xxxxxx/PMC8983056.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
101215604
32338
Nat Methods
Nat Methods
Nature methods
1548-7091
1548-7105
34194050
8983056
10.1038/s41592-021-01186-4
NIHMS1790732
Article
The ENIGMA Toolbox: Multiscale neural contextualization of multisite neuroimaging datasets
Larivière Sara 1
Paquola Casey 23
Park Bo-yong 45
Royer Jessica 6
Wang Yezhou 7
Benkarim Oualid 8
de Wael Reinder Vos 9
Valk Sofie L. 1011
Thomopoulos Sophia I. 12
Kirschner Matthias 13
ENIGMA Consortium14
Lewis Lindsay B. 1516
Evans Alan C. 1718
Sisodiya Sanjay M. 1920
McDonald Carrie R. 21
Thompson Paul M. 22
Bernhardt Boris C. 23
1 Multimodal Imaging and Connectome Analysis Laboratory, McConnell Brain Imaging Centre, Montreal Neurological Institute and Hospital, McGill University, Montreal, Quebec, Canada
2 Multimodal Imaging and Connectome Analysis Laboratory, McConnell Brain Imaging Centre, Montreal Neurological Institute and Hospital, McGill University, Montreal, Quebec, Canada
3 Institute of Neuroscience and Medicine (INM-1), Research Center, Jülich D-52425 Jülich, Germany
4 Multimodal Imaging and Connectome Analysis Laboratory, McConnell Brain Imaging Centre, Montreal Neurological Institute and Hospital, McGill University, Montreal, Quebec, Canada
5 Department of Data Science, Inha University, Incheon, South Korea
6 Multimodal Imaging and Connectome Analysis Laboratory, McConnell Brain Imaging Centre, Montreal Neurological Institute and Hospital, McGill University, Montreal, Quebec, Canada
7 Multimodal Imaging and Connectome Analysis Laboratory, McConnell Brain Imaging Centre, Montreal Neurological Institute and Hospital, McGill University, Montreal, Quebec, Canada
8 Multimodal Imaging and Connectome Analysis Laboratory, McConnell Brain Imaging Centre, Montreal Neurological Institute and Hospital, McGill University, Montreal, Quebec, Canada
9 Multimodal Imaging and Connectome Analysis Laboratory, McConnell Brain Imaging Centre, Montreal Neurological Institute and Hospital, McGill University, Montreal, Quebec, Canada
10 Otto Hahn Research Group for Cognitive Neurogenetics, Max Planck Institute for Cognitive and Brain Sciences, Leipzig, Germany
11 INM-7, FZ Jülich, Jülich, Germany
12 Imaging Genetics Center, Mark & Mary Stevens Neuroimaging and Informatics Institute, Keck School of Medicine, University of Southern California, Marina del Rey, California, United States
13 Department of Psychiatry, Psychotherapy and Psychosomatics, Psychiatric Hospital, University of Zurich, Switzerland
14 http://enigma.ini.usc.edu
15 McConnell Brain Imaging Centre, Montreal Neurological Institute and Hospital, McGill University, Montreal, Quebec, Canada
16 McGill Centre for Integrative Neuroscience, McGill University, Montreal, Quebec, Canada
17 McConnell Brain Imaging Centre, Montreal Neurological Institute and Hospital, McGill University, Montreal, Quebec, Canada
18 McGill Centre for Integrative Neuroscience, McGill University, Montreal, Quebec, Canada
19 Department of Clinical and Experimental Epilepsy, UCL Queen Square Institute of Neurology, London, United Kingdom
20 Chalfont Centre for Epilepsy, Bucks, United Kingdom
21 Department of Psychiatry, Center for Multimodal Imaging and Genetics, University of California San Diego, La Jolla, California, United States
22 Imaging Genetics Center, Mark & Mary Stevens Neuroimaging and Informatics Institute, Keck School of Medicine, University of Southern California, Marina del Rey, California, United States
23 Multimodal Imaging and Connectome Analysis Laboratory, McConnell Brain Imaging Centre, Montreal Neurological Institute and Hospital, McGill University, Montreal, Quebec, Canada
CORRESPONDENCE TO: Boris C. Bernhardt, [email protected]; Sara Larivière, [email protected]
Author contributions
Core developers: S.L., B.C.B.
Toolbox beta testing: B.-Y.P., J.R., C.P., S.L.V., Y.W, M.K.
Writing: S.L., B.C.B.; revised and approved by other listed co-authors.
25 3 2022
7 2021
01 7 2022
18 7 698700
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
pmcAmong big data neuroscience initiatives, the ENIGMA (Enhancing NeuroImaging Genetics through Meta-Analysis) Consortium—a worldwide alliance of over 2,000 scientists diversified into over 50 Working Groups—has yielded some of the largest studies of the healthy and diseased human brain. Through harmonized procedures, and by sharing site-specific brain metrics (e.g., cortical thickness) or aggregated statistical maps, ENIGMA has set the stage for large-scale analyses comparing findings across different topics or disorders 1, 2. In parallel, increasingly available resources offer opportunities to contextualize findings across multiscale brain organization. Examples include the Allen Human Brain Atlas3 (AHBA; microarray-derived postmortem gene expression), the BigBrain Project4 (3D postmortem human brain histology), and the Human Connectome Project5 (HCP; high-definition in vivo functional and structural connectomics). Here we introduce the ENIGMA Toolbox, an open ecosystem for integration and visualization of multisite ENIGMA results and their multiscale neural contextualization. Our Toolbox relies on an efficient codebase for exploring and analyzing big data, aiming to facilitate and homogenize follow-up analyses of ENIGMA, or other, MRI datasets around the globe.
To advance and simplify cross-disorder analysis and multiscale neural contextualization of neuroimaging findings, the ENIGMA Toolbox offers the ability to access over 100 ENIGMA-derived statistical maps (Fig. 1A), to visualize and manipulate cortical and subcortical surface data and generate publication-ready figures (Fig. 1B), and to contextualize neuroimaging findings at the microscale (postmortem gene expression and cytoarchitecture; Fig. 1C) and macroscale (structural and functional connectome properties; Fig. 1D). To increase generalizability and usability, our Toolbox is compatible with most neuroimaging data and supports the mapping between parcellation maps and vertexwise surface space. To ensure cross-software compatibility, Toolbox users can also export data results to a range of file formats.
As of release v1.1.0, our Toolbox includes 100+ case-control summary statistics from several ENIGMA Working Groups, including: 22.q11.2 deletion syndrome, attention deficit hyperactivity disorder, autism spectrum disorder, bipolar disorder, epilepsy, major depressive disorder, obsessive-compulsive disorder, and schizophrenia. These datasets, obtained from standardized and quality-controlled protocols, represent morphological (e.g., subcortical volume, cortical thickness, surface area) case-control effect sizes from previously published meta- and mega-analyses. Summary statistics from other Working Groups will be continuously added to the Toolbox as they are published.
To deepen our understanding of the molecular and cellular underpinnings of healthy and diseased brain organization, our Toolbox provides microscale neural contextualization workflows to cross-reference neuroimaging findings against the AHBA brain-wide gene expression data. Toolbox users can import microarray expression datasets, visualize brain maps of gene expression levels, relate gene expression to properties of brain organization, and identify genes that are spatially correlated with a given brain map. The cell body-stained BigBrain data4 and the von Economo and Koskinas cytoarchitectural atlas6 are also accessible within our Toolbox. These digitized brain maps are invaluable for linking brain microstructure to functional localization and enable users to speculate on the underlying cytoarchitectural composition of, for instance, structurally abnormal areas in specific diseases. When combined, transcriptomic and cytoarchitectonic decoding can embed neuroimaging findings in a rich neurobiological context and yield insights into the etiology of several brain disorders.
At the macroscopic level, network connectivity offers a vantage point to quantify brain reorganization in diseases that are increasingly being conceptualized as network disorders. Our Toolbox provides tools to relate surface maps to normative connectome properties derived from functional and structural HCP data. Building on prior neurodegenerative7, 8, psychiatric9, and epilepsy10 research, Toolbox users can build hub susceptibility models to assess the vulnerability of highly connected network hubs to disease-related effects and disease epicenter models to identify regions whose connectivity profiles herald disease-related effects. Combined, these two network models can advance our understanding of how connectome architecture relates to morphological abnormalities across a range of disorders. Network models can be further enriched with microstructural properties to inject multiscale information into cortical and subcortical morphometric findings.
By bridging the gap between pre-established data processing protocols and analytic workflows, we hope that the ENIGMA Toolbox facilitates neuroscientific contextualization of results and cross-consortia initiatives. We are eager for researchers and clinicians to test hypotheses beyond traditional case-control comparisons. We hope that our platform will lead to novel and harmonized analyses in global neuroimaging initiatives.
Acknowledgments
Many scientists around the world contributed to ENIGMA but did not take part in the writing of this report. A full list of contributors to ENIGMA is available here: http://enigma.ini.usc.edu/about-2/consortium/members/.
The authors would like to express their gratitude to the open science initiatives that made this work possible: (i) The ENIGMA Consortium (core funding for ENIGMA was provided by the NIH Big Data to Knowledge (BD2K) program under consortium grant U54 EB020403 to P.M.T.), (ii) The Allen Human Brain Atlas, (iii) BigBrain/HIBALL, and (iv) The Human Connectome Project (Principal Investigators: David Van Essen and Kamil Ugurbil; 1U54MH091657) funded by the 16 NIH Institutes and Centers that support the NIH Blueprint for Neuroscience Research; and by the McDonnell Center for Systems Neuroscience at Washington University.
S.L. acknowledges funding from Fonds de la Recherche du Québec – Santé (FRQ-S) and the Canadian Institutes of Health Research (CIHR). C.P. was funded through a postdoctoral FRQ-S fellowship. O.B. was funded by a Healthy Brains for Healthy Lives (HBHL) postdoctoral fellowship. B.-y.P. was funded by the National Research Foundation of Korea (NRF-2020R1A6A3A03037088), Molson Neuro-Engineering fellowship from the Montreal Neurological Institute and Hospital, and FRQ-S. J.R. was supported by CIHR. R.V.d.W. was funded by studentships from the Savoy Foundation for Epilepsy and the Richard and Ann Sievers award. S.L.V. was supported by the Otto Hahn award of the Max Planck Society. M.K. acknowledges funding from the Swiss National Science Foundation (P2SKP3_ 178175). S.M.S. was supported by the Epilepsy Society, UK. Part of this work was undertaken at University College London Hospitals, which received a proportion of funding from the NIHR Biomedical Research Centres funding scheme. C.R.M. acknowledges funding from the National Institutes of Health (NINDS R01NS065838 and R21 NS107739). B.C.B. acknowledges research funding from the SickKids Foundation (NI17-039), the Natural Sciences and Engineering Research Council of Canada (NSERC; Discovery-1304413), CIHR (FDN-154298), Azrieli Center for Autism Research (ACAR), an MNI-Cambridge collaboration grant, salary support from FRQ-S (Chercheur-Boursier), BrainCanada, the Helmholtz BigBrain Analytics and Learning Lab (Hiball), and the Canada Research Chairs (CRC) Program.
Figure 1. Overview of the ENIGMA Toolbox.
(a) World map of a subset of Working Groups. 100+ summary statistics from published studies are accessible within the ENIGMA Toolbox. (b) Surface visualization tools are provided to project cortical and subcortical data results to the surface. As an example, we displayed gray matter atrophy in left focal epilepsy. (c) To contextualize neuroimaging data with respect to microscale brain organization, Toolbox users can fetch disease-related gene expression data (here, we displayed the average expression levels of focal epilepsy genes (left)), stratify atrophy (or other effect maps) according to BigBrain statistical moments (middle left) and gradient (middle right), and stratify atrophy patterns according to cytoarchitectonic classes (right). (d) To contextualize neuroimaging data with respect to macroscale brain organization, Toolbox users can load preprocessed functional and structural connectivity data (left), build hub susceptibility models to assess relationships between hub regions and atrophy patterns (middle left) and assess statistical significance of two surface maps using spin permutation testing (middle right), and identify cortical and subcortical disease epicenters that herald patterns of atrophy.
Code availability
Our Toolbox is available in Python and Matlab and complemented with expandable online documentation (http://enigma-toolbox.readthedocs.io). Derivative data (e.g., summary statistics from published ENIGMA studies) and codes are openly accessible under the terms of the BSD-3-Clause license at http://github.com/MICA-MNI/ENIGMA. Requests to work on a project with subject-level data can be proposed to the Working Group via the chairs (http://enigma.ini.usc.edu/). Users seeking help are encouraged to subscribe and post their questions to the ENIGMA Toolbox mailing list at https://groups.google.com/g/enigma-toolbox.
Ethics declaration
The authors declare no competing interests. P.M.T. received partial grant support from Biogen, Inc., and consulting payments from Kairos Venture Capital, Inc., for work unrelated to ENIGMA and this manuscript.
References
1. Patel Y Jama Psychiatry (2020).
2. Boedhoe PS American Journal of Psychiatry 177 , 834–843 (2020).
3. Hawrylycz MJ Nature 489 , 391–399 (2012).22996553
4. Amunts K Science 340 , 1472–1475 (2013).23788795
5. Van Essen DC Neuroimage 62 , 2222–2231 (2012).22366334
6. von Economo CF & Koskinas GN , Vol. (J. Springer, 1925).
7. Zhou J , Gennatas ED , Kramer JH , Miller BL & Seeley WW Neuron 73 , 1216–1227 (2012).22445348
8. Zheng Y-Q PLoS biology 17 , e3000495 (2019).31751329
9. Gollo LL Nature neuroscience 21 , 1107–1116 (2018).30038275
10. Larivière S Sci Adv 6 (2020).
|
PMC008xxxxxx/PMC8983059.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
9712381
20688
Int J Clin Pract
Int J Clin Pract
International journal of clinical practice
1368-5031
1742-1241
34490963
8983059
10.1111/ijcp.14758
NIHMS1793364
Article
Impact of malignancy on In-hospital mortality, stratified by the cause of admission: An analysis of 67 million patients from the National Inpatient Sample
Kobo Ofer http://orcid.org/0000-0001-7418-498X
12
Brown Sherry-Ann 3
Nafee Tarek 4
Mohamed Mohamed O. http://orcid.org/0000-0002-9678-5222
2
Sharma Kamal 5
Istanbuly Sedralmontaha 6
Roguin Ariel 12
Cheng Richard K. 7
Mamas Mamas A. 28
1 Department of Cardiology, Hillel Yaffe Medical Center, Hadera, Israel
2 Keele Cardiovascular Research Group, Keele University, Stoke on Trent, United Kingdom
3 Cardio-Oncology Program, Division of Cardiovascular Medicine, Medical College of Wisconsin, Milwaukee, WI, USA
4 Department of Medicine, Roger Williams Medical Center, Boston University School of Medicine, Boston, MA, USA
5 U.N. Mehta ICRC, B. J. Medical College, Ahmedabad, India
6 Faculty of Medicine, University of Aleppo, Aleppo, Syrian Arab Republic
7 Division of Cardiology, University of Washington Heart Institute, Seattle, WA, USA
8 Institute of Population Health, University of Manchester, Manchester, United Kingdom
Correspondence Mamas A. Mamas, Keele Cardiovascular Research Group, Keele University, Stoke-on-Trent, United Kingdom. [email protected]
3 4 2022
11 2021
14 9 2021
05 4 2022
75 11 e14758e14758
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Objective:
To describe the patient characteristics and the reason for admission of patients with malignancy by malignancy, and to study mortality rates for the different causes of admissions among the different types of cancer.
Patients and Methods:
Using the nationwide Inpatient Sampling (2015–2017) we examined the cause of admission and associated in-hospital mortality, stratified by presence and type of malignancy. Multivariable logistic regression models were used to examine the association between in-hospital mortality and malignancy sites for different primary admission causes.
Results:
Out of 67 819 693 inpatient admissions, 8.8% had malignancy. Amongst those with malignancy, haematological malignancy was the most common (20.2%). The most common cause of admission amongst all cancers were malignancy-related admissions, where up to 57% of all colorectal admissions were malignancy-related. The most common non-malignancy cause of admission was infectious causes, which were most frequent among patients with haematological malignancy (18.4%). Patients with malignancy had higher crude mortality rates (5.7% vs 1.9%). Mortality rates were highest among patients with lung cancer (8.7%). Among all admissions, the adjusted rates of mortality were higher for patients with lung (OR 3.65, 95% CI [3.59–3.71]), breast (OR 2.06, 95% CI [1.99–2.13]), haematological (OR 1.79, 95% CI [1.76–1.82]) and colorectal (OR 1.71, 95% CI [1.66–1.76]) malignancies compared with patients with no malignancy.
Conclusion:
Our work highlights the need to consider the burden of cancer on our hospital services and consider how the prognostic impact of different types of admissions may relate to the type of cancer diagnosis and understand whether these differences relate to disparities in clinical care/treatments.
pmc1 | INTRODUC TION
Cancer is a leading cause of mortality in the United States; responsible for ~600 000 deaths per annum.1,2 Approximately 60% of patients with cancer who present to emergency departments are admitted as inpatients to the hospital.3,4 The most common presentations include both cancer-related (pain, metabolic disturbance, nausea and vomiting, malaise) and non-cancer-related causes (pneumonia and pulmonary embolism), which contribute to a 6% 30-day mortality.3–5 While much work has focussed on specific causes of admissions and their outcomes in patients with different types of cancer in emergency departments6–8 there is limited contemporary national data for patients admitted as inpatients. Furthermore, the impact of different cancer types on in-hospital mortality remains unknown at the national level, with little data to inform whether in-hospital mortality associated with the type of malignancy is modified by the specific reason for that particular hospital admission.
Causes of death among cancer patients have been previously described in the literature.9–13 Commonly reported causes of overall mortality (in-hospital and out-of-hospital) among cancer patients include cardiovascular complications and infection.9,11,13 Several advances have been made in recent years in the management of comorbid cardiovascular disease or risk factors in the setting of active or previous malignancy. Despite our expanded understanding, it remains unknown if any specific malignancy carries an increased risk of cause-specific in-hospital mortality. In fact, there is limited data on in-hospital mortality in cancer patients outside of a surgical or critical care setting.14–17 Disentangling the effect of certain malignancies on inpatient mortality associated with an admitting diagnosis would provide clinicians with pertinent prognostic data at the point of admission and may generate hypotheses regarding interventions that may mitigate this risk.
In this study, we aim to describe the patient characteristics and characterise the reason for admission by malignancy type among cancer patients enrolled in the National Inpatient Sample database in the United States from 2015 to 2017. Furthermore, we aim to study mortality rates for the different causes of admissions amongst the different types of cancer. We hypothesise that the presence and type of malignancy increase the incidence of in-hospital mortality, stratified by the cause of admission.
2 | METHODS
2.1 | Data source
The National Inpatient Sample (NIS) is the largest all-payer inpatient health care database in the United States, developed by the Healthcare Cost and Utilization Project (HCUP) and sponsored by the Agency for Healthcare Research and Quality (AHRQ). The NIS dataset contains hospital information on from 7 to 8 million yearly hospital discharges from 2004 onwards. Since 2012, the NIS samples discharge from all hospitals participating in HUCP, approximating a 20% stratified sample of all discharges from US community hospitals.
2.2 | Study design and population
We analysed all adult (≥18 years) patients hospitalised from 1 October 2015 to December 2017. This particular period was chosen because of the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) diagnosis codes that were implemented in October 2015 which would provide more granular data as opposed to the previous ICD-9 coding. Patient characteristics, malignancy status and primary admission cause were extracted using ICD-10 codes provided in Table S1. Information on patient demographics was recorded for each hospital discharge including age, gender, race, admission day (week-day or weekend), expected primary payer and median household income according to ZIP code. Missing records for age, gender, elective and weekend admission, and mortality status were excluded from the analysis (Figure S1 for study flow diagram). Each discharge record had information on up to 40 diagnoses and 25 procedures. The main outcome measured was in-hospital mortality between the cancer groups and the no cancer groups, stratified by primary cause of admission.
2.3 | Statistical analysis
Statistical analysis was performed on IBM SPSS version 25. Continuous variables are presented as a median (25th percentile and 75th percentile) range, because of skewed data, and categorical data are presented as frequencies and percentages. Categorical variables were compared using Pearson’s chi-square test, while continuous variables were compared using the Mann-Whitney U test, as appropriate. Sampling weights were used to calculate the estimated total discharges as specified by AHRQ. Multivariable logistic regression models were used to examine the association between in-hospital mortality and any of the most frequent five malignancy sites (haematological, breast, prostate, lung, and colon malignancies) for different primary admission causes. All models were adjusted for baseline differences between the groups, controlling for the following covariates: age, gender, weekend and elective admission, median zip income, hospital region, bed size, location and teaching status, renal failure, hypothyroidism, obesity, dyslipidaemia, coagulopathy, hypertension, heart failure, chronic ischaemic heart disease, chronic lung disease, smoking status, anaemia, chronic liver disease, diabetes mellitus type 2, thrombocytopenia and dementia.
3 | RESULTS
A total of 67 819 693 inpatient admissions were recorded to hospitals in the United States from 1 October 2015 to December 2017, of which almost 9% (8.8%) had malignancy. Among those with malignancy, haematological malignancy was the most common (20.2%) followed by lung (14.7%), colorectal (9.1%), prostate (7.7%) and breast (6.4%) malignancies. Baseline characteristics of patients with haematological, lung, colorectal, prostate, breast, other malignancy or no malignancy are shown in Table 1. Patients with malignancy were older and were more likely to be men (excluding breast malignancy) and medicare patients. A larger proportion of patients with malignancy were treated in large, teaching hospitals. Generally, patients with malignancy had higher rates of hypertension, dyslipidaemia, chronic renal failure, anaemia, thrombocytopenia and liver disease and lower rates of dementia, obesity and lung disease (excluding lung malignancy). The risk factor profile differed by malignancy, with the prevalence of ischaemic heart disease greatest in those patients with prostate cancer and lowest in those with breast cancer. Similarly, anaemia was most prevalent in patients with haematological malignancies and lowest in prostate cancer. Key differences in the comorbidity profile of the different cancers can be seen in Table 1.
3.1 | Primary causes of admission
Table 2 (and Figure 1) presents the primary cause of admission stratified by cancer type. The most common cause of admission amongst all cancers were malignancy-related admissions, where up to 57% of all colorectal admissions were malignancy-related, whilst this accounted for only 19% of haematological admissions. The most common non-malignancy-related cause of admission was infectious causes, which were most frequent among patients with haematological malignancy (18.4%), followed by lung (16.9%), breast (13.3%), other malignancies (12.3%) and prostate cancer (11.2%) compared with only 11.3% of admissions in patients without cancer. Finally, cardiovascular causes for admissions were less common in patients with cancer, compared with those without cancer (16%) in which cardiovascular admissions were observed in 4.7% of patients with colorectal cancers, 9.7% in patients with lung cancers and 9.2% in patients with breast cancer. When stratified by the presence or absence of metastases (Table S2), similar patterns of admissions were observed, although infectious causes were more commonly observed in the presence of metastases whilst cardiovascular admissions were less common, for all of the cancer types studied. When stratified by the causes of admission (Table S3), haematological malignancies were the most common cause for cardiovascular admissions, accounting for 25.7% of cardiovascular admissions followed by lung malignancies (16.9%).
3.2 | In-hospital mortality
Patients with malignancy had higher crude mortality rates compared with patients with no malignancy (5.7% vs 1.9%) (Table 3, Figure 2). Mortality rates were highest among patients with lung cancer (8.7%), followed by haematological malignancy (5.3%), colorectal cancer (4.2%), breast cancer (4%) and prostate cancer (2.8%). Admission because of infectious causes were associated with the highest mortality among both cancer and no cancer groups. The highest crude mortality rate was observed in patients with lung malignancy (15.2%) followed by colorectal (12.8%), haematological (10.4%), breast (8.4%), prostate (7.3%) and no malignancy (5.4%). Similarly, cardiovascular mortality was greatest in patients with cancer (5.5%) compared with those without malignancy (3.3%), with the highest rates seen in the lung malignancy group (7.9%) and the lowest seen in the breast malignancy group (3.6%).
Among all admissions, the adjusted rates of mortality were higher for patients with lung (OR 3.65, 95% CI [3.59–3.71]), breast (OR 2.06, 95% CI [1.99–2.13]), haematological (OR 1.79, 95% CI [1.76–1.82]) and colorectal (OR 1.71, 95%CI [1.66–1.76]) malignancies compared with patients with no malignancy, and lower in patients with prostate malignancy (OR 0.89, 95% CI [0.86–0.92]) (Table 4, Figure 3). Similar results were observed in patients admitted because of infectious cause (lung malignancy OR 2.89, 95% CI (2.84–2.93), colorectal malignancy OR 1.91, 95% CI (1.86–1.96), breast malignancy OR 1.6, 95% CI (1.55–1.65), haematological malignancy OR 1.47, 95% CI (1.45–1.49), and prostate malignancy OR 0.89, 95% CI [0.86–0.92]). For cardiovascular admissions, patients with lung (OR 1.97, 95%CI [1.92–2.02]), haematological (OR 1.17, 95% CI [1.14–1.2]) and colorectal malignancy (OR 1.13, 95% CI [1.07–1.2]) had an increased odds for mortality compared with patients without malignancy while patients with breast (OR 0.88, 95% CI [0.82–0.93]) and prostate malignancy (OR 0.86, 95% CI [0.81–0.9]) had lower rates of mortality. Among patients admitted with a renal cause, adjusted mortality rates were higher in all five common malignancies compared with patients without malignancy, and among gastrointestinal admissions adjusted mortality rates were significantly higher in patients with lung, breast colorectal and haematological malignancy but similar in patients with prostate malignancy (Table 4, Figure 3).
4 | DISCUSSION
In this analysis of more than 67 million hospitalised patients, we report significant differences in both the causes of admission in patients with different cancers, as well as differences in in-hospital mortality for individual admission types. Overall, the most common cause of admission amongst patients with cancer was malignancy-related, while the most common causes of non-malignancy-related admission were infectious causes followed by cardiovascular diseases. Patients with malignancy were more likely to be older men (excluding breast malignancy) and Medicare was found to be the most common payer type. Patients with malignancy also had increased likelihood of receiving treatment at larger teaching hospitals. Overall, lung cancer had the highest odds of mortality of all malignancies reported, whilst prostate cancer carried the lowest odds of mortality. Infectious admissions were more commonly reported in patients with cancer, with the greatest odds of mortality observed in patients with lung cancer followed by colorectal malignancies. Whilst admission with cardiovascular diseases was less common amongst patients with cancer than in those without, the odds of mortality from a cardiovascular admission was greater in the cancer cohort, with the worst outcomes reported in patients with lung cancer.
Our findings are consistent with an analysis of the US Nationwide Emergency Department Sample in which 29.5 million admissions in patients with active cancer were studied6 in which the most common cancer-related diagnosis was pneumonia, followed by non-specific chest pain, urinary tract infections and septicaemia. The conditions most often resulting in inpatient hospitalisation included septicaemia (98.0%), intestinal obstruction (91.2%), congestive heart failure (90.8%), and pneumonia (88.6%). Similarly, previous national studies derived from the United States report that one in five hospitalisations with sepsis was associated with malignancy,18 with cancer-related-sepsis hospitalisations associated with higher in-hospital mortality (27.9% vs 19.5%, P < .001), with haematological malignancies associated with the greatest odds of mortality (OR 1.4 95% CI 1.3–1.4) compared with solid tumours. When solid tumours were stratified by cancer type, lung cancer had amongst the highest odds of mortality (OR 95% CI 2.2–2.3) of those reported. The increased prevalence of infectious admissions in patients with cancer is likely to be multi-factorial. Cancer treatments (eg, chemotherapy, radiation, surgery, bone marrow transplantation, and blood products) are known to increase the risk for infections, whilst cancer is known to be associated with immunosuppression.
Interestingly we report lower prevalence of cardiovascular-related admissions in patients with cancer, but higher mortality for many of the cancers studied. Previous ED studies have suggested that circulatory diseases as a principal diagnosis were more commonly observed for cancer patients (7% vs 4.1%), although it wasn’t clear what proportion of these patients with a circulatory disease were admitted to hospital or died whilst in the ED. Our previous work reporting outcomes of 27 million admissions to ED with cardiovascular disease in the United States suggests that 60% of deaths occur in the ED department, prior to hospital admission with the commonest cause being cardiac arrest.19
Patients with cancer are known to have an increased risk of cardiovascular events compared with the general population, with substantial variations between cancer sites,20,21 related to shared risk factors and cardiotoxicity from cancer therapies.22 Our previous work has shown that patients with cancer admitted with acute myocardial infarction23 or ST-elevation myocardial infarction24 have a significantly greater odds of in-hospital mortality, compared with patients without cancer with prognosis related to both cancer type and stage, with worse outcomes seen particularly amongst patients with lung or colon cancer23 or haematological malignancies.25 One of the factors that may contribute to this is that patients with cancer are less likely to be invasively treated, even though the potential benefits may be similar24 Similarly, patients admitted with acute heart failure syndrome and comorbid cancer have worse outcomes,26 were less likely to be treated with ACEi and B-blockers or to receive invasive management such as respiratory support, haemodialysis or intubation. A recent analysis from the US Nationwide Inpatient Sample has suggested that patients with heart failure are also less likely to be offered CRT or CIED devices, with worse in-hospital outcomes.27 Cancer patients admitted with a principal diagnosis of heart failure, are not only at greater risk of mortality but also have a greater risk of sustaining in-hospital adverse events including pneumonia, urinary tract infections and sepsis.26 Care must be taken to help these patients avoid infection and hospital admission. One group has suggested the use of a risk score to identify patients with malignancy at high risk for in-hospital mortality.28 Perhaps such a score could be further developed, validated and implemented to determine utility in curbing in-hospital mortality in these patients.
There are a number of limitations to this analysis. Firstly, we have used broad categories to classify admissions including cardiovascular, infections, renal, etc, and within these categories, there will be multiple different principal diagnoses with differing prognostic impacts. It is, therefore, possible that patients with cancer were admitted with more severe phenotypes within each category and therefore this is what is driving their worse outcomes. Nevertheless, It is beyond the scope of this analysis to study individual principle diagnoses within each category, although there is previous literature showing worse outcomes associated with cancer in sepsis,18 heart failure,26 AMI23 and pneumonia.29 Secondly, the NIS does not capture pharmacology/drugs treatments and therefore differences between the cohorts studied, or whether the patients had received chemotherapy recently which may impact on clinical outcomes, or haematological indices around platelet counts, clotting or white cell counts that are important determinants of outcomes. Thirdly, the NIS does not capture granular data around stage of cancer, or how long patients have been diagnosed with cancer in relation to the index admission, nor whether there are ceilings of care in a place that limit the invasiveness of management. Nevertheless, our analysis suggests that similar patterns of admissions were observed in individual cancers stratified by the presence or absence of metastases. Finally, we did not discriminate between unplanned and planned admissions, nor did we determine where admissions originated (eg, emergency department and outpatient clinic).
In conclusion, in this analysis of 67 million hospitalisations of patients over a 2-year period, we show that 9% of all hospital admissions occur in patients with comorbid cancer. We show that the principal cause of hospitalisation for cancer patients are because of malignancy-related conditions, followed by infection and cardiovascular conditions. We show that different cancers have different patterns of admissions and are associated with different mortality outcomes for each of the common causes for admission. Our work highlights the need to consider the burden of cancer on our hospital services and consider how the prognostic impact of different types of admissions may relate to the type of cancer diagnosis and understand whether these differences relate to disparities in clinical care/treatments.
Supplementary Material
Supplementary Material
FIGURE 1 Cause of admission stratified by cancer type
FIGURE 2 Mortality rates by cancer type and admission cause. GI, Gastrointestinal
FIGURE 3 Adjusted* Mortality odds by type of malignancy. *For all admissions. Reference group: no malignancy. Adjusted to: Age, gender, weekend and elective admission, median zip income, hospital region, bed size, location and teaching status, renal failure, hypothyroidism, obesity, dyslipidaemia, coagulopathy, hypertension, heart failure, chronic ischaemic heart disease, chronic lung disease, smoking status, anaemia, chronic liver disease, diabetes mellitus type 2, thrombocytopenia and dementia
TABLE 1 Patients’ demographics, record characteristics and comorbidities for included hospital records, stratified by malignancy status and cancer type
Malignancy (N = 5 947 883) 8.8%
Haematological n-1 201 500 (20.2%) Lung n-877 110 (14.7%) Colorectal n-543 480 (9.1%) Breast n-378 020 (6.4%) Prostate n-456 330 (7.7%) Other n-2 491 444 (41.9%) No malignancy (N = 61 871 810) 91.2%
Age (y), median (IQR) 67 (53, 77) 69 (62, 77) 66 (56, 76) 63 (53, 73) 71 (63, 80) 65 (55, 74) 59 (38, 74)
Females, % 44% 48.7% 47.1% 98.9% 0% 49.2% 59%
Ethnicity
White 72.2% 77.7% 70.6% 67.2% 69.8% 71.3% 67.1%
Black 12.1% 12.5% 13.3% 17.2% 17.8% 12.3% 15.4%
Hispanic 8.9% 4.5% 9.1% 8.9% 7.1% 9.2% 11.2%
Asian/Pacific Islander 2.7% 2.8% 3.5% 3.1% 2.2% 3.4% 2.7%
Native American 0.4% 0.4% 0.5% 0.4% 0.3% 0.4% 0.6%
Other 3.7% 2.2% 3% 3.2% 2.8% 3.4% 3%
Hospital location
Northeast 21.3% 20.4% 19.2% 21.8% 20.3% 21.2% 18.6%
Midwest 23.1% 24.2% 22.4% 21.3% 23.9% 22.1% 22.2%
South 36.2% 40.6% 39.3% 38% 35.5% 36.6% 39.7%
West 19.4% 14.8% 19.1% 19% 20.3% 20.1% 19.5%
Hospital size
Small 15.1% 16.9% 17.3% 18.2% 17.5% 14.3% 19.8%
Medium 25.1% 28.4% 28.5% 28.9% 28.8% 25% 29.6%
Large 59.7% 54.7% 54.2% 52.9% 53.7% 60.6% 50.6%
Hospital location/teaching status
Rural 6.2% 9% 8.9% 7.6% 7.1% 5.6% 9.6%
Urban non-teaching 18.8% 23.4% 22.6% 22.6% 21.5% 18.1% 25.4%
Teaching 75% 67.6% 68.5% 69.8% 71.4% 76.3% 65.1%
Weekend admission 18.7% 19.6% 14.5% 18% 14.9% 17.4% 20.7%
Elective admission 21.3% 19% 38.6% 26.9% 39.9% 28.1% 23.8%
Median ZIP income
1st quartile 24.9% 30.5% 28.6% 26.4% 25.8% 26.6% 31.1%
2nd quartile 24.8% 26.8% 26% 24.4% 24.4% 24.9% 26%
3rd quartile 25.5% 23.5% 24.2% 24.5% 24.9% 24.8% 23.6%
4th quartile 24.8% 19.3% 21.2% 24.6% 24.9% 23.7% 19.3%
Expected primary payer
Medicare 60.1% 65.8% 54.4% 50.5% 65.6% 54.4% 46.2%
Medicaid 9.4% 10.1% 11.1% 12.9% 4.5% 11.7% 19.3%
Private 26.2% 19.8% 29.8% 32.9% 26.2% 28.9% 27.2%
Uninsured 1.8% 1.6% 2.2% 1.5% 1.1% 2.1% 4.1%
No charge 0.2% 0.1% 0.2% 0.2% 0.1% 0.2% 0.4%
Other 2.3% 2.6% 2.2% 2% 2.6% 2.5% 2.9%
Comorbidities
Ischaemic heart 19.9% 23.9% 15.7% 10.8% 24.3% 16% 18.3%
Heart failure 18.1% 15.4% 10% 11.9% 14.7% 10.3% 15.9%
Hypertension 59.1% 63.1% 57.1% 54% 66% 58.3% 52.1%
Dyslipidaemia 31.2% 36% 28.5% 26.2% 38.9% 29.4% 26.5%
Diabetes 24.3% 23.5% 23.7% 22.6% 24.5% 25.1% 23.1%
Smoking 7.5% 21.7% 10.7% 7.8% 9.1% 11.4% 13.2%
Chronic lung disease 19.3% 54.7% 15.7% 18.4% 16.7% 18.5% 20.8%
Chronic renal failure 22.2% 12.9% 12.3% 10.9% 20.1% 15.3% 8.9%
Obesity 9.6% 7.3% 11.1% 13.1% 9.6% 10.7% 13.6%
Anaemia 54.2% 36.3% 46.1% 35.2% 32.8% 39.3% 22.1%
Thrombocytopenia 17.3% 7.3% 5.1% 7% 6.1% 7.9% 4%
Coagulopathy 2.7% 1.3% 1.8% 1.4% 1.4% 2.5% 1.7%
Dementia 4.8% 4.1% 3.8% 4.1% 7.2% 3.6% 6.2%
Chronic liver disease 4.7% 3.5% 5.6% 4.5% 2.9% 8.7% 3.6%
Hypothyroidism 14.2% 12% 10.3% 15.4% 7.6% 12.6% 10.6%
Note: All P < .001.
TABLE 2 Cause of admission stratified by cancer type
Malignancy (N = 5 947 883) 8.8%
Haematological n-1 201 500 (20.2%) Lung n-877 110 (14.7%) Colorectal n-543 480 (9.1%) Breast n-378 020 (6.4%) Prostate n-456 330 (7.7%) Other n-2 491 444 (41.9%) No malignancy (N = 61 871 810) 91.2%
Infectious 18.4% 16.9% 9.1% 13.3% 11.2% 12.3% 11.3%
Cardiovascular 10.7% 9.7% 4.7% 9.2% 10.6% 7.1% 16%
Malignancy-related 19% 39.4% 57.1% 34.6% 43.9% 42.9% N/A
Respiratory 2.7% 6.5% 0.8% 2.6% 2% 2% 3.7%
Haematological 5.4% 2.1% 1.3% 4.3% 1.7% 1.9% 1%
Gastrointestinal 5.4% 3.9% 8.8% 6.9% 4.8% 8.5% 9.5%
Renal 3.2% 2.1% 3.1% 3% 4.5% 3.6% 2.9%
Rheumatology 3.2% 1.6% 0.6% 3.7% 3% 1.6% 7.9%
Endocrine 1.4% 1.5% 0.7% 1.4% 1% 1.3% 3.1%
Neurology 1.7% 2.4% 1.5% 2.8% 2.3% 2.7% 2.5%
Note: All P < .001.
TABLE 3 Mortality rates by cancer type and admission cause
Haematological malignancy Lung malignancy Colorectal malignancy Breast malignancy Prostate malignancy All malignancies No malignancy
Malignancy-related admission 7.4% 7.8% 3.1% 4.3% 1.5% 4.9% N/A
Infectious admission 10.4% 15.2% 12.8% 8.4% 7.3% 12.6% 5.4%
Cardiovascular admission 5.2% 7.9% 5.1% 3.6% 3.7% 5.5% 3.3%
Gastrointestinal admission 3.1% 5.5% 3% 2.7% 2.3% 3.8% 1.2%
Renal admission 4% 3.3% 4.3% 4.2% 3.2% 4.6% 1.5%
Any admission 5.3% 8.7% 4.2% 4% 2.8% 5.7% 1.9%
Note: All P < .001.
TABLE 4 Adjusteda Mortality odds by type of malignancy and admission cause
Malignancy OR (95% CI) P
All admissions
Haematological 1.79 (1.76–1.82) <.001
Lung 3.65 (3.59–3.71) <.001
Colorectal 1.71 (1.66–1.76) <.001
Prostate 0.89 (0.86–0.92) <.001
Breast 2.06 (1.99–2.13) <.001
Infectious admissions
Haematological 1.47 (1.45–1.49) <.001
Lung 2.89 (2.84–2.93) <.001
Colorectal 1.91 (1.86–1.96) <.001
Prostate 0.89 (0.86–0.92) <.001
Breast 1.6 (1.55–1.65) <.001
Cardiovascular admissions
Haematological 1.17 (1.14–1.2) <.001
Lung 1.97 (1.92–2.02) <.001
Colorectal 1.13 (1.07–1.2) <.001
Prostate 0.86 (0.81–0.9) <.001
Breast 0.88 (0.82–0.93) <.001
GI admissions
Haematological 1.64 (1.56–1.72) <.001
Lung 3.71 (3.53–3.9) <.001
Colorectal 2.01 (1.9–2.13) <.001
Prostate 1.01 (0.91–1.1) .935
Breast 2.06 (1.9–2.24) <.001
Renal admissions
Haematological 1.7 (1.61–1.79) <.001
Lung 3.5 (3.29–3.73) <.001
Colorectal 2.44 (2.26–2.64) <.001
Prostate 1.18 (1.08–1.28) <.001
Breast 2.79 (2.54–3.1) <.001
a Reference group: no malignancy. Adjusted to: Age, gender, weekend and elective admission, median zip income, hospital region, bed size, location and teaching status, renal failure, hypothyroidism, obesity, dyslipidaemia, coagulopathy, hypertension, heart failure, chronic ischaemic heart disease, chronic lung disease, smoking status, anaemia, chronic liver disease, diabetes mellitus type 2, thrombocytopenia and dementia.
What’s known
Cancer is a leading cause of mortality in the United States.
Approximately 60% of patients with cancer who present to emergency departments are admitted as inpatients.
Most common presentations include both cancer-related and non-cancer-related causes.
What’s new
Nine per cent of all hospital admissions occur in patients with comorbid cancer.
The principal cause of hospitalisation for cancer patients is because of malignancy-related conditions, followed by infection and cardiovascular conditions.
Different cancers have different patterns of admissions and are associated with different mortality outcomes for each of the common causes for admission.
DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from HCUP. Restrictions apply to the availability of these data, which were used under license for this study.
SUPPORTING INFORMATION
Additional supporting information may be found in the online version of the article at the publisher’s website.
REFERENCES
1. Heron M , Anderson RN . Changes in the leading cause of death: recent patterns in heart disease and cancer mortality. NCHS Data Brief. 2016;254 :1–8.
2. Siegel RL , Miller KD , Jemal A . Cancer statistics, 2019. CA Cancer J Clin. 2019;69 (1 ):7–34.30620402
3. Caterino JM , Adler D , Durham DD , Analysis of diagnoses, symptoms, medications, and admissions among patients with cancer presenting to emergency departments. JAMA Netw Open. 2019;2 (3 ):e190979.
4. Numico G , Cristofano A , Mozzicafreddo A , Hospital admission of cancer patients: avoidable practice or necessary care? PLoS One. 2015;10 (3 ):e0120827.
5. Sadik M , Ozlem K , Huseyin M , AliAyberk B , Ahmet S , Ozgur O . Attributes of cancer patients admitted to the emergency department in one year. World J Emerg Med. 2014;5 (2 ):85–90.25215154
6. Rivera DR , Gallicchio L , Brown J , Liu B , Kyriacou DN , Shelburne N . Trends in adult cancer-related emergency department utilization: an analysis of data from the nationwide emergency department sample. JAMA Oncol. 2017;3 (10 ):e172450.
7. Panattoni L , Fedorenko C , Greenwood-Hickman MA , Characterizing potentially preventable cancer- and chronic disease-related emergency department use in the year after treatment initiation: a regional study. J Oncol Pract. 2018;14 (3 ):e176–e185.29452549
8. Hsu J , Donnelly JP , Moore JX , Meneses K , Williams G , Wang HE . National characteristics of Emergency Department visits by patients with cancer in the United States. Am J Emerg Med. 2018;36 (11 ):2038–2043.29573899
9. Ambrus JL , Ambrus CM , Mink IB , Pickren JW . Causes of death in cancer patients. J Med. 1975;6 (1 ):61–64.1056415
10. Lin L , Yan L , Liu Y , Yuan F , Li H , Ni J . Incidence and death in 29 cancer groups in 2017 and trend analysis from 1990 to 2017 from the Global Burden of Disease Study. J Hematol Oncol. 2019;12 (1 ):96.31511035
11. Oh CM , Lee D , Kong HJ , Causes of death among cancer patients in the era of cancer survivorship in Korea: Attention to the suicide and cardiovascular mortality. Cancer Med. 2020;9 (5 ):1741–1752.31960609
12. Prakash G , Bakhshi S , Raina V , Characteristics and pattern of mortality in cancer patients at a tertiary care oncology center: report of 259 cases. Asian Pac J Cancer Prev. 2010;11 (6 ):1755–1759.21338228
13. Zaorsky NG , Churilla TM , Egleston BL , Causes of death among cancer patients. Ann Oncol. 2017;28 (2 ):400–407.27831506
14. Finlayson EV , Goodney PP , Birkmeyer JD . Hospital volume and operative mortality in cancer surgery: a national study. Arch Surg. 2003;138 (7 ):721–725; discussion 6.12860752
15. Hill JS , Zhou Z , Simons JP , A simple risk score to predict in-hospital mortality after pancreatic resection for cancer. Ann Surg Oncol. 2010;17 (7 ):1802–1807.20155401
16. Taccone FS , Artigas AA , Sprung CL , Moreno R , Sakr Y , Vincent JL . Characteristics and outcomes of cancer patients in European ICUs. Crit Care. 2009;13 (1 ):R15.19200368
17. Wong SL , Revels SL , Yin H , Variation in hospital mortality rates with inpatient cancer surgery. Ann Surg. 2015;261 (4 ):632–636.24743604
18. Hensley MK , Donnelly JP , Carlton EF , Prescott HC . Epidemiology and outcomes of cancer-related versus non-cancer-related sepsis hospitalizations. Crit Care Med. 2019;47 (10 ):1310–1316.31356477
19. Kwok CS , Wong CW , Ravindran R , Location of death among patients presenting with cardiovascular disease to the emergency department in the United states. Int J Clin Pract. 2021;75 (4 ):e13798.
20. Strongman H , Gadd S , Matthews A , Medium and long-term risks of specific cardiovascular diseases in survivors of 20 adult cancers: a population-based cohort study using multiple linked UK electronic health records databases. Lancet. 2019;394 (10203 ):1041–1054.31443926
21. Sturgeon KM , Deng L , Bluethmann SM , A population-based study of cardiovascular disease mortality risk in US cancer patients. Eur Heart J. 2019;40 (48 ):3889–3897.31761945
22. Lenihan DJ , Cardinale DM . Late cardiac effects of cancer treatment. J Clin Oncol. 2012;30 (30 ):3657–3664.23008297
23. Bharadwaj A , Potts J , Mohamed MO , Acute myocardial infarction treatments and outcomes in 6.5 million patients with a current or historical diagnosis of cancer in the USA. Eur Heart J. 2020;41 (23 ):2183–2193.31800032
24. Mohamed MO , Van Spall HGC , Kontopantelis E , Effect of primary percutaneous coronary intervention on in-hospital outcomes among active cancer patients presenting with ST-elevation myocardial infarction: a propensity score matching analysis. Eur Heart J Acute Cardiovasc Care. 2021. zuaa032. 10.1093/ehjacc/zuaa032
25. Mohamed MO , Lopez-Mattei JC , Parwani P , Management strategies and clinical outcomes of acute myocardial infarction in leukaemia patients: nationwide insights from United States hospitalisations. Int J Clin Pract. 2020;74 (5 ):e13476.
26. Kaneko H , Itoh H , Yotsumoto H , Association of cancer with outcomes in patients hospitalized for heart failure. Circ J. 2020;84 (10 ):1771–1778.32921679
27. Mohamed MO , Barac A , Contractor T , Prevalence and in-hospital outcomes of patients with malignancies undergoing de novo cardiac electronic device implantation in the USA. Europace. 2020;22 (7 ):1083–1096.32361739
28. Ramchandran K , Shega J , Schumacher M , A predictive model to identify cancer patients at risk for in-hospital mortality based on admission criteria. J Clin Oncol. 2011;29 (15 _suppl):6045.
29. Hespanhol V , Barbara C . Pneumonia mortality, comorbidities matter? Pulmonology. 2020;26 (3 ):123–129.31787563
|
PMC008xxxxxx/PMC8983060.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
7609829
6055
Neuropathol Appl Neurobiol
Neuropathol Appl Neurobiol
Neuropathology and applied neurobiology
0305-1846
1365-2990
34388852
8983060
10.1111/nan.12758
NIHMS1790382
Article
A systems-level analysis highlights microglial activation as a modifying factor in common epilepsies
Altmann Andre 1†
Ryten Mina 2†
Di Nunzio Martina 3†
Ravizza Teresa 3
Tolomeo Daniele 3
Reynolds Regina H 2
Somani Alyma 4
Bacigaluppi Marco 5
Iori Valentina 3
Micotti Edoardo 3
Di Sapia Rossella 3
Cerovic Milica 3
Palma Eleonora 6
Ruffolo Gabriele 6
Botía Juan A. 27
Absil Julie 8
Alhusaini Saud 910
Alvim Marina K. M. 11
Auvinen Pia 1213
Bargallo Nuria 1415
Bartolini Emanuele 1617
Bender Benjamin 18
Bergo Felipe P. G. 11
Bernardes Tauana 11
Bernasconi Andrea 19
Bernasconi Neda 19
Bernhardt Boris C. 1920
Blackmon Karen 2122
Braga Barbara 11
Caligiuri Maria Eugenia 23
Calvo Anna 14
Carlson Chad 2124
Carr Sarah J. 25
Cavalleri Gianpiero L. 926
Cendes Fernando 11
Chen Jian 27
Chen Shuai 2829
Cherubini Andrea 23
Concha Luis 30
David Philippe 8
Delanty Norman 92631
Depondt Chantal 32
Devinsky Orrin 21
Doherty Colin P. 2633
Domin Martin 34
Focke Niels K. 3536
Foley Sonya 37
Franca Wendy 11
Gambardella Antonio 2338
Guerrini Renzo 1617
Hamandi Khalid 3940
Hibar Derrek P. 41
Isaev Dmitry 41
Jackson Graeme D. 4243
Jahanshad Neda 41
Kalviainen Reetta 1213
Keller Simon S. 44
Kochunov Peter 45
Kotikalapudi Raviteja 1835
Kowalczyk Magdalena A. 42
Kuzniecky Ruben 46
Kwan Patrick 47
Labate Angelo 2338
Langner Soenke 34
Lenge Matteo 16
Liu Min 19
Martin Pascal 35
Mascalchi Mario 4849
Meletti Stefano 50
Morita-Sherman Marcia E. 11
O’Brien Terence J. 4751
Pariente Jose C. 14
Richardson Mark P. 2552
Rodriguez-Cruces Raul 30
Rummel Christian 53
Saavalainen Taavi 1354
Semmelroch Mira K. 42
Severino Mariasavina 55
Striano Pasquale 56
Thesen Thomas 2122
Thomas Rhys H. 3940
Tondelli Manuela 50
Tortora Domenico 55
Vaudano Anna Elisabetta 50
Vivash Lucy 4757
von Podewils Felix 58
Wagner Jan 59
Weber Bernd 6061
Wiest Roland 53
Yasuda Clarissa L. 11
Zhang Guohao 62
Zhang Junsong 2829
ENIGMA-Epilepsy Working Group
Leu Costin 636465
Avbersek Andreja 65
EpiPGX Consortium
Thom Maria 465
Whelan Christopher D 941
Thompson Paul 41
McDonald Carrie R 6667
Vezzani Annamaria 3*
Sisodiya Sanjay M 6568*
1 Centre for Medical Image Computing, University College London, London, UK.
2 Department of Neurodegenerative Disease, UCL Queen Square Institute of Neurology, London, UK.
3 Department of Neuroscience, Istituto di Ricerche Farmacologiche Mario Negri IRCCS, Milano, Italy.
4 Division of Neuropathology, UCL Queen Square Institute of Neurology, London, UK.
5 Department of Neurology, San Raffaele Scientific Institute and Vita Salute San Raffaele University, Milan, Italy.
6 Department of Physiology and Pharmacology, University of Rome, Sapienza
7 Departamento de Ingeniería de la Información y las Comunicaciones. Universidad de Murcia, Murcia, Spain.
8 Department of Radiology, Hôpital Erasme, Universite Libre de Bruxelles, Brussels 1070, Belgium.
9 Department of Molecular and Cellular Therapeutics, Royal College of Surgeons in Ireland, Dublin, Ireland.
10 Department of Neurology and Neurosurgery, Montreal Neurological Institute, McGill University, Montreal, Quebec, Canada.
11 Department of Neurology, University of Campinas, Campinas, Brazil.
12 Epilepsy Center, Department of Neurology, Kuopio University, Kuopio, Finland.
13 Institute of Clinical Medicine, Neurology, University of Eastern Finland, Kuopio, Finland.
14 Magnetic Resonance Image Core Facility, IDIBAPS, Barcelona, Spain.
15 Centre de Diagnostic Per la Imatge (CDIC), Hospital Clinic, Barcelona, Spain.
16 Pediatric Neurology Unit, Children’s Hospital A. Meyer-University of Florence, Italy.
17 IRCCS Stella Maris Foundation, Pisa, Italy.
18 Department of Diagnostic and Interventional Neuroradiology, University of Tübingen, Tübingen, Germany.
19 Neuroimaging of Epilepsy Laboratory, Montreal Neurological Institute and Hospital, McGill University, Montreal, Quebec, Canada.
20 Multimodal Imaging and Connectome Analysis Lab, Montreal Neurological Institute and Hospital, McGill University, Montreal, Quebec, Canada.
21 Comprehensive Epilepsy Center, Department of Neurology, New York University School of Medicine, New York, USA.
22 Department of Physiology, Neuroscience and Behavioral Science, St. George’s University, Grenada, West Indies.
23 Institute of Molecular Bioimaging and Physiology of the National Research Council (IBFM-CNR), Catanzaro, Italy.
24 Medical College of Wisconsin, Department of Neurology, Milwaukee, WI, USA.
25 Department of Basic and Clinical Neuroscience, Institute of Psychiatry, Psychology and Neuroscience, King’s College London, UK.
26 FutureNeuro Research Centre, RCSI, Dublin, Ireland.
27 Department of Computer Science and Engineering, The Ohio State University, USA.
28 Cognitive Science Department, Xiamen University, Xiamen, China.
29 Fujian Key Laboratory of the Brain-like Intelligent Systems, China.
30 Instituto de Neurobiología, Universidad Nacional Autónoma de México. Querétaro, Querétaro, México.
31 Division of Neurology, Beaumont Hospital, Dublin 9, Ireland.
32 Department of Neurology, Hôpital Erasme, Universite Libre de Bruxelles, Brussels 1070, Belgium.
33 Neurology Department, St. James’s Hospital, Dublin 8, Ireland.
34 Functional Imaging Unit, Institute of Diagnostic Radiology and Neuroradiology, University Medicine Greifswald, Greifswald, Germany.
35 Department of Neurology and Epileptology, Hertie Institute for Clinical Brain Research, University of Tübingen, Tübingen, Germany.
36 Department of Clinical Neurophysiology, University Medicine Göttingen, Göttingen, Germany.
37 Cardiff University Brain Research Imaging Centre, School of Psychology, Wales, UK.
38 Institute of Neurology, University ‚ “Magna Græcia”, Catanzaro, Italy.
39 Institute of Psychological Medicine and Clinical Neurosciences, Hadyn Ellis Building, Maindy Road, Cardiff, UK.
40 Department of Neurology, University Hospital of Wales, Cardiff, UK.
41 Imaging Genetics Center, Mark and Mary Stevens Neuroimaging and Informatics Institute, University of Southern California, Los Angeles, California, USA.
42 The Florey Institute of Neuroscience and Mental Health, Austin Campus, Melbourne, VIC, Australia.
43 Florey Department of Neuroscience and Mental Health, The University of Melbourne, Melbourne, VIC, Australia.
44 Department of Molecular and Clinical Pharmacology, Institute of Translational Medicine, University of Liverpool, UK.
45 Maryland Psychiatric Research Center, Department of Psychiatry, University of Maryland School of Medicine, Maryland, USA.
46 Department of Neurology, Zucker Hofstra School of Medicine, New York, NY 10075, USA.
47 Department of Neurology, Royal Melbourne Hospital, Parkville, 3050, Australia.
48 Neuroradiology Unit, Children’s Hospital A. Meyer, Florence, Italy.
49 “Mario Serio” Department of Experimental and Clinical Biomedical Sciences, University of Florence, Italy.
50 Department of Biomedical, Metabolic, and Neural Science, University of Modena and Reggio Emilia, NOCSE Hospital, Modena, Italy.
51 Department of Medicine, University of Melbourne, Parkville, VIC, 3052, Australia.
52 Department of Neurology, King’s College Hospital, London, UK.
53 Support Center for Advanced Neuroimaging (SCAN), University Institute for Diagnostic and Interventional Neuroradiology, Inselspital, University of Bern, Bern, Switzerland.
54 Central Finland Central Hospital, Medical Imaging Unit, Jyväskylä, Finland.
55 Neuroradiology Unit, Department of Head and Neck and Neurosciences, Istituto Giannina Gaslini, Genova, Italy.
56 Pediatric Neurology and Muscular Diseases Unit, Department of Neurosciences, Rehabilitation, Ophthalmology, Genetics, Maternal and Child Health, University of Genoa, Genova, Italy.
57 Melbourne Brain Centre, Department of Medicine, University of Melbourne, Parkville, VIC, 3052, Australia.
58 Department of Neurology, University Medicine Greifswald, Greifswald, Germany.
59 Department of Neurology, University of Ulm and Universitäts- and Rehabilitationskliniken Ulm, Germany.
60 Department of Epileptology, University Hospital Bonn, Bonn, Germany.
61 Department of Neurocognition / Imaging, Life & Brain Research Centre, Bonn, Germany.
62 Department of Computer Science and Electrical Engineering, University of Maryland, Baltimore County, USA.
63 Genomic Medicine Institute, Lerner Research Institute, Cleveland Clinic, Cleveland, OH, USA.
64 Stanley Center for Psychiatric Research, Broad Institute of Harvard and MIT, Cambridge, MA, USA.
65 Department of Clinical and Experimental Epilepsy, UCL Queen Square Institute of Neurology, London, UK.
66 Multimodal Imaging Laboratory, University of California San Diego, San Diego, California, USA.
67 Department of Psychiatry, University of California San Diego, San Diego, California, USA.
68 Chalfont Centre for Epilepsy, Bucks, UK.
† Contributed equally
Author contributions
A. Altmann, M.R., A.V., S.M.S. conceived and designed the research and wrote the paper. Authors J.A. to J.Z. and C.D.W., P.T., C.R.M., S.M.S. contributed to generation and analysis of human neuroimaging data as part of ENIGMA-Epilepsy working group, which is co-ordinated by C.R.M. and S.M.S., with overall leadership of the ENIGMA consortia by P.T. A. Altmann analysed AIBS data. C.L. and A. Avbersek and EpiPGX generated and analysed the human drug resistance GWAS data. J.A.B. generated co-expression network data. R.H.R. performed EWCE analyses. A.S. and M.T. generated and analysed the human post mortem data. M.D.N., T.R., D.T., M.B., V.I., E.M., and A.V. generated and analysed the murine experimental model data. All authors contributed to the editing of the manuscript.
* To whom correspondence may be addressed
27 3 2022
2 2022
05 9 2021
01 2 2023
48 1 e12758e12758
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Aims:
The causes of distinct patterns of reduced cortical thickness in the common human epilepsies, detectable on neuroimaging and with important clinical consequences, are unknown. We investigated the underlying mechanisms of cortical thinning using a systems-level analysis.
Methods:
Imaging-based cortical structural maps from a large-scale epilepsy neuroimaging study were overlaid with highly spatially-resolved human brain gene expression data from the Allen Human Brain Atlas. Cell-type deconvolution, differential expression analysis and cell-type enrichment analyses were used to identify differences in cell-type distribution. These differences were followed up in post-mortem brain tissue from humans with epilepsy using Iba1 immunolabelling. Furthermore, to investigate a causal effect in cortical thinning, cell-type specific depletion was used in a murine model of acquired epilepsy.
Results:
We identified elevated fractions of microglia and endothelial cells in regions of reduced cortical thickness. Differentially-expressed genes showed enrichment for microglial markers, and in particular, activated microglial states. Analysis of post-mortem brain tissue from humans with epilepsy confirmed excess activated microglia. In the murine model, transient depletion of activated microglia during the early phase of the disease development prevented cortical thinning and neuronal cell loss in the temporal cortex. Although the development of chronic seizures was unaffected, the epileptic mice with early depletion of activated microglia did not develop deficits in a non-spatial memory test seen in epileptic mice not depleted of microglia.
Conclusions:
These convergent data strongly implicate activated microglia in cortical thinning, representing a new dimension for concern and disease modification in the epilepsies, potentially distinct from seizure control.
MRI
cortical thinning
gene expression
post mortem
pmcIntroduction
Significant progress is being made in understanding disease processes in the epilepsies. Many genetic variants causing or associated with rare and common epilepsies have been reported [1,2], with discovery continuing apace [3]. Numerous direct structural causes of epilepsy have been revealed by brain MRI. Several structural abnormalities are now themselves known to have a genetic basis [4]. As a result, the proportion of causally-explicable epilepsies is growing rapidly. Conversely, the mechanisms whereby these identified causes promote epileptogenesis and seizures remain obscure for most human epilepsies. Moreover, beyond causation and epileptogenesis, the epilepsies involve many other processes: some lead to clinically-apparent consequences, such as developmental delay or memory dysfunction, whereas others, without necessarily obvious symptoms, may be detected only on investigation - cerebellar atrophy is one example. The natural history of any given epilepsy need not be a single linear dynamic from causation to a unique, predictable, final outcome: for example, the epilepsies are associated with shortened longevity (even if seizures stop) [5] and increased risk of particular comorbidities [6]. Known causes per se may not explain all the observed outcomes, suggesting that many epilepsies could be conceptualised as intricate matrices of causation, processes and outcomes [7], with complex inter-dependencies, such as a likely link between reduction in cortical thickness and disease duration [8].
Through the ENIGMA-Epilepsy consortium, we recently showed that across a wide range of common human epilepsies, which are known to have both distinct and shared genetic architecture [2,3,9,10], there are also shared, pan-syndrome, and distinct, syndrome-specific, regional patterns of altered cortical thickness and altered subcortical grey matter volumes [8]. The causes of the structural changes in these epilepsies are not known. The findings suggest structural losses may reflect an initial insult, subsequent epileptogenesis or progressive neurodegeneration, or some combination, and show robustly that the common epilepsies cannot necessarily be considered entirely benign at the structural level. We sought to identify processes underlying the structural findings.
The pathophysiology of neurological disease has been successfully revealed using powerful combinations of brain MRI findings, regionalised brain-specific gene expression and gene co-expression networks, in a systems biology framework to implicate candidate genes [11–14]. Here, we used the findings from the in vivo ENIGMA-Epilepsy imaging study [8] in combination with the post mortem atlas of gene expression in the brain from healthy subjects, curated by the Allen Institute [15,16] to direct interrogation of regionalised brain cell-type composition and generic biological processes which might underlie thickness or volume reductions across the studied epilepsy syndromes. We hypothesised that this approach could suggest disease mechanisms causing the observed structural changes. We further explored the findings with a series of additional experiments in both human and animal tissues. We demonstrated experimentally in a murine model of acquired epilepsy that depletion of the implicated cell-type, microglia, can successfully avert cortical thinning and the concomitant neuronal cell loss and cognitive deficit, without modifying spontaneous seizures. Microglia have been shown to have various roles in a few rare, severe human epilepsies. Our new results implicate microglia in the widespread, but largely unstudied, reduction in cortical thickness that accompanies the numerically far more important common human epilepsies and point to the potential for prevention of such thinning by manipulation of microglia.
Materials and Methods
In order to explore mechanisms underlying cortical thinning in the epilepsies, we designed the study as shown in Fig. 1. We obtained statistical maps from a large structural MRI study comparing people with epilepsy to healthy controls conducted by the ENIGMA-Epilepsy consortium [8] (Table S1). To determine cell-type composition differences that spatially correlate with the reported structural changes on MRI, we used a healthy control dataset, the Allen Human Brain Atlas (AHBA), comprising densely-sampled gene expression across the cortex [15,16]. The statistical maps were mapped onto the AHBA using MNI-coordinates. This enabled us to investigate how the spatial distribution of gene expression in the healthy brain correlates with regional vulnerability in the diseased brain. Our primary hypothesis was that cell-types causing structural changes observed on human brain MRI could be identified using the design shown in Fig. 1A. Briefly, cortical brain regions were classified as vulnerable to, or relatively protected from, cortical thinning based on the previous imaging results [8]. Then two complementary approaches were used to identify cell-types with association to reduced cortical thickness: (i) AHBA microarray expression data were de-convolved into cell-type fractions and these fractions were tested for differences between brain regions; (ii) a standard differential expression analysis was carried out for each gene expression probe and cell-types were identified using enrichment analyses (for details on pre-processing, atlas mapping, quality control and the statistical approach see Supplementary Methods, section A). The analysis code for the gene expression analyses is available (https://github.com/andrealtmann/AHBA_Epilepsy).
We hypothesised that cell-types identified from the in silico analysis of data on the spatial distribution of gene expression in the brain from non-epileptic donors would also show compatible differences in spatial distribution in tissue samples obtained from people with epilepsy. Therefore, immunostaining of microglia in post mortem human brain tissue was carried out to substantiate the cell-type findings from the in silico work. The human post mortem cases were classified into different epilepsy groups and control groups according to clinical and pathology criteria. The final sample size was based on availability of tissues for certain epilepsies (as characterised by clinical data including EEG and MRI) and whether regional tissue paraffin-embedded block samples were available in each case from multiple brain regions (see Supplementary Methods, section C). Briefly, the labelling index (LI) sums overall microglia presence in terms of cell bodies as well as microglial processes (as cell bodies may not be present in every section, but processes are likely to be important in microglial roles), overcoming issues of microglial clustering, overlap and aggregation around vessels that can confound individual cell counts. LI also covers all microglial cell-types regardless of morphology, size and shape. LI was measured across 14 ROIs in post mortem brain tissue derived from 55 individuals, comprising individuals with non-lesional epilepsy (EP-NL, n=18), lesional epilepsy (EP-L, n=21) and non-epilepsy controls (NEC, n=16). No sample size calculations were performed since this was not a discovery sample, but a previous similar study of regional cortical pathology achieved a significant statistical result compared to controls with nine post mortem cases, whereas this study included 55 individuals [17]. Staining was repeated on the same cases in different immunohistochemistry runs for comparison.
Further, in order to assess whether cortical thinning can be prevented by manipulating microglia, we used a well-established murine model of epilepsy induced by convulsive status epilepticus (SE) provoked by intra-amygdalar injection of kainic acid in C57/BL6N adult male mice (30 g; Charles River, Italy)[18–21]. This model mimics features of MTLE with hippocampal sclerosis, with neuronal damage also observed in extrahippocampal areas [19–21]. To explore the importance of microglia in brain structural changes and pathologic outcomes, mice were treated with the CSF1R inhibitor PLX3397 in a controlled experiment. We assessed the effect of PLX3397 treatment on cognitive performance using the novel object recognition test (NORT), regional brain volume estimates from post mortem MRI, abundance of microglia and neurons quantified using immunohistochemistry and Nissl staining, respectively. The sample size was determined a priori based on previous experience with the same epilepsy model [18,20]. A simple random allocation was applied to assign a subject to a particular experimental group. Group 1: medicated diet (supplemented with PLX3397), n=5 and placebo (non-medicated) diet, n=5; Group 2: medicated diet (supplemented with PLX3397), n=6 and placebo (non-medicated) diet, n=5 (experimental design in Fig. S1A). Respective sham mice (n=8) were prepared for behavioural testing and post mortem analyses. These groups were run 15 days apart. Group 3: placebo diet mice exposed to SE, n=8 (run in parallel with group 1, n=4 and group 2, n=4) and respective sham mice, n=8. Group 3 mice were prepared to compensate for potential drop-outs during the longitudinal experiment (Fig. 3C). Limited PLX3397-supplemented diet availability prevented further experiments. To verify the reproducibility of the experimental findings, we compared several EEG measures between the different placebo-diet mouse cohorts, namely SE onset and duration, temporal distribution of spikes during SE, number and duration of spontaneous seizures. For all these measures, statistical analysis showed that replication was successful. The data were collected by anonymising samples and the assessor was blinded to diagnosis (for both human tissue and experimental mouse model data). For detailed Materials and Methods and mouse treatment protocols see Supplementary Methods (section D).
Results
Cortical regions at most risk of reduced thickness are characterised by higher density of microglia and endothelial cells
The AHBA provided gene expression profiles for all cortical regions of interest [15,16]. This gene expression atlas is unique in that from 158 to 348 different brain structures have been sampled from each control individual (n=6), none of whom had epilepsy, covering in total 414 different structures, and that sampling has been precisely mapped to MNI space, enabling linkage of the expression data to external MRI-brain maps in MNI space (see Supplementary Methods, section A). Thus, the gene expression profiles used in this analysis were highly regionally specified. However, only two out of the six brains have been sampled in both hemispheres: therefore, we restricted our analysis to the left hemisphere, which was available for all subjects [22].
We focussed on the regions where thickness was reduced and sought to identify mechanisms that underlie this cortical regional ‘vulnerability’ to damage; areas without significant loss of thickness compared to controls were considered ‘relatively protected’, remaining like normal cortex. To identify the molecular basis of this regional vulnerability in the broad spectrum of epilepsies, we focused our analyses on the shared pan-syndrome MRI findings [8]. Of the 34 cortical regions in the left hemisphere profiled within the ENIGMA-Epilepsy data set, eight were considered vulnerable to reduced cortical thickness, while 26 were considered relatively protected [8] (Table S1).
Using linear mixed effects models, we compared these two types of cortical regions for significant differences in cell-type fractions, inferred from gene expression using Scaden [23]. We identified increased ratios of endothelial cells (T=4.4; P=1×10−5) and microglia (T=4.1; P=4.5×10−5) in vulnerable regions, along with a decreased ratio of excitatory neurons (T=−3.9; P=1.2×10−4) in vulnerable regions (Table 1). We confirmed these cell-type enrichments by a complementary two-step approach. The first step was a gene-wise association screen, showing that, out of 14,138 tested genes, 3,182 genes were more highly expressed in vulnerable regions (at PFDR<0.05) and 2,223 genes were expressed significantly less in vulnerable regions (at PFDR<0.05; Dataset S1). In the second step, these gene-wise results were subject to threshold-free (i.e., not relying on lists of significantly differentially expressed genes) gene set enrichment analyses using cell-type-specific gene sets from different data sources (Dataset S2). This analysis confirmed the enrichment of endothelial cells and microglia in vulnerable regions, and neurons in protected regions. We noted that both analyses provided qualitatively the same result when using all 1,628 AHBA samples from both hemispheres, where 18 of 68 ROIs were considered vulnerable (Table S2; Figure S15). Thus, regions vulnerable to cortical thinning were characterized by elevated proportion of microglia and endothelial cells.
Selected processes and microglial signatures are implicated amongst genes associated with reduced cortical thickness
Using our gene-level association results, we next investigated the biological processes that could underpin regional differences in vulnerability to reduced cortical thickness in epilepsy. In the first instance, we investigated enrichment of differentially expressed genes in Gene Ontology (GO) terms and KEGG and REACTOME pathways using a threshold-free approach based on the area under the receiver operator characteristics curve (AUC) [24]. Amongst genes with higher expression in relatively protected cortical regions, the most significant terms we identified related to RNA processing (REACTOME: RNA polymerase II transcription, AUC=0.429, PFDR=1.19×10−11) and synaptic function (GO: postsynaptic specialization organization, AUC=0.270, PFDR=3.95×10−4; Dataset S3). Conversely, amongst genes with higher expression in vulnerable cortical regions, the most significant terms related to electron transport (GO: electron transport chain, AUC=0.74, PFDR=3.45×10−18) and immune function and regulation (REACTOME: innate immune system, AUC=0.59, PFDR=1.46×10−15; GO: antigen processing and presentation, AUC=0.65, PFDR=7.98×10−13; Dataset S3), the latter being consistent with our results obtained using cell-specific gene sets. There was also strong enrichment for the KEGG pathways related to neurodegenerative diseases (Alzheimer’s disease: AUC=0.65 PFDR=3.53×10−6; Parkinson’s disease: AUC=0.75, PFDR=4.55×10−12; Huntington’s disease: AUC=0.66, PFDR=1.80×10−7). Moreover, this approach also enabled us to obtain more specific process-related information and suggested the importance of the interferon gamma signalling pathway (GO: Response to interferon gamma, AUC=0.597, PFDR=1.42×10−3; GO: interferon gamma mediated signalling pathway, AUC=0.621, PFDR=6.71×10−3; Dataset S3).
A number of cell-types showed enrichment in vulnerable regions. But both cell-type analysis and the observed pathway enrichments implicated microglia and immune processes, respectively, prompting us to investigate microglia in more detail, given existing evidence for a role for microglia in epilepsy [25] and in neurodegeneration in general [26]. More precisely, since microglia can exist in a range of activation states within the context of epilepsy [25], we sought to identify the microglial cell states of greatest importance in reduced cortical thickness. We collated gene signatures for distinct microglial states from the existing literature and also inferred signatures of microglial state through co-expression network analyses (see Supplementary Methods, section A). While there were significant overlaps in gene membership across the 16 microglial signatures used (Fig. S2), each of the gene lists was distinct. We identified a significant enrichment for 13 of the 16 signatures with genes overexpressed in vulnerable cortex. This included the microglial signature generated by Srivastava et al. [27] (AUC=0.68, PFDR=1.45×10−12; Dataset S4), which was positively correlated with seizure frequency in a particular mouse model of chronic epilepsy. Strong enrichments were also identified for an inferred human microglial signature enriched for type 1-like microglial markers (AUC=0.73, PFDR=7.27×10−18; grey60; Dataset S4), as well as signatures for aged, late activation, and de-activated microglia. However, we saw little evidence for enrichment within signatures of early activation (“Early Response” signature [28], AUC=0.53, PFDR=0.062). Similar microglial states have recently been implicated in various forms of chronic neurodegeneration [28–30], but more data are needed to definitively determine whether similar underlying processes and microglial states are indeed involved.
Genetic evidence supports immune activation as a modifying, but not causal, factor in the common epilepsies
In the ENIGMA-Epilepsy imaging study, region-specific reduced cortical thickness across the common epilepsies was correlated with disease duration and age of onset of epilepsy [8]. Given that the analyses of gene expression data from healthy donors described above suggest the importance of microglia responses in vulnerability to reduced cortical thickness, we hypothesized that genetic variants affecting microglial responses would also impact upon the severity of epilepsy. While no microglial eQTL data set exists to date, eQTL analyses have been performed using monocytes at rest, and also monocytes treated with IFN-γ or LPS [31]. We postulated that these eQTLs would be enriched for heritability of risk loci for epilepsy that is drug-resistant (one surrogate for disease severity) but not for epilepsy per se. To investigate this, we used GWAS data on epilepsy susceptibility [2] and separate GWAS data on the phenotype of drug-resistant epilepsy from the EpiPGX consortium (www.epipgx.eu). Using LD score regression, we sought enrichment in heritability for these phenotypes within state-specific monocyte eQTLs [31] (Supplementary Methods, section B). We found no significant enrichment in the heritability of epilepsy in this form of annotation, suggesting that susceptibility to common epilepsies is less likely to be driven by immune processes (Fig. 1B, Table 2). However, eQTLs regulating the response to IFN-γ were highly enriched for drug-resistant epilepsy vs drug-responsive epilepsy loci (P=0.00045; PFDR=0.0095, Table 2). This result was particularly striking given the small size of the EpiPGX drug-resistant epilepsy GWAS (2423 drug-resistant cases vs 1626 drug-responsive cases). Thus, we provide genetic evidence in support of microglial-mediated responses as a modifying factor in severity of epilepsy, but not its susceptibility.
Widespread regionalised over-representation of microglia is present in brain tissue from people with epilepsy
Based on the analyses of regional gene expression patterns in healthy brains above, we hypothesised that brain tissue from individuals with various forms of epilepsy would have regionalized higher densities of microglia as compared to tissue from non-epilepsy controls and that this would be apparent beyond the context of acute seizure activity. Using Iba-1 immunolabelling, we found that single ramified microglia and processes were found throughout the cortex; scattered perivascular macrophages were also labelled (Fig. 2A). Enlarged and more complex/branching microglia, focal aggregates and amoeboid/macrophage forms were noted in some ROIs (Fig. 2A). Consistent with our hypothesis, the Iba1 LI was significantly higher in all epilepsy (EP-NL and EP-L together) than NEC for all ROIs (P=4.0×10−13) and for both subgroup comparisons (EP-NL [P=3.7×10−13] and EP-L [P=3.5×10−13]) against NEC. Regional differences were noted within the epilepsy groups, and for individual ROIs compared between epilepsy and control groups (see Fig. 2, Fig. S16 and Table S3). We noted that the Iba1 LI was similar across EP-NL, EP-L and NEC in BA17, as compared to pulvinar and BA22 where the Iba1 labelling index was higher in EP-NL and EP-L groups than in NEC. Thus, these results are consistent with the view that there is an over-representation of microglial footprint in brain tissue from people with chronic epilepsy, and that such microglial responses may occur in a regionally-specific manner. This observation was supported by evidence of region-specific microglia expansion in epileptic mice as reflected by the increased number of Iba1-positive cells in entorhinal cortex but not in perirhinal cortex (Fig. S3). Moreover, in sham mice we also observed that the temporal cortices have a higher microglia density than the hippocampus, denoting regionalised microglia enrichment at baseline (Fig. S3).
Experimental evidence supports microglial activation as a modifying factor for cortical thickness in a mouse model of acquired epilepsy
The analyses so far have shown (i) elevated expression of microglia-related genes in brain regions, deemed ‘vulnerable’ from the ENIGMA-Epilepsy imaging study, in healthy brains and (ii) widespread regionalised over-representation of microglia in brain tissue from people with epilepsy compared to tissue from non-epileptic controls. To provide proof-of-concept evidence causally linking microglia activation to cortical thinning, we used a mouse model of acquired epilepsy where convulsive seizures originate and spread in the limbic system, and also involve the neocortex [18–21]. Spontaneous seizures develop a few days after the acute insult (mean±S.E.M., onset, 6.2±0.5 days, n=21 mice of Fig. 3C) and recur for months (Fig. S4B,C), and are drug-resistant [20]. Microglia are morphologically activated (CD11b-positive area, hippocampus: mm2, sham, 0.09±0.01; 1 week post-SE, 1.67±0.48, P=0.0002, Mann-Whitney test; n=10–11 mice) and proliferate by 2.0-fold on average within one week after SE as assessed in a different cohort of mice by analysis of Iba1-positive cells in the forebrain (number of cells, hippocampus: sham, 614.4±7.1; 1 week post-SE, 1347±68.6, P=0.0011, Mann-Whitney test; n=10–11 mice).
First, we studied whether the thickness and volume of selected cortical brain regions were reduced, and those of the lateral ventricles increased (directions of change as predicted by the ENIGMA-Epilepsy findings in humans) in placebo (non-medicated)-diet fed epileptic mice (n=10) vs sham mice (not exposed to SE, n=16) as assessed by post mortem MRI (experimental design in Fig. S1A). In the post mortem MRI analysis (Fig. 3A, Figs. S5 and S6), we also included additional placebo diet-fed epileptic mice (n=8) run in parallel with mice depicted in Fig. 3C (n= 10). These additional mice were video-EEG monitored only in the terminal disease phase (day 55–90), and their seizure frequency (0.91±0.26, n=8) was similar to mice depicted in Fig. 3C (1.65±0.75, n=10; p=0.41 by Mann-Whitney test). These additional epileptic mice were also included in the histopathological brain analysis (Fig. 3B) and for behavioural testing (Fig. 3D). We found that the lateral ventricles were enlarged by 2-fold (P=0.0006 by ANOVA followed by Tukey’s test; Fig. S5A): this effect was associated with a significant reduction in the volume of the entorhinal (P=0.043; Fig. S5E) and perirhinal (P=0.025 by ANOVA followed by Tukey’s test; Fig. S5F) cortices. No significant changes were observed in other brain areas such as the hippocampus, caudato-putamen and the thalamus, although their average volumes trended lower than the corresponding values in sham mice (Fig. S5B–D). Notably, a significant reduction in the thickness of entorhinal (P=0.0001 by ANOVA followed by Tukey’s test; Fig. 3A) and perirhinal (P=0.046 by ANOVA followed by Tukey’s test; Fig. S6) cortices was also measured in the same mice.
Next, we measured brain region volumes and cortical thickness in mice fed with medicated diet, supplemented with PLX3397, in order to deplete microglia by >90% (Fig. S7) during the initial disease development (i.e., until day 7 after spontaneous seizure onset; experimental design in Fig. S1A). Importantly, this decrease in microglia was induced transiently, since microglia re-populated the forebrain within one week after switching mice to a placebo non-medicated diet [32]. This protocol was followed in order to assess the impact of microglia activation in the initial phases of the disease on the structural brain changes detected in the epileptic mice. We found that the decrease in thickness of the entorhinal cortex of epileptic mice under placebo diet vs sham mice was prevented in mice that were depleted of microglia in the early disease phase (P=0.022 vs placebo by ANOVA followed by Tukey’s test; Fig. 3A). This indicates that early microglia depletion prevented the entorhinal cortical thinning determined in the chronic disease phase. Microglia depletion did not affect the thinning of the perirhinal cortex (Fig. S6) or the ventricle and subcortical volume changes occurring in epileptic mice (Fig. S5). Evaluation of neuronal cell number in the entorhinal cortex of epileptic mice receiving placebo diet showed a significant decrease in both neuronal cell density (P=0.0001 vs sham by ANOVA followed by Tukey’s test; Fig. 3B) and their average cell body size (P=0.037 vs sham by ANOVA followed by Tukey’s test; Fig. 3B). The reduction in cell density, but not of the average cell body size, in epileptic mice was prevented by microglia depletion in the early disease phase (P=0.006 vs placebo diet by ANOVA followed by Tukey’s test; Fig. 3B). Similarly, microglia depletion affected the reduction in discrimination index (a measure of non-spatial memory deficit dependent on entorhinal cortex function) [33] observed in epileptic mice fed with placebo diet (P=0.0004 epileptic mice vs sham by ANOVA followed by Tukey’s test; P=0.049 epileptic mice on the PLX3397 diet vs placebo diet by Mann-Whitney test) (Fig. 3D). The total exploration time of objects during the familiarisation phase did not differ in the various experimental groups (sham, 24.07±2.6 sec; SE+placebo, 28.4±4.7 sec; SE+PLX3397, 29.7±5.9 sec; P=0.25 sham vs placebo; P=0.5 SE+placebo vs SE+PLX3397). Notably, microglia depletion in the early disease phase did not modify the onset, duration and severity of SE (Fig. S4A). Moreover, both frequency and duration of spontaneous seizures was unaltered by early microglia depletion as compared to placebo non-medicated diet-fed mice (Fig. 3C; S4C vs B).
Discussion
The epilepsies are complex conditions with multiple facets including various causes, differing responses to treatment, and unpredictable outcomes. Most attention has been paid to causation and processes of epileptogenesis across the broad constituent spectrum of syndromes. In contrast, disease progression has not been a primary focus of research even though for some rare epilepsies (the DEE), the window of opportunity to ameliorate disease may be open for longer than expected, as disrupted patterns of gene expression last into adulthood [34]. Here, we show that across the broad spectrum of the more common epilepsies (specifically excluding DEE), gene expression-predicted microglia density spatially correlated with reduced cortical thickness; the related genes newly implicate innate immunity, and, particularly, microglial activation, as contributors to the underlying cause of cortical thinning. We also show that this molecular signature of innate immunity activation is significantly enriched for a gene set already causally linked to seizure frequency in a mouse model of chronic epilepsy [27], though we note our model differs from the model in that study. Further, our data add new evidence supporting the general concept that microglial activation is associated with at least some of the structural changes seen in brain areas involved in seizure circuitry, in that microglial depletion in mice early during disease development can directly prevent associated cortical thinning in the entorhinal cortex. Notably, microglia depletion in mice also prevented neuronal cell loss in the entorhinal cortex and this neuroprotective effect was associated with improvement of cognitive deficit measured by the non-spatial memory test, that is sensitive to entorhinal cortex function [33].
Furthermore, we tested the hypothesis that microglial/monocyte activation is a key modulator of the severity of epilepsy using both genetic and functional approaches. We used the availability of GWAS data for resistance to anti-epileptic drug treatment, a marker of disease severity, to investigate enrichment in heritability at genetic loci already known to influence the expression of genes involved in monocyte activation. We find a highly significant enrichment in the heritability of epilepsy severity amongst these immunomodulatory loci, despite the absence of a significant enrichment for heritability of epilepsy per se. Finally, in keeping with these observations, experimental microglial depletion timed to coincide with a period of epileptogenesis in a murine model of acquired epilepsy can prevent regionalised cortical thinning, but does not influence the eventual development of seizures themselves in our model. Data from the experimental model also demonstrate that the cortical thinning is at least partly due to reduced neuronal cell density and average neuronal size: reduction of neuronal density can be prevented by appropriately-timed microglial depletion while neuronal size changes were not rescued, which may explain the observation that entorhinal volume changes are not completely prevented by microglial depletion. We, and others [35–40], find activated microglia are present in excess in brain tissue from people with epilepsy, compared to brain tissue from healthy controls, providing evidence for translation to human epilepsies of our assertions from the experimental model data. We selected Iba1 as a robust immunomarker for microglia in formalin-fixed tissue [41]. Like HLA-DR and CD68, Iba1 labels all phenotypes of microglia from ramified and amoeboid forms to macrophages and is therefore suited to structural studies of normal cortex in the absence of focal pathology [42]. We recently reported increased Iba1 labelling in central autonomic cortical regions in SUDEP, which also associated with increased seizures prior to death [43]. Together, these findings separate important processes occurring in the course of the epilepsies, and incriminate potentially modifiable microglial activation states in the hitherto largely-ignored feature of cortical thinning in the common human epilepsies. Unsurprisingly, our results also suggest other factors are likely to be involved, which we have not explored further yet.
Importantly, we note as limitations that we assume a high degree of similarity in the genetics of gene expression in monocytes and microglia and that the clue to the possible role for microglial activation in cortical thinning came from a cross-sectional study of chronic human epilepsy: although reduced cortical thickness correlated with disease duration [8], we could not distinguish whether the structural difference had developed at disease onset (e.g., with causation), during epileptogenesis, during the course of the disease, or a combination of these epochs. Our multimodal data, and especially the experimental model results, allow us to begin to address this question. Notably, we used spatially-resolved whole-brain gene expression data from healthy controls, rather than from the brains of people with epilepsy specifically to avoid confounding by secondary effects: some such effects (e,g., compensatory changes) may be worth exploring, but that was not our purpose here. The murine model in our study relates to early processes in epileptogenesis, and shows a clear separation for microglial roles in cortical thinning and seizure occurrence, whilst data from other models relate to the chronic disease state, and shows an effect of microglial manipulation on seizure frequency in that chronic state [27,44] (cortical thinning was not assessed in those models). The experimental and human data are not directly compatible, and we cannot test hypotheses arising from the chronic mouse model in data from human ENIGMA-Epilepsy. However, taken together, the data suggest microglia may have multiple modifying roles during epileptogenesis and progression of disease across common human epilepsies, though we find no evidence that they contribute to the actual occurrence of these common forms of epilepsy (from either our human or murine data). That seizure frequency and cortical thinning may be separable processes adds to important epidemiological evidence that seizure frequency is not the only contributor to morbidity in people with a history of epilepsy [5]. Microglia have many roles in specific types of epilepsy, demonstrated clearly in a variety of animal models. Such roles include phagocytosis, which may link consumption of synapses with cognitive changes in long-term active epilepsy [45], providing another possible mechanism for actual loss of brain volume in epilepsy: ‘time is brain’ [46].
Dysregulation of innate immunity is considered to possibly contribute to brain pathology and seizures in some severe human epilepsies. Microglial activation is seen in Rasmussen’s encephalitis and in tissue from epilepsies due to hippocampal sclerosis and mesial temporal lobe sclerosis, focal cortical dysplasia and tuberous sclerosis [35–39]. The latter two conditions are known to have genetic, rather than inflammatory, causes, but the extent of microglial activation in the chronic disease phase correlates with severity (seizure frequency) and disease duration in these studies [35,39] and not just cause, pointing again to distinctions between processes related to the initial cause (e.g. genetic disorder) and others that manifest during active disease. Importantly, resected human brain tissue is only available from a few cases of a few types of epilepsies (mostly surgical specimens from MTLE and focal cortical dysplasia [47]), so that it is impossible to otherwise evaluate the role of microglia using neuropathological data in the majority of common human epilepsies, from which brain tissue cannot be obtained in life. Brain imaging in animal models using a label (TSPO) for microglial activation shows dynamic upregulation during epileptogenesis, with persistent, although declining, activity in the chronic phase and correlation with spontaneous seizures [48]; in chronic human temporal lobe epilepsy, there is increased TSPO binding ipsilateral and contralateral to seizure foci [49] and in the interictal phases [50]. Using immunolabelling, we demonstrate that there is over-representation of activated microglia in human and experimental brain tissue, compared to controls. The animal data also highlight that microglia expansion in epileptic mice occurs in the entorhinal but not in perirhinal cortex, and this effect is mirrored in the rescue of cortical thinning by microglia depletion (Fig. S3). However, neither TSPO imaging nor neuropathological study is realistically applicable to large numbers of people with epilepsy, especially common epilepsies, whilst MRI is, providing a readily-available means of evaluating clinical translation of the implications of our findings. We propose, using MRI-derived patterns and correlation with gene expression, that activated microglia-associated functions drive the important, but as yet largely neglected, phenotype of cortical thinning in a broad swathe of common human epilepsies. Subsequent experimental intervention in an animal model suggests that early manipulation of microglia has the capacity to rescue disease-related cortical thinning, neuronal cell loss and cognitive deficits, opening up new areas for attention and treatment in common human epilepsies. We note that other cell-types and processes are also implicated: these also need consideration, and suggest the possibility of multiple players in cortical thinning: we have focussed on one cell-type, which does not diminish the potential utility of its manipulation. Other processes and cell-types will be the subject of future investigations.
Our results point to important roles for neuroinflammatory pathways and potentially specific molecular actors, such as IFN-γ. However, the diversity of microglial states and functions, and the complex, dynamic, interactions between neurons, astroglia and microglia, that at the very least can promote epileptogenesis [25] have yet to be fully resolved. Clinical translation of our key observation of the widespread role of microglial activation across the breadth of types of epilepsy into therapeutic options to prevent irreversible loss of brain substance will require definition of the time course of thinning in the different types of epilepsy. Translation will also require the development of safe, effective and tolerable treatments that target precise mechanisms without compromising immune surveillance of brain tissue, a need across diverse neurological disorders [51].
Supplementary Material
Supplementary material
Supplementary material_S1-S6
Acknowledgements
We thank Angela Richard-Londt, Francesca Launchbury and Matthew Ellis for assistance with human neuropathology data collection, Mojgansadat Borghei, Ilaria Lagorio and Yi Yao for assistance with data collection and Rafal Kaminski and Jonathan van Eyll for helpful discussions. We thank Dr L Porcu (Department of Oncology, Istituto di Ricerche Farmacologiche Mario Negri IRCCS, Milano, Italy) for statistical advice.
Funding
A. Altmann holds a Medical Research Council eMedLab Medical Bioinformatics Career Development Fellowship. This work was supported by the Medical Research Council (grant number MR/L016311/1). M.R. holds an Medical Research Council Clinician Scientist Fellowship (grant number MR/N008324/1). R.H.R. was supported through the award of a Leonard Wolfson Doctoral Training Fellowship in Neurodegeneration. The work was supported by grants from the European Union (7th Framework Programme Grants 279062, EpiPGX and 602102, EPITARGET). This work was partly undertaken at UCLH/UCL, which received a proportion of funding from the Department of Health’s NIHR Biomedical Research Centres funding scheme. The Epilepsy Society through the Katy Baggott Foundation supports the Epilepsy Brain and Tissue Bank at UCL and ERUK support the Corsellis Epilepsy brain collection. The work was also supported by the Epilepsy Society, UK (C.L., S.M.S.). Associazione Italiana Contro L’Epilessia (FIRE-AICE) and Fondazione Monzino (A.V.) and by NIH grants R01 NS097719 (M.E.M.) U54 EB020403 (P.T.) and NIH/NINDS R01 NS065838 (C.R.M.).
Data Availability Statement
The complete normalized microarray gene expression data from the Allen Institute for Brain Science that support the findings of this study are available from the institute’s website at http://human.brain-map.org/static/download. The eQTL summary statistics on state-specific monocytes that support the findings of this study are available in the supplementary tables at DOI: 10.1126/science.1246949. The summary results from the EpiPGX GWAS data that support the findings of this study are available from http://www.epigad.org. Data from the experimental model that support the findings of this study are available from the corresponding author upon reasonable request.
Abbreviations
AHBA Allen Human Brain Atlas
BA Brodmann Area
DEE developmental and epileptic encephalopathies
eQTL expression Quantitative Trait Locus
EP-L Lesional Epilepsy
EP-NL Non-Lesional Epilepsy
FDR False Discovery Rate
GO Gene Ontology
GWAS Genome Wide Association Study
IFN-γ Interferon
KEGG Kyoto Encyclopaedia of Genes and Genomes
LD Linkage Disequilibrium
LI Labelling Index
LPS Lipopolysaccharide
MTLE Mesial Temporal Lobe Epilepsy
MNI Montreal Neurological Institute
NEC Non-Epilepsy Controls
ROI region of interest
SE status epilepticus
Fig. 1. Analysis overview
Panel A: The ENIGMA-Epilepsy study identified “vulnerable” and “relatively protected” brain regions indicated in red and blue, respectively (second column; top) [8]. Cortical samples of the AIBS dataset (purple dots; first column, bottom) were marked as either “vulnerable” (red dots) or “relatively protected” (blue dots) depending on their location (second column; middle). Brain cell-type fractions were estimated from the gene expression data and the differential analysis showed an increased fraction of microglia and endothelial cells in “vulnerable” compared to “relatively protected” regions. Differential gene expression analysis between the two groups followed by pathway analysis confirmed the enrichment for marker genes for microglia as well as immune activation related pathways.
Panel B: LD score regression estimating the enrichment of immune response eQTL signatures in different epilepsy GWAS finds strong enrichment in disease severity (drug-resistant vs drug-susceptible) but not in disease risk (cases vs controls).
Fig. 2. Presence of excess activated microglia in post mortem brain tissue from people with epilepsy
Panel A: High magnification of morphological types of Iba1-labelled cells including (a) ‘rod’ cells, (b) ramified microglia, (c) perivascular macrophage and (d) amoeboid forms. Fixation time in illustration of ramified microglia was 467 days (bar = 30 microns). Right column shows Iba1 labelling in randomly selected from cases in the study (representing all three groups) with a range of immunostaining quantified from 0.5 to 6.5% field fraction with progressive increase in complexity, number and size of ramified microglial (all taken at X20).
Panel B: Scatter graph of all data points from 709 sections including all brain regions showing mean and standard deviation for labelling index in the four main groups: Epilepsy-ALL (EP-ALL), Epilepsy Non-Lesional (EP-NL), Epilepsy-Lesional (EP-L) and non-epilepsy controls (NEC). EP-ALL, EP-NL, EP-L are all significantly greater than NEC (*respective P-values: 4.0×10−13, 3.7×10−13, 3.5×10−13).
Panel C. Iba1 immunolabelling shown in 10 Brodmann areas and thalamus in each hemisphere, colour coded for the mean percentage labelling index in the three groups as in B.
Panel D. Scatter graph of the mean Iba1 LI in the same Brodmann areas and thalamus (averaged over both hemispheres) in the three groups as in B.
Fig. 3. Effects of microglia depletion in the early disease phase on entorhinal cortex thickness and neuronal cell loss, and on cognitive deficit in epileptic mice
The experimental design is depicted in Fig. S1A. Grey symbols represent sham (n=6–8) and epileptic mice fed with placebo diet (n=8) run in parallel with experimental mice of Fig 3C (see text for details).
Panel A: box-and-whisker plots depicting median, minimum, maximum and single values related to the entorhinal cortex thickness, as assessed by quantitative post mortem MRI analysis performed in epileptic mice at the end of EEG monitoring (placebo are mice fed with non-medicated diet: n=18; PLX3397 are mice fed with medicated diet supplemented with PLX3397: n=7), and in sham mice (not exposed to status epilepticus; n=16). MRI images depict representative slices showing the ROI used to quantify the cortical thickness. Four mice in the PLX3397 group did not undergo MRI analysis and therefore they were not included in the subsequent histological (B) and behavioral analyses (D). The white line within the ROI was manually drawn to measure the cortical thickness. **P=0.0001 vs sham; °P=0.022 vs placebo by ANOVA followed by Tukey’s test. Scale bar: 1 cm.
Panel B: representative Nissl-stained sections (top row) of the entorhinal cortex in the experimental groups (top row; sham, n=14; placebo, n=15; PLX3397, n=7), and the relative quantification of the number and the average size of Nissl-stained neurons (bottom row). Two sham and three placebo mice were excluded from the analysis due to poor quality of Nissl staining. Data are shown by box-and-whisker plots depicting median, minimum, maximum and single values *P=0.0037, **P=0.0001 vs sham; °°P=0.006 vs placebo diet by ANOVA followed by Tukey’s test. Scale bars: 100 μm.
Panel C: box-and-whisker plots depicting median, minimum, maximum and single values of the number of spontaneous seizures/day and their average duration during days 1–7, 8–15 and 55–90 from epilepsy onset (day 1) in the placebo (n=10) and PLX3397-supplemented diet (n=11) experimental groups (protocol in Fig. S1A). Friedman’s two-way nonparametric ANOVA (p=0.041) followed by post-hoc multiple comparisons test with Bonferroni correction: P-values for Number of seizures/day: p=0.363, days 1–7; p=0.339, days 8–15; p=0.965, days 55–90; P-values for Seizures duration: p=0.799, days 1–7; p=0.325, days 8–15; p=0.262, days 55–90.
Outliers were identified only for the Number of seizures/day in the placebo group (n=1 in days 1–7) and in the PLX3397 group (n=2 in days 8–15 and n=2 in days 55–90), however, their omission did not change the results of the primary statistical analysis therefore the values were not removed from the corresponding data set (P-values for sensitivity analysis: p=0.831, days 1–7; p=0.375, days 8–15; p=0.084, days 55–90).
