Adaptive Immune Resistance to Intravesical BCG in Non–Muscle Invasive Bladder Cancer: Implications for Prospective BCG-Unresponsive Trials

Bladder cancer is the fifth most common cancer in the United States, and 70% of patients will harbor non–muscle invasive disease (NMIBC; refs. 1, 2). Transurethral resection of bladder tumor (TURBT) followed by Bacillus Calmette-Guerin (BCG) immunotherapy is the standard first-line treatment. Although up to 35% of patients are cured with intravesical BCG, 40% to 60% will have tumor recurrence within 2 years (3). The evidence supporting BCG as a first-line treatment for patients with NMIBC is clear, and based on several randomized trials (4–6). However, the high recurrence rate after BCG has led to a sustained effort to find biomarkers that predict tumor recurrence and BCG response to identify patients that may be better served with new or alternative therapies (7–9).

Immune checkpoint inhibitors have recently been FDA approved in the second-line setting for patients with metastatic and locally advanced urothelial carcinoma, and as first-line therapies among cisplatin-ineligible patients. In 2007, Inman and colleagues demonstrated that PD-L1 is expressed in NMIBC in addition to more advanced disease, and that PD-L1 expression may be a marker of tumor recurrence after BCG (10). This report and others have led to interest in utilizing systemic immune checkpoint blockade alone or with BCG for NMIBC. Three large trials (SWOG 1605 NCT02844816, Keynote 057 NCT02625961, ADAPT NCT03317158) have recently been undertaken examining systemic immune checkpoint inhibition for patients with BCG unresponsive and relapsing NMIBC. However, the mechanism of PD-L1 as a marker of higher tumor stage or BCG resistance has not been established.

Across the spectrum of human malignancy, immune checkpoint expression is associated with adaptive immune resistance (11). CD8⁺ T cells, while present in the tumor microenvironment (TME), are unable to mount a cytotoxic response against the tumor in the presence of immune checkpoint activation. Our group and others have previously shown that BCG initiates widespread T-cell infiltration into the tumor microenvironment, but that these T cells do not appear to be “activated” (12, 13). It remains unclear whether PD-L1 expression correlates with BCG response, and whether BCG regulates PD-L1 expression during T-cell expansion. We thus sought to characterize immune cell infiltration and immune checkpoint expression in a discovery cohort of patients with NMIBC treated with BCG, and validated baseline PD-L1 expression patterns in a second institutional cohort.

We have recently put together an international consortium of urologists to study mechanisms of BCG resistance. In this first study under this group's guidance, patients with treatment-naïve histologically confirmed non–muscle invasive urothelial carcinoma of the bladder treated with TURBT and induction intravesical BCG between 2008 and 2015 were identified from an Institutional Review Board–approved Johns Hopkins bladder cancer database. Studies were performed in accordance with the Declaration of Helsinki. All patient with T1 disease underwent a restaging TURBT. Patients were selected for having available bladder tissue specimen pre- and post-BCG with enough tumor available to obtain a core. A total of 63 of 279 patients treated during this time frame with high-risk NMIBC met this criteria. Specifically, pre-BCG treatment samples were obtained from primary, incident tumors 1–3 months prior to starting a 6-week induction course of BCG. Cystoscopic evaluations were performed 3 months after the first BCG instillation, and those responding to treatment were then evaluated with cystoscopy and urine cytology every 3 months for 2 years, 6 months to 5 years, and annually to 10 years at the discretion of the treating urologist. Posttreatment tumors were obtained 3 months after the first BCG instillation in those with immediate recurrences, or at the time of transurethral resection for a delayed recurrence. All pathologic evaluations were performed at Johns Hopkins Hospital by a genitourinary pathologist (A.S. Baras). All patients included were required to have a minimum follow up of 2 years after diagnosis to define the patients as a “BCG Responder.” A second cohort (n = 9) of “late relapsers” was identified comprising patients whose bladder cancer recurred more than 2 years after initial induction BCG treatment.

