Microexons and autism: a hidden layer of genetic risk

Microexons and autism are more closely connected than previously appreciated. A Nature study has identified a class of rare genetic variants that may interfere with RNA processing in neurons and contribute to genetic susceptibility to autism. What it did not do is explain “85% of autism cases,” as some viral summaries have claimed.

The 85% figure is real, but it means something different. High-confidence autism-associated genetic variants can currently be identified in only about 15% of autistic individuals. The remaining roughly 85% are described in the paper as idiopathic cases, meaning that no established genetic explanation has been found.

The researchers asked whether some of that missing genetic signal might lie in regions that conventional gene-focused analyses tend to overlook.

Tiny pieces with large consequences

A gene is not copied from DNA directly into one finished RNA molecule. Cells first produce a longer precursor and then remove some segments while joining others together. This editing process is called splicing.

Microexons are among the smallest pieces that can be inserted during this process. They are only 3 to 27 nucleotides long. Yet their inclusion can alter protein interactions, change protein properties, or even determine whether an RNA molecule survives long enough to be translated.

Regulated microexons are particularly important in the nervous system. Previous studies had already found abnormal microexon splicing in brain tissue from autistic individuals. What remained unclear was how many neuronal microexons had gone undetected and whether human genetic variation could directly interfere with their regulation.

The team analysed RNA sequencing from 4,790 human tissue samples, including brain samples, together with data from 47,380 individual nuclei from human cortical regions. They identified 1,514 microexons that had not previously been annotated in major databases. Of these, 292 showed different splicing patterns between neural and non-neural cells or tissues.

Some of the newly identified microexons can do more than subtly modify a protein. Their inclusion is predicted to introduce premature stop signals that trigger nonsense-mediated RNA decay, a cellular quality-control pathway that destroys particular RNA molecules. A sequence only a few nucleotides long can therefore influence whether a transcript produces a protein at all.

Searching for disruptive variants

The next challenge was to determine which DNA variants might interfere with this process.

The researchers developed a machine-learning model that learned the sequence features surrounding neuronally regulated microexons. These features included splice signals and binding sites for proteins that help control RNA processing.

They then applied the model to whole-genome sequencing data from autistic individuals and their relatives.

The family-based analysis ultimately included 25,132 people from 7,391 families across three major autism cohorts: SSC, SPARK and MSSNG. Rare, and especially ultrarare, variants produced the strongest predicted effects on microexon regulation.

The researchers defined a set of 72 potentially impactful variants associated with autism status. These variants were detected in 2.5% of autistic individuals in the analysed data and were absent from their unaffected siblings in the corresponding comparison.

That is very different from 85%.

The study does not explain 85% of autism. Instead, it identifies a possible source of genetic risk within the much larger group of cases for which conventional genetic analyses have not produced a clear explanation.

There was another notable result. Among carriers of these 72 variants for whom previous genetic findings were available, 82% had been classified as idiopathic: no earlier candidate genetic finding had been recorded that could account for their autism.

From prediction to biological effect

Most of the variant effects identified in the study are computational predictions, so the researchers experimentally tested selected examples in induced human neurons.

For the variants examined, the DNA changes did alter microexon splicing. One example involved MAP4K4, where a variant affecting the microexon itself changed how that sequence was incorporated into RNA.

These experiments support the idea that the model can identify genuine splicing effects. They do not establish that every one of the 72 variants causes autism on its own. The study demonstrates an association with autism status and shows that selected variants can alter RNA processing, but it does not determine the contribution of each variant to an individual person's traits.

The broader result is therefore not simply a new list of autism-associated mutations. Genetic studies often concentrate on variants that alter protein-coding regions directly. This work points to another layer: very small exons and the regulatory DNA surrounding them, which controls how neuronal RNA is assembled.

A microexon may contain only a handful of genetic letters. Whether those letters are included or skipped can nevertheless reshape a much larger molecular program. The study makes that overlooked layer of RNA regulation a new place to search for part of autism's still-unexplained genetic architecture.