Panel D: Novel object recognition test (NORT) in epileptic mice fed with placebo- (n=13) or PLX3397-supplemented diet (n=7), and sham controls (n=13). Three sham and five placebo diet epileptic mice were excluded from the analysis since they showed a total exploration time <6 sec during the familiarisation phase. Memory was evaluated by measuring the discrimination index, which was calculated as time spent (sec) exploring the familiar (F) and the novel (N) object as follows: (N - F)/(N + F). Data are shown by box-and-whisker plots depicting median, minimum, maximum and single values, differences significant at **P=0.0004 vs sham by ANOVA followed by Tukey’s test; *P=0.049 vs placebo by Mann-Whitney test.
Table 1: Association between inferred cell type fractions and reduced cortical thickness.
Cell type T-value P-value PFDR Pperm
Astrocytes 2.080608 0.037 0.043 0.17
Endothelial 4.426344 1.0×10 −5 7.67×10 −5 0.026
ExNeurons −3.855666 1.15×10−4 3.08×10−4 0.051
InNeurons 0.684069 0.49 0.49 0.37
Microglia 4.081655 4.5×10 −5 1.79×10 −4 0.032
Oligodendrocytes 2.456355 0.014 0.019 0.053
OPC 2.758717 5.8×10−3 0.014 0.068
Unknown 2.690987 0.007 0.011 0.066
Columns represent the cell type, the t-value from the association analysis, the corresponding P-value and the corrected P-value using FDR. The least column lists a P-value obtained from 1,000 permutations of the vulnerable/protected status of 34 ROIs in the left hemisphere.
Table 2: Results from stratified LD score regression estimating the enrichment of immune response eQTL signatures (rows) in different epilepsy GWAS (columns).
GWAS
ILAE (epilepsy vs HC) Drug-resistant vs HC Drug-resistant vs drug responders
eQTL Type P PFDR P PFDR P PFDR
Naïve monocytes 8.73×10−2 1.63×10−1 3.43×10−2 8.12×10−2 2.85×10−3 2.00×10 −2
INF- γ-treated 1.56×10−2 8.12×10−2 5.09×10−2 1.06×10−1 4.53×10−4 9.52×10 −3
LPS-treated (2 hours) 1.43×10−1 2.32×10−1 2.70×10−1 3.90×10−1 3.48×10−2 8.12×10−2
LPS24–treated (24 hours) 3.17×10−2 8.12×10−2 1.97×10−2 8.12×10−2 1.27×10−3 1.33×10 −2
GWAS, genome-wide association study; eQTL, expression quantitative trait locus; ILAE, International League Against Epilepsy study (The International League Against Epilepsy Consortium on Complex Epilepsies, 2014); HC, healthy controls; LPS, lipopolysaccharide. Bold font marks significant enrichments at PFDR<0.05.
Ethical Approval
The project has ethical approval (under NRES 17/SC/0573) and post mortem epilepsy cases had research consent, compliant with the UK Human Tissue Act 2014. Furthermore, in accordance with the ARRIVE guidelines, procedures involving animals and their care were conducted in conformity with the institutional guidelines that are in compliance with national (D.L. n.26, G.U. March 4, 2014) and international guidelines and laws (EEC Council Directive 86/609, OJ L 358, 1, December 12, 1987, Guide for the Care and Use of Laboratory Animals, U.S. National Research Council, 1996), and were reviewed and approved by the intramural ethical committee.
Conflict of Interest statements
C.D.W. is now an employee of Biogen Research and Early Development (Cambridge, Massachusetts, 02142, USA).
References
1. Orsini A , Zara F , Striano P . Recent advances in epilepsy genetics. Neurosci Lett. 2018;667 :4–9.28499889
2. The International League Against Epilepsy Consortium on Complex Epilepsies. Genetic determinants of common epilepsies: a meta-analysis of genome-wide association studies. Lancet Neurol. 2014;13 (9 ):893–903.25087078
3. The International League Against Epilepsy Consortium On Complex Epilepsies. Genome-wide mega-analysis identifies 11 new loci and highlights diverse biological mechanisms in the common epilepsies. Nat Commun. 2018;9 :5269.30531953
4. Barkovich AJ , Dobyns WB , Guerrini R . Malformations of Cortical Development and Epilepsy. Cold Spring Harb Perspect Med. 2015;5 (5 ):a022392.25934463
5. Bell GS , Neligan A , Giavasi C , Keezer MR , Novy J , Peacock JL , Outcome of seizures in the general population after 25 years: A prospective follow-up, observational cohort study. J Neurol Neurosurg Psychiatry. 2016;87 (8 ):843–50.26780937
6. Novy J , Bell GS , Peacock JL , Sisodiya SM , Sander JW . Epilepsy as a systemic condition: Link with somatic comorbidities. Acta Neurol Scand. 2017;136 (4 ):352–9.28573736
7. Klein P , Dingledine R , Aronica E , Bernard C , Blümcke I , Boison D , Commonalities in epileptogenic processes from different acute brain insults: Do they translate? Epilepsia. 2018;59 (1 ):37–66.29247482
8. Whelan CD , Altmann A , Botía JA , Jahanshad N , Hibar DP , Absil J , Structural brain abnormalities in the common epilepsies assessed in a worldwide ENIGMA study. Brain. 2018;141 (2 ):391–408.29365066
9. Vadlamudi L , Milne RL , Lawrence K , Heron SE , Eckhaus J , Keay D , Genetics of epilepsy: The testimony of twins in the molecular era. Neurology. 2014;83 (12 ):1042–8.25107880
10. Corey LA , Pellock JM , Kjeldsen MJ , Nakken KO . Importance of genetic factors in the occurrence of epilepsy syndrome type: A twin study. Epilepsy Res. 2011;97 (1–2 ):103–11.21885256
11. Heidari M , Johnstone DM , Bassett B , Graham RM , Chua ACG , House MJ , Brain iron accumulation affects myelin-related molecular systems implicated in a rare neurogenetic disease family with neuropsychiatric features. Mol Psychiatry. 2016;21 (11 ):1599–607.26728570
12. Zhang B , Gaiteri C , Bodea LG , Wang Z , McElwee J , Podtelezhnikov AA , Integrated systems approach identifies genetic nodes and networks in late-onset Alzheimer’s disease. Cell. 2013;153 (3 ):707–20.23622250
13. Voineagu I , Wang X , Johnston P , Lowe JK , Tian Y , Horvath S , Transcriptomic analysis of autistic brain reveals convergent molecular pathology. Nature. 2011;474 (7351 ):380–4.21614001
14. Johnson MR , Shkura K , Langley SR , Delahaye-Duriez A , Srivastava P , Hill WD , Systems genetics identifies a convergent gene network for cognition and neurodevelopmental disease. Nat Neurosci. 2016;19 (2 ):223–32.26691832
15. Hawrylycz MJ , Lein ES , Guillozet-Bongaarts AL , Shen EH , Ng L , Miller JA , An anatomically comprehensive atlas of the adult human brain transcriptome. Nature. 2012;489 (7416 ):391–9.22996553
16. Hawrylycz M , Miller JA , Menon V , Feng D , Dolbeare T , Guillozet-Bongaarts AL , Canonical genetic signatures of the adult human brain. Nat Neurosci. 2015;18 (12 ):1832–44.26571460
17. Blanc F , Martinian L , Liagkouras I , Catarino C , Sisodiya SM , Thom M . Investigation of widespread neocortical pathology associated with hippocampal sclerosis in epilepsy: A postmortem study. Epilepsia. 2011;52 (1 ):10–21.
18. Frigerio F , Pasqualini G , Craparotta I , Marchini S , van Vliet EA , Foerch P , n-3 docosapentaenoic acid-derived protectin D1 promotes resolution of neuroinflammation and arrests epileptogenesis. Brain. 2018;141 (11 ):3130–43.30307467
19. Jimenez-Mateos EM , Engel T , Merino-Serrais P , McKiernan RC , Tanaka K , Mouri G , Silencing microRNA-134 produces neuroprotective and prolonged seizure-suppressive effects. Nat Med. 2012;18 (7 ):1087–94.22683779
20. Iori V , Iyer AM , Ravizza T , Beltrame L , Paracchini L , Marchini S , Blockade of the IL-1R1/TLR4 pathway mediates disease-modification therapeutic effects in a model of acquired epilepsy. Neurobiol Dis. 2017 Mar;99 :12–23.27939857
21. Gu B , Huang YZ , He XP , Joshi RB , Jang W , McNamara JO . A Peptide Uncoupling BDNF Receptor TrkB from Phospholipase Cγ1 Prevents Epilepsy Induced by Status Epilepticus. Neuron. 2015;88 (3 ):484–91.26481038
22. French L , Paus T . A FreeSurfer view of the cortical transcriptome generated from the Allen Human Brain Atlas. Front Neurosci. 2015;
23. Menden K , Marouf M , Oller S , Dalmia A , Magruder DS , Kloiber K , Deep learning--based cell composition analysis from tissue expression profiles. Sci Adv. 2020;6 (30 ):eaba2619.
24. Weiner J 3rd , Domaszewska T . tmod: an R package for general and multivariate enrichment analysis. PeerJ. 2016;4.
25. Devinsky O , Vezzani A , Najjar S , De Lanerolle NC , Rogawski MA . Glia and epilepsy: Excitability and inflammation. Trends Neurosci. 2013;36 (3 ):174–84.23298414
26. Hickman S , Izzy S , Sen P , Morsett L , El Khoury J . Microglia in neurodegeneration. Nat Neurosci. 2018;21 (10 ):1359–69.30258234
27. Srivastava PK , van Eyll J , Godard P , Mazzuferi M , Delahaye-Duriez A , Van Steenwinckel J , A systems-level framework for drug discovery identifies Csf1R as an anti-epileptic drug target. Nat Commun. 2018;9 (1 ):3561.30177815
28. Mathys H , Adaikkan C , Gao F , Young JZ , Manet E , Hemberg M , Temporal Tracking of Microglia Activation in Neurodegeneration at Single-Cell Resolution. Cell Rep. 2017;21 (2 ):366–80.29020624
29. Holtman IR , Raj DD , Miller JA , Schaafsma W , Yin Z , Brouwer N , Induction of a common microglia gene expression signature by aging and neurodegenerative conditions: a co-expression meta-analysis. Acta Neuropathol Commun. 2015;3 :31.26001565
30. Grabert K , Michoel T , Karavolos MH , Clohisey S , Baillie JK , Stevens MP , Microglial brain region-dependent diversity and selective regional sensitivities to aging. Nat Neurosci. 2016;19 (3 ):504–16.26780511
31. Fairfax BP , Humburg P , Makino S , Naranbhai V , Wong D , Lau E , Innate immune activity conditions the effect of regulatory variants upon monocyte gene expression. Sci (New York, NY). 2014;343 (6175 ):1246949.
32. Elmore MRP , Najafi AR , Koike MA , Dagher NN , Spangenberg EE , Rice RA , Colony-stimulating factor 1 receptor signaling is necessary for microglia viability, unmasking a microglia progenitor cell in the adult brain. Neuron. 2014;82 (2 ):380–97.24742461
33. Wilson DIG , Langston RF , Schlesiger MI , Wagner M , Watanabe S , Ainge JA . Lateral entorhinal cortex is critical for novel object-context recognition. Hippocampus. 2013;23 (5 ):352–66.23389958
34. Delahaye-Duriez A , Srivastava P , Shkura K , Langley SR , Laaniste L , Moreno-Moral A , Rare and common epilepsies converge on a shared gene regulatory network providing opportunities for novel antiepileptic drug discovery. Genome Biol. 2016;17 (1 ):245.27955713
35. Boer K , Spliet WGM , Van Rijen PC , Redeker S , Troost D , Aronica E . Evidence of activated microglia in focal cortical dysplasia. J Neuroimmunol. 2006;173 (1–2 ):188–95.16483671
36. Aronica E , Gorter JA , Redeker S , Ramkema M , Spliet WGM , Van Rijen PC , Distribution, characterization and clinical significance of microglia in glioneuronal tumours from patients with chronic intractable epilepsy. Neuropathol Appl Neurobiol. 2005;31 (3 ):280–91.15885065
37. Ravizza T , Gagliardi B , Noé F , Boer K , Aronica E , Vezzani A . Innate and adaptive immunity during epileptogenesis and spontaneous seizures: Evidence from experimental models and human temporal lobe epilepsy. Neurobiol Dis. 2008;29 (1 ):142–60.17931873
38. Iyer A , Zurolo E , Spliet WGM , Van Rijen PC , Baayen JC , Gorter JA , Evaluation of the innate and adaptive immunity in type i and type II focal cortical dysplasias. Epilepsia. 2010;51 (9 ):1763–73.20345941
39. Ravizza T , Boer K , Redeker S , Spliet WGM , van Rijen PC , Troost D , The IL-1beta system in epilepsy-associated malformations of cortical development. Neurobiol Dis. 2006;24 (1 ):128–43.16860990
40. Morin-Brureau M , Milior G , Royer J , Chali F , LeDuigou C , Savary E , Microglial phenotypes in the human epileptic temporal lobe. Brain. 2018;141 (12 ):3343–60.30462183
41. Boche D , Perry VH , Nicoll JAR . Activation patterns of microglia and their identification in the human brain. Neuropathol Appl Neurobiol. 2013;39 (1 ):3–18.23252647
42. Hendrickx DAE , van Eden CG , Schuurman KG , Hamann J , Huitinga I . Staining of HLA-DR, Iba1 and CD68 in human microglia reveals partially overlapping expression depending on cellular morphology and pathology. J Neuroimmunol. 2017;309 :12–22.28601280
43. Somani A , El-Hachami H , Patodia S , Sisodiya S , Thom M . Regional microglial populations in central autonomic brain regions in SUDEP. Epilepsia. 2021;62 (6 ):1318–28.33942290
44. Wu W , Li Y , Wei Y , Bosco DB , Xie M , Zhao M-G , Microglia depletion aggravates the severity of acute and chronic seizures in mice. Brain Behav Immun. 2020;
45. Weinhard L , Di Bartolomei G , Bolasco G , Machado P , Schieber NL , Neniskyte U , Microglia remodel synapses by presynaptic trogocytosis and spine head filopodia induction. Nat Commun. 2018;9 (1 ):1228.29581545
46. Caciagli L , Bernasconi A , Wiebe S , Koepp MJ , Bernasconi N , Bernhardt BC . A meta-Analysis on progressive atrophy in intractable temporal lobe epilepsy: Time is brain? Neurology. 2017;89 (5 ):506–16.28687722
47. Blumcke I , Spreafico R , Haaker G , Coras R , Kobow K , Bien CG , Histopathological Findings in Brain Tissue Obtained during Epilepsy Surgery. N Engl J Med. 2017;377 (17 ):1648–56.29069555
48. Amhaoul H , Hamaide J , Bertoglio D , Reichel SN , Verhaeghe J , Geerts E , Brain inflammation in a chronic epilepsy model: Evolving pattern of the translocator protein during epileptogenesis. Neurobiol Dis. 2015;82 :526–39.26388398
49. Gershen LD , Zanotti-Fregonara P , Dustin IH , Liow J-S , Hirvonen J , Kreisl WC , Neuroinflammation in Temporal Lobe Epilepsy Measured Using Positron Emission Tomographic Imaging of Translocator Protein. JAMA Neurol. 2015;72 (8 ):882–8.26052981
50. Butler T , Li Y , Tsui W , Friedman D , Maoz A , Wang X , Transient and chronic seizure-induced inflammation in human focal epilepsy. Epilepsia. 2016;57 (9 ):e191–4.27381590
51. Gupta N , Shyamasundar S , Patnala R , Karthikeyan A , Arumugam TV ., Ling EA , Recent progress in therapeutic strategies for microglia-mediated neuroinflammation in neuropathologies. Expert Opin Ther Targets. 2018;22 (9 ):765–81.30138572
|
PMC008xxxxxx/PMC8983092.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
9805460
21529
Urol Oncol
Urol Oncol
Urologic oncology
1078-1439
1873-2496
31495569
8983092
10.1016/j.urolonc.2019.08.006
NIHMS1783868
Article
Evaluation of MSKCC Preprostatectomy nomogram in men who undergo MRI-targeted prostate biopsy prior to radical prostatectomy
Glaser Zachary A. M.D. a*
Gordetsky Jennifer B. M.D. ab
Bae Sejong PhD c
Nix Jeffrey W. M.D. a
Porter Kristin K. M.D., PhD d
Rais-Bahrami Soroush M.D. ad
a Department of Urology, University of Alabama at Birmingham, Birmingham, AL
b Department of Pathology, University of Alabama at Birmingham, Birmingham, AL
c Division of Preventive Medicine, University of Alabama at Birmingham, Birmingham, AL
d Department of Radiology, University of Alabama at Birmingham, Birmingham, AL
Author contributions: Zachary A. Glaser: Conceptualization, data curation, formal analysis, investigation, methodology, resources, validation, visualization, writing - original draft, and writing - review and editing; Jennifer B. Gordetsky: Conceptualization, data curation, investigation, methodology, project administration, supervision, validation, visualization, writing - original draft, and writing - review and editing; Sejong Bae: Formal analysis, software, validation, investigation; Jeffrey W. Nix: Data curation, funding acquisition, supervision, validation, visualization, writing - original draft, and writing - review and editing; Kristin K. Porter: Conceptualization, data curation, supervision, validation, visualization, writing - original draft, and writing - review and editing; Soroush Rais-Bahrami: Conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, software, supervision, validation, visualization, writing - original draft, and writing - review and editing.
* Corresponding author. Tel.: +1-314-580-4008; fax: +1-205-934-4933. [email protected] (Z.A. Glaser).
10 3 2022
12 2019
05 9 2019
05 4 2022
37 12 970975
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Introduction:
The Memorial Sloan Kettering Cancer Center (MSKCC) Preprostatectomy nomogram is a widely used resource that integrates clinical factors to predict the likelihood of adverse pathology at radical prostatectomy. Adoption of magnetic resonance imaging targeted biopsy (TB) permits optimized detection of clinically-significant cancer over systematic biopsy (SB) alone. We aim to evaluate the prognostic utility of the MSKCC Preprostatectomy nomogram with TB pathology results.
Methods:
Men who underwent SB and magnetic resonance imaging TB who later underwent radical prostatectomy at our institution were included. Patient information was entered into the MSKCC Preprostatectomy nomogram using 5 biopsy reporting schemes with TB reported by both individual core (IC) and aggregate group (AG) methods. The likelihood of extraprostatic extension, seminal vesicle invasion, and lymph node involvement as predicted by the nomogram for each biopsy reporting schema were compared to radical prostatectomy pathology.
Results:
We identified 63 men from January 2014 to November 2017. On receiver operating characteristic analysis, IC-TB, AG-TB, SB plus IC-TB, and SB plus AG-TB exhibited similar, if not improved, area under the curve compared to SB alone in predicting extraprostatic extension (0.671, 0.674, 0.658, and 0.6613 vs. 0.6085). This was similarly observed for seminal vesicle invasion prediction using SB plus IC-TB compared to SB alone (0.727 vs. 0.733). For lymph node involvement, superior but nonsignificant area under the curve was observed for AG-TB (0.647) compared to IC-TB (0.571) and SB alone (0.524)
Conclusions:
Using TB pathology results either alone or combined with SB pathology results as input to the MSKCC Preprostatectomy nomogram appears comparable for prognosticating adverse pathology on radical prostatectomy compared to SB alone, but robust validation is warranted prior to adoption into clinical practice.
Magnetic resonance imaging
Radical prostatectomy
Fusion biopsy
Prostate cancer
Risk calculator
pmc1. Introduction
In 2018, nearly 165,000 American men will be newly diagnosed with prostate cancer (CaP) [1]. Among men with newly diagnosed clinically-significant localized disease, radical prostatectomy is considered a standard-of-care approach to primary definitive treatment with curative intent [2]. However, surgical intervention is not without recognized potential morbidity and may not yield durable cure in select circumstances. As such, clinical risk stratification tools such as the Memorial Sloan Kettering Cancer Center (MSKCC) Preprostatectomy nomogram are widely relied upon to guide shared decision-making in this pretreatment setting [3]. While historically useful, these nomograms were developed and validated on systematic extended-sextant biopsy schema but not extensively studied or validated in the era of advanced imaging for prostate cancer detection as well as image-targeted biopsy (TB) of suspicious lesions.
Multiparametric magnetic resonance imaging (mpMRI) with fusion-TB has become increasingly utilized for the detection and localization of CaP. Robust evidence suggests this approach superiorly detects clinically-significant cancer foci using fewer cores compared to systematic biopsy (SB) [4–8]. While recently proposed nomograms that account for degree of cancer suspicion on mpMRI and TB pathology have been developed, incorporation of TB pathology into a previously validated, and widely accepted clinical tool such as the MSKCC Preprostatectomy nomogram has not been evaluated [9–13] Since many urologists currently refer to this nomogram in routine clinical practice, understanding the prognostic utility of TB in place of, or in conjunction with, SB pathology from which the nomogram was largely derived has potentially broad and impactful implications. Furthermore, urologists would not be tasked with incorporating the often complex and sometimes unavailable lesion parameters on MRI that are required for use of the novel TB nomograms [9–10].
Additionally, defining the best method of pathologic reporting for prostate biopsy cores sampled from a MRI-targeted lesion for use with this nomogram has yet to be determined. It has been proposed that reporting the aggregate group (AG) of pathologic findings from multiple cores sampled from a single target as a composite grade and percentage of cancer involvement better correlates with true tumor volume and the finding of extraprostatic extension (EPE) on radical prostatectomy than an individual core (IC) method [14]. Herein, we aim to evaluate the MSKCC Preprostatectomy nomogram with inclusion of TB pathology according to both IC and the proposed AG histologic reporting schemas in conjunction with SB findings and in isolation compared to SB pathology findings alone.
2. Methods
2.1. Patient selection and imaging
After obtaining institutional review board approval, our prospectively maintained patient database was accessed for men who underwent SB and TB and later elected for radical prostatectomy at our institution. A mpMRI was performed on all men using a 3 Tesla MRI scanner. Images were reviewed and interpreted by an abdominopelvic radiologist with at least 3 years of experience and subspecialization in genitourinary MRI. A multidisciplinary prostate imaging conference comprised of abdominopelvic radiologists, genitourinary pathologists, and urologic oncologists then reviewed each suspicious lesion, and assigned a Prostate Imaging Reporting and Data System (PI-RADS) v2 suspicion score for target designation as previously described [13].
2.2. Biopsy protocol and surgery
For biopsy naïve men, SB was performed concurrently at the time of TB. For men with a prior SB history, repeat SB was conducted at the time of TB in some cases based upon shared decision-making model between the patient and urologist. TB acquisition was performed using the UroNav system (Philips Medical Systems/InVivo, Gainesville, FL). At least 2 needle cores were obtained for each suspicious focus. Biopsy pathology was discussed in clinic where the options of active surveillance (AS), surgery, radiation therapy, and other treatment modalities were presented. All men who opted for surgery underwent robotic assisted laparoscopic radical prostatectomy with or without bilateral pelvic lymph node dissection based on preoperative clinical staging, patient counselling, and available adjunct data. TB, preoperative counseling, and surgery were performed by 1 of 2 fellowship-trained urologic oncologists who performed all TB cases. The decision to perform pelvic lymph node dissection was based on shared decision-making as well as a ≥2% chance of lymph node involvement (LNI) on the MSKCC Preprostatectomy nomogram as recommended by the NCCN guidelines [15]. All biopsy and final surgical pathology were reviewed and reported by a single fellowship-trained genitourinary surgical pathologist. The TB cores were given Gleason score/Grade Groups and percentage of tissue involved with CaP from 1 of 2 reporting methods as previously described [14]. In brief, the IC technique reported Gleason score/Grade Group and percent core involvement for each individual biopsy core irrespective of the number of cores samples from any given MRI-targeted lesion. The AG method reported the composite Gleason score/Grade Group and percent tissue involved with CaP assuming that all biopsy cores from the same MRI-targeted lesion was 1 tissue sample (Fig. 1).
2.3. Nomogram prognostication
In order to compare the prognostic utility of entering TB pathology into the MSKCC Preprostatectomy nomogram either in place of, or in conjunction with, SB pathology, 5 biopsy pathology schemas were generated based on SB and TB core data: SB alone, IC from TB alone, AG from TB alone, SB plus IC from TB, and SB plus AG from TB. Patient clinical and demographic information was entered into the MSKCC Preprostatectomy nomogram using each biopsy schema [3]. The likelihood of organ-confined disease, EPE, seminal vesicle invasion (SVI), and LNI as predicted by the nomogram were compared to final surgical pathology from radical prostatectomy.
2.4. Statistical analysis
Descriptive statistics were formulated for demographic and other clinical/pathologic characteristics. The receiver operating characteristic (ROC) curves, area under the curve (AUC) analyses, were used to compare the prognostic accuracy of the nomogram with input from each of the 5 biopsy reporting schemas compared to final surgical pathology from radical prostatectomy. DeLong methodology was performed for statistical comparisons of the ROCs [16]. Statistical significance was considered if P values were less than a predetermined threshold of 0.05. All analyses were conducted using SAS v9.4 (SAS Institute, Cary NC).
3. Results
From January 2014 to November 2017, 644 men underwent prostate mpMRI followed by fusion TB. Among these, 63 (9.8%) pursued radical prostatectomy in the same time frame at our institution. Available clinical and demographic information are summarized in Table 1. Median age was 65 years. Our cohort consisted of 51/63 (81%) white men and 11 (17%) African American men. Preoperative disease characteristics included a median prostate specific antigen (PSA) of 7.43 (IQR 5.2–10.1), 54/63 (85.7%) with cT1c disease, and no men had undergone previous oncologic intervention.
All but 2 men underwent SB prior to or at time of TB. On SB, cancer was not identified in 24/61 (39.3%) men, and Grade Group 1, 2, and 3 CaP was identified in 22 (36.1%), 8 (13.1%), and 6 (9.8%) men, respectively (Table 2). All 63 men underwent mpMRI and subsequent TB. A majority of men (88.7%) had 1 to 3 suspicious prostatic foci which were sampled by a median of 4 (IQR 3.5–6) total target cores. De novo cancer detection or Grade Group upgrade occurred in 46 (73%) of men following TB. All men included in this study analysis underwent radical prostatectomy. A majority of men harbored Grade Group 2 (n = 26, 41.3%) or 3 (n = 23, 36.5%) disease on final surgical pathology (Table 3). EPE and SVI were identified in 34 (54%) and 6 (9.5%) of men, respectively. LNI was identified in 7 (11.1%) of the 49 (77.8%) men who underwent lymph node dissection.
On ROC analysis, likelihood of EPE, SVI, and LNI as predicted by the MSKCC Preprostatectomy nomogram using 1 of 5 biopsy pathology reporting schemas were individually compared to final surgical pathology. For prediction of EPE, using IC alone, AG alone, SB plus IC, and SB plus AG exhibited higher absolute AUCs to SB alone (0.671, 0.674, 0.658, 0.661 vs. 0.609, though not statistically significant: P > 0.05, Fig. 2). For SVI, the AUC for AG plus SB was comparable to SB alone (0.727 vs. 0.733, P > 0.05), and IC alone, AG alone, IC plus SB exhibited inferior absolute AUCs (0.428, 0.630, 0.699, P > 0.05, Fig. 2). Following ROC analysis of LNI, IC alone, and AG alone demonstrated higher absolute AUCs to SB alone (0.571, 0.647 vs. 0.524, P > 0.05), and IC plus SB as well as AG plus SB had inferior absolute AUCs (0.472 and 0.472, P> 0.05, Fig. 2).
4. Discussion
Performing mpMRI and subsequent TB during the diagnostic workup of CaP has become commonplace, and is recommended by the National Comprehensive Cancer Network in the setting of a prior negative biopsy and/or suspicion of aggressive disease [15,17]. Recent evidence demonstrates this diagnostic modality optimally detects significant cancer requiring fewer cores than traditional SB [4,7,18]. Moreover, men eligible for AS may be more likely to pursue this when TB is obtained in addition to SB, thus avoiding a potentially unnecessary radical intervention [19].
Among the 644 men who underwent mpMRI and TB, just 9.8% pursued radical prostatectomy during the same study period at our institution. This is likely multifactorial and could be due to having persistently benign prostatic tissue on TB following a prior negative SB, which is a common occurrence [19,20]. Alternatively, a significant proportion (>50% in this study) of men considering or already enrolled in AS elected to continue with this surveillance strategy following a stable risk assessment rendered based upon TB results as previously observed by Lai et al. [13] Lastly, men may have elected to pursue definitive surgical or nonsurgical intervention at another center.
Prognostic tools such as the MSKCC Preprostatectomy nomogram facilitate the discussion of surgery with patients and offer insight as to the likelihood of adverse pathological outcomes [3]. The likelihood of biochemical recurrence and disease progression using this formula was designed, validated, and revalidated in an era prior to the widespread adoption of MRI-directed TB sampling [21–23]. While providers may already enter targeted core pathology into the Preprostatectomy nomogram, evaluation of the algorithm in this high-yield biopsy setting is warranted to ensure patients are not receiving misguided preoperative counseling.
While several preoperative nomograms incorporating mpMRI and TB pathology do exist, robust validation and incorporation into routine practice present considerable barriers [9,10,24]. Moreover, detailed imaging information required for these nomograms such as lesion size and clinical stage on MRI may not readily be available to the clinical urologist, can vary across institutions based on the MRI magnet strength, sequences used and the experience of the reporting radiologist. This is the first study to assess the prognostic reliability of an already existing, previously externally validated, widely used and user-friendly nomogram with TB pathology. These findings suggest using TB pathology alone or in conjunction with SB cores predicts final surgical pathology at least comparably to SB pathology alone.
In order to adequately sample a region of interest during a TB session, multiple cores are usually acquired from each target. For nomograms originally designed to prognosticate standard template biopsies, this could be problematic as more than 1 positive core from the same suspicious foci could skew the algorithm. Therefore, reporting the overall yield from an aggregate of cores from 1 suspicious imaging focus may better reflect the true pathologic cancer burden. Gordetsky et al. recently evaluated this reporting method in a cohort of men who underwent TB and subsequent radical prostatectomy. Interestingly, the AG method demonstrated greater concordance with tumor lesion volume, lesion density, and the presence of EPE on mpMRI compared to standard IC reporting [14]. This study is among the first to utilize this novel AG reporting method in a standardized manner [14]. Superior prediction of surgical pathology using the AG method was not demonstrated in the present study. However, this should bring to light the lack of a discipline-wide consensus on TB pathology reporting, and the potential impact on patient care given the variability in reporting methodologies across institutions and even individual pathologists.
Our study has several acknowledged limitations. This represents a single institution, retrospective study lacking long-term follow-up to evaluate biochemical recurrence and overall survival which were also prognostic read-outs of the MSKCC Preprostatectomy nomogram. The cohort size precludes the establishment of both superiority and inferiority regarding biopsy core reporting, and is considerably smaller than the patient populations used to generate MSKCC Preprostatectomy and more recent Gandaglia et al. nomograms [3,9]. In addition, the clinical course of men who underwent mpMRI and subsequent TB but did not pursue radical prostatectomy at our institution was not captured in the present study. Lastly, LNI could not be evaluated in the 14 men where a pelvic lymph node dissection was not performed. Whether or not the lymph node pathology would have impacted the overall findings of this study is unknown. Future multi-institutional studies may allow for larger patient populations and normalization across institutional variability to potentially validate the current study findings. Subsequent long-term follow-up of these patients will permit evaluation of important outcomes such as biochemical recurrence and survival.
5. Conclusions
Using TB pathology either alone or combined with SB pathology results for the MSKCC Preprostatectomy nomogram appears comparable for prognosticating adverse pathologic features on radical prostatectomy. Continued use of this nomogram may be a pragmatic alternative to the complex nomograms recently proposed for this setting, but robust validation is warranted before adopting this into routine clinical practice.
Funding:
This work was funded in part by Junior Faculty Development Grant (ACS-IRG 001–53) and by Developmental funds from the UAB Comprehensive Cancer Center Support Grant (NCI P30 CA 013148) to Soroush Rais-Bahrami.
Fig. 1. Overview of biopsy acquisition and pathologic reporting schemas for a prostate with a single target (T) identified via mpMRI (A). Biopsy reporting consisted of systematic cores only (red, B), target cores only (white, C), or a combination (D). Target cores were then reported as individual cores (IC) or as aggregate groups (AG, E).
Fig. 2. ROC curves for predicting (A) EPE, (B) SVI, and (C) LNI using various biopsy reporting schemas compared to final surgical pathology according to the MSKCC Preprostatectomy nomogram.
Table 1 Baseline patient characteristics.
No. of patients 63
Median age (IQR) 65 (60–69)
Ethnicity (%)
White 51 (81.0)
Black 11 (17.5)
Other 1 (1.5)
Median PSA (IQR) 7.43 (5.2–10.1)
Clinical stage (%)
T1c 54 (85.7)
T2a 8 (12.7)
T2b 1 (1.5)
Table 2 Standard and targeted prostate biopsy pathology.
Systematic biopsy Grade Group (%)
No cancer 24 (39.3)
1 22 (36.1)
2 8 (13.1)
3 6 (9.8)
4 0 (0)
5 1 (1.6)
No systematic biopsy 2 (3.2)
No. targets (%)
1 20 (31.7)
2 24 (38.0)
3 12 (19.0)
4 2 (3.2)
>4 5 (7.9)
Median no. cores (IQR) 4 (3.5–6)
Target biopsy Grade Group (%)
No cancer 3 (4.8)
1 11 (17.5)
2 20 (31.7)
3 13 (20.6)
4 12 (19.0)
5 4 (6.3)
De novo CaP or Grade Group upgrade (%) 46 (73)
CaP = prostate cancer.
Table 3 Final radical prostatectomy surgical pathology.
Median prostate weight (g) 48.15
Grade Group (%)
1 2 (3.2)
2 26 (41.3)
3 23 (36.5)
4 5 (7.9)
5 7 (11.1)
Extraprostatic extension (%) 34 (54.0)
Seminal vesicle invasion (%) 6 (9.5)
Lymph node dissection performed (%) 49 (77.8)
Lymph node invasion (%) 7 (11.1)
Conflicts of interest: Jeffrey W. Nix and Soroush Rais-Bahrami serve as consultants to Philips/InVivo Corp.
References
[1] Siegel RL , Miller KD , Jemal A . Cancer statistics, 2018. CA Cancer J Clin 2018;68 :7–30.29313949
[2] Sanda Martin G. ; Crispino Tony ; Freedland Stephen ; Greene Kirsten ; Klotz Laurence H. ; Makarov Danil V. ; Nelson Joel B. ; Reston James ; Rodrigues George ; Sandler Howard M. ; Taplin Mary Ellen ; Cadeddu Jeffrey A. . CLINICALLY LOCALIZED PROSTATE CANCER: AUA/ASTRO/SUO GUIDELINE. In: Association AU , editor. AUA Guidelines 2017.
[3] Memorial Sloan Kettering Cancer Center. Prediction Tools/Prostate Cancer Nomograms: Pre- Radical Prostatectomy. Memorial Sloan Kettering Cancer Center; 2018.
[4] Yarlagadda VK , Lai WS , Gordetsky JB , MRI/US fusion-guided prostate biopsy allows for equivalent cancer detection with significantly fewer needle cores in biopsy-naive men. Diagn Interv Radiol 2018;24 :115–20.29770762
[5] Siddiqui MM , George AK , Rubin R , Efficiency of prostate cancer diagnosis by MR/ultrasound fusion-guided biopsy vs standard extended-sextant biopsy for MR-visible lesions. J Natl Cancer Inst 2016;108. https://academic.oup.com/jnci/article/108/9/djw039/2499548.
[6] Schoots IG , Roobol MJ , Nieboer D , Magnetic resonance imaging-targeted biopsy may enhance the diagnostic accuracy of significant prostate cancer detection compared to standard transrectal ultrasound-guided biopsy: a systematic review and meta-analysis. Eur Urol 2015;68 :438–50.25480312
[7] Siddiqui MM , Rais-Bahrami S , Turkbey B , Comparison of MR/ultrasound fusion-guided biopsy with ultrasound-guided biopsy for the diagnosis of prostate cancer. JAMA 2015;313 :390–7.25626035
[8] Kasivisvanathan V , Rannikko AS , Borghi M , MRI-targeted or standard biopsy for prostate-cancer diagnosis. N Engl J Med 2018;378 :1767–77.29552975
[9] Gandaglia G , Ploussard G , Valerio M , A novel nomogram to identify candidates for extended pelvic lymph node dissection among patients with clinically localized prostate cancer diagnosed with magnetic resonance imaging-targeted and systematic biopsies. Eur Urol 2018;75 :506–14.30342844
[10] Rayn KN , Bloom JB , Gold SA , Added value of multiparametric magnetic resonance imaging to clinical nomograms in predicting adverse pathology in prostate cancer. J Urol 2018;200 :1041–7.29852182
[11] Bjurlin MA , Rosenkrantz AB , Sarkar S , Prediction of prostate cancer risk among men undergoing combined MRI-targeted and systematic biopsy using novel pre-biopsy nomograms that incorporate MRI findings. Urology 2018;112 :112–20.29155186
[12] Truong M , Wang B , Gordetsky JB , Multi-institutional nomogram predicting benign prostate pathology on magnetic resonance/ultrasound fusion biopsy in men with a prior negative 12-core systematic biopsy. Cancer 2018;124 :278–85.28976544
[13] Lai WS , Gordetsky JB , Thomas JV , Factors predicting prostate cancer upgrading on magnetic resonance imaging-targeted biopsy in an active surveillance population. Cancer 2017;123 :1941–8.28140460
[14] Gordetsky JB , Schultz L , Porter KK , Defining the optimal method for reporting prostate cancer grade and tumor extent on magnetic resonance/ultrasound fusion-targeted biopsies. Hum Pathol 2018;76 :68–75.29551676
[15] Prostate Cancer. NCCN Clinical Practice Guidelines in Oncology. Version 4.2018 ed.
[16] DeLong ER , DeLong DM , Clarke-Pearson DL . Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988;44 :837–45.3203132
[17] Siegel RL , Miller KD , Jemal A . Cancer statistics, 2017. CA Cancer J Clin 2017;67 :7–30.28055103
[18] Radtke JP , Schwab C , Wolf MB , Multiparametric magnetic resonance imaging (MRI) and MRI-transrectal ultrasound fusion biopsy for index tumor detection: correlation with radical prostatectomy specimen. Eur Urol 2016;70 :846–53.26810346
[19] Gordetsky JB , Saylor B , Bae S , Prostate cancer management choices in patients undergoing multiparametric magnetic resonance imaging/ultrasound fusion biopsy compared to systematic biopsy. Urol Oncol 2018;36 :241 e7–e13.
[20] Bjurlin MA , Renson A , Rais-Bahrami S , Predicting benign prostate pathology on magnetic resonance imaging/ultrasound fusion biopsy in men with a prior negative 12-core systematic biopsy: external validation of a prognostic nomogram. Eur Urol Focus 2018. 10.1016/j.euf.2018.05.005.
[21] Kattan MW , Wheeler TM , Scardino PT . Postoperative nomogram for disease recurrence after radical prostatectomy for prostate cancer. J Clin Oncol 1999;17 :1499–507.10334537
[22] Kattan MW , Vickers AJ , Yu C , Preoperative and postoperative nomograms incorporating surgeon experience for clinically localized prostate cancer. Cancer 2009;115 :1005–10.19156928
[23] Stephenson AJ , Scardino PT , Eastham JA , Postoperative nomogram predicting the 10-year probability of prostate cancer recurrence after radical prostatectomy. J Clin Oncol 2005;23 :7005–12.16192588
[24] Feng TS , Sharif-Afshar AR , Wu J , Multiparametric MRI improves accuracy of clinical nomograms for predicting extracapsular extension of prostate cancer. Urology 2015;86 :332–7.26194289
|
PMC008xxxxxx/PMC8983093.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
8003538
2808
Cardiovasc Intervent Radiol
Cardiovasc Intervent Radiol
Cardiovascular and interventional radiology
0174-1551
1432-086X
31044292
8983093
10.1007/s00270-019-02226-5
NIHMS1783877
Article
Percutaneous Cryoablation of Stage T1b Renal Cell Carcinoma: Safety, Technical Results, and Clinical Outcomes
Gunn Andrew J. http://orcid.org/0000-0001-9081-446X
1
Joe Winston B. 2
Salei Aliaksei 1
El Khudari Husameddin 1
Mahmoud Khalid H. 1
Bready Eric 1
Keasler Eric M. 1
Patten Patrick P. 1
Gordetsky Jennifer B. 3
Rais-Bahrami Soroush 4
Abdel Aal Ahmed K. 1
1 Division of Vascular and Interventional Radiology, Department of Radiology, University of Alabama at Birmingham, 619 19th St S, Birmingham, AL 35249, UK
2 University of Alabama at Birmingham School of Medicine, 1720 2nd Ave. S., Birmingham, AL 35294, UK
3 Department of Pathology, University of Alabama at Birmingham, 619 19th St S, Birmingham, AL 35249, UK
4 Department of Urology, University of Alabama at Birmingham, 1720 2nd Ave. S., Birmingham, AL 35294, UK
✉ Andrew J. Gunn, [email protected]
10 3 2022
7 2019
01 5 2019
05 4 2022
42 7 970978
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Purpose
The use of percutaneous cryoablation for T1b (4.1–7.0 cm) renal cell carcinoma, has not yet been widely adopted. The purpose of this study was to describe our experience in the cryoablation of stage T1b tumors with an emphasis on safety, technical results, and clinical outcomes.
Materials and Methods
A retrospective review of hospital records identified 37 patients who underwent cryoablation for T1b lesions from 2008 to 2018. Patient demographics, comorbidities, tumor characteristics, technical parameters, and outcomes were recorded and analyzed. Recurrence-free, overall, and cancer-specific survival rates were estimated using the Kaplan–Meier method.
Results
Thirty-seven patients (22 males, 15 females; mean age 66.5 ± 11.3) with 37 T1b tumors (mean diameter 47.3 ± 6.3 mm) were included. A median of 3 probes were used (range: 1–7). Angio-embolization was used in 3/37 (8.1%) and 2/37 patients (5.4%) required hydrodissection. The mean number of total cryoablation procedures for each patient was 1.5 (median 1; range: 1–4). Technical success was achieved in 88.2% of patients. Recurrence-free survival was 96.5%, 86.1%, and 62.6% at 1, 2, and 3 years respectively. Cancer-specific survival was 100% at 1, 2, and 3 years respectively. Overall survival was 96.7%, 91.8%, and 77.6% at 1, 2, and 3 years respectively. Complications classified as CIRSE grade 2 or higher occurred in 6/37 (16.2%) patients.
Conclusion
T1b cryoablation is associated with high rates of technical success, excellent cancer-specific survival, and an acceptable safety profile.
Level of Evidence
Level 4, Case Series.
Ablation
Cryoablation
Renal cell carcinoma
Kidney
pmcIntroduction
Widespread reliance on cross-sectional imaging in the diagnosis of intra-abdominal pathology has led to a rise in the global incidence of localized renal cell carcinoma (RCC) in recent decades [1–4]. As such, an estimated 65,340 new cases of RCC were diagnosed in the United States in 2018 alone [5]. While partial nephrectomy (PN) remains the standard of care for patients with early-stage, non-centralized lesions, there is an increasing role for ablative techniques in the management of RCC, particularly in patients with multiple comorbidities or those who wish to avoid traditional extirpative surgery [6, 7]. Standard image-guided thermal ablation techniques include radiofrequency ablation (RFA), microwave ablation (MWA), and cryoablation, all of which have been shown to achieve good oncologic outcomes in patients with smaller, T1a tumors (≤ 4.0 cm) [8–12]. This is reflected in recent guidelines set by the National Comprehensive Cancer Network (NCCN) and American Urological Association (AUA), which present thermal ablation as a viable therapy for localized T1a tumors [6, 7].
Use of percutaneous ablation for larger tumors, such as T1b (4.1–7.0 cm), however, has not yet been adopted by clinical practice guidelines. This is, in part, due to concern that ablation of higher T stage RCC may be associated with a greater likelihood of recurrence [6, 7, 13–15]. Yet, early experience suggests percutaneous cryoablation may be an option for the treatment of T1b lesions in certain settings. In a subset analysis, cryoablation for T1b tumors achieved recurrence-free and metastasis-free survival comparable to that of PN [8]. Studies including both T1a and T1b lesions have shown rates of tumor recurrence at short-term follow-up are low and not necessarily related to tumor size [16, 17]. Furthermore, cryoablation of T1b tumors does not appear to be limited by technical feasibility [18]. In fact, compared to percutaneous RFA, cryoablation may attain a higher rate of tumor coverage by the ablative zone resulting in lower rates of retreatment [19]. Although complications following cryoablation of larger tumors remain a concern [16, 20], early experience suggests that ablation of T1b tumors is safe in the setting of thorough periprocedural care [18], a result which was corroborated in a recent case series [21]. Despite these promising results, further studies are necessary to establish the potential of percutaneous cryoablation in the treatment of stage T1b RCC. The purpose of this study was to describe our experience in the percutaneous cryoablation of stage T1b RCC with an emphasis on safety, technical results, and clinical outcomes.
Materials and Methods
Patient Population and Data Collection
This single-center retrospective review was approved by an Institutional Review Board (IRB) and was compliant with the Health Insurance Portability and Accountability Act (HIPAA). Patient informed consent was not required given the retrospective nature of this study. A review of our radiology information system (RIS) identified 37 adult patients with 37 RCC lesions measuring 4.1–7.0 cm who underwent percutaneous cryoablation from 2008 to 2018. Patient demographics (age, sex, race, and body mass index) and the presence of comorbid conditions were identified. Multiple comorbidities were summarized for each patient using the Charlson Comorbidity Index with age.
Tumor Classification
For all tumors, pre-operative multiplanar imaging was reviewed to evaluate anatomical features including size, location, tumor geometry, nearness to collecting system, and involvement of the renal sinus. Based on these characteristics, tumor complexity was graded according to both R.E.N.A.L. (radius, exophytic/endophytic properties, nearness to the collecting system, anterior/posterior, location relative to the polar lines) nephrometry scores [22] and PADUA scores (Preoperative Aspects and Dimensions Used for an Anatomical score) [23]. R.E.N.A.L. nephrometry scores were used to stratify tumors to low (score ≤ 6), intermediate (score 7–9), and high complexity groups (score > 9). PADUA scores were used similarly to stratify tumors to low (score 6–7), intermediate (score 8–9), and high complexity groups (score > 9).
Technical Aspects
Pre-procedural evaluation and planning was carried out by a combination of interventional radiologists and urologists with accompanying input from a multidisciplinary team when indicated. In all instances, percutaneous cryoablation was performed by a board-certified interventional radiologist. Routine prophylactic antibiotics were administered, and anticoagulation was held prior to the procedure according to published guidelines [24, 25]. Sites were prepared and draped in standard sterile fashion following mass localization and identification of the percutaneous route by preliminary computed tomography (CT) scans, cryoablation probes were advanced under CT guidance. All cryoprobes utilized in this series relied on rapidly-expanding, highly-pressurized argon gas to induce tumor freezing and subsequent necrosis. With cryoablation probes in place, patients typically underwent two freeze thaw cycles consisting of a 10-minute initial freeze with an 8-minute passive thaw followed by an additional 10-minute freeze followed by active thaw. Intermittent CT scans were performed during the freeze cycles in order to monitor the progression of the ablation. In this series, the need for pre-ablation angio-embolization was based on lesion size and/or the suggestion of a highly vascular RCC on pre-procedural imaging per review of the notes. In the instance that tissue displacement was required in order to avoid injury to adjacent structures either hydrodissection or pneumodissection was performed. Confirmatory biopsy was performed concurrently with cryoablation when possible [26].
The number and types of ablation probes required for each procedure was recorded. The mean number of total cryoablation procedures per patient was recorded. Follow-up consisted of a clinic visit approximately 2–3 weeks after the procedure to evaluate for early complications. In general, imaging of the ablation zone was performed 3 months after the procedure then again 9 months later (12 months post-procedure). Patients subsequently underwent yearly surveillance imaging. Follow-up imaging intervals were shortened if there was suspicion for persistent or new disease. Contrast-enhanced CT was the preferred modality for follow-up imaging, however the use of other modalities, such as magnetic resonance imaging (MRI) or contrast-enhanced ultrasound (CEUS), was dictated by specific patient characteristics such as allergies to iodinated contrast or compromised renal function. Post-procedural measures included technical success and local recurrence. Technical success was defined by coverage of the entire lesion with the ice ball and the absence of new contrast enhancement and/or tumor enlargement within 3 months of ablation. Staged ablation procedures were considered technically successful if planned at or before the index procedure, in which case technical success and local recurrence were measured from the last staged procedure. Local recurrence was defined as new contrast enhancement within the ablation zone or enlargement of the ablated tumor greater than or equal to 3 months after the procedure. Any complications that occurred from the time of the procedure until last known follow-up or death were identified. These were graded according to the Cardiovascular and Interventional Radiological Society of Europe (CIRSE) classification system for complications [27]. Other parameters such as reasons for interventional radiology referral and relevant laboratory values (pre- and post-procedural) were collected. Oncologic outcomes including recurrence-free, overall, and cancer-specific survival were assessed for all patients. Oncologic outcomes were measured from the index procedure if a single ablation session was performed or from the date of the last procedure if a staged ablation was performed.
Statistical Analysis
Continuous variables were summarized as means and standard deviations or medians with ranges. Categorical variables were expressed as frequencies and percentages. Univariate analyses were performed to assess whether tumor characteristics, patient demographics, comorbidities, or technical aspects (probe number, use of hydrodissection or embolization) correlated with observed outcomes. Due to small size, the Kruskal–Wallis test and Fisher’s exact test were used to compare continuous and categorical variables, respectively. Any variables with a p value ≤ 0.1 on univariate analysis were subsequently included in a multivariable logistic regression. Odds ratios (OR) with two-sided 95% confidence intervals (CI) were provided for any associations (p ≤ 0.05) from the stepwise selection method. Cancer-specific, recurrence-free, and overall survival were estimated at 1, 2, and 3 years using the Kaplan–Meier (KM) method. Time to event was defined from the date of procedure to the date of event or last follow-up. Subjects were censored if no event occurred or if lost to follow-up. A p-value of ≤ 0.05 was considered to indicate statistical significance. Statistical analyses were performed using the SAS software package, version 9.4 (SAS, Cary, North Carolina).
Results
Thirty-seven patients were treated with percutaneous cryoablation between 2008–2018. The decision to undergo cryoablation vs PN was based on poor surgical candidacy in 30/37 cases (81%) and patient preference in 7/37 cases (19%). Data regarding technical success and local recurrence were available for 34/37 (91.9%) patients. Detailed demographic and comorbidity data are summarized in Table 1. Tumor characteristics including renal nephrometry scores are presented in Table 2.
Technical success was achieved in 88.2% (30/34) of patients (Fig. 1). Among the patients not achieving technical success, one underwent repeat cryoablation for tumor enhancement noted 2 months after the index procedure. Worsening enhancement was noted approximately 6 months after the second ablation and, due to the patient’s advanced age and comorbidities, a multidisciplinary decision was made to proceed with chemotherapy. The patient passed away due to respiratory failure 1 year later. A separate patient was noted to have residual enhancement on CEUS approximately 3-months after the index procedure. This patient underwent repeat ablation; however, the residual disease persisted after the second ablation. The patient subsequently decided to proceed with radical nephrectomy. A third patient was noted to have residual enhancement on CT 2 months after the index procedure and was treated with repeat cryoablation. Approximately 14 months after the second procedure, follow-up imaging showed persistent tumor with new renal vein thrombosis. Given the patient’s advanced age and poor surgical candidacy, a multidisciplinary decision was made to palliate the patient’s hematuria with trans-arterial embolization. This patient passed away 3 months after embolization. In a fourth patient, marginal enhancement was noted on follow-up CT. This was treated with repeat cryoablation and the patient has completed four months of follow-up without evidence of residual disease.
The median number of probes used was 3, (range 1–7). While it is not common practice to use a single probe for tumors ≥ 4 cm, this occurred in one instance as the patient was not able to tolerate additional probe insertion during the index procedure. Thus, the patient was scheduled for a second procedure as part of a staged ablation. The second ablation was technically successful and the patient has now undergone 33 months of follow-up with no evidence of residual or recurrent disease. Information regarding probe brand was available for 16/37 (43.2%) patients. Probes used included IceFORCE (n = 6, 2.1 mm, Galil Medical Inc., Arden Hills, Minnesota), combined IceFORCE and IcePearl probes (n = 4, 2.1 mm, Galil Medical Inc.), Endocare Perc-24 probes (n = 3, 2.4 mm, Endocare, Austin, Texas), IceSphere probes (n = 2, 1.5 mm, Galil Medical Inc.), and combined IceSphere and IceRod probes (n = 1, 1.5 mm, Galil Medical Inc.). In general, the decision to use specific cryoablation systems and probes was left to the discretion of individual providers. The mean number of total cryoablation procedures per patient was 1.5, (median 1; range 1–4). Pre-ablation biopsy was performed in 62.2% (n = 23) of patients (Table 2). Angio-embolization was used in 3 (8.1%) patients prior to ablation. Two patients (5.4%) required hydrodissection.
Mean follow-up was 26.4 ± 28 months (median: 16.4, range: 0.03–102.6). Among the 34 patients for whom local recurrence could be adequately assessed, 8 patients (23.5%) experienced recurrence, at a mean of 26.5 ± 14.9 months (median: 25.2, range: 4.4–53). Recurrence-free survival was 96.5%, 86.1%, and 62.6% at 1, 2, and 3 years respectively (Fig. 2). None of the evaluated clinical variables including total nephrometry scores or their component parts were associated with an increased risk of local recurrence (p > 0.05), and no variables met the significance threshold for inclusion in multivariable analysis. Cancer-specific survival was 100% at 1, 2, and 3 years respectively. Overall survival was 96.7%, 91.8%, and 77.6% at 1, 2, and 3 years respectively. Charlson Comorbidity Index was the only variable associated with overall survival on logistic regression analysis (OR = 1.55, 95% CI 1.01–2.39).
CIRSE grade 1 complications occurred in 11 patients (29.7%) including perinephric hematoma (n = 9) (Fig. 3), vasovagal episode (n = 1), and flank pain (n = 1). Complications classified as CIRSE grade ≥ 2 occurred in 6 (16.2%) patients (Table 3). One patient experienced post-procedural hematuria which resolved spontaneously following overnight observation (CIRSE grade 2). One patient developed a persistent abscess which was managed with percutaneous drainage (CIRSE grade 3). A single instance of pseudoaneurysm with arterio-venous fistula was recognized approximately 1 day after ablation and successfully treated with coil embolization (CIRSE grade 3). Two patients experienced short hospital stays of ≤ 48 h, one for urinary tract infection with altered mental status, and another for unexplained leukocytosis. Both resolved following management with appropriate antibiotics (CIRSE grade 3). One patient developed an abscess with colo-ureteral fistula approximately two weeks after the initial procedure, ultimately requiring surgical intervention (CIRSE grade 4). No significant change in creatinine was seen after ablation (p = 0.85). Among tested clinical variables including both total RENAL and PADUA scores and their component parts, only endophytic tumors (p = 0.005), involvement of the renal sinus (p = 0.045), and displacement or infiltration of the collecting system (p = 0.022) were associated with an increased risk of complications. On multivariable analysis, tumors with endophytic/mixed as opposed to exophytic location (OR = 12.76; 95% CI 2.10–77.4) and tumors involving the renal sinus (OR = 6.33; 95% CI 1.08–37.3) were independently associated with complications.
Discussion
Cryoablation of T1b RCC presents unique technical challenges to the interventional radiologist [18]. Despite this, the present study indicates percutaneous cryoablation of these tumors can be achieved with a high rate of technical success (88.2%). Our results compare favorably to those of Hebbadj et al. [21] who reported the same measure (which authors termed “technical efficacy”) as 87.5% in a smaller series of 27 patients with T1b tumors. Moreover, Atwell et al. [18] also demonstrated excellent rates of technical success in their series of 46 T1b tumors treated with percutaneous cryoablation. These overall high rates of technical success may be explained by a number of factors. First, unlike other forms of thermal ablation, cryoablation permits real-time visualization of ice ball formation within the ablative zone. This enables the practitioner to monitor for complete coverage of the target zone with adequate margins [28]. Second, cryoablation also enables simultaneous use of multiple probes to create spherical, ellipsoid, or cylindrical ablation zones [29]. Once in place, probes work synergistically and can be manipulated independently to ensure large, irregular tumors are adequately treated while minimizing damage to nearby structures [30]. Taken together, these advantages suggest percutaneous cryoablation may be uniquely suited for the treatment of T1b tumors from a technical standpoint. To this end, a recent multi-center review comparing outcomes following RFA (n = 23) vs cryoablation (n = 23) for T1b tumors noted cryoablation was associated with better initial tumor coverage by the ablative zone and fewer instances of retreatment than RFA [19].
Current guidelines note that thermal ablation of larger tumors should be utilized with caution due to higher rates of local recurrence compared to partial nephrectomy. These guidelines, however, are based largely on experience with laparoscopic cryoablation and RFA rather than percutaneous cryoablation [12–14]. In one of the largest retrospective analyses of percutaneous cryoablation of T1b tumors to date, authors demonstrated favorable local tumor control rates with an estimated progression-free survival of 96.4% at 1-, 3- and 5-year follow-ups [18]. A more recent case series reported rates of local tumor control of 82.6% and 60.3% at 1 and 3 years respectively [21]. These authors defined local tumor control as absence of contrast enhancement and ablation area enlargement at imaging follow-up ≥ 6 months after the initial procedure. Similarly, 8 patients (23.5%) in our cohort experienced local recurrence with recurrence-free survival of 96.5%, 86.1%, and 62.6% at 1, 2, and 3 years respectively. The results of case series specific to T1b cryoablation, which now include a total of 110 patients, suggest that local recurrence following percutaneous cryoablation remains an important consideration in patients with T1b lesions. Despite this, the association between tumor size and local recurrence remains unclear [16, 17, 21, 31]. In our univariate analysis, none of the tested variables were associated with local recurrence, including tumor size. Nevertheless, local recurrence rates may not diminish the clinical utility of cryoablation for T1b tumors, given that repeat ablation following recurrence has been shown to be highly efficacious. For example, in a recent meta-analysis comparing thermal ablation to partial nephrectomy, any significant differences in recurrence free survival between surgery and thermal ablation disappeared when multiple ablations were considered [32]. Indeed, this may provide an explanation for the excellent cancer-specific survival noted in the present series (100% at 5 years) despite the rates of recurrence, as a majority of recurrences (63%) were treated with repeat ablation.
One major benefit of percutaneous cryoablation versus laparoscopic or traditional surgery appears to be a lower rate of complications [33]. It is important to assess whether these benefits persist in the treatment of larger tumors, as some studies have indicated complication rates increase with greater tumor size and number of probes used [16, 20]. In our cohort, we had an overall low rate of complications CIRSE grade 2 or above (16.2%). Complications were not associated with tumor size or the number of cryoablation probes on univariate analysis. The only variables associated with complications included: endophytic/mixed tumors (p = 0.005), involvement of renal sinus (p = 0.045) and displacement/infiltration of the collecting system (p = 0.022). On multivariable analysis, tumors with endophytic/mixed as opposed to exophytic location were greater than 12 times more likely to be associated with complications, OR = 12.7 (95% CI 2.10–77.4, p = 0.006). In addition, patients with tumors involving the renal sinus were over 6 times more likely to experience an adverse event, OR = 6.33 (1.08–37.3, p = 0.041). This result is similar to other studies which have identified centrally-located tumors to be associated with higher rates of complications, as these tumors may require more invasive probe placement and longer duration of freezing necessary to achieve tumor control [20, 34]. These findings may help guide patient selection and the need for close periprocedural monitoring in the future. Overall, the rate of clinically significant complications (CIRSE grade ≥ 2) observed in this study was similar to those reported by Hebbadj et al. and Atwell et al. in their experience with T1b cryoablation (11.1% and 15.2%, respectively). It should be noted that in both these instances major complications were reported according to Clavien-Dindo classification rather than the CIRSE classification [18, 21]. In the present study, CIRSE grade 1 complications, while not uncommon (29.7%), were largely limited to clinically insignificant perinephric hematomas (Fig. 3). Notably, a small degree of self-resolving perinephric bleeding is to be expected following cryoablation, and some authors do not consider it to be a true complication [18].
As the data presented here represents a decade of experience with percutaneous cryoablation of T1b RCC, it is important to acknowledge the ways in which practice patterns at our institution have changed over time. When this service line was first introduced, providers were chiefly concerned with the risk of complications associated with ablating larger lesions. This led to a conservative approach in which some providers preferred to stage ablations rather than use a large number of probes in the same session or employ adjunctive measures such as hydrodissection or embolization. We recognize that this is not a standard practice at other institutions. This factor contributed to the mean number of procedures performed in this cohort being > 1. Improved comfort and experience in treating larger lesions has enabled a more aggressive approach in our current practice. For example, in an aforementioned procedure performed earlier in our experience, a 4.2 cm RCC was treated in two separate ablation sessions with a single probe at each session. More recently, in 2018, a 6.7 cm RCC was treated using 7 probes in a single session. Regardless, it is possible that the more conservative approaches of our early practice, including the low utilization of displacement maneuvers intra-procedurally, could have resulted in under-sized ablation sizes thereby leading to recurrence. Despite this, recurrence rates were overall low, especially within the first two years after ablation, and similar to prior series. Nevertheless, the size and heterogeneity of the current cohort limits further analysis along these lines.
This study is not without limitations. First, probe brand was not systematically recorded within the electronic medical record, and therefore it was not possible to reliably assess the impact of specific probe type on measured outcomes. Second, the retrospective nature of this study did not allow us to control for important variables such as variations in technique that may exist between different providers at our institution or changing practice patterns over time. Third, in the literature specific to T1b ablation, definitions of technical success are discordant [18, 21]. Our definition aligns more closely with that offered by Atwell et al. although the present study allows for planned staged procedures to be considered technically successful [21]. This discordance highlights the need for outcomes to be classified in a more uniform manner as future experience with T1b RCC ablation is elaborated. Fourth, despite the fact that this study adds an additional 37 patients to the literature and is comparable in size to other studies that focused specifically on T1b RCC cryoablation [18, 21], it remains limited by small sample size with relatively few patients having follow-up > 60 mo. Fifth, this study was also limited by the fact that pre-procedural confirmatory biopsy was not performed universally, although it was carried out in a majority of patients (23/37, 62.2%). Standard practice at our institution involves placement of ablation probes prior to introduction of the biopsy needle. If bleeding is noted at this stage, the ablation may proceed without biopsy in order to minimize blood loss. Additionally, the use of multiple probes for large, irregular tumors may obscure biopsy needles on CT guidance, lowering the success rate of biopsy. Although histologic confirmation is in general an important component of interventional oncology, the optimal timing of percutaneous biopsy in the setting of thermal ablation has yet to be determined and remains an area of ongoing research [26].
Future directions include data collection in a multi-institutional, prospective fashion which would allow for more robust analysis of oncologic outcomes. Once sufficient data are available, a meta-analysis is likely to lend additional insight into the efficacy of percutaneous cryoablation in the treatment of higher T stage RCC. Ultimately, however, larger comparative effectiveness studies are needed in order to assess the relative merit of thermal ablation vs PN in the treatment of T1b RCC.
Conclusion
Percutaneous cryoablation is a viable option for T1b RCC with low rates of high grade complications. Local recurrence remains a concern in the cryoablation these tumors, however high rates of technical success may be achieved with excellent cancer-specific survival at 1, 2, and 3 years.
Acknowledgements
The authors would like to acknowledge the assistance of Dr. Yufeng Li, Ph.D, from the University of Alabama at Birmingham in performing the statistics.
Fig. 1 Example of technically successful T1b cryoablation as demonstrated by (A) pre-ablation axial contrast-enhanced CT image depicting exophytic T1b RCC (dashed arrow) of the right kidney (maximum diameter 46 mm) (B) ablation zone immediately post-procedure (solid white arrow) demonstrating ice ball with complete tumor coverage and appropriate margin (C) subtracted CT image showing ablation bed (dashed arrow) with no residual enhancement at 3-month follow-up
Fig. 2 Kaplan Meier curve demonstrating recurrence-free survival of 100% at 0 months (at risk = 34), 96.5% at 12 months (at risk = 22), 86.1% at 24 months (at risk = 13), 62.6% at 36 months (at risk = 6), and 37.6% at 60 months (at risk = 2)
Fig. 3 Axial CT images depicting (A) patient in prone position immediately prior to cryoablation of a 4.8 cm T1b renal tumor (solid white arrow) (B) ablation bed immediately after procedure surrounded by a small perinephric hematoma (dashed white arrow)
Table 1 Summary of patient demographics (N = 37)
Age
Mean ± SD 66.5 ± 11.3
Median (min–max) 65 (39.3–90.7)
Body mass index
Mean ± SD 34.8 ± 8.8
Median (min–max) 34 (15.5–54.3)
Comorbility index
Mean ± SD 7.1 ± 2.4
Median (min–max) 7 (2–12)
Sex n (%)
Male 22 (59.5)
Female 15 (40.5)
Race n (%)
White 23 (62.2)
Black 9 (24.3)
Asian 2 (5.4)
Other 3 (8.1)
Chronic kidney disease stage n (%)
< 3 17 (45.9)
3A 8 (21.6)
3B 8 (21.6)
4 1 (2.7)
5 3 (8.1)
Coronary artery disease n (%)
No 25 (67.6)
Yes 12 (32.4)
Diabetes mellitus n (%)
No 19 (51.4)
Yes 18 (48.6)
Hypertension n (%)
No 8 (21.6)
Yes 29 (78.4)
Heart failure n (%)
No 34 (91.9)
Yes 3 (8.1)
Table 2 Summary of tumor characteristics (N = 37)
Tumor size (mm)
Mean ± SD 47.3 ± 6.3
Median (min–max) 45 (41–67)
Total RENAL score
Low n (%) 10 (27.0)
Mean ± SD 5.6 ± 0.5
Median (min–max) 6 (5–6)
Intermediate n (%) 21 (56.8)
Mean ± SD 8.1 ± 0.8
Median (min–max) 8 (7–9)
High n (%) 6 (16.2)
Mean ± SD 10.3 ± 0.5
Median (min–max) 10 (10–11)
Overall
Mean ± SD 7.8 ± 1.7
Median (min–max) 8 (5–11)
Total PADUA Score*
Low n (%) 3 (8.1)
Mean ± SD 7 ± 0
Median (min–max) 7 (7–7)
Intermediate n (%) 17 (45.9)
Mean ± SD 8.5 ± 0.5
Median (min–max) 8 (8–9)
High n (%) 17 (45.9)
Mean ± SD 10.9 ± 1.0
Median (min–max) 11 (10–13)
Overall
Mean ± SD 9.5 ± 1.6
Median (min–max) 9 (7–13)
Stage n (%)
T1b 37 (100)
Anterior/posterior n (%)
Anterior 18 (48.6)
Posterior 19 (51.4)
Histologic type n (%)
Clear cell 12 (32.4)
Papillary 5 (13.5)
Unclassified 6 (16.2)
No biopsy 14 (37.8)
Table 3 Complications and laboratory values, all patients (N = 37)
Complications n (%)
None 20 (54.1)
≤ 24 h 14 (37.8)
≤ 30 days 2 (5.4)
> 30 days 1 (2.7)
CIRSE Grade n (%)
1 11 (29.7)
2 1 (2.7)
3 4 (10.8)
4 1 (2.7)
5 0
6 0
Creatinine before procedure
Mean ± SD 1.6 ± 1.3
Median (min–max) 1 (0.6–6.4)
INR before procedure
Mean ± SD 1.1 ± 0.3
Median (min–max) 1 (0.9–2.4)
Platelets before procedure
Mean ± SD 207.6 ± 84.8
Median (min–max) 189 (47–374)
GFR before procedure
Mean ± SD 50.2 ± 20.8
Median (min–max) 53 (8–126)
Creatinine after procedure
Mean ± SD 1.8 ± 1.7
Median (min–max) 1 (0.7–7.9)
GFR after procedure
Mean ± SD 47.6 ± 17.8
Median (min–max) 56 (6–70)
Compliance with Ethical Standards
Conflict of interest Dr. Gunn is a consultant and speaker for BTG International, and Dr. Abdel Aal is a consultant for Abbott Laboratories, Baxter Intl., W.L. Gore & Associates, C.R. Bard Inc., Sirtex Medical, and Boston Scientific Corporation. None of the other authors report a potential conflict of interest associated with this work and there is no grant support associated with this work.
Ethical Approval All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. For this type of study formal consent is not required. This study was approved by the Institutional Review Board (IRB) of the University of Alabama at Birmingham.
Human and Animal Rights This article does not contain any studies with animals performed by any of the authors.
Informed Consent For this type of study informed consent is not required.
References
1. Znaor A , Lortet-Tieulent J , Laversanne M , Jemal A , Bray F . International variations and trends in renal cell carcinoma incidence and mortality. Eur Urol. 2015;67 (3 ):519–30.25449206
2. Sun M , Thuret R , Abdollah F , Lughezzani G , Schmitges J , Tian Z , Age-adjusted incidence, mortality, and survival rates of stage-specific renal cell carcinoma in North America: a trend analysis. Eur Urol. 2011;59 (1 ):135–41.21035250
3. Capitanio U , Montorsi F . Renal cancer. Lancet. 2016;387 (10021 ):894–906.26318520
4. Hollingsworth JM , Miller DC , Daignault S , Hollenbeck BK . Rising incidence of small renal masses: a need to reassess treatment effect. J Natl Cancer Inst. 2006;98 (18 ):1331–4.16985252
5. Siegel RL , Miller KD , Jemal A . Cancer statistics, 2018. CA-Cancer J Clin. 2018;68 (1 ):7–30.29313949
6. Kidney Cancer. In: NCCN Clinical Practice Guidelines in Oncology. National Comprehensive Cancer Network. 2018. https://www.nccn.org/professionals/physician_gls/pdf/kidney.pdf. Accessed 21 Dec 2018.
7. Campbell S , Uzzo RG , Allaf ME , Bass EB , Cadeddu JA , Chang A , Renal mass and localized renal cancer: AUA guideline. J Urol. 2017;198 (3 ):520–9.28479239
8. Thompson RH , Atwell T , Schmit G , Lohse CM , Kurup AN , Weisbrod A , Comparison of partial nephrectomy and percutaneous ablation for cT1 renal masses. Eur Urol. 2015;67 (2 ):252–9.25108580
9. Takaki H , Yamakado K , Soga N , Arima K , Nakatsuka A , Kashima M , Midterm results of radiofrequency ablation versus nephrectomy for T1a renal cell carcinoma. Jpn J Radiol. 2010;28 (6 ):460–8.20661697
10. Yang Y , Chen S , Chen F , Zhu K , Deng Q , Luo L , Outcome of radiofrequency ablation over partial nephrectomy for small renal mass (< 4 cm): a systematic review and meta-analysis. Int J Clin Exp Med. 2015;8 (11 ):20670.26884989
11. Dandrea D , Shariat SF , Klatte T . Update on ablative therapies of renal tumors. Curr Opin Urol. 2016;26 (5 ):410–6.27214581
12. Moreland AJ , Ziemlewicz TJ , Best SL , Hinshaw JL , Lubner MG , Alexander ML , High-powered microwave ablation of t1a renal cell carcinoma: safety and initial clinical evaluation. J Endourol. 2014;28 (9 ):1046–52.24846329
13. Tanagho YS , Roytman TM , Bhayani SB , Kim EH , Benway BM , Gardner MW , Laparoscopic cryoablation of renal masses: single-center long-term experience. Urology. 2012;80 (2 ):307–14.22857748
14. Gervais DA , McGovern FJ , Arellano RS , McDougal WS , Mueller PR . Radiofrequency ablation of renal cell carcinoma: part 1, Indications, results, and role in patient management over a 6-year period and ablation of 100 tumors. AJR Am J Roentgenol. 2005;185 (1 ):64–71.15972400
15. Best SL , Park SK , Yaacoub RF , Olweny EO , Tan YK , Trimmer C , Long-term outcomes of renal tumor radio frequency ablation stratified by tumor diameter: size matters. J Urol. 2012;187 (4 ):1183–9.22335865
16. Schmit GD , Thompson RH , Kurup AN , Weisbrod AJ , Carter RE , Callstrom MR , Percutaneous cryoablation of solitary sporadic renal cell carcinomas. BJU Int. 2012;110 (11b ):E526–31.22584010
17. Schmit GD , Atwell TD , Callstrom MR , Farrell MA , Leibovich BC , Patterson DE , Percutaneous cryoablation of renal masses ≥ 3 cm: efficacy and safety in treatment of 108 patients. J Endourol. 2010;24 (8 ):1255–62.20380510
18. Atwell TD , Vlaminck JJ , Boorjian SA , Kurup AN , Callstrom MR , Weisbrod AJ , Percutaneous cryoablation of stage T1b renal cell carcinoma: technique considerations, safety, and local tumor control. J Vasc Interv Radiol. 2015;26 (6 ):792–9.25824313
19. Hasegawa T , Yamanaka T , Gobara H , Miyazaki M , Takaki H , Sato Y , Radiofrequency ablation versus cryoablation for T1b renal cell carcinoma: a multi-center study. Jpn J Radiol. 2018;36 (9 ):551–8.29968201
20. Atwell TD , Carter RE , Schmit GD , Carr CM , Boorjian SA , Curry TB , Complications following 573 percutaneous renal radiofrequency and cryoablation procedures. J Vasc Interv Radiol. 2012;23 (1 ):48–54.22037491
21. Hebbadj S , Cazzato RL , Garnon J , Shaygi B , Buy X , Tsoumakidou G , Safety considerations and local tumor control following percutaneous image-guided cryoablation of T1b renal tumors. Cardiovac Intervent Radiol. 2018;41 (3 ):449–58.
22. Kutikov A , Uzzo RG . The RENAL nephrometry score: a comprehensive standardized system for quantitating renal tumor size, location and depth. J Urol. 2009;182 (3 ):844–53.19616235
23. Ficarra V , Novara G , Secco S , Macchi V , Porzionato A , De Caro R , Preoperative aspects and dimensions used for an anatomical (PADUA) classification of renal tumours in patients who are candidates for nephron-sparing surgery. Eur Urol. 2009;56 (5 ):786–93.19665284
24. Venkatesan AM , Kundu S , Sacks D , Wallace MJ , Wojak JC , Rose SC , Practice guideline for adult antibiotic prophylaxis during vascular and interventional radiology procedures. J Vasc Interv Radiol. 2010;21 (11 ):1611–30.21029949
25. Patel IJ , Davidson JC , Nikolic B , Salazar GM , Schwartzberg MS , Walker TG , Consensus guidelines for periprocedural management of coagulation status and hemostasis risk in percutaneous image-guided interventions. J Vasc Interv Radiol. 2012;23 (6 ):727–36.22513394
26. Chung MS , Maxwell AW , Wang LJ , Mayo-Smith WW , Dupuy DE . Should renal mass biopsy be performed prior to or concomitantly with thermal ablation? J Vasc Interv Radiol. 2018;29 (9 ):1240–4.30075976
27. Filippiadis DK , Binkert C , Pellerin O , Hoffmann RT , Krajina A , Pereira PL . Cirse quality assurance document and standards for classification of complications: the cirse classification system. Cardiovasc Interv Radiol. 2017;40 (8 ):1141–6.
28. Higgins LJ , Hong K . Renal ablation techniques: state of the art. AJR Am J Roentgenol. 2015;205 (4 ):735–41.26204112
29. Patel N , King AJ , Breen DJ . Percutaneous image-guided cryoablation of small renal masses. Abdom Radiol (NY). 2016;41 (4 ):754–66.26915090
30. Ahmed M , Brace CL , Lee FT Jr , Goldberg SN . Principles of and advances in percutaneous ablation. Radiology. 2011;258 (2 ):351–69.21273519
31. Maxwell AW , Baird GL , Iannuccilli JD , Mayo-Smith WW , Dupuy DE . Renal cell carcinoma: comparison of RENAL nephrometry and PADUA scores with maximum tumor diameter for prediction of local recurrence after thermal ablation. Radiology. 2017;283 (2 ):590–7.27875105
32. Pierorazio PM , Johnson MH , Patel HD , Sozio SM , Sharma R , Iyoha E , Management of renal masses and localized renal cancer: systematic review and meta-analysis. J Urol. 2016;196 (4 ):989–99.27157369
33. Taylor BL , Stavropoulos SW , Guzzo TJ . Ablative therapy for small renal masses. Urol Clin North Am. 2017;44 (2 ):223–31.28411914
34. Schmit GD , Schenck LA , Thompson RH , Boorjian SA , Kurup AN , Weisbrod AJ , Predicting renal cryoablation complications: new risk score based on tumor size and location and patient history. Radiology. 2014;272 (3 ):903–10.24814178
|
PMC008xxxxxx/PMC8983094.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
101674571
44756
Abdom Radiol (NY)
Abdom Radiol (NY)
Abdominal radiology (New York)
2366-004X
2366-0058
27847998
8983094
10.1007/s00261-016-0967-5
NIHMS1783893
Article
Quantitative iodine content threshold for discrimination of renal cell carcinomas using rapid kV-switching dual-energy CT
Zarzour Jessica G. 1
Milner Desmin 1
Valentin Roberto 1
Jackson Bradford E. 2
Gordetsky Jennifer 34
West Janelle 5
Rais-Bahrami Soroush 41
Morgan Desiree E. 1
1 Department of Radiology, University of Alabama at Birmingham, 619 19th Street South, JTN 357, Birmingham, AL 35294, USA
2 Department of Preventative Medicine, University of Alabama at Birmingham, Birmingham, AL, USA
3 Department of Pathology, University of Alabama at Birmingham, Birmingham, AL, USA
4 Department of Urology, University of Alabama at Birmingham, Birmingham, AL, USA
5 School of Medicine, University of Alabama at Birmingham, Birmingham, AL, USA
Correspondence to: Jessica G. Zarzour; [email protected]
10 3 2022
3 2017
05 4 2022
42 3 727734
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Purpose:
Determine iodine content threshold discriminating papillary renal cell carcinomas (pRCC) from complex cysts (CCs) using rapid kV-switching dual-energy CT (rsDECT).
Materials and methods:
IRB-approved retrospective study of 72 consecutive patients with pathologic diagnosis of renal cell carcinoma, who underwent rsDECT from 2011 to 2015. Controls included consecutive patients with CC during same period. Iodine content of each pRCC (n = 27) was measured on rsDECT workstation for arterial (n = 15) or nephrographic phase (n = 12), and compared to iodine content for clear cell renal cell carcinomas (ccRCC, n = 46) and complex cysts (n = 54). An optimal iodine content threshold was estimated using logistic regressions and Youden’s J based on maximum specificity and sensitivity.
Results:
Iodine threshold of 1.28 mg/cc was optimal to discriminate between pRCCs and CCs for nephrographic phase (sens 1.0, spec 0.96, PPV 0.92, and NPV 1.0, AUC 0.997, acc 0.97, p < 0.0001). Iodine threshold of 1.22 mg/cc was the optimal cutoff value to discriminate between pRCCs and CCs in the arterial phase (sens 0.67, spec 0.97, PPV 0.91, NPV 0.85, AUC 0.76, and acc 0.84, p = 0.006). The optimal threshold to discriminate between ccRCCs and pRCCs was 1.85 mg/cc in the arterial phase (sens 0.87, spec 0.92, PPV 0.87, NPV 0.92, p < 0001) and 2.71 mg/cc in the nephrographic phase (sens 1.0, spec 1.0, PPV 1.0, NPV 1.0, p < 0.0001).
Conclusions:
Quantitative iodine values on rsDECT discriminate between papillary RCC and complex cysts, and between papillary RCC and clear cell RCC, the former addressing an important clinical challenge particularly when an unenhanced series has not been performed. These rsDECT thresholds differ from values derived from dual-source DECT technology.
Rapid switching dual energy CT
Papillary renal cell carcinoma
Clear cell renal cell carcinoma
Complex cysts
Iodine quantification
pmcThe diagnosis of neoplastic renal enhancement requires an increase in attenuation of 15–20 Hounsfield units (HU) between unenhanced and contrast-enhanced images on conventional multidetector CT (MDCT) [1]. Unlike clear cell renal carcinomas which typically show vivid enhancement in clinical practice, the distinction between less enhancing papillary renal carcinomas and complex cysts can be problematic, particularly when a lesion detected in the setting of a postcontrast-only CT measures >20 HU. In this situation, a second examination is required to determine enhancement as the lesion could represent a high attenuation (complex) benign renal cyst or an enhancing solid renal lesion [2]. This leads to increased cost, radiation exposure, and patient anxiety. Dual-energy CT (DECT) material-specific images have been utilized to calculate quantitative iodine concentration in focal lesions after contrast, and this quantification of iodine has been suggested as a surrogate for classic enhancement denoting neoplasm [3–5].
Four types of commercially available DECT are now available in the United States, but clinical evidence is substantial for only two technologies, the dual-source dual-energy CT (dsDECT) scanner and the single source rapid kV-switching DECT (rsDECT) [6, 7]. Previously published iodine threshold levels for distinguishing enhancing from non-enhancing lesions have varied significantly between rapid kV-switching and dual-source DECT. In 2011, Kaza et al. reported a 2.0 mg/cc iodine threshold to discriminate between enhancing and non-enhancing renal lesions utilizing rsDECT [8]. Also in 2011, Chandarana et al. first reported a threshold of 0.5 mg/cc [5], a value since confirmed with high accuracy for detection of malignancy in subsequent studies utilizing dsDECT [5, 9, 10]. Importantly, a recent phantom study comparing rsDECT and dsDECT at a single institution revealed significant differences in simulated monochromatic CT numbers and iodine concentration measurements between the two technologies [11]. To our knowledge, no larger series has confirmed the threshold for iodine quantification within renal neoplasms utilizing rsDECT since the original paper in 2011.
Although DECT enables quantitative measurement of iodine and diagnosis of an enhancing renal mass in a single postcontrast-only clinical setting, problems associated with variability in quantification of iodine between the two most commonly utilized types of DECT scanners, as well as variable investigative methods with early DECT renal lesion (such as combining all types of enhancing renal lesions) studies make further investigation imperative. The primary objective of this study was to determine the optimal quantitative iodine threshold to address the clinically relevant problem of distinguishing between papillary renal cell carcinomas and complex cysts utilizing a rsDECT scanner. Distinguishing between types of renal cell carcinomas was a secondary objective.
Materials and methods
This was a retrospective HIPAA compliant case–control study that was approved by our Institutional Review Board, with a waiver of informed consent.
Population
A search of the department of pathology surgical pathology database and the radiology information system identified consecutive patients with a pathologic diagnosis of papillary or clear cell renal cell carcinoma who underwent multiphasic abdominal rsDECT from 2011 to 2015. A control group comprised consecutive patients with complex cysts ≥1 cm demonstrating greater than that 1-year stability was identified through a search of the radiology information system over the same dates. Lesions classified as complex cysts were those not meeting strict criteria for a simple cyst (<10 HU on unenhanced image or <10 HU enhancement between unenhanced and postcontrast images) [1, 12–15]. Exclusions included lesions less than 1-cm-long axis diameter to mitigate the effect of volume averaging. Demographic data including age, gender, and weight were obtained from the electronic medical record.
Dual-energy CT technique
All subjects were scanned using a rapid kV-switching dual-energy 64 detector MDCT scanner (Discovery CT750 HD scanner, General Electric Healthcare, Waukesha, WI, USA) as part of a multiphasic abdominal examination. Standard unenhanced images were obtained using a conventional 120 kVp polychromatic beam multidetector CT technique (scan type, helical; detector coverage, 36 mm; slice thickness, 2.5 mm; interval, 2.5 mm; pitch, 1.375:1; speed, 55; and gantry rotation time, 0.8 s). Intravenous contrast was administered (Iohexol: Omnipaque 350, General Electronic Healthcare, Princeton NJ, USA) using a standardized weight-based dose injected at 2.8–4.0 cc/s rate for a fixed 30-s injection interval, followed by a 20 cc normal saline bolus injected at the same rate as the contrast. Images were acquired utilizing dual-energy technique during either the arterial (25- to 30-s delay) or nephrographic (100-s delay) phase. Note that in this retrospective study, arterial phase DECT images were acquired in patients suspected of having liver or pancreatic pathology, and venous/nephrographic phase DECT images were acquired in patients suspected of having a renal mass. The DECT images were acquired at 0.625 mm slice thickness, reconstructed to 2.5 mm slices, and sent to the departmental pictorial archive and communication system (PACS) (iSite version 4.0, Philips Medical Systems Best, Netherlands) for routine clinical viewing. The dual-energy 0.625 and 2.5 mm images were also sent to the independent dual-energy workstation (the Gemstone Spectral Image (GSI) Viewer, ADW 4.5, General Electric Healthcare, Milwaukee, WI) for further evaluation.
Image analysis
Lesion attenuation in HU was measured on the unenhanced conventional 120 kVp images and 70 keV simulated monoenergetic rsDECT images on the departmental PACS, using uniform ROI placement in matching locations. Lesion iodine content in mg/cc was recorded using the rsDECT workstation during simultaneous viewing of the PACS to ensure matching placement of ROIs in the lesions. For heterogeneous lesions, the ROI excluded necrotic areas and was placed on the portion of the lesion that was most avidly enhancing. For homogeneous lesions, the ROI included as much of the lesion as possible. Lesion size was measured on PACS. Measurements were made by radiology residents and were verified by a fellowship-trained abdominal imaging radiologist with 6 years of experience. Each DECT evaluation for this study took less than 5 min per reader; training for a radiologist already accustomed to placing ROIs is estimated at 30 min to learn how to use the independent workstation to perform this analysis.
Pathologic analysis
A board certified fellowship-trained, genitourinary pathologist reviewed all tumors and provided RCC subtype classification based on hematoxylin and eosin preparation and immunohistochemical staining. Fuhr-man grade for each tumor was analyzed to include low-grade RCC (Fuhrman I–II) and high-grade RCC (Fuhrman III–IV).
Statistical analysis
The frequencies and percentages for categorical variables as well as the mean, standard deviation, minimum, and maximum values were calculated for continuous variables. Box and whisker plots were generated to visualize group differences for the distributions of quantitative iodine content. To estimate an optimal iodine content threshold, logistic regressions were used with Firth’s penalized likelihood approach. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy, and the area under the curve (AUC) were calculated. The AUC was further evaluated by comparison with the null value of 0.5. In addition, Youden’s J optimum threshold criteria based on maximum specificity and sensitivity were used to determine the cutoff for distinguishing pRCC from CC and pRCC from ccRCC [16]. Receiver operating characteristic (ROC) curves were generated for each measurement. The correlation between the concentration of tumor iodine and tumor grade was investigated using Kruskal–Wallis one-way analysis of variance. Statistical analysis was performed using SAS version 9.4 (2013, Cary, NC) with a two-sided 0.05 level of significance. No formal power calculations were performed as this was largely an exploratory study for threshold identification across multiple variables of interest. While small sample size may present an issue, the purpose of the analysis was descriptive in nature and intended to be exploratory and hypothesis generating. A dedicated faculty statistician performed all statistical analyses.
Results
The study population included 30 women and 42 men, mean age 60.5 years (range, 26–81 years) and mean weight 86.8 kg (range, 48.5–131.5 kg). Flow chart of participants is shown in Table 1. Twenty-seven subjects had pRCC with the mean size of 2.7 cm (±0.97 cm). Forty-five subjects had clear cell RCC with the mean size of 5.1 cm (±3.4 cm). Of the patients with papillary RCCs, 15 had arterial phase and 12 had nephrographic phase rsDECT acquisition. Of the patients with clear cell RCC, 26 had arterial phase and 20 had nephrographic phase rsDECT acquisition.
One patient with a clear cell RCC underwent rsDECT evaluation in both the arterial and nephrographic phases. The mean ± standard deviation time between the rsDECT examination and pathologic diagnosis was 49.9 ± 89.5 days. Thirty-five patients underwent radical nephrectomy, 31 had partial nephrectomy and 6 patients had a biopsy.
The control group included 36 women and 18 men, mean age 64.9 years (range, 41–88) and mean weight 82.1 kg (range, 49.0–127.5 kg). Fifty-four CC were evaluated, mean size 2.2 cm ± 1.6 cm. Twenty-nine had arterial phase and 25 had nephrographic phase rsDECT acquisition. All included CC demonstrated size stability for greater than 1 year. The mean Hounsfield Unit attenuation of the CC was 32.8 ± 20.6 on the unenhanced series, 46.1 ± 22.5 in the arterial phase, and 41.6 ± 23.0 in the nephrographic phase.
The mean quantitative iodine content for pRCC (Fig. 1) was 1.30 mg/cc ± 0.80 in the arterial phase and 1.9 mg/cc ± 0.51 in the nephrographic phase. The mean quantitative iodine content for ccRCC (Fig. 2) was 4.56 mg/cc ± 2.2 in the arterial phase and 5.71 mg/cc ± 4.3 in the nephrographic phase. The mean quantitative iodine content for CC (Fig. 3) was 0.71 mg/cc ± 0.35 in the arterial phase and 0.82 mg/cc ± 0.31 in the nephro-graphic phase. These differences were statistically significant (p < 0.0001). The enhancement and quantitative iodine values are summarized in Tables 2, 3 and Figure 4.
The optimal quantitative iodine content threshold for distinguishing pRCC from CC differed for the arterial and venous phases. In the arterial phase, the iodine content threshold was 1.22 mg/cc (sensitivity 0.67, specificity 0.97, PPV 0.91, NPV 0.85, accuracy 0.84, AUC 0.76, p = 0.006). There were three false positive cysts with iodine content of 1.53, 1.27 and 1.26 mg/cc and five false-negative papillary RCCs with iodine content of 0.61, 0.64, 0.69, 0.95 and 1.05 mg/cc in the arterial phase (Fig. 5). Higher accuracy was found in the nephrographic phase, where the optimal iodine threshold was 1.28 mg/cc (sensitivity 1.0, specificity 0.96, PPV 0.92, NPV 1.0, accuracy 0.97%, AUC 0.997, p < 0.0001). There was 1 false positive cyst (iodine content 1.29 mg/cc) in the nephrographic phase and no false-negative papillary RCCs. Various thresholds were investigated and are shown in Table 4.
The quantitative iodine content threshold for distinguishing papillary RCC from clear cell RCC was also investigated. An quantitative iodine content threshold of 1.85 mg/cc in the arterial phase and 2.71 mg/cc in the nephrographic phase distinguished papillary RCC from clear cell RCC with a high degree of accuracy (sensitivity 0.87, specificity 0.92, PPV 0.87, NPV 0.92, accuracy 0.90, p < 0.0001 and sensitivity 1.0, specificity 1.0, PPV 1.0, NPV 1.0, accuracy 1.0, p < 0.0001, respectively).
There was no significant correlation between quantitative iodine content and Fuhrman grade. For ccRCC in the arterial phase, the mean iodine content in Fuhrman grade I/II was 4.86 mg/cc ± 2.44 and in Fuhrman grade III/IV was 4.14 mg/cc ± 1.7 (p = 0.39). For ccRCC in the nephrographic phase, the mean iodine content in Fuhrman grade I/II was 4.91 mg/cc ± 3.71 and in Fuhrman III/IV was 6.51 mg/cc ± 4.89 (p = 0.42). For pRCC in the arterial phase, the mean iodine content in Fuhrman grade I/II was 1.41 mg/cc ± 1.01 and in Fuhrman grade III/IV was 1.35 mg/cc ± 2.39 (p = 0.82). For pRCC in the nephrographic phase, the mean iodine content in Fuhrman grade I/II was 2.0 mg/cc ± 0.61 and in Fuhrman grade III/IV was 1.75 mg/cc ± 0.58 (p = 0.94).
Discussion
Papillary renal cell carcinoma, representing 15–20% of renal cell carcinomas (RCCs) [17, 18], may only enhance minimally between pre- and postcontrast series. Therefore, while clear cell RCCs rarely pose a diagnostic dilemma, papillary RCCs may be difficult to distinguish from complex cysts with only a postcontrast single-energy MDCT series.
Our results suggest that rsDECT is highly accurate in distinguishing papillary RCCs from complex cysts, particularly in the nephrographic phase. Confidently, identifying minimally enhancing tumors with postcontrast-only examinations using iodine material density quantification overcomes an important limitation of single-energy MDCT. However, a recent phantom study showed a significant variance in iodine concentration values and monochromatic CT numbers between the two major dual-energy hardware implementations: dual-source and rapid kV-switching DECT [11]. The greatest discrepancies were at lower lesion iodine concentrations, lower monochromatic energies, and in smaller lesions, with higher values observed on the rsDECT vs. the dsDECT systems [11]. This is particularly important for evaluating papillary RCCs, which generally have lower iodine concentrations than clear cell RCCs. Small renal lesions are often incidentally discovered on routine or emergency single-phase CTs, so determining reliable iodine content thresholds for each DECT platform is essential to validating their use in these settings [19].
Variable iodine quantification between the two major DECT platforms is the likely reason that the previous studies have shown differing optimal iodine content enhancement thresholds. Initial work on an rsDECT unit evaluated several iodine content enhancement thresholds for lesions (defined as > 20 HU increase on single-energy CT) [8]. A 2.0 mg/cc threshold was concluded as optimal for identifying enhancement (sensitivity 90%, specificity 93.7%, PPV 81.8%, NPV 96.7, accuracy 92.8%) [8]. In this study, there were 20 enhancing lesions, including only two RCCs (subtype unspecified), eight other enhancing lesions with no histologic diagnosis, six oncocytomas, three AMLs, and one Bosniak category III lesion [8]. In our larger study, including 72 patients with pathologically proven renal cancers, we could subcategorize levels of enhancement based on histology. The mean iodine content was significantly higher in clear cell RCC than papillary RCC, no matter the phase of contrast. For the 27 histologically proven papillary RCCs, we found a lower threshold of 1.28 mg/cc for the nephrographic phase for accurately differentiating papillary RCCs from complex cysts, although the arterial phase threshold of 1.22 mg/cc was less accurate. The delayed, progressive enhancement of papillary RCCs likely accounts for this difference [20]. This same factor likely accounts for the five false-negative cases of papillary RCC in the arterial phase acquisition group; there were no false-negatives in the nephrographic phase acquisition group. We believe that including more strongly enhancing clear cell RCCs, which are not as challenging to diagnose, skews the threshold in other series and that lowering threshold in our more uniform population leads to better characterization of more clinically challenging lesions. In our series, only 8 of the 27 papillary RCCs had iodine content greater than 2.0 mg/cc and therefore 19 would have been incorrectly diagnosed as non-neoplastic. Because distinguishing papillary RCCs from complex cysts presents the greatest diagnostic challenge, we believe that the previously reported 2.0 mg/cc should not be used for renal masses where papillary RCC is considered.
Research on dsDECT has shown an optimal iodine enhancement threshold of 0.5 mg/cc [5, 9, 10] for distinguishing neoplastic from non-neoplastic lesions. In a study of 32 malignant renal tumors and 48 benign renal cysts, the iodine threshold of 0.5 mg/cc had a sensitivity of 100%, a specificity of 97.7%, a PPV of 97.2%, and a NPV of 100% [9]. Another study of 36 malignant renal tumors and 36 benign tumors also used 0.5 mg/cc as the iodine content threshold (sensitivity 100.0%, specificity 94.4%, PPV 94.7%, NPV 100%) [10]. Our study has established a threshold for rsDECT and has shown that published dsDECT values cannot be used with rsDECT. Further confirmation of our thresholds by other investigators with rsDECT platforms will be helpful. To our knowledge, there are no known studies that have addressed software version differences in quantitative iodine concentrations within a single vendor’s technology, and this may also be an important avenue of study for all users of DECT.
Papillary and clear cell subtypes of RCC differ in biologic behavior, prognosis, and response to targeted therapies [21, 22]. Papillary RCCs enhance at a low level and progressively, while clear cell carcinomas enhance avidly [20,23]. One study using dsDECT showed a threshold for distinguishing papillary from clear cell RCC as 0.9 mg/cc [24]. Our study using rsDECT showed that a threshold of 1.85 mg/cc for the arterial phase and 2.71 mg/cc for the nephrographic phase differentiated between these two subtypes, again illustrating the higher quantitative iodine thresholds on rsDECT than on dsDECT.
The limitations of this study include its retrospective nature. While the RCCs in our study were histologically proven, the complex cysts were concluded as benign based on stability of over one year rather than histology, and a further limitation is that one-year stability may not be enough to prove benign nature in slow growing cystic renal tumors. The lack of correlation with Fuhrman grade might have been affected by the varied postcontrast timing in our population, an unavoidable aspect of our retrospective study where the clinical indication dictated the phase of acquisition for which DECT was applied. Finally, our results do not apply to dsDECT, so radiologists should be careful to apply values appropriate to their own platforms. While our study was limited to a single commercially available platform, we were able to control for extraneous between-platform variability. Future studies should consider the identification and application of thresholds across multiple platforms.
In conclusion, rsDECT accurately discriminates between papillary RCCs and complex cysts and between papillary and clear cell RCCs, with the former being the more important clinical challenge. A threshold of 1.28 mg/cc can distinguish papillary RCCs from complex cysts with a single nephrographic phase postcontrast series. A threshold above 2.7 indicates clear cell subtype rather than papillary subtype of renal cell carcinoma. These thresholds differ substantially from thresholds previously reported with dsDECT and an earlier rsDECT study of a smaller population with more heterogeneous lesion types.
Funding
No funding was received for this study.
Conflict of interest
Desiree E. Morgan has received research support from GE Healthcare and was a one-time consultant for educational materials for DECT. The other authors declare that they have no conflict of interest.
Fig. 1. 60-year-old man with pathologically proven papillary RCC. A Axial CT (0.6 cm slice thickness) in the nephro-graphic phase shows the homogeneously hypoenhancing mass to measure 77 HU. B Axial rsDECT (0.6 cm slice thickness) iodine density image shows the quantitative iodine content with the papillary RCC to be 2.5 mg/cc.
Fig. 2. 67-year-old man with pathologically proven clear cell RCC. A Axial CT (0.6 cm slice thickness) urogram demonstrates a characteristic enhancing mass measuring 89 HU. B Axial rsDECT (0.6 cm slice thickness) iodine density image shows the quantitative iodine content within the clear cell RCC to be 2.9 mg/cc.
Fig. 3. 58-year-old female with a complex cyst that has been stable for 3 years. A Axial CT (0.6 cm slice thickness) in the nephrographic phase demonstrates a well-circumscribed lesion in the mid-right kidney that measures 68 HU. B Axial rsDECT (0.6 cm slice thickness) iodine density image shows the quantitative iodine content within the complex cyst to be 0.92 mg/cc.
