Polygenic risk scores for severe psychiatric disorders in bipolar disorders: associations with the clinical and dimensional expression, interactions with childhood maltreatment and mediation models

Abstract

Polygenic risk scores (PRSs) for several psychiatric disorders have been associated with the clinical presentation of bipolar disorder (BD). PRSs have also been suggested to moderate the associations between childhood maltreatment and BD severity. In this study, we investigated how PRSs for BD, schizophrenia, major depressive disorders (MDD) and attention-deficit/hyperactivity disorder (ADHD) might disentangle the clinical and dimensional heterogeneity of BD in a sample of 852 affected individuals. We used logistic and linear regressions, moderation and mediation models to test the associations between PRSs, dimensions in childhood/adulthood and clinical indicators of severity of BD. All models were adjusted for age, sex, BD type and depressive symptoms. None of the PRSs were significantly associated with the clinical expression of BD when considered in terms of mode of onset, course, or psychiatric comorbidities. Nevertheless, the PRS-ADHD significantly and positively correlated with the levels of childhood maltreatment, childhood ADHD symptoms, and of some adulthood measures (affective lability, impulsivity and hostility) with p values ranging from 3.10^-8-4.10^-4. None of the PRSs moderated the effects of childhood maltreatment on the clinical or dimensional variables. Mediation model suggested paths from both PRS-ADHD and PRS-MDD to childhood ADHD symptoms and childhood maltreatment. The links between PRS-ADHD to all adulthood dimensions were mediated by childhood ADHD symptoms (p < 0.002). In turn, some adulthood dimensions (mainly affect intensity and affective lability) were associated with the clinical severity of BD, as defined by rapid cycling, suicide attempts and anxiety disorders. In conclusion, this study disentangles the associations between the genetic liability for four psychiatric disorders and the clinical/dimensional heterogeneity of BD. We suggest a continuum from the genetic risk for ADHD and MDD through dimensions in childhood/adulthood to a severe/complex clinical expression of BD.

Fig. 1. Associations of each dimensional assessment in association with the four PRSs in multivariable analyses adjusted for age, se, bipolar disorder type, depressive symptoms and the six first components of the population’s genetic substructure.

Beta represent unstandardized coefficients obtained from the linear regressions of each dimension in association with the four PRS as dependent variables. adjusting for age, sex, BD type and MADRS score. All models were adjusted for the first six principal components of the population’s genetic substructure. For a question of clarity. unstandardized coefficients for age, sex, BD type, MADRS, and PCs are not displayed into the figures. Nominal p-values are indicated in brackets. A CTQ childhood trauma questionnaire (Log: log10 transformed), B WURS wender utah rating scale, C AIM affect intensity measure, D ALS affective lability scale, E BIS barrat impulsivity scale, F BDHIAtt buss durkee hostility inventory attitudinal component; G BDHIMot buss durkee hostility inventory motor component. PRS polygenic risk score, BD bipolar disorder, SZ schizophrenia, MDD major depressive disorder, ADHD attention deficit with hyperactivity disorder.

Fig. 2. Path analysis diagram including PRS-BD. PRS-ADHD. childhood measures. adulthood dimensions and clinical variables. adjusted for age. sex. BD type and MADRS score (n = 533).

Single-headed arrows represent regression paths and double-headed (straight and curved) arrows represent correlations. All path coefficients and correlations are reported as standardized estimates. The level of significance of path coefficients and correlations is given both by the type of arrow and the number of stars. For reasons of clarity. the paths corresponding to p values > 0.05 were not included in the path diagram. For reasons of clarity. the paths between adulthood dimensions are not displayed (all p < 0.001). For reasons of clarity. only significant associations in-between clinical variables are displayed (p < 0.005). PRS polygenic risk score, BD bipolar disorder, ADHD attention deficit with hyperactivity disorder, CTQ childhood trauma questionnaire, WURS wender utah rating scale (childhood ADHD symptoms), AIM affect intensity measure, ALS affective lability scale, BDHI Att buss durkee hostility inventory attitudinal component, BDHI Mot buss durkee hostility inventory motor component, BIS barrat impulsivity scale, RMSEA root mean square error of approximation, CFI comparative fit index, TLI tucker and lewis index.