Enhancing the Ability of Experimental Autoimmune Encephalomyelitis to Serve as a More Rigorous Model of Multiple Sclerosis through Refinement of the Experimental Design

Abstract

Advancing the understanding of the mechanisms involved in the pathogenesis of multiple sclerosis (MS) likely will lead to new and better therapeutics. Although important information about the disease process has been obtained from research on pathologic specimens, peripheral blood lymphocytes and MRI studies, the elucidation of detailed mechanisms has progressed largely through investigations using animal models of MS. In addition, animal models serve as an important tool for the testing of putative interventions. The most commonly studied model of MS is experimental autoimmune encephalomyelitis (EAE). This model can be induced in a variety of species and by various means, but there has been concern that the model may not accurately reflect the disease process, and more importantly, it may give rise to erroneous findings when it is used to test possible therapeutics. Several reasons have been given to explain the shortcomings of this model as a useful testing platform, but one idea provides a framework for improving the value of this model, and thus, it deserves careful consideration. In particular, the idea asserts that EAE studies are inadequately designed to enable appropriate evaluation of putative therapeutics. Here we discuss problem areas within EAE study designs and provide suggestions for their improvement. This paper is principally directed at investigators new to the field of EAE, although experienced investigators may find useful suggestions herein.

Introduction

A need for better therapeutics for multiple sclerosis.

Multiple sclerosis (MS) is a progressive neurologic disorder of the CNS that results in motor, sensory, and cognitive impairment. There are two distinct disease processes: acute relapses and progression of disability.¹⁹ Clinical presentation of MS is variable and can be classified broadly as relapsing–remitting, primary–progressive, and secondary–progressive. Although therapeutic options are available that can decrease the frequency of relapses in the relapsing–remitting form of MS, the disease still progresses, and its long-term effects are not prevented. Furthermore, there are no effective treatments for the more progressive forms of MS. An incomplete understanding of the disease mechanisms contributes to the inability to develop treatments that effectively slow, much less prevent, the long-term effects of MS. Therefore, continued, comprehensive, pathophysiologic study of MS is necessary to assist the development of new and enhanced therapeutics.

The development of an animal model of MS.

Over the past decade, MRI studies and the analysis of autopsy tissue from MS patients have made invaluable contributions to the field and revolutionized how the disease is viewed. For example, studies establishing that axonal and neuronal degeneration are key pathologic components have helped refocus the view of MS as a chronic degenerative disease.¹³ Although important contributions have come from the study of MS subjects, much of the current knowledge about this disease is based on research involving animal models.

The most commonly used animal model is experimental autoimmune (or allergic) encephalomyelitis (EAE). Historically, EAE was observed after the use of Pasteur's rabies vaccine in humans. Some subjects showed signs of encephalitis and polyneuritis—symptoms not usually associated with vaccinations. In addition, demyelination was present near blood vessels in the CNS.⁶ Ensuing studies found that myelin antigens in vaccines containing spinal cord or brain material were responsible for triggering an immune response that was directed against myelin in the CNS of the recipient.¹⁶⁴ A more complete history of the chronologic events leading to the development of EAE as a model of MS recently has been presented.⁶

Currently, a variety of methods are used to induce EAE including the injection of an encephalitogenic myelin protein, an encephalitogenic peptide of a myelin protein, or injection of spinal cord homogenate. The adoptive (or passive) transfer of spleen and lymph node cells from an immunized animal into a naïve animal can also induce the disease. Today the model is used for a wide array of studies including those centered on pathogenesis, therapeutic interventions, immune responses, stress responses, genetics, cellular repair, and other areas.

Is EAE a good model of MS?

The value of EAE as a model of MS has been discussed since shortly after its inception and remains an active topic of debate.^{55,61,160,163-165} Some researchers contend that EAE is not a suitable research model because of its inability to mimic some of the pathologic, immunologic, and chronic features of MS. For example, in a review in 2005,¹⁶⁰ the authors discuss how EAE tends to be an acute, monophasic disorder whereas MS tends to display chronic relapsing characteristics. This difference suggests that EAE may be more suitable for studying acute demyelinating encephalomyelitis than it is for studying MS. They¹⁶⁰ mention that radiologic and pathologic data from patients with MS should direct researchers to investigate issues other than the autoimmune components that appear to be foremost in EAE. These authors point to histologic and MRI studies that have shown considerable axonal and cortical damage in MS but not in some models of EAE.¹⁶⁰

Some pathologic features that are present in various EAE models are not apparent in MS. For example, the extravasation of red blood cells into the CNS of SJL mice^53,144 and Lewis rats⁶⁶ with EAE is not a typical pathologic feature associated with MS. Moreover, encephalitogenic regions associated with myelin basic protein (MBP) or proteolipid protein (PLP) induce more CD4⁺ than CD8⁺ T cells in EAE, but inflammatory MS tends to contain more CD8⁺ T cells.^17,69,96 Comparison of gene expression profiles between EAE and MS tissues suggests greater alteration of immunologically relevant genes in EAE than in MS but greater alteration of CNS genes in MS than in EAE.²⁵ This finding led to the view that EAE is an immunologic disease involving relatively healthy brain and spinal cord tissue, whereas MS involves brain and spinal cord tissue that is functionally altered. The EAE model has been proposed to represent a reductionist approach that inevitably leads to exclusion of relevant features of MS.⁶ For example, inducing EAE with MBP fails to account for other myelin proteins that may play a role in MS. Furthermore, critics note that many interventions that successfully ameliorate EAE do not have similar value for treatment of MS.^145,160

Due to the lack of efficacy in predicting effective MS treatments, as well the variety and complexity of the EAE models, some authors suggest that EAE cannot be relied on for drug screening for MS treatments.¹⁴⁵ In fact, some treatments that have benefit in EAE actually worsened MS. For example, IFNγ showed promise in EAE by protecting against chronic–progressive EAE,⁵⁶ but this compound exacerbated disease in clinical trials with MS patients.¹²⁹

Benefits of the EAE model include: it has similar clinical (for example, paralysis, incontinence) and pathologic (for example, demyelination, infiltration of inflammatory cells into the CNS) traits to MS; it enables testing of experimental treatments without subjecting humans to compounds that have an inherent risk of toxicity; it can be used to perform detailed studies on pathogenesis (for example, animals can be euthanized at different stages of disease); and it can be used to perform mechanistic studies not possible in humans (for example, antibodies against galactocerebroside,¹¹⁵ sulfatide and sphingomyelin⁸² are present in sera or cerebrospinal fluid of MS patients, and administration of a sulfatide-specific antibody to EAE animals was found to worsen disease⁸²).

