Human neurodiversity may be rooted in neuronal evolution. A cross-species analysis identifies a human-accelerated cortical neuron subtype whose gene-expression shifts overlap with autism-linked genes, suggesting that recent changes that made our brains unique also increased variability in neurodevelopmental outcomes [1]. The work, posted August 3, 2024, analyzed more than one million neocortical neurons across six species and tied human-specific regulatory changes to natural selection on the human lineage [3].

The candidate neuron type—layer 2/3 intratelencephalic (IT) neurons—showed the fastest transcriptomic evolution in humans, with marked down-regulation of genes implicated in autism spectrum conditions, including 83% of known haploinsufficient ASD genes in this set [3]. The authors also report polygenic positive selection signals and allele-specific expression in human brain organoids, building a mechanistic case for human-unique regulatory shifts with complex trade-offs for cognition and neurodevelopment [1].

Key Takeaways

– Shows human-accelerated L2/3 IT neurons among 1,000,000+ cells across six species, highlighting a distinct trajectory in neuronal evolution tied to human lineage [1]. – Reveals 83% of ASD haploinsufficient genes are downregulated in these neurons, pointing to a targeted regulatory shift in the human cortex [3]. – Demonstrates polygenic positive selection acting on human regulatory alleles, with convergence across three cortical regions (MTG, DLPFC, M1) [1]. – Indicates replication across three snRNA-seq datasets, each profiling 300,000–500,000 nuclei, reinforcing robustness and cross-region consistency [3]. – Suggests autism’s heterogeneity aligns with four genetic subtypes (37%, 34%, 19%, 10%) from 5,000+ participants, supporting a polygenic framework [4].

How neuronal evolution accelerated a cortical subtype

A central claim of the 2024 preprint is that L2/3 IT neurons—projection neurons that interconnect cortical areas—are “human-accelerated,” meaning their gene-expression profiles diverged more rapidly on the human lineage than in other species examined [1]. These neurons reside in superficial cortical layers integral to association cortices, which expanded with human cognition, making them a plausible substrate for recent adaptive changes [1].

Within this neuron class, autism-linked genes showed striking directional shifts. The study reports down-regulation of 83% of haploinsufficient ASD genes—genes where one functional copy is insufficient for typical function—within human L2/3 IT neurons [3]. This is not a single-gene effect; rather, the pattern implicates coordinated regulatory changes across many loci, consistent with a polygenic mode of evolution [1].

The authors argue that human cognitive advantages may have been accompanied by trade-offs—specifically, increased variability and susceptibility in neurodevelopmental pathways that shape connectivity and synaptic function [1]. Importantly, the work is a preprint and has not yet undergone peer review, so conclusions should be viewed as provisional pending replication and formal validation [1].

The data: 1,000,000+ neurons across six species

The analysis integrates three large single-nucleus RNA-sequencing datasets from human cortex—middle temporal gyrus (MTG), dorsolateral prefrontal cortex (DLPFC), and primary motor cortex (M1)—each including roughly 300,000–500,000 nuclei, summing to more than one million neurons across species [3]. Sampling multiple cortical regions helps ensure that observed human-specific shifts are not artifacts of a single area but instead reflect a broader cortical pattern [3].

Cross-species comparisons establish that the acceleration is lineage-specific: human L2/3 IT neurons evolved faster relative to homologous neuron types in nonhuman primates and rodents, aligning with the hypothesis of recent selection on human cortical circuitry [1]. To probe mechanism, the authors used brain organoids to test for allele-specific expression (ASE), observing biases that indicate cis-regulatory differences between human alleles—evidence that genetic variation plausibly drives the transcriptomic shifts [1].

The Fraser Laboratory’s publication record corroborates the project’s scope and authorship, including links to the preprint, methods summaries, and contacts, supporting transparency and reproducibility expectations for large-scale genomic work [5].

What neuronal evolution suggests about autism prevalence

If neuronal evolution concentrated regulatory changes in a neuron subtype that underpins long-range cortical communication, it could help explain why autism is relatively prevalent in humans compared with its rarity in other species [1]. The logic is not that autism was “selected for,” but that selection for traits like advanced association cortex functions might have inadvertently increased the variance in neurodevelopmental pathways where small disruptions yield autism traits [1].

The 83% down-regulation figure among haploinsufficient ASD genes in human L2/3 IT neurons underscores how many autism-linked loci converge on this cell type’s regulatory state [3]. Polygenic positive selection signals along the human lineage reinforce the notion that numerous small effect alleles collectively shifted expression programs, a pattern consistent with adaptive fine-tuning rather than a single hard sweep [1]. Taken together, these findings outline a trade-off model: recent adaptive gains in cognition coexist with a wider tail of neurodevelopmental diversity [1].

Neuronal evolution, developmental timing, and human risk

Any account of human-specific neurodevelopment must grapple with timing. A 2023 Nature Reviews Neuroscience synthesis emphasizes that human neurons mature more slowly, with extended developmental windows and species-specific synaptic gene-expression programs that differ from other primates and rodents [2]. This prolonged schedule can heighten sensitivity periods when gene-regulatory perturbations have outsized effects on circuit assembly and plasticity [2].

