Cortical differences across psychiatric disorders and associated common and rare genetic variants

Abstract

Genetic studies have identified common and rare variants increasing the risk for neurodevelopmental and psychiatric disorders (NPDs). These risk variants have also been shown to influence the structure of the cerebral cortex. However, it is unknown whether cortical differences associated with genetic variants are linked to the risk they confer for NPDs. To answer this question, we analyzed cortical thickness (CT) and surface area (SA) for common and rare variants associated with NPDs, in ~33000 individuals from the general population and clinical cohorts, as well as ENIGMA summary statistics for 8 NPDs. Rare and common genetic variants increasing risk for NPDs were preferentially associated with total SA, while NPDs were preferentially associated with mean CT. Larger effects on mean CT, but not total SA, were observed in NPD medicated subgroups. At the regional level, genetic variants were preferentially associated with effects in sensorimotor areas, while NPDs showed higher effects in association areas. We show that schizophrenia- and bipolar-disorder-associated SNPs show positive and negative effect sizes on SA suggesting that their aggregated effects cancel out in additive polygenic models. Overall, CT and SA differences associated with NPDs do not relate to those observed across individual genetic variants and may be linked with critical non-genetic factors, such as medication and the lived experience of the disorder.

Figure 1.. Graphical abstract.

Study overview. We aimed to compare effect sizes on CT and SA for three primary categories: i) 8 NPDs (attention deficit hyperactivity disorder (ADHD); autism spectrum disorder (ASD)^,; bipolar disorder (BD); clinical high-risk for psychosis (CHR-PS); conduct disorder (CD); major depressive disorder (MDD); obsessive-compulsive disorder (OCD); and schizophrenia (SCZ)), considering medicated and unmedicated subgroups where available; ii) common variants^– associated with NPDs; and iii) 18 different CNV and aneuploidy rare variants associated with NPDs. Towards this, we aggregated multiple datasets as well as published summary statistics from ENIGMA consortium to compare case control effect sizes (Cohen’s d) across 8 psychiatric disorders and associated common and rare variants on global and regional CT and SA. Global and regional effect sizes were compared, and spatial patterns of variations were evaluated using four sets of consensus maps including: mean absolute effect size maps, significance maps, variance maps, and principal component analysis. Our findings across genetic variants increasing the risk for NPDs align with twin and SNP heritability estimates, which are higher for surface area and sensorimotor regions, suggesting that these are general properties of the genetic architecture of the cerebral cortex. Overall, our study suggests that the neuroimaging alterations observed in NPDs are distinct from those observed across genetic variants increasing the risk for NPDs. Brain and cortex maps were generated using the ggseg package in R. Common and rare variant illustrations are from the NIAID NIH BIOART Source (https://bioart.niaid.nih.gov/bioart/170 and https://bioart.niaid.nih.gov/bioart/204)

Figure 2:. Global cortical differences across NPDs and associated genetic variants.

A-D) Comparison between neurodevelopmental and psychiatric disorders (NPDs) associated rare genetic liability (CNVs) and diagnosis on A) mean cortical thickness (CT) effect sizes; B) total surface area (SA) effect sizes; C) the ratio of CT and SA effect sizes; and D) paired CT and SA effect sizes. Case-control differences were adjusted for age, sex, and site. *: FDR significant (q<0.05), across all pairs of comparisons. Absolute effect sizes (Y-axis) are plotted on a log10 scale. E) Effects on CT and SA of common variants associated with NPDs. We tested if NPD genome-wide significant SNPs were enriched in SNPs associated with SA or CT. We ranked independent NPD-associated SNPs based on their p-value association with CT and SA. Median rankings are indicated using dotted red lines. * and arrows represent significant (FDR q<0.05) median ranking compared to permutation-based null distribution. Abbreviations: Adult-Adolescence-Pediatric sample abbreviations: -y=young; -p=pediatric; Abs=absolute; ADHD=attention deficit hyperactivity disorder; ASD=autism spectrum disorder; BD=bipolar disorder; CD: conduct disorder; CHRPS: clinical high risk for psychosis; CHRPSn: CHR who did not develop a psychotic disorder; CHRPSp: CHR who later developed a psychotic disorder; CNV=copy number variant; CT=cortical thickness; Del=deletion; Dup=duplication; GWAS: genome-wide association study; MDD=major depressive disorder; NPD=neurodevelopmental and psychiatric disorders; OCD=obsessive-compulsive disorder; prox.=proximal; SA=surface area; SCZ=schizophrenia; SNP=single nucleotide polymorphism; TS=Turner syndrome; WBS=Williams-Beuren syndrome;

Figure 3:. Global cortical differences across medication sub-groups.

