Pathogenic impact of isoform switches in 1209 cancer samples covering 27 cancer types using an isoform-specific interaction network

Abstract

Under normal conditions, cells of almost all tissue types express the same predominant canonical transcript isoform at each gene locus. In cancer, however, splicing regulation is often disturbed, leading to cancer-specific switches in the most dominant transcripts (MDT). But what is the pathogenic impact of these switches and how are they driving oncogenesis? To address these questions, we have analyzed isoform-specific protein-protein interaction disruptions in 1209 cancer samples covering 27 different cancer types from the Pan-Cancer Analysis of Whole Genomes (PCAWG) project of the International Cancer Genomics Consortium (ICGC). Our study revealed large variations in the number of cancer-specific MDT (cMDT) between cancer types. While carcinomas of the head and neck, or brain, had none or only a few cMDT, cancers of the female reproduction organs showed the highest number of cMDT. Interestingly, in contrast to the mutational load, the number of cMDT was tissue-specific, i.e. cancers arising from the same primary tissue had a similar number of cMDT. Some cMDT were found in 100% of all samples in a cancer type, making them candidates for diagnostic biomarkers. cMDT showed a tendency to fall at densely populated network regions where they disrupted protein interactions in the proximity of pathogenic cancer genes. A gene ontology enrichment analysis showed that these disruptions occurred mostly in enzyme signaling, protein translation, and RNA splicing pathways. Interestingly, no significant correlation between the number of cMDT and the number of coding or non-coding mutations could be identified. However, some transcript expressions correlated with mutations in non-coding splice-site and promoter regions of their genes. This work demonstrates for the first time the large extent of cancer-specific alterations in alternative splicing for 27 different cancer types. It highlights distinct and common patterns of cMDT and suggests novel pathogenic transcripts and markers that induce large network disruptions in cancers.