A new wave of research is putting autism evolution on the scientific map. In 2025, a Molecular Biology and Evolution study comparing primate brains reported that human layer 2/3 intratelencephalic (IT) neurons evolved unusually fast, coinciding with shifts in the expression of genes linked to autism—evidence consistent with an evolutionary trade-off that favored cognition while elevating autism susceptibility [1]. The work aims to contextualize autism’s high prevalence with a cellular target—abundant L2/3 IT neurons that underpin higher-order cortical functions—rather than a single “autism gene” [2].

Key Takeaways

– shows 3.2% of U.S. children are on the autism spectrum, underscoring high prevalence that rapid cortical neuron evolution may help contextualize. – reveals human layer 2/3 IT neurons evolved unusually fast, with autism-risk gene expression shifts reported in a 2025 Molecular Biology and Evolution paper. – demonstrates 6.5-fold higher odds of de novo HAR copy-number variants in 2,100 autism cases versus siblings, linking human-accelerated regions to risk. – indicates two human-specific genes that expand the cerebrum affect progenitor proliferation and fate, offering mechanistic clues for cognition–neurodevelopment trade-offs. – suggests selection for language and cognition favored postnatal developmental slowing in humans, a September 2025 inference consistent with a trade-off in autism susceptibility.

How the study connects autism evolution to human cortical neurons

The focal point of the 2025 analysis is a common cell type: layer 2/3 IT neurons, abundant excitatory neurons that form many of the cortex’s long-range, integrative circuits [2]. By comparing primate brains, the authors found these neurons underwent accelerated evolution in humans and that many autism-associated genes changed their expression in this cell population, highlighting a plausible cellular substrate for autism’s frequency in modern populations [2].

This same team emphasized that the pattern is consistent with selection for advanced cognition—especially language—despite accompanying risks. Lead author Alexander L. Starr characterized the result as a trade-off: evolutionary pressure favored complex cognitive traits even if it raised the likelihood of neurodevelopmental divergence [1]. The press materials note that a subset of high-risk autism genes showed down-regulation in human L2/3 IT neurons, suggesting a coordinated shift as these cells evolved [3].

Importantly, the research is about correlation across evolutionary time rather than individual diagnosis. The findings do not reduce autism to a single neuron class; rather, they identify where in the cortex evolutionary changes and autism-linked gene expression shifts most strongly intersect [1]. That focus provides a testable hypothesis for future single-cell and circuit-level experiments across species and developmental stages [2].

What 3.2% prevalence means in population terms

Autism’s prevalence provides a reality check on the scale of the question. The authors point to a 3.2% rate among U.S. children, underscoring that autism is both relatively common and heterogeneous—a pattern difficult to explain solely by rare, highly penetrant mutations [1]. The proposed trade-off framework situates this number within human brain evolution: if selection favored traits that expanded and refined cortical integration, the population-level outcome could reasonably include a higher baseline of neurodevelopmental diversity [1].

This interpretation does not deny environmental or polygenic contributions. Instead, it adds a macro-level context—how selection on brain circuitry might shape distributions of cognitive and social traits across a population [1]. The significance of the 3.2% figure, in this telling, is that it is neither an anomaly nor a paradox when viewed through the lens of cortical neuron evolution and gene regulation [1].

Genetic mechanisms linking autism evolution and brain growth

Two lines of genetic evidence bolster the plausibility of the autism evolution hypothesis. First, independent work identified two human-specific genes that jointly expand the cerebrum, with one gene boosting neural progenitor proliferation and the other altering progenitor fate—mechanisms that directly scale cortical size and complexity [4]. These kinds of human-specific genetic innovations likely set the stage on which later neuron-class-specific expression shifts, such as those seen in L2/3 IT neurons, could matter for cognition and risk [4].

Second, regulatory evidence from human-accelerated regions (HARs)—short DNA elements that changed rapidly in humans—links evolution to autism risk at the level of gene control. In an analysis of 2,100 children from the Simons Simplex Collection, de novo copy-number variants overlapping HARs were 6.5 times more likely in autism cases than in unaffected siblings, pointing to evolutionary regulatory elements with functional impact on brain development [5]. Because HARs are enriched near genes active in neurodevelopment, their disruption offers a route by which evolutionary changes in control regions could shape autism liability [5].

The 2025 study’s observation that many high-risk autism genes are down-regulated in human L2/3 IT neurons adds a third thread: a coordinated expression-level signature in a neuron class central to human cognition [3]. Together, these genetic and regulatory signals—human-specific brain growth genes, HAR-linked regulatory risk, and neuron-class-specific expression shifts—provide converging, testable mechanisms linking brain evolution and autism susceptibility [3].