BCG response was defined according to standard definitions: patients were deemed BCG unresponsive if they had persistent high-grade T1 disease at the initial 3-month cystoscopy, or had relapsed high-grade NMIBC with or without CIS within 6 months of last exposure to BCG. BCG relapsing patients had recurrent high-risk (high-grade) NMIBC after prior complete response and did not fulfill the BCG-unresponsive definition (3, 14). Patients who received an inadequate BCG induction (less than five of six courses) and those with incomplete follow up information were excluded.

Tissue microarrays were constructed at the Johns Hopkins tissue microarray facility utilizing 1.0 mm cores from marked tumor as described previously (15). Antibodies were acquired from commercial sources, and IHC staining was performed at the IHC pathology core facility as follows: IHC for PD-1 (CellMarque; dilution 1:100), PD-L1 (Spring Bioscience; for clone SP142 dilution 1:100 – DAKO; for clone 22C3 predilute) and markers of tumor-infiltrating lymphocytes (TIL) CD3 (DAKO; dilution 1:100), CD8 (CellMarque; predilute), CD4 (Ventana; predilute), FoxP3 (Abcam; dilution 1:100) was performed on 4-μm–thick paraffin sections of the tissue microarray masterblock on the Ventana Discovery Autostainer (Ventana Medical Systems), Ventana Benchmark Ultra Autostainer (Ventana). Staining was scored in blinded manner with respect to BCG response status by a genitourinary pathologist (A. Matoso). Positive staining cells were scored as a percentage of tumor cells.

For PD-L1 staining, due to the reported discordance among commercial PD-L1 antibodies, both the SP-142 and 22C3 clones were utilized in this study as follows. The initial TMA was stained for PD-L1 using the Sp-142 clone; however, given known patchy/focal distribution of PD-L1 staining, a full slide review was performed to investigate expression in the tumor microenvironment and stroma (16). On this full slide review, both the SP-142 and 22C3 stains were performed. To maintain consistency across antibodies and scoring methods, >5% positive staining was considered positive for both SP-142 and 22C3 (17–19). Given that this definition is utilized primarily for SP-142, whereas >1% has been reported for 22C3, data were also reported for 22C3 using >1% positive staining as the cutoff for positivity (20, 21). Both tumor cells and tumor-associated immune cells were scored as positive according to the combined positive score (CPS) methodology. Briefly, the CPS methodology is a validated scoring algorithm that combines the total number of PD-L1⁺ cells (tumor cells, lymphocytes, and macrophages) in relation to total tumor cells and is reported as percent (22, 23).

CD8 and PD-L1 IHC for immune cell localization was performed by the Johns Hopkins Hospital tumor microenvironment core. CD8 (Leica Biosystems) antibodies were diluted in antibody dilution buffer according to the Bond Polymer Refine Red Kit. PD-L1 (SP142) Spring Biosciences M4420 1:100 diluted in Dako block with an overnight incubation 4°C. A validation cohort was then obtained from Columbia University Medical Center, comprised of 57 BCG-naïve NMIBC. PD-L1 (22C3) and SP142 IHC was performed on these cases to confirm baseline PD-L1⁺ in this cohort.

RNA were isolated from FFPE cores using AllPrep DNA/RNA FFPE Kit (Qiagen) following the manufacturer's instructions. RNA purity and integrity were measured by Nanodrop and Agilent Tapestation, respectively. Transcriptome mRNA sequencing was performed using Ion Torrent's AmpliseqRNA platform (Thermo Fisher Scientific) with a S5XL sequencer (Thermo Fisher Scientific). Briefly, 20 ng of purified RNA were transcribed into cDNA using SuperScript VILO Kit. Then cDNA were amplified using IonAmpliseq Transcriptome human Gene Expression Core panel, followed by the ligation of adapters and barcode to amplicons and purification. Purified libraries were quantified using Ion Library Quantification Kit (Thermo Fisher Scientific) according to the manufacturer's instructions. Libraries were diluted to 100 pmol/L and pooled in set of 8. Pooled libraries were amplified on the ion sphere particles (ISP) using emulsion PCR and enriched on the IonChef (Thermo Fisher Scientific). Template positive ISPs were loaded into Ion 540 chip and run on the S5XL instrument. Primary analysis of RNA-sequencing data was performed using AmpliSeqRNA analysis plugin in the Torrent Suite Software. This plugin aligned the raw sequence reads to human reference genome that contains 20,802 RefSeq transcripts (hg19 Ampliseq Transcriptome_ERCC_V1.fasta) using Torrent Mapping Alignment Program (TMAP). Then, the number of read mapped per gene was counted to generate raw counts file and normalized reads per gene per million mapped reads (RPM) file. To perform the test for differential expression (DE) between responders and nonresponders, the Bioconductor package DESeq2 was used (24).

