5Advancing the 3Rs in Neuroscience Research

Repair Strategies

After an injury, formation of glial scars inhibit central nervous system repair by creating both physical (e.g., cyst) and biochemical (e.g., inhibitory signals) barriers to axonal growth. The goal of any repair strategy is to fill any cysts, maintain glial/neuronal survival, limit scar formation, promote axonal regeneration, and make functional reconnections. Using animal models, researchers are studying injecting growth factors, blocking inhibitory signals (e.g., anti-Nogo [described by Lemon]), transplanting cells, bridging the gap using biodegradable scaffolds to align the axons, and promoting plasticity/sprouting of any remaining intact fibers. No one treatment alone is capable of repairing the spinal cord, Barnett noted. Current thought is that a combination of strategies will be required.

Three main laboratory strategies are currently used to treat a damaged spinal cord. The first, neuroprotection, is to protect what is left and minimize further damage. Second, especially for incomplete injuries, the strategy includes remyelination or making the most of what remains. Repair is the third strategy, which includes restoring communication, axonal regeneration, and reconnection, often by cell transplantation or pharmacological intervention.

Spinal Cord Research in Animals

Spinal cord research in humans is difficult and in some cases impossible. There is no ability to biopsy tissue, imaging is limited, and studies cannot be done on large groups of people with similar pathology. The only way to investigate spinal cord injury, Barnett said, is to use animal models or primary cells from animal tissue.

An example of an animal model of a spinal cord lesion is a wire knife lesion, generated by inserting the knife into the dorsal column and pulling up a piece of tissue. Barnett noted that this method is clean, accurate, and consistent, resulting in a cavity and glial scarring that mimics human spinal cord injury. By tracing regenerating axons using fluorescent labeling techniques, Barnett has observed that while many axons enter and fill the lesion site, they have limited ability to grow through the lesion, and few exit and find their target. This, Barnett explained, is the major problem with many of the spinal cord injury repair therapies.

One aspect of spinal cord injury that researchers want to mimic is the glial scar. A useful model would have a lesion surrounded by reactive astrocytes that express molecules of interest; axons would be inhibited from entering or exiting the scar and would become demyelinated; and there would be activated microglia.

Several disadvantages to rat models of spinal cord injury include the need for large numbers of animals, the severity of the procedure, and the distress and discomfort to the animals, Barnett said. Additionally, there is a long time frame for results and the experiments are expensive and time consuming. To address this, Barnett is working to replace animals in her experiments.

Replacing Animals with Cell Culture

Barnett described her in vitro model of spinal cord injury in which disassociated embryonic spinal cord cells from rats are layered on top of an astrocyte monolayer derived from embryonic tissue (Sorensen et al., 2008). Growth in culture over time leads to complex axonal/glial interactions resulting in myelinated neurons. This system allows for the study of contact between astrocytes and how they communicate with the axons, which is necessary for understanding these problems in spinal cord injury. Barnett and colleagues next induced lesions in the cell culture by cutting with a scalpel to studying axon density and myelination adjacent to the lesion and cell growth into the damaged area. To validate the model, Barnett has studied several molecules previously tested in vivo to see if they could promote outgrowth or repair in the vitro system.

Overall, the findings from the in vitro model of spinal cord injury correlate with in vivo findings, including the formation of features typical of a glial scar, neurites that do not cross the boundary of the scar, and myelination and neurite density that is decreased adjacent to the lesion. The cells in culture respond to reagents that have been reported to promote axonal growth in rat models of spinal cord injury. This model also could be used to prescreen combinations of biological and pharmacological agents for potential therapy for repair of spinal cord injury. Barnett noted that getting the model published so others can become aware of it has been successful, but also challenging (Boomkamp et al., 2012).

Animal Models of Epilepsy

In the United Kingdom, 450,000 people, or 0.5 to 1 percent of the population, suffer from epilepsy, with approximately 30,000 new cases diagnosed each year. One-third of patients do not respond to any of the currently available drugs and 20 to 30 percent do not improve with surgery. In the developing world, 60 to 90 percent of epilepsy is undiagnosed or untreated.

A variety of in vivo and in vitro models of epilepsy exist, including spontaneously induced epileptic mouse strains, chemically or physically induced models, and cultured neurons. Woodhall’s research relies on a long-established technique called lithium-pilocarpine epileptogenesis, which uses a chemical insult to provoke development of epilepsy over an extended period of time and results in a chronic epileptic syndrome in an animal. Brain slices are then obtained from the animals for testing. After injection of the drugs, the rodent goes into acute status epilepticus defined as continuous seizures with very short gaps in between. In many laboratories, this phase is allowed to continue anywhere from 90 minutes to 6 or 7 hours, Woodhall said. Seizures are then arrested with a sedative. The animal enters a quiescent period that lasts 1 or 2 weeks before they begin to exhibit spontaneous recurrent seizures. A conservative estimate of mortality from this approach is 5 to 50 percent; however, in some laboratories mortality rates are more than 80 percent, Woodhall noted. This extremely high mortality rate prompted Woodhall to focus on how this model could be refined and survival improved.

