Immune gene expression, survival outcome and response to PD-1/PD-L1 blockade: A TCGA pan-cancer analysis.

3033 Background: Immune infiltration predicts survival in particular cancer-types. However, differences in immune profiling and associations with patient outcomes among multiple cancer-types have not been fully explored. Methods: Clinical and RNA-seq data from 9,761 tumors representing 30 cancer-types were obtained from TCGA project. The expression of 35 immune-related gene signatures tracking various cell-types was evaluated, including PD1, CTLA4, CD8A and CD4 genes. Association of each signature with overall survival (OS) was evaluated using Cox models. Correlations between each signature and the overall response rates (ORR) following anti-PD1/PDL1 monotherapy in each cancer-type (as reported in the literature) were evaluated across 16 cancer-types. Results: In the combined dataset, 15 out of 35 signatures (43%) mostly tracking cytotoxic T-cell response (e.g. CD8-T-cell, PD1 and CD8A) were highly correlated (mean coefficient [cc] = 0.69). The top 5 cancer-types with the highest percentages of samples expressing cytotoxic T-cell response signatures were lung adenocarcinoma (LUAD), thymoma, testicular cancer, melanoma and basal-like breast cancer. The top 5 cancer-types with the lowest percentages were chromophobe renal cell, pheochromocytoma/paraganglioma, prostate cancer, adrenocortical cancer and uveal melanoma. In terms of OS, expression of T-cell response signatures (as a continuous variable) was consistently associated with better outcomes in melanoma, endometrial adenocarcinoma, head and neck carcinoma, hepatocarcinoma, bladder cancer and LUAD, and with worse outcomes in low grade glioma, uveal melanoma, glioblastoma and clear cell renal carcinoma. Finally, high expression of CD8A gene, CD8-T-cell signature and PD1 gene was found highly correlated (cc = 0.68, 0.62 and 0.77, P< 0.001) with ORR following anti-PD1/PDL1 therapy. Conclusions: Significant differences in the expression of immune signatures exist within and across cancer-types. Whereas the association of cytotoxic T-cell response with OS might be dependent on cancer-type, response to anti-PD1/PDL1 checkpoint inhibitors seems associated with the absolute levels of pre-existing cytotoxic T-cell response.