Regulation and Biological Functions of Alternative Splicing in Neurons of the Adult Mouse Visual Cortex

King’s Undergraduate Research Fellowship, summer 2022, at the Centre for Developmental Neurobiology, King’s College London, with Prof. Eugene Makeyev and day-to-day bioinformatics mentoring from Dr Fursham Hamid. The project was one I conceptualised and took to the KURF board for funding.

The question

Neurons produce many RNA isoforms from a single gene by alternative splicing. Some splicing events introduce a premature stop codon, so the transcript is destroyed by nonsense-mediated decay, which makes splicing a mechanism for tuning expression rather than only for diversifying protein products.

The idea the project rests on is a consequence of where NMD happens. Because it operates after export to the cytoplasm, transcripts targeted by NMD should be relatively enriched in the nucleus. So comparing single-nucleus against single-cell RNA-seq of the same neuron types should flag candidate NMD-regulated exons, without needing to block NMD experimentally.

Data and processing

Matched single-cell and single-nucleus full-length RNA-seq from adult mouse primary visual cortex (Bakken et al. 2018, Allen Institute), as four count matrices: cell and nucleus, exon and intron. Processing in Seurat v3 covered quality control, normalisation, variable feature selection, PCA, clustering, UMAP and marker identification. Splicing was quantified with Whippet, using an index and wrapper provided by Dr Hamid.

Findings

Exon counts from cells and nuclei integrated cleanly, and longer exons were enriched in nuclei, consistent with incomplete splicing of nascent transcripts still in the nucleus.

Comparing neuron subclasses by gene ontology: L4 neurons were enriched for glutamate receptor activity and myosin binding, L5 for scaffold protein and actin binding. Among GABAergic interneurons, Pvalb cells were more enriched for GABAergic functions and Sst for glutamatergic, with both enriched for cytoskeletal anchoring.

The splicing comparison is the part the project was built for. Of all splicing event types, cassette exons showed the strongest positive relationship between splicing change and nuclear enrichment. That is the expected direction, since cassette exons generate the most NMD-targeted transcripts. Sf3b1 is the worked example: a cassette exon between exons 4 and 5 that introduces a premature stop codon, in a splicing factor that regulates its own expression this way.

What this does and does not show

This is a candidate list rather than a result. The splicing analysis used 80 randomly sampled cells and nuclei, too few points for a correlation test, so the relationship is reported as a visible trend with no coefficient attached. Nuclear enrichment also has explanations other than NMD, including nuclear retention and slow export, which this comparison cannot separate.

I also helped run a Centre workshop for Year 12 students using the Allen Brain Atlas.