Revisiting Evidence for Epigenetic Control of Alternative Splicing

Abstract

Alternative splicing is crucial for increasing eukaryotic cell transcriptome and proteome diversity. Changes in alternative splicing play a key role in cell differentiation and tissue development, and aberrations in this process have been associated with diseases. Despite its importance, the exact mechanisms for regulating alternative splicing are poorly understood. Several epigenetic marks, such as histone modification H3K36me3, have previously been associated with changes in alternative splicing. Here, we leverage the EpiATLAS data set to systematically re-evaluate evidence for epigenetic control of alternative splicing. We used SUPPA2 to calculate percentage spliced-in (PSI) values for skipped exons and retained introns and integrated this information with histone ChIP-seq and DNA methylation data. In addition to genome-wide association analysis with partial correlation and machine learning, we perform locus-specific PSI modeling on individual alternative splicing events. The latter represents a new contribution enabled by the unprecedented number of uniformly processed datasets in EpiATLAS. Our results confirm previously reported global associations of DNAm and H3K36me3 for exon inclusion and emphasize the importance of intrinsic features for genome-wide associations. On an event-specific level, trying to identify co-transcriptionally spliced events with epigenetic influence, we show that overall gene expression biases locus-specific analyses. Further, we show that epigenetic signal can predict PSI value in cis and trans, indicating sample-specific epigenetic fingerprints that distort across-sample analyses. Specifically, generalized linear models select epigenetic features predictive for PSI values not only in their genomic vicinity but also on different chromosomes. Our study demonstrates that epigenetic marks are associated with the alternative inclusion of exons and introns. Without a mechanistic explanation for these associations, our work emphasizes the need for more detailed research into the relationship between epigenetic changes and transcriptome diversity.