23 Oct 2014
October 24, 2014. Even as researchers gain traction on the genes involved in schizophrenia, turning these discoveries into clinically actionable information faces plenty of challenges. Drug development for psychiatric conditions takes longer and fails more often than that for other conditions, prompting the exit of drug companies from psychiatry. Despite these difficulties, researchers in two plenaries discussed ways to smooth the road for getting genetic discoveries into the clinic.
One way to reduce risk to drug companies may be through public-private partnerships, such as the one forged by NEWMEDS in Europe. On Monday, October 14, Tine Stensbol of Lundbeck in Denmark described this project, a collaboration of 19 different pharmaceutical and academic institutions that have drawn up a plan to clear bottlenecks in drug development for schizophrenia and depression. NEWMEDS was born in 2009 with €1 billion contributed by the European Union and €1 billion by drug companies, and an understanding that any intellectual property that came out of the collaboration would be shared. Slated to end in 2015, the project is focused on improving animal models for psychiatric disease, which often do not predict drug efficacy, and finding biomarkers that could subtype disorders, which could give cleaner results in clinical trials.
One achievement Stensbol noted was to get the CANTAB suite of cognitive tests for rodents up and running across different sites. "It's a little boring because it's just replicating, but it's really important," Stensbol said. The collaboration also made a database of past clinical trials for psychiatric drugs, with analyses showing that outcomes at four weeks are equivalent to those at six weeks, suggesting that answers to trials can be obtained sooner—and more cheaply—than previously thought. She also noted that women in clinical trials had less of a placebo effect than men, suggesting that a drug's true effect size may be more easily obtained in women.
The group is also developing three mouse models of copy number variations (CNVs) associated with increased risk for schizophrenia: 1q21.1, 15q13.3, and 22q11.2. Some of this risk might be mediated by cognitive and brain abnormalities, according to their study published last year of unaffected human carriers of these CNVs (see SRF related news report).
Panelists discuss pathways to therapy
A plenary panel on Wednesday, October 15, took up the "bench to bedside" question again. Raimund Buller of Lundbeck outlined the problems that made drug companies skittish: poor animal models (too often based on how current drugs work), no biomarkers for the disorders, heterogeneity within a diagnosis, and the increasing placebo response. Despite all that, he said that pharma has an interest in genetics, which could help remedy these problems. He also proposed collaborations among drug companies, academia, and regulators with clear roadmaps and honest timetables.
"We have a harder problem than anywhere else in medicine," said Steve Hyman of the Broad Institute, Cambridge, Massachusetts, noting the high polygenicity, heterogeneity in etiology, and small effect sizes of signals coming out of genomewide association studies (GWAS), combined with the inability to access the brain, let alone the circuits, involved. Calling the latest schizophrenia GWAS "a difficult beginning," he pointed optimistically to new and ever cheaper ways of interrogating the genome. He also suggested that in vitro systems of human cells, such as induced pluripotent stem cells (iPSCs) or organoid culture systems, may be more useful than transgenic mice for understanding the function of the hundreds of variants that increase risk for diseases such as schizophrenia or autism. For example, even if it were possible to swap the more than 100 schizophrenia variants into a mouse, it would still not be a schizophrenia mouse because it lacks the exact brain regions and cell types found in humans, he said.
Peter Falkai of the University of Gottingen, Germany, presented the clinician's perspective, citing pharmacogenetics—the genetics of drug response—as the most promising bridge to the clinic. Yet how strong the effects and the evidence should be before introducing genetic tests to the clinic remain unclear. He also noted ethical issues in making genetic counseling routine, communicating realistic expectations about the meaning of test results, and putting into place legal and institutional guidelines for genetic counseling.
Societal challenges were also noted by Kari Stefansson of deCODE genetics in Reykjavik, Iceland. For example, he noted that deCODE's database contains information on women in Iceland who carry a mutation that dramatically increases risk for breast cancer, but given the right not to know, the company can do nothing to contact these people. "How do you convince society to make use of this information?" he asked. He also noted the delicate nature of targeting the brain, because compounds that normalize biochemical pathways disrupted in psychiatric conditions may also shift a person's sense of self or personality.
Scientifically, Stefansson highlighted a role for rare variants in leading researchers to the pathways perturbed in psychiatric disorders, as well as the possibility for somatic mutations in the brain. He also noted pleiotropism, in which genes affect more than one trait, which may mean that the same biochemical pathway may be manipulated to treat different disorders.
Working in both cancer and depression, Richard Weinshilboum of the Mayo Clinic, Rochester, Minnesota, argued that the genetic subtyping that has helped subclassify kinds of cancer and predict drug response will also work for psychiatric disorders. He foresees combining genotype information with drug responses from human cell lines to individualize treatment. "That's the future of psychiatry," he said.
In the following discussion, several panelists agreed about the need to train more genetic counselors. Falkai advocated telling patients the current state of the knowledge, even if it is not fully formed, and also argued that current diagnostic categories can still be of some use. Hyman noted that some kind of explanation, be it genetics, environment, or both, can satisfy patients, even if that information does not yet guide treatment. Stefansson called the challenges in using genetic testing in clinics "formidable, but doable."
When Buller reiterated the importance of defining reasonable timetables for drug development (saying, "otherwise, these ideas are just music"), Stefansson disagreed, saying it was too early to force discoveries. "We are still just poking in the dark," he said. For pharmacogenomics, however, it might not be too early for a roadmap, Falkai said.
When pressed by audience member Francis McMahon about whether he suggested getting rid of animal models in favor of iPSCs, Hyman said both were needed, with animal models particularly useful for basic science questions, rather than for modeling human illness. He advised using the best model for a particular question and not pursuing behaviors because they somewhat resemble those in humans. "Behavior is a moral hazard," he said, adding that it can be useful when it gives a readout of a pathophysiology.
Traction on TCF4
In this murky time, a few intrepid biologists have begun to investigate the function of genes implicated by GWAS. An early favorite has been transcription factor 4 (TCF4), which controls expression of other genes. Just which genes these are was the focus of a talk on Tuesday, October 14, by Joseph McClay of Virginia Commonwealth University, Richmond. McClay described genes to which TCF4 bound, as revealed from chromatin immunoprecipitation and sequencing in neural cell lines. This led to nearly 400 genes of interest, which included those involved in cell adhesion and axon guidance. Analysis in postmortem brain tissue suggested that these TCF4 targets were differentially expressed in schizophrenia compared to controls.
In the same session, Brady Maher of the Lieber Institute in Baltimore, Maryland, presented results from manipulations of TCF4 expression in cortex. Knockdown of TCF4 in embryonic rats resulted in cortical neurons that did not sustain spiking during depolarization. To understand why this might be, Maher purified the "translatome"—RNA connected to ribosomes—with a focus on rat ionic channel genes. This revealed significant upregulations of KCNQ1, a potassium channel, and SCN10A, a sodium channel. Blocking KCNQ1 could rescue the TCF4 knockdown phenotype, whereas agonists to SCN10A could mimic it. Together, the findings suggest that TCF4 can alter information processing between cortical neurons.—Michele Solis.