Animal models have significantly increased our understanding of human nervous system disorders and their treatment, but many drugs which have proved effective with animal models have not been successful in human clinical trials (Tricklebank & Garner 2012). There is thus a need to improve the use of animal models for the discovery of effective new drugs for human disorders (Paul et al. 2010). The workshop on ‘Improving translation of animal models for nervous system disorders' held at the National Academy of Sciences, Institute of Medicine, in Washington, DC, 28–29 March 2012, was organized to discuss the issues that contribute to the poor translation of results from animal models to human nervous system disorders, to consider strategies to increase the scientific rigor of preclinical testing, to identify methods to maximize bidirectional translation between basic and clinical research, and to determine the next steps for improvement of the development and testing of animal models of nervous system disorders. Once again, the search is on for better mouse models (Beckers et al. 2009) and the proceedings of this workshop will be of great interest to those doing research in genes, brain and behaviour.
In the first session, Steven Paul discussed the expectations of animal models in drug development, including their role in understanding disease biology, interpreting the relationships between genotype and phenotype and testing the safety, toxicity and side effects of new drugs (Paul et al. 2010). Mark Tricklebank emphasized the importance of testing the validity of animal models of central nervous system (CNS) disorders before using them as targets for novel drugs. He examined the ways in which animal models may fail, including problems with experimental design and statistical analyses (Nieuwenhuis et al. 2011), the mismatch between preclinical animal data and human disease mechanisms, and the species gap in CNS disorders such as schizophrenia, which does not naturally occur in rats and mice (Morgan et al. 2012). Katrina Kelner spoke on the impact of publication bias and suggested that negative results should be published as should the results of clinical trials when products fail (Clozel 2011). She suggested that research standards should be increased by journals and that statistical methods should be scrutinized carefully (Young et al. 2008). There was considerable discussion on the role of journal editors in improving the quality of publication, a topic that is not new to GBB (Crusio et al. 2009).
The second session consisted of case studies which focused on the problems in translational research using animal models of neurodegeneration (Kim et al. 2011); Alzheimer's disease (Chin 2011; de Calignon et al. 2012); stroke (Manaenko et al. 2011); schizophrenia (Moore 2010); addiction (Sora et al. 2010) and pain (Mogil 2009). The post-breakout discussion focused on potential causes for translational failures with respect to the NIH white paper on new approaches to drug discovery (Sorger & Allerheiligen 2011). Emphasis was placed on the analysis of cases in which there are known translational failures and on attending to guidelines for best practices in preclinical medicine (NINDS 2011). There are also preclinical research guidelines for Alzheimer's research (Shineman et al. 2011), toxicology (Stokes 2011), stroke (Dirnagl & Lauritzen 2011) and amyotrophic lateral sclerosis (Ludolph et al. 2010).
The third session examined the key components of animal model research that would benefit from standar dization. Andrew Holmes discussed the issue of standardization in testing mouse models of anxiety. The issue of standardization in mouse behavioural phenotyping studies remains controversial, with some labs working to develop standardized procedures (Fuchs et al. 2011) and others critical of the attempt to standardize behavioural tasks (Richter et al. 2009). Others, such as Kafkafi et al. (2005) have suggested ways of examining data from genotype by laboratory interactions to extract replicable behavioural measures from the data from different laboratories. Tim Bussey discussed the use of the touch screen cognitive testing method for rodents (Bussey et al. 2008) which has the advantages of being automated, high throughput, non-aversive and translational as results can be compared with humans tested on the CANTAB apparatus (Levaux et al. 2007). Lennert Mucke discussed the risk/benefit ratio of standardization in animal models of Alzheimer disease (AD) and suggested that we must look for functionally relevant endpoints, such as the direct comparison of rodent models of visuo-spatial learning (Cisséet al. 2011) with assessment of navigational skills of patients (deIpolyi et al. 2007). As for improving animal models, he suggested the use of best practices for preclinical animal studies (Shineman et al. 2011) and concluded that, while standardization was important when principles of knowledge are well developed, it is premature when the disease mechanisms are unknown (Huang & Mucke 2012).
Session four facilitated development of corresponding animal and clinical endpoints. Neal Swerdlow compared auditory startle and pre-pulse inhibition (PPI) of startle in human patients and animal models. PPI has predictive validity, the response is comparable across species and drug effects are the same, but there are subgroups which may respond differently in both humans and mice (Swerdlow et al. 2009). There are common endpoints across species, and the neurobiological substrates appear homologous (van den Buuse 2010), but because PPI deficits are not specific to any one disease, PPI cannot be used to diagnose any specific disease (Powell et al. 2012). Larry Steinman spoke on the successes and failures of using the experimental autoimmune encephalomyelitis (EAE) model of multiple sclerosis (MS). Based on research with this animal model, the drug Natalizumab was developed (Steinman 2005). Unfortunately, some patients taking natalizumab developed progressive multifocal leukoencephaly (PML) and died (Langer-Gould & Steinman 2006) and Natalizumab was withdrawn. What went wrong? First, there is no animal model of PML, so it was undetected in preclinical trials. Second, the risk of having PML in addition to MS is very low, so when natalizulab was reinstated, it was with a warning that patients should be tested for PML (Steinman et al. 2012). One question arising from the controversy was the appropriateness of the EAE model for studying MS (Mix et al. 2010; Croxford et al. 2011). Michela Gallagher discussed the rat hippocampus model of age-related memory loss (Gallagher et al. 2010) and described how hippocampal hyperactivation is associated with hippocampal thinning in Alzheimer's patients (Putcha et al. 2011). This results in a loss of pattern separation and a shift to pattern completion with more incorrect and fewer correct memories (Yassa & Stark 2011). These counter-intuitive results with animal models and human subjects have led to the use of Levetiracitam as a possible treatment for AD (Koh et al. 2010).
