Impact Of Polygenic And Poly-Environmental Risk Factors On A Psychosis Risk Phenotype Explained Through Brain Structure

Abstract Background While single (genetic and environmental) risk factors for psychosis have been studied for their impact on brain structure and function, there is little understanding of how they interact to generate psychosis liability on the neural level. Direct associations between cumulative genetic risk scores and risk phenotypes are often weak, and analyses of G×E interactions are scarce. We developed and tested a multivariate model, in which the effects of cumulative environmental and genetic risk on a dimensional phenotype are mediated by brain structural variation. Methods In a data set of 440 non-clinical subjects, we tested a moderated mediation model with an interaction of an environmental (ERS) and a polygenic risk score (PRS) for schizophrenia, impacting on the subclinical psychosis spectrum phenotype schizotypy. We propose this effect to be mediated by grey matter volume variation, derived from voxel-based morphometry. In addition, cognitive function (CF) was considered as a potential moderator. Results Firstly, in a whole-brain analysis, we detected a significant interaction effect of PRS×ERS in a cluster (k=910, x/y/z=-4/-50/33, p=0.024 FWE cluster-level corrected) including the left precuneus (Pc, 64%) and posterior cingulate gyrus (pcG, 33%). Secondly, cluster values were extracted and entered into a multivariate moderated mediation model. This model was significant, showing that Pc/pcG volume mediated the impact of a PRS×ERS interaction on positive schizotypy (R2=10.91%, p=4.9×10–5). In predicting Pc/pcG variation (R2=51.69%), neither PRS (b=0.638, p=0.830) nor ERS had a main effect on grey matter variation, but their interaction was significant (b=-3.13, p=0.002): The intensity and direction of the PRS effect is moderated by the level of ERS, with a positive slope for low ERS (i.e., low environmental risk), and a negative slope for high ERS. In predicting positive schizotypy, the direct effects of PRS (b=6.116, p=0.477) and ERS (b=0.006, p=0.068) were not significant. However, we demonstrate an indirect effect through brain structural variation, showing a significant mediation (index=0.223, bootstrapped confidence interval 0.004–0.542). Cluster variation had a significant main effect on positive schizotypy (b=-0.277, p=0.049), but was modulated by the level of cognitive function, with a positive slope for low CF, and a negative slope for high CF, showing a second significant interaction (b=-0.070, p=0.027). Discussion Our finding is the first to integrate polygenic and poly-environmental markers with MRI parameters to demonstrate that the interaction of these cumulated risk factors leads to the emergence of subclinical symptoms through changes in brain structure. Furthermore, our model confirms cognition as a protective factor, indicating that above-average levels of cognitive function can compensate for dysfunctional processes that arise from altered neurodevelopment. Such compensatory mechanisms are crucial for understanding resilience, explaining high (positive) symptom load in unaffected individuals. Conventional diathesis-stress models propose increased vulnerability specifically to adverse events – our model extends this to suggest an inverted effect for high PRS and low ERS subjects. Under favourable environmental conditions, an increased genetic load might paradoxically result in low psychopathology outcomes or gain of function, supporting the notion of genes associated with schizophrenia as “plasticity genes” rather than simple risk factors. In sum, the present study provides proof for a multivariate model predicting the impact of genetic and environmental risk on a psychosis risk phenotype, extendable to other clinical spectra.