Email Icon Facebook icon Twitter Icon GooglePlus Icon Contact

User Top Menu

Schizophrenia Genetics 2015—Part 3, Rare Allure

16 Aug 2015

In SRF's five-part 2015 schizophrenia genetics update, reporter Michele Solis surveys leaders in the field about milestones, challenges, and current research.

See Part 1, Renaissance; Part 2, From Discovery to Understanding; Part 4, Rethinking Diagnoses; and Part 5, Plan of Action.

Download a PDF of the entire series.


August 17, 2015. Beyond the common genetic variants that all people carry, rare variants are also expected to contribute to schizophrenia risk. To geneticists, rare means occurring in fewer than one out of 100 people, and finding them requires reading out genomes letter by DNA letter in thousands of people. But this fine-combing may be rewarded by findings that are simpler to interpret.

"Rare contributions are incontrovertible because we have the mutations in our hands," said David Goldstein of Columbia University in New York City, who leads several sequencing projects for different human diseases.

Unlike SNPs in genomewide association studies (GWAS), which only flag a region that contains a risk variant, rare variants themselves are the nucleotides of interest. They are also more likely to have a large effect on risk, given that their rarity reflects a relatively recent change to the human genome, with little time for suppression by natural selection.

Yet so far, the dividends remain elusive. Rare variants are indeed found in people with schizophrenia, but they are rarer than expected and scattered across the genome, only occasionally hitting the same gene in different people. This, combined with an unexpected amount of rare variation in everyone, has made it hard to statistically peg any one genetic mishap to schizophrenia.

"The large sequencing studies published to date have relatively modest results—not so dissimilar from the results in the first rounds of GWAS," said Mark Daly of the Broad Institute in Cambridge, Massachusetts.

This leaves rare variant hunters with the same mantra as their GWAS counterparts—increase sample size.

"I think there's just a heck of a lot of loci for schizophrenia," Goldstein said. "So if a lot of different genes can confer risk, you need really large sample sizes in order to implicate any of them."

As the search for rare variants intensifies, more traction has come from contributions made by another type of genetic rarity called copy number variants (CNVs). CNVs involve the loss or gain of DNA segments containing multiple genes and elevate risk for diverse psychiatric disorders (Malhotra and Sebat, 2012).

Yet even these potent rare variants, CNVs or otherwise, may not lead exclusively to schizophrenia. Instead, certain combinations of factors—genetic and environmental—could put someone on the path for a specific disorder.

"Risk for a disorder is rare variant plus common variant plus environment, and under environment you can put in chance," said Michael O'Donovan of Cardiff University in Wales.

Sequencing sense

Sequencing data for schizophrenia has begun to surge in the past five years, with results reported from several groups. So far, all studies have restricted themselves to the exome, the protein-coding parts that comprise 1 percent of the genome. Exome mutations may damage the protein building blocks of a cell and lead directly to insights about the biology of schizophrenia.

But even for these, interpretation can get tricky. We all carry rare, probably protein-damaging, mutations, with no obvious untoward consequences. This complicates making a connection between a certain rare variant and disease. For this reason, studies have focused on "de novo" mutations—non-inherited mutations that spontaneously occur in sperm or egg cells. If found in people with schizophrenia, but not in their unaffected parents, then chances would seem to be better that they have something to do with the illness. Plus, de novo mutations are more likely to include damaging variants because natural selection hasn't had a chance to temper their actions.

This approach has successfully identified risk variants for autism (see SRF related news report), but the jury is still out on whether they contribute similarly to a disorder like schizophrenia, which typically starts during late adolescence and early adulthood. De novo variants do turn up in schizophrenia, often at a rate higher than in controls, but these are distributed across the genome. Two studies published last year hint that some of these might pile up on the same genes: one reported deleterious mutations in SET1DA in two different people (see SRF related news report), and another reported 18 different genes hit twice by de novo mutations, including TAF13 (see SRF related news report). Both SET1DA and TAF13 have roles in transcription, but more mutations hitting them need to be found to turn them into statistically definitive risk factors.

Another approach forgoes de novo finding and instead sequences lots of cases and controls, with the idea that a true risk variant would turn up more often in schizophrenia. The largest effort of this kind, led by Shaun Purcell of Mount Sinai School of Medicine in New York City, sequenced nearly 5,000 people, yet found only a small enrichment of rare variants in schizophrenia, and no pileups of mutations hitting the same gene (see SRF related news report).

At the very least, sequencing efforts have been large enough to rule out contributions by moderately rare mutations, occurring in 0.5 to 1 percent of people. Thought to inhabit a genetic sweet spot in that they are not so rare as to be hard to find, yet not so common as to have weak effects, these "medium rares" are not enriched in schizophrenia, according to the Purcell paper and an earlier paper published by Goldstein's group (see SRF related news report).

Yet the ragtag crew of rare variants found so far in schizophrenia may act on some of the same biological processes, including early brain development (see SRF related news report), chromatin modification (see SRF related news report and SRF news report), or synaptic machinery (see SRF related news report). Researchers also point to the overlap with GWAS-implicated genes, including those involved in glutamate and calcium signaling.

"I think the main message from the rare variant stuff is that, while we've seen convergence on certain biological pathways, identifying individual genes in significant numbers, much less individual mutations, will require much larger samples than currently published," O'Donovan said. "And that's an expensive endeavor. It will take a bit of time."

Sequencing families enriched for mental illness might expedite the discovery of repeat instances of the same mutation.

