For the October issue of Nature Methods I did a story about convergence in neuropsychiatric research.
In this context, I interviewed Steve Hyman and a few of his comments are in my story. But there was more. What follows is a slightly edited version of my exchange with him. And I have more in the works to share from my reporting. But first, this.
A few questions for
Steve Hyman, Harvard University Department of Stem Cell and Regenerative Biology, Broad Institute of MIT and Harvard, Director Stanley Center for Psychiatric Research at the Broad Institute
Q: Neuropsychiatric disorders such as schizophrenia, autism spectrum disorder (ASD), bipolar disorder, attention deficit/hyperactivity disorder (ADHD) are tough to parse. In conversations for my story, I heard that in vitro assays, genomic analysis, animal models, human brain imaging have not yielded definitive insight what underpins them. What’s next?
Given the highly polygenic nature of all common psychiatric disorders it is invariably challenging to design informative neurobiological experiments to follow up on results from genetics to investigate disease mechanisms or to nominate candidate biomarkers or drug targets.
Given the difficulty, we have a duty to the students and postdocs who will perform the work to ensure that such studies are based on well powered, rigorously designed genetic studies. Follow-up studies based on putative findings from such discredited approaches as candidate gene methods, candidate gene-environment interaction designs or other underpowered or ill-designed genetic studies, create a disservice to everyone, especially the trainees who still waste years on them.
Since such biological follow-up studies still get published, often in clinical or neurobiological journals that lack editorial expertise or ready access to appropriate reviewers, the upshot is the addition of noise to the literature that many basic neurobiologists and clinical investigators are not in a position to filter out.
It is challenging for neurobiologists to select genes worth working on that have been shown to be disease-associated with a high degree of statistical confidence—as opposed to the fool’s gold of biological plausibility.
This difficulty is a direct result of the yawning cross-disciplinary chasm that lies between the human genetics of common diseases, compiled by hypothesis-free, highly computational projects yielding large datasets comprised of myriad low-penetrance loci and traditional laboratory-based neurobiology, which is hypothesis-driven and reductionist.
In my view, what we need most to bridge this chasm successfully is a cadre of young scientists who have split their training years between labs focused on ‘wet’ neurobiology and labs focused on the generation and analysis of large, complex datasets that capture the complexity and heterogeneity of human phenotypes including disease risk.
I am pessimistic about brief training sessions for established PIs. It is hard to learn a vastly different language and initially counterintuitive approaches in a few weeks. For now, sound selection of genes for biological follow-up studies of common, and thus polygenic, psychiatric disorders might be based on good long-term collaborations between the labs of neurobiologists and statistical geneticists.
There is a “yawning cross-disciplinary chasm
between the human genetics of common diseases,
compiled by hypothesis-free, highly computational
projects yielding large datasets comprised of myriad
low-penetrance loci and traditional
which is hypothesis-driven and reductionist.”
With loci from genome-wide association studies (GWAS) it is also important to know whether the locus has been fine-mapped and that there is reasonable certainty about the gene that is being tagged by significant SNPs-- and to ask whether the directionality of effect has been established. In fairness, the field is deeply in arrears on such mapping, which is difficult and it gets little credit. But without it, many still-born experiments have been undertaken.
For a neurobiologist, the small effect size of GWAS loci in psychiatric disorders should serve a warning not to “Mendelize;” not to think that a knockout tells you what is happening in the disease. Although, a knockout, if studied in more than one inbred mouse strain, might help elucidate what the gene does. The small effect size is also a warning to not think that a significant locus, assuming it is mapped appropriately to a gene, automatically nominates a drug target.
A locus might ultimately lead to a target, but that is many experiments down the road after the role of the gene’s product in a pathway in the right cell types is understood. These extra difficulties should not dissuade labs from following up on psychiatric genetics.
It is useful to think of significant GWAS loci as ‘finding tools’ for genes involved in causal processes, albeit sometimes quite indirectly. Human disease cohorts can hide confounding variables such as in a GWAS study related to lung cancer yielded genes encoding nicotinic receptors. That was because the cohort studied was predominantly made up of smokers.