Studies in animals have led to the advancement of the understanding of the immunopathogenesis in MS. For example, recent studies in mice have revealed that Th17 cells, which are IL17-producing CD4+ T cells that are distinct from Th1 and Th2 cells, play a crucial role in the pathogenesis of EAE,^87,95,131 and there is a reciprocal pathway between the development of pathogenic Th17 cells and protective regulatory T (Treg) cells.¹⁰ Although data are accumulating implicating Th17 cells in the pathogenesis of MS,^3,113 there are differences in the development and role of Th17 cells between animals and humans. For example, IL6 and TGFβ initiate the differentiation of Th17 cells in mice, but these factors are ineffective for the induction of these cells in humans.²⁴ Furthermore, the development of Th17 cells can be inhibited by IFNβ⁶⁵ and IFNγ^67,131 in mice, and these IFNs can suppress EAE,^56,181,196 but only IFNβ is effective for the treatment of MS, whereas IFNγ exacerbates disease.¹²⁹ Some discrepancies also exist for Treg cells between EAE findings and the predicted outcomes regarding these cells in MS subjects. However, many of the human studies¹²⁵ examined Treg cells derived from the blood, which may show key differences from those derived from the site of inflammation (that is, the CNS).

Despite several attributes of EAE, caution needs to be exercised before moving to studies on patients with MS, because toxicities for some compounds may not be revealed until human studies are performed. For example, natalizumab (Tysabri) was developed to treat MS after showing promise in EAE (discussed in reference 165). However, in an unforeseen complication, three patients developed progressive multifocal leukoencephalopathy (PML).⁸⁰ The animals used for the EAE studies on natalizumab are not susceptible to PML, which raises the possibility of performing additional screens on models of PML susceptibility before testing in humans.¹⁶⁵ Another example is Linomide, an orally administered immunomodulatory quinoline 3-carboxamide, which showed promise in EAE and phase II clinical trials in MS but eventually presented an unacceptably high rate of myocardial infarction.^124,152 Despite these shortcomings, EAE has been used to make valuable contributions. For example, of the 6 main FDA-approved medications for treating MS, 3—glatiramer acetate, mitoxantrone, and natalizumab—were pursued after showing signs of success in EAE models.¹⁶⁵

Explanations for why treatments that ameliorate EAE fail to improve MS include incorrect use of the model and inaccurate interpretation of data. Indeed, a recent study has indicated that one reason for poor translation of findings from animal studies to human interventions is the use of flawed study designs that can lead to erroneous conclusions;¹³⁵ this concern has been raised for the EAE model.¹⁶ We concur with this position. To increase the usefulness of the information obtained from EAE studies, we contend that investigators must 1) have a thorough understanding of the limitations and benefits of the EAE model and 2) optimally design experiments so that the data are meaningful and conclusions are pertinent to MS.

Designing Studies to Maximize the Information Obtained from an EAE Study

Many studies use EAE animals to investigate the expression of genes of the immune response and the roles of their products, whereas other studies serve to use EAE specifically as a model of MS to test compounds for therapeutic value or to gain insights into disease mechanisms. Here we focus on using EAE specifically as a model of MS.

Having a well-defined research question or hypothesis is a first step for designing a study that will provide optimal conditions to relate outcomes to MS. In the following subsections, parameters that can influence the outcome are identified and suggestions, which may particularly be useful for the new investigator, are provided for developing an optimal study design. In addition to the topics discussed below that apply to the EAE model, an overview of general issues relative to study designs involving animals has been published in the Institute for Laboratory Animal Research Journal.⁷⁸ This journal issue includes articles that cover appropriate controls, factors accounting for variability, group assignments, and statistics, as well as other topics that can affect the outcome of a study.^{31,48,49,58,74,81,157}

Selecting an EAE model.

Various presentations of pathologic lesions and clinical signs that represent an acute monophasic, relapsing–remitting, or chronic–progressive disease can be generated in EAE models, depending on the species and method of EAE induction that are selected.^6,60,115,178 In addition, EAE can be used to study the different arms of the immune response as it relates to MS (Table 1). Because recent reviews^6,61 address various EAE models across different species, we only briefly discuss models in mice (due to their popularity) and nonhuman primates (due to their proximity to several human traits).

Table 1.

Examples of immunologic components studied in murine EAE models

Immune component	Murine straina	Encephalitogen	References
CD4+ T cells	B10.PL	MBPAc1-9	143
	Biozzi AB/H	PLP56-70	133
	C57BL/6	MOG35-55	39, 159
	SJL	PLP139-151	12
		PLP178-191	137
		MBP87-99	72
CD8+ T cells	B10.PL	MBPNac1-9 and adoptive transfer	79
	C3HeB/FeJ	MBP79-87 and adoptive transfer	75
	C57BL/6	Adoptive transfer with T cells from MOG35-55-immunized mice	52, 168
		Active EAE with MOG35-55	52, 159, 168
		Active EAE with MOG37-50, MOG37-46, or MOG42-50	52
Treg cells	B10.PL	MBPAc1-11	99, 109
	CBA x C57BL/6 - F1	MOG35-55, MBP37-47, whole MBP, or adoptive transfer	70
	C57BL/6	MOG35-55	99, 188, 197
	SJL	PLP139-151	47
Th17 cells	C57BL/6	MOG35-55	65, 169, 197
	SJL	PLP139-151	73
B cells/antibodies	B10.PL	MBPAc1-11	109
	BALB/c	PLP180-199 or PLP185-206	105
	C57BL/6	MOG35-55 and adoptive transfer	183
Dendritic cells	CBA x C57BL/6 - F1	MOG35-55, MBP37-47, whole MBP, or adoptive transfer	70
	C57BL/6	MOG35-55	98, 195
	SJL	PLP139-151	195
Microglia or macrophages	B10.PL	Adoptive transfer of MBP T cells	141
	Biozzi AB/H	Spinal cord homogenate	57
	C57BL/6	MOG35-55	156
	SJL	Passive transfer of MBP reactive T cells	177
Mast cells	W/Wv	MOG35-55 and adoptive transfer	63
Complement	B10.D2/nSnJ	Guinea pig myelin	27
	C57BL/6	MOG35-55	170, 183

^aKnockout, congenic, or transgenic strains are not always denoted.

