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 3, CNVs, Part 4, Bigger Genetics. Read a PDF of the entire series.
25 March 2010. The genetic evidence amassed to date points to the conclusion that schizophrenia arises from a mixture of common variants of small effect and rarer variants of large or small effect. From linkage to candidate studies to genomewide association and copy number variant studies, dozens of suspects have arisen. Names have been named, and the debate will continue over the strength of the evidence for particular genes and how to best catalog the full spectrum of genetic variation in people with schizophrenia (you can follow the scoring for common variants at SZGene). But there are enough genes of interest now that the questions increasingly on researchers’ minds go to the next step: how can that information lead us toward understanding the biology of schizophrenia, and, ultimately, to new therapies?
Endophenotypes: bridging genes and disease
“Genes do not encode for complex behaviors,” said Daniel Weinberger of the National Institute of Mental Health, Bethesda, Maryland. They do not encode for hallucinations or delusions, panic attacks or sadness. “Genes encode for simple molecules in cells. And ultimately, those simple molecules in cells affect the element of brain circuits and systems.”
In fact, Weinberger writes in a recent editorial in Nature, “Finding specific genes for mental illness now seems a pipe dream. A more realistic endeavor for the next 10 years is to look for genes that code for basic cellular and brain functions that modulate our responses to the environment and that come together in particular ways in individuals at increased risk.”
Those basic cellular and brain functions are what Weinberger calls intermediate phenotypes and others commonly refer to as endophenotypes (for more, see SRF Live Discussion led by Irving Gottesman and Mayada Akil). Rather than behavioral readouts, intermediate phenotypes are heritable traits of brain biology that are under direct genetic control. Thus, instead of psychosis, an intermediate phenotype would be activation of brain networks during a memory task. The phenotype can be a biochemical measure, such as dopamine levels, an imaging readout, such as the size of a brain structure or a functional MRI measure. The goal of studying intermediate phenotypes is to allow researchers to define the proximal action of genetic variation on the brain and to eventually understand how that variation contributes to the development of schizophrenia.
Because endophenotypes are allegedly closer to gene function than the disease diagnosis, they should show a stronger association with genotypes. Proposed examples of this notion include the association of variants in dopamine-regulating COMT and MTHFR genes with activation of the dorsolateral prefrontal cortex during memory tasks (e.g., see SRF related news story). In newer findings, a schizophrenia-associated polymorphism in the ZNF804A gene was associated with altered functional connectivity in the brains of healthy people (see SRF related news story and Walter et al., 2010). Those studies found significant associations of the SNP with brain function in a sample of just more than 100 people.
Not everyone agrees with the endophenotype approach. Some argue that the proposed intermediate phenotypes themselves are genetically complex, perhaps no less so than schizophrenia itself. Furthermore, the links between genes and endophenotypes may be hard to discern. For example, Mike Owen makes this critique of the ZNF804A study: “They looked at three tests of cognition, but shouldn't they have looked at more before drawing conclusions? Who knows what other associated endophenotypes they might find?” Finally, there is yet no evidence that places endophenotypes on the causal path to disease. They could simply be indexing disease risk, Owen said.
Weinberger disagrees. “Many studies have shown that cortical cognitive dysfunctions are present before the emergence of the diagnostic symptoms of schizophrenia, suggesting that they are on the path, just as high cholesterol levels are risk factors on the path to cardiovascular disease. The same is true for [velo-cardio-facial syndrome], in which cognitive deficits precede the emergence of psychosis in early adolescence and clearly are linked to the chromosomal hemi-deletion.”
The question for Weinberger is not whether intermediate phenotypes are a worthwhile strategy. “They are a critical strategy,” he said. “The real debate is about how we identify intermediate biological traits that will be maximally informative in understanding the mechanisms of these genes.”
Another question related to the value of intermediate phenotypes is whether genetics may be more strongly correlated with different behavioral or cognitive manifestations of schizophrenia. The disease is not a monolith—patients display a varying mixture of symptoms ranging from psychosis to cognitive symptoms, negative symptoms, and mood disorders. These symptom domains could be another way to break patients into subgroups that may more strongly associate with particular genes. “There are many ways to sub-classify samples using different methods at the clinical level or at the phenotype level to identify much stronger contributions to the genotype in specific individuals,” said Markus Nöthen, Bonn University, Germany.
Those kinds of studies require not only many subjects, but also much clinical detail on each subject. Nöthen pointed out that one aim of the Psychiatric GWAS Consortium is to carry out genomewide association studies in a more detailed way with symptom dimensions. “But the problem is whether all of the samples included in the study have this information available. Probably the common information will be rather crude, for example, age at onset, or so on. Still it might be very important information, and we hope that these will actually lead to new loci,” he said.