Questions persist as to whether these models are good models of temporal lobe epilepsy, or of epilepsy in general, and whether the pathology is similar to that seen in humans. There are also concerns about reproducibility, as measurement of key indicators can be highly variable. For example, γ-amino butyric acid (GABA)–mediated levels, an indicator of inhibitory action, in this model have been shown to decrease, increase, or remain unchanged. In addition, Sloviter (2005) showed that when animals are allowed to remain in acute status epilepticus for 6 or 7 hours, large areas of hemorrhage and damage were visible in brain slices. This raises questions about the seizures the model elicits, specifically whether these seizures as a result of gross global damage are a true model of human epilepsy, said Woodhall.

Refinement

Woodhall raised several questions regarding refinement of the current epilepsy model: whether the severity of this approach can be reduced; whether acute status epilepticus can be avoided altogether; whether more “ethical value” can be gained from the model; and whether other approaches could be used.

Seizure activity feeds from the cortex, through the basal ganglia, and back into the cortex, to create a positive feedback loop during epileptogenesis. Seizures then become uncontrolled and spread to the brainstem, killing the animal, Woodhall explained. Use of the central muscle relaxant, xylazine, reduces the intensity of the seizure activity and instantly reduces mortality rates. The other critical point in the process is arrest of the seizures. A massive dose of diazepam is currently used, which can stop the heart. Instead, a cocktail of very low doses of synergistic drugs, acting at different receptor systems, can more controllably terminate the seizures, Woodhall explained. It turns out, he said, that acute status epilepticus can be avoided. To make the most of the model, Woodhall identified several ways to increase the use of the fragile brain slices obtained from animals. Enlisting multiple researchers on one day to extract as much data as possible from each individual rat reduces the number of animals needed during the experiment. Methods for production and storage of slices were also improved. Together, these refinements led to development of a new model, low-dose lithium-pilocarpine-xylazine epileptogenesis, with a very brief period of acute status epilepticus, much longer quiescent period, and less than 2 percent mortality. The new model, which mimics the unique features of pediatric epilepsy, was validated using brain slices from children who had surgery for intractable epilepsy, Woodhall noted. In addition to refining the models themselves, Woodhall said that data sharing among researchers is another aspect of refinement and overall reduction as well.

Refinement presents some challenges, Woodhall noted. The new epilepsy model, for example, takes longer to achieve recurrent seizures and is therefore more expensive, and there is more variability. Woodhall concurred with Barnett that it can be challenging to publish refinements to methods that have been broadly used for decades.

Can Systematic Reviews Assist with the 3Rs?

Systematic reviews could also assist with replacement, refinement, and reduction, Murphy suggested. Preclinical systematic reviews could potentially:

• Replace animal use by
  • providing evidence of the validity of studies by comparing in vitro, invertebrate, or in silico data with data from traditional animal studies.
• Refine experimental procedures by
  • highlighting how differing methodologies affect measures of efficacy.
  • providing a platform for setting a standard for the methodology of a particular model and unifying the reporting requirements.
  • providing evidence of the effectiveness of refinements.
• Reduce the ineffective use of animals by
  • avoiding duplication, preventing further studies of ineffective interventions.
  • providing a more precise estimate of treatment effect, thereby informing future power analysis.

Systematic reviews do have weaknesses, however, noted Murphy. The value of the review for the development of therapeutics for humans depends on the quality of included preclinical studies. Systematic reviews can become outdated rather quickly and must be regularly updated as new data become available. This requires some sort of repository or electronic ware-house for the data so that modifications can readily be made. Systematic reviews can still be susceptible to bias in the selection of studies, especially if the predefined rules are not followed. Finally, there is the challenge of obtaining unpublished data; in particular, negative data are difficult to collect.

Preclinical Studies

A fundamental problem with the use of animals in research is that efficacy in animal models of disease does not necessarily equal efficacy in humans, Murphy noted. Many compounds come through animal studies only to fail in the clinic. Clinical trials fail for a variety of reasons. For example, they may be underpowered or they may underestimate the variability of the endpoint measures, leading to inconclusive results. However, sometimes the treatment regimen for humans differs from that of the animal model. In stroke, for example, the majority of preclinical data suggested a short therapeutic window; however, it generally is not possible to see a patient within 15 minutes after the start of a stroke. This, Murphy suggested, is one of the reasons that many stroke compounds have failed in the clinic.

