The emergence of genetic variants linked to brain and cognitive traits in human evolution

Abstract

Human evolution involved major anatomical transformations, including a rapid increase in brain volume over the last 2 million years. Examination of fossil records provides insight into these physical changes but offers limited information on the evolution of brain function and cognition. A complementary approach integrates genome dating from the Human Genome Dating Project with genome-wide association studies to trace the emergence of genetic variants linked to human traits over 5 million years. We find that genetic variants underlying cortical morphology (~300,000 years, P = 4 × 10-28), fluid intelligence (~500,000 years, P = 1.4 × 10-4), and psychiatric disorders (~475,000 years, P = 5.9 × 10-33) emerged relatively recently in hominin evolution. Among psychiatric phenotypes, variants associated with depression (~24,000 years, P = 1.6 × 10-4) and alcoholism-related traits (~40,000 years, P = 5.2 × 10-12) are the youngest. Genes with recent evolutionary modifications are involved in intelligence (P = 1.7 × 10-6) and cortical area (P = 3.5 × 10-4) and exhibit elevated expression in language-related areas (P = 7.1 × 10-4), a hallmark of human cognition. Our findings suggest that recently evolved genetic variants shaped the human brain, cognition, and psychiatric traits.

Keywords: Paleogenomics; brain; evolution; genetics; neuropsychiatry.

Fig. 1

Genetic timeline of SNPs associated with human phenotypic variation. A) Density timeline (normalized count; y-axis) of phenotype-associated SNPs from the GWAS atlas (2500 GWAS, 36,506 unique SNPs) and all SNPs from the HGD over the last 2 million years (x-axis). HGD SNPs with MAF equal to or above the lowest MAF in phenotype-associated SNPs are shown (MAF ≥ 0.001; ~ 7.2 million SNPs). B) Density timeline of phenotype-associated SNPs grouped by MAF bins (0–0.01, 0.01–0.1, 0.1–0.2, 0.2–0.3, 0.3–0.4, 0.4–0.5). C) Validation analysis comparing the GWAS atlas phenotype-associated SNPs timeline to an alternative dataset from the EBI Catalog, showing high concordance (rho = 0.98, P = 2.8 × 10⁻⁷²). D) Absolute SNP count per time bin (100 bins of ~ 20,000 years) for phenotype-associated SNPs and randomly selected SNPs. Asterisks denote bins where the number of phenotype-associated SNPs deviated significantly from the null model (MAF-controlled, 100 tests, Bonferroni P < 5 × 10⁻⁴). MAF, minor allele frequency.

Fig. 2

Genetic timeline of human traits. A) Genetic variants associated with human phenotypic variation were extracted from the GWAS atlas, which organizes phenotypes hierarchically into domains, chapters, subchapters, and trait levels. Expected trait ages (dots; n = 361) and domain ages (stars; n = 24) were estimated under a null model (see evolutionary age of human phenotypes in Materials and methods). Z-scores beyond the dotted black line indicate traits with a median age of SNPs significantly younger (negative z-scores) or older (positive z-scores) than expected by chance (controlling for polygenicity and MAF; Bonferroni P < 1.4 × 10⁻⁴). Gray dotted lines indicate nominal significance thresholds. Labels of nominally significant traits are displayed, with bold labels indicating Bonferroni significant effect. Colors denote each domain. B) Timeline of SNPs linked to human phenotypes at the domain level. Histogram showing the density (y-axis) of SNPs emerging over time (shown up to 2 Mya; x-axis). Domains with a significantly younger median age include “psychiatric”, “activities”, and “environment”, while “metabolic” and “neoplasms” are significantly older (controlling for polygenicity and MAF; n = 24 domains, Bonferroni P < 2.1 × 10⁻³). C) Timeline of SNPs linked to human phenotypes at the chapter level. Histogram showing the density (y-axis) of SNPs emerging over time (shown up to 2 Mya; x-axis). Chapters with a significantly younger median age are “mental and behavioral disorders” and “major life areas”, while “digestive, metabolic and endocrine systems” and “malignant neoplasms” are significantly older (controlling for polygenicity and MAF; n = 31 chapters, Bonferroni P < 1.6 × 10⁻³). ACT, activities; BMI, body mass index; BODY, body structures; CAR, cardiovascular; CONN, connective tissue; DER, dermatological; END, endocrine; ENVT, environment; ENVL, environmental; GI, gastrointestinal; HEM, hematological; IMM, immunological; MET, metabolic; MORT, mortality; NEO, neoplasms; NEU, neurological; NUT, nutritional; OPH, ophthalmological; PSY, psychiatric; REP, reproduction; RES, respiratory; SKE, skeletal; SOC, social interactions.

Fig. 3

Enrichment analysis and transcriptomic brain map of evolutionarily recent genes. A) Evolutionary timeline of genes (n = 18,328) with age estimates ranging from 2,965,600 to 3803 years ago. The y-axis represents density distribution, and the x-axis shows gene age bins (only genes younger than 2 million years are displayed). B) Evolutionary timeline of genes ranked by intolerance to LoF mutations. The y-axis shows the median evolutionary age of the most LoF-intolerant genes and LoF-tolerant genes. LoF-intolerant genes are significantly younger (t-test, P = 8.1 × 10⁻⁶). C) Evolutionary timeline of genes associated with 5 major neuropsychiatric disorders. Solid lines indicate phenotypes significantly younger than expected under the null model (Bonferroni P < 5 × 10⁻³) for SCZ, BD, and AD. The insert panel shows null distributions for each condition. D) MAGMA gene-set analysis testing for enrichment of genes related to brain, cognition, and neuropsychiatric among the oldest and youngest genes. The x-axis shows P values, while the y-axis lists tested phenotypes. The dotted line indicates the Bonferroni significance threshold (P < 2.5 × 10⁻³). Young genes show strong enrichment for intelligence and CA, and nominal associations for AD and SCZ. No enrichment is observed for brain-related traits or neuropsychiatric disorders in old genes. E) Functional annotation of young genes identified 21 out of 42 significant biological functions related to the brain (q < 0.05, FDR). Bold text indicates significant enrichment for brain-related biological processes. F) Normative expression levels of the young genes across brain areas, ranging from relatively low to high expression levels (z-scores). The figure shows higher normalized expression in Broca’s area (pars triangularis: Z = 3.4, P = 7.1 × 10⁻⁴; pars opercularis: Z = 2.6, P = 8.5 × 10⁻³), a critical region for language processing, as well as in the posterior cingulate (z = 2.7, P = 7.0 × 10⁻³), a key hub of the default mode network. G) Young genes show significant overexpression (P < 0.05) in regions linked with 5 cognitive domains (n = 111 total terms) from the Neurosynth database (meta-analysis of over 10,000 functional brain studies). Cognitive domains related to language are highlighted in bold. H) Young genes exhibit significantly higher expression in prenatal stages (8 to 37 weeks post-conception) compared to postnatal stages (4 months to 40 years old).