Evolutionarily young brain-regulatory variants disproportionately carry schizophrenia heritability across European and East Asian ancestries

Background. Schizophrenia is a heritable psychiatric disorder (twin h² 60–80%) with reduced fecundity, yet common risk alleles persist at substantial population frequencies, posing a long-standing evolutionary paradox. Prior work has examined evolutionary signatures at gene, region, and genome-wide variant levels; however, whether fine-mapped credible-set variants show heterogeneous evolutionary structure that maps onto heritability within loci has not been comprehensively tested. The PGC3 schizophrenia GWAS [1] reported 287 independent loci with FINEMAP-derived posterior inclusion probabilities for 20,766 candidate variants, enabling variant-level evolutionary inference in schizophrenia.

Aim. To test whether fine-mapped PGC3 schizophrenia credible-set variants partition into distinct evolutionary sub-populations that disproportionately carry common-variant heritability, evaluating cross-ancestry portability and disease specificity.

Methods. We assembled variant-level annotations across all 20,766 PGC3 credible-set variants, integrating: (i) Atlas of Variant Age coalescent allele ages [2]; (ii) ancestral-polarised standardised |iHS| recomputed on 1000 Genomes EUR haplotypes; (iii) GTEx v10 brain–blood cis-eQTL specificity across 13 brain and 3 blood/immune tissues; (iv) single-cell brain cis-eQTLs across 8 cell types [3]; and (v) ancient-DNA-derived directional selection coefficients [4]. The 17q21.31 MAPT inversion (1,742 variants) was excluded a priori. Gaussian mixture modelling on standardised joint features (log₁₀ allele age, brain–blood specificity, |iHS|) yielded three sub-populations (Young/Mid/Old; k=3 selected for parsimony, with 100% cluster-assignment stability under re-prediction). Cluster-level partitioned LD score regression conditional on the 53-category baseline-LD model was applied to PGC3 European schizophrenia, PGC3 East Asian schizophrenia [1], and PGC-ALZ2 Alzheimer’s disease.

Results. The Young sub-population (cluster C0; n=1,712 variants; 0.029% of HapMap3 SNPs) captured 1.74% of common-variant schizophrenia heritability, with a 60.5-fold enrichment exceeding the 53-category Bonferroni threshold (Table 1). The EUR-derived annotation replicated as a pre-specified single test in East Asian schizophrenia (Table 1), was dissociable from Alzheimer’s disease (opposite-direction, inconsistent with a power-limited shared late-onset signal of comparable magnitude), and was robust to long-range-LD masking (61.4-fold). Across the 8 brain cell types [3], cluster C0 carried the highest cis-eQTL coverage uniformly (96.9% in microglia; 0.1–3.1 percentage-points over Old, 4–17 over Mid), indicating broad regulatory engagement.

Conclusions. Stratifying fine-mapped PGC3 credible-set variants by their evolutionary signatures identifies a small, evolutionarily young, brain-regulatory subset that disproportionately carries common-variant schizophrenia heritability, replicating across European and East Asian samples and absent in Alzheimer’s disease. The pattern is consistent with non-antagonistic pleiotropy [5] preserving ancient risk alleles, but does not rule out background-selection or mutation–selection-balance alternatives. Cluster definitions are EUR-derived; cross-ancestry and cross-disorder replication remain important next steps.

Table 1. Cluster-level partitioned LD score regression enrichments across PGC3 EUR SCZ, PGC3 EAS SCZ and PGC-ALZ2.

Trait	Cluster	Enrichment	s.e.	95% CI	p	Z
PGC3 EUR SCZ (h² = 0.327)	C0 (Young)	60.5×	16.6	[28.0, 93.0]	4.3×10⁻⁴	+3.55
	C1 (Mid)	15.5×	4.5	[6.7, 24.3]	1.4×10⁻³	+3.42
	C2 (Old)	14.8x	14.1	[−12.8, 42.4]	0.331	+0.92
PGC3 EAS SCZ (h² = 0.380)	C0 (Young)	69.8x	20.9	[28.8, 110.8]	1.8×10⁻³	+3.14
PGC-ALZ2 (h² = 0.0101)	C0 (Young)	−5.3×	14.7	[−34.1, +23.5]	0.667	−0.55

*All regressions condition on the 53-category baseline-LD model. Coefficient Z is the conditional regression coefficient. 95% CIs are computed as enrichment ± 1.96 × s.e.

References

1. Trubetskoy, V., Pardiñas, A.F., Qi, T., et al., 2022. Mapping genomic loci implicates genes and synaptic biology in schizophrenia. Nature 604 (7906), 502-508.
2. Albers, P.K., McVean, G., 2020. Dating genomic variants and shared ancestry in population-scale sequencing data. PLoS Biol. 18 (1), e3000586.
3. Bryois, J., Calini, D., Macnair, W., et al., 2022. Cell-type-specific cis-eQTLs in eight human brain cell types identify novel risk genes for psychiatric and neurological disorders. Nat. Neurosci. 25 (8), 1104-1112.
4. Akbari, A., Perry, A., Barton, A.R., et al., 2026. Ancient DNA reveals pervasive directional selection across West Eurasia. Nature, https://doi.org/10.1038/s41586-026-10358-1.
5. Gonzále-Peñas, J., de Hoyos, L., Díaz-Caneja, C.M., et al., 2023. Recent natural selection conferred protection against schizophrenia by non-antagonistic pleiotropy. Sci. Rep. 13, 15500.