REVIEWS Psychiatric genetics: progress amid controversy Margit Burmeister*‡§, Melvin G. McInnis‡ and Sebastian Zöllner‡|| Abstract | Several psychiatric disorders — such as bipolar disorder, schizophrenia and autism — are highly heritable, yet identifying their genetic basis has been challenging, with most discoveries failing to be replicated. However, inroads have been made by the incorporation of intermediate traits (endophenotypes) and of environmental factors into genetic analyses, and through the identification of rare inherited variants and novel structural mutations. Current efforts aim to increase sample sizes by gathering larger samples for case–control studies or through meta-analyses of such studies. More attention on unique families, rare variants, and on incorporating environment and the emerging knowledge of biological function and pathways into genetic analysis is warranted. candidate genes Heritability The hereditary basis for psychiatric disorders was positional and recent genome-wide The proportion of phenotypic already recognized at the turn of the nineteenth century association (GWA) studies. Although it is not clear variation that is explained by by Emil Kraepelin. Twin and adoption studies consist- which of the hundreds of reported associations will be genetic variation. ently demonstrate a genetic influence on all major confirmed, throughout this Review we emphasize the psychiatric disorders1, confirming work that started lessons that emerge from some convincingly replicated in the 1930s2. In fact, estimated heritability for bipolar findings. We demonstrate how alternative phenotypes disorder, schizophrenia and autism (80% to >90%)3–7 is — such as intermediate traits or endophenotypes — or much higher than that of breast cancer (5% to 60%)8,9 unique families with rare disorders have helped gene and Parkinson disease (13% to 30%)10, for which several identification. The incorporation of environmental fac- genetic risk factors are now well established1. In many tors from longitudinal studies has led to some signifi- respects, psychiatric disorders are similar to other com- cant and reproducible genetic findings, with variants *Molecular and Behavioral 11,12 Neuroscience Institute, plex traits that have been studied genetically : stud- in the serotonin system being implicated in depression University of Michigan, 5061 ies are complicated by locus heterogeneity, imprecisely and impulsive-aggressive behaviours. Recent reports BSRB, 109 Zina Pitcher Place, specified traits, incomplete penetrance and interaction of rare large chromosomal deletions in autism and Ann Arbor, Michigan with non-genetic factors, resulting in a low contri- schizophrenia suggest that rare mutations might need 48109-2200, USA. odds ratios ‡Department of Psychiatry bution of each individual risk allele ( <2). to be considered in common psychiatric disorders and Depression Center, Linkage studies have low power to detect such low-risk as well. We end by discussing some of the biological University of Michigan, alleles, and association studies are only powerful if the pathways that are involved in psychiatric disorders, and Rachel Upjohn Building, risk allele is common (that is, with a minor allele fre- how genome-wide expression data, animal models 4250 Plymouth Road; quency >0.05), or if the odds ratio can be increased by and bioinformatics help in deciphering the biochem- Ann Arbor, Michigan 48109-5734, USA. focusing on a disease subtype or by incorporating envi- istry and neuroscience of the disorders. Functional §Department of Human ronmental factors. Furthermore, the winner’s curse leads knowledge in turn might be useful in the dissection of Genetics, 4909 Buhl Building, to overestimation of effect sizes, thereby handicapping further genetic risk factors. 1241 East Catherine Street, replication13. Because some genetic factors apparently influence Ann Arbor, Michigan 48109-5618, USA. Here, we focus on progress that has been made behaviour or susceptibility for psychiatric disorders ||Department of Biostatistics, in psychiatric disorders, primarily concentrating on without falling into a specific diagnostic category, inte- University of Michigan, major depression, bipolar disorder, schizophrenia and grating genetic risk alleles with phenomenology is likely M4140 SPH II, 109 South autism. The current debate centres on the nature of the to markedly change the current classification scheme Observatory Sreet, Ann Arbor specific biological influences, the best ways to identify in psychiatry. In addition, we have only limited knowl- Michigan 48019-2029, USA. Correspondence to M.B. them, and how the environment interacts with genes edge of the pathological mechanisms that lead to any e-mail: [email protected] and their products