Multi-omics causal inference of nuclear-encoded mitochondrial genes in autism spectrum disorder

Mitochondrial dysfunction is increasingly implicated in autism spectrum disorder (ASD), yet its causal genetic basis remains unclear. Mitochondria are maternally inherited organelles essential for neurodevelopment and cellular energy homeostasis, while most mitochondrial proteins are nuclear-encoded and follow Mendelian inheritance. Clarifying how genetically regulated mitochondrial gene activity relates to ASD risk may provide new mechanistic insight.

We applied a multi-omics Mendelian randomization (MR) framework integrating methylation (mQTL), expression (eQTL; blood and 12 GTEx brain regions), and protein (pQTL) datasets. We used summary-data-based MR (SMR) with HEIDI to exclude LD-driven signals and Bayesian colocalization (PPH4 > 0.70) to require a shared causal variant. Where independent cis instruments were available, two-sample MR estimated effects and assessed robustness. ASD outcomes came from IEU-802, IEU-806, and FinnGen GWAS.

Convergent evidence highlighted three mitochondria-related genes. CRAT and PRDX6 showed cross-layer support in specific datasets (mQTL/eQTL/pQTL) with overall protective associations. TMEM177 was supported across mQTL and eQTL and exhibited tissue-specific divergence-risk-increasing associations in cerebellar/cortical regions but protective associations in peripheral blood. TMEM177's biology is consistent with a role in complex IV (COX2) assembly, CRAT regulates acetyl-CoA buffering and metabolic flexibility, and PRDX6 contributes to redox homeostasis and membrane repair. Locus-specific CpG variation was directionally aligned with gene expression and ASD risk.

Our findings support a structure-metabolism-redox axis-TMEM177, CRAT, and PRDX6-linking mitochondrial regulation to ASD susceptibility.