DISC1 as a genetic risk factor for schizophrenia and related major mental illness: response to Sullivan

In a recent guest editorial¹, Patrick Sullivan questioned several aspects of the original Scottish t(1;11) family through which DISC1 was discovered^2–4. He challenged the wider significance of the original genetic finding and questioned the relevance of DISC1 biology to psychiatry, casting doubt on the validity of the ‘integrative’ approach. Sullivan prefaced his strong views on DISC1 by welcoming the ‘uncompromising statistical rigour and replication’ that is now being applied to his area of expertise in genome wide association studies (GWAS). In his opinion, the evidence in support of DISC1 falls short of the mark, but the methods he champions are based on the assumption of the ‘common disease; common variant’ hypothesis and are not valid tests of the ‘common disease; rare variant’ hypothesis of which DISC1 is a prime example. If we were to apply his criteria universally, we would, for example, be obliged to ignore everything learnt about Alzheimer’s Disease through APP, PSEN1 and PSEN2, none of which show up in GWAS, yet these clear-cut, family-based rare variant discoveries have provided profound insights into the disorder.⁵ We would likewise have to ignore much of the insights that have been gained from copy number variant analysis in autism^{6, 7} and, indeed, schizophrenia^{8, 9} Denying an important role for rare variants makes no sense and contradicts theoretical considerations¹⁰: insights from exceptional cases and families have repeatedly served medical research well across the full spectrum of conditions generally considered ‘common and complex’.

So why does Sullivan pick out DISC1 for special criticism? This is not the time or place to reiterate the wealth of evidence in favour of DISC1 and of the DISC1 pathway, but readers might find it useful to consult recent reviews^11–14 in addition to those he cites¹. Here, we summarise his Table 1, answer his main questions and expand upon these in the main text.

Table 1.

Answers to questions from Sullivan about DISC1

The pedigree The classical karyotyping methods developed in Edinburgh by Jacobs et al42 are robust for tracking the t(1;11) translocation2, were used effectively again in Blackwood et al, 2001, but for convenience, efficiency and utility, have since been supplemented with fluorescence in situ hybridisation and now PCR methods: all methods give entirely congruous results. There is no evidence from follow up studies for any significant change from the original report that the t(1;11) tracks susceptibility to a spectrum of schizophrenia, bipolar disorder and major depressive disorder with genome wide significance. A comprehensive follow up on the t(1;11) family, including brain imaging, has recently been completed and will be submitted for publication in due course. The phenotypes These are not ‘worrying’. The presence of major depressive disorder along with bipolar disorder and schizophrenia is supported by GWAS-based co-heritability analysis. The absence of mental retardation is unsurprising: mental retardation is a common co-morbid feature of many deletion syndromes, with or without other psychiatric features, but not of balanced translocations. The focus on DISC1 The focus on DISC1 derives from the remarkable properties attributable to the protein – interaction and modulation of multiple proteins that have been independently linked genetically to psychiatric and neurodevelopmental disorders and biologically to fundamental neurological processes of signalling, trafficking, synaptic function and neurogenesis. Other molecular genetic consequences of the t(1;11) have been considered and are published. None challenge the primacy of the DISC1 disruption. Genetic results The examples cited as the ‘most rigorously conducted genomic studies’ are selective and biased towards meta-analyses of linkage and association, blunt tools to test for rare variants. Positive published findings with genome wide significance have been ignored.