We compared immune cell populations between BCG responders and nonresponders, and between pretreatment and posttreatment tumor samples of nonresponders. Statistical analysis was performed using Prism 5 (GraphPad). One-way ANOVA tests were conducted and results were considered statistically significant at P ≤ 0.05.

A total of 63 patients with paired pre- and post-BCG samples were included in the TMA. This cohort included 31 BCG responders and 32 BCG nonresponders (95 total samples). These nonresponders could further be subdivided into 14 BCG unresponsive, 13 BCG relapsing, and 5 BCG progressors to muscle invasive disease. Of those who recurred 12 went on to cystectomy, 18 went on to further intravesical therapy, 1 had chemoradiation, and 1 patient progressed to nodal disease and went on to have systemic chemotherapy. No differences in age, gender, or tumor stage at initial TURBT were identified between responders and nonresponders (Table 1). Of the total cohort of 63 patients, 15 (24%) experienced an adverse event, of which 6 were flu-like symptoms, and 9 were lower urinary tract symptoms (dysuria, frequency, urgency).

IHC of the TMA is demonstrated in Fig. 1. In the initial comparison of pretreatment samples from BCG responders versus nonresponders within the TMA, no significant differences in immune cell expression were noted (Table 2). However, given overall weak PD-L1 expression on the TMA, a full slide review to assess tumor microenvironment was performed utilizing the combined positivity score (CPS), a validated scoring system that combines all PD-L1⁺ cells (tumor cells, lymphocytes, and macrophages) in relation to total tumor cells. The full slide review revealed a greater proportion (25% for SP142 clone and 28% for 22C3 clone) of positive PD-L1 expression among pretreatment samples from nonresponders than previously shown in the TMA (13%), with poor concordance (60%) between the TMA and full slide review. PD-L1 expression was significantly increased among pretreatment samples from BCG nonresponders compared with responders in both the SP-142 and 22C3 clones with 96% agreement between the two clones when using the <5% cutoff for positivity (P = 0.001; Fig. 2A and B). A more inclusive definition of PD-L1 (22C3) positivity increased the total number of samples with positive expressing cells but lessened the separation between responders and nonresponders (Fig. 2C). There was no PD-L1⁺ samples among tumors that were pure CIS histology. A separate cohort of 57 BCG-naïve high-risk NMIBC cases from another institution (CUMC) confirmed baseline PD-L1 expression (22C3) to be 26%. Notably, RNA-seq of cored tumor tissue (without associated stroma) did not identify any gene expression changes of immunologic markers between responders and nonresponders (Supplementary Fig. S1).

Among BCG nonresponders, tumor tissue prior to BCG treatment was compared with tumor tissue after BCG (Fig. 3). RNA-seq performed on cored pre/posttreatment tumor pairs demonstrated no changes in immune-associated gene expression (CD274, CD4, CD68, CD8, FOXP3, PDCD1) after BCG (Fig. 3A), with consistent expression within each paired samples (Fig. 3B). Similarly, no changes in the number of CD4⁺ cells, FOXP3⁺ cells, and PD-L1 (SP142) or PD-1 expression were observed. Across all specimens, increased number of CD8⁺ cells was observed after BCG (P = 0.017; Fig. 3C). On a full slide rereview of the TMA, no differences in PD-L1 expression were observed before and after BCG (Fig. 3D).

