Combined high-resolution H3K27ac epigenomic and single-cell transcriptional profiling as a signature predictive of response to neoadjuvant immune checkpoint inhibitors (ICI) in urothelial cancer (UC).

4595 Background: Multiple molecular predictors have been associated with response to ICI in metastatic UC (mUC), including clonal TMB, APOBEC mutagenesis and multiple RNA signatures. Tumor-intrinsic and tumor-extrinsic (microenvironmental) properties can drive resistance to PD-1/PD-L1 inhibitors, although the factors that determine resistance specifically in MIBC remains incompletely characterized. Epigenomic analysis combined with single cell (sc) technology can reveal mechanisms of response involving both cancer cells and the microenvironment. Methods: We identified two MIBC patients’ cohorts from Dana-Farber Cancer Institute (DFCI) and from Hospital Trias i Pujol in Spain treated with neoadjuvant therapy with ICI with tissue available pre- and post- treatment. Molecular profiling was performed with H3K27ac in Formalin-Fixed Paraffin-Embedded (FFPE) blocks and scRNAseq from Fresh Frozen (FF) and FFPE tissues. For the analysis, we performed peak calling, unsupervised analysis (using our internal pipeline), comparison of enhancers with a published datasets (using Cistrome toolkit) and sc analysis using Seurat. Results: 35 patients with pre and post available samples were included in the analysis. The H3K27ac enhancer analysis revealed a signature predictive of treatment response to immunotherapy. Comparison with publicly available ChIP-seq datasets showed that enhancers more active in pre-treatment samples in responders were highly represented in lymphoid lineages. In contrast, the enhancers less active in responders were associated with squamous differentiation. When translated to a gene expression signature, we found enhancers at novel and known genes described in UC that included CYTOR, TGFBR2 and IKZF1( involved in immunotherapy response) and FYN, PDE4B and GPR174 involved in cell migration, microenvironment and EMT among the top ten genes. We projected the gene signature over a scRNAseq atlas resulting from 20 patients that includes all inflammatory lineages, and found an increase within specific T cell populations and a reduction of B and plasma cells in responders as compared to non-responders in the pre-treatment samples. Conclusions: Our combined epigenetic and scRNAseq strategy revealed the existence of specific microenvironment and tumor states in pre-treatment samples from MIBC. We also provided a novel gene signature predictive of response. Prospective validation is ongoing.