Attenuating the experimental autoimmune encephalomyelitis model improves preclinical evaluation of candidate multiple sclerosis therapeutics

Abstract

Background

Multiple sclerosis (MS) is a chronic disease of the central nervous system (CNS), exhibiting hallmarks of both inflammation and neurodegeneration and with limited treatment options. The intricate nature of MS pathophysiology and its variable progression pose severe challenges for the development of effective therapies. The experimental autoimmune encephalomyelitis (EAE) MS model, in its most common form, is an aggressive disease, which is not representative of the MS course and offers a limited time window for drug evaluation. This study aimed to generate an attenuated EAE variant, which extends the clinical testing window while preserving the high incidence of the standard EAE model.

Methods

Components of the EAE induction protocol were titrated to develop a milder disease profile. In a subsequent drug trial using the MS medication fingolimod hydrochloride (FTY, Gilenya), the new variant was validated under prophylactic and therapeutic treatment regimens.

Results

The attenuated EAE variant retains the standard hallmarks of neuroinflammation and, crucially, significantly extends the time frame for clinical drug testing. Unlike the standard variant, where FTY efficacy could only be demonstrated by prophylactic treatment, the attenuated variant facilitated differentiation of drug effects by therapeutic treatment initiated early in the acute phase of disease.

Conclusion

The new EAE variant is suitable for use in preclinical assessment of candidate therapeutics and the identification of targetable molecular mechanisms underpinning disease development and progression. This study illustrates the importance of optimizing and refining the experimental tool to enhance the translational success of the candidate therapeutics for MS.

This graphical abstract presents a summary of the research conducted in our laboratory. Our goal was to optimize the experimental autoimmune encephalomyelitis (EAE) model of multiple sclerosis (MS), which is typically highly aggressive, to enhance its use in therapeutic drug trials. To achieve this, we carefully titrated key reagents used in EAE induction, such as myelin oligodendrocyte glycoprotein peptide 35–55 (MOG_35–55), pertussis toxin (PTx), and Freund's adjuvant (FA) containing Mycobacterium tuberculosis (Mt), to develop a variant with an extended treatment window while maintaining high disease incidence. After testing eight different protocols, we successfully established a suitable variant, which was then used to conduct both a therapeutic and a prophylactic drug trial using fingolimod hydrochloride (FTY, Gilenya). These trials demonstrated the effectiveness of the new model in providing a better assessment of disease progression during therapeutic treatment.

1. INTRODUCTION

Multiple sclerosis (MS) is an inflammatory and neurodegenerative disease that affects the central nervous system (CNS). The cause of MS is still unknown, but evidence suggests that it results from a combination of genetic predisposition and environmental factors, ¹ for example, low vitamin D levels, ² , ³ viral infections, including Epstein–Barr virus, ⁴ , ⁵ smoking, ⁶ , ⁷ or early‐life obesity. ⁸ , ⁹ The hallmarks of MS pathology include blood–brain barrier (BBB) loss of function, lymphocytic infiltration, demyelination, and neuro/axonal loss. These features manifest differentially between the gray matter (GM) and white matter (WM) CNS compartments. Adding to the disease complexity, these destructive processes evolve over time. ¹⁰ Currently available treatments are predominantly immunomodulatory and adequately address the inflammatory component of the disease. However, the mechanisms driving neurodegeneration remain unclear, thereby making the accumulation of neurodegeneration unpreventable. ¹¹ , ¹² As a result, enhancing treatment options for MS is of major importance.

Unraveling the mechanisms that underlie the initial destructive events and their progression in MS is a challenging task. Ethical considerations prevent the collection of CNS tissue samples from patients, ¹³ making postmortem tissue from brain banks a valuable alternative source. Nevertheless, given the typically prolonged course of the disease, postmortem material is often seen as biased toward end‐stage pathology. ¹⁴ Investigators have therefore turned to animal models, but this approach has its own challenges, as no known animal species spontaneously develops an MS‐like disease. Furthermore, due to the complexity of MS, it has not been possible to date to create a model that accurately recapitulates the genetic, neurodegenerative, and immunopathological elements that collectively drive MS initiation and development. Thus, various in vivo models have been designed, whereby each replicates a particular facet of MS. ¹⁵ , ¹⁶ For instance, virus‐induced murine models using Theiler's encephalomyelitis virus (TMEV) or mouse hepatitis virus (MHV) mimic the potential viral etiology of the disease. ¹⁷ , ¹⁸ Toxic models of demyelination/remyelination induced in rodents, using ethidium bromide or the copper chelator cuprizone, are also available. The cuprizone model, which has seen increasing usage over the past decade, acts by inducing oligodendrocyte apoptosis through oral administration of cuprizone, leading to severe demyelination, but upon removal of the toxin from the diet, spontaneous remyelination occurs. This model has proven very valuable in investigations of myelin and axonal damage, restoration of populations of the oligodendrocyte lineage, and the intricate signaling, which drives remyelination. ¹⁹ , ²⁰ , ²¹

The most widely used model for MS is known as experimental autoimmune encephalomyelitis (EAE). ²² This model is generated in rodents and nonhuman primates by immune‐mediated demyelination using crude CNS homogenates, purified myelin proteins, or peptides derived from these proteins. It exhibits certain major hallmarks of MS, namely inflammation, widespread demyelination, axonal and neuronal loss, alongside astrocytic and microglial/macrophage activation, leading to motor function impairment and ascending paralysis. Other MS‐like symptoms observed include vision and balance issues, bladder and bowel dysfunction, and neuropsychological deficits. ²³ , ²⁴ The consensus is that the EAE model has played a significant role in the elucidation of multiple mechanisms underlying MS, including BBB loss of function, neuro/axonal injury and loss, the relationship between GM pathology and progression and between MS and gut dysbiosis. ²⁵ Furthermore, EAE has been immensely valuable for understanding the distinct roles of cytokine/chemokine networks and contributions of immune subsets, for example, by demonstrating the differential expression pattern and potential roles of Th1 and TH17 subsets between the brain and spinal cord, which have translational implications. ²⁶

