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 1, Linkage; Part 2, GWAS, Part 4, Bigger Genetics, Part 5, From Genes to Biology…and Therapies. Read a PDF of the entire series.
19 March 2010. Automated genotyping on single-nucleotide polymorphism (SNP) chips provided researchers with an unexpected view of genetic variation when they began to notice that the signal for some of these SNPs was less intense than expected, or in some cases stronger. The reason, they found, was that the genome is peppered with submicroscopic deletions and duplications that have the potential to disrupt genes or add extra copies. Too small to be detected by the traditional cytogenetic techniques that led to DISC1 and other gene candidates (see SRF related news story), these copy number variants (CNVs) were just recently recognized as a major source of genetic variation among people (see SRF related news story), and have been strongly linked to autism (see SRF related news story).
In the last two years, data have accumulated quickly, suggesting an important role for CNVs in schizophrenia: people with schizophrenia are reported to have elevated incidence of CNVs (see SRF related news story), including CNVs that are not inherited but have arisen in individuals (de novo CNVs; see SRF related news story). Last year, genomewide association studies found a handful of recurrent CNVs that were linked to schizophrenia in large samples (see SRF related news story). The CNVs in those studies showed effect sizes that far outstrip the modest contributions of common variants.
Nevertheless, the operative word is “rare”—so far, CNVs that appear to be causative have only been found in a fraction of a percent of schizophrenia cases. Still, they have grabbed attention as possible windows into pathophysiology and as a potential explanation for the missing portion of the genetic risk for schizophrenia that could not be explained by common variants.
The involvement of CNVs in schizophrenia is, in fact, old news: gross genomic changes were associated with the disease nearly two decades ago, when David St. Clair and colleagues at the University of Edinburgh, Scotland, found a balanced translocation between chromosomes 1 and 11 that invariably led to schizophrenia and other mental disorders in a Scottish family (St. Clair et al., 1990). David Porteous’s group at the University of Edinburgh subsequently identified a novel gene, disrupted-in-schizophrenia 1 (DISC1), at the breakpoint (Millar et al., 2000), and DISC1 is now the most researched gene in the field (e.g., see SRF meeting report from Neuroscience 2009).
Also, schizophrenia occurs frequently in velo-cardio-facial syndrome (VCFS, also called DiGeorge syndrome), which results from deletion of multiple genes at 22q11.2, a region implicated by several early schizophrenia linkage scans. VCFS can include cleft palate, heart defects, characteristic facial appearance, learning disorders, and speech and feeding problems. Children with VCFS have a 25-times increased risk of schizophrenia (see SRF related news story), and about one-third of all babies born with the deletion will go on to develop schizophrenia (see, e.g., Murphy et al., 1999 ).
The DISC1 translocation and the 22q11.2 deletion were both found by classic cytogenetics techniques, and Porteous's group has since found other genes using the same methods. DISC1 interacts with PDE4B, which his lab identified as disrupted by translocation in another individual with schizophrenia and a cousin with psychosis. Using cytogenetics, Porteous and colleagues also linked schizophrenia with disruptions in the genes for NPAS3 (a brain-enriched transcription factor) and for GRIK4 (a glutamate receptor gene; see SRF related news story), two additional genes that are known to have important roles in neurogenesis and in neurosignaling. “So using cytogenetics, our lab alone has identified four clear-cut gene hits,” said Porteous. That number is now up to five, with their recent findings on the ABCA13 gene (see SRF related news story).
But are those genes relevant to the wider population of people with schizophrenia who do not carry the rare mistakes? Porteous argues that they are. “Each one of those has then been confirmed by doing a genewide association analysis looking in the general population of schizophrenia and bipolar disorder," he said. "We asked, if we look specifically at SNPs in the DISC1 locus or the PD4B locus or NPAS3 or GRIK 4, do we find evidence of association in the general population? The answer in each and every case has been, yes.”
The bottom line, Porteous said, is “We’ve got some real candidates, and genes that are beyond candidates. Certainly for DISC1, there is no question for most people who know the field well that this gene is causally related to psychosis, and there’s a wealth of information now about the biology that backs up that supposition.” Recent work in a mouse model of DISC1 deletion has revealed the gene’s role in neurogenesis (see, e.g., SRF related news story), cortical development (see, e.g., SRF related news story), and in biochemical pathways related to depression (see SRF related news story). In addition, it has pointed to the kinase GSK3, downstream from DISC1, as a candidate therapeutic target for depression and schizophrenia.
