Genetic factors influencing a neurobiological substrate for psychiatric disorders

Abstract

A retrospective meta-analysis of magnetic resonance imaging voxel-based morphometry studies proposed that reduced gray matter volumes in the dorsal anterior cingulate and the left and right anterior insular cortex-areas that constitute hub nodes of the salience network-represent a common substrate for major psychiatric disorders. Here, we investigated the hypothesis that the common substrate serves as an intermediate phenotype to detect genetic risk variants relevant for psychiatric disease. To this end, after a data reduction step, we conducted genome-wide association studies of a combined common substrate measure in four population-based cohorts (n = 2271), followed by meta-analysis and replication in a fifth cohort (n = 865). After correction for covariates, the heritability of the common substrate was estimated at 0.50 (standard error 0.18). The top single-nucleotide polymorphism (SNP) rs17076061 was associated with the common substrate at genome-wide significance and replicated, explaining 1.2% of the common substrate variance. This SNP mapped to a locus on chromosome 5q35.2 harboring genes involved in neuronal development and regeneration. In follow-up analyses, rs17076061 was not robustly associated with psychiatric disease, and no overlap was found between the broader genetic architecture of the common substrate and genetic risk for major depressive disorder, bipolar disorder, or schizophrenia. In conclusion, our study identified that common genetic variation indeed influences the common substrate, but that these variants do not directly translate to increased disease risk. Future studies should investigate gene-by-environment interactions and employ functional imaging to understand how salience network structure translates to psychiatric disorder risk.

Fig. 1. Generation of the component of the common neurobiological substrate (CCS) and genome-wide association study (GWAS) analysis workflow.

A, B Comparison between the CCS and the three individual volumes (A) and the residuals of the three volumes after correction for covariates (B). AIC anterior insula cortex, dACC dorsal anterior cingulate cortex. C Histograms of the three extracted volumes and the CCS. Note that A–C show combined data from all five GWAS cohorts. Correlations were left and right AIC: r = 0.65, left AIC and dACC: r = 0.52, right AIC and dACC: r = 0.46. D GWAS analysis workflow. All measures were extracted using NLO-based Jacobian modulation. All GM volumes were corrected for age, age², and sex as covariates; handedness was used as an additional covariate for the three samples 1000BRAINS, CONNECT100, and BiDirect. PCA: principal component analysis, LD: linkage disequilibrium.

Fig. 2. Presentation of the genome-wide association study (GWAS) results.

A Manhattan plot showing the strength of evidence for an association (p value) in the discovery stage component of the common neurobiological substrate (CCS) GWAS. Each variant is shown as a dot, with alternating shades of blue according to chromosome; the top-associated locus 5q32.2 is labeled with a red diamond. The red line marks the genome-wide significance level. B Matrix of the pairwise linkage disequilibrium (LD) pattern (1000 Genomes population CEU) between the 12 variants that reached genome-wide significance in the discovery GWAS. The two variants rs17076061 and rs72088023 (r² = 0.267) showed the strongest support for an association in their respective LD blocks and were analyzed in the replication stage. All other variants had pairwise LD > 0.5 with either of these two variants, their association strengths are provided for comparison only. P_Disc. discovery stage GWAS p value, p_Repl.(1s) one-sided p value in the replication cohort, Mbp mega base pair. C Regional association plot of the top-associated locus after pooled analysis of the discovery stage GWAS and the replication sample. The dot color indicates LD with the lead variant (rs17076061; pink). Gray dots represent signals with missing LD r² values. cM: centimorgan. D Forest plot of the pooled analysis of the replicated variant rs17076061 in discovery and replication cohorts. D. P.: pooled analysis of discovery stage cohorts, Repl.: replication, Pool.: pooled analysis of the discovery GWAS and the replication cohort SHIP-Trend.

Fig. 3. Analyses of age-by-diagnosis and age-by-SNP effects on the component of the common neurobiological substrate (CCS).

A A significant smaller CCS was observed in MDD (BiDirect, MPIP, FOR2107), BD (FOR2107), SZA (FOR2107), and SCZ (FOR2107) patients, affirming the transdiagnostic finding by Goodkind et al. (Supplementary Table S15). B Age² trajectories of the patient/control groups plotted for each cohort. A non-linear, quadratic age dependency was observed in MDD (pooled MPIP, BiDirect, FOR2107), but no other diagnostic group. Data points represent the CCS after residualization against all covariates except for age and age² (separate fit for patients and controls (Supplementary Table S15). C Age-stratified analyses of the association between diagnosis and the CCS using five age groups in the combined patient/control cohorts (with cohort as a covariate). No significant heterogeneity was observed. Size and color of the effect sizes per bin are proportional to the sample size. D Age-stratified analyses of the association between the top SNP (rs17076061) and the CCS using five age groups in the combined five population-based GWAS cohorts (with cohort modeled as a covariate). No significant heterogeneity between the age groups was observed. Size and color of the effect sizes per bin are proportional to the sample size.