The active induction of EAE, a method developed decades ago, requires the use of potent adjuvants that, in combination, enhance the immunogenicity of the given antigen. ²⁷ These adjuvants include complete Freund's adjuvant (CFA) supplemented with heat‐inactivated Mycobacterium tuberculosis (Mt), which serves as an antigen depot, ensuring continuous stimulation of the immune system due to slow release from the injection site. Another commonly used adjuvant is the toxic protein from the bacterium Bordetella pertussis, known as pertussis toxin (PTx), the specific purpose of which is to transiently increases BBB permeability. ²⁸ However, it has also been demonstrated that CFA, as well, causes disruption of BBB integrity. ²⁹ Rodent EAE is generally favored over the nonhuman primate counterpart due to considerations of cost, care requirements, and ethical issues related to experimenting on species with higher cognitive and emotional capabilities. ³⁰ Additionally, mice present the advantage of being eminently suitable for genetic manipulation. ³¹ , ³² The variant generated using the peptide 35 to 55 of the myelin oligodendrocyte glycoprotein (MOG_35–55, a low‐abundance myelin component) in the C57BL/6 mouse strain is by far the most used, as it exhibits both an encephalitogenic T‐cell response and a demyelinating autoantibody response. ³³

Like MS, EAE is a multifaceted disease, whereby each specific antigen to mouse/rat strain combination leads to a unique clinical profile, manifesting as chronic progressive, chronic relapsing, or monophasic EAE. ²³ Disease severity and clinical profiles are strongly influenced by antigen: host combination, environmental conditions such as the microbiome and even minor variations in protocols used. This is because it is a disease that heavily relies on the preexisting immune status of the animal. ³⁴ , ³⁵ The standard EAE protocol routinely used in most laboratories, namely MOG_35–55‐induced C57BL/6 mice, results in a severe disease with a chronic progressive clinical course and a high disease incidence that is extremely reproducible. On the contrary, the rapid progression of the disease presents a challenge for preclinical drug evaluation. ³⁶ In this variant, symptoms manifest from about 10 to 12 days postinduction (DPI), and humane endpoint is reached between 16 and 21 DPI, leaving a limited window for evaluation of drug effect.

To address this issue, we generated a novel MOG_35–55‐induced C57BL/6 EAE variant with a significantly milder progression and longer survival by titrating concentrations of the various induction reagents. We also investigated whether the classical hallmarks of inflammatory demyelination were preserved in the variant. Finally, we addressed the validity of this novel variant through a systematic drug trial employing FTY, a sphingosine 1‐phosphate receptor modulator and well‐established utilized therapeutic for MS, ³⁷ by comparing drug efficacy in the novel variant against the standard counterpart in modifying disease course.

2. METHODS

2.1. Animal housing and ethics compliance

C57BL/6J mice were sourced from Ozgene ARC (Murdoch, Western Australia, Australia). They were used at a minimum of 22 g at baseline and between 12 and 16 weeks of age. Only female mice were used to reflect the MS prevalence in females. ³⁸ The mice were housed in the La Trobe Animal Research and Training Facility in open‐top cages at 23°C and 41% humidity, with a 12:12‐h light: dark cycle. They were fed standard rodent dry pellets (Barastoc rat and mouse feed, Ridley AgriProducts, Australia) and sterilized tap water ad libitum. All procedures were approved by the La Trobe Animal Ethics Committee and strictly followed guidelines set by the National Health and Medical Research Council of Australia. ³⁹

2.2. EAE Induction and clinical profiling

MOG_35–55‐induced EAE in C57bl/6 mice was performed as described. ²⁷ On day 0, mice received two subcutaneous injections in the inguinal regions, of 100 μL each, containing MOG_35–55 peptide (Modpep, Melbourne, Australia) emulsified in Freund's adjuvant (Sigma‐Aldrich, Saint Louis, MO, USA). Two different preparations of Freund's adjuvant were used, namely incomplete Freund's adjuvant (IFA) or CFA, which is normally supplied with 1 mg/mL of heat‐inactivated Mt. Both were supplemented with 4 mg/mL Mt. (DIFCO, Detroit, MI, USA). The mice also received two doses of PTx (Sigma Aldrich) in 300‐μL phosphate‐buffered saline (PBS, 0.01 M, 154 mM sodium chloride, 7.68 mM disodium hydrogen phosphate, 2.67 mM sodium dihydrogen phosphate, pH 7.4), administered via the intraperitoneal route at 0 and 2 DPI. In the vehicle‐only control mice (VOC), MOG_35–55 was substituted with PBS. Multiple protocols were assessed during this study (Table 1). From 10 DPI, the mice were weighed and clinically scored daily by visual assessment of ambulatory difficulties according to the following scale: 0 = no symptoms, 1 = limp tail, 2 = hind limb weakness, 3 = hind limb paralysis, 4 = ascending paralysis, and 5 = moribund. Where intermediate symptoms were observed, an additional value of 0.5 was added to the score. Once a mouse reached a clinical score (CS) of 3–3.5 or had lost 20% of its initial weight, it was humanely killed by carbon dioxide inhalation according to animal ethics guidelines.

TABLE 1.

Experimental autoimmune encephalomyelitis (EAE) induction protocols.

Protocol	MOG35–55 peptide (μg)	PTx (ng)	Mt (μg)
SP	200	350	500*
1	150	150	400
2	150	250	400
3	100	200	400
4	100	250	500*
5	200	250	400
6	200	250	500*
7	200	300	400
8	80	300	400

Note: EAE protocols used for this study are summarized in this table. The standard protocol used routinely in our laboratory is shown in the top line. The titration of each induction reagent, namely MOG_35–55 peptide, PTx, and Mt., was carried out, where * indicates the use of complete Freunds adjuvant (CFA), as opposed to incomplete Freunds adjuvant (IFA). Values shown indicate the amount administered per mouse. Protocols 1, ⁴⁰ 3, ⁴¹ and 5 ⁴² were derived from the literature.