“There will be many and much more detailed studies to come,” Nöthen said. “We have been extremely interested in the different phenotype dimension for many years, and Marcella Rietschel at the Institute of Mental Health in Mannheim has set up databases for all of our patients, which incorporate more than 2,000 items per patient on a lifetime basis. We are actually very much looking forward to correlating this with the genes identified, and I think it will be extremely interesting to see if symptom dimensions or symptom clusters are associated with specific genes.”
David Porteous, University of Edinburgh, United Kingdom, likes the idea of drilling down into existing datasets to mine the genetic heterogeneity for clues to biology. While it is clear there is no single gene for schizophrenia or bipolar disorder with a large population effect, it could be that only a very small handful of the long list of possible candidates are important and relevant to any one individual. “That question has not yet been addressed and should be before we move on. Going for ever-bigger numbers of samples just adds more ‘noise’ to the analysis,” he said.
Phenotypic parcellation could also lead to better modeling in animals, Nöthen said. “Symptom dimensions might be much more appropriate actually to model in a mouse than schizophrenia. Establishing genotype correlations is extremely important to guide neurobiology studies using other systems.”
Among other big questions yet to be approached comprehensively are gene-gene interactions (epistasis), gene-environment interactions, and the question of epigenetics, or heritable changes in the genome that are not necessarily reflected in the DNA sequence. All of these may contribute to the risk of schizophrenia, and with the exception of a handful of studies on epistasis and gene-environment interactions, these areas are still largely unexplored.
In any case, gene discovery will lead to new therapies, according to Nicholas Brandon at Pfizer Neuroscience, Princeton, New Jersey. Starting at Merck and then at Wyeth (now Pfizer), Brandon has been a long-time collaborator of Porteous, working on the DISC1 pathway. Brandon points out that DISC1 came with a high level of biological plausibility, because of strong genetic evidence that it segregates with mental illness in an extended family. That helped the field move quickly from genes to biology, and studies of DISC1’s partners, starting with PDE4 (see SRF related news story), and lately GSK3 and others (see SRF related news story and SfN 2009 meeting coverage) have yielded several potential drug targets.
Brandon’s main job is target validation, the assessment of whether or not a particular protein will make a plausible drug target. And that critical analysis is something he sees lacking so far in the wake of the large genomewide association studies and copy number variant discoveries.
“I’m thinking beyond that new genetic hit and the first excitement and asking if something has the legs to carry on through to discovery,” Brandon said. “That’s where I think a lot of discussion has to be had now, how much effort you want to sink into these immediately.” For example, there are issues common to all drugs, such as potential side effects, and others unique to psychiatric illness, such as, Will a target affect positive or negative symptoms?
Compared to the DISC1 experience, Brandon has found the genomewide association results a bit thin so far. “A lot of money has been spent on these studies, and the output so far has been fairly limited,” he said. “I’m sure there will be a lot of meta-analyses of all the datasets, and perhaps we’ll parse out some other important findings. But I have a feeling we’ll end up with maybe one or two new candidates, and mostly we’ll end up with molecules which are part of pathways we’ve already identified using more ’old-school’ techniques. I’d love to be proved wrong, and I’d love to see some of the new findings really open up areas of biology we haven’t thought about before.”
Even though the crop of genes so far has not been overwhelming, Brandon said, they will prove to be a key platform for moving the field forward. One area where genetics will help considerably will be to define populations of patients that may benefit from differential therapy. For example, Brandon hypothesized, some patients may respond best to glutamate-based therapies, others to dopamine-targeting drugs, or even to GSK3-pathway interventions. “I think that’s where the industry’s moving now, trying to classify, at the genetic level, responders, non-responders. I think the current wave of studies, even though they may not be bringing up potentially new targets, can be providing new critical information for their part of the discovery process,” he said.
It all comes down to the details, Brandon stressed. “At the end of the day it’s still going to be people taking forward distinct pathways, whether it’s a DISC1 pathway or the neuregulin pathway, and exploring the biology in a great detail. That takes time, and that’s where I think the major breakthroughs will come.”
Optimism rules the day
To a person, all the researchers SRF spoke with professed optimism for the next five to 10 years. Whether geneticists or biologists, high-throughput proponents or those enamored of cell-based approaches, all see the next few years as a time of great opportunity to go from genes to pathways, from pathways to biology, and from biology to interventions.
We hope that another series of genetics articles circa 2015 will justify that optimism, but we humbly recognize that the puzzle of schizophrenia has resisted solution for more than a century. To paraphrase Winston Churchill, who spoke of a different battle, closing in on the genetics of schizophrenia will not be the end. It will not even be the beginning of the end. But maybe it will be the end of the beginning.—Pat McCaffrey.
See Part 1, Linkage; Part 2, GWAS, Part 3, CNVs, Part 4, Bigger Genetics. Read a PDF of the entire series.