Abstract

Growing evidence indicates that innate immune pathway activation is critical for responses to immunotherapy and overall cancer prognosis. It has been posited that innate immunity in the tumor microenvironment can be driven by derepression of endogenous repetitive element RNA. The ability to characterize these species can potentially provide novel predictive biomarkers for tumor immune responses and a mechanistic basis for elements of innate activation by tumors. We first compared total RNA and poly(A)-capture protocols applied to tumor RNA-sequencing to detect non-RNA coding transcriptomes in the TCGA. While the poly(A) protocol efficiently detects coding, most non-coding genes, and much of the LINE/SINE/ERV repeat repertoire, we found that it fails to capture overall repeat expression and co-expression. The probing of total RNA expression reveals distinct repetitive co-expression subgroups. Secondly, we found that total repeat element expression delivers the most dynamic changes in samples, which may serve as more robust biomarkers of clinical outcomes. Finally, we show that while expression of ERVs, but not other immunostimulatory repeats such as HSATII, is associated with response to immunotherapy in a cohort of patients with urothelial cancer treated with anti-PD-L1 therapy, global repeat derepression strongly correlates with an immunosuppressive phenotype in the microenvironment of colorectal and pancreatic tumors. We validate in situ in human primary tumors, associating the immunosuppressive phenotype with HSATII expression. In conclusion, we demonstrate the importance of analyzing repetitive element RNAs as potential biomarkers of response to immunotherapy and the need to better characterize these features in next generation sequencing protocols.