Fig. 4. Box and whisker plots of the quantitative iodine content in papillary RCC vs. clear cell RCC vs. complex cysts in the arterial and nephrographic phases.
Fig. 5. 52-year-old man with pathologically proven papillary RCC. The rsDECT acquisition was performed in the arterial phase and lead to a false-negative result. The quantitative iodine content of the lesion measured 0.69 mg/cc. The maximal enhancement of papillary RCCs occurs in the nephro-graphic phase.
Table 1. Flow chart of participants
568 patients with pathologic diagnosis of Renal Cell Carcinoma 383 patients with “complex” or “hyperdense” cysts evaluated on rsDECT
↓ ↓
90 patients with rsDECT prior to resection 164 patients with cysts not meeting criteria for simple cysts and >1cm
↓ ↓
45 Clear cell RCC
27 Papillary RCC
4 Unclassified
4 Clear cell papillary RCC
4 Chromophobe
5 Oncocytoma 54 patients with greater than 1 year follow up
↓
Clear cell RCC and Papillary RCCs included in analysis (n=72)
Table 2. Quantitative iodine content (mg/cc) in clear cell RCC (ccRCC), papillary RCC (pRCC), and complex cysts (CCs)
ccRCC (mean ± SD) Number of ccRCC pRCC (mean ± SD) Number of pRCC CCs (mean ± SD) Number of CCs p value
Arterial phase 4.56 ± 2.2 26 1.30 ± 0.80 15 0.71 ± 0.35 29 <0.0001
Nephrographic phase 5.71 ± 4.3 20 1.90 ± 0.51 12 0.82 ± 0.31 25 <0.0001
Table 3. Enhancement in HU in clear cell RCC (ccRCC), papillary RCC (pRCC), and complex cysts (CCs)
UE ccRCC HU (mean ± SD) ccRCC enhancement (mean ± SD) Number of ccRCC UE pRCC HU (mean ± SD) pRCC enhancement (mean ± SD) Number of pRCC UE CC HU (mean ± SD) CCs enhancement (mean ± SD) Number of CCs
Arterial phase 140.0 ± 49.5 110.7 ± 51.0* 26 58.3 ± 19.4 27.6 ± 20.9* 15 46.1 ± 22.5 8.9 ± 6.8* 29
Nephrographic phase 121.3 ± 34.9 96.4 ± 26.1* 20 73.6 ± 13.3 44.3 ± 13.8* 12 41.6 ± 23.0 8.0 ± 5.9* 25
UE, unenhanced
* The differences between these three groups was statistically significant with p < 0.0001
Table 4. Evaluation of varying iodine content thresholds in distinguishing papillary RCC from complex cysts
Threshold (mg/cc) Sensitivity Specificity PPV NPV
Arterial phase 0.95 0.73 0.66 0.52 0.83
1.09 0.67 0.93 0.63 0.82
1.22 0.67 0.97 0.91 0.85
1.27 0.40 0.97 0.86 0.76
1.62 0.33 1.00 1.00 0.74
Nephrographic phase 1.01 1.00 0.72 0.63 1.00
1.15 1.00 0.84 0.75 1.00
1.28 1.00 0.96 0.92 1.00
1.39 0.83 1.00 1.00 0.93
2.03 0.50 1.00 1.00 0.81
Bold value signifies the optimal thresholds for iodine quantification in the arterial phase and nephrographic phase
Compliance with ethical standards
Ethical approval All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.
Informed consent Informed consent was obtained from all individual participants included in the study.
References
1. Israel GM , Bosniak MA (2005) How I do it: evaluating renal masses. Radiology 236 (2 ):44l–450
2. Suh M , (2003) Distinction of renal cell carcinomas from high-attenuation renal cysts at portal venous phase contrast-enhanced CT. Radiology 228 (2 ):330–334 12819331
3. Feuerlein S , (2012) Iodine quantification using dual-energy multidetector computed tomography imaging: phantom study assessing the impact of iterative reconstruction schemes and patient habitus on accuracy. Invest Radiol 47 (11 ):656–661 22996313
4. Neville AM , (2011) Detection of renal lesion enhancement with dual-energy multidetector CT. Radiology 259 (1 ):173–183 21292866
5. Chandarana H , (2011) Iodine quantification with dual-energy CT: phantom study and preliminary experience with renal masses. AJR Am J Roentgenol 196 (6 ):W693–W700 21606256
6. Morgan DE (2014) Dual-energy CT of the abdomen. Abdom Imaging 39 (1 ):108–134 24072382
7. Marin D , (2014) State of the art: dual-energy CT of the abdomen. Radiology 271 (2 ):327–342 24761954
8. Kaza RK , (2011) Distinguishing enhancing from nonenhancing renal lesions with fast kilovoltage-switching dual-energy CT. AJR Am J Roentgenol 197 (6 ):1375–1381 22109292
9. Mileto A , (2014) Accuracy of contrast-enhanced dual-energy MDCT for the assessment of iodine uptake in renal lesions. AJR Am J Roentgenol 202 (5 ):W466–W474 24758682
10. Ascenti G , (2013) Distinguishing enhancing from nonenhancing renal masses with dual-source dual-energy CT: iodine quantification versus standard enhancement measurements. Eur Radiol 23 (8 ):2288–2295 23479222
11. Mileto A , (2015) Virtual monochromatic images from dual-energy multidetector CT: variance in CT numbers from the same lesion between single-source projection-based and dual-source image-based implementations. Radiology 279 :269–277 26536403
12. Coleman BG , (1984) Hyperdense renal masses: a computed tomographic dilemma. AJR Am J Roentgenol 143 (2 ):291–294 6611058
13. Jonisch AI , (2007) Can high-attenuation renal cysts be differentiated from renal cell carcinoma at unenhanced CT? Radiology 243 (2 ):445–450 17456870
14. Birnbaum BA , (2002) Renal cyst pseudoenhancement: evaluation with an anthropomorphic body CT phantom. Radiology 225 (1 ):83–90 12354989
15. Silverman SG , (2007) Hyperattenuating renal masses: etiologies, pathogenesis, and imaging evaluation. Radiographics 27 (4 ):1131–1143 17620471
16. Ruopp MD , (2008) Youden Index and optimal cut-point estimated from observations affected by a lower limit of detection. Biom J 50 (3 ):419–430 18435502
17. Amin MB , (1997) Papillary (chromophil) renal cell carcinoma: histomorphologic characteristics and evaluation of conventional pathologic prognostic parameters in 62 cases. Am J Surg Pathol 21 (6 ):621–635 9199639
18. Delahunt B , Eble JN (1997) Papillary renal cell carcinoma: a clinicopathologic and immunohistochemical study of 105 tumors. Mod Pathol 10 (6 ):537–544 9195569
19. Silverman SG , (2008) Management of the incidental renal mass. Radiology 249 (1 ):16–31 18796665
20. Kim JK , (2002) Differentiation of subtypes of renal cell carcinoma on helical CT scans. AJR Am J Roentgenol 178 (6 ): 1499–1506 12034628
21. Teloken PE , (2009) Prognostic impact of histological subtype on surgically treated localized renal cell carcinoma. J Urol 182 (5 ): 2132–2136 19758615
22. Leibovich BC , (2010) Histological subtype is an independent predictor of outcome for patients with renal cell carcinoma. J Urol 183 (4 ):1309–1315 20171681
23. Zhang J , (2007) Solid renal cortical tumors: differentiation with CT. Radiology 244 (2 ):494–504 17641370
24. Mileto A , (2014) Iodine quantification to distinguish clear cell from papillary renal cell carcinoma at dual-energy multidetector CT: a multireader diagnostic performance study. Radiology 273 (3 ):813–820 25162309
|
PMC008xxxxxx/PMC8983095.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
101300401
37336
J Robot Surg
J Robot Surg
Journal of robotic surgery
1863-2483
1863-2491
32112185
8983095
10.1007/s11701-020-01056-9
NIHMS1783860
Article
Perioperative outcomes of laparoscopic, robotic, and open approaches to pheochromocytoma
Fang Andrew M. 1http://orcid.org/0000-0002-2390-7575
Rosen Jennifer 1
Saidian Ava 1
Bae Sejong 23http://orcid.org/0000-0001-9982-3223
Tanno Fabio Y. 4
Chambo Jose L. 4
Bloom Jonathan 5
Gordetsky Jennifer 167http://orcid.org/0000-0002-0745-2160
Srougi Victor 4
Phillips John 5
Rais-Bahrami Soroush 138http://orcid.org/0000-0001-9466-9925
1 Department of Urology, University of Alabama at Birmingham, Faculty Office Tower 1107, 510 20th Street South, Birmingham, AL 35233, USA
2 Division of Preventative Medicine, Department of Medicine, University of Alabama at Birmingham, Birmingham, AL, USA
3 O'Neal Comprehensive Cancer Center at UAB, University of Alabama at Birmingham, Birmingham, AL, USA
4 Department of Urology, Hospital das Clínicas de São Paulo, University of São Paulo Medical School, São Paulo, Brazil
5 Department of Urology, New York Medical College, Valhalla, NY, USA
6 Department of Pathology, University of Alabama at Birmingham, Birmingham, AL, USA
7 Present Address: Department of Pathology, Vanderbilt University, Nashville, TN, USA
8 Department of Radiology, University of Alabama at Birmingham, Birmingham, AL, USA
Andrew M. Fang and Jennifer Rosen equally contributed to this study.
Author contributions AMF: conception and design of the study, acquisition of data, interpretation of the data, drafting the article, final approval. JR: conception and design of the study, acquisition of data, interpretation of the data, drafting the article, final approval. AS: revising, final approval. SB: analysis and interpretation of data, revising, final approval. FYT: acquisition of data, revision, final approval. JLC: acquisition of data, revision, final approval. JB: acquisition of data, revision, final approval. JG: conception and design of the study, revision, final approval. VS: conception and design of the study, revision, final approval. JP: conception and design of the study, revision, final approval. SR-B: conception and design of the study, analysis and interpretation of the data, revision, final approval.
✉ Soroush Rais-Bahrami, [email protected]
10 3 2022
12 2020
28 2 2020
05 4 2022
14 6 849854
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
While multiple studies have demonstrated that minimally invasive surgical (MIS) techniques are a safe and efficacious approach to adrenalectomy for pheochromocytomas (PC), these studies have only been small comparative studies. The aim of this multi-institutional study is to compare perioperative outcomes between open and MIS, stratified by robotic and conventional laparoscopic, techniques in the surgical management of PC. We retrospectively evaluated patients who underwent adrenalectomy for PCs from 2000 to 2017 at three different institutions. Clinical, perioperative, and pathologic parameters were analyzed using t test, Chi square, and Fisher exact statistical measures. Of the 156 adrenalectomy cases performed, 26 (16.7%) were with an open approach and 130 (83.3%) using MIS techniques. Of the MIS procedures, 41 (31.5%) were performed robotically and 89 (68.5%) performed laparoscopically without robotic assistance. Demographic and clinical parameters were similar between the open and MIS groups. Patients, who underwent MIS procedure had a lower complication rate (p = 0.04), shorter hospitalization (p = 0.02), shorter operative time (p < 0.001), and less blood loss (p = 0.002) than those who underwent open surgical resection. Conventional laparoscopic and robotic operative approaches resulted in similar complication rates, length of hospitalization, and blood loss. Our study is one of the largest cohorts comparing the perioperative outcomes between conventional laparoscopic and robotic adrenalectomies in patients with PC. Our results support that MIS techniques have potentially lower morbidity compared to open techniques, while laparoscopic and robotic approaches have similar perioperative outcomes.
Pheochromocytoma
Laparoscopy
Robotic
Minimally invasive surgery
Adrenal
pmcIntroduction
Pheochromocytomas (PC) are metabolically active tumors derived from the neural crest cells in the adrenal medulla. Excessive catecholamine release from these tumors can present with a wide range of symptomology ranging anywhere from asymptomatic to polypharma-resistant hypertension, tachycardia, and palpitations.
Since the 1990s, laparoscopic adrenalectomy has largely replaced open surgery as the standard for adrenalectomy in well-selected patients at many centers [1]. Despite the decreased length in hospitalizations and fewer complications that laparoscopic surgeons offer, there are still concerns regarding its ability to safely resect large adrenal tumors and pheochromocytomas [2, 3]. Over the last 10 years, the advent of robotic adrenalectomy utilizing the da Vinci robotic surgical platform (Intuitive Surgical, Sunnyvale, CA, USA) has provided surgeons with a three-dimensional operative view and increased maneuverability using the EndoWrist® technology further modifying laparoscopic surgical approaches. Initial reports cited the learning curve of robotic adrenalectomy, longer operative time, and increased healthcare costs as potential disadvantages when compared to traditional laparoscopy. There has also been suggestion that the lack of tactile feedback when using robotic-assisted laparoscopy can potentially result in undue tumor manipulation and inadvertent catecholamine release. This could increase the risk of hemorrhage, cardiovascular instability, and intraoperative hypertensive crisis [4].
While multiple studies have demonstrated comparable safety and efficacy between the robotic and laparoscopic approaches; these have all been small comparative cohort studies [5–9]. The aim of this multi-institutional study is to compare perioperative outcomes between open and minimally invasive surgery (MIS), furthermore stratifying MIS as robotic and conventional laparoscopic techniques, in the operative management of preoperatively recognized PCs.
Methods
Study population and parameters
After IRB approval was obtained at all participating centers, patients who underwent adrenalectomy for pathologically confirmed PC (PC) were retrospectively evaluated at University of Alabama at Birmingham (Birmingham, AL), New York Medical College (Valhalla, NY) and Hospital das Clínicas de São Paulo (São Paulo, Brazil), from 2000 to 2017. No partial adrenalectomies were included in our multi-institutional cohort or analysis. We obtained all available demographic, familial, and clinical data. Adjusted Charlson comorbidity index (CCMI) scores were calculated using an Excel-based application. Operative report, intraoperative anesthesia reports, Pheochromocytoma of the Adrenal gland Scaled Scores (PASS), inpatient complications, postoperative hospitalization, and pathologic reports were examined [10]. Hemorrhage was defined as a patient requiring transfusion in the setting of acute blood loss anemia. High-grade perioperative complications were defined as Clavien–Dindo III and greater. Hypertensive adrenal crisis was defined as symptomatic ≥ stage 2 hypertension and hemodynamic compromise requiring control with vasoactive agents. Pathologic specimen size was defined as the maximum diameter of the surgical specimen (cm), while tumor size was defined as the maximum diameter of the tumor (cm) itself. Our primary outcomes were perioperative morbidity that we assessed with variables including estimated blood, prevalence of high-grade perioperative complications, and length of hospitalization.
Statistical analysis
Continuous variables were assessed with unpaired Student t tests between two groups. Chi square and Fisher exact tests were used to assess categorical variables among different groups. All p values were two-sided with a predetermined alpha level set at 0.05.
Results
Between 2000 and 2017, 153 patients underwent 156 adrenalectomy cases for inclusion in this study (Table 1). Patients who underwent MIS and open adrenalectomies were similar in terms of age, gender, CCMI, tumor side, and BMI (Table 2). However, differences were noted between the conventional laparoscopic and robotic surgery groups. Patients who underwent conventional laparoscopic surgery tended to be younger (46.2 ± 17.9 versus 55.9 ± 15.4 years old, respectively, p = 0.004), more likely to have history of hypertensive crisis (39.3% versus 19.5%, respectively, p = 0.03), less likely to have a cardiovascular history (9.1% versus 24.4%, respectively, p = 0.02), and have a lower body mass index (BMI) (24.5 ± 4.9 versus 29.8 ± 6.5, respectively, p < 0.001) (Table 3).
In evaluation of perioperative parameters, the MIS group had faster operative time (174.6 ± 76.4 versus 260.4 ± 117.6 min, respectively, p < 0.001), less blood loss (150 ± 318 versus 439 ± 423 ml, respectively, p = 0.002) and required less intraoperative fluid (3466 ± 1079 versus 4843 ± 1957 ml, respectively, p = 0.006) when compared to patients undergoing open surgical approach to adrenalectomy. Intraoperative hemodynamics were consistent between MIS and open surgical cohorts, the exception being the MIS group that had a higher mean ending systolic blood pressure than the open surgical group (88.1 ± 14.7 versus 77.0 ± 10.6 mmHg, p < 0.02) (Table 2). Compared to the robotic approach, laparoscopy had a faster operative time (157.9 ± 53.1 versus 210.4 ± 103.0 min, respectively, p = 0.004). Otherwise, laparoscopic and robotic surgical approaches had similar blood, intraoperative fluid use, and intraoperative hemodynamics (Table 3).
While the MIS group had a lower perioperative complication compared to an open technique rate (26.2% versus 46.2%, respectively, p = 0.04), the laparoscopic and robotic approaches had similar perioperative complication rates. The rate of high-grade complications was similar regardless of approach. Additionally, patients who underwent MIS procedures also had shorter intensive care unit (ICU) stays (1.2 ± 1.5 versus 2.1 ± 2.0 days, respectively, p = 0.06) and shorter hospitalizations (3.5 ± 3.4 versus 5.2 ± 3.0 days, respectively, p = 0.02). The laparoscopic group had longer ICU stays than the robotic group (1.5 ± 1.6 versus 0.7 ± 1.2 days, respectively, p = 0.002), but no significant difference in length of hospitalization (3.7 ± 3.8 versus 3.0 ± 2.25 days, respectively, p = 0.203). Regarding tumor size, the MIS group had smaller tumors than those undergoing an open approach (79.9 ± 106.6 versus 215 ± 178.1 g, respectively, p < 0.04), but similar specimen size (7.6 ± 2.1 versus 8.7 ± 3.0 cm, respectively, p = 0.21). Of the MIS group, the laparoscopic approach had smaller specimen size than robotic (6.9 ± 1.9 versus 8.1 ± 2.1 cm, respectively, p = 0.03), but similar tumor size (4.6 ± 2.5 versus 6.2 ± 8.4 cm, respectively, p = 0.25). Positive margin on pathology was similar between both MIS and open (6.7% vs 25%, respectively, p = 0.1) and between robotic and laparoscopic approaches (5.6% versus 7.4%, respectively, p = 1). No association was found in PASS between the MIS and open groups (4.4 ± 3.2 versus 4.6 ± 4.1, respectively, p = 0.864) and laparoscopic and robotic subgroups (4.8 ± 3.6 versus 2.7 ± 2.1, respectively, p = 0.7).
Overall, there were a total of 46 complications (30%) throughout the cohort. 10% of these complications were high-grade (Clavien ≥ 3a). Three Clavien 4a complications were noted, two of which stemmed from postoperative myocardial infarction (MI). However, there were no Clavien 4b complications or deaths in our multi-institutional cohort. Postoperative hypertension was the most common post-adrenalectomy complication (30% of all complications) followed by pulmonary-related complications (26%), and hypotension (17%). Two adrenalectomies were converted from a laparoscopic to open approach. One of these conversions was complicated by a non-ST segment elevation MI (Clavien 4a), while the other patient’s postoperative course was complicated by respiratory failure and hypotension (Clavien 2).
Discussion
In this study of 153 patients with 156 PCs, we found that there was no significant difference in age, gender, CCMI, tumor size, or BMI among the patients that were treated with an open surgical approach to adrenalectomy compared to those treated with an MIS approach. The MIS group was found to have faster operative times (p < 0.001), less blood loss (p = 0.002), and shorter hospitalizations (p = 0.021). Overall, the MIS group had a lower perioperative complication rate of any grade than the open group (p = 0.041). However, most complications that occurred were either Clavien 1 or 2 including postoperative hypertension and minor pulmonary-related complications. Both groups had similar rates of high-grade complications.
On the other hand, patients undergoing a laparoscopic approach had different preoperative characteristics compared to robotic in terms of age, history of hypertensive crisis, cardiovascular history and BMI. The lower age and BMI seen in the laparoscopic group could be attributed to younger and less obese patients being deemed less complex surgical candidates and more amenable to a laparoscopic approach. Furthermore, the higher incidence of cardiovascular history and hypertensive crises in the laparoscopic group could be attributed to surgeon preference for the tactile feedback of laparoscopy to limit catecholamine spillage during resection. Furthermore, patients undergoing laparoscopic adrenalectomy had faster operative time compared to robotic (p = 0.004). This difference could be attributed either to surgeon experience or the docking time required for the robotic approach. Nevertheless, both laparoscopic and robotic approaches had similar rates of complications and length of hospitalization.
Our findings are congruent with the established literature on the advantages that laparoscopic adrenalectomy offers over an open approach [11–13]. In Lee et al.’s investigation of 669 patients in the Veterans Affairs National Surgical Quality Improvement Program, who underwent either laparoscopic or open adrenalectomy, they reported that open procedures had increased operative times (3.9 ± 1.8 h versus 2.9 ± 1.3 h, p < 0.001), greater length of stay (9.4 ± 11.0 days versus 4.1 ± 4.7 days, p < 0.0001), and increased 30-day morbidity rates (17.4% versus 3.6%, p < 0.0001) compared to the laparoscopic approach [12]. While these national databank analyses also provided excellent descriptions of patient-specific factors associated with perioperative complications after adrenal surgery including American Society of Anesthesiology (ASA) score and diabetes, they failed to address disease-specific data that could potentially affect perioperative outcomes. To address these shortcomings, Chen et al. analyzed 653 laparoscopic adrenalectomies performed across a 24-year period. They found that diagnosis of PC [odds ratio (OR), 4.31 (95% confidence interval (CI) 1.43–13.05), p = 0.01] and tumor size of 6 cm or greater [OR 2.47 (95% CI 1.05–5.78), p = 0.04] were independent risk factors associated with increased perioperative compilations [3]. Nevertheless, despite the risks associated with operating on PCs, our data further support an MIS approach as a safer option for adrenalectomy.
With the advent of the da Vinci system, the popularity of robotic surgery across many traditionally laparoscopic surgeries has risen in popularity. However, the feasibility and safety of robotic adrenalectomy have only been studied in small cohort studies. Brandao et al. pooled 9 studies that included 277 robotic-assisted and 323 laparoscopic adrenalectomies into their meta-analysis. They found no statistical difference in postoperative complication rate between the two groups, but noted a significantly longer length of hospitalization [weighted mean difference (WMD) − 0.43; 95% CI − 0.56 to − 0.30; p < 0.01] and higher estimated blood loss (WMD − 18.21; 95% CI − 29.11 to − 7.32; p < 0.01) in the traditional laparoscopic group [14]. In the first study comparing laparoscopic to robotic adrenalectomy for PCs, Aliyev et al. compared 88 patients, who underwent either procedure. They noted no difference in the morbidity or mortality between the two techniques, but the laparoscopic approach was associated with a longer length of hospitalization than the robotic (1.8 ± 0.1 versus 1.2 ± 0.1 days, respectively, p = 0.036) [15]. Our analysis of 130 patients, who underwent either MIS approach, further contributes to this series as evidence of the safety and efficacy of a robotic approach compared to a laparoscopic approach.
Laparoscopic resection of large tumors was historically thought to be challenging due to concerns of malignancy, technical difficulty, and increased perioperative complications. However, Agcaoglu et al. postulated that the robotic approach’s increased range of motion and three-dimensional view may provide significant advantages to adrenalectomy. In their examination of 63 adrenal tumors ≥ 5 cm, they reported lower conversion to open rate (4% vs 11%, p < 0.043), shorter operative times (159.4 ± 13.4 versus 187.2 ± 8.3 min, p = 0.043) and shorter hospitalization stays (1.4 ± 0.2 versus 1.9 ± 0.1 days, p = 0.009) using a robotic compared to a laparoscopic approach. Interestingly, they were able to achieve 28 min of time saving and zero morbidity with the robotic technique. They attributed their success to surgeons already having robotic surgery experience for other various general surgery procedures [6]. This preference for utilizing the robot for large adrenal tumors is reflected in our cohort, with patients undergoing the robotic adrenalectomy having a larger specimen size than those who underwent a laparoscopic approach (p = 0.031).
With the addition of robotic adrenalectomy over the last several years, there have been concerns with the cost of robotic surgery with some studies estimating an increase of $950 compared to the laparoscopic approach [5]. In Probst et al.’s analysis of 28 matched pairs of open and robotic adrenalectomies, the cost of robotic procedure was higher than that of open surgery. However, the overall cost of hospitalization was lower for patients in the robotic group. They attributed this overall cost reduction to the decreased length of stay associated with robotic procedures (p < 0.01) [16]. Additionally, Feng et al. compared 122 patients who underwent either laparoscopic or robotic adrenalectomy. They found that both procedures had similar relative costs, operative times, and length of stay. Their cost analysis revealed that limiting extraneous robotic instruments and surgical team experience can greatly influence the cost of robotic adrenalectomy [17].
There were several limitations in our study that need to be addressed. First, as a multi-institutional non-randomized retrospective cohort study, our study is limited by selection bias. Operative approaches and postoperative care were made by each participating institution’s surgeon after consideration of patient factors including BMI, tumor size, location and comorbidities. Surgeon’s operative proficiency and each institution’s experience in each of the three approaches further contribute to the selection bias. Second, recall bias from incomplete institutional reporting on all perioperative and pathologic variables could have also influenced the results. This can be seen in the positive margin rate in the open group. While no difference was seen between the MIS and open approaches in their rate of positive margin, only 46% of patients in the open group reported their margin rate. Lastly, the small sample size of the open group limited the power of our results.
To our knowledge, this is the first multi-institutional study investigating perioperative outcomes in patients undergoing open, laparoscopic, robotic approaches to adrenalectomies in patients with PC. Our findings further corroborate the safety and efficacy of MIS approaches and the promising potential that robotic adrenalectomy has on the management of PCs. Future randomized studies with matched controlled cohorts are needed to help further elucidate the role that robotic surgery has in adrenalectomies, specifically in patients with PCs.
Conclusion
There is growing evidence that MIS techniques offer lower perioperative morbidity and decreased length of hospitalization compared to an open approach in the surgical management of PC. Furthermore, laparoscopic and robotic adrenalectomy offer similar perioperative outcomes with robotic adrenalectomy having a place in the management of large PC. However, further studies examining the surgical approaches to PC are still needed to corroborate these findings.
Table 1 Study population demographic and clinical data
Population characteristics
Patients 153
Adrenalectomy cases 156
No. male/female 70/83
Mean age ± SD (years) 48.34 ± 18.27
Mean BMI ± SD 26.54 ± 6.29
% Cardiac history 15.83
% Hypertension 59.72
% Crisis 33.97
% Family history 34.55
% Pediatric 5.00
Median CCMI (range) 1 (0–8)
Mean MAP at presentation (mmHg) 97.32 (12.71)
Average MAP at discharge MAP (mmHg) 91.41 (12.25)
Mean prescribed HTN meds admit ± SD 1.35 (1.38)
Mean prescribed HTN meds discharge ± SD 1.17 (1.16)
Mean prescribed meds admit ± SD 4.00 (3.56)
Mean prescribed meds discharge ± SD 6.45 (4.86)
Any complication (%) 46 (29.4)
High-grade complication (%) 15 (9.62)
Table 2 Comparison of clinical and perioperative parameters between MIS and open adrenalectomy
Parameters MIS (n = 130, mean ± SEM) Open (n = 26, mean ± SEM) p value
Age (years) 49.3 ± 17.7 43.7 20.7 0.153
Gender (male/female) 60/70 11/15 0.719
CCMI 1.8 ± 1.94 1.7 ± 2.4 0.888
Hypertensive crisis 43 (33.1%) 10 (38.5%) 0.597
CV history 17 (14.4%) 5 (23.8%) 0.329
Side (L/R/Bil) 51/71/8 6/17/1 0.337
BMI 26.4 ± 6.1 27.2 ± 7.8 0.608
OR fluids (ml) 3466 ± 1079 4843 ± 1957 0.006
EBL (ml) 150 ± 318 439 ± 423 0.002
OR time (min) 174.6 ± 76.4 260.4 ± 117.6 < 0.001
OR start mean BP (mmHg) 90.4 ± 19.4 94.0 ± 22.9 0.566
OR max SBP (mmHg) 171.8 ± 24.6 173.8 ± 25.4 0.798
OR min SBP (mmHg) 81.5 ± 17.7 83.3 ± 16.2 0.754
OR end BP (mmHg) 88.1 ± 14.7 77.0 ± 10.6 0.016
Any complication (%) 34 (26.2%) 12 (46.2) 0.041
High-grade complication (%) 10 (7.7%) 5 (19.2) 0.135
ICU stay (days) 1.2 ± 1.5 2.1 ± 2.0 0.056
Hospital stay (days) 3.5 ± 3.4 5.2 ± 3.0 0.021
Mass (g) 79.9 ± 106.6 215.0 ± 178.1 0.042
Specimen size (cm) 7.6 ± 2.1 8.7 ± 3.0 0.206
Tumor size(cm) 5.1 ± 5.12 7.5 ± 3.7 0.001
Positive margin 3 (6.7%) 3 (25%) 0.101
PASS 4.4 ± 3.2 4.6 ± 4.1 0.864
Familial disease 33 (27.8%) 5 (27.8%) 0.596
Table 3 Comparison of clinical and perioperative parameters between laparoscopic and robotic adrenalectomy
Parameters Lap (n = 89, mean ± SEM) Robo (n = 41, mean ± SEM) p value
Age (years) 46.2 ± 17.9 55.9 ± 15.4 0.004
Gender (male/female) 38/51 22/19 0.244
CCMI 1.6 ± 1.8 2.2 ± 2.2 0.127
Hypertensive crisis 35 (39.3%) 8 (19.5%) 0.026
CV history 7 (9.1%) 10 (24.4%) 0.024
Side (L/R/Bil) 33/48/8 18/23/0 0.133
BMI 24.5 ± 4.9 29.8 ± 6.5 < 0.001
OR fluids (ml) 3468 ± 981 3462 ± 1237 0.980
EBL (ml) 134 ± 243 173 ± 404 0.584
OR time (min) 157.9 ± 53.1 210.4 ± 103.0 0.004
OR start mean BP (mmHg) 95.2 ± 17.6 87.3 ± 20.1 0.138
OR max SBP (mmHg) 175.3 ± 25.4 169.8 ± 24.2 0.404
OR min SBP (mmHg) 86.1 ± 16.7 78.8 ± 17.9 0.124
OR end BP (mmHg) 88.4 ± 16.0 88.0 ± 14.1 0.917
Any complication (%) 23 (25.8%) 11 (26.8%) 0.905
High-grade complication 6 (6.7%) 4 (9.8%) 0.724
ICU stay (days) 1.5 ± 1.6 0.7 ± 1.2 0.002
Hospital stay (days) 3.7 ± 3.8 3.0 (2.3) 0.203
Mass (g) 75.9 ± 105.4 82.0 ± 108.9 0.847
Specimen size (cm) 6.9 ± 1.9 8.1 ± 2.1 0.031
Tumor size (cm) 4.6 ± 2.5 6.2 ± 8.4 0.245
Positive margin 1 (5.6%) 2 (7.4%) 1
PASS 4.8 ± 3.6 3.7 ± 2.1 0.699
Familial disease 29 (56.9%) 4 (9.8%) < 0.001
Compliance with ethical standards
Conflict of interest Soroush Rais-Bahrami, M.D. serves as a consultant for Philips/InVivo Corp, Intuitive Surgical, Genomic Health Inc, Bayer Healthcare, and Blue Earth Diagnostics. AM Fang, J Rosen, A Saidian, S Bae, FY Tanno, JL Chambo, J Bloom, J Gordetsky, V Srougi, and J Phillips declare that they have no conflicts of interest.
Ethical approval All procedures performed in studies involving human participants were approved by the Institutional Review Board of each participating institution and with the 1964 Helsinski Declaration and its later amendments or comparable ethical standards.
Informed consent Informed consent was obtained from all individual participants included in the study.
References
1. Germain A , Klein M , Brunaud L (2011) Surgical management of adrenal tumors. J Visc Surg 148 :250–261. 10.1016/j.jviscsurg.2011.06.003
2. Murphy MM , Witkoski ER , Ng SC (2019) Trends in adrenalectomy: a recent national review. Surg Endosc 24 :2518–2526. 10.1007/s00464-010-0996-z
3. Chen Y , Scholten A , Chomsky-Higgins K (2018) Risk factors associated with perioperative complications and prolonged length of stay after laparoscopic adrenalectomy. JAMA Surg 153 :1036–1041. 10.1001/jamasurg.2018.2648 30090934
4. Morris LF , Perrier ND (2012) Advances in robotic adrenalectomy. Curr Opin Oncol 24 :1–6. 10.1097/CCO.0b013e32834da8e1 22080946
5. Agcaoglu O , Aliyev S , Karabulut K , Siperstein A , Berber E (2012) Robotic vs. laparoscopic posterior retroperitoneal adrenalectomy. Arch Surg 147 :272–275. 10.1001/archsurg.2011.2040 22430911
6. Agcaoglu O , Aliyev S , Karabulut K , Mitchell J , Siperstein A , Berber E (2012) Robotic versus laparoscopic resection of large adrenal tumors. Ann Surg Oncol 19 :2288–2294. 10.1245/s10434-012-2296-4 22396002
7. Aksoy E , Taskin HE , Aliyev S , Mithcell J , Siperstein A , Berer E (2013) Robotic versus laparoscopic adrenalectomy in obese patients. Surg Endosc 27 :1233–1236. 10.1007/s00464-012-2580-1 23073684
8. Brunaud L , Ayav A , Zarnegar R , Rouers A , Klein M , Boissel P , Bresler L (2008) Prospective evaluation of 100 robotic-assisted unilateral adrenalectomies. Surgery 144 :995–1001. 10.1016/j.surg.2008.08.032 19041009
9. Karabulut K , Agcaoglu O , Aliyev S , Siperstein A , Berber E (2012) Comparison of intraoperative time use and perioperative outcomes for robotic versus laparoscopic adrenalectomy. Surgery 151 :537–542. 10.1016/j.surg.2011.09.047 22142558
10. Thompson LD (2012) Pheochromocytoma of the Adrenal gland Scaled Score (PASS) to separate benign from malignant neoplasms: a clinicopathologic and immunophenotypic study of 100 cases. Am J Surg Pathol 26 :551–566. 10.1097/00000478-200205000-00002
11. Kiernan CM , Shinall MC Jr , Menez W , Peters ME , Broome JT , Solorzano CC (2014) Influence of adrenal pathology on perioperative outcomes: a multi-institutional analysis. Am J Surg 208 :619–625. 10.1016/j.amjsurg.2014.06.002 25129428
12. Lee J , El-Tamer M , Schifftner T (2008) Open and laparoscopic adrenalectomy: analysis of the National Surgical Quality Improvement Program. J Am Coll Surg 206 :953–959. 10.1016/j.jamcollsurg.2008.01.018 18471733
13. Elfenbein DM , Scarborough JE , Speicher PJ , Scheri RP (2013) Comparison of laparoscopic versus open adrenalectomy: results from American College of Surgeons-National Quality Improvement Project. J Surg Res 184 :216–220. 10.1016/j.jss.2013.04.014 23664532
14. Brandao LF , Autorino R , Laydner H (2014) Robotic versus laparoscopic adrenalectomy: a systematic review and metaanalysis. Eur Urol 65 :1154–1161. 10.1016/j.eururo.2013.09.021 24079955
15. Aliyev S , Karabulut K , Agcaoglu O , Wolf K , Mitchell J , Siperstein A , Berber E (2014) Robotic versus laparoscopic adrenalectomy for pheochromocytoma. Ann Surg Oncol 20 :4190–4194. 10.1245/s10434-013-3134-z
16. Probst KA , Ohlmann CH , Saar M , Simer S , Stoeckle M , Janssen M (2016) Robot-assisted vs open adrenalectomy: evaluation of cost-effectiveness and peri-operative outcome. BJU Int 118 :952–957. 10.1111/bju.13529 27170225
17. Feng Z , Feng MP , Feng Rice MJ , Solorzano CC (2018) A cost-conscious approach to robotic adrenalectomy. J Robot Surg 12 :607–611. 10.1007/s11701-018-0782-9 29388004
|
PMC008xxxxxx/PMC8983098.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
9918367185806676
51469
Divers High Educ
Divers High Educ
Diversity in higher education
1479-3644
35387430
8983098
10.1108/s1479-364420190000022006
NIHMS1785524
Article
All for One And One for All: Coordinating the Resources of Individual Student Research Training Initiatives in Biomedical Sciences at Xavier University of Louisiana
Foroozesh Maryam 1*
Giguette Marguerite 2
Birdwhistell Teresa 1
Morgan Kathleen 1
Johanson Kelly 1
Coston Tiera S. 3
Wilkins-Green Clair 4
1 Department of Chemistry, Xavier University of Louisiana, New Orleans, LA 70125
2 Office of the Provost, Xavier University of Louisiana, New Orleans, LA 70125
3 Center for the Advancement of Teaching and Faculty Development, Xavier University of Louisiana, New Orleans, LA 70125
4 Office of Institutional Research and Decision Support, Xavier University of Louisiana, New Orleans, LA 70125
* Corresponding author: Maryam Foroozesh, Department of Chemistry, Xavier University of Louisiana, New Orleans, LA 70125, [email protected], Phone number: 504-520-5078
25 3 2022
28 2 2019
05 4 2022
22 129149
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Xavier University of Louisiana has a national reputation for producing science, technology, engineering, and mathematics (STEM) graduates who go on to obtain MD and PhD degrees. According to a 2013 National Science Foundation report, Xavier is ranked first in producing African American graduates who go on to receive life sciences PhD degrees, fifth in the nation in producing African American graduates who go on to receive science and engineering PhD degrees, and seventh in producing African American graduates who go on to receive physical sciences PhD degrees. Xavier is currently third among the nation’s colleges and universities in the number of African American graduates enrolled in medical school, according to data compiled by the Association of American Medical Colleges, and ranked first in the number of African American alumni who successfully complete their medical degrees. The success of Xavier’s graduates is due to a combination of university-based student support initiatives and externally funded programs, in particular, the Building Infrastructure Leading to Diversity (BUILD), Maximizing Access to Biomedical Research Careers (MARC) U*STAR, and Research Initiative in Scientific Enhancement (RISE) programs. These three programs, funded by the Training, Workforce Development, and Diversity (TWD) Division at the National Institutes of Health (NIH), offer select trainees undergraduate research opportunities, support mechanisms, and a variety of activities designed to improve their potential for success in graduate school. The BUILD, MARC U*STAR, and RISE programs work closely together and with the University to leverage the resources provided by each in order to provide the best experience possible for their students with a minimum of redundancy of effort. This chapter focuses on the program components and how the programs work together.
Student research training
underrepresented
biomedical workforce
supportive environment
mentoring
cultural responsiveness
pmcINTRODUCTION AND BACKGROUND
National statistics point to the decreasing numbers of US students, especially minorities, enrolling in science, technology, engineering, and mathematics (STEM)-related fields of study at both the undergraduate and the graduate levels, resulting in a decreasing number entering the related research workforce. As a result, the country is increasingly falling behind other nations in growing its STEM workforce. This represents a serious challenge to our economic productivity as the advanced technology and science sectors, requiring large numbers of STEM workers, have been the fastest growing sectors of the world’s economy (U.S. Department of Education, 2016).
Given their percentage of the US population, African Americans obtain far fewer science degrees than expected. Based on the most recent US census estimates (2016), African Americans constitute about 13.3% of the United States’ population; however, they earn less than 9% of the bachelor’s degrees, 8% of the master’s degrees, and 3% of the doctoral degrees conferred in STEM disciplines in the country (National Science Foundation, National Center for Science, & Engineering Statistics, 2015). This disparity is due to many socioeconomic factors, including a lack of access to high-performing school systems and thus receiving substandard preparation in STEM disciplines, and unfamiliarity with STEM careers other than the ones related to medicine.
Xavier University of Louisiana (Xavier) has a strong commitment to enrolling students who have not had access to the educational resources required for a solid foundation in the STEM disciplines. Along with serving the nation’s best and brightest students attracted to Xavier by its reputation in STEM, the University aims to provide all of its students with exceptional educational opportunities inside and outside of the classroom, including hands-on research training. Xavier works to overcome unequal educational backgrounds and socioeconomic challenges to develop student talent through expanding STEM and biomedical training opportunities and building on an established reputation in science education. A diverse scientific workforce taps into unique perspectives, broadening the scope of inquiry into unexplored or neglected areas, challenging disparities and promoting equality.
THEORETICAL UNDERPINNINGS AND ASSUMPTIONS
While over the years, many individual, innovative courses, workshops, student support systems, and faculty development programs have been developed across the country, most have been limited in their focus and reach, and implemented to function in isolation from others. This chapter reports attempts at Xavier to foster cooperation between programs.
Xavier recognizes the value of hands-on research training and direct engagement of students in research projects and has a long history of providing research training opportunities to its students through programs funded by various federal agencies or private foundations. The University’s success in securing funding for these programs, and the institution-wide participation in them, demonstrates faculty and staff’s grant-writing and management skills, their knowledge in designing and implementing innovative interventions, and their deep commitment to improving student training and educational opportunities.
Xavier’s suite of student training programs in biomedical disciplines has been designed to respond to (1) extensive reviews of the national literature regarding factors that lead to, or become barriers to, African Americans achieving terminal degrees in STEM/biomedical fields, (2) surveys of Xavier graduates who went on to graduate schools in such fields, to determine which Xavier educational and/or support programs especially helped or hindered them in being successful in their graduate school programs, and (3) surveys of their graduate school faculty to determine perceived strengths and weaknesses. Based on the data obtained from these reviews and surveys, Xavier redesigned often unrelated programs (e.g., career advising, academic advising, tutoring, graduate placement, faculty and student development, academic courses) and integrated and focused them on the major goal of moving students along the STEM/biomedical research pathway.
INSTITUTIONAL CONTEXT AND STRATEGY OF CHANGE
Xavier University of Louisiana is a historically Black and Catholic university (HBCU) that is nationally recognized for its STEM curricula. A Xavier education also features a solid liberal arts foundation and broad learning experiences that extend beyond the classroom and campus (Hannah-Jones, 2015). Of the 3,044 students enrolled at Xavier in Fall 2017, approximately 72% were African Americans, and about 72% of the 2,345 undergraduates majored in the biomedical sciences. Xavier is a national leader in the number of STEM majors who go on to receive MD and PhD degrees in science and engineering. The University is best known for its reputation in the health professions, ranking first in the number of African Americans who earn bachelor’s degrees in the physical sciences and second in the number earning bachelor’s degrees in the biological and biomedical sciences (Diverse Issues in Higher Education, 2016). A report released by the National Science Foundation in 2013 confirms Xavier’s success in educating science graduates. According to the report, Xavier is ranked first in producing African American graduates who go on to receive life sciences PhD degrees, fifth in the nation in producing African American graduates who go on to receive science and engineering PhD degrees, and seventh in producing African American graduates who go on to receive physical sciences PhD degrees (Fiegener & Proudfoot, 2013). Additionally, Xavier is currently third among the nation’s colleges and universities in the number of African American graduates enrolled in medical school, according to data compiled by the Association of American Medical Colleges, and ranked first in the number of African American alumni who successfully complete their medical degrees (American Association of American Medical Colleges, Applicants, & Matriculants Data, 2016). Xavier alumni contribute significantly to the number of underrepresented minority physicians, pharmacists, dentists, and other professionals in medically related fields who practice throughout the United States. The success of Xavier’s pre-medical program was also recently highlighted in The New York Times (Hannah-Jones, 2015).
Although many well-prepared, highly motivated students are attracted by Xavier’s reputation in the sciences, other students, though bright and capable, come from underperforming public school systems and thus have received substandard preparation in STEM disciplines. Xavier has a mission and history of inclusion in the education of underserved populations. Grounded in the Christian principles exemplified by its founders, Saint Katharine Drexel and the Sisters of the Blessed Sacrament, Xavier has the distinction of being the only Catholic university founded by an American-born saint (Griffin & Hurtado, 2010). True to its HBCU roots, Xavier’s mission continues to be to create a more just and humane society by preparing its students to assume roles of leadership and service in a global society. Xavier is dedicated to a curriculum that focuses on the liberal arts, ethical and moral values, critical thinking, and career preparation for further graduate or professional school education.
Approximately 55.1% of Xavier’s students are from Louisiana, primarily from the New Orleans area. The remaining 44.9% come from other states, the District of Columbia, the Virgin Islands and abroad. While Xavier’s student body is predominantly African American, the University is open to all and strongly supports diversity among its students, faculty, and staff.
As previously mentioned, due to its exemplary research and education training programs, Xavier has been successful in graduating large numbers of Black scientists and health professionals, thereby addressing a national need. The Building Infrastructure Leading to Diversity (BUILD), Maximizing Access to Biomedical Research Careers (MARC U STAR), and Research Initiative in Scientific Enhancement (RISE) programs are three student training programs funded by the National Institutes of Health’s (NIH) National Institute of General Medical Sciences (NIGMS) under the Training, Workforce Development, and Diversity (TWD) umbrella. These programs have significantly contributed to Xavier’s success by providing cohorts of students with knowledge of, and preparation for, biomedical research careers. Xavier’s TWD programs have focused on several initiatives proven to increase student success in graduate school, including research skills training and enhancement of scientific communication skills (Burelli & Rapoport, 2008; Doyle, 2000; Hakim, 2000; Lopatto, 2004; Rothman & Narum, 1999). It is important to note that there are a number of other successful student training programs at Xavier; however, we are focusing only on the TWD programs in this document.
PROGRAM GOALS, COMPONENTS, AND IMPLEMENTATION
The overarching goal of the TWD programs at Xavier University of Louisiana is to increase the number of individuals from underrepresented groups who choose careers in biomedical research and pursue terminal degrees in these disciplines. Toward this end, Xavier’s TWD programs were designed with the following goals:
stimulate interest of Xavier students in biomedical research careers;
increase retention in the biomedical disciplines at Xavier; and
increase the number of Xavier students entering graduate programs in bio-medical research areas.
By implementing isolated initiatives, rather than looking at the whole inclusive picture, educational institutions run the risk of fragmenting the student experience into smaller and smaller pieces and further isolating students from the greater scientific community (Longden, 2006). The TWD programs at Xavier work together to offer the successful programming that they have been providing over their past project cycles, improving the academic and research competitiveness of Xavier’s undergraduates, while putting emphasis on effectively integrating and evaluating initiatives. In order to achieve a greater impact on a larger cohort of students, leverage the available funding, and prevent duplication of efforts, the programs intersect and closely collaborate with each other. In this chapter, rather than focusing on the differences between the three aforementioned programs, their integrated best practices and how the programs are informing and improving each other are discussed in detail.
The TWD programs at Xavier are highly innovative programs designed to broaden the career interests of students early on and engage them in activities that entice them to continue their education toward biomedical terminal degrees and research careers. Strategies used involve transformation of Xavier’s academic and non-academic programs through the redesign, supplementation, and integration of academic advising (BUILD, MARC, RISE), tutoring and supplemental instruction (SI) (BUILD), career services (BUILD, MARC, RISE), research skills training (BUILD, MARC, RISE), on- and off-campus hands-on research opportunities (BUILD, MARC, RISE), mentor training (BUILD, MARC, RISE), mentee training (BUILD, MARC, RISE), and development of new biomedical and research skills courses (BUILD, RISE). The programs also focus on providing faculty members with opportunities to improve their teaching skills (BUILD, RISE) and increase collaborations and communication among faculty engaged in STEM course/curriculum development (BUILD, RISE), as well as their research competitiveness (BUILD). In addition to the wide range of activities supported by these programs within the institution, Xavier is partnering with a number of major research universities across the nation to provide the students with summer research opportunities on their campuses and Xavier faculty with more research collaboration prospects.
Addressing a National Need
Minorities currently represent an expanding portion of the US population, and yet many groups are underrepresented in the STEM and biomedical research workforce. As reported in the 2016 US News/Raytheon STEM Index, the number of White students who earned STEM degrees grew 15% in the prior five-year period, while the number of Black students fell by about the same percentage (Neuhauser & Cook, 2016). In general, the Index showed that women, Blacks, and Hispanics still lag far behind White and Asian men in earning degrees and securing jobs in STEM fields. Thus, unless scientific education becomes more inclusive, the US will be denied the talents of a large segment of its population.
Xavier is already a leader in placing African American students in professional and graduate programs despite the facts that it is a relatively small school and that the majority of its students and their families are socioeconomically challenged. Xavier does not require its students to report their financial status to the University. However, institutional data from the past five years supplied by Xavier’s Office of Planning, Institutional Research and Assessment, indicate that an average of 43% of its students receives Pell grants. Notably, in 2017, Xavier was highlighted in a New York Times article for its sixth-place ranking among all colleges and universities nationwide for the upward economic mobility of its students from the bottom fifth of the income distribution to the top three-fifths (Leonhardt, 2017). Even with this positive economic trend for Xavier’s graduates, there is always room for improvement.
Consistently, the Higher Education Research Institute (HERI) Cooperative Institutional Research Program (CIRP) Freshman Survey indicates that approximately 50% of incoming freshmen enter Xavier with a “strong sense of belonging to a community of scientists” and approximately 65% of incoming students indicate that they will definitely or probably pursue a STEM-focused career. Further, less than 5% of Xavier’s incoming freshmen say that the highest degree they plan to attain is a bachelor’s degree; of the remainder, 10% intend to go on to a master’s, 34% to a medical degree, 26% to a PhD and a further 19% to a professional doctorate. Seventy-one percent of Xavier’s current undergraduates are majoring in a STEM discipline, and if we include all biomedical disciplines, this number is closer to 83%. Therefore, Xavier, with its unique culture and brand in STEM and biomedical fields and the large proportion of its students with interest in these disciplines, has the potential to lead the way in making a difference.
The literature identifies unique barriers faced by African Americans and other individuals from underrepresented groups entering into biomedical research careers, including the lack of (1) Early Awareness of and Deepening Exposure to biomedical research careers and the type of rewards associated with them, (2) Supportive Relationships, particularly those related to faculty advising and mentoring, (3) Suitable Educational Infrastructure, namely (a) innovative STEM courses that engross students in activities that promote a scientific mindset and (b) the supplemental help needed when they face educational challenges, and (4) Active Engagement in meaningful biomedical research experiences and the presence of faculty and institutional resources needed to do this (Butts et al., 2012; Tabak & Collins, 2011; Baxter, Botelho, & O’Donnell, 2015; Pender, Marcotte, Domingo, & Maton, 2010). Surveys of Xavier STEM alumni and the graduate faculty at other institutions who teach them (made possible by an NIH-funded BUILD planning grant) echoed many of these findings and pointed to areas in Xavier’s curricula and support services that needed to be revised or supplemented.
Three Distinctive Programs, One Common Goal
Within the confines of each funding opportunity announcement, Xavier’s TWD programs have been designed to maximize the impact of the resources on the Institution and its students. The three programs use a multi-pronged approach with many concurrent and sequential activities and initiatives that aim to enhance the academic and research environment at Xavier.
The youngest and most extensive of these programs is the BUILD Program. Xavier is one of only 10 institutions across the country receiving support from this experimental funding mechanism at the NIH (2014–2019, and now with a competitive renewal for 2019–2024). The Xavier BUILD Program, Project Pathways, provides resources for key offices and centers across the campus that assists students with academic support, professional development, and undergraduate research activities. This program has contributed to the expansion of both personnel and support of the Student Academic Success Office (SASO), the Office of Career Services (OCS), the Center for Undergraduate Research and Graduate Opportunity (CURGO), and the Center for the Advancement of Teaching and Faculty Development (CAT+FD). These expanded and significantly improved offices and centers have greatly enhanced the quality of student and faculty support services at Xavier and have positively impacted the entire institution.
The BUILD Program coordinates a number of activities designed to educate all interested freshman and sophomore students about the variety of possible biomedical research careers and helps them identify careers that match their interests. Beginning in the freshman year, through a series of broad-based workshops, CURGO and OCS assist Xavier students in gaining a better understanding of the nature of biomedical research careers. Such activities allow students to gradually and voluntarily increase their involvement, based on their level of interest in biomedical research. Early exposure to biomedical research career paths allows students the opportunity to build the skills necessary for success. Also, when motivated to pursue a specific career, students are more likely to engage in their courses, which leads to enhanced retention and improved graduation rates (Jimerson & Ferguson, 2007). These offices guide students through a process that results in their ability to identify a specific biomedical research interest, obtain experiences that advance their research skills, and develop the tools needed to be accepted to and succeed in graduate programs. They also offer assistance to students in creating Individual Development Plans (IDPs) as well as assistance in preparing applications, curriculum vitae, and personal statements. CURGO hosts the annual Festival of Scholars and Summer Research Symposium, which encourage Xavier students to present their research and creative work to the broader campus community.
Both freshmen and sophomores are given opportunities through BUILD to explore research labs at Xavier through the Freshman Open House held during the Biomedical Week at Xavier and also through Research Shadowing Experiences, during which they shadow a current research student on campus and interact with a research group. The BUILD Program also organizes a series of peer-led discussion groups where interested students, primarily sophomores, discuss current scientific topics in the news and the various careers of people involved in the studies. They also hear research-active students discuss their specific research projects and talk about their experiences with the graduate school application process. This peer approach is critical to building and sustaining interest in science careers (Snyder, Sloane, Dunk, & Wiles, 2016).
Each year, the Program chooses 12 sophomore students from among the participants in these different activities who apply to become BUILD Scholars starting the summer before their junior year. The student’s grade point average (GPA) is just one factor considered during the selection, and the process is holistic in nature. The Scholars receive two years of mentored research, including a summer at a research-intensive partner institution. In order to increase the reach of the Program beyond that already included in the funded proposal, each year, several additional students are chosen as BUILD Research Students. Like the Scholars, the BUILD Research Students are matched with mentors and participate in hands-on research; however, due to budget limitations, a BUILD Research Student’s appointment is for a period of 12 months at a lower funding level but with potential for competitive continuation for an additional year.
The oldest of the TWD programs at Xavier, the MARC Program, supports 10 students each year in their junior and senior years and provides them with hands-on research experiences, including a summer at a research-intensive institution. The students are selected at the end of their sophomore year and enter the program in the fall of their junior year. The MARC Program is an honors program with a GPA requirement of 3.0 or above.
The RISE Program accepts students in their freshman, sophomore, and occasionally junior years (while MARC and BUILD programs only accept rising juniors as Scholars, BUILD accepts students at any level as BUILD Research Students who participate in most but not all program activities) and has a minimum GPA requirement of 2.75. This program supports 12 students each year. Earlier involvement with support programming through RISE increases the reach of the TWD student training programs and provides earlier access to hands-on research and longer periods of research training to the selected students. This program also requires one summer at a research-intensive university.
To improve efficiency, the three TWD programs at Xavier use a common online application process coordinated through the Center for Undergraduate Research and Graduate Opportunity. All program applicants are required to work with the Office of Career Services to complete a basic IDP and submit it with their application materials. The qualified applicants are interviewed by a selection committee comprised of faculty members from the biomedical departments and the Program Directors of the three programs. The selected students are placed in the different programs for which they are the best fit.
It is important to note that all three programs have submitted renewal proposals in 2018. Due to the timing of this publication, the descriptions in this document are based on the current programs and do not include changes proposed in the new cycles. If funded and implemented, information and results from the new funding cycles will be shared with the greater academic community through future publications.
Programs’ Components and Implementation
Student interest in biomedical research careers, science identity, self-efficacy, and motivation is measured on entry to Xavier by the HERI CIRP Freshman Survey (as a baseline) and again upon entry into the programs, with a survey that measures interest in biomedical research careers, motivation, science identification, and self-efficacy. Graduating students also complete a final post-test, which captures these same measures. The objective is a significant increase in the different measures from pre- to post-test for the program students.
Entering Research Workshop (Currently BUILD; Going Forward, also RISE)
Program students complete the “Entering Research at Xavier University of Louisiana” (ER XULA) Program modeled after the program developed at the University of Wisconsin-Madison by Drs Branchaw, Pfund, and Rediske (Handelsman, Pfund, Miller Lauffer, & Maidl Pribbenow, 2005; Branchaw, Pfund, & Rediske, 2010). These faculty members served as consultants for Xavier’s BUILD Program in 2015, assisting in the implementation of the workshops at Xavier, which have subsequently been adapted to better fit the needs of Xavier students. ER XULA consists of 10 workshops that focus on preparing students to be effective mentees and researchers. The curriculum involves helping students become aware of and acquire a number of soft skills necessary for becoming contributing members of a research group. Topics covered include but are not limited to choosing a mentor (based on students’ interests), mentor mentee communication, searching the literature, reading scientific articles, establishing goals and expectations, defining hypotheses or research questions, and designing scientific experiments. As part of these workshops, the students also learn how to effectively update and follow their IDPs. To increase effectiveness, the first three introductory ER XULA sessions are offered in the spring semester, soon after program students are selected. The other seven modules are covered over seven consecutive weeks during the students’ first summer in the programs. Each session is 1.5 hours long.
Grant-writing Workshops (Currently BUILD; Going Forward, also RISE)
At the end of their first summer, program students complete a series of three grant-writing workshops and are provided an opportunity to write at least one small proposal for CURGO funding. These workshops are offered following the ER XULA Program and are delivered by CURGO Staff. The combination of these workshops, the Science and Technical Writing course, and the Science Communication course (described below) prepares the students for writing small proposals. Each fall semester, CURGO, releases a request for proposals specifically to students. These proposals are to secure funds for research supplies or travel to scientific meetings. To practice and improve their proposal writing skills, program students are expected to submit at least one small proposal while in the programs. Mentors review the proposals and provide feedback before submission.
Intramural Research (BUILD, MARC, RISE)
All program students participate in an intramural research project with an active investigator at Xavier (or another local university as appropriate). It is very important for students to work for an extended period of time with the same research mentor which results in the students becoming truly engaged in the different aspects of their research project and developing a sense of ownership and commitment to it. Students can then have the opportunity to follow changes in the research direction, learn several research protocols, troubleshoot problems, collect their own data, and evaluate the significance of their results. In addition, this engagement results in the development of a true mentor–mentee relationship. An article in the Council on Undergraduate Research Quarterly identifies the relationship between students and mentors as most important, stating, “Most minority students who end up being scientists have typically had outstanding mentors. We suggest that establishing strong mentoring programs will be the key to any successful effort in recruiting and retaining minority students in the sciences” (Wubah et al., 2000). Mentors are required to complete the Preparing for Mentoring and Advising at Xavier (P-MAX) Program, as described in detail later, and register with the National Research Mentoring Network (NRMN). All research staff directly working with the program students are also required to complete P-MAX. Program students are also required to register with the NRMN and use the resources provided by that national network.
Progress Reports (BUILD, MARC, RISE)
All program students write annual progress reports on their intramural research projects. These progress reports are technical in nature and include experiments attempted and completed, as well as a complete record of collected data. The program students are required to complete the Science and Technical Writing course (described below), which includes instruction on writing technical reports.
Research Poster (BUILD, MARC, RISE)
Each program student presents at least one research poster at a national scientific meeting. The students prepare and submit a research abstract following the conference guidelines. Before submission, the research mentor reviews the abstract and provides input. Poster presentations provide students with experience explaining their research projects to scientists and other students in a more relaxed atmosphere than a formal seminar. Program students also are expected to make poster presentations at Xavier’s annual Festival of Scholars, Summer Research Symposium, or other local meetings.
Library Research Techniques Workshop (BUILD, MARC, RISE)
All program students receive instruction in library research techniques. Students learn about the organization of the library, basic library search processes, hard copy indices available, various science citation manuals, and select print and online primary and secondary scientific resources. They also learn how to search databases, including American Chemical Society Publications (ACS Web Editions), Medline, Scifinder Scholar, Chemica, Science Direct, and Biological Abstracts online. Xavier librarians provide this instruction. At the conclusion of this training, students complete a bibliographic exercise to measure their ability to search the scientific literature competently. To reinforce these skills, the students are asked to search for background information about their research topics and obtain research papers related to upcoming seminars for the journal club meetings. This activity is now wholly supported by University funds and is fully sustainable. Xavier librarians are also available to meet with the students one-on-one as needed to assist them in performing literature searches for their specific research projects.
Extramural Research Experience (BUILD, MARC, RISE)
All program students spend at least one summer at a research-intensive summer program. This experience provides the students with a first-hand perspective of performing research at a large majority institution. An article in the Council of Undergraduate Research Quarterly stated, “SROP (summer research opportunity programs) are designed to not only give minority undergraduates valuable research experience; they also allow them to see what life as a graduate student at a CIC (Committee on Institutional Cooperation) institution would be like” (Foertsch, Alexander, & Penberthy, 1997). These experiences allow participants to be better informed in their decisions regarding graduate school selection and better prepared for graduate work once they enroll. In addition, Xavier’s Center for Undergraduate Research and Graduate Opportunity provides students with summer research experience information. Xavier’s BUILD Program currently has 17 research-intensive partner institutions. MARC and RISE Programs have also formed relationships with a number of research-intensive universities over the years and are also leveraging the relationship between Xavier and BUILD Partner Institutions to provide more summer research program placement opportunities to their students. By participating in both a long-term intramural research experience and an extramural summer program, the students are provided with a more complete perspective on biomedical research. Note that BUILD students are placed in an external lab where the research is in some way related to their research project at Xavier; the student ideally can carry out experiments that cannot be done at Xavier or can learn a new technique to bring back to their Xavier lab.
Science and Technical Writing Course (ENGL 3001) (BUILD, MARC, RISE)
All program students complete the Science and Technical Writing course offered by the English Department at Xavier. Although many undergraduate science majors are technically prepared to pursue advanced graduate degrees, they too often have less than adequate technical writing skills. ENGL 3001 is a one-credit-hour course that meets once each week for one hour and 15 minutes. The course was recently fully revised. Writing topics covered include but are not limited to letters of application, curriculum vitae, research summaries or abstracts, article reviews, lab reports, research documentation, and proposals. Grammar and syntax are taught on an “as needed” basis. All writing assignments relate to the students’ majors and problems of scientific interest. The course also explores the use of graphics for communicating information. Pre- and post-tests are administered by the Instructor to evaluate the progress of the students. If the pre-test shows a significant deficiency in writing skills, the student is also required to use the services of the Xavier Writing Resource Center. This course is now taught on regular basis and is open to all Xavier students. Writing skills are further reinforced throughout the students’ period of participation in the programs through the preparation of research progress reports, research abstracts, a senior thesis, and a short research proposal.
Science Communication Course (CMST 3125) (Currently BUILD; Going Forward, also RISE)
Program students complete the Science Communication course offered by the Communication Studies Department. An understanding of and ability to engage in science communication, where scientists present scientific information to expert and non-expert audiences, is key to success in scientific careers, graduate and professional schools and in generating financial and policy support. In this course, students gain the skills to (1) demonstrate the ability to dispense scientific information in face-to-face, group, and interpersonal settings in long and short formats, (2) deliver in-person and electronically mediated scientific presentations with a measure of poise and confidence, (3) give scientific presentations and respond to questions extemporaneously, (4) understand how language use and style are powerful tools in effectively delivering ideas and that this power comes with ethical, professional, and social responsibility, (5) utilize effective language and style to deliver understandable and persuasive scientific messages and arguments, (6) make use of clearly designed and meaningful visual aids to inform or persuade during presentations and as stand-alone representations of work, (7) analyze and speak to various audiences, and (8) transfer scientific knowledge for the purposes of generating support and informing decision-making. Course assignments include “Basic Scientific Presentation,” “Elevator Pitch,” “Narrative Research Presentation,” “Video News Research Presentation,” and “Poster Presentation.” The Course is now taught on regular basis and is open to all Xavier students.
Ethical Conduct in Scientific Research (PHIL 3400) (BUILD, MARC, RISE)
All program students complete the Ethical Conduct in Scientific Research course offered by the Philosophy Department. The students practice recognizing ethical problems in research and resolving them in a well-reasoned manner. Upon completion of the course, students should be able to (1) identify and explain the prevailing legal and professional norms relevant to resolving ethical problems in research, (2) relate these norms to the broader values they reflect, (3) recognize when an ethical problem in research has arisen and identify the legal and professional norms relevant for resolving it, (4) generate ethically sound, well-reasoned answers to ethical problems in research, both orally and in writing, (5) explain the roles virtue and integrity play in research ethics, and (6) identify and explain practical strategies they can use to develop the character traits and habits of mind characteristic of virtuous researchers. The Course is now taught on regular basis and is open to all Xavier students.
Career Roundtable (BUILD, MARC, RISE)
An annual career roundtable is organized at which students can compare and contrast biomedical research careers in academia, industry, and government. This has been a highly successful approach and one that is easily sustainable. Three biomedical researchers from underrepresented groups are invited to participate in each roundtable. These researchers are asked to give a short presentation on the type of research they perform, their job responsibilities, and how their profession affects their personal life. They are asked to talk to the students about the journey to their current positions, the ups and downs, and the pros and cons. The students are then provided the opportunity to ask questions and have a dialogue with the researchers. The roundtables are widely publicized and open to all students. All program students are expected to participate. At the end of the meeting, the students in attendance are asked to evaluate the roundtable and the individual speakers.
Journal Club (RISE)
Journal club meetings are organized in advance of certain program-sponsored seminars. Students are asked to review and discuss two or more publications by the incoming seminar speaker in order to familiarize themselves with the speaker’s research topics and techniques (technical scientific reading comprehension training). Students are informed of the seminar dates and titles in advance and are asked to conduct a literature search for recent articles by the speaker, reinforcing their literature search skills. They are asked to read the articles and be ready to discuss the topic and the scientific techniques used at the journal club meeting (scheduled one week before the seminar). During the meeting, a faculty member with expertise in the field leads the discussion. All program students are expected to participate in the discussion. At the end of the academic year, students are asked to evaluate the journal club meetings and the associated seminars.
Seminars (BUILD, MARC, RISE)
Established biomedical researchers are invited to present research seminars and spend time with students discussing graduate school opportunities and programs. This has been another successful activity which is easily sustainable, especially in that many speakers are eager to recruit Xavier students and pay their own travel expenses. Speakers from academia also have the opportunity to recruit students for the graduate programs at their institutions. Speakers are asked in advance to direct their talks toward undergraduate students and to emphasize the use and importance of various research techniques. They also meet with students to discuss career goals, issues faced by underrepresented groups, and graduate school expectations and environment. While attempting to expose students to a variety of biomedical fields, all efforts are made to choose speakers whose research interests are similar to those expressed by the students. Researchers are invited from a variety of research environments. Many of these researchers are members of underrepresented groups who serve as excellent role models for Xavier students. The seminars are advertised widely. Program students are expected to participate.
Scientific Conferences (BUILD, MARC, RISE)
Program students are required to attend scientific meetings and present their research results while networking with other scientists. In general, many undergraduates are unaware of the larger scientific community and, even when involved in research, are still not aware of how various research projects at different institutions are parts of a puzzle to solve a specific scientific problem. They often have a very myopic view of science. Attending a national scientific meeting is an enlightening experience for any undergraduate student. Attending national meetings such as the Annual Biomedical Research Conference for Minority Students (ABRCMS) is even more impactful, since for many underrepresented minority students, seeing a large number of scientists from underrepresented groups in one location provides a sense of community and belonging.
Research Seminar (Currently BUILD; Going Forward, also MARC and RISE)
All program students are required to present one formal seminar to their departments or at Xavier’s annual Festival of Scholars in their senior year. Festival of Scholars is an on-campus research symposium held in April of each year, where Xavier students present their research outcomes in the form of a poster or oral presentations. Mentors actively help the students prepare for their presentations. The students are evaluated on their performance by their peers and faculty members.
Senior Thesis (BUILD, MARC, RISE)
All program students write a senior thesis. The intention is that these theses are technically complete and well-written. Mentors review the theses and provide feedback, giving the students an opportunity to revise before submitting to the programs for evaluation. All theses include an introduction describing the background and significance of the research problem, an experimental methods section, the data collected, a statement of results, a discussion of the significance of the results, a conclusion, and a bibliography.
Curriculum Enhancement (BUILD and RISE)
The BUILD and RISE programs also focus on strengthening the supportive environment needed for Xavier students to overcome barriers to success through curricular enhancement. Several new courses have been developed with support from these programs including the above-mentioned Science and Technical Writing, Scientific Communication, and Ethical Conduct in Scientific Research courses. These courses are required for the program students, but open to all.
In addition, to increase the retention and success of Xavier students in the biomedical sciences and to assure courses taken by students are providing effective learning experiences, the programs provide funding and support for evaluation and revision/development of science and mathematics courses. Results from curricular assessment instruments are used by the various departments to identify the areas of weakness in the biomedical curricula at Xavier. The programs provide curriculum development mini-grants for faculty to develop new courses or modify existing courses in order to fill gaps in the biomedical curricula at Xavier.
Academic Support (BUILD)
The mission of Xavier’s SASO is to improve retention and graduation rates of Xavier students by offering academic support through tutoring, SI, and academic success workshops. These services were originally offered only to students taking introductory courses in biology, chemistry, physics, mathematics, writing, and reading. Using BUILD Program support, SASO has expanded its tutoring services and SI to include sophomore- and junior-level courses. SASO also provides academic monitoring of all students using an alert system through which faculty report any student academic, attendance, or behavioral issues. The alerts are then provided to academic advisors who, in turn, meet with the students to assist them in overcoming the underlying issues, ideally, before their grades suffer.
Post-baccalaureate Technician Program (BUILD)
An aspect of Project Pathways showing early success, the BUILD Technician Program, is a post-baccalaureate research training program which provides recent Xavier graduates with the additional training and research experience needed to successfully enter and complete graduate work. It is not uncommon for senior-level undergraduates to express the first-time interest in graduate school, and this program gives such students the time to become competitive applicants for admission to and completion of biomedical graduate programs. Others joining the Program as BUILD Technicians are students who have applied unsuccessfully to graduate school or who graduate mid-year. BUILD Technicians participate in the Program for approximately one year, during which time they are considered full-time staff employees receiving all staff benefits. Additionally, they participate in BUILD enrichment activities such as mentor training and GRE workshops and attend and present at scientific meetings (Fig. 1).
Preparing Mentors and Advisors at Xavier (BUILD, MARC, RISE)
To improve the quality of undergraduate research mentoring at Xavier, the programs provide training activities for faculty and staff, which educate Xavier research mentors on the challenges faced by their mentees and equip them to provide the necessary support both in and out of the laboratory. Xavier’s CAT+FD works closely with the NRMN to present workshops on topics such as cultural awareness, stereotype threat, setting reasonable goals and expectations, and using technology to foster the mentor–mentee relationship. P-MAX is based on the Entering Mentoring seminar developed at the University of Wisconsin Madison by Drs Handelsman, Pfund, Miller Lauffer, and Pribbenow and is designed to provide participating faculty and staff with the knowledge and skills needed to mentor and advise undergraduate students, particularly those engaged in research (Handelsman et al., 2005). The Program addresses topics such as mentor–mentee communication, goal- and expectation-setting, cultural competence, issue identification and resolution, and best practices for good mentoring and advising. Faculty are also provided a repository of resources to support their growth as mentors. P-MAX begins with an intensive, day-long, workshop followed by at least three additional one-hour workshops during each of the subsequent fall and spring semesters. Case studies (found in literature or based on faculty experiences) are routinely used to stimulate discussion on the topic to be addressed, and faculty are encouraged to bring their own real-world experiences for discussion. P-MAX programming is made accessible, relevant, and inviting and is offered in varying formats (panel discussions, hands-on workshops, seminars), at different locations (program locations vary across campus and online) and by expert presenters/facilitators (outside speakers and consultants are frequently brought in to conduct workshops and seminars and advice CAT+FD staff). All research staff working directly with undergraduate research students – including the post-baccalaureate technicians are also expected to participate in the P-MAX workshops. Efforts are underway to take P-MAX online, so that faculty from other institutions who mentor Xavier students can participate in training.
Faculty Research Funding (BUILD)
To assure engagement of Xavier students in quality research experiences, it is also important to have faculty members who are research competitive. Consequently, it is necessary to address the need for junior faculty mentoring and release time, to expand Xavier’s collaborative research network, and to expose Xavier’s undergraduates to a broader range of research opportunities mentored and guided by faculty at Xavier and at other institutions. Pilot project funding is provided through the BUILD Program to address these needs. Project Pathways partners with two other faculty research programs at Xavier (Louisiana Cancer Research Consortium (LCRC) and NIH-funded Research Centers at Minority Institutions (RCMI)) to leverage available funds and provide funding to a larger group of faculty members than would be possible without the partnership. Applicants/awardees fall into one of the following three categories: (1) recently hired faculty who need start-up funding, (2) more experienced investigators with current or recently terminated external grant support who are seeking bridge funds to keep their laboratories functioning while they seek longer-term external funding, or (3) faculty who need seed funding to pursue new directions in their research. Faculty with BUILD Program funding are expected to expand the network of research opportunities by working with Xavier undergraduates seeking research experiences. They are also expected to develop a well-defined student research experience and complete the P-MAX Program.
Research Opportunities for Non-program Students (BUILD)
The BUILD Program provided funding to expand the Center for Undergraduate Research and Graduate Opportunity (CURGO) at Xavier. To provide Xavier students with information about undergraduate research opportunities on- and off-campus, CURGO utilizes an online career services platform, as well as one-on-one support in identifying and applying to research opportunities. CURGO also sponsors student workshops on applying to summer programs, preparing for summer research experiences, and proposal writing. In addition to workshops and seminars, CURGO provides funding to students and faculty for research and travel to conferences.
The Science Education Research Group (BUILD, RISE)
Science Education Research Group (SERG) is an open forum for faculty to meet bi-weekly to bring questions, concerns, or suggestions related to teaching and learning for discussion with other faculty. SERG meetings are informal and multidisciplinary to encourage and support collaboration and communication among and between science and non-science faculty. Topics discussed are suggested by the faculty participants, and discussions are facilitated by the Project Pathways’ Education Improvement Specialist (EIS), who is housed in CAT+FD. The EIS also provides support in the form of topic research and presentation of pedagogical trends. SERG is fully supported by the institution (Fig. 2).
PROGRAM OUTCOMES AND EVOLUTION
Looking at the programs holistically, it is evident that their various initiatives touch every aspect of a student’s journey at Xavier from freshman year to graduation and beyond. The strong focus on evaluation makes these projects data-driven experiments, which are expected to lead to the determination of effective practices, which could be shared with the greater academic community with the ultimate goal of effectively increasing diversity in the biomedical workforce. Data on the successfulness of the various initiatives are being collected and analyzed, and will be reported at a future time.
LESSONS LEARNED AND RECOMMENDATIONS
The strategies developed by the TWD programs at Xavier are designed to address the challenges and barriers Xavier students face as they work toward graduate studies and entering the biomedical workforce. Xavier University of Louisiana has a long history of providing high-quality, rigorous education to African American students in a very supportive environment with highly dedicated faculty and staff. The initiatives highlighted here could be used by other institutions as model initiatives for assisting students in STEM and other biomedical fields of study to successfully matriculate through college and graduate school and develop their research careers.
FUNDING ACKNOWLEDGMENT
Research reported in this publication was supported by the National Institute Of General Medical Sciences of the National Institutes of Health under Award Numbers UL1GM118967 8TL4GM118968, RL5GM118966, 5T34GM07716, and 2R25GM060926. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Fig. 1. Leveraging Institutional, BUILD, MARC, and RISE Resources for Trainees. Shown are Selected Trainee Activities from the BUILD, MARC, and RISE Programs. Green Boxes Indicate that the Activity is Provided by that Program, while Gold Boxes Indicate that the Activity is Provided by Shared University and Program Funding through a University Office or Center, such as the Center for the Advancement of Teaching and Faculty Development (CAT+FD), Center for Undergraduate Research and Graduate Opportunity (CURGO) or the Student Academic Success Office (SASO), or an Academic Course. Arrows Indicate Activities that are Utilized by the Program, but Provided by the TWD Program from which the Arrow Originates.
Fig. 2. Leveraging Institutional, BUILD, MARC, and RISE Resources for Faculty. Shown are Selected Faculty Activities from the BUILD, MARC, and RISE Programs. Green Boxes Indicate that the Activity is Provided by that Program, while Gold Boxes Indicate that the Activity is Provided by Shared University and Program Funding through a University Office or Center, such as the Center for the Advancement of Teaching and Faculty Development (CAT+FD), Office of Research and Sponsored Programs (ORSP) or Academic Departments. Arrows Indicate Activities that are Utilized by the Program but Provided by the TWD Program from which the Arrow Originates.
REFERENCES
American Association of American Medical Colleges, Applicants and Matriculants Data. (2016). Undergraduate institutions supplying 15 or more Black or African-American applicants to US Medical Schools, 2016–2017 [Data File] Retrived from https://www.aamc.org/download/321446/data/factstablea2-1.pdf
Baxter S , Botelho J , & O’Donnell K (2015). Increasing student success in STEM: A guide to systemic institutional change Retrieved from https://www.aacu.org/sites/default/files/files/pkalkeck/casestudies.pdf
Branchaw J , Pfund C , & Rediske R (2010). Entering research: A facilitator’s manual. Workshop for Students Beginning Research in Science New York, NY: W.H. Freeman and Company.
Burrelli J , & Rapoport A (2008, August). Role of HBCUs as baccalaureate-origin institutions of Black S&E doctorate recipients. National Science Foundation Info Brief, (NSF 08–319) Retrieved from https://files.eric.ed.gov/fulltext/ED502482.pdf
Butts GC , Hurd Y , Palermo A-GS , Delbrune D , Saran S , Zony C , & Krulwich TA (2012). Role of institutional climate in fostering diversity in biomedical research workforce: A case study. Mount Sinai Journal of Medicine, New York, 79 (4 ), 498–511.
Diverse Issues in Higher Education. (2016). Top 100 degree producers 2016: baccalaureate [Data File] Retrieved from http://diverseeducation.com/top100/BachelorsDegreeProducers2016.php?AppKey=38d31000g2h3a9a0b3f1b9i5j0e0&ComparisonType1_1=%3D&MatchNull1_1=N&school=Xavier+University+of+Louisiana&ComparisonType2_1=%3D&MatchNull2_1=N&state=zip&ComparisonType3_1=%3D&MatchNull3_1=N&major=zip&ComparisonType4_1=%3D&MatchNull4_1=N&race=African+American
Doyle MP (2000). Academic excellence: The role of research in the physical sciences at undergraduate institutions Tucson, AZ: Research Corporation.
Fiegener MK , & Proudfoot SL (2013, April). Baccalaureate origins of US-trained S&E doctorate recipients. National Science Foundation, National Center for Science and Engineering Statistics (NSF 13–323) Retrieved from https://www.nsf.gov/statistics/infbrief/nsf13323/nsf13323.pdf
Foertsch J , Alexander BB , & Penberthy D (1997). Summer research opportunity programs (SROPs) for minority undergraduates – A longitudinal study of program outcomes, 1986–1996 Retrieved from http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.583.5609&rep=rep1&type=pdf
Griffin KA , & Hurtado S (2010). Institutional diversity in American higher education. In Schuh JH , Jones SR , & Harper SR (Eds.), Student services: A handbook for the profession (pp. 24–42). San Francisco, CA: Jossey-Bass.
Hakim TM (2000). How to develop and administer institutional undergraduate research programs Washington, DC: Council on Undergraduate Research.
Handelsman J , Pfund C , Miller Lauffer S , & Maidl Pribbenow C (2005). Entering mentoring: A seminar to train a new generation of scientists Madison, WI: University of Wisconsin Press.
Hannah-Jones N (2015, September 9). A prescription for more Black doctors: How does Tiny Xavier University in New Orleans manage to send more African-American students to medical school than any other college in the country? The New York Times Magazine Retrieved from https://www.nytimes.com
Jimerson SR , & Ferguson P (2007). A longitudinal study of grade retention: Academic and behavioral outcomes of retained students through adolescence. School Psychology Quarterly, 22 (3 ), 314–339.
Leonhardt D (2017, January 18). America’s great working-class colleges. The New York Times Retrieved from https://www.nytimes.com
Longden B (2006). An institutional response to changing student expectations and their impact on retention rates. Journal of Higher Education Policy and Management, 28 (2 ), 173–187.
Lopatto D (2004). Survey of undergraduate research experiences (SURE): First findings. Cell Biology Education, 3 (4 ), 270–277.15592600
National Science Foundation, National Center for Science and Engineering Statistics. (2015, June 30). Science and engineering degrees: 1966–2012. Detailed statistical tables NSF 15–326 Retrieved from https://www.nsf.gov/statistics/2015/nsf15326/
Neuhauser A , & Cook L (2016, May 17). 2016 US News/Raytheon STEM Index shows uptick in hiring, education. US News and World Report Retrieved from https://www.usnews.com/
Pender M , Marcotte DE , Domingo MRS , & Maton KI (2010). The STEM pipeline: The role of summer research experience in minority students’ PhD aspirations. Education Policy Analysis Archives, 18 (30 ), 1–36.21841903
Rothman FG , & Narum JL (1999). Then, now, and in the next decade: A commentary on strengthening undergraduate science, mathematics, engineering and technology education Washington, DC: Project Kaleidoscope.
Snyder JJ , Sloane JD , Dunk RDP , & Wiles JR (2016). Peer led team learning helps minority students succeed. PLoS Biology, 14 (3 ), 1–7. e1002398. doi:10.1371/journal.pbio.1002398
Tabak LA , & Collins FS (2011). Weaving a Richer Tapestry in biomedical science. Science, 333 (6045 ), 940–941.21852476
U.S. Department of Education. (2016, March 16). Fact sheet: Spurring African American STEM degree completion Retrieved from https://www.ed.gov/news/press-releases/fact-sheet-spurring-african-american-stem-degree-completion
Wubah D , Schaefer D , Gasparich G , Brakke D , McDonald G , & Downey D (2000). Retention of minority students through research. Council on Undergraduate Research Quarterly, 20 (3 ), 120–125.
|
PMC008xxxxxx/PMC8983099.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
9421547
4136
Hum Pathol
Hum Pathol
Human pathology
0046-8177
1532-8392
30179687
8983099
10.1016/j.humpath.2018.08.021
NIHMS1783882
Article
Histologic findings associated with false-positive multiparametric magnetic resonance imaging performed for prostate cancer detection☆
Gordetsky Jennifer B. MD ab*
Ullman David MD a
Schultz Luciana MD c
Porter Kristin K. MD d
Pena Maria del Carmen Rodriguez MD a
Calderone Carli E. MD b
Nix Jeffrey W. MD b
Ullman Michael MD e
Bae Sejong PhD f
Rais-Bahrami Soroush MD bd
a Department of Pathology, University of Alabama at Birmingham, Birmingham, AL 35249, USA
b Department of Urology, University of Alabama at Birmingham, Birmingham, AL 35249, USA
c Instituto de Anatomia Patológica, Piracicaba, Santa Bárbara d’Oeste 13419-160, Brazil
d Department of Radiology, University of Alabama at Birmingham, Birmingham, AL 35249, USA
e MedStar Georgetown University Hospital, Washington, DC 20007, USA
f Division of Preventive Medicine, University of Alabama at Birmingham, Birmingham, AL 35249, USA
* Corresponding author at: Department of Pathology, University of Alabama at Birmingham, NP 3550, 1802 6th Avenue South, Birmingham, AL 35249. [email protected] (J. B. Gordetsky).
10 3 2022
1 2019
01 9 2018
05 4 2022
83 159165
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Summary
Magnetic resonance imaging (MRI)/ultrasound fusion–targeted biopsy (TB) has been shown to more accurately identify higher-grade prostate cancers compared with standard-of-care systematic sextant prostate biopsy (SB). However, occasional false-positive imaging findings occur. We investigated the histologic findings associated with false-positive prostate MRI findings. A retrospective review was performed on our surgical pathology database from 2014 to 2017 selecting patients with no cancer detected on TB with concurrent SB after at least 1 prior benign SB session. Histologic features evaluated included percentage of core involvement by chronic inflammation, percentage of core composed of stroma, percentage of glands involved by atrophy, and presence of the following features: acute or granulomatous inflammation, stromal nodular hyperplasia, adenosis, squamous metaplasia, basal cell hyperplasia, and presence of skeletal muscle. Histologic findings were compared between TB and concurrent SB. We identified 544 patients who underwent TB. Of these, 41 patients, including 62 targeted lesions, met criteria. Compared with SB tissue, the mean percentage of stroma was increased in TB (P = .02). Basal cell hyperplasia was also found to be more common on TB (P = .02). Both high percentage of stroma (P = .046) and presence of basal cell hyperplasia (P = .038) were independent predictors on multivariate analysis. The combination of high chronic inflammation, high stroma, acute inflammation, and basal cell hyperplasia was associated with TB (P = .001). Atrophic glands and chronic inflammation showed a positive correlation (r = 0.67, P = .003), which was especially seen in high prostate imaging reporting and data system lesions. Specific benign histologic entities are associated with false-positive findings on prostate MRI.
Multiparametric MRI
Prostatitis
Prostate biopsy
False-positive imaging
Pathology
Histology
Cancer
pmc1. Introduction
Historically, prostate cancer has been difficult to identify on imaging. Although transrectal ultrasound (US) was used to guide placement of the biopsy needle, it did not accurately localize lesions suspicious for malignant disease. Thereby, detection of prostate cancer classically depended on systematic sampling of the prostate gland using a sextant approach. These biopsies represented a random tissue sampling rather than biopsies targeting a specific lesion suspicious for harboring cancer. To date, prostate cancer remains the only solid organ malignancy standardly diagnosed in this fashion.
Magnetic resonance imaging has shown improvement in the identification of prostate cancer on imaging. Indeed, studies have shown equivalent detection of prostate cancer using an MRI/US fusion–targeted biopsy (TB) approach compared with the systematic extended-sextant biopsy (SB) technique while limiting the number of cores sampled [1–3]. TB has also been shown to detect more clinically significant cancers [4–6]. This has been shown in patients undergoing initial biopsy for clinical suspicion of prostate cancer as well as men who have had a prior negative standard biopsy [7].
Despite the optimization in prostate cancer detection, there remain limitations to this new technology. One of these limitations is the presence of false-positive lesions identified as suspicious for malignancy on MRI. Studies have shown that there are certain entities that can lead to false-positive reads on MRI. Some examples include inflammation, glandular hyperplasia, and stromal hyperplasia [8–12]. However, understanding the histologic findings behind false-positive MRI findings for the detection of prostate cancer is still being explored. Herein, we investigate benign prostate gland histology associated with false-positive lesions on prostate MRI.
2. Materials and methods
A retrospective institutional review board–approved search was performed on our surgical pathology database from 2014 to 2017. Image processing and targeting of lesions at the time of biopsy was performed using the DynaCad software and UroNav fusion biopsy system, respectively (Phillips/InVivo Corp, Gainsville, FL). Prostate imaging reporting and data system (PIRADS) version 2 scoring was assigned by a multidisciplinary consensus conference with fellowship-trained radiologists and urologic oncologists specializing in prostate MRI, all with more than 4 years of experience with prostate MRI. Two fellowship-trained urologie oncologists performed all TBs. Each targeted lesion was sampled by at least 2 needle cores as recommended based on prior publication [13]. Patients were selected that had no history of prostate cancer and at least 1 previous SB session. Patients were then filtered for those who underwent an additional biopsy session with TB and concurrent repeat SB with all cores negative for adenocarcinoma. As patients underwent concurrent TB and SB, each patient was able to act as their own control for the purposes of analyses performed comparing TB tissue with SB tissue.
Prostate tissue sampled was organized by sextants. Tissue from the area of the TB was compared with a section of tissue from a SB taken during the same biopsy procedure that was separated by at least 2 sextants (Fig. 1). Patients were limited to those who had a maximum of 2 targeted lesions on TB to allow for enough sextant distance between the standard and targeted tissue cores. Histologic features were evaluated including the percentage of tissue involvement by chronic inflammation, the percentage of tissue composed of stroma, and the percentage of glands involved by atrophy. A high degree of chronic inflammation was defined according to a cutoff of 10% of tissue involved. A high degree of stroma was defined according to a cutoff of 60% of tissue involved. In addition, the presence or absence of the following features was assessed: acute inflammation, granulomatous inflammation (necrotizing or nonspecific granulomatous prostatitis), stromal nodular hyperplasia, adenosis, squamous metaplasia, basal cell hyperplasia, and skeletal muscle. Histologic findings on TB were compared with SB. Additional clinical and radiologic information was gathered, including patient age, prostate-specific antigen (PSA), and corresponding PIRADSv2 score for each targeted lesion detected on MRI (Fig. 2). Evaluation of all prostate biopsy pathology was performed by 2 urological pathologists (J. B. G. and L. S.).
Statistical analyses were done using STATA (StataCorp 2005. Stata Statistical Software: Release 9.2; StataCorp, College Station, TX). The χ2 test was used for categorical variables and Student t test for continuous variables. Categorical and continuous variables were compared using Kruskal-Wallis 1-way analysis of variance. Linear dependences were calculated using the Spearman correlation.
3. Results
We identified 544 patients who underwent MRI/US TB. Of these, 96 patients had no tumor on TB and concurrent 12-core extended-sextant SB. Of this cohort, 75 patients had a maximum of 2 targeted lesions on MRI for appropriate separation of tissue. Of these, 47 patients had at least 1 prior negative SB session. From this group, a final cohort of 41 patients had concurrent SB tissue at least 2 sextants away from the concurrent TB. The mean number of previous negative SB for our final cohort was 1.6 (range, 1–5). Since the time of the concurrent TB with SB, 3 patients underwent an additional follow-up prostate biopsy, all negative for prostatic adenocarcinoma.
Tissues from 62 targeted lesions were compared with tissues from 41 SBs (Table 1). The mean PSA was 8.5 ± 4.2 ng/mL, and the mean age was 63.2 ± 6.2 years. Mean PIRADS score was 3.3 ± 0.7. Twenty-one (51.2%) of 41 patients had 1 targeted lesion. The remaining 20 (48.8%) of 41 patients had 2 targeted lesions. The mean number of targeted cores biopsied per patient was 3.4 ± 1.2. In terms of location of the targeted lesions, 32 (51.6%) of 62 lesions were located in the right lobe and 30 (48.4%) of 62 in the left lobe. Lesions were most common in the mid-gland (25/62 [40.3%]) followed by the base (21/62 [33.9%]) and apex (16/62 [25.8%]). The posterior gland (44/62; 71%) was more common than the anterior gland (18/62 [29%]), and the peripheral zone (38/62 [61.3%]) was more common than the central zone (24/62 [38.7%; Table 2).
TB tissue, as examined in this patient cohort, had increased mean percentage of stroma compared with SB tissue (67.0% versus 59.4%, respectively; P = .02; Table 3). In addition, high percentage of stroma remained an independent predictor of TB tissue over SB tissue on multivariate analysis (P = .046). Basal cell hyperplasia was also more frequent in TB tissue compared with SB tissue (13/62 [21.0%] versus 2/62 [3.2%], respectively; P = .02). This finding maintained statistical significance on multivariate analysis (P = .038; Table 4). Acute inflammation was also increased in TB tissue (18/62 [29.0%] versus 6/41 [14.6%]), which trended toward statistical significance (P = .09). TB tissue had increased mean percentage of chronic inflammation compared with SB tissue (10.3% versus 7.1%); however, this was not statistically significant (P = .14). The presence of skeletal muscle was not significantly different between the 2 groups, seen in 7 (17.5%) of 41 SBs and 10 (16.1%) of 62 TBs (P = .90). Presence of granulomatous inflammation was not statistically different between the 2 groups, seen in 1 (2.5%) of 41 SBs and 2 (3.2%) of 62 TBs (P = .82). Similarly, differences in squamous metaplasia were not significant, seen in 1 (2.5%) of 41 SBs and 3 (4.8%) of 62 TBs (P = .53). Stromal nodular hyperplasia was only found in 2 cases and was present only on TB tissue from those cases (P = .25). Adenosis was only found in 2 cases, both identified in the TB tissue cores (P = .25). Both of these lesions had a high PIRADSv2 score of 4. Atrophic glands and chronic inflammation showed a positive correlation among all lesions (r = 0.39, P < .0001) and correlated more strongly among lesions with high PIRADS suspicion scores (r = 0.66, P = .0034). We found that the combination of high chronic inflammation, high percentage of stroma, presence of acute inflammation, and presence of basal cell hyperplasia correlated with TB tissue (P = .001; Figs. 3 and 4).
4. Discussion
Multiple studies have shown the superiority of multiparametric MRI and MRI/US fusion-targeted biopsies over the standard-of-care extended-SB approach in the detection of clinically significant prostate cancer. Indeed, TB is now recommended for patients with a prior negative SB and continued clinical suspicion for prostate cancer, including elevated PSA [7,14]. In addition, MRI/US TB has been shown to be useful in other situations, such as patients on active surveillance [15]. With the growing acceptance and utilization of this technology, more patients will become candidates for TB. To date, most patients who undergo MRI/US TB are not biopsy naïve. In other words, they have previously undergone a SB, and they had either a benign pathology result or they are pursuing active surveillance [16]. Many previous studies evaluating the utility of MRI in the diagnosis of prostate cancer have focused on the high negative predictive value of this technique. Having a low false-negative rate is logically of utmost importance so as not to miss cases of clinically significant cancer. However, achieving a very low false-negative rate comes with the inevitable problem of false-positive imaging suspecting prostate cancer, which is not detected on biopsy sampling. Our goal as a part of this study was to help define the pathologic causes of false suspicion on prostate MRI to ultimately help guide methods of reducing the number of unnecessary biopsies.
It has previously been shown that MRI can result in false-positive imaging results when using this technique for the detection of cancer in different organs. For example, in screening for breast cancer, MRI added to mammography can increase the screening sensitivity for women at high risk for malignancy [17]. However, the false-positive rates are also increased using this method. Similarly, MRI has been studied for the detection for pulmonary nodules representing lung cancer, but again, a significant number of false-positive diagnoses were observed [18]. A similar phenomenon has been described in MRI used for prostate cancer screening. In general, the stroma of the anterior prostate gland and of central zone have low signal on T2-weighted MRI, mimicking prostate cancer [11]. However, contrary to this phenomenon, we found that most of our suspicious lesions with benign pathology were present in the posterior aspect of the gland and the peripheral zone. In addition, although benign prostatic hyperplasia typically has increased T2-weighted signal, stromal nodular hyperplasia can mimic transition zone tumors [11]. Furthermore, inflammation can present as a lesion on MRI suspicious for prostate cancer. We previously described granulomatous prostatitis presenting as lesions highly suspicious for prostate cancer (PIRADSv2 scores 4 and 5) [8]. Other studies have demonstrated that prostatitis can appear similar to prostate cancer with low signal intensity on T2-weighted images and early enhancement on contrast-enhanced MRI [9,10]. Interestingly, in our current study, we did not find a difference between TB tissue and SB tissue in terms of granulomatous inflammation. However, this is likely due to the small number of cases (n = 3) with granulomatous inflammation present in our study population.
Diffusion-weighted MRI depends on the movement of water molecules in tissue. The restriction of diffusion, which can be caused by any increase in cellularity, results in decreased signal. Thus, prostate cancer, which causes increased cellularity due to the increased number of more densely packed glands, reveals altered water diffusion patterns on MRI. However, stromal nodular hyperplasia may also cause increased relative cellularity. In our study, we found that TB tissue had increased mean percentage of stroma compared with tissue examined from SB (P = .02). In addition, a high percentage of stroma was an independent predictor of TB tissue on multivariate analysis (P = .046). This increase in stromal density partially helps explain the false suspicion for cancer on MRI. Basal cell hyperplasia was also found to be associated with TB tissue and remained an independent predictor on multivariate analysis (P = .038). It is possible that the increased cellularity due to basal cell hyperplasia could similarly create an increased suspicion for prostatic adenocarcinoma on MRI. In addition, we found that stromal nodular hyperplasia and adenosis, although small in numbers (n = 4), were only found on TB. The adenosis lesions both had a PIRADS score of 4, highlighting this particular imaging pitfall.
Dynamic contrast-enhanced MRI is another parameter used in the detection of prostate cancer. Prostate cancer shows earlier enhancement and greater signal on dynamic contrast enhanced images [19]. The increased enhancement is thought to be due to increased angiogenesis and nonphysiologic neovascularity associated with tumor tissues. Because inflammatory and other reparative processes can also stimulate vessel growth and promote increased blood flow, this presents another confounding issue for the diagnostic accuracy of MRI. We found that high chronic inflammation and high stroma, in combination with the presence of acute inflammation and basal cell hyperplasia, were correlated with TB over SB (P = .001). This is a logical finding in that inflammation could potentially cause an increase in angiogenesis, whereas basal cell hyperplasia and increased stroma could lead to increased tissue density.
One limitation of our study is the possibility of a pathologically false-negative patient included in our tissue analysis. To attempt to limit this possible bias, all patients selected had at least 1 prior negative SB (range, 1–5) as well as an additional negative TB with concurrent repeat SB. It is important that this particular population of patients is evaluated because they represent a select population of men that are undergoing multiple unnecessary surgical procedures with minimal benefit after at least 2 total SB and an additional augmented sampling via TB. Each biopsy poses a small but significant morbidity including patient discomfort, risk for infection, bleeding, and possible hospitalization not to mention the emotional and financial burden on patients and the health care system. Meanwhile, it is estimated that only 3% to 4% of patients will be diagnosed with prostate cancer after 2 negative SBs [20,21]. It is likely that the number of false negatives is even lower after 2 negative SBs and an additional negative TB, which has been shown to have higher detection rates than SB alone [22].
Finally, it may be argued that our population does not represent a true “false-positive” MRI population, as the average PIRADS score for our population was 3.3. Lesions on prostate MRI with a PIRADS score of 3 have been shown to be associated with a low risk for harboring clinically significant prostate cancer [23]. However, it should be mentioned that these were patients with enough suspicion on imaging to proceed with a TB in a true clinical setting. This is exactly the population that needs to be addressed, as there is a high degree of unnecessary biopsies in hindsight. Future endeavors need to be taken to better delineate imaging markers that are associated with these newly identified histologic mimickers of prostate cancer seen on MRI, given the increased utilization of this imaging modality for the screening of prostate cancer.
5. Conclusions
There are specific benign histologic entities associated with lesions suspicious for prostatic adenocarcinoma on prostate MRI. Increased stroma and basal cell hyperplasia are associated with lesions falsely suspicious for harboring prostatic adenocarcinoma. High chronic inflammation and high stroma, in combination with the presence of acute inflammation and basal cell hyperplasia, correlated strongly with TB. Additional qualitative features and quantitative measures on MRI need to be investigated to help decrease the amount of unnecessary prostate biopsies driven by false-positive imaging findings corresponding to the benign histology identified.