Some animal studies also have methodological bias. As an example of empirical bias in the design of experimental stroke studies, Murphy noted that studies are generally done in young, healthy, male animals, while humans who have strokes are generally older with comorbidities (Crossley et al., 2008).

The quality of preclinical studies is highly variable. A recent survey found that 40 percent of 271 randomly chosen articles did not state a hypothesis or objective, or the number and characteristics of animals (e.g., species, strain, sex, age, weight). The survey also found that more than 85 percent of studies did not report randomization or blinding and 30 percent did not report statistical methods (Kilkenny et al., 2009). Study quality influences measures of efficacy. The assessment of study quality is an inherent part of a systematic review, Murphy noted.

Murphy suggested that preclinical systematic reviews could help address some of these issues. A systematic review by Perel and colleagues (2007) comparing treatment effects in animal experiments and clinical trials found systematic reviews of preclinical data could identify low-quality animal studies and better predict success or failure of compounds in clinical trials.

Integrating Systematic Review into Preclinical Translational Research

In summary, Murphy said, systematic review could be applied to preclinical data in order to improve the overall quality and value of animal studies, support the 3Rs, and inform clinical trials. The path to implementation of systematic reviews as a matter of routine potentially includes the Food and Drug Administration, pharmaceutical companies, research institutions, and publishers.

Murphy suggested two strategies for supporting systematic reviews of preclinical research. The first is to raise awareness of the power of applying systematic reviews to animal studies. Conducting reviews can inform and improve the timing, design, and quality of studies and better inform subsequent clinical trials. The second is to secure support from publishers and journal editors. Access to useful data might increase with more rigorous application of requirements for publication and rejection of low-quality or incomplete studies. In addition, the support for the publication of negative data would enable increased sharing of primary data, regardless of the outcome.

Refinement Stemming from Target Validation in Humans

In drug development, animal research plays a role in target validation, screening of compounds to optimize pharmacokinetics and efficacy, and safety and toxicology testing. Advances in technologies, however, are enabling increased target validation in man, potentially reducing the need for animals. Studies of the genetics of rare diseases, imaging studies, genome-wide association studies (GWAS), pharmacogenomics, and stem cell research are informing industry decisions to pursue particular drug targets.

For example, researchers are modeling schizophrenia using human-induced pluripotent stem cells, identifying new pathways and potential drug targets that have not been previously associated with schizophrenia (Brennand et al., 2011). Studies of mutations in individuals with rare diseases or isolated syndromes who exhibit a gain or loss of function also can help focus drug discovery efforts. For complex disorders involving multiple genes, GWAS are beginning to cluster pathways, identifying convergent nodes on these pathways that may be important in terms of disease progression.

While increased target validation in humans is unlikely to replace animal models, it will allow refinement of the questions asked of the models, Hunter said. For example, knowledge from human validation studies could lead to an increased focus on models of mechanism, rather than models of disease. Refined models could also help identify unwanted target-related effects, allowing a target to be invalidated early in the process.

Animal Models of Mechanism Versus Models of Disease

Few animal models faithfully represent the full complexity of the disease being modeled. This is especially true for nervous system disorders and diseases, for which animal models are limited in predicting drug efficacy, Hunter noted. Animal models may provide conflicting data in terms of exposures of drugs required, and result in false negatives. As a result, products are frequently tested in multiple models.

Hunter suggested a need to move toward more mechanistic models. Increasing disease knowledge allows for better identification of key mechanisms. The focus then should be on developing mechanistic in vivo assays that can be translated to humans. Such assays could demonstrate compound effects on the mechanism, define the exposures required for efficacy on the mechanism, and allow comparison of pharmacodynamics with pharmaco kinetics. This could lead to a reduction in the number of models and experiments needed, Hunter opined.

This approach requires a different mindset, Hunter said. For progression into human trials, if a molecule works in an animal model of disease, it is often necessary to show that it works in several models. On the other hand, if a molecule works on a particular mechanism, only one experiment may be needed to take it forward.

Precompetitive Collaborations

The current economics of drug development are not sustainable, Hunter commented. One approach to help move discovery forward is the concept of precompetitive collaborations. A number of efforts are underway globally to share more data and information. One example is the Innovative Medicines Initiative (IMI),¹ a public–private partnership between the European Federation of Pharmaceutical Industries and Associations and the European Union. Large consortiums facilitated by IMI share information on existing animal data, developing new models, and standardizing models across different companies and institutions. The NEWMEDS Consortium, for example, is working to develop both new preclinical models and translational experimental medicinal models for schizophrenia and depression.

In summary, Hunter stressed that advances in technology and creative approaches to precompetitive collaboration and data sharing are providing real opportunities to refine animal models.