Genome-wide maps of polyadenylation reveal dynamic mRNA 3′-end formation in mammalian cell lineages

Post-transcriptional regulation, often mediated by miRNAs and RNA-binding proteins at the 3' untranslated regions (UTRs) of mRNAs, is implicated in important roles in the output of transcriptome. To decipher this layer of gene regulation, it is essential to measure global mRNA expression quantitatively in a 3'-UTR-specific manner. Here we establish an experimental and bioinformatics pipeline that simultaneously determines 3'-end formation by leveraging local nucleotide composition and quantitatively measures mRNA expression by sequencing polyadenylated transcripts. When applied to purified mouse embryonic skin stem cells and their daughter lineages, we identify 18,060 3' UTRs representing 12,739 distinct mRNAs that are abundantly expressed in the skin. We determine that similar to 78% of UTRs are formed by using canonical A[A/U]UAAA polyadenylation signals, whereas similar to 22% of UTRs use alternative signals. By comparing to relative and absolute mRNA abundance determined by qPCR, our RNA-seq approach can precisely measure mRNA fold-change and accurately determine the expression of mRNAs over four orders of magnitude. Surprisingly, only 829 out of 12,739 genes show differential 3'-end usage between embryonic skin stem cells and their immediate daughter cells, whereas the numbers increase to 933 genes when comparing embryonic skin stem cells with the more remotely related hair follicle cells. This suggests an evolving diversity instead of switch-like dynamics in 3'-end formation during development. Finally, core components of the miRNA pathway including Dicer, Dgcr8, Xpo5, and Argonautes show dynamic 3'-UTR formation patterns, indicating a self-regulatory mechanism. Together, our quantitative analysis reveals a dynamic picture of mRNA 3'-end formation in tissue stem cell lineages in vivo.