Funding/Support:
This work was funded by a Junior Faculty Development Grant (ACS-IRG 001-53) and by developmental funds from the UAB Comprehensive Cancer Center Support Grant (NCI P30 CA 013148) to Soroush Rais-Bahrami, MD.
Fig. 1 Prostate gland extended-sextant template with a needle representing tissue examined from a TB site and a circle representing comparative tissue examined from a standard biopsy site at least 2 quadrants from the targeted lesion.
Fig. 2 Left posterior mid-gland peripheral zone focal moderate hypointensity on ADC (red arrow, D) with corresponding mild hyperintensity on high b-value diffusion-weighted image (red arrow, B) corresponds to a circumscribed moderate hypointense focus on T2-weighted images (red arrow, A). There is also asymmetric perfusion (red arrow, C). This lesion is high suspicion for clinically significant prostate cancer (PIRADS 4).
Fig. 3 Hematoxylin-eosin–stained slides of prostate core biopsy tissue showing benign prostate glands with a typical gland-to-stroma ratio (A), atrophic prostate glands with increased stroma (B), stromal nodular hyperplasia (C), and basal cell hyperplasia (D).
Fig. 4 Hematoxylin-eosin–stained slides of prostate core biopsy tissue showing nonspecific granulomatous prostatitis (A), adenosis (B), acute inflammation (C), and squamous metaplasia (D).
Table 1 Patient clinical and imaging characteristics
Patient characteristic n %
Race
African American 5 12
White 24 59
Unknown 12 29
Age (y)
51–60 15 36
61–70 22 54
71–80 4 10
Mean, 63.2 ± 6.2 y
PSA
<10 31 76
>10.1 10 24
Mean, 8.5 ± 4.2 ng/mL
Targeted lesions per patient
1 lesion 21 51.2
2 lesions 20 48.8
PIRADS
I 2 3
II 12 19
III 31 50
IV 16 26
V 1 2
Mean, 3.3 ± 0.7
Table 2 Location of targeted lesions
Location n %
Right 32 51.6
Left 30 48.4
Base 21 33.9
Mid 25 40.3
Apex 16 25.8
Anterior 18 29
Posterior 44 71
Central 24 38.7
Peripheral 38 61.3
Table 3 Histologic features comparing targeted prostate biopsy tissue suspicious for cancer on MRI with standard biopsy tissue
Targeted biopsy (%) Standard biopsy (%) P
Percentage of stroma 67.0 59.4 .02
Basal cell hyperplasia 21.0 3.2 .02
Acute inflammation 29.0 14.6 .09
Chronic inflammation 10.3 7.1 .14
Granulomatous inflammation 3.20 2.50 .82
Skeletal muscle 16.1 17.5 .90
Squamous metaplasia 4.80 2.50 .53
Table 4 Multivariate analysis predicting benign targeted biopsy tissue suspicious for cancer on MRI
Parameter Odds ratio 95% Confidence limits P
High chronic inflammation 1.616 0.561 4.651 .3735
Acute inflammation 0.535 0.178 1.61 .2659
Basal cell hyperplasia 5.369 1.095 26.327 .0383
High stromal component 2.386 1.016 5.601 .0459
☆ Competing interest: Jeffrey W. Nix and Soroush Rais-Bahrami serve as consultants for Philips/InVivo Corp. All other authors report no conflicts of interest or financial disclosures that were pertinent to this study.
References
[1] Siddiqui MM , George AK , Rubin R , Efficiency of prostate cancer diagnosis by MR/ultrasound fusion–guided biopsy vs standard extended-sextant biopsy for MR-visible lesions. J Natl Cancer Inst 2016; 108 :djw039.27130933
[2] Borkowetz A , Platzek I , Toma M , Comparison of systematic transrectal biopsy to transperineal magnetic resonance imaging/ultrasound-fusion biopsy for the diagnosis of prostate cancer. BJU Int 2015;116 :873–9.25523210
[3] Ahmed HU , El-Shater Bosaily A , Brown LC , Diagnostic accuracy of multi-parametric MRI and TRUS biopsy in prostate cancer (PROMIS): a paired validating confirmatory study. Lancet 2017;389 : 815–22.28110982
[4] Gordetsky JB , Thomas JV , Nix JW , Higher prostate cancer grade groups are detected in patients undergoing multiparametric MRI-targeted biopsy compared with standard biopsy. Am J Surg Pathol 2017;41 : 101–5.27574875
[5] Siddiqui MM , Rais-Bahrami S , Turkbey B , Comparison of MR/ultrasound fusion–guided biopsy with ultrasound-guided biopsy for the diagnosis of prostate cancer. JAMA 2015;313 :390–7.25626035
[6] Truong M , Rais-Bahrami S , Nix JW , Perineural invasion by prostate cancer on MR/US fusion targeted biopsy is associated with extraprostatic extension and early biochemical recurrence after radical prostatectomy. HUM PATHOL 2017;66 :206–11.28705708
[7] Kasivisvanathan V , Rannikko AS , Borghi M , MRI-targeted or standard biopsy for prostate-cancer diagnosis. N Engl J Med 2018;378 : 1767–77.29552975
[8] Rais-Bahrami S , Nix JW , Turkbey B , Clinical and multiparametric MRI signatures of granulomatous prostatitis. Abdom Radiol (NY) 2017; 42 :1956–62.28238033
[9] Nagel KN , Schouten MG , Hambrock T , Differentiation of prostatitis and prostate cancer by using diffusion-weighted MR imaging and MR-guided-biopsy at 3 T. Radiology 2013;267 :164–72.23329653
[10] Esen M , Onur MR , Akpolat N , Utility of ADC measurement on diffusion-weighted MRI in differentiation of prostate cancer, normal prostate and prostatitis. Quant Imaging Med Surg 2013;3 :210–6.24040617
[11] Quon JS , Moosavi B , Khanna M , False positive and false negative diagnoses of prostate cancer at multi-parametric prostate MRI in active surveillance. Insights Imaging 2015;6 :449–63.
[12] Coker MA , Glaser ZA , Gordetsky JB , Thomas JV , Rais-Bahrami S .Targets missed: predictors of MRI-targeted biopsy failing to accurately localize prostate cancer found on systematic biopsy. Prostate Cancer Prostatic Dis 2018. 10.1038/s41391-018-0062-9.
[13] Lai WS , Zarzour JG , Gordetsky JB , Co-registration of MRI and ultrasound: accuracy of targeting based on radiology-pathology correlation. Transl Androl Urol 2017;6 :406–12.28725582
[14] Rosenkrantz AB , Verma S , Choyke P , Prostate magnetic resonance imaging and magnetic resonance imaging targeted biopsy in patients with a prior negative biopsy: a consensus statement by AUA and SAR. J Urol 2016;196 :1613–8.27320841
[15] Lai WS , Gordetsky JB , Thomas JV , Factors predicting prostate cancer upgrading on magnetic resonance imaging–targeted biopsy in an active surveillance population. Cancer 2017;123 :1941–8.28140460
[16] Gordetsky JB , Saylor B , Bae S , Prostate cancer management choices in patients undergoing multiparametric magnetic resonance imaging/ultrasound fusion biopsy compared to systematic biopsy. Urol Oncol 2018;36 :241.e7–241.e13.
[17] Othman E , Wang J , Sprague BL , Comparison of false positive rates for screening breast magnetic resonance imaging (MRI) in high risk women performed on stacked versus alternating schedules. Springerplus 2015;4 :77.25741458
[18] Cieszanowski A , Lisowska A , Dabrowska M , MR imaging of pulmonary nodules: detection rate and accuracy of size estimation in comparison to computed tomography. PLoS One 2016;11 :e0156272.27258047
[19] Verma S , Turkbey B , Muradyan N , Overview of dynamic contrast-enhanced MRI in prostate cancer diagnosis and management. AJR Am J Roentgenol 2012;198 :1277–88.22623539
[20] Djavan B , Zlotta A , Remzi M , Optimal predictors of prostate cancer on repeat prostate biopsy: a prospective study of 1,051 men. J Urol 2000; 163 :1144–8.10737484
[21] Keetch DW , Catalona WJ , Smith DS . Serial prostatic biopsies in men with persistently elevated serum prostate specific antigen values. J Urol 1994;151 :1571–4.7514690
[22] Hong CW , Walton-Diaz A , Rais-Bahrami S , Imaging and pathology findings after an initial negative MRI-US fusion-guided and 12-core extended sextant prostate biopsy session. Diagn Interv Radiol 2014;20 : 234–8.24509182
[23] Liddell H , Jyoti R , Haxhimolla HZ . mp-MRI prostate characterised PIRADS 3 lesions are associated with a low risk of clinically significant prostate cancer—a retrospective review of 92 biopsied PIRADS 3 lesions. Curr Urol 2015;8 :96–100.26889125
|
PMC008xxxxxx/PMC8983100.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
9421547
4136
Hum Pathol
Hum Pathol
Human pathology
0046-8177
1532-8392
31075299
8983100
10.1016/j.humpath.2019.04.016
NIHMS1783875
Article
PTEN and ERG detection in multiparametric magnetic resonance imaging/ultrasound fusion targeted prostate biopsy compared to systematic biopsy
Baumgartner Erin M.D. a
del Carmen Rodriguez Pena Maria M.D. a
Eich Marie-Lisa M.D. a
Porter Kristin K. M.D., Ph.D. b
Nix Jeffrey W. M.D. c
Rais-Bahrami Soroush M.D. bc1
Gordetsky Jennifer M.D. d*1
a Department of Pathology, University of Alabama at Birmingham, Birmingham, AL 35249, USA
b Department of Radiology, University of Alabama at Birmingham, Birmingham, AL 35249, USA
c Department of Urology, University of Alabama at Birmingham, Birmingham, AL 35249, USA
d Vanderbilt University Medical Center, Departments and Pathology and Urology, Nashville, TN 37232
1 Co-Senior Authors.
* Corresponding author at: Vanderbilt University Medical Center, Departments and Pathology and Urology, C-3321A MCN 1161 21st Avenue South, Nashville, TN 37232. [email protected] (J. Gordetsky).
10 3 2022
8 2019
07 5 2019
05 4 2022
90 2026
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Summary
Multiparametric magnetic resonance imaging (MRI)/ultrasound fusion targeted prostate biopsy has been shown to outperform systematic biopsy in the detection of clinically significant prostate cancer. Aside from tumor grade, tumor biomarkers such as phosphatase and tensin homolog (PTEN) and ETS-related gene (ERG) have prognostic significance in prostate cancer and may help direct management of patients with low-grade tumors. Our objective was to compare the detection of PTEN and ERG expression in MRI-targeted versus systematic prostate biopsies. We compared immunohistochemical expression for PTEN and ERG on prostate biopsy cores from patients with Grade Group (GG) 1 or GG2 prostate cancer who had undergone systematic biopsy with concurrent targeted biopsy. Fifty-three cases had both systematic and MRI-targeted prostate tissue available for staining for PTEN; and 52 cases, for ERG. ERG positivity was seen in 37/52 (71.2%) cases, and PTEN loss was seen in 15/53 (28.3%) cases. The detection of ERG expression was not significantly different between MRI-targeted and systematic biopsy (P = .4). Targeted biopsy was superior to systematic biopsy in the detection of PTEN loss (P = .02). MRI-targeted cores detected 14/15 (93.3%) cases of PTEN loss compared to 7/15 (46.7%) cases detected by systematic cores. Most cases with PTEN loss showed heterogeneous expression in both systematic and targeted cores. In 14/15 (93.3%) cases with PTEN loss, GG was the same between targeted and systematic biopsy. Targeted biopsy is superior to systematic biopsy in the detection of PTEN loss in GG1 and GG2 tumors. Inclusion of targeted cores may be helpful for evaluation of certain prognostic biomarkers.
Prostate adenocarcinoma
Phosphatase and tensin homolog
ETS-related gene
ERG
Magnetic resonance imaging
Pathology
pmc1. Introduction
In the United States, prostate cancer remains the most common malignancy in men, with the exception of skin cancer [1]. Multiparametric magnetic resonance imaging (MRI)/ultrasound (US) fusion targeted prostate biopsy has been shown in previous studies to outperform systematic biopsy in the detection of clinically significant prostate cancer [2–5]. Targeted biopsy has been shown to detect higher-grade tumors as well as poor prognostic features, such as perineural invasion and possibly extraprostatic extension [4,6–8]. In addition, the implementation of this new technique may affect how urologists manage patients with prostate cancer, with potentially an increased utilization of active surveillance [9,10]. Factors that influence selection of active surveillance include cancer grade, imaging findings, serum prostate specific antigen (PSA) level, and patient preference as part of a shared decision-making model with their physicians.
With the increased utilization of active surveillance and MRI/US fusion targeted biopsy, additional prognostic biomarkers can be useful in helping to determine which patients require definitive therapy. Phosphatase and tensin homolog (PTEN) and ETS-related gene (ERG) expression in prostate cancer has been shown to have prognostic significance and may help direct the management of patients with low- and intermediate-risk tumors [11]. In cases with tumor of the same grade involving multiple cores, tumor heterogeneity might influence the results of any biomarker study.
Although studies have shown that MRI-targeted prostate biopsy better detects higher-grade tumors, whether this technique better samples cancer tissue in terms of detecting prognostic biomarkers is undetermined. To date, there have been limited investigations of prognostic biomarkers in prostate cancer detected by MRI-targeted biopsy compared to systematic biopsy. Our objective was to compare the detection of PTEN and ERG expression in MRI-targeted versus systematic prostate core biopsies to determine if the targeted approach better detects expression of these prognostic biomarkers.
2. Materials and methods
A retrospective review was performed of our prospectively maintained, institutional review board-approved, prostate cancer database, searching for patients who underwent MRI/US fusion targeted prostate biopsy and concurrent systematic extended-sextant biopsy from 2014 to 2018. Post–image processing of multiparametric MRI and targeted biopsy of 3D segmented suspicious lesions was performed using the DynaCad and UroNav systems, respectively (Phillips/InVivo, Gainsville, FL). prostate imaging reporting and data system (PIRADS) v2 scoring was assigned by a multidisciplinary consensus conference with fellowship-trained radiologists and urologic oncologists specializing in prostate MRI. Two fellowship-trained urologic oncologists performed all targeted biopsies. For each patient, a minimum of 8 cores was sampled via the standard systematic approach using the conventional extended-sextant template. Each MRI-targeted lesion was sampled by at least 2 needle cores, and all prostate biopsy cores were evaluated for Gleason score and percent tumor involvement, per each MRI-targeted lesion, by the “aggregate of cores methods” as previously described [12].
All histologic evaluation was performed by 2 fellowship-trained genitourinary pathologists (J. B. G. and M. R. P.). Prostate cancer grading was assessed in accordance with the current standard criteria [13,14]. Statistical analyses were done utilizing JMP 13.1.0. Categorical and continuous variables were compared for statistical significance using χ2 test and Student t tests, respectively. The detection rate of ERG positivity and PTEN loss between systematic and MRI-targeted cores was compared using the McNemar test.
Patients were selected that had Grade Group (GG) 2 as the highest grade of prostate cancer detected among all cores sampled during the biopsy session, including both MRI-targeted and systematic cores. The presence of either GG1 or GG2 tumor on either targeted or systematic cores was acceptable for inclusion. No cases with GG3–5 or cases with intraductal carcinoma detected in any cores were included. For inclusion in this data analysis, it was required that prostate cancer was detected on both MRI-targeted cores as well as systematic cores within the case, although the same GG was not required to be present on both biopsy approaches.
Immunohistochemistry (IHC) for PTEN and ERG was performed on all cores with tumor available in each case. Criteria for PTEN loss were defined as per previously published studies [15] (Figure). ERG positivity was determined by nuclear staining on IHC. For each patient, prostate cancer IHC expression was compared between systematic and MRI-targeted cores.
3. Results
A total of 53 cases with GG1 or GG2 tumors were identified where patients had undergone systematic biopsy with concurrent MRI/US fusion targeted biopsy (Table 1). A total of 66 cancer-positive cores from the systematic approach and a total of 58 cancer-positive cores from the MRI-targeted approach were evaluated with IHC. The average number of total cores sampled was significantly higher for the systematic approach (11.8 ± 0.9) compared to the targeted approach (4.2 ± 2.0), P < .0001. An average of 2.2 ± 1.3 systematic biopsy cores and 2.0 ± 0.9 MRI-targeted biopsy cores per case contained cancer (P = .93). The average age was 66 years (range 44–81), and the average PSA was 6.2 ng/mL (range 1.5–17.2). Races present in our cohort included 41 whites, 10 African Americans, and 2 others. GG1 prostate cancer was the highest GG overall in 28/53 (52.8%) cases, and GG2 prostate cancer was the highest GG overall in 25/53 (47.2%) cases. In our cohort, 30/53 (56.6%) patients went on to choose active surveillance, 8/53 (15.1%) patients were treated with radical prostatectomy, 9/53 (17.0%), patients were treated with radiation therapy, and 6/53 (11.3%) patients were lost to follow-up.
PTEN staining and interpretation were performed on all 53 cases. Staining and interpretation for ERG status were performed on 52 cases; 1 case had tissue depleted after staining with PTEN, and therefore, ERG staining was unable to be performed on both the MRI-targeted and systematic cores.
ERG positivity was seen in 37/52 (71.2%) cases when assessing all biopsy cores available (Table 2). MRI-targeted biopsy detected 29/37 (78.4%) cases of ERG positivity compared to 32/37 (86.5%) cases detected by systematic biopsy. The detection of ERG expression was not significantly different between targeted and systematic biopsy (P = .4).
PTEN loss was seen in a total of 15/53 (28.3%) cases. Patients with PTEN loss consisted of 11 whites, 2 African Americans, and 2 others. MRI-targeted biopsy detected 14/15 (93.3%) cases of PTEN loss compared to 7/15 (46.7%) cases detected by systematic biopsy cores. MRI-targeted biopsy was superior to systematic biopsy in the detection of PTEN loss (P = .02) (Table 3). Of the 15 cases that had PTEN loss, 14/15 cases had tumor detected in the same regions on targeted and systematic approaches. Only 1 case had tumor detected in a location that was different between the standard and targeted biopsy. In this particular case, the targeted biopsy was positive in the left mid to left lateral base, whereas the standard biopsy was positive in the right apex and right mid gland. Of the 15 cases with PTEN loss, 12/15 (80%) also had gain of ERG expression. Heterogeneous expression was seen in most cases with PTEN loss. PTEN loss was found to be heterogeneous on systematic biopsy in 5/7 (71.4%) cases and on MRI-targeted biopsy in 11/14 (78.6%) cases.
In the 14/15 (93.3%) cases with PTEN loss, GG was the same between the MRI-targeted and systematic biopsy. In 1 case with PTEN loss, the GG was higher in the MRI-targeted cores (GG2) versus the systematic cores (GG1) sampled during the same biopsy session. PTEN loss was seen more often in GG2 tumors in both the systematic and MRI-targeted approach. For the systematic approach, PTEN loss was seen in GG1 tumors in 3/32 (9.4%) cases and in GG2 tumors in 4/21 (19.0%) cases (P = .31). For the MRI-targeted approach, PTEN loss was seen in GG1 tumors in 5/34 (14.7%) cases and in GG2 tumors in 9/19 (47.4%) cases (P = .01) (Table 4). There was no significant difference in race, age, PSA, grade group, or PIRADS score when comparing those patients with and without PTEN loss (Table 5). In the group where PTEN was retained, more patients (65.8%) went on to choose active surveillance versus those with PTEN loss (33.3%). However, this finding was not statistically significant (P = .053) and, importantly, was retrospective in nature, and biomarker information was not used to direct patient management.
Gain of ERG expression was seen more often in GG1 tumors than GG2 tumors in both the MRI-targeted and systematic approaches. For the systematic approach, ERG expression was seen in GG1 tumors in 22/31(71.0%) cases and in GG2 tumors in 10/21 (47.6%) cases (P = .7). For the MRI-targeted approach, ERG expression was seen in GG1 tumors in 19/33 (57.6%) cases and in GG2 tumors in 10/19 (52.6%) cases (P = .09). There was no significant difference in the detection of ERG expression in GG1 versus GG2 tumors.
4. Discussion
Prostate cancer is the most common noncutaneous cancer in the United States and the second leading cause of cancer-related deaths in American men. Improvements in MRI now allow for the more precise detection of prostate cancer by imaging [3,5,16,17]. Most notably, the combination of multiparametric MRI and superimposed real-time transrectal US allows for the improved targeting of suspicious lesions [2–4]. Multiple studies have shown the improved sensitivity and specificity of MRI/US fusion-targeted biopsies over the systematic biopsy approach in the detection of clinically significant prostate cancer, while requiring fewer cores [2–5]. In addition to the accurate detection of higher GGs, other clinically significant tumor characteristics such as extraprostatic extension, seminal vesicle invasion, and perineural invasion have also been demonstrated to be more frequently detected on MRI/US fusion biopsy [6–8]. However, despite the diagnostic improvements brought about by MRI targeted biopsy, the question of whether systematic biopsy can be eliminated entirely is still widely debated because of the potential risk of missing clinically significant disease by random sampling [18–20].
In 2014, the University of Alabama at Birmingham implemented the utilization of MRI/US fusion targeted prostate biopsies, with subsequent studies demonstrating optimization of detection of clinically significant prostate cancers [4,9]. In 2018, our group additionally assessed the management choices in patients who undergo MRI-targeted prostate biopsy compared to patients who undergo systematic biopsy [9]. It was found that patients who undergo MRI-targeted biopsy are 3.93 times more likely to choose active surveillance over early definitive treatment compared to men diagnosed on systematic biopsy alone ;, a preference which remained statistically significant after adjusting for age, PSAD, prior biopsy history, provider, tumor grade, and race [9]. As MRI-targeted biopsy becomes increasingly embraced by practitioners globally, it is important to see how prognostic biomarkers perform within this new biopsy technique.
PTEN and ERG mutations are two of the most common mutations found in prostate cancer, occurring in approximately 60% and 40% of prostate cancers, respectively [11]. As the most commonly deleted tumor suppressor gene in prostate cancer, PTEN loss is one of the most promising prognostic and predictive tissue-based biomarkers in prostate cancer [15,21]. A study by Lotan et al found that PTEN loss in biopsy specimens with Gleason score 3 + 3 = 6 prostate cancer is associated with an increased risk of Gleason score upgrading in the final radical prostatectomy specimen. This finding suggests that PTEN mutations may predict which tumors are undergraded on biopsy, thus helping guide clinical decision making [22]. ERG and PTEN mutations tend to occur together, with mouse studies implicating a symbiotic relationship in tumor oncogenesis [23,24]. These findings have led to the hypothesis that there may be a synergistic effect of ERG expression and PTEN loss on prostate cancer progression; several retrospective studies examining PTEN and ERG in human prostate cancer have established that PTEN mutations tend to occur following a preceding ERG mutation, with PTEN deletion occurring more commonly in ERG-rearranged prostate tumors [ 23,24]. Similar to this concept, in our study, we found that 80% of cases with PTEN loss also showed ERG expression on IHC.
We found a gain of ERG expression in 71.2% of cases in our cohort. This is somewhat higher than the results of previous studies, which report that the ERG gene is rearranged in approximately half of all prostate tumors [21]. This finding is likely the result of multiple factors and may reflect our particular patient demographic. Interestingly, we found no difference in ERG expression between the 2 cohorts, perhaps reflecting its high frequency in our population. ERG expression was seen more frequently in GG1 tumors; however, this was not statistically significant. As gain of ERG expression was found to be a much more prevalent finding, using MRI-targeted tissue for ERG expression may not hold an advantage over tissue acquired by a systematic approach.
As the PTEN gene is almost always lost by deletion in prostate cancer, fluorescence in situ hybridization (FISH) has traditionally been considered the criterion standard assay to detect in situ PTEN loss in tumor tissue [15]. However, the detection of PTEN loss also lends itself well to immunohistochemical assays. A multi-institutional cohort study conducted by Lotan et al investigated the sensitivity and specificity of PTEN immunohistochemistry relative to FISH for detection of PTEN gene deletion in prostate cancer and found that IHC staining had a 97% concordance with homozygous PTEN deletions detected by FISH, showing that automated PTEN immunohistochemistry assay is a sensitive method for the detection of homozygous PTEN gene deletions [15]. In our study, of the 7 systematic cores in which there was PTEN loss, 2 showed homozygous loss (28.6%), with the remaining 5 showing heterozygous loss (71.4%). Of the 14 MRI-targeted cores in which there was PTEN loss, 3 showed homozygous loss (21.4%), with the remaining 11 showing heterozygous loss (78.6%). This finding highlights the issue of tumor heterogeneity and its potential influence on the results of biomarker studies. A study conducted by Lotan et al in 2016 demonstrated that PTEN loss is commonly subclonal and heterogeneous in primary prostate tumors [15,21]. In general, previous PTEN FISH studies have shown that hemizygous deletions are more prevalent than homozygous deletions, a finding illustrated in our cohort, in which most PTEN loss cases were hemizygous [22]. Additional studies by Lotan et al have shown that protein loss is most strongly associated with shorter recurrence-free survival if the loss is homogeneous in all tumor cores sampled and that heterogeneous PTEN loss is a weaker prognostic indicator when compared to homogeneous loss [15,21].
In our study of patients with GG1 and GG2 tumors, we saw PTEN loss in 15/53 (28.3%) cases, which is higher than some other studies but may also reflect the presence of intermediate-risk tumors. Lotan et al showed that PTEN loss by immunohistochemistry is expected to be present in ~11% of Gleason score 6 biopsies overall, making it a rare finding in lower GGs [22]. In addition, they demonstrated that finding PTEN loss in a GG1 tumor may reflect unsampled higher-grade tumor elsewhere [22]. In our study, we also found an association between PTEN loss and higher grade. For the MRI-targeted approach, PTEN loss was seen in GG1 tumors in 14.7% of cases and in GG2 tumors in 47.4% of cases (P = .01). This association was only statistically significant in the MRI-targeted group. In addition, of the 15 cases in which PTEN protein was lost, there was a single case in which the GG differed between the systematic and targeted approaches. In this single case, PTEN loss was found in the target core, which showed GG2 tumor, whereas the systematic core showed GG1 tumor. Interestingly, PTEN loss was recorded as hemizygous in this case. These results can perhaps be explained by the combination of several factors at play. Clinically significant prostate cancer is more likely detected by MRI/US targeted biopsy, and PTEN loss is often associated with clinically significant tumor. In addition, in the targeted approach, more cores are sampled per lesion than the systematic approach, which relies on random sampling. Thereby, our finding of PTEN loss occurring more frequently in targeted cores suggests the ability to identify and target the more significant prostate cancer and sample it more thoroughly to account for the heterogeneity of PTEN expression. Indeed, of the 15 cases that had PTEN loss, 14/15 cases had tumor detected in the same regions on targeted and systematic approaches. Therefore, the difference in PTEN detection between the targeted biopsy and systematic biopsy in most cases was likely explained by tumor heterogeneity and the advantage of better sampling by a targeted approach. Only 1 case had tumor detected in a location that was different between the standard and targeted biopsy, in which the targeted biopsy was positive in the left mid to left lateral base whereas the standard biopsy was positive in the right apex and right mid gland. In this case, the difference in detection of PTEN may reflect multifocal prostate cancer. As MRI-targeted biopsy tissue successfully detected PTEN loss in 93.3% of cases, one could consider adapting a triage method for biomarker testing. One could start with testing of MRI-targeted tissue and then proceed to systematic cores as needed.
Notable limitations of this study are the relatively small sample size and retrospective nature. In the group where PTEN was retained, more patients (65.8%) went on to choose active surveillance versus those with PTEN loss (33.3%). However, this finding was not statistically significant (P = .053). This is expected because the study was retrospective in nature and clinicians did not have biomarker information to direct patient management. Future studies will need to further investigate if hemizygous PTEN deletions detected via IHC on biopsy accurately predict upgrading to allow for recommendations in terms of active surveillance. Other confounding factors include racial and ethnic backgrounds of the patient populations studied. Similar to other centers, at our institution, a lower number of African American men have undergone an MRI-targeted biopsy sessions compared to white men. Larger numbers would be required to evaluate the impact of prognostic biomarker expression in a race-stratified manner. Another potential factor may be access to health care by means of the referral patterns to our tertiary care center and distances associated with the regional area that our institution serves.
5. Conclusions
MRI/US fusion targeted biopsy is superior to systematic biopsy sampling in the detection of PTEN loss by IHC in GG1 and GG2 prostate cancer. Detection of ERG expression by IHC was equivalent between the MRI-targeted and systematic approach. These findings suggest that inclusion of tissue from MRI-targeted cores may be helpful for the assessment of some prognostic biomarkers.
Funding/Support:
This work was funded in part by a Junior Faculty Development Grant (ACS-IRG 001-53), by a pilot grant from the Young Supporters Board of the UAB Comprehensive Cancer Center, and by developmental funds from the UAB Comprehensive Cancer Center Support Grant (NCI P30 CA 013148) to Soroush Rais-Bahrami.
Figure Immunohistochemical stain for PTEN at high-power magnification showing (A) weak but positive PTEN staining in tumor cells, (B) strong positive PTEN staining in tumor cells, (C) heterogeneous loss of PTEN staining in tumor cells, and (D) homogeneous loss of PTEN staining in tumor cells.
Table 1 Clinical demographics of patient population that underwent MRI-targeted and systematic prostate needle core biopsy
Variable n (%)
No. of cases 53
Age (y), mean (range) 66 (44–81)
Race
African American 10 (18.9%)
White 41 (77.3%)
Other 2 (3.8%)
Family history
Yes 14 (26.4%)
No 38 (71.7%)
Unknown 1 (1.9%)
GG
1 28 (52.8%)
2 25 (47.2%)
PSA, mean ± SD 6.2 ± 3.4
PIRADS
3 10 (18.9%)
4 27 (50.9%)
5 16 (30.2%)
No. of cores obtained during biopsy, mean ± SD
Systematic 11.8 ± 0.9
Targeted 4.2 ± 2.0
No. of cancer-positive cores, mean ± SD
Systematic 2.2 ± 1.3
Targeted 2.0 ± 0.9
Management
Active surveillance 30 (56.6%)
Radiation therapy 9 (17.0%)
Radical prostatectomy 8 (15.1%)
Unknown 6 (11.3%)
Table 2 ERG status detected on systematic and MRI-targeted prostate needle core biopsies
MRI-targeted biopsy Systematic biopsy
ERG negative ERG positive Total
ERG negative 15 8 23
ERG positive 5 24 29
Total 20 32
NOTE. Total cases n = 52. P = .4.
Table 3 PTEN status detected on systematic and MRI-targeted prostate needle core biopsies
MRI-targeted biopsy Systematic biopsy
PTEN Loss PTEN Retained Total
PTEN loss 6 8 14
PTEN retained 1 38 39
Total 7 46
Total cases n = 53. P = .02.
Table 4 PTEN and ERG status on systematic and MRI-targeted biopsy by GG
GG Total ERG cases Total PTEN cases PTEN loss ERG gain
Systematic biopsy
1 31 32 3 22
2 21 21 4 10
MRI-targeted biopsy
1 33 34 5 19
2 19 19 9 10
Table 5 Comparative Clinical and Pathologic Characteristics of Patients with and without PTEN Loss.
Cases with PTEN loss Cases with no PTEN loss P
No. of cases 15 38
Detection on biopsy method, n (%)
Targeted cores 14 (93%)
Standard cores 7 (47%)
Targeted and standard cores 6 (40%)
Age (y), mean ± SD 66 ± 8 65 ± 8 .66
Race, n (%) 1.0
African American 2 (15%) 8 (21%)
White 11 (85%) 30 (79%) .36
GG, n (%)
1 6 (40%) 22 (58%)
2 9 (60%) 16 (42%)
PSA, mean ± SD 6.3 ± 4.0 6.1 ± 3.1 .79
PIRADS, n (%) .36
3 1 (7%) 9 (24%)
4 8 (53%) 19 (50%)
5 6 (40%) 10 (26%) .053
Treatment, n (%)
Active surveillance 5 (33.3%) 25 (65.8%)
Radiation therapy 5 (33.3%) 4 (10.5%)
Radical prostatectomy 3 (20.0%) 5 (13.2%)
Unknown 2 (13.3%) 4 (10.5%)
Competing interests: Jeffrey W. Nix and Soroush Rais-Bahrami serve as consultants for Philips/InVivo Corp. Soroush Rais-Bahrami also serves as a consultant to Blue Earth Diagnostics and Genomic Health Inc. All other authors report no conflicts of interest or financial disclosures that were pertinent to the following study.
References
[1] Siegel RL , Miller KD , Jemal A . Cancer statistics, 2019. CA Cancer J Clin 2019;69 (1 ):7–34.30620402
[2] Yarlagadda VK , Lai WS , Gordetsky JB , MRI/US fusion-guided prostate biopsy allows for equivalent cancer detection with significantly fewer needle cores in biopsy-naive men. Diagn Interv Radiol 2018;24 (3 ):115–20.29770762
[3] Siddiqui MM , Rais-Bahrami S , Turkbey B , Comparison of MR/ultrasound fusion-guided biopsy with ultrasound-guided biopsy for the diagnosis of prostate cancer. JAMA 2015;313 (4 ):390–7.25626035
[4] Gordetsky JB , Thomas JV , Nix JW , Rais-Bahrami S . Higher prostate cancer grade groups are detected in patients undergoing multiparametric MRI-targeted biopsy compared with standard biopsy. Am J Surg Pathol 2017;41 (1 ):101–5.27574875
[5] Kasivisvanathan V , Rannikko AS , Borghi M , MRI-targeted or standard biopsy for prostate-cancer diagnosis. N Engl J Med 2018; 378 (19 ):1767–77.29552975
[6] Truong M , Rais-Bahrami S , Nix JW , Messing EM , Miyamoto H , Gordetsky JB . Perineural invasion by prostate cancer on MR/US fusion targeted biopsy is associated with extraprostatic extension and early biochemical recurrence after radical prostatectomy. HUM PATHOL 2017;66 :206–11.28705708
[7] Gordetsky JB , Nix JW , Rais-Bahrami S . Perineural invasion in prostate cancer is more frequently detected by multiparametric MRI targeted biopsy compared with standard biopsy. Am J Surg Pathol 2016;40 (4 ):490–4.26523543
[8] Baumgartner EM , Porter KK , Nix JW , Rais-Bahrami S , Gordetsky JB . Detection of extraprostatic disease and seminal vesicle invasion in patients undergoing magnetic resonance imaging-targeted prostate biopsies. Trans Androl Urol 2018;7 (Suppl. 4 ):S392–6.
[9] Gordetsky JB , Saylor B , Bae S , Nix JW , Rais-Bahrami S . Prostate cancer management choices in patients undergoing multiparametric magnetic resonance imaging/ultrasound fusion biopsy compared to systematic biopsy. Urol Oncol 2018;36 (5 ):241.e7–241.e13.
[10] Lai WS , Gordetsky JB , Thomas JV , Nix JW . Rais-Bahrami S. Factors predicting prostate cancer upgrading on magnetic resonance imaging-targeted biopsy in an active surveillance population. Cancer 2017;123 (11 ):1941–8.28140460
[11] Ullman D , Dorn D , Rais-Bahrami S , Gordetsky J . Clinical utility and biologic implications of phosphatase and tensin homolog (PTEN) and ETS-related gene (ERG) in prostate cancer. Urology 2018;113 :59–70.29225123
[12] Gordetsky JB , Schultz L , Porter KK , Nix JW , Thomas JV . Del Carmen Rodriguez Pena M, et al. Defining the optimal method for reporting prostate cancer grade and tumor extent on magnetic resonance/ultrasound fusion-targeted biopsies. HUM PATHOL 2018;76 :68–75.29551676
[13] Epstein JI , Egevad L , Amin MB , The 2014 International Society of Urological Pathology (ISUP) consensus conference on Gleason grading of prostatic carcinoma: definition of grading patterns and proposal for a new grading system. Am J Surg Pathol 2016;40 (2 ):244–52.26492179
[14] Gordetsky J , Epstein J . Grading of prostatic adenocarcinoma: current state and prognostic implications. Diagn Pathol 2016;11 :25.26956509
[15] Lotan TL , Wei W , Ludkovski O , Analytic validation of a clinical-grade PTEN immunohistochemistry assay in prostate cancer by comparison with PTEN FISH. Mod Pathol 2016;29 (8 ):904–14.27174589
[16] Gordetsky JB , Ullman D , Schultz L , Histologic findings associated with false-positive multiparametric magnetic resonance imaging performed for prostate cancer detection Hum Pathol. 2019;83 :159–65.30179687
[17] Engelhard K , Labanaris AP , Bogner K , Lübke L , Dworak O , Kühn R . How good is post-biopsy multiparametric magnetic resonance imaging in detecting and characterising the index lesion of localised prostate cancer? Scand J Urol 2014;48 (6 ):499–505.24754780
[18] Ma TM , Tosoian JJ , Schaeffer EM , The role of multiparametric magnetic resonance imaging/ultrasound fusion biopsy in active surveillance. Eur Urol 2017 Feb;71 (2 ):174–80.27236496
[19] Hansen NL , Kesch C , Barrett T , Multicentre evaluation of targeted and systematic biopsies using magnetic resonance and ultrasound image-fusion guided transperineal prostate biopsy in patients with a previous negative biopsy. BJU Int 2017 Nov;120 (5 ):631–8.27862869
[20] Borkowetz A , Platzek I , Toma M , Direct comparison of multiparametric magnetic resonance imaging (MRI) results with final histopathology in patients with proven prostate cancer in MRI/ultrasonography-fusion biopsy. BJU Int 2016 Aug;118 (2 ):213–20.26935133
[21] Lotan TL , Wei W , Morais CL , PTEN loss as determined by clinical-grade immunohistochemistry assay is associated with worse recurrence-free survival in prostate cancer. Eur Urol Focus 2016;2 (2 ):180–8.27617307
[22] Lotan TL , Carvalho FL , Peskoe SB , PTEN loss is associated with upgrading of prostate cancer from biopsy to radical prostatectomy. Mod Pathol 2015;28 (1 ):128–37.24993522
[23] Carver BS , Tran J , Gopalan A , Aberrant ERG expression cooperates with loss of PTEN to promote cancer progression in the prostate. Nat Genet 2009;41 :619–24.19396168
[24] King JC , Xu J , Wongvipat J , Cooperativity of TMPRSS2-ERG with PI3-kinase pathway activation in prostate oncogenesis. Nat Genet 2009;41 :524–6.19396167
|
PMC008xxxxxx/PMC8983101.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
9815755
30139
Prostate Cancer Prostatic Dis
Prostate Cancer Prostatic Dis
Prostate cancer and prostatic diseases
1365-7852
1476-5608
30413806
8983101
10.1038/s41391-018-0107-0
NIHMS1783880
Article
Comparison of biparametric MRI to full multiparametric MRI for detection of clinically significant prostate cancer
Sherrer Rachael L. 1
Glaser Zachary A. 1
Gordetsky Jennifer B. 12
Nix Jeffrey W. 1
Porter Kristin K. 3
Rais-Bahrami Soroush http://orcid.org/0000-0001-9466-9925
13
1 Department of Urology, University of Alabama at Birmingham, Birmingham, AL, USA
2 Department of Pathology, University of Alabama at Birmingham, Birmingham, AL, USA
3 Department of Radiology, University of Alabama at Birmingham, Birmingham, AL, USA
✉ Soroush Rais-Bahrami, [email protected]
10 3 2022
5 2019
09 11 2018
05 4 2022
22 2 331336
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Background
Multiparametric magnetic resonance imaging (MP-MRI) and MRI/ultrasound (US) fusion-guided biopsy are becoming more widely used techniques for prostate cancer (PCa) diagnosis and management. However, their widespread adoption and use, where available, are limited by cost and added time. These limitations could be minimized if a biparametric MRI (BP-MRI) focusing on T2-weighted and diffusion-weighted imaging is performed. Herein we report the cancer detection rate of BP-MRI compared with full MP-MRI.
Methods
Biopsy-naive and prior negative biopsy patients with clinical suspicion for PCa underwent MP-MRI with an imaging protocol incorporating narrow field-of-view T2-weighted, diffusion-weighted, and DCE pelvic MRI. Then patients underwent MRI/US fusion-guided biopsy of target lesions between November 2013 and October 2017. The pathology results were compared to the positivity of the DCE sequence compared to the BP-MRI findings alone.
Results
There were 648 targeted lesions biopsied in 344 patients. We defined biparametric screen filter positivity as both T2-weighted and diffusion-weighted imaging positivity for the same lesion. The majority of target lesions (552/648, 85%) were screen filter positive. For those that were screen filter negative, a minority (14/96, 15%) had DCE-positive findings. Of these, 2/3 (67%) cancer-positive cases were seen on T2-weighted imaging. For those 82 that were screen filter negative and DCE negative, the DCE phase would not have added imaging suspicion. Only 3/82 (3.7%) were cancer positive; 2 with low risk, grade group 1 cancer and 1 with intraductal carcinoma, all identified on targeted T2-weighted MRI positivity.
Conclusions
BP-MRI for the evaluation of PCa and for guiding MRI/US fusion-targeted biopsy has the advantages of reducing cost, time, and contrast exposure of MP-MRI by eliminating the DCE phase. These benefits are realized without forfeiting valuable diagnostic information, as shown by similar cancer detection rates of BP-MRI and MP-MRI in this study, particularly for clinically significant cases of PCa.
pmcIntroduction
Prostate cancer (PCa) is the most common solid organ malignancy in American men and the leading cause of cancer mortality in this population, with 29,430 deaths projected in 2018 [1]. PCa is traditionally most often diagnosed on systematic transrectal ultrasound (TRUS)-guided extended-sextant prostate biopsy performed owing to suspicion from elevated prostate-specific antigen (PSA) or abnormal findings on digital rectal exam (DRE). Multiparametric magnetic resonance imaging (MP-MRI) and MRI/ultrasound (US) fusion-targeted biopsy (TBx) are relatively new techniques for detection of PCa that allow for targeted, image-guided sampling of lesions suspicious for cancer and reduction of sampling error, as well as evaluation of regional structures such as seminal vesicles, bones, and lymph nodes for PCa involvement. Studies have shown that TBx enables superior detection of clinically significant cases of PCa when compared to systematic biopsy alone [2, 3].
However, the use of MP-MRI to guide TBx is limited by several factors, including cost, time of image acquisition and interpretation, and use of intravenous, gadolinium contrast. MP-MRI of the prostate is comprised of anatomical T2-weighted imaging (T2W) in addition to functional imaging techniques, commonly diffusion-weighted imaging (DWI) and dynamic contrast enhancement (DCE) sequences. This typically requires 30–45 min of MRI gantry time, placement of intravenous access, and contrast administration [4, 5]. The use of contrast contributes additional cost and time, as well as the potential risk of gadolinium-based contrast administration complications. It has been proposed that these clinical risks and burdens could be reduced by eliminating the DCE sequence altogether. Instead, performing a more abbreviated biparametric-MRI (BP-MRI) would require less time (approximately 15–30 min less per case), cost less, and obviate the need for contrast administration and associated risks [4, 6]. As such, this diagnostic modality could significantly improve patient access while limiting many of the financial and time constraints institutions face with traditional MP-MRI [7].
While these are compelling reasons to eliminate DCE in this setting, the potential of reducing diagnostic accuracy must be considered. Therefore, BP-MRI must first be shown not to overlook clinically significant PCa that a DCE phase would detect. It has been previously reported that BP-MRI in men with elevated PSA has a similar diagnostic accuracy compared to full MP-MRI [4]. Herein we aimed to determine the cancer detection rate (CDR) of BP-MRI compared to traditional MP-MRI in men who were either biopsy naive or had one or more prior negative prostate biopsies to further investigate the necessity of DCE for reliable detection of clinically significant PCa.
Methods
An Institutional Review Board-approved, HIPAA (Health Insurance Portability and Accountability Act)-compliant retrospective review of prostate MP-MRI and MRI/US fusion-TBx records from November 2013 to October 2017 was performed. Patients meeting our inclusion criteria were those who underwent MP-MRI for clinical suspicion of PCa based on elevated serum PSA level and/or abnormal DRE without prior PCa diagnosis. This included both biopsy naive men as well as those with a prior negative biopsy. Patients with suspicious lesions on MP-MRI underwent TBx. All patients also underwent concurrent 12-core extended-sextant biopsy at the time of TBx if they had not undergone a separate standard template biopsy within a 12-month period. For patients with more than one MRI/US fusion-TBx, only the initial MP-MRI and TBx session data were included in this study. Our inclusion criteria are summarized in Fig. 1.
Our standard prostate MP-MRI protocol consisted of a baseline non-contrast T1 phase, triplanar T2W imaging, DCE imaging, and DWI with multiple b-values, including a high b-value sequence from which an apparent diffusion coefficient map was derived [8]. All images were reviewed by a body radiologist with subspecialization in genitourinary MRI on an Intellispace Portal (Philips Medical systems, Eindhoven, The Netherlands) Picture Archiving Communications System. These were secondarily reviewed in a multidisciplinary prostate imaging conference consisting of body radiologists, genitourinary pathologists, and urologic oncologists. In this setting, each suspicious lesion was discussed and assigned a PI-RADS v.2 suspicion score, with denotation of imaging parameter positivity and three-dimensional segmentation of the prostate volume and regions of interest for TBx guidance in the DynaCad postimage processing software platform (Philips/InVivo Corp, Gainesville, FL, USA) as previously described [9].
Next, the results were interpreted according to BP-MRI screen filter and DCE positivity. BP-MRI screen filter positivity was defined as a foci having both T2W and DWI positivity. The zonal and anatomical location of BP-MRI screen filter-negative lesions that were positive on traditional MP-MRI were recorded.
Results
A total of 344 men met the inclusion criteria for this study that afforded 648 target lesions for analysis. Our study population comprised a median age of 65 years (SD 7.69, interquartile range 60–70), 204 (59%) of which were white, 65 (19%) black, 2 (1%) Hispanic/Latino, 2 (1%) Asian, and 71 (21%) where this demographic was not available (Table 1). The study population had a median PSA of 7.61 ng/mL (SD 9.96, interquartile range 5.29–10.7). Ninety-five patients (28%) were biopsy naive, while 249 (72%) had a history of one or more prior negative prostate biopsy sessions. Of those who had prior negative biopsy sessions before undergoing MRI and TBx, 145 men had 1 prior biopsy session, and 104 had ≥2 (range 1–9, SD 1.14).
The imaging characteristics and subsequent surgical pathology of the 648 targeted lesions are recorded in Fig. 2. The majority of target lesions (552/648, 85%) were BP-MRI screen filter positive. Among these lesions, 201 (36%) were DCE positive.
For the 96 (15%) target lesions that were BP-MRI screen filter negative, imaging characteristics, anatomical location, and final surgical pathology were reviewed. A small minority (14/96, 15%) exhibited DCE positivity: 4 (29%) in the central gland (CG) and 10 (71%) in the peripheral zone (PZ). Among these, 3 (21%) harbored cancer. While two lesions were more favorable risk PCa (Gleason 3 + 3 = 6, grade group 1 and low volume Gleason 3 + 4 = 7, grade group 2), one did contain 4 + 5 = 9 disease. All but one cancer-positive case was appreciated on the T2W phase (Fig. 2).
Among the 82/96 (85%) BP-MRI screen filter and DCE-negative lesions, 26 (32%) were located in the CG, 53 (65%) in the PZ, 1 (1%) located at the CG/PZ junction, and 1 (2%) in the seminal vesicles. As such, the DCE phase would not have provided any additional clinical benefit nor upgraded PI-RADS v.2 suspicion score [10]. In these screen-negative cases, where only T2W or DWI was positive, surgical pathology found PCa in 3/82 (4%) lesions. All three lesions were detected on T2W imaging in isolation. Two cases were low-risk disease (Gleason 3 + 3 = 6, grade group 1) and one case harbored variant intraductal carcinoma, which could not formally be graded. The comprehensive pathology findings of screen-negative TBx lesions are illustrated in Fig. 2.
Discussion
MP-MRI and TBx are modern, ever-expanding techniques for optimized detection of PCa compared to traditional TRUS-guided systematic or extended-sextant biopsy. These novel modalities are becoming widely available and integrated into daily clinical practice [11]. Nevertheless, the substantial costs and experience necessary for conducting high-fidelity imaging protocols and interpreting these imaging studies continues to pose a challenge for many providers [12]. More specifically, traditional MP-MRI protocols are costly and time-consuming, and controversy exists surrounding the use of contrast. These are modifiable factors that have limited access to certain patient populations due to regulations imparted by insurance providers and health-care payer systems [13, 14]
Currently, most prostate biopsies are prompted by an elevated serum PSA level or abnormal DRE. However, PSA screening and systematic biopsy alone often find clinically insignificant PCa, which has been shown to lead to over-treatment with limited benefit in cancer-specific or overall survival measures [15, 16]. With the widespread use of active surveillance for lower-risk PCa, prostate MRI leads to better risk stratification and increased confidence for safe cancer surveillance and is supported in current American Urological Association and the Society for Abdominal Radiology guidelines [17–19]. Additionally, differentiation of clinically significant cases of cancer in patients initially diagnosed with possible active surveillance-eligible low-risk disease has been quantified and calculated with MRI-based risk calculators [8]. Evidence supporting the role of MRI as a screening tool in biopsy-naive men had been limited to smaller series and relied largely on the negative predictive value of the diagnostic imaging study [9, 20]. However, recent investigation by an international, multicenter randomized trial demonstrated the benefit of MRI and MRI-directed biopsy over the standard TRUS-guided systematic biopsy approach in a population of biopsy-naive men [21]. As MRI finds more wide-spread adoption and implementation in the workflow of PCa detection, even prior to first biopsy, it is imperative to be cost-conscious to allow for wider public health reach of benefit.
One potential strategy for reducing the cost, study time, and contrast-associated risks is elimination of the DCE MRI phase, thus relying solely on BP-MRI that consists of just the T2W and DWI MRI sequences. This would require less time and less overall imaging and radiologic interpretation-associated expenses, as well as obviating the need for intravenous catheter insertion and contrast administration. However, up to this point, there is a paucity of evidence in support of omitting the DCE phase and only obtaining sequences in a BP-MRI study to safely detect PCa in this setting.
BP-MRI is a less-invasive imaging modality that has potential as a more rapid, affordable screening tool for PCa coupled with standard screening with PSA and DRE. A multidisciplinary group at the National Cancer Institute evaluated 143 biopsy-naive men who underwent full MP-MRI. They observed that using the BP-MRI sequences alone superiorly predicted the presence of PCa compared to PSA level or PSA density (PSAD) alone. When analyzed in conjunction with PSA and PSAD, integrating BP-MRI data significantly improved the sensitivity and specificity of clinically significant PCa detection [6]. These findings were further validated by the same center in a subsequent cohort of 59 biopsy-naive patients [22].
Kuhl et al. reported similar findings in their evaluation of 542 men with an elevated serum PSA (≥3 ng/mL) who underwent MP-MRI with a separate BP-MRI interpretation [4]. In their study, complete MP-MRI with contrast enhancement only affected the diagnosis of one clinically significant case of PCa out of a total of 139 (<1%), which would not have been otherwise detected with BP-MRI. In the current study, performing the DCE phase of the MP-MRI would permit detection of three screen-negative lesions that were PCa positive on TBx. However, two of these lesions would have been identified and targeted based upon T2W. Kuhl et al. found that the diagnostic accuracy of BP-MRI for clinically significant PCa (89.1%, 483 of 542) was comparable, if not slightly better, than that of the full MP-MRI with DCE (87.2%, 473 of 542) for evaluation and suspicion scoring prior to biopsy. In their study, BP-MRI was achieved in <9 min while providing “diagnostic accuracy and CDRs that are equivalent to those of conventional full multiparametric contrast-enhanced MR imaging protocols,” which is a much longer diagnostic study requiring 30–45 min to perform [4].
It should be mentioned that radiologists were not blinded to the DCE phase of each case, and therefore it is possible that their interpretation of the T2W and DWI sequences and PI-RADS scoring may have been affected by their viewing of the DCE series. Furthermore, one would consider whether a prospective analysis with only BP-MRI studies as a screening in biopsy-naive men would render different CDRs for each suspicion score level compared to those we have previously reported for our full MP-MRI protocol of imaging [9]. In our patient series, we were not able to identify any specific patient-related features where the data from DCE MRI would augment the decision-making process and detection of PCa, consistent with the PI-RADS v.2 suspicion scoring system where DCE findings play only a minor role in determining the presence of clinically significant PCas, only pertinent in cases where otherwise equivocal lesions scored a PI-RADS 3 based on T2W and DWI features would benefit from DCE data [23, 24]. Additionally, the number of cases with a single MRI sequence prompting TBx and PCa detection was very limited and did not allow for a well-powered statistical analysis of the accuracy of each independent MRI parameter.
Depending on an institution’s policies for renal function and gadolinium contrast, adding the DCE phase to MP-MRI may increase patient preparation time for glomerular filtration rate (GFR) assessment or possibly even exclude them. Even if an institution does not perform GFR assessment with Group II agents, patients must be screened for prior contrast reactions and potentially premedicated, if indicated [25]. Furthermore, intravenous access must be obtained that in itself can be a time-consuming task. In total, this likely adds at least 20 min to the imaging study time, which has been estimated at approximately 45 min based on prior publications [6, 7]. Interpreting the DCE phase adds an average 5 min per MP-MRI for a subspecialty body imager. When taken together, the time savings of a BP-MRI when compared to MP-MRI has been estimated by two independent institutions to be approximately three-fold, potentially reducing prostate MRI volume/access by 50–67% when performing these studies during normal business hours [4, 7].
Additional benefits of reduced scan duration and table time include patient comfort and satisfaction. Anxiety and claustrophobia during prostate MRI can be significant barriers for some patients, and increased patient compliance during the scan reduces motion and artifacts, improving the study’s image quality. Improved patient experience leads to increased compliance both during the scan itself and for follow-up imaging, which is particularly important for patients undergoing active surveillance.
Immediate hypersensitivity reactions to gadolinium contrast are very rare, occurring in approximately 0.079% of all administrations. Although most immediate hypersensitivity reactions are mild, the incidence of moderate and severe reactions together is 0.013% per MR contrast media dose and 0.021% per person, which is rare but not low enough to be disregarded [26]. Furthermore, residual gadolinium has been shown to be deposited in tissues, particularly in the brain, and the effects of this deposition are currently unknown [27]. While gadolinium retention has not been directly linked to adverse health effects in patients with normal kidney function, in 2017 the Food and Drug Administration (FDA) issued a new class warning to alert the public about the potential for it to remain in the body for months or years after receiving the gadolinium contrast agent. Resultantly, the FDA now mandates that patients receive proper counseling on this matter before receiving a gadolinium contrast agent [28]. Our study suggests that BP-MRI has equitable PCa detection compared to traditional MP-MRI, so using this curtailed MRI study would eliminate these possible gadolinium-associated risks while still maintaining the diagnostic benefits.
Conclusions
BP-MRI for the detection of PCa and for guiding MRI/US fusion-targeted prostate biopsy has the advantages of reducing cost, time, and contrast exposure compared to traditional MP-MRI by eliminating the DCE phase. Of those with biparametric screen-negative findings, DCE did not in isolation find any clinically significant PCa. This abbreviated imaging study does not significantly sacrifice diagnostic yield and may permit greater access to the diagnostic benefits of prostate MRI in this setting.
Acknowledgements
This work was funded in part by Junior Faculty Development Grant (ACS-IRG 001-53) and by Developmental funds from the UAB Comprehensive Cancer Center Support Grant (NCI P30 CA 013148) to Soroush Rais-Bahrami.
Fig. 1 Inclusion and exclusion criteria for MP-MRIs included in the study analysis
Fig. 2 Pathology results of MRI/US fusion-targeted lesions biopsied
Table 1 Patient demographics
All Biopsy naive Prior negative
Age (years) 65 (7.69) 65 (7.76) 65 (7.67)
<50 12 5 7
51–60 75 16 59
61–70 177 45 132
71+ 80 29 51
Race
White 204 69 135
Black/African 65 12 53
American
Hispanic/Latino 2 0 2
Asian 2 2 0
Patient refused/declined 71 12 59
PSA (ng/mL) 7.61 (9.96) 6.00 (7.90) 8.40 (10.6)
Compliance with ethical standards
Conflict of interest JWN and SR-B serve as consultants to Philips/InVivo Corp. The other authors declare that they have no conflict of interest.
References
1. Siegel RL , Miller KD , Jemal A . Cancer statistics, 2018. CA Cancer J Clin. 2018;68 :7.29313949
2. Schoots IG , Roobol MJ , Nieboer D , Bangma CH , Steyerberg EW , Hunink MG . Magnetic resonance imaging-targeted biopsy may enhance the diagnostic accuracy of significant prostate cancer detection compared to standard transrectal ultrasound-guided biopsy: a systematic review and meta-analysis. Eur Urol. 2015;68 :438.25480312
3. Siddiqui MM , Rais-Bahrami S , Turkbey B , George AK , Rothwax J , Shakir N , Comparison of MR/ultrasound fusion-guided biopsy with ultrasound-guided biopsy for the diagnosis of prostate cancer. JAMA. 2015;313 :390.25626035
4. Kuhl CK , Bruhn R , Krämer N , Nebelung S , Heidenreich A , Schrading S . Abbreviated biparametric prostate MR imaging in men with elevated prostate-specific antigen. Radiology. 2017;285 :493–505.28727544
5. Bjurlin MA , Meng X , Nobin JL , Wysock JS , Lepor H , Rosenkrantz AB , Optimization of prostate biopsy: the role of magnetic resonance imaging targeted biopsy in detection, localization and risk assessment. J Urol. 2014;192 :648–58.24769030
6. Rais-Bahrami S , Siddiqui MM , Vourganti S , Turkbey B , Rastinehad AR , Stamatakis L , Diagnostic value of biparametric magnetic resonance imaging (MRI) as an adjunct to prostate-specific antigen (PSA)-based detection of prostate cancer in men without prior biopsies. BJU Int. 2015;115 :381–8.24447678
7. Porter KK , King A , Galgano S , Financial implications of biparametric prostate MRI. In preparation (2018).
8. Lai WS , Gordetsky JB , Thomas JV , Nix JW , Rais-Bahrami S , Factors predicting prostate cancer upgrading on magnetic resonance imaging-targeted biopsy in an active surveillance population. Cancer. 2017;123 :1941–8.28140460
9. Yarlagadda VK , Lai WS , Gordetsky JB , Porter KK , Nix JW , Thomas JV , MRI/US fusion-guided prostate biopsy allows for equivalent cancer detection with significantly fewer needle cores in biopsy-naive men. Diagn Interv Radiol. 2018;24 :115–20.29770762
10. Gaur S , Harmon S , Mehralivand S , Bednarova S , Calio BP , Sugano D , Prospective comparison of PI-RADS version 2 and qualitative in-house categorization system in detection of prostate cancer. J Magn Reson Imaging. 2018;48 :1326–35.29603833
11. Muthigi A , Sidana A , George AK , Kongnyuy M , Maruf M , Valayil S , Current beliefs and practice patterns among urologists regarding prostate magnetic resonance imaging and magnetic resonance-targeted biopsy. Urol Oncol. 2017;35 :32. e1–7.
12. Vourganti S , Starkweather N , Wojtowycz A . MR/US fusion technology: what makes it tick? Curr Urol Rep. 2017;18 :20.28233228
13. Gorin MA , Walsh PC . Magnetic resonance imaging prior to first prostate biopsy-are we there yet? Eur Urol. 2018. 10.1016/j.eururo.2018.05.018
14. Faria R , Soares MO , Spackman E , Ahmed HU , Brown LC , Kaplan R , Optimising the diagnosis of prostate cancer in the era of multiparametric magnetic resonance imaging: a cost-effectiveness analysis based on the Prostate MR Imaging Study (PROMIS). Eur Urol. 2018;73 :23–30.28935163
15. Wilt TJ , Brawer MK , Jones KM , Barry MJ , Aronson WJ , Fox S , Radical prostatectomy versus observation for localized prostate cancer. N Engl J Med. 2012;367 :203–13.22808955
16. Glaser ZA , Gordetsky JB , Porter KK , Varambally S , Rais-Bah-rami S . Prostate cancer imaging and biomarkers guiding safe selection of active surveillance. Front Oncol. 2017;7 :256.29164056
17. Loeb S , Byrne N , Makarov DV , Lepor H , Walter D . Use of conservative management for low-risk prostate cancer in the Veterans Affairs Integrated Health Care System from 2005–15. JAMA. 2018. 10.1001/jama.2018.5616
18. Gordetsky JB , Saylor B , Bae S , Nix JW , Rais-Bahrami S . Prostate cancer management choices in patients undergoing multiparametric magnetic resonance imaging/ultrasound fusion biopsy compared to systematic biopsy. Urol Oncol. 2018;36 :241. e7–13.
19. Rosenkrantz AB , Verma S , Choyke P , Eberhardt SC , Eggener SE , Gaitonde K , Prostate magnetic resonance imaging and magnetic resonance imaging targeted biopsy in patients with a prior negative biopsy: a Consensus Statement by AUA and SAR. J Urol. 2016;196 :1613–8.27320841
20. Ahmed HU , El-Shater Bosaily A , Brown LC , Gabe R , Kaplan R , Parmar MK , Diagnostic accuracy of multi-parametric MRI and TRUS biopsy in prostate cancer (PROMIS): a paired validating confirmatory study. Lancet. 2017;389 :815–22.28110982
21. Kasivisvanathan V , Rannikko AS , Borghi M , Panebianco V , Mynderse LA , Vaarala MH , MRI-targeted or standard biopsy for prostate-cancer diagnosis. N Engl J Med. 2018;378 :1767–77.29552975
22. Fascelli M , Rais-Bahrami S , Sankineni S , Brown AM , George AK , Ho R , Combined biparametric prostate magnetic resonance imaging and prostate-specific antigen in the detection of prostate cancer: a validation study in a biopsy-naïve patient population. Urology. 2016;88 :125–34.26680244
23. Weinreb JC , Barentsz JO , Choyke PL , Cornud F , Haider MA , Macura KJ , PI-RADS prostate imaging - reporting and data system: 2015, Version 2. Eur Urol. 2016;69 :16–40.26427566
24. Sheridan AD , Nath SK , Syed JS , Aneja S , Sprenkle PC , Weinreb JC , Risk of clinically significant prostate cancer associated with prostate imaging reporting and data system category 3 (equivocal) lesions identified on multiparametric prostate MRI. AJR Am J Roentgenol. 2017;210 :347–57.29112469
25. American College of Radiology. Manual on contrast media, version 10.3. American College of Radiology; 2017.
26. Jung JW , Kang HR , Kim MH , Lee W , Min KU , Han MH , Immediate hypersensitivity reaction to gadolinium-based MR contrast media. Radiology. 2012;264 :414–22.22550309
27. Malayeri AA , Brooks KM , Bryant LH , Evers R , Kumar P , Reich DS , National Institutes of Health perspective on reports of gadolinium deposition in the brain. J Am Coll Radiol. 2016;13 :237–41.26810815
28. FDA in brief: FDA requires new class warning and additional research on retention in the body of gadolinium from gadolinium-based contrast agents used in magnetic resonance imaging. 2017. https://www.fda.gov/NewsEvents/Newsroom/FDAInBrief/ucm589604.htm
|
PMC008xxxxxx/PMC8983102.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
101190326
33174
J Am Coll Radiol
J Am Coll Radiol
Journal of the American College of Radiology : JACR
1546-1440
1558-349X
30447932
8983102
10.1016/j.jacr.2018.09.026
NIHMS1783861
Article
Patient Demographics and Referral Patterns for [F-18]Fluciclovine-PET Imaging at a Tertiary Academic Medical Center
Galgano Samuel J. MD Department of Radiology, University of Alabama at Birmingham, Birmingham, Alabama.
Calderone Carli E. MD Department of Radiology, University of Alabama at Birmingham, Birmingham, Alabama.
McDonald Andrew M. MD Department of Radiation Oncology, University of Alabama at Birmingham, Birmingham, Alabama.
Nix Jeffrey W. MD Department of Urology, University of Alabama at Birmingham, Birmingham, Alabama.
deShazo Mollie MD Division of Medical Oncology, Department of Medicine, University of Alabama at Birmingham, Birmingham, Alabama.
Yang Eddy S. MD, PhD Department of Radiation Oncology, University of Alabama at Birmingham, Birmingham, Alabama.
McConathy Jonathan E. MD, PhD Department of Radiology, University of Alabama at Birmingham, Birmingham, Alabama.
Rais-Bahrami Soroush MD Department of Radiology, University of Alabama at Birmingham, Birmingham, Alabama.
Department of Urology, University of Alabama at Birmingham, Birmingham, Alabama.
Samuel J. Galgano, MD: Department of Radiology, University of Alabama at Birmingham, 619 19th St S, JTN 325, Birmingham, AL 35249; [email protected].
10 3 2022
3 2019
14 11 2018
05 4 2022
16 3 315320
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
pmcINTRODUCTION AND MOTIVATION
Prostate cancer is the most common noncutaneous malignancy in men in the United States with an estimated 164,690 new cases diagnosed in 2018 [1]. Approximately one in nine men will be diagnosed with prostate cancer in their lifetime, but many do not die from the disease [1]. There are an estimated 2.9 million men living in the United States diagnosed with prostate cancer at some point in their lives either on surveillance or after treatment [1]. However, approximately 29,430 deaths from prostate cancer occurred in 2018 (1 in 41 men), making it the second leading cause of cancer-related death in American men [1].
In May 2016, the US FDA approved the use of fluciclovine in patients with elevated prostate-specific antigen (PSA) in the setting of suspected biochemically recurrent or persistent prostate cancer. Fluciclovine targets the transmembrane amino acid transporters ACST2 and LAT1, both of which are overexpressed in prostate cancer cells [2,3]. Fluciclovine has been shown to improve the ability to detect metastatic disease when compared with conventional imaging (CT abdomen and pelvis or MRI pelvis and skeletal scintigraphy with 99mTc-labeled methylene-diphosphonate) and choline-PET/CT [4–7]. The most recent National Comprehensive Cancer Network (NCCN) Clinical Practice Guidelines in Oncology: Prostate Cancer Version 2.2018 included fluciclovine-PET/CT and PET/MRI as a consideration in patients who experience PSA persistence or recurrence after either radical prostatectomy or radiation therapy [8].
To date, there have been very little published data on utilization trends or referral patterns for standard of care fluciclovine-PET/CT and PET/MRI examinations. Thus, the primary aim of the article is to report and discuss patient demographics and referral patterns for fluciclovine-PET/CT and PET/MRI at a tertiary care academic medical institution that was an early adopter of this PET imaging modality.
PATIENT IDENTIFICATION
An institutional review board–approved, HIPAA-compliant search of the institutional database was performed to identify all patients who underwent fluciclovine-PET/CT or PET/MRI at our institution between January 1, 2017, and June 30, 2018. Patients who underwent fluciclovine-PET/CT or PET/MRI scan for research purposes were excluded, and only patients who underwent the scan as part of the clinical workup were included in the final analysis. After identification of the patient cohort, the medical record was accessed to obtain patient age, race, prior treatment history, PSA level at time of fluciclovine-PET scan order, zip code of primary residence, and insurance status. Distance between the patient zip code of primary residence and our institution was determined through the use of Google Maps. All graphs and statistical analyses were performed through the use of Microsoft Excel (Microsoft Corporation, Redmond, Washington) and JMP v.10 (SAS, Cary, NC).
RESULTS
A total of 72 fluciclovine-PET scans were performed at our institution from January 1, 2017 through June 30, 2018 (Fig. 1). Patient demographics and insurance coverage are reported in Table 1. Two patients opted to pay independently for their fluciclovine-PET scan, with an estimated out-of-pocket cost of over $4,000 per case. Patient prescan PSA values are also illustrated in Table 1. The large range in serum PSA values is due to several patients with known biochemical recurrence who underwent fluciclovine-PET as part of radiation treatment planning to distant metastatic disease. Distances traveled for the examination are illustrated in Figure 2. Treatment histories before the examination are summarized in Figure 3. The specialties of the referring physicians and internal and external referrals are summarized in Figure 4. The number of clinical examinations performed per quarter is demonstrated in Figure 5, which indicates progressive increase in utilization since January 2017.
DISCUSSION
The volume of [18F]fluciclovine-PET examinations at our institution continues to increase; we currently perform at least one study per week on a regular basis. Several additional private hospitals in the metropolitan area are offering fluciclovine-PET studies, and hence the overall volume of examinations performed in our patient catchment region is likely much higher. The majority of these studies will continue to be performed as PET/CT, but increasing consideration is being given to PET/MRI owing to its superior soft-tissue characterization within the pelvis [9–11]. However, reimbursements for clinical PET/MRI remain a challenge, because no existing Current Procedural Terminology (CPT) code exists for PET/MRI examinations and instead the examination is submitted as two separate CPT codes (78813 for whole-body PET and 72197 for pelvic MRI with and without contrast). Several attempted PET/MRI examinations have been denied by insurance companies because of CPT code 78813, not including anatomic imaging for localization, although these precertifications were submitted simultaneously with the MRI precertification.
As expected, the vast majority of patients are adult men greater than 65 years old. The majority of patients who underwent fluciclovine-PET examinations were white (69%), with only 11% of scans performed on African American patients. This is out of proportion to our racial breakdown for all PET examinations performed at our institution, where the demographic makeup is approximately 76% white and 16% African American. Additionally, this is far less than the proportion of African Americans who undergo multiparametric prostate MRI (32%) and who undergo MRI/US fusion biopsy (29%) at our institution [12]. This leads to the question of whether there are racial disparities in utilization or simply early adoption of this advanced oncologic PET imaging technology, which deserves further study in a larger population.
A major challenge in the implementation and increase in volume of [18F]fluciclovine-PET examinations at our institution has been preauthorization and precertification by private insurance companies as health care payers. The majority of our patients who have undergone [18F]fluciclovine-PET examinations claim Medicare as their primary insurance coverage. Coverage of this imaging modality in the setting of its FDA approval has allowed for the performance of these examinations in patients carrying Medicare coverage without any precertification difficulty. However, for patients with private insurance coverage, there has been considerable difficulty obtaining precertification for the examination despite FDA approval for its use in the biochemically recurrent prostate cancer setting posttreatment. Ultimately, this has led to some patients not obtaining preauthorization from their commercially held health care payers, for which a few have proceeded with the study by accepting financial responsibility and out-of-pocket cost for the PET/CT examination and interpretation of the PET portion of the study. As of recently, given the inclusion of [18F] fluciclovine-PET/CT and PET/MRI in the NCCN Guidelines, the drug has been reclassified and is considered to be part of standard of care for patients with suspected biochemically recurrent prostate cancer, although some commercial health care payers have not yet approved payment for its use in this setting.
Median PSA values for patients with suspected biochemical recurrence referred for [18F]fluciclovine-PET examinations were within the expected range. With the exception of one patient with a disparately elevated PSA > 1,500 ng/mL, many of these patients presented for imaging at the earlier times of recurrence based on PSA biochemical recurrence without any evidence of recurrent or metastatic disease on conventional imaging. Several of the patients imaged with this PET modality were known to have metastatic disease, but the clinical picture was confounded by the presence of other known malignancies, including squamous cell lung cancer and small lymphocytic lymphoma where [18F]fluciclovine-PET/CT was ordered to attempt to differentiate which primary malignancy was likely the source of the adenopathy seen on prior imaging.
Despite serving a known large catchment area, the distance traveled by patients for [18F]fluciclovine-PET examinations at our institution was intriguing. Over 50% of the patients who underwent [18F] fluciclovine-PET examinations at our institution traveled greater than 51 miles for their study. Even with a geographically large metropolitan area, it seems that the majority of our patients are traveling from beyond the metropolitan region. This suggests a lack of access to fluciclovine-PET examinations in more rural areas of the state and adjacent states and likely indicates prompt referral to the centralized tertiary academic center. Approximately 75% of fluciclovine-PET examinations were ordered by physicians within our hospital system, which again suggests that patients are traveling further within our catchment area to get clinical and imaging services that are not being offered in more rural areas of the state.
The most common specialty of the referring physicians for fluciclovine-PET examinations was urology. This is not a surprising observation, given that urologists are frequently the first clinicians seen by patients with prostate cancer and are maintained throughout their treatment course often to the level of recurrent or metastatic disease. At our institution, medical oncology is the second most frequent ordering provider, owing to the large number of referrals obtained for patients that have experienced biochemical recurrence and failed both radical prostatectomy and pelvic salvage radiotherapy. This poses an interesting question if fluciclovine-PET examinations will be utilized in the medical oncology setting for monitoring of disease response and progression in a fashion similar to fluorine-18–2-fluoro-2-deoxy-D glucose PET examinations. If this presents a potential role similar to fluorine-18–2-fluoro-2-deoxy-D glucose PET imaging, we would expect to see a substantial increase in volume of fluciclovine-PET examinations as more patients are being followed with serial examinations. Radiation oncologists stand to benefit from fluciclovine-PET examinations by establishing eligibility for salvage or image-targeted stereotactic radiation therapy and for potentially improved delineation of tumor volumes and boundaries for contouring of radiation treatment planning.
A known limitation of our study is its retrospective, single-institution analysis. It is likely that observations made in this article may not reflect utilization of fluciclovine-PET in the community as a whole. However, the authors feel that this article is an early stepping stone for further research into utilization and disparities in fluciclovine-PET utilization.
CONCLUSIONS
[18F]Fluciclovine-PET/CT and PET/MRI are being increasingly ordered by clinicians, and use is expected to further increase based on recent inclusion in NCCN guidelines and associated support for payment for this FDA-approved diagnostic study by health care payers. Potential geographic and racial barriers to utilization may exist, and further research on the topic is warranted.
S.J.G. and J.E.M. receive research support from Blue Earth Diagnostics and the Radiological Society of North America. S.R.B. is a member of the advisory board for Blue Earth Diagnostics. The other authors state that they have no conflict of interest related to the material discussed in this article.
Fig 1. Flowchart demonstrating number of both clinical and research fluciclovine-PET examinations performed at our institution.
Fig 2. Distance traveled by patients for fluciclovine-PET at our institution.
Fig 3. Treatments previously undergone by patients undergoing fluciclovine-PET at our institution. ADT = androgen-deprivation therapy.
Fig 4. Distribution of referring clinician specialties for fluciclovine PET examinations.
Fig 5. Number of clinical fluciclovine-PET examinations performed by quarter at our institution.
Table 1. Patient demographics and serum PSA values (ng/mL) of patients undergoing [18F]flucidovine-PET
Age
Mean 70.4
SD 7.6
Range 49–91
Race Number of patients
White 50
African American 8
Decline to respond 5
Not available 9
Insurance Number of patients
Medicare 43
Blue Cross Blue Shield 16
VA 4
Other 7
Self-pay 2
PSA ng/mL
Median 2.95
Average 31.23
SD 213.10
Min 0.00
Max 1812.87
Min = minimum; Max = maximum; PSA = prostate-specific antigen; VA = Veterans Administration.
REFERENCES
1. American Cancer Society. Facts and figures 2018. Available at https://www.cancer.org/research/cancer-facts-statistics/allcancer-facts-figures/cancer-facts-figures2018.html. Accessed June 3, 2018.
2. Segawa A , Nagamori S , Kanai Y , L-type amino acid transporter 1 expression is highly correlated with Gleason score in prostate cancer. Mol Clin Oncol 2013;1 : 274–80.24649160
3. Oka S , Okudaira H , Ono M , Differences in transport mechanisms of trans-1-amino-3-[18F]fluorocyclobutanecarboxylic acid in inflammation, prostate cancer, and glioma cells: comparison with L-[methyl-11C]methionine and 2-deoxy-2-[18F]flu-oro-D-glucose. Mol Imaging Biol 2014;16 : 322–9.24136390
4. Suzuki H , Inoue Y , Fujimoto H , Diagnostic performance and safety of NMK36 (trans-1-amino-3-[18F] fluorocyclobutanecarboxylic acid)-PET/CT in primary prostate cancer: multicenter phase IIb clinical trial. Jpn J Clin Oncol 2017;47 :283.27920097
5. Schuster DM , Nieh PT , Jani AB , Anti-3-[(18)F]FACBC positron emission tomography-computerized tomography and (111)In-capromab pendetide single photon emission computerized tomography-computerized tomography for recurrent prostate carcinoma: results of a prospective clinical trial. J Urol 2014;191 :1446–53.24144687
6. Odewole OA , Tade FI , Nieh PT , Recurrent prostate cancer detection with anti-3-[(18)F]FACBC PET/CT: comparison with CT. Eur J Nucl Med Mol Imaging 2016;43 :1773–83.27091135
7. Nanni C , Zanoni L , Pultrone C , (18)F-FACBC (anti1-amino-3-(18)F-fluorocyclobutane-1-carboxylic acid) versus (11)C-choline PET/CT in prostate cancer relapse: results of a prospective trial. Eur J Nucl Med Mol Imaging 2016;43 :1601–10.26960562
8. National Comprehensive Cancer Network. NCCN Clinical Practice Guidelines in Oncology: Prostate Cancer Version 2. 2018. Available at: https://www.nccn.org/professionals/physician_gls/PDF/prostate.pdf. Accessed June 3, 2018.
9. Muller BG , Kaushal A , Sankineni S , Multiparametric magnetic resonance imaging-transrectal ultrasound fusion-assisted biopsy for the diagnosis of local recurrence after radical prostatectomy. Urol Oncol 2015;33 :425.e1–6.
10. Valle LF , Greer MD , Shih JH , Multiparametric MRI for the detection of local recurrence of prostate cancer in the setting of biochemical recurrence after low dose rate brachytherapy. Diagn Interv Radiol 2018;24 :46–53.29317377
11. Patel P , Mathew MS , Triliski I , Multiparametric MR imaging of the prostate after treatment of prostate cancer. Radiographics 2018;38 :437–49.29373089
12. Lai WS , Gordetsky JB , Thomas JV , Factors predicting prostate cancer upgrading on magnetic resonance imaging-targeted biopsy in an active surveillance population. Cancer 2017;123 : 1941–8.28140460
|
PMC008xxxxxx/PMC8983103.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
0366151
7907
Urology
Urology
Urology
0090-4295
1527-9995
31152764
8983103
10.1016/j.urology.2019.05.022
NIHMS1783872
Article
[18F]Fluciclovine-PET Guided Salvage Lymph Node Dissection Following Radical Prostatectomy
Galgano Samuel J.
Calderone Carli E.
Nix Jeffrey W.
Rais-Bahrami Soroush
Department of Radiology, University of Alabama at Birmingham, Birmingham, AL; the Department of Urology, University of Alabama at Birmingham, Birmingham, AL; and the O’Neal Comprehensive Cancer Center at UAB, University of Alabama at Birmingham, Birmingham, AL
Address correspondence to: Soroush Rais-Bahrami, M.D., Department of Urology, University of Alabama at Birmingham, Faculty Office Tower 1107, 510 20th Street South, Birmingham 3529, AL. [email protected]
10 3 2022
10 2019
29 5 2019
05 4 2022
132 2832
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
pmcINTRODUCTION
Despite advances in standard-of-care treatment for prostate cancer, including pelvic radiotherapy and radical prostatectomy with lymph node dissection, biochemical recurrence remains a problem with some studies reporting up to 30% of patients with biochemical recurrence at 10-years following treatment. Upon detection of a rise in serum prostate-specific antigen following treatment, clinicians are left with the dilemma of how best to treat these patients. Frequently, these patients are restaged with CT of the abdomen and pelvis or MRI of the pelvis in addition to a nuclear medicine bone scan. However, detection of recurrent disease both within the prostate gland/prostatectomy bed and in the region of pelvic lymph nodes remains challenging and is limited by size criteria per RECIST 1.1. [18F]Fluciclovine PET imaging has improved detection of these previously occult metastases and helped guide treatment planning. In some select patients, a targeted lymphadenectomy may be the treatment of choice if the PET-avid disease is limited to a solitary lymph node site of spread. However, this concept has not been well described in the literature and here we provide 3 cases and short-term outcomes of patients that underwent [18F]fluciclovine PET guided surgical excision of their biochemically recurrent prostate cancer.
CASE PRESENTATIONS
Patient 1
A 62-year-old male with biopsy Gleason 3 + 4 = 7, Grade Group 2, prostate cancer underwent a robotic-assisted radical prostatectomy with pelvic lymph node dissection in 2009. Postoperative pathology demonstrated pathologic T2, Gleason 3 + 4 = 7, Grade Group 2, surgical margin negative, and lymph node negative prostate cancer. Postoperative serum prostate specific antigen (PSA) was undetectable at a level of <0.01 ng/mL. Beginning in 2013, routine surveillance revealed a slowly rising serum PSA which escalated to 1.23 ng/mL by April 2017 with doubling time of approximately 6 months. Conventional staging imaging obtained at this time failed to demonstrate site of recurrent disease. The patient then underwent [18F]fluciclovine PET/CT, which revealed a PET-avid lymph node in the perirectal space (Fig. 1). The patient subsequently underwent targeted surgical excision in June 2017. Postoperative pathology confirmed metastatic nodal focus of prostate cancer spread and serum PSA once again returned to undetectable levels at first assessment postoperatively and has remained undetectable for 18 months thereafter.
Patient 2
A 61-year-old male with biopsy Gleason 4 + 3 = 7, Grade Group 3, prostate cancer underwent a robotic-assisted radical prostatectomy with lymph node dissection in 2009. Postoperative pathology demonstrated pathologic T3, Gleason 4 + 3 = 7, Grade Group 3, surgical margin negative, and lymph node negative prostate cancer. Initially following surgery, serum PSA became undetectable to <0.01 ng/mL, however it began to rise again in 2011. The patient then underwent salvage radiation therapy to the pelvis for biochemical recurrence with a return to undetectable PSA levels. In 2015, PSA surveillance revealed a second biochemical recurrence, with levels reaching 3.2 ng/mL in December 2017. At this time, the patient underwent [18F]fluciclovine PET/CT, which demonstrated a solitary suspicious right external iliac chain lymph node (Fig. 2). The patient was subsequently underwent a salvage pelvic lymph node dissection, from which the resultant pathology confirmed metastatic prostate cancer spread to one right external iliac lymph node. Serum PSA returned to undetectable levels postoperatively.
Patient 3
A 64-year-old male with biopsy Gleason 4 + 3 = 7, Grade Group 3, prostate cancer underwent a robotic-assisted radical prostatectomy with lymph node dissection in 2007. Postoperative pathology demonstrated pathologic T2, Gleason 4 + 3 = 7, Grade Group 3, surgical margin negative, and lymph node negative prostate cancer. Postoperative serum PSA was undetectable at <0.01 ng/mL. In 2011, the patient developed a biochemical recurrence with a rise in PSA to a level of 0.4 ng/mL. Patient then underwent salvage radiation therapy to the pelvis, achieving a secondary nadir serum PSA of 0.2 ng/mL. In 2013, a subsequent rise in serum PSA to 0.9 ng/mL was observed and conventional staging imaging at this time was negative for recurrence. The patient was intermittently placed on androgen deprivation therapy, but was unable to tolerate this treatment approach due to side effects. A subsequent CT obtained demonstrated a prominent 11 mm perirectal lymph node. The patient then underwent follow-up [18F]fluciclovine-PET/CT in 2017 (Fig. 3A and B), which showed marked tracer activity in the suspicious lymph node seen on CT. Subsequently, the patient underwent an image-targeted lymphadenectomy. Postoperative pathology confirmed malignancy within the lymph node specimen, which stained positive for immunohistochemical markers suggestive of possible metastatic melanoma. Postoperatively, the patient’s serum PSA continued to rise. Therefore, a follow-up [18F] fluorodeoxyglucose PET/CT was performed and showed minimal activity in the lymph node (Fig. 3C and D) and follow-up [18F]fluciclovine-PET/CT was obtained, which redemonstrated intense activity in a perirectal nodule (Fig. 3E and F). These findings are most consistent with inadequately resected lymph node harboring metastatic prostate cancer. The patient was then referred for consideration of additional salvage radiation therapy.
DISCUSSION BY SOROUSH RAIS-BAHRAMI, MD
[18F]Fluciclovine was approved in May 2016 by the United States Food and Drug Administration for patients with biochemically recurrent cases of prostate cancer based upon rising PSA. Conventional imaging modalities used for staging indication in the setting of prostate cancer biochemical recurrence including CT, MRI, and [99mTc] nuclear medicine bone scan have been limited in sensitivity and specificity, as detection of cancer persistence or recurrence in the prostate bed and/or pelvic lymph nodes relies on size and morphologic criteria of anatomic findings in the postoperative field. PET imaging offers the advantage of earlier detection and localization of recurrent malignancy, particularly in subcentimeter lymph nodes. Thus, fluciclovine-PET/CT and PET/MRI have become increasingly utilized for the restaging of men with biochemically recurrent prostate cancer.1
Several PET radiotracers have been studied for evaluation of prostate cancer. The earliest tracer, [18F]fluorodeoxyglucose (FDG), is a mainstay in oncologic imaging, but use is limited in prostate cancer with only poorly differentiated and/or aggressive tumors consistently demonstrating substantial FDG uptake.2 Subsequently, [11C]choline and [18F]choline were developed and evaluated for use in patients with prostate cancer. Choline analogs function as a substrate for cell membrane metabolism, a process that is upregulated in prostate cancer cells. Although results were improved from studies with [18F]FDG, sensitivity and specificity remained limited.3–5 Additionally, very few clinical centers offer choline-PET imaging, which significantly limits availability to patients and providers.
[18F]Fluciclovine is an amino acid analog that targets the ASCT2 and LAT1 transporters, both of which are upregulated in prostate cancer cells. These transporters are involved with glutamine uptake at the cell membrane and fluciclovine acts as a substrate for the transporters. Importantly, [18F]fluciclovine is not metabolized intracellularly or incorporated into proteins.6 Several studies thus far have evaluate the diagnostic performance of fluciclovine-PET/CT in the setting of biochemical recurrence. A prospective study by Nanni et al demonstrated superior diagnostic performance of [18F]fluciclovine-PET/CT for patients with biochemically recurrent prostate cancer when compared to [11C]choline-PET/CT.7 Fluciclovine has also been evaluated in comparison to a historical prostate cancer SPECT imaging agent, [111In]capromab (Prostascint) and was shown to be superior in the detection of both intraprostatic and extraprostatic biochemical recurrence.8,9 Many of the studies in patients with biochemically recurrent prostate cancer have focused on the impact [18F]fluciclovine-PET/CT had on patients for salvage radiotherapy planning. A study by Akin-Akintayo et al demonstrated that [18F]fluciclovine-PET changed management decisions in 40.5% of patients with biochemically recurrent prostate cancer.10 However, only 2 case reports exist in the literature that describe [18F]fluciclovine-PET/CT guided lymph node dissection in patients with clinically biochemically recurrent prostate cancer.11,12
There is an evolving paradigm shift in the treatment of patients with oligometastatic prostate cancer. Based on the results from the STAMPEDE trial, patients with oligometastatic prostate cancer (limited node positive, visceral metastasis negative) demonstrated greater than expected failure-free survival when treated with radiotherapy combined with ADT vs ADT alone.13 However, the extent of both nodal and distant metastatic disease is underestimated on conventional imaging with CT, MRI, and nuclear medicine bone scan. In patients with both newly diagnosed prostate cancer and biochemically recurrent disease, [18F]fluciclovine demonstrated lymph node metastases not detected on conventional, standard-of-care staging imaging.14,15 This information is important to the treating radiation oncologist, as underestimation of the metastatic disease burden could potentially result in undertreatment. Additionally, there is emerging data on performing radical prostatectomy and pelvic lymph node dissection in patients with suspicion of limited lymph node involvement preoperatively, with several studies reporting positive lymph nodes outside the standard dissection template and up to 13% of metastatic lymph nodes found outside the template of an extended pelvic lymph node dissection.16–20 Advanced preoperative imaging with molecular or functional imaging, including [18F]fluciclovine, shows great potential for improving pretreatment staging of these patients to improve a durable oncologic response.21 There are currently ongoing studies of [18F]fluciclovine PET imaging in the primary staging setting (NCT03264456 and NCT03081884), particularly for those with high-risk prostate cancer, to better define potential extraprostatic spread or oligometastatic lymph node involvement not appreciated on conventional staging scans. Although these studies are examining the potential role of these imaging studies in changing clinical decision-making and potentially directing targeted lymph node sampling or dissection templates, oncologic outcomes proving the value of this added imaging modality for primary staging is not yet mature.
A large study prior to salvage lymph node dissection demonstrated a risk of early recurrence following salvage lymph node dissection of 25% with preoperative PET imaging serving an important role of risk stratification prior to salvage lymphadenectomy in predicting early recurrence.22 Future studies with longer-term clinical follow-up are needed to determine the potential oncologic benefit of PET imaging in patients with prostate cancer beyond the routine serum PSA follow-up. We recognize this as prostate cancer, even in cases of limited regional spread one or very few pelvic lymph nodes, may have an indolent clinical course. As such, there may be a clinically impactful role in targeted salvage lymph node removal or radiotherapy as the more prostate cancer specific PET imaging develops and is widely employed in the staging of patients with biochemically recurrent prostate cancer.
Finally, the development of several PET radiotracers for prostate cancer may enable surgeons to perform intraoperative molecular imaging to assist in targeting areas of recurrent or metastatic cancer foci. The concept of fluorescent molecular probes for intraoperative identification of malignancy has been studied in several malignancies in the preclinical setting, including colorectal and head and neck cancers.23,24 A [68Ga]prostate specific membrane antigen (PSMA)-11-derived dual-labeled radiopharmaceutical was developed that had properties of a PET-imaging agent combined with a fluorescent dye conjugate was studied and demonstrated feasibility and high potential for use in the preoperative, intraoperative, and postoperative detection of prostate cancer.25 Additionally, PSMA-directed PET radiotracers have been studied in a potential theranostic setting, where the diagnostic PET would be performed with a [68Ga]-tagged PET radiotracer and the therapeutic targeted radiation therapy would be provided by substituting the [68Ga] with a beta emitter, lutetium-177.26 This theranostic approach is currently FDA-approved for clinical use in patients with gastroentero-pancreatic neuroendocrine tumors utilizing [68Ga]DOTA-TATE and [177Lu]DOTATATE. Thus, it is conceivable that the theranostic approach may have a significant role in the management of patients with metastatic prostate cancer as PSMA radiopharmaceuticals seek FDA-approval.
Funding:
None.
Figure 1. Maximum intensity projection (A), fused [18F]fluciclovine PET/CT (B), and CT (C) image of patient 1 which demonstrates a fluciclovine-avid lymph node in the perirectal space. (Color version available online.)
Figure 2. Fused [18F]fluciclovine PET/CT (A) and CT (B) image of patient 2 which demonstrates a fluciclovine-avid lymph node in the right external iliac chain (arrow). (Color version available online.)
Figure 3. Initial fused [18F]fluciclovine PET/CT (A) and CT (B) images which demonstrate a fluciclovine-avid lymph node in the presacral space. Follow-up fused [18F]FDG PET/CT (C) and CT (D) images demonstrate minimal tracer activity in the presacral lymph node. Follow-up fused [18F]fluciclovine PET/CT (E) and CT (F) images demonstrate persistent activity in the presacral lymph node following attempted surgical resection. (Color version available online.)
Conflicts of Interest: Samuel Galgano receives research support from Blue Earth Diagnostics. Jeffrey Nix serves as a consultant to Philips/InVivo Corp. Soroush Rais-Bahrami serves as a consultant to Philips/InVivo Corp, Genomic Health Inc, Blue Earth Diagnostics, and Bayer Healthcare.
References
1. Galgano SJ , Calderone CE , McDonald AM , Patient demographics and referral patterns for [F-18]fluciclovine-Pet imaging at a tertiary academic medical center. J Am Coll Radiol. 2019;16 : 315–320.30447932
2. Effert PJ , Bares R , Handt S , Wolff JM , Bull U , Jakse G . Metabolic imaging of untreated prostate cancer by positron emission tomography with 18fluorine-labeled deoxyglucose. J Urol. 1996;155 :994–998.8583625
3. Parker WP , Davis BJ , Park SS , Patterns of recurrence following primary radiation therapy for prostate cancer using C-11 choline positron emission tomography/computed tomography: unique identification of sites of recurrence impacting clinical management. Int J Radiat Oncol Biol Phys. 2016;96 :S112.
4. Schiavina R , Scattoni V , Castellucci P , 11C-choline positron emission tomography/computerized tomography for preoperative lymph-node staging in intermediate-risk and high-risk prostate cancer: comparison with clinical staging nomograms. Eur Urol. 2008;54 : 392–401.18456393
5. Evangelista L , Briganti A , Fanti S , New clinical indications for (18)F/(11)C-choline, new tracers for positron emission tomography and a promising hybrid device for prostate cancer staging: a systematic review of the literature. Eur Urol. 2016;70 :161–175.26850970
6. Okudaira H , Shikano N , Nishii R , Putative transport mechanism and intracellular fate of trans-1-amino-3-18F-fluorocyclobutanecarboxylic acid in human prostate cancer. J Nucl Med. 2011;52 : 822–829.21536930
7. Nanni C , Zanoni L , Pultrone C , (18)F-FACBC (anti1-amino-3-(18)F-fluorocyclobutane-1-carboxylic acid) versus (11)C-choline PET/CT in prostate cancer relapse: results of a prospective trial. Eur J Nucl Med Mol Imaging. 2016;43 :1601–1610.26960562
8. Schuster DM , Nieh PT , Jani AB , Anti-3-[(18)F]FACBC positron emission tomography-computerized tomography and (111)In-capromab pendetide single photon emission computerized tomography-computerized tomography for recurrent prostate carcinoma: results of a prospective clinical trial. J Urol. 2014;191 :1446–1453.24144687
9. Schuster DM , Savir-Baruch B , Nieh PT , Detection of recurrent prostate carcinoma with anti-1-amino-3–18F-fluorocyclobutane-1-carboxylic acid PET/CT and 111In-capromab pendetide SPECT/CT. Radiology. 2011;259 :852–861.21493787
10. Akin-Akintayo OO , Jani AB , Odewole O , Change in salvage radiotherapy management based on guidance with FACBC (Fluciclovine) PET/CT in postprostatectomy recurrent prostate cancer. Clin Nucl Med. 2017;42 :e22–e28.27749412
11. Schiavina R , Concetti S , Brunocilla E , First case of 18F-FACBC PET/CT-guided salvage retroperitoneal lymph node dissection for disease relapse after radical prostatectomy for prostate cancer and negative 11C-choline PET/CT: new imaging techniques may expand pioneering approaches. Urol Int 2014;92 :242–245.24334968
12. Cancian M , Pereira J , Renzulli JF 2nd. Salvage pelvic lymph node dissection after fluciclovine positron emission tomography/computed tomography detected prostate cancer recurrence. J Endourol Case Rep. 2018;4 :59–61.29682612
13. James ND , Spears MR , Clarke NW , Failure-free survival and radiotherapy in patients with newly diagnosed nonmetastatic prostate cancer: data from patients in the control arm of the STAMPEDE trial. JAMA Oncol. 2016;2 :348–357.26606329
14. Selnaes KM , Kruger-Stokke B , Elschot M , (18)F-fluciclovine PET/MRI for preoperative lymph node staging in high-risk prostate cancer patients. Eur Radiol. 2018;28 :3151–3159.29294158
15. Andriole GL , Kostakoglu L , Chau A , The Impact of positron emission tomography with 18F-fluciclovine on the treatment of biochemical recurrence of prostate cancer: results from the LOCATE trial. J Urol. 2019;201 :322–331.30179618
16. Allaf ME , Palapattu GS , Trock BJ , Carter HB , Walsh PC . Anatomical extent of lymph node dissection: impact on men with clinically localized prostate cancer. J Urol. 2004;172 :1840–1844.15540734
17. Burkhard FC , Studer UE . The role of lymphadenectomy in prostate cancer. Urol Oncol. 2004;22 :198–202. discussion 202–194. 15271316
18. McDowell GC 2nd , Johnson JW , Tenney DM , Johnson DE . Pelvic lymphadenectomy for staging clinically localized prostate cancer. Indications, complications, and results in 217 cases. Urology. 1990;35 :476–482.2353374
19. Joniau S , Van den Bergh L , Lerut E , Mapping of pelvic lymph node metastases in prostate cancer. Eur Urol. 2013;63 :450–458.22795517
20. Schuessler WW , Pharand D , Vancaillie TG . Laparoscopic standard pelvic node dissection for carcinoma of the prostate: is it accurate? J Urol. 1993;150 :898–901.8345606
21. Zarzour JG , Galgano S , McConathy J , Thomas JV , Rais-Bahrami S . Lymph node imaging in initial staging of prostate cancer: an overview and update. World J Radiol. 2017;9 :389–399.29104741
22. Fossati N , Suardi N , Gandaglia G , Identifying the optimal candidate for salvage lymph node dissection for nodal recurrence of prostate cancer: results from a large, multi-institutional analysis. Eur Urol 2019;75 :176–183.30301694
23. Marston JC , Kennedy GD , Lapi SE , Panitumumab-IRDye800CW for fluorescence-guided surgical resection of colorectal cancer. J Surg Res. 2019;239 :44–51.30798171
24. Prince AC , Moore LS , Tipirneni KE , Evaluation of optical imaging agents in a fluorescence-guided surgical model of head and neck cancer. Surg Oncol. 2018;27 :225–230.29937175
25. Baranski AC , Schafer M , Bauder-Wust U , PSMA-11-derived dual-labeled PSMA inhibitors for preoperative PET imaging and precise fluorescence-guided surgery of prostate cancer. J Nucl Med. 2018;59 :639–645.29191856
26. Yadav MP , Ballal S , Tripathi M , (177)Lu-DKFZ-PSMA-617 therapy in metastatic castration resistant prostate cancer: safety, efficacy, and quality of life assessment. Eur J Nucl Med Mol Imaging. 2017;44 :81–91.27506431
|
PMC008xxxxxx/PMC8983104.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
0366151
7907
Urology
Urology
Urology
0090-4295
1527-9995
31866522
8983104
10.1016/j.urology.2019.12.008
NIHMS1783866
Article
High-volume Concurrent Polypoid Ureteritis and Ureteritis Cystica Manifesting With Ureteral Obstruction
Glaser Zachary A.
Fougerousse Joseph A.
Galgano Samuel J.