…or a whole flock?
The advent of SNP chips opened a window on genetic variation that was hidden to cytogenetics. Using the raw signal intensity data from the chips, researchers could detect submicroscopic insertions and deletions. Genomewide analysis of the CNVs identified this way, or with dedicated CNV chips, in schizophrenia patients led to some key insights. First, several studies consistently found a higher load of copy number variation in individuals with schizophrenia compared to unaffected individuals (see SRF related news story on Walsh et al., 2008). The excess was seen not only in familial cases, but also in de novo CNVs in sporadic disease (see SRF related news story). The increase was small, but appeared specific to schizophrenia compared to bipolar disorder. In total, the data support the idea that rare mutations, either inherited or newly arisen, have a role in the etiology of schizophrenia.
Second, while CNVs are in aggregate more common in people with schizophrenia, individually they are rare and occur all over the genome, consistent with the apparent genetic heterogeneity of the disease. Some larger deletions appear more frequently, and two studies looking at genomewide association of CNVs with schizophrenia came up with recurrent deletions in four regions that were statistically associated with the disease: 22q11.2, (the VCFS deletion region), 1q21.1 (previously linked to schizophrenia), 15q13.3, and 15q11.2 (see SRF related news story).
Finally, genomewide analysis of CNVs suggests that some rare mutations confer a high risk for schizophrenia. Compared to common variants, which carry odds ratios between 1.08 and 1.5, the effect sizes of these recurrent CNVs range from 5 to 20.
Jonathan Sebat, University of California, San Diego, sees these insights as fundamental lessons, but maybe not so surprising. “In hindsight, when we look back on it, we might say we’ve rediscovered something we already knew. When you introduce a severe mutation in a gene involved in brain development, you produce a severe cognitive phenotype. Once you have technologies that allow you to find these severe mutations, you have really impressive power to learn something about the biology of this disorder.”
Chasing down biology might mean going back to patients who have rare deletions to look for more defined phenotypes or endophenotypes (for example, see SRF related news story on the 1q21.1 deletion and head circumference and SRF related news story). Or, researchers might create animal models, like the mouse models of VCFS (see SRF related news story) or DISC1 knockdown (Kim et al., 2009, and see SRF related news story).
“This is something that you can’t do with a common risk allele that has an odds ratio of 1.1 and is present on 60 percent of chromosomes, Sebat said. “I’m not sure that it makes sense to study the biology of that genetic variant in patients. You might as well go around the lab and collect samples, because it’s the most common polymorphism in the population, and its effect on disease risk in an individual is small.” On the other hand, CNV discovery leads to genes and to plausible testable biological hypotheses, Sebat said. “That’s what makes [CNVs] such an excellent approach.”
The best example of a CNV hit in schizophrenia so far is the neurexin gene. Both seminal studies on CNVs in schizophrenia identified single cases of schizophrenia with deletions involving the NRXN1 gene (Kirov et al., 2008 and Walsh et al., 2008, and see SRF related news story). Later, the large SGENE Consortium study found a statistically significant association with deletions in the coding region of the NRXN1 gene and schizophrenia (see SRF related news story).
“I think the field would regard this as still a preliminary finding, but we’ve replicated it and others have replicated it, too, so it will not be preliminary much longer,” said Sebat. “This is a site where you have deletions of a single gene, and they have a very predictable effect on the function of that gene. This gets you into biology very quickly, because the function of this gene has been studied for quite some time now.”
Porteous is looking for that kind of biological follow-up to the other CNV findings. “I think that the onus now is on the CNV practitioners to take some of their findings and seek additional supportive evidence, either to demonstrate that if you do, indeed, increase or decrease the level of expression of one of these genes in a model system, that it has a biological effect consistent with it being causally related to schizophrenia.” he said. “But it may need people other than those who are doing the discovery studies just now, because very often the labs that do the handle-turning, large-scale GWAS and CNV work are not the ones who are best suited to doing this next exercise in biology,” he added.
