A classification system for urothelial carcinoma (UC) defined by genomic drivers and the tumor microenvironment (TME) is predictive of immunotherapy response.

Abstract

4591 Background: UC is associated with high recurrence rates, progression, and resistance to platinum-based therapy. Checkpoint inhibitors (CPIs) are often used for treating UC, but predictive biomarkers that characterize response are lacking in the majority of patients. Defining the TME is essential to understanding patient response to CPIs. Employing a transcriptome-based classification platform, we sought to identify predictive and prognostic subtypes of UC using malignant cell and TME features. Methods: We collected a metacohort of 2,418 UC samples from 14 publicly available datasets, one of which had atezolizumab response data (IMvigor210). Using the methodology described in Bagaev et al. 2021, we selected 28 signatures composed of specific gene sets reflecting distinct cellular processes. Analysis of signature expression and unsupervised clustering was performed. Results: We identified 7 recurring novel UC subtypes (Table 1) with unique genomic and molecular characteristics. The subtypes and key findings include: an immune desert (D) subtype characterized by genomic instability high HER2 expression; an immune desert, FGFR-altered (D-FGFR) subtype with FGFR alterations; an immune enriched (IE) subtype with an enriched TME and high CPI response; a fibrotic (F) subtype with a mesenchymal TME, and strong TGFβ signaling; an immune enriched, fibrotic (IE-F) subtype with a mixed TME and a high CPI response rate; a fibrotic, basal (F-B) subtype with a mesenchymal TME and minimal genomic targets, and a neuroendocrine-like (NE-L) subtype with a high CPI response rate. Conclusions: UC can be classified into 7 subtypes with distinct prognoses, CPI response rates, and druggable targets using malignant cell and TME profiling. Patients with IE, IE-F, and NE-L UCs may be good candidates for CPIs. D UCs may benefit from HER2-i and PARP-i, while FGFR-i might be more suitable for D-FGFR UCs. TGFβi and PARPi may be effective for F UCs, but F-B UCs have no targetable findings. These findings warrant additional investigation for clinical translation. [Table: see text]