The administration of MBP to B10.PL mice¹¹⁸ or the adoptive transfer of T-cell clones specific for MBP into mice^45,120,198 results in chronic relapsing EAE. Typically, myelin oligodendrocyte glycoprotein (MOG) induces relapsing EAE in SJL¹⁸⁰ and Biozzi ABH² mice, but when given to C57BL/6 mice, MOG induces chronic progressive EAE.¹¹⁵ Exposing mice to UV light immediately before subcutaneous injection of MOG peptides in complete Freund adjuvant results in a secondary progressive model.¹⁷⁹ Injection of SJL mice with an encephalitogenic peptide of PLP or adoptive transfer of PLP peptide specific T cells induces a relapsing-remitting disease.¹¹⁴ Spontaneous relapses and remissions also occur in SJL mice injected with a mixture of mouse spinal cord, pertussis vaccine, and complete Freund adjuvant containing Mycobacterium tuberculosis. These mice have a pathologic profile of both acute and chronic inflammatory demyelination.¹⁰⁴ However, this model requires approximately 6 mo, on average, for the initial signs of disease to develop,¹⁰⁴ making for a long, drawn-out experiment.

A model that has spontaneous disease induction, which may be similar to what occurs in MS, has been generated.³⁸ This model uses a humanized transgenic mouse that strongly expresses HLA-DR15 with an MS T cell receptor. Not only does EAE develop spontaneously, but the model also replicates many features of MS, including paralysis, demyelination, and axonal degeneration.

EAE in mice has been used to study specific features of MS. For example, vertigo can be a symptom of MS, and a spontaneous form of EAE in transgenic mice leads to a high degree of inflammatory cell infiltration into the brainstem and cerebellum resulting in head tilt and spinning.^93,189 In another example, knock-in techniques were used to create a TCR-MOG × IgH-MOG double-transgenic strain of mice that exhibit a spontaneous model known as opticospinal EAE, which is very similar to the Devic variant of human MS.⁸⁹ Because plaques in Devic disease are restricted to the optic nerve and spinal cord, this strain originally was thought to best model neuromyelitis optica,^9,89 but the model may prove more useful in studies of MS.¹⁴⁶

The nonhuman primate model (for example, common marmoset Callithrix jacchus) of EAE has particular relevance to MS, because the marmoset immune system is similar to that of humans, and when whole white-matter or MOG is used as the encephalitogen source, a chronic– relapsing remitting model develops that closely approximates MS.⁵⁹ In addition, the brain lesions in the marmoset model appear similar to MS, including axonal damage,¹⁰⁸ and lesions can be readily monitored by using comparable tools (for example, MRI) as for MS.¹⁸²

NonEAE models provide useful information relative to MS.

In addition to EAE, other animal models can be used to characterize the disease or recovery process more completely. Theiler murine encephalomyelitis virus has been used to advance the understanding of the mechanisms of demyelination related to MS.⁴⁴ The validity of a viral model of MS is supported by epidemiological evidence that suggests the possibility that a virus may play a role in, or even be an etiologic factor for the development MS.⁹⁰ Theiler disease has two common, yet fairly distinct forms: one causes an acute attack on the immune system that is usually fatal, and the other (Theiler original) initially causes an acute infection of the gray matter and later a chronic demyelinating condition of the white matter.¹²⁶ Although not all strains of mice are susceptible to infection with Theiler murine encephalomyelitis virus, some researchers suggest that the chronic disease may be the best available model of MS.¹²⁶ Other relevant viral models include the murine coronavirus-induced demyelinating disease,⁹⁴ certain strains of the murine hepatitis virus,¹¹² and canine distemper virus.¹⁸⁶

In addition to viral-induced demyelination, various chemically induced models of demyelination have been studied. The models can share specific features of the proposed pathophysiology of EAE or MS and have a very predictable timeframe of lesion development. In addition, remyelination can be studied when the offending agent is removed from the system. These models use metabolic disruptors or gliotoxic agents, like cuprizone, ethidium bromide, and lysolecithin.^{15,111,121,150}

Pharmacology versus genetics.

Depending on the approach (for example, pharmacologic, transgenic, knockout), the results from the study can lead to different interpretations regarding the role of a molecule in disease. Studies that have examined the role of TNF in EAE are a good illustration of how different methods can lead to different outcomes, and the TGN1412 story provides a good illustration of the limitations of pharmacologic approaches in animal models.

Administration of TNF reduced the duration and severity of pathology in both TNF^−/− and TNF^+/+ mice with EAE,¹⁰³ suggesting that treatment with TNF may limit pathology in MS patients. However, when TNF was expressed retrovirally by means of MBP-specific T-cells, which were adoptively transferred into mice, EAE severity was exacerbated.²⁸

The use of antiTNF factors has shown more promise than TNF in EAE. Treatment with type 1 soluble TNF receptor in SJL/J mice with EAE induced by adoptive transfer lessened disease and protected the subjects from reacquiring EAE.¹⁵⁴ Similar to the soluble receptor, the use of an antiTNF antibody ameliorated EAE in SJL/J mice.¹⁵⁵ Furthermore, when overexpressed in mice transgenic at the MBP promoter, TNF caused an increase in the severity of EAE compared with that in nontransgenic controls.¹⁷³ However, the use of TNF-deficient mice showed that TNF is not required for demyelination in EAE⁸⁸ but clarified its involvement in disease initiation.^83,88,167

In MS clinical trials, a therapy designed to suppress TNFα was unsuccessful, and in some cases the therapy actually worsened MS.^185,191,192 Furthermore, it has been shown that the use of TNF antagonists, for example, adalimumab and etanercept used to treat rheumatoid arthritis, may actually induce MS in patients.^1,8

The fiasco involving the compound TGN1412 highlights how differences in a drug's mechanism of action between humans and animals can lead to erroneous conclusions with potentially dire consequences. Preclinical studies conducted in rats with JJ316, a superagonist of CD28, revealed that JJ316 ameliorated EAE by inducing Treg cells, which acted to suppress the immune response.¹¹ TGN1412 was identified as the functional equivalent in humans to JJ316 in rats, but unlike results in the rat, TGN1412 caused a massive cytokine storm, resulting in multiorgan failure in 6 volunteers (reviewed in reference 161). When tested in nonhuman primates, TGN1412 elicited only modest levels of proinflammatory cytokines despite higher doses than those tested in humans. Therefore, results from rats and monkeys both failed to predict the outcome in humans, that is, massive cytokine storm. Differences in receptor structures and disparate molecular mechanisms between animals and humans argue for a model that more accurately reflects the responses in humans, and attempts have been made to ‘humanize’ the immune system for EAE models.^38,76

Variables within subjects affect the disease course.

Sex and age of EAE subjects can influence the disease course. For example, among B.10.S × SJL/J F₂ intercross mice with EAE, females were more likely to develop brain lesions than males.¹⁷⁵ SJL and NZW female mice had a higher incidence of EAE after immunization, but in B.10/PL and PL/J strains, males developed more severe EAE than females.¹³⁰ Of note, women are more than twice as likely as men to develop MS, but men tend to develop more severe cases.³⁷

EAE severity and susceptibility increase with age in mouse models,¹⁷⁵ and increasing age in MS patients is correlated to disease severity.¹⁰¹

Environmental factors affect the disease course.