Magi-Galluzzi Cristina
Rais-Bahrami Soroush
Department of Urology, University of Alabama at Birmingham, Birmingham, AL; the Department of Radiology, University of Alabama at Birmingham, Birmingham, AL; the Department of Pathology, University of Alabama at Birmingham, Birmingham, AL; and the O’Neal Comprehensive Cancer Center at UAB, University of Alabama at Birmingham, Birmingham, AL
Authors’ Contributions: Zachary A. Glaser: Conceptualization, formal analysis, investigation, resources, validation, visualization, writing—original draft, and writing—review and editing; Joseph A. Fougarousse: Conceptualization, formal analysis, investigation, validation, visualization, and writing—review and editing; Samuel J. Galgano: Conceptualization, data curation, supervision, validation, visualization, writing—original draft, and writing—review and editing; Cristina Magi-Galluzzi: Conceptualization, data curation, supervision, validation, visualization, writing—original draft, and writing—review and editing; Soroush Rais-Bahrami: Conceptualization, data curation, funding acquisition, investigation, project administration, resources, supervision, visualization, writing—original draft, and writing—review and editing.
Address correspondence to: Soroush Rais-Bahrami, M.D., Department of Urology, University of Alabama at Birmingham, Faculty Office Tower 1107, 510 20th Street South, Birmingham, AL 35294. [email protected]
10 3 2022
2 2020
19 12 2019
05 4 2022
136 e7e11
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
pmcPolypoid ureteritis, or fibroepithelial ureteral polyps, are a rare but benign neoplasm of unknown etiology in the upper urinary tract and may occur at any age.1 Ureteritis cystica is also a rare but benign phenomenon that may develop following inflammatory conditions of the ureter and/or renal pelvis.1 Due to their appearance on imaging and frequently accompanying symptoms of gross hematuria, the masses are often clinically indistinguishable from upper tract urothelial carcinoma (UTUC) in adults.2 Children may manifest with signs and symptoms of ureteropelvic junction obstruction.3 Complete resection is often necessary for symptom relief and to definitively address partial or complete upper urinary tract obstruction.
We present a case of a middle-aged female incidentally found to have a partially obstructing left midureteral mass and prominent regional lymph node on imaging workup of a nonurologic diagnosis. Ureteroscopic evaluation and biopsy revealed simultaneous polypoid ureteritis and surrounding ureteritis cystica. Subsequent staged, endoscopic holmium laser ablation was performed to definitively address her resultant hydroureteronephrosis.
CASE PRESENTATION
A 67-year-old white female with no significant past medical history was admitted to the hospital for new onset jaundice. Her symptoms consisted of yellow skin and conjunctiva, diffuse pruritus, and elevated hepatic enzymes. As a part of her workup, she underwent a contrasted computed tomography scan that incidentally demonstrated a roughly 5 cm left midureteral mass with frond-like intraluminal projections, proximal ureteral dilation, and a paraaortic 1.1 cm lymph node with high concern for regionally advanced UTUC (Fig. 1). She was not experiencing any flank pain or gross hematuria. Beyond elevated liver function tests, her laboratory workup was unremarkable. She had a creatinine of 0.8 mg/dL, a normal estimated glomerular filtration rate, no microscopic hematuria, and a negative urine culture. On further interrogation, she did not have any past urologic history such as urolithiasis, recurrent urinary tract infections, or prior urologic instrumentation. She never smoked or experienced chemical exposure, was a retired school teacher, and did not have any ties to geographic regions with associated endemic nephropathy.
Her acute jaundice was ultimately attributed to an azithromycin-related, drug-induced liver injury that resolved with conservative management. She was discharged from the hospital 3 days following her admission with plans for a confirmatory diagnostic workup of her ureteral mass once her hepatic function fully recovered.
Three weeks later, she was taken to the operating room for cystoscopy, left ureteral barbotage, retrograde pyelogram, and a diagnostic ureteroscopy. Her retrograde pyelogram redemonstrated the midureteral intraluminal appearing mass with associated proximal hydroureteronephrosis (Fig. 2). Semirigid ureteroscopy revealed numerous intraluminally projecting rounded, tan-colored, pedunculated masses (Fig. 3). The ureteroscope could be advanced proximally to these lesions allowing appreciation of her proximally dilated ureteral lumen with normal mucosa. A repeat retrograde pyelogram through the semirigid ureteroscope did not demonstrate any additional filling defects in the proximal ureter or renal collecting system. Numerous biopsies were then obtained using a transureteroscopic biopsy forceps and a basket device. An indwelling double-J ureteral stent was left in place and she was discharged home on the same day.
Gross and microscopic pathologic analysis revealed polypoid ureteritis with marked ureteritis cystica. To alleviate her visualized proximal obstruction, she was taken back to the operating room for a second staged ureteroscopic procedure to perform complete ablation of these benign lesions using a holmium laser and double-J stent replacement to allow scaffolded healing of the ablated ureteral lumen (Fig. 3). She was again discharged home the same day without incident. Her stent was uneventfully removed 4 weeks later in the office.
DISCUSSION BY SOROUSH RAIS-BAHRAMI, MD
Benign ureteral tumors are seldom encountered. Fibroepithelial ureteral polyps were first described in 1932 with only several hundred cases reported to date.4,5 Ureteritis cystica was first characterized by Morgagni the mid-18th century by their appearance of suburothelial cysts with fingerlike projections into the ureteral lumen.1,6,7 Due to the rarity of both lesions, proposed etiologies and incidence for both are quite variable in the literature.3,8–11
In children, ureteral polyps are thought to be congenital in origin and primarily occur in the proximal third of the ureter and renal pelvis.7,12,13 They are seldom detected in utero presumably due to their indolent growth pattern, and instead manifest during early childhood. Symptoms and signs are similar to those of ureteropelvic junction obstruction caused by more common etiologies such as flank pain, recurrent infection, hematuria, and/or hydronephrosis.3 For unclear reasons, the left ureter is more commonly affected than the right, with a reported 2:1 left:right predominance.14 Following discovery on imaging and confirmatory diagnosis via retrograde pyelogram or ureteroscopy, complete excision for symptom relief and treatment of proximal obstruction can be achieved with a dismembered pyeloplasty.3,15,16 Alternatively, successful endoscopic resection has been reported.5,17 Ureteritis cystica is less commonly reported in the pediatric urology literature.
Ureteral polyps in adults typically present in the third or fourth decades of life with a reported male predominance.1,18 Individual polyp size can span anywhere from 5 mm-12 cm in size.1 While pediatric ureteral polyps are thought to be congenital in origin, the etiology is less clear in adult patients. Chronic inflammatory states such as infection, Schistosomiasis, prior trauma, and/or instrumentation may promote their development.9,19 They may also develop in the absence of a known cause. In 1 retrospective series by Childs et al, 9/22 (41%) of patients with histologically confirmed benign ureteral polyps had no prior urologic history and were presumed to have developed the tumors idiopathically. Ureteritis cystica may also form in adults following a previous inflammatory condition involving urothelial mucosa such as urolithiasis, infection, or prior endoscopic instrumentation.1,11,20 Presenting symptoms of both conditions are similar to those experienced by children, but some adults may be asymptomatic and they are instead discovered incidentally when abdominal imaging is performed for a nonurologic indication.21
On imaging, ureteral polyps and ureteritis cystica may be detected secondary to ureteral obstruction or an apparent intraluminal mass. On contrast-enhanced computed tomography, ureteral polyps and ureteritis cystica appear as an intermediate density mass within the ureter that may result in hydroureteronephrosis. On delayed excretory phase imaging with excreted iodinated contrast in the ureter, they can manifest as multiple intraluminal filling defects with a frond-like appearance, analogous to their appearance on retrograde pyelogram. Currently, little data exist about the appearance of ureteritis cystica on magnetic resonance imaging due to the rarity of this entity. Given the appearance of ureteral polyps and ureteritis cystica on imaging along with frequently accompanying hematuria, adult patients are initially managed with suspicion that the mass represents urothelial carcinoma of the ureter. In the contemporary setting, patients then undergo ureteroscopic evaluation and transureteroscopic biopsy for pathologic confirmation.22 Due to the sometimes difficult nature of upper tract instrumentation, cytology and inadequate tissue acquired via biopsy, a definitive diagnosis of polypoid ureteritis and/or ureteritis cystica may only be available after radical nephroureterectomy.2,23,24
Gross pathologic examination of fibroepithelial ureteral polyps reveals a tan or brown smooth surface with round intraluminal projections. Microscopically, the polyps exhibit a normal urothelial mucosal surface with a loose fibrovascular stroma underneath. Ureteritis cystica similarly has a normal urothelial mucosa but with cystically dilated urothelial cysts, known as van Brunn nests, embedded below their mucosal lining. Both differ from the grossly apparent papillary projections, microscopic nuclear pleomorphism, and stromal infiltration evident in UTUC.1 The incidence of concurrent ureteral polyps and ureteritis cystica is not known but estimated to be exceedingly rare with only a few previously reported cases.21
Management of both polypoid ureteritis and ureteritis cystica is aimed at relief of symptoms as well as alleviation of proximal ureteral obstruction to optimally preserve renal function. Ureteroscopic resection with loop electrocautery or laser ablation is safe and well-tolerated with acceptable outcome in affected patients.1,7,8,10 Alternative endoscopic treatment modalities for obstructed patients include balloon dilation, silver nitrate instillation, or long-term antibiotic therapy with varying long-term success.8,11 In an asymptomatic individual without evidence of obstruction, intervention may not be necessary at all.8,11 Radical nephroureterectomy or segmental ureterectomy are not indicated unless suspicion for UTUC remains following a thorough diagnostic workup and tissue sampling.2
The role of long-term surveillance after definitive resection or ablative therapy is unclear. Association of polypoid ureteritis and ureteritis cystica with urothelial carcinoma or other malignant processes is very uncommon with only a few rare cases reported.1,5,8,25 Imaging and/or ureteroscopic surveillance may benefit patients with a large tumor burden on initial presentation who may be at risk for symptomatic recurrence. In those instances, repeat resection or ablation procedures can be performed.5 Postoperative ureteral stricture formation is another potential complication with an incidence similar to that of other endoscopic procedures for manipulation of upper tract tumors, and may require additional procedures to address the stricture to maintain patency of the ureter.5,10,26
In the present case, the etiologies both fibroepithelial ureteral polyp and ureteritis cystica formation are unclear. They may have occurred idiopathically, the polyps could have propagated secondary ureteritis cystica formation by causing inflammatory changes to nearby urothelium over time. Periodic imaging surveillance with ultrasonography is planned to ensure hydroureteronephrosis does not recur.
CONCLUSION
Ureteral fibroepithelial polyps and ureteritis cystica are two distinct, benign and rare conditions of the upper urinary tract with an incredibly low malignant association though recognized as clinical mimickers of UTUC. Underlying etiologies of both processes are unclear, and patients may present with quite varying symptomatology. Following confirmatory diagnostic workup, endoscopic resection or ablation provides symptom relief and alleviates obstruction of the affected renal moiety.
Funding: This work was funded in part by Junior Faculty Development Grant (ACS-IRG 001-53) and by Developmental funds from the UAB Comprehensive Cancer Center Support Grant (NCI P30 CA 013148) to Soroush Rais-Bahrami.
Figure 1. Coronal CT scan with intravenous contrast of the abdomen and pelvis. (A) Arterial phase (B) delayed urogram. CT, computed tomography.
Figure 2. Left retrograde pyelogram redemonstrating a midureteral filling defect and proximally dilated system. Several air-bubbles were present in the proximal ureter.
Figure 3. Appearance of ureteral fibroepithelial polyps and smaller ureteritis cystica lesions on (A) initial semirigid ureteroscopy and (B) subsequent holmium laser ablation.
Conflicts of Interest: Soroush Rais-Bahrami serves as a consultant for Philips/InVivo Corp, Intuitive Surgical, Genomic Health Inc, Bayer Healthcare, and Blue Earth Diagnostics.
References
1. Zhou MN G , Epstein J . Uropathology: High-Yield Pathology. Philadelphia, PA: Elsevier; 2012.
2. Hong S , Kwon T , You D , Incidence of benign results after laparoscopic radical nephroureterectomy. JSLS. 2014;18 :e2014.00335.
3. Adey GS , Vargas SO , Retik AB , Fibroepithelial polyps causing ureteropelvic junction obstruction in children. J Urol. 2003;169 : 1834–1836.12686857
4. Melicow M F HV . Primary benign tumors of ureter: review of literature and report of case. Surg Gynecol Obstet. 1932:680–689.
5. Childs MA , Umbreit EC , Krambeck AE , Sebo TJ , Patterson DE , Gettman MT . Fibroepithelial polyps of the ureter: a single-institutional experience. J Endourol. 2009;23 :1415–1419.19715398
6. Morgagni G . Jo. Baptistæ Morgagni de Sedibus, Et Causis Morborum Per Anatomen Indagatis: Libri Quinque. Neapoli: Simoniana; 1762.
7. Wein Alan J , Kavoussi Louis R , Partin Alan W , Peters Craig A . Campbell-Walsh Urology. Philadelphia, PA: Elsevier; 2016.
8. Poturalski MJ , Purysko AS , Herts BR . Ureteritis cystica. J Urol. 2015;193 :1379–1380.25598136
9. Karmo BT , Lim K , Santucci RA , Mansour S . Robot-assisted right ureteral polypectomy: a case report. Can Urol Assoc J. 2013;7 :E426–E429.23826056
10. Georgescu D , Multescu R , Geavlete BF , Fibroepithelial polyps - a rare pathology of the upper urinary tract. Rom J Morphol Embryol. 2014;55 :1325–1330.25611262
11. Menendez V , Sala X , Alvarez-Vijande R , Sole M , Rodriguez A , Carretero P . Cystic pyeloureteritis: review of 34 cases. Radiologic aspects and differential diagnosis. Urology. 1997;50 :31–37.9218015
12. Gomez Fraile A , Aransay A , Matute JA , Lopez F , Olcoz F , Muley R . Pyelic benign fibroepithelial polyp in childhood: a case report. J Pediatr Surg. 1993;28 :948–949.8229575
13. Kanamori Y , Iwanaka T , Sugiyama M , Antenatally diagnosed, intermittently worsened hydronephrosis caused by a ureteral polyp. Pediatr Int. 2010;52 :e11–e13.20158636
14. Bhalla RS , Schulsinger DA , Wasnick RJ . Treatment of bilateral fibroepithelial polyps in a child. J Endourol. 2002;16 :581–582.12470466
15. Shive ML , Baskin LS , Harris CR , Bonham M , MacKenzie JD . Ureteral fibroepithelial polyp causing urinary obstruction. J Radiol Case Rep. 2012;6 :23–28.
16. Bian Z , Liu X , Hua Y , Liu F , Lin T , He D . Laparoscopic management of multiple ureteral polyps in children. J Urol. 2011;186 :1444–1449.21855936
17. Lam JS , Bingham JB , Gupta M . Endoscopic treatment of fibroepithelial polyps of the renal pelvis and ureter. Urology. 2003;62 :810–813.14624899
18. Williams TR , Wagner BJ , Corse WR , Vestevich JC . Fibroepithelial polyps of the urinary tract. Abdom Imaging. 2002;27 :217–221.11847584
19. Macfarlane MT , Stein A , Layfield L , deKernion JB . Preoperative endoscopic diagnosis of fibroepithelial polyp of the renal pelvis: a case report and review of the literature. J Urol. 1991;145 :549–551.1997707
20. Hanafy HM , Saad SM , Al-Ghorab MM . Bilharzial (schistosomal) ureteritis calcinosa. Eur Urol. 1981;7 :161–164.7202453
21. Janeiro C , Oliveira F , Andrade G , Ureteritis cystica and ureteral polyp-case report. AME Case Rep. 2018;2 :32.30264028
22. Network NCC. NCCN Guidelines Version 4.2019 Bladder Cancer. 2019.
23. Guarnizo E , Pavlovich CP , Seiba M , Carlson DL , Vaughan ED Jr. , Sosa RE . Ureteroscopic biopsy of upper tract urothelial carcinoma: improved diagnostic accuracy and histopathological considerations using a multi-biopsy approach. J Urol. 2000;163 :52–55.10604312
24. Gennaro KH , Gordetsky J , Rais-Bahrami S , Selph JP . Ureteral endometriosis: preoperative risk factors predicting extensive urologic surgical intervention. Urology. 2017;100 :228–233.27542859
25. Richmond HG , Robb WA . Adenocarcinoma of the ureter secondary to ureteritis cystica. Br J Urol. 1967;39 :359–363.6027772
26. Complications of Urologic Surgery: Prevention and Management. 4th ed Philadelphia, PA: Elsevier; 2001.
|
PMC008xxxxxx/PMC8983105.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
101674571
44756
Abdom Radiol (NY)
Abdom Radiol (NY)
Abdominal radiology (New York)
2366-004X
2366-0058
28677001
8983105
10.1007/s00261-017-1237-x
NIHMS1783887
Article
Incidental findings on multiparametric MRI performed for evaluation of prostate cancer
Sherrer Rachael L. 1
Lai Win Shun 1
Thomas John V. 2
Nix Jeffrey W. 1
Rais-Bahrami Soroush http://orcid.org/0000-0001-9466-9925
12
1 Department of Urology, University of Alabama at Birmingham, Faculty Office Tower 1107, 510 20th Street South, Birmingham, AL 35294, USA
2 Department of Radiology, University of Alabama at Birmingham, Jefferson Tower N354, 619 19th Street South, Birmingham, AL 35294, USA
Correspondence to: Soroush Rais-Bahrami, [email protected]
10 3 2022
3 2018
05 4 2022
43 3 696701
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Purpose:
Multiparametric magnetic resonance imaging (mp-MRI) and MRI/Ultrasound (US) fusion-guided biopsy are relatively new techniques for improved detection, staging, and active surveillance of prostate cancer (PCa). As with all imaging modalities, MRI reveals incidental findings (IFs) which carry the risk of increased cost, patient anxiety, and iatrogenic morbidity due to workup of IFs. Herein, we report the IFs from 684 MRIs for evaluation of PCa and consider their characteristics and clinical significance.
Methods:
Patients underwent mp-MRI prostate protocol incorporating triplanar T2-weighted, diffusion-weighted, and dynamic contrast-enhanced pelvic MRI as well as a post-contrast abdominopelvic MRI with the primary indication of detection or evaluation of PCa. A total of 684 consecutive prostate MRI reports performed in a series of 580 patients were reviewed. All extraprostatic findings reported were logged and then categorized by organ system and potential clinical significance.
Results:
There were 349 true IFs found in 233 (40%) of the 580 patients. One hundred nineteen additional extraprostatic findings were unsuspected but directly related to PCa staging, while the 349 IFs were unrelated and thus truly incidental beyond study indication. While the majority of true IFs were non-urologic, only 6.6% of IFs were considered clinically significant, non-urologic findings, and more than a third of MRI reports had urologic IFs not related to PCa.
Conclusions:
Rates of incidental findings on prostate indication MRI are similar to other abdominopelvic imaging studies. However, only 6.6% of the IFs were considered to be clinically significant non-urologic findings. Further investigations are needed to assess downstream workup of these IFs and resulting costs.
Prostate adenocarcinoma
Cancer staging
Incidental findings
Public health
pmcProstate cancer (PCa) is the most common solid-organ malignancy diagnosed in American men, estimated at 161,360 new cases diagnosed in the United States in 2017 [1]. Since multiple management options exist for early staged PCa, early detection with proper risk stratification is critical for selecting the most appropriate treatment options [2]. Currently, PCa is commonly diagnosed using systematic transrectal ultrasound (TRUS)-guided prostate biopsies, usually prompted by elevated serum prostate-specific antigen (PSA) or abnormal digital rectal exam (DRE) findings. However, this traditional approach can result in high rates of false negative results, undergrading, or understaging of the disease, especially with cancer located in occult regions of the gland: distal apical, midline, subcapsular, and anterior areas [3–7]. The use of multiparametric magnetic resonance imaging (mp-MRI) and MRI/US fusion-guided prostate biopsy is gaining traction at many centers [8]. This technology allows for image-guided targeting of cancer suspicious lesions with proven improvement in detection of PCa, particularly clinically significant cancer foci, compared to random TRUS guided biopsy [9, 10]. This ultimately allows for more accurate detection, localization, staging, risk stratification, and treatment counseling for men with PCa [7].
As with all imaging modalities, MRI has the potential for revealing both clinically significant and indolent incidental findings (IFs). In addition to the prostate gland, mp-MRI performed for evaluation of prostate cancer can also identify possible pathologies in the field of view: abdominal and pelvic organs, bones, and vasculature. This has the potential benefit of improving PCa staging through evaluation of the seminal vesicles, bones, and regional pelvic lymph nodes for potential PCa involvement. However, there is also the risk of patient anxiety, increased cost burden, and iatrogenic morbidity due to work up prompted by IFs that are not clinically significant, potentially leading to an increased public health burden [11]. Previous studies have elicited the costs of IFs in various imaging modalities [12–20]. However, there are no previous studies comprehensively evaluating and reporting IFs found on mp-MRI done for evaluation of PCa. Herein, we report the incidence, anatomic distribution, and clinical significance of incidental findings found on 684 consecutive mp-MRI studies performed in a series of 580 patients for the evaluation of prostate cancer at a single center.
Methods
An IRB-approved, HIPAA-compliant retrospective review of prostate mp-MRI records spanning a three-year period from July 2013–June 2016 was performed. All patients included in the study underwent mp-MRI with the primary indications of detecting or risk-stratifying cases of PCa; this included patients with clinical suspicion of prostate cancer based upon elevated PSA or abnormal DRE who were either biopsy naïve or had prior negative prostate biopsy sessions as well as men with prostate cancer diagnoses on active surveillance or seeking MRI for pre-treatment planning. Prostate indication mp-MRI studies for men who had already undergone primary definitive therapy were excluded. Figure 1 shows how these inclusion criteria were used to select the mp-MRI records used for final analysis in this study.
The mp-MRI prostate protocol included triplanar T2-weighted, diffusion-weighted, and dynamic contrast-enhanced pelvic MRI as well as a post-contrast abdominopelvic T1-weighted MRI as previously described [21]. All MRIs were reviewed by an MRI specialized body radiologist on an Intellispace Portal (Philips Medical systems, Eindhoven, The Netherlands) Picture Archiving Communications System (PACS).
The radiologist’s written report for each MRI was then reviewed and all extraprostatic findings were recorded. An incidental finding was defined as an asymptomatic, unrelated imaging abnormality discovered during workup for prostate cancer detection and risk stratification. Additionally, PCa-related findings important for staging were also recorded and analyzed separately. For patients who underwent more than one prostate MRI, IFs were only counted for the first report in which they were mentioned. IFs were then categorized by clinical significance based on a classification system used in previous similar studies [12–14].
Subsequently, a retrospective review of the medical record was also performed for patients who had IFs reported documenting patient demographics. Patients were also stratified by biopsy status, specifically biopsy naïve, prior negative biopsy, or pre-MRI biopsy-proven PCa.
Results
There were 580 male patients who underwent 684 MRI studies that were included in this study, with a mean age of 63.3 years (range 40–84 years) and mean PSA of 8.42 ng/mL (range 0.3–76.34 ng/mL). For each MRI report, the patient’s prostate biopsy and PCa history were recorded. In 87 (13%) reports, the patient was biopsy naïve, 219 (32%) had at least one previous negative biopsy, 372 (54%) had needle core biopsy-proven PCa, and 6 (1%) had PCa diagnosed in a manner other than biopsy (i.e., TURP specimen). Ninety-three (16%) patients underwent more than one prostate indication mp-MRI study, whether for continued monitoring of PCa on AS or because of a previously terminated incomplete mp-MRI study. The average number of MRIs in patients who had more than one was 2.12, and there were 104 repeat MRIs total. The patient’s age at exam correlated with the number of IFs (p = 0.016). The number of MRI studies performed per patient did not correlate with the number of IFs found (p = 0.584). However, prior biopsy history (biopsy naïve, prior negative biopsy, and PCa-proven cases) at time of mp-MRI was significantly associated with number of IFs (p = 0.025); patients with prior PCa-positive biopsies had a significantly higher average of IFs (0.87) compared to those with negative biopsies (0.57) and who were biopsy naïve (0.60).
There were 349 incidental findings total in 233 (40%) of the 580 patients. 347 (60%) patients had no IFs, 145 (25%) patients had one IF, and 88 (15%) patients had two or more IFs (Fig. 2). For MRI reports in which the patient was biopsy naïve at the time of the study, there was an average of 0.60 findings per MRI, patients with a previous negative biopsy had an average of 0.57 findings per MRI, and patients with prior diagnoses of PCa had an average of 0.87 findings per MRI. In addition to the 349 truly incidental findings, there were also 119 extraprostatic findings directly related to PCa staging (suspected cancer spread to seminal vesicles, lymph nodes, bones, and any other metastases), which were not truly incidental outside the indication of the study performed. 170 (49%) of the non-PCa-related extraprostatic IFs were considered related to the genitourinary system. A summary of PCa-related findings and urologic non-PCa-related IFs is reported in Table 1.
Table 2 shows the non-urologic IFs organized by clinical significance. Of these 179 non-urologic IFs, 23 (13%) were classified as bearing high clinical significance, and 156 (87%) were classified as low to moderate clinical significance.
The most frequent of the 119 imaging findings related to PCa staging was suspected seminal vesicle invasion, which accounted for 40 (34%) of PCa staging-related findings reported. The most common of the 170 urologic IFs not related to PCa staging were bladder wall thickening/trabeculation (n = 89, 52%) and renal cysts (n = 39, 23%). Indeterminate liver lesion (n = 4, 17%), abdominal aortic aneurysm greater than 3 cm (n = 3, 13%), osteonecrosis (n = 3, 13%), and bowel-containing inguinal hernia (n = 3, 13%) were the most common of the 23 non-urologic IFs considered to carry higher clinical significance. The most frequent IFs of low to moderate clinical significance, which were also the most frequent IFs overall, were colonic diverticulosis (n = 71, 46%) and fat-containing patent inguinal hernia (n = 62, 40%).
Of the 684 MRI reports, 58 (8.5%) included a total of 71 specific recommendations for follow-up by the radiologist, including bone scan (n = 24, 34%), clinical correlation (n = 15, 21%), ultrasound (n = 9, 13%), CT (n = 8, 11%), plain film radiography (n = 6, 8%), correlation with prior imaging (n = 3, 4%), follow-up MRI (n = 2, 3%), clinical follow-up (n = 2, 3%), colonoscopy (n = 1, 1%), and urinalysis (n = 1, 1%). Twenty-five (3.7%) imaging studies were considered incomplete, most were prematurely stopped prior to completion of the scan due to residual hemorrhage from previous prostate biopsy (n = 18, 72%) or patient intolerance of the MRI study related to discomfort or claustrophobia (n = 5, 20%).
Discussion
Mp-MRI and MRI/US fusion-guided biopsies are relatively new techniques for improved detection, staging, and active surveillance of PCa [2, 21, 22]. As with all imaging modalities, MRI has the potential for detecting clinically insignificant IFs. This can prompt additional clinical encounters and imaging studies, which in turn can lead to increased costs, patient anxiety, and potentially iatrogenic morbidity.
Previous studies have investigated such downstream effects of IFs found on various imaging modalities. Yee et al. reported extracolonic findings in 63.0% of patients undergoing CT colonography of which 9.0% were considered clinically impactful IFs. In our study, 42% of mp-MRIs had incidental findings and 6.6% of the IFs were clinically significant. Our study populations were similar: all male patients with a similar age distribution (mean age 62.5 years, range 30–90 in the study of CT colonography and mean age 63.3 years, range 40–84 in the current study of mp-MRI of the prostate). Additionally, we used a similar system of categorization of IFs as reported by Yee and colleagues: findings not related to the study indication were sorted into two categories of clinical relevance (high clinical significance and low to moderate clinical significance). Yee et al. recommended this system of categorization instead of a three-tier system used in other previous studies of IFs due to its simplicity and relevance to clinical management [19]. The primary difference found was the percentage of clinically significant findings identified: 9.0% in the CT colonography study vs. 6.6% in the current study. The increased visualization of the lungs in CT colonography vs. mp-MRI of the prostate may have contributed to this higher proportion of IF detection in the series of CT colonography, as approximately 20% of the IFs were pulmonary nodules.
Lai et al. assessed the IFs discovered on abdominopelvic imaging for another urologic indication: CT urography for evaluation of asymptomatic microscopic hematuria (AMH). One major difference between our studies lies in the expected yield of the imaging study. While they reported that the etiology of AMH is often not elicited on CT urogram, mp-MRI for the evaluation of PCa frequently yields clinically useful information for the indication intended: intraprostatic lesions for biopsy targeting and/or PCa staging by means of assessing seminal vesicle invasion, lymph node spread, and bony metastatic lesions in the region of the pelvis. On the other hand, the CT urograms in the study reported by Lai and colleagues revealed a higher percentage of clinically significant findings: 14.8% vs. our series with only 6.6%. While this may be in part due to the inclusion of female patients and therefore female pathology in their imaging reports (4 of 30, 13% of their highly clinically significant findings were related to female pelvic anatomy), they reported that overall there was no significant difference in the number of IFs between males and females. Additionally, they found IFs in a higher overall percentage of their patients. Only 11.38% of patients in their study had no IFs at all, whereas 394 of 684 (58%) of MRIs in our study had no IFs. Renal cysts were one of the most common urologic IFs in both studies (68/172, 39.5% in Lai et al.; 39/170, 22.9% in the current study). This may be due to the limited imaging series performed (a single post-contrast T1 W scan for staging purposes) which images the upper abdomen as part of the mp-MRI studies with prostate indication evaluated in this current study.
Further workup of IFs can be of great benefit to the patient. One patient in our current population was found to have a pedunculated lesion incidentally discovered along the posterior left bladder wall protruding into the bladder luminal cavity. This patient was then found to have high-grade, non-muscle invasive bladder carcinoma and subsequently underwent surgical resection and intravesical BCG therapy with surveillance follow-up negative for malignancy.
In another patient, an indeterminate lesion within the left kidney was incidentally noted on mp-MRI for evaluation of PCa with a recommendation to have dedicated renal imaging. On follow-up renal ultrasound, a non-obstructing renal calculus as well as a solid heterogeneously mass with internal vascularity was seen, consistent with neoplasm in the region of the upper pole of the left kidney and left adrenal gland. While on mp-MRI the mass was reported to arise from the kidney, on ultrasound it appeared to be more likely adrenal in origin, and a multiphasic CT or dedicated MRI was recommended for further evaluation and characterization of the mass. CT imaging was performed which revealed a 4.7 cm heterogeneous enhancing left adrenal mass. On clinical follow-up, the patient had elevated normetanephrine and metanephrine levels and long-standing panic attacks. Considering the likelihood of pheochromocytoma, he was premedicated with alpha-blockade and then underwent a laparoscopic left adrenalectomy with pathology confirmed pheochromocytoma. His clinical symptoms and serum markers normalized after surgical resection.
These two patients benefitted from the additional workup, demonstrating the potential benefit of following up on incidental findings. These cases also show how an incidental finding can lead to a cascade of further workup such as imaging, referrals, labs, and procedures. While these patients did benefit from IF workup and treatment, this is not always the case. One patient with biopsy-proven Gleason 4 + 3 = 7 PCa underwent mp-MRI, in which there was a thickened, nodular appearance of the inferior bladder wall. The patient was asymptomatic at the time, but underwent cystoscopy for workup of the imaging findings which was negative for gross tumor. Cytology from bladder washings was also negative for malignancy. In this case, the further diagnostic workup prompted by the IF was not clinically beneficial to the patient. Bladder wall thickening may be a non-specific IF more frequently encountered in this patient population undergoing MRI most often due to elevated PSA or known history of PCa. In these patients, bladder outlet obstruction via prostate enlargement or non-compliant nodularity can increase the likelihood of bladder muscle trabeculations seen as diffuse wall thickening on imaging. This most common urologic IF may be of low yield on further workup, and hence, clinical correlation with assessment of hematuria, prostate volume, urinary flow rate, and post-void residual urine volume may potentially be of high utility in risk stratification.
Hemorrhage due to prostate biopsy was mentioned in MRI reports as limiting the study quality in some cases. Of the 25 studies that had to be discontinued and rescheduled, 18 were due to residual hemorrhage from recent prostate biopsy. While this only makes up 2.6% of the total MRIs in this study, these account for 72% of all the incomplete exams. In the studies that were completed, 29 (4.4%) specifically mentioned that residual blood products limited the quality of the exam. Therefore, hemorrhage from recent prostate biopsy negatively impacted the feasibility or quality of 7.9% (54/684) of the MRIs. Since MRI is an expensive imaging modality, it seems sensible to wait for post-biopsy hemorrhage to clear when possible in order to attain the most informative, clinically useful, sensitive, and specific study possible. However, Feng and colleagues reported that while hemorrhage is prevalent post-prostate biopsy, present in 52% of 36 cases even after 6 weeks, there was no difference in detection of extracapsular extension on mp-MRI as compared with pathology regardless of time elapsed since prostate biopsy [23]. Larger studies may be beneficial for evaluating the impact of post-biopsy hemorrhage on mp-MRI diagnostic value particularly for detection of intraprostatic pathology since a recent consensus panel recommended MRI to be postponed at least 6 weeks though the majority of panelists thought imaging quality would still be impaired at this timepoint [24].
One of the limitations of this study was the interpretation of the imaging studies by multiple (n = 15) radiologists. Each radiologist used different wording to report his or her findings, making systematic classification of these findings challenging. This could be prevented by having one or two radiologists interpret each MRI used in the study, but this was not possible for this study due to the retrospective nature of the chart review and high volume of imaging studies reviewed for this analysis. Also, whereas some studies reviewed the actual images in addition to imaging reports, we reviewed reports only possibly limiting our findings [14, 15, 19]. Additionally, there is some variation in previous studies in classification of IFs by clinical significance. Since this is not standardized, it may introduce some bias in the comparison of various studies, such as the comparisons discussed above. A standardized classification method for IFs would be beneficial for more accurate comparison between studies.
In addition to patient anxiety and iatrogenic morbidity, increased costs are also a potential consequence of IF workup. In order to evaluate the financial burden on both the public health system and the patient, more research is needed examining the cost of follow-up imaging, diagnostic studies, and treatment. A cost analysis would allow for more informed follow-up choices in which the benefits merit the costs required. Such a study looking at the results of IF workup and consequent costs, morbidity, and mortality could help identify what follow-up actually helps the patient, and which interventions cause more morbidity. With this new information on IFs from prostate MRIs and more research on the costs and results of workup of these IFs, improved guidelines for follow-up of IFs could be created and implemented to improve both patient outcomes and efficient use of resources.
Conclusions
Incidental findings are inevitable when employing any imaging. Prostate indication MRI demonstrates rates of incidental findings in 42% of cases, commensurate with other abdominal and pelvic imaging studies done for other indications. However, only 6.6% of these IF were considered to be highly clinically significant.
Abbreviations
AMH asymptomatic microscopic hematuria
AS active surveillance
CT computed tomography
DRE digital rectal examination
IF incidental finding
MRI magnetic resonance imaging
mp-MRI multiparametric magnetic resonance imaging
PCa prostate cancer
PSA prostate-specific antigen
TRUS transrectal ultrasound
US ultrasound
Fig. 1. Inclusion and exclusion criteria for mp-MRIs included in study.
Fig. 2. Graphic representation of the distribution of incidental findings in multiparametric MRI studies with a prostate indication.
Table 1. PCa-related extraprostatic imaging findings and urologic incidental findings found on multiparametric MRI studies with a prostate indication
Extraprostatic finding related to prostate cancer 119
Suspected osseous metastasis 31
Pelvic lymphadenopathy, suspected metastases 20
Borderline pelvic lymphadenopathy, indeterminate for metastases 17
Suspected seminal vesicle invasion 40
Suspected rectal invasion 2
Suspected bladder invasion 4
Soft tissue nodule anterior to bladder suspicious for metastasis 1
Suspected invasion of urethra 2
Suspected invasion of ejaculatory ducts 2
Non-prostate cancer related urologic incidental findings 170
Bladder diverticulum 8
Bladder wall thickening/trabeculation 89
Bladder/kidney stones 3
Bladder mass 2
Ureteral diverticulum 1
Ureterocele 2
Renal cysts 39
Solid renal masses 5
Atrophic transplant kidney 1
Atrophic native kidneys 1
Hydroureteronephrosis 2
Benign scrotal pathology (hydrocele, varicocele, epididymal cyst) 6
Prostatic utricle cyst, ejaculatory duct cyst, Mullerian duct cyst 10
Urachal remnant/cyst 1
Table 2. Non-urologic incidental findings found on multiparametric MRI studies with a prostate indication stratified by clinical significance
Non-urologic incidental findings (high clinical significance) 23
Mesenteric mass 1
Psoas mass 1
Possible colonic neoplasm 1
Indeterminate liver lesion 4
Abdominal aortic aneurysm > 3 cm 3
Common iliac aneurysms 1
Femoral artery aneurysms 1
Panniculitis 1
Diverticulitis 2
Osteonecrosis 3
Bowel-containing inguinal hernia 3
Bowel-containing ventral hernia 1
Bladder-containing inguinal hernia 1
Non-urologic incidental findings (low to moderate clinical significance) 156
Diverticulosis 71
Fat-containing inguinal hernia 62
Hiatal hernia 1
Fat-containing ventral hernia 1
Gallstones 2
Hepatic cysts 3
Adrenal cystic lesion 1
Ectasia of abdominal aorta 1
Double IVC variant 1
Pelvic venous varices 1
Femoral artery filling defect (prior vascular closure device) 1
Tarlov cyst 3
Enchondroma 1
Small amount of pelvic free fluid 4
Lipomatosis 1
Lipoma 2
Compliance with ethical standards
Conflicts of interest Soroush Rais-Bahrami and Jeffrey W. Nix are consultants for Philips/InVivo Corp. Rachael L. Sherrer, Win Shun Lai, and John V. Thomas have no conflicts of interest to declare.
Ethical approval All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.
References
1. Siegel RL , Miller KD , Jemal A (2017) Cancer statistics. CA Cancer J Clin 67 :7 28055103
2. Rais-Bahrami S , Siddiqui MM , Turkbey B , (2013) Utility of multiparametric magnetic resonance imaging suspicion levels for detecting prostate cancer. J Urol 190 :1721 23727310
3. Nix JW , Turkbey B , Hoang A , (2012) Very distal apical prostate tumours: identification on multiparametric MRI at 3 Tesla. BJU Int 110 :E694 23035719
4. Muthigi A , Sidana A , George AK , (2016) Midline lesions of the prostate: role of MRI/TRUS fusion biopsy and implications in Gleason risk stratification. Int Urol Nephrol 48 :1445 27305918
5. Sankineni S , George AK , Brown AM , (2015) Posterior subcapsular prostate cancer: identification with mpMRI and MRI/TRUS fusion-guided biopsy. Abdom Imaging 40 :2557 25916869
6. Volkin D , Turkbey B , Hoang AN , (2014) Multiparametric magnetic resonance imaging (MRI) and subsequent MRI/ultrasonography fusion-guided biopsy increase the detection of anteriorly located prostate cancers. BJU Int 114 :E43 24712649
7. Hoeks CMA , Barentsz JO , Hambrock T , (2011) Prostate cancer: multiparametric MR imaging for detection, localization, and staging. Radiology 261 :46 21931141
8. Muthigi A , Sidana A , George AK , (2017) Current beliefs and practice patterns among urologists regarding prostate magnetic resonance imaging and magnetic resonance-targeted biopsy. Urol Oncol 35 (32 ):e1
9. Schoots IG , Roobol MJ , Nieboer D , (2015) Magnetic resonance imaging-targeted biopsy may enhance the diagnostic accuracy of significant prostate cancer detection compared to standard transrectal ultrasound-guided biopsy: a systematic review and meta-analysis. Eur Urol 68 :438 25480312
10. Siddiqui Mm , Rais-Bahrami S , Turkbey B , (2015) Comparison of MR/ultrasound fusion-guided biopsy with ultrasound-guided biopsy for the diagnosis of prostate cancer. JAMA 313 :390 25626035
11. Westbrook JI , Braithwaite J , McIntosh JH (1998) The outcomes for patients with incidental lesions: serendipitous or iatrogenic? AJR 171 :1193 9798845
12. Lai WS , Ellenburg J , Lockhart ME , (2016) Assessing the costs of extraurinary findings of computed tomography urogram in the evaluation of asymptomatic microscopic hematuria. Urology 95 :34 27318261
13. Morgan AE , Berland LL , Ananyev SS , (2015) Extraurinary incidental findings on CT for hematuria: the radiologist’s role and downstream cost analysis. AJR Am J Roentgenol 204 :1160 26001224
14. Yee J , Kumar NN , Godara S , (2005) Extracolonic abnormalities discovered incidentally at CT colonography in a male population. Radiology 236 :519 16040909
15. Priola AM , Priola sM , Giaj-Levra M , (2013) Clinical implications and added costs of incidental findings in an early detection study of lung cancer by using low-dose spiral computed tomography. Clin Lung Cancer 14 :139 22868220
16. MacHaalany J , Yam Y , Ruddy TD , (2009) Potential clinical and economic consequences of noncardiac incidental findings on cardiac computed tomography. JACC 54 :1533 19815125
17. Lumbreras B , Donat L , Hernández-Aguado I (2010) Incidental findings in imaging diagnostic tests: a systematic review. Br J Radiol 83 :276 20335439
18. Hara AK , Johnson CD , MacCarty RL , (2000) Incidental extracolonic findings at CT conolonography. Radiology 215 :353 10796907
19. Gleucker TM , Johnson CD , Wilson LA , (2003) Extracolonic findings at CT colonography: evaluation of prevalence and cost in a screening population. Gastroenterology 124 :911 12671887
20. Lee CI , Tsai EB , Sigal BM , (2010) Incidental extracardiac findings at coronary CT: clinical and economic impact. AJR Am J Roentgenol 194 :1531 20489093
21. Lai WS , Gordetsky JB , Thomas JV , (2017) Factors predicting prostate cancer upgrading on magnetic resonance imaging-targeted biopsy in an active surveillance population. Cancer 123 :1941 28140460
22. Wang L , Hricak H , Kattan MW , (2006) Prediction of organ-confined prostate cancer: incremental value of MR imaging and MR spectroscopic imaging to staging nomograms. Radiology 238 :597 16344335
23. Feng T , Koopman S , Kim HL (2015) Impact of post prostate biopsy hemorrhage on multiparametric magnetic resonance imaging. Can J Urol 22 :5
24. Scheltema MJ , Tay KJ , Postema AW , (2017) Utilization of multiparametric prostate magnetic resonance imaging in clinical practice and focal therapy: report from a Delphi consensus project. World J Urol 35 :695 27637908
|
PMC008xxxxxx/PMC8983113.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
8701680
1585
Am J Health Promot
Am J Health Promot
American journal of health promotion : AJHP
0890-1171
2168-6602
33949215
8983113
10.1177/08901171211012951
NIHMS1789531
Article
Theoretical Mediators of Diabetes Risk and Quality of Life Following a Diabetes Prevention Program for Latino Youth With Obesity
Soltero Erica G. PhD http://orcid.org/0000-0002-7202-9262
1
Ayers Stephanie L. PhD 2
Avalos Marvyn A. MS 2
Peña Armando MS 3
Williams Allison N. MSW 3
Olson Micah L. MD 34
Konopken Yolanda P. BS 5
Castro Felipe G. PhD, MSW 3
Arcoleo Kimberly J. PhD 6
Keller Colleen S. PhD 3
Patrick Donald L. PhD, MSPH 7
Jager Justin PhD 28
Shaibi Gabriel Q. PhD http://orcid.org/0000-0002-6890-2903
3
1 Children’s Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine, Houston, TX, USA
2 Southwest Interdisciplinary Research Center, Arizona State University, Phoenix, AZ, USA
3 Center for Health Promotion and Disease Prevention, Arizona State University, Phoenix, AZ, USA
4 Department of Pediatric Endocrinology and Diabetes, Phoenix Children’s Hospital, Phoenix, AZ, USA
5 Family Wellness Program, St. Vincent De Paul Medical and Dental Clinic, Phoenix, AZ, USA
6 School of Nursing, University of Rochester, Rochester, NY, USA
7 Department of Health Services, School of Public Health, University of Washington, Seattle, WA, USA
8 T. Denny Sanford School of Social and Family Dynamics, Arizona State University, Tempe, AZ, USA
Corresponding Author: Erica G. Soltero, Children’s Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine, 1100 Bates Avenue, Houston, TX 77030, USA: [email protected]
25 3 2022
9 2021
05 5 2021
01 9 2022
35 7 939947
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Purpose:
This study tested self-efficacy and social support for activity and dietary changes as mediators of changes in type 2 diabetes related outcomes following a lifestyle intervention among Latino youth.
Setting and Intervention:
Latino adolescents (14–16 years) with obesity (BMI% = 98.1 ± 1.4) were randomized to a 3-month intervention (n = 67) that fostered self-efficacy and social support through weekly, family-centered sessions or a comparison condition (n = 69).
Measures:
Primary outcomes included insulin sensitivity and weight specific quality of life. Mediators included self-efficacy, friend, and family social support for health behaviors. Data was collected at baseline, 3-months, 6-months, and 12-months.
Analysis:
Sequential path analysis was used to examine mediators as mechanisms by which the intervention influenced primary outcomes.
Results:
The intervention had a direct effect on family (β = 0.33, P < .01) and friend social support (β = 0.22, P < .001) immediately following the intervention (3-months). Increased family social support mediated the intervention’s effect on self-efficacy at 6-months (β = 0.09, P < .01). However, social support and self-efficacy did not mediate long-term changes in primary outcomes (P > .05) at 12-months.
Conclusions:
Family social support may improve self-efficacy for health behaviors in high-risk Latino youth, highlighting the important role of family diabetes prevention. Fostering family social support is a critical intervention target and more research is needed to understand family-level factors that have the potential to lead to long-term metabolic and psychosocial outcome in vulnerable youth.
Latino health
adolescents
family
diabetes prevention
theory
pmcPurpose
Obesity and type 2 diabetes (T2D) disproportionately impact Latino populations. The prevalence of T2D is twice as high in Latino adults (22.6%) compared to non-Hispanic whites (11.3%).1 These disparities emerge early as Latino youth exhibit higher rates of prediabetes (22.9%) compared to white youth (15.1%).2 Given that the estimated yearly costs of T2D is ~$327 billion and that Latino youth are the fastest growing pediatric subpopulation in the U.S., preventing T2D in high-risk Latino youth is a public health priority.3
Theory-based lifestyle interventions are the first-line approach for the prevention and management of pediatric obesity. Social Cognitive Theory (SCT) is the most widely applied theory in pediatric obesity interventions and one of the most commonly used theoretical frameworks in interventions designed for Latino youth.4,5 Previous literature has demonstrated that SCT constructs like social support and self-efficacy are associated with changes in obesity-related health behaviors in youth, but limited evidence has extended to changes in disease outcomes.6 Social support is defined as the provision of material and/or interpersonal resources through social connections or linkages.7 The primary processes for fostering social support within obesity interventions among youth include providing informational support through knowledge on healthy eating and activity, instrumental support by providing resources for engaging in health behaviors, modeling support from instructors and family members by demonstrating or participating in health behaviors with youth, and emotional support by providing praise or encouragement for behavior change.6,7 Social support for youth primarily comes from family and friends and it is hypothesized that different sources of support may have different influences on health.8 Family support is strongly linked to health behaviors in youth, and this construct may be particularly salient within the context of Latino culture where familismo (familism) is a core value.9 Self-efficacy is defined as confidence in one’s ability to perform a behavior.10 The primary processes within obesity prevention interventions for fostering self-efficacy include building mastery experiences through activities like cooking demonstrations and exercises sessions, developing health knowledge and behavior change skills, and addressing beliefs about one’s ability to make healthy lifestyle changes.4 According to the SCT, interpersonal processes such as social support can lead to increased self-efficacy.11,12 For example, social supports that include role modeling, verbal encouragement, or engaging together in behaviors, provide opportunities to observe others and have the mastery experiences needed to build one’s confidence in their ability to perform that behavior.13
Given the associations of social support and self-efficacy with health behaviors, these constructs are hypothesized as theoretical mechanisms that contribute to health outcomes and have been used to develop successful diabetes prevention programs for adults.14 Given the effectiveness of diabetes prevention programs for reducing diabetes risk in adults, similar programs grounded in SCT have been proposed for high-risk youth.15 However, there is a lack of prospective empirical studies in both adults and youth that test whether changes in self-efficacy and social support are the underlying mechanisms by which diabetes prevention programs are efficacious.16,17 This gap limits the field from identifying intervention inputs that drive and support changes in disease outcomes in high-risk populations like Latino youth.
Following participation in a 12-week, culturally-grounded, lifestyle intervention, we observed significant short-term improvements (12-week) in insulin sensitivity and short- and long-term (12-months) improvements in weight-specific quality of life (QoL-W). The intervention impact on primary outcomes has been previously published.18 Therefore, the purpose of the current study is to test self-efficacy and social support for health behaviors as mediators of long-term changes in diabetes risk and weight-specific quality of life (QoL-W) following a culturally-grounded diabetes prevention program for Latino youth with obesity. We hypothesize that the intervention will lead to increases in self-efficacy and social support, which will mediate long-term changes in insulin sensitivity and QoL-W.
Methods
Study Design
Participants.
Latino youth were recruited through clinical, community, and media outlets. Clinical referral sites included Latino-serving ambulatory pediatric clinics, Federally Qualified Health Centers, school-based health clinics, safety net clinics, and a tertiary care children’s hospital. Community recruitment efforts included health coalitions, health fairs, churches, and community-based organizations. The media strategy used Spanish-language advertisements in local magazines and newspapers. Collectively, these strategies yielded 913 referrals who were screened initially by phone and then in person for the following inclusion criteria: 1) self-identify as Latino, 2) age 14–16, and 3) obese (BMI >95th percentile for age and sex). Exclusion criteria included: 1) taking medication(s) or diagnosed with a condition that influenced carbohydrate metabolism, physical activity (PA), and/or cognition, 2) diagnosed with T2D, 3) enrolled in a weight loss program, or 4) diagnosed with depression or any other condition that may impact quality of life. A total of 160 Latino youth met all inclusion / exclusion criteria and agreed to participate. From this sample, 24 youth were found to have prediabetes and, by protocol, were automatically assigned to the lifestyle intervention arm of the study. Since these youth were not included in the randomization schedule, the current analysis includes 136 Latino youth with obesity who were randomized to either the 3-month lifestyle intervention or comparison control condition (described below).
Procedures.
All study procedures and materials were approved by the Institutional Review Board at Arizona State University. Written informed consent and assent were obtained from parents and youth prior to study procedures. This study is registered at www.clinicaltrials.gov (Clinicaltrials.gov Identifier: NCT02039141) and the protocol has been previously published.19 Data collection was performed in the clinical research unit at Arizona State University by trained research staff using identical procedures as baseline (T1), 3-months (T2), 6-months (T3), and 12-months (T4). Obesity was assessed as BMI and BMI percentiles from height and weight measures and percent fat using bioelectrical impedance analysis.
Primary Outcomes
Diabetes Risk.
A 2-hour 75-gram Oral Glucose Tolerance Test (OGTT) was used to estimate insulin sensitivity using insulin and glucose concentrations collected every 30 minutes. This measure of insulin sensitivity calculates a range from 0–12 with smaller values corresponding to lower levels of insulin sensitivity or higher risk for diabetes.20
Quality of life.
QoL-W was a mean scale calculated from 26-items that measured quality of life in the domains of self, social relationships, and environment as they pertain to weight-related concerns.21 This mean scale has been validated in adolescents in community and clinic settings with Latino youth.21 Examples of questions include, “I feel depressed about how much I weigh;” “Because of my weight other people think I am unattractive;” and “Because of my weight I avoid being seen in a swim suit.” Response categories ranged from (0) “Not at all” to (10) “Very much”21 with higher scores reflecting greater QoL. At T4, quality of life showed good reliability (Cronbach’s α = 0.968).
Mediators
Self-efficacy.
Self-efficacy was assessed using the previously validated Physician-based Assessment & Counseling for Exercise (PACE+): Physical Activity and Diet Surveys for Adolescents.22 This survey assesses participant’s confidence in their ability to change behaviors relating to physical activity (6-items) and fruit, vegetable, and fat intake (18-items), using a 5-point Likert scale with (1) ‘I’m sure I can’t’ to (5) ‘I’m sure I can’ with higher scores indicating greater self-efficacy. This 24-item mean scale included questions like, “Rate how sure you are that you can do physical activity when you feel sad” and “Rate how sure you are that you can choose low-fat foods when you are craving high fat?”23 Self-efficacy had high reliability at both T2 (Cronbach’s α = 0.947) and T3 (Cronbach’s α = 0.937).
Social support.
Social support for physical activity and healthy eating habits was also assessed using the PACE+ Physical Activity and Diet Survey for Adolescents.22 This instrument measures 2 separate mean scales-social support from family (13-items) and social support from friends (9-items). These items ask the frequency with which a family member or friend encourages, participates, or provides assistance (e.g. transportation) for physical activity in a typical week. Response categories for all questions were on a 5-point Likert scale from (1) “never” to (5) “every day” with greater scores indicating greater social support.22 Social support from family members included questions like, “During a typical week, how often has a member of your household provided fruits and vegetables as a snack?” Social support from family members demonstrated good reliability at T2 (Cronbach’s α = 0.904) and T3 (Cronbach’s α = 0.890). An example of friend social support questions include “During a typical week, do your friends tell you that you are doing well in physical activity or sports?” Social support from friends also showed high reliability at T2 (Cronbach’s α = 0.829) and T3 (Cronbach’s α = 0.863).
Group Assignment
Intervention group.
The intervention was delivered to families at a local YMCA by bilingual/bicultural community health educators. The sessions were delivered in both Spanish and English as the majority of the parents preferred communicating in Spanish while the majority of the youth preferred communicating in English. The intervention consisted of nutrition and health education, exercise, and behavior change strategies. The curriculum was informed by the SCT, mainly self-efficacy and social support,19 and was culturally-grounded, integrating Latino cultural values such as familismo (familism) and respeto (respect) into activities and content.19 Self-efficacy was integrated into the curriculum with families participating in weekly goal-setting, self-monitoring, and through mastery experiences such as preparing a meal and exercising together as a family. Intervention implementers provided participants with various forms of social support by discussing roles and responsibilities, skills development for building confidence for making healthy changes in physical activities and dietary behaviors. Social support within and between families was fostered through activities such as discussing how to overcome challenges and barriers to meeting behavior change goals.19 Following the 3-month intervention, youth participated in monthly booster sessions, which were designed to reinforce and celebrate behavior changes, address challenges in maintaining healthy behaviors, and provide ongoing social support and encouragement for behavior change.
Comparison group.
Youth in the comparison group received their lab results and a one-page handout on healthy lifestyle behaviors. Comparison youth also received monthly contact from a research team member and were given a 1-year YMCA membership and the opportunity to participate in an abridged version of the intervention at the conclusion of the study.
Statistical Analyzes
Using Mplus 7.0, sequential multivariate linear regression path analyzes were conducted to examine the direct effects of the intervention on social support and self-efficacy and the indirect, mediation effects of the intervention on insulin sensitivity and quality of life through social support and self-efficacy. Global model fit was assessed using the chi-square goodness of fit statistic, the root mean square error of approximation (RMSEA), and the comparative fit index (CFI). To reduce bias of the parameter estimates, 95% bias corrected confidence intervals were calculated based on 10,000 bootstrap samples. The standardized betas (β) are reported for the direct and indirect paths, and all models control for T1 outcomes of interest, age, and sex. Overall, missingness at T4 was low (19%); nonetheless, to avoid list-wise deletion and maximize available data, we utilized the full-information maximum likelihood (FIML)24 estimation to handle missing data. FIML accounts and adjusts for attrition over time and allows for intent-to-treat analyzes of the data (N = 136).
At each sequential stage, we performed model optimization, trimming non-significant paths. Paths were deemed superfluous if there was no significant contribution to model fit.25 To compare the more complex model to the trimmed parsimonious model, a chi-square difference test was performed.25 A non-significant chi-square difference test indicated the trimmed paths were not contributing to overall model fit and could be omitted from the model without loss of model fit. The process of model optimization aims to find a more parsimonious model that still explains relationships in the data, 6-months post-intervention.
Results
A total of 136 Latino adolescents (14–16 years old) with obesity (BMI% = 98.1 ± 1.4), were randomized to the intervention (n = 67) or comparison (n = 69). Table 1 presents demographic, anthropometric, metabolic, quality of life, self-efficacy, and social support data. There were no significant differences on any of these variables between youth in the intervention and comparison group.
The direct effect of the intervention on T2 family social support, friend social support, and self-efficacy are presented in Figure 1. Compared to the comparison group, youth in the intervention group reported significant increases in family (β = 0.33, P < .01) and friend social support for health behaviors (β = 0.22, P < .001) at T2. In contrast, the intervention had no direct effect on self-efficacy for health behaviors (β = 0.05, P > .05) immediately following the intervention. This model exhibited good fit (χ2[11] = 14.18, P = .22; CFI = .986; RMSEA = 0.046). A more parsimonious model that did not include self-efficacy at T2 (χ2[6] = 4.27) indicated that self-efficacy for health behaviors at T2 can be omitted from the model (Δ χ2[5] = 9.91, P = .08).
A mediation analysis (Figure 2) to test the indirect effect of the intervention on T3 self-efficacy demonstrated that changes in self-efficacy for health behaviors at T3 were mediated by family social support for health behaviors at T2 (β = 0.09, P < .01). In contrast, the indirect path for changes in self-efficacy at T3 mediated through friend social support at T2 was not significant (β = −0.03, P > .05). This revised model exhibited a good fit (χ2[12] = 11.61, P = .48; CFI = 1.00; RMSEA = 0.00). The chi-square difference test indicates that friend social support at T2 could be omitted from the model (Δ χ2[5] = 3.70, P = .59).
The final model tested direct and indirect effects of the intervention on changes in insulin sensitivity and QoL-W at T4. Although the intervention demonstrated significant direct effects on increasing insulin sensitivity (β = 0.15, P < .05) and QoL-W (β = 0.25, P < .001), there were no mediated pathways through family social support or self-efficacy on insulin sensitivity or QoL-W (P’s > .05) at T4 (Figure 3). This model exhibited a good fit χ2[21] = 15.03, P = .82; CFI = 1 .00; RMSEA = 0.00).
Discussion
Few studies have tested the mechanisms that underpin successful diabetes prevention interventions. This study tested social support and self-efficacy for healthy behaviors as mediators of long-term changes in insulin sensitivity and QoL-W following a culturally grounded diabetes prevention program. Although the intervention had a direct effect on family and friend social support for health behaviors, neither social support nor self-efficacy mediated improvements in health outcomes. We did observe that family social support increased self-efficacy for health behaviors in high-risk Latino youth, which is in line with SCT and affirms that social support is an important mechanism for fostering self-efficacy in this population. More research is needed to elucidate the role that these theoretical constructs play in mediating long-term changes in metabolic and psychosocial outcomes in youth.
The intervention did not lead to direct improvements in self-efficacy for health behaviors. Previous health promotion and diabetes prevention interventions have reported increased self-efficacy following a lifestyle intervention.26 However, these interventions were focused on individuals whereas the current intervention is family-focused with group-based activities.26 Self-efficacy is an intrapersonal factor used to improve health behavior change at the individual level, which may explain why we did not observe a direct effect of the intervention on self-efficacy.27 While enhancing self-efficacy for healthy eating and physical activity was a behavioral change target in designing the curriculum, the group-based, family-focused nature of the intervention may have limited the intervention’s direct effect on self-efficacy. Our findings are consistent with other group-based physical activity and obesity prevention interventions among adolescents that also report no direct intervention effect on self-efficacy.28–30 In contrast, the intervention significantly improved family and friend social support for health behaviors. These findings are consistent with previous obesity prevention interventions that have reported increases in perceived support from family and friends.31 The intervention enhanced social support within families by delivering content in a family-oriented, collectivist manner, and through activities like team sports and preparing a meal together.19 Fostering a supportive social environment allowed family and peers to serve as role models for one another, build social connections with each other, and share encouragement.32
Interestingly, family social support mediated changes in self-efficacy after the intervention. SCT holds that social processes such as social support can facilitate reciprocal determinism, observational learning, and modeling as mechanisms to increase self-efficacy and, in turn support behavior change.10 Our findings are consistent with multiple studies among youth and young adults that have demonstrated increases in social support contribute to subsequent increases in self-efficacy.33–36 Findings from these studies and ours suggest that there may be a sequential order from enhanced social support to increased self-efficacy and that social support from family may be the pathway by which self-efficacy for health behaviors is increased among Latino adolescents.35 This finding provides novel information about the timing by which theoretical constructs within SCT may operate in this vulnerable, high-risk population.10,35 While increased family social support led to increased self-efficacy, there was no relationship between friend social support and self-efficacy. Previous studies have shown that family social support for health behaviors is more influential in early adolescence compared to late adolescence, which is traditionally marked by a shift in social ties toward friends.37 However, our data suggest that family support remains an important leverage point for enhancing self-efficacy and promoting healthy lifestyle behaviors in Latino youth. Because parental support for health behaviors typically declines during this developmental period,37 enhancing family social support during adolescence is critical, particularly among high-risk Latino youth who report lower levels of family social support for health behaviors compared to Non-Hispanic white youth. Fostering and sustaining family social support represents an important intervention input given that familism is a strong Latino cultural value and that social support is consistently associated with improved dietary habits and physical activity in Latinos.38
It is important to note that while the intervention led to increased family social support, which in turn led to increased self-efficacy, this pathway did not mediate long-term changes in insulin sensitivity or QoL-W. Recent reviews of obesity prevention interventions have reported that there is a lack of evidence supporting theoretical constructs as mediators of intervention effects.39 Using this study as an example, we highlight critical next steps for behavioral interventions to promote the testing of theoretical mediators that guide the intervention. Greater consideration for the operationalization and evaluation of theoretical constructs is needed to adapt and refine constructs so that intervention inputs can be better aligned with intervention strategies.40 In the current study, retrospective intervention mapping revealed that social support was more deeply integrated throughout all intervention components compared to self-efficacy. This may explain why we did not observe a direct intervention effect on self-efficacy. Future iterations of the intervention should improve the operationalization of this construct to ensure that intervention strategies effectively target this mediator to avoid misinterpretations of the effectiveness of theoretical mediators.41
Future studies should also more rigorously evaluate theoretical mediators using formal mediation analyzes to identify the underlying mechanisms by which lifestyle interventions reduce disease outcomes.16 However, for underrepresented populations in particular, this also includes the need for researchers to obtain a deeper understanding of the appropriateness of theoretical mediators within the sociocultural context of the population of interest.42 While our measures of social support and self-efficacy were validated in Latino youth, these validation studies took place before the current sample of youth was born and these measures alone may not fully capture contemporary issues or cultural nuances that influence these relevant constructs within this population. For example, theoretical constructs may be influenced by broader social factors like acculturation or other social determinants experienced by Latino families.43 Thus, the ongoing evaluation and adaptation of mediators should take contemporary issues and broader social influences into consideration. Embedding qualitative and mixed-methods approaches may be helpful in gaining a deeper understanding of theoretical mediators, potential mechanisms, and other important constructs that may be operational within intervention trials.44
In addition to providing a deeper understanding of theoretical mediators, the rich, contextual insight gained from qualitative methods may point to the need to adapt evidence-based interventions to better meet the sociocultural needs of ethnic minority groups. For example, the SCT emphasizes individual and intrapersonal factors as opposed to broader social and environmental variables.10 Health disparities research has shown that individual-level strategies may not be efficacious within vulnerable communities where there is a clear need to leverage social and familial connectedness for health promotion and disease prevention.45 In fact, the most effective strategies for vulnerable, minority populations recognize that health behaviors occur in the context of relationships with family and friends.46 Thus, interventions grounded in the SCT may need to be adapted to account for social and familial factors in order to maximize intervention effectiveness.32 Culturally adapting evidence-based interventions has been identified as a key strategy for addressing health disparities.47 Frameworks like the Ecologic Validity Model provide a systematic process for culturally adapting and rigorously evaluating evidence-based interventions.42 Systematically documenting the adaptation process is critical for evaluating the pathways by which mediators and adapted intervention components interact with and influence biological systems.
Conclusion
Advancing diabetes prevention science focusing on minority youth is particularly urgent considering the growing costs of T2D in the U.S. and the widening disparities among high-risk youth. The recent vision report published by the National Advisory Council on Minority Health and Health Disparities underscored the importance of advancing the knowledge base for effective disease prevention interventions and the mechanisms by which these interventions affect disease outcomes.46 Therefore, the purpose of this study was to test self-efficacy and friend and family social support for healthy eating and physical activity as mediators of long-term changes in diabetes risk and QoL-W following a culturally-grounded diabetes prevention program for Latino youth. While social support and self-efficacy did not mediate long-term changes in insulin sensitivity or QoL-W, our findings suggest that family support plays an important role in the context of health promotion and disease prevention for Latino youth. Coordinated efforts to enhance social support within and between high-risk families living in vulnerable and underserved communities is an important framework and potential mechanism to increase the efficacy and effect size of prevention programs that aim to advance health equity.
Acknowledgments
We are grateful to our collaborators from the Family Wellness Program at the St. Vincent de Paul Medical and Dental Clinic and the Valley of the Sun YMCA. We are indebted to the children and families who participated in this study.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by the National Institutes of Health through the National Institute on Minority Health and Health Disparities (P20MD002316; U54MD002316) and the National Institute on Diabetes and Digestive and Kidney Diseases (R01DK107579). This work is also a publication of the United States Department of Agriculture, Agricultural Research Service (USDA/ARS), Children’s Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine, Houston, Texas, and funded in part with federal funds from the USDA/ARS under Cooperative Agreement No. 58-3092-0-001.
Figure 1. Direct effects of the intervention on self-efficacy for health behaviors and friend and family social support for health behaviors.
Figure 2. The indirect effect of the intervention on self-efficacy for health behaviors through family and friend social support.
Figure 3. The indirect effect of the intervention on insulin sensitivity (T4) and weight-specific quality of life (T4) through family social support for health behaviors (T2) and self-efficacy for health behaviors (T3).
Table 1. Baseline Participant Characteristics.
Variable Control (M ± SD) Intervention (M ± SD) P
Age (years) Sex (%) 15.1 ± 0.9 15.4 ± 1.0 .10
Boys 50.7% 40.3% .22
Girls 49.3% 59.7%
County of origin—youth (%)
U.S-born 73.9% 81.8% .27
Foreign-born 26.1% 18.2%
County of origin-parent (%)
U.S-born 9.0% 12.3% .53
Foreign-born 91.0% 87.7%
Body mass index (%) 98.3 ± 1.2 98.1 ± 1.4 .36
Body fat (%) 44.7 ± 7.6 45.2 ± 7.2 .67
Insulin sensitivity 1.6 ± 1.2 1.7 ± 1.0 .60
Weight-specific quality of life 64.6 ± 25.7 63.9 ± 24.0 .86
Self-efficacy for health behaviors 3.6 ± 0.8 3.8 ± 0.8 .11
Family social support for health behaviors 3.0 ± 0.8 3.1 ± 0.9 .56
Friend social support for health behaviors 2.2 ± 0.8 2.3 ± 0.8 .77
So What?
What is already known on this topic?
Latino youth and families are disproportionately impacted by type 2 diabetes. Interventions that are theoretically-driven are a first-line approach for diabetes prevention. The Social Cognitive Theory (SCT) has been widely applied and associated with improved health behaviors.
What does this article add?
This study examined SCT constructs (social support and self-efficacy) as mediators of diabetes related outcomes in Latino youth following a lifestyle intervention. SCT constructs did not mediate long-term outcomes; however, increased family social support led to increased self-efficacy, suggesting social support may be needed to increase self-efficacy.
What are the implications for health promotion practice or research?
Fostering family social support for health behaviors is a critical intervention target for diabetes prevention programs in high-risk Latino youth. More research is needed to understand theoretical constructs like family-level social factors and their mediating role on disease outcomes in Latino families.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
References
1. Menke A , Casagrande S , Geiss L , Cowie CC . Prevalence of and trends in diabetes among adults in the United States, 1988–2012. JAMA. 2015;314 (10 ):1021–1029.26348752
2. Menke A , Casagrande S , Cowie CC . Prevalence of diabetes in adolescents aged 12 to 19 years in the United States, 2005–2014. JAMA. 2016;316 (3 ):344–345.27434447
3. Rhodes ET , Goran MI , Lieu TA , Health-related quality of life in adolescents with or at risk for type 2 diabetes mellitus. J Pediatr. 2012;160 (6 ):911–917.22217471
4. Contento IR , Koch PA , Lee H , Calabrese-Barton A . Adolescents demonstrate improvement in obesity risk behaviors after completion of choice, control & change, a curriculum addressing personal agency and autonomous motivation. J Am Diet Assoc. 2010; 110 (12 ):1830–1839.21111093
5. Bandura A Social cognitive theory of self-regulation. Organ Behav Hum Decis Process. 1991;50 (2 ):248–287.
6. Donnelly R , Springer A . Parental social support, ethnicity, and energy balance-related behaviors in ethnically diverse, low-income, urban elementary schoolchildren. J Nutr Educ Behav. 2015;47 (1 ):10–18.25106093
7. Thompson RA , Flood MF , Goodvin R . Social support and developmental psychopathology. In: Cicchetti D , Cohen DJ , eds. Developmental Psychopathology: Volume Three: Risk, Disorder, and Adaptation. John Wiley & Sons; 2006:1–37.
8. Nelson EE , Leibenluft E , McClure EB , Pine DS . The social reorientation of adolescence: a neuroscience perspective on the process and its relation to psychopathology. Psychol Med. 2005; 35 (2 ):163–174.15841674
9. Hu J , Amirehsani KA , Wallace DC , McCoy TP , Silva Z . A family-based, culturally tailored diabetes intervention for Hispanics and their family members. Diabetes Educ. 2016; 42 (3 ):299–314.26957533
10. Bandura A Self-efficacy: toward a unifying theory of behavioral change. Psychol Rev. 1977;84 (2 ):191–215.847061
11. Bandura A Health promotion by social cognitive means. Health Educ Behav. 2004;31 (2 ):143–164.15090118
12. Welk GJ , Almeida J , Morss G . Laboratory calibration and validation of the Biotrainer and Actitrac activity monitors. Med Sci Sports Exerc. 2003;35 (6 ):1057–1064.12783056
13. Dishman RK , Saunders RP , Motl RW , Dowda M , Pate RR . Self-efficacy moderates the relation between declines in physical activity and perceived social support in high school girls. J Pediatr Psychol. 2009;34 (4 ):441–451.18812410
14. Knowler WC , Barrett-Connor E , Fowler SE , Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. N Engl J Med. 2002;346 (6 ):393–403.11832527
15. Franks PW , Huang TT , Ball GD . Lifestyle intervention for type 2 diabetes risk reduction: using the diabetes prevention program to inform new directions in pediatric research. Can J Diabetes. 2007;31 (3 ):242–251.
16. Katzmarzyk PT , Baur LA , Blair SN , International conference on physical activity and obesity in children: summary statement and recommendations. Int J Pediatr Obes. 2008; 3 (1 ):3–21.
17. Bowen PG , Clay OJ , Lee LT , Vice J , Ovalle F , Crowe M . Associations of social support and self-efficacy with quality of life in older adults with diabetes. J Gerontol Nurs. 2015;41 (12 ): 21–29; quiz 30–31.
18. Soltero EG , Olson ML , Williams AN , Effects of a community-based diabetes prevention program for Latino youth with obesity: a randomized controlled trial. Obesity (Silver Spring). 2018;26 (12 ):1856–1865.30426694
19. Williams AN , Konopken YP , Keller CS , Culturally-grounded diabetes prevention program for obese Latino youth: rationale, design, and methods. Contemp Clin Trials. 2017;54 : 68–76.28104469
20. Matsuda M , DeFronzo RA . Insulin sensitivity indices obtained from oral glucose tolerance testing: comparison with the euglycemic insulin clamp. Diabetes Care. 1999;22 (9 ):1462–1470.10480510
21. Patrick DL , Skalicky AM , Edwards TC , Weight loss and changes in generic and weight-specific quality of life in obese adolescents. Qual Life Res. 2011;20 (6 ):961–968.21188537
22. Norman GJ , Sallis JF , Gaskins R . Comparability and reliability of paper- and computer-based measures of psychosocial constructs for adolescent physical activity and sedentary behaviors. Res Q Exerc Sport. 2005;76 (3 ):315–323.16270708
23. Hagler AS , Norman GJ , Radick LR , Calfas KJ , Sallis JF . Comparability and reliability of paper- and computer-based measures of psychosocial constructs for adolescent fruit and vegetable and dietary fat intake. J Am Diet Assoc. 2005; 105 (11 ):1758–1764.16256760
24. Graham JW . Missing data analysis: making it work in the real world. Ann Rev Psychol. 2009;60 :549–576.18652544
25. Kline RB . Principles and Practice of Structural Equation Modeling. The Guilford Press; 1998.
26. Geria K , Beitz JM . Application of a modified diabetes prevention program with adolescents. Public Health Nurs. 2018;35 (4 ): 337–343.29285793
27. Jurkowski JM , Lawson HA , Green Mills LL , Wilner PG III , Davison KK . The empowerment of low-income parents engaged in a childhood obesity intervention. Fam Community Health. 2014;37 (2 ):104–118.24569157
28. Lubans DR , Morgan PJ , Callister R , Collins CE , Plotnikoff RC . Exploring the mechanisms of physical activity and dietary behavior change in the program x intervention for adolescents. J Adolesc Health. 2010;47 (1 ):83–91.20547296
29. van Stralen MM , Yildirim M , te Velde SJ , What works in school-based energy balance behaviour interventions and what does not? A systematic review of mediating mechanisms. Int J Obes (Lond). 2011;35 (10 ):1251–1265.21487398
30. Wieland ML , Biggs BK , Brockman TA , Club fit: development of a physical activity and healthy eating intervention at a boys & girls club after school program. J Prim Prev. 2020;41 (2 ): 153–170.32096111
31. Mendonça G , Cheng LA , Mélo EN , de Farias Júnior JC . Physical activity and social support in adolescents: a systematic review. Health Educ Res. 2014;29 (5 ):822–839.24812148
32. Huang TT , Goran MI . Prevention of type 2 diabetes in young people: a theoretical perspective. Pediatr Diabetes. 2003; 4 (1 ):38–56.14655523
33. Wang Y , Hager ER , Magder LS , Arbaiza R , Wilkes S , Black MM . A dyadic analysis on source discrepancy and a mediation analysis via self-efficacy in the parental support and physical activity relationship among black girls. Child Obes. 2019;15 (2 ):123–130.30653347
34. Trost SG , Sallis JF , Pate RR , Freedson PS , Taylor WC , Dowda M . Evaluating a model of parental influence on youth physical activity. Am J Prev Med. 2003;25 (4 ):277–282.14580627
35. Rovniak LS , Anderson ES , Winett RA , Stephens RS . Social cognitive determinants of physical activity in young adults: a prospective structural equation analysis. Ann Behav Med. 2002; 24 (2 ):149–156.12054320
36. Middelweerd A , te Velde SJ , Abbott G , Timperio A , Brug J , Ball K . Do intrapersonal factors mediate the association of social support with physical activity in young women living in socioeconomically disadvantaged neighbourhoods? A longitudinal mediation analysis. PLoS One. 2017;12 (3 ):e0173231.28301538
37. Davison KK , Jago R . Change in parent and peer support across ages 9 to 15yr and adolescent girls’ physical activity. Med Sci Sports Exerc. 2009;41 (9 ):1816–1825.19657287
38. Marquez B , Elder JP , Arredondo EM , Madanat H , Ji M , Ayala GX . Social network characteristics associated with health promoting behaviors among Latinos. Health Psychol. 2014;33 (6 ): 544–553.24884908
39. Bagherniya M , Taghipour A , Sharma M , Obesity intervention programs among adolescents using social cognitive theory: a systematic literature review. Health Educ Res. 2018;33 (1 ):26–39.29293954
40. Michie S , Johnston M , Francis J , Hardeman W , Eccles M . From theory to intervention: mapping theoretically derived behavioural determinants to behaviour change techniques. Appl Psychol. 2008;57 (4 ):660–680.
41. Cerin E , Barnett A , Baranowski T . Testing theories of dietary behavior change in youth using the mediating variable model with intervention programs. J Nutr Educ Behav. 2009;41 (5 ): 309–318.19717113
42. Bernal G , Sáez-Santiago E . Culturally centered psychosocial interventions. J Community Psychol. 2006;34 (2 ):121–132.
43. Castro FG , Barrera M Jr , Holleran Steiker LK . Issues and challenges in the design of culturally adapted evidence-based interventions. Annu Rev Clin Psychol. 2010;6 :213–239.20192800
44. Creswell JW , Creswell JD . Research Design: Qualitative, Quantitative, and Mixed Methods Approaches. 5th ed. Sage Publications, Inc.; 2017.
45. Kennedy L , Pinkney S , Suleman S , Mâsse LC , Naylor PJ , Amed S . Propagating change: using RE-FRAME to scale and sustain a community-based childhood obesity prevention initiative. Int J Environ Res Public Health. 2019;16 (5 ):736.
46. Alvidrez J , Stinson N Jr . Sideways progress in intervention research is not sufficient to eliminate health disparities. Am J Public Health. 2019;109 (S1 ):S102–104.30699028
47. Jones NL , Breen N , Das R , Farhat T , Palmer R . Cross-cutting themes to advance the science of minority health and health disparities. Am J Public Health. 2019;109 (S1 ):S21–s24.30699031
|
PMC008xxxxxx/PMC8983114.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
101231934
33303
Heart Fail Clin
Heart Fail Clin
Heart failure clinics
1551-7136
35341541
8983114
10.1016/j.hfc.2021.12.002
NIHMS1788547
Article
Advances in Machine Learning Approaches to Heart Failure with Preserved Ejection Fraction
Ahmad Faraz S. MD, MS abc
Luo Yuan PhD bc
Wehbe Ramsey M. MD, MSAI ac
Thomas James D. MD ac
Shah Sanjiv J. MD ac
a Division of Cardiology, Department of Medicine, Northwestern University Feinberg School of Medicine, Chicago, IL
b Department of Preventive Medicine, Northwestern University Feinberg School of Medicine, Chicago, IL
c Bluhm Cardiovascular Institute Center for Artificial Intelligence, Northwestern Medicine, Chicago, IL
Corresponding Author: Sanjiv J. Shah, MD, Stone Professor of Medicine, Director of Research, Bluhm Cardiovascular Institute, Division of Cardiology, Department of Medicine, Director, Center for Deep Phenotyping and Precision Therapeutics, Institute for Augmented Intelligence in Medicine, Northwestern University Feinberg School of Medicine, 676 N. St. Clair St., Suite 730, Chicago, IL 60611, Phone: 312-926-8294; Fax: 312-253-4470, [email protected]
27 3 2022
4 2022
04 3 2022
01 4 2023
18 2 287300
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Heart failure
artificial intelligence
machine learning
deep learning
natural language processing
pmcIntroduction
Heart failure (HF) is a common condition with a profound impact on patients and the health care system. Heart failure with preserved ejection fraction (HFpEF) comprises nearly half of all cases of HF, affects over 3 million US adults, and is underdiagnosed.1 Multiple risk factors and conditions, such diabetes, hypertension, obesity, metabolic syndrome, and aging, contribute to inflammation and endothelial dysfunction and ultimately leads to the multiorgan, systemic syndrome of HFpEF in many patients.2 Others can develop the syndrome of HFpEF for other reasons, including due to genetic variants (e.g., hypertrophic cardiomyopathy), infiltration of the myocardium (such as cardiac amyloidosis), or other toxic-metabolic myocardial insults. Regarding of its etiology, HFpEF remains a highly morbid condition that negatively impacts quality of life, is marked by frequent hospitalizations and high mortality rate, and has few therapeutic options despite numerous clinical trials testing various medications and devices. Machine learning (ML) has the potential to improve diagnosis and treatment of patients with HFpEF, although its impact thus far has been limited.
ML, a domain that arose from the fields of statistics and computer science, focuses on teaching computers to learn from data, interpret it, and make predictions. It enables internet search, speech recognition, image identification, and human-computer interactions by learning from large datasets.3–5 ML is not a new field but has been gaining considerable attention within the health care and cardiovascular research communities over the past several years. The reasons for this trend are likely multifactorial, in part due to improved processing power and the availability of large datasets with a large number of variables (features).6,7 Indeed, the inability to interpret the increasingly high density of data coming from diverse sources in the clinical setting is one of the key reasons why ML may be uniquely poised to help clinicians and researchers avoid a loss of potentially valuable information that could improve clinical decision-making and patient care.
HFpEF is a prototypical cardiovascular condition that may benefit from ML because of its inherent heterogeneity, the need for improved classification, and the challenges in HFpEF diagnosis and treatment.8 Here we first briefly highlight several unmet needs within the HFpEF field that may benefit from ML approaches. Next, we provide an overview of key types of ML and concepts within the field. We describe several challenges and pitfalls of ML and provide a roadmap for the evaluation of ML studies. Finally, we discuss future directions of ML in HFpEF and the broader field of precision cardiovascular medicine.
The unmet need for better approaches to heart failure diagnosis and management
HFpEF is a heterogenous condition with multiple pathways that lead to its development. Identifying less common etiologies of HFpEF, such as cardiac amyloidosis, hypertrophic cardiomyopathy, and cardiac sarcoidosis, can be challenging. Earlier diagnosis of cardiomyopathies with specific treatment options may lead to improved quality of life and survival. For example, cardiac amyloidosis is likely far more common than previously realized and may comprise up to 5–7% of HFpEF cases.9,10 Prior studies have revealed the presence of early clinical clues of cardiac amyloidosis, including symptoms, laboratory abnormalities, and changes on echocardiograms and electrocardiograms (ECG).11–13 Identifying early signs is essential given the emergence of novel therapies that can halt cardiac amyloidosis disease progression and improve prognosis.14,15
The majority of patients with HFpEF develop the condition in the setting of one of more risk factors and conditions—such as diabetes, hypertension, obesity, metabolic syndrome, sedentary lifestyle, chronic kidney disease, and ischemic heart disease. Numerous clinical trials in patients with HFpEF have failed to show a difference in their primary outcome.2,16 This may be in part due to the heterogeneity of patients with HFpEF and the need for better sub-classification. ML can be used to uncover patterns in diverse data and identify subgroups of patients with distinct pathophysiologic profiles and potential differential responses to therapy. ML approaches are designed to leverage the vast amounts of available data and account for higher-order interactions, multi-dimensionality, and non-linear effects, and may complement conventional statistical methods.17 ML algorithms can be applied to either data from HFpEF observational cohorts for identifying novel subgroups or predictors of adverse outcomes or data from HFpEF clinical trial participants for hypothesis-generating, post-hoc analyses to identify subgroups of patients with HF who benefited from the intervention. While there is some debate about whether responder analyses are possible in clinical trials, it is possible that lack of that targeting may explain why numerous clinical trials in HFpEF have failed to show a difference in their primary outcome.
Identifying appropriate patients and enrolling in HF clinical trials has become increasingly challenging, particularly in the United States (US). Despite the high prevalence of HF, in North America and Western Europe the typical clinical trial enrollment rate is a dismal 1–2 patients per year per site in HF trials.18 For example, in the Treatment of Preserved Cardiac Function Heart Failure with an Aldosterone Antagonist (TOPCAT) trial, which randomized HFpEF patients to spironolactone vs. placebo, the average US enrollment rate was only 1.4 patients per site per year even though HFpEF is a leading cause of hospitalization in the US. This problem could be addressed by better identification of eligible patients though applying ML to EHR data.
A substantial literature exists on risk prediction for hospitalization and mortality in patients with HFpEF. Although the majority of risk models for patients with HFpEF were developed in cohorts with all types of HF, there are a few models developed specifically for patients with HFpEF (such as the I-PRESERVE Score and ARIC Score).19 The majority of models have modest discrimination and calibration at best.19 ML has to potential to lead to the development of either higher performing models using more diverse data sources (i.e., EHR, imaging, ECG, sensors) or a more parsimonious set of variables that will facilitate deployment with EHR systems. For example, the Machine Learning Assessment of RisK and EaRly mortality in Heart Failure (MARKER-HF) risk model uses eight laboratory variables to predict 90-day and 1-year mortality and with similar performance to other risk models that require more data, much of which are not readily available in the EHR.20,21 Thus, ML has the potential to improve prediction for patients with HFpEF, though trials studying the implementation of these algorithms as part of routine care will be needed prior to wider adoption.
Overview of machine learning techniques
Artificial intelligence (AI) broadly refers to computer systems designed to perform tasks that usually require human intelligence (Figure 1). ML enables AI computer systems through learning from data without explicit programming. ML algorithms can be broadly categorized, depending on the nature of the data that are used in the training process. Supervised learning algorithms (Figure 2) are given labeled training data, so that each training instance consists of features (e.g. age, gender, blood pressure, left ventricular ejection fraction) and a label (e.g., HF diagnosis, mortality, HF hospitalization,) indicating the class to which the instance belongs.4 The algorithms then learn to predict the outcome or label based on the aforementioned features, with the goal of making accurate predictions in new patients. Unsupervised learning algorithms (Figure 2) are given unlabeled data and aim to discover intrinsic patterns in the data (e.g., sub-populations such as sub-phenotypes, or clusters). A hybrid of these two approaches is called semi-supervised learning, which uses a combination of labeled and unlabeled data.22 Semi-supervised algorithms can be particularly useful in building predictive models when limited amounts of labeled data are available. Lastly, reinforcement learning a form of ML that seeks to find an optimal set of sequential actions in a prespecified environment/domain to maximize a defined reward or goal.23 Reinforcement learning—which has been used to successfully teach computers how to play and win games (by sequentially playing the game and learning from mistakes)—can also be applied to the healthcare setting (e.g., if we want to ask how the sequence and timing of interventions affects outcomes).
Types of machine learning approaches
Supervised machine learning
In supervised learning, algorithms are developed using labeled training data (Figure 2). This means that for each item in the training set, we know the features (variables) and the outcome to be estimated. For example, several studies have attempted to use supervised ML to predict re-hospitalization in HF patients. In these studies, the computer is told who was re-hospitalized and who was not, and the algorithm may estimate for each participant whether he or she will be re-hospitalized based on predefined features. There are several different supervised ML techniques, and each has its relative strengths and weakness Example of these approaches include linear and logistic regression, support vector machines,24 and random forests.25
Unsupervised machine learning
Unsupervised learning differs from supervised learning in that the algorithms are developed based on intrinsic patterns in the data (i.e., the data are unlabeled). Unsupervised ML can be used to identify subgroups (i.e., sub-phenotypes) within a heterogeneous clinical syndrome. For example, suppose we wanted to identify subgroups within the clinical syndrome of HFpEF with the goal of designing a trial to test the effect of a targeted therapy on reducing the risk of HF hospitalization. We can cast this as an unsupervised learning problem in which the training dataset does not come with pre-assigned labels (i.e., “outcomes”). Unsupervised learning algorithms evaluate intrinsic relationships and patterns in the data to identify subgroups and reduce dimensionality, and typically include clustering, principal components analysis (dimensionality reduction), association rules (pattern mining), or factorization algorithms (Figure 2). In healthcare, clustering is one of the most popular approaches.
Clustering creates groups of similar patients. Clustering algorithms have multiple variations, among which the most popular ones include k-means clustering, hierarchical clustering, and model-based clustering. The algorithms generally measure the distance between individual features to identify clusters, and there are several different methods for determining the optimal number of clusters.26,27 Additional details on different types of clustering are included in Supplemental Table 1. Importantly, clustering is a way of dividing the population into groups so that patients are different across groups and similar within each group.
Beyond clustering, there are many other unsupervised learning techniques. For example, if we have a population of patients with HFpEF, we may expect a sub-group of patients to vary along one dimension (e.g. general health status) that is reflected in many different aspects of their medical history (and thus many different features within the dataset). In this case dimensionality reduction methods like principal component analysis or factor analysis28 are often used.
Deep learning, natural language processing, and additional machine learning techniques
Deep learning
Deep learning, which is an extension of neural networks, is a popular approach that attempts to overcome a major disadvantage of conventional ML techniques: their inability to effectively interpret and process high-dimensional raw signal data, such as audio and images.5,29 As a result of this limitation, attempts to apply ML to echocardiographic data have mostly relied upon measured and annotated quantitative measurements by highly experienced echocardiographers.30 Instead, deep learning can often learn from raw data (such as the pixel values from an image or raw text from the EHR) using a hierarchy of increasingly complex and discriminative layers of processing. A key benefit of deep learning is its ability to learn features directly from data without the need for painstaking feature annotation by domain experts, a process known as feature engineering. Deep learning applications can have varying architectures, each with different strengths and limitations. These include these include convolutional neural networks, recurrent neural networks, autoencoders, and generative adversarial network.5
Deep learning is a technique that can be applied for very large datasets with either structured or unstructured data (e.g., imaging or ECG repositories). Deep learning can also be applied to a smaller dataset (e.g. medical images), if there exists a similar large dataset (e.g., an archive of conventional photographic images) on which the analytical model can be pre-trained in an approach called transfer learning.31
Natural language processing (NLP)
Natural language processing (NLP) applies computational algorithms to the problem of recovering structure from text and speech. These algorithms can be based on hand-coded rules, machine-learned models, or (commonly) a blend of both approaches. In the clinical domain, NLP can be used to extract clinically relevant information from a large number of clinical notes that are commonly found in the EHR.32,33 In fact, deep learning algorithms have been applied with great success to many NLP tasks.34 At the lowest level, NLP algorithms are used to segment text into words (or “tokens”), which can be further analyzed for part of speech (e.g., noun, verb, adjective), and other lexical features. Other NLP algorithms are then applied to recover higher-level linguistic structure, such as sentences, phrases, and sentential parses. These structures can subsequently be used as features for recognizing and extracting semantically meaningful chunks of information from text, such as entities that are mentioned and relations between these entities. In the clinical domain, such entities can be diseases, drugs, symptoms, and medications. They could also be events and their temporal properties, such as an adverse reaction occurring after treatment was performed or a drug prescribed.35,36
Transformers
Transformers are a relatively novel class of deep learning methods for analyzing sequence data based on self-attention that have been leveraged with remarkable success for complex NLP tasks.37 For example, researchers at Google (Mountain View, CA)38 developed Bidirectional Encoder Representations from Transformers (BERT), and multiple groups have extended this work by pre-training BERT models on corpuses of biomedical and clinical text and then testing those models on biomedical and health care tasks, including readmission prediction.39–42 OpenAI (San Francisco, CA), one of the leaders in developing generalized AI models, created the Generative Pre-trained Transformer 3 (GPT-3),43, which is a highly sophisticated language model that is able to generate text and conversations mimicking humans, though its adoption into healthcare settings will require additional research and thoughtful exploration of its potential and pitfalls.44
In the realm of clinical medicine, NLP can be an effective method to derive structured data from the EHR due to the large amount of unstructured, free text entries in hospital and clinic notes, discharge summaries, imaging reports, procedure reports, and the like. In HFpEF, NLP can serve a variety of purposes such as detecting HFpEF and cardiomyopathies to assist with earlier diagnosis or providing data for use in developing strategies to reduce hospital readmissions in patients with HFpEF.45–49 Leveraging NLP algorithms could have a significant impact on HFpEF clinical trials. For example, NLP-assisted identification of suitable clinical trial patients could save time and money as it often takes study coordinators large amounts of time to screen single patients for HF trials, which often have complex inclusion and exclusion criteria. Applied in this manner, NLP could reduce the time needed for HF clinical trial screening from the EHR from several days (when done by humans) to just a few minutes when done by a computer.
When designing ML systems, the choice of the method—which depends on several factors including the size of the dataset, the number and informativeness of the features, and the ultimate research question or goal—is central.
Feature Selection, Hyperparameters, Overfitting, and Regularization
Feature selection is one of the most critical aspects of ML and depends on the task. For image classification, having large, labeled, and balanced datasets is key to successful model development. For other tasks, such as subtyping HFpEF with clinical data, selecting features that are informative and orthogonal increasing the likelihood of identifying meaningful clusters.
Aside from feature selection, there are design choices that must be made by the ML scientist when training a ML model. These model parameters that are explicitly set to control the learning process are called “hyperparameters” (e.g., model algorithm/architecture and size, learning rate, number of clusters in unsupervised learning). Theoretically, these hyperparameters could be tuned such that a given model would fit the training data perfectly. However, such a model would be unlikely to generalize well to new, unseen datasets because it would have modelled noise inherent in the training dataset in addition to the true relationship between data inputs and outputs/labels. This is an example of overfitting and is a consequence of the concept of the bias-variance tradeoff.
A model that has overfit to training data but does not generalize well to external datasets would be said to have low bias and high variance – bias is an expression of the error of the model in approximating a complex function that governs real-life processes, while variance is an expression of the sensitivity of model performance to subtle changes from one dataset to the next. Conversely, a model that is underfit is said to have high bias and low variance, likely because it is too simplistic to approximate the true underlying relationship between data inputs and outputs. Such a model would perform poorly on the given task, but performance would be robust to subtle changes from one dataset to the next. The goal in training any model is to attempt to resolve this trade-off and minimize both bias and variance as much as possible.
There are several strategies to resolve this trade-off on a given dataset including acquiring additional training data, adjusting the complexity of the model, and methods such as model ensembling, transfer learning, and regularization. Regularization refers to a family of methods that impose some sort of penalty or constraint on the model to prevent overly confident predictions and overfitting on the training data. Examples include penalized regression, data augmentation, dropout, and early stopping of the training process. The idea of regularization is foundational for most ML techniques.50
Train-Test Paradigm
During the training process, it is crucial that the ML scientist monitor for overfitting. This is done via a validation dataset. Validation datasets can be a hold-out partition of the training set or alternatively when data size is limited can take the form of k-fold cross-validation, where the training set is iteratively partitioned into k number of subsets, with one subset being used as the validation dataset during each iteration. It is important to note that this is distinct from the concept of “validation datasets” used to validate a given model in the risk prediction literature. In contrast, validation datasets in ML are strictly used during model development as a litmus test to monitor for overfitting (evident when performance on the training set significantly exceeds performance on the validation set). For this reason, the “validation dataset” is also often referred to as a “development dataset” in the ML literature. A validation dataset does nothing to “validate” a model given information leak can occur during the training process while adjusting hyperparameters and one can begin to overfit to the validation dataset itself. While cross-validation can help mitigate this somewhat, there is no substitute for a separate hold-out test dataset. In order to accurately assess a given model’s performance, the model must be evaluated on a test set that the model was not exposed to during the training process – this can either be a distinct hold-out partition from the original dataset or, ideally, a completely separate, external dataset. The careful choice of a testing strategy is essential for any ML approach, and ideally should be developed during the planning stages of comprehensive ML projects.
Case studies in machine learning
Detection of rare causes of heart failure with preserved ejection fraction
The widespread adoption of EHRs and the availability of repositories of digitized cardiovascular diagnostic testing, such as echocardiograms, cardiac MRIs, and ECGs, have enabled the development of algorithms for a range of tasks, including automation of measurements (i.e. left ventricular ejection fraction, diastolic dysfunction), enhancing image quality, disease diagnosis, and risk prediction for disease development or prognosis.51–53 In HFpEF, underlying etiologies remain underdiagnosed, especially for rare causes, such as cardiac amyloidosis.
Several studies have examined the use different types of data to identify patients with specific etiologies of HFpEF. In 2018, Zhang et al. used convolutional neural networks to train a model to accurately identify different echocardiographic views and predict specific diseases related to HFpEF (hypertrophic cardiomyopathy, cardiac amyloidosis, and pulmonary arterial hypertension) with good discrimination (C-statistics=0.85–0.93).54 Tison et al. developed a deep learning model for ECG data to predict cardiac structure and function as well as specific diseases with good discrimination for hypertrophic cardiomyopathy, cardiac amyloidosis, and pulmonary arterial hypertension (C-statistics=0.86–0.94).55 Grant et al. developed an end-to-end pipeline to automatically quantify left ventricular hypertrophy and predict its etiology with an area under the curve of 0.83 for cardiac amyloidosis and 0.98 for hypertrophic cardiomyopathy.56 Using ECG and echocardiographic data to identify those with cardiac amyloidosis, Goto et al. demonstrated that the stepwise use of deep learning initially ECG data to identify those at high risk and then on echocardiographic data for those increased the positive predictive value from 33% to 74–77% in cohorts from two institutions.57 Huda et al. trained a ML model using ICD codes to identify patients with wild-type transthyretin cardiac amyloidosis and found good to excellent performance in four external validation cohorts, including a large, integrated health system.58
Taken together, these studies demonstrate the promise of using ML to identify rare, subtypes of HFpEF on different types of data. Future areas of research HFpEF etiology detection includes exploring the use of state-of-the-art NLP models, incorporating longitudinal changes in diverse data types, and fusing multiple data types into a single model. Lastly, although the above studies have shown promising, “in silico” results, implementation studies are needed to understand whether these technologies can be deployed to impact the care and outcomes with patients with HFpEF.
Phenomapping of heart failure with preserved ejection fraction
Shah et al.30 hypothesized that the application of ML techniques to dense phenotypic data (“phenomapping”) would yield a novel classification system for HFpEF, and that the identified “pheno-groups” would have unique pathophysiological profiles and differential outcomes. Phenotypic features utilized in the study included clinical variables, physical characteristics, laboratory data, ECG parameters, and echocardiographic variables. Figure 3 details the key steps of the study. The investigators identified 3 distinct classes (pheno-groups) of HFpEF patients with differing clinical characteristics and profiles and differential rates of cardiovascular hospitalization and death. The 3 clusters were replicated in an independent, prospective validation cohort. This approach highlights the importance of having an a priori hypothesis; using high quality, quantitative data; and validating in a separate cohort.
Several other studies of applied similar phenomapping techniques to patients with HFpEF and are reviewed in detail along with HFpEF therapeutic implications by Galli et al.59 The overlap in similar phenotypes across multiple cohorts of patients using different unsupervised learning approaches suggests that certain pheno-groups may represent a distinct pathophysiological profile and have a differential response to therapeutics. A recent study by Pandey et al. used deep-learning models with a limited number of echocardiographic measurements of systolic and diastolic function to characterize the severity of diastolic dysfunction and identify subgroups with differential risks of adverse events and potential response to spironolactone.60 For this study and other similar studies, the identification of pheno-groups should represent the starting point for future investigations, such as mechanistic studies of the underlying pathogenies of HFpEF or clinical trials testing targeted interventions. Moreover, future studies will ideally move beyond applying ML algorithms to selected groups of echocardiographic measurements to using ML to analyze the echocardiographic images themselves, along with other types of unstructured data, such as ECGs and clinical notes, in combination with other structured demographic and clinical data. Advantages of these approaches are (1) increased throughput, reproducibility, and scalability (due to the lack of needing human echocardiographic measurements), and (2) use of multi-modal datasets, which increase likelihood of orthogonal (and therefore more informative) input features into ML models.
