28 Apr 2015
April 29, 2015. The future of animal models in psychiatric research, particularly for schizophrenia and other psychotic disorders, was discussed at a workshop on Tuesday evening, March 31, at the International Congress on Schizophrenia Research in Colorado Springs. Acknowledging the perceived failure of animal model findings to translate to humans, the eight-member panel and about 30 audience members grappled for two hours with what animal models could and could not teach researchers. Several panelists referred to the maxim that "Animal models are all wrong, but some of them are useful," and urged researchers to be more precise in describing their models. This kind of precision could align animal models with the Research Domain Criteria recently adopted by the National Institute of Mental Health (NIMH), which moves toward classifying mental disorders according to the neural systems disrupted. Others saw more reasons for hope in that animal models may help parse the varied etiologies of these disorders and even offer researchers insight into why clinical trials failed.
A wealth of schizophrenia-related models has been generated, ranging from developmental and drug-induced to genetic and genetic-by-environmental factors. Since 2007, the relevant findings have been curated for the Animal Models for Schizophrenia Research database at the SRF website by Jim Koenig and colleagues at the Maryland Psychiatric Research Center. Opening the workshop, one of the session organizers, SRF's own Hakon Heimer, asked the audience to think about who should maintain this resource in the future, now that Koenig has moved to the National Institutes of Health, and what its scope and criteria should be.
Modeling processes related to schizophrenia may be especially complicated, given that the disorder is thought to have its roots in early development, yet diagnostic symptoms only emerge much later. Kicking off the brief presentations, Kim Do of the University of Lausanne, Switzerland, said that animal modelers needed to heed the timing of their manipulations because of this fact, and suggested that the heterogeneity of symptoms among people with schizophrenia may reflect different timing of perturbations during the prolonged process of brain development.
Later in the discussion, John Waddington of the Royal College of Surgeons in Ireland, Dublin, echoed the necessity to follow the evolution of the consequences of any perturbation researchers might make. Though it was important to map this trajectory, he said there are logistical and financial constraints in keeping rodents in experimental protocols for many months at a time to allow prospective studies.
Trapped by drug-induced models?
Some of the most prominent animal models involve giving a drug such as amphetamine to mimic the enhanced dopamine signaling found in psychosis. Such treatments lead to psychotic-like hyperactivity, impaired prepulse inhibition (PPI), and attention deficits—phenotypes that can be reversed by antipsychotic drugs (APDs). These acute drug-induced models have become the go-to standard in preclinical research, but they have misled the field, asserted Anthony Grace of the University of Pittsburgh. Though they may simulate components of the schizophrenia disease state, he said that they locked people into thinking that dopamine was the root problem of schizophrenia and ignoring other etiologies. At best, he said, these were models of drug action.
Mark Geyer of the University of California at San Diego further noted that developing drugs to reverse drug-induced phenotypes led only to drugs that worked through the same mechanisms, a situation he referred to as "receptor tautology." He acknowledged that these models had some value in identifying new APDs, but that by design they ensured that researchers found basically the same kinds of drug. On the other hand, using developmental manipulations has greater potential to identify novel treatments for aspects of schizophrenia. Amanda Law of the University of Colorado in Denver pointed out that it was likely unrealistic to expect truly novel drugs to reverse all of the classical drug-induced phenotypes, and that we should be examining novel drugs in new ways, such as reversing learning or memory dysfunction.
An audience member questioned whether anyone truly believed that drug-induced models were still viable. A case in point is the great expectations in the field for NMDA receptor blockade models (see the glutamate hypothesis of schizophrenia by Moghaddam), but clinical trials of metabotropic glutamate receptor agonists based on work in these models have not panned out (see SRF related news report). However, workshop co-organizer Patricio O'Donnell of Pfizer in Cambridge, Massachusetts, explained that these clinical trials weren't necessarily conclusive. The trials took any person with schizophrenia, without any effort to see if glutamate disruptions were part of the person's disorder.
Several people agreed with this concern—one panelist said that if "all comers" were admitted to cancer trials, that field wouldn't have seen so much progress, and an audience member said that, given the diversity of genetic clues emerging for schizophrenia, looking at schizophrenia as a unitary concept was hopeless. Geyer said that part of the reason for the current bad reputation for animal modeling lies in schizophrenia's different etiologies. "If we're asked to predict clinical efficacy in all comers with different etiologies, we should refuse," he said. "Our mistake, as animal modelers, was to accept that challenge."
Grace also argued that the field needed animal models to reveal what is wrong with clinical trials. These trials conflate people with very different medication histories, yet researchers know that antipsychotic medications can have lasting effects on the brain that a brief period of wash-out may not change. He suggested that people early in their illness might provide a better comparison to animal models because their brains have not yet been changed by years of medication.