The cohort of pretreatment samples from BCG nonresponders was studied according to their PD-L1 expression status to understand the coexpression and localization of T cells in the tumor microenvironment. The number of pretreatment CD4⁺ T cells was very low among PD-L1⁺ nonresponders (0% for Sp142, 12% for 22C3) and high among PD-L1⁻ nonresponders (60% for Sp142, 50% for 22C3, P < 0.01; Fig. 4A–C). No major changes in the number of pretreatment CD8⁺ cells was observed between PD-L1⁺ and PD-L1⁻ nonresponders, as CD8⁺ T cells were widespread across the majority of cases. To assess whether tumor recurrence after BCG was associated with adaptive immune resistance, we used IHC to analyze coexpression of CD8 and PD-L1 in pretreatment tissue among BCG nonresponders (Fig. 5). All 7 PD-L1⁺ (Sp142) pretreatment samples had evidence of PD-L1 and CD8 colocalization with increased density of CD8⁺ cells in areas of PD-L1 expression (Fig. 5).

A different cohort of late relapsers (n = 9), defined as a tumor recurrence more than 2 years after completing BCG therapy included 9 paired samples pre- and post-BCG treatment. Similar to complete responders without late recurrences, no pretreatment tumors had PD-L1 expression present (Fig. 2D), and a widespread influx of CD8 T cells was observed. However, two of the 9 patients had evidence of PD-L1 expression late relapsing tumor, and these PDL1⁺ cells also appeared to colocalize with CD8⁺ T cells (Supplementary Fig. S2).

This study provides evidence that intrinsic resistance to BCG may be due to a pretreatment adaptive immune response and immune exhaustion characterized by PD-L1–expressing T cells and regulatory CD4⁺ T cells within the tumor infiltrates. Our study supports the concept that PD-L1–mediated BCG resistance may account for bladder cancer recurrence in 25% of patients. Improving BCG antitumor response has been a major emphasis of intravesical treatment of NMIBC for many years. To improve BCG immunologic effects on urothelial cancer, efforts to elucidate the mechanism of action of BCG have been met with challenges, as nearly all major immune cell subsets have been implicated in BCG activity (25–28). Although it has long been known that BCG initiates recruitment of T cells into the TME, how these T cells have interacted directly with the tumor has not been fully demonstrated (12, 13). Prior work has specifically demonstrated the role of effector and regulatory T cells in predicting response to intravesical BCG; however, few prior analyses have assessed the relationship of these T cells to the immune checkpoint (27). Our group recently described the TME of stage for stage urothelial cancer in the TMA of a different patient cohort (29). This found increased PD-L1⁺ staining with stage and FOXP3/CD8 expression ratio of >1 in intratumoral lymphocytes had lower risk of grade progression (29). However, our previous report did not evaluate mechanisms of BCG response or resistance, nor was staining performed on a whole slide review using multiple PD-L1 antibodies. In this study, we demonstrate that among nearly all tumor samples, BCG elicits a widespread influx of CD8⁺ T cells. However, among nonresponders, 25%–30% of cases show pretreatment colocalization of PD-L1⁺ in areas of high density of CD8⁺ cells. The majority of tumor areas with colocalization of PD-L1⁺ cells and CD8⁺ lymphocytes lacked any CD4⁺ T cells. In contrast, PD-L1 expression was nearly absent among BCG responders, which were enriched with CD8⁺ and CD4⁺ T cells. Finally, we observed a lack of pretreatment CD4⁺ T cells in PD-L1⁺ nonresponders. This highlights another potential mechanism of BCG resistance. The priming of CD4⁺ T cells from naïve to effector T cells induces changes in expression of surface proteins and endothelial receptors. These include chemokines specific to lymphocyte trafficking, notably CXCR3, CXCL9, and CXCL10. Inefficient trafficking of CD4⁺ cells into the tumor microenvironment may be due to downregulation of adhesion molecules on endothelial cells or mismatching of chemokine receptors on T cells and tumor secreted chemokines (27). The mechanism of inefficient CD4⁺ trafficking in BCG nonresponders may another mechanism of BCG resistance and warrants further investigation.

