Schizophrenia Research Forum - A Catalyst for Creative Thinking

Schizophrenia Genetics 1: Linkage Studies—What’s Under Your Peak?

In SRF's schizophrenia genetics overview, writer Pat McCaffrey surveys the range of experimentation and opinion in the field in a five-part series.

See Part 2, Linkage; Part 3, CNVs, Part 4, Bigger Genetics, Part 5, From Genes to Biology…and Therapies. Read a PDF of the entire series.

Editor's note: From the large-scale collaboration of the Psychiatric Genomewide Association Study Consortium, to the very public debate about priorities and funding, schizophrenia researchers had a lot to talk and argue about in 2009. To quote a popular metaphor in this field, some describe the genetics glass as half empty, a vessel of as-yet unfulfilled promise, while others see the cup running over with new discoveries and possibilities. Undaunted by the roiling waters, SRF writer Pat McCaffrey dove in and asked a handful of prominent researchers such questions as, What is going on in schizophrenia genetics? What have we learned about the genetic landscape and the genomic substrates of schizophrenia? What paths are available to researchers, and what is the best way to proceed?

What follows is the first in a series of five articles that will air a spectrum of views (see also the interview with Lars Bertram on the new meta-analysis results from SZGene). Our molecular genetics story starts in the 1980s, with the early linkage studies that produced the first flock of gene candidates. Then, it moves on to the controversial common disease/common variant hypothesis and the large genomewide association studies (GWAS) it inspired, followed by a look at the hunt for rare mutations in schizophrenia. Finally, we gaze ahead at the next steps and ask how the accumulated wisdom is contributing to the search for new treatments. As always, we hope these articles will stimulate comments or corrections (not to mention commendations for Pat!) from our readers.

12 March 2010. The hunt for the genetic basis of schizophrenia started well before the Human Genome Project or the International HapMap. Buoyed by the evidence of inherited risk and the hope of identifying single-gene Mendelian forms of the disease, in the late 1980s researchers began to deploy new molecular genetic techniques in linkage studies to identify large chunks of DNA that segregated with schizophrenia in affected families. The first crop of candidate genes from beneath the linkage peaks—dysbindin (DTNBP1) at chromosome 6p, neuregulin (NRG1) at 8p, and the D-amino acid oxidase activator (DAOA, formerly G72/G30) at 13q—were all identified in 2002 by linkage scans in relatively small family samples, followed by linkage disequilibrium mapping (Straub et al., 2002; Stefansson et al., 2002; Chumakov et al., 2002). Many subsequent studies bore out these linkages, and meta-analyses in the SchizophreniaGene (SZGene) database continued to support the role of these genes when this story went to press (see the most up-to-date results of SchizophreniaGene meta-analyses, revised 29 January 2010 to include criteria that assess the quality of the data; see interview with Lars Bertram on SZGene).

In addition to linkages, the candidate gene approach yielded some early successes by focusing studies on genes with a biologically plausible link to schizophrenia, such as those encoding dopamine system molecules. These early linkage and candidate gene finds are now among the best-supported candidate genes for schizophrenia, though they are not universally accepted.

A promising start
At the time, debates sometimes grew contentious about specific candidate genes, but the linkage studies proved that schizophrenia geneticists could collaborate for the greater good. "As a group, we talked very intensely at meetings, we collaborated well prior to publication, and we fed each other our best markers and regions," said Richard Straub of the National Institute of Mental Health, Bethesda, Maryland, one of the first researchers to do linkage studies in schizophrenia. This spirit of cooperation culminated in a meta-analysis of 20 different studies, which identified genomewide significant linkage at 2q, and an additional nine locations that the researchers agreed were strong candidate gene regions (Lewis et al., 2003).

In retrospect, linkage studies in pools of unrelated families should not have worked, if indeed schizophrenia springs from many genes of small effect size or from rare variants in many different genes (as is discussed in Part 2 and Part 3 of this series). Small effect sizes and genetic heterogeneity make it hard to separate linkage peaks from noise. The definition of schizophrenia presented another difficulty. A clinical syndrome with variable manifestations, schizophrenia has no biological diagnostic criteria. Is it one disease or many? A problematic phenotype just added to the uncertainty around the search for genes responsible for the disorder.