Discussion
We have highlighted some of the unmet needs in HFpEF and ways in which ML may be used to address these challenges, include rare etiology detection, identification of subtypes with different outcomes and potentially differential response to therapeutics, and increased efficiency for clinical trial recruitment through using NLP. Although studies using ML are becoming increasingly common in the healthcare field, as with any newly applied methodology, a healthy amount of skepticism is warranted. It is unlikely that ML alone will be able to solve the problem of disentangling the heterogeneity of and enhancing therapeutic targeting in HFpEF. For example, one could argue that even more important than any analytical technique is the availability of orthogonal (i.e., uncorrelated) features that provide a more comprehensive viewpoint of the patients and outcomes under investigation. Furthermore, the “hype” of any new (or newly applied) analytical technique can often lead to errors in its application and interpretation, as detailed below. Finally, ML could fall prey to the lifecycle of several other novel diagnostic and analytical techniques in medicine.
Several groups have published or are in the process of developing guidelines for the design, reporting, and evaluation of ML studies in health care.61–67 Below, we offer complementary recommendations to help guide future work in applying ML to HFpEF investigations (Table 1).
We should not apply ML algorithms to every large dataset or research question; rather, we must think critically about how ML would be better than a traditional approach for the specific research question at hand. Developing a priori hypotheses and analytical plans, informed by prior research and clinical insights (from clinical domain experts), may be helpful to reduce spurious findings. For many research questions, a conventional statistical approach or study design will be sufficient. ML cannot overcome threats to validity that occur with subgroup analyses in smaller data sets or traditional, post-hoc clinical trial analyses. Findings from these analyses, although potentially important, are hypothesis-generating and need to be viewed with the same level of caution as other observational studies.
Bigger data does not necessarily imply better data. First, if many of the variables included in the dataset are highly correlated with each other, the addition of all these variables to the analysis is unlikely to be helpful. Orthogonal data types (i.e., data that are not highly correlated with each other and yet meaningful to the disease or outcome of interest) are essential. Second, the size of the dataset cannot overcome poor data quality or bias introduced from the data or study design. For example, data from EHRs and other clinical systems are often of poor quality, incomplete, and biased.68,69 Although longitudinal and/or multiple imputation can be used,70,71 the missing data are likely not random. Any imputation of data or exclusion of participants with incomplete data will likely introduce bias into the results. Lastly, ML has the potential to detect and/or perpetuate structural racism, and there is a growing body of research on ways to ensure that ML, especially when applied in health care settings, increases equity.
Concurrent with the continued growth in the amount of data, the systems to collect, store, and organize data—and ML methods themselves—are rapidly evolving. For example, as discussed earlier, deep learning has dramatically evolved since the mid-2000s, leading to significant improvements in speech and visual object recognition.5,29 Incorporating the latest advances in data management and ML into existing and future studies will be critical to finding meaningful application of ML to HFpEF. Large-scale HF clinical trials should be designed with ML in mind such that dense phenotypic, -omic, and physical activity (e.g., accelerometer) data are collected and training and validation subsets should be outlined in advance.
For both unsupervised and supervised ML approaches, validation in an external test dataset, similar to traditional risk prediction models,72,73 is critically important to establish the accuracy of the model. Given the fact that the ideal datasets for ML will be unique due to the requirement for large amounts of variables from multiple domains, validation for unsupervised learning analyses will not always be immediately possible due to the lack of an appropriate test dataset. However, any ML study without a test cohort should be viewed as hypothesis-generating only, with a requirement for future validation prior to further research of identified sub-groups or integration into clinical practice.
In the case of supervised learning, validation is a requirement given the need to test the accuracy of the algorithm.74 For supervised models, measures of discrimination and calibration should therefore be reported. Also, whenever using a complex ML algorithm, the results should be compared to those of a simpler algorithm.
In general, care must be taken in the evaluation of the quality and promise of ML algorithms. For models intended to inform clinical practice, the development of a new classification system or prediction model is not sufficient to change practice. These models must be deployed clinically, with formal testing of their performance and clinical impact; not surprisingly, the implementation of risk tools into clinical practice may not change clinical practice,75 and additional studies for the use of ML models in clinical practice are needed, including hybrid-effectiveness studies.76 “Dataset shift” is a mismatch between the data on which the model is deployed and on which it was trained and can occur for numerous reasons, including changes in technology, the patient population, clinician behavior, or clinical workflows.77 This mismatch can lead to change in model performance and highlights the need for a health system governance infrastructure to monitor and address model performance and impact over time.77,78
Future advancements in HFpEF precision medicine may require a combination of the ML techniques. Difference ML approaches, such as NLP, deep learning, unsupervised ML, and supervised ML, could be used for specific tasks as part of a larger framework in order to efficiently identify novel approaches to the classification and treatment of HFpEF in clinical trials (Figure 4).
Conclusions
HFpEF is a morbid and costly clinical syndrome with significant heterogeneity. ML is an exciting tool that may help address some of the major challenges in HFpEF and in cardiovascular medicine. However, as with the application of any new technique, methodical and thoughtful application of ML to specific, hypothesis-driven questions will be essential if it is to make a lasting impact on the field of HFpEF.
Supplementary Material
1
Acknowledgments
We would like to thank Mark Berendsen, MLIS, and Linda O’Dwyer, MA, MSLIS, AHIP, for their assistance with the literature review for this paper.
Funding Sources
Dr. Ahmad was supported by grants from the Agency for Healthcare Research and Quality (K12HS026385), National Institutes of Health/National Heart, Lung, and Blood Institute (K23HL155970) and the American Heart Association (AHA number 856917). Dr. Luo was supported by grants from National Institutes of Health (U01TR003528, 1R01LM013337). Dr. Thomas was supported by a grant from the Irene D. Pritzker Foundation. The statements presented in this work are solely the responsibility of the author(s) and do not necessarily represent the official views of the Patient-Centered Outcomes Research Institute® (PCORI®), the PCORI® Board of Governors or Methodology Committee, the Agency for Healthcare Research and Quality, the National Institutes of Health, or the American Heart Association.
Disclosure Statement:
FA receives consulting fees from Amgen. Pfizer, and Livongo Teladoc outside of this work. SJS has received research grants from the National Institutes of Health (R01 HL107577, R01 HL127028, R01 HL140731, R01 HL149423), Actelion, AstraZeneca, Corvia, Novartis, and Pfizer; and has received consulting fees from Abbott, Actelion, AstraZeneca, Amgen, Aria CV, Axon Therapies, Bayer, Boehringer-Ingelheim, Boston Scientific, Bristol-Myers Squibb, Cardiora, CVRx, Cytokinetics, Edwards Lifesciences, Eidos, Eisai, Imara, Impulse Dynamics, Intellia, Ionis, Ironwood, Lilly, Merck, MyoKardia, Novartis, Novo Nordisk, Pfizer, Prothena, Regeneron, Rivus, Sanofi, Shifamed, Tenax, Tenaya, and United Therapeutics. JT receives consulting fees from Edwards, Abbott, GE, and Caption Health and reports spouse employment with Caption Health
Figure 1. Artificial Intelligence and Machine Learning
Schema of the relationship between artificial intelligence, machine learning, and “deep” learning. SVM = Support vector machines, RF = random forests, KNN = k-nearest neighbor, PCA = principal components analysis, DNN = deep neural network, CNN = convolutional neural network, RNN = recurrent neural network, GAN = generative adversarial network, VAE = variational autoencoder.
From Wehbe RM, Khan SS, Shah SJ, Ahmad FS. Predicting High-Risk Patients and High-Risk Outcomes in Heart Failure. Heart Fail Clin. 2020 Oct;16(4):387–407.
Figure 2. Types of Machine Learning
In supervised machine learning, the outcomes (labels) are provided so that the machine learning algorithm can be trained to identify features that can successfully predict these outcomes (or labels) in external datasets. In the example provided, deep phenotypic data from the active treatment arm of a HFpEF clinical trial could be analyzed using a statistical learning approach in order to identify features (“signature”) of treatment response. The resultant classifier could then be used to determine whether it could successfully predict treatment responders in a similar previously completed clinical trial (post-hoc analysis) or in a new prospective trial of “all comers.” Alternatively, patients could be screened by the classifier and only enrolled in a new, prospective, targeted treatment trial if they met criteria for being a “treatment responder.” Unsupervised machine learning can be applied to a heterogeneous clinical syndrome such as HFpEF in order to identify homogeneous sub-groups of patients by looking for intrinsic patterns within data from these patients. The identified sub-groups can then be examined to determine whether they have differential outcomes or responses to therapies. See Supplementary Table 1 for definitions of the examples of machine learning algorithms shown in the figure. HFpEF = heart failure with preserved ejection fraction; RCT = randomized controlled trial.
Figure 3. Example of an Unsupervised Machine Learning Workflow for the Identification of Novel Heart Failure with Preserved Ejection Fraction (HFpEF) Subtypes
(A) The workflow for unsupervised machine learning begins with data processing, including examination of missing values, multiple imputation, transformation, scaling, and correlation testing, to understand the relationship between the features (variables) to be selected for the analysis. Shown here is an example of a heatmap displaying the correlation between laboratory, electrocardiographic, and echocardiographic features that were used in phenomapping of HFpEF (the color key corresponds to the range of correlation coefficients shown in the heatmap). The point of this type of analysis is to remove highly correlated features prior to the unsupervised learning analyses. (B) Next, various types of unsupervised machine learning analyses can be used to identify clusters within the data. In the displayed example, agglomerative hierarchical clustering was used to create an initial heatmap to demonstrate the heterogeneity of the HFpEF patients and the presence of potential clusters. (C) Concomitant with unsupervised learning analyses is the need to determine the optimal number of clusters. In the example shown here, model-based clustering was used on the HFpEF dataset, and the optimal number of clusters was determined using the Bayesian information criterion (BIC) analyses. The lowest BIC value corresponds to 3 clusters, which is the optimal number in this example. (D) Once the optimal number and composition of the clusters is identified, differences in clinical characteristics, including outcomes, can be compared among the clusters. In the example shown here, the 3 HFpEF clusters differed significantly in survival free of cardiovascular hospitalization or death, and represented 3 distinct clinical profiles of HFpEF. In the example shown here, pheno-group #1 was characterized by younger patients with relatively normal B-type natriuretic peptide and moderate diastolic dysfunction; pheno-group #2 included obese, diabetic patients with the worst left ventricular relaxation and high prevalence of obstructive sleep apnea; and pheno-group #3 was composed of older patients with significant right ventricular dysfunction, pulmonary hypertension, chronic kidney disease, and electrical and myocardial remodeling.
Adapted from Shah SJ, Katz DH, Selvaraj S, Burke MA, Yancy CW, Gheorghiade M, Bonow RO, Huang CC, Deo RC. Phenomapping for novel classification of heart failure with preserved ejection fraction. Circulation. 2015 Jan 20;131(3):269–79.
Figure 4. Conceptual Framework for the Integration of Machine Learning Approaches for Novel Classification and Treatment of Heterogeneous Clinical Syndromes such as Heart Failure
In this precision medicine conceptual framework, four areas of machine learning discussed here (unsupervised learning, supervised learning, deep learning, and natural language processing) are used in combination to classify and provide targeted treatment for a heterogeneous clinical syndrome such as heart failure. NLP = natural language processing.
Table 1. Evaluative Framework for Machine Learning Studies
Categories Evaluation Criteria
Study question and design ✓ Does machine learning offer specific advantages over other statistical modeling approaches?
✓ If yes, then why? Potential criteria include the data source, feature learning, outcomes, or combination of these.
Data ✓ Are data being collected primarily for (1) research or (2) clinical/administrative purposes?
✓ Are the issues of biases and data quality (i.e. completeness, heterogeneity, density, accuracy, and representativeness) described and addressed?
Approach ✓ Were the reasons for the selected approaches specified (i.e. supervised vs unsupervised vs semi-supervised vs reinforcement learning, model selection)?
✓ If complex models are used, are we sure simpler models would not do better? Was the model chosen a priori?
✓ Is the internal validation convincing?
✓ Was the testing dataset separated prior to model training and was there an external validation?
✓ Is the causality convincing or could features used for prediction have been produced by the outcome?
✓ Is performance properly quantified both for internal and external validation?
✓ Was the model performance compared to other models addressing the same clinical question?
Clinical relevance and generalizability ✓ Do the results have clinical relevance or provide mechanistic insight into the pathophysiology of the clinical syndrome of interest?
✓ How well should we expect the study population to generalize to the target population?
Synopsis:
Heart failure (HF) with preserved ejection fraction (HFpEF) represents a prototypical cardiovascular condition in which machine learning may improve targeted therapies and mechanistic understanding of pathogenesis. Machine learning, which involves algorithms that learn from data, has the potential to guide precision medicine approaches for complex clinical syndromes such as HFpEF. It is therefore important to understand the potential utility and common pitfalls of machine learning so that it can be applied and interpreted appropriately. Unmet needs in HFpEF that could benefit from machine learning include: (1) earlier diagnosis of HFpEF and its etiology; (2) improved classification (sub-phenotyping) of HF; (3) improved management of HFpEF through prediction of therapeutic responsiveness and HF hospitalization; and (4) improved identification of HFpEF patients for clinical trials. There are two main settings of machine learning, supervised learning and unsupervised learning. In supervised learning, predefined outcomes or labels are used to develop models to help predict future outcomes. In unsupervised learning, algorithms are applied to unlabeled data to understand intrinsic patterns within the data (e.g., for novel classification of HF subtypes). These techniques can be applied to many different data types, including clinical characteristics, cardiovascular diagnostic testing (i.e., imaging, labs, electrocardiograms, sensors), and clinical notes. While machine learning holds considerable promise for HFpEF, it is subject to several potential pitfalls (e.g., bias, confounding, overfitting, and lack of external validation), which are important factors to consider when interpreting machine learning studies.
Key Points:
Heart failure with preserved ejection fraction is a heterogenous, morbid condition with several unmet needs.
Machine learning has the potential to guide precision medicine approaches for heart failure with preserved ejection fraction, such as identification of rare etiologies, sub-phenotyping, and increasing the efficiency of clinical trial enrollment.
Understanding the strengths, limitations, and pitfalls of machine learning approaches is critical to realizing the potential of machine learning to impact the health of patient with heart failure with preserved ejection fraction.
Clinical Care Points:
Machine learning has promise to improve identifying and treating different subtypes of patients with HFpEF but also pitfalls.
Machine learning approaches may ultimately contribute improvements in care for patients with HFpEF, but applications must be rigorously developed and tested prior to implementation into clinical care.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
REFERENCES
1. Virani SS , Alonso A , Aparicio HJ , Heart Disease and Stroke Statistics-2021 Update: A Report From the American Heart Association. Circulation. 2021;0 (0 ):CIR0000000000000950.
2. Shah SJ , Kitzman DW , Borlaug BA , Phenotype-Specific Treatment of Heart Failure With Preserved Ejection Fraction: A Multiorgan Roadmap. Circulation. 2016;134 (1 ):73–90.27358439
3. Murphy KP . Machine Learning. MIT Press; 2012.
4. Bishop CM . Pattern Recognition and Machine Learning. Springer; 2006.
5. Goodfellow I , Bengio Y , Courville A . Deep Learning. Cambridge, MA: MIT Press; 2016.
6. Obermeyer Z , Emanuel EJ . Predicting the Future - Big Data, Machine Learning, and Clinical Medicine. N Engl J Med. 2016;375 (13 ):1216–1219.27682033
7. Altman RB , Ashley EA . Using “big data” to dissect clinical heterogeneity. Circulation. 2015;131 (3 ):232–233.25601948
8. Shah SJ . Precision Medicine for Heart Failure with Preserved Ejection Fraction: An Overview. J Cardiovasc Transl Res. 2017;10 (3 ):233–244.28585183
9. Kittleson MM , Maurer MS , Ambardekar AV , Cardiac Amyloidosis: Evolving Diagnosis and Management: A Scientific Statement From the American Heart Association. Circulation. 2020;142 (1 ):e7–e22.32476490
10. Kazi DS , Bellows BK , Baron SJ , Cost-Effectiveness of Tafamidis Therapy for Transthyretin Amyloid Cardiomyopathy. Circulation. 2020;141 (15 ):1214–1224.32078382
11. Aus dem Siepen F , Hein S , Prestel S , Carpal tunnel syndrome and spinal canal stenosis: harbingers of transthyretin amyloid cardiomyopathy? Clin Res Cardiol. 2019;108 (12 ):1324–1330.30953182
12. Falk RH , Alexander KM , Liao R , Dorbala S . AL (Light-Chain) Cardiac Amyloidosis: A Review of Diagnosis and Therapy. J Am Coll Cardiol. 2016;68 (12 ):1323–1341.27634125
13. Geller HI , Singh A , Alexander KM , Mirto TM , Falk RH . Association Between Ruptured Distal Biceps Tendon and Wild-Type Transthyretin Cardiac Amyloidosis. JAMA. 2017;318 (10 ):962–963.28898370
14. Donnelly JP , Hanna M . Cardiac amyloidosis: An update on diagnosis and treatment. Cleve Clin J Med. 2017;84 (12 Suppl 3 ):12–26.29257735
15. Wechalekar AD , Gillmore JD , Hawkins PN . Systemic amyloidosis. Lancet. 2016;387 (10038 ):2641–2654.26719234
16. Butler J , Fonarow GC , Zile MR , Developing therapies for heart failure with preserved ejection fraction: current state and future directions. JACC Heart Fail. 2014;2 (2 ):97–112.24720916
17. Goldstein BA , Navar AM , Carter RE . Moving beyond regression techniques in cardiovascular risk prediction: applying machine learning to address analytic challenges. Eur Heart J. 2016.
18. Gheorghiade M , Vaduganathan M , Greene SJ , Site selection in global clinical trials in patients hospitalized for heart failure: perceived problems and potential solutions. Heart Fail Rev. 2014;19 (2 ):135–152.23099992
19. Wehbe RM , Khan SS , Shah SJ , Ahmad FS . Predicting High-Risk Patients and High-Risk Outcomes in Heart Failure. Heart Fail Clin. 2020;16 (4 ):387–407.32888635
20. Adler ED , Voors AA , Klein L , Improving risk prediction in heart failure using machine learning. European Journal of Heart Failure. 2020;22 (1 ):139–147.31721391
21. Greenberg B , Adler E , Campagnari C , Yagil A . A machine learning risk score predicts mortality across the spectrum of left ventricular ejection fraction. European Journal of Heart Failure. 2021;23 (6 ):995–999.33724626
22. Zhou X , Belkin M . Chapter 22 - Semi-Supervised Learning. In: Diniz PSR , Suykens JAK , Chellappa R , Theodoridis S , eds. Academic Press Library in Signal Processing. Vol 1 . Elsevier; 2014:1239–1269.
23. Sutton RS , Barto AG . Reinforcement Learning: An Introduction, 2nd edition. Cambridge, Massachusetts: MIT Press; 2017.
24. Vapnik V , Golowich SE , Smola A . Support vector method for function approximation, regression estimation, and signal processing. Advances in Neural Information Processing Systems 9. 1997;9 :281–287.
25. Liaw A , Wiener M . Classification and Regression by randomForest. R News. 2002;2/3 :18–22.
26. Hartigan JA , Wong MA . A k-means clustering algorithm. Appl Stat. 1979;28 :100–108.
27. Ng A , Jordan M , Weiss Y . On spectral clustering: analysis and an algorithm. In: Dietterich T , Becker S , Ghahramani Z , eds. Advances in Neural Information Processing Systems 14. Vol 14 . Cambridge: MIT Press; 2002:849–856.
28. Jolliffe IT . Principal Component Analysis and Factor Analysis. In: Principal Component Analysis. New York, NY: Springer New York; 1986:115–128.
29. LeCun Y , Bengio Y , Hinton G . Deep learning. Nature. 2015;521 (7553 ):436–444.26017442
30. Shah SJ , Katz DH , Selvaraj S , Phenomapping for novel classification of heart failure with preserved ejection fraction. Circulation. 2015;131 (3 ):269–279.25398313
31. Shin HC , Roth HR , Gao M , Deep Convolutional Neural Networks for Computer-Aided Detection: CNN Architectures, Dataset Characteristics and Transfer Learning. IEEE Trans Med Imaging. 2016;35 (5 ):1285–1298.26886976
32. Hirschberg J , Manning CD . Advances in natural language processing. Science. 2015;349 (6245 ):261–266.26185244
33. Zeng Z , Deng Y , Li X , Naumann T , Luo Y . Natural Language Processing for EHR-Based Computational Phenotyping. IEEE/ACM Trans Comput Biol Bioinform. 2019;16 (1 ):139–153.29994486
34. Goldberg Y A primer on neural network models for natural language processing. arXivorg. 2015.
35. Nadkarni PM , Ohno-Machado L , Chapman WW . Natural language processing: an introduction. Journal of the American Medical Informatics Association : JAMIA. 2011;18 (5 ):544–551.21846786
36. Luo Y , Thompson WK , Herr TM , Natural Language Processing for EHR-Based Pharmacovigilance: A Structured Review. Drug Saf. 2017;40 (11 ):1075–1089.28643174
37. Vaswani A , Shazeer NM , Parmar N , Attention is All you Need. ArXiv. 2017;abs/1706.03762.
38. Devlin J , Chang M-W , Lee K , Toutanova K . BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. Paper presented at: NAACL2019.
39. Alsentzer E , Murphy J , Boag W , Publicly Available Clinical BERT Embeddings. ArXiv. 2019;abs/1904.03323.
40. Huang K , Altosaar J , Ranganath R . ClinicalBERT: Modeling Clinical Notes and Predicting Hospital Readmission. ArXiv. 2019;abs/1904.05342.
41. Lee J , Yoon W , Kim S , BioBERT: a pre-trained biomedical language representation model for biomedical text mining. Bioinformatics. 2019;36 (4 ):1234–1240.
42. Rasmy L , Xiang Y , Xie Z , Tao C , Zhi D . Med-BERT: pretrained contextualized embeddings on large-scale structured electronic health records for disease prediction. npj Digital Medicine. 2021;4 (1 ):86.34017034
43. Brown TB , Mann B , Ryder N , Language Models are Few-Shot Learners. ArXiv. 2020;abs/2005.14165.
44. Korngiebel DM , Mooney S . Considering the possibilities and pitfalls of Generative Pre-trained Transformer 3 (GPT-3) in healthcare delivery. NPJ Digital Medicine. 2021;4 .
45. Byrd RJ , Steinhubl SR , Sun J , Ebadollahi S , Stewart WF . Automatic identification of heart failure diagnostic criteria, using text analysis of clinical notes from electronic health records. Int J Med Inform. 2014;83 (12 ):983–992.23317809
46. Friedlin J , McDonald CJ . A natural language processing system to extract and code concepts relating to congestive heart failure from chest radiology reports. AMIA Annu Symp Proc. 2006;Annual Symposium Proceedings/AMIA Symposium.:269–273.17238345
47. Ho JC , Ghosh J , Steinhubl SR , Limestone: high-throughput candidate phenotype generation via tensor factorization. J Biomed Inform. 2014;52 :199–211.25038555
48. Peissig PL , Santos Costa V , Caldwell MD , Relational machine learning for electronic health record-driven phenotyping. J Biomed Inform. 2014;52 :260–270.25048351
49. Vijayakrishnan R , Steinhubl SR , Ng K , Prevalence of heart failure signs and symptoms in a large primary care population identified through the use of text and data mining of the electronic health record. J Card Fail. 2014;20 (7 ):459–464.24709663
50. Schölkopf B , Smola AJ . Learning with Kernels: Support Vector Machines, Regularization, Optimization, and Beyond. Cambridge: MIT Press; 2001.
51. Litjens G , Ciompi F , Wolterink JM , State-of-the-Art Deep Learning in Cardiovascular Image Analysis. JACC: Cardiovascular Imaging. 2019;12 (8_Part_1 ):1549–1565.31395244
52. Seetharam K , Brito D , Farjo PD , Sengupta PP . The Role of Artificial Intelligence in Cardiovascular Imaging: State of the Art Review. Frontiers in Cardiovascular Medicine. 2020;7 (374 ).
53. Luo Y , Mao C , Yang Y , Integrating hypertension phenotype and genotype with hybrid non-negative matrix factorization. Bioinformatics. 2019;35 (16 ):2885.30753340
54. Zhang J , Gajjala S , Agrawal P , Fully Automated Echocardiogram Interpretation in Clinical Practice. Circulation. 2018;138 (16 ):1623–1635.30354459
55. Tison GH , Zhang J , Delling FN , Deo RC . Automated and Interpretable Patient ECG Profiles for Disease Detection, Tracking, and Discovery. Circ Cardiovasc Qual Outcomes. 2019;12 (9 ):e005289.31525078
56. Duffy G , Cheng P , Yuan N , High-Throughput Precision Phenotyping of Left Ventricular Hypertrophy with Cardiovascular Deep Learning. ArXiv. 2021;abs/2106.12511.
57. Goto S , Mahara K , Beussink-Nelson L , Artificial intelligence-enabled fully automated detection of cardiac amyloidosis using electrocardiograms and echocardiograms. Nature Communications. 2021;12 (1 ):2726.
58. Huda A , Castaño A , Niyogi A , A machine learning model for identifying patients at risk for wild-type transthyretin amyloid cardiomyopathy. Nature Communications. 2021;12 (1 ):2725.
59. Galli E , Bourg C , Kosmala W , Oger E , Donal E . Phenomapping Heart Failure with Preserved Ejection Fraction Using Machine Learning Cluster Analysis: Prognostic and Therapeutic Implications. Heart Failure Clinics. 2021;17 (3 ):499–518.34051979
60. Pandey A , Kagiyama N , Yanamala N , Deep-Learning Models for the Echocardiographic Assessment of Diastolic Dysfunction. JACC Cardiovascular imaging. 2021;14 (10 ):1887–1900.34023263
61. Collins GS , Dhiman P , Andaur Navarro CL , Protocol for development of a reporting guideline (TRIPOD-AI) and risk of bias tool (PROBAST-AI) for diagnostic and prognostic prediction model studies based on artificial intelligence. BMJ Open. 2021;11 (7 ):e048008.
62. Liu Y , Chen P-HC , Krause J , Peng L . How to Read Articles That Use Machine Learning: Users’ Guides to the Medical Literature. JAMA. 2019;322 (18 ):1806–1816.31714992
63. Sengupta PP , Shrestha S , Berthon B , Proposed Requirements for Cardiovascular Imaging-Related Machine Learning Evaluation (PRIME): A Checklist: Reviewed by the American College of Cardiology Healthcare Innovation Council. JACC: Cardiovascular Imaging. 2020;13 (9 ):2017–2035.32912474
64. Vasey B , Clifton DA , Collins GS , DECIDE-AI: new reporting guidelines to bridge the development-to-implementation gap in clinical artificial intelligence. Nature Medicine. 2021;27 (2 ):186–187.
65. Stevens LM , Mortazavi BJ , Deo RC , Curtis L , Kao DP . Recommendations for Reporting Machine Learning Analyses in Clinical Research. Circulation: Cardiovascular Quality and Outcomes. 2020;13 (10 ):e006556.33079589
66. Liu X , Faes L , Calvert MJ , Denniston AK . Extension of the CONSORT and SPIRIT statements. The Lancet. 2019;394 (10205 ):1225.
67. Norgeot B , Quer G , Beaulieu-Jones BK , Minimum information about clinical artificial intelligence modeling: the MI-CLAIM checklist. Nature Medicine. 2020;26 (9 ):1320–1324.
68. Bae CJ , Griffith S , Fan Y , The Challenges of Data Quality Evaluation in a Joint Data Warehouse. EGEMS (Wash DC). 2015;3 (1 ):1125.26290882
69. Hersh WR , Weiner MG , Embi PJ , Caveats for the use of operational electronic health record data in comparative effectiveness research. Med Care. 2013;51 (8 Suppl 3 ):S30–37.23774517
70. Wells BJ , Chagin KM , Nowacki AS , Kattan MW . Strategies for handling missing data in electronic health record derived data. EGEMS (Wash DC). 2013;1 (3 ):1035.25848578
71. Luo Y , Szolovits P , Dighe AS , Baron JM . 3D-MICE: integration of cross-sectional and longitudinal imputation for multi-analyte longitudinal clinical data. Journal of the American Medical Informatics Association : JAMIA. 2018;25 (6 ):645–653.29202205
72. Lloyd-Jones DM . Cardiovascular risk prediction: basic concepts, current status, and future directions. Circulation. 2010;121 (15 ):1768–1777.20404268
73. Steyerberg EW , Moons KG , van der Windt DA , Prognosis Research Strategy (PROGRESS) 3: prognostic model research. PLoS Med. 2013;10 (2 ):e1001381.23393430
74. Saeb S , Lonini L , Jayaraman A , Mohr DC , Kording KP . Voodoo Machine Learning for Clinical Predictions. bioRxiv. 2016.
75. Peiris D , Usherwood T , Panaretto K , Effect of a computer-guided, quality improvement program for cardiovascular disease risk management in primary health care: the treatment of cardiovascular risk using electronic decision support clusterrandomized trial. Circ Cardiovasc Qual Outcomes. 2015;8 (1 ):87–95.25587090
76. Curran GM , Bauer M , Mittman B , Pyne JM , Stetler C . Effectiveness-implementation Hybrid Designs: Combining Elements of Clinical Effectiveness and Implementation Research to Enhance Public Health Impact. Medical Care. 2012;50 (3 ):217–226.22310560
77. Finlayson SG , Subbaswamy A , Singh K , The Clinician and Dataset Shift in Artificial Intelligence. New England Journal of Medicine. 2021;385 (3 ):283–286.
78. Guo LL , Pfohl SR , Fries J , Systematic Review of Approaches to Preserve Machine Learning Performance in the Presence of Temporal Dataset Shift in Clinical Medicine. Appl Clin Inform. 2021;12 (4 ):808–815.34470057
|
PMC008xxxxxx/PMC8983117.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
101573691
39703
Cell Rep
Cell Rep
Cell reports
2211-1247
35172163
8983117
10.1016/j.celrep.2022.110376
NIHMS1781189
Article
Independent host- and bacterium-based determinants protect a model symbiosis from phage predation
Lynch Jonathan B. 12
Bennett Brittany D. 13
Merrill Bryan D. 4
Ruby Edward G. 1
Hryckowian Andrew J. 567*
1 Pacific Biosciences Research Center, University of Hawai’i at Manoa, Honolulu, HI 96822, USA
2 Department of Integrative Biology and Physiology, University of California, Los Angeles, Los Angeles, CA 90095, USA
3 Department of Biological Sciences, University of Southern California, Los Angeles, CA 90089, USA
4 Department of Microbiology and Immunology, Stanford University School of Medicine, Stanford, CA 94304, USA
5 Department of Medicine, University of Wisconsin School of Medicine and Public Health, Madison, WI 53792, USA
6 Department of Medical Microbiology and Immunology, University of Wisconsin School of Medicine and Public Health, Madison, WI 53706, USA
7 Lead contact
* Correspondence: [email protected]
AUTHOR CONTRIBUTIONS
J.B.L., B.D.B., B.D.M., and A.J.H. performed experiments/computational analyses and analyzed the data. J.B.L., B.D.B., B.D.M., and A.J.H. prepared the display items. E.G.R. provided key insights, tools, and reagents. J.B.L., B.D.B., and A.J.H. wrote the paper. All authors edited the manuscript prior to submission.
20 2 2022
15 2 2022
05 4 2022
38 7 110376110376
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
SUMMARY
Bacteriophages (phages) are diverse and abundant constituents of microbial communities worldwide, capable of modulating bacterial populations in diverse ways. Here, we describe the phage HNL01, which infects the marine bacterium Vibrio fischeri. We use culture-based approaches to demonstrate that mutations in the exopolysaccharide locus of V. fischeri render this bacterium resistant to infection by HNL01, highlighting the extracellular matrix as a key determinant of HNL01 infection. Additionally, using the natural symbiosis between V. fischeri and the squid Euprymna scolopes, we show that, during colonization, V. fischeri is protected from phages present in the ambient seawater. Taken together, these findings shed light on independent yet synergistic host- and bacterium-based strategies for resisting symbiosis-disrupting phage predation, and we present important implications for understanding these strategies in the context of diverse host-associated microbial ecosystems.
Graphical abstract
In brief
Lynch et al. isolate and characterize a bacteriophage that infects the marine bacterium Vibrio fischeri. They show that the mutualism between V. fischeri and the Hawaiian bobtail squid Euprymna scolopes is not disrupted by phages present in the environment and highlight both bacterium- and squid-based determinants of protecting this mutualism against phage predation.
pmcINTRODUCTION
Bacteriophages (phages) are viruses that infect bacteria, influencing broad areas of bacterial physiology and ecology. These effects can manifest through bacterial lysis, development of bacterial resistance to phage infection, and horizontal gene transfer and can have positive, negative, or neutral effects on bacterial fitness (Brum et al., 2015; Matilla et al., 2014; Roossinck, 2011; Suttle, 2007) . Phages are abundant across diverse environments, including the mammalian gastrointestinal tract, rhizosphere, and oceanic plankton, where they dramatically influence bacterial physiology and community function (Al-Shayeb et al., 2020; Breitbart et al., 2018; Dion et al., 2020; Reyes et al., 2010).
Given the known impacts of phages in microbial communities (microbiotas), there is great interest in the roles that phages play in eukaryote-associated microbiotas in particular. In these associations, phages can select for or against certain bacteria, alter bacterial gene expression, or influence spatial organization of the microbiota (Barr et al., 2013), suggesting that “phage therapy” could be used to selectively alter the composition and function of diverse host-associated microbiotas.
Phage biology is commonly studied using a variety of well-described model bacterium-phage pairs, such as Escherichia coli (phages T4, T7, and λ (Kutter et al., 2018)), Vibrio cholerae (phages ICP1, ICP2, and ICP3 (Yen and Camilli, 2017)), and others (Ofir and Sorek, 2018). Despite these efforts, we lack knowledge of how most phages interact with their bacterial hosts in their natural environments. This knowledge gap is due both to the high diversity of phages and their target bacteria and to the difficulty of inferring these relationships from sequence homology or other bioinformatic techniques (Dion et al., 2020; Reyes et al., 2010). This difficulty is largely driven by the fact that phages can be restrictive or promiscuous, with host ranges spanning from specific strains to multiple bacterial species (Beumer and Robinson, 2005; Flores et al., 2011; de Jonge et al., 2019). In addition, bacterially encoded phage resistance can be either broadly or narrowly protective (e.g., alteration of extracellular structures or CRISPR spacer acquisition, respectively), and phages have evolved many ways to counter these bacterial defenses (Black, 1988; Fan et al., 2018; Morona and Henning, 1984; Parent et al., 2014; Porter et al., 2020; Wang et al., 2019; Xu et al., 2014; Zborowsky and Lindell, 2019). Finally, bacterial evasion strategies can be constitutive or differentially expressed under particular environmental conditions (Reyes-Robles et al., 2018), further complicating an understanding of phage-bacterium interactions and highlighting the need for diverse and tractable model systems to study them.
In animal-bacterial symbioses, the animal host also plays key, but understudied, roles in phage dynamics. For example, phages can adhere to mucus, creating regions of high and low phage abundance and activity (Barr et al., 2013, 2015). Animals can also produce effectors like immunoglobin A that bind bacteria and protect them from phage infection (Rollenske et al., 2021). Indirectly, the host environment influences expression of surface molecules by bacterial symbionts, which changes their susceptibility to phages (Porter et al., 2020).
While many animal-associated microbiotas are highly diverse communities, rendering them difficult to study, some animals develop mutualisms with narrowly defined sets of bacteria that enable controlled experimental approaches. One well-studied example is the bioluminescence-based mutualism between the Hawaiian bobtail squid, Euprymna scolopes, and the marine γ-proteobacterium Vibrio (Aliivibrio) fischeri. Within hours after the squid hatch from their eggs, V. fischeri cells from the surrounding seawater colonize the crypt spaces of a specialized symbiotic tissue (the light organ [LO]), where the bacteria multiply and reside extracellularly for the approximately 9-month duration of the squid’s life (Essock-Burns et al., 2020; McFall-Ngai, 2014). This colonization process is marked by an extreme bottlenecking, in which ~103–106 cells/mL of V. fischeri in seawater are whittled down to a few cells that initiate the colonization (Wollenberg and Ruby, 2009). These successfully colonizing bacteria are further sorted into the six distinct crypts of the LO, so that each crypt ordinarily contains one or two clonal populations of V. fischeri cells (Bongrand and Ruby, 2019; Speare et al., 2018; Sun et al., 2016). To ensure successful colonization of symbiosis-competent V. fischeri while simultaneously restricting LO access of non-V. fischeri microbes, E. scolopes employs several strategies, such as cilia-derived flow patterns, that physically control LO access (Nawroth et al., 2017).
To leverage this tractable partnership for the study of phages in beneficial host-microbe associations, we isolated and characterized a phage with marked specificity for a symbiosis-competent V. fischeri strain, as well as spontaneous phage-resistant mutants of this strain. We hypothesized that the near-clonality of V. fischeri in the LO would predispose these symbionts to decimation upon the introduction of a virulent phage, potentially by reducing or removing the resident V. fischeri population and undoing the previously stable host-symbiont relationship. Surprisingly, our data demonstrate that, although phages prey on populations of V. fischeri in the ambient seawater, LO-resident V. fischeri are protected from phage predation. We posit that host-mediated protection of symbionts from phages provides a fitness advantage by discouraging phage sweeps of the symbiont population (e.g., no light production in the case of the squid-Vibrio symbiosis) or where general microbiota stability, rather than persistent phage-driven oscillations in community structure, promotes host homeostasis. This work identifies how animal hosts and their symbiotic bacteria can respond to and protect highly specific mutualisms from phages, and it establishes a framework to understand the mechanisms governing these relationships in this and other host-associated microbial ecosystems.
RESULTS
Isolation and genomic characterization of HNL01 and comparative analysis with Vibrio cholerae phage ICP1
Using a previously reported protocol (Hryckowian et al., 2020), we isolated HNL01 from coastal Hawaiian seawater (Figure 1A). HNL01 was propagated on V. fischeri ES114, the well-studied LO isolate of this species (Figure 1B). We assayed the bacterial-host range of HNL01 by performing plaque assays on 22 additional strains of V. fischeri, as well as five previously described mutants of ES114 (Table S1). As isolates from the ES114 background were the only strains that formed visible plaques, we conclude that HNL01 has a restricted host range within V. fischeri strains and that several common extracellular macromolecules do not impact its ability to infect. For example, mutation of neither the O-antigen of lipopolysaccharide (LPS), a well-studied symbiosis polysaccharide (Syp), nor the dominant outer membrane protein (OmpU) affected the ability of HNL01 to infect ES114; similarly, the central trait of the symbiosis, bioluminescence, was not required for phage infection (Table S1). These results suggest that the specificity of HNL01 for ES114 is driven by other effectors.
Transmission electron microscopy of HNL01 demonstrated that this phage is a myovirus (e.g., tailed phage with a contractile tail; Figures 1A and 1B), and whole-genome sequencing of HNL01 revealed a 111,792-bp genome consisting of 38% guanine-cytosine (GC) content and 166 predicted protein-coding genes, 33 of which have conserved domains (Figure S1 and Table S2). The virion morphology and genome size of HNL01 resemble the well-studied Vibrio cholerae-infecting phage ICP1, a myovirus with a 125,956-bp genome (Seed et al., 2011). Sequence similarity at the amino acid level was apparent between these two phages (Figure 1C and Table S2), with multiple conserved proteins involved in nucleotide metabolism and virion structure (Table S2 and Figure S1). Together, this resemblance suggests a distant evolutionary relationship between HNL01 and ICP1, as has been observed for phages that infect Bacteroidetes (Guerin et al., 2018; Hryckowian et al., 2020).
Phage predation does not decrease V. fischeri colonization of the E. scolopes LO
Because phages are ubiquitous in nature and are increasingly recognized to shape host-associated ecosystems (Koskella and Brockhurst, 2014), we sought to understand how HNL01 impacts the establishment and maintenance of LO colonization. Newly hatched, uncolonized juvenile squid were colonized with V. fischeri ES114 for 24 h (Figure 2A, left), and then phages were added to the ambient seawater to a final concentration of 107 plaque-forming units per milliliter (PFU/mL). After 24 h of phage exposure (i.e., 48 h after initiation of bacterial colonization), viable V. fischeri in the LOs were quantified. The results highlighted that phages in the ambient environment do not impact the maintenance of symbionts (Figure 2A, right; “no phage”: 8.25 × 104 colony-forming units per LO (CFU/LO); “plus phage”: 1.55 × 105 CFU/LO, p = 0.07). Importantly, due to the possibility of phage killing during plating, the CFU counts from the plus phage condition are potentially conservative; there may be higher CFU counts than we measured in the phage-exposed squid, emphasizing that ambient phage does not decrease LO colonization. To mimic conditions that might be encountered in coastal seawater, we simultaneously exposed juvenile squid to both excess phage (at ~107 PFU/mL) and V. fischeri cells (at ~8 × 104 CFU/mL) and allowed colonization to proceed for 24 h. We similarly observed no difference in colonization between the two colonization groups (Figure 2B; no phage: 2.1 × 105 CFU/LO; plus phage: 1.75 × 105CFU/mL, p = 0.78). Subsequently, homogenates from either 10 V. fischeri-colonized squid or 10 phage-exposed, V. fischeri-colonized squid from the colonizations represented in Figure 2B were tested for phage sensitivity in a soft agar overlay. V. fischeri extracted from LOs was similarly sensitive to phage regardless of prior phage exposure during colonization (Figure 2C). These data indicate that LO-colonizing bacteria are protected from ambient phages but that these bacteria do not acquire permanent resistance to phages in the LO.
Phage resistance mutations in ES114 do not impact LO colonization
In addition to the phage protection provided to V. fischeri by the LO (Figure 2), we asked whether phage resistance could be acquired in vitro to protect V. fischeri outside the squid. To address this question, we added HNL01 at a range of multiplicities of infection (MOIs) to cultures of V. fischeri and monitored changes in culture density (OD600) over time for 16 h. Independently of the inoculum size or MOI, a decrease in OD600 was observed within the first ~1 h of phage exposure in cultures where phage was added, followed by ~5 h, during which bacteria were undetectable, and a subsequent outgrowth of presumably resistant bacteria (Figure 3A). We collected bacteria from these cultures, re-challenged them with HNL01, and compared their phage sensitivity with that of bacteria from the phage-free cultures. As expected, bacteria that had not been previously exposed to HNL01 were susceptible, whereas bacteria that had survived exposure to HNL01 were not cleared by phage, suggesting selection for heritable mutations that result in phage resistance (Figure 3B).
We then retrieved four individual V. fischeri ES114 colonies that grew inside HNL01 plaques. After re-streaking them to isolation, we confirmed that they were resistant to HNL01 lysis via plaque assay (similar to Figure 3B). We performed PCR on these isolates with HNL01-specific primers to test for lysogeny, and did not observe any of the tested phage genes in the genomes of these V. fischeri cells (data not shown). To identify mutation(s) involved in resistance to HNL01, we performed whole-genome sequencing on these HNL01-resistant clones. A mutation shared among these resistant clones relative to wild type was a missense mutation encoding an E183V substitution in the glycosyltransferase gene VF_0175, which is part of the locus (VF_0157-80) responsible for production of exopolysaccharide (EPS) by strain ES114 (Bennett et al., 2020). Homologs of VF_0175 are predicted in many, but not all, V. fischeri genomes (Bongrand et al., 2020), suggesting that this gene may be under selection in different environmental contexts. Targeted deletion of VF_0175 or VF_0157-80, together with the provision of VF_0175 or VF_0175(E183V) in trans, confirmed that a functional copy of this gene in the context of a full-length EPS locus is necessary for HNL01 infection (Figure 3C). We will subsequently refer to the VF_0175(E183V) “phage resistance” allele as VF_0175PR.
V. fischeri EPS is associated with HNL01 susceptibility
To evaluate the extent to which the protein encoded by VF_0175 and the wild-type E183 residue contribute to EPS production, we used Alcian blue to stain EPS in liquid culture supernatants of V. fischeri ES114-derived strains. Targeted deletion of VF_0175 or VF_0157-80, along with providing VF_0175 or VF_0175PR in trans, confirmed that, as was found for phage lysis, a functional copy of VF_0175 in the context of the full-length EPS locus is necessary for wild-type levels of EPS production (Figure 3D).
Autonomous and squid-derived phage resistance contributes to symbiont colonization and maintenance
We next asked whether the VF_0175PR allele impacted competition with phage-sensitive V. fischeri. We co-cultured wild-type V. fischeri ES114 and its resistant VF_0175PR derivative in the presence or absence of HNL01. In the presence of phage, wild-type ES114 was outcompeted by the VF_0175PR strain, consistent with our observations of V. fischeri ES114 mutants in Figure 3 (Figure 4A). In addition, we observed that the VF_0175PR strain monocolonized E. scolopes LOs equally well as wild-type V. fischeri in both the presence and the absence of phage in the ambient water (Figure S2), suggesting that, rather than impacting LO colonization, the variation in this locus observed across V. fischeri strains is important for evading phages like HNL01.
To extend our previous observations that LO-resident bacteria are protected from phages, we conducted co-colonization experiments using differentially labeled wild-type and resistant VF_0175PR V. fischeri. Both strains colonized LOs similarly well in the absence of phages at both 24 and 48 h post-colonization (Figure 4B; median wild type/VF_0175PR competition index, 1.28 at 24 h and 2.68 at 48 h). However, when squid were simultaneously exposed to bacteria and phage, as in Figure 2B, the wild-type strain was outcompeted by the VF_0175PR strain (Figure 4B; median wild type/VF_0175PR competition index, 2.24 × 10−4 at 24 h and <1 × 10−4 at 48 h, p < 0.0001 compared with no phage at either time point, Mann-Whitney test; Figure 4B). Finally, we conducted a sequential-exposure experiment, where squid that were initially co-colonized with wild type and VF_0175PR were then exposed to phage at 24 h post-inoculation. Although phages present during the initiation of colonization led to the dominance of phage-resistant VF_0175PR V. fischeri in the symbiosis, equal ratios of resistant and susceptible bacteria were present in LOs initially colonized without phage, regardless of later phage addition (Figure 4B, median wild type: VF_0175PR competitive index without phage, 2.68; median competitive index with phage after co-colonization, 1.37, p = 0.187, Mann-Whitney test).
DISCUSSION
We describe two synergistic means by which a mutualistic bacterium can be protected from phage predation: a cell-intrinsic, exopolysaccharide-based resistance factor and a host-derived shielding effect. Using the highly specific mutualism between E. scolopes and V. fischeri, we have shown that these strategies protect the normal host-microbe association despite high symbiont clonality and a continuous connectivity between the symbiont culture and ambient seawater. We have also shown that these types of phage resistance are important at different stages of colonization: bacterium-determined resistance is important during the transition from free living to the initial colonization, while symbiotic host-determined resistance is important during the maintenance of an already established colonization.
We were surprised that HNL01 shared genomic features with ICP1, a phage that is prevalent and abundant in the stool of humans with cholera infections and may play a role in the seasonality of cholera through selection against the V. cholerae O-antigen, the primary receptor for ICP1 (Seed et al., 2011, 2012). While we found that some membrane components are not required for HNL01 infection (e.g., mutations in waaL (the O-antigen ligase) and eptA, both of which are involved in LPS decoration, do not affect ES114 susceptibility to HNL01; Table S1), we did identify other surface-associated infection resistance genes (e.g., mutations in the EPS gene cluster), suggesting that surface-exposed polysaccharides are important for phage infection in both V. cholerae and V. fischeri. Because V. fischeri ES114 is the only HNL01 host we identified, broader studies of V. fischeri phages around different areas of Hawai’i and other geographic regions where V. fischeri is endemic will reveal whether related phages exist for other V. fischeri strains and the ways in which V. fischeri evades phages among its various symbiotic relationships (Bongrand et al., 2020).
Envelope and extracellular matrix modifications are major determinants of host-phage relationships at the individual and population levels (Díaz-Pascual et al., 2019; Dunsing et al., 2019; Hryckowian et al., 2020; Kim et al., 2019; Porter et al., 2020; Simmons et al., 2020), emphasizing the importance of the cell surface in phage susceptibility and community dynamics. Resistance factors are important with respect to the survival of individual bacterial cells, but they likely become even more critical in areas with high densities of related and susceptible cells, where even inefficient infections could quickly sweep through a population. Scenarios like this not only include highly specific symbioses like the squid-Vibrio system but may also include spatially organized, complex communities like biofilms and microcolonies.
Given the ubiquity of phages in the biosphere and their enrichment at mucosal surfaces (Barr et al., 2013, 2015), we asked whether associating with E. scolopes promoted infection of V. fischeri by HNL01, perhaps as a regulator of symbiont density or function. Instead, our work demonstrated that E. scolopes is a safe haven from viral infection and that LO colonization protects V. fischeri from phages. This host-mediated protection may most directly benefit V. fischeri, but it also is likely to benefit the squid by protecting its mutualist asset. This type of shielding may be especially important when a specific symbiont is not easily replaceable, because of a colonization refractory period, niche competition, or ecological dynamics with the rest of the microbiota (Koch et al., 2014; Rao et al., 2021; Speare et al., 2018), as has been reported with the squid-Vibrio symbiosis. Similarly, it was previously hypothesized that the distribution of gastrointestinal bacteria between phage-accessible and phage-inaccessible sites is a driver of long-term co-existence of phage and bacteria (Lourenço et al., 2020). Conversely, whereas protection of some mutualists from phages could benefit the host, it could similarly backfire if pathogens were protected, especially if they were desired to be targeted via phage therapy.
Although the exact means by which the squid protects V. fischeri remains undefined, we speculate that several processes play a role. First, the LO epithelium is proficient at uptake of small particles, which might allow non-specific consumption or transport of phages before they can infect V. fischeri in the crypts, as has been shown with other eukaryotic cells (Bichet et al., 2021; Cohen et al., 2020; Nguyen et al., 2017). Second, mucus has been shown in other models to adsorb phages. V. fischeri both stimulates mucus production in the squid and aggregates on this mucus at the onset of colonization; this mucus might be a means of sequestering phages (Barr et al., 2013; Nyholm and McFall-Ngai, 2003; Nyholm et al., 2002). Third, E. scolopes generates flow fields through the beating of cilia on the exterior of the LO. This promotes aggregation of V. fischeri-sized particles around LO entry points. It is possible that these flow patterns may actively exclude phage-sized particles from the LO surface, mechanically protecting V. fischeri as it initiates LO colonization (Nawroth et al., 2017). These processes of symbiont protection hold broad implications for phage infections in other symbioses, such as the mammalian gastrointestinal tract, where hosts might non-specifically protect symbionts by using similar strategies. Future work with HNL01, such as visually tracking phage particles engineered to produce fluorescent proteins or phage particles labeled via fluorescence in situ hybridization as they interact with V. fischeri and the squid during colonization, will help to define specific protection mechanisms deployed during this and other symbioses.
This work also shows that, in V. fischeri, mutation-derived resistance to phages is readily selected for under phage predation (Figures 3A and 3B). In four independent HNL01-resistant clones, we identified a point mutation in VF_0175, which encodes a putative glycosyltransferase. Although we did not identify any fitness costs imposed by the VF_0175PR mutation during in vitro growth or experimental LO colonization, it is possible that such naturally arising strains would be outcompeted by wild-type V. fischeri in the absence of phages in nature, but random or specifically induced mutation to VF_0175 is strongly selected for during phage attacks.
Wild-type VF_0175 is necessary for plaque formation by HNL01 on V. fischeri ES114, but introducing VF_0175 in the absence of the surrounding VF_0157–0180 locus does not confer phage sensitivity, suggesting that carriage of a VF_0175 homolog is not sufficient for infection (Figure 3C). Interestingly, these mutant strains produce a lower level of EPS than wild-type ES114 (Figure 3D), suggesting that EPS itself may be necessary for infection by HNL01-like phages. However, similar changes in EPS production have differing effects on bacteriaprey interactions in other systems (Deveau et al., 2002; Roach et al., 2013), and another polysaccharide type produced by V. fischeri—the Syp polysaccharide—is required during LO aggregation and colonization (Yip et al., 2006), demonstrating that V. fischeri needs to balance multiple selective pressures in symbiosis. Importantly, EPS production appears to be upregulated by genes expressed by V. fischeri in the LO (Bennett et al., 2020), where the bacteria are at high density and susceptible to a population-crashing phage sweep. These regulatory genes are downregulated soon after V. fischeri is vented into the water column (Thompson et al., 2017), highlighting the importance of host-mediated protection of bacteria in the LO.
Using a binary symbiotic system, we show that host- and bacteria-derived features protect a host-microbe mutualism in the face of phage exposure. This work illuminates key approaches that members of a host-associated microbiota can use to escape or avoid phage predation, providing a resilience mechanism for these communities. This work answers long-standing questions about the maintenance of symbiont populations, and it serves as a foundation for investigating specific parameters that lead to protection in the LO. More broadly, this helps us to understand rules that modulate phage susceptibility in a specific host-associated niche, which is necessary for understanding the roles that phages play in microbiotas and for harnessing phages therapeutically.
Limitations of the study
Our work demonstrates that the mutualism between V. fischeri and E. scolopes is protected from phages in the ambient seawater but is by no means a comprehensive description of the ways the squid-Vibrio mutualism—or systems like it—are maintained in the presence of phages. As such, we did not define the mechanism by which the squid protects its symbionts from phages. In part, this is due to technical challenges of using Euprymna hosts. For example, there is a lack of genetic tools that would allow for the inhibition of specific host effectors to probe relevant interactions. In addition, autofluorescence in the squid LO conflicts with imaging techniques like SYBR Gold labeling, limiting our understanding of whether phages are excluded from the LO.
Although we identified a mutation (VF_0175PR) in V. fischeri ES114 that successfully confers resistance to HNL01 and identified that this mutation reduced EPS production, it remains to be determined whether EPS is a receptor for HNL01 or whether the EPS machinery (encoded by VF_0157-80) modifies another extracellular epitope that is the phage receptor. We partially addressed this question by using mutants that lack some surface-exposed epitopes, including the LPS O-antigen and dominant outer-membrane protein OmpU, which were still phage sensitive. This work narrows the list of possible HNL01 receptors/co-receptors but does not conclusively identify them.
STAR★METHODS
RESOURCE AVAILABILITY
Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Andrew Hryckowian ([email protected]).
Materials availability
HNL01 and strains generated as part of this study are available upon request from the Lead Contact ([email protected]).
Data and code availability
The HNL01 genome sequence and the genome sequences of the 4 HNL01-resistant ES114 clones are deposited in GenBank and are publicly available as of the date of publication. Accession numbers are listed in the key resources table. All data reported in this paper will be shared by the lead contact upon request.
This paper does not report original code.
Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.
EXPERIMENTAL MODEL AND SUBJECT DETAILS
Animal husbandry
Adult Euprymna scolopes were collected from the waters of Maunalua Bay, O‘ahu, Hawai‘i, and maintained at Kewalo Marine Laboratory, where they would lay eggs as previously described (Montgomery and McFall-Ngai, 1993). Juvenile squid were collected <3 hours after hatching and placed in filter-sterilized ocean water (FSOW) and maintained on a 12:12-h light-dark photic cycle for the duration of experiments.
Bacterial strains and culture conditions
The LO symbiont V. fischeri strain ES114 (Boettcher and Ruby, 1990) was used for all experiments unless otherwise noted. Bacteria were grown in Lysogeny broth-salt medium (LBS) (10 g tryptone L−1, 5 g yeast extract L−1, 342 mM NaCl, and 20 mM Tris, pH 7.5) (Stabb et al., 2001) or seawater-tryptone medium (SWT) (70% ocean water, 32.5 mM glycerol, 5 g tryptone L−1,3 g yeast extract L−1) (Boettcher and Ruby, 1990) at 28°C, either on agar plates or in liquid culture shaking at 225 rpm. E. coli WM3064 was grown in lysogeny broth (Bertani, 1951) supplemented with 300 μM diaminopimelic acid at 37°C.Where appropriate, chloramphenicol (2.5 μg mL−1) or kanamycin (50 μg mL−1) was added to the media.
METHOD DETAILS
Isolation of Vibrio fischeri-infecting bacteriophage HNL01
Approximately 700 mL of coastal ocean water was collected each from Kewalo Basin and Kāne’ohe Bay on O’ahu, Hawai’i, USA, and was centrifuged at 5,500 × g for 10 minutes at room temperature to precipitate solids. The supernatants were then sequentially filtered through 0.45 μm and 0.22 μm pore polyvinylidene fluoride (PVDF) filters. The filtered water was concentrated 750-fold via 100 kDa PVDF size exclusion columns. The initial screening of plaques was performed using a soft agar overlay method in which 35 μL of filtered and concentrated seawater was combined with 200 μL of overnight V. fischeri ES114 LBS culture and with 4.5 mL molten SWT top agar (cooled to ~50°C, 3.5 g L−1 agar) and poured onto a warmed SWT agar plate (15 g L−1 agar). Soft agar overlays were incubated aerobically at room temperature overnight.
Single isolated plaques were picked into 100 μL autoclaved Instant Ocean (IO; Spectrum Brands, Madison, WI, USA), and serial dilutions were prepared and spotted onto a solidified top agar overlay. Isolated plaques were picked after overnight growth. This procedure was repeated for a total of 3 times to purify HNL01.
A high-titer stock of HNL01 was generated by flooding a soft-agar overlay plate that yielded a ‘lacy’ pattern of bacterial growth (near confluent lysis). After overnight incubation of the plate, 5 mL of sterile IO were added to the plate to resuspend the phage. After at least 2 hours of incubation at room temperature, the lysate was extracted and filter sterilized through a 0.22 μm PVDF filter.
Plaque assays
Isolated colonies from freshly streaked −80°C V. fischeri stocks (Table S1) were used to inoculate liquid LBS cultures. Strains used were either generated in this work (see plasmid and mutant construction, below) or in previous studies (Ast et al., 2009; Bose et al., 2008, 2011; Dunn et al., 2006; Fidopiastis et al., 1998; Hussa et al., 2007; Lee and Ruby, 1992, 1994; Mandel et al., 2008; Nishiguchi et al., 1998; Nishiguchi and Nair, 2003; Post et al., 2012; Preheim et al., 2011; Ruby and Lee, 1998; Schwartzman et al., 2019).
After ~16 h, 200 μL of culture were mixed with 4 mL molten SWT top agar and poured onto warmed SWT agar plates. 1 μL of serial 1:10 dilutions of phage lysate was spotted onto solidified top agar, and plates were incubated at 28°C for 3–5 h before counting plaques. Alternatively, 500 μL of bacteria (either from a stationary-phase culture of V. fischeri in SWT as in Figure 3B or pooled squid homogenates as in Figure 2C) were added to 4.5 mL of molten SWT top agar, then spread onto a warmed SWT plate and allowed to solidify. 5 μL of diluted phage (109 PFU/mL unless otherwise noted) were spotted onto plates, and after drying, plates were incubated at 28°C for 24 hours to allow for active growth of V. fischeri before checking for lysis. PFU counts and phage susceptibility assays were performed on SWT plates, as plaques were noticeably smaller or non-observable on LBS plates.
HNL01 genome sequencing
DNA was extracted from a high-titer HNL01 lysate and sequencing libraries were prepared using the Ultra II FS Kit (New England Biolabs, Ipswich, MA, USA). Libraries were quantified using a BioAnalyzer (Agilent, Santa Clara, CA, USA) and subsequently sequenced using 150-base single-end reads (Illumina MiSeq). Reads were imported into Geneious Prime (2021.1.1) and quality-trimmed at an error rate of 0.001%. The Geneious assembler was used to assemble 30% of the surviving reads ≥ 150bp, with the options “Medium sensitivity/Fast” and "circularize contigs" selected. Coverage for the genome assembly reported by Geneious was 170 +− 21.
HNL01 genome annotation and comparative analysis with Vibrio cholerae phage ICP1
A 383-bp short direct terminal repeat was identified in HNL01 by visualizing read pileups in Geneious Prime (2021.1.1), and the genome was arranged to place this repeat at the 5’ end of the genome. Protein-coding genes and tRNAs were predicted and annotated using DNA Master default parameters (http://cobamide2.pitt.edu), which incorporates Genemark (Besemer and Borodovsky, 2005), Glimmer (Delcher et al., 1999), and tRNAscan-SE (Lowe and Eddy, 1997). Phage genomes were annotated and compared on the basis of shared gene phamily (pham) membership within Phamerator using default parameters (Cresawn et al., 2011). Phams are groups of related protein-encoding genes where pham membership is built and expanded when a candidate protein shares ≥32.5% identity or a BLASTp e-value ≤ 1e−50 with one or more existing members of the pham. Genome maps of HNL01 and ICP1 were visualized in Phamerator using default parameters. Amino acid sequences for HNL01 and ICP1 were concatenated and visualized as a dot plot using word size 5 in Geneious Prime (2021.1.1)
Plasmid and mutant construction
Primers used to create the V. fischeri gene deletion and expression plasmids used in this study are listed in Tables S1 and S3, respectively. In-frame deletion of VF_0175 from the V. fischeri genome was performed as previously described (Bennett et al., 2020; Saltikov and Newman, 2003). Briefly, ~1kb fragments surrounding VF_0175 were fused via an internal EcoRI site and inserted via BamHI and SacI sites into pSMV3 (Coursolle and Gralnick, 2010), which has kanamycin resistance and sacB cassettes. Plasmids were introduced to V. fischeri ES114 through conjugation with E. coli WM3064, and counter-selection to remove pSMV3 and VF_0175 was performed on low-salt, high-sucrose LB plates (Bennett et al., 2020) at room temperature. The wild-type VF_0175 complementation plasmid was constructed by cloning VF_0175 from the V. fischeri ES114 genome, and inserting it into the expression plasmid pVSV105 (Dunn et al., 2006) via SphI and KpnI sites. The VF_0175PR allele was ordered as a gblock from Integrated DNA Technologies (Coralville, IA, USA), and inserted into pVSV105 via SphI and KpnI sites.
Alcian blue detection of extracellular polysaccharides
EPS was detected as previously described (Bennett et al., 2020). Briefly, overnight bacterial cultures grown in minimal salts medium supplemented with 6.5 mM N-acetylneuraminic acid and 0.05% (wt/vol) casamino acids were centrifuged for 15 min at 12,000 × g and 4°C. 250 μL culture supernatant was incubated with 1 mL Alcian blue reagent on a rocker for one hour at room temperature, then centrifuged 10 min at 10,000 rpm and 4°C. Pellets were washed with 1 mL 100% ethanol and centrifuged 10 min at 10,000 × g and 4°C. Pellets were solubilized in 200 μL SDS-acetate and the absorbance read at OD620.
Identification of phage-resistance genes in Vibrio fischeri
Four colonies from were picked from the center of HNL01 phage spots containing at least 105 PFU. Colonies were streaked for isolation over two more passages on LBS plates. Isolates were grown overnight in LBS medium and were used to make glycerol stocks and confirm phage insensitivity. An overnight LBS culture was used for DNA extraction using the Qiagen Blood and Tissue Kit (Qiagen, Germantown, MD, USA). Sequencing libraries were prepared from genomic DNA from each HNL01-resistant isolate, as well as the parental V. fischeri ES114 strain, using the Ultra II FS Kit (New England Biolabs). Libraries were quantified using a BioAnalyzer (Agilent) and subsequently sequenced using 151-base single-end reads (Illumina MiSeq). Differences between wild-type and resistant V. fischeri ES114 was identified through genomic comparison using CLC Genomics Workbench (Qiagen).
Vibrio fischeri colonization and phage exposure
Juvenile squid were incubated with ~104–105 CFU/mL of wild-type V. fischeri ES114 and/or the VF_0175PR mutant (as noted) for 24 h. For co-colonizations, ~1:1 inocula of each strain were used, and competitive index was measured relative to strain ratio in the inoculum (competitive index = (wildtype:mutant ratio in LO)/(wild-type: mutant ratio in inoculum). When appropriate, animals were transferred to fresh FSOW after 24 h. Animals were frozen at −80°C in FSOW immediately before dark (i.e., at maximal colonization), then thawed and homogenized, and serial dilutions were plated on LBS-agar plates to determine V. fischeri CFU counts. GFP- and RFP-positive CFU were counted using a fluorescence dissecting microscope. When appropriate, phage was added to the water at ~107 PFU/mL.
In vitro phage exposures
HNL01-sensitive and resistant V. fischeri ES114 strains received pVSV102 (GFP, Kanr), pVSV208 (RFP, Camr), or pVSV104-H (Kanr) (Dunn et al., 2006) via conjugation with donor strain E. coli WM3064 (Lynch et al., 2019). Cultures of wild-type and/or HNL01-resistant V. fischeri ES114 strains were streaked onto LBS plates with the appropriate antibiotic and then cultured in LBS overnight. Cells were then diluted in SWT from 1:10–1:10,000. 100μL of diluted cells were placed in wells of a 96-well flat-bottom, transparent plate with or without HNL01 at the noted dilutions of a ~1011 PFU/mL stock, and were incubated at 28°C with periodic shaking for 24–48 h. A Tecan GenIOS plate reader (Tecan, Baldwin Park, CA, USA) was used to measure GFP and RFP fluorescence and/or overall cell density (OD600).
PCR detection of phage genes
Taq DNA polymerase was used perform to PCR against HNL01-specific genes (gp47, gp61, gp96, gp147; see Table S3) on glycerol stocks of HNL01-resistant V. fischeri isolates to assess their potential lysogen status (see Table S3 for primer sequences). Cycling conditions were: 95°C for 5 min, 35 cycles of 95°C for 30 sec → 55°C for 30 sec → 68°C for 1 min, then 68 °C for 5 min followed by a 4°C hold. PCR was also run directly on phage samples and on wild-type ES114 as controls. Products were run on a 1% agarose gel to determine PCR product presence/absence, and lack of integrated prophages was confirmed by whole genome sequencing.
Electron microscopy
Serial dilutions of phages were spotted onto V. fischeri ES114 + SWT top-agar lawns as described above and allowed to form plaques overnight. Dilutions with the most easily distinguishable plaques (i.e., the lowest dilution that still produced plaques) were briefly covered with 10 μL of FSOW to resuspend phage, then removed to a clean PCR tube. A 4 μL portion of the suspension was spotted onto a charged formvar-coated copper grid, negatively stained with uranyl acetate, and imaged with a Hitachi HT7700 transmission electron microscope at 100kV at the University of Hawai’i MICRO Facility.
QUANTIFICATION AND STATISTICAL ANALYSIS
Statistical analysis was performed using Graphpad Prism 9.1.0. Details of specific analyses, including statistical tests used, are found in applicable figure legends. All experiments were performed in at least triplicate. * = p < 0.05, ** = p < 0.005, *** = p < 0.0005, **** = p < 0.0001.
Supplementary Material
1
2
ACKNOWLEDGMENTS
We thank Randall Scarborough for assistance in collecting seawater samples, Daniel Russell and Rebecca Garlena for sequencing phage and bacterial genomes, and Tina Carvalho for assistance with transmission electron microscopy. This work was funded by NIH grant F32 GM119238 and a Ford Foundation Postdoctoral Fellowship to J.B.L., a National Science Foundation Graduate Research Fellowship DGE-114747 to B.D.M., NIH R01 GM135254–02 and R01 AI050661 to E.G.R., and startup funding from the University of Wisconsin-Madison to A.J.H. Microscopy was performed at the University of Hawai‘i’s MICRO facility, supported under COBRE grant P20 GM125508. The graphical abstract was created with biorender.com.
Figure 1. Isolation and genomic characterization of HNL01
(A and B) Transmission electron micrographs of (A) HNL01 and (B) V. fischeri ES114 and HNL01 (dotted box surrounds phage).
(C) The amino acid sequences of all annotated protein-coding open reading frames (ORFs) in HNL01 were concatenated and compared with the concatenated protein-coding ORFs in V. cholerae phage ICP1 with a dot plot. Numbers denote position within the concatenated amino acid sequence for each phage. See also Figure S1, Tables S1, and S2.
Figure 2. Ambient phage does not reduce squid colonization despite symbiont maintenance of viral sensitivity
(A) Colony-forming units (CFU) of V. fischeri ES114 from squid light organ (LO) homogenates. Squid were colonized for 24 h (left) in the absence of phage and then exposed to phage, and CFU were measured at 48 h; n = 19–20 for each experimental condition.
(B) Squid were exposed to HNL01 and V. fischeri at the same time, and CFU/LO was measured at 24 h; n = 10 for each condition.
(C) Left of dotted line: schematic showing phage spotting results. Light gray indicates growth of V. fischeri lawn in top agar; dark gray indicates lysis of cells by phage. Right of dotted line: HNL01 spotted on top agar overlay of V. fischeri from an overnight culture (left), from V. fischeri extracted from colonized LOs (middle), or from LOs of squid concurrently exposed to V. fischeri and HNL01 (right). Each point represents CFU from the LO of one squid. Bars represent median ± 95% confidence interval (CI). See also Table S1. Statistical significance determined by Mann-Whitney test.
Figure 3. V. fischeri rapidly displays phage resistance during exposure to phage during in vitro growth
(A) Growth curve of V. fischeri ES114 with varying dilutions of phage (0–1% by volume of ~1011 PFU/mL stock) and initial V. fischeri inoculum (1:10–1:10,000 of an overnight culture). Points with error bars represent mean ± SEM.
(B) Phage suspensions were spotted onto top agar lawns subcultured from V. fischeri cultures grown without (left) or with (right) phage. Arrows indicate location where phages were placed onto V. fischeri-containing top agar. Note: the white rectangle on the right side of each image is included as an image contrast reference, emphasizing lack of V. fischeri clearance in the righthand image.
(C) Plaque assay measurements of HNL01 infection of wild-type V. fischeri ES114 (black), ΔVF_0175 (blue), and ΔVF_0157-80 (orange) carrying empty vector (open symbols), vector with wild-type VF_0175 (closed symbols), or vector with VF_0175PR (half-filled symbols). Points along the x-axis represent results falling below the limit of detection (1 PFU/mL).
(D) Alcian blue was used to stain EPS in supernatants of 24-h liquid cultures grown in minimal salts medium supplemented with 6.5 mM N-acetylneuraminic acid 0.05% (wt/vol) casamino acids for wild-type V. fischeri ES114, ΔVF_0175, and ΔVF_0157-80 carrying empty vector, vector with wild-type VF_0175, or vector with VF_0175PR (same labeling scheme as 3C). Points represent biological replicates, bars represent means ± SEM. See also Tables S1 and S3.
Figure 4. Bacteria-derived phage resistance is beneficial to V. fischeri during planktonic lifestyle, but host-derived protection dominates after colonization
(A) Growth curves from co-cultured, differentially labeled wild-type (WT) V. fischeri ES114 and the VF_0175PR mutant with or without phage exposure. WT was labeled with GFP, and the VF_0175PR mutant was labeled with RFP to distinguish them during co-culture. Data presented are normalized relative light units (RLU): (RLU – RLUt=0)/(maximum RLU range for that strain). Points represent means ± SEM.
(B) Relative ratios of CFU from WT or strain VF_0175PR from LO homogenates of co-colonized squid. Phages were introduced at time = 0, time = 24 h, or neither, and homogenates were measured at t = 24 h (left two columns) or t = 48 h (right three columns); n = 10–38 for each condition. Bars represent median ± 95% CI. ****p < 0.0001, Mann-Whitney test. See also Figure S2, Tables S1, and S3.
KEY RESOURCES TABLE
REAGENT or RESOURCE SOURCE IDENTIFIER
Bacterial and virus strains
Vibrio fischeri ES114 Boettcher and Ruby, 1990 BioProject: PRJNA12986
V. fischeri ES213 (Boettcher, 1994) BioProject: PRJNA311605
V. fischeri MJ11 Mandel et al., 2008 BioProject: PRJNA19393
V. fischeri SR5 Fidopiastis et al., 1998 BioProject: PRJNA73351
V. fischeri KB2B1 Wollenberg and Ruby, 2009 BioProject: PRJNA675878
V. fischeri KB4B5 Wollenberg and Ruby, 2009 BioProject: PRJNA311605
V. fischeri MB13B1 Wollenberg and Ruby, 2009 BioProject: PRJNA311605
V. fischeri MB13B2 Wollenberg and Ruby, 2009 BioProject: PRJNA311605
V. fischeri MB13B3 Wollenberg and Ruby, 2009 BioProject: PRJNA311605
V. fischeri MB14A3 Wollenberg and Ruby, 2009 BioProject: PRJNA311605
V. fischeri MB15A4 Wollenberg and Ruby, 2009 BioProject: PRJNA311605
V. fischeri MB14A3 Wollenberg and Ruby, 2009 BioProject: PRJNA311605
V. fischeri MB11B1 Wollenberg and Ruby, 2009 BioProject: PRJNA311605
V. fischeri EM17 Lee and Ruby, 1994, Ruby and Lee, 1998 BioProject: PRJNA212806
V. fischeri EM24 Lee and Ruby, 1994, Ruby and Lee, 1998 GenBank: JF509755
V. fischeri EM30 Lee and Ruby, 1994 GenBank: JF509756
V. fischeri ET101 Nishiguchi et al., 1998 GenBank: AY292923
V. fischeri ET401 Nishiguchi and Nair, 2003 GenBank: AY292943
V. fischeri 9CS99 Wollenberg 2012, Preheim et al., 2011 GenBank: JF717859
V. fischeri H905 Lee and Ruby, 1992 BioProject: PRJNA316342
V. fischeri VLS2 Lee and Ruby, 1994, Bose et al., 2011 BioProject: PRJNA311605
V. fischeri ZF73 Gift from Martin Polz, MIT BioProject: PRJNA318805
V. fischeri 5F7 Gift from Martin Polz, MIT BioProject: PRJNA318805
V. fischeri emors.6.1 Ast et al., 2009 BioProject: PRJNA590810
V. fischeri emors.6.2 Ast et al., 2009 BioProject: PRJNA590810
V. fischeri JAS340 Schwartzman et al., 2019 N/A
V. fischeri KV1787 Hussa et al., 2007 N/A
V. fischeri JG01 Lynch et al., 2019 N/A
V. fischeri MB06859 Post et al., 2012 N/A
V. fischeri EVS102 Bose et al., 2008 N/A
V. fischeri BDB127 Bennett et al., 2020 N/A
V. fischeri BDB233 This paper N/A
V. fischeri BDB234 This paper N/A
V. fischeri BDB129 Bennett et al., 2020 N/A
V. fischeri BDB186 Bennett et al., 2020 N/A
V. fischeri BDB235 This paper N/A
V. fischeri BDB236 This paper N/A
V. fischeri BDB222 This paper N/A
V. fischeri BDB227 This paper N/A
V. fischeri BDB235 This paper N/A
V. fischeri BDB236 This paper N/A
Escherichia coli WM3064 Saltikov and Newman, 2003 N/A
Vibrio fischeri phage HNL01 This paper BioProject: PRJNA741526
Deposited data
Vibrio fischeri phage HNL01 genome sequence This paper BioProject: PRJNA741526
Complete genome sequences of HNL01-resistant V. fischeri clones This paper; BioProject: PRJNA751213 BioSample: SAMN20511716
BioSample: SAMN20511717
BioSample: SAMN20511718
BioSample: SAMN20511719
Experimental models: Organisms/strains
Euprymna scolopes Raised in house N/A
Oligonucleotides
See Table S3 for oligonucleotide information N/A N/A
Recombinant DNA
pVSV105 (Cloning vector; Camr) Dunn et al., 2006 N/A
pVSV105::VF_1075 (VF_0175, 17 bp upstream, 49 bp downstream; Camr) This paper N/A
pVSV105::VF_0175(E183V) (VF_0175 with codon 183 GAG→GTT, 17 bp upstream, 49 bp downstream; Camr) This paper N/A
pSMV3 (Deletion vector, Kanr, sacB) Coursolle and Gralnick, 2010 N/A
pVSV102 (GFP expression vector; Kanr) Dunn et al., 2006 N/A
pVSV208 (RFP expression vector; Camr) Dunn et al., 2006 N/A
pVSV105-H (Empty experssion vector; Kanr) Dunn et al., 2006 N/A
Software and algorithms
Geneious Prime (2021.1.1) N/A https://www.geneious.com/prime/
DNA Master N/A http://cobamide2.pitt.edu
Genemark Besemer and Borodovsky, 2005 http://topaz.gatech.edu/GeneMark/license_download.cgi
Glimmer Delcher et al., 1999 https://sourceforge.net/projects/glimmer/
tRNAscan-SE Lowe and Eddy, 1997 http://lowelab.ucsc.edu/tRNAscan-SE/
Phamerator Cresawn et al., 2011 https://phamerator.org/
Graphpad Prism 9.1.0 https://www.graphpad.com/ https://www.graphpad.com/scientificsoftware/prism/
Highlights
A Vibrio fischeri-infecting bacteriophage, HNL01, is isolated and characterized
HNL01 does not disrupt the mutualism between V. fischeri and Euprymna scolopes
HNL01-resistant V. fischeri mutants rapidly emerge in vitro
When HNL01 is present, squid colonization by HNL01-resistant V. fischeri is favored
SUPPLEMENTAL INFORMATION
Supplemental information can be found online at https://doi.org/10.1016/j. celrep.2022.110376.
DECLARATION OF INTERESTS
The authors declare no competing interests.
INCLUSION AND DIVERSITY
One or more of the authors of this paper self-identifies as an underrepresented ethnic minority in science. One or more of the authors of this paper received support from a program designed to increase minority representation in science.
REFERENCES
Al-Shayeb B , Sachdeva R , Chen LX , Ward F , Munk P , Devoto A , Castelle CJ , Olm MR , Bouma-Gregson K , Amano Y , (2020). Clades of huge phages from across Earth’s ecosystems. Nature 578 , 425–431.32051592
Ast JC , Urbanczyk H , and Dunlap PV (2009). Multi-gene analysis reveals previously unrecognized phylogenetic diversity in aliivibrio. Syst. Appl. Microbiol 32 , 379–386.19481895
Barr JJ , Auro R , Furlan M , Whiteson KL , Erb ML , Pogliano J , Stotland A , Wolkowicz R , Cutting AS , Doran KS , (2013). Bacteriophage adhering to mucus provide a non-host-derived immunity. Proc. Natl. Acad. Sci. U. S. A 110 , 10771–10776.23690590
Barr JJ , Auro R , Sam-Soon N , Kassegne S , Peters G , Bonilla N , Hatay M , Mourtada S , Bailey B , Youle M , (2015). Subdiffusive motion of bacteriophage in mucosal surfaces increases the frequency of bacterial encounters. Proc. Natl. Acad. Sci. U. S. A 112 , 13675–13680.26483471
Bennett BD , Essock-Burns T , and Ruby EG (2020). HBTR, a heterofunctional homolog of the virulence regulator TCPP, facilitates the transition between symbiotic and planktonic lifestyles in Vibrio fischeri. mBio 11 , e01624–e01720.32873761
Bertani G (1951). Studies on lysogenesis. I: the mode of phage liberation by lysogenic Escherichia coli. J. Bacteriol 62 , 293–300.14888646
Besemer J , and Borodovsky M (2005). GeneMark: web software for gene finding in prokaryotes, eukaryotes and viruses. Nucleic Acids Res. 33 , W451–W454.15980510
Beumer A , and Robinson JB (2005). A broad-host-range, generalized transducing phage (SN-T) acquires 16S rRNA genes from different genera of bacteria. Appl. Environ. Microbiol 71 , 8301–8304.16332816
Bichet MC , Chin WH , Richards W , Lin YW , Avellaneda-Franco L , Hernandez CA , Oddo A , Chernyavskiy O , Hilsenstein V , Neild A , (2021). Bacteriophage uptake by mammalian cell layers represents a potential sink that may impact phage therapy. iScience 24 , 102287.33855278
Black PN (1988). The fadL gene product of Escherichia coli is an outer membrane protein required for uptake of long-chain fatty acids and involved in sensitivity to bacteriophage T2. J. Bacteriol 170 , 2850–2854.3286621
Boettcher R (1994). Occurrence of plasmid DNA in the sepiolid squid symbiont Vibrio fischeri. Curr. Microbiol 29 , 279–286.
Boettcher KJ , and Ruby EG (1990). Depressed light emission by symbiotic Vibrio fischeri of the sepiolid squid Euprymna scolopes. J. Bacteriol 172 , 3701–3706.2163384
Bongrand C , and Ruby EG (2019). Achieving a multi-strain symbiosis: strain behavior and infection dynamics. ISME J. 13 , 698–706.30353039
Bongrand C , Moriano-Gutierrez S , Arevalo P , McFall-Ngai M , Ruby EG , Visick KL , and Polz M (2020). Using colonization assays and comparative genomics to discover symbiosis behaviors and factors in Vibrio fischeri. mBio 11 , e03407–e03419.32127462
Bose JL , Rosenberg CS , and Stabb EV (2008). Effects of luxCDABEG induction in Vibrio fischeri: enhancement of symbiotic colonization and conditional attenuation of growth in culture. Arch. Microbiol 190 , 169–183.18521572
Bose JL , Wollenberg MS , Colton DM , Mandel MJ , Septer AN , Dunn AK , and Stabb EV (2011). Contribution of rapid evolution of the luxR - luxI intergenic region to the diverse bioluminescence outputs of Vibrio fischeri strains isolated from different environments. Appl. Environ. Microbiol 77 , 2445–2457.21317265
Breitbart M , Bonnain C , Malki K , and Sawaya NA (2018). Phage puppet masters of the marine microbial realm. Nat. Microbiol 3 , 754–766.29867096
Brum JR , Cesar Ignacio-Espinoza J , Roux S , Doulcier G , Acinas SG , Alberti A , Chaffron S , Cruaud C , De Vargas C , Gasol JM , (2015). Ocean plankton. patterns and ecological drivers of ocean viral communities. Science 348 , 1261498.25999515
Cohen SK , Aschtgen MS , Lynch JB , Koehler S , Chen F , Escrig S , Daraspe J , Ruby EG , Meibom A , and McFall-Ngai M (2020). Tracking the cargo of extracellular symbionts into host tissues with correlated electron microscopy and nanoscale secondary ion mass spectrometry imaging. Cell. Microbiol 22 , e13177.32185893
Coursolle D , and Gralnick JA (2010). Modularity of the Mtr respiratory pathway of shewanella oneidensis strain MR-1. Mol. Microbiol 77 , 995–1008.20598084
Cresawn SG , Bogel M , Day N , Jacobs-Sera D , Hendrix RW , and Hatfull GF (2011). Phamerator: a bioinformatic tool for comparative bacteriophage genomics. BMC Bioinformatics 12 , 395.21991981
Delcher AL , Harmon D , Kasif S , White O , and Salzberg SL (1999). Improved microbial gene identification with GLIMMER. Nucleic Acids Res. 27 , 4636–4641.10556321
Deveau H , Van Calsteren MR , and Moineau S (2002). Effect of exopolysaccharides on phage-host interactions in Lactococcus lactis. Appl. Environ. Microbiol 68 , 4364–4369.12200288
Díaz-Pascual F , Hartmann R , Lempp M , Vidakovic L , Song B , Jeckel H , Thormann KM , Yildiz FH , Dunkel J , Link H , (2019). Breakdown of Vibrio cholerae biofilm architecture induced by antibiotics disrupts community barrier function. Nat. Microbiol 4 , 2136–2145.31659297
Dion MB , Oechslin F , and Moineau S (2020). Phage diversity, genomics and phylogeny. Nat. Rev. Microbiol 18 , 125–138.32015529
Dunn AK , Millikan DS , Adin DM , Bose JL , and Stabb EV (2006). New rfp- and pES213-derived tools for analyzing symbiotic Vibrio fischeri reveal patterns of infection and lux expression in situ. Appl. Environ. Microbiol 72 , 802–810.16391121
Dunsing V , Irmscher T , Barbirz S , and Chiantia S (2019). Purely polysaccharide-based biofilm matrix provides size-selective diffusion barriers for nanoparticles and bacteriophages. Biomacromolecules 20 , 3842–3854.31478651
Essock-Burns T , Bongrand C , Goldman WE , Ruby EG , McFall-Ngai MJ , and Graf J (2020). Interactions of symbiotic partners drive the development of a complex biogeography in the squid-vibrio symbiosis. mBio 11 , e00853–e00920.32457244
Fan F , Li X , Pang B , Zhang C , Li Z , Zhang L , Li J , Zhang J , Yan M , Liang W , (2018). The outer-membrane protein TolC of Vibrio cholerae serves as a second cell-surface receptor for the VP3 phage. J. Biol. Chem 293 , 4000–4013.29259138
Fidopiastis PM , Von Boletzky S , and Ruby EG (1998). A new niche for Vibrio logei, the predominant light organ symbiont of squids in the genus Sepiola. J. Bacteriol 180 , 59–64.9422593
Flores CO , Meyer JR , Valverde S , Farr L , and Weitz JS (2011). Statistical structure of host-phage interactions. Proc. Natl. Acad. Sci. U. S. A 108 , E288–E297.21709225
Guerin E , Shkoporov A , Stockdale SR , Clooney AG , Ryan FJ , Sutton TDS , Draper LA , Gonzalez-Tortuero E , Ross RP , and Hill C (2018). Biology and Taxonomy of crAss-like bacteriophages, the most abundant virus in the human gut. Cell Host Microbe 24 , 653–664.e6.30449316
Hryckowian AJ , Merrill BD , Porter NT , Van Treuren W , Nelson EJ , Garlena RA , Russell DA , Martens EC , and Sonnenburg JL (2020). Bacteroides thetaiotaomicron-infecting bacteriophage isolates inform sequence-based host range predictions. Cell Host Microbe 28 , 371–379.e5.32652063
Hussa EA , O’Shea TM , Darnell CL , Ruby EG , and Visick KL (2007). Two-component response regulators of Vibrio fischeri: identification, mutagenesis, and characterization. J. Bacteriol 189 , 5825–5838.17586650
de Jonge PA , Nobrega FL , Brouns SJJ , and Dutilh BE (2019). Molecular and evolutionary determinants of bacteriophage host range. Trends Microbiol. 27 , 51–63.30181062
Kim H , Kim M , Bai J , Lim JA , Heu S , and Ryu S (2019). Colanic acid is a novel phage receptor of Pectobacterium carotovorum subsp. carotovorum phage POP72. Front. Microbiol 10 , 143.30837957
Koch EJ , Miyashiro T , McFall-Ngai MJ , and Ruby EG (2014). Features governing symbiont persistence in the squid-vibrio association. Mol. Ecol 23 , 1624–1634.24118200
Koskella B , and Brockhurst MA (2014). Bacteria-phage coevolution as a driver of ecological and evolutionary processes in microbial communities. FEMS Microbiol. Rev 38 , 916–931.24617569
Kutter E , Bryan D , Ray G , Brewster E , Blasdel B , and Guttman B (2018). From host to phage metabolism: hot tales of phage T4’s takeover of E. coli. Viruses 10 , 387.
Lee KH , and Ruby EG (1992). Detection of the light organ symbiont, Vibrio fischeri, in Hawaiian seawater by using lux gene probes. Appl. Environ. Microbiol 58 , 942–947.16348678
Lee KH , and Ruby EG (1994). Effect of the squid host on the abundance and distribution of symbiotic Vibrio fischeri in nature. Appl. Environ. Microbiol 60 , 1565–1571.16349257
Lourenço M , Chaffringeon L , Lamy-Besnier Q , Pédron T , Campagne P , Eberl C , Bérard M , Stecher B , Debarbieux L , and De Sordi L (2020). The spatial heterogeneity of the gut limits predation and fosters coexistence of bacteria and bacteriophages. Cell Host Microbe 28 , 390–401.32615090
Lowe TM , and Eddy SR (1997). tRNAscan-SE: a program for improved detection of transfer RNA genes in genomic sequence. Nucleic Acids Res. 25 , 955–964.9023104
Lynch JB , Schwartzman JA , Bennett BD , McAnulty SJ , Knop M , Nyholm SV , and Ruby EG (2019). Ambient pH alters the protein content of outer membrane vesicles, driving host development in a beneficial symbiosis. J. Bacteriol 201 , e00319–19.31331976
Mandel MJ , Stabb EV , and Ruby EG (2008). Comparative genomics-based investigation of resequencing targets in Vibrio fischeri: focus on point miscalls and artefactual expansions. BMC Genomics 9 , 138.18366731
Matilla MA , Fang X , and Salmond GPC (2014). Viunalikeviruses are environmentally common agents of horizontal gene transfer in pathogens and biocontrol bacteria. ISME J. 8 , 2143–2147.25118075
McFall-Ngai MJ (2014). The importance of microbes in animal development: lessons from the squid-Vibrio symbiosis. Annu. Rev. Microbiol 68 , 177–194.24995875
Montgomery MK , and McFall-Ngai M (1993). Embryonic development of the light organ of the sepiolid squid Euprymna scolopes berry. Biol. Bull 184 , 296–308.29300543
Morona R , and Henning U (1984). Host range mutants of bacteriophage Ox2 can use two different outer membrane proteins of Escherichia coli K-12 as receptors. J. Bacteriol 159 , 579–582.6378883
Nawroth JC , Guo H , Koch E , Heath-Heckman EAC , Hermanson JC , Ruby EG , Dabiri JO , Kanso E , and McFall-Ngai M (2017). Motile cilia create fluid-mechanical microhabitats for the active recruitment of the host microbiome. Proc. Natl. Acad. Sci 114 , 9510–9516.28835539
Nguyen S , Baker K , Padman BS , Patwa R , Dunstan RA , Weston TA , Schlosser K , Bailey B , Lithgow T , Lazarou M , (2017). Bacteriophage transcytosis provides a mechanism to cross epithelial cell layers. mBio 8 , e01874–e01917.29162715
Nishiguchi MK , and Nair VS (2003). Evolution of symbiosis in the Vibrionaceae: a combined approach using molecules and physiology. Int. J. Syst. Evol. Microbiol 53 (Pt 6 ), 2019–2026.14657139
Nishiguchi MK , Ruby EG , and McFall-Ngai MJ (1998). Competitive dominance among strains of luminous bacteria provides an unusual form of evidence for parallel evolution in sepiolid squid-vibrio symbioses. Appl. Environ. Microbiol 64 , 3209–3213.9726861
Nyholm SV , and McFall-Ngai MJ (2003). Dominance of Vibrio fischeri in secreted mucus outside the light organ of Euprymna scolopes: the first site of symbiont specificity. Appl. Environ. Microbiol 69 , 3932–3937.12839763
Nyholm SV , Deplancke B , Gaskins HR , Apicella MA , and McFall-Ngai MJ (2002). Roles of Vibrio fischeri and nonsymbiotic bacteria in the dynamics of mucus secretion during symbiont colonization of the Euprymna scolopes light organ. Appl. Environ. Microbiol 68 , 5113–5122.12324362
Ofir G , and Sorek R (2018). Contemporary phage biology: from classic models to new insights. Cell 172 , 1260–1270.29522746
Parent KN , Erb ML , Cardone G , Nguyen K , Gilcrease EB , Porcek NB , Pogliano J , Baker TS , and Casjens SR (2014). OmpA and OmpC are critical host factors for bacteriophage Sf6 entry in Shigella. Mol. Microbiol 92 , 47–60.24673644
Porter NT , Hryckowian AJ , Merrill BD , Fuentes JJ , Gardner JO , Glowacki RWP , Singh S , Crawford RD , Snitkin ES , Sonnenburg JL , (2020). Phase-variable capsular polysaccharides and lipoproteins modify bacteriophage susceptibility in bacteroides thetaiotaomicron. Nat. Microbiol 5 , 1170–1181.32601452
Post DMB , Yu L , Krasity BC , Choudhury B , Mandel MJ , Brennan CA , Ruby EG , McFall-Ngai MJ , Gibson BW , and Apicella MA (2012). O-antigen and core carbohydrate of Vibrio fischeri lipopolysaccharide. J. Biol. Chem 287 , 8515–8530.22247546
Preheim SP , Boucher Y , Wildschutte H , David LA , Veneziano D , Alm EJ , and Polz MF (2011). Metapopulation structure of Vibrionaceae among coastal marine invertebrates. Environ. Microbiol 13 , 265–275.20819104
Rao C , Coyte KZ , Bainter W , Geha RS , Martin CR , and Rakoff-Nahoum S (2021). Multi-kingdom ecological drivers of microbiota assembly in preterm infants. Nature 591 , 633–638.33627867
Reyes-Robles T , Dillard RS , Cairns LS , Silva-Valenzuela CA , Housman M ,Ali A , Wright ER , and Camilli A (2018). Vibrio cholerae outer membrane vesicles inhibit bacteriophage infection. J. Bacteriol 200 , 1–9.
Reyes A , Haynes M , Hanson N , Angly FE , Heath AC , Rohwer F , and Gordon JI (2010). Viruses in the faecal microbiota of monozygotic twins and their mothers. Nature 466 , 334–338.20631792
Roach DR , Sjaarda DR , Castle AJ , and Svircev AM (2013). Host exopolysaccharide quantity and composition impact erwinia: amylovora bacteriophage pathogenesis. Appl. Environ. Microbiol 79 , 3249–3256.23503310
Rollenske T , Burkhalter S , Muerner L , von Gunten S , Lukasiewicz J , Wardemann H , and Macpherson AJ (2021). Parallelism of intestinal secretory IgA shapes functional microbial fitness. Nature 598 , 657–661.34646015
Roossinck MJ (2011). The good viruses: viral mutualistic symbioses. Nat. Rev. Microbiol 9 , 99–108.21200397
Ruby EG , and Lee KH (1998). The Vibrio fischeri-Euprymna scolopes light organ association: current ecological paradigms. Appl. Environ. Microbiol 64 , 805–812.16349524
Saltikov CW , and Newman DK (2003). Genetic identification of a respiratory arsenate reductase. Proc. Natl. Acad. Sci 100 , 10983–10988.12939408
Schwartzman JA , Lynch JB , Flores Ramos S , Zhou L , Apicella MA , Yew JY , and Ruby EG (2019). Acidic pH promotes lipopolysaccharide modification and alters colonization in a bacteria–animal mutualism. Mol. Microbiol 112 , 1326–1338.31400167
Seed KD , Bodi KL , Kropinski AM , Ackermann HW , Calderwood SB , Qadri F , and Camilli A (2011). Evidence of a dominant lineage of Vibrio cholerae-specific lytic bacteriophages shed by cholera patients over a 10-year period in Dhaka, Bangladesh. mBio 2 , e00334–10.21304168
Seed KD , Faruque SM , Mekalanos JJ , Calderwood SB , Qadri F , and Camilli A (2012). Phase variable O antigen biosynthetic genes control expression of the major protective antigen and bacteriophage receptor in Vibrio cholerae O1. Plos Pathog. 8 , e1002917.23028317
Simmons EL , Bond MC , Koskella B , Drescher K , Bucci V , and Nadell CD (2020). Biofilm structure promotes coexistence of phage-resistant and phage-susceptible bacteria. mSystems 5 , e00877–19.32576653
Speare L , Cecere AG , Guckes KR , Smith S , Wollenberg MS , Mandel MJ , Miyashiro T , and Septer AN (2018). Bacterial symbionts use a type VI secretion system to eliminate competitors in their natural host. Proc. Natl. Acad. Sci. U. S. A 115 , E8528–E8537.30127013
Stabb EV , Reich KA , and Ruby EG (2001). Vibrio fischeri genes hvnA and hvnB encode secreted NAD+-glycohydrolases. J. Bacteriol 183 , 309–317.11114931
Sun Y , LaSota ED , Cecere AG , LaPenna KB , Larios-Valencia J , Wollenberg MS , and Miyashiro T (2016). Intraspecific competition impacts Vibrio fischeri strain diversity during initial colonization of the squid light organ. Appl. Environ. Microbiol 82 , 3082–3091.27016564
Suttle CA (2007). Marine viruses - major players in the global ecosystem. Nat. Rev. Microbiol 5 , 801–812.17853907
Thompson LR , Nikolakakis K , Pan S , Reed J , Knight R , and Ruby EG (2017). Transcriptional characterization of Vibrio fischeri during colonization of juvenile Euprymna scolopes. Environ. Microbiol 19 , 1845–1856.28152560
Wang C , Tu J , Liu J , and Molineux IJ (2019). Structural dynamics of bacteriophage P22 infection initiation revealed by cryo-electron tomography. Nat. Microbiol 4 , 1049–1056.30886360
Wollenberg MS , and Ruby EG (2009). Population structure of Vibrio fischeri within the light organs of Euprymna scolopes squid from two oahu (Hawaii) populations. Appl. Environ. Microbiol 75 , 193–202.18997024
Xu D , Zhang J , Liu J , Xu J , Zhou H , Zhang L , Zhu J , and Kan B (2014). Outer membrane protein OmpW is the receptor for typing phage VP5 in the Vibrio cholerae O1 el tor biotype. J. Virol 88 , 7109–7111.24719419
Yen M , and Camilli A (2017). Mechanisms of the evolutionary arms race between Vibrio cholerae and Vibriophage clinical isolates. Int. Microbiol 20 , 116–120.29446802
Yip ES , Geszvain K , DeLoney-Marino CR , and Visick KL (2006). The symbiosis regulator RscS controls the syp gene locus, biofilm formation and symbiotic aggregation by Vibrio fischeri. Mol. Microbiol 62 , 1586–1600.17087775
Zborowsky S , and Lindell D (2019). Resistance in marine cyanobacteria differs against specialist and generalist cyanophages. Proc. Natl. Acad. Sci. U. S. A 116 , 16899–16908.31383764
|
PMC008xxxxxx/PMC8983122.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
2984705R
2786
Cancer Res
Cancer Res
Cancer research
0008-5472
1538-7445
35064018
8983122
10.1158/0008-5472.CAN-21-2691
NIHMS1775826
Article
Development of novel aptamer-based targeted chemotherapy for bladder cancer
Wang Yao 14#
Zhang Yang 24#
Li Pengchao 3#
Guo Jiajie 14
Huo Fan 2
Yang Jintao 14
Jia Ru 2
Wang Juan 5
Huang Qiju 2
Theodorescu Dan 67*
Yu Hanyang 14*
Yan Chao 24*
1 State Key Laboratory of Coordination Chemistry, Department of Biomedical Engineering, College of Engineering and Applied Sciences, Jiangsu Key Laboratory of Artificial Functional Materials, Nanjing University; Nanjing, Jiangsu, China.
