A pan-cancer interrogation of intronic polyadenylation and its association with cancer characteristics

Abstract

mRNA cleavage and polyadenylation is an integral 2-step process in the generation of protein-encoding mRNA or noncoding transcripts. More than 60% of human genes have multiple polyadenylation sites either in the 3 untranslated region (3UTR-APA) or in the intronic/exonic region, resulting in expression of isoforms with alternative polyadenylation (APA) under different physiologic conditions. The 3UTR-APAs have been extensively studied, but the biology of intronic polyadenylations (IPA) remain largely unexplored. Here we characterized the IPA profiles of 9,679 patient samples across 32 cancer types from the Cancer Genome Atlas (TCGA) cohort. Overall, we identified 22,260 detectable IPA sites; 9,014 (40.5%) occurred in all 32 cancer types and 11,676 (52.5%) occurred in 2 to 31 cancer types. By comparing tumors and their paired normal tissues, we identified 180 to 4,645 dysregulated IPAs in 132 to 2,249 genes in each of 690 patient tumors from 22 cancer types that showed consistent patterns within individual cancer types. Furthermore, across all cancer types, IPA isoforms and their gene regulation showed consistent pan-cancer patterns, and cancer types with similar histologic features were clustered at higher levels of hierarchy. We selected 2,741 genes that were consistently regulated by IPAs across cancer types, including 1,834 pan-cancer tumor-enriched and 907 tumor-depleted IPA genes. Pan-cancer tumor-enriched IPA genes were amply represented in the functional pathways such as cilium assembly and DNA damage repair. Expression of IPA isoforms in DNA damage repair genes was associated with tumor mutation burdens. Expression of IPA isoforms of tumor-enriched IPA genes was also associated with patient characteristics (e.g., sex, race, cancer stages, and subtypes) in cancer-specific and feature-specific manners. Importantly, IPA isoform expression for some genes could be a more accurate prognostic marker than gene expression (summary of all possible isoforms). In summary, our study reveals the roles and the clinical relevance of tumor-associated IPAs in cancer.