1 November 2012. The search for the common variants that contribute to psychiatric disorders using genomewide association studies (GWAS) continues (see SRF genetics review), with researchers also refining their hypotheses to ask whether variants found in one population apply to other ethnicities, or whether some variants might modify individual features of disease. Although the combined effects of validated common variants are small so far (the so-called “missing heritability” problem), researchers seemed optimistic that there were more variants to be found, and that an ensemble of variants—rare and common alike—would combine to influence risk for psychiatric disorders.
Wednesday’s plenary speakers at the 20th World Congress of Psychiatric Genetics held in Hamburg, Germany, 14-18 October, emphasized this polygenic nature, with Peter Visscher of the University of Queensland in Brisbane, Australia, by arguing that liability for a disorder like schizophrenia would stem from a combination of rare and common variants, de novo variants, and environment, and that different cases would carry different “portfolios” of risk factors. Although rare and common variants contribute to schizophrenia risk, Visscher remarked on a strange absence of the “low frequency” variants in between (with minor allele frequency between 0.5 and 5 percent) that may be revealed through larger sample sizes and/or sequencing efforts (see SRF related news story). By reviewing how heritability is estimated in the first place, he suggested that a recent estimate for schizophrenia heritability based on national registry data may be more representative of the samples used in GWAS (Wray et al., 2012); though lower than other estimates (e.g., 0.67 compared to 0.81 from Sullivan et al., 2003), this is still a substantial amount of heritability, which he argued could not be accounted for by de novo events, gene-gene interactions, or epigenetics.
To commemorate his receipt of the Snow and Ming Tsuang Lifetime Achievement Award, Raymond Crowe of the University of Iowa, Iowa City, recounted highlights from his career, which began at a time when American psychiatrists were not convinced of the heritability of mental illness. Describing his first forays into the genetic basis of panic disorder, he said that even these early family studies pointed to a polygenic, complex trait. As new findings confirm this with an unimaginable diversity of variants, he quipped, “This is not your mom and dad’s polygenic model.”
Supporters of GWAS insist that larger sample sizes are required to get common variants for psychiatric disorders up and over the very high bar set for genomewide significance (p <5 x 10-8), and this has been the case for other complex traits like diabetes or height. Now, schizophrenia fulfills this prediction, too, according to Stephan Ripke of the Broad Institute, Cambridge, Massachusetts, who presented results from the largest yet GWAS for schizophrenia organized by the Psychiatric Genomics Consortium (no longer called the "Psychiatric GWAS Consortium," but retaining the acronym "PGC"). Since their previous study, which found seven signals (see SRF related news story), the consortium has more than doubled the size of their discovery set, reaching over 50,000 samples. This “second wave” produced 62 genomewide significant hits, and 80 percent of these variants have the same effect in a replication dataset. Though the effect sizes of the 62 signals were small, together accounting for only 3 percent of risk for schizophrenia, Ripke said that the polygenic score used to estimate the collective contributions of all promising variants—even those not reaching genomewide significance—put their contribution even higher. He declined to mention specific genes, but instead emphasized the collaboration that made the detection of these signals possible, with some researchers even contributing data before publishing on it themselves. “The best parties are when everyone brings something to share,” he said.
A key question is whether these GWAS-detected signals, which are largely derived from samples of European descent, generalize to other ethnicities. Do disruptions to the same sets of genes increase risk for schizophrenia among diverse populations? Several posters at the Congress presented efforts to replicate GWAS and candidate gene results in far-flung, though small, samples, with mixed results in Indonesia, Japan, and Bosnia.
In a talk on Monday, Teresa de Candia of the University of Colorado, Boulder, advised some caution about generalizing GWAS results, based on her estimates of the genetic risk for schizophrenia shared between European-Americans and African-Americans. Using data from the Molecular Genetics of Schizophrenia (MGS) sample, she reasoned that if the same SNPs predict schizophrenia in both populations, the genetic similarity among those with schizophrenia would be greater than the genetic similarity between cases and controls, regardless of ethnicity. This comparison gave a middling answer, with evidence for both unique and shared risk factors, with the shared portion decreasing as African ancestry increased (the African-American genotypes indicate mixing with populations of European descent). The lack of shared variance may reflect unique causal variants in the two populations, or it may be due to the different genome structures of different ethnicities, which here means that an SNP marking a causal variant in European-Americans may not be associated with that causal variant (i.e., not be in linkage disequilibrium) in African-Americans.
Dimensions of disease
Despite their diagnostic categories, different mental illnesses resemble each other somewhat, and studies find that they run together in families (see SRF related news story; SRF news story). Noting phenotypic similarities between schizophrenia and bipolar disorder, and the emerging genetic overlaps between them, Ripke of the Broad Institute presented results in a talk on Tuesday from the PGC cross disorder group, which aims to identify shared genetic loci across distinct psychiatric disorders. A GWAS on combined schizophrenia and bipolar disorder samples (n = 19,800) compared to controls (n = 19,400) revealed several genomewide significant hits, flagging known genes (e.g., CACNA1C) and new ones (e.g., PIK3C2A, a kinase within an intracellular signaling pathway). To identify disease-specific signals, the researchers compared the two disorders directly, but no genomewide significant hits emerged, possibly because of the limited sample size. Ripke also explored whether the combined effects of variants fingered in bipolar disorder GWAS (quantified with a polygenic score) could explain symptoms in schizophrenia, like mania, depression, positive symptoms, and negative symptoms. Only one correlation emerged, with polygenic scores of bipolar-detected risk alleles measured in people with schizophrenia increasing with mania severity.
This suggests that common variants can describe more than just risk for a particular disorder—they might also modulate severity of symptoms, even across disease categories. In the poster aisles, interest for this dimensional picture of common variant action also seemed to be building. For example, a poster from Ayman Fanous of Georgetown University in Washington, DC, described a search for genetic markers in schizophrenia cases that varied with age at onset, positive symptoms, negative symptoms, mania, and depression. No genomewide significant hits came out of the subset of the PGC sample he used, but several promising ones (p <10-5) highlighted nervous system genes that have not turned up before in GWAS focused on risk alone. This suggests these genes may work independently from risk genes to modify aspects of the disorder.
GWAS meets cytomegalovirus
In a talk on Tuesday, Anders Børglum of Aarhus University in Denmark presented data on gene-environment interactions that may contribute to schizophrenia, based on a unique Danish resource that banks bloodspots of every person born in Denmark since 1981. From the blood, researchers not only gather SNP data about the people, but can also evaluate their mothers' infection status at the time of birth by measuring antibody titers to certain pathogens. This allowed researchers to directly test a hypothesis about maternal infection during pregnancy, specifically infection by cytomegalovirus, contributing to schizophrenia risk (see SRF related news story). Though no genomewide significant signals were found in a GWAS of 915 cases and 915 controls, Børglum did report a significant interaction between a SNP flagging CTNNA3, a gene involved in cell adhesion, and maternal cytomegalovirus infection, which increased risk fivefold. Børglum suggested that cytomegalovirus may disconnect complexes involved in cell adhesion, which could include alterations to synaptic function in schizophrenia.—Michele Solis.