What Have We Learned About the Genetics of Obsessive-Compulsive and Related Disorders in Recent Years?

As noted previously (4), OCRDs, like all neuropsychiatric disorders (e.g., schizophrenia and major depression) and most common nonpsychiatric disorders (e.g., non–insulin-dependent diabetes and asthma), are viewed as genetically complex, multifactorial disorders. The term “complex” refers to the fact that it is unlikely that a major single genetic variation predisposes to or directly causes the disorder; rather, many changes in the genome may make minor contributions to disease risk, and the combination of such changes is likely to vary between individuals and populations. The term “multifactorial” refers to the fact that changes in the genetic code alone do not cause disease; rather, a combination of genetic and environmental (risk) factors most likely underlies causation.

Schizophrenia is a good example of a complex, multifactorial neuropsychiatric illness for which the genetic contribution is slowly being elucidated. Most recently, a whole-genome (i.e., all the DNA/genes in an organism) study identified approximately 150 different genomic regions as contributing to the disease (6). In almost all instances, the exact mutations in these regions have not yet been identified. Although these results are promising, the approximately 150 regions collectively explain only about 20% of the risk of schizophrenia. The remaining more than 80% may involve additional genes, genetic processes, or environmental or other shared or nonshared (random) risk factors. Some of these have recently been identified among so-called rare coding variants (7) and copy number variants (8), both representing a different group of genetic variation compared with so-called common variation (see below). The goal of all genetic studies is to tease out what proportion of the representation of illness (including risk of onset and other clinically relevant parameters) is likely to be genetic (as opposed to environmental or random) and, ultimately, to identify specific causative variations.

Historically, twin and family studies have been conducted to demonstrate that there are genetic contributions to a disorder. Genomic studies in which the DNA sequence is examined directly (or indirectly) are used to identify genomic regions shared by affected versus nonaffected relatives or, in studies of unrelated cases and unrelated controls, by cases more than controls. A case-control whole-genome design was used in the aforementioned schizophrenia study (6). In all such studies, the disease entity (i.e., phenotype) of interest must be specified. For example, some studies use only a narrow disease definition (e.g., a single disorder, such as OCD or TS), whereas others use a broad definition, often an entire spectrum (e.g., all OCRDs together). These types of studies not only can identify associated genes/regions for a specific disorder but also can help elucidate the relationship of genetic liability for one disorder to another (e.g., does TS share the same genes/regions as OCD?).

Another approach to phenotypic description involves specific features or disorder dimensions (e.g., neuropsychological, neuroanatomical, and neurocognitive) that cut across traditional disease boundaries and that are sometimes referred to as endophenotypes. The focus is on objective and heritable traits that are part of the etiology or representation of a trait but not necessarily specific to it. An endophenotype is presumed to be an intermediate in the causal chain between the observed behavioral phenotype and the underlying (biological) cause. For example, performance on neurocognitive tests may be used to tease out some of the differences that may be attributable to genetic variability between patients with OCD, their first-degree unaffected relatives, and unrelated healthy controls. These phenotypes will be an increasingly important component of genomic studies as more stable, heritable endophenotypic traits are identified (9).

Twin and family studies as well as genome-wide studies have been conducted in OCD, TS/chronic tics, and other related disorders. Most studies have been performed on the OCD or TS/chronic tics phenotypes (few studies using symptom dimensions or endophenotypic end points have been conducted to date). In addition to genome-wide studies, analyses of particular genes that may be plausibly involved in disease causation or that may influence the effectiveness of pharmacologic treatments (i.e., “candidate genes”) have been performed. Animal models derived from gene knockouts, biochemical manipulations, or innate characteristics (i.e., naturally occurring) have also been investigated. Excellent reviews of the literature on the genetics of OCD and TS, including twin, family, and genome-wide case-control studies, were recently published (10–12). Likewise, d’Angelo et al. (13) authored an excellent review of OCD animal models. Below we present a summary of recent findings.

