A Model Is a Model Is a Model of Mental Illness?
October 31, 2013. Apologies to Gertrude Stein and her rose, but the paraphrase of the title is apt when discussing genetically altered mice asserted to be models of mental illness. Two new, high-profile papers bring this issue to the forefront.
In a study appearing October 16 in Neuron, researchers in Susumu Tonegawa's laboratory at the Massachusetts Institute of Technology in Cambridge report that ensembles of hippocampal neurons in a transgenic mouse lacking calcineurin in its forebrain do not appear to "replay" recently learned maze information as wild-type mice do. They suggest that this might underlie some of the memory and behavioral abnormalities of the mouse, but more importantly, that it represents a mouse model of cognitive deficits in schizophrenia, as they had previously published evidence for calcineurin as a susceptibility gene for schizophrenia.
In another new study, published October 23 in Nature, Huda Zoghbi and colleagues at Baylor College of Medicine in Houston, Texas, suggest that they have found an animal model of bipolar mania in "hyperkinetic" mice who carry an extra copy of the gene for Shank3, causing them to overexpress the protein. To this evidence, they add two patients with duplications of the gene for Shank3 who have complex neuropsychiatric disorders that include hyperactivity or bipolar disorder.
The two studies elicit quite different reactions from other researchers. In a Neuron Preview of the calcineurin transgenic mouse, Patricio O'Donnell of Pfizer, Inc., in Cambridge, Massachusetts, writes that "the loss of awake replay in calcineurin knockout mice provides a glimpse into what could be fundamental mechanisms perhaps relevant to cognitive deficits in schizophrenia and related disorders."
Francis McMahon of the National Institute of Mental Health in Bethesda, Maryland, takes a different view of the "bipolar mania" model in a comment for Schizophrenia Research Forum. He writes that despite the informative characterization of Shank3 overexpressing mice, "the paper suffers from a too strong reliance on partial behavioral resemblances with mania in humans."
The ripple replay
The Tonegawa lab produced the original research suggesting a link between calcineurin and schizophrenia risk, and collaborator David J. Foster of Johns Hopkins University in Baltimore, Maryland, supplied the technique that allowed the researchers to look for neural processing abnormalities in the transgenic mice. Among other researchers over the past decade, Foster has helped to describe neuronal ensembles in the hippocampus that encode spatial maps—so-called "place cells." The key clues have been electrophysiologic patterns, both on EEG and in single neuron recordings, that appear to represent the brain replaying spatial information, presumably to store it in memory.
However, in his editorial on the paper, O'Donnell writes that such replay "relates to more than memory but to pondering of different scenarios evaluated in the learning process." As such, he suggests it could be a model for how cognition works more generally in the brain and thus be a useful probe for cognitive deficits such as those in schizophrenia.
In their study, Foster, co-first author Junghyup Suh of MIT, and their colleagues found that the calcineurin forebrain knockout mice have apparently normal levels of neural activity in the hippocampus as they navigate a maze, yet overactive hippocampi during awake resting periods. This was true both for a hippocampus-wide EEG pattern, termed "sharp-wave ripple," linked to the replay of spatial memory and for the firing of individual neurons. Moreover, Suh and colleagues found that the hippocampal neurons in the calcineurin knockout mice lacked the ordered repetition of place cell firing patterns seen in wild-type mice which apparently replay spatial information while resting.
"Thus, we present a novel form of disruption of neural information processing in an animal model of schizophrenia," write the authors. A critical open question, which the authors do not address, is whether variation in the gene for calcineurin will be confirmed as a risk factor for the disease in the era of genomewide association as the generally accepted standard.