Because they can influence experimental outcome, environmental conditions should be kept consistent during an experiment, as well as between experiments within a study. The season during which experiments are done can affect disease susceptibility.^175,176 Disruption of the mother–infant bond (for example, separating neonatal rodents from their mothers) resulted in increased disease activity or susceptibility.^{32,92,110,174} Early weaning increased the severity of EAE,⁹² whereas gentling decreased susceptibility.⁹¹ In addition, sound stress increased disease susceptibility.³²

Exposure to different types of bacteria rendered mice refractory to disease induction,⁹⁷ whereas disease could be reactivated in EAE animals exposed to various enterotoxins.¹⁵¹ In other autoimmune conditions, the gut flora (or lack thereof) influenced susceptibility to disease.^14,134

Selecting and assigning animals to study groups.

Animals within an experiment should have common characteristics (that is, strain, sex, and age), unless the research question addresses 1 of these factors. If examining a transgenic or knockout strain, where low numbers of animals are obtained, it may be necessary to include both males and females, but the ratio of sexes should be maintained between control and experimental groups, and evaluation of data for each sex as well as both sexes should be performed.

Consideration of the kind, size, and direction of the effects expected in the study will allow planning for an adequate number of animals to detect those effects (adequate power). If the research question proposes effects in only 1 direction, then a 1-tailed statistical test, which often requires fewer animals than does a 2-tailed test, may be appropriate.

Assignment of an animal subject to a group (for example, EAE–drug, EAE–vehicle, nonEAE–drug, nonEAE–vehicle) should occur through a randomization process, and scoring of animals should be performed in a blinded manner (see Ensuring correct collection of data). Studies on other models that did not use randomization or blind scoring had substantially more positive results, i.e., detected differences between groups, than did studies that used these procedures.⁷ Sometimes the randomization procedure needs to be designed to help ensure that the starting groups are comparable. For example, if the weights of animals vary widely, stratifying the animals based on weight before the random selection process may be helpful. If the animals are to be divided between 2 groups, then the 2 heaviest animals would be randomly selected to go to group 1 or group 2, and the next 2 heaviest animals would be randomly selected, and so on. If both sexes are to be used, females can be divided from males, and then the animals for each sex can be assigned randomly to groups.

Methods of disease induction.

The induction of EAE is based on the administration of an encephalitogenic emulsion that initiates a T cell-mediated immunity against 1 or more myelin antigens. The type of MS being modeled will direct the selection of the myelin antigen, animal species, and strain. Myelin-related peptides or proteins should be solubilized in buffer and mixed with an adjuvant, by using a homogenizer to create a unified emulsion. Usually complete Freund adjuvant (or incomplete Freund adjuvant with the addition of a tailored dose of triturated M. tuberculosis) is used to promote the immune response. Once the emulsion is prepared, the experimental animal is injected subcutaneously. Historically, hindpaw injections were used under the premise that they afford high access to the draining lymphatics. However, the current standard for veterinary care does not accept hindpaw injections, due to the determination that an injection of a noxious material, like complete Freund adjuvant, into a sensitive area, such as the hindpaw, leads to prolonged pain or distress. Current EAE models typically use dorsal injections of encephalitogen at the nape, thorax, or above the base of the tail. The clinical manifestations associated with EAE usually develop within 1 to 2 wk after administration of encephalitogen.

Usually intravenous or intraperitoneal pertussis toxin is coadministered on the day of encephalitogen administration, and 1 or 2 boosters often are given over the following days. Pertussis toxin amplifies the immune-response–promoting pathology.^23,71 Pertussis toxin also was thought to act by increasing the permeability of the blood–brain barrier, but this view has been questioned.¹²² Because pertussis toxin stimulates the immune response and may increase the number of lymphoid cells in the CNS, it is imperative that nonEAE controls receive pertussis toxin injections, as well as the initial injections of emulsion but without the encephalitogen, to ensure that observed effects are due to EAE and not to a nonspecific reaction to the ancillary components that are used to facilitate disease induction.

Adoptive or passive transfer of EAE can be achieved through a variety of methods. For example, T-cell clones or T-cells sensitized to a myelin antigen can be injected into naïve or vaccinated recipients.^107,166 Alternatively, spleen or lymph node cells that have been collected from immunized animals and then stimulated in vitro with antigen or mitogen can be injected into naïve or vaccinated recipients.^62,107

Selecting the route of drug administration.

Systemic administration of a drug can be achieved by means of injection, drinking water, gavage, food, or patch. Injection is a straightforward method for controlling the amount of a drug that is administered but the route of injection can make a difference to drug availability in the body. For example, intravenous into the mouse tail vein can be difficult to achieve and, in situations of repeated administration, damage to the tail vein can occur. Intraperitoneal injection can overcome this problem, but the drug will undergo first-pass biotransformation through the liver, which can reduce the bioavailability of drugs that are metabolized rapidly. Subcutaneous injections avoid this effect but may not result in the entire dose of some compounds to enter into the circulation.

Systemic drug administration can be achieved by the oral route, which is the preferred mode of delivery for compounds in humans and therefore warrants examination in EAE when suitable drugs are being tested. Dissolving compounds in drinking water or grinding the agent into powdered chow (particularly for poorly soluble compounds) may be advantageous for some chemicals. However, administering an agent in drinking water or food prevents direct control of how much drug the subject takes in, and therefore, how much drug is delivered to the blood stream. This drawback can be a significant concern as EAE progresses, because motor and behavioral signs can result in loss of appetite or anorexia.¹³⁸ In addition, administration by food or water may not achieve periodic peak plasma concentrations of drug, which could be an important component of the mechanism of drug action in humans following an injection or pill. Gavage can be used to overcome some of the shortcomings of drug delivery by drinking water or food, but similar to the situation with intraperitoneal injections, first-pass metabolism is a concern.

Topical application of an agent may be advantageous when a drug can be delivered by a cream, ointment, or patch. The drug must have good contact with the skin to achieve adequate absorption and be placed such that the animal subject or a cagemate cannot remove the compound by licking or scratching. Covering the application with plastic wrap may protect the delivery system, and applying it to abraded tissue may facilitate drug absorption.^21,149

Intracerebroventricular injection enables testing of the direct effects of an agent on the pathophysiologic processes affecting the CNS and is useful for agents that are expensive or difficult to obtain, because smaller amounts of drug or injection volumes can be given for a more concentrated effect. In addition, this route would be appropriate for a chemical that does not cross the blood–brain barrier. However, intracerebroventricular injection has several drawbacks: it requires anesthesia of the animals, it is traumatic to the brain, and it usually is used to deliver only a single dose, unless an intracerbroventricular cannula is installed that can be fed by a tube connected to an osmotic pump implanted under the skin.^142,187

Preparing drugs for administration.

