High-intensity exercise training alters gut microbiota to mitigate the development of experimental autoimmune encephalomyelitis

Abstract

Exercise training (ET) has demonstrated beneficial effects in autoimmune and neurological disorders, including multiple sclerosis and its animal model, experimental autoimmune encephalomyelitis (EAE). ET modulates the gut microbiota, which influences neuroimmune interactions via the microbiota-gut-brain and microbiota-gut-immune system axes. However, the role of gut microbiota in mediating ET’s protective effects in autoimmune neuroinflammation remains unclear. We investigated whether gut microbiota mediates the beneficial effects of ET on EAE development. Healthy mice underwent high-intensity continuous training (HICT). Fecal microbiota from HICT and sedentary mice were transplanted into naïve recipients, followed by proteolipid protein (PLP) immunization to induce EAE. Disease severity, gut microbial composition (16S rDNA sequencing), short-chain fatty acid (SCFA) levels (LC–MS), and autoreactive T-cell proliferation (flow cytometry) were assessed. Faecal microbiota transplantation (FMT) from HICT donors significantly reduced EAE severity, delaying onset and decreasing CNS inflammation, demyelination, and axonal damage. These effects correlated with distinct microbial signatures, including increased Faecalimonas and Escherichia genera, and decreased Mucispirillum genus. HICT-FMT mice exhibited higher Faecalimonas abundance and reduced serum SCFA levels. PLP-reactive T-cell proliferation was suppressed in HICT-FMT recipients. Gut microbiota from HICT mice confers protection against EAE development, associated with microbial-metabolic shifts and modulation of autoreactive T-cell responses.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-48522-2.

Introduction

Exercise training (ET) exerts beneficial effects on neurological diseases, including Multiple Sclerosis (MS)^1,2. While the favorable impact of ET on brain health is widely accepted, the mechanisms underlying these benefits remain poorly understood. Experimental autoimmune encephalomyelitis (EAE), an animal model of MS, has provided insight into potential mechanisms through which ET exerts its therapeutic effects¹. In a series of studies, we utilized the passive transfer EAE model that enabled us to differentiate between systemic immunomodulatory effects and direct neuroprotective effects of ET and examined the effects of various training paradigms on the progression of EAE. Among various ET protocols, high-intensity continuous training (HICT) has shown optimal effects in slowing disease progression by inducing both systemic immunomodulation and direct protective effects on the central nervous system (CNS)^3–8. However, the factors mediating these effects are yet to be defined. Potential mediators include the gut microbiota.

Recent research has revealed the profound influence of the gut microbiota and its metabolites on physiological functions, including those of the immune and nervous systems⁹. Gut bacteria and their metabolic products, such as short-chain fatty acids (SCFAs), cytokines, and neurotransmitters (e.g., catecholamines, GABA), can access the brain through the circulatory system or the vagus nerve and modulate various processes, including signal transduction, neurotransmission, enzyme expression, and cytokine production by CNS microglia^10,11. This dynamic interplay is referred to as the “microbiota-gut-brain axis”. Additionally, gut microbiota dysbiosis, defined as an imbalance in microbial composition, plays a crucial role in numerous diseases, including autoimmune and neurological conditions^12–15. In the context of MS and EAE, studies suggest that gut microbiota contributes to critical processes such as inhibiting Th17 cell formation, regulating blood–brain barrier (BBB) permeability, modulating microglial and astrocyte function, and influencing myelin gene expression^16–18.

Furthermore, ET has been shown to modulate gut microbiota composition and its associated metabolites^19–21. Positive changes include increased microbial diversity and a favorable balance between beneficial and pathogenic bacterial communities. ET also exerts systemic effects on the immune system²², including enhanced antioxidant enzyme activity (e.g., catalase, glutathione peroxidase), increased anti-inflammatory cytokines (IL-10), and reduced pro-inflammatory cytokines (TNF-α, IL-17). However, the effects of ET on the gut microbiome depend on the intensity of training. While mild exercise has beneficial impacts²³, strenuous exercise (≥ 60–70% VO₂max) may compromise gut barrier integrity and exacerbate inflammation^23,24. Thus, the intensity of training must be thoroughly considered when evaluating the gut microbiome.

Fecal microbiota transplantation (FMT) has emerged as a powerful tool to investigate causal relationships between gut microbiota and disease. Recent studies have demonstrated the therapeutic potential of FMT in gastrointestinal and systemic autoimmune disorders^25,26, as well as in MS^27,28.

Transferring microbiota from ET-trained donors to naïve recipients, followed by EAE induction, offers a novel approach to determine whether ET-induced microbial shifts can modulate disease susceptibility and progression.

