Sex-stratified genome-wide association meta-analysis of Major Depressive Disorder

Abstract

There are striking sex differences in the prevalence and symptomology of Major Depressive Disorder (MDD). We conducted the largest sex-stratified genome wide association and genotype-by-sex interaction meta-analyses of MDD to date (Females: 130,471 cases, 159,521 controls. Males: 64,805 cases, 132,185 controls). We found 16 and eight independent genome-wide significant SNPs in females and males, respectively, including one novel variant on the X chromosome. MDD in females and males shows substantial genetic overlap with a large proportion of MDD variants displaying similar effect sizes across sexes. However, we also provide evidence for a higher burden of genetic risk in females which could be due to female-specific variants. Additionally, sex-specific pleiotropic effects may contribute to the higher prevalence of metabolic symptoms in females with MDD. These findings underscore the importance of considering sex-specific genetic architectures in the study of health conditions, including MDD, paving the way for more targeted treatment strategies.

Figure 1.

Miami plot of sex-stratified GWAS meta-analysis of MDD, with female and male meta-analyses shown on the top and bottom, respectively. The –log10 p values for each SNP are shown with positions according to human genome build 37 (GRCh37 assembly). Chromosome 23 is the X chromosome. The darker grey and lighter grey dotted horizontal lines indicate genome-wide significance ($\text{P}=5\times {10}^{-8}$) and nominal significance ($\text{P}=1\times {10}^{-6}$), respectively. SNPs in dark purple indicate the lead independent genome-wide significant SNPs, and any SNPs in linkage disequilibrium with them.

Figure 2.

Comparisons of genetic architecture and polygenic overlap across the two sexes. A) Autosomal SNP-based heritability (${h^2}_{\text{SNP}}$) on the liability scale using a population prevalence of 0.1 in males and 0.2 in females, B) Polygenicity, and C) Selection parameter using our sex-stratified meta-analysis results in SBayesS. D) Autosomal SNP-based genetic correlation ($r_g$) between males and females using our sex-stratified meta-analysis results, and combination of all Markov chain Monte Carlo (MCMC) samples of $r_g$ estimated in all six male-female within cohort, 30 male-female across cohort, 15 male-male across cohort and 15 female-female across cohort combinations. E) Pearson correlation of the MDD effect sizes of SNPs known to be associated with sex-combined MDD for males vs females using our sex-stratified meta-analysis results, and meta-analysis of Pearson correlations estimated in all six male-female within cohort, 30 male-female across cohort, 15 male-male across cohort and 15 female-female across cohort combinations. F) Venn diagram depicting the number of causal variants explaining 90% of MDD ${h^2}_{\text{SNP}}$ in females only, males only, or both sexes, as identified by MiXeR. G) Venn diagram depicting the number of genomic regions that contain a causal variant for MDD in females only, males only or both sexes, as identified by gwas-pw. Females or female-female comparisons are in yellow, males or male-male comparisons in dark purple and female-male comparisons in green. For A) – C), as a Bayesian framework was used the points represent the mean posterior value, error bars are the 95% highest posterior density interval and percentages are the posterior probability that female value > male value. For D) and E), as frequentist statistics were used the points represent the mean while error bars are the 95% confidence interval and stars represent the $r_g$ / R being significantly different to 1 (after p-value correction for multiple tests).

Figure 3.

Manhattan plot of genome-wide genotype-by-sex interaction meta-analysis for MDD. The –log10 p-values for each SNP are shown with positions according to human genome build 37 (GRCh37 assembly). Chromsome 23 is the X chromosome. The darker grey and lighter grey dotted horizontal lines indicate genome-wide significance ($\text{P}=5\times {10}^{-8}$) and nominal significance ($\text{P}=1\times {10}^{-6}$), respectively. SNPs in dark purple indicate the lead independent nominally significant SNPs, and any SNPs in linkage disequilibrium with them.

Figure 4.

A) Genetic correlations of MDD in females and males with other psychiatric disorders, metabolic and substance use traits. B) Genetic correlations of MDD in females and males with BMI in females and males. C) Venn diagrams depicting the number of causal variants explaining 90% of ${h^2}_{\text{SNP}}$ in MDD only, BMI/metS only, or both traits as identified by MiXeR in females (yellows) and males (purples) D) Venn diagrams depicting the number of genomic regions that contain a causal variant for both MDD and BMI/metabolic syndrome in females only, both sexes, or males only as identified in gwas-pw. Females are in yellow and males in dark purple. * = significantly different genetic correlation between MDD and the other trait across sexes. PTSD = post-traumatic stress disorder, ADHD = attention deficit hyperactivity disorder, BMI = body mass index, Waist/hip ratio = waist to hip ratio adjusted for BMI, metS = metabolic syndrome.