Structural brain imaging studies offer clues about the effects of the shared genetic etiology among neuropsychiatric disorders

Neuropsychiatric disorders have substantial heritability, as shown by many studies of twins and families. Genomewide association studies (GWAS) have shown that common genetic variants account for some of this heritability, and that some of this heritability is shared across neuropsychiatric disorders. The genetic overlap across disorders may partly explain why these disorders tend to co-occur with one another in both clinical and community samples. Subcortical brain volumes and cortical thickness/surface area dynamically change across the lifespan, with heritabilities of sMRI change phenotypes ranging from approximately 5% to 42% and varying with age. ENIGMA studies have characterized MRI-derived phenotypes across psychiatric and neurological disorders and reported significant case-control differences for ADHD, ASD, BD, common epilepsy syndromes, MDD, OCD, and SCZ. Here, the study estimates the degree of similarity in sMRI phenotypes among these disorders and evaluates whether such similarities are influenced by corresponding similarities in common genetic variant architectures.

- Data sources: Summary statistics (covariate-adjusted Cohen’s d standardized mean differences, SMDs) for regional subcortical volumes and cortical thickness/surface area from 12 multisite ENIGMA analyses for ADHD, ASD, BD, epilepsy, MDD, OCD, and SCZ. Segmentations were performed locally using ENIGMA protocols with FreeSurfer and ENIGMA QC. The analysis included 41 regions (7 subcortical, 34 cortical; left/right averaged). ADHD and ASD samples included youth and adults; other disorders were adults only. The epilepsy cohort included temporal lobe epilepsy, genetic generalized epilepsy, and extratemporal epilepsy. - Genetic data: Public GWAS summary statistics for each disorder (primarily PGC; MDD from PGC+UK Biobank; epilepsy from epiGAD). Variants were filtered (INFO ≥0.90, exclusion of MHC region), merged with HapMap3 (hg37) retaining MAF ≥5%. - Genetic correlations: Cross-disorder SNP-based genetic correlations estimated via LD-score regression using HapMap3 LD-scores. - sMRI similarity: Pairwise correlations of disorder-specific SMD vectors across 41 regions were computed (primary analyses report Pearson correlations across regions for each disorder pair; additional mention of Spearman rank correlations for sensitivity). Empirical permutation p values were generated by shuffling Cohen’s d values 10,000 times per disorder pair to derive null distributions. Multiple testing controlled using Bonferroni correction (significance threshold p = 0.00227). Leave-one-out analyses assessed influence of specific disorder pairs. Binomial sign tests assessed homogeneity of effect direction across disorders per region. Cochran’s Q tests evaluated heterogeneity of SMD magnitudes across disorders per region. Statistical analyses used R v3.5.2.

- Case-control sMRI effect sizes: Largest mean reductions across regions were observed in SCZ (mean Cohen’s d = −0.22, SE = 0.014), epilepsy (−0.12, SE = 0.017), and BD (−0.097, SE = 0.011); smallest in MDD (−0.018, SE = 0.006). All regions except caudate and putamen showed significant between-disorder variability in effect size magnitudes (Cochran’s Q p values = 0.012 to 2.8×10^−32). Eighteen sMRI phenotypes showed consistent direction of effect across disorders (binomial sign test p < 0.05), including multiple cortical thickness regions (e.g., insula, rostral anterior cingulate) and hippocampal volume. - Cross-disorder sMRI correlations (Pearson’s r across 41 regions): Highest positive correlation SCZ–BD r = 0.81 (df = 73, Bonferroni p = 2.38×10^−17). Other significant positive correlations after correction: BD–MDD r = 0.69 (Bonferroni p = 2.55×10^−10); OCD–SCZ r = 0.65 (Bonferroni p = 1.16×10^−8); MDD–SCZ r = 0.58 (Bonferroni p = 1.17×10^−6); BD–OCD r = 0.50 (Bonferroni p = 9.95×10^−5); MDD–OCD r = 0.46 (Bonferroni p = 6.89×10^−4); ASD–BD r = 0.38 (Bonferroni p = 0.02); ASD–SCZ r = 0.36 (Bonferroni p = 0.03). A significant negative correlation after correction: ADHD–BD r = −0.53 (Bonferroni p = 2.48×10^−5). Several nominal correlations did not survive correction (e.g., MDD–epilepsy r = −0.37; MDD–ADHD r = −0.33; SCZ–ADHD r = −0.32; ADHD–epilepsy r = −0.36; ASD–MDD r = 0.26). - Concordance with genetics: Cross-disorder sMRI correlations were positively associated with cross-disorder SNP-based genetic correlations. Overall association reported as r ≈ 0.49 (abstract) and Spearman’s ρ = 0.44, p = 0.049 (Figure 3). Leave-one-out analyses yielded ρ in the range 0.35–0.58 (lowest when removing SCZ–BD, ρ = 0.35), indicating robustness of the positive association. - Permutation analyses: Empirical permutation p values supported the strongest observed sMRI correlations (often <1×10^−4), acknowledging constraints due to spatial coherence and overlapping cortical maps.

The study demonstrates that patterns of structural brain alterations across cortical and subcortical regions are substantially similar for several neuropsychiatric disorders, notably SCZ with BD, OCD, and MDD, and that these similarities align with known cross-disorder genetic correlations. This convergence suggests that shared common variant architectures contribute to overlapping sMRI phenotypes, providing a biological substrate for observed clinical comorbidity across disorders. The negative correlation between ADHD and BD sMRI patterns aligns with differing neurodevelopmental trajectories. The moderate but significant association between sMRI cross-disorder correlations and SNP-based genetic correlations indicates that common genetic variation accounts for part of the shared neuroanatomical signatures, though environmental influences, rare variants, and disorder-specific factors likely also contribute.

This work aggregates ENIGMA-derived sMRI case-control effect sizes across seven neuropsychiatric disorders and links cross-disorder similarity in brain structure alterations to shared common genetic architectures. Key contributions include mapping robust sMRI similarity matrices across disorders and demonstrating their concordance with LD-score-based genetic correlations. Future research should incorporate spatially informed permutation frameworks, harmonize age and demographic profiles across cohorts, integrate multimodal imaging and longitudinal data, and assess contributions of rare variants and environmental exposures to shared and distinct neuroanatomical patterns.

- Use of summary-level data prevented proper adjustment for nonindependence due to sample overlap across imaging studies and for spatial coherence among brain regions, potentially biasing p values. - Inability to implement a spatial permutation framework because cortical thickness and surface area maps were jointly analyzed and overlapped. - Cohort heterogeneity: ADHD and ASD included youth and adults, whereas other disorders included adults only; ethnicity data were incomplete; epilepsy cohort comprised mixed syndromes. - GWAS constraints: MDD GWAS excluded 23andMe data; variant filtering and LD reference choices may influence rg estimates. - Summary statistics preclude adjustment for individual-level covariates beyond those already included in ENIGMA studies, and residual confounding may remain.