Nonetheless, linkage studies turned out to be surprisingly good at finding evidence of genes of small effect size, says Straub, probably because some peaks span multiple susceptibility genes that sum up to a stronger signal. That is the advantage of linkage over association studies, because a genomic region can “get credit” for the disease in different families, even if the actual genes involved are different. In addition, a signal can sum up the effects of different alleles of the same gene, said Straub. In linkage, common variation and rare changes all contribute to the linkage signal, so that despite the small size and low resolution of the early studies, gene candidates did emerge as researchers looked for biologically plausible genes under the peaks.

The early linkage studies were “shockingly successful,” said Straub, and really opened up the field. “Having one gene allowed nomination of other candidate genes,” and in that way neuregulin led to ErbB4, and dysbindin set researchers onto other potential genes. For instance, Straub and colleagues found that dysbindin is part of an eight-protein macromolecular complex that is involved in vesicle trafficking and dopamine receptor stability (the BLOC-1 complex; see SRF related news story). That finding led them to another candidate gene. “We did a few SNPs in the gene for MUTED, one of the three known dysbindin binding partners at the time," said Straub. "We got a beautiful signal in the 3’ UTR that may involve a microRNA binding site.”

Two studies from other labs on MUTED have not been positive (see SZGene entry for MUTED), but Straub stands by his analysis. He argues that in a genetically heterogenous disease like schizophrenia, replications in different samples should not be the gold standard and are, in fact, statistically unlikely. “It is so much harder for competent investigators to generate a false positive than a false negative in the detection of genes in complex disorders that negative studies should never be weighted as strongly as positive,” he said. In fact, he added, “I’ve maintained for many years that a negative really means nothing.” Instead, researchers need to look at the cumulative evidence for a gene’s involvement, be it genetic or biological.

Not everyone agrees on the strength of the evidence for genes implicated by linkage studies. In his recent talk at the 2009 World Congress on Psychiatric Genetics, Michael Owen of Cardiff University in the United Kingdom said that while some of the genes (NRG1, DTNBP1, and DAOA) had strong support, there were still some holes in the data. For one, there were no clear risk alleles identified for the genes. “These findings are better than we had previously, but they fall short of the strength and consistency one would hope to see and that have been seen in other diseases,” he said.

Biology: help or hindrance?
Owen also spoke critically of the "contamination” of genetics research by biology. “Some say that geneticists are obsessed by statistics, and what you need is biology. If you are looking at biologically plausible things, you don’t need huge samples, and you should focus on a biology-based interpretation of the data, they argue. That’s a view I happen to profoundly disagree with,” he said, because researchers do not know what biology underpins schizophrenia.

Straub thinks the weight of the evidence justifies a stronger conclusion. “The chances of me analyzing eight proteins related to dysbindin and then getting a beautiful signal out of MUTED, if dysbindin wasn’t a schizophrenia gene, is basically zero,“ he says. “It turns out that within two protein binding partners of dysbindin are five schizophrenia genes, and that immediately validates dysbindin."

Kenneth Kendler, of Virginia Commonwealth University in Richmond, was also involved in the early linkage studies, and he, like Owen, is reserving judgment on whether neuregulin and dysbindin are susceptibility genes. He is also skeptical about the candidate gene approach, which has nominated the dopamine receptor genes and other genes in the dopamine pathway. “The candidate gene approach is basically a Bayesian approach where researchers take prior knowledge about genes that are involved to up the chances that P values produced will be true and not false positives,” Kendler said.

With schizophrenia, however, that approach makes little sense to him, “because we don’t have a clue what causes the illness.” He noted, for example, that while drugs targeting the dopamine system produce a treatment response, the evidence that dopamine is involved in pathophysiology is not strong.

Though the debate continues on the results of linkage and candidate gene studies, the work done in this arena has left its imprint on the field. The prospect of causative genes drew neurobiologists into schizophrenia research when they could see, for the first time, a path to molecular and cellular studies. And because of the uncertainty in the genetic studies, geneticists needed biological support to complement their own work. “These discoveries ushered in a new era in schizophrenia research,” said Straub. “Because they were not definitive, they encouraged scientists to assemble other evidence—both genetic and biological—to support genetic evidence that on its own might not be compelling.” That situation led to an influx of physiologists and biologists into the field, pursuing studies aimed at building a biological rationale for a candidate gene, he said.

Wherever studies of the early candidate genes lead in the future, linkage studies kicked off the transformation of the field of psychiatric research from clinical studies and epidemiology to biology. “Genes are the first objective signs to point us to the cause of these illnesses,” said Daniel Weinberger, also of the National Institute of Mental Health. “These are the first clues to the biological mechanisms of causation, which we never had before. That’s the sea change we have seen.”—Pat McCaffrey.