Point mutations and deletions in the three Shank genes have also been linked to schizophrenia, as well as autism and intellectual disability (see related SRF related news story). There are also a few duplications of the genomic region 22q13, containing Shank3, reported in people with mental illness. First author Kihoon Han and colleagues decided to see what effect Shank3 overexpression would have. Their transgenic mice produce ~50 percent more Shank3 than wild-type animals do and are overactive. This would seem to indicate that they might be modeling some aspect of ADHD, but when they gave the mice amphetamine, which can have a calming effect in ADHD, they instead produced even more activity. This suggested to the researchers that they might instead be modeling mania, which is increased by amphetamine.
Various animal tests and behaviors proposed to model mania—activity during tail suspension, acoustic startle abnormalities, abnormal circadian rhythms, increased appetite—pointed in the predicted direction, leading the researchers to propose that they were modeling the mania of human bipolar disorder. One drug effective in some cases of mania, valproate, was effective in normalizing some of these measures, but lithium was not.
The transgenic mice had a number of biological abnormalities, including alterations in the ratio of excitatory to inhibitory signaling (E/I balance) in cultured hippocampal neurons toward more excitation (which has been reported to underlie seizures). When Han and colleagues explored the large number of proteins interacting with Shank3, they zeroed in on proteins regulating the actin cytoskeleton as their main suspects for the E/I alterations. They write that their data suggest that Shank3 acts as a scaffold that allows other proteins to form F-actin.
More on modeling
Han and colleagues found supporting evidence for the idea that the mice model mania in the form of two human patients with multiple neurobehavioral conditions but very circumscribed 22q13 duplications, including Shank3, encompassing two and three genes, respectively. One was diagnosed with ADHD and other neurobehavioral deficits, while the other had a bipolar diagnosis, and both had seizure disorders. They assert, "These findings indicate that Shank3 overexpression causes a hyperkinetic neuropsychiatric disorder."
However, the question of whether this represents a model of mania runs up against data that do not conform: McMahon points out that "Shank3 overexpressing mice also displayed decreased social interaction, decreased ultrasonic vocalization, and lack of response to lithium, which are strongly inconsistent with mania. Hedonic behavior and aggression were not reported."
In the quest to explain the molecular basis of mental illness, McMahon is placing his bets on in-vitro models based on induced pluripotent stem cells derived from people with psychiatric disorders. This, he maintains, may allow us "to make a clean break from anthropomorphic dependence on behavioral 'resemblances' and move toward the molecular components which are truly conserved across organisms."
Vis-à-vis the study by Han and colleagues, on the other hand, O'Donnell argues that it represents a credible model for schizophrenia research in part because it is not purported to represent the primary diagnostic phenotype of hallucinations and delusions. Instead, the researchers are examining neural activity underlying a specific memory function, one that may help explain some aspects of disordered thought in psychotic disorders. "[M]odels should be viewed … as reagents to probe specific questions about etiological and pathophysiological pathways," he writes.
Traffic is increasing on two avenues for research into psychiatric disorders—animal and in-vitro models are increasing in numbers and ease of use, and will soon intersect with a virtual rush hour of many, perhaps hundreds of risk genes of very small effect. It will be interesting to see whether the two merge seamlessly to produce new insights and treatments, or produce a pile-up of too much uninterpretable data.—Hakon Heimer.