A-C) Comparing the effect sizes on mean cortical thickness (CT) and total surface area (SA) of neurodevelopmental and psychiatric disorders (NPDs) diagnosis sub-groups with and without medications using A-B) violin plots; and C) paired boxplots. Case-control differences were adjusted for age, sex, and site. *: FDR significant (q<0.05), across all pairs of comparisons. Absolute effect sizes (Y-axis) are plotted on a log10 scale. Abbreviations: Adult-Adolescence-Pediatric sample abbreviations: -y=young; -p=pediatric; Abs=absolute; ADHD=attention deficit hyperactivity disorder; ASD=autism spectrum disorder; BD=bipolar disorder; CD: conduct disorder; CHRPS: clinical high risk for psychosis; CHRPSn: CHR who did not develop a psychotic disorder; CHRPSp: CHR who later developed a psychotic disorder; CT=cortical thickness; Del=deletion; Dup=duplication; DZ=di-zygotic; MDD=major depressive disorder; MZ=mono-zygotic; NPD=neurodevelopmental and psychiatric disorders; OCD=obsessive-compulsive disorder; SA=surface area; SCZ=schizophrenia;

Medication abbreviations: -m=medicated; -u=un-medicated; BD-m1=lithium medication; BD-m2=antiepileptics medication; SCZ-m1=1st generation; SCZ-m2=2nd generation; SCZ-m12=1st & 2nd generation;

Figure 4.. Consensus maps of regional cortical differences.

A) Number of FDR significant cortical regions (out of 34) per CNV/NPD for cortical thickness (up, red) and surface area (down, blue). B) Regional profiles of twin heritability, and mean absolute effect sizes across common and rare genetic variants and NPDs for cortical thickness and surface area across 34 Desikan cortical regions. Each point represents: i) First two rows: the twin heritability and 95% CI; ii) third row: mean estimate from linear regression for NPD associated common variants (SA NPD-SNPs); and iii) bottom four rows: mean absolute effect size (Cohen’s d), with error bars showing the standard error of the mean. Y-axis: heritability estimates or effect sizes. X-axis: cortical regions ordered according to the cortical gradient from sensorimotor to association regions. Dotted line: correlation with the cortical gradient. Each panel displays Pearson correlation and *:spin-permutation significant, p-spin < 0.05. Abbreviations, Abs=absolute; ADHD=attention deficit hyperactivity disorder; ASD=autism spectrum disorder; BD=bipolar disorder; CD: conduct disorder; CHRPS: clinical high risk for psychosis; CHRPSn: CHR who did not develop a psychotic disorder; CHRPSp: CHR who later developed a psychotic disorder; CNV=copy number variant; CT=cortical thickness; Del=deletion; Dup=duplication; MDD=major depressive disorder; NPD=neurodevelopmental and psychiatric disorders; OCD=obsessive-compulsive disorder; SA=surface area; SCZ=schizophrenia; TS=Turner syndrome; Adult-Adolescence-Pediatric abbreviations: -y=adolescence/young; -p=pediatric).

Figure 5.. Latent dimensions of regional cortical differences.

A) The variance explained by the first principal component (PC1) for neurodevelopmental and psychiatric disorders (NPDs), their associated common and rare genetic variants. B) Correlation between latent dimensions. Pairwise spatial correlations between PC1 for CT and SA across NPDs, and associated common and rare genetic variants. C) Correlation between the cortical gradient and the latent dimension of cortical differences. Plots are arranged along the sensorimotor-association axis based on correlation with cortical gradient (AHBA gene expression principal component). Positive and negative correlation values indicate greater similarity with association and sensorimotor cortical regions, respectively. Each plot displays Pearson correlation and *:spin-permutation significant, p-spin < 0.05. D) PC1 variable loadings and variance explained for NPDs and genetic variants. Abbreviations, Abs=absolute; ADHD=attention deficit hyperactivity disorder; ASD=autism spectrum disorder; BD=bipolar disorder; CD: conduct disorder; CHRPS: clinical high risk for psychosis; CHRPSn: CHR who did not develop a psychotic disorder; CHRPSp: CHR who later developed a psychotic disorder; CNV=copy number variant; CT=cortical thickness; Del=deletion; Dup=duplication; MDD=major depressive disorder; NPD=neurodevelopmental and psychiatric disorders; OCD=obsessive-compulsive disorder; PC: principal component; p-spin: spin permutation based p-value; r: Pearson correlation; SA: surface area; SCZ=schizophrenia; SNPs: single nucleotide polymorphism; TS=Turner syndrome; % variance exp.= percentage of variance explained;