Psychotic-like experiences, polygenic risk scores for schizophrenia, and structural properties of the salience, default mode, and central-executive networks in healthy participants from UK Biobank

Abstract

Schizophrenia is a highly heritable disorder with considerable phenotypic heterogeneity. Hallmark psychotic symptoms can be considered as existing on a continuum from non-clinical to clinical populations. Assessing genetic risk and psychotic-like experiences (PLEs) in non-clinical populations and their associated neurobiological underpinnings can offer valuable insights into symptom-associated brain mechanisms without the potential confounds of the effects of schizophrenia and its treatment. We leveraged a large population-based cohort (UKBiobank, N = 3875) including information on PLEs (obtained from the Mental Health Questionnaire (MHQ); UKBiobank Category: 144; N auditory hallucinations = 55, N visual hallucinations = 79, N persecutory delusions = 16, N delusions of reference = 13), polygenic risk scores for schizophrenia (PRS_SZ) and multi-modal brain imaging in combination with network neuroscience. Morphometric (cortical thickness, volume) and water diffusion (fractional anisotropy) properties of the regions and pathways belonging to the salience, default-mode, and central-executive networks were computed. We hypothesized that these anatomical concomitants of functional dysconnectivity would be negatively associated with PRS_SZ and PLEs. PRS_SZ was significantly associated with a latent measure of cortical thickness across the salience network (r = -0.069, p = 0.010) and PLEs showed a number of significant associations, both negative and positive, with properties of the salience and default mode networks (involving the insular cortex, supramarginal gyrus, and pars orbitalis, p_FDR < 0.050); with the cortical thickness of the insula largely mediating the relationship between PRS_SZ and auditory hallucinations. Generally, these results are consistent with the hypothesis that higher genetic liability for schizophrenia is related to subtle disruptions in brain structure and may predispose to PLEs even among healthy participants. In addition, our study suggests that networks engaged during auditory hallucinations show structural associations with PLEs in the general population.

Fig. 1. Diagram of nodes of interest and their linking pathways derived from the connectome for the networks of interest.

Colours were randomly assigned, and width of the links are proportional to mean FA values across all participants. Axial and coronal views of the networks with colour-coded FA gradient. L: left, R: right. A list of abbreviations is provided in Supplementary Table 2.

Fig. 2. Diagrams of structural equation models (SEM).

a Multiple indicators, multiple causes (MIMIC) model for neurostructural properties of each network. A separate model was applied to FA, grey matter thickness (CT), and grey matter volume (GMV). From each individual bilateral node (L: left; R: right) or pathway, a latent score was calculated for FA, CT, GMV controlling for sex and whole-brain structural properties at the manifest level (i.e., whole brain FA/CT/GMV). Relation between FA/CT/GMV and PRS_SZ is indicated by path β; path γ represents the association between the interaction of age and PRS_SZ and FA/CT/GM factors; path δ represents the association between age and the latent factor. PRS_SZ was corrected for sex and population stratification (paths not shown). The dashed line represents a possible residual correlation between nodes. b Path diagram of mediation model, where the ε coefficient represents the regression of X on M, λ coefficient the regression of M on Y, and σ coefficient the direct path of X on Y. The product of the ε and λ coefficients describes the indirect path of X on Y through M (σ′).

Fig. 3. Diagrams of MIMIC and independent pathway models of the salience network.

a Common + independent pathways model for latent factor of grey matter volume. b Common + independent pathways model for latent factor of grey matter cortical thickness. Dashed lines represent independent pathways while dotted lines residual correlations between nodes. Black lines represent statistically significant pathways (p < 0.050) while grey lines non-significant pathways (p > 0.050).

Fig. 4. Raincloud plots representing differences in cortical thickness (CT) and volume (GMV) between participants reporting PLEs (blue) and matched controls (green).

a Bilateral CT of the insula and auditory hallucinations. b GMV of right supramarginal gyrus and delusions of reference. c GMV of the left pars orbitalis and persecutory delusions. d CT of the right supramarginal gyrus and persecutory delusions. Note: CT is measured in cm and GMV in cm³. Asterisks represent significant differences between groups (*p < 0.05, **p < 0.01).

Fig. 5. Path diagram of the mediation model, where the ε coefficient represents the coefficient of regressions of PRS_SZ on cortical thickness (CT) of the right insula, and λ the coefficient of the regression of CT of the right insula on auditory hallucinations, and σ coefficient of the direct path of PRS_SZ and auditory hallucinations.

Coefficient of σ′ represents the indirect path of PRS_SZ on auditory hallucinations through CT of the right insula. Note: Superscript a represents p values < 0.01 and superscript b refers to confidence intervals not including zero.