Sullivan queries the Robertsonian t(13q;14q) translocation reported in a small branch of the Scottish t(1;11) pedigree, and the chromosome 1 constriction that is more-prevalent but predominantly in branches of the family that do not carry the t(1;11). The former was in fact discussed in the Lancet study², and neither materially alters the conclusion that the t(1;11) robustly tracks the high risk of psychiatric illness in this family. The LOD score for schizophrenia alone meets stringent genome wide significance; for schizophrenia plus bipolar disorder and major depressive disorder it substantially exceeds genome wide significance (MLOD=7.1). Regarding the alleged disparity between the pedigree in the original 1970 report¹⁵ and the follow up report in 2001³, there is nothing suspicious here. Any clinician who has followed up families in their care over such a long period of time will fully understand that not all of the original members will still be willing or able to participate and that new members and new cases may become available. That the follow up study confirmed and substantiated the original study should give comfort not concern. For the avoidance of any further doubt, we can confirm that the pedigree presented in the 2001 study focuses only upon branches of the family carrying the t(1;11), that the t(1;11) shows unequivocal linkage to the high burden of major mental illness in this family and that the t(1;11) has been scored both by classical chromosome banding and by fluorescence in situ hybridisation¹⁶. The molecular cloning of the t(1;11) breakpoint⁴ has more recently allowed us to develop a PCR-based assay spanning the breakpoint with sequence confirmation: we routinely apply this test to validate samples in contemporary studies (Thomson, unpublished). Finally, regarding the Scottish t(1;11) pedigree, a major third wave of follow up has recently been completed with brain imaging added to the clinical phenotype: this will be submitted shortly for publication.

Sullivan finds the spectrum of psychiatric diagnoses seen in the t(1;11) family ‘worrying’, but what the t(1;11) family showed par example is evidence for genetic and biological overlap between schizophrenia, bipolar disorder and major depressive disorder. The epidemiological evidence is now clear on this^{17, 18} as too is the accumulating GWAS evidence for shared genetic liability across DSM-defined diagnoses^{19, 20}. The presence of both schizophrenia and major depressive disorder in the t(1;11) family is entirely consistent with recent GWAS-derived estimates of co-heritability: schizophrenia and bipolar disorder (0.68); bipolar disorder and major depressive disorder (0.47); schizophrenia and major depressive disorder (0.43)²¹. The variability in presentation in the Scottish t(1;11) family may be accounted for by genetic (independently segregating modifiers) or non-genetic (experiences and exposures) factors: these are under active investigation. In a similar vein, there is nothing ‘worrying’ about the absence of mental retardation: mental retardation is seen frequently in chromosome deletion conditions without any other psychiatric features²². Finally, it is indeed the case that the original propositus was a male with adolescent conduct disorder: schizophrenia is known to increase in prevalence with an increasing number of conduct problems in childhood²³.

With regards to the focus on DISC1, other possible explanations – a functional effect of the disrupted chromosome 11 locus, an independent or co-dependent effect of the long non-coding RNA antiparallel to DISC1, named DISC2, and/or a dominant negative effect of DISC1 fusion proteins - have not been ignored. We and others have considered and actively investigated all of these alternative hypotheses^24–26, and will continue to do so as the necessary tools, reagents and assays are developed. As part of on-going work, we are carrying out whole-genome sequencing of t(1;11) family members and will report the findings in due course.