See Part 2, GWAS; Part 3, CNVs, Part 4, Bigger Genetics, Part 5, From Genes to Biology…and Therapies. Read a PDF of the entire series.

Comments on Related News

Related News: Studies Suggest Potential Roles for Dysbindin in Schizophrenia

Comment by:  Philip Seeman (Disclosure)
Submitted 29 November 2007
Posted 29 November 2007
  I recommend the Primary Papers

The publication by Iizuka and colleagues is an important advance toward unraveling the basic biology of psychosis in general, and schizophrenia in particular. This is because they have found that a pathway known to be genetically associated with schizophrenia can alter the surface expression of dopamine D2 receptors. D2 continues to be the main target for all antipsychotic drugs (including aripiprazole and even the new Lilly glutamate agonists that have a potent affinity for D2High receptors).

In fact, the authors of this excellent study may do well to go one step further by testing whether the downregulation of dysbindin actually increases the proportion of D2 receptors that are in the high-affinity state, namely D2High. This is because all schizophrenia animal models markedly increase the proportion of D2High receptors by 100 to 900 percent (Seeman et al., 2005; Seeman et al., 2006). This generalization holds for animal models based on brain lesions, sensitization by amphetamine, phencyclidine, cocaine, caffeine or corticosterone, birth injury, social isolation, and more than 15 gene deletions in pathways for glutamate (NMDA), dopamine, GABA, acetylcholine, and norepinephrine. Although the proportion of D2High receptors invariably increases markedly, the total number of D2 receptors is generally unchanged, slightly reduced, or modestly elevated.

This publication for the first time bridges the hitherto wide gap between genetics and the antipsychotic targeting of the main cause of psychotic signs and symptoms, which is excessive D2 activity, presumably that of D2High, the functional component of D2.


Seeman P, Weinshenker D, Quirion R, Srivastava LK, Bhardwaj SK, Grandy DK, Premont RT, Sotnikova TD, Boksa P, El-Ghundi M, O'dowd BF, George SR, Perreault ML, Männistö PT, Robinson S, Palmiter RD, Tallerico T. Dopamine supersensitivity correlates with D2High states, implying many paths to psychosis. Proc Natl Acad Sci U S A. 2005 Mar 1;102(9):3513-8. Epub 2005 Feb 16. Abstract

Seeman P, Schwarz J, Chen JF, Szechtman H, Perreault M, McKnight GS, Roder JC, Quirion R, Boksa P, Srivastava LK, Yanai K, Weinshenker D, Sumiyoshi T. Psychosis pathways converge via D2high dopamine receptors. Synapse. 2006 Sep 15;60(4):319-46. Review. Abstract

View all comments by Philip Seeman

Related News: Studies Suggest Potential Roles for Dysbindin in Schizophrenia

Comment by:  Christoph Kellendonk
Submitted 4 December 2007
Posted 4 December 2007

The study by Iizuka and colleagues is indeed very interesting. It suggests that one of the most promising risk genes for schizophrenia, the dysbindin gene, may functionally interact with dopamine D2 receptors. The D2 receptor itself is an old candidate in the study of schizophrenia, mostly because until very recently all antipsychotic medication had been directed against D2 receptors. But in addition, PET imaging studies have shown that the density and occupancy of D2 receptors is increased in drug-free and drug-naïve patients with schizophrenia.

How could this increase arise? In a subpopulation of patients it may be due to a polymorphism in the D2 receptor gene, the C957T polymorphism. The C-allele increases mRNA stability and has been found to be associated with schizophrenia, though obviously not all patients carry the C-allele. Iizuka and colleagues found an independent way in which the genetic risk factor dysbindin may upregulate D2 receptor signaling. Because dysbindin is downregulated in the brains of patients with schizophrenia, they used siRNA technology to study the molecular consequences of decreased dysbindin levels in cell culture.

They found that downregulation of dysbindin increases D2 receptor density in the outer cell membrane, suppresses dopamine-induced D2 receptor internalization, and increases D2 receptor signaling. The study is very promising but requires further confirmation.

How specific are the observed effects for D2 receptors? Because dysbindin is involved in both membrane trafficking and degradation of synaptic vesicles, knocking down dysbindin in growing cells may affect many physiological processes, one of them being D2 receptor signaling. Does quinpirole treatment differentially affect GTPgS incorporation in siRNA and control cells? This would be a more immediate way of looking at D2 signaling than measuring CREB phosphorylation. And, of course, the most important question is, What will happen in vivo? Maybe the sandy mouse, which carries a deletion in the dysbindin gene, could be of help here. Using these mice for a similar line of experiments may answer this question.

