Are Polygenic Risk Scores for Major Mental Disorders Associated with General or Specific Psychosis Symptom dimensions?

Background Psychotic symptoms can be conceptualised as dimensions of psychopathology cutting across diagnostic boundaries. Thus, they might be considered enhanced quantitative phenotypes to relate to genetic variants as summarised by Polygenic Risk Scores (PRSs) for Major Mental Disorders (MMDs), including Schizophrenia (SZ), Bipolar Disorder (BP), and Major Depressive Disorder (MDD). The objectives of this study were to: 1) identify the dimensional structure of symptoms at First Episode Psychosis (FEP), testing whether a bi-factor model statistically fits the conceptualization of psychosis as a single common construct (general psychosis factor) while also recognising multidimensionality (positive, negative, disorganized, manic, and depressive symptom factors); 2) examine the extent to which MMD PRSs indexed the phenotypic variance due to the general psychosis construct and to the specific symptom dimensions. Methods The sample included 1182 FEP patients recruited as part of the EUGEI study. The MRC Sociodemographic Schedule and the OPerational CRITeria (OPCRIT) were used to collect sociodemographic information and assess psychopathology. DNA was extracted from blood or saliva samples collected from 940 participants. The following analysis steps were performed: 1) OPCRIT psychopathology items were analysed using multidimensional item response modelling in Mplus to estimate unidimensional, multidimensional, and bi-factor models of psychosis. Models’ fit statistics were compared using Log-Likelihood, and Akaike and Bayesian Information Criteria. 2) SZ, BP, and MDD PRSs were built using the results from large mega-analyses from Working Groups of the Psychiatric Genomics Consortium. In PRSice, individuals’ number of risk alleles in the target sample was weighted by the log odds ratio from the discovery samples, and summed into the three PRSs. 3) For the best data fitting psychosis model, linear regressions were estimated to predict symptom dimensions as a continuous outcome from the three PRSs, accounting for population stratification. Results The best model fit statistics was observed for the bi-factor model including one general and five specific symptom factors compared with the other models. This indicated that there was a broad latent structure underlying the whole range of psychosis symptoms among five latent specific symptom dimensions. PRSs for SZ, BP, and MDD were calculated at the best model fitting P-value threshold. As expected, there was a substantial difference in discrimination of case-control status between SZ PRS and BP and MDD PRSs. A significant positive linear regression equation was observed for BP PRS and mania dimension severity (t(864)=2.74, p<0.01), explaining 5% of the variance; whereas a significant negative linear regression equation was found for MDD PRS and the negative dimension severity (t(864)=-1.75, p=0.05), explaining 3% of the variance. No significant association was found for SZ, BP, or MDD PRSs and the general psychosis trait score. Discussion These results suggest that a) psychosis at illness onset can be conceptualised as being composed of one general factor and five specific symptom dimensions, b) there is an association between mania dimension score and BP PRS. Despite the need to both replicate these findings also using PGC new released GWAS to build better powered PRSs, psychosis symptom dimensions have clearly been shown to be a valid and a useful continuous quantitative phenotype across categorical disorders.