2.3. Histology and lesion load quantification

At experimental end point, transcardiac perfusion was carried out with PBS followed by fixation with 4% w/v paraformaldehyde (PF, Sigma‐Aldrich) in PBS, with n = 5 mice/group. Spinal cord tissues were dissected from the vertebral column from the second cervical (C2) region to the sixth cervical region (C6). The tissue samples (C2–C6) were processed for paraffin embedding using a standard protocol. Triplicate sections were collected at 7 μm thickness, then stained with preoxidized Mayer's hematoxylin solution (Amber Scientific, Midvale, Australia) and eosin solution (1% w/v, Amber Scientific) (hematoxylin and eosin [HE]) following a standard protocol. Sections were scanned using the Axioscan 7 slide scanner (Zeiss, Baden‐Württemberg, Germany). To quantify lesioned areas, lesions were outlined using the “freehand sections” tool on ImageJ, and lesion load was calculated as a percentage of total tissue area.

2.4. Immunofluorescence staining

C2–C6 spinal cord biopsies from a separate cohort of mice, with n = 3 mice/group, were placed in cryoprotectant (30% sucrose [w/v] in PBS) overnight and then snap frozen. Horizontal sections were cut at 14 μm on a cryostat (CM1950, Leica Biosystems, Nussloch, Germany) and collected on positively charged slides. Immunostaining was performed as described, ⁴³ in triplicates, using antibodies to CD3 (1:100, Abcam, Cambridge, UK), ionized calcium‐binding adapter molecule 1 (Iba1, 1:200, Novus Biologicals, Centennial, CO, USA), myelin basic protein (MBP, 1:200, Abcam, UK), hypophosphorylated neurofilament high (NF, 1:100, Novus Biologicals), and glial fibrillary acidic protein (GFAP, 1:100, Abcam). Detection was performed using Alexa 488 or 594‐conjugated secondary antibodies (Thermo Fisher Scientific, Waltham, MA, USA), and nuclei were stained with 4′,6 diamidino‐2‐phenylindole (DAPI) in Fluoroshield mounting medium (Sigma‐Aldrich). Negative controls included the absence of primary antibody or substitution of primary antibody with isotype control. Immunofluorescence was detected using an LSM 780 confocal microscope (Zeiss), and images were captured using ZEN 2011 Black edition software version 3.1 (Zeiss) [38/39]. To quantify myelination, ³⁷ fluorescence intensity (in arbitrary units) was measured using QuPath version 0.4.3 and ImageJ (32‐bit for Windows, LOCI, University of Wisconsin) from 30‐μm‐thick sections of the L1–L5 lumbar spinal cord, immunostained with anti‐MBP and with n = 5 mice/group and three sections/mouse. A defined region of interest (500 × 200 μm) was reproducibly applied across all images within each experimental series. Fluorescence intensity was calculated using the formula: integrated density – (area × mean fluorescence of background). The background fluorescence value was obtained from a corresponding negative control section.

2.5. Drug treatment

FTY (MW 343.93, Sigma‐Aldrich) was used at a dose of 0.3 mg/kg body weight. ⁴⁴ Treatment was initiated at two different time points, namely with a 10‐DPI presymptomatic initiation [38/39], or alternatively, therapeutically at CS 1–1.5. During the therapeutic trial, each mouse was treated individually. Upon observation of first clinical symptoms, each mouse was transferred to another cage, and FTY was administered orally via the drinking water. Water bottles were weighed daily to monitor water consumption (Figure S1). The experiment was conducted in a blinded manner.

2.6. RNA Isolation, cDNA synthesis, and qPCR analysis

After transcardiac perfusion with PBS, the C2–C6 spinal cord region was collected with n = 4 mice/group. RNA was extracted using the RNeasy lipid tissue mini kit (Qiagen, Netherlands) followed by complementary DNA (cDNA) synthesis with the Tetro cDNA Synthesis kit (Bioline, Boston, MA, USA). Primer pairs and reference gene (β‐actin) are provided in Table 2. Quantitative real‐time PCR (qPCR) reactions were run on a CFX96 real‐time system (Bio‐Rad, Hercules, CA, USA) using Luna Universal qPCR Master mix (M300, New England Biolabs, Ipswich, MA, USA) and 0.25 μM of primers in a 20‐μL volume. Each reaction contained 75 ng of template DNA. All reactions began with initial denaturation at 95°C for 1 min and 15 s followed by 45 cycles of denaturation at 95°C for 1 min and then annealing/extension at 60°C for 30 s. Melt curve temperature was between 60 and 95°C. The reaction was terminated at 4°C. All samples were analyzed in triplicates, and each plate included a no‐template control. The Ct values were generated by the Bio‐Rad CFX Manager software and imported into Microsoft Excel (Microsoft Corporation, Redmond, WA, USA) for conversion to fold expression relative to unit mass (1 μg total RNA). ³⁹

TABLE 2.

Primer sequence of target genes and reference gene.

Target gene	Forward primer sequence (5′–3)	Reverse primer sequence (5′–3)
TNF‐α	AAGCCTGTAGCCCACGTCGTA	GGCACCACTAGTTGGTTGTCTTTG
IFN‐γ	TCATGGCTGTTTCTGGCTGT	CCCAGATACAACCCCGCAAT
IL‐6	GCCTTCTTGGGACTGATGCT	TGCCATTGCACAACTCTTTTC
Β‐Actin	AGTGTGACGTTGACATCCGT	GCAGCTCAGTAACAGTCCGC

Abbreviations: IFN‐γ, interferon‐γ; IL‐6, interleukin‐6; TNF‐α, tumor necrosis factor α.

2.7. Data processing and analysis

Statistical analyses were conducted using Prism (version 10, GraphPad Software, Boston, MA, USA) or the Statistical Package for the Social Sciences (SPSS, version 29, IBM, Armonk, NY, USA). Graphs were produced with error bars denoting ±standard error of the mean (SEM).

Power analysis to test between‐group differences was conducted using an estimated effect size (f = 0.619) derived from both our published and pilot studies. ⁴⁵ Based on this, at alpha of 0.05 and power of 0.8, calculations suggested that six mice per group were required to detect an effect, while minimizing type 2 error. ⁴⁶ Due to differences in within‐group sizes in previous studies, and at the time of the current analysis, data report both the actual power (1 − β) associated with the analysis of variance (ANOVA) and the partial eta squared (partial η ²) to demonstrate that we had more than sufficient power to detect a statistical difference between groups with n = 5 mice per group. ⁴⁷ To account for the specific requirements of Prism analysis software, all mice were assigned a score of 4 at end point. This allowed for the accurate calculation of overall average CS across all groups. Significance between two groups was assessed with a Welch's t‐test. For comparisons involving more than two groups, one‐way ANOVA was employed followed by Tukey's multiple comparisons post‐hoc test. Survival analysis entailed generating a survival graph (Prism) and determining significance using a log‐rank (Mantel‐Cox) test. All data with a p‐value (≤0.05) were deemed significant (*); p ≤ 0.01 is indicated as (**), p ≤ 0.001 as (***), and a p ≤ 0.0001 as (****). The number of mice (n) used for each experiment is specified in the figure legends.

