Genome-Wide Association Study of Cognitive Decline in Schizophrenia

To the Editor: Substantial evidence suggests that many patients with schizophrenia experience a decline in intellectual functioning. Approximately 50% of patients with schizophrenia show cognitive deterioration, with an IQ decline of 10 points from premorbid IQ (1), and cognitive decline in schizophrenia remains stable regardless of baseline characteristics and changes in clinical state (2). As there is considerable interindividual variation in the degree of decline, it is conceivable that genetic influences have a role in determining the severity of cognitive decline in schizophrenia.

We conducted a genome-wide association study (GWAS) of cognitive decline in 166 patients with schizophrenia. The mean estimated premorbid IQ of these individuals was 101.2 (SD=10.0) using the Japanese Adult Reading Test (JART), and the mean current full-scale IQ was 85.1 (SD=16.8) using the WAIS; the difference score (subtraction of JART score from full-scale IQ) was −16.1 (SD=13.1). We performed a multiple linear regression analysis to compare the difference scores in major allele homozygous genotypes and minor allele carriers, with gender and years of education as covariates, using the PLINK software package, version 1.07. Detailed information regarding the participants and methods is provided in the data supplement that accompanies the online edition of this letter (see “Supplementary Methods and Data” and Table S1). Although we did not observe any association at a widely used benchmark for genome-wide significance (p=5×10⁻⁸), the strongest association was observed at rs7157599 on chromosome 14, a missense polymorphism (Asn8Ser) in the DEGS2 gene (p=5.4×10⁻⁷). The most significant 10 markers are listed in Table 1, and the top 200 markers are listed in Table S2 in the online data supplement. rs17069667 is an intronic single-nucleotide polymorphism (SNP) in the CSMD1 gene, which has been identified as a new risk gene for schizophrenia by a large-scale GWAS (3). Associations between the 10 SNPs and the estimated premorbid IQ were not observed (p>0.3 in all cases); however, associations between the 10 SNPs and full-scale IQ were observed in all of the SNPs (Table 1). Analyses using an additive model and with age, gender, illness duration, and antipsychotic dosage as covariates also revealed slightly reduced associations, but these remained significant (see Table S3 in the online data supplement).

Our results suggest associations at the p<1×10⁻⁵ level between difference score in schizophrenia and four genes, one of which has been identified as a new locus for schizophrenia (CSMD1). Replication analysis using the National Institute of Mental Health Clinical Brain Disorders Branch sample revealed a directionally consistent trend association of genotype for a proxy of the top SNP, rs7157599 (rs3783332: r²=0.63, one-tailed p=0.03). Although the study should be replicated with a larger sample size, our results indicate that the measurement of cognitive decline in schizophrenia as a quantitative phenotype (in conjunction with GWAS) could be a gene discovery tool. We should note, however, that we cannot rule out an interaction of our gene variants with the effect of antipsychotic drugs on cognition.