Following the meta-analysis of the combined mixed-ancestry schizophrenia, schizoaffective, and bipolar cohorts, 40 variants surpassed genome wide significance (P-values < 5 × 10−8) (). The 40 SNPs mapped to 6 loci, 5 of which had been previously associated with susceptibility to SCZ and/or BP (ITIH1, SDCCAG8, MHC, MAD1L1 and CSMD1; ). Two SNPs mapped to a novel locus containing a gene of unknown function, TSNARE1 (t-SNARE domain containing 1) on chr8q24.3 (top SNP rs10098073: fixed effects P 1.28 × 10−9 random effects P 1.28 × 10−9 OR 1.108; rs4129585 fixed effects P 2.38 × 10−8 random effects P 1.28 × 10−9 OR 1.108). In addition to the two genome-wide significant SNPs, multiple other SNPs in LD showed a trend towards association at the locus (). Odds ratios for the most significantly associated SNP (rs10098073) across the 16 cohorts ranged from 1.003 to 1.258 (SD 0.06) with one outlier, the Dublin cohort, crossing 1 at OR 0.97. The genomic inflation factor of the meta-analysis was 1.08 indicating that population stratification had been adequately controlled. Analyzing the Caucasian cases and controls alone (n = 11,681 cases and 24,498 controls), the top TSNARE1 SNP, rs10098073, remains genome-wide significant (rs10098073 P-val 3.947 × 10−8 OR 1.12; rs4129585 P-val 1.061 × 10−7 OR 1.17).
Genome wide significant variants following SCZ meta analysis
Manhattan plot of the SCZ-BP meta-analysis.
TSNARE1 regional association plot.
While variants at TSNARE1 have not reached genome-wide significance in previous analyses, rs10098073 showed a consistent trend towards association with schizophrenia12
. For the PGC discovery and replication analyses, the P
-value for rs10098073 was 2.09 × 10−5
; OR 1.077; Analysis of the same SNP in the CLOZUK10
study produced a P
-value of 0.0165; OR 1.101. After combining the PGC and CLOZUK studies, analysis of rs10098073 resulted in P
1.19 × 10−6
; OR 1.101. The CLOZUK sample (n = 2652) has not been included in any other GWAS apart from the primary study12
. There are no overlapping samples between the CLOZUK study and our sample set and therefore the CLOZUK sample can serve as an interdependent replication of the association.