Highlight: The Intertwined Evolution of Human Cognition and Autism

The evolution of cognition played a major role in defining our self-perception as unique among organisms. Creative thought, innovation, tool-building, establishment of intricate social networks, and comprehending complex concepts (Laland and Seed 2021), all contribute to our definition of humanity and our assessment of intelligence in other species The neocortex—a six-layered section of nerve cells exclusive to mammalian brains—is particularly associated with these cognitive traits and other higher brain functions, and it is disproportionately large in humans compared to other primates. The complexity of the neocortex is marked by major heterogeneity at the cellular and molecular levels, with important implications for the evolution and expression of cognitive phenotypes. Additionally, this diversity of cell types, each with its characteristic gene expression profile, can shed light on how natural selection shapes phenotypes.

It is widely recognized that a protein's rate of evolution is related to its level of expression across tissues, since highly expressed proteins tend to evolve more slowly and may face higher selective constraint. But does this general principle of protein evolution apply across levels of biological organization? Specifically, does the abundance of a given cell type within a tissue pose a selective constraint on how its gene expression patterns diverge across speciesve (Fig. 1)? This hypothesis suggests that overall changes to gene expression patterns in abundant cell types result in a larger negative effect on fitness; abundant cell types would thus experience slower evolutionary rates when compared to less abundant cell types. This fundamental, yet unresolved, question motivated a new study by Alexander L. Starr and Hunter B. Fraser, now published in Molecular Biology and Evolution. Divergence in gene expression patterns of homologous cell types across species is a good proxy for their evolutionary divergence; yet, few tissues hold sufficient diversity of cell types to test such hypotheses. Starr, a PhD student in Fraser's lab at Stanford University (USA), explains what made brain cells a good model for their study: “There are thousands of different types of neurons in the brain, whereas most other organs have at most a few dozen cell types, and their relative proportions are highly conserved in many parts of the brain.” According to Starr, the abundance of single-cell RNA-seq data across species of neuronal tissues also influenced this choice.

Fig. 1.

Conceptual model illustrating the hypothesis of Starr and Fraser (2025): within a tissue, the evolvability of the different cell types is constrained by their relative abundances. Since gene expression changes in abundant cell types are more likely to result in a larger negative effect on fitness, these cell types should experience slower evolutionary rates when compared to rare cell types. In their study, the authors test this hypothesis by comparing gene expression profiles of neuronal cells in humans, nonhuman primates, and other mammals. Illustration provided by Starr and Fraser (2025).

Starr and Fraser (2025) began their study by collecting neuronal single-nuclei RNA sequencing data from previous studies focusing on humans and other mammals. After an initial comparison of expression patterns of orthologous genes in humans and marmosets, the authors observed a strong negative correlation between neuronal abundance and gene expression divergence, i.e. the more abundant types of neurons showed higher conservation of expression patterns across species. This relationship held across regions of the neocortex and between other pairs of species beyond human and marmoset, and was driven by highly expressed and cell-type-specific genes (instead of highly expressed genes active across multiple cell types).

With multiple lines of evidence supporting their initial hypothesis—that highly abundant cell types are more constrained in their rates of divergence in gene expression across species—the authors also found a compelling pattern in their data. “The human-chimpanzee divergence didn’t fit the pattern we observed in the other species comparisons as well,” mentioned Starr, drawing attention to how the positive correlation between abundance and evolutionary constraint was particularly weak when comparing human and chimpanzee neocortical neurons from layers 2 and 3 (L2/3 neurons, which integrate signals across cortical areas). In other words, human L2/3 neurons showed much higher divergence from chimpanzee neurons than would be expected given their high abundance. “When we saw that layer 2/3 neurons—which are thought to have played a particularly important role in the evolution of human cognition—were the largest outlier, we were intrigued and decided to investigate the potential functional consequences of this,” adds Starr. Their results suggest this pattern resulted from positive selection for high divergence in L2/3 neurons in humans, rather than relaxed selection in our branch.

These results make sense. It is reasonable that cell types underlying cognition were selectively favored during human evolution. To further explore this, Starr and Fraser tested for phenotype enrichment in the genes that differed between human and chimpanzee in L2/3 neurons. Their findings at this stage were striking: genes associated with autism spectrum disorder (ASD), as well as schizophrenia, had a consistent bias toward lower expression in humans, while this pattern was not consistent in genes associated with other phenotypes. Comparisons of the human/chimpanzee data to gorilla gene expression data further confirmed that this pattern of down-regulation is specific to the human lineage. To explore whether these changes were linked to selection on genetic variation, rather than nongenetic effects, the authors retrieved published expression data from hybrid human-chimpanzee cortical organoids. In the hybrid cells, the human alleles of the ASD genes were consistently downregulated compared to the chimpanzee alleles, demonstrating an effect of cis-regulatory mutations. These genes also show lower between-individual variability of expression in humans, further suggesting a functional constraint on their expression when compared to other species.

The expression of ASD in humans, a highly heritable condition, is broadly linked to reduced transcription of synaptic genes associated with the condition (e.g. Gandal et al. 2022); Starr and Fraser (2025) compellingly argue that the same selective pressures that promoted the evolution of human cognition also increase our susceptibility to neurodevelopmental conditions such as autism, through shared cellular and molecular mechanisms. According to Starr, these findings raise new questions on the evolutionary links between cognition and autism: “did the downregulation of ASD-linked genes push humans closer to an ASD-like phenotype, or did it make gene regulatory or neural networks more sensitive, so that genetic or environmental perturbations have larger effects on cognition in humans than the same perturbation would in other species?”

The rise in autism diagnoses in recent decades has been frequently linked to improvements in diagnostic practices and higher societal awareness of the condition. Many of our readers are aware of family or close friends who exhibit ASD traits, maybe even themselves. Molecular and cellular evidence suggests that this evolutionary framework inextricably ties cognition and neural disorders together in humans. Neurodiversity thus seems an integral feature of how our human brain has evolved.

Want to learn more? Check out these other articles focusing on neurodevelopment and evolution recently published in Molecular Biology and Evolution:

• “Brain gene regulatory networks coordinate nest construction in birds” (Fang et al. 2024)

• “Deciphering the role of rapidly evolving conserved elements in primate brain development and exploring their potential involvement in Alzheimer's Disease” (Hu et al. 2024)

• “Evolutionary history of bilaterian FoxP genes: complex ancestral functions and evolutionary changes spanning 2R-WGD in the vertebrate lineage” (Tsai et al. 2025)