3. RESULTS

3.1. Optimization of the EAE protocol to generate a variant with an attenuated disease profile

Our aim was to generate a protocol where the minimum requirements would consist of a mean CS of 2 or above at peak of disease, along with a high incidence of at least 90%, low spontaneous resolution, and an extended experimental time window. This would generate a variant allowing drug effect to be determined not only by clinical profiling over a reasonable period but also permit tissue analyses so that quantification of parameters of recovery could be conducted. All protocols tested were induced using the method outlined in Section 2.2, with data shown in summarizing the results of each protocol shown in Table 3. The parameters assessed were the dosage of the MOG_35–55 peptide and PTx and the amount of Mt. supplementation of Freund's adjuvant. We began by reproducing three published protocols (1, 3, and 5, Table 1), which varied from our standard protocol (SP; Figure 1A–C). First, these protocols used IFA instead of CFA, resulting in a reduced concentration of 400 μg Mt/mouse compared to 500 μg Mt/mouse in SP. Second, in these published protocols, the PTx dose ranged from 150 ng (protocol 1), or 200 ng (protocol 3), to 250 ng (protocol 5), versus 350 ng/mouse in the SP. Third, mice received a MOG_35–55 dosage ranging from 100 μg (protocol 3) to 150 μg (protocol 1), but was similar at 200 μg/mouse (protocol 5) to our standard dose. However, we were unable to reproduce the profiles of all the three published protocols (Figure 1D–F,J–L,P–R). Protocols 1 and 3 (Figure 1D–F,J–L) showed great disparity in clinical profiles between subjects, ranging from fast‐progressing mice to mice that showed disease resolution or did not develop disease. This resulted in high survival rates of 80%, but a low mean CS of 1. Protocol 5 with a higher MOG_35–55 and PTx dosage compared to protocols 1 and 3, but equivalent Mt., also showed large discrepancy between individuals. More than 50% of mice spontaneously recovered from disease symptoms, although the mean CS reached the value of 1.9 (Figure 1P–R). Because of the large intersubject variation and high proportion of asymptomatic mice, the mean CS is not a true reflection of the disease, and these protocols are unsuitable for drug evaluation. We therefore sought to develop a novel EAE variant by modification of each of these published protocols.

TABLE 3.

Parameters of experimental autoimmune encephalomyelitis (EAE) protocols evaluated to generate attenuated EAE.

Protocol	Mean clinical score	DI (%)	SR (5)	DO (%)	Survival (%)	Endpoint (DPI)
SP	4	100	0	0	0	19
1	1	60	20	20	80	35
2	2.4	80	20	0	40	35
3	1	800	60	20	80	35
4	2.5	80	0	20	40	35
5	1.9	100	60	20	60	35
6	2.4	80	20	20	60	35
7	4	100	0	0	0	25
8	2.4	60	0	60	40	35

Note: The overall mean clinical score (CS), disease incidence (DI), percentage showing spontaneous recovery (SR), percentage with delayed onset (DO), percentage survival at experimental end point, and the experimental end point for each protocol are shown for each protocol. The higher survival percentage and longevity seen in protocols 1 to 6 and 8 are associated with lower disease incidence and/or higher spontaneous recovery and delayed onset relative to standard protocol (SP) and protocol 7.

FIGURE 1.

Experimental autoimmune encephalomyelitis (EAE) progression and survival in candidate protocols. Based on the protocols listed in Table 1, standard protocol (SP) and eight other EAE protocols were compared with a sample size of n = 5/group. In A to AA, three datasets for each protocol are shown. The left‐hand graph depicts the concentrations of the EAE induction reagents as a bar chart, namely MOG_35–55 peptide, Mt., and PTx. The right‐hand graph shows clinical score (CS) for each individual mouse (black lines) and mean CS (identified by a unique color for each protocol). Note that some protocols resulted in fewer than five mice exhibiting disease development (Table 3). The bottom graph shows the survival probability for each protocol. The top row (A–C) shows results using SP. Protocols 1 (D–F), 3 (J–L), and 5 (P–R) are from the literature. Protocols 2 (G–I), 4 (M–0), 6 (S–U), and 7 (V–X) were modified from these published protocols. Protocol 8 (Y–AA) was subsequently modified from protocol 7 to test MOG_35–55 effect. AB shows a comparison of mean clinical scores between protocols (SP to 8) and AC shows a comparison between survival curves. AD is a violin plot at 17 DPI showing the distribution of CS across the protocols and frequency of mice at each CS.

Protocol 2 (Figure 1G–I) was a modification of protocol 1, where the effect of increasing the PTx dosage to 250 ng/injection, while maintaining the lower MOG_35–55 and Mt. doses relative to SP, was investigated. This modification increased the percentage of mice developing disease above CS 2 compared to protocol 1, resulting in a mean CS of 2.4 at the peak of disease and 40% survival by experimental end point. However, 40% of mice were asymptomatic or spontaneously recovered.

Protocol 4 (Figure 1M–O) was a modification of protocol 3. However, given that increasing the PTx dosage in protocol 2 only showed a partial effect, we maintained the PTx dosage at 250 ng/injection, whereas the Mt. concentration was raised to 500 μg/mouse by substituting IFA with CFA. This resulted in a mean CS of 2.5 at the peak of disease and 40% survival by experimental end point. However, similar to protocol 3, large intersubject variation was observed within the cohort, with 20% asymptomatic mice and 20% with delayed clinical onset.

Protocol 6 (Figure 1S–U) was a modification of protocol 5, where increasing the Mt. dosage to 500 μg/mouse was evaluated, while maintaining the higher PTx dosage of 250 ng/injection and MOG_35–55 level of 200 μg/mouse. This protocol resulted in a mean CS of 2.4 at the peak of disease and 60% survival by experimental end point, but 20% of mice had spontaneous resolution and 20% had a delayed clinical onset.

