A schizophrenia missense variant in slc39a8 is associated with poorer cognitive ability: a cross sectional study of rs13107325

rs13107325 is a non-synonymous single nucleotide polymorphism within the gene SLC39A8, which encodes a ZIP8 transporter. The minor allele at rs13107325 is associated with schizophrenia, a psychiatric disorder marked by hallucinations, delusions, and impaired cognitive ability, and has a high probability (>99%) of being the causal variant at the associated genome-wide association study locus. We tested whether the number of schizophrenia-risk alleles at rs13107325 was associated with schizophrenia-related phenotypes within a sample of participants with schizophrenia. We hypothesised that schizophrenia-risk alleles will be associated with worse symptom severity, younger age of onset, poorer cognitive ability, lower IQ, and worse educational attainment. Our second aim was to test for equivalent phenotypic associations in a large sample of the general population. Participants with schizophrenia or schizoaffective disorder (depressed) were included from cross-sectional, multi-ancestry clinical samples: Cardiff COGS (N=662), Cardiff F-Series (N=422), and Cardiff SibPairs (N=148). Phenotypes included symptom dimensions for positive, negative, and disorganised symptoms and current cognitive ability (derived using confirmatory factor analysis of SAPS, SANS, and MATRICS scores); premorbid IQ; age at psychosis onset; years in education; and highest educational qualification. The MATRICS-based symptom dimension of current cognitive ability and premorbid IQ were only available in Cardiff COGS. Participants without a psychotic spectrum disorder (ICD-10 F20-F29) and of European ancestry, who either attempted the cognitive tasks or mental health questionnaire were included from the UK Biobank, a population-based cohort (N=364,014). Phenotypes included reported psychotic experiences (any, distressing, reoccurring); a measure of generalised cognitive function ‘g’; fluid intelligence, years in education; and highest educational qualification. We used principal component analysis to calculate g – using tests of numeric memory, reaction time, pair matching, and trail making. We performed regression analyses, adjusted for age at interview, gender, and genetic principal components and corrected for multiple testing within each sample. In meta-analyses of the schizophrenia samples, no associations were statistically significant. In individual samples, two nominally significant associations were observed, but neither survived correction for multiple testing: participants who carried more schizophrenia-risk alleles had lower premorbid IQ in Cardiff COGS (β = -0.21; P = 0.020) and a higher severity of positive symptoms in Cardiff SibPairs (β = 0.11; P = 0.022). In the UK Biobank non-schizophrenia sample, after correction for multiple testing, participants who carried more schizophrenia-risk alleles had a lower g score (β = -0.05; P FDR = 3.08 × 10-5); a lower fluid intelligence score (β = -0.05; P FDR = 1.98 × 10-1); were less likely to obtain GCSEs (OR = 0.97; P FDR = 0.011) or a degree-level qualification (OR = 0.96; P FDR = 1.28 × 10-4); and had fewer years in education (β = -0.03; P FDR = 0.020). Although an individual SNP is expected to contribute only a minimal degree of risk to any phenotype, we found that carrying more schizophrenia-risk alleles was associated with poorer cognitive ability, but not psychotic symptoms, in the general population. Larger sample sizes are required to test whether this is replicated in patients with schizophrenia.