Although most definitions use 2 years as a marker of durable response to BCG in NMIBC, there is a small group that recurs after this period (30). Among these “late relapsers,” similar to complete responders without late recurrences, no pretreatment tumors had PD-L1 expression present in our study. However, 2 of the 9 patients displayed both PD-L1 expression with CD8 colocalization on their late relapsing tumor. Thus, although the late relapsers initially “appear” to be similar to the durable complete responders, and when they recur a portion of them elicit evidence of adaptive immune resistance. That 22% of the late responders demonstrate this is consistent with a similar percentage of nonresponders also demonstrating this resistance phenotype.

The CD8/PD-L1 colocalization has been used as a marker of adaptive immune resistance in which CD8⁺ T cells are recruited to the tumor and secrete IFNγ to lyse tumor cells. In response to IFNγ signaling, PD-L1 is upregulated to resist cytotoxic T-cell activity (11, 18). Taken together, these data suggest that in a subset of BCG nonresponders, CD8⁺ T cells are inactive or exhausted due to an upregulated immune checkpoint.

Adaptive immune resistance is a process first described by Taube and colleagues, during which cytotoxic T cells are recruited to the tumor, and secrete IFNγ as they initiate a cytotoxic antitumor response; these signaling pathways in turn upregulate PD-L1 to evade recognition by cytotoxic, proinflammatory immune cells (31, 32). The concept of PD-1/PD-L1 pathway induction as a key mechanism of adaptive immune resistance led to the development of anti-PD-1/PD-L1 immunotherapy. In particular, the finding that PD-L1⁺ tumors are associated with elevated numbers of TILs has been demonstrated in several other tumor subtypes, and has been used to suggest that immune checkpoint inhibition may especially benefit patients with PD-L1⁺–expressing tumors (11, 33, 34). If one mechanism of BCG failure is PD-L1 upregulation and adaptive immune resistance, then efforts to block the immune checkpoint and “reactivate” BCG-specific T cells, may embody promising therapeutic solutions. Currently, three prospective trials are underway to evaluate immune checkpoint blockade alone and combined with BCG for patients with BCG unresponsive disease. Our data would suggest that 25%–30% of the patients in these trials will have PD-L1+ expressing cells. Our results reflect predominantly papillary/invasive tumors that made up the majority of this study. Although no PDL1 positivity was found in the few pure CIS tumors included in this study, the role of the immune landscape as it relates to pure CIS will be the focus of additional explorations by this consortium.

It is of utmost importance to elucidate whether PD-L1⁺ is a predictive biomarker in NMIBC that can be used to guide therapy or a negative prognostic biomarker that reflects inherently more aggressive biology compared with PD-L1⁻ counterparts. PD-L1⁺ has been shown to be a negative prognostic marker for most cancers (i.e., Merkel cell carcinoma, NSCLC, and breast cancer), with the notable exception of melanoma (35). The role of PD-L1 as a biomarker of disease response in bladder cancer remains controversial. For patients with metastatic disease, Imvigor 2010 did not find improved response rates among patients whose tumors had increased PD-L1 expression on tumor-infiltrating cells (36). In Keynote-045, which used the CPS scoring methodology employed in our analysis, expression of 10% did not correlate with response rates or survival (22). Some early reports across all bladder cancer stages (with majority muscle invasive) found very weak or no checkpoint expression on tumor cells after BCG treatment (10, 37). Like this study, Hashizumi and colleagues relegated their study of PD-L1 expression to NMIBC pairs, and found that among 22 such pairs there was an increase in CD8 and PD-L1 (E1L3N antibody) after BCG (38). That study relied solely on a TMA for PD-L1 expression, which we found to be unreliable in our tumor cohort. Nevertheless, although these increases in immune cell populations after BCG were not found by our larger study that employed multiple PD-L1 antibodies and a full slide review, the concept of immune escape proposed by these authors was validated in our analysis by demonstrating colocalization of PD-L1 and CD8 cells, which is necessary to prove this hypothesis.