Methods for the preparation and storage of drugs can influence efficacy. For drugs that are chemically stable in their vehicle and not prone to degradation after a freeze–thaw cycle, stock solutions can be made, aliquoted, and stored. Then working solutions can be prepared for each administration. For agents that are unstable, fresh preparations are necessary for each administration.

The choice of vehicle depends on the chemical nature of the test agent and delivery route. Agents that are hydrophilic should be dissolved in a saline solution (for example, PBS pH 7.4). This method minimizes animal discomfort and is amenable to intravenous and intraperitoneal injection. Dimethyl sulfoxide is a vehicle that dissolves both polar and nonpolar test agents and easily penetrates cell membranes. Although the toxicity of DMSO historically has been controversial, its use in research remains popular due to its ability to dissolve a broad range of compounds. However, even using low doses can give rise to undesired biological effects, for example, induction of protein expression in the liver.⁸⁵ Depending on the route of administration, the mouse LD₅₀ for DMSO ranges from 3.1 g/kg (intravenous) to 50 g/kg (topical) (MSDS, Sigma-Aldrich).¹⁵⁸

Lipophilic compounds are amenable to the transdermal or subcutaneous injection route due to lipophilicity-enhancing absorption through the subcutaneous tissue and fat.^21,128 In cases where a test agent is insoluble in a biologically suitable vehicle, the appropriateness of further examination is questioned.

The constituents in a particular preparation can influence the effect of the drug it contains. For example, EAE was inhibited in mice given an intracranial injection of fibroblasts transformed with a retroviral vector to express IL10. However, intracranial administration of an adenovirus encoding IL10 or recombinant IL10 did not inhibit EAE, even though IL10 was still present in the CNS.²⁶

Drugs preparations can be made on the basis of either weight or molarity, and the route of administration influences which is selected. For example, if a drug must be mixed with powdered food, weight is usually the method of preparation. Basing dosage on weight also can be appropriate if only 1 drug is being tested or if extrapolation of the dose to other species is an objective. However, concentration based on molarity allows for comparison of the efficacy of multiple drugs with different molecular weights. Regardless of the preparation method, doses of agents should be given to the animals based on individual body weight (for example, mg/kg). This method is more accurate than assuming that all animals of a given age have the same body weight. However, during EAE animals can lose a substantial percentage (that is, as much as 20% or more) of their initial body weight as the disease progresses.^40,41 This disease characteristic raises a legitimate question as to whether the dose should be based on the initial body weight or adjusted to match the daily body weight. For long-term studies (that is, 2 mo or longer), the dose can be based on the body weight at the start of each week, whereas using a dose that is based on initial body weight could be considered for a shorter study.

Factors affecting the dosing of animal subjects.

Ideally, the basic pharmacokinetic characteristics of test agents are known or would be assessed before their use in the EAE model. These parameters include drug half-life, plasma protein-binding profile, metabolism (bioinactivation or bioactivation), and excretion rate.³³ All of these factors contribute to the effective plasma concentration of the drug. Knowledge of temporal alterations in plasma concentration facilitates development of an appropriate dosing schedule. For example, drugs that are cleared slowly are likely to achieve high plasma levels with repeated administration, whereas a drug that is cleared quickly will need to be administered more often to maintain an efficacious in vivo level. The need for repeated administration would disfavor study designs involving excessive drug injections and might indicate implantation of an osmotic pump to achieve required drug levels.

In addition, knowledge about the drug's metabolites is advantageous. If the metabolites are active, therapeutic effects will be maintained for prolonged periods of time, and dosing is adjusted accordingly.²¹ Further, studies testing the toxicologic profile of active metabolites may be warranted.

Pilot studies involving limited numbers of animals and testing a range of doses can be an effective way to identify candidate optimal doses. In addition, the results of pilot studies can be used ascertain the number of animals necessary to provide sufficient statistical power. Power calculations should be adjusted for formal studies that involve comparison of multiple dosages.

A compound that is effective at a reasonable dose and that has a suitable toxicologic profile may warrant future testing in humans. However, researcher need to consider safe dosing strategies for possible testing of new drugs in humans, for example, using doses in animal studies that can be realistically translated to humans. A 2005 FDA draft of dosing guidelines for testing studies provides conversion factors and an equation to derive the human equivalent dose for a drug based on the dose given to another species.¹⁸⁴ Although the need for a high dose to yield the desired effect (that is, disease suppression) may preclude future clinical use of a drug in humans, research of that compound can still provide useful insight into its mechanism(s) of action or the pathophysiologic processes at work in EAE. However, investigators should be cautious about extrapolating high-dose drug effects in animals to humans or advocating these particular agents as possible treatments.

Matching the dosing regimen to the study question.

The design of the schedule for the dosing regimen should take into account the proposed therapeutic mechanism relative to the pathogenic course. For example, if the drug is proposed to inhibit disease induction, dosing may begin just before, at the same time as, or shortly after encephalitogen administration. However, the test compound may prevent adequate exposure of the immune system to the encephalitogen, a situation analogous to suboptimal administration of the encephalitogen. Therefore, investigators should consider delaying the delivery of test compounds until 2 to 3 d after encephalitogen injections, when immune cells will have started to respond to the encephalitogen. Alternatively, studies can be performed by using the adoptive transfer model of EAE, in which spleen and activated T cells are extracted from EAE mice and cultured.^106,107,166 These cells could then be treated with the test agent prior to administration into a naïve subject, in which EAE development would be assessed to determine whether the drug inhibits passive EAE development. Instead, the test compound might be given to the donor animal prior to the collection of cells to reveal whether the agent suppressed the ability to transfer disease, or the recipient animal could be given the test compound after injection of donor cells.

If a drug is proposed to abate pathologic processes that directly lead to clinical signs or relapse, drug administration could begin after development of these features.¹¹⁹ However, pathology could be well underway once treatment starts because clinical evidence of disease follows the development of pathology, and clinical signs can progress rapidly in EAE. Therefore, drugs that require even a modest time to induce effects may not be assessed accurately when administration is delayed until the appearance of clinical signs. An alternative starting time can be selected that is representative of when the disease phase would begin. A disadvantage of this approach is that individual animals begin to show disease or relapse at different times, leading to treatment either before or after the onset or relapse of disease. Further, pathology can occur in animals that do not go on to develop disease.¹⁹⁴

Ensuring correct collection of data.

