Allelic expression patterns of imprinted and non-imprinted genes in cancer cell lines from multiple histologies.

Abstract

Background: Imprinted genes are epigenetically regulated in normal tissues to follow monoallelic expression according to the parent of origin of each allele. Some of these patterns are dysregulated in cancer.

Results: We developed a novel computational multi-omic pipeline to evaluate monoallelic and biallelic expression patterns based on matched RNA-seq expression data, whole-exome sequencing information, and copy number data. We analyzed allelic expression of the entire genes, individual isoforms, and each exon of 59,283 autosomal protein-coding and ncRNA genes, with a focus on 94 genes previously reported to be imprinted. We analyzed 108 cell lines from 9 different tumor histologies using molecular data from the DepMap Portal for the Cancer Cell Line Encyclopedia. Allelic expression patterns of imprinted genes and isoforms in tumor cells were variable. We also identified additional genes and isoforms with predominantly monoallelic expression due to a variety of potential mechanisms. We provide a novel public dataset of transcriptome-wide allelic expression patterns in cell lines from diverse tumor categories, which can serve as a resource for future cancer studies. We examined associations of in vitro cell line response to antitumor agents and repurposed drugs with allelic patterns and overall levels of isoform expression of imprinted genes and of additional genes with predominantly monoallelic expression. Drug response was associated with isoform expression patterns of multiple imprinted genes including CPA4, DGCR6, DNMT1, GNAS, GRB10, H19, NAA60, OSBPL5, PHACTR2, and ZFAT, predominantly monoallelically expressed MAP2K5 and BCLAF1, and additional predominantly monoallelically expressed genes. Multiple associations may be related to mechanisms of drug activity, including associations between the response to the DNA damaging agents and allelic expression of ZFAT, CDC27, and BCLAF1 isoforms, and the response to inhibitors of multiple signaling pathways with expression patterns of GNAS isoforms.

Conclusions: Tumor cells have a range of monoallelic and biallelic expression patterns in both imprinted and non-imprinted genes and are likely affected by the complex interplay among changes in allelic expression, sequence variants, copy number changes, and expression changes of biologically important genes. Multiple isoform-specific patterns of allelic expression were associated with drug response, indicating complex mechanisms of cancer chemoresistance.