Iizuka and colleagues found an exciting new functional interaction between two major molecules involved in schizophrenia. I believe that these are the kind of interactions we have to look for if we want to understand complex genetic disorders such as schizophrenia.

View all comments by Christoph Kellendonk

Related News: Policy Brief: A $2 Billion Proposal for Research in Neuropsychiatric Disease

Comment by:  Michael Owen, SRF Advisor
Submitted 30 March 2010
Posted 30 March 2010

By and large, I agree that this is how the big picture looks.

Regarding the genetics: I disagree that we now know that “private” mutations are responsible for disease in many cases. This is conjecture, since no pathogenic point mutations have yet been identified. Moreover, a number of the rare pathogenic variants that have been identified to date (CNVs) are not private, and I see no reason to suppose that the situation will be different for point mutations. However, I very much agree that the focus is going to have to be on very large samples to get full traction on the genetics. Some people think that as we move to sequencing over the next few years, the focus should shift away from very large samples to studies of families. However, given the findings from linkage, where single families with cast iron linkages are extremely rare, and the huge statistical challenges to identifying association with individually rare variants, the samples required will be just as large as those that will be required to identify common risk variants by GWAS, i.e., tens of thousands. So the focus now needs to be on amassing the sample base for these large studies. As we wait for the cost of sequencing to come down over the next few years, it will be possible to subject these samples to GWAS with the current SNP chips (the costs of these are falling); this will identify more common risk variants. And also bear in mind that as we get data from the 1000 Genomes Project and other sequencing studies, it will be increasingly possible to impute a greater proportion of the rare variants from the SNP-chip data. This is not to say that studies of smaller, well-characterized samples and families will not be useful; by delivering greater homogeneity (phenotypic and genetic, respectively), they might well be fruitful, but even if robust findings can be obtained in this way, these studies will only reveal a very small proportion of the picture. There will still be a need to test the generality of the findings in larger, more representative cohorts.

I also agree that the role of computational biology is likely to be critical in interpreting genetic findings. This will be required to distinguish true risk mutations from the many others that will be neutral with regard to the phenotype under study, and to interpret the data in individual cases where several rare mutations and many common variants are likely to be playing a role.

Regarding the circuit analysis: There seems to be a bit more hand waving here. However, I do think that we need systematic efforts to define the relationships between brain phenotypes and psychopathology. One point not mentioned in the article is that this will need to be directed against an understanding of psychopathology at the level of symptoms and syndromes and not diagnoses. Most of the studies to date are small scale, and use a variety of different measures, and they tend to focus on brain phenotypes in very simplistic ways. We need to know which of these relate to which aspects of the rich psychopathology seen in patients and which are endpoints in themselves (i.e., they index the same risk factors as the psychiatric phenotypes but do not mediate risk). Defining causal pathways in humans is difficult and requires sophisticated statistical approaches and large samples, but it is possible. However, there will also be a need to complement the human work on neural circuits with studies of experimental animals where mechanistic insights are easier to glean. Here what is needed is work aimed at establishing correspondence across species and being able to move programs of work from animal to humans and vice versa.

Finally, I agree that the integration of genetics with neuroscience is going to be key to understanding psychiatric illness. A genetic risk variant that is robustly associated with a disorder allows studies of neural circuits to be firmly anchored to the etiology of the illness, though I stress again the need to be cautious about inferring that specific brain phenotypes mediate disease risk.

View all comments by Michael Owen

Related News: Schizophrenia Genetics 2: The Rise of GWAS

Comment by:  Chris Carter
Submitted 7 April 2010
Posted 8 April 2010

I wonder whether the relative lack of success in schizophrenia GWAS may be because the origin of schizophrenia may lie not so much in the genetic make-up of people with schizophrenia themselves, but in their prenatal experience, and possibly with the genes of the mother rather than with those of the offspring. Famine, rubella, influenza, herpes (HSV1 and HSV2), and poliovirus infection as well as high fever during pregnancy have all been listed as risk factors for the offspring developing schizophrenia in later life, as have maternal preeclampsia and obstetric complications. (See page at Polygenic Pathways for the many references.)

Maternal resistance to these effects is likely to be gene-dependent. Is it worth considering GWAS in the mothers rather than in the offspring?

View all comments by Chris Carter