As noted previously (4), twin and family studies are commonly used tools to describe and dissect the etiology of (common) disorders. Twin studies are used to study the question of nature versus nurture—i.e., what proportion of the phenotype variation is attributable to genetic factors in contrast to environmental factors? To do that, twin studies examine the concordance of disease in monozygotic (identical) twins versus dizygotic (fraternal) twins. Because monozygotic twins share all their genes and dizygotic twins share only one-half of their genes on average (similar to any other pair of siblings), rates of co-occurrence of disease in both monozygotic twins compared with dizygotic twins can be used to estimate heritability, i.e., the proportion of phenotypic variation that can be attributed to genetic factors. Family studies, on the other hand, are used to study recurrence risks in family members—i.e., the likelihood that a disease or trait present in one family member will occur again in other family members. Although inherently similar concepts, heritability and recurrence risk have different clinical utility.

After an era of so-called candidate gene studies that overall was found to be rather disappointing in terms of new insights into disease biology of OCRDs, genome-wide studies have become more popular, with decreasing costs and increasing technological advancements. As described above for the schizophrenia example, at least three different types of genetic variation are studied: common variation, copy number variation, and rare (coding) variation. Different study types have been used over the years to examine the different types of variation. Case-control study samples make use of cohorts of nonrelated individuals that are usually defined by presence (cases) or absence (controls) of a phenotype of interest. Family-based studies, on the other hand, are usually centered around a so-called index case (i.e., someone showing the phenotype of interest) and include that person’s relatives, who can either be affected by the disorder or who present themselves without the phenotype. A simple example for a family-based study sample is a so-called trio, an index case together with the parents. There are advantages and disadvantages to both types of studies. In brief, case-control studies allow for a cost-effective identification of disease-associated regions in the genome, whereas family-based studies are sometimes better suited to study disease pathophysiology through the eyes of inheritance. What is common to both approaches is that they usually do not come with a specific hypothesis about the involvement of particular genes or genomic regions; rather, these studies survey the genome for disease association. This is why they are sometimes referred to as “hypothesis-free approaches,” an unfortunate characterization of their nature, because they clearly test a hypothesis—that is, whether genetic markers are associated with disease status.

As noted previously (4), in addition to direct changes in DNA sequence that may disrupt normal gene activity, other types of variation can affect how genes play their role in the etiology of some disorders (including OCRDs). The term “epigenetics” usually refers to external modifications to DNA that can alter when and how much of a gene’s product is produced (i.e., expressed). It is important to understand that epigenetic changes do not alter the actual DNA sequence itself but do affect gene regulation and expression. Types of epigenetic mechanisms are of long-term or short-term consequence and include, among others, X-chromosome inactivation, genomic imprinting, chromatin remodeling, and gene expression regulation through microRNAs. Among different types of chromatin remodeling are histone modifications. Similar to variations in the DNA sequence, epigenetic modifications can be heritability transmitted; however, in many cases they are the consequence of lifestyle factors (e.g., diet, smoking, and physical activity) and environmental influences (e.g., different kinds of pollution), as well as downstream consequences of other illnesses (e.g., inflammation). Data on epigenetic changes in OCRDs are emerging and suggest that expression of certain genes in the brains of patients with OCD may be affected, but no statistically significant, replicated findings have been published.

In the above sections, we have focused on studies that were designed to identify disease-associated genetic markers with one disorder in mind. Although inherently capable of identifying genetic markers that are associated with more than one disorder, these studies typically focus on designs that include cases or family members with one specific disorder. Cross-disorder analyses, on the other hand, usually focus on more than one disorder and are designed to look across diagnostic boundaries. They achieve this by including either individuals who are diagnosed with more than one disorder (so called comorbid cases) or cohorts that are each focusing on a different disorder of interest. Cross-disorder analyses can be conducted with disorders that are closely related (e.g., OCRDs) or with disorders that have a known (clinical or epidemiological) relationship but are not closely related by means of a diagnostic manual (e.g., internalizing disorders and asthma). It is of note that results from current cross-disorder analyses typically do not take any causal relationship between disorders into account, nor can causal relationships in the genetic etiology for these disorders be typically inferred from the analyses. However, there is a growing trend to conduct studies employing approaches that allow for these interpretations (see end of section).

