12 November 2009. In a special plenary session of the World Congress of Psychiatric Genetics 8 November 2009 in San Diego, Pablo Gejman of NorthShore University HealthSystem and Northwestern University, Evanston, Illinois, gave the first public presentation of data from the largest genomewide analysis of common variation in schizophrenia to date. The news was good: the analysis of 12,200 cases and 9,300 controls yielded six loci with genomewide significant p values. The results confirmed previously reported associations of three regions (MHC, 18q, and 11q) and revealed three more (7p, 8q, 10q).
As Chair of the schizophrenia working group of the Psychiatric GWAS Consortium, Gejman took care not to hype the data. “The process is going very well, and we are making incremental progress,” he said.
Helping that progress, Gejman said, is the fact that the PGC has recently incorporated 30,000 additional samples from the SGENE consortium, which the investigators will use for a replication study on this first round of findings.
The PGC, first proposed in 2005, now involves more than 165 member scientists from 68 institutions in 19 countries. All together, the members have amassed genotyping data on 69,000 subjects with schizophrenia, bipolar disorder, major depression, ADHD, and autism, along with healthy controls. The “Freeze 1” data that Gejman presented included 10 GWASs incorporating 17 independently collected samples form Europe, North America, and Australia. The study, which collected and analyzed primary data (genotypes), is far larger than any of the previously published works, each of which had roughly 3,000 cases. It also differs from the meta-analyses published last summer, which pooled top p values from three smaller GWASs (see SRF related news story on the SGENE, International Schizophrenia Consortium, and Molecular Genetics of Schizophrenia samples).
Power is good!
The boost in power paid off with multiple genomewide significant hits. Three repeat loci include an extended region in the major histocompatibility (MHC) locus on chromosome 6p21.3-22.1. In that area, which spans eight megabases and 230 genes, 129 SNPs met criteria for genomewide significance (p values <5 x 10-8), with the top SNP checking in at 5.7 x 10-11 and an odds ratio of 1.19. Significant SNPs in other loci had p values in the range of 2-3 x 10-8, including one near the TCF4 gene on chromosome 18q21.2, and another in 11q24.2, two regions also picked out in the SGENE meta-analysis (Stefansson et al., 2009). The 11q SNP in the SGENE paper was near the neurogranin (NRGN) gene, a postsynaptic protein, and a schizophrenia candidate gene. The PGC analysis found a different SNP, near a different, but equally interesting gene, PKNOX2. The gene encodes a transcription factor that was significantly associated with substance abuse in European women (Chen et al., 2009). However, the report was subsequently retracted because the authors violated an embargo on the use of data they obtained from a public database (see Holden, 2009), which leaves the finding in limbo until another group can repeat the analysis.
Regions of new association included 7p22.3-22.2, 8q21.3, and 10q24.32. The 7p region includes the MAD1L1 gene, which functions in mitotic checkpoint control and has been previously implicated in cancer.
Gejman concluded that the results are consistent with a polygenic architecture of schizophrenia, and indicate that further increases in sample size will probably reveal additional significant loci. That raised a question from the audience of what would be the necessary sample size to make the investigators feel they had done a complete analysis. Gejman replied, “It is not what I want; it is what there is. Even with our best efforts, we may not be able currently to greatly expand our sample size. There are another 30,000 subjects for replication that are not in our GWAS—that gives us the ceiling as we know it now.”
Gejman did not show data for the SNP in the ZNF804A zinc finger gene reported by O’Donovan and colleagues (see SRF related news story on O’Donovan et al., 2008), but he told SRF their data were supportive of a role for that SNP. (The ZNF804A association was recently replicated [Riley et al., 2009], and the gene was the subject of much work presented at the conference. New findings on ZNF804A will be the focus of anther installment of our conference report).
Bipolar and cross-disorder analyses
The other disease groups had their day, too. John Kelsoe of the University of California, San Diego, reported on four genomewide significant findings in bipolar disorder, based on analysis of 7,481 samples and 9,250 controls. The study replicated ANK3 (see SRF related news story on Ferreira et al., 2008) and found potential new loci on chromosome 6, 11, and 12. On chromosome 6, the analysis zeroed in on the SYNE1 gene, which encodes a nuclear membrane protein and has been implicated in several Mendelian ataxias. Part of that locus also includes the CPG2 gene, which encodes a protein involved in spine formation and glutamate receptor expression, and which Kelsoe called “a novel and interesting candidate.” The chromosome 12 locus was the home of many genes, while on chromosome 11, Kelsoe highlighted ODZ4, a brain-expressed gene of unknown function. For the other diseases (depression, autism, attention deficit-hyperactivity disorder), none gave any genomewide significant results, but none had as many subjects as schizophrenia or bipolar disorder.
In addition to identifying common variants that play a role in individual diseases, the PGC is also pursuing cross-disorder studies, with the goal of understanding shared genetic risk for multiple psychiatric diseases. Jordan Smoller of Harvard Medical School presented some preliminary analysis from that group, which looked at the association of CACNA1c in a mixed group of schizophrenia and bipolar subjects. A SNP in the gene, which encodes an ion channel, has been linked to both bipolar disorder (Ferreira et al., 2008) and schizophrenia (Green et al., 2009). In the PGC analysis, the SNP did not reach genomewide significance in either group, but did when the two were combined. Moreover, the odds ratio for the combined group (1.11) was the same as for each individual disease. This suggests that the gene has equal influence on the two disorders, Smoller said.
What does it mean?
At the end of the session, SRF caught up with Pamela Sklar, director of genetics at the Broad Institute and Co-chair with Kelsoe of the PGC bipolar disorder working group. “The major message from this meeting is that there are genes for schizophrenia and bipolar. We have observations that are not false positives,” Sklar said. Also, it’s clear that both rare and common variants are involved, she said, and results to date support the idea that the field needs to build up sample sizes, and go forward by applying whatever techniques are appropriate for each of the individual diseases. “There’s a lot more to do. We now need to tease out how these genes contribute to produce disease in one individual,” she concluded.
For a complicated collaboration, the PGC has gone surprisingly smoothly, said Patrick Sullivan of the University of North Carolina at Chapel Hill, who chairs the group studying major depressive disorder. “Almost everyone we approached saw the logic, and the degree of cooperation has been high. A lot of us thought it was going to be more trouble, but it has been relatively straightforward. I think most people get the concept that this is required for us to move forward.”
If you were at the session, we welcome your comments, additions, or corrections to this story.—Pat McCaffrey.