Pediatric pan-central nervous system tumor analysis of immune-cell infiltration identifies correlates of antitumor immunity

Abstract

Immune-therapy is an attractive alternative therapeutic approach for targeting central nervous system (CNS) tumors and the constituency of the Tumor Immune Microenvironment (TIME) likely to predict patient response. Here, we describe the TIME of >6000 primarily pediatric CNS tumors using a deconvolution approach (methylCIBERSORT). We produce and validate a custom reference signature defining 11 non-cancer cell types to estimate relative proportions of infiltration in a panCNS tumor cohort spanning 80 subtypes. We group patients into three broad immune clusters associated with CNS tumor types/subtypes. In cohorts of medulloblastomas (n = 2325), malignant rhabdoid tumors (n = 229) and pediatric high-grade gliomas (n = 401), we show significant associations with molecular subgroups/subtypes, mutations, and prognosis. We further identify tumor-specific immune clusters with phenotypic characteristics relevant to immunotherapy response (i.e. Cytolytic score, PDL1 expression). Our analysis provides an indication of the potential future therapeutic and prognostic possibilities of immuno-methylomic profiling in pediatric CNS tumor patients that may ultimately inform approach to immune-therapy.