Model symptoms, not diseases
In his presentation, Geyer emphasized precision when talking about animal models, with explicit statements about both the manipulation and what is measured. For example, "PPI models"—which refer to disruptions in prepulse inhibition—are common in the field, but they arise from different perturbations. Each model's utility can be measured according to its face validity (looks like some aspect of schizophrenia), construct validity (assesses the construct of interest, such as "gating" or "attention"), etiological validity (mimics a hypothesized brain or genetic pathology), and predictive validity (responds to parametric manipulations or treatments as seen in humans). Focusing on the neurobiology of these different concepts would shift efforts toward understanding the neural systems involved in schizophrenia and away from simulating a humanized version of the disorder in rodents.
This kind of systems-based perspective points to some convergence across animal models, said Grace, citing parallels between the ventral hippocampal lesion model and the MAM (methylazoxymethanol acetate) model he developed, which both produce hypersensitivities in the same brain circuits.
He also reminded the workshop members that schizophrenia was more than psychosis. "Schizophrenia was originally conceived of as a cognitive problem, but with the development of antipsychotic drugs, it was cast as psychotic," he said, adding that psychosis isn't the most troubling symptom for many people with the disorder.
Some people in the audience suggested that animal models should focus on finding drugs to treat the cognitive and motivational deficits in schizophrenia that go untouched by APDs. However, another audience member countered that APDs work well in only 40 percent of people, many of whom would rather not take the medications. "I'd argue it is still an important problem," he said. Another audience member spoke to the utility of modeling adjunctive treatments that could work hand in hand with current APDs.
Given that it is hard to model the etiology of a disease whose pathophysiology remains largely unknown, Law maintained in her presentation that genetics provides the strongest clues, though that field is still in its infancy. Animal models, she said, would be most useful as and when researchers work out how a particular genetic risk variant alters gene expression or function. Animal models should then recreate more etiologically the disease-related molecular mechanisms—for example, mimicking a boost in a particular gene's expression found in association with risk for schizophrenia. She argued that it was not enough to simply create a knockout of a gene and that modeling a disease's molecular phenotype has more validity. O'Donnell replied that this kind of information was still useful. Law agreed that gene knockouts do provide useful information, but mainly in the context of "gene function." How these models relate to disease risk or pathophysiology is an entirely different question.
Waddington remarked that because psychosis is "disrespectful" of diagnostic boundaries, occurring in 12 different disorders, and that risk genes for one disorder also apply to others, models needed to reflect this broader view of psychiatric biology. "We need to abandon the diagnostic silos in animal models," he said. He also said there was room for more sophisticated animal models of epistatic interactions between genes (see SRF related conference report).
The final presenter, Mikhail Pletnikov of Johns Hopkins University, Baltimore, Maryland, agreed, saying that animal models have a lot of untapped potential for exploring gene-gene or gene-environment interactions. He encouraged researchers to report effect sizes of their manipulations in order to facilitate comparisons between models. He also reminded the workshop attendees that sex-dependent effects were underappreciated, and a careful accounting of results in males and females should become more common.
One problem with animal models may stem from the inadequacy of their human counterparts, Geyer noted. He argued that animal modelers needed to put pressure on clinical colleagues to provide quantitative measures of human behavior or biomarkers, rather than scores on questionnaires that require language. Improved human experimental medicine studies would provide measures that would be more translatable in animals and perhaps bring to light subgroups that lend themselves to more precise animal modeling and more personalized medicine.
One example of this approach might be the CNTRICS (Cognitive Neuroscience measures of Treatment Response of Impaired Cognition in Schizophrenia) battery of tests designed to pinpoint the parts of cognition affected in schizophrenia. These specific tests, Law said, may uncover aspects of cognitive function that have clearer parallels in animals. She also said that animal modelers need not restrict themselves to behavioral readouts: imaging, electrophysiology, neuroanatomy, and biochemical markers may also highlight neural systems that are similarly perturbed in animal models and schizophrenia.
This cross-validation between animals and humans could be an iterative process, and one audience member suggested it could help falsify some animal models. Kim Do said that this kind of back and forth between species led her to a new mechanism linking low glutathione to reduced myelination in the brain (Monin et al., 2014).
An audience member made the point that animal models could be relevant to postmortem studies. Though postmortem studies can be difficult to interpret—for example, decreases in a type of neuron could be a cause or consequence of a disorder—animal models could help parse this, she said.
Grace said that it was important to track intact abilities as well as deficits to get a fuller picture of what was going on in the brains. "We have to make sure that we are not just looking at what we expect to see," he said.
However, focusing on behaviors shared across species may miss informative phenotypes, Waddington said. Naturalistic behaviors of rodents, such as nest building, may give clearer signals than behaviors that are not in the normal rodent repertoire. Other panelists offered that such ethological considerations had led them to reconsider animal housing, or to reverse their light cycle so that nocturnal rodents are active at the same time as their diurnal researchers.