Cnsc-40. harnessing evolutionary constraint identifies potential non-coding drivers in brain tumors

Abstract The somatic mutational landscape of brain tumors has primarily focused on coding mutations for driver gene discovery. Non-coding mutations can drive cancer by targeting regulatory elements and be valuable biomarkers and therapeutic targets. However, their identification and functional impact have been elusive. Here, we apply evolutionary constraint (phyloP scores, a constraint metric from sequence alignment of 241 mammals) to whole genome sequencing data from three different brain tumors; pilocytic astrocytoma (PA), medulloblastoma (MB), and glioblastoma (GBM) to identify non-coding constraint mutations (NCCMs) with regulatory potential. NCCM rates (phyloP ≥ 1.2) differ significantly between these cancers, with low rates of NCCMs in PA reflecting its benign feature and high rates in MB and GBM, in support of their aggressivity and heterogeneity. Many of the NCCMs affect genes that are developmentally regulated and expressed in the brain, hence they could be potential drivers. In PA, a high NCCM frequency only affects the BRAF locus, while in MB, over 500 genes have high levels of NCCMs. Intriguingly, many of these genes are associated with different age of onset, such as NUAK1 in adult patients and HOXB1 in young patients, pointing to different molecular pathways in patient groups. NCCMs in the HOXB locus altered the expression of multiple genes of the cluster in MB cells. In GBM NCCMs with regulatory potential cluster as hotspots in PTEN, LRFN5 and MIR99AHG. We observe an enrichment of NCCMs in the regulatory regions of transcription factors including FOXA1, HOXA9 and EBF1. We conclude that since evolutionary constraint is agnostic to cell type or disease stage, it is a powerful predictor of function. Understanding changes in the non-coding cancer genomes will lead to more robust patient stratification of future treatment.