Hundreds of New Brain Cell "Microexons" Found, Pointing to Autism-Linked Variants
Researchers at the University of Toronto expanded the map of tiny exons that are inserted only in neurons, used machine learning to read the rules that regulate them, and found variants in the genomes of autistic people that are predicted to disrupt splicing.
When a gene is turned into a protein blueprint, the cell picks only the pieces it needs (exons) and joins them together. Among these, very short pieces of 3 to 27 bases are called microexons. They are used especially often in brain neurons, and it was known that their processing is frequently disrupted in the brains of people with autism spectrum disorder (ASD). A team led by Professor Benjamin Blencowe at the Donnelly Centre at the University of Toronto in Canada published in Nature a study that greatly expands the list of these pieces and the rules by which they operate.
Decoding the "grammar of insertion" with machine learning
The researchers analysed the transcriptomes of many neuron types on a large scale and newly found hundreds of previously unknown microexons. Some were used only in particular neuron subtypes, and some acted as switches that regulate gene expression levels by inducing nonsense-mediated mRNA decay. Next, they built a sequence rule with machine learning to predict "microexons inserted in neurons" and identified the key regulatory elements around them.
Applying this rule to the genomes of autistic people revealed variants predicted to change microexon splicing. The effect of rare variants was largest, most were found in idiopathic cases in which the causal gene had not been identified, and they were statistically associated with autism. The variants were concentrated in synapse-related genes, in known autism-associated genes and in genes not yet linked to autism.
Significance and limits
Autism genetics has focused on variants that change protein-coding sequences, and this study shows that variants hidden in regulatory sequences also matter. However, the effects of most variants are computer predictions, and an association does not mean causation. This is basic research based on human cell and genome data and is not yet at the stage of direct application to diagnosis or treatment.
- Hundreds of new neuron-specific microexons found, some acting as gene expression switches
- Machine learning defined the splicing regulation rules (a sequence code)
- Variants predicted to disrupt splicing found in genomes of autistic people, with rare variants having the largest effect
Limitations The effects of most variants are computer predictions from an association analysis. This is basic research based on human genome and transcriptome data and is not at the stage of use in diagnosis or treatment.
Original research
- Parada GE, Bretschneider H, Li JD, … Blencowe BJ · Donnelly Centre, University of Toronto (Canada) and others. A microexon regulatory code reveals autism-linked genetic variation. Nature. DOI: 10.1038/s41586-026-11119-w
Copyright in the original research belongs to its publisher and authors.