Common variants for schizophrenia ascertained through genome-wide association of a cognitive endophenotype

Abstract

BACKGROUND: Schizophrenia is genetically complex, likely involving 1000s of common variants. To overcome small individual effect sizes, one approach is to create a polygenic risk score combining risk across SNPs (including those below the significance threshold) and weighted individually by their effect sizes. Here, we directly tested whether polygenic variants associated with an endophenotype for schizophrenia – verbal memory performance – in healthy controls (HC) predicted memory performance in independent HC and schizophrenia patient (SZ) samples. We also explicitly tested the implications of the common variant, common disease model by examining the relationship between endophenotype-generated polygenic scores and schizophrenia-related polygenic risk scores in patients with schizophrenia.

METHODS: Caucasian healthy controls (N = 649) and schizophrenia patients (N = 58) recruited for the Consortium for Neuropsychiatric Phenomics, as well as SZ (N = 61) and HC (N = 63) subjects recruited from the Swedish Twin Registry, completed the California Verbal Learning Task, a list-learning measure of verbal memory. A bootstrapping analysis of associations between CVLT performance and common variants was run (1000 iterations) in the CNP control sample and multiple cut-points (SNPs present in 700, 800, and 900 of 1000 associations) were used to calculate CVLT polygenic scores. Schizophrenia risk scores were calculated using variants identified by the Psychiatric Genomics Consortium in a large case-control sample.

RESULTS: Polygenic memory scores derived from weights based on the CNP control sample revealed significant associations with CVLT performance in the Swedish HC sample, as well as the CNP SZ sample. Higher schizophrenia risk scores were associated with lower CVLT polygenic scores (CVLT scores and performance are positively correlated) in both patient samples.

CONCLUSIONS: Polygenic memory scores reflecting variation in CVLT performance in healthy controls significantly predicted CVLT performance in independent HC and SZ samples. This is the first demonstration, to our knowledge, of polygenic scores derived from an endophenotype-based, genome-wide association study in schizophrenia successfully predicting the same phenotype (i.e., memory performance) in an independent sample. Furthermore, the relationship between schizophrenia risk scores and CVLT-derived scores provides the first genome-wide, molecular genetic evidence of the shared etiology between episodic memory performance and schizophrenia.