Protocol 7 (Figure 1V–X) was also a modification of protocol 5, with a higher PTx concentration of 300 ng/injection but the same dosage of PTx and Mt. This resulted in a 100% incidence rate with a mean peak CS of 4, meaning that there was 0% spontaneous resolution. Additionally, this cohort exhibited an extended survival time till 25 DPI compared to SP.

Finally, to determine whether the same result as in protocol 7 could be achieved by reducing the MOG_35–55 dosage, we attempted protocol 8 (Figure 1Y–AA), where mice were administered only 80 μg of the antigen. We observed delayed clinical onset in 60% of the mice, whereas the other 40% were asymptomatic. The data from protocols SP to 8 were further compared, where the average clinical score (Figure 1AB) and the probability of survival (Figure 1AC) show that only P7 is associated with a mean clinical score above 2 coupled with extended longevity. To determine whether there were significant differences between the end points of the protocols, a one‐way ANOVA was conducted. This revealed a significant difference between end point (CS 3–3.5 or 34 DPI) across protocols (F(8, 36) = 2.852, p = 0.015, partial η ² = 0.388, 1 − β = 0.890). Post‐hoc tests showed that the end point for SP was significantly different from protocols 1, 3, and 8, which did not develop the disease.

Violin plots were generated to represent the distribution of CS for each protocol and frequency of mice at each CS (Figure 1AD). This analysis shows the 17 DPI time point, which corresponds to 80% mortality with SP but 100% survival with protocol 7. The plots illustrate the rapid disease progression associated with SP compared to the large intersubject variation in protocols 2, 4, 5, and 6, as well as the low mean clinical score in protocols 1, 3, and 8. Protocol 7, on the contrary, exhibits a more even CS distribution. Changes in CS over the 15–20 DPI period (Figure S2) confirm the more uniform progression in protocol 7. Main effect analyses for MOG_35–55, PTx, and Mt. were conducted using the 17, 20, and 25 DPI data (Tables S1–S3). They show significantly higher effect on CS with MOG_35–55 at 200 μg (protocols SP, 5, 6, and 7) versus 80 μg (protocol 8) at all three time points. PTx at 350 ng (SP) resulted in a higher mean CS at 17 DPI relative to all other protocols and at 20 DPI relative to 200 and 300 ng (protocols 3, 7, and 8), but there was no difference at 25 DPI. Mt. at 500 μg (protocols SP, 4, and 6) resulted in a higher mean CS at 17 and 20 DPI but not 25 DPI. These data confirm the relationship between the higher concentrations of these components and rapid disease progression.

3.2. The attenuated EAE variant has a significantly increased clinical window

Protocol 7, from here on named the attenuated protocol (AP), therefore best fits the required criteria. Further comparison between SP and AP of mean CS (Figure 2A) and cumulative CS (Figure 2B) shows that the attenuated EAE variant is less aggressive than SP. Thus, between 13 and 26 DPI, based on mean CS, clinical progression of AP compared to SP is significantly decreased (p < 0.01). Disease severity in terms of cumulative CS over time is also significantly decreased in AP (p < 0.05). Weight loss (Figure 2C), which is an additional measure of disease progression, is more severe in SP overall (p < 0.05) as demonstrated with the more drastic weight loss in the acute phase of the disease. Additionally, the survival curve (Figure 2D) shows that AP mice survived longer till 25 DPI compared to SP that had 100% mortality by 19 DPI (p < 0.05).

FIGURE 2.

The attenuated protocol compared to standard protocol. (A) The attenuated protocol (AP) has a significantly decreased mean clinical score (CS) compared to the standard protocol (SP) from 13 to 26 DPI (**p < 0.01). (B) Cumulative CS is significantly higher in SP compared to AP from 17 to 35 DPI (*p < 0.05). (C) Percentage weight change relative to 0 DPI shows a decrease in weight in the AP group, but overall SP shows a more severe weight change (*p < 0.05). (D) Probability of survival with AP is improved significantly (*p < 0.05) by around 6 days from 19 DPI for SP to 25 DPI. n = 5 mice/group.

Therefore, in all subsequent experiments, AP was used, because it provided the highest degree of incidence and a slower progression to CS 3 and survival till 25 DPI, thereby extending the treatment period by 6 days.

3.3. The attenuated variant exhibits all the pathological hallmarks of an EAE model

To ascertain that parameters of neuroinflammation were maintained in the new AP variant, a series of histological and immunopathological assays were performed at experimental end point. All qualitative and quantitative evaluations were performed using spinal cord tissue (Figure 3A). VOC mice, which do not develop disease and do not exhibit lesions, acted as controls. We first identified lesions by HE staining (Figure 3Ba–c) followed by scanning of whole tissue sections. No alterations in lesion topography were observed between SP and AP, whereby lesions were confined to WM regions and never observed in GM in both cases. Lesion load was calculated by outlining lesioned area using ImageJ as described in Section 2.3 and calculating lesioned area as a proportion of total tissue area. There was no significant difference in lesion area between the two protocols (Figure 3C). Lesion severity was determined using an antibody against the pan T‐cell marker CD3. Results showed that although severe lesions were detectable in the AP experimental group, these appeared to be generally less aggressive than those observed in SP (Figure 3Bd,e). Demyelination, a key pathological hallmark of neuroinflammation, was evaluated by immunochemistry using an antibody against the major myelin component MBP. Visual inspection demonstrated large unstained areas that were not evident in the VOC group but appeared equally extensive irrespective of protocol used, as in Figure 3Bg–i. This was confirmed by quantitative confocal microscopic analysis, which showed no significant difference between protocols (Figure 3D).

FIGURE 3.

