Questions about DISC1 as a Genetic Risk Factor for Schizophrenia

A welcome change in psychiatric genetics has been the widespread recognition of the essential role of uncompromising statistical rigor and replication. Put simply, the genome is a big place and it is trivially easy to find false leads, non-significant but “suggestive” genomic findings that an integrative scientist might find “intriguing”.

Indeed, due to advances in sequencing technology, the next few years are certain to see an explosion in observations of unique events in people with schizophrenia and other psychiatric disorders. Some of these will be claimed to be causal. However, the paucity of results from exome sequencing of sizable samples in autism ^1-3 and schizophrenia ^{4, 5} combined the surprisingly high rates of deleterious exonic variation in apparently normal people ⁶ indicate that it will be very challenging to delineate disease-related variants from background noise. For example, even with the improbably optimistic assumptions that 1% of schizophrenia cases are caused by fully penetrant mutations in one gene that has no confounding background variation, observing 10 deleterious mutations in 1,000 cases and 0 in 1,000 controls would not be clearly delineated from the distribution of test statistics across 15-20,000 genes. ⁷ In reality, locus heterogeneity, incomplete penetrance, and realistic background variation will make this task markedly more difficult.

As there is already an influential example of a unique genomic event, it is timely to review the genomics of “Disrupted in Schizophrenia 1” (DISC1), a t(1;11) (q42.1;q14.3) ^† structural variant identified using cytogenetic methods. ^{8, 9} Over 20 years after the initial report, the status of DISC1 as a risk factor for schizophrenia remains unclear and perhaps polarized: some researchers are convinced that it is a proven etiological factor in schizophrenia and others convinced it is not. Other groups await empirical data to resolve its role. Indeed, my group has found non-significant but “intriguing” results about DISC1 twice, and both times its salience faded with more data. ^{10, 11}

The purpose of this editorial is to review the genetic evidence for the involvement of DISC1 in schizophrenia. There are important unanswered questions that need to be resolved for DISC1 to be established as a bona fide genetic risk factor for schizophrenia.

Views of DISC1 in the Literature

Some consider DISC1 as a proven risk factor for schizophrenia. ^{12, 13} Examples of statements about DISC1 include: “this private mutation has revealed important mechanisms of disease”, ¹⁴ “a key susceptibility gene for schizophrenia is DISC1”, ¹⁵ “a susceptibility gene for schizophrenia”, ¹⁶ “a convincing candidate gene”, ¹⁷ and “DISC1, a major susceptibility factor for several mental disorders”. ¹⁸ Some psychiatric disorders have been termed “DISC1opathies”, ¹⁹ and DISC1 has been referred to as the “special gene”.

The DISC1 Pedigree

The pedigree was initially reported in 1970, and identified via an 18 year-old male karyotyped in a cytogenetic study of boys sentenced to a youth prison in Scotland. ²⁰ The propositus had conduct disorder, and none of his first-degree relatives had a psychotic disorder.

Three cytogenetic abnormalities were reported to segregate in this pedigree: a balanced translocation between chr1 and a group C chromosome (chr6-12), a separate chr1 “unusually large secondary constriction”, and a Robertsonian translocation between two group D chromosomes (chr13-15). To my knowledge, the most recent report of the phenotypes in the pedigree was in 2001, ²¹ but the 2001 pedigree is considerably smaller than the 1970 report. Diagnoses were established using a structured diagnostic interview by psychiatrists blind to genotype, and of 29 individuals with t(1;11) (q42.1;q14.3): 11 (37.9%) had no diagnosis, an anxiety disorder, conduct disorder, or alcohol dependence; 10 (34.4%) had recurrent major depressive disorder (rMDD); and 8 (27.6%) had a psychotic disorder (7 schizophrenia and 1 bipolar disorder). Parametric linkage analyses under a dominant model maximized at 7.1 when rMDD, schizophrenia, and bipolar disorder were considered affected. The next largest LOD of 4.5 was for mood disorders (rMDD and bipolar disorder), and schizophrenia alone had a LOD of 3.6.

These reports do not answer multiple questions of interest to the research community (Table 1). First, it is possible that t(1;11) (q42.1;q14.3) status was based on laboratory assessments done over 40 years ago. This should give any researcher pause, particularly if the key linkage analyses in Blackwood et al. ²¹ are based on the Jacobs et al. ²⁰ structural variant assignments. Second, I could find no published explanation or analysis of why the researchers focused on one of the three structural variants reported to segregate in this pedigree. Third, critically, sensitivity analyses were not reported (i.e., systematically changing diagnoses within the pedigree and reevaluating linkage evidence). The importance of these analyses was amply illustrated by the old-order Amish linkage studies in the late 1980s where a LOD of 4.9 faded to non-significance with a few changes in the pedigree. ²² It is possible that the reported LOD scores are fragile and sensitive to changes in diagnostic status.