2 State Key Laboratory of Pharmaceutical Biotechnology, School of Life Sciences, Nanjing University; Nanjing, Jiangsu, China.
3 Department of Urology, the First Affiliated Hospital of Nanjing Medical University; Nanjing, Jiangsu, China.
4 Chemistry and Biomedicine Innovation Center (ChemBIC), Nanjing University; Nanjing, Jiangsu, China.
5 State Key Laboratory of Coordination Chemistry, School of Chemistry and Chemical Engineering, Nanjing University; Nanjing, Jiangsu, China.
6 Samuel Oschin Comprehensive Cancer Institute, Cedars-Sinai Medical Center; Los Angeles, CA, USA.
7 Department of Surgery (Urology), Cedars-Sinai Medical Center; Los Angeles, CA, USA.
# These authors contributed equally.
Author Contributions:
Conceptualization: DT, HY, CY
Methodology: YW, YZ, PL, FH, HY, CY
Investigation: YW, YZ, PL, JG, FH, JY, RJ, JW, QH
Visualization: YW, YZ, PL
Funding acquisition: DT, HY, CY
Project administration: PL, HY, CY
Supervision: PL, HY, CY
Writing – original draft: YW, YZ, PL
Writing – review & editing: DT, HY, CY
* Corresponding authors: Dan Theodorescu, 8700 Beverly Blvd, SCCT OCC Mezz C2001, Los Angeles, CA 90048, +1-310-423-8431, [email protected]; Hanyang Yu, 163 Xianlin Ave., College of Engineering and Applied Sciences, Nanjing, Jiangsu 210023, +86-17714530121, [email protected]; Chao Yan, 163 Xianlin Ave., School of Life Sciences, Nanjing, Jiangsu 210023, +86-25-89681662, [email protected].
11 3 2022
15 3 2022
15 9 2022
82 6 11281139
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Bladder cancer (BC) is common worldwide, with most patients presenting with non-muscle invasive disease. Multiple intravesical recurrences lead to reduced quality of life and high costs for patients with this form of BC. Intravesical chemotherapy aimed at reducing recurrence is the standard-of-care but has significant side effects from non-specific cytotoxicity to normal urothelium. Importantly, toxicity limits doses that can be administered. Thus, tumor-specific drug targeting could reduce toxicity and enhance effectiveness by allowing higher doses. Here, using cell internalization systematic evolution of ligands by exponential enrichment (SELEX), we identify a novel bladder cancer-specific, chemically modified nucleic acid aptamer that can be preferentially internalized into tumor cells but not normal urothelial cells. The 35-nucleotide B1 aptamer is internalized into bladder cancer cells through clathrin-mediated endocytosis and macropinocytosis. As proof of principle, a B1-guided DNA nanotrain delivery vehicle for epirubicin was constructed as a targeted intravesical chemotherapy. The B1-nanotrain-epirubicin construct exhibited selective cytotoxicity towards bladder cancer cells and outperformed epirubicin in murine orthotopic xenograft models of human bladder cancer. This aptamer-based delivery system makes targeted chemotherapy possible for bladder cancer, providing a compelling rationale for clinical development.
bladder cancer
aptamer
intravesical therapy
chemotherapy
epirubicin
pmcINTRODUCTION
Bladder cancer (BC) is one of the most common malignancies worldwide, with an estimated 550,000 new cases and 200,000 deaths annually (1–3). Approximately 75% of patients have non-muscle invasive bladder cancer (NMIBC) at initial diagnosis. All such patients undergo an attempt at complete transurethral tumor resection. For those where complete resection is not possible or whose tumors have high-risk pathological features, resection is followed by adjuvant intravesical chemotherapy or immunotherapy aimed at reducing tumor recurrence and progression (4–6). However, recurrence and progression remain a problem in many patients despite adjuvant therapy. Such therapy also has side effects due to non-specific cytotoxicity to the normal urothelium, limiting the doses of agents delivered. Thus, tumor-specific delivery of drugs would not only reduce toxicity but also enhance effectiveness by allowing use of higher drug doses. Antibody-drug conjugate (ADC) is a promising strategy for enhancing tumor-specific therapeutic targeting (7,8). Several systemically administrated ADCs are currently in clinical trials for muscle invasive bladder cancer (MIBC) but none have been tested intravesically for treating NMIBC, partly due to the high cost (9). Therefore, there is an unmet need for effective and affordable, tumor-specific therapy for bladder cancer.
Aptamers are single-stranded oligonucleotides that bind target molecules by forming distinct stable tertiary structures (10–12). Aptamer sequences towards various molecular targets can be easily identified in the laboratory by in vitro selection (also known as systematic evolution of ligands by exponential enrichment or SELEX) (13,14). In addition to affinity and selectivity comparable to that of antibodies, aptamers possess unique advantages such as ease of synthesis and modification and low immunogenicity (15). Aptamers can bind a wide variety of targets ranging from metal ions and small molecules to large proteins and even whole cells (16–20). Aptamers with cancer cell specificity are particularly intriguing because they provide an attractive platform for both diagnosis and therapy (21–25). Aptamers with selectivity for acute T lymphoblastic leukemia, Burkitt’s lymphoma, colon cancer, glioblastoma, non-small cell lung cancer and ovarian adenocarcinoma cancer cell lines have been identified (26).
Here we report not only the first bladder cancer-selective aptamer but show that it can be linked to a nanotrain construct for targeted delivery of chemotherapeutics intravesically. Intravesical instillation of the aptamer-guided nanotrain coupled with epirubicin showed selective anti-tumor activity in an orthotopic xenograft model of human bladder cancer. This work lays the foundation for major improvement in selectivity and efficacy of intravesical chemotherapy for bladder cancer.
MATERIALS AND METHODS
Cell culture
Human bladder cancer cell lines T24 (CLS Cat# 300352/p619_T24, RRID: CVCL_0554) (27), TCCSUP (ATCC Cat# HTB-5, RRID: CVCL_1738), J82 (KCLB Cat# 30001, RRID: CVCL_0359), SW780 (ATCC Cat# CRL-2169, RRID: CVCL_1728), UM-UC-3 (KCB Cat# KCB 2014053YJ, RRID: CVCL_1783), RT4 (CLS Cat# 300326/p10281_RT-4, RRID: CVCL_0036) and Biu87 (KCB Cat# KCB 2006103YJ, RRID: CVCL_6881) were purchased from the National Collection of Authenticated Cell Cultures (Shanghai, China) between 2018 and 2021. The normal bladder epithelial cell line SV-huc-1 (BCRC Cat# 60358, RRID: CVCL_3798) (28) and human bladder cancer cell line 5637 (CLS Cat# 300105/p667_5637, RRID: CVCL_0126) were purchased from American Type Culture Collection (ATCC). Human bladder cancer cell line KU-7 (RRID: CVCL_4714) engineered to stably express firefly luciferase and green fluorescent protein (GFP) was purchased from Caliper Life Sciences (Hopkinton, MA). Detailed information of the cell lines was summarized in Supplementary Table 3. The cells were cultured in either RPMI-1640 medium (T24, TCCSUP, J82, SW780, UM-UC-3, RT4, Biu87, 5637 and KU-7) or F12K (SV-huc-1) medium supplemented with 10% fetal bovine serum (FBS, Hyclone) and 100 units/mL penicillin-streptomycin (Gibco), and incubated at 37°C with 5% CO2 and 95% humidity. All cells were authenticated by STR and checked for mycoplasma contamination every 3 months. Cells were discarded after 15 passages from thawing.
Preparation of the initial chemically modified nucleic acid library
The initial single-stranded DNA library contained 5’ biotin and was purchased from Biosyntech (Suzhou, China). This library contained a central 40-nucleotide (nt) randomized region flanked by an upstream 19-nt and a downstream 20-nt primer binding site. This DNA library was used as a template in primer extension reactions to prepare the chemically modified nucleic acid library for selection. In the primer extension reaction, primer 1R (Table S1) and biotin-labeled template were annealed in the reaction buffer (20 mM Tris-HCl, pH 8.4, 50 mM KCl and 5 mM MgCl2) containing chemically modified nucleoside triphosphates (2’-F pyrimidines and 2’-deoxy purines). SFM4–3 (a variant of Stoffel fragment of Taq DNA polymerase) (29) and MnCl2 were added to initiate the reaction (30). After primer extension reaction at 50°C for 12 h, the double-stranded DNA (dsDNA) was immobilized on streptavidin-coated agarose beads via biotin-streptavidin interaction and the non-biotin-labeled complementary strand was recovered by eluting with 100 μL NaOH (200 mM) for 3 min for 5 times.
Cell internalization SELEX
One day prior to selection, 5 × 106 T24 cells were seeded in a 100-mm dish. On the next day, when cells grew to a density of 80%, the medium were removed and the wells were blocked with RPMI-1640 media containing 100 μg/mL of salmon sperm DNA and 100 μg/mL yeast tRNA for 30 min (all rounds expect rounds 1 and 2 during selection). The prepared chemically modified nucleic acid library was dissolved in selection buffer (PBS supplemented with 1 mM MgCl2 and 1 mM CaCl2) and allowed to fold by heating at 95°C for 5 min and cooling on ice for 10 min. The library was then incubated with T24 cells at 37°C for 1 h. After incubation, cells were washed with PBS thrice followed by digestion with trypsin. Surface-bound nucleic acid sequences were removed by treating with a nuclease blend (0.8 U/μL DNase I, 1 U/μL RNase I, 0.5 U/μL RNase T1/A, Thermo) at 37°C for 10 min (31–33). Cells were washed again with PBS thrice. Sequences that were internalized were recovered by extracting total RNA with Trizol. And the complementary DNA (cDNA) sequences were synthesized using primer 1F and MMLV-RT (Moloney murine leukemia virus reverse transcriptase, NEB). The cDNA was amplified by PCR using primers 2F and 3R. The dsDNA PCR products were separated on 12% denatured polyacrylamide gels. The biotin-labeled strand migrated faster and was excised and recovered from the gel. This strand was then used as a template in a primer extension reaction to prepare a chemically modified nucleic acid library for the next round of selection.
In order to enrich for nucleic acid aptamer sequences that were selective to T24 cells, negative selection steps were introduced in round 3 through round 14 to remove sequences bound to SV-huc-1 cells. In order to enrich for aptamer sequences that were capable of high-affinity binding and fast internalization, the selection stringency was incrementally increased by: i) decreasing the number of target T24 cells from 5 × 106 in a 100-mm dish to 1 × 106 in a 30-mm dish; ii) reducing the incubation time of library with target cells from 60 min to 30 min; iii) increasing the digestion time of nuclease blend from 10 min to 30 min. After 14 rounds of in vitro selection and amplification, the enriched DNA pool was cloned into pEasy-T1 vector (TransGen, Cat#CT101–01) and sent to Sangon Biotech (Shanghai, China) for sequencing.
Flow cytometry
To monitor the selection progress or to test the internalization effects of individual aptamers, cells were seeded in 12-well plates at 3 × 105 cells/well and incubated at 37°C overnight. Cells were then blocked by incubation with 150 μL 2× selection buffer and 150 μL RPMI-1640 medium supplemented with 100 μg/mL yeast tRNA and 100 μg/mL salmon sperm DNA for 30 min. The Cy5.5-labeled pool from each round or individual aptamer was folded in 150 μL 2× selection buffer by heating at 95°C for 5 min and cooling on ice for 10 min. Nucleic acid solution (150 μL) and RPMI-1640 medium (150 μL) were added to incubate with cells for 1 h. Following incubation, cells were washed with PBS for three times and treated with 200 μL 0.05% trypsin. Trypsin digestion was stopped by addition of 300 μL complete media and cells were transferred to a 1.5 mL tube for centrifugation. Cells were re-suspended with 1 mL PBS and analyzed on an Attune NxT Flow Cytometer (Invitrogen) recording 10,000 events. Data were analyzed with the FlowJo software (FlowJo, RRID: SCR_008520).
To investigate the effect of various endocytosis inhibitors, T24 cells were seeded and prepared similarly. After washing with PBS twice, cell were treated with different endocytosis inhibitors for 1 h. The inhibitors used in this study included clathrin-associated endocytosis inhibitors (100 μM chlorpromazine; 200 μM dynasore) and non-clathrin-associated endocytosis inhibitors (200 μM amiloride; 500 μM omeprazole; 2.5 mM cyclodextrin). Chlorpromazine is a cationic amphiphilic drug that can inhibit clathrin-mediated endocytosis by preventing clathrin-coated pit formation (34). Dynasore is a potent inhibitor of dynamin GTPase activity, which is required during vesicle fission, and can therefore effectively block clathrin-mediated endocytosis (35). Amiloride causes acidification in submembranous region at sites of macropinocytosis through inhibition of Na+/H+ exchanger pump, and interferes with micropinocytosis (36). Omeprazole is a specific H+/K+ ATPase proton pump inhibitor (37). Cyclodextrin is an inducer of cholesterol depletion (38). After treatment with endocytosis inhibitors, nucleic acid aptamers were added and cells were treated and analyzed as described above.
Bladder cancer patient samples
Primary human bladder tumor tissue and normal urothelium tissues were collected from the surgical specimen of patients who received radical cystectomy for high-risk non-muscle invasive bladder cancer. Informed written consent was obtained from each patient, the collection of tissue specimens was approved by the Internal Review and Ethics Boards at the First Affiliated Hospital of Nanjing Medical University (2021-SR-083). Briefly, freshly collected bladder tumor tissue or normal para-tumor urothelium tissue was digested with 1 mg/mL collagenase for 30 min, filtered through a 70 μm cell sieve, and resuspended to 1 × 106 cells/mL in cell culture medium supplemented with 10 μM Rock inhibitor (Selleck). Subsequently, cells were incubated with Cy5.5-labeled aptamer B1 and analyzed with flow cytometry as describe above. For immunoblotting analysis of AP2B1, 20 mg of tissue samples were homogenized in 150–250 μL cold RIPA buffer on ice, then centrifuged at 14,000 g for 5 minutes at 4°C, proteins in the supernatant were then separated by SDS-PAGE, transferred to PVDF membrane, and subjected to Western blotting with AP2B1 antibody (Abcam Cat# 151961, RRID: AB_2721072).
Confocal Imaging
T24 and SV-huc-1 cells were plated at 2 × 105 cells/well in 12-well plates and incubated overnight. Cy5.5-labeled nucleic acid sequences (200 nM individual aptamer or nanotrain) were added to each well and incubated with cells for 1 h. After washing with PBS thrice, cell nuclei were stained with Hoechst 33342 for 10 min. Finally, cells were imaged on a confocal microscope (Zeiss LSM 880 with Airyscan, Zeiss, Germany). To determine the subcellular localization of aptamers, cells were stained with 1 mM LysoTracker Green DND-26 (Beyotime, Shanghai, China) for 30 min before confocal microscopy.
Target identification and validation
To explore the potential roles of membrane proteins during aptamer internalization, T24 cells were digested with 200 μL 0.05% trypsin/0.53 mM EDTA or 0.1 mg/mL proteinase K at 37°C for 10 min, respectively. FBS was then added to quench the reaction. After washing with PBS twice, cells were incubated with 200 nM Cy5.5-modified aptamer in 300 μL binding buffer at 37°C for 1 h, and analyzed by flow cytometry as described above.
To identify the possible protein targets involved in aptamer internalization, T24 cells were washed with PBS twice and lysed in 300 μL native lysis buffer for 10 min. T24 cell lysates were collected in an Eppendorf tube and incubated on ice for 30 min. After centrifugation the supernatant was recovered for further analysis. Separation of the possible target proteins by pull-down was performed as previously described (39). Briefly, a biotin-labeled aptamer was incubated with streptavidin-coated magnetic beads at room temperature for 30 min. Then the aptamer-bead conjugates were mixed with cell lysate supernatant to allow the interaction of aptamer with possible target proteins at room temperature for 30 min. After washing the beads with buffer (20 mM Tris, pH 7.5, 10 mM NaCl, and 0.1% Tween-20) for three times to remove unbound proteins, the beads were boiled in PBS and SDS-PAGE loading buffer for 2 min. The eluted proteins in supernatant were analyzed with 12% SDS-PAGE and stained with Brightein protein gel stain (US Everbright, Suzhou, China). A scrambled nucleic acid sequence was used in a parallel experiment as negative control. The protein gel electrophoresis results of the aptamer and control groups were compared, and the protein bands specifically present in the aptamer group was excised, digested, and analyzed by LC-MS/MS. Data were analyzed using the ProteinPilot software (ThermoFisher, RRID:SCR_018681).
After identification of AP2B1 as a potential target protein, AP2B1-specific siRNA (100 nM) was transfected into T24 cells with 2 μL Enhancer-4000 transfection reagent. The AP2B1 siRNA gene silencing effect was validated by immunoblotting with anti-AP2B1 rabbit monoclonal antibody (Abcam) and dylight-800-modified secondary antibody (Cell Signaling Technology Cat# 5151, RRID:AB_10697505). In T24 cells treated with AP2B1 siRNA, the aptamer mediated internalization was evaluated by flow cytometry 72 h after transfection.
Preparation and characterization of nanotrain
Aptamer B1 with a trigger sequence, sequences M1 and M2 (Table S1) were individually folded in PBS supplemented with 5 mM MgCl2 by heating at 95°C for 5 min and cooling on ice for 10 min. Subsequently, these three sequences were mixed, and allowed to anneal for 24 h. The assembly of the nanotrain was evaluated by 2% agarose gel electrophoresis. A scrambled B1 with a trigger sequence was also used to construct a nonselective nanotrain as a negative control in flow cytometry and confocal imaging analysis.
Drug-loaded nanotrain (NT-EPI) was prepared by adding epirubicin (16 μM) into the nanotrain solution and incubating at room temperature for 24 h. The drug loading and releasing events were monitored by fluorescence spectrometry (excitation: 475 nm; emission: 590 nm), using a SpectraMax ID3 microplate reader (Molecular Devices). To determine the stability of the NT-EPI construct, a drug diffusion assay was performed. The prepared NT-EPI was transferred into a dialysis unit (3,500 molecular weight cut off). The exterior chamber contained PBS supplemented with 5 mM MgCl2. At specific time points, a 100 μL solution was collected from the exterior chamber and the fluorescence intensity was measured to determine the released drug concentration (excitation: 475 nm; emission: 590 nm). To simulate in vivo release of epirubicin, NT-EPI was treated with DNase I or low pH conditions, epirubicin release from the nanotrain was determined by fluorescence spectrometry.
In vitro cytotoxicity assay
Cells were seeded at 5 × 103 cells/well in 96-well plates in 100 μL complete medium and incubated overnight. Cells were then treated with 100 μL EPI or NT-EPI (from 0.04 to 0.63 μM) for 48 h. Cell viability was determined by addition of MTT ([3-(4,5)-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide, Sigma) to cells and incubation for 4 h. The absorbance at 550 nm was recorded using a SpectraMax M4 microplate reader (Molecular Devices) following solvation of formazan with 100 μL DMSO.
Establishment of mouse orthotopic xenograft bladder tumor model
All animal care and handling procedures were approved by the Ethics Committee of Nanjing University and were performed in accordance with NIH guidelines. Female nu/nu mice (8 weeks old, average 21-gram body weight), were purchased from SLAC Laboratory Animals (Shanghai, China). Mice were maintained on a 12 h light/dark cycle with free access to food and water. KU-7 cancer cells stably expressing luciferase were prepared and concentrations adjusted to 2 × 107 cells/mL in PBS. Mice were anaesthetized with 5% sodium pentobarbital and residual urine in the bladder were removed by gentle pressing. An indwelling needle (0.7 × 19 mm, BD 8234041) was gently inserted into mouse bladder through the urethra for injection of fluids and cells. Silver nitrate (AgNO3, 5 μL, 0.3 M) was first injected intravesically to erode the mouse bladder epithelium for 20 s, followed by injection of 50 μL KU-7 cell suspension. Mice were returned to cages after 2 h. Two weeks after tumor inoculation, tumor growth was examined by IVIS after intraperitoneal injection of the sodium fluorescein luciferase substrate.
In vivo therapeutic experiments
On day 14 after KU-7 tumor implantation, all mice with positive tumor formation were randomly divided into three groups for subsequent treatment studies. Mice were intravesically instilled with 0.4 mg/mL of free epirubicin, nanotrain-epirubicin and PBS, respectively. Mice were treated every 5 days for a total of 7 times. Body weights were recorded every day. At the end of experiment, tumor fluorescence signal was determined using the IVIS system.
Immunohistochemistry
The whole mouse bladders were fixed in 4% paraformaldehyde solution for 72 h, embedded in paraffin and cut into 5 μm thick slices. The sections were placed on a glass slide, deparaffinized and hydrated. Then the sections were permeabilized with 2% Triton X-100 for 1 h, incubated with anti-eGFP antibody (Abcam Cat# ab111258, RRID: AB_10861262), and developed by DAB (Servicebio, Wuhan, China) for 1–15 min. The slides were photographed with an optical microscope (Olympus, BX53M).
Bladder histology
Mouse bladders were collected and processed for standard H&E staining. The slides were photographed with an optical microscope (Olympus, BX53M). Bladder inflammation status was scored by two pathologists in a blinded manner based on infiltration of inflammatory cells in the lamina propria and the presence of interstitial edema as previously described (+: mild infiltration with no or mild edema, ++: moderate infiltration with moderate edema, and +++: moderate to severe infiltration with severe edema) (40).
Statistical analysis
All statistical tests were performed using the open-source statistics package R or using GraphPad Prism software 8 (San Diego, CA; GraphPad Prism, RRID:SCR_002798). Data are presented as the mean ± SD or SEM. Differences were considered statistically significant at p < 0.05. Normality and equal variances between group samples were assessed using the Shapiro-Wilk test and Brown–Forsythe tests, respectively. When normality and equal variance were achieved between sample groups, one-way ANOVA (followed by Bonferroni’s multiple comparisons test), two-way ANOVA (followed by Bonferroni’s multiple comparisons test) or t-tests were used. Where normality or equal variance of samples failed, Kruskal–Wallis one-way ANOVA (followed by Dunn’s correction) or Mann–Whitney U tests were performed.
Data availability
All data are available within the article and supplementary materials.
RESULTS
Identification of chemically modified aptamers that selectively internalize into BC cells
To specifically target drugs to bladder cancer cells, we tested if AS1411, a widely used cancer cell-targeting aptamer, could selectively recognize bladder cancer cells in vitro. After incubation, AS1411 accumulated similarly in both T24 human bladder cancer cells and immortalized SV-huc-1 human bladder urothelial cells (Supplementary Fig. 1A). Hence, we sought to identify novel BC cell-specific nucleic acid aptamers using the cell internalization SELEX approach. Briefly, a DNA template library was first synthesized, which contained a 40-nucleotide randomized region flanked by two constant primer-binding sites (Supplementary Table 1). A chemically modified nucleic acid library containing 1014 distinct sequences was then generated by a primer extension reaction on the DNA template library using 2’-modified nucleotide substrates (2’-deoxy purines and 2’-fluoro pyrimidines, dRfY) and an engineered Taq DNA polymerase variant, SFM4–3. Such modifications endow nucleic acid sequences with resistance to nuclease digestion and thus promote stability (41).
In each round (R1-R14) of cell internalization SELEX, the library was first incubated in vitro with SV-huc-1 normal bladder urothelial cells for negative selection to remove sequences that can internalize into SV-huc-1 cells. In the subsequent positive selection step, the nucleic acid pool was incubated with T24 bladder cancer cells, followed by extensive washing and nuclease treatment to remove the weak binders. Only sequences that were internalized into T24 cells were recovered, reverse transcribed and amplified to initiate the next round of selection (Fig. 1A). To enrich for aptamers with high binding affinities and fast binding kinetics, we gradually increased the selection stringency by incrementally reducing target cell number and incubation time during the selection process (Supplementary Fig. 1B). To monitor for aptamer enrichment, nucleic acid pools from select rounds were amplified with Cy5.5-labeled primers, incubated with T24 cells and analyzed by flow cytometry. The results showed a clear fluorescence shift with selection, indicating progressive enrichment of high-affinity aptamers (Supplementary Fig. 1C).
Since fluorescence signals plateaued after 14 rounds of selection (Supplementary Fig. 1C), the R14 library was cloned and Sanger-sequenced, 48 unique aptamer sequences were identified (Supplementary Table 2), and homology analysis revealed four major families. All families were dominated by guanine-rich sequences and these guanine residues seemed highly conserved (Supplementary Fig. 1D). Nine sequences were predicted to form distinct complex secondary structures by sfold (42) and were thus chosen for flow cytometry analysis (Supplementary Fig. 1E). Of these, aptamer clone 26 (C26) was the most effective in entering T24 cells and resulted in a higher fluorescence signal than others (Supplementary Fig. 1F). To confirm this cellular internalization was due to specific tertiary structure of aptamer C26, the naive nucleic acid library (R0) was labeled with fluorescence and used as a control. As expected, the R0 library showed minimal internalization into T24 cells in both flow cytometry and confocal microscopy analysis compared with C26 (Supplementary Fig. 2A–C). C26 had minimal internalization into SV-huc-1 cells yet was readily internalized into 5637 and KU-7 human bladder cancer cell lines (Supplementary Fig. 2D–E) suggesting its selectivity is not limited to the T24 discovery cell line.
Solid phase synthesis of the 87-nucleotide C26 is expensive and technically challenging. Moreover, C26 might contain nonessential regions that could be removed without compromising its BC cell affinity and selectivity. We therefore constructed two truncated versions of C26, C26.1 contained two stem-loop motifs and C26.2 retained only one stem-loop motif (Supplementary Fig. 2F). Both versions have improved internalization into T24 compared with full-length C26 (Supplementary Fig. 2G). C26.2, with only 35 nucleotides, was the most potent in internalizing bladder cancer cells, and was renamed as aptamer B1 and utilized in all subsequent studies (Fig. 1B).
Characterization of aptamer B1 internalization into BC cells
To verify that B1 maintained BC cell selectivity, we prepared Cy5.5-labeled aptamer B1 and evaluated its interaction with tumor and normal cells. After T24 cells were incubated with B1, flow cytometry analysis showed a concentration-dependent increase in fluorescence intensity, whereas incubation with an all-DNA version or a scrambled sequence of B1 resulted in minimal fluorescence increase (Fig. 1C, Supplementary Fig. 3A), suggesting that B1 internalization into T24 cells was dependent on its discrete tertiary structure, which in turn relied on both the nucleic acid sequence and specific chemical modifications. The internalization of B1 in T24, KU-7 and SV-huc-1 cells were also tested using confocal microscopy. After 1-h incubation, B1 was internalized into T24 and KU-7 cells, but not SV-huc-1 cells (Fig. 1D). Additionally, to test the universality of B1 for bladder cancer, we assessed B1 internalization into a panel of bladder cancer cell lines including TCCSUP, J82, SW780, UMUC3, BIU87 and RT4, B1 could readily internalize into all these lines (Supplementary Fig. 3B–G). We further compared the internalization capability of B1 between primary human tumor cells and normal urothelial cells, harvested from NMIBC patients. Freshly collected bladder tumor tissue or normal para-tumor urothelium tissue was digested into single cells, incubated with Cy5.5-labeled aptamer B1 for 1 h and analyzed with flow cytometry. Incubation of primary human bladder tumor cells with Cy5.5-labeled B1 resulted in a dose-dependent increase in fluorescence intensity which plateaued around 800 nM. In contrast, much weaker fluorescence signal was detected in cells derived from the histologically normal tissues from NMIBC patients (Fig. 1E, Supplementary Fig. 3H–I). These results demonstrated that B1 could be easily prepared using standard phosphoramidite chemistry and such preparation maintained its bladder cancer-specific internalization.
Mechanistic investigation of aptamer B1 internalization into BC cells
To explore the mechanism of B1 internalization into BC cells, Cy5.5-labeled aptamer B1 was incubated with T24 cells and analyzed by confocal microscopy. B1 accumulated primarily in the cytosol and partially co-localized with lysosomes, implying that a fraction of aptamer B1 escapes lysosomal degradation (Fig. 2A). To identify the potential endocytic pathways involved in B1 internalization, we used five different endocytosis inhibitors to pre-treat T24 cells before incubation with B1 and monitored fluorescence change via flow cytometry. Three of the five inhibitors, chlorpromazine, dynasore and amiloride partially reduced aptamer B1 internalization (Fig. 2B and C, Supplementary Fig. 4A). Based on the known functions of these inhibitors (see Methods), these results suggest that clathrin-mediated endocytosis and macropinocytosis are the major pathways involved in B1 internalization into T24 cells.
In order to identify the potential molecular targets involved in B1 internalization, we treated T24 cells with proteinase K or trypsin prior to incubation with B1. Such treatment reduced B1 internalization in both cases, implying that cell membrane protein targets were likely involved in the internalization process (Supplementary Fig. 4B). Next, we attempted to isolate such protein targets in an affinity pull-down assay using biotin-labeled B1 as a bait. A distinct protein band with an apparent molecular weight around 100 kDa was clearly enriched in the B1 treatment group (Fig. 2D). Subsequent mass spectrometry analysis identified AP2B1 as a candidate protein. AP2B1 is the beta subunit component of the adaptor protein complex 2 (AP2) which is involved in clathrin-dependent endocytosis. The canonical isoform of AP2B1 contains 937 amino acid residues and has a molecular weight of 104 kDa. The binding between B1 and AP2B1 was then confirmed by a pull-down assay using biotin-labeled B1 followed by immunoblotting with anti-AP2B1 antibody (Supplementary Fig. 4C). We further knocked down AP2B1 expression using siRNA and observed a reduced internalization of B1 into T24 cells (Fig. 2E and Supplementary Fig. 4D). These results suggested that the AP2B1 protein, and the associated clathrin-mediated endocytosis, played an important role in aptamer B1 internalization. Importantly, a comparison of AP2B1 expression between primary human tumor tissue and normal urothelial tissue harvested from NMIBC patients showed that AP2B1 is highly expressed in tumor tissues but not normal tissues (Supplementary Fig. 4E), providing a possible explanation for the selectivity of B1 towards bladder tumor cells.
Construction of B1-tethered DNA nanotrain for BC targeted drug delivery
Given that aptamer B1 possessed selective internalization capabilities towards BC cells, we reasoned that B1 could be used to construct a targeted drug delivery system for intravesical treatment of BC. Thus, an aptamer-tethered DNA nanotrain (NT) design originally devised by Tan et. al (43) was employed (Fig. 3A). Briefly, B1 was conjugated to a DNA trigger probe sequence, which could initiate a hybridization chain reaction of two monomer strands M1 and M2. In the absence of B1 trigger, the two monomers formed individual intramolecular hairpin structures. In the presence of B1 trigger, the hairpin structures were opened. M1 and M2 were hybridized to form a long double helical structure. Since such a long dsDNA could serve as boxcars for loading chemotherapeutic drugs, this aptamer-guided drug delivery system could potentially induce selective cytotoxicity in BC cells. The nanotrain assembly process was first analyzed on agarose gel electrophoresis, and the results showed that the nanotrain structure could only form when M1, M2 and B1 trigger strands were all present (Fig. 3B). Additionally, with increasing equivalents of M1 and M2, nanotrains with longer boxcars could be constructed (Supplementary Fig. 5A). We chose a B1 trigger to each monomer ratio of 1:10, taking into account of issues associated with nanotrain assembly and stability, drug loading capacity and cell entry efficiency (43). To verify that the nanotrain retained the targeting selectivity of B1, we constructed a nanotrain using FAM-labeled M1 and M2 strands and showed that B1-nanotrain could selectively internalize into T24 bladder cancer cells but not SV-huc-1 normal urothelial cells, and that the observed cell selectivity was afforded by B1, since nanotrain with a scrambled aptamer sequence could not effectively discriminate T24 and SV-huc-1 cells (Fig. 3C and 3D).
Next, we evaluated the drug loading and release capabilities of B1 DNA-nanotrain. Epirubicin (EPI) was chosen because it is an anthracycline chemotherapeutic drug commonly used in intravesical bladder cancer treatment (44). The intrinsic fluorescence of epirubicin and fluorescence quenching upon intercalation into duplex DNA enabled the in vitro monitoring of drug loading and intracellular imaging of drug release (Fig. 4A). The nanotrain boxcar region was designed to be GC-rich for maximal epirubicin intercalation. Successful drug loading was observed after epirubicin was incubated with nanotrains with varying boxcar lengths. The intrinsic fluorescence intensity of EPI decreased gradually with an increasing molar ratio equivalent of M1/M2 strands (Fig. 4A, 1:160 means one B1-guided nanotrain was loaded with 160 EPI molecules, assuming that each M1/M2 pair could accommodate 16 EPI molecules), indicating drug intercalation into dsDNA. Such drug-loaded nanotrain was then assessed for its stability and payload release capacity. Ideally, the drug-loaded nanotrain should remain stable under physiological conditions and start to release the drug inside the cytoplasm of target cancer cells. We first monitored epirubicin release from nanotrain in phosphate-buffered saline (PBS) using a dialysis assay. EPI itself can freely diffuse across the dialysis membrane but once intercalated into the nanotrain, the complex should be trapped inside the inner chamber. The fluorescence change in the exterior chamber could therefore reflect nanotrain stability. The results showed that EPI-loaded nanotrains (NT-EPI) remained intact for the duration of experiment (48 hours), while EPI itself diffused freely to the exterior chamber (Fig. 4B). Next, EPI release from the nanotrain was also monitored under simulated cytoplasmic and lysosomal conditions. In the presence of DNase I, the DNA nanotrain was readily digested and EPI was released, restoring its fluorescence intensity (Fig. 4C). Similarly, at pH 4.5, but not 5.5 or 7.5, the DNA nanotrain was degraded and epirubicin was effectively unloaded (Supplementary Fig. 5B). These results suggest that EPI-loaded DNA nanotrain is stable under mild conditions and could release its payload under simulated cytoplasmic and lysosomal conditions.
Selective cytotoxicity of aptamer B1 NT-EPI in BC cells
We next assessed EPI-loaded B1-guided nanotrain for its selective cell targeting and drug delivery capabilities using confocal microscopy. Fluorescence signals were observed in T24 and KU-7 cells treated with NT-EPI, but not in SV-huc-1 cells, indicating that the drug-loaded nanotrain maintained its cell-selective targeting capability (Fig. 4D). In contrast, treatment with free EPI resulted in indiscriminate drug accumulation in all lines. We further investigated the dynamics of drug uptake and release in cells treated with free EPI or EPI-loaded nanotrain, respectively. Treatment with free EPI resulted in rapid and indiscriminate uptake and accumulation of drugs in cell nucleus in all lines within 1 h (Supplementary Fig. 6A–C). In contrast, when delivered via B1-tethered nanotrain system EPI was released much slower in BC cells (Supplementary Fig. 6D–F). Fluorescence signals from both FAM-labeled nanotrain and EPI were observed 30 min after treatment, indicating that the complex had already been taken up by cells and the drug had begun to dissociate. The released EPI was initially colocalized with nanotrain in the cytoplasm, and gradually translocated into the nucleus after 2–3 h (Supplementary Fig. 6D–F). These results suggested that B1-guided DNA nanotrain could selectively deliver EPI into BC cells.
We then evaluated the cytotoxicity of this drug delivery system in cancerous and normal cells using the MTT assay. B1 NT-EPI exhibited similar cytotoxicity profile to that of free EPI in both T24 and KU-7 cells after 48 h (Fig. 4E and 4F). In contrast, B1 NT-EPI was much less toxic to SV-huc-1 compared to EPI alone (Fig. 4G). To exclude the possibility that selective cytotoxicity was due to B1 itself, we incubated KU-7, T24 and SV-huc-1 cells with B1 or a scrambled 35-nt control sequence in the absence of EPI and no growth inhibition was observed (Supplementary Fig. 7A–C). These data demonstrate that the B1 NT-EPI complex is more toxic against BC cells in vitro.
Inhibition of tumor growth by B1 NT-EPI in orthotopic xenograft models of human bladder cancer
Given the above results, we next evaluated the anti-tumor effect of NT-EPI in vivo. An orthotopic xenograft bladder cancer model was established by intravesical implantation of the luciferase-labeled KU-7 cells in nude mice (45). After 14 days, mice with positive bioluminescence were randomly divided into three groups of 9 mice for different treatment regimens: free EPI (0.4 mg/ml, 50 μl/instillation), aptamer B1 NT-EPI (EPI concentration equivalent to 0.4 mg/ml) and PBS control. Mice were instilled every 5 days for 7 cycles and tumor growth was measured by luminescence at the end of treatment (Fig. 5A). In the PBS group, all mice had bladder tumors; in the free EPI group 3 mice still had bladder tumors; in the NT-EPI treatment group, 8 mice were tumor-free with one showing minimal signal (Fig. 5B and C). The therapeutic effect of NT-EPI was also confirmed by H&E staining of mice bladder from different groups (Fig. 5D and Supplementary Fig. 7D. Notably, compared to the free EPI group which showed slight decrease in body weight indicating toxicity, no body weight loss was observed in the NT-EPI group during the treatment period (Supplementary Fig. 7E). Most importantly, EPI-induced cystitis, which was apparent in the free EPI group, was essentially absent in the NT-EPI group suggesting reduced damage to normal urothelium (Fig. 5D and Supplementary Fig. 7F). These data demonstrated both the efficacy and safety of the B1 NT-EPI.
DISCUSSION
Aptamers have been extensively explored for targeted therapy of many diseases, either as an agonist/antagonist on their own or as a targeting moiety for other drugs (46). Despite numerous preclinical studies showing efficacy of aptamer-based therapy, only one aptamer drug has been approved by the FDA in the past 30 years. Pegaptanib (commercial name Macugen) was approved in 2004 for the treatment of age-related macular degeneration (AMD) via direct targeting of VEGF in the eye (47). A key obstacle for the translation of aptamer drugs into the clinic is the poor pharmacokinetics and pharmacodynamics in vivo. Like siRNA drugs, unprotected aptamers are degraded in the circulation or bind to serum proteins. They also have multiple biological barriers to cross to reach the desired organ and recognize their specific targets. The intravesical approach to bladder cancer therapy represents a unique opportunity for the clinical translation of aptamer-based therapy for several reasons: aptamer-based agents can be delivered in high concentrations, in a relatively neutral carrier (saline) into a minimally hostile environment (urine) and make direct contact with the tumor cells. To our knowledge, our work represents the first bladder cancer-specific aptamer and the first example of effective aptamer-based therapy in an intravesical setting.
Efforts have been made to improve the effectiveness of intravesical chemotherapy for NMIBC (48). The major strategies include increasing intravesical dwell time or increasing drug uptake. For example, using carriers that allow localized diffusion of drug over time, a slow-releasing intravesical version of gemcitabine could increase the dwell period of gemcitabine up to 21 days (49). Mixing mitomycin C or gemcitabine in hydrogels to achieve sustained release have shown promise in phase II trials (50,51). To increase drug uptake by cancer cells following intravesical administration, albumin-bound nanoparticles were tested to deliver paclitaxel (52) or rapamycin (53). Photodynamic therapy to increase drug uptake is also being tested in trials (54,55). However, none of these strategies discriminates between BC cells and normal urothelial cells which is a central feature of our aptamer nanotrain delivery system.
Although both the T24 cell line used for aptamer discovery and the KU-7 cell line used in mouse orthotopic models are of NMIBC origin, their biological behaviors are quite different (27,56) . Moreover, we found that aptamer B1 could be internalized into a wide range of bladder cancer cell lines with different history and biological behavior (the detailed characteristics of the cell lines used in this study are listed in Supplementary Table 3). We reason that having a broad effectiveness against urothelial cancers is a strength of our approach given the recently clearly documented tumor heterogeneity that is present in human bladder cancer (57).
Mitomycin C (MMC), gemcitabine and epirubicin are the three drugs commonly used for bladder cancer intravesical chemotherapy. Here we used epirubicin as a proof-of-principle for the aptamer technology for two reasons. First, epirubicin has very strong auto-fluorescence that disappears when intercalating into the DNA and reappears upon release, making it a perfect “reporter” cargo for the molecular characterization of the aptamer-based drug delivery system. Second, compared with MMC or gemcitabine, epirubicin is not very effective at the doses that it is used currently in the clinic, although it has a very good side effect profile (58–61). The coupling with aptamer B1 to increase its specificity therefore allowing higher doses to be used should improve its efficacy without increasing its side effects. Therefore, our approach may alter the indications for use of epirubicin in NMIBC. Importantly, our technology is “cargo agnostic” so other agents such as MMC or gemcitabine can also be used as cargo in this system when desired.
Our work has some limitations. First, only a relatively small number of putative aptamer sequences were retrieved by Sanger sequencing. It is likely that more aptamer sequences, potentially with higher affinity and selectivity, could be identified by deep sequencing of the R14 library. Second, although we showed that B1 preferentially internalized BC cells but not normal urothelial cells, a larger number of cell lines and human samples should be examined. Third, although we implicated specific endocytosis pathways used by B1 and its putative target AP2B1, the latter may not be the direct target of B1. This is because AP2B1 is found on the cytoplasmic face of the plasma membrane so other mechanisms that would place the B1 on this protein might be involved and warrant further investigation. Although we are only using B1 as a delivery vehicle, for preclinical development and clinical use its protein target needs to be thoroughly investigated.
In conclusion, our aptamer-based drug delivery system provides the conceptual framework and ideal toolbox for development of novel targeted intravesical chemotherapy for BC. This unique approach provides compelling preclinical data that merits further translational development.
Supplementary Material
1
ACKNOWLEDGEMENTS
Funding:
National Key Research and Development Program of China 2019YFA0904000 (H. Yu)
National Natural Science Foundation of China 21877060 (C. Yan)
National Natural Science Foundation of China 21977046 (H. Yu.)
The Fundamental Research Funds for the Central Universities 0213–14380192, 0213–14380181 (H. Yu)
The Fundamental Research Funds for the Central Universities 0211–14380173 (C. Yan)
The Program for Innovative Talents and Entrepreneur in Jiangsu Province (H. Yu)
National Institutes of Health grant R01CA075115 (D. Theodorescu)
Fig. 1. Selection of bladder cancer cell-specific aptamers by cell internalization SELEX.
(A) Selection scheme of chemically modified RNA aptamers that can internalize into T24 human bladder cancer cells but not SV-huc-1 normal bladder urothelial cells. R: purine. Y: pyrimidine. (B) Predicted secondary structure of aptamer B1 using sfold. (C) Binding curve of aptamer B1 with T24 cells. An all-DNA version of B1 and a B1-derived scrambled 35-nt sequence were used as control. Representative data of three independent experiments are shown. (D) Confocal microscopy analysis showed that aptamer B1 internalization is cell type-specific. T24 and KU-7 are two bladder cancer cell lines, SV-huc-1 is a normal bladder urothelial cell line. Scale bars: 40 μm. Representative data of three independent experiments are shown. (E) Comparison of the binding curves of aptamer B1 with primary human tumor cells and normal urothelial cells generated from surgical specimens harvested from a NMIBC patient. Freshly collected bladder tumor tissue or normal para-tumor urothelium tissue was digested into single cells, incubated with Cy5.5-labeled aptamer B1 for 1 h and analyzed with flow cytometry. Raw data are shown in Supplementary Fig. 3B.
Fig. 2. Mechanistic investigation of aptamer B1 internalization into T24 cell.
(A) Cells were incubated with Cy5.5-labeled aptamer B1, lysosomes and nuclei were stained with LysoTracker Green and Hoechst 33342, respectively. Scale bar: 40 μm. (B and C) Cells were first treated with an endocytosis inhibitor, and then incubated with Cy5.5-labeled aptamer B1, followed by flow cytometry analysis. Chlorpromazine inhibits clathrin-mediated endocytosis. Amiloride inhibits pinocytosis. (D) SDS-PAGE analysis of proteins recovered from pull-down experiments using biotin-labeled aptamer B1. Asterisk indicates a protein band specific to aptamer B1 treatment. (E) Knockdown of AP2B1 with siRNA reduced aptamer B1 internalization into T24 cells. Representative data of three independent experiments are shown.
Fig. 3. Design and characterization of aptamer B1-tethered DNA nanotrain.
(A) Schematic of the self-assembled aptamer-tethered DNA nanotrain for cellular delivery of drugs. Aptamer directs drug-loaded nanotrain for selective internalization into bladder cancer cells. Chemotherapeutic drugs are unloaded inside the cell and induce cytotoxicity. Drug intrinsic fluorescence can be used to monitor its intercalation into DNA duplex and intracellular release. (B) Agarose gel analysis of nanotrain assembly. Lane1: M1+M2; Lane2: M1; Lane3: M2; Lane4: aptamer-tethered; Lane5: aptamer-tethered DNA nanotrain. (C) Flow cytometry analysis of nanotrain internalization into T24 and SV-huc-1 cells. (D) Confocal microscope images of nanotrain internalization into T24 and SV-huc-1 cells. Scale bars: 40 μm. Representative data of three independent experiments are shown.
Fig. 4. Aptamer B1-tethered nanotrains stably carry and selectively deliver epirubicin (EPI) into bladder cancer cells.
(A) Characterization of the loading of epirubicin into the nanotrain. The intrinsic fluorescence of EPI decreased with increasing equivalents of nanotrain boxcar component strands, as it was quenched upon intercalation into DNA duplex. (B) Stability of the EPI-loaded nanotrain in PBS was evaluated by dialysis. Free EPI molecules diffused across the dialysis membrane and resulted in a high fluorescence intensity in the exterior chamber. Interaction of EPI with nanotrain (NT-EPI) restricted its diffusion. (C) Release of EPI in the presence of DNase. Intercalation of EPI into boxcar DNA duplex of NT resulted in fluorescence quenching. Treatment with DNase Ⅰ degraded dsDNA and restored the intrinsic fluorescence of EPI. (D) Confocal microscope images showing targeted delivery of NT-EPI into T24 and KU-7 bladder cancer cells. Cells were treated with free EPI or NT-EPI for 1 h. The nuclei were stained with DAPI. Scale bar: 40 μm. (E and F) Aptamer B1 NT-EPI exhibits similar cytotoxicity to that of free EPI against T24 (E) and KU-7 (F) bladder cancer cells. (G) Aptamer B1 NT-EPI was much less toxic to SV-huc-1 cells compared with free EPI. Cells were treated for 48 h and cell viability was measured by MTT assay. Data are mean ± SD. Representative data of three independent experiments shown.
Fig. 5. Inhibition of bladder cancer growth by B1 NT-EPI in vivo.
(A) Schematic of orthotopic xenograft model establishment and treatment regimen. Luciferase-labeled KU-7 human bladder cancer cells were implanted into the bladder of nude mice. Fourteen days later, mice with positive bladder fluorescence signal were randomly divided into three groups and treated with intravesical instillation of PBS, free EPI (0.4 mg/ml, 50 μl/instillation) or aptamer B1 NT-EPI (equal molar EPI molecules). (B) Luminescence imaging and (C) quantitative analysis of tumors in mouse bladder at the end of treatment. Data are mean ± SD (n = 9 mice/group). ** P < 0.01. (D) Representative H&E staining images of bladder tumor sections from each group. Sections of whole bladder was shown in the middle (scale bar: 500 μm), with magnification on both sides (scale bar: 100 μm). The NT-EPI group showed much less cystitis compared to free EPI, as indicated by reduced infiltration of inflammatory cells and interstitial edema, suggesting reduced damage to normal urothelium. Quantification of all nine mice in each group are summarized in Supplementary Fig. 6.
Significance statement:
These findings identify a bladder cancer-specific aptamer that can be used for targeted delivery of chemotherapy, potentially reducing toxicity and enhancing therapeutic efficacy.
Conflict of Interest Statement: Y.W., Y.Z. H.Y. and C.Y. are beneficiaries of a pending patent for the aptamers and delivery vehicles described in this manuscript. All other authors declare no potential conflicts of interest.
References
1. Bray F , Ferlay J , Soerjomataram I , Siegel RL , Torre LA , Jemal A . Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2018;68 :394–424 30207593
2. Miller KD , Nogueira L , Mariotto AB , Rowland JH , Yabroff KR , Alfano CM , Cancer treatment and survivorship statistics, 2019. CA Cancer J Clin 2019;69 :363–85 31184787
3. Kamat AM , Hahn NM , Efstathiou JA , Lerner SP , Malmström P-U , Choi W , Bladder cancer. The Lancet 2016;388 :2796–810
4. Babjuk M , Burger M , Comperat EM , Gontero P , Mostafid AH , Palou J , European Association of Urology Guidelines on Non-muscle-invasive Bladder Cancer (TaT1 and Carcinoma In Situ)-2019 Update. Eur Urol 2019;76 :639–57 31443960
5. Sathianathen NJ , Regmi S , Gupta S , Konety BR . Immuno-Oncology Approaches to Salvage Treatment for Non-muscle invasive Bladder Cancer. Urol Clin North Am 2020;47 :103–10 31757293
6. van Rhijn BWG , Burger M , Lotan Y , Solsona E , Stief CG , Sylvester RJ , Recurrence and progression of disease in non–muscle-invasive bladder cancer: from epidemiology to treatment strategy. Eur Urol 2009;56 :430–42 19576682
7. Vlachostergios PJ , Jakubowski CD , Niaz MJ , Lee A , Thomas C , Hackett AL , Antibody-Drug Conjugates in Bladder Cancer. Bladder cancer 2018;4 :247–59 30112436
8. Coats S , Williams M , Kebble B , Dixit R , Tseng L , Yao NS , Antibody-Drug Conjugates: Future Directions in Clinical and Translational Strategies to Improve the Therapeutic Index. Clin Cancer Res 2019;25 :5441–8 30979742
9. Hansel TT , Kropshofer H , Singer T , Mitchell JA , George AJ . The safety and side effects of monoclonal antibodies. Nat Rev Drug Discov 2010;9 :325–38 20305665
10. Wilson DS , Szostak JW . In vitro selection of functional nucleic acids. Annu Rev Biochem 1999;68 :611–47 10872462
11. Dunn MR , Jimenez RM , Chaput JC . Analysis of aptamer discovery and technology. Nat Rev Chem 2017;1 :0076
12. Rothlisberger P , Hollenstein M . Aptamer chemistry. Adv Drug Deliv Rev 2018;134 :3–21 29626546
13. Ellingtion AD , Szostak JW . In vitro selection of RNA molecules that bind specific ligands. Nature 1990;346 :818–22 1697402
14. Tuerk C , Gold L . Systematic evolution of ligands by exponential enrichment: RNA ligands to bacteriophage T4 DNA polymerase. Science 1990;249 :505–10 2200121
15. Zhou J , Rossi J . Aptamers as targeted therapeutics: current potential and challenges. Nat Rev Drug Discov 2017;16 :181–202 27807347
16. Sun H , Tan W , Zu Y . Aptamers: versatile molecular recognition probes for cancer detection. Analyst 2016;141 :403–15 26618445
17. Yan A , Levy M . Cell internalization SELEX: in vitro selection for molecules that internalize into cells. Methods Mol Biol 2014;1103 :241–65 24318899
18. Rodríguez-Dorantes M , Cortés-Ramírez SA , Cruz-Burgos JM , Reyes-Grajeda JP , Losada-García A , González-Covarrubias V , Cell-Internalization SELEX of RNA Aptamers as a Starting Point for Prostate Cancer Research. Methods Mol Biol 2021;2174 :245–54 32813254
19. Tanaka K , Okuda T , Kasahara Y , Obika S . Base-modified aptamers obtained by cell-internalization SELEX facilitate cellular uptake of an antisense oligonucleotide. Mol Ther Nucleic Acids 2021;23 :440–9 33473329
20. Uemachi H , Kasahara Y , Tanaka K , Okuda T , Yoneda Y , Obika S . Discovery of cell-internalizing artificial nucleic acid aptamers for lung fibroblasts and targeted drug delivery. Bioorg Chem 2020;105 :104321 33074117
21. Marshall ML , Wagstaff KM . Internalized Functional DNA Aptamers as Alternative Cancer Therapies. Front Pharmacol 2020;11 :1115 32848740
22. Yoon S , Rossi JJ . Aptamers: Uptake mechanisms and intracellular applications. Adv Drug Deliv Rev 2018;134 :22–35 29981799
23. Shangguan D , Li Y , Tang ZW , Cao ZHC , Chen HW , Mallikaratchy P , Aptamers evolved from live cells as effective molecular probes for cancer study. Proc Natl Acad Sci 2006;103 :11838–43 16873550
24. Nelissen FHT , Peeters WJM , Roelofs TP , Nagelkerke A , Span PN , Heus HA . Improving Breast Cancer Treatment Specificity Using Aptamers Obtained by 3D Cell-SELEX. Pharmaceuticals-Base 2021;14 :349
25. Kruspe S , Mittelberger F , Szameit K , Hahn U . Aptamers as Drug Delivery Vehicles. Chemmedchem 2014;9 :1998–2011 25130604
26. Xiong X , Lv Y , Chen T , Zhang X , Wang K , Tan W . Nucleic acid aptamers for living cell analysis. Annu Rev Anal Chem 2014;7 :405–26
27. Zuiverloon TCM , de Jong FC , Costello JC , Theodorescu D . Systematic Review: Characteristics and Preclinical Uses of Bladder Cancer Cell Lines. Bladder cancer 2018;4 :169–83 29732388
28. Nickerson ML , Witte N , Im KM , Turan S , Owens C , Misner K , Molecular analysis of urothelial cancer cell lines for modeling tumor biology and drug response. Oncogene 2017;36 :35–46 27270441
29. Chen T , Hongdilokkul N , Liu Z , Adhikary R , Tsuen SS , Romesberg FE . Evolution of thermophilic DNA polymerases for the recognition and amplification of C2’-modified DNA. Nat Chem 2016;8 :556–62 27219699
30. Chen T , Hongdilokkul N , Liu Z , Adhikary R , Tsuen SS , Romesberg FE . Evolution of thermophilic DNA polymerases for the recognition and amplification of C2’-modified DNA. Nat Chem 2016;8 :556–62 27219699
31. Sefah K , Shangguan D , Xiong X , O’Donoghue MB , Tan W . Development of DNA aptamers using Cell-SELEX. Nat Protoc 2010;5 :1169–85 20539292
32. Yan A , Levy M . Cell internalization SELEX: in vitro selection for molecules that internalize into cells. Methods Mol Biol 2014;1103 :241–65 24318899
33. Xiao Z , Shangguan D , Cao Z , Fang X , Tan W . Cell-specific internalization study of an aptamer from whole cell selection. Chem Eur J 2008;14 :1769–75 18092308
34. Rejman J , Bragonzi A , Conese M . Role of clathrin- and caveolae-mediated endocytosis in gene transfer mediated by lipo- and polyplexes. Mol Ther 2005;12 :468–74 15963763
35. Clemente LP , Rabenau M , Tang S , Stanka J , Cors E , Stroh J , Dynasore Blocks Ferroptosis through Combined Modulation of Iron Uptake and Inhibition of Mitochondrial Respiration. Cells-Basel 2020;9 :2259
36. Vandenbeuch A , Kinnamon SC . Is the Amiloride-Sensitive Na+ Channel in Taste Cells Really ENaC? Chem Senses 2020;45 :233–4 32099995
37. Oosterhuis B , Jonkman JHG . Omeprazole - Pharmacology, Pharmacokinetics and Interactions. Digestion 1989;44 :9–17 2691315
38. Mahjoubin-Tehran M , Kovanen PT , Xu S , Jamialahmadi T , Sahebkar A . Cyclodextrins: Potential therapeutics against atherosclerosis. Pharmacol Ther 2020;214 :107620
39. Wu XQ , Liu HL , Han DM , Peng B , Zhang H , Zhang L , Elucidation and Structural Modeling of CD71 as a Molecular Target for Cell-Specific Aptamer Binding. J Am Chem Soc 2019;141 :10760–9 31185171
40. Liu W , Evanoff DP , Chen X , Luo Y . Urinary bladder epithelium antigen induces CD8+ T cell tolerance, activation, and autoimmune response. J Immunol 2007;178 :539–46 17182594
41. Pieken WA , Olsen DB , Benseler F , Aurup H , Eckstein F . Kinetic characterization of ribonuclease-resistant 2’-modified hammerhead ribozymes. Science 1991;253 :314–7 1857967
42. Ding Y , Chan CY , Lawrence CE . Sfold web server for statistical folding and rational design of nucleic acids. Nucleic Acids Res 2004;32 :W135–41 15215366
43. Zhu G , Zheng J , Song E , Donovan M , Zhang K , Liu C , Self-assembled, aptamer-tethered DNA nanotrains for targeted transport of molecular drugs in cancer theranostics. Proc Natl Acad Sci 2013;110 :7998–8003 23630258
44. Chou R , Selph S , Buckley DI , Fu R , Griffin JC , Grusing S , Intravesical Therapy for the Treatment of Nonmuscle Invasive Bladder Cancer: A Systematic Review and Meta-Analysis. J Urol 2017;197 :1189–99 28027868
45. Theodorescu D , Cornil I , Fernandez BJ , Kerbel RS . Overexpression of normal and mutated forms of HRAS induces orthotopic bladder invasion in a human transitional cell carcinoma. Proc Natl Acad Sci U S A 1990;87 :9047–51 2247480
46. He F , Wen N , Xiao D , Yan J , Xiong H , Cai S , Aptamer-Based Targeted Drug Delivery Systems: Current Potential and Challenges. Curr Med Chem 2020;27 :2189–219 30295183
47. Yu AM , Choi YH , Tu MJ . RNA Drugs and RNA Targets for Small Molecules: Principles, Progress, and Challenges. Pharmacol Rev 2020;72 :862–98 32929000
48. Tran L , Xiao JF , Agarwal N , Duex JE , Theodorescu D . Advances in bladder cancer biology and therapy. Nat Rev Cancer 2021;21 :104–21 33268841
49. Safety and Tolerability of TAR-200 mg in Subjects With Non-Muscle-Invasive Bladder Cancer. In: Medicine. UNLo, editor2016.
50. A Phase 2b Study of UGN-102 for Low Grade Intermediate Risk Non-Muscle-Invasive Bladder Cancer (OPTIMA II). 2018.
51. Safety of Pre-TURBT Intravesical Instillation of Escalating Doses of TC-3 Gel and MMC in NMIBC Patients. 2014.
52. McKiernan JM , Holder DD , Ghandour RA , Barlow LJ , Ahn JJ , Kates M , Phase II trial of intravesical nanoparticle albumin bound paclitaxel for the treatment of nonmuscle invasive urothelial carcinoma of the bladder after bacillus Calmette-Guerin treatment failure. J Urol 2014;192 :1633–8 24996128
53. Seager CM , Puzio-Kuter AM , Patel T , Jain S , Cordon-Cardo C , Mc Kiernan J , Intravesical delivery of rapamycin suppresses tumorigenesis in a mouse model of progressive bladder cancer. Cancer Prev Res 2009;2 :1008–14
54. Intravesical Photodynamic Therapy (PDT) in BCG Refractory/Intolerant Non-Muscle Invasive Bladder Cancer (NMIBC) Patients. 2019.
55. Monro S , Colon KL , Yin H , Roque J , 3rd, Konda P, Gujar S, Transition Metal Complexes and Photodynamic Therapy from a Tumor-Centered Approach: Challenges, Opportunities, and Highlights from the Development of TLD1433. Chem Rev 2019;119 :797–828 30295467
56. Gildea JJ , Golden WL , Harding MA , Theodorescu D . Genetic and phenotypic changes associated with the acquisition of tumorigenicity in human bladder cancer. Genes Chromosomes Cancer 2000;27 :252–63 10679914
57. Gouin KH 3rd , Ing N , Plummer JT , Rosser CJ , Ben Cheikh B , Oh C , An N-Cadherin 2 expressing epithelial cell subpopulation predicts response to surgery, chemotherapy and immunotherapy in bladder cancer. Nat Commun 2021;12 :4906 34385456
58. Witjes JA , Brausi M , Malmstrom PU . How to Treat Multifocal Ta High-grade Disease if Bacillus Calmette-Guerin Is Unavailable. Eur Urol Oncol 2020;3 :705–9 31324506
59. Lee S , Lim B , You D , Hong B , Hong JH , Kim CS , Association of Bacillus Calmette-Guerin shortages with bladder cancer recurrence: A single-center retrospective study. Urol Oncol 2020;38 :851 e11- e17
60. Fankhauser CD , Teoh JY , Mostafid H . Treatment options and results of adjuvant treatment in nonmuscle-invasive bladder cancer (NMIBC) during the Bacillus Calmette-Guerin shortage. Curr Opin Urol 2020;30 :365–9 32141939
61. Zamboni S , Baumeister P , Mattei A , Mordasini L , Antonelli A , Simeone C , Single postoperative instillation for non-muscle invasive bladder cancer: are there still any indication? Transl Androl and Urol 2019;8 :76–84
|
PMC008xxxxxx/PMC8983123.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
0410462
6011
Nature
Nature
Nature
0028-0836
1476-4687
34880500
8983123
10.1038/s41586-021-04137-3
NIHMS1791693
Article
A hormone complex of FABP4 and nucleoside kinases regulates islet function
Prentice Kacey J. 1
Saksi Jani 1
Robertson Lauren T. 1
Lee Grace Y. 1
Inouye Karen E. 1
Eguchi Kosei 1
Lee Alexandra 1
Cakici Ozgur 1
Otterbeck Emily 1
Cedillo Paulina 1
Achenbach Peter 2
Ziegler Anette-Gabriele 2
Calay Ediz S. 1
Engin Feyza 13
Hotamisligil Gökhan S. 14*
1 Sabri Ülker Center for Metabolic Research, Harvard T.H. Chan School of Public Health, Department of Molecular Metabolism; Boston, MA, USA
2 Institute of Diabetes Research, Helmholtz Zentrum Munchen, German Research Center for Environmental Health, Munich-Neuherberg, Germany
3 Departments of Biomolecular Chemistry and Medicine, Division of Endocrinology, Diabetes and Metabolism, University of Wisconsin-Madison School of Medicine and Public Health, Madison, WI, USA
4 Broad Institute of Harvard and MIT, Cambridge, MA, USA.
Author Contributions
K.J.P. designed and performed the in vitro and in vivo experiments, analyzed the data, prepared the figures, and wrote and revised the manuscript. F.E., K.E.I., A.L., P.C. and L.T.R. designed and performed in vivo experiments, analyzed data, and revised the manuscript. J.S., Y.L., K.E., E.O., and E.S.C. designed and performed in vitro experiments, analyzed data, and revised the manuscript. O.C. analyzed the crystal structure and generated recombinant proteins used in this study, analyzed the data, and revised the manuscript. P.A., and A-G.Z. collected and analyzed human serum samples and revised the manuscript. G.S.H. conceived, supervised and supported the project, designed experiments, interpreted results, and revised the manuscript.
* To whom correspondence should be addressed: Gökhan S. Hotamisligil, 665 Huntington Ave. SPH1-617, Boston, MA, 02115, USA, Phone: 617 432-1950, [email protected]
Correspondence and requests for materials should be addressed to G.S.H at [email protected].
26 3 2022
12 2021
08 12 2021
08 6 2022
600 7890 720726
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Summary
Liberation of energy stores from adipocytes is critical to support survival in times of energy deficit, however, uncontrolled or chronic lipolysis associated with insulin resistance and/or insulin insufficiency, disrupts metabolic homeostasis1,2. Coupled to lipolysis is the release of a recently identified hormone, fatty acid-binding protein 4 (FABP4)3. While circulating FABP4 levels have been strongly associated with cardiometabolic diseases in both preclinical models and humans4–7, no mechanism of action has yet been described8–10. Here, we show that hormonal FABP4 forms a novel functional hormone complex with Adenosine Kinase (ADK) and Nucleoside Diphosphate Kinase (NDPK) to regulate extracellular ATP and ADP levels. We identify a substantial impact of this hormone on beta-cells and given the central role of beta-cell function in both the control of lipolysis and development of diabetes, postulate that hormonal FABP4 is a key regulator of an adipose-beta-cell endocrine axis. Antibody-mediated targeting of this hormone complex improves metabolic outcomes, enhances beta-cell function, and preserves beta-cell integrity to prevent both type 1 and type 2 diabetes. Thus, the FABP4-ADK-NDPK complex, Fabtin, represents a previously unknown hormone and mechanism of action integrating energy status with the function of metabolic organs, representing a promising target against metabolic disease.
pmcFABP4 targeting enhances beta-cell mass
This project was inspired by the observation of an increase in islet number by gross examination of the pancreata of lean FABP4−/− mice in vivo (Figure 1a), with the presence of islets being confirmed by dithizone staining (Figure S1a). Detailed analysis revealed that lean FABP4−/− mice exhibited significantly higher beta-cell mass and pancreatic insulin content compared to wild type (WT) littermates (Figure 1b–d). There was not a general increase in endocrine cells, as there was no difference in glucagon positive area (Figure S1b,c). Functionally, islets isolated from FABP4−/− mice demonstrated significantly increased glucose-stimulated insulin secretion (GSIS) (Figure 1e). Importantly, FABP4 is not expressed in islet endocrine cells or the clonal beta-cell line INS1 (Figure S1d,e). Thus, this cell type provides an opportunity to examine the specific role of hormonal FABP4 and the mechanisms underlying its actions.
To further support the paracrine/endocrine role of FABP4 in vivo, we used a monoclonal antibody (CA33, herein active antibody/a-Ab), the efficacy of which was determined in diet-induced obesity (DIO) as improving metabolic parameters11. Treatment of WT DIO mice with a-Ab for three weeks reduced 6hr fasting glucose levels and improved glycemic excursion during glucose tolerance test (GTT) (Figure 1f,g). Examination of a-Ab-treated DIO mice pancreata revealed increased islet number, consistent with lean FABP4−/− mice, and a trend toward increased beta-cell mass (Figure 1h–j). Thus, hormonal FABP4 influences beta-cell mass in vivo.
Hormonal FABP4 is elevated in T1D
It is well-established that FABP4 is elevated in T2D4 and correlates with BMI. We investigated if hormonal FABP4 is also regulated in T1D, independent of adiposity and insulin resistance, contributing to beta-cell dysfunction directly. We evaluated serum from normoglycemic (NGT) individuals with or without autoantibody positivity and new-onset T1D patients (<1 year) from the BABYDIAB and DIMELLI cohorts12–14. Serum FABP4 was significantly elevated (~1.6-fold) in new-onset T1D individuals compared to both NGT groups (Figure 1k; Table S1). In a second population (BRI Cohort) of older T1D patients with various duration of disease, serum FABP4 levels significantly correlated with HbA1c (r2=0.16, p=0.005), suggesting FABP4 is associated with glycemic control (Figure 1l; Table S2,S3). To complement these analyses, we quantified serum FABP4 from WT female non-obese diabetic (NOD) mice, comparing levels one week before T1D onset and at the time of diagnosis to age-matched non-diabetic controls. FABP4 was significantly increased both shortly before and in new-onset T1D mice (Figure 1m). Thus, circulating FABP4 levels are regulated before and immediately after impairment of glycemic control, suggesting that hormonal FABP4 may have a role in beta-cell failure and pathogenesis of both T1D and T2D.
To explore the effect of antibody targeting of secreted FABP4 in T1D, we utilized WT female NOD mice matched for age, weight, blood glucose, plasma insulin, and circulating FABP4 (Figure S1f–i). Mice were treated twice weekly with a-Ab, PBS, or a control antibody of the same isotype (c-Ab) starting at 10 weeks of age, when insulitis had already been established. A-Ab significantly protected against development of T1D (Figure 1n). Among diabetic mice, a-Ab-treated animals had significantly reduced blood glucose and higher plasma insulin levels, suggesting a less severe diabetes phenotype (Figure 1o,p). This protection is similarly observed in FABP4−/− NOD mice15. Throughout the study there were no differences in blood glucose, body weight, or plasma insulin levels among non-diabetic mice (Figure S1j–l). Consistent with FABP4−/− and a-Ab treated DIO mice, non-diabetic NOD mice treated with a-Ab exhibited improved GTT, corresponding to improved GSIS both in vivo and ex vivo (Figure 1q–s), and a significant increase in islet number and beta-cell mass compared to controls (Figure 1t–v). Alpha cell area was significantly reduced with a-Ab treatment in both NOD and DIO models, suggesting a normalization of islet endocrine populations that may also contribute to improved glycemia (Figure 1w–y, S1m). Collectively, a-Ab-mediated protection against T1D and improvement of glycemic control in a DIO model of T2D are both associated with improved beta-cell mass and function.
Discovery of the FABP4-ADK-NDPK complex (Fabtin)
Despite the influence of hormonal FABP4 on beta-cell mass and function in vivo recombinant FABP4 (rFABP4) had no effect on mouse islet GSIS, even at supraphysiological doses (Figure S2a). Previous studies have also been unable to demonstrate a consistent impact of rFABP4 alone on GSIS in vitro16,17. However, acute rFABP4 administration to WT NOD mice significantly increased blood glucose and reduced plasma insulin compared to vehicle-injected controls at the peak of plasma FABP4 levels (Figure S2b–d). The discordance of in vitro and in vivo activity suggests that hormonal FABP4 may partner with other proteins to elicit its biological function. This possibility was supported by the structural properties of the FABP4-a-Ab interaction, which exhibits a weak affinity for rFABP411. FABP4 binding by a-Ab is mediated through limited interactions on the light chain (Figure S2e), unlike prototypical antibody-antigen interactions mediated by both the heavy and light chains, or predominantly the heavy chain. Thus, the heavy chain of a-Ab may remain available for additional binding partners. Such a complex would define a novel mechanism of hormone action and shed light into the diverse metabolic functions of hormonal FABP4, including its effects on beta-cells.
To identify potential FABP4 binding partner(s), we performed pulldown and mass spectrometry (MS) using serum from WT glucose intolerant DIO mice11 and matched DIO FABP4−/− controls. Comparing proteins pulled-down specifically by a-Ab in WT, but not FABP4−/− serum, we identified Adenosine Kinase (ADK) as a top hit (Figure 2a,b). This was intriguing as genetic beta-cell-specific ADK deficiency produces an islet phenotype similar to FABP4-deficiency18,19. Pulldowns comparing a-Ab with c-Ab identified another nucleoside kinase, Nucleoside Diphosphate Kinase (NDPK, also known as nm23/nme23/nme1/nme2) as the top hit (Figure 2a,c). NDPK was bound to a-Ab, but not c-Ab, in both WT and FABP4−/− serum, suggesting a-Ab may independently recognize this protein. To define the a-Ab binding site, we generated nested peptides covering NDPK-A and evaluated interaction with a-Ab by MicroScale Thermophoresis (MST) (Figure S3a). Peptides 2 and 3 had low-affinity binding with a-Ab (μM), while peptide 8 and full-length NDPK-A had comparable high affinity binding (nM) (Figure S3b). No binding was detected between a-Ab and any other peptide, thus the non-overlapping central 5 amino acids of peptide 8 (aa 76–80) is likely the primary epitope for a-Ab (Figure S3c). Examination of the crystal structure for NDPK-A (PDB: 3L7U) places the proposed binding region on the surface, amenable for a-Ab interaction (Figure S3d). Furthermore, the tertiary folding of NDPK-A places peptides 3 and 8 in close-proximity, which may facilitate binding. Together, these findings support the presence of a complex between FABP4-ADK-NDPK in vivo, which is targeted by a-Ab through interaction with both FABP4 and NDPK.
MST was also used to validate interactions between the proposed complex components. The strongest interactions were observed between FABP4 and ADK (Kd ~7nM) and ADK and NDPK (Kd ~1.8nM) (Figure 2d; Figure S4a,b). ADK, however, has a relatively weak affinity for a-Ab directly (Kd ~1.9μM; Figure 2d; Figure S4c). We also detected NDPK binding to FABP4 with an affinity of ~700nM (Figure 2d; Figure S4d) and to a-Ab with a Kd of ~150nM (Figure 2d; Figure S4e). Furthermore, using GST-tagged NDPK we were readily able to immunoprecipitate all three components of the proposed complex, as well as the complex with a-Ab (Figure 2e). Interestingly, addition of FABP4 significantly reduced the abundance of ADK bound to GST-NDPK, suggesting FABP4 may act by disrupting the NDPK-ADK interaction (Figure S4f,g). These data provide strong evidence that FABP4 forms a complex with ADK and NDPK, driven by the strong affinities of FABP4 to ADK and ADK to NDPK.
We next explored the function of this complex in vivo with an on-bead kinase assay to measure activity of endogenous NDPK and ADK using a-Ab or c-Ab conjugated beads. Both NDPK and ADK are bi-directional kinases that produce ATP or ADP depending on the substrates. Consistent with the MS, ADK activity to generate ATP was specifically observed with a-Ab but not c-Ab (Figure 2f). Furthermore, ADK activity was significantly lower in FABP4−/− serum than WT, supporting that ADK binds to FABP4, and that FABP4 regulate kinase activity in the complex (Figure 2f). To test this, we assessed the ATP and ADP generating activity of recombinant ADK (rADK) with the proposed complex components. FABP4 both alone and in combination with NDPK significantly increased ATP generation by ADK, while NDPK alone had no effect (Figure 2g, Figure S4h). There was no difference in the capacity to generate ADP, suggesting the alteration in activity is uni-directional (Figure S4i).
The on-bead serum pulldown assays also revealed NDPK activity to produce ADP only with a-Ab, but not c-Ab (Figure 2h). Furthermore, we observed increased activity in FABP4−/− serum compared to WT, despite no difference in NDPK abundance, supporting the notion that FABP4 regulates the activity of this complex. Consistent with a high affinity interaction between NDPK and ADK, addition of ADK significantly reduced the ADP-producing activity of NDPK. FABP4 also suppressed NDPK activity, but to a lesser extent, and had no further effect over addition of ADK, suggesting that ADK is the primary component regulating NDPK activity (Figure 2i; Figure S4j). There was no effect on the capacity of NDPK to produce ATP with addition of FABP4 and/or ADK, suggesting the effect is also uni-directional (Figure S4k).
Addition of a-Ab to either ADK or NDPK alone had no effect on activity (Figure S4l,m), indicating the functional importance of the entire FABP4-ADK-NDPK complex in vivo. Addition of a-Ab also had no effect on ADK activity to produce ATP compared to complex alone (Figure 2j). Conversely, a-Ab rescued the ADP-producing capacity of NDPK in the complex, while c-Ab had no effect (Figure 2k). This suggests a-Ab is preventing ADK-mediated suppression of NDPK activity. These observations allowed us to generate a model (Figure 2l) regarding the function of the FABP4-ADK-NDPK complex. Under WT conditions, the FABP4-ADK-NDPK complex has high capacity for ATP production (via ADK) and low ADP production (via NDPK). In the absence of FABP4, ADK activity is reduced, resulting in lower ATP, and inhibition of NDPK is lost, resulting in higher ADP. This relative increase in ADP compared to ATP is phenocopied by targeting the complex using a-Ab.
FABP4-ADK-NDPK impacts GSIS via P2Y1
Unlike rFABP4, which had no effect on GSIS in vitro (Figure S2a), addition of NDPK-ADK significantly enhanced GSIS, comparable to the increase observed from FABP4−/− islets (Figure 3a). Further, addition of FABP4 to NDPK-ADK significantly suppressed this response. This was recapitulated in both primary human islets (Figure 3b) and INS1 cells (Figure 3c). Addition of any of the proposed complex components or antibodies alone had no effect (Figure 3b, S5a). In both human islets and INS1 cells, a-Ab rescued FABP4-ADK-NDPK GSIS to levels observed with NDPK-ADK alone, while c-Ab had no effect (Figure 3b,c). This consistent response, in mouse, rat, and human beta-cells supports the conserved relevance of this complex.
At least some functions of intracellular FABP4 have been attributed to its lipid-binding capacity20. Importantly, activity of FABP4 to inhibit GSIS was not dependent on lipid binding, as both a lipid binding mutant (LBM) FABP4, and FABP4 treated with the inhibitor BMS-309403, which competes with fatty acid interactions, exhibited the same effect on suppressing NDPK-ADK-potentiated GSIS as WT FABP4 (Figure S5b). Therefore, hormonal FABP4 exhibits a distinct biology, likely independent of its intracellular lipid-binding function as determined by these measures.
Extracellular ATP and ADP signal through cell-surface purinergic G-protein coupled receptors (GPCRs). The predominant receptor on beta-cells is P2Y1, which is potently agonized by ADP and antagonized by ATP, and regulates cAMP, Ca2+ fluxes and insulin secretion in both rodents and humans21,22 (outlined in Figure 3d). Importantly, all substrates required for NDPK and ADK, including ATP, ADP, GTP, and GDP, are released from beta-cells within insulin granules23. Endogenous extracellular ATP and ADP regulate GSIS, as made clear upon apyrase treatment, which selectively degrade extracellular ATP (high ratio; producing excess ADP), or ATP and ADP (low ratio; producing excess AMP). Increasing extracellular ADP while eliminating ATP (high ratio) potently enhanced GSIS from WT mouse islets, while degradation of both ATP and ADP (low ratio) significantly diminished GSIS (Figure 3e). To determine if modulating extracellular nucleoside concentrations could influence complex activity on beta-cell function, we supplemented the complex with ATP or ADP. Addition of exogenous ATP blocked the increase in GSIS by NDPK-ADK to levels observed with FABP4-ADK-NDPK (Figure 3f), consistent with previous findings24. This inhibition was more pronounced with ATPγS, a non-metabolizable ATP analog, in-line with metabolism of ATP being part of the complex activity. Conversely, ADP supplementation potentiated GSIS in the presence of the FABP4-ADK-NDPK complex to levels observed with NDPK-ADK alone (Figure 3g).
To elucidate the role of P2Y1 receptors in complex activity, we utilized a potent and selective inhibitor, the nucleotide analog MRS217925. ADP-stimulated GSIS was completely inhibited by MRS2179, confirming inhibitor activity (Figure 3h). Treatment with MRS2179 abolished NDPK-ADK potentiation of GSIS, with no effect in the presence of FABP4-ADK-NDPK, consistent with receptor inhibition by ATP under these conditions (Figure 3i). To confirm that kinase activity of both NDPK and ADK are required to modulate GSIS, we generated a kinase-dead H118N NDPK-A mutant that exhibited no capacity to generate ADP (Figure 3j). Treatment with H118N NDPK and WT ADK had no effect on GSIS, confirming requirement of NDPK kinase activity (Figure 3k). We modeled that GSIS inhibition by the FABP4-ADK-NDPK complex is dependent on the production of ATP by ADK, as activated by FABP4. To test this, we utilized an ADK inhibitor to block the ability of ADK to produce ATP (Figure 3l). Pre-treatment of the complex with the ADK inhibitor improved GSIS equivalent to NDPK-ADK alone, indicating that ATP production by ADK mediates this activity (Figure 3m). Quantification of INS1 supernatant nucleoside abundance also supported this model. NDPK-ADK increased the ADP/ATP ratio, with a corresponding increase in the GTP/GDP ratio, suggesting GDP as the acceptor substrate for an NDPK-dependent reaction (Table S4). Consistently, H118N NDPK with WT ADK failed to alter the ADP/ATP or GTP/GDP ratio. Conversely, addition of ADK inhibitor to the NDPK-ADK-FABP4 complex increased the extracellular ADP/ATP and GTP/GDP ratios, suggesting ADK modulates NDPK activity. Thus, we propose that in the absence of FABP4, or with a-Ab, NDPK increases production of extracellular ADP, resulting purinergic receptor activation and increased GSIS (Figure 3d). When FABP4-ADK-NDPK are in complex, there is a decrease in the extracellular ADP/ATP ratio, which inhibits purinergic receptors, reducing GSIS.
FABP4-ADK-NDPK alters calcium dynamics
P2Y1 activates two downstream signaling pathways, phospholipase C (PLC), which promotes IP3 generation, and Gi-mediated inhibition of cAMP generation21,26. These directly influence Ca2+ flux, which regulates both GSIS and survival. Addition of NDPK-ADK reduced pPKA substrate phosphorylation downstream of cAMP, while FABP4-ADK-NDPK enhanced the response (Figure 4a). Further, INS1 cells treated with FABP4-ADK-NDPK exhibit increased phosphorylation of the endoplasmic reticulum (ER) Ca2+ efflux transporter IP3R (Figure 4b).
Consistent with reduced GSIS, acute treatment with the FABP4-ADK-NDPK complex significantly decreased glucose-induced extracellular Ca2+ influx compared to control or NDPK-ADK (Figure 4c,d). Addition of NDPK-ADK increased Ca2+ influx in-line with enhanced GSIS. Activation of IP3R with FABP4-ADK-NDPK treatment suggests there may be increased ER Ca2+ efflux into the cytosol. To evaluate this, we performed experiments in the absence of extracellular Ca2+, such that changes in Ca2+ could only be sourced from the ER. Upon addition of Thapsigargin (Tg), a sarco/endoplasmic reticulum Ca2+ ATPase (SERCA) inhibitor that prevents Ca2+ uptake from the cytosol into the ER, ER Ca2+ efflux was significantly enhanced in cells with FABP4-ADK-NDPK compared to controls (Figure 4e,f). In both cases, co-treatment with a-Ab was able to rescue Ca2+ flux to control levels. Co-treatment with a specific P2Y1 receptor agonist MRS2365 (an ADP mimetic) rescued both Ca2+ influx (Figure S6a,b) and ER Ca2+ efflux to control levels (Figure S6c,d). The FABP4-ADK-NDPK-induced alteration in ER Ca2+ efflux was also rescued by co-treatment with NKY80, an adenylyl cyclase inhibitor which blocks cAMP formation (Figure S6e). Altogether, FABP4-ADK-NDPK signals through P2Y1 to modulate cAMP, resulting in alterations in Ca2+ flux and beta-cell dysfunction.
ER Ca2+ homeostasis is essential for a variety of processes and reduced ER Ca2+ can induce ER stress, impair metabolic processes and compromise survival27,28. We observed a significant increase in ER stress marker CHOP with FABP4-ADK-NDPK treatment, as well as increased cleaved caspase 3 (CC3) and pJNK, indicating that chronic exposure to this complex induces beta-cell stress and death (Figure 4g; Figure S6f). In the pathogenesis of diabetes, beta-cells are exposed to numerous stressors including ER dysfunction, lipotoxicity or cytotoxic stress from invading immune cells, which potentiate beta-cell apoptosis. Upon stimulation with FABP4-ADK-NDPK and ER stress-inducer Tg, there was a significant increase in Chop and Bip (Grp78), which were rescued to control levels with a-Ab (Figure 4h, S6g). This corresponded to higher CC3/7 activity, and potentiated cell death (Figure S6h). Additionally, incubation with a cytokine cocktail mimicking immune infiltration in T1D resulted in significantly enhanced CC3/7 activity in the presence of the complex compared to cytokines alone (Figure 4i). Thus, the FABP4-ADK-NDPK complex alters ER Ca2+ homeostasis, resulting in ER dysfunction, increased sensitivity to environmental stress, and potentiation of beta-cell death in vitro. These mechanisms are critical in both T1D and T2D pathogenesis.
Blocking the FABP4-ADK-NDPK complex preserves beta-cell mass
Given our findings that targeting FABP4-ADK-NDPK with a-Ab improves beta-cell stress resistance and functionality in vitro, we wanted to confirm that a-Ab treatment reduces beta-cell death in vivo. We observed significant increases in islet number and beta-cell mass with a-Ab treatment in both NOD and DIO models (Figure 1h–j, t-v). The conserved phenotype suggests a direct effect on beta-cells, as there is limited immune influence on beta-cell mass in DIO mice. This is supported by the lack of significant impact on the immune cell profile in pancreata of non-diabetic a-Ab-treated NOD mice (Figure S6i–o, Table S5). To first investigate the contribution of proliferation to the increase in beta-cell mass, NOD mice were given BrdU along with a-Ab, c-Ab, or PBS for 5 weeks beginning at 10 weeks of age. There was no significant difference in the number of BrdU+ beta-cells between a-Ab and c-Ab (Figure 4j,k). In the DIO model, however, three weeks of a-Ab treatment was associated with a small but significant increase in beta-cell proliferation, as assessed by Ki67 (Figure 4l,m). This is likely associated with the pro-proliferative environment of DIO as compared to NOD. Conversely, and consistent with our in vitro findings, we observed a profound reduction in CC3+ beta-cells upon a-Ab treatment compared to controls in both the NOD and DIO models (Figure 4n–q). Thus, targeting the FABP4-ADK-NDPK complex with a-Ab preserves beta-cell mass and enhances beta-cell function to protect against diabetes in multiple models.
Discussion
Hormones are traditionally investigated as solitary entities, with a single hormone released to act through a defined receptor. Here we identified a novel mechanism of hormone biology, wherein circulating FABP4 forms a functional complex with two extracellular nucleoside kinases, NDPK and ADK, to mediate its biological activity. This activity is independent of a defined receptor and is instead regulated through metabolite signaling via purinergic receptors to produce downstream effects. This unique mechanism of action allows for FABP4, the only known hormone to be released from adipose tissue upon stimulation of lipolysis, to have a diverse activity profile, coupling energy fluxes with metabolic response, and establishing a novel endocrine axis of metabolic regulation. We suggest that the FABP4-ADK-NDPK complex likely targets other tissues to influence multiple immunometabolic pathologies where FABP4 is dysregulated and/or purinergic receptor activity is relevant. These may include inflammation, insulin resistance, and various cardiac pathophysiologies that are linked to FABP4 and relevant for purinergic signaling29,30. For example, FABP4-ADK-NDPK complex likely acts on hepatocytes to regulate glucose production in obesity, an activity previously attributed to a-Ab treatment11. We propose the name Fabtin for this new hormone complex formed by circulating FABP4-ADK-NDPK to indicate its unique constitution.
Supplementary Material
1
Acknowledgements
The authors thank members of the Hotamisligil laboratory and the Sabri Ülker Center, past and present, for their contributions to our understanding of FABP4 and their helpful discussions. We thank the Islet Core and Clinical Islet Laboratory (Alberta Islet Distribution Program, University of Alberta) for providing us with human islets from review board-approved donors. UCB generated and purified antibodies used in this study and independently reproduced the interaction data between NDPK, FABP4, and a-Ab. The Hotamisligil lab is supported by grants from the National Institutes of Health (NIH DK123458) and Juvenile Diabetes Research Foundation (JDRF; 2-SRA-2019-660-S-B). A JDRF Postdoctoral Fellowship (3-PDF-2017-400-A-N) funds K.J.P. The Sigrid Juselius Foundation, the Finnish Foundation for Cardiovascular Research, and the Otto A. Malm Foundation support J.S.. F.E. was supported by grants from JDRF (JDRF-5-CDA-2014-184-A-N) and NIH (NIH 5K01DK102488-03).
Declaration of Interests
The Hotamisligil lab has generated intellectual property (assigned to Harvard University) related to hormonal FABP4 and its therapeutic targeting and receives funding for this project from Lab1636, LLC, an affiliate of Deerfield Management. G.S.H. is in the Scientific Advisory Board of Crescenta Pharmaceuticals and holds equity. Other authors have no conflicts of interest to declare.
Figure 1. Depletion of FABP4 increases beta-cell mass and function
(a) Gross pancreas images in lean wild type (WT) and FABP4−/− mice. (b) Insulin immunohistochemistry (IHC) in pancreatic sections from 7-wk-old WT or FABP4−/− mice (N=4/group). (c) Quantification of percentage insulin positive area per total pancreatic area based on IHC (N=4/group; P=0.0318). (d) Total pancreatic insulin content from 7-wk-old WT and FABP4−/− mice (N=3/group; P=0.0198). (e) Glucose-stimulated insulin secretion (GSIS) from islets ex vivo under low glucose (2.8mM; LG) and high glucose (16.7mM; HG) conditions (N=8/group; P=0.006008). (f) 6hr fasting blood glucose from diet-induced obese (DIO) mice before treatment (wk 0) and following a-Ab for 3 wks (N=10/group; P=0.000064). (g) Glucose tolerance test (GTT) in DIO mice treated for 2 wks with PBS or a-Ab (N=10/group). (h) Insulin IHC in pancreatic sections from DIO mice treated with PBS or a-Ab for 3 wks (N=6/group) with (i) quantification of total islet number per pancreatic section (N=8/group; P=0.0157), and (j) percentage of insulin positive area per total pancreatic area (N=4/group). (k) Plasma FABP4 levels in autoantibody positive and negative normal glucose tolerant (NGT) individuals compared to new-onset T1D patients (<1-year duration; BABYDIAB and DiMELLI cohorts) (N=30/group; Ab+ vs. T1D P=0.0049; Ab- vs. T1D P=0.0047). (l) Correlation of plasma FABP4 with HbA1c percentage in established T1D patients (BRI cohort; N=50/group). (m) Plasma FABP4 levels in NOD mice while NGT, one week prior to hyperglycemia (Prior), or at time of T1D onset (N=35 (NGT), 16 (Prior), 10 (T1D); NGT vs. Prior P<0.00001; NGT vs. T1D P=0.0193). (n) Incidence curve for NOD model of T1D following treatment with PBS, a-Ab, or c-Ab beginning at 10 wks of age (N=36/group; P=0.0079). (o) Average blood glucose of NOD mice at the time of T1D diagnosis (N=23 (PBS), 11 (a-Ab), 19 (c-Ab); PBS vs. a-Ab P=0.0491; a-Ab vs. c-Ab P=0.0072). (p) Plasma insulin levels prior to T1D diagnosis in NOD mice (N=22 (PBS), 10 (a-Ab), 18 (c-Ab); PBS vs. a-Ab P=0.0350; a-Ab vs. c-Ab P=0.0055). (q) GTT and (r) corresponding plasma insulin values in non-diabetic NOD mice treated with PBS or a-Ab (N=6/group). (s) GSIS from islets isolated from NOD mice treated with PBS or a-Ab for 15 wks (N=4/group; P=0.0452). (t) Insulin IHC and quantification of (u) percentage of insulin positive area per total pancreatic area (P=0.0125), and (v) islet number per pancreatic section from mice treated with a-Ab or c-Ab for 5 wks (N=5/group; P=0.0085). (w) Quantification of insulin and glucagon immunofluorescence (IF) in pancreatic sections of NOD mice treated with a-Ab or c-Ab for 5 wks (N=4/group; P=0.0409). (x) IF and (y) quantification of insulin (green) and glucagon (red) staining in pancreatic sections of DIO mice treated with a-Ab or PBS for 3 wks (N=26 (PBS) and 30 (a-Ab) islets from 5 mice/group; P=0.0044). Scale bars for panels 500um (b,h,t) and 200um (x). *P<0.05, **P<0.01, ***P<0.001. Data are mean +/− SEM. Two-tailed unpaired t-test (c-f,I,j,u,v,w,y); One-way ANOVA (k,m,o,p); Two-way ANOVA (g,q,r,s); Simple linear regression correlation (l); Log-rank Mantel-Cox test (n).
Figure 2. Circulating FABP4 forms a hormonal complex with NDPK and ADK to regulate extracellular nucleosides.
(a) Results of three independent immune-precipitation (IP) and mass spectrometry experiments from WT or FABP4−/− DIO serum with a-Ab or c-Ab with top hits based on enrichment and spectral counts. Spectral counts of (b) ADK and (c) NDPK from Screen 1 (N=3/group). (d) Summary of MST among proposed complex components (N=6/interaction). (e) IP of GST-tagged NDPK with each of the proposed complex components and a-Ab (LC, light chain; HC, heavy chain; N=4). (f) On-bead ADK kinase assay for ATP production from WT or FABP4−/− mouse serum (N=5/group). (g) ATP production from recombinant ADK in the presence of complex components (N=3/group; ADK vs. ADK FABP4 P=0.0113; ADK vs FABP4-ADK-NDPK P=0.0154; ADK-NDPK vs. ADK FABP4 P=0.0360). (h) On-bead NDPK kinase assay for ADP production from WT or FABP4−/− mouse serum (N=5/group). (i) ADP production from recombinant NDPK in the presence of complex components (N=3/group; NDPK vs. NDPK ADK P=0.0001; NDPK vs. NDPK FABP4 P=0.0221; NDPK vs. FABP4-ADK-NDPK P=0.0001). (j) ADK activity to produce ATP in the presence of complex components with a-Ab or c-Ab (N=3/group; ADK vs. FABP4-ADK-NDPK P=0.0076; FABP4-ADK-NDPK vs. a-Ab P=0.0340; ADK vs. c-Ab P<0.0001; FABP4-ADK-NDPK vs. c-Ab P<0.0001; a-Ab vs. c-Ab P<0.0001). (k) NDPK activity to produce ADP in the presence of complex components with a-Ab or c-Ab (N=3/group; NDPK vs. FABP4-ADK-NDPK P<0.0001; NDPK vs. c-Ab P=0.0002; FABP4-ADK-NDPK vs. a-Ab P=0.0002; a-Ab vs. c-Ab P=0.0003). (l) Proposed model of extracellular nucleoside regulation from the complex. *P<0.05, **P<0.01, ***P<0.001. Data are mean +/− SEM. Two-way ANOVA (f,h); One-way ANOVA (g,i-k).
Figure 3. The FABP4-ADK-NDPK complex inhibits GSIS and is neutralized by a-Ab.
GSIS from (a) WT primary mouse islets (N=3/group; P<0.0001), (b) human islets (N=3/group; Control vs. NDPK-ADK P=0.0038; NDPK-ADK vs. FABP4-ADK-NDPK P=0.0049; Control c-Ab vs. NDPK-ADK c-Ab P=0.0043; NDPK-ADK c-Ab vs. FABP4-ADK-NDPK c-Ab P=0.0230; Control a-Ab vs. NDPK-ADK a-Ab 0.0031; NDPK-ADK a-Ab vs. FABP4-ADK-NDPK a-Ab P=0.0067) and (c) INS1 cells with complex components (N=3/group; Control vs. NDPK-ADK P=0.0012; NDPK-ADK vs. FABP4-ADK-NDPK P=0.0002; FABP4-ADK-NDPK vs. a-Ab P=0.0009; c-Ab vs. a-Ab P=0.0066). (d) Proposed model for the activity of FABP4-ADK-NDPK on P2Y1. (e) GSIS from WT mouse islets treated with high ratio (ATP degrading) or low ratio (ATP and ADP degrading) apyrase (N=3/group). GSIS from INS1 cells treated with control, NDPK-ADK, or FABP4-ADK-NDPK in the presence of (f) 5uM ATP or non-metabolizable ATPγS (N=7/group) and (g) 5uM of ADP (N=3/group; Control vs. NDPK-ADK P=0.0132; NDPK-ADK vs. FABP4-ADK-NDPK P=0.0307; FABP4-ADK-NDPK vs. FABP4-ADK-NDPK ADP P=0.0273). (h) GSIS from INS1 cells with ADP in the presence or absence of P2Y1 antagonist MRS2179 (N=4/group). (i) GSIS from INS1 cells treated with NDPK-ADK or FABP4-ADK-NDPK in the presence of MRS2179 (N=3/group; Control vs. NDPK-ADK P=0.0232; NDPK-ADK vs. FABP4-ADK-NDPK P=0.0038; NDPK-ADK vs. NDPK-ADK MRS2179 P=0.0026). (j) Kinase activity of WT NDPK or kinase dead H118N NDPK to produce ADP (N=4/group). (k) GSIS from INS1 cells treated with WT NDPK or H118N NDPK in combination with ADK and FABP4 (N=4/group). (l) Kinase activity of ADK to generate ATP in the presence of ADK inhibitor (N=4/group). (m) GSIS from INS1 cells treated with NDPK-ADK or FABP4-ADK-NDPK in the presence of ADK inhibitor (N=3). In all cases, low glucose (LG; 2.8mM), high glucose (HG; 16.7mM). *P<0.05, **P<0.01, ***P<0.001. Data are mean +/− SEM. Two-way ANOVA (a,b,e-m); One-way ANOVA (c).
Figure 4. The FABP4-ADK-NDPK complex alters beta-cell calcium dynamics and promotes cell death.
(a) Western blot (WB) and quantification of pPKA substrate phosphorylation in human islets treated with NDPK-ADK or FABP4-ADK-NDPK (N=6; Control vs. NDPK-ADK P<0.0001; NDPK-ADK vs. FABP4-ADK-NDPK P=0.0007; Control vs. FABP4-ADK-NDPK P=0.0095). (b) WB and quantification of pIP3R in INS1 cells treated with FABP4-ADK-NDPK (N=4; P=0.0034). Cytosolic calcium flux in INS1 cells from (c,d) the extracellular space in response to glucose (N=3 coverslips, 150 cells/condition except NDPK-ADK 144 cells) and (e,f) the ER in response to 1uM thapsigargin by Fura-2 AM staining (N=3 coverslips, 75 cells/condition except NDPK-ADK 100 cells). (g) WB for cleaved caspase 3 (CC3), pJNK, Chop, and B-Tubulin from INS1 cells treated with FABP4-ADK-NDPK for 24hrs (N=4). (h) Gene expression of ER stress marker Chop following 2hr treatment with complex components in the presence or absence of Tg (N=3; Control vs. FABP4-ADK-NDPK P<0.0001; NDPK-ADK vs. FABP4-ADK-NDPK P=0.0012; FABP4-ADK-NDPK vs. a-Ab P=0.0003). (i) Cleaved caspase 3/7 activity in INS1 cells treated with cytokine cocktail (TNFa, IFNg and IL-1b) with or without FABP4-ADK-NDPK (N=4/group). (j) IF for insulin (green), BrdU (red), and nuclei (DAPI; blue) and (k) quantification of pancreatic sections from NOD mice treated with BrdU and a-Ab or c-Ab for 5wks (N=5 mice/group; 48 (c-Ab), 46 (a-Ab) islets). (l) IF for insulin (green), Ki67 (red), and nuclei (DAPI; blue) and (m) quantification of pancreatic sections from DIO mice treated with PBS or a-Ab for 3 weeks (N=5 mice/group; 39 (PBS), 45 (a-Ab) islets; P=0.0437). (n) IF staining of pancreatic sections from NOD mice treated with a-Ab or c-Ab for insulin (red), CC3 (green) and nuclei (DAPI; blue) with (o) quantification of CC3 and insulin co-stained cells (N=5 mice/group; 33 (c-Ab), 45 (a-Ab) islets; P=0.0010). (p) IF of pancreatic sections from DIO mice treated with PBS or a-Ab for insulin (green), CC3 (red) and nuclei (DAPI; blue) with (q) quantification of CC3 and insulin co-stained cells (N= 5 mice/group; 25 (PBS), 34 (a-Ab) islets; P<0.0001). Scale bars are 150um (j,l,p) and 100um (n). *P<0.05, **P<0.01, ***P<0.001. Data are mean +/− SEM. One-way ANOVA (d,f); Two-tailed unpaired t-test (a,b,k,m,o,q); Two-way ANOVA (h,i).
Supplementary Information is available for this paper.