"I think that the degrees of heterogeneity we now all agree exist tell you that you've got to do something differently. Included in that would be a return to family-based studies," said David Porteous of University of Edinburgh in Scotland. Porteous and colleagues discovered the gene disrupted-in-schizophrenia-1 (DISC1) in a large Scottish family beset by mental illness, including schizophrenia (see SRF related news report).

Targeted sequencing of DISC1 is now underway to probe its role in mental illness beyond the original family. This has revealed a huge amount of variation in the gene. So far, these variants have been tied more to major depressive disorder than to schizophrenia (Thomson et al., 2014).

Though DISC1's status as a schizophrenia gene is contentious (Sullivan, 2013; Porteous et al., 2014), it remains a favorite among biologists because many are convinced that it could reveal biological mechanisms that are subverted in psychiatric disorders beyond the Scottish family. Last year, a high-profile paper reported that stem cell-derived neurons from two people with DISC1 mutations and psychiatric disorders had weakened synapses; this deficit could be rescued in vitro by correcting the DISC1 mutation through genome editing (see SRF related news report).

"There's little doubt in my mind that DISC1 is a real risk gene that has probably taught us more about the biology of schizophrenia than anything we've found in GWAS so far," said Francis McMahon of the National Institute of Mental Health (NIMH) in Bethesda, Maryland.

Despite the murky picture of rare variant contributions to schizophrenia, researchers are already bringing exome sequencing to the clinic. For example, Anna Need of Imperial College London is sequencing the exomes of children with a variety of psychiatric illnesses, including schizophrenia, in order to find some genetic explanation for their disorders. In some cases, the children carry mutations in genes associated with a Mendelian disorder that doesn't match their symptoms. This kind of genetic diagnosis could inform treatment plans and help chart the connections between a disrupted gene and its resulting phenotypes.

"We thought we had a reasonable idea of what the genes causing known Mendelian disorders did. But they're actually turning out to be linked to a much more heterogeneous spectrum of conditions and symptoms," said Need, who co-authored a recent survey of the diverse outcomes that can stem from disruptions to the same gene (Zhu et al., 2014).

We are the 99 percent

The bulk of the risk conferred by rare variants may well reside in non-coding regions, which comprise 99 percent of the genome. Whole genome sequencing (WGS) to probe these stretches of DNA has begun for schizophrenia, but researchers know that they are unprepared for interpreting the onslaught of data.

"Going out and trying to find the risk factors anywhere in the genome for schizophrenia now is an intimidating proposition, since our ability to interpret mutations outside of the exome is currently limited," Goldstein said. "Our hope is to help build up the science of interpreting regulatory variation by integrating whole genome sequence data with comprehensive transcriptomic data."

The Whole Genome Sequencing for Psychiatric Disorders (WGSPD) consortium has been formed among those collecting WGS data for schizophrenia and other disorders, with one aim being to come up with interpretation aids. Goldstein, Daly, and Matthew State of the University of California, San Francisco, are building a framework for recognizing the most important non-coding variants through analysis of transcriptome data from blood cells and neurons derived from stem cells. For example, a variant of interest could be identified when changes in how a gene is spliced or expressed occur in tandem with a variant in a splice site or regulatory region for that gene.

CNV signposts

In contrast to the incipient returns from sequencing, CNVs are more established as risk factors. Although rare, CNVs in certain locations of the genome turn up in multiple cases of schizophrenia, and significantly less often in controls. For example, 25 percent of people carrying a large 3 Mb deletion on chromosome 22—a CNV first identified in 1995—will develop psychosis. Since then, new technologies, including SNP arrays, have detected other CNVs in either case-control or de novo studies (see SRF related news report). CNVs are found in a minority of schizophrenia cases, but they attract interest because they are associated with big effects on risk, with odds ratios about 10 and beyond.

The Psychiatric Genomics Consortium (PGC) has combined some of the array data used for its GWAS efforts to look for schizophrenia-associated CNVs. Ongoing analyses of 41,000 people suggest that researchers have, for the most part, found the big ones.

"The CNVs we know and love—they stand out like a sore thumb," said Jonathan Sebat of the University of California, San Diego, who co-leads the analysis. "But there are some additional genetic factors in there that we haven't yet identified."

But even with such strong associations, CNVs evade straightforward interpretation. Most involve numerous genes, making it hard to pin down exactly which contribute to schizophrenia. Analyses of the long list of potential culprits have at least highlighted post-synaptic machinery (see SRF related news report).

"Knowing that the synapse is involved in schizophrenia—that's not really earth-shattering news," Sebat said. "But I think it's more interesting to see that we're now starting to get down to components of the synapse that appear to be driving that signal."

Another wrinkle is that CNVs are not terribly specific. For example, a duplication at chromosome 16p11.2 is associated with increased risk for schizophrenia, bipolar disorder, and intellectual disability. Interestingly, other CNVs associated with psychiatric disorders can be carried with apparently no ill effects, though closer examination has revealed subtle effects on cognition and brain structure (see SRF related news report). This suggests that CNVs must work with other factors to produce a particular phenotype.

Psychosis may be an extreme outcome of a CNV, Sebat suggested.

"The typical presentation of the CNV may not be schizophrenia but may be the unfortunate kid in the back of the classroom who is struggling with reading comprehension and having difficulties in mathematics, and maybe not socializing to the extent the other kids are socializing. Whether that kid goes on to develop psychosis later in life involves a number of factors, both nature and nurture," he said.

Such non-specificity seems to be a rule even for rare point mutations and common variants. Time and again, genetics declines to follow the diagnostic categories set forth by psychiatrists. There is much speculation about what else is going on to lead to specific mental disorder phenotypes. We will explore this in story 4 of the series, "Rethinking diagnoses."—Michele Solis.

See Part 4, Rethinking Diagnoses.