Pathomechanisms and Signatures in the Longitudinal Course of Psychosis



016_ Polygenic Risk Scores across the extended psychosis spectrum

Research Question and Aims

Psychosis is a debilitating psychiatric condition, with highly polygenic nature, being a severe clinical hallmark of schizophrenia (SCZ) and bipolar disorder (BD). Both are characterized by a substantial genetic overlap (Smeland et al., 2019) and a positive clinical response to antipsychotic drugs (Singh, Chen, & Canuso, 2012). However, it is largely unclear how exactly the genetics contributes to this disease phenotype. In addition to that, recently, the categorical diagnostic nosology has started emphasizing the trans-diagnostic aspects (Krueger & Eaton, 2015) and the continuum-like spectrum of increasing symptom severity and persistence (Van Os, Linscott, Myin-Germeys, Delespaul, & Krabbendam, 2009; van Os & Reininghaus, 2016), suggesting some intermediate forms and a possible phenotype gradation. The polygenic risk score (PRS) cumulatively estimates the genome-wide effects of genetic variants and has been repeatedly shown to be associated with various forms of cognition, traits and disorders in clinical and non-clinical populations (Krapohl et al., 2016; Power et al., 2015). Following this rationale, the aim of the proposed project is to investigate whether there is evidence for a differential polygenic burden across the extended spectrum of psychotic symptoms. This question will be tested in a sample consisting of the following well-characterized subgroups: patients diagnosed with SCZ, patients diagnosed with BP, at-risk state for psychosis, at-risk state for BP, subclinical psychosis-like experiences (schizotypy), and non-psychiatric healthy controls. This project should help to elucidate questions around the genetic foundation of a psychosis spectrum from non-clinical to clinical forms, with a special consideration of a latent liability and genetic risk.

Analytic Plan

Phenotype definition
The analysis will use the DNA samples collected in Switzerland (Zurich area) and Germany (multi-center), all genotyped on the Illumina Infinium Psych-Array using ~ 600,000 markers. The Swiss samples were obtained in the framework of two research projects conducted at the University Hospital of Psychiatry Zurich, the ZInEP study (Zurich Program for Sustainable Development of Mental Health services) (Theodoridou et al., 2014) and the EE study ("Exceptional Experiences") (Unterrassner et al., 2017). The German samples derive from the PsyCourse database (Budde et al., 2019) and will include schizophrenia patients, bipolar disorder patients and non-psychiatric healthy controls.
The following groups will be analyzed:
- schizophrenia patients (F20.x diagnosis according to the ICD-10; 295.10/295.20/295.30/295.60/295.90 diagnosis according to the DSM-IV): n = 491;
- bipolar disorder patients (F31.x diagnosis according to the ICD-10; 296.0x/296.4x/296.5x/296.6x/296.8x diagnosis according to the DSM-IV): n = 463;
- individuals at-risk state (high and ultra-high) for psychosis (criteria for high-risk status: at least one cognitive-perceptive basic symptom or at least two cognitive disturbances according to the Schizophrenia Proneness Interview (Schultze-Lutter, Addington, Ruhrmann, & Klosterkötter, 2007); criteria for the ultra-high-risk status: at least one attenuated psychotic symptom, or at least one brief limited intermittent psychotic symptom according to the Structured Interview for Prodromal Syndromes (McGlashan et al., 2001), or state-trait criteria [reduction in global assessment of functioning (GAF) of >30% in the past year plus either schizotypal personality disorder or a first degree relative with psychosis]): n = 182;
- individuals at-risk state for bipolar disorder (score ?14 in the Hypomania Checklist (Angst et al., 2005)): n = 127 (partially overlapping with the at-risk state for psychosis);
- healthy individuals with psychosis-like experiences (based on the Schizotypal Personality Questionnaire (Raine, 1991), Magical Ideation Scale (Eckblad & Chapman, 1983), Exceptional Experiences Questionnaire-Revised (Fach, Atmanspacher, Landolt, Wyss, & Rössler, 2013) and Revised Symptom-Checklist-90 (Derogatis, 1996)): n = 269; - non-psychiatric healthy controls (without history of affective or psychotic illness, as assessed with the Short Diagnostic Interview for Mental Disorders (Margraf, 2013): n = 250.
Polygenic Risk Scores will be calculated for each individual by summing the number of risk alleles in multiple gene loci and integrating weights from the Genome-Wide Association Studies output libraries. The analysis will implement the following pre-imputation quality control criteria:
 98% subject call rate, 98% SNP call rate, Hardy-Weinberg equation >0.001, heterozygosity ?3 S.D, minor allele frequency (MAF) >0.1% and relatedness <0.9. The pre-phasing imputation will be done using SHAPEIT and IMPUTE2.
The post-imputed data will be filtered by means of the following thresholds:
INFO >0.7% and MAF >0.5%. The scoring for the polygenic risk scores will apply the latest SCZ and BD GWAS summary statistics within the PLINK 1.90.
The PRSs will be statistically compared for group means, as well as in a binary logistic regression analysis with group as target and testing two models, one with covariates alone (ancestry principal components - age, sex, recruitment center and age*sex) and one with covariates plus the PRS.
The deriving adjusted R2 or other measures of variance explained will be compared statistically.

The first hypothesis assumes the significant contribution of the PRS to the phenotype-specific variance explained.
The second hypothesis assumes the gradation of PRS group means (from negative to positive) from controls, psychosis-like experiences (schizotypy), at-risk state for psychosis and SCZ; and for controls, at-risk state for BP, and BP.
We will also test the linear association between the PRS and a symptom measure across the whole population, when available (PANSS).

Resources needed

The variables for analysis will be: polygenic risk scores, group/diagnosis, study center, age, sex and PANSS (if available). The necessary analytic support involves primarily calculating the PRS according to the state-of-the-art approach.