One reason why there may exist conflicting results in the association of PD-L1 expression with treatment response is that choices of expression, scoring, and staining procedures all can impact real-world correlations (36). This study attempted to mitigate such variation by performing a whole slide review, and utilizing multiple antibody stains (22C3 and SP142) rather than rely on a single metric. In particular, our finding that TMA cannot be relied upon for PD-L1 expression estimates has been demonstrated by others, particularly because TMA cores may isolate only PD-L1 on tumor cells, and not reflect expression on infiltrating lymphocytes (39). This is one reason we have transitioned to whole section analysis in this and in future studies. Nevertheless, NGS evaluation of PD-L1 positivity did not correlate with the whole slide review of our patient cohort. This may be due to the fact that IHC characterizes tumor and associated stroma, whereas RNA-seq data were obtained from cores of marked tumor and could not fully evaluate the tumor microenvironment. Although there is a compendium of data evaluating the role of PD-L1 IHC as predictive and prognostic biomarkers, how IHC correlates with next-generation sequencing is an important unsolved question with clinical implications. Efforts to clarify the prognostic and predictive roles of PD-L1 in bladder cancer utilizing combined pathologic and NGS will be major translational components of ongoing anti-PD-L1 clinical trials in NMIBC.

One strategy to improve BCG efficacy and encourage T-cell trafficking to the tumor involves priming T cells so that at the time of intravesical induction therapy there are preexisting BCG-specific T cells in the TME. Preclinical studies have shown that parental exposure to BCG before intravesical instillation triggers a profound inflammatory response and accelerates T-cell entry into the bladder urothelium (40). In these preclinical studies, BCG priming translated into dramatic differences in murine survival. As an outgrowth of these discoveries, S1602 (“PRIME”) is a randomized trial comparing the effect of intradermal priming prior to intravesical instillation (41). Additional efforts using polyinosinic:polycytidylic acid [poly(I:C)], a TLR3 agonist, in bladder cancer cells lines and ex vivo T cells and macrophages (data not shown) demonstrate increased MHC class I molecule presentation and augmented BCG responsiveness (42). These important clinical trials and preclinical studies will help clarify the relationship between primed T cells and PD-L1–expressing cells in the TME.

This study does have limitations that should be noted. First, many of the tumor samples were very small and duplicates or triplicates from different parts of the tumor were unable to be obtained for the TMA. Molecular heterogeneity is a potential concern in this study, as it is possible that specific tumor sections were PD-L1⁻ in this study, whereas PD-L1⁺ in undersampled sections of tumor (43). For this reason, we would not suggest interpreting our results as evidence to withhold BCG for patients with PD-L1⁺ tumors. Rather we view this as major next step to understanding BCG resistance mechanisms and in doing so enhancing our ability to overcome them. Furthermore, our finding that PD-L1⁺-expressing cells were colocalizing with CD8⁺ cells and devoid of CD4⁺ cells is not affected by potential PD-L1 heterogeneity and undersampling. Thus, our main conclusions would not change and are valid. Furthermore, the SP142 (Ventana) assay is the most utilized in bladder cancer clinical trials; however, it has been criticized for its inferior sensitivity compared with other PD-L1 antibodies (17–19, 44). Given these known differences among the different clones of PD-L1 antibodies, we conducted the project using two different clones and obtained similar results with both. Furthermore, our results were validated by a separate external cohort.

In conclusion, in a carefully annotated cohort of patients with NMIUC undergoing intravesical BCG, we identified a subset of tumors expressing PD-L1 from patients with poor response to BCG therapy. Our data suggest that the mechanism of BCG nonresponse among PD-L1⁺-expressing tumors could be related to adaptive immune resistance. These results suggest that a subset of patients with BCG nonresponsive bladder cancer may benefit from simultaneous immune checkpoint inhibitors and BCG therapy.