to cause disease. We review link- psychiatric disorder, and the identification of genetic doi:10.1038/nrg2381 age and association findings, including biological and risk factors might provide a better understanding of NATURE REVIEWS | GENETICS VOLUME 9 | JULY 2008 | 527 © 2008 Macmillan Publishers Limited. All rights reserved. REVIEWS the aetiology of psychiatric disorders. Although phar- and another to chromosome 11p in an Amish pedigree, macogenetics is not discussed here, genetic findings quickly faded when these results turned out to be false Locus heterogeneity might lead to novel treatments and/or personalized positives15–18. The linkage era (1980–2005) for psychiat- When variation in different medicine. With highly controversial genetic tests for ric disorders failed to identify any single locus that was genes affects the same 14 phenotype; it is also known as psychiatric disorders now launched , but initial GWA unequivocally replicated across multiple independent genetic heterogeneity. This studies of psychiatric disorders unable to confirm pre- samples. Scientists suspected that the individual studies should be contrasted with vious findings or each other, it is timely to take stock were affected by low power, and so it organized col- allelic heterogeneity, in which of the field, to outline the progress that has been made, laborative efforts and meta-analyses. Nevertheless, a multiple variants in the same gene affect the same disease. and to candidly review the controversies and prospects meta-analysis of schizophrenia linkage studies found for the future of psychiatric genetics. only one genome-wide significant linkage peak — in Penetrance a region never before implicated in this disorder19. The probability of observing a Psychiatric diagnosis Findings in meta-analyses of bipolar disorder depended specific phenotype for A condition of any genetic analysis is a valid and accu- on the method that was used to combine data. Non- individuals carrying a particular parametric linkage analysis rank genotype. If this probability is rate phenotype. However, there is no biochemical or using -based methods 20 smaller than 1 for all genotypes physiological test available for psychiatric disorders detected no genome-wide significant linkage findings , of a variant then the variant has that is equivalent to, for example, the one that measures whereas joint analysis of all linkage data sets identified incomplete penetrance. glucose level for type 2 diabetes. A psychiatric diagnosis two genome-wide significant peaks — on chromosome 21 Odds ratio is made through clinical examination, usually by direct 6q and on 8q . Linkage studies of autism have been A measure of effect size. interview, and summarized on the basis of diagnostic somewhat more consistent, with chromosomes 2q, 7q, Defined as the ratio of the criteria (TABLE 1; Supplementary information S1 (table)). 15q and 16p implicated reproducibly22–24, as well as less odds of a disease being Problems arise in the overlap of categories; for exam- reproducible peaks25. Astonishingly, the first two link- observed in one group of ple, psychosis (illusions or hallucinations) can be part age studies of obsessive-compulsive disorder identified genotypes and the odds of a (REFS 26,27) disease being observed in of at least three different diagnoses — schizophrenia, a peak in the same area of chromosome 9 , another group. bipolar disorder and psychotic depression (FIG. 1). In although this region was not confirmed by a later and addition, the boundaries of diagnostic categories can much larger study28. The existing linkage results are Linkage study be blurred when symptoms in patients are not clear. now incorporated in positional candidate gene associa- A family-based method to search for a chromosomal For this reason, many genetic studies report on their tion studies and pathway analyses. Positional cloning location of a gene by findings using ‘narrow’ criteria: in the case of bipolar starting with linkage can lead to gene identification in demonstrating co-segregation disorder, for example, this would indicate that the sub- rare families with unique, essentially Mendelian forms of the disease with genetic ject had documented evidence of acute mania (required of mental illness (see below), but this approach is not markers of known for a diagnosis of bipolar I disorder), a clinical state that currently being pursued as the major research avenue chromosomal location. is strikingly similar across patients. Most studies use in common heterogeneous psychiatric disorders. Association study a ‘final best estimate’ process that uses data from the Genetic study looking for interview as well as clinical records and family history. Genetic association studies association between a disease