Cross-disorder analyses have been performed within the OCRD spectrum, most prominently for OCD and TS/chronic tics. The most recent analyses suggest a relationship in which there are shared but also distinct genetic components to each disorder (38, 39). This is an observation that has been shared across a variety of mental illnesses—i.e., many (closely related) disorders share part of their genetic risk architecture (see also above in the sections about CNVs and rare [coding] variants). However, their risk architecture also involves genetic loci that are uniquely associated and are not shared with other mental illnesses. For OCD and TS, the analysis by Yu et al. (39) showed that OCD with co-occurring TS or chronic tics may have a different underlying genetic susceptibility, compared with OCD alone. This is in line with a recent analysis in a Danish population-based sample that identified distinct patterns of genetic correlations for other disorders depending on whether individuals with an OCD diagnosis had also been diagnosed as having autism, ADHD, or major depressive disorder (40).

On a larger scale, recent cross-disorder analyses have tried to identify shared and unique associations across multiple mental illnesses, including OCRDs (usually OCD and TS). They have identified a number of genome-wide significant genetic loci that seem to be shared between mental illnesses, such as schizophrenia, autism, and ADHD, but also between OCD and TS (41). However, because of their own limited sample size (see above), the evidence for the association for OCD and TS with these loci still requires replication or at least future samples that include more individuals diagnosed as having OCRDs (see below). Nevertheless, one recent set of cross-disorder analyses borrowed its statistical approaches from other fields (such as epidemiology and sociology). The advantage of these approaches lies in their ability to model complex relationships between the genetic etiologies of the disorders under study, up to the point that the causal relationship in their respective genetic risk architectures can be inferred. An example of such an analysis was reported by Grotzinger et al. (42), who studied a compulsive disorder factor (among other factors) in their recent GWAS. One genome-wide significant genomic locus on chromosome 3p21.31 was identified to be associated with the compulsive disorder factor. This region was also among the genetic loci that the above-mentioned cross-disorder analyses (41) identified across multiple mental illnesses, including TS and OCD.

The complexity of OCD and its relationship to other OCRDs, such as TS/chronic tics, HD, BDD, TTM, and SPD, have been formally recognized in DSM-5; however, identifying the nature of the underlying genetic predisposition to these disorders may be both helped and hampered. Heterogeneity within and between the disorders (e.g., specific symptoms, symptom severity, comorbidities, and early versus late onset), population variation (e.g., sex and race-ethnicity), probable involvement of many genes (possibly hundreds) each with small effects (“polygenicity”) provide just some challenges for investigators. The interplay of known and unknown environmental factors with each other or with genetic factors (or both) creates additional hurdles for investigators. One clear principle, however, guides the field: the larger the sample and the more phenotypic descriptors available (e.g., comorbidity or endophenotypes), the more likely the identification of genetic contributions and ultimately the unraveling of the etiology. This has been demonstrated for schizophrenia, in that GWASs in recent years have found six genome-wide significant hits with approximately 10,000 cases, 62 hits with 25,000 cases, and 128 at the current 36,989 cases with schizophrenia (43–45). It must be noted, however, that although this relationship usually holds true across mental illnesses, the sample size at which successful identification of genome-wide significant loci is possible varies between disorders. The same is true for the pace (i.e., the number of genome-wide significant hits per added constant number of samples) at which we learn more about the genetic risk factors that influence risk the risk of developing a disorder of interest. For major depressive disorder, for example, an earlier GWAS with 9,240 cases was not able to identify a genome-wide significant association (46), and a recent analysis of 135,458 cases identified 44 genome-wide significant loci (47), falling significantly short of the 128 loci identified for schizophrenia, with a sample size roughly four times smaller. This is not to say that these disorders are, per se, less “genetic” (although that might be the case for some); it merely could mean that their “risk architecture” might be different and that other approaches—e.g., those looking for rare (coding) variation, CNVs, or epigenetic factors—might require an increase in sample size to learn more about the etiology of the respective disorder. The PGC TS/OCD working group and other groups around the globe are narrowing in on sample sizes that are required to identify genome-wide associated loci for OCRDs (TS, OCD, and others).

