Associations between antagonistic SNPs for neuropsychiatric disorders and human brain structure

Abstract

A previously published genome-wide association study (GWAS) meta-analysis across eight neuropsychiatric disorders identified antagonistic single-nucleotide polymorphisms (SNPs) at eleven genomic loci where the same allele was protective against one neuropsychiatric disorder and increased the risk for another. Until now, these antagonistic SNPs have not been further investigated regarding their link to brain structural phenotypes. Here, we explored their associations with cortical surface area and cortical thickness (in 34 brain regions and one global measure each) as well as the volumes of eight subcortical structures using summary statistics of large-scale GWAS of brain structural phenotypes. We assessed if significantly associated brain structural phenotypes were previously reported to be associated with major neuropsychiatric disorders in large-scale case-control imaging studies by the ENIGMA consortium. We further characterized the effects of the antagonistic SNPs on gene expression in brain tissue and their association with additional cognitive and behavioral phenotypes, and performed an exploratory voxel-based whole-brain analysis in the FOR2107 study (n = 754 patients with major depressive disorder and n = 847 controls). We found that eight antagonistic SNPs were significantly associated with brain structural phenotypes in regions such as anterior parts of the cingulate cortex, the insula, and the superior temporal gyrus. Case-control differences in implicated brain structural phenotypes have previously been reported for bipolar disorder, major depressive disorder, and schizophrenia. In addition, antagonistic SNPs were associated with gene expression changes in brain tissue and linked to several cognitive-behavioral traits. In our exploratory whole-brain analysis, we observed significant associations of gray matter volume in the left superior temporal pole and left superior parietal region with the variants rs301805 and rs1933802, respectively. Our results suggest that multiple antagonistic SNPs for neuropsychiatric disorders are linked to brain structural phenotypes. However, to further elucidate these findings, future case-control genomic imaging studies are required.

Fig. 1. Schematic overview of the four-folded approach.

Our analyses systematically characterize the eleven antagonistic SNPs with regard to their link to brain structure and brain-related traits: In (1) we perform a SNP to brain image-derived phenotype (IDP) analysis. In (2) we investigate if implicated IDPs are altered in patients with neuropsychiatric disorders compared to controls using the ENIGMA datasets. In (3) we assess if antagonistic SNPs are part of eQTLs for brain tissue, and if antagonistic SNPs are associated with additional cognitive and behavioral traits. In (4) we investigate if there are further associations of the antagonistic SNPs with brain structure at the voxel-wise level that might have been missed in the SNP-to-IDP analysis. SNP single-nucleotide polymorphism.

Fig. 2. Significant associations between the antagonistic SNPs and IDPs.

A presents the SNP-to-IDP associations (p_FDR < 0.05, red line) color-coded by SNP. Note that we replaced rs1933802 using the proxy rs314280 as described in the Materials and Methods. Brainplots present significant SNP-to-IDP associations for CT (B) and SA (C). CT cortical thickness, FDR false discovery rate, IDP image-derived phenotype, SA surface area, SNP single-nucleotide polymorphism, sup superior, temp temporal, Vol Volume.

Fig. 3. Association of the SNP rs9329221 with SCZ and SA measures of the superior temporal region.

The T allele of the SNP rs9329221 was linked to SCZ risk [24] and was associated with a decrease of SA within the Desikan-Killiany region superior temporal [35]. This region showed prominent decrease of SA in patients with SCZ compared to controls (Table S5a in [64]). We note that Fig. 3 displays association results of the different individual investigations and does not represent a separate mediation analysis. FDR false discovery rate, HC healthy control, SA surface area, SCZ schizophrenia, SNP single-nucleotide polymorphism.

Fig. 4. Associations of rs301805 and rs1933802 with gray matter volume in the FOR2107 study.

Associations of the G allele dosage of rs301805 (A) and the G allele dosage of rs1933802 (B) with GMV with p_uncorrected < 0.001 and k > 10. The peak voxels and anatomical labels of the GMV clusters are provided in Table S6. Furthermore, associations of rs301805 and rs1933802 with GMV that remained significant at p_FWE < 0.05 are shown in Figure S2. FWE family-wise error, GMV gray matter volumes.