Why an evolutionary trade-off can persist

A core question in evolutionary medicine is why traits with potential costs persist. The authors propose that selection favored developmental programs that lengthened key postnatal windows and enhanced capacities such as language—features that integrate information across distributed cortical areas heavily populated by L2/3 IT neurons [2]. If these same programs modulate expression of autism-associated genes, they could increase trait variance, including autism, as a byproduct of the cognitive gains [2].

Starr summarized the logic plainly: selection can favor a net benefit even amid increased risk elsewhere in the phenotype space [1]. SciTechDaily’s coverage frames this as a rise in neurodiversity accompanying improved cognition—a population-level spread of developmental outcomes that reflects both benefits and trade-offs [3]. In evolutionary terms, the fitness advantages of enhanced communication, abstraction, and social coordination could outweigh the increased incidence of neurodevelopmental differences, allowing the trait architecture to persist [1].

The trade-off model is not deterministic. It does not predict that “more intelligence equals more autism” at an individual level; instead, it explains how selection on cortical systems can shift population distributions while leaving individual outcomes to the complex interplay of many genes and environments [3]. That perspective helps align apparent contradictions between high prevalence and the assumption that selection should purge costly variants [1].

Caveats, limits, and what to watch next

The new findings are cellular and regulatory correlations grounded in comparative biology; they do not establish causal pathways for any one person’s diagnosis [1]. Although L2/3 IT neurons show a striking human acceleration signature and autism-linked expression shifts, many other cell types and pathways also contribute to autism, and the field must reconcile these layers across development and brain regions [2]. Moreover, “down-regulation” of risk genes in a specific neuron class does not directly translate into clinical outcomes without functional validation [3].

Sample scope and generalizability are also considerations. Cross-primate comparisons are powerful but necessarily limited by species availability, tissue quality, and developmental stage alignment; adding more species and developmental time points will be essential [2]. Regulatory evidence from HARs and human-specific brain-growth genes supports the broader framework, yet these data sets must be integrated into mechanistic studies that track how specific regulatory changes alter neuron development and circuit function [4]. Readers should treat “trade-off” as a population-level evolutionary hypothesis that complements, not replaces, molecular and environmental models of autism [1].

Finally, communication matters. Framing autism evolution as part of human neurodiversity and cognitive emergence helps avoid simplistic value judgments while recognizing the lived realities of autistic people and families—a balance the authors emphasize by focusing on biology rather than deficit-based narratives [2].

How the autism evolution hypothesis can be tested next

Several data-driven paths can probe—and potentially falsify—the hypothesis. First, single-cell transcriptomic and epigenomic atlases across additional primate species and developmental stages can test whether human L2/3 IT neurons uniquely concentrate autism-risk expression shifts and when those shifts emerge [2]. Second, functional studies that perturb the two human-specific cerebrum-expansion genes in organoids or animal models can quantify downstream effects on progenitor dynamics, cortical layering, and circuit integration relevant to autism [4].

Third, systematic mapping of HAR activity in developing cortex—combined with CRISPR-based disruption or duplication—can test whether the 6.5-fold case enrichment of HAR copy-number variants translates into predictable changes in gene networks implicated in social communication and sensory integration [5]. Fourth, integrating the 2025 neuron-specific expression findings with longitudinal postnatal data can examine the proposed slowing of development and its relationship to language acquisition and circuit maturation [2]. Finally, cross-referencing these datasets with the molecular signatures identified in the MBE paper can reveal whether down-regulated high-risk genes in human L2/3 IT neurons occupy regulatory neighborhoods shaped by HARs or human-specific growth genes, providing a mechanistic bridge across levels of analysis [3].

Taken together, these next steps would move the field beyond correlation toward causal circuit maps that link evolutionary changes in neuron classes and regulatory elements to measurable developmental trajectories and behavioral phenotypes [1].

Sources:

[1] Phys.org (Oxford University Press) – How evolution explains autism rates in humans: https://phys.org/news/2025-09-evolution-autism-humans.html

[2] EurekAlert! / Oxford University Press – How evolution explains autism rates in humans: www.eurekalert.org/news-releases/1096746″ target=”_blank” rel=”nofollow noopener noreferrer”>https://www.eurekalert.org/news-releases/1096746 [3] SciTechDaily – Autism’s High Prevalence Could Be an Evolutionary Trade-Off: https://scitechdaily.com/autisms-high-prevalence-could-be-an-evolutionary-trade-off/

[4] ScienceDaily / Deutsches Primatenzentrum – How did the large brain evolve?: www.sciencedaily.com/releases/2025/03/250326154423.htm” target=”_blank” rel=”nofollow noopener noreferrer”>https://www.sciencedaily.com/releases/2025/03/250326154423.htm [5] Harvard Medical School – Autism and Evolution: https://hms.harvard.edu/news/autism-evolution

Image generated by DALL-E 3