During initial testing of an EAE model, a nonEAE control group should be included. These animals receive all components of the emulsion except for the encephalitogen, any follow-up procedures (such as pertussis toxin boosters), similar handling as EAE animals, and injections of test compounds. These control animals will serve to determine whether a test drug acts to depress or increase the activity of normal animals and to establish baseline effects for assays of biochemical or pathologic markers.

Scorers should be blinded to any grouping of subjects, and each animal should have a group-independent identifier to allow tracking of scores. The process of clinical sign scoring should be applied equally to control and experimental groups and must be well-defined and consistently followed. Scorers should be trained with identical sets of instructions and show greater than 97% accuracy with the trainer. The characteristics of the clinical signs should be readily apparent, and each subject should be evaluated for approximately the same amount of time (for example, 2 to 3 min) with a minimal amount of handling, because subjects may interpret external stimuli as stressors and respond through activation of sympathetic responses or increased muscle tone.

Ideally 2 independent scorers should evaluate animals at each time point, and their scores should be averaged. In cases of large scoring discrepancies, the scorers could discuss what evidence guided their assessments and confirm that the standards for scoring were followed; an independent referee might then make a judgment. However, both scores, although discrepant, may be accurate (for example, an animal became fatigued due to exertion that followed the first but preceded the second scorer evaluation).

Because clinical signs can vary throughout the day depending on wake–sleep cycles,⁴² the time of scoring must remain consistent throughout the course of the experiment, and scorers should perform their evaluations as close together temporally as possible. Furthermore, a rapidly changing course of disease would argue for multiple scoring sessions during a 24-h period.

Methods for scoring clinical signs can affect the study outcome.

The design or selection of an appropriate scoring scale greatly influences the usefulness of data generated in a given study. A scoring system can range from a very broad evaluation of disease activity to a highly sensitive measure. In extreme cases, the scale may not accurately depict the clinical signs that are displayed and therefore would fail to adequately gauge disease activity. In the following sections, we offer several suggestions for optimizing the system used to score clinical signs.

Advantages and disadvantage of existing clinical scoring scales.

No single scale is used in the EAE field. Some of this variability reflects different disease profiles between various EAE models, however some scales include design features that increase their sensitivity as measures of disease activity. Existing scales vary in length from 5 to 9 intervals, with the most common being a scale containing 6 intervals (that is, scores of 0 to 5).⁵¹ Typically, a score of 0 represents a normal-appearing subject and the highest rank of the scale indicates a moribund state or death (an undesirable outcome due to ethical considerations and loss of tissue specimens). Even comparable scoring systems can yield differences in results between labs, given variability between animals, environments, and experimental procedures.

The analytical power of using the most suitable scale for a given study is considerable. For example, statistically significant differences between treatment and control groups of animals might be achieved with a sensitive scale (for example, scores of 0 to 8), whereas data collected from the same animals by using a cruder scale (for example, scores of 0 to 5) may prevent detection of significant differences. A scale with more discrete intervals typically has more power to detect statistical differences and is a more sensitive measure of mild disease activity than is a scale with fewer intervals.⁵¹ In a hypothetical example (Figure 1), the earliest clinical sign (which appeared on day 8 after encephalitogen injection) went undetected when a common 0-to-5 scale was used but was apparent with more sensitive scales. Using the same 0-to-5 scale, recovery to a normal level was identified at day 20, whereas the other scoring systems (with more intervals) still indicated low-level disease activity (Table 2; Figure 1). In addition to being unable to detect mild presentations of disease, the 0-to-5 scale was unable to distinguish between some levels of high disease activity. For example, the 0-to-5 scale yielded a score of 4 for both days 12 and 13, whereas the 0-to-8 scale differentiated the disease activity between these days with scores of 6 and 7, respectively (Figure 1).

Figure 1.

Daily scores of clinical signs from a hypothetical animal with EAE were generated by using five different scoring scales. Note the 0–5 point scale reveals that days 12 and 13 both have equivalent peak maximal disease activity, whereas all of the other scales show that day 13 has a greater level of clinical signs than day 12. In addition, the 0–5 point scale indicates that days 15–18 all have the same score (a score of 2), whereas the 0–8 scale has a decrease between days 15 and 16 and then equivalent scores for days 16–18. In contrast, the nonweighted and both weighted scales reveal that disease activity is progressively decreasing between days 15–18. Other differences are noted in the text.

Table 2.

Comparisons between different scales

	Day (after encephalitogen injection)
	7	8	9	10	11	12	13	14	15	16	17	18	19	20	Cumulative total
Nonweighted	0	1	4	9	13	14	16	12	9	8	5	4	3	1	99
Weighted factor 1	0	0.5	2.3	5	6.5	7	7.7	5.5	3.7	3.3	2	1.5	1.3	0.5	46.8
Weighted factor 2	0	0.5	3	7.17	10.3	11.5	13.5	10.5	5.34	4.67	2.67	2.17	1.5	0.5	73.6
0–5 scale	0	0	1	2	3	4	4	3	2	2	2	2	1	0	26
0–8 scale	0	1	2	4	6	6	7	6	5	4	4	4	1	1

Scoring scales are associated with several inherent problems. For example, the assignment of clinical signs to scale intervals is not necessarily based on an equivalent progression of steps. This characteristic is emphasized by the fact that although scale intervals are numerically equivalent, the progression from 1 clinical sign to the next is not made up of equivalent levels over the entire range of the scale. In addition, clinical signs do not always build upon one another; therefore, a clinical sign that is assigned a high value can skip a clinical sign assigned to a low value. For example, an advanced score (for example, hindlimb paralysis) may bypass signs assumed to be part of a milder presentation (for example, incontinence). Another inherent problem is that individual clinical signs that are represented in separate intervals can be due to a single pathologic lesion, and this presentation may overrepresent the pathologic profile of the disease. Alternatively, many small pathologic lesions or lesions in various CNS regions may not result in a measurable clinical sign due to an insensitive scale or failure to evaluate the sign that the animal presents.

An alternative method for scoring clinical signs.

A scoring strategy that minimizes some of the problems associated with the traditional scoring scales (for example, skipping over clinical signs) while maintaining or enhancing statistical power is one that scores each clinical sign separately (Table 3). In other words, instead of matching an animal's overall clinical profile to a point on a scale, clinical signs are individually evaluated and tabulated. This practice avoids the assumption that the appearance of a sign is dependent upon the presentation of one listed lower on a scale. Statistics then can be run for each clinical sign (for example, dragging 2 hindlimbs) or for category of signs (for example, gait; Table 3). The number of days a subject has a clinical sign over the duration of the study, or a portion thereof (for example, clinically active period, relapse) can be used for statistical comparisons. In addition, the various clinical signs within a group of animals (for example, presence of signs associated with the hindlimbs versus signs associated with the forelimbs) can be compared statistically.

Table 3.

