A New Era in Psychiatric Genetics?

Several major psychiatric disorders, including schizophrenia, bipolar disorder, and autism, have been shown to have strong underlying genetic components and are among the most heritable of genetically complex medical illnesses. However, 30 years of psychiatric genetics research demonstrates that high heritability does not mean ease of finding genetic causes. Indeed, the history of psychiatric genetics is largely a story of unreplicated discoveries and unrealized expectations. At the risk of raising expectations once again, this disappointing history may be about to change, and a new chapter of discovery may be about to begin.

The recent completion of the International HapMap Project has led to the identification of a large set of single nucleotide polymorphisms (SNPs) that capture much of the common variation in the human genome (the International HapMap Consortium 2005). With the advent of high-throughput genotyping chips that can survey more than 500,000 of these SNPs, it is now feasible to map individual variation quickly to compare large numbers of cases and control subjects. For the first time, whole genome association studies can be undertaken with the potential for linking common SNPs as well as copy number variations to vulnerability to mental disorders. The Foundation for the National Institutes of Health (NIH), a nonprofit organization that supports public–private partnerships, recently announced funding of six whole genome association projects (http://www.fnih.org) via the Genetic Association Information Network (GAIN). Of note, four of these six projects were mental disorders: schizophrenia, bipolar disorder, major depressive disorder, and attention-deficit/hyperactivity disorder. In part, these four were selected from more than 30 applications across medical disciplines because of the availability of several thousand DNA samples in the National Institute of Mental Health (NIMH) Genetics Repository (http://www.NIMHGenetics.org). Additional whole genome association studies are underway for autism, major depressive disorder, and bipolar disorder.

The march into whole genome association comes with great expectations because of two recent successes. In adult-onset macular degeneration (DeWan et al 2006, Edwards et al 2005, Haines et al 2005, Klein et al 2005 and Yang et al 2006), which until recently was not considered an obvious candidate for finding genes of large effect, SNPs in two genes have been shown to predict over 50%of the variance. Moreover, these are both complement factor genes, suggesting a novel pathophysiology and entirely new approaches to prevention and treatment. Similarly, whole genome association studies of inflammatory bowel disease (Crohn’s disease and ulcerative colitis) have revealed high levels of association with SNPs in the IL23R gene on chromosome 1p31 (Duerr et al. 2006). This gene encodes a subunit of the inflammatory cytokine interleukin 23 and, in what may become a pattern with “disease genes,” includes a less common variant that confers protection against inflammatory bowel disease (Nadeau and Topol 2006).

Will we see similar breakthroughs with whole genome association studies in mental disorders? Perhaps. If genetic variations are associated with these disorders, finding them will require 1) large sample sizes, 1) well-defined phenotypes, and 3) robust analytic tools. For some mental disorders, we have the DNA and clinical data already in the NIMH Genetics Repository, but the requisite numbers may be greater than we once assumed. Attempts to model sample size suggest that the power for detecting a modest association (roughly odds ratios of 1.4 or 1.5) may require more than 2000 cases and 2000 control subjects (Donnelly, presented at the GAIN Analysis Workshop, Dulles, Virginia, 2006). Most current collections include fewer than 2000 cases for any mental disorder. For some disorders, such as obsessive–compulsive disorder and anorexia, we know of no collection of even 1000 samples. The future will belong to data sharing so that samples of adequate size with appropriate consents can be aggregated. The NIMH Genetics Repository continues to expand its collections, including DNA from ethnic minority populations, to serve as a resource for psychiatric genetics.

The issue of phenotypes is recurrent within psychiatric genetics and is often used as an explanation for nonreplication. A simple rule for identifying phenotypes with an underlying genetic basis is to follow patterns of inheritance. Although this has worked for breast cancer and Alzheimer’s disease, genetic epidemiology remains a relatively underdeveloped field in psychiatry despite the success of this approach in other complex genetic disorders. Much has also been written about intermediate or endophenotypes, but the value of this approach for finding genes remains to be demonstrated. This is an era in which further research in the phenome of mental disorders (to understand the components of clinical phenotypes) and the underlying genetic epidemiology (to understand the familial patterns of segregation of clinical phenotypes and their change over time) is needed. There is little confidence that DSM-IV will track with patterns of heritability or with genetic variation, but in the absence of anything better, clinical diagnosis will be one landmark for defining phenotype. Medication responses, either clinical improvement or sensitivity to adverse events, might be one of the phenotypes to yield a genetic signature of practical value, but here again psychiatry lags behind other areas of medicine. Genomic approaches may be able to identify who will develop metabolic syndrome or tardive dyskinesia on antipsychotics or who has a form of depression that will not respond to selective serotonin reuptake inhibitors. It is possible that whole genome association projects will provide a panel of candidate SNPs that could be tested in large-scale prospective clinical trials for personalizing treatment even before genomics provides a predictive test of autism or schizophrenia.

Much will depend on having powerful analytical tools to sift through the ocean of data emerging from these studies. Several recent articles have described the challenge of analyzing whole genome association data (Hirschhorn and Daly 2005, Pe’er et al 2006 and Skol et al 2006), but the field is still in its infancy. The NIMH is committed to expanding this important area of research (see RFA MH-07-060; Available at: http://www.nimh.nih.gov/grants/rfa.cfm) with additional initiatives in the planning stages. Optimal progress will depend on broad access to both genotyping and phenotyping data. For the GAIN studies, both genotypes and phenotypes will be available immediately following cleanup of the genotyping results. The team submitting the DNA and phenotypes will have 9 months’ exclusive rights to publication, but it is our expectation that many additional analyses will be undertaken by others to test out their favorite hypothesis and that these studies can begin as soon as the data are available via a data access portal. This ambitious data-sharing strategy, which resembles practices in physics and computer science, involves a culture change for biomedical science. We believe that the need to find genetic signatures is of urgent public health significance and that data collected with public funds should be accessible to anyone who can assist in this search.

Is whole genome association the latest flash in the pan? We don’t think so, but we caution against premature optimism. There are three concerns. First, the current approach focuses on common SNPs. The assumption is that complex diseases such as schizophrenia or depression result from an unfortunate alignment of common SNPs, each contributing a small increase in vulnerability but interacting in aggregate and, together with other factors (environment, epigenetic modification, etc.), to pass the threshold for disease. It is certainly possible that SNPs that are uncommon in the general population (HapMap population) play a much larger role as the source of vulnerability for mental disorder. These could be missed by the current approach leading to false-negative results even with a whole genome association study of more than 500,000 common SNPs. Second, even with whole genome association and large enough samples, we are still stuck with the heterogeneity of the population. The recent evidence that as many as 10% of an autism sample has an illness that can be explained by recombinant events (such as Fragile X or duplications of MeCP2) reminds us that de novo events and epigenetic modifications likely contribute to the pathophysiology of many of those clinically diagnosed with these disorders (Beaudet and Zoghbi 2005) This, of course, is an opportunity as well as a challenge, but it will not improve the resolution of whole genome association. Finally, it is worth noting that finding a list of candidate SNPs is the beginning or perhaps the end of the beginning of the genetic search. The ultimate goal is to identify a molecular lesion, understand the cellular pathway for the affected protein, and then translate this discovery into a diagnostic test or novel intervention. For mental disorders with many associated alleles, we will need a wave of sequencing to identify molecular lesions and a generation of cell biology to understand their significance. If we are entering a new era in the search for genetic association, this is only the first step on a long road to mapping the risk for mental disorders.