Autism’s Genetic Edge: Evolution’s Hidden Bargain

A new study argues that the prevalence of autism in human populations may reflect an evolutionary trade-off embedded within our genetic heritage. Researchers say that rapid changes in certain brain neuron types and shifts in autism-linked genes likely promoted advanced cognition at the cost of increased neurodiversity.

The analysis, published in Molecular Biology and Evolution, focuses on a class of outer-layer neurons called L2/3 intratelencephalic cells, which appear to have experienced accelerated evolution along the human lineage. Those changes coincided with dramatic modifications in genes already associated with autism, a pattern consistent with positive selection specifically in humans. The authors propose that this evolutionary pressure helped extend the developmental window of postnatal brain maturation—a trait linked to cognitive flexibility and language capacity—but introduced vulnerability to autism as a byproduct.

Lead author Alexander L. Starr and co-researcher Hunter B. Fraser argue that the same genetic shifts that sculpted uniquely human neural circuitry might also have heightened sensitivity to developmental perturbations. Under this view, autism does not represent a purely deleterious mutation cluster, but rather one face of the broader variability in human neural function.

The team analysed single-nucleus RNA sequencing datasets across three brain regions in multiple mammalian species, comparing humans with other apes. They found that L2/3 IT neurons stood out for their unusually rapid divergence in gene expression and regulation. At the same time, dozens of autism-associated loci showed transcriptional alterations that aligned with the evolutionary changes in those neurons. The coincident patterns, the authors contend, exceed what one would expect under neutral drift.

Because humans display a prolonged period of brain growth and synaptic pruning after birth—longer than in chimpanzees or other primates—the authors hypothesise that the genetic architecture supporting that slow maturation also created a landscape in which disruptions could more easily tip a child toward neurodevelopmental divergence. In effect, mutations or perturbations that would otherwise be benign might push developing brains into autistic phenotypes.

The study reframes earlier theories linking cognition and neuropsychiatric risk. Some past models posited that autism and schizophrenia might be counterbalancing by-products of complex brain evolution; this work gives molecular specificity to that idea, placing one neuron type as a focal node. The authors caution, however, that the precise selective advantage remains speculative—whether the changes favoured language, social cognition, memory, adaptability, or alternative traits is not resolved.

Independent experts have expressed measured enthusiasm. One neuroscientist noted that the study’s integration of gene expression evolution with disease-linked genes is ambitious, but that further replication—especially in developmental time courses—is needed. Others highlight that autism is highly heterogeneous, so any evolutionary model must coexist with multiple etiologies including environment, somatic mutation, and epigenetics.

This evolutionary model gains traction in parallel with other recent advances in autism research. A major autism study published earlier this year using data from over 5,000 children identified four biologically distinct subtypes of autism, unveiling more granular genotype–phenotype relationships. The work underscores that autism is unlikely reducible to a single mechanism or pathway. Meanwhile, novel methodological advances such as graph transformer models applied to brain connectivity data have pushed classification accuracy upward, underscoring the growing interplay between computational tools and neurobiology.