Systematic Analysis of Alternative Polyadenylation from High-throughput RNA Sequencing Data

Alternative polyadenylation (APA) of pre-messenger RNA (pre-mRNA) results in the formation of multiple mature mRNA transcripts from a single gene with distinct 3′ untranslated regions (3′ UTRs). It has been estimated that as many as 80% of mammalian protein-coding genes contain multiple polyadenylation (poly[A]) sites, thus contributing extensively to eukaryotic transcriptome diversity and complexity. In this thesis, I describe a new computational method, ‘Quantification of APA’ (QAPA), for profiling APA using conventional bulk RNA-seq data that accurately infers changes in APA by estimating differential 3′ UTR isoform expression. To demonstrate QAPA’s utility, I analyzed a longitudinal RNA-seq study of differentiation of mouse embryonic stem cells (ESCs) into glutamatergic neurons. By characterizing global changes of APA during this developmental process, I found a large number of genes with lengthening 3′ UTRs. These genes do not significantly overlap with those with steady-state gene expression or alternative splicing (AS) changes, indicating that APA represents a largely distinct regulatory layer. Modelling of poly(A) site usage using machine learning techniques further revealed cis-elements that are predictive of this process. I next used QAPA along with additional RNA-seq analysis tools that I helped develop to characterize the transcriptomes of different neuronal cell subtypes of the mouse hippocampus. This study revealed transcriptomic variation between different subtypes involving distinct programs of co-ordinated regulation between gene expression, AS, and APA. Among these programs, in comparing hippocampal neurons from the proximal and distal axis versus other regions, I identified a program of genes displaying differential expression of a subset of ribosomal proteins that are co-regulated with 3′ UTR lengthening of genes known to localize to the synapse. This observation suggests an important role for the coordination of local translation and mRNA localization in neurons. This also highlights the importance of profiling different post-transcriptional regulatory layers to begin to understand their contributions to cell type diversity. My thesis thus contributes new methods for profiling transcriptomes using RNA-seq data and demonstrates the utility of these methods in defining multilayered regulatory networks that underlie the development and specification of neuronal subtypes.Ph.D