Sex differences in brain protein expression and disease

Abstract

Most complex human traits differ by sex, but we have limited insight into the underlying mechanisms. Here, we investigated the influence of biological sex on protein expression and its genetic regulation in 1,277 human brain proteomes. We found that 13.2% (1,354) of brain proteins had sex-differentiated abundance and 1.5% (150) of proteins had sex-biased protein quantitative trait loci (sb-pQTLs). Among genes with sex-biased expression, we found 67% concordance between sex-differentiated protein and transcript levels; however, sex effects on the genetic regulation of expression were more evident at the protein level. Considering 24 psychiatric, neurologic and brain morphologic traits, we found that an average of 25% of their putatively causal genes had sex-differentiated protein abundance and 12 putatively causal proteins had sb-pQTLs. Furthermore, integrating sex-specific pQTLs with sex-stratified genome-wide association studies of six psychiatric and neurologic conditions, we uncovered another 23 proteins contributing to these traits in one sex but not the other. Together, these findings begin to provide insights into mechanisms underlying sex differences in brain protein expression and disease.

Fig. 1. Summary of main analyses.

We investigated sex differences in protein expression and its genetic regulation using brain proteomic and genetic data. Next, we compared effects of sex on gene expression and its genetic regulation at both the mRNA and protein levels using genetic and brain transcriptomic and proteomic data. Subsequently, we examined the intersection between psychiatric and neurologic causal genes and genes with sex-biased protein expression or sb-pQTLs. Finally, we integrated sex-stratified GWAS with pQTLs to identify sex-specific causal genes in psychiatric and neurologic disorders. TMT, tandem mass tag.

Fig. 2. Sex-biased pQTLs.

a, Operational definition of sb-pQTLs: an SNP needs to meet both of the following two criteria to be declared a sb-pQTL. First, it is a pQTL in males or females or both at FDR < 0.05. Second, it has a significant genotype-by-sex interaction in protein expression at FDR < 0.05. We identified 166 index sb-pQTLs corresponding to 150 unique proteins in human brain. b, Genomic site-type enrichment of 166 index sb-pQTLs. Data are presented as OR ± 95% confidence interval. Fisher’s exact test was used to calculate the ORs. Error bars reflect a 95% confidence interval. ncRNA, non-coding RNA; UTR, untranslated region.

Fig. 3. Causal genes in psychiatric, neurologic and brain morphologic traits with sex-biased protein abundance.

a, Percentage of causal genes with sex-biased protein abundance among the identified causal genes for each brain trait. b, Causal genes with sex-biased expression at both the mRNA and protein levels in concordant directions (n = 28). Full results are in Supplementary Table 17. PTSD, post-traumatic stress disorder.

Fig. 4. Sex-specific causal genes and proteins.

a, Causal genes with sb-pQTLs (n = 12). Among these, three also had sex-biased protein expression: CNTN2, ERLEC1 and GIGYF2 (in bold). Detailed results are in Supplementary Table 18. b, Causal proteins in females only (n = 11). Among these 11 proteins, three also had sex-biased protein expression: DOC2A, ITIH3 and DLST (in bold). c, Causal proteins in males only (n = 12). Of these 12, two also had sex-biased protein expression: CADM2 and ZZEF1 (in bold). Cross-disorder refers to major depression, bipolar disorder and schizophrenia. Asterisks indicate significant P values in the sex-specific PWAS. Detailed results are in Supplementary Table 20.

Extended Data Fig. 1. Proteins with sex-biased expression per chromosome.

Ratio of percent of proteins with sex-differentiated abundance over percent of proteins profiled in the brain proteomes for each chromosome. This enables comparison of the number of genes with sex-differentiated protein abundance across the chromosomes, taking into consideration the number of genes encoded on each chromosome and the number of these proteins profiled in the brain proteomes.

Extended Data Fig. 2. Quantile–quantile (QQ) plots.

QQ plots for the P values of the genotype-by-sex interaction term in the regression model protein ~ sex + SNP + sex × SNP + 56 SVs + 10 PCs for all autosomes (a) and the X chromosome (b). The genomic inflation factor lambda was 1.011 for all chromosomes and 1.032 for the X chromosome.