Comprehensive molecular profiling of metastatic melanoma to predict response to monotherapy and combination immunotherapy.

9511 Background: Several factors have been proposed as biomarkers for response to PD1 therapy, including tumor mutational burden (TMB), immune gene expression, PD-L1 expression and TILs, while few specific mechanisms of resistance have been identified. The relative importance of these factors or detailed examination of biomarkers of response to combination immunotherapy have yet to be explored. Methods: Cutaneous metastatic melanoma (MM) patients (pts) treated with anti-PD-1 (PD1) +/- anti-CTLA-4 (CTLA4) were selected. Pre-treatment tumors underwent whole genome sequencing (WGS), RNA sequencing (RNAseq) and immunohistochemistry (IHC; TILs and PD-L1). Results: Tumors from 77 pts treated with PD1 (n = 53) or PD1+CTLA4 (n = 24) underwent WGS. Higher TMB (p = 0.0001), lower structural variant (SV) burden (p = 0.001) and higher neoantigen load (p = 0.001) were associated with response. There was no difference in the expression of specific genes reported to confer resistance (JAK1/2, PTEN or BAF/PBAF complex members) or response (SERPINB3/4, ARID) in responders vs non-responders. RNAseq was performed on 53/77 samples; IFN and TCR signalling pathways were enriched in responders. Cytolytic activity (CYT, p = 0.002), T cell proportion estimated by CIBERSORT (p = 0.002) and confirmed by IHC (p = 0.033), and PD-L1 expression (IHC, p = 0.026) were also higher in responders. Multivariate analysis including DNA (TMB, SV count), RNA (six gene IFN expression signature - IFNG.6; effector T cell gene expression signature; chemokine gene expression; CYT), IHC (PD-L1, TILs) and clinical factors (sex, age, RECIST Sum of diameters, LDH) identified TMB and IFNG.6 as independent predictors of response (AUC = 0.83). 15 outliers with discordant molecular features and clinical outcomes had varying profiles, including 5 non-responders with high TMB, but low IFNG.6 expression, suggestive of a failure of immune activation. Conclusions: Comprehensive clinical and genomic analysis demonstrated that TMB and IFNG expression independently predict response, suggesting defects in both immune recognition or activation in non-responders.