ADC target signatures in urothelial carcinoma: A transcriptomic framework for precision therapy.

Abstract

821 Background: Nectin-4 targeting antibody drug conjugate (ADC) enfortumab vedotin (EV), in combination with pembrolizumab, is the first-line treatment for patients with metastatic or locally advanced urothelial cancer (UC). However, optimal treatment strategies for non-responders remain an unmet clinical need. We analyzed tumor transcriptomes to identify ADC target expression profiles which may inform treatment selection for these patients. Methods: We conducted a literature search to identify ADC targets approved or under investigation for UC. Molecular targets of interest included Nectin-4 ( NECTIN4 ), Her2 ( ERBB2 ), Trop-2 ( TACSTD2 ), DLL3, B7-H3 ( CD276 ), FGFR3, and PDL1 ( CD274 ). We interrogated The Cancer Genome Atlas (TCGA) dataset, comprising tumors from 434 patients with UC and 19 normal bladder tissue samples. Hierarchical clustering was used to identify recurrent ADC target expression profiles. To validate our findings, we repeated the analysis on an independent cohort of 478 patients with UC from the Oncology Research Information Exchange Network (ORIEN) using R version 4.5.1. Results: Two mutually exclusive patterns of ADC target expression were identified in the TCGA dataset. Cluster 1 (70.2%, n = 318) was characterized by overexpression of ERBB2 , FGFR3 , NECTIN4 and TACSTD2 . Cluster 2 (29.8%, n = 135) was characterized by overexpression of DLL3 , CD276 , and CD274 . The two clusters of ADC targets demonstrated a significant negative correlation (Spearman rho = -0.40, p < 0.0001). Normal bladder tissue did not cluster according to ADC gene expression profiles. In the ORIEN dataset, a similar pattern was observed: where the seven target genes were separated into two clusters. One cluster showed overexpression of FGFR3 , NECTIN4 , and TACSTD2 , while the other showed overexpression of DLL3 , CD276 , CD274 and ERBB2 . The enrichment scores of the two clusters were also negatively correlated (Spearman rho = -0.79, p < 0.0001) in the ORIEN dataset. Notably, ERBB2 clustered with the second group in ORIEN, suggesting biological variability. Conclusions: Through interrogation of TCGA data and validation in an independent ORIEN cohort, we identified two mutually exclusive ADC target expression profiles in UC. Tumors lacking overexpression of NECTIN4 , TACSTD2 , or FGFR3 were enriched for DLL3 , CD276 and CD274 expression, suggesting alternative ADC targets for non-responders to current therapies. The differential ERBB2 clustering highlights potential biological variability of this target. These findings warrant prospective proteomic validation and may inform rational treatment selection beyond EV.