The neurexin example is the rarest of the rare, where deletions affect a single gene. Most deletions and insertions are not as informative because they span multiple genes. In the case of the 22q11.2 deletion syndrome, no fewer than 27 genes are involved. Although several genes in this region have some positive single-candidate results (see SchizophreniaGene Chromosome 22 overview), and the COMT gene has a positive meta-analysis at the time of this writing, genomewide association studies of the region found no common variants that affect disease risk in any of those genes (International Schizophrenia Consortium, Purcell et al., 2009). Despite a large amount of work aimed at dissecting the biological functions of those genes, and the nomination of multiple candidate genes, the path from genes to schizophrenia remains unclear (see SRF related news story). Recent work only adds to the complexity of the problem, by suggesting that CNVs may affect the expression of genes at a distance from the affected regions (Henrichsen et al., 2009).
Another fundamental insight that has come from CNV research is the shared susceptibility for psychiatric diseases. For example, the studies have complemented recent epidemiological findings of a shared inheritance between schizophrenia and bipolar disorder (see SRF related news story) and helped to provide concrete evidence that genetic risk can span several diseases. Neurexin deletions were originally tied to autism, and the family in which DISC1 was identified has a high incidence of both schizophrenia and unipolar depression. Other common effects of some of the “schizophrenia” CNVs include mental retardation, attention deficit-hyperactivity disorder (ADHD), and epilepsy.
That leads some researchers to question whether some of the common deletions, such as 1q21.1 or 22q11, are truly risk factors for schizophrenia. They argue that the defects could cause an endophenotype common to several diseases, or even a pervasive brain developmental disorder that manifests as any one of a number of conditions depending on environmental or other genetic factors (for more on this topic, see SRF related news story, SRF news story, SRF news story, and SRF news story with related interview with author Evan Eichler, University of Washington, Seattle).
“There are a bunch of these deletion syndromes like VCFS, and it looks like overall maybe as many as 3 to 5 percent of people who carry the diagnosis 'schizophrenia' have one of these diagnosable, definable chromosomal syndromes,” said Daniel Weinberger, National Institute of Mental Health, Bethesda, Maryland. “Are these cases schizophrenia, or are these cases with mild mental deficiency as the fundamental effect of these chromosomal abnormalities, which have accessory symptomatology that we can’t differentiate from schizophrenia?"
As Kenneth Kendler, Virginia Commonwealth University, Richmond, put it, “The least exciting interpretation of the CNVs is that you have broad syndromal cases where something quite large is disrupted in the brain, giving an increased risk of schizophrenia. If you reach your hand into a computer, and grab some wires and yank them out, the thing doesn’t work right, but that’s not very informative. Are deletions that take out 30 or 50 genes at a time biologically significant, or are they just like reaching in and yanking out the wires? I don’t think that’s been decided yet.”
Getting the whole picture
What remains to be seen is how much of the risk of schizophrenia in the population is tied up in rare, or even unique, structural variants. The CNVs discovered to date account for just a few percent of schizophrenia cases, but David St. Clair, now at the University of Aberdeen, Scotland, estimates that that number may increase to 10-20 percent, as more novel loci are discovered (St. Clair, 2008). By some estimates, at least half of the larger CNVs (50-100 kb) in the genome cannot be detected by the SNP technology currently used. And as yet, there are no data on other types of rare variants, such as small insertions and deletions of fewer than 1,000 bases, which are thought to be 10 times more common than CNVs. This situation directly contrasts with that involving common variants, where St. Clair writes that researchers have probably seen what there is to be seen.
A deeper understanding of genetic risk will require a full cataloging of disease-causing alleles, said Markus Nöthen, Bonn University, Germany. “We now have a genomewide picture of common variants and, through the same chip-based technology, have the first glimpse of rare variants that confer higher penetrance. But that is only part of the picture,” he said. “From a genetic point of view, what’s really important will be to see the whole allelic spectrum.”
That spectrum includes yet-to-be-discovered CNVs, but also smaller structural changes, right down to rare SNPs and point mutations. There is only one way to get that big picture, and that is large-scale sequencing of individual genomes, an effort some researchers are calling for (see SRF related news story), and some are already starting. Is the field on the verge of its own genomic revolution? For more on that question, stay tuned for Part 4 of the SRF genetics series.—Pat McCaffrey.
See Part 1, Linkage; Part 2, GWAS, Part 4, Bigger Genetics, Part 5, From Genes to Biology…and Therapies. Read a PDF of the entire series.