Validation of the attenuated experimental autoimmune encephalomyelitis (EAE) variant. (A) The spinal cord region was used for investigation of parameters of neuroinflammation. The outline identifies the white matter (WM) and gray matter (GM) regions of the cervical spinal cord and the central canal (CC). The black box identifies the white matter region used for qualitative and quantitative data. (B) Images of standard protocol (SP) and attenuated protocol (AP) mice were compared to vehicle‐only control mice (VOC). (Bac) Hematoxylin and eosin (HE)‐stained images with multiple white matter lesions (dotted lines) in both SP and AP. No lesions were detected in VOC or GM of SP and AP mice. (Bd–f) Immunofluorescence images of cryostat sections immunostained with anti‐CD3 and anti‐Iba1 depicting T‐cell infiltration and microglial reactivity. The lesioned areas in SP and AP are outlined by the dotted lines, and white arrows show microglial activation. (Bg–i) Clear demyelination was seen in both SP and AP stained with anti‐myelin basic protein (anti‐MBP) and indicated by white arrows. (Bj–l) Tissue was stained for axonal injury using anti‐NF (green). The white arrow identifies elevated hypophosphorylated NF expression. (Bm–o) Anti‐GFAP positive staining in both SP and AP shows an increase in astrocytic reactivity compared to control as pointed out by the white arrows. Sample size = 3 mice/group x 3 sections per mouse. Magnification: 50x. Scale bars = 150 μm in (Ba‐c); Magnification: 400x. Scale bars = 20 μm in (Bd‐o). (C) Lesion load quantification of HE samples. No significant difference was found between AP and SP. (D) Quantification of MBP revealed no significant difference in myelin loss between SP and AP. n = 5 mice/group x 3 sections per mouse.

Additional pathological hallmarks of neuroinflammation, microglial and astrocytic reactivity, and axonal injury were also investigated by immunofluorescence staining (Figure 3Bd–f, j–o). Microglial reactivity, which was identified with anti‐Iba1, was less prominent in the AP compared to SP but more prominent than in VOC group (Figure 3Bf). Axonal injury was detected with NF. Loss of phosphorylation of the axonal cytoskeletal NF is a well‐established feature of CNS inflammation. Here, we show that NF levels were apparently decreased in AP compared to SP. However, both AP and SP NF levels were elevated relative to VOC mice (Figure 3Bj–l). Finally, astrocytic reactivity was identified with GFAP. Reactive astrocytes, as identified by higher GFAP reactivity and more complex branching of processes, are observed in both AP and SP relative to the VOC group (Figure 3Bm–o).

To obtain further quantitative estimation of differences, if any, between SP and AP protocols, qPCR evaluation of the pro‐inflammatory markers interleukin‐6 (IL‐6), tumor necrosis factor α (TNF‐α), and interferon γ (IFN‐γ) was performed (Figure S3). As in the previous analyses, VOC mice were used as controls. Data are expressed in terms of fold change (2^delta CT) and normalized against normal samples. There were no significant differences between SP and AP in any of these markers, again confirming that AP exhibits typical parameters of neuroinflammation.

To investigate a potential differential effect between SP and AP in brain versus spinal cord lesion development, HE‐stained brain sections were compared. The results, shown in Figure S4, reveal no identifiable differences in lesion topography and severity between the two protocols in the brain.

Taken together, we conclude that alteration of the EAE induction protocol to generate an attenuated clinical disease has retained the essential hallmarks of neuroinflammation. Although differences were observed in lesion severity and glial reactivity, these did not translate into significant differences in lesion load and the inflammatory environment. The alterations in components of disease induction, however, have altered the clinical profile in AP, resulting in a less‐severe acute phase and extended longevity.

3.4. The attenuated protocol better illustrates drug efficacy

The efficacy of FTY was tested in both SP and AP to validate the attenuated EAE variant as a suitable model to use in preclinical drug trials. FTY was tested first at the presymptomatic time of 10 DPI. This time point was selected based on our previous intracellular cytokine staining evidence that it immediately precedes detectable accumulation of autoreactive T cells and appearance of symptoms. ³⁸ Alternatively, therapeutic intervention was initiated immediately upon clinical presentation at CS 1–1.5. Both EAE variants showed a response to FTY treatment, but with differences depending on the protocol used. Using SP, presymptomatic intervention was found to be significantly more effective than treatment initiation at disease onset, with mice reaching a peak mean CS of 3 with the former by 17 DPI, instead of 3.75 in the case of the therapeutic regimen (p < 0.05; Figure 4A). There was no significant difference between the therapeutic regimen and no treatment controls, showing that drug efficacy was demonstrable only when treatment was initiated presymptomatically. A similar relationship was observed in the case of AP (Figure 4B), where presymptomatic treatment was more effective than treatment upon clinical inset (p < 0.01). However, in the case of AP, there was a highly significant difference between the therapeutic regimen and no treatment control (p < 0.0001), showing that drug effect could be demonstrated by attenuation of disease aggressiveness. Comparison of the two presymptomatic interventions (Figure 4C) showed no significant differences in disease severity, with overlapping profiles during the acute phase of disease and no significant difference at end point. On the contrary, comparison of the therapeutic regimens (Figure 4D) revealed significantly higher efficacy of the drug in AP (p < 0.0001), which can be appreciated from the significantly lower mean CS at peak of disease with AP compared to SP, as well as at experimental end point, demonstrating that AP better illustrates drug efficacy under the therapeutic drug regimen.

FIGURE 4.

Drug trial of fingolimod hydrochloride (FTY) in two different experimental autoimmune encephalomyelitis (EAE) variants. The experimental end point was set as 34 DPI. Drug treatment initiation is shown by arrows. (A) Comparison between FTY administered at 10 DPI in SP or from CS 1–1.5 and to the no‐treatment control. CS mean was significantly higher (*p < 0.05), with the therapeutic regimen compared to treatment initiated at 10 DPI but not significantly different from the no‐treatment control. (B) With AP, a significant difference is observed from 15 to 34 DPI between the therapeutic regimen and 10 DPI initiation (**p < 0.01), but both are significantly different from the no‐treatment control (****p < 0.0001). (C) There was no significant difference between SP/FTY and AP/FTY administration from 10 DPI. (D) There was a significant increase in mean CS in the SP/FTY group when drug treatment was initiated at CS 1–1.5 compared to AP/FTY from 15 to 34 DPI (****p < 0.0001). n = 7/group.

4. DISCUSSION

EAE is the most widely accepted model of MS and has been extensively used to generate proof of concept for mechanisms underlying MS immunopathogenesis. However, despite the undoubtedly valuable similarities between EAE and MS, important differences have also been identified, which need to be kept in mind, particularly with respect to preclinical studies. A salient difference relates to lesion topography, where inflammation is most prominent in the spinal cord, while simultaneously being quantitatively minor in the brain, which is distinctly different from MS. Additionally, EAE is a rapidly progressive disease unlike MS. Therefore, the model needs to be constantly revised and refined to improve its translational value.