Hypothetical scoring system for clinical signs

Starting date:																Nonweighted			Weighted factor 1				Weighted factor 2
Animal no.:		Day (after encephalitogen injection)														Separate totals	Grouped totals	Cumulative	Weighting factor	Separate totals	Grouped totals	Cumulative	Weighting factor	Separate totals	Grouped totals	Cumulative
Category	Clinical signs	7	8	9	10	11	12	13	14	15	16	17	18	19	20
	None															0	0	99	0	0	0	46.76	0	0	0	73.6
Weight	Loss ≥0.4 g first day; ≥0.1 g thereafter			x	x	x	x	x								5	5		1	5	5		1	5	5
Skin	Piloerection		x	x	x	x	x	x	x	x	x	x	x	x	x	13	21		0.5	6.5	10.5		0.5	6.5	10.5
	Matted fur				x	x	x	x	x	x	x			x		8			0.5	4			0.5	4
Tail	Loss of tone in distal half of tail or in a tail segment			x	x	x	x	x	x	x	x	x	x			10	26		0.33	3.3	8.6		0.5	5	13
	Loss of tone in entire tail					x	x	x	x	x	x					6			0.33	2			0.5	3
	Diminished lifting or diminished curling of tail				x	x	x	x	x	x	x	x	x	x		10			0.33	3.3			0.5	5
Bladder	Incontinence				x	x	x	x	x							5	5		1	5	5		1.5	7.5	7.5
Righting	Difficulty righting when placed on back			x	x	x	x	x	x	x	x	x				9	16		0.5	4.5	8		1	9	16
	Inability to right within 5 s after placed on back				x	x	x	x	x	x	x					7			0.5	3.5			1	7
Gait	Clumsy				x	x	x	x	x	x	x	x	x			9	18		0.33	3	6		0.67	6	12.1
	Dragging 1 hindlimb					x	x	x	x	x						5			0.33	1.7			0.67	3.4
	Dragging 2 hindlimbs					x	x	x	x							4			0.33	1.3			0.67	2.7
Paresis	Reduced range of forelimb abduction when placed on back					x	x	x	x							4	6		0.5	2	3		1.25	5	7.5
	No forelimb abduction when placed on back						x	x								2			0.5	1			1.25	2.5
Advanced signs	Side resting position							x								1	2		0.33	0.33	0.66		1	1	2
	Near-complete or complete plegia							x								1			0.33	0.33			1	1
	Rapid, slow, or deep breathing															0			0.33	0			1	0

The nonweighted, separate totals reflect counting the number of days with a sign, whereas the nonweighted, grouped totals represent the sum of the separate totals within a category. Weighting factor 1 uses weighting factors whose sum is equal to 1 for each category, and the separate totals represent the number of days with a sign times this factor while the grouped totals represent the sum of the separate totals within a category. Weighting factor 2 uses weighting factors whose sum for each category increases with more severe clinical signs, and the separate totals represent the number of days with a sign multiplied by this factor, whereas the grouped totals represent the sum of the separate totals within a category.

Although the individually tabulated scoring system can be designed to associate equal values with the different signs assessed, the individual clinical signs may not contribute equally to the overall disease presentation. Therefore, summing data would still lead to the problem of numerically equivalent scale intervals not necessarily representing equivalent changes in disease progression, as described earlier for standard scoring scales. One way to attempt to rectify this deficiency is to apply weighted values to the different clinical signs. To assign the appropriate weighting factor to a given sign, one partial, albeit imperfect, correction is to break a category of traits (for example, gait) into smaller components and then to apply a weight to each component such that the sum of the weights is equal to 1 (Table 3, weighted factor 1). In this manner, a subject does not obtain a full value of what would be equivalent to 1 interval jump on a standard scale unless all features of the overall trait were present. Dissecting traits into smaller components increases the sensitivity to detect differences between subjects. Alternatively, greater weight can be applied to clinical signs that represent more advanced disease processes (Table 3, weighted factor 2).

To reflect the overall clinical disease profile of an animal on a given day, the scores for clinical signs can be summed (with or without weighted adjustments) for each day (for example, the center columns of Table 2 represent the sum of daily nonweighted or weighted scores from Table 3). Cumulative analyses of these summed daily scores over the duration of the study or a portion thereof (for example, clinically active period, relapse) can be generated and used for statistical comparisons between groups or within groups. Statistical analyses can be performed on the maximal clinical score, area under the clinical score curve, day of onset, day of recovery, etc. (also see Evaluating clinical data).⁵¹

Some potential advantages of using an individually tabulated scoring system over scales with fewer or even an expanded number of intervals, are revealed through the data in Table 2 and Figure 1. For example, the 0-to-5 and 0-to-8 scales yielded the same scores for both days 10 and day 18, whereas the nonweighted and both weighted scales all showed greater than 2-fold differences between the scores for these 2 days (Table 2; Figure 1). A similar, albeit milder, situation occurred for days 16 and 17. Even more subtle differences were observed for days 11 and 14, but nonetheless, each of the nonweighted and weighted scales displayed differences between these days, whereas the 0-to-5 and 0-to-8 point scales returned equivalent values (Table 2; Figure 1). In addition, the nonweighted and weighted scales discerned differences between days 19 and 20, whereas the 0-to-8 scale assigned equal values to these days, and the 0-to-5 scale failed to detect disease on day 20 (Table 2; Figure 1).

Investigators can tailor the list of clinical signs to match the features of the EAE model that is being studied to maximize the usefulness of the data obtained. For example, if the model showed pronounced infiltration of inflammatory cells into the brainstem and cerebellum, then the scale could reflect spinning behavior and head tilting.^93,189

Behavioral testing can be an objective measure of disease activity.

To date, few behavioral tests have been used to evaluate subjects with EAE, but these tools may offer a more quantitative and accurate measure of disease activity than do traditional scoring scales. Behavioral tests have been used for a localized EAE model that causes lesions specifically in the spinal cord.^20,84 They also have been used for other models of spinal cord injury.¹¹⁷ Few studies involving common EAE models have applied behavior tests.^132,138-140

The open-field locomotion test rates a subject's hindlimb disability. The subject is placed on a flat, open area with a nonslip floor and observed for 3 to 5 min. A score of 0 to 21 is assigned based on hindlimb movement, with 21 indicating no disability and 0 indicating no hindlimb movement.^20,84

The grid-walk and narrow-beam tests are designed to gauge subjects’ motor skills. The grid-walk test involves counting how many times a subject's foot drops below the plane of a grid support, and the narrow-beam test measures plantar placement and how far a subject traverses a dowel or narrow beam.^20,84

