Purpose: Enfortumab vedotin (EV) is an antibody–drug conjugate (ADC) targeting NECTIN4 (encoded by the PVRL4/NECTIN4 gene) approved for treatment-refractory metastatic urothelial cancer. Factors that mediate sensitivity or resistance to EV are unknown. In this study, we sought to (i) examine heterogeneity of NECTIN4 gene expression across molecular subtypes of bladder cancer and (ii) determine whether NECTIN4 expression mediates EV sensitivity or resistance. Experimental Design: Molecular subtyping and NECTIN4 expression data from seven muscle-invasive bladder cancer clinical cohorts (n = 1,915 total specimens) were used to assess NECTIN4 expression across molecular subtypes. The outcome of the transcriptomic analysis was relative NECTIN4 expression in the consensus molecular subtypes of bladder cancer. Expression of NECTIN4 was validated in bladder cancer cell lines. NECTIN4 was stably overexpressed or knocked down in basal and luminal bladder cancer cell lines and EV drug sensitivity assays were performed, as measured by cell proliferation and clonogenic assays. Results: NECTIN4 expression is heterogenous across molecular subtypes of bladder cancer and significantly enriched in luminal subtypes. NECTIN4 expression is positively correlated with luminal markers GATA3, FOXA1, and PPARG across all cohorts. NECTIN4 expression is both necessary and sufficient for EV sensitivity in luminal and basal subtypes of urothelial bladder cancer cells. Downregulation of NECTIN4 leads to EV resistance. Conclusions: Sensitivity to EV is mediated by expression of NECTIN4, which is enriched in luminal subtypes of bladder cancer. These findings may have implications for biomarker development, patient selection, and the inclusion of molecular subtyping in ongoing and future EV clinical trials. See related commentary by Teo and Rosenberg, p. 4950

Purpose: To characterize immune cell expression among patients with non–muscle invasive bladder cancer (NMIBC) treated with Bacillus Calmette-Guerin (BCG). Experimental Design: Patients with NMIBC treated with intravesical BCG (2008–2015) were identified, and a tissue microarray was constructed using paired pre- and post-BCG bladder samples. Among patients undergoing BCG, cystoscopic evaluation began 3 months after initiating BCG treatment to determine therapeutic response. IHC was performed for CD8, CD4, FoxP3, PD-L1 (SP-142 and 22C3), and PD-1. A full slide review of PD-L1+ staining tumors was performed to characterize PD-L1 and CD8 colocalization. RNA-seq was performed on cored tumors from available specimens. We compared immune cell populations between BCG responders and nonresponders, and between pretreatment and postreatment tumor samples. Baseline PD-L1 staining in the BCG naïve population was then validated in a separate cohort. Results: The final cohort contained 63 pretreatment NMIBC cases, including 31 BCG responders and 32 BCG nonresponders. No differences in CD4, CD8, or FoxP3 expression were identified between responders and nonresponders. Baseline PD-L1 expression (22C3 and SP-142) was observed in 25% to 28% of nonresponders and 0% to 4% of responders (P < 0.01). PD-L1+ cells in BCG nonresponders colocalized with CD8+ T cells. In addition, BCG therapy did not increase PD-L1 gene expression (RNA-seq) or protein levels (IHC). The number of pretreatment CD4+ T cells was very low among PD-L1+ nonresponders (12%) and high among PD-L1− nonresponders (50%, P < 0.01). In a separate cohort of 57 patients with NMIBC undergoing BCG, baseline PD-L1 (22C3) staining was similar (26%). Conclusions: One mechanism of BCG failure may be adaptive immune resistance. Baseline tumor PD-L1 expression predicts an unfavorable response to BCG and if validated, could be used to guide therapeutic decisions.

Visual inspection of histopathology images of stained biopsy tissue by expert pathologists is the standard method for grading of prostate cancer (PCa). However, this process is time-consuming and subject to high inter-observer variability. Machine learning-based methods have the potential to improve efficient throughput of large volumes of slides while decreasing variability, but they are not easy to develop because they require substantial amounts of labeled training data. In this paper, we propose a deep learning-based classification technique and data augmentation methods for accurate grading of PCa in histopathology images in the presence of limited data. Our method combines the predictions of three separate convolutional neural networks (CNNs) that work with different patch sizes. This enables our method to take advantage of the greater amount of contextual information in larger patches as well as greater quantity of smaller patches in the labeled training data. The predictions produced by the three CNNs are combined using a logistic regression model, which is trained separately after the CNN training. To effectively train our models, we propose new data augmentation methods and empirically study their effects on the classification accuracy. The proposed method achieves an accuracy of 92 % in classifying cancerous patches versus benign patches and an accuracy of 90 % in classifying low-grade (i.e., Gleason grade 3) from high-grade (i.e., Gleason grades 4 and 5) patches. The agreement level of our automatic grading method with expert pathologists is within the range of agreement between pathologists. Our experiments indicate that data augmentation is necessary for achieving expert-level performance with deep learning-based methods. A combination of image-space augmentation and feature-space augmentation leads to the best results. Our study shows that well-designed and properly trained deep learning models can achieve PCa Gleason grading accuracy that is comparable to an expert pathologist.