All materials produced by the Hotamislgil lab will be made available upon reasonable request. Third party materials are subject to approval from the originating institutions.
References
1 Jensen MD , Caruso M , Heiling V & Miles JM Insulin regulation of lipolysis in nondiabetic and IDDM subjects. Diabetes 38 , 1595–1601, doi:10.2337/diab.38.12.1595 (1989).2573554
2 Kim JY Increased lipolysis, diminished adipose tissue insulin sensitivity, and impaired beta-cell function relative to adipose tissue insulin sensitivity in obese youth with impaired glucose tolerance. Diabetes 66 , 3085–3090, doi:10.2337/db17-0551 (2017).28887312
3 Cao H Adipocyte lipid chaperone AP2 is a secreted adipokine regulating hepatic glucose production. Cell Metab 17 , 768–778, doi: 10.1016/j.cmet.2013.04.012 (2013).23663740
4 Prentice KJ , Saksi J & Hotamisligil GS Adipokine FABP4 integrates energy stores and counterregulatory metabolic responses. J Lipid Res 60 , 734–740, doi:10.1194/jlr.S091793 (2019).30705117
5 Saksi J Low-expression variant of fatty acid-binding protein 4 favors reduced manifestations of atherosclerotic disease and increased plaque stability. Circ Cardiovasc Genet 7 , 588–598, doi:10.1161/CIRCGENETICS.113.000499 (2014).25122052
6 Tuncman G A genetic variant at the fatty acid-binding protein aP2 locus reduces the risk for hypertriglyceridemia, type 2 diabetes, and cardiovascular disease. Proc Natl Acad Sci U S A 103 , 6970–6975, doi:10.1073/pnas.0602178103 (2006).16641093
7 Zhao W Identification of new susceptibility loci for type 2 diabetes and shared etiological pathways with coronary heart disease. Nat Genet 49 , 1450–1457, doi:10.1038/ng.3943 (2017).28869590
8 Ertunc ME Secretion of fatty acid binding protein aP2 from adipocytes through a nonclassical pathway in response to adipocyte lipase activity. J Lipid Res 56 , 423–434, doi:10.1194/jlr.M055798 (2015).25535287
9 Schlottmann I , Ehrhart-Bornstein M , Wabitsch M , Bornstein SR & Lamounier-Zepter V Calcium-dependent release of adipocyte fatty acid binding protein from human adipocytes. Int J Obes (Lond) 38 , 1221–1227, doi:10.1038/ijo.2013.241 (2014).24352293
10 Villeneuve J Unconventional secretion of FABP4 by endosomes and secretory lysosomes. J Cell Biol 217 , 649–665, doi:10.1083/jcb.201705047 (2018).29212659
11 Burak MF Development of a therapeutic monoclonal antibody that targets secreted fatty acid-binding protein aP2 to treat type 2 diabetes. Sci Transl Med 7 , 319ra205, doi:10.1126/scitranslmed.aac6336 (2015).
12 Thumer L German new onset diabetes in the young incident cohort study: DiMelli study design and first-year results. Rev Diabet Stud 7 , 202–208, doi:10.1900/RDS.2010.7.202 (2010).21409312
13 Warncke K Does diabetes appear in distinct phenotypes in young people? Results of the diabetes mellitus incidence Cohort Registry (DiMelli). PLoS One 8 , e74339, doi:10.1371/journal.pone.0074339 (2013).24023937
14 Ziegler AG , Hummel M , Schenker M & Bonifacio E Autoantibody appearance and risk for development of childhood diabetes in offspring of parents with type 1 diabetes: the 2-year analysis of the German BABYDIAB Study. Diabetes 48 , 460–468, doi:10.2337/diabetes.48.3.460 (1999).10078544
15 Xiao Y Fatty acid binding protein 4 promotes autoimmune diabetes by recruitment and activation of pancreatic islet macrophages. JCI Insight 6 , e141814, doi: 10.1172/jci.insight.141814 (2021).
16 Kralisch S Circulating adipocyte fatty acid-binding protein induces insulin resistance in mice in vivo. Obesity (Silver Spring) 23 , 1007–1013, doi:10.1002/oby.21057 (2015).25865078
17 Wu LE Identification of fatty acid binding protein 4 as an adipokine that regulates insulin secretion during obesity. Mol Metab 3 , 465–473, doi:10.1016/j.molmet.2014.02.005 (2014).24944906
18 Annes JP Adenosine kinase inhibition selectively promotes rodent and porcine islet beta-cell replication. Proc Natl Acad Sci USA 109 , 3915–3920, doi:10.1073/pnas.1201149109 (2012).22345561
19 Navarro G Genetic disruption of adenosine kinase in mouse pancreatic beta-cells protects against high-fat diet-induced glucose intolerance. Diabetes 66 , 1928–1938, doi:10.2337/db16-0816 (2017).28468960
20 Furuhashi M Treatment of diabetes and atherosclerosis by inhibiting fatty-acid-binding protein aP2. Nature 447 , 959–965, doi:10.1038/nature05844 (2007).17554340
21 Khan S Autocrine activation of P2Y1 receptors couples Ca (2+) influx to Ca (2+) release in human pancreatic beta-cells. Diabetologia 57 , 2535–2545, doi:10.1007/s00125-014-3368-8 (2014).25208758
22 Leon C The P2Y1 receptor is an ADP receptor antagonized by ATP and expressed in platelets and megakaryoblastic cells. FEBS Lett 403 , 26–30, doi:10.1016/s0014-5793(97)00022-7 (1997).9038354
23 Detimary P , Jonas JC , & Henquin JC Stable and diffusible pools of nucleotides in pancreatic islet cells. Endocrinology 137 , 4671–4676, doi: 10.1210/en.137.11.4671 (1996).8895332
24 Bauer C ATP mediates a negative autocrine signal on stimulus-secretion coupling in mouse pancreatic beta-cells. Endocrine 63 , 270–283, doi:10.1007/s12020-018-1731-0 (2019).30229397
25 Baurand A Inhibition of platelet function by administration of MRS2179, a P2Y1 receptor antagonist. Eur J Pharmacol 412 , 213–221, doi:10.1016/s0014-2999(01)00733-6 (2001).11166284
26 Seo JB , Jung SR , Hille B & Koh DS Extracellular ATP protects pancreatic duct epithelial cells from alcohol-induced damage through P2Y1 receptor-cAMP signal pathway. Cell Biol Toxicol 32 , 229–247, doi:10.1007/s10565-016-9331-3 (2016).27197531
27 Engin F Restoration of the unfolded protein response in pancreatic beta-cells protects mice against type 1 diabetes. Sci Transl Med 5 , 211ra156, doi:10.1126/scitranslmed.3006534 (2013).
28 Engin F , Nguyen T , Yermalovich A & Hotamisligil GS Aberrant islet unfolded protein response in type 2 diabetes. Sci Rep 4 , 4054, doi:10.1038/srep04054 (2014).24514745
29 Burnstock G , Vaughn B & Robson SC Purinergic signalling in the liver in health and disease. Purinergic Signal 10 , 51–70, doi:10.1007/s11302-013-9398-8 (2014).24271096
30 Burnstock G Purinergic signaling in the cardiovascular system. Circ Res 120 , 207–228, doi:10.1161/CIRCRESAHA.116.309726 (2017).28057794
|
PMC008xxxxxx/PMC8983234.txt
|
LICENSE: This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
2985117R
4816
J Immunol
J Immunol
Journal of immunology (Baltimore, Md. : 1950)
0022-1767
1550-6606
35304420
8983234
10.4049/jimmunol.2100700
VAPA1777263
Article
NeoScore integrates characteristics of the neoantigen:MHC class I interaction and expression to accurately prioritize immunogenic neoantigens
Borden Elizabeth S. *†
Ghafoor Suhail ‡
Buetow Kenneth H. ‡§
LaFleur Bonnie J. ¶
Wilson Melissa A. ‡§
Hastings K. Taraszka *†
Pubmed index as Hastings KT
* Department of Basic Medical Sciences, College of Medicine-Phoenix, University of Arizona, Phoenix, AZ, United States
† Phoenix Veterans Affairs Health Care System, Phoenix, AZ, United States
‡ Center for Evolution and Medicine, Arizona State University, Tempe, AZ, United States
§ School of Life Sciences, Arizona State University, Tempe, AZ, United States
¶ BIO5 Institute, University of Arizona, Tucson, AZ, United States
Author contributions: Conceptualization E.S.B., K.H.B., M.A.W., and K.T.H. Methodology E.S.B, S.G., B.J.L., M.A.W., and K.T.H. Software E.S.B. Formal Analysis E.S.B. and B.J.L. Investigation E.S.B. Resources K.H.B., M.A.W., and K.T.H. Writing -- Original Draft E.S.B. and K.T.H. Writing -- Reviewing and Editing E.S.B., K.H.B., B.J.L., M.A.W., and K.T.H. Visualization E.S.B. Supervision K.T.H. Funding Acquisition K.T.H.
Corresponding author: K. Taraszka Hastings, 602-827-2106 (phone), 602-827-2127 (fax), [email protected]
9 3 2022
01 4 2022
18 3 2022
01 10 2022
208 7 18131827
This file is available for text mining. It may also be used consistent with the principles of fair use under the copyright law.
Accurate prioritization of immunogenic neoantigens is key to developing personalized cancer vaccines and distinguishing those patients likely to respond to immune checkpoint inhibition. However, there is no consensus regarding which characteristics best predict neoantigen immunogenicity, and no model to date has both high sensitivity and specificity and a significant association with survival in response to immunotherapy. We address these challenges in the prioritization of immunogenic neoantigens by 1) identifying which neoantigen characteristics best predict immunogenicity, 2) integrating these characteristics into an immunogenicity score, the NeoScore, and 3) demonstrating a significant association of the NeoScore with survival in response to immune checkpoint inhibition. One thousand random and evenly split combinations of immunogenic and non-immunogenic neoantigens from a validated dataset were analyzed using a regularized regression model for characteristic selection. The selected characteristics, the dissociation constant and binding stability of the neoantigen:MHC class I complex and expression of the mutated gene in the tumor, were integrated into the NeoScore. A web application is provided for calculation of the NeoScore. The NeoScore results in improved, or equivalent, performance in four test datasets as measured by sensitivity, specificity, and area under the receiver operator characteristics curve compared to previous models. Among cutaneous melanoma patients treated with immune checkpoint inhibition, a high maximum NeoScore was associated with improved survival. Overall, the NeoScore has the potential to improve neoantigen prioritization for the development of personalized vaccines and contribute to the determination of which patients are likely to respond to immunotherapy.
pmcIntroduction:
Cancers arise through mutations in the genome of healthy human cells. As these mutations occur, some will produce mutated proteins, which have the potential to be processed into neoantigens that bind MHC class I and are presented on the cell surface. These neoantigens then act as tumor-specific targets with the potential to elicit a cytotoxic CD8+ T cell response (1–4). Tumor-specific neoantigens have strong potential to be targets of T cell-mediated destruction, because they are not subject to immune tolerance or non-reactivity to self. However, there are two key ways in which the above mechanism may fail in tumor destruction. For one, recent evidence suggests that most neoantigens do not elicit an immune response in their natural state, termed immunologic ignorance (5). Second, once a T cell response is mounted to a neoantigen, that response may become exhausted over time due to inhibitory signals from the tumor microenvironment (6).
Circumventing these limitations to re-invigorate the host immune response has been the goal of many recent cancer therapies. To address immunologic ignorance, personalized cancer vaccines have been created and have demonstrated early success (7–12). These vaccines have taken several forms, including direct exposure to neoantigens (11), neoantigen-encoding RNA vaccines (12), neoantigen-loaded dendritic cell vaccines (7), and adoptive transfer of neoantigen-specific T cells (2, 13). Each of these methods requires accurate knowledge of the neoantigens presented by the tumor cell with the potential to elicit an immune response. In silico prioritization methods have been used to prioritize which set of neoantigens should be experimentally tested, but the ability to prioritize the immunogenicity of each neoantigen, with high sensitivity and specificity, is still limited (14–18).
Exhaustion of T cells and attenuation of T cell activation can be overcome by immune checkpoint inhibition, such as monoclonal antibodies against PD-1 or CTLA-4. Immune checkpoint inhibition blocks inhibitory signals to the T cells to enhance T cell-mediated tumor destruction (19–21). However, immune checkpoint inhibition is only effective in a subset of patients (6), and there is no consensus on how to prioritize which patients will respond (18, 21–26). In a pan-cancer analysis, Yarchoan et al. demonstrated that cancer types with a higher mutational burden, such as melanoma, had improved response to anti-PD-1 therapy compared to cancer types with a lower mutational burden (27). There are limitations to mutational burden as a predictor of response to immune checkpoint inhibition. First, in multiple myeloma, there was an association of increased tumor mutational burden with decreased response to immune checkpoint inhibition (28). Second, in melanoma, the association between the tumor mutational burden and response to immune checkpoint inhibition was confounded by the melanoma subtype (29). Finally, in lung cancers that progressed after treatment with immune checkpoint inhibitors, there was an increase in the tumor mutational burden compared to the pretreatment state (30), thus, contradicting the expectation that tumor cells resistant to immune checkpoint inhibition would have a low number of neoantigens. Together, these findings suggest that tumor mutational burden is not sufficient for predicting response to immune checkpoint inhibition.
Several recent papers have been dedicated to predicting neoantigen immunogenicity based on the characteristics of validated immunogenic neoantigens (16–18, 23, 31). However, there is no consensus regarding which neoantigen characteristics are important for the prioritization of immunogenic neoantigens or the best way to combine the characteristics into an overall immunogenicity score. To identify the characteristics that best prioritize the immunogenicity of the neoantigens, a model-based approach was applied to evaluate neoantigen characteristics encompassing expression, processing, presentation, and T cell recognition in prioritizing neoantigen immunogenicity. The selected characteristics were then combined into an overall logistic regression model called the “NeoScore.” The development of the NeoScore has largely focused on melanoma, except for one lung cancer patient included in the Tumor Neoantigen Selection Alliance (TESLA) consortium dataset (16).
Immune checkpoint inhibition and personalized neoantigen vaccines are particularly effective in mutation-rich melanoma (25, 32–34). However, even in melanoma, a positive outcome from these interventions is not assured (6). To assess the clinical utility of the NeoScore and its ability to improve assessment of outcome in melanoma, the relationship of the NeoScore to survival in response to immunotherapy was tested using the datasets of Van Allen et al. 2015 and Liu et al. 2019 (21, 29).
Materials and Methods:
Datasets
For training of the neoantigen prioritization model, whole-exome sequencing (WES) and RNA sequencing (RNAseq) data were obtained from the TESLA consortium database on Synapse (16). The TESLA consortium data came from four patients with melanoma and a single lung cancer patient. The TESLA consortium provided RNAseq data, WES data, and clinically determined HLA types for each of these patients to 28 teams and used the neoantigen rankings from 25 of the teams to prioritize which neoantigens were experimentally validated. The neoantigens validated for their ability to elicit a T cell response were selected by two guidelines: 1) the top 5 neoantigens ranked by each team were tested and 2) the neoantigens that came up most frequently in the top 50 ranked neoantigens for each team were selected. The neoantigens tested were also constrained by HLA restriction requirements for the validation experiments. The final available dataset includes 5 patients with a total of 347 neoantigens that had been tested for their ability to elicit a T cell response using multimer staining. The TESLA consortium found that 26 of the 347 tested neoantigens elicited a T cell response in an unvaccinated state (16). Neoantigens were used for model creation in this manuscript if they were 1) tested for immunogenicity by the TESLA consortium, 2) identified by either GATK Mutect2 or Strelka, and 3) had expression data. Accession information for the TESLA consortium dataset is included in Table I.
For testing the neoantigen prioritization model, lists of tested neoantigens from melanoma tumors were obtained (7, 11, 35, 36). These datasets contain 357 tested neoantigens (n=7 immunogenic neoantigens) (35), 149 tested neoantigens (n=18 immunogenic neoantigens) (11), 57 tested neoantigens (n=11 immunogenic neoantigens) (36) and 21 tested neoantigens (n=9 immunogenic neoantigens) (7). The neoantigens were tested for immunogenicity with tetramer staining and cytokine secretion (35), ELISPOT (11), and multimer staining (7, 36). Strønen, Ott, and Carreno also provided the expression data quantified as either fragments per kilobase of transcript per million mapped reads (FPKM), reads per kilobase of transcript per million mapped reads (RPKM), or transcripts per million (TPM) (7, 11, 35, 36). For the Cohen dataset, no expression data was provided. The RNAseq data from the Cohen dataset was obtained and analyzed as described below for read count quantification. Accession information is in Table I.
Finally, for survival analysis, WES and RNAseq data were obtained from the Van Allen et al., Liu et al., and Rizvi et al. cohorts (21, 25, 29) (all accession information is in Table I). The inclusion criteria for the Van Allen dataset were as follows: 1) both WES and RNAseq data available, 2) cutaneous melanoma as the primary lesion. All patients in the Van Allen cohort underwent treatment with an anti-CTLA-4 monoclonal antibody (ipilimumab). Inclusion criteria for the Liu et al dataset were as follows: 1) both WES and RNAseq data available, 2) cutaneous melanoma as the primary lesion, 3) no prior treatment with an anti-CTLA-4 monoclonal antibody (37). All patients in the Liu cohort underwent treatment with an anti-PD-1 monoclonal antibody (nivolumab or pembrolizumab). All samples from the Rizvi et al. dataset were included in the analysis including adenocarcinoma, squamous cell carcinoma, and non-small cell lung cancer. All patients in the Rizvi cohort underwent treatment with an anti-PD-1 monoclonal antibody (pembrolizumab only).
Data Preparation
WES and RNAseq FASTQ files from the TESLA consortium, Liu et al., and Van Allen et al., RNAseq FASTQ files from Cohen et al., and WES FASTQ files from Rizvi et al. were visualized for quality using FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). FASTQ files were trimmed for quality using Trimmomatic (38) IlluminaClip with the following parameters: seed_mismatches = 2, palindrome_clip_threshold = 30, simple_clip_threshold = 10, leading = 10, trailing = 10, winsize = 4, winqual = 15. Quality was then re-visualized after trimming. Trimmed WES reads were mapped to the GRCh38.p7 reference genome, from 1000 genomes (39), and read group labels were added using BWA-mem (40). SAM files were converted to BAM and coordinate sorted (41). The BAM files were then converted to pileup format using SAMtools v. 1.4 (41).
Identification of somatic mutations and neoantigen generation
Single nucleotide variants (SNVs) and small insertions/deletions (indels) were identified using four programs, along with their recommended filters: GATK Mutect2 version 4.1.7.0 with default parameters, Varscan2 version 2.3.9 with minimum coverage of 10, minimum variant allele frequency of 0.08, and somatic p-value of 0.05, Strelka version 2.9.2 with default parameters, and LoFreq version 2.1.1 with default parameters (42–45). Results for GATK Mutect2 were filtered with the recommended FilterMutectCalls, and Varscan2 results were filtered using the Perl false-positive filter (https://github.com/ckandoth/variant-filter). Results from all four programs were examined with and without their respective filters. LoFreq results were not filtered to allow maximal potential to identify the missing mutations. Matched normal samples were used as the reference for each sample. The overlap of GATK Mutect2 and Strelka was used for the identification of SNVs and indels in the Liu, Van Allen, and Rizvi datasets for assessing the association of the NeoScore model with survival in response to immune checkpoint inhibition in melanoma.
Somatic mutations were separated from those SNVs that fell within 1 bp of an indel position, as these are likely to be false positives due to alignment errors. The mutations were annotated using the Variant Effect Predictor (VEP) tool from Ensembl version 90.9 (46). Then, 21mer amino acid sequences were generated for each mutation using pVAC-Seq tools version 3.0.5 (47). Finally, the 21mers were split into all possible 9mers and 10mers where the mutation of interest was in every location. The full pipeline from read mapping through to the identification of somatic mutations is available at https://github.com/SexChrLab/Cancer_Genomics.
Because the clinical samples for Liu et al. and Van Allen et al. were comprised of formalin-fixed, paraffin embedded (FFPE) samples, we considered applying an FFPE filter to remove false positive mutations introduced by FFPE storage. However, the characteristic G→A and C→T mutations introduced by FFPE processing overlap with the mutational signature introduced by ultraviolet radiation. Removal of the FFPE signature also removed the characteristic bias towards G→A and C→T mutations in the samples, and therefore, an FFPE filter was not applied.
Calculation of neoantigen characteristics
For each of the validated neoantigens, neoantigen characteristics with potential significance in predicting expression, processing, MHC binding, and T cell receptor (TCR) recognition probability were calculated as described here. The full pipeline for calculation and processing of the neoantigen characteristics can be found at https://github.com/ElizabethBorden/Process_peptide_lists. A log10 transformation was applied if the distribution of the characteristic had a large degree of skewness. The difference in the values for each of the neoantigen characteristics between immunogenic and non-immunogenic neoantigens was assessed using a two-sample, two-sided t-test. Correlation coefficients were calculated using Spearman correlation coefficients. P-values below 0.05 were considered statistically significant.
Expression:
For the datasets from Cohen et al., the TESLA consortium, and Liu et al., transcriptome assembly and read count quantifications were completed with Salmon version 0.11.3, using the Ensembl GRCh38.p7 reference genome (48, 49). mRNA expression in units of TPM was log10-transformed, and a constant of 0.1 was added to all values before the transformation to avoid taking the log of zero. To account for the different units used across each dataset, expression values were centered and normalized by subtracting the mean and then dividing by the standard deviation.
Clonality:
Copy number variation was calculated with sequenza (50), and clonality was calculated using the deconvolution software, FastClone (51).
Variant allele frequency:
Variant allele frequency (VAF) as calculated by GATK Mutect2 (42). Since the VAF was only calculated by GATK Mutect2, but not for Strelka, 14 missing data points were estimated as the average of the rest of the data. No statistically significant difference in the VAF was observed between immunogenic and non-immunogenic neoantigens, either with or without these data points.
Processing:
Cleavage and TAP transport potentials were calculated for each of the available neoantigens using NetCTLpan1.0 (52).
Dissociation constants:
Dissociation constants of the neoantigen:MHC class I complex were calculated in nanomolar (nM) units using NetMHCpan4.0 (53). These values were log10-transformed before inclusion in the model.
Binding stability:
Binding stability of the neoantigen:MHC class I complex was then calculated as the half-life in units of hours using NetMHCstabpan1.0 (53, 54). These values were log10-transformed and adjusted by a factor of 0.01 to avoid taking the log of zero before inclusion in the model.
Hydrophobicity method from the TESLA consortium:
The number of hydrophobic residues was divided by the total number of residues in the neoantigen to create a “hydrophobicity fraction” (16). In keeping with the methods of the TESLA consortium, hydrophobic residues here were isoleucine, leucine, phenylalanine, methionine, tryptophan, valine, and cysteine.
Hydrophobicity with empirical prevalence:
The hydrophobicity fraction was calculated as described for the TESLA consortium using the amino acids found to be empirically prevalent by Chowell et al. (55). Proline, leucine, and methionine were considered to have a high probability and given a score of +2. Glycine, tryptophan, phenylalanine, isoleucine, and valine were found to have a medium probability and given a score of +1. All others were given a score of 0.
Hydrophobicity Łuksza method:
A neoantigen was given a score of zero if the mutation at the anchor residue changed from a hydrophobic to a hydrophilic residue (23). All other neoantigens were given a score of one. Hydrophobic neoantigens here were defined as alanine, isoleucine, leucine, methionine, phenylalanine, tryptophan, tyrosine, and valine.
TCR recognition:
Prediction of TCR recognition probability, R, was calculated as described (23). A BLOSUM62 similarity matrix was used to assess the sequence similarity between a neoantigen and the closest matched known T cell epitope from the Immune Epitope Database (IEDB) (56). The sequence similarity was then used in place of binding energies, and the TCR recognition probability was calculated as: R=Zk−1∑e∈IEDBexp[−ka−s,e]
Where |s,e| is the sequence similarity, a is the horizontal displacement of the binding curve, and k sets the steepness of the curve at a. Based on the model fit by Łuksza et al., the parameters a and k were set to 26 and 4.87 respectively (23). These parameters were optimized for both melanoma and lung cancer patients, the two cancer populations included in our training and test datasets. Finally, Z(k) is the partition function over the unbound state and all bound states, calculated as follows: Zk=1+∑e∈IEDBexp[−ka−s,e]
Sequence Similarity:
The closest matched human peptide was identified using Blast v. 2.10.1 (57), and the sequence similarity of the potential neoantigen to the closest matched human peptide was calculated using a BLOSUM62 matrix, as described (17). The sequence similarity was normalized across neoantigen length by dividing by the number of amino acids.
Amplitude:
Dissociation constants for the neoantigen:MHC class I complex were then adjusted by the dissociation constants for the closest matched normal human peptide:MHC class I complex, a characteristic called amplitude. The ratio was taken of the dissociation constant of the closest matched human peptide:MHC class I (KdWT) to the dissociation constant of the neoantigen:MHC class I (KdMT) as follows (23): A=KdWT/KdMT
Analysis of anchor vs. non-anchor residue mutations:
Neoantigens were separated by their mutation position (anchor vs. non-anchor residues). Then, both the amplitude and the dissociation constant of the neoantigen:MHC class I complex were compared between the immunogenic and non-immunogenic neoantigens within each group. Comparisons were done using a two-sample, two-sided t-test. P-values below 0.05 were considered statistically significant.
HLA typing
HLA types were identified on normal tissue samples for each patient in the Liu et al. dataset using HLA-LA (58). HLA types provided in the supplementary data were used for each patient in the Van Allen et al. and Rizvi et al. datasets (21, 25).
Elastic net regression modeling
The TESLA consortium data was split into 1000 random subgroups containing all the immunogenic neoantigens (n=26) and an equal number of randomly selected non-immunogenic neoantigens. An elastic net regression was performed using the glmnet package in R for each of these splits (59). The selected coefficients and area under the receiver operator characteristics curve (AUC) from each model were tabulated across the 1000 splits. An optimal threshold for the model was selected in the TESLA dataset to optimize both sensitivity and specificity by the Youden Index and was implemented through the optimal cutpointr package from R statistical software (https://github.com/thie1e/cutpointr).
Logistic regression modeling
Logistic regression modeling was performed, and optimism values were calculated with the RMS package in R statistical software (https://cran.r-project.org/web/packages/rms/rms.pdf).
Survival analysis
To avoid scaling expression values on a patient-level basis, coefficients were determined on the TESLA consortium data for log10-transformed, non-scaled TPM expression data. The intercept changed to −1.7951 and the expression coefficient to 1.2868, but there was no change in the coefficients of the dissociation constant or binding stability and no effect on the performance of the model. The optimal threshold for each of the mutational burden, neoantigen burden, and maximum NeoScore were determined by optimizing based on an adjusted log rank test implemented in the MaxStat package in R (60). Then, the survival analysis was performed between scores above the selected threshold and below, using Kaplan Meier estimation and the log-rank test statistic. P-values below 0.05 were considered statistically significant.
Results:
Overlap of Strelka and GATK Mutect2 identifies the maximum number of validated immunogenic mutations
The first step in effective prioritization of neoantigens is to identify a high-fidelity list of somatic mutations. Isolating somatic mutations in cancers is more difficult than variant calling in normal tissue since cancers do not follow the typical rules of copy number or heterozygosity and often consist of multiple clonal populations with normal tissue contamination. The TESLA consortium validated neoantigens derived from mutations identified by 25 teams. The methods used for identifying the somatic mutations were not reported, making it difficult to reproduce all neoantigens (16). To maximize the immunogenic mutations identified, three highly rated programs were used to identify single nucleotide variants (SNVs) and small insertions and deletions (indels) for the data from the TESLA consortium: VarScan2, GATK Mutect2, and Strelka (42–44). A large degree of overlap was found in the ability of each program to identify the 34 validated immunogenic mutations. As shown in Figure 1, GATK Mutect2 and Strelka successfully identified 27/34 mutations, while VarScan2 identified 24/34. The mutations from each of these programs overlapped, such that 27 was the maximum number of immunogenic mutations identified. To ensure that the success of VarScan2 in identifying the immunogenic mutations was not hindered by over-filtering, the unfiltered output was assessed and only one immunogenic mutation had been eliminated by filtration steps (data not shown). GATK Mutect2 and Strelka identified the same number of immunogenic mutations with or without their respective filters (data not shown). LoFreq was also tested for the identification of somatic mutations, as LoFreq is optimized to call low-frequency mutations (45). However, LoFreq did not identify any additional immunogenic mutations (data not shown). The neoantigens derived from unidentified immunogenic mutations may be due to the reference genome version, peptide generation steps, or mutations identified by programs that were not tested. Overall, the combination of GATK Mutect2 and Strelka identified greater than or equal to the number of validated immunogenic neoantigens as in all other reported pipelines (16), while simultaneously decreasing the total number of potential neoantigens by 89.44%.
Dissociation constant and binding stability of the neoantigen:MHC class I complex as well as expression are significantly different between immunogenic and non-immunogenic neoantigens
A set of computationally predicted neoantigen characteristics were calculated for the expression, processing, presentation, and T cell receptor (TCR) recognition of SNV and small indel-derived neoantigens (Figure 2A). Each of these characteristics was included in the development of the NeoScore model. Figure 2B demonstrates that the set of characteristics included in model development is inclusive of all of those considered by the three models to which the NeoScore is compared. To begin, the distribution of each of these characteristics was assessed for immunogenic and non-immunogenic neoantigens from the TESLA consortium dataset.
Expression was included in the NeoScore model development in two ways: 1) mRNA expression level and 2) variant allele frequency (VAF). Expression level from RNAseq data was calculated as the transcripts per million (TPM) expression of the gene from which the neoantigen is derived. Immunogenic neoantigens had a significantly higher expression level than non-immunogenic neoantigens (p=2.875×10−4, Figure 2C). The clonality of the neoantigen was calculated by FastClone (51), but FastClone did not converge for 4/6 of the samples, so the VAF calculated by GATK-mutect2 was included in the development of the NeoScore model instead. No significant difference in the VAF was observed between immunogenic and non-immunogenic neoantigens (p=0.359, Figure 2D).
Processing steps for the neoantigen were included in the development of the NeoScore model as both the proteasomal cleavage potential and TAP (transporter associated with antigen processing) potential scores. While there was a higher average for both characteristics in the immunogenic than non-immunogenic neoantigens, no statistically significant difference was observed (p=0.817 and p=0.836, respectively) (Figure 2E–F). The binding to the MHC class I molecule was then considered by both the dissociation constant and stability of the neoantigen:MHC class I interaction. NetMHCpan was selected as the software for predicting the MHC class I dissociation constants due to its enhanced performance compared to other binding prediction methods (53). NetMHCpan achieved over a 0.90 AUC across 6 independent test datasets (61). The MHC class I dissociation constants were significantly lower in immunogenic than non-immunogenic neoantigens, indicating a higher binding affinity of immunogenic neoantigens to MHC class I (p=2.088×10−7, Figure 2G). The binding stability showed significantly higher values for immunogenic over non-immunogenic neoantigens. (p=3.895×10−5, Figure 2H).
The ability of the neoantigen to stimulate a T cell response was included in the NeoScore model development using the model created by Łuksza et al. to calculate a characteristic referred to as the TCR recognition probability (23). The TCR recognition probability is a probabilistic model that considers the sequence similarity between the neoantigen and a known T cell epitope from the Immune Epitope Database (IEDB) as a proxy for the binding affinity of the neoantigen:MHC class I-TCR interaction. The TCR recognition probability showed no statistically significant difference between the immunogenic and non-immunogenic neoantigens (p=0.636, Figure 2I).
Two methods to account for T cell development were also included in NeoScore development. During maturation in the thymus, T cells expressing TCRs with high avidity to normal human peptides undergo apoptosis. Thus, a neoantigen with a high degree of sequence similarity to a normal human peptide is less likely to elicit a T cell response. The first method used to account for T cell development was the sequence similarity to the closest matched normal human peptide. No statistically significant difference was observed in the sequence similarity for immunogenic and non-immunogenic neoantigens (p=0.511, Figure 2J). The second method considered was the amplitude, where the dissociation constant of the neoantigen:MHC class I complex is adjusted by the dissociation constant of the closest matched human peptide:MHC class I complex. The amplitude adjusts for the regulation of TCR specificities during T cell maturation but also considers that only a normal human peptide capable of binding an MHC class I molecule will significantly impact the immunogenicity of the neoantigen. The amplitude was not significantly different between immunogenic and non-immunogenic neoantigens (p=0.209, Figure 2K).
One final characteristic considered in the development of the NeoScore model was the hydrophobicity of the neoantigen. The hydrophobicity of the neoantigen has been proposed to be associated with greater neoantigen immunogenicity because of the hydrophobicity of key binding pockets in the MHC class I binding groove (55). Mixed results have been reported for the association of hydrophobicity and immunogenicity to date (16, 18, 23, 55). One reason is the use of different methods for hydrophobicity, three of which are considered here. In the first method, the hydrophobicity is calculated as a fraction of the neoantigen residues that are hydrophobic (16). The TESLA consortium reported a significantly lower hydrophobicity of immunogenic neoantigens compared to non-immunogenic neoantigens, which is in the opposite direction as expected. While there was not a statistically significant difference in the neoantigens included in our analysis, the hydrophobicity fraction still had a lower average for immunogenic than non-immunogenic neoantigens (Figure 3, p=0.204). No difference in hydrophobicity was seen in additional datasets from Carreno or Strønen et al., and a significantly higher hydrophobicity of immunogenic neoantigens was observed in the dataset from Ott et al. (Figure 3, p=0.0168) (7, 11, 36). The second method calculated the hydrophobicity fraction using the empirical observations from Chowell et al. to determine which amino acids would increase the likelihood of immunogenicity (55). The method using the observations from Chowell et al. considers both hydrophobicity and other chemical properties such as side chain bulkiness and polarity. No differences in hydrophobicity were seen across the four datasets using the empirical observations from Chowell et al. (Figure 3). The final method considered the hydrophobicity at the anchor residues. A mutation that changed a previously hydrophobic anchor residue to a hydrophilic residue was given a score of zero, while all other changes or no change were given a score of one (23). While there was no statistically significant difference in hydrophobicity for any of the four datasets using the method of Łuksza et al., three out of four datasets showed a greater percentage of immunogenic neoantigens without a loss of hydrophobicity at an anchor residue (Figure 3). Overall, since the method from Łuksza showed the greatest consistency, only this method was included in development of the NeoScore model.
Given the inconsistent association of hydrophobicity with immunogenicity, we explored the association of hydrophobicity and immunogenicity when the data was separated out by HLA allele. Published motifs of the peptides that bind to different HLA alleles have demonstrated distinct differences in the conserved amino acids, suggesting that hydrophobicity may play a greater role in determining binding depending on the HLA allele. For example, published motifs for peptides that bind HLA-A01:01 and HLA-A03:01 contain conserved polar and charged amino acids whereas motifs for peptides that bind HLA-A02:01 contain several conserved hydrophobic amino acids (62). When we assessed the association of hydrophobicity and immunogenicity for each HLA allele independently, immunogenic neoantigens were significantly less hydrophobic than immunogenic neoantigens for HLA-A01:01 using the hydrophobicity method from the TESLA consortium, consistent with the polar and charged conserved amino acids in the peptides that bind this allele (Supplementary Table 1). This data suggests that there might be an advantage to separately evaluating the role of the hydrophobicity characteristic in predicting neoantigens for different HLA alleles. However, due to the small sample sizes available, we were not able to incorporate the effect of hydrophobicity on predicting immunogenicity based on the HLA allele.
Next, the degree of correlation between the characteristics calculated for each neoantigen was assessed. Only two characteristics were significantly correlated: the dissociation constant and the binding stability (Figure 4A). Despite their correlation, there is evidence that the two characteristics both contribute to accurate neoantigen prioritization. While the dissociation constant assesses the affinity of the interaction between the neoantigen and the MHC class I molecule, the stability predicts the length of time that the neoantigen will remain bound. The importance of binding stability is demonstrated in Figure 4B, where clustering of the immunogenic neoantigens in the upper left-hand corner can be observed, indicating the influence of both characteristics in determining immunogenicity.
Capietto et al. recently found that a different set of neoantigen characteristics will influence immunogenicity, if the mutation occurs in an anchor residue compared to mutations in non-anchor residues (63). They demonstrated that, if a mutation occurred in an anchor residue, the amplitude had a greater predictive value than the dissociation constant of the neoantigen:MHC class I complex alone. To assess the influence of the mutation position, the distribution of the amplitude and the unadjusted dissociation constant for neoantigens with a mutation in an anchor or a non-anchor residue were analyzed. To maximize the chances of detecting a significant difference with the relatively low number of immunogenic neoantigens derived from mutations in anchor residues (n=13), the analysis of the impact of mutation’s position was performed across the combination of four datasets (7, 11, 16, 36). As shown in Supplementary Figure 1, no statistically significant difference was observed for the amplitude with either anchor or non-anchor residue mutations. While the anchor-residue mutations did have a higher average amplitude in immunogenic neoantigens, the difference was not statistically significant. In contrast, the dissociation constant was significantly lower in immunogenic neoantigens than non-immunogenic neoantigens. Therefore, there was not a compelling reason to fit separate models for neoantigens with mutations in anchor residues and those with mutations in non-anchor residues. Furthermore, because the amplitude is mathematically dependent on the dissociation constant, the amplitude was not included in the subsequent steps of model development.
Regularized regression approach creates a neoantigen prioritization model, NeoScore, that outperforms existing models that score each neoantigen
While analysis of the distribution of each characteristic determined those with a statistically significant difference between immunogenic and non-immunogenic neoantigens, we next assessed the full set of characteristics to determine which influenced the ability to optimally prioritize SNV and small indel-derived immunogenic neoantigens. A regularized regression is a model-based approach to determine the group of characteristics that are important for discriminating between immunogenic and non-immunogenic neoantigens, whether or not each characteristic is statistically significant. A logistic model using an elastic net-based regularization was unable to identify individual characteristics that are most predictive of immunogenicity by optimizing shrinkage penalties; a consequence that is often seen with a small effective sample size. The effective sample size for these regularized regression methods using models from the binomial family is based on the class with the smallest number of observations, which is the immunogenic neoantigens (n=26). Consequently, a cross-validation approach was applied to select the best subset of neoantigen characteristics. One thousand combinations of 26 non-immunogenic and 26 immunogenic neoantigens were randomly selected, and the elastic net regularized regression was fit on each combination (Figure 5A). Performing 1000 random samples allowed for the examination of the impact of neoantigen characteristics on immunogenicity while adjusting for the small effective sample size. The number of times each characteristic was selected out of the 1000 random combinations was tracked. The dissociation constant, binding stability, and mRNA expression were each selected over 700 of the 1000 times, while no other characteristic was selected over 500 times (Figure 5B). These fits were consistently able to distinguish immunogenic and non-immunogenic neoantigens, as demonstrated by the high density of area under the receiver operator characteristics curve (AUC) values around the mean of 0.861 (25th percentile 0.828, 75th percentile 0.895) (Figure 5C).
Based on the results of the model-based regression approach, a logistic regression model was fit with the dissociation constant, binding stability, and expression in the TESLA consortium data. The final logistic regression model will be called the “NeoScore.” The equation for the model is as follows: NeoScore=−4.1922+2.2958×E−1.8478×Kd+0.7698×S
Where E is the scaled, log10-transformed expression value, Kd is the log10-transformed dissociation constant in units of nanomolar (nM), and S is the log10-transformed stability measured as the half-life of the binding interaction in units of hours. The coefficients for expression and stability were both positive, as expected since higher expression and longer binding times likely increase the chance of a neoantigen to elicit an immune response. The coefficient for the dissociation constant was negative, as expected since a lower dissociation constant indicates a higher binding affinity. These coefficients match the observed directions of change from Figure 2. Raw data from NetMHCpan, NetMHCstabpan, and Salmon can be processed and combined to return a set of neoantigens prioritized by their NeoScore using the following web application: https://bordene.shinyapps.io/MHCI_neoantigen_prioritization/.
Given the large discrepancy between the number of immunogenic and non-immunogenic neoantigens, we assessed the impact of changing the ratio of immunogenic to non-immunogenic neoantigens on the performance of the model. This was done by fitting the logistic regression model on 100 random down-sampled datasets from the TESLA consortium data at 10 different ratios of immunogenic to non-immunogenic neoantigens. Each of the logistic regression models was then applied to the Cohen dataset, and the optimism was calculated by subtracting the AUC on the Cohen dataset from the AUC on the TESLA dataset. The data demonstrated that decreasing the sample size of either immunogenic or non-immunogenic neoantigens increased the optimism, suggesting a less generalizable model. The lowest optimism was obtained when the maximal number of both immunogenic and non-immunogenic neoantigens were used, even though there were not equivalent numbers of immunogenic to non-immunogenic neoantigens. (Supplementary Figure 2).
Once the NeoScore model was fit in the TESLA dataset, model performance was assessed in both the TESLA training dataset and four independent test datasets. In the TESLA consortium dataset, the NeoScore had an AUC of 0.845, which exceeds the AUC of 0.70 needed to be considered a discriminatory model (64). The NeoScore also outperformed the AUC of the Łuksza model (AUC=0.615) (Table II; Figure 6A). The NeoScore was then tested in four additional datasets. The NeoScore outperformed the Łuksza model in the Cohen (0.832 AUC vs. 0.689 AUC) and Strønen datasets (0.681 AUC vs. 0.620 AUC) (35, 36) (Table II; Figure 6B,C). In the Carreno dataset, the NeoScore slightly outperformed Łuksza and the pTuneos hydrophobicity model (0.704 AUC for NeoScore, 0.696 AUC for pTuneos hydrophobicity, and 0.657 AUC for Łuksza) (Table II; Figure 6D). Published results of the immunogenicity scores from the pTuneos model were used (18), as the model was not able to be successfully run with the other datasets. Both the hydrophobicity-only model and the full model provided by pTuneos are included, as the hydrophobicity model outperformed the full model. Similarly, in the Ott dataset, the NeoScore slightly outperformed the Łuksza model (0.609 AUC for NeoScore and 0.575 AUC for Łuksza) (11) (Table II; Figure 6E).
Since the model by the TESLA consortium consists of a single set of recommended thresholds for the neoantigen:MHC class I dissociation constant, neoantigen:MHC class I binding stability, and expression, the NeoScore could not be compared to the TESLA consortium model in terms of the AUC. Therefore, an optimal threshold for the NeoScore was selected that maximized the sum of the sensitivity and specificity in the TESLA dataset and classified the NeoScore into high and low immunogenicity. The reported sensitivity and specificity for each dataset are based on the threshold optimized in the TESLA dataset (−2.478). In the TESLA and Cohen datasets, the NeoScore obtained a greater sensitivity with a lower specificity at the optimal cutpoint compared to the model by the TESLA consortium. Across all the remaining datasets, the sensitivity and specificity of the NeoScore were statistically equivalent to that achieved by the TESLA consortium, as demonstrated by the overlap of the 95% confidence interval (Table II).
Since only a subset of neoantigens were tested for immunogenicity by the TESLA consortium, we assessed the full set of neoantigens predicted to be immunogenic by the NeoScore model. All possible 9 and 10mer neoantigens were generated from each mutation across the five tumors in the TESLA dataset, and the immunogenicity of each neoantigen was prioritized by the NeoScore model. To make the analysis comparable to that done by the TESLA consortium, the neoantigens were assessed for their predicted immunogenicity in the context of the HLA types that were tested by the TESLA consortium. The sensitivity and specificity for the validated immunogenic neoantigens were as reported in Table I. The NeoScore predicted 740 additional immunogenic neoantigens across the five tumors that had not been tested by the TESLA consortium (range 54–310 per patient, data not shown). The large number of untested neoantigens is consistent with the low overlap observed between groups in the original TESLA consortium submissions. Among the 25 submissions, a median of 13% overlap was observed between the neoantigens predicted to be immunogenic (16). The low overlap and large number of untested candidates supports the need for further validation of neoantigen prioritization models.
Despite the high performance of the NeoScore in the TESLA and Cohen datasets, there is a marked decrease in the performance in the Carreno, Strønen, and Ott datasets. A likely cause for the decreased performance is how the immunogenicity was tested in these datasets. TESLA and Cohen both tested for reactive T cells present in the patient with no additional T cell stimulation. In contrast, Carreno et al. administered a dendritic cell vaccine with each of the predicted neoantigens and subsequently tested for an immune response to each neoantigen (7). Strønen et al. exposed PBMCs from healthy patients to dendritic cells transfected with the neoantigen of interest. They then tested for an immune response to those neoantigens (36). Ott et al. immunized patients with pools of long synthetic peptides and then tested for an immune response to each neoantigen that could be generated from the long peptides (11). None of these three methods rely on the expression of the neoantigen in the tumor to activate neoantigen-specific T cells. Therefore, a logistic regression model was fit to the TESLA dataset using only the neoantigen:MHC class I binding stability and dissociation constant. The following abbreviated NeoScore model was obtained: NeoScoreabbreviated=−1.2364−1.2146×Kd+0.9903×S
The coefficients obtained for stability and dissociation constant are comparable to those obtained in the full NeoScore model. The threshold for the abbreviated NeoScore was optimized in the TESLA dataset (−2.856) and then tested in the Cohen, Strønen, Carreno, and Ott datasets. As expected, the abbreviated NeoScore underperformed compared to the full NeoScore model in Cohen and TESLA but outperformed the full NeoScore model in both Carreno and Strønen (Table II; Figure 6). These results suggest that expression predicts neoantigen immunogenicity when priming of T cell responses is dependent on expression of the neoantigen by the tumor. However, when the T cells are stimulated independently of expression by the tumor, the predictive benefit of expression is undermined. Similarly, the performance of the TESLA consortium model, which relies on neoantigen expression, distinctly drops in the Carreno and Strønen datasets (Table II; Figure 6). The poor performance in Ott dataset across models remains unexplained. The Ott dataset did not significantly differ from the other datasets in terms of the distribution of the location of the mutations (anchor vs. non-anchor residues) or the general distribution of the characteristics (data not shown). Overall, elimination of expression enhanced the performance of the NeoScore model when considering the immune response stimulated independently of the tumor.
To further reaffirm the subset of neoantigen characteristics selected, a logistic regression model was fit using all nine neoantigen characteristics from Figure 5B, which did not notably improve the AUC (0.853 for the model with all nine characteristics compared to 0.845 for the NeoScore, data not shown). Furthermore, the optimism is much higher for the model that included all neoantigen characteristics (8.64%) than the NeoScore (2.47%). The optimism indicates the likelihood that the model is overfitting the training data, which would make a less generalizable model. The lack of benefit from the additional characteristics adds support that the subset of characteristics selected is optimal for prioritizing immunogenic neoantigens. Overall, the model with the neoantigen:MHC class I dissociation constant and binding stability, and mRNA expression showed the best potential to consistently separate immunogenic and non-immunogenic neoantigens in validated datasets.
A high maximum NeoScore has a significant association with improved survival in cutaneous melanoma patients treated with immunotherapy
Once the ability of the NeoScore to discriminate immunogenic from non-immunogenic neoantigens with high sensitivity and specificity was established, the association of a single, strongly immunogenic neoantigen with survival in response to immune checkpoint inhibition was assessed. The survival analysis was performed across two datasets, one with treatment with an anti-CTLA-4 monoclonal antibody (21) and the other with anti-PD-1 monoclonal antibodies (29). In both datasets, the cohort was restricted to immunotherapy-naive patients with cutaneous melanoma (21, 29), which allowed us to assess the factors that drive response to immune checkpoint inhibition in melanoma with a high tumor mutational burden and no previous immunoediting. While the range of mutational burden in cutaneous melanoma is wide (1,368–33,591 for the Van Allen dataset and 864–24,292 for the Liu dataset), all samples have a high mutational burden due to UV-induced, DNA mutations. The final cohort sizes were 34 for Van Allen and 53 for Liu. However, the statistical power of the survival analysis is limited by the number of deaths observed in each dataset, resulting in an effective sample size of 22 for Van Allen and 20 for Liu. The NeoScore for each individual neoantigen per tumor was calculated. Since each patient has many neoantigens with a wide range of NeoScores, there are several potential ways to summarize the neoantigen profile for the sake of comparison between patients. Three ways were selected here: 1) the mutational burden, 2) the neoantigen burden, and 3) the highest NeoScore for a neoantigen from the patient, referred to as the “maximum NeoScore.” The mutational burden and neoantigen burden were attempted for the sake of comparison to the literature, while the maximum NeoScore was used in accordance with the principle that a single immunogenic neoantigen can drive the immune response. Survival analysis was then performed separately for the mutational burden, neoantigen burden, and the maximum NeoScore.
Although an optimal NeoScore threshold was determined for the prediction of neoantigen immunogenicity, every patient had at least one neoantigen that exceeded the optimized threshold, indicating that the presence of a predicted immunogenic neoantigen was not sufficient to differentiate the response to immune checkpoint inhibition. Therefore, unique optimal thresholds for survival analysis were determined for the maximum NeoScore using maximally ranked statistics implemented in the MaxStat package and R statistical software (60). Maximally ranked statistics methods implement a search over all possible log-rank test statistics based on thresholds of a predictor variable (here the maximum NeoScore) for the largest standardized log rank test statistic. Since maximally ranked statistic methods optimizes the threshold to detect a difference, performance of the NeoScore and optimal threshold will need to be validated in an independent dataset.
Next, the association of mutational burden and neoantigen burden with survival in response to immune checkpoint inhibition were evaluated. For the sake of consistency, maximally ranked statistics were also used to select the optimal threshold for the mutational burden and neoantigen burden. In the Van Allen dataset, there was no association between mutational burden and progression-free survival in response to immune checkpoint inhibition (p=0.37) (Figure 7A). In the Liu dataset, a high tumor mutational burden was significantly associated with poor progression-free survival (p=0.047) (Figure 7B), which is consistent with the finding of Liu et al. 2019 (29). The neoantigen burden was strongly correlated with the mutational burden in both datasets (Figure 7C and D). On survival analysis, the same results were observed for neoantigen burden as for mutational burden where a high neoantigen burden was not associated with improved progression-free survival in the Van Allen dataset and was associated with decreased progression-free survival in the Liu dataset (data not shown).
When tested with the optimal threshold, patients in the Van Allen dataset with a high maximum NeoScore (above −0.525) had significantly improved progression-free survival (p = 3.5 × 10−4) (Figure 7E). Similarly, in the Liu dataset, patients with a high maximum NeoScore (above −0.152) had significantly improved progression-free survival (p = 8.2 × 10−4) (Figure 7F). In both the Van Allen and Liu datasets a high maximum NeoScore was associated with significantly improved overall survival at the same thresholds (p = 0.002 and p = 0.013, respectively, data not shown). Overall, these results suggest an improved association of the NeoScore with survival following treatment with immunotherapy, compared with tumor mutational burden, in tumors with high tumor mutational burden.
Given the added expense of RNAseq data in a clinical setting, we assessed whether the abbreviated NeoScore would have a similar association with progression-free survival in response to immunotherapy. Therefore, we analyzed the association of the maximum abbreviated NeoScore with progression-free survival in response to treatment in the Liu and Van Allen dataset, as well as an additional lung cancer dataset with no available RNAseq data from Rizvi et al. While the abbreviated NeoScore demonstrated a significant association with progression-free survival in the Van Allen dataset, there was no significant association of the abbreviated NeoScore with progression-free survival in either of the other test datasets (Supplementary Figure 3). The loss of a significant association in the absence of the expression characteristic further emphasizes the importance of expression to the clinical relevance of the NeoScore model.
Discussion:
Prioritization of immunogenic neoantigens is critical for applications to both the development of personalized vaccines and the identification of patients that are likely to benefit from treatment with immune checkpoint inhibition. However, many neoantigen characteristics have been suggested in the literature to date with no consensus on which characteristics impact whether the neoantigen generates a T cell response. Additionally, no model has demonstrated both the ability to score each neoantigen with high sensitivity and specificity and significant association with survival in response to immune checkpoint inhibition. The successes of this study are as follows: 1) identification of those neoantigen characteristics of greatest importance in determining neoantigen immunogenicity, 2) the combination of these characteristics into a single overall immunogenicity score, the NeoScore, with practical applications to personalized vaccine development, 3) integration of the NeoScore into a web application for easy use, and 4) demonstration of the clinical significance of the NeoScore in melanoma.
A model-based statistical prediction approach was used to select the characteristics of SNV and small indel-derived neoantigens that were most predictive of immunogenicity. The dissociation constant and binding stability of the neoantigen:MHC class I complex, and the expression were the combination of neoantigen characteristics best able to discriminate between immunogenic and non-immunogenic neoantigens. While the identification of these three characteristics is consistent with the recent findings of the TESLA consortium (16), it is important to note that the approaches taken by our group and the original group differed in several key ways. First, a completely agnostic approach was applied to the selection of characteristics, including all characteristics that have been suggested in the literature to date; whereas the TESLA consortium began with those shown to have statistical significance. Taking a completely agnostic approach ensured the greatest potential to select the combination of characteristics that maximized the separation of immunogenic and non-immunogenic neoantigens. Second, additional characteristics were included that were not considered by the TESLA consortium, including variant allele frequency and sequence similarity to the closest matched human peptide. Finally, these results were expanded by combining the characteristics into an overall immunogenicity score, the NeoScore. A web application has been made available to calculate the NeoScore or the abbreviated NeoScore and provide a list of neoantigens prioritized by their predicted immunogenicity. The web application is expected to streamline the application of NeoScore for research purposes.
One of the key advantages of the NeoScore is that it allows for the prioritization of neoantigens that may not exceed the single set of thresholds provided by the TESLA consortium. While a threshold for the NeoScore was optimized in the TESLA dataset and demonstrated strong performance across the test datasets, the optimized threshold is necessarily conservative for two reasons. First, all datasets used to build and test the NeoScore consisted of neoantigens that had already been prioritized by the original group. Thus, the NeoScore is trained to discriminate between top candidates. Second, the NeoScore is trained on the natural T cell responses to neoantigens with no stimulation by a therapeutic agent. Treatment with immunotherapy or personalized vaccines may be able to elicit an immune response to neoantigens with a lower NeoScore, Therefore, all neoantigens can be ranked in order of their NeoScore. Researchers applying NeoScore to a new dataset can first rank the predicted neoantigens, then decide on the optimal number of neoantigens to test for a patient based on the unique neoantigen profile of the given patient.
One consideration for future applications of the NeoScore is that it is based on a combination of predictive tools that each come with their own sensitivity and specificity. While we have selected highly rated tools for MHC class I dissociation constants and binding stability, these tools are constantly improving. One example is that the MHC class I binding stability tool that we employed is trained on stability data for 25,000 peptides, which is over an order of magnitude less than the training data available for MHC class I dissociation constant predictions (54). As expanded training data becomes available, the predictive value of each tool is likely to increase, which in turn, will further enhance the predictive value of integrated models such as the NeoScore. Additionally, some recent work has suggested the potential for combining predictions from multiple tools in either a consensus or aggregate approach to further enhance the predictive value of each individual tool (65). Further research is needed into how to optimally aggregate the scores from multiple tools to enhance their application. However, the high performance of the NeoScore in predicting immunogenicity of individual neoantigens and significant association with survival in response to immune checkpoint inhibitors indicates the high performance of each of the individual tools combined to create the NeoScore.
A surprising finding is the lack of association between mutational burden and progression-free survival in the Van Allen dataset and the association of increased mutational burden with decreased progression-free survival in the Liu dataset. These results are inconsistent with the association of increased mutation burden with increased response rate to immune checkpoint inhibition observed across cancer types (27). An explanation for the lack of association of mutation burden with survival following treatment with immune checkpoint inhibition is that the cutaneous melanoma subset consists of all tumors with a high mutational burden. A tumor with a particularly low mutational burden may not have any neoantigens, causing it to have a poor response to immune checkpoint inhibition. However, a high mutational burden alone does not guarantee a good response to immune checkpoint inhibition. As demonstrated by our work, a high maximum NeoScore has an improved association with progression-free survival compared to mutational burden.
The literature to date supports that mutational burden is associated with response to therapy across cancers (27) or across cancer sub-types, but not within tumors with a high mutational burden. Within non-small cell lung cancer, there was a significant increase in response rates and survival in patients with a higher mutational burden than those with a lower mutational burden (25). These results reflect a split between the patients with a mutational signature from smoking carcinogens and those with no evidence of exposure to smoking carcinogens. In contrast, within small cell lung cancers, which are nearly universally associated with smoking, there was a weaker association of mutational burden with response to therapy (22). Three independent studies demonstrated a significant association between high mutational burden and improved response to immune checkpoint inhibition with either anti-CTLA-4 or anti-PD-1 monoclonal antibody treatment in melanoma patients (21, 26, 29). However, these studies included cutaneous, occult, acral, and mucosal melanoma, which may have different mutational profiles. As demonstrated here, there is no association between high mutational burden and improved progression-free survival in the Van Allen and Liu datasets when restricted to cutaneous melanoma. As noted by Snyder et al., the patient in their dataset with the highest number of mutations had minimal or no benefit from anti-CTLA-4 monoclonal antibody treatment. Overall, in combination with prior studies, our results highlight the importance of considering the immunogenicity of the neoantigen in predicting the response to immune checkpoint inhibition.
Despite the successes of our work to date, there are several limitations that highlight areas for future research. For one, the maximum immunogenicity score is not able to account for the response to therapy of all patients. One possible reason that the NeoScore is not able to fully predict treatment response is that the model did not include neoantigens derived from gene fusions, products of noncanonical open reading frames, canonical open reading frames with a frameshift, or large indels. To our knowledge, there is no available dataset that has validated neoantigens derived from these sources. Given that recent work has demonstrated that a single neoantigen from a gene fusion product can drive complete tumor regression (66) and non-canonical proteins disproportionately generate MHC class I binding neoantigens (67), patients misclassified by our model may have had neoantigens from one of these classes of mutations. Additionally, since each of the validated datasets tested the immunogenicity of neoantigens that had already been prioritized by the original group, there may be classes of neoantigens that are immunogenic but were not included in any test dataset. Consideration of additional classes of neoantigens is particularly important given recent evidence that there are classes of neoantigens that have very low binding to MHC class I (68). The inclusion of neoantigens derived from a broader set of mutations may alter the neoantigen characteristics selected as important by our regularized regression approach. For example, neoantigens derived from gene fusions, large indels, or frameshifts are likely to have less sequence similarity to normal human peptides than an SNV or small indel-derived neoantigen, causing characteristics such as the sequence similarity or amplitude to be of greater importance. The need to consider additional types of mutations underscores the importance of generating a validated dataset, including these additional mutations, and repeating characteristic selection.
A second reason that the NeoScore may not be able to account for the response to therapy of all patients is that the model does not consider whether the neoantigens ranked as immunogenic were able to elicit a CD4+ T cell response. Studies have observed that the most effective vaccines are those with neoantigens that elicit combined CD4+ and CD8+ T cell responses (69–73). Additionally, work by Alspach et al. demonstrated the need for MHC class II-restricted neoantigen expression by tumor cells to elicit CD4+ T cell responses and generate effective anti-tumor immune responses both in the absence of therapy and in response to immunotherapy. What is still unknown is if there is any advantage to having a single neoantigen that elicits both a CD8+ and CD4+ T cell-mediated response compared to two independent neoantigens that elicit CD8+ and CD4+ T cell responses, independently. If the ideal circumstance is a single neoantigen capable of stimulating a combined response, an efficient approach may be to test the top MHC class I-restricted neoantigens for their potential to elicit a CD4+ T cell response. However, if two separate neoantigens are equally effective, a separate model for the prioritization of MHC class II-restricted neoantigens may be of greater clinical utility. Overall, there is a need for improved understanding of the interplay between MHC class I- and II-restricted neoantigens in stimulating an effective anti-tumor immune response.
This work has successfully identified the key neoantigen characteristics associated with neoantigen immunogenicity using an agnostic approach that considered all the characteristics suggested by the literature to date. These characteristics have been integrated into a single, overall score, the NeoScore, that predicts the immunogenicity of each neoantigen with high sensitivity and specificity. Finally, a high maximum NeoScore has a significant association with improved survival in response to treatment with immune checkpoint inhibition in cutaneous melanoma. The NeoScore is anticipated to improve neoantigen prioritization for the development of personalized vaccines and the determination of which patients are likely to respond to immunotherapy.
Supplementary Material
1
Acknowledgments:
We thank Dr. Tanya N. Phung for her advice and assistance in comparing the software for the identification of SNVs and indels and integrating the pipeline from raw whole-exome sequencing data to potential neoantigens. We thank Dr. Chi Zhou for his assistance with the pTuneos software. We thank Anngela Adams for her assistance in polishing figures.
Financial Support:
This work was supported in part by the Springboard Initiative from the University of Arizona College of Medicine-Phoenix (K.T.H), the University of Arizona College of Medicine-Phoenix M.D./Ph.D. Program (E.S.B), and a Medical Student Award from the Melanoma Research Foundation Medical Student Award (E.S.B.). This work benefited from support for a related project by the Merit Review Award I01-BX005336 from the United States Department of Veterans Affairs (VA), Biomedical Laboratory Research and Development Service (K.T.H.). The contents do not represent the views of the VA or the United States Government.
Figure 1. Maximum Number of Validated Immunogenic Mutations Identified by the Overlap of Strelka and GATK Mutect2.
Comparison of the number of neoantigens derived from single nucleotide variants (SNVs) and small insertions and deletions (indels) by each of three different programs (Varscan2, GATK Mutect2, and Strelka) in the TESLA consortium dataset. Red dots represent the validated immunogenic neoantigens from each patient. The small circle represents validated immunogenic neoantigens that were not identified by the three programs.
Figure 2. Expression, Dissociation Constant, and Stability are Significantly Different Between Immunogenic and Non-Immunogenic Neoantigens.
(A) Steps in the formation of an MHC class I-restricted immunogenic neoantigen. (B) Computationally predicted neoantigen characteristics included in development of the NeoScore model in comparison to three other models. Expression consists of the gene-level mRNA expression and clonality or variant allele frequency of the mutation. Processing consists of proteasomal cleavage and TAP transport potential. MHC class I binding is assessed through the MHC class I dissociation constant, binding stability, and hydrophobicity (Figure 3) of the neoantigen. The potential to stimulate a T cell response is assessed by the similarity to known T cell epitopes, the sequence similarity to normal human peptides, and the relative MHC class I dissociation constant compared to the dissociation constant of the closest matched normal peptide. All ten characteristics listed under the NeoScore model were considered for inclusion, and those in bold (expression, dissociation constant, and binding stability) were the characteristics selected for inclusion in the final model. (C-K) Comparison of the distribution of the characteristics for immunogenic and non-immunogenic neoantigens. (C) Log10-transformed distribution of mRNA expression. (D) Distribution of variant allele frequency (VAF) calculated by GATK Mutect2. (E-F) Distribution of proteasomal cleavage and TAP transport potential calculated by NetCTLpan. (G) Log10-transformed distribution of the MHC class I to neoantigen dissociation constant calculated by NetMHCpan. (H) Log10-transformed distribution of the MHC class I to neoantigen binding stability calculated by NetMHCstabpan. (I) T cell recognition (TCR) recognition probability calculated using the model from Łuksza et al. (J) Log10-transformed distribution of the BLOSUM62 sequence similarity between the closest matched human peptide and the neoantigen of interest, divided by the length of the neoantigen to normalize. (K) Log10-transformed distribution of the amplitude calculated as the ratio of the dissociation constant of the closest matched human peptide to MHC class I to the dissociation constant of the neoantigen to MHC class I. **:p<0.001, ***:p<10−5
Figure 3. No Method for Calculating Neoantigen Hydrophobicity is Consistently Associated with Immunogenicity.
Comparison of the distribution of hydrophobicity values for the TESLA consortium, Carreno, Strønen, and Ott datasets between immunogenic and non-immunogenic neoantigens. Hydrophobicity was calculated by three methods. The TESLA consortium method is calculated as the number of hydrophobic amino acids divided by the total number of amino acids. The method based on the Chowell et al. empirical data uses the same calculation scheme but considers both hydrophobicity and other chemical properties such as side chain bulkiness and polarity in ranking the neoantigens. The Łuksza et al. method considers the change of an amino acid at the anchor residues compared to the closest matched human peptide. *:p<0.05
Figure 4. Binding Stability and Dissociation Constant are Significantly Correlated but Both Contribute to Immunogenicity.
(A) Correlation between all calculated neoantigen characteristics: mRNA expression (Exp.), Variant Allele Frequency (VAF), proteasomal cleavage potential (Cle.), TAP transport potential (TAP), MHC class I to neoantigen dissociation constant (Kd), MHC class I to neoantigen binding stability (Stab.), TCR recognition probability (TCR), amplitude, defined as the ratio of the MHC class I binding of the closest matched human peptide and the neoantigen (Amp.), and normalized sequence similarity score for closest matched human peptide and the neoantigen (Sim.). The size of the circles indicates the absolute magnitude of the correlation. (B) Correlation of MHC class I dissociation constant with MHC class I binding stability of the neoantigen. Lines show the linear relationship between the characteristics for immunogenic (black) and non-immunogenic (gray) neoantigens, respectively. *:p<0.05
Figure 5. Regularized Regression Approach Selects Dissociation Constant, Expression, and Stability as the Characteristics of Greatest Importance in Prioritizing Immunogenicity.
Regularized (elastic net) regression approach to the selection of neoantigen characteristics. (A) Process used for characteristic selection. 26 immunogenic and 26 randomly selected non-immunogenic neoantigens were isolated 1000 times and fit with an elastic net regression. For every fit, the characteristics selected by the model as well as its performance were tracked. (B) Number of times each characteristic was selected out of 1000 runs. The characteristics included are as follows: MHC class I to neoantigen dissociation constant (Kd), mRNA expression (Expression), MHC class I to neoantigen binding stability (Stability), Variant Allele Frequency (VAF), hydrophobicity of the anchor residues (Hydro), TAP transport potential (TAP), T cell recognition potential (TCR), normalized sequence similarity score for closest matched human peptide and the neoantigen (Similarity), and proteasomal cleavage potential (Cleavage). (C) Distribution of the area under the receiver operator characteristics curve (AUC) values for each of the 1000 fits. The solid black line indicates the mean AUC, and the dashed black lines represent the 25th and 75th percentile, respectively.
Figure 6. NeoScore Outperforms the Model by Łuksza and Performs Equivalently to the Model by the TESLA Consortium.
Comparison of the area under the receiver operator characteristics curves (AUC) for the NeoScore and abbreviated NeoScore compared to the models by Łuksza, the TESLA consortium, and pTuneos. A) Performance of the NeoScore and abbreviated NeoScore in the TESLA consortium dataset (the dataset in which the model was fit). Comparison of the NeoScore and abbreviated NeoScore with existing models in the (B) Cohen dataset, (C) Strønen dataset, (D) Carreno dataset, and (E) Ott dataset.
Figure 7. High Maximum NeoScore has Improved Association with Survival Compared to Mutational Burden.
Comparison of progression-free survival probability within treatment-naive, cutaneous melanoma patients. All splits between the groups were determined using maximally ranked statistics, and p-values were calculated using a log-rank test. A high tumor mutational burden is (A) not significantly associated with survival in the Van Allen data and (B) significantly associated with decreased survival in the Liu data. Mutational burden directly correlates with neoantigen burden in (C) the Van Allen data and (D) the Liu data. Patients with a high maximum (max.) NeoScore have (E) significantly increased survival in the Van Allen data and (F) significantly increased survival in the Liu dataset.
Table I: Summary of datasets used, materials used, and accession information
Dataset Materials used Accession
TESLA Consortium (16) Raw RNAseq and WES data https://www.synapse.org/#!Synapse:syn21048999/wiki/603788
Lists of validated neoantigens and the T cell response Supplementary Table S4
Cohen (35) Raw RNAseq data https://trace.ncbi.nlm.nih.gov/Traces/sra/?study=SRP062169
Lists of validated neoantigens and the T cell response Supplementary Table 2
Strønen (36) Lists of validated neoantigens and the T cell response Supplementary Table S8
Carreno (7) Lists of validated neoantigens and the T cell response Supplementary Tables S1-S3
Ott (11) Lists of validated neoantigens and the T cell response Supplementary Table 4
Van Allen (21) Raw RNAseq and WES data dbGaP accession number (phs000452.v3.p1) https://www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/study.cgi?study_id=phs000452.v3.p1
Liu (29) Raw RNAseq and WES data
Rizvi (25) Raw WES data dbGaP accession number (phs000980.v1.p1) https://www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/study.cgi?study_id=phs000980.v1.p1
Table II: Comparison of the AUC and optimal threshold sensitivities and specificities for the prediction of immunogenic neoantigens using NeoScore vs. other models.
Dataset Model Sensitivity [95% C.I.] Specificity [95% C.I.] AUC
TESLA consortium (n = 347) NeoScore 0.846 [0.769–1.00] 0.738 [0.523–0.844] 0.845
Abbreviated NeoScore 1 0.923 [0.731–1.00] 0.567 [0.530–0.760] 0.772
Łuksza 0.692 [0.462–0.885] 0.561 [0.439–0.763] 0.615
TESLA consortium 0.385 0.941 --------
Cohen (n = 357) NeoScore 0.857 [0.571–1.00] 0.546 [0.494–0.597] 0.832
Abbreviated NeoScore 1 1.00 [1.00–1.00] 0.363 [0.311–0.414] 0.744
Łuksza 1.00 [1.00–1.00] 0.254 [0.211–0.303] 0.689
TESLA consortium 0.571 0.857 ---------
Strønen (n = 57) NeoScore 0.364 [0.091–0.636] 0.889 [0.800–0.978] 0.681
Abbreviated NeoScore 1 0.909 [0.727–1.00] 0.644 [0.511–0.778] 0.794
Łuksza 0.727 [0.455–1.00] 0.422 [0.289–0.556] 0.620
TESLA consortium 0.364 0.867 ---------
Carreno (n = 21) NeoScore 0.444 [0.111–0.778] 0.750 [0.500–1.00] 0.704
Abbreviated NeoScore 1 0.778 [0.556–1.00] 0.833 [0.583–1.00] 0.935
Łuksza 0.778 [0.553–1.00] 0.583 [0.333–0.833] 0.657
TESLA consortium 0.222 1.00 --------
pTuneos hydrophobicity 0.500 [0.125–1.00] 0.857 [0.429–1.00] 0.696
pTuneos full model 0.250 [0.125–1.00] 1.00 [0.143–1.00] 0.536
Ott (n = 165) NeoScore 0.222 [0.056–0.389] 0.878 [0.823–0.932] 0.609
Abbreviated NeoScore 1 0.389 [0.167–0.611] 0.782 [0.714–0.844] 0.597
Łuksza 0.626 [0.544–0.701] 0.500 [0.278–0.722] 0.575
TESLA consortium 0.056 0.959 --------
1. The abbreviated NeoScore omits expression.
Key Points:
MHC class I binding, stability, and expression predict neoantigen immunogenicity.
These characteristics were integrated into the NeoScore to rank immunogenicity.
A high NeoScore is associated with improved response to melanoma immunotherapy.
Declaration of interests: The authors declare no competing interests.
References:
1. Schumacher TN , and Schreiber RD . 2015. Neoantigens in cancer immunotherapy. Science 348 : 69–74.25838375
2. Tran E , Turcotte S , Gros A , Robbins PF , Lu YC , Dudley ME , Wunderlich JR , Somerville RP , Hogan K , Hinrichs CS , Parkhurst MR , Yang JC , and Rosenberg SA . 2014. Cancer immunotherapy based on mutation-specific CD4+ T cells in a patient with epithelial cancer. Science 344 : 641–645.24812403
3. Ward JP , Gubin MM , and Schreiber RD . 2016. The Role of Neoantigens in Naturally Occurring and Therapeutically Induced Immune Responses to Cancer. Adv Immunol 130 : 25–74.26922999
4. Yarchoan M , Johnson BA 3rd , Lutz ER , Laheru DA , and Jaffee EM . 2017. Targeting neoantigens to augment antitumour immunity. Nat Rev Cancer 17 : 209–222.28233802
5. Linette GP , Becker-Hapak M , Skidmore ZL , Baroja ML , Xu C , Hundal J , Spencer DH , Fu W , Cummins C , Robnett M , Kaabinejadian S , Hildebrand WH , Magrini V , Demeter R , Krupnick AS , Griffith OL , Griffith M , Mardis ER , and Carreno BM . 2019. Immunological ignorance is an enabling feature of the oligo-clonal T cell response to melanoma neoantigens. PNAS 116 : 23662–23670.31685621
6. Rausch MP , and Hastings KT . 2017. Immune Checkpoint Inhibitors in the Treatment of Melanoma: From Basic Science to Clinical Application. In Cutaneous Melanoma: Etiology and Therapy. Ward WH , and Farma JM , eds, Brisbane (AU) Chapter 9, Codon Publications.
7. Carreno BM , Magrini V , Becker-Hapak M , Kaabinejadian S , Hundal J , Petti AA , Ly A , Lie WR , Hildebrand WH , Mardis ER , and Linette GP . 2015. Cancer immunotherapy. A dendritic cell vaccine increases the breadth and diversity of melanoma neoantigen-specific T cells. Science 348 : 803–808.25837513
8. Garon EB , Rizvi NA , Hui R , Leighl N , Balmanoukian AS , Eder JP , Patnaik A , Aggarwal C , Gubens M , Horn L , Carcereny E , Ahn MJ , Felip E , Lee JS , Hellmann MD , Hamid O , Goldman JW , Soria JC , Dolled-Filhart M , Rutledge RZ , Zhang J , Lunceford JK , Rangwala R , Lubiniecki GM , Roach C , Emancipator K , Gandhi L , and K.−. Investigators. 2015. Pembrolizumab for the treatment of non-small-cell lung cancer. Europe PMC 372 : 2018–2028.
9. Hilf N , Kuttruff-Coqui S , Frenzel K , Bukur V , Stevanovic S , Gouttefangeas C , Platten M , Tabatabai G , Dutoit V , van der Burg SH , Thor Straten P , Martinez-Ricarte F , Ponsati B , Okada H , Lassen U , Admon A , Ottensmeier CH , Ulges A , Kreiter S , von Deimling A , Skardelly M , Migliorini D , Kroep JR , Idorn M , Rodon J , Piro J , Poulsen HS , Shraibman B , McCann K , Mendrzyk R , Lower M , Stieglbauer M , Britten CM , Capper D , Welters MJP , Sahuquillo J , Kiesel K , Derhovanessian E , Rusch E , Bunse L , Song C , Heesch S , Wagner C , Kemmer-Bruck A , Ludwig J , Castle JC , Schoor O , Tadmor AD , Green E , Fritsche J , Meyer M , Pawlowski N , Dorner S , Hoffgaard F , Rossler B , Maurer D , Weinschenk T , Reinhardt C , Huber C , Rammensee HG , Singh-Jasuja H , Sahin U , Dietrich PY , and Wick W . 2019. Actively personalized vaccination trial for newly diagnosed glioblastoma. Nature 565 : 240–245.30568303
10. Keskin DB , Anandappa AJ , Sun J , Tirosh I , Mathewson ND , Li S , Oliveira G , Giobbie-Hurder A , Felt K , Gjini E , Shukla SA , Hu Z , Li L , Le PM , Allesoe RL , Richman AR , Kowalczyk MS , Abdelrahman S , Geduldig JE , Charbonneau S , Pelton K , Iorgulescu JB , Elagina L , Zhang W , Olive O , McCluskey C , Olsen LR , Stevens J , Lane WJ , Salazar AM , Daley H , Wen PY , Chiocca EA , Harden M , Lennon NJ , Gabriel S , Getz G , Lander ES , Regev A , Ritz J , Neuberg D , Rodig SJ , Ligon KL , Suva ML , Wucherpfennig KW , Hacohen N , Fritsch EF , Livak KJ , Ott PA , Wu CJ , and Reardon DA . 2019. Neoantigen vaccine generates intratumoral T cell responses in phase Ib glioblastoma trial. Nature 565 : 234–239.30568305
11. Ott PA , Hu Z , Keskin DB , Shukla SA , Sun J , Bozym DJ , Zhang W , Luoma A , Giobbie-Hurder A , Peter L , Chen C , Olive O , Carter TA , Li S , Lieb DJ , Eisenhaure T , Gjini E , Stevens J , Lane WJ , Javeri I , Nellaiappan K , Salazar AM , Daley H , Seaman M , Buchbinder EI , Yoon CH , Harden M , Lennon N , Gabriel S , Rodig SJ , Barouch DH , Aster JC , Getz G , Wucherpfennig K , Neuberg D , Ritz J , Lander ES , Fritsch EF , Hacohen N , and Wu CJ . 2017. An immunogenic personal neoantigen vaccine for patients with melanoma. Nature 547 : 217–221.28678778
12. Sahin U , Derhovanessian E , Miller M , Kloke BP , Simon P , Lower M , Bukur V , Tadmor AD , Luxemburger U , Schrors B , Omokoko T , Vormehr M , Albrecht C , Paruzynski A , Kuhn AN , Buck J , Heesch S , Schreeb KH , Muller F , Ortseifer I , Vogler I , Godehardt E , Attig S , Rae R , Breitkreuz A , Tolliver C , Suchan M , Martic G , Hohberger A , Sorn P , Diekmann J , Ciesla J , Waksmann O , Bruck AK , Witt M , Zillgen M , Rothermel A , Kasemann B , Langer D , Bolte S , Diken M , Kreiter S , Nemecek R , Gebhardt C , Grabbe S , Holler C , Utikal J , Huber C , Loquai C , and Tureci O . 2017. Personalized RNA mutanome vaccines mobilize poly-specific therapeutic immunity against cancer. Nature 547 : 222–226.28678784
13. Zacharakis N , Chinnasamy H , Black M , Xu H , Lu YC , Zheng Z , Pasetto A , Langhan M , Shelton T , Prickett T , Gartner J , Jia L , Trebska-McGowan K , Somerville RP , Robbins PF , Rosenberg SA , Goff SL , and Feldman SA . 2018. Immune recognition of somatic mutations leading to complete durable regression in metastatic breast cancer. Nat Med 24 : 724–730.29867227
14. Bjerregaard AM , Nielsen M , Hadrup SR , Szallasi Z , and Eklund AC . 2017. MuPeXI: prediction of neo-epitopes from tumor sequencing data. Cancer Immunol Immunother 66 : 1123–1130.28429069
15. Kim S , Kim HS , Kim E , Lee MG , Shin EC , Paik S , and Kim S . 2018. Neopepsee: accurate genome-level prediction of neoantigens by harnessing sequence and amino acid immunogenicity information. Ann Oncol 29 : 1030–1036.29360924
16. Wells DK , van Buuren MM , Dang KK , Hubbard-Lucey VM , Sheehan KCF , Campbell KM , Lamb A , Ward JP , Sidney J , Blazquez AB , Rech AJ , Zaretsky JM , Comin-Anduix B , Ng AHC , Chour W , Yu TV , Rizvi H , Chen JM , Manning P , Steiner GM , Doan XC , Alliance TTNS , Merghoub T , Guinney J , Kolom A , Selinsky C , Ribas A , Hellmann MD , Hacohen N , Sette A , Heath JR , Bhardwaj N , Ramsdell F , Schreiber RD , Schumacher TN , Kvistborg P , and Defranoux NA . 2020. Key parameters of tumor epitope immunogenicity revealed through a consortium approach improve neoantigen prediction. Cell 183 : 818–834.33038342
17. Wood MA , Paralkar M , Paralkar MP , Nguyen A , Struck AJ , Ellrott K , Margolin A , Nellore A , and Thompson RF . 2018. Population-level distribution and putative immunogenicity of cancer neoepitopes. BMC Cancer 18 : 414.29653567
18. Zhou C , Wei Z , Zhang Z , Zhang B , Zhu C , Chen K , Chuai G , Qu S , Xie L , Gao Y , and Liu Q . 2019. pTuneos: prioritizing tumor neoantigens from next-generation sequencing data. Genome Med 11 : 67.31666118
19. Gubin MM , Zhang X , Schuster H , Caron E , Ward JP , Noguchi T , Ivanova Y , Hundal J , Arthur CD , Krebber WJ , Mulder GE , Toebes M , Vesely MD , Lam SS , Korman AJ , Allison JP , Freeman GJ , Sharpe AH , Pearce EL , Schumacher TN , Aebersold R , Rammensee HG , Melief CJ , Mardis ER , Gillanders WE , Artyomov MN , and Schreiber RD . 2014. Checkpoint blockade cancer immunotherapy targets tumour-specific mutant antigens. Nature 515 : 577–581.25428507
20. Lommatzsch M , Bratke K , and Stoll P . 2018. Neoadjuvant PD-1 Blockade in Resectable Lung Cancer. N Engl J Med 379 : e14.
21. Van Allen EM , Miao D , Schilling B , Shukla SA , Blank C , Zimmer L , Sucker A , Hillen U , Foppen MHG , Goldinger SM , Utikal J , Hassel JC , Weide B , Kaehler KC , Loquai C , Mohr P , Gutzmer R , Dummer R , Gabriel S , Wu CJ , Schadendorf D , and Garraway LA . 2015. Genomic correlates of response to CTLA-4 blockade in metastatic melanoma. Science 350 : 207–211.26359337
22. Hellmann MD , Callahan MK , Awad MM , Calvo E , Ascierto PA , Atmaca A , Rizvi NA , Hirsch FR , Selvaggi G , Szustakowski JD , Sasson A , Golhar R , Vitazka P , Chang H , Geese WJ , and Antonia SJ . 2019. Tumor Mutational Burden and Efficacy of Nivolumab Monotherapy and in Combination with Ipilimumab in Small-Cell Lung Cancer. Cancer Cell 35 : 329.30753829
23. Łuksza M , Riaz N , Makarov V , Balachandran VP , Hellmann MD , Solovyov A , Rizvi NA , Merghoub T , Levine AJ , Chan TA , Wolchok JD , and Greenbaum BD . 2017. A neoantigen fitness model predicts tumour response to checkpoint blockade immunotherapy. Nature 551 : 517–520.29132144
24. McGranahan N , Furness AJ , Rosenthal R , Ramskov S , Lyngaa R , Saini SK , Jamal-Hanjani M , Wilson GA , Birkbak NJ , Hiley CT , Watkins TB , Shafi S , Murugaesu N , Mitter R , Akarca AU , Linares J , Marafioti T , Henry JY , Van Allen EM , Miao D , Schilling B , Schadendorf D , Garraway LA , Makarov V , Rizvi NA , Snyder A , Hellmann MD , Merghoub T , Wolchok JD , Shukla SA , Wu CJ , Peggs KS , Chan TA , Hadrup SR , Quezada SA , and Swanton C . 2016. Clonal neoantigens elicit T cell immunoreactivity and sensitivity to immune checkpoint blockade. Science 351 : 1463–1469.26940869
25. Rizvi NA , Hellmann MD , Snyder A , Kvistborg P , Makarov V , Havel JJ , Lee W , Yuan J , Wong P , Ho TS , Miller ML , Rekhtman N , Moreira AL , Ibrahim F , Bruggeman C , Gasmi B , Zappasodi R , Maeda Y , Sander C , Garon EB , Merghoub T , Wolchok JD , Schumacher TN , and Chan TA . 2015. Cancer immunology. Mutational landscape determines sensitivity to PD-1 blockade in non-small cell lung cancer. Science 348 : 124–128.25765070
26. Snyder A , Makarov V , Merghoub T , Yuan J , Zaretsky JM , Desrichard A , Walsh LA , Postow MA , Wong P , Ho TS , Hollmann TJ , Bruggeman C , Kannan K , Li Y , Elipenahli C , Liu C , Harbison CT , Wang L , Ribas A , Wolchok JD , and Chan TA . 2014. Genetic basis for clinical response to CTLA-4 blockade in melanoma. N Engl J Med 371 : 2189–2199.25409260
27. Yarchoan M , Hopkins A , and Jaffee EM . 2017. Tumor Mutational Burden and Response Rate to PD-1 Inhibition. N Engl J Med 377 : 2500–2501.29262275
28. Miller A , Asmann Y , Cattaneo L , Braggio E , Keats J , Auclair D , Lonial S , Network MC , Russell SJ , and Stewart AK . 2017. High somatic mutation and neoantigen burden are correlated with decreased progression-free survival in multiple myeloma. Blood Cancer J 7 : e612.28937974
29. Liu D , Schilling B , Liu D , Sucker A , Livingstone E , Jerby-Arnon L , Zimmer L , Gutzmer R , Satzger I , Loquai C , Grabbe S , Vokes N , Margolis CA , Conway J , He MX , Elmarakeby H , Dietlein F , Miao D , Tracy A , Gogas H , Goldinger SM , Utikal J , Blank CU , Rauschenberg R , von Bubnoff D , Krackhardt A , Weide B , Haferkamp S , Kiecker F , Izar B , Garraway L , Regev A , Flaherty K , Paschen A , Van Allen EM , and Schadendorf D . 2019. Integrative molecular and clinical modeling of clinical outcomes to PD1 blockade in patients with metastatic melanoma. Nat Med 25 : 1916–1927.31792460
30. Anagnostou V , Smith KN , Forde PM , Niknafs N , Bhattacharya R , White J , Zhang T , Adleff V , Phallen J , Wali N , Hruban C , Guthrie VB , Rodgers K , Naidoo J , Kang H , Sharfman W , Georgiades C , Verde F , Illei P , Li QK , Gabrielson E , Brock MV , Zahnow CA , Baylin SB , Scharpf RB , Brahmer JR , Karchin R , Pardoll DM , and Velculescu VE . 2017. Evolution of Neoantigen Landscape during Immune Checkpoint Blockade in Non-Small Cell Lung Cancer. Cancer Discov 7 : 264–276.28031159
31. Schmidt J , Smith AR , Magnin M , Racle J , Devlin JR , Bobisse S , Cesbron J , Bonnet V , Carmona SJ , Huber F , Ciriello G , Speiser DE , Bassani-Sternberg M , Coukos G , Baker BM , Harari A , and Gfeller D . 2021. Prediction of neo-epitope immunogenicity reveals TCR recognition determinants and provides insight into immunoediting. Cell Reports Medicine 2 : 100194.33665637
32. Hodi FS , Chiarion-Sileni V , Gonzalez R , Grob JJ , Rutkowski P , Cowey CL , Lao CD , Schadendorf D , Wagstaff J , Dummer R , Ferrucci PF , Smylie M , Hill A , Hogg D , Marquez-Rodas I , Jiang J , Rizzo J , Larkin J , and Wolchok JD . 2018. Nivolumab plus ipilimumab or nivolumab alone versus ipilimumab alone in advanced melanoma (CheckMate 067): 4-year outcomes of a multicentre, randomised, phase 3 trial. Lancet Oncol 19 : 1480–1492.30361170
33. Shemesh CS , Hsu JC , Hosseini I , Shen BQ , Rotte A , Twomey P , Girish S , and Wu B . 2021. Personalized Cancer Vaccines: Clinical Landscape, Challenges, and Opportunities. Mol Ther 29 : 555–570.33038322
34. Solomon BJ , Beavis PA , and Darcy PK . 2020. Promising Immuno-Oncology Options for the Future: Cellular Therapies and Personalized Cancer Vaccines. Am Soc Clin Oncol Educ Book 40 : 1–6.
35. Cohen CJ , Gartner JJ , Horovitz-Fried M , Shamalov K , Trebska-McGowan K , Bliskovsky VV , Parkhurst MR , Ankri C , Prickett TD , Crystal JS , Li YF , El-Gamil M , Rosenberg SA , and Robbins PF . 2015. Isolation of neoantigen-specific T cells from tumor and peripheral lymphocytes. J Clin Invest 125 : 3981–3991.26389673
36. Strønen E , Toebes M , Kelderman S , van Buuren MM , Yang W , van Rooij N , Donia M , Boschen ML , Lund-Johansen F , Olweus J , and Schumacher TN . 2016. Targeting of cancer neoantigens with donor-derived T cell receptor repertoires. Science 352 : 1337–1341.27198675
37. Orenbuch R , Filip I , Comito D , Shaman J , Pe’er I , and Rabadan R . 2020. arcasHLA: high-resolution HLA typing from RNAseq. Bioinformatics 36 : 33–40.31173059
38. Bolger AM , Lohse M , and Usadel B . 2014. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics 30 : 2114–2120.24695404
39. Fairley S , Lowy-Gallego E , Perry R , and Flicek P . 2019. The International Genome Sample Resource (IGSR) collection of open human genomic variation resources. Nucleic Acids Res 48 : D941–D947.
40. Li H 2013. Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM. arXiv.
41. Li H , Handsaker B , Wysoker A , Fennell T , Ruan J , Homer N , Marth G , Abecasis G , Durbin R , and Genome S Project Data Processing. 2009. The Sequence Alignment/Map format and SAMtools. Bioinformatics 25 : 2078–2079.19505943
42. Benjamin D , Sato T , Cibulskis K , Getz G , Stewart C , and Lichtenstein L . 2019. Calling Somatic SNVs and Indels with Mutect2. bioRxiv.
43. Koboldt DC , Zhang Q , Larson DE , Shen D , McLellan MD , Lin L , Miller CA , Mardis ER , Ding L , and Wilson RK . 2012. VarScan 2: somatic mutation and copy number alteration discovery in cancer by exome sequencing. Genome Res 22 : 568–576.22300766
44. Saunders CT , Wong WS , Swamy S , Becq J , Murray LJ , and Cheetham RK . 2012. Strelka: accurate somatic small-variant calling from sequenced tumor-normal sample pairs. Bioinformatics 28 : 1811–1817.22581179
45. Wilm A , Aw PP , Bertrand D , Yeo GH , Ong SH , Wong CH , Khor CC , Petric R , Hibberd ML , and Nagarajan N . 2012. LoFreq: a sequence-quality aware, ultra-sensitive variant caller for uncovering cell-population heterogeneity from high-throughput sequencing datasets. Nucleic Acids Res 40 : 11189–11201.23066108
46. McLaren W , Gil L , Hunt SE , Riat HS , Ritchie GR , Thormann A , Flicek P , and Cunningham F . 2016. The Ensembl Variant Effect Predictor. Genome Biol 17 : 122.27268795
47. Hundal J , Carreno BM , Petti AA , Linette GP , Griffith OL , Mardis ER , and Griffith M . 2016. pVAC-Seq: A genome-guided in silico approach to identifying tumor neoantigens. Genome Med 8 : 11.26825632
48. Cunningham F , Amode MR , Barrell D , Beal K , Billis K , Brent S , Carvalho-Silva D , Clapham P , Coates G , Fitzgerald S , Gil L , Giron CG , Gordon L , Hourlier T , Hunt SE , Janacek SH , Johnson N , Juettemann T , Kahari AK , Keenan S , Martin FJ , Maurel T , McLaren W , Murphy DN , Nag R , Overduin B , Parker A , Patricio M , Perry E , Pignatelli M , Riat HS , Sheppard D , Taylor K , Thormann A , Vullo A , Wilder SP , Zadissa A , Aken BL , Birney E , Harrow J , Kinsella R , Muffato M , Ruffier M , Searle SM , Spudich G , Trevanion SJ , Yates A , Zerbino DR , and Flicek P . 2015. Ensembl 2015. Nucleic Acids Res 43 : D662–669.25352552
49. Patro R , Duggal G , Love MI , Irizarry RA , and Kingsford C . 2017. Salmon provides fast and bias-aware quantification of transcript expression. Nat Methods 14 : 417–419.28263959
50. Favero F , Joshi T , Marquard AM , Birkbak NJ , Krzystanek M , Li Q , Szallasi Z , and Eklund AC . 2015. Sequenza: allele-specific copy number and mutation profiles from tumor sequencing data. Ann Oncol 26 : 64–70.25319062
51. Xiao Y , Wang X , Zhang H , Ulintz PJ , Li H , and Guan Y . 2020. FastClone is a probabilistic tool for deconvoluting tumor heterogeneity in bulk-sequencing samples. Nature Communications 11 : 4469.
52. Stranzl T , Larsen MV , Lundegaard C , and Nielsen M . 2010. NetCTLpan: pan-specific MHC class I pathway epitope predictions. Immunogenetics 62 : 357–368.20379710
53. Reynisson B , Alvarez B , Paul S , Peters B , and Nielsen M . 2020. NetMHCpan-4.1 and NetMHCIIpan-4.0: improved predictions of MHC antigen presentation by concurrent motif deconvolution and integration of MS MHC eluted ligand data. Nucleic Acids Res 48 : W449–W454.32406916
54. Rasmussen M , Fenoy E , Harndahl M , Kristensen AB , Nielsen IK , Nielsen M , and Buus S . 2016. Pan-Specific Prediction of Peptide-MHC Class I Complex Stability, a Correlate of T Cell Immunogenicity. J Immunol 197 : 1517–1524.27402703
55. Chowell D , Krishna S , Becker PD , Cocita C , Shu J , Tan X , Greenberg PD , Klavinskis LS , Blattman JN , and Anderson KS . 2015. TCR contact residue hydrophobicity is a hallmark of immunogenic CD8+ T cell epitopes. PNAS 112 : E17540E11762.
56. Vita R , Overton JA , Greenbaum JA , Ponomarenko J , Clark JD , Cantrell JR , Wheeler DK , Gabbard JL , Hix D , Sette A , and Peters B . 2015. The immune epitope database (IEDB) 3.0. Nucleic Acids Res 43 : D405–412.25300482
57. Camacho C , Coulouris G , Avagyan V , Ma N , Papadopoulos J , Bealer K , and Madden TL . 2009. BLAST+: architecture and applications. BMC Bioinformatics 10 : 421.20003500
58. Dilthey AT , Mentzer AJ , Carapito R , Cutland C , Cereb N , Madhi SA , Rhie A , Koren S , Bahram S , McVean G , and Phillippy AM . 2019. HLA*LA-HLA typing from linearly projected graph alignments. Bioinformatics 35 : 4394–4396.30942877
59. Engebretsen S , and Bohlin J . 2019. Statistical predictions with glmnet. Clin Epigenetics 11 : 123.31443682
60. Hothorn T 2007. MaxStat: maximally selected rank statistics.
61. Jurtz V , Paul S , Andreatta M , Marcatili P , Peters B , and Nielsen M . 2017. NetMHCpan-4.0: Improved Peptide-MHC Class I Interaction Predictions Integrating Eluted Ligand and Peptide Binding Affinity Data. J Immunol 199 : 3360–3368.28978689
62. Bassani-Sternberg M , Chong C , Guillaume P , Solleder M , Pak H , Gannon PO , Kandalaft LE , Coukos G , and Gfeller D . 2017. Deciphering HLA-I motifs across HLA peptidomes improves neo-antigen predictions and identifies allostery regulating HLA specificity. PLoS Comput Biol 13 : e1005725.28832583
63. Capietto A-H , Jhunjhunwala S , Pollock SB , Lupardus P , Wong J , Hänsch L , Cevallos J , Chestnut Y , Fernandez A , Lounsbury N , Nozawa T , Singh M , Fan Z , de la Cruz CC , Phung QT , Taraborrelli L , Haley B , Lill JR , Mellman I , Bourgon R , and Delamarre L . 2020. Mutation position is an important determinant for predicting cancer neoantigens. Journal of Experimental Medicine 217 : e20190179.
64. Hosmer DW , Lemeshow S , and Sturdivant RX . 2013. Applied logistic regression, third edition. John Wiley & Sons, Inc.
65. Bonsack M , Hoppe S , Winter J , Tichy D , Zeller C , Kupper MD , Schitter EC , Blatnik R , and Riemer AB . 2019. Performance Evaluation of MHC Class-I Binding Prediction Tools Based on an Experimentally Validated MHC-Peptide Binding Data Set. Cancer Immunol Res 7 : 719–736.30902818
66. Yang W , Lee KW , Srivastava RM , Kuo F , Krishna C , Chowell D , Makarov V , Hoen D , Dalin MG , Wexler L , Ghossein R , Katabi N , Nadeem Z , Cohen MA , Tian SK , Robine N , Arora K , Geiger H , Agius P , Bouvier N , Huberman K , Vanness K , Havel JJ , Sims JS , Samstein RM , Mandal R , Tepe J , Ganly I , Ho AL , Riaz N , Wong RJ , Shukla N , Chan TA , and Morris LGT . 2019. Immunogenic neoantigens derived from gene fusions stimulate T cell responses. Nat Med 25 : 767–775.31011208
67. Ruiz Cuevas MV , Hardy M-P , Hollý J , Bonneil É , Durette C , Courcelles M , Lanoix J , Côté C , Staudt LM , Lemieux S , Thibault P , Perreault C , and Yewdell JW . 2021. Most non-canonical proteins uniquely populate the proteome or immunopeptidome. Cell Reports 34 : 108815.33691108
68. Duan F , Duitama J , Al Seesi S , Ayres CM , Corcelli SA , Pawashe AP , Blanchard T , McMahon D , Sidney J , Sette A , Baker BM , Mandoiu II , and Srivastava PK . 2014. Genomic and bioinformatic profiling of mutational neoepitopes reveals new rules to predict anticancer immunogenicity. J Exp Med 211 : 2231–2248.25245761
69. Alspach E , Lussier DM , Miceli AP , Kizhvatov I , DuPage M , Luoma AM , Meng W , Lichti CF , Esaulova E , Vomund AN , Runci D , Ward JP , Gubin MM , Medrano RFV , Arthur CD , White JM , Sheehan KCF , Chen A , Wucherpfennig KW , Jacks T , Unanue ER , Artyomov MN , and Schreiber RD . 2019. MHC-II neoantigens shape tumour immunity and response to immunotherapy. Nature 574 : 696–701.31645760
70. Bennett SR , Carbone FR , Karamalis F , Miller JF , and Heath WR . 1997. Induction of a CD8+ cytotoxic T lymphocyte response by cross-priming requires cognate CD4+ T cell help. J Exp Med 186 : 65–70.9206998
71. Kreiter S , Vormehr M , van de Roemer N , Diken M , Lower M , Diekmann J , Boegel S , Schrors B , Vascotto F , Castle JC , Tadmor AD , Schoenberger SP , Huber C , Tureci O , and Sahin U . 2015. Mutant MHC class II epitopes drive therapeutic immune responses to cancer. Nature 520 : 692–696.25901682
72. Ossendorp F , Mengede E , Camps M , Filius R , and Melief CJ . 1998. Specific T helper cell requirement for optimal induction of cytotoxic T lymphocytes against major histocompatibility complex class II negative tumors. J Exp Med 187 : 693–702.9480979
73. Xu M , Kallinteris NL , and von Hofe E . 2012. CD4+ T-cell activation for immunotherapy of malignancies using Ii-Key/MHC class II epitope hybrid vaccines. Vaccine 30 : 2805–2810.22386748
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.