A footprint analysis is useful in rating a subject's support and gait. This test consists of coating the subject's paws with ink and then allowing it to walk along a track on white paper. The footprints are analyzed for the distance between the hindpaws (a measure of base support), the length of the stride, and the angle of foot placement.²⁰

A spasticity test has been used to evaluate the ability of a test drug to ameliorate this impairment in a Lewis rat model of EAE.³⁰ This test measures the amount of force required to bend the joints in subjects.⁶⁸

Many of the described behavior tests rely on scoring by visual observation, which poses potential problems with equivalent training of personnel, scorer subjectivity, scoring variability, and scoring errors. Converting these tests to computer-based monitoring may be more accurate and objective.¹¹⁷ An early study used a basic system to electronically monitor the movement of subjects and, as expected, found that decreased movement was a feature of EAE.⁴³ A more accurate version of the original electronic movement monitor was developed (the force-plate actometer).⁵⁴ This device measures movement in spatial and temporal dimensions and permits a more indepth analysis than can be obtained from methods that detect movements by means of photo beams.⁵⁴ The actometer can track and quantify locomotive activity, gait, tremor, and rotational behavior,^117,162 all of which are features that can be relevant to EAE and MS.^86,189 Other computer-based systems [for example, the DigiGait Imaging System (Mouse Specifics, Quincy, MA)] also can perform measures of gait.^100,193

Animals with EAE can have decreases in muscle mass and strength as compared with control animals,²⁹ indicating that fatigue may need to be considered as a covariable when interpreting results from behavioral tests. In addition, various mental and psychologic features are commonly associated with EAE and MS; for example, symptoms similar to depression are typical of acute inflammation in both humans with MS and animals with EAE. In addition, EAE animals can display weight loss, decreased consumption of sucrose solution, and hypersomnia,¹³⁸ and a correlation has been found between mental alterations and the inflammatory process in EAE.^139,140 Accordingly, investigators might consider study design modifications, such as shortening the length of time for a given test in order to minimize fatigue or performing tests during periods of normally high activity. In addition, studies might be designed in which a stimulus (for example, prodding or even caffeine) is provided to promote an increase in activity level in otherwise depressed animals.

Various Ways to Evaluate Clinical Data

Clinical score data can be analyzed in numerous ways, and interpretation is not always straightforward, for example, a treatment may not affect disease onset but might temper disease severity. Often data from all subjects in a group are combined at each time point and then used for analyses of the ‘extent of disease’ profile; examples include the area under the curve of daily clinical scores and the number of days with a clinical score above a certain level.⁵¹ In addition, analysis of maximal, mean, and median clinical scores is often useful to determine the most extreme disease severity observed or the central tendency of a group. In other cases, scores during discrete phases of disease (for example, onset, recovery, relapse) or associated with particular clinical signs [for example, incontinence (Table 3)] can be evaluated separately. Another parameter of disease requiring analysis is time to event, for example, time of disease onset or time of relapse.

Weight loss, measured in grams or as a percentage, can be a useful indicator of disease. Using percentage weight loss rather than grams of weight loss minimizes differences in starting weight among subjects. As EAE progresses, subjects will be less likely or unable to maintain a normal food and fluid intake, and assistance in accessing food and water (for example, by placing food on the cage floor and using a water bottle with a stem that is reached easily) may be required.

MRI Can Assess Disease Activity in EAE

Magnetic resonance imaging is an important tool that is used in the diagnosis of MS and monitoring of disease activity. Before MRI, the identification of pathologic features required postmortem analysis, but now many of these features can be detected in living persons. This technology can detect signs of MS including inflammatory demyelinating lesions, breakdown of the blood–brain barrier, cortical inflammatory lesions, and tissue damage or loss over space and time.^{18,22,123,153} In addition, sensitive MRI techniques have shown potential to detect preinflammatory indications of disease, such as decreases in the magnetization transfer ratio and increases in tissue water diffusion in what initially appears to be normal white matter but that eventually develops lesions.^{46,50,147,190} In addition, MRI has been used to detect iron deposition in gray matter structures (for example, thalamus and putamen) of MS subjects;^{5,35,36,64,14} these deposits may have a pathogenic role. The ability to detect iron has led to the use of small iron particles as labels to track macrophage infiltration into the CNS, which was associated with the disruption to the blood–brain barrier, by using high-resolution MRI techniques.^34,102 Magnetic resonance spectroscopy can detect reduction in concentrations of N-acetylaspartate, an indicator of neuronal health,⁴ during the early progression of disease.⁷⁷ A more detailed list of the applications of MRI and magnetic resonance spectroscopy in MS recently has been presented.¹²³

Many of the pathologic features of MS that we have described elsewhere in this review can be detected in EAE subjects by using MRI. Nonhuman primates can readily be scanned in instruments with standard field-strength magnets, whereas high-magnetic-field scanners enable the collection of high-resolution MRI, which is ideal for species with small brains, such as mice and rats, as well as for imaging the spinal cord. In addition to advancing the understanding of pathology independent of euthanasia of animals, MRI is being used for evaluating potential therapeutics,^136,171,172 and MRI clearly will have a growing presence in EAE studies. Further, MRI may actually be a more sensitive measure of disease activity than visually assessing clinical signs, and eventually MRI findings might be used to direct clinical evaluations (for example, demonstration of optic nerve involvement would be followed by visual evoked potential studies).

Future Directions for Refinement of the EAE Model

Many of the pathologic features that are characteristic of MS can be induced in EAE models. These features include demyelination, axonal transection, neuronal loss, infiltration of T cells and macrophages into the CNS, activation of microglia, and astrocyte gliosis, among others. In addition, EAE can be used to study immune and inflammatory mechanisms. However, these measures may differ between EAE models and MS. Efforts have been pursued to incorporate components important to the disease in humans, such as components of the human immune system, into the EAE model,^38,76 so that it more accurately reflects MS.

Recently, bioinformatics¹²⁷ and microarray²⁵ approaches have been used to help identify whether associations or differences for various disease parameters are present between EAE and MS subjects. The outcomes from these and other studies likely will guide determinations regarding the relevance of EAE findings for application in MS and identify which models are most closely aligned with MS.

Conclusion

Although criticisms have been levied against EAE as an accurate representation of MS, having an animal model enables completion of studies that are not possible in humans, and if the studies are designed appropriately, the relevance of the results to MS is increased. However, even an appropriately designed study will have limitations, for example, disparate structures of a pharmaceutical target in animals compared with that in humans and toxicities specific to humans. The data from a given EAE study represent only a fragment of information in the overall context of MS, therefore multiple additional studies will be required to confirm the initial findings. The literature is filled with findings that were touted as having promise as an intervention strategy which were negated in subsequent studies. Given these restrictions, investigators should avoid overly optimistic or enthusiastic speculations about the implications of their results.