Abstract
Human tumors are characterized by extensive intratumoral transcriptional variability within the cancer cell and stromal compartments. This variation drives phenotypic heterogeneity, producing cell states with differential pro- and anti-tumorigenic properties. While bulk RNA sequencing cannot achieve cell-type-specific transcriptional granularity, single-cell sequencing has permitted an unprecedented view of these cell states. Despite this knowledge, we lack an understanding of the mechanistic drivers of this transcriptional and phenotypic heterogeneity. 3' untranslated region alternative polyadenylation (3' UTR-APA) drives gene expression alterations through regulation of 3' UTR length. These 3' UTR alterations modulate mRNA stability, protein expression and protein localization, resulting in cellular phenotypes including differentiation, cell proliferation, and migration. Therefore, we sought to determine whether 3' UTR-APA events could characterize phenotypic heterogeneity of tumor cell states. Here, we analyze the largest single-cell human pancreatic ductal adenocarcinoma (PDAC) dataset and resolve 3' UTR-APA patterns across PDAC cell states. We find that increased proximal 3' UTR-APA is associated with PDAC progression and characterizes a metastatic ductal epithelial subpopulation and an inflammatory fibroblast population. Furthermore, we find significant 3' UTR shortening events in cell-state-specific marker genes associated with increased expression. Therefore, we propose that 3' UTR-APA drives phenotypic heterogeneity in cancer.