Suh J, Foster DJ, Davoudi H, Wilson MA, Tonegawa S. Impaired hippocampal ripple-associated replay in a mouse model of schizophrenia. Neuron. 2013 Oct 16; 80(2):484-93. Abstract
O'Donnell P. Of Mice and Men: What Physiological Correlates of Cognitive Deficits in a Mouse Model of Schizophrenia Tell Us about Psychiatric Disease. Neuron. 2013 Oct 16; 80(2):265-6. Abstract
Han K, Holder Jr JL, Schaaf CP, Lu H, Chen H, Kang H, Tang J, Wu Z, Hao S, Cheung SW, Yu P, Sun H, Breman AM, Patel A, Lu HC, Zoghbi HY. SHANK3 overexpression causes manic-like behaviour with unique pharmacogenetic properties. Nature. 2013 Oct 23. Abstract
Comments on News and Primary Papers
Primary Papers: SHANK3 overexpression causes manic-like behaviour with unique pharmacogenetic properties.Comment by: Francis McMahon, SRF Advisor
Submitted 31 October 2013
Posted 31 October 2013
Huda Zoghbi has been a major contributor over the years to our understanding of genes and gene regulation in brain diseases, such as Rett syndrome. Here she and collaborators report on the neurobiological and behavioral impact of increased expression of the gene SHANK3, which has been implicated in autism, schizophrenia, and neurodevelopmental disorders. Patients with 22q13.3 deletion syndrome—involving SHANK3 and a variable number of contiguous genes—are occasionally diagnosed with bipolar disorder in the context of diverse neuropsychiatric symptoms, but little is known about the even rarer 22q13 duplications which presumably increase gene dosage. Here, Han et al. report that transgenic mice engineered to overexpress Shank3 exhibit abnormal behavior and seizures that are responsive to valproate. Han et al. further show that Shank3 directly interacts with the Arp2/3 complex to increase F-actin levels at the synapse.
This paper is a nice example of how animal models and systems biology can be combined to illuminate the extremely complex effects of key proteins in the brain. I do have a problem with the authors' conclusion that the results "suggest that ~50 percent increase in Shank3 level causes a hyperkinetic phenotype in mice that resembles mania." The overexpressing mice did show hyperactivity, decreased immobility on tail suspension, unusual circadian activity, and response to valproate—all of which are typically reported in rodent models of mania and are broadly consistent with manic episodes in humans. However, Shank3 overexpressing mice also displayed decreased social interaction, decreased ultrasonic vocalization, and lack of response to lithium, which are strongly inconsistent with mania. Hedonic behavior and aggression were not reported.
So this paper tells us a lot of valuable new information about the role of Shank3 on neuronal function, but the paper suffers from a too strong reliance on partial behavioral resemblances with mania in humans. As the field moves toward cellular models enabled by induced pluripotent stem cell technology, it may finally be possible to make a clean break from anthropomorphic dependence on behavioral "resemblances" and move toward the molecular components which are truly conserved across organisms. Meanwhile, genetic studies in human patients are still the best way to form strong hypotheses about genes involved in disease and treatment response.
View all comments by Francis McMahonComment by: Kevin J. Mitchell
Submitted 7 November 2013
Posted 11 November 2013
Instead of asking whether a particular animal model is "valid" as a proxy for a particular psychiatric disorder, we should be asking, Is it useful? Can it tell us something we can't learn in humans? If we base that solely on supposed behavioral similarities, we haven't gotten very far—we might as well just be doing rodent psychoanalysis. What we are interested in is elucidating the underlying neurobiological abnormalities and the pathways from etiological factors to resultant pathophysiological states. Such states should be expected to affect behaviors in a species-specific manner—maybe there will be some surface similarity in the results between rodents and humans, but maybe not. Certainly, expecting any animal model to recapitulate the full profile of human symptoms associated with a particular psychiatric diagnostic category is asking too much—does any human patient model the entire spectrum of disease? If these diagnostic categories are really umbrella terms for hundreds of distinct genetic conditions, each with variable outcomes, then the focus in models should be more on the expression of particular symptom domains than on entire disease profiles.
Starting with strong etiological factors is a proven route to discovery of pathogenic mechanisms. As such, the SHANK3 duplication mice are more inherently relevant to disease than the calcineurin mice, which are an artificial transgenic line not directly representative of any human patient. Indeed, the genetic evidence implicating calcineurin in schizophrenia risk has effectively been superseded by negative results from very large GWASs (unless it has popped up again in the unpublished results of the PGC). It is, nevertheless, a very interesting genetic preparation that can be used to dissect circuit mechanisms of memory, which clearly are of relevance to several disease states. That really ought to be enough to garner a wide readership without resorting to claims of direct disease model validity.