MOG_35–55‐induced C57BL/6 EAE causes an ascending paralysis, typically evaluated using a scoring system that assesses ambulatory difficulties by visual inspection alone. A score of 1 is defined by tail flaccidity, corresponding to inflammation confined to the lumbar and sacral spinal cord regions. By score 2, characterized by slight‐to‐severe hind limb weakness, inflammation has become widespread and is observed in the thoracic and cervical spinal cord, as well as the cerebellar WM. At score 3, paralysis is evident in one or both hind limbs, and inflammation has intensified, although remaining mostly confined to the spinal cord and cerebellum, with little involvement of the cerebral hemispheres. ⁴⁸ Therefore, our first prerequisite was to develop an EAE variant exhibiting a disease pathology that was sufficiently severe to allow clear differentiation of drug effect, not only by clinical observations but also by quantitative measures such as gait and strength analyses, as well as estimations of hallmarks of neuroinflammation via confocal microscopy in an eventual preclinical drug evaluation. Therefore, our first requirement was that the protocol resulted in a minimum of CS 2 for each mouse. The second requirement was a high degree of disease incidence and low levels of spontaneous disease resolution. This was important not only to avoid wastage of animals but also to allow drug trials to be initiated early in the acute phase with the confidence that evidence of low CS could be solely attributed to drug effect. Lastly, it was important to extend the time frame of experimentation before reaching a humane end point to fully allow evaluation of tissue repair, for example, when evaluating candidate remyelinating drugs.

Our initial efforts to reproduce published protocols failed to meet these requirements. Notably, we encountered considerable variability in disease manifestations among individual animals. Some of these variations were unacceptable, such as approximately 50% of spontaneous resolution in protocols 3, 5, and 6. Within those protocols, instances of spontaneous resolution were observed following a peak CS of 1–1.5. In a preclinical trial, if using these protocols, the mean CS would not accurately represent the effect of the drug on at least 50% of the cohort. Such variations within each protocol are consistent with the fact that the SP is tailored to achieve a high incidence rate (>90%). It has been shown that deviations from these conditions generate a range of disease profiles, which can in the C57BL/6 mouse strain manifest as chronic progressive, or chronic relapsing, or monophasic. ⁴⁹ Additionally, our inability to reproduce published protocols may be attributable to environmental differences, particularly in terms of the gut microbiome. It is known that the composition of the gut microbiome profoundly influences EAE development, as mice housed in bacterial pathogen‐free environments develop a significantly milder form of EAE. Research has also shown that modulation of the microbiome can profoundly affect the disease profile. ⁵⁰ For example, dietary interventions, such as intermittent fasting and calorie restriction before the induction of EAE, produced a less‐aggressive disease course. ⁵¹ , ⁵²

The optimization of the protocol posed challenges due to the necessity of modifying multiple components, namely PTx and the Mt. content of Freund's adjuvant. These adjuvants have overlapping effects where PTx has complex functions, the predominant one being a transient increase in BBB permeability. It is believed to also stimulate the Th1 response via IL‐12 induction. ⁵³ Mt. enhances the adjuvanticity of CFA by promoting the Th1 response to myelin antigens and acts as a potent inducer of IL‐12. It is established that Mt. also contributes to BBB permeability. ³³ , ⁵⁴ Thus, in the C57/BL/6 mouse strain, observations over a 4‐week time course following CFA treatment alone demonstrated increased perivascular extravasation of serum immunoglobulin G (IgG), albumin, IgM, and exogenous horseradish peroxidase (but to varying degrees), most prominently in the brainstem and cervical spinal cord. ²⁹ Therefore, we optimized a protocol by modification of one or two components relative to each published protocol. Comparison of protocols 1 and 2 showed that increasing the PTx dosage was effective in increasing disease incidence and reducing the proportion of mice with delayed onset, consequently raising the mean CS to 2.4, but this occurred without change in spontaneous recovery. In protocol 4, the PTx level was therefore used at the same level as in protocol 2, with increased Mt. dosage relative to protocol 3. This reduced the percentage of spontaneous resolution and raised the mean CS to 2.5 but had no effect on disease incidence and survival to experimental end point of 35 DPI. The effect on disease incidence in this case may also be due to the lower MOG_35–55 dosage of 100 μg/mouse. Therefore, in protocols 6 and 7, we increased the MOG_35–55 dosage to 200 μg/mouse and alternatively increased either the Mt. dosage to 500 μg/mouse (protocol 6) or the PTx dosage at 300 ng/mouse (protocol 7). This established that PTx at 300 ng and MOG_35–55 as 200 μg were the appropriate dosages to increase the disease incidence and reduce the percentage of spontaneous recovery, which was confirmed by comparison between protocols 7 and 8 where MOG_35–55 alone was reduced. These doses are consistent with the main effect analyses, which showed that all three components of EAE induction contribute to disease aggressiveness. By lowering PTx and Mt dosages, milder progression was achieved, but reducing MOG_35–55 dosage had the effect of delaying disease development. Therefore, the MOG_35–55 level had to be maintained at a higher concentration. Consequently, the optimized protocol demonstrated the most normal distribution and consistent responses and generated the most predictable disease course for drug evaluation. Concerns have been raised regarding the artificial disruption of the BBB during EAE induction, facilitating immune cell infiltration. Alternative approaches, such as humanized models, which utilize myelin‐specific human T‐cell receptors and human histocompatibility complex antigens, ⁵⁵ or EAE mouse strains not reliant on pertussis toxin, like the EAE variant generated with the use of peptide 139–151 from the major myelin component proteolipid protein (PLP_139–151) in the SJL/J mouse strain, which develops a relapsing–remitting disease profile, offer promising avenues to better replicate MS pathophysiology and refine EAE models. ⁵⁶ However, some argue that BBB breakdown is an integral part of MS pathology, suggesting that its inclusion in models could be relevant. ³¹