Table 1.

Unanswered questions about DISC1.

Unanswered questions about DISC1
The pedigree
Have the karyotypes from the late 1960s been updated with modern methods? Were the key Blackwood et al. (2001) linkage analyses based on the Jacobs et al. (1970) karyotypes?
Three structural variants were reported to segregate in this pedigree: which can be verified with modern methods? Which segregate with psychiatric phenotypes? What was the justification for focusing solely on t(1;11) (q42.1;q14.3)? Why was the rest of the pedigree not reported?
The most recent phenotype reports are from 2001. How have the diagnoses changed? What effect do changes in diagnosis have on the linkage results? Given that linkage results can be sensitive to influential subjects, what do sensitivity analyses show?
The phenotypes that appear to track with t(1;11) (q42.1;q14.3) are dissimilar to other rare structural variants where schizophrenia, autism, epilepsy, and/or mental retardation are associated. The prominence of rMDD is worrying. Why is this pedigree different? Does the absence of these other conditions suggest that DISC1 is not a true schizophrenia risk factor?
The focus on DISC1
It is possible that the chr1 DISC1 side of the breakpoint is not centrally important: what role does the chr11 side of the breakpoint play (e.g., the predicted lincRNA)?
Much rests on the assumption that the translocation that impacts DISC1 is causal. However, efforts to falsify this genomic hypothesis are few. How can genomic data be used to more clearly implicate or exclude DISC1?
Genetic results
The DISC1 translocation is private to a single pedigree. The largest and most rigorously conducted genomic studies of common variation, rare variation, and copy number variation provide no support for a role of DISC1 in schizophrenia, bipolar disorder, autism, and MDD. A rigorous analysis of pleomorphic effects similarly found no evidence for a role for DISC1. Do these negative exclude DISC1 with confidence?

Fourth, the logical connections of t(1;11) (q42.1;q14.3) with schizophrenia are not compelling. The propositus and his immediate relatives have conduct disorder. The linkage analyses are more consistent with a mood disorder phenotype. The high prevalence of recurrent MDD is disconcerting given the predominant role of environmental risk factors in its etiology. ^{23, 24} Of greatest concern is that mental retardation, autism spectrum disorders, and epilepsy have not been reported to segregate with t(1;11) (q42.1;q14.3) in this pedigree. This is atypical for rare SVs of strong effect that tend to increase risk for multiple neuropsychiatric disorders. ⁷

DISC1 proponents have argued that the lack of a uniform connection to a single psychiatric phenotype is expected and consistent with genetic risk factors having pleomorphic effects. Empirical data have suggested that pleomorphic effects are indeed the case; ²⁵ however, this does not appear to be a cleanly falsifiable argument in this pedigree. Indeed, if this argument were true, the authors make the case that “disrupted in schizophrenia” is a misnomer.

The Focus on the Chr1 Translocation Region

The t(1;11) (q42.1;q14.3) structural variant was identified as disrupting a novel gene that was given the name DISC1. ⁸ Although this was a standard medical genetics approach, there are additional unanswered questions. First, the chr11 side of the breakpoint disrupts a predicted long intergenic non-coding RNA (lincRNA, ENST00000562245.1 or RP11-660M18.2). Such RNA molecules are expressed, do not code for protein, but can have important regulatory roles. Second, as noted above, t(1;11) (q42.1;q14.3) is one of three structural variants reported in this pedigree, and other structural variants could be relevant. Third, the members of this pedigree share considerable amounts of the genome identical-by-descent; have the relevance of other genetic variants been excluded? Is a gene-disrupting translocation in DISC1 merely a red herring for causal variation elsewhere in the genome? For example, some translocations are not copy number neutral, causal genetic variation in the vicinity of the breakpoints could be “hitchhiking” due to limited recombination within the pedigree, and the disease status could result a mechanism with no connection to DISC1.

Finally, causal environmental effects can also cluster in extended pedigrees. The high prevalences of conduct disorder and recurrent MDD in this pedigree are notable. As these can emerge from the “matrix of disadvantage”, it is possible that non-genetic effects play an etiological role in this pedigree.

