Director’s Blog: Mapping the Risk Architecture of Mental Disorders
By Thomas Insel on
It’s difficult to overstate the impact that genomic medicine is having on biomedical research and practice. For cancer diagnostics, rare disease therapeutics, and fields like microbiomics and infectious diseases, the advent of cheap, fast, precise genomic sequencing has been a game changer. What about mental disorders? There has been a lot of hype about genomics revolutionizing diagnosis or treatment of mental disorders, but is there any real hope that the kind of advances that have helped patients in the rest of medicine will help people with autism or schizophrenia or mood disorders?
The history of psychiatric genomics has been, until recently, disappointing. The search for candidate genes—such as those, like the serotonin transporter gene, suspected to be contributors to risk because of their role in medication response—led to many papers but few replications and no actionable findings. Unbiased scans of the whole genome were challenging because there is so much variation in the genome, most of which is unrelated to risk or resilience. To detect a signal from all of this background noise, one would need many thousands of samples. Over the past five years, as the field realized the need for larger numbers of samples, investigators from around the world have worked together to share results in the hope of attaining the statistical power needed to find variants associated with schizophrenia or autism. New findings demonstrate that sharing data does indeed lead to exciting results.
A report in Nature this week from the Psychiatric Genomics Consortium, a team of investigators in more than 80 institutions across 25 countries, looks at common variation (variation present in 10 percent of the general population) in nearly 37,000 cases of schizophrenia and over 113,000 controls.1 This genome wide association study revealed 108 different loci where variations were associated with schizophrenia; 83 of these had not been reported previously. Note, these are not “108 genes for schizophrenia.” These are areas of the genome where variations in sequence are associated with schizophrenia. Most of these are not in or even near genes. And any one of these 108 regions contributes only a tiny fraction of risk in the population. Nevertheless, this is a major step forward in describing the genetic risk for schizophrenia.
Beyond the huge number of new associations, what makes this such an exciting study? First, many of the common variants overlap with rare mutations identified in previous reports as associated with risk of schizophrenia, confirming that some regions are worth a deeper dive. Second, some of the previous “suspects” for schizophrenia such as the dopamine D2 receptor and specific glutamate receptor subtypes which are targets for anti-psychotic medications, are in the list of 108 loci. Finding known targets in this list begs the question of how many other loci might contain targets for medications yet to be developed. And third, these genetic findings give us a new opportunity to look at risk. A polygenic risk score developed with these data can be used to stratify individuals for vulnerability to schizophrenia. In case-control studies of three populations, those in the top 10 percent of risk had as much as a 20.3-fold increase in risk for schizophrenia. While the use of the polygenic risk score needs to be validated in a general population, these kinds of risks could conceivably be helpful for detecting the earliest phases of the disorder, years before psychosis.
This new report adds to the growing evidence that common variation will be a major contributor to risk for schizophrenia, just as it is for diabetes, hypertension, and inflammatory bowel disease. The genetics of autism, by contrast, has been mostly about rare variants. Many of these rare variants have been spontaneous or de novo changes in the genome and some have a considerable impact on risk. This simple picture of schizophrenia as a common variant disorder versus autism as a rare variant disorder was overturned this week by a report in Nature Genetics demonstrating that many common variants must be associated with autism.2 As in the early days of schizophrenia genetics, the previous absence of common variants in autism appears to be a function of numbers. With a population sample including over 5,000 strictly defined autism cases, common variation emerges as the major source of genetic risk for this disorder. The precise components of this risk need to be determined in other studies, but the new picture suggests that the risk architecture of autism may not be profoundly different from schizophrenia.
It may seem as if psychiatric genetics has gone from too few clues to too many. In fact, these new findings in schizophrenia and autism place these disorders squarely in the field of complex genetic disorders, disorders in which scores or hundreds of variants, both common and rare, contribute to risk. This recent progress is indeed a giant step forward for the field, but it is one step on a long journey. In truth, we do not have a rapid means to pivot from a genomic association to a target for treatment development. And complex genetic disorders, by definition, will not yield a simple genetic test for diagnosis. But these findings do suggest a way forward. By identifying the molecular pathways of risk, using cell-based studies with those pathways manipulated, and filling in the gap between molecular neuroscience and brain function, these new findings become part of the foundation for translational science.