Modification of the EAE protocol can also result in changes in lesion severity and topography. ⁵⁷ Therefore, it was important to ascertain whether any such changes had occurred. Alterations in lesion topography were not observed between SP and AP. Histological comparison between SP and AP at end point revealed observable differences in T‐cell intensity, glial reactivity, and axonal injury, which did not translate into significant difference in lesion load and severity of the inflammatory environment, nor altered lesion topography. It is well established that spatiotemporal lesion development is unique to each EAE variant, ²⁷ which is a valuable property in terms of time and cost savings when selecting regions of interest for evaluation of drug efficacy. The data here show that alteration of clinical profile has not significantly affected histological hallmarks of neuroinflammation between SP and AP. In this context, it is also important to note that attenuation of the protocol was not associated with the development of atypical EAE, defined as a deviation from the classical EAE ascending paralysis associated with spinal cord inflammation, to ataxia, tremors, or asymmetric paralysis, resulting from brainstem, cerebellar and/or cerebral hemispheres. This is not an unexpected result, as this form of EAE has been associated with models of adoptive myelin‐reactive T‐cell transfer, or with less commonly used EAE variants such as EAE induced in PL/J mice by MOG peptides ⁵⁸ or in BALB/c mice by PLP_178–191. ⁵⁹ Furthermore, the absence of atypical EAE suggests that the modification of the protocol did not significantly alter the differential expression pattern Th1 and TH17 subsets between the brain and spinal cord, where it was demonstrated that transfer of Th17 cells induced mainly brain inflammation into recipient mice, resulting in atypical EAE, whereas transfer of Th1 cells led to the development of classical EAE with only spinal cord inflammation. ²⁶ Th17 cells are now recognized as key contributors during the early stages of EAE/MS; however, much remains to be done to improve appreciation of the relationship between the Th1 and Th17 subsets. The less‐aggressive profile of the AP may be more conducive to transfer experiments and investigations of early immunopathogenetic mechanisms relevant to MS.

The extension in survival time in the AP that prolonged the evaluation period for drug efficacy from 19 to 25 days is primarily attributable to a significant decrease in disease severity during the acute phase. This appears to be the most critical element in the improvement in the capacity to conduct therapeutic drug trials, as demonstrated in our comparison of FTY efficacy in the two EAE variants under different regimens. First, data in Figure 4 demonstrated that using a therapeutic approach it was not possible to identify a drug effect with SP, whereas a clear effect was evident with AP. This shows that the aggressiveness of the SP protocol is such that a drug effect, which is real as shown by AP, cannot be demonstrated by SP. On the contrary, there was no difference between SP and AP when treatment was initiated at 10 DPI. This shows that prophylactic treatment always appears beneficial irrespective of induction protocol. However, such a regimen has no clinical relevance, as at this point in time mice are not symptomatic.

Before FTY entered the market, it underwent testing in the standard EAE model, though primarily via the use of a prophylactic regimen, namely with initiation of drug administration at 0 DPI [43/44, 61]. This approach is potentially problematic as our intracellular cytokine staining data have shown absence of autoreactive T cells in the blood, lymphoid, and CNS tissue before 10 DPI [38/39]. Additionally, the mechanism of action of FTY, which targets sphingosine‐1‐phosphate receptors, directly interferes with T‐cell trafficking. ⁶⁰ This suggests that prophylactic treatment of EAE before 10 DPI has the potential to alter or prevent disease development, thereby invalidating the comparison with the control groups. Although FTY successfully advanced to the MS drug market, several other drugs that followed a similar approach, such as the LINGO antagonist (Opicinumab), ⁶¹ failed to progress beyond phase two clinical trials. ⁶² , ⁶³ Overall, these findings highlight the importance of careful considerations such as appropriate timing of drug administration, which has the potential to alter the outcome of the study. Comprehensive clinical and pathological evaluations are also essential, as seen in this context with the timing of T‐cell infiltration, to precisely determine the therapeutic administration of immunomodulatory drug FTY.

A limitation of the study is that despite the attenuation of disease progression, the protocol is still aggressive, because mice continue to progress and need to be humanely killed, albeit at a later time. On the contrary, the extended time window, even if modest, was sufficient to generate a protocol, which enabled therapeutic intervention. Additionally, it would be of interest to quantify the inflammatory environment over the 16–20 DPI window to further understand parameters of the less‐aggressive disease. The same analysis should be performed during experiments aimed at drug evaluation, particularly with respect to candidate neuroprotective drugs.

5. CONCLUSION

Although EAE is a distinct disease from MS, it is accepted as a representative model of the neuroinflammatory component of the human counterpart of the disease. Preclinical drug evaluation will always require EAE. If addressed properly, this model can generate important proof‐of‐concept for mechanisms underlying the disease, the identification of targetable mechanisms, and evaluation of candidate therapeutics. The studies reported here highlight the importance of optimization of the appropriate variant within the environment where drug trials are to be conducted for improved evaluation of drug effect.

AUTHOR CONTRIBUTIONS

Vernise J. T. Lim: Conceptualization; data curation; investigation; methodology; writing – original draft; writing – review and editing. Melanie J. Murphy: Data curation; methodology. W. Stephen Penrose: Data curation; methodology. Coral Warr: Supervision. M. Cristina Keightley: Supervision. Jacqueline M. Orian: Conceptualization; funding acquisition; methodology; project administration; resources; supervision; writing – review and editing.

FUNDING INFORMATION

This work was funded by private donations and grants from La Trobe Research Focus Areas and Multiple Sclerosis Australia, Grant/Award Number: 20‐032. Vernise J. T. Lim was the recipient of a scholarship from La Trobe University.

CONFLICT OF INTEREST STATEMENT

The authors have no conflicts of interest to declare.

ETHICS STATEMENT

All procedures received full approval from La Trobe University Animal Ethics Committee and were performed strictly in accordance with guidelines set by the National Health and Medical Research Council of Australia (Approval number: AEC21010).

Supporting information

Data S1.

Lim VJT, Murphy MJ, Penrose WS, Warr C, Keightley MC, Orian JM. Attenuating the experimental autoimmune encephalomyelitis model improves preclinical evaluation of candidate multiple sclerosis therapeutics. Anim Models Exp Med. 2025;8:1428‐1440. doi: 10.1002/ame2.70071

DATA AVAILABILITY STATEMENT

Data available on request from the authors.