The shared genetic basis of mood instability and psychiatric disorders: A cross-trait genome-wide association analysis

Abstract

Recent genome-wide association studies of mood instability (MOOD) have found significant positive genetic correlation with major depression (DEP) and weak correlations with other psychiatric disorders. We investigated the polygenic overlap between MOOD and psychiatric disorders beyond genetic correlation to better characterize putative shared genetic determinants. GWAS summary statistics for schizophrenia (SCZ, n = 105,318), bipolar disorder (BIP, n = 413,466), DEP (n = 450,619), attention-deficit hyperactivity disorder (ADHD, n = 53,293), and MOOD (n = 363,705) were analyzed using the bivariate causal mixture model and conjunctional false discovery rate methods. MOOD correlated positively with all psychiatric disorders, but with wide variation in strength (r_g = 0.10-0.62). Of 10.4 K genomic variants influencing MOOD, 4 K-9.4 K influenced psychiatric disorders. Furthermore, MOOD was jointly associated with DEP at 163 loci, SCZ at 110, BIP at 60 and ADHD at 25. Fifty-three jointly associated loci were overlapping across two or more disorders, seven of which had discordant effect directions on psychiatric disorders. Genes mapped to loci associated with MOOD and all four disorders were enriched in a single gene-set, "synapse organization." The extensive polygenic overlap indicates shared molecular underpinnings across MOOD and psychiatric disorders. However, distinct patterns of genetic correlation and effect directions may relate to differences in the core clinical features of each disorder.

Keywords: ADHD; bipolar disorder; genetic overlap; major depression; mood instability; schizophrenia.

FIGURE 1

(a) Total number of shared variants between mood instability (MOOD, blue) and schizophrenia (SCZ), bipolar disorder (BIP), major depression (DEP), attention‐deficit hyperactivity disorder (ADHD), and height, as estimated by MiXeR. Venn diagrams representing the proportion of unique and shared variants associated with MOOD and each of SCZ, DEP, BIP, ADHD, and height. Polygenic overlap is represented in gray. The numbers indicate the estimated quantity of variants in thousands per component that explains 90% of SNP heritability for each phenotype. The size of the circle reflects the extent of polygenicity for each trait. Genetic correlation (r _g) is represented in the horizontal bars beneath the Venn diagrams. Right of the central bar (red) indicates positive r _g and left of the central bar (blue) indicates negative r _g. (b) MiXeR density plots illustrating the number of variants (color scale, blue to yellow) with a given MiXeR‐modelled effect size (β), for MOOD (y axis) and each of SCZ, BIP, DEP, ADHD, and height (x axis). Extensive polygenic overlap of concordant and discordant variants is observed for MOOD and SCZ and BIP. The plots of MOOD and ADHD and DEP also illustrate extensive polygenic overlap, but most variants have concordant effects. The plot of MOOD and height indicates that most variants influencing each trait have little to no effect on the other trait.

FIGURE 2

The distribution of transdiagnostic mood instability loci. Venn diagram showing the numbers of diagnosis‐specific and transdiagnostic loci across each MOOD and psychiatric disorder conjFDR analysis. ADHD, attention‐deficit hyperactivity disorder; BIP, bipolar disorder; DEP, major depression; SCZ, schizophrenia

FIGURE 3

Manhattan plot showing ‐log10 transformed conjunctional FDR (conjFDR) values (y axis) for mood instability (MOOD) and (a) schizophrenia (SCZ, blue), (b) bipolar disorder (BIP, orange), (c) major depression (DEP, green), and (d) attention‐deficit hyperactivity disorder (ADHD, red) against chromosomal position (x axis) for each SNP. The dotted line represents conjFDR <0.05 significance threshold. Black circles represent lead SNPs. Lead SNPs from transdiagnostic loci across three or more disorders are annotated with mapped genes. N.B. Not all mapped genes for each locus are presented due to space limitations. Credible genes (bold) were prioritized followed by protein‐coding genes and then nonprotein‐coding genes. *Locus spans 8p23 inversion region with complex linkage disequilibrium. This biases gene‐mapping strategies so only a single mapped gene is presented.