Moreover, genes identified by GWAS can also be seen as finding tools for pathways and molecular complexes involved in pathogenesis. In copy number variant-associated disorders such ‘nominated’ complexes include the postsynaptic density in excitatory neurons and versions of the inflammasome, among others.
For example, schizophrenia GWAS results point to--albeit not exclusively--to molecular complexes in excitatory synapses as being disease-associated, for example by tagging many synaptic genes including voltage-gated ion channels and also to developmental and experience-dependent synaptic plasticity, for example by tagging not only ion channels and glutamate-receptor subunits, but also at least two genes in complementary pathways.
Once it is possible to identify pathways and molecular complexes and in a disciplined manner, one is on the way to address well-posed neurobiological questions and hopefully many informative experiments.
Q: As you think about how labs working on neuropsychiatric disorders might or might not ‘travel’ methodologically in their research from the genetic wealth of data about disease-causing/influencing genes to insight about which mutations cause changes in proteins or in tissue that can be imaged, in say, schizophrenia or bipolar disorder or autism or other neuropsychiatric disorders, how do you train scientists, also related to the methods they might apply.
I am training and also haranguing anyone in translational psychiatry to reflect deeply on how best to use the many exciting tools that technology has bestowed on us during the last decade or so—and to understand likely reasons for the failure, more than a half century after the pharmacologic revolution of the 1950s. I have reviewed far too many papers in which advanced technology is put in the service of ill-posed questions or poor experimental designs.
“I have reviewed far too many papers
in which advanced technology is put in
the service of ill-posed questions or
poor experimental designs. “
What approaches do I advocate? Nomination of genes and pathways for translational studies absolutely must come from human genetics, and when we know more, they might come from other aspects of human brain biology. These are genes selected from rigorous large-scale studies of human genetics that have been properly fine-mapped can be worked on with confidence.
Having selected a gene, basic science must follow, since we know so little about what most genes associated with psychiatric disorders actually do. One potential advantage of analyzing polygenic disease compared with monogenic disorders—at least in theory—is that there may be genetic loading within a few of the hundreds, or perhaps thousands, of cell types in the brain.
The large datasets—human brain single cell nuclear transcriptomes across development and across phenotypes—and the computational methodologies to exploit such data are still under construction. This study is a good start, in my view.
Now, once you have selected one or several rigorously identified human genes, if you are fortunate you might know about a few relevant cell types in which those genes are expressed together with other GWAS loci. Perhaps you want to understand interaction partners in the relevant cell types. In any case, there is much basic science that will be needed to put the genetics to work in the service, ultimately, of improving the lives of patients.
A true bottleneck in our field right now is advancing the utility of diverse experimental systems in which to interrogate the function and dysfunction of selected genes and pathways and developing criteria for how to appropriately match the lab’s questions with model systems.
Ideally, some of that work needs to be done in human cellular systems, be that in a two- dimensional system or in organoids, but it must be recalled that for regulatory loci, informative patterns of gene expression depend not only on species but cell type, which for the nervous system is daunting.
Go ahead and work in fibroblasts if you want to study skin and in lymphoblastic cell lines if you are a hematologist, but too often scientists rationalize the use of cell types chosen for convenience. Work in model organisms will be necessary, especially if the genes in question influence circuit development, cognition, or behavior. In doing so it is critical to be clear-eyed about what is conserved from the chosen species to humans.
Some consortia in neuropsychiatry
A consortium of 800 investigators in 38 countries. The data access portal is here
Scientists at 15 research institutes are studying the regulatory elements in the genomes of individuals with neuropsychiatric disorders.
An group of scientists mapping the networks underlying neuropsychiatric disorders such as autism spectrum disorder, intellectual disability and schizophrenia.
Enhancing neuro-imaging genetics through meta-analysis.
Information about data access is here.
Comorbidity and Synapse Biology in Clinically Overlapping Psychiatric Disorders. A European Consortium with a focus on the mechanisms of intellectual disability, autism spectrum disorder and schizophrenia.