View all comments by Kevin J. MitchellComment by: Barbara K. Lipska
Submitted 13 November 2013
Posted 15 November 2013
There is a classic catch-22 in an attempt to model schizophrenia (and other major mental disorders) as, on the one hand, the main purpose of creating a model is to discover the cause of illness (e.g., a genetic defect and the subsequent pathological processes underlying the disease), and on the other hand, it is unclear what to model because the etiology of schizophrenia is still not well understood. Many new models focus on genetic causes because of the strong evidence for heritability of mental illness and the recent discoveries of particular predisposing genes. It is also becoming clear that in most cases, no single gene is necessary or sufficient to cause the disease, but instead, common, low-penetrance genetic variants in more than one susceptibility gene, each contributing a small effect, act in combinations to increase the risk of illness. In some other cases it is possible that rare, but highly penetrant, mutations (i.e., point mutations, translocations, deletions) in single genes are responsible. It is also increasingly clear that there are interactions among susceptibility genes, and between genes and environmental factors that contribute to the risk for mental illness. Given all this, there is no doubt that the task of modeling schizophrenia in animals is formidably difficult.
It is further complicated by the fact that a gene-based animal model 1) may have to be related to a specifically human transcript and/or protein variant or variants artificially introduced into the animal; 2) will not exhibit abnormalities in all schizophrenia-related phenotypes (as animals will not have hallucinations or delusions); and 3) may require additional environmental manipulations to become fully penetrant at the behavioral level. We should thus perhaps accept the fact that a mouse model for an individual candidate gene will never be representative of the entire disorder, and at best it will reproduce either a subtype of the disorder or a particular aspect of a given phenotype. In that context, human cell-based models and studies of human brain tissues obtained postmortem from patients with mental illness (severely underutilized resources!) are perhaps better alternatives to gain insight into the origins and pathophysiology of these specifically human, challenging disorders.
View all comments by Barbara K. LipskaComment by: Karoly Mirnics, SRF Advisor
Submitted 12 November 2013
Posted 15 November 2013
I recommend the Primary Papers
I think that we are often making a mistake if we directly declare what disease are we modeling. There are no "valid" animal models of human schizophrenia or other major psychiatric disorders, and most likely, there will never be—the mouse is not a human, and has a quite different lifestyle! Furthermore, the mouse and the human genetic diversity are quite distinct. Thus, talking about modeling physiological and pathophysiological processes is much more correct. Understanding behavioral modulation by the various interneuronal subtypes, evaluating the role of gene X on cortical lamination, or assessing the effects of factor Y on neuronal outgrowth are all disease-relevant, essential studies.
View all comments by Karoly Mirnics
Comments on Related News
Related News: Genetics and Schizophrenia—Calcineurin Connection GrowsComment by: Mary Reid
Submitted 26 February 2007
Posted 27 February 2007
Tom Fagan mentions that calcineurin regulates phosphorylations elicited by both glutamatergic and dopaminergic signaling. The activity of D-amino acid oxidase is increased in schizophrenia, and this also affects signaling through both these pathways. Is there any clinical benefit with the use of sodium benzoate which inhibits DAO activity?
He also mentions that EGR3 can be regulated by the activity of neuregulin. Interestingly, Roberts et al. suggest that BDNF, which is decreased in first-episode psychosis (Buckley et al., 2007), induces synthesis of EGR3 to regulate activity of GABRA4. Ma and colleagues (Ma et al., 2005) conclude that GABRA4 is involved in the etiology of autism and it has also has been implicated in nicotine dependence (Saccone et al., 2007).
Glorioso and colleagues (Glorioso et al., 2006) report changes in genes encoding early-immediate genes such as EGR1 and EGR2 and RGS4 which is involved in cellular signaling and has been implicated in schizophrenia following BDNF gene ablation. Several studies report that zinc increases BDNF expression. Does this give support to the zinc deficiency theory of schizophrenia?
View all comments by Mary Reid