Session five explored methods for increasing bidirectional application of research findings between basic and clinical researchers. Richard Ransohoff discussed the importance of looking at the role of cortical demyelination in the early stages of multiple sclerosis (Lucchinetti et al. 2011). He criticized the EAE model of MS and suggested that new mouse models of MS that spontaneously develop EAE (Krishnamoorthy et al. 2009) should be investigated using standardized procedures involving special facilities for conducting research and drug testing on demyelination/remyelination models. Deanna Barch discussed the development of the MATRICS program to guide preclinical research in schizophrenia (Young et al. 2009). The aim of this program is to develop a consensus on the improvement of predictive validity of studies on the enhancement of cognition in schizophrenia (August et al. 2012). Gerry Dawson talked about the P1 Vital approach to bridging the gap between animal models and clinical trials which combines academic and pharmacological industry expertise (Dawson et al. 2011) to develop new treatments for neuropsychiatric disorders (Koychev et al. 2012). Hugo Geerts suggested that the pharmacological industry should utilize the methods developed by the aeronautics and microelectronics industries for dealing with complex systems by formalizing ‘collective knowledge,’ embracing complexity and analyzing failures and using the methods of quantitative testable modelling of new drugs using computer-aided research and development. Since humans are more complex than animal models, different methods may be needed to discover new drugs for complex multifactorial diseases (Swinney & Anthony 2011). He proposed a system called mechanistic disease modelling (Geerts 2011) which uses computational methods to integrate preclinical data, human pathology data, CNS drug pharmacology, receptor physiology and data from clinical trials to help develop new drugs for Alzheimer's, schizophrenia, Parkinson's, and other neurological disorders. Thomas Steckler discussed the formation of NEWMEDS, a public–private partnership for the discovery of new drugs to treat schizophrenia and depression. He suggested that adaptive design principles are important in the development of new drugs (Wang et al. 2011) and should be used to improve the predictive validity of drugs developed with animal models (Gilmour et al. 2012).
The objectives of section six on future directions and next steps were to define important, yet practical expectations for animal models in nervous system disorders and to identify opportunities for the success of improving translation of results from animal models to human patients. In this session, the session chairs from each of the previous sessions summarized the results of their sessions. What did we learn from this workshop? In all diseases there is a use of animal models to understand the underlying neurobiology and develop new drugs, but many of these drugs are not effective in human patients. There were many suggestions as to how the predictive validity of animal models could be improved. The presentations from the workshop, ‘Improving translation of animal models for nervous system disorders' are available on the workshop homepage. < http://www.iom.edu/Activities/Research/NeuroForum/2012-MAR-28.aspx >. The slides are located on the right hand side of the page under the dropdown menu entitled ‘presentations’ and listed by the speaker's last name.
What was not discussed at the meeting? Many of the confounds in animal studies discussed by Schellinck et al. (2010) were not discussed at the workshop. These include developmental effects, housing and vivarium issues, social experience, the test room environment, equipment and the factors in the lab environment which lead to idiosyncratic results. Another issue not discussed were the institutional barriers against good research: th refusal of animal care committees to support research which replicates published work; the 1 and 2 year grants which do not enable adequate time to conduct studies and the expectation for rapid publication of preliminary results in order to obtain academic promotion and grants, the lack of funding for the design and testing of new equipment, the lack of resources to develop standardized tests, etc. Finally, there was the implicit assumption that the environmental or laboratory effects are simply noise to be eliminated (Crabbe et al. 1999); but what if the environmental effects are real? There was no mention during the meeting of epigenetic effects and their interaction with genes, although there is considerable research on genetic–epigenetic interactions in Alzheimer's Disease (Daniilidou et al. 2011); schizophrenia (Gavin & Sharma 2010); drug addiction (Wong et al. 2011); chronic pain (Denk & McMahon 2012); Parkinson's disease (Habibi et al. 2011); cognitive processes (Day & Sweatt 2011) and autoimmune disorders (Costenbader et al. 2012). Consideration of new epigenetic approaches to the study of nervous system disorders using animal models may result in the development of new treatments for these disorders (Hamm & Costa 2011).