Genetic Findings in Other Samples

To the best of my knowledge, t(1;11) (q42.1;q14.3) is private to this Scottish pedigree and has never been found elsewhere, and there are no copy number variants in the DISC1 region significantly associated with schizophrenia, bipolar disorder, or autism. ⁷

Genome-wide linkage meta-analyses for schizophrenia and bipolar disorder do not provide support for DISC1 or for chr11 side of the translocation. ^{26, 27}

For common genetic variation, candidate gene studies have reported genetic associations with various psychiatric disorders in DISC1. However, these small studies are known to have issues with quality control. The largest and most comprehensive studies show no common SNP association signal in the DISC1 region. The PGC schizophrenia GWAS mega-analysis (9,394 cases and 12,462 controls) had a minimum P=0.02 in DISC1, a level of significance about 1 million times from than required for genome-wide significance. ²⁸ A separate meta-analysis of DISC1 variants from 10 candidate gene and three GWAS (11,626 schizophrenia cases and 15,237 controls) found no significant associations even at a liberal gene-wise significance level. ²⁹ Similarly, a yet larger GWAS shows no DISC1 evidence (Sullivan, in preparation). There are also no notable findings on the chr11 side of the translocation.

Some papers have hypothesized that the effects of DISC1 are pleomorphic in the sense of predisposing to multiple psychiatric disorders. The PGC cross-disorders group has conducted an integrated GWAS mega-analysis of 61,220 subjects including cases with schizophrenia, bipolar disorder, MDD, autism, and ADHD. ²⁵ This analysis directly and systematically evaluated pleomorphic effects of common genetic variation in DISC1, effectively testing whether any common SNP was associated with more than one disorder. There were no notable associations in the DISC1 region (minimum P=0.02, a million times larger than that required for genome-wide significance).

There are few published resequencing studies of DISC1, and larger and more comprehensive studies are in progress. To date, the largest published study of rare exonic variation was negative (discovery in 727 schizophrenia cases and 733 controls, replication in 2,191 cases and 2,659 controls). ¹¹ Some smaller studies have claimed association although replication efforts were absent or negative. The strong assertion that 2% of the attributable risk for schizophrenia was due to rare DISC1 variants ³⁰ has not replicated.

The Genetic Evidence for DISC1 is Not Strong

Confident associations in human genetics require evidence of statistical association beyond chance and replication in multiple independent samples. ³¹ Moreover, we have come to expect exceptional quality control and vigorous efforts to understand the impact of many different types of bias. In my view, the central goal of psychiatric genetics is now the identification of high-confidence associations and not the potential confusion engendered by lists of “intriguing” findings.

The published genetic evidence for an association of DISC1 with schizophrenia does not meet a high standard. The genetic evidence is limited solely to cytogenetic abnormalities within a single pedigree. There are no independent lines of genetic evidence (e.g., structural variation in other pedigrees, evidence for increased exonic deleterious mutations in cases, or common variant associations). The apparent absence of autism, mental retardation, and epilepsy and the presence of recurrent MDD and conduct disorder in this pedigree are perplexing and atypical.

It is certainly possible that the questions in Table 1 are readily addressed or have already been answered via analyses of which I am unaware. However, one cannot escape the conclusion that the genetic findings for DISC1 do not now meet community standards in human genetics. DISC1 stands apart: the genetic evidence in support of other rare variants of strong effect has increased in the past decade whereas the genetic evidence for DISC1 has not. ⁷

What about biology? DISC1 proponents argue that its fascinating roles in the development and function of the brain trump the genetic findings. This argument is not accepted in mainstream human genetics: biology does not play a role in establishing a genetic association (but only later in understanding its role). Invoking biology to cover up deficiencies in the genetic evidence is a slippery slope. Most genes have a direct or indirect role in CNS biology and any integrative scientist worth his or her salt could make an “intriguing” case for a large fraction of human genes. Indeed, I recently calculated that 61% of the genes in the human genome are of legitimate interest in psychiatric genetics. Thus, to connect DISC1 to any psychiatric disorder requires ironclad genetic associations, and these are currently lacking.

Names are powerful things, and, at present, one could reasonably posit that “disrupted in schizophrenia” is a misnomer and prone to misinterpretation. The official HUGO gene name unmistakably implies a highly certain role in the etiology of schizophrenia. Unless the genetic evidence improves in the near future, wouldn’t it be scientifically responsible to change the name of DISC1 to a more neutral descriptor?