Abstract H3K27-altered diffuse midline glioma (DMG) is a pediatric-type high-grade glioma (pHGG) diagnosed in children and young adults with dismal prognosis. The somatic mutational landscape is defined by driver mutations in histone H3 genes (H3.1K27M, H3.3K27M) and partner mutations in p53, PI3K/RTK, and MAPK signalling genes. Rare germline variation in cancer predisposition genes has been recently discovered in DMG patients, yet functional genetic variation in the non-coding genome and its impact on tumorigenesis is so far unknown. Here, we aim to investigate genome-wide genetic variation in DMG and study the functional impact of genetic variation on transcriptional gene regulation. To identify and characterise cis-regulatory variation in cancer, we developed novel computational pipelines to measure allele-specific gene expression (ASE) in cancer cell lines. We decoupled the impact of somatic DNA copy number variation on gene dosage and estimated ASE driven by cis-regulatory variation. We studied 30 H3K27-altered DMG and 10 H3WT pHGG cell lines and identified high rates of ASE for protein-coding genes (10-20%). Somatic copy number alteration was the primary mechanism of transcriptional dysregulation and explained up to 80% of ASE in DMGs. After accounting for genome instability, the rate of ASE was significantly reduced (9-11%) and not different between DMG and pHGG cell lines. This suggests that genome-wide loss of H3K27me3 does not unmask additional cis-regulatory variation in DMGs. We discovered ASE for DMG driver genes in 67% of cell lines and observed an enrichment of ASE in known imprinted genes and genes affected by loss-of-function mutations. We are currently performing quantitative trait locus (cis-eQTL) and allele-specific chromatin mapping to discover the mechanisms that drive ASE and to fine-map functional non-coding genetic variation in DMG/pHGG. Our study demonstrates pervasive allelic imbalance of gene expression in pHGGs driven by genome instability, genetic variation in protein coding genes, and cis-regulatory variation.
Delevaux et al. (Tue,) studied this question.