D.J. McConkey reports receiving commercial research grants from AstraZeneca and Ranier Pharmaceuticals and reports receiving speakers bureau honoraria from Janssen and H3 Biomedicine. A.M. Kamat is an employee/paid consultant for IBCG, TMC Innovation, Theralase, BioClin Therapeutics, Cepheid, Medac, Asieris, Pfizer, Abbott Molecular, US Biotest, Ferring, Imagin, Cold Genesys, Roviant, Sessen Bio, Merck, Bristol-Myers Squibb, Eisai, Arquer, MDx Health, AstraZeneca, and Photocure; reports receiving commercial research grants from Merck, Bristol-Myers Squibb, Photocure, and FKD Industries; and holds ownership interest (including patents) in CyPRIT with UT MD Anderson Cancer Center. R.S. Svatek is an employee/paid consultant for GoldGenesys, Ferring Pharmaceuticals, and Photocure. J.J. Meeks is an employee/paid consultant for Merck, AstraZeneca, Ferring, and Janssen, and reports receiving commercial research grants from Epizyme, AbbVie, and Tesaro. S.P. Lerner is an employee/paid consultant for Anchiano Therapeutics, Roche/Genentech, Ferring, Merck, QED, Verity, and Vaxiion; reports receiving commercial research grants from Roche/GNE, JBL, Viventia, Vaxiion, UroGen, FKD, and ENDO; reports receiving speakers bureau honoraria from Nucleix, Dava Oncology, and MSD Korea; and holds ownership interest (including patents) in the Broad Institute. C.P. Dinney is an employee/paid consultant for FKD Therapies OY and NCI; reports receiving commercial research grants from NCI; and reports receiving speakers bureau honoraria from ASCO-SITC, Johns Hopkins Hospital, and Dana Farber Harvard Cancer Center. P.C. Black is an employee/paid consultant for Roche, Merck, AstraZeneca, Bristol-Myers Squibb, Ferring, and Janssen, and reports receiving commercial research grants from iProgen and GlaxoSmithKline. C.G. Drake is an employee/paid consultant for Bayer, Bristol-Myers Squibb, Compugen, F-Star, Genocea, Janssen, Merck, Merck-Serono, Pfizer, Pierre Fabre, Roche/Genentech, Shattuck Labs, Tizona Therapeutics, and Werewolf; reports receiving speakers bureau honoraria from Bristol-Myers Squibb; and holds ownership interest (including patents) in Harpoon, Compugen, Tizona, Shattuck Labs, Urogen, and Kleo. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M. Kates, A. Matoso, D.J. McConkey, A.M. Kamat, S.P. Lerner, C.G. Drake, T.J. Bivalacqua

Development of methodology: M. Kates, A. Matoso, D.J. McConkey, T.J. Bivalacqua

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Kates, A. Matoso, A.S. Baras, M.J. Daniels, A. Brant, N. Mikkilineni, D.J. McConkey, S.P. Porten, P.C. Black, J.M. McKiernan, C. Anderson, T.J. Bivalacqua

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Kates, A. Matoso, W. Choi, A. Brant, N. Mikkilineni, D.J. McConkey, R.S. Svatek, P.C. Black, T.J. Bivalacqua

Writing, review, and/or revision of the manuscript: M. Kates, A. Matoso, W. Choi, A.S. Baras, M.J. Daniels, A. Brant, N. Mikkilineni, D.J. McConkey, A.M. Kamat, R.S. Svatek, S.P. Porten, J.J. Meeks, S.P. Lerner, C.P. Dinney, P.C. Black, J.M. McKiernan, C. Anderson, C.G. Drake, T.J. Bivalacqua

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Kates, M.J. Daniels, K. Lombardo, N. Mikkilineni, R.S. Svatek

Study supervision: D.J. McConkey, T.J. Bivalacqua

M. Kates was supported by a Clinician Scientist Development Grant, CSDG-19-001-01-CCE, from the American Cancer Society.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.