Computational analyses of non-canonical architectural and structural features associated with alternative splicing

Abstract

Splicing of nuclear introns is catalysed by the spliceosome, one of the most complex macromolecular machines currently known. Even though the canonical splicing signals that drive the precise recognition of splice sites are well-characterised, recent advances in transcriptome profiling technologies and computational method development have enabled widespread identification of non-canonical splicing features. Non-canonical splicing is highly associated with dynamic splicing regulation, and occurs most prevalently in neuronal tissues. In this present work, I have investigated two types of non-canonical features that are related to atypical exon-intron structures and DNA/RNA conformations. First, I studied a group of extremely small exons, known as microexons (≤30 nucleotides), which were shown to be part of an evolutionarily conserved network of neuronal alternative splicing events that play essential roles in neuronal development. Since standard RNA-seq tools cannot efficiently detect microexon splice sites, I developed MicroExonator, a novel pipeline for reproducible de novo discovery and quantification of microexons. As a proof of principle, I analysed microexon alternative inclusion patterns across 289 RNA-seq samples coming from eighteen different tissues across a wide range of mouse embryonic and adult stages. I detected 2,938 microexons, 343 of which are differentially spliced throughout mouse embryonic development, including 35 that are not present in mouse transcript annotation databases. Unsupervised clustering of microexons alone segregates brain tissues by developmental time and further analysis suggest a key function for microexon inclusion in axon growth and synapse formation. Moreover, I developed a module to adapt MicroExonator splicing analysis to single-cell RNA-seq samples that I used to analyse data from the mouse visual cortex. As a result, I found 39 microexons that are differentially included between glutamatergic and gabaergic neurons, fifteen of which are found in genes that encode synaptic proteins. The second type of non-canonical features that I studied are sequences associated with non-B DNA structures and possibly atypical RNA conformations. I analysed the enrichment of different non-B DNA motifs across splice site sequences. The strongest and most consistent enrichments were found for G-quadruplex motifs, which are enriched ~3-fold both upstream and downstream of splice junctions. Further analysis of G4-seq experiments corroborated the enriched motifs detected at splice sites leads to in-vitro G-quadruplex formation. Moreover, enrichment analyses of G-quadruplex motifs and G4-seq experiments across multiple species suggest that the association of G-quadruples to splice sites is a property restricted to mammals and birds. Interestingly, I found stronger enrichment of G-quadruplexes associated with weak splice sites, suggesting that they could function as cis-regulatory elements of alternative splicing events. Finally, to explore if microexons and exons flanked by intronic G-quadruplexes were involved in dynamic splicing changes, I analyse alternative splicing events induced by depolarisation treatments in human and mouse neurons. I found a widespread cassette exon skipping response after neuronal depolarization, which was particularly enriched in microexons and exons flanked by G-quadruplexes motifs. Taken together, these results suggest that non-canonical splicing features are an important regulatory mechanism of alternative splicing. Further characterisation of non-canonical splicing might provide a better understanding of fine-tuned alternative splicing mechanisms, in particular in the context of neuronal development and heterogeneity.