With regards to independent genetic results, the failure of meta-analyses of genome wide linkage and association to replicate DISC1 is misleading: it is well understood that the presence and impact of rare or recent genetic events go undetected by these methods⁵. GWAS depends upon genotyped alleles tagging risk loci that are then implicated indirectly by a significant difference in allele frequency between cases and controls. The statistical tests may be highly significant, but the population level odds ratios are typically less than 1.2. Given the reliance on strong linkage disequilibrium at a whole-sample level between the genotyped common allele and the risk allele, this method is neither designed nor suited to the detection of rare risk alleles: this is true even for those of high penetrance and impact, for example structural or coding variants, for which the odds ratios are typically greater than 2¹⁰. Family-based linkage studies in relatively isolated populations have the potential to bridge the gap between ‘common’ and ‘rare’ variant detection, and this is true for DISC1. In a population based study of 221 Finnish families with schizophrenia, D1S2709, an intragenic marker for DISC1, gave a Z(max) of 2.71 in the combined cohort increasing to 3.21 when an internal isolate was excluded. Linkage to DISC1 was replicated in an independent set of 70 Finnish families with schizophrenia²⁷. Analysis of the combined Finnish samples exceeds genome-wide significant linkage for DISC1 and schizophrenia (LOD=3.6)²⁸. When conditioned on the DISC1 risk haplotype, the NDE1 locus was identified as a strong linkage peak (LOD = 3.17), a finding that was replicated by association (P<0.011 after correction for multiple testing)²⁸. Although NDE1 does not appear as a top GWAS ‘hit’, deletions and duplications spanning NDE1 are among the most common copy number variants in schizophrenia²⁹. NDE1 is critical for multiple neurodevelopmental processes: through protein-protein interaction, DISC1 regulates NDE1 function³⁰. Thus, consistent with current neurodevelopmental concepts in schizophrenia³¹, the genetic and biological evidence for DISC1 and NDE1 provide evidence for a shared ‘risk’ pathway. NDE1 is just one of a large number of proteins for which there is firm evidence for direct interaction with DISC1: several likewise have biological functions that have been independently linked to psychiatric illness^11–14. None independently meet PGC criteria for genome wide significance, but some meet gene-wide significance. Of note, the DISC1 interactor PDE4B encodes a regulator of cAMP signalling that is targeted by the antidepressant/antipsychotic rolipram and the PDE4B gene is directly disrupted by a t(1;16) translocation in two cousins diagnosed with schizophrenia or a psychotic disorder^{13, 30, 32}.

What have we learnt so far from sequencing DISC1? Perhaps more than Sullivan suggests¹. An early sequencing study of DISC1 identified six ultra-rare non-synonymous amino-acid substitutions in 288 schizophrenia cases but not 288 matched controls³³. Moens et al³⁴ reported an increased burden of rare missense mutations in a Swedish cohort. Green et al³⁵ reported an excess of exon 11 rare variants in schizotypy. Crowley et al³⁶ found nothing of statistical significance on sequencing 2.7kb of DISC1 in 727 schizophrenia cases and 733 controls. In a recently published study, we sequenced the full 528kb of the genomic DISC1 locus in 1,562 cases and controls³⁷. We found no significant increase in the burden of non-synonymous coding variants in cases, but did find evidence for gene-wide enrichment of conserved, regulatory variants in major depressive disorder. We also reported the ultra-rare variant R37W in a case of major depressive disorder, transmitted to two affected offspring, one of the ultra-rare variants first reported by Song et al³³ in a case of schizophrenia. We now know that R37W alters the cellular distribution of DISC1, the modulatory role of DISC1 in cellular stress³⁸ and also mitochondrial trafficking³⁹.

Sullivan’s dictum that ‘biology does not have a role in establishing a genetic association (but only later in understanding its role)’ is not one that we or a large section of the genetics community would agree with, but ironically it does apply to the case for the Scottish t(1;11) family: only once the breakpoint was cloned and DISC1 identified could the biology begin⁴. It is worth remembering that in 2000, DISC1 was, and still is, a singleton protein with no close homology to any other and thus no neat and meaningful descriptor. Naming the protein in the context of its discovery (genome wide significance for schizophrenia in the t(1;11) family) made sense then and no change is warranted now.

There is no harm and much to gain from reasoned and vigorous debate over alternative approaches to cracking a scientific problem as pressing and challenging as understanding the genetic and biological basis of major mental illness. Indeed, we should take as clear evidence of progress that the relative merits of case, family and population approaches can be debated on the basis of evidence. The value of a biological pathway based approach is very well established in other fields: in our view, shared by others^{31, 40, 41}, it is critical to the advancement of psychiatric research. DISC1 has already made a valuable contribution to this endeavour. We believe strongly that DISC1 and the DISC1 pathway paradigm have served the field well, will continue to do so and that the foundational evidence is sound. We will continue to take this systems approach as the one most likely to provide actionable biological insight.