When layered onto a human-accelerated neuron subtype, extended developmental timing could amplify the functional impact of small expression shifts, increasing the range of outcomes without requiring large-effect mutations [2]. In other words, the human brain’s protracted development may magnify the phenotypic footprint of polygenic regulatory changes that accumulated during recent neuronal evolution [2].

Genetic heterogeneity aligns with a polygenic model

Autism is not a single condition but a spectrum with multiple genetic routes. A 2025 Nature Genetics study of more than 5,000 SPARK participants decomposed autism heterogeneity into four genetic subtypes, accounting for 37%, 34%, 19%, and 10% of the cohort, each tied to distinct genetic programs [4]. These proportions highlight that no single pathway dominates; instead, diverse polygenic architectures shape different clinical profiles within the spectrum [4].

That heterogeneity dovetails with a selection narrative grounded in many small regulatory shifts rather than single-gene causes [1]. If thousands of alleles with modest effects collectively nudge expression in human-accelerated neurons, we should expect multiple subtypes and trajectories—precisely what large cohorts are now resolving via integrative genetics and transcriptomics [4].

Signals of selection in neuronal evolution

The preprint’s statistical tests support polygenic positive selection, indicating that allele frequency shifts on the human lineage align with the observed expression changes in L2/3 IT neurons [1]. This selection footprint suggests adaptation across numerous loci rather than isolated sweeps, a pattern expected for complex cognitive traits influenced by many genes of small effect [1].

Organoid allele-specific expression offers a key validation step: if cis-regulatory variants bias expression in vitro, that strengthens the causal link between genotype and neuron-type-specific transcriptomic differences seen in vivo [1]. While ASE does not prove downstream physiology, it increases confidence that regulatory DNA changes, not just epigenetic state, underpin the human-accelerated profile [1].

Methods, caveats, and what to watch next

Methodologically, the study leverages single-nucleus RNA-seq across MTG, DLPFC, and M1, integrating 300,000–500,000 nuclei per dataset and exceeding one million neurons across species, which improves power to detect subtle, cell-type-restricted shifts [3]. Cross-region replication reduces the risk that findings are region-specific artifacts, and the six-species design enables lineage-aware inference rather than simple human–mouse contrasts [3].

As a preprint, the work awaits peer review, replication, and functional studies to connect expression shifts to circuit dynamics and behavior [1]. The Fraser Lab’s public documentation of the project and methods increases transparency, but formal validation in independent datasets and longitudinal human cohorts will be crucial next steps [5]. Future priorities include single-cell multi-omics across development, in vivo perturbations of candidate regulatory variants, and cross-ethnic analyses to refine estimates of selection and effect sizes on the human lineage [1].

A key limitation is inferential: selection signals and ASE establish plausibility for genetic regulation of expression but do not quantify how much each variant contributes to phenotype or prevalence [1]. Moreover, prolonged human neuronal maturation means that developmental environment and timing can modulate outcomes, complicating attempts to partition variance into “genetic” and “non-genetic” bins [2]. Integrating the SPARK subtype framework with neuron-type-resolved eQTL and regulatory maps may help bridge this gap by linking subtype-specific genetic programs to specific cell types and developmental windows [4].

Why this matters now

The convergence of cross-species transcriptomics, organoid genetics, and large human cohorts is beginning to turn abstract hypotheses about human brain uniqueness into testable models grounded in numbers [1]. Evidence that a specific neuron subtype underwent accelerated regulatory evolution in humans—while concurrently shifting expression of 83% of haploinsufficient ASD genes—sharpens the search for mechanisms behind both advanced human cognition and increased neurodiversity [3].

This line of inquiry reframes autism not as an evolutionary anomaly but as part of the expected variance produced when complex traits evolve via many small changes in a rapidly adapting system [1]. As datasets grow and methods mature, researchers can move from correlation to causation, mapping which regulatory variants and developmental periods most strongly influence outcomes, and designing more targeted supports aligned with the spectrum’s diverse biological narratives [4].

Sources:

[1] bioRxiv / PMC – A general principle of neuronal evolution reveals a human-accelerated neuron type potentially underlying the high prevalence of autism in humans: https://pmc.ncbi.nlm.nih.gov/articles/PMC11312593/

[2] Nature Reviews Neuroscience – Human neuronal maturation comes of age: cellular mechanisms and species differences: https://www.nature.com/articles/s41583-023-00760-3 [3] PubMed / NCBI – A general principle of neuronal evolution reveals a human-accelerated neuron type potentially underlying the high prevalence of autism in humans (PMID 39131279): https://pubmed.ncbi.nlm.nih.gov/39131279/

[4] ScienceDaily – Decomposition of phenotypic heterogeneity in autism reveals underlying genetic programs: www.sciencedaily.com/releases/2025/07/250724040455.htm” target=”_blank” rel=”nofollow noopener noreferrer”>https://www.sciencedaily.com/releases/2025/07/250724040455.htm [5] Stanford University / Fraser Lab – Publications – Fraser Laboratory: https://web.stanford.edu/~hbfraser/publications/

Image generated by DALL-E 3