Studies of the genetic underpinnings of OCRDs have come a long way in recent years. Although the field still needs to catch up with scientific efforts in phenotypes, such as schizophrenia, bipolar disorder, and major depressive disorder, much has happened. Coming from an era in which genetic research in OCRDs was badly underfunded, we now see multiple large-scale studies funded for common and rare variation (48, 49). These efforts, together with existing large-scale biobanks and cohort studies, will help us within the next year to reach approximately 60,000 cases enrolled in genetic studies of OCD and it is hoped within the next 5 years even 100,000 cases. This will also be made possible with an extension of efforts from U.S.- and Europe-centric data collections into South America and countries in South and East Asia as well as in Africa. Currently, some of these efforts are either on the brink of receiving funding or are funded through both institutional and national programs. This seems important not only because of cultural influences on phenotypes such as OCD (50) but also because of the necessity to make sure that we are not falling further behind on the principles of equity, diversity, and inclusion (EDI). Among other articles, a recent overview article has outlined that the principles of EDI are not self-serving but will mark the cornerstones of our efforts to translate current genetic findings into treatments and clinical care (51). This holds true not only for GWASs but also for studies that target rare (coding) variation, CNVs, epigenetic changes, and other risk factors for OCRDs.

With regard to these translation efforts, pharmacogenomics is leading the field in OCRDs. The influence of genetic markers on antidepressant response in OCD has long been demonstrated (52). Although earlier studies of treatment response prediction focused on well-described markers that are known to affect the metabolization of antidepressants (e.g., CYP2D6), newer studies focus on larger sets of markers (53). Beyond pharmacogenomic approaches, the interplay of genes and the environment has also been used to study the response to treatment in OCD (54). On the genetic side, so called polygenic risk scores (PRS) have been included in these studies (54). PRS are a measure of the genetic vulnerability a person carries to develop a certain phenotype. In contrast to single SNPs, the effects of which are usually small on their own, PRS represent the accumulated risk across a group of SNPs that are associated with a phenotype. We are only at the beginning of a long road that leads from PRS as a research tool to clinical instruments (55, 56), but some argue that this road leads to nowhere. First studies, however, suggest that although not informative in their own right, PRS might have the ability to aid decision making in settings where other sources of information (e.g., diagnostic tools, classical pharmacogenetic testing, and imaging) have already set the tone for downstream treatment and care (55). Before this can happen, however, at least two steps need to be taken or completed: diversification of current study samples and clarification of studied phenotypes.

We have discussed the need to follow EDI principles, but we would like to reiterate the importance of doing so. Without proper inclusion and diversification, we are at an increased risk of entering a new era of stigmatization and discrimination toward those in our society who are already at an increased risk of developing mental illnesses and a decreased likelihood of receiving proper care.

In an environment of classification systems and strict guidelines for diagnoses, the request for clarification of studied phenotypes might sound like a request for more accurate phenotypes—it is not. Generally, individuals who have been diagnosed as having an OCRD and who have been included in a genetic study do have an OCRD. However, this “diagnosis” of an OCRD is not the only characteristic of their illness. They might be mildly affected or severely affected, their illness might have started at a relatively early age or they might have developed their illness in adulthood, and they might have been diagnosed as having another mental illness or they might not show any comorbidities. All these characteristics of their illnesses have roots in their genes. As a consequence, results from studies that focus, for example, on individuals diagnosed as having severe OCD can likely in a new study distinguish those with more severe OCD from controls. However, such studies will be less well equipped to distinguish those with a milder form of OCD from controls. Recent studies have shown exactly this issue for other mental illnesses, such as schizophrenia (57) and major depressive disorder (47). Therefore, it seems of outmost importance to be clear on all characteristics of the case and control cohorts enrolled in a study. Furthermore, this issue of severity points to the need to carry out genetic research across the phenotypic spectrum in OCRDs and not only include individuals who are severely affected by the disorders. Studies of healthy individuals with trait phenotypes related to mental illnesses (e.g., captured through self-report questionnaires) are as important as studies into the most severely affected (e.g., patients without treatment response in need of deep brain stimulation) if we want to achieve our goals of early detection and prevention. First studies have been conducted on the overlap between obsessive-compulsive trait phenotypes and OCD (26, 27); more studies across the OCRD spectrum (and in diverse samples) will need to follow in order to bring us closer to personalized prevention, care, and medicine.

We encourage all clinicians to stay abreast of research efforts through organizations and worldwide research efforts, such as the International OCD Foundation and the PGC. Furthermore, we encourage clinicians to inform themselves and their patients of resources and research studies available through such organizations and efforts (www.iocdf.org and www.pgc-ts-ocd.info).