Considering the evidence linking gut microbiota dysbiosis to neurodegenerative diseases and the modulatory effects of ET on both the gut microbiome and neurodegeneration, it is logical to triangulate the three and investigate whether ET can influence autoimmune neurodegeneration via alterations in gut microbiota. While previous research has explored microbiota involvement in EAE²⁸, our work uniquely investigates the protective effects of ET-conditioned microbiota through FMT in a controlled experimental setting. The current study aims to determine whether the gut microbiota from HICT donors is associated with protection against EAE, and to characterize accompanying microbial and metabolic shifts. The specific objectives of this research are: (1) characterize ET-induced changes in gut microbiota composition and metabolite profiles in trained versus sedentary mice; (2) determine the association of the gut microbiota to ET-mediated attenuation of EAE; and (3) assess the impact of ET-modified microbiota on the encephalitogenic potential of autoimmune T cells.

Materials and methods

Experimental animals

This study is reported in accordance with the ARRIVE guidelines (https://arriveguidelines.org). All experimental procedures were reviewed and approved by the Institutional Animal Care and Use Committee (approval no. IL-210-10-20) and conducted in accordance with the United States Public Health Service Policy on Humane Care and Use of Laboratory Animals. Female SJL/JCrHsd mice, 6–7 weeks (wks) old, were sourced from Envigo Inc. (Israel). The animals were housed in standard cages, with 2–3 animals per cage, at a controlled temperature of 22 ± 1 °C, under a 12-h light/dark cycle (lights on at 07:00), with unrestricted access to food and water, with ad libitum access to food (standard rodent chow diet) and water. Bedding, water, treadmill and food were not sterilized prior to use.

Experimental design

To investigate the role of gut microbiota in mediating the beneficial effects of ET on EAE development, HICT protocol, proteolipid protein (PLP)_139–151 EAE model and FMT paradigms were utilized (Fig. 1). Separate experiments were conducted for each endpoint, and different cohorts of mice were used for the clinical assessment, histopathology, and immunological analyses. To that end, healthy donor mice were subjected to a HICT treadmill-running program and sedentary (SED) mice served as controls. Thereafter, fecal samples from HICT and SED donor mice were transplanted into two groups of naïve recipient mice via oral gavage (HICT fecal microbiome transplanted—HICT-FMT and SED fecal microbiome transplanted—SED-FMT groups, respectively). Each donor mouse was paired with a single antibiotic-treated recipient. Next, the HICT-FMT and SED-FMT mice were immunized with PLP_139–151 peptide to induce EAE (EAE-induced HICT fecal microbiome transplanted—HICT-FMT-EAE and EAE-induced SED fecal microbiome transplanted—SED-FMT-EAE groups, respectively). EAE clinical (n = 10/group, control saline treated n = 5) and pathological (n = 6/group) severity were compared between groups. To evaluate the effects of ET on gut microbiome composition and systemic metabolites, feces and sera were collected from HICT and SED mice at the end of the training period (n = 5/group) and from HICT-FMT and SED-FMT mice 3 wks post FMT (n = 5/group), and the gut microbiota composition and serum short- chain fatty acids (SCFAs) levels were compared between the experimental and control groups. Additionally, to compare the effects of the FMT after HICT on the systemic autoimmune system, the proliferation of encephalitogenic lymph node cell (LNC) from HICT-FMT and SED-FMT mice was evaluated (n = 8/group).

Fig. 1.

Experimental design to investigate the impact of high-intensity continuous training (HICT) on gut microbiota composition, serum and fecal short- chain fatty acids (SCFA) levels, autoimmune cell proliferation and experimental autoimmune encephalomyelitis (EAE) development. Healthy donor mice were subjected to a 6 week- HICT treadmill running program or a sedentary (SED) period. Their feces were collected and transplanted to naïve recipient mice (HICT-FMT and SED-FMT groups, respectively), followed by proteolipid (PLP)_139–151 peptide immunization to induce EAE. EAE clinical and pathological severity were compared between groups. To evaluate the effects of HICT on the gut microbiota composition and systemic metabolites, feces and sera from HICT and SED mice at the end of the training period and from HICT-FMT and SED-FMT mice 3 wks after FMT were analyzed for gut microbiome profile and serum SCFA levels. Encephalitogenic lymph node cell (LNC) proliferation assay from PLP- immunized HICT-FMT and SED-FMT mice was performed in vitro.

Treadmill exercise training

Healthy mice underwent 6-wks of HICT treadmill running, including pre- and post- training performance tests on a 5-lane treadmill designed for mice (Panlab Harvard Apparatus, USA) as we previously described^4–7. The 6 wks of training started with a 3 wk preparatory period. The running speed was based on exhaustion speed performance tests⁵. HICT was defined as 70–75% of exhaustion speed. HICT significantly improves both maximal speed and exercise tolerance⁵.

Gut microbiota analysis and transplantation

Gut microbiome analyses, identification of unique bacterial taxa and FMT were performed as previously described^15,29,30.

Fecal sample collection

Fecal samples were collected 48 h (hr) after the final exercise session, in the morning following an overnight fast. Additional samples were obtained from naïve mice three wks after FMT, also in the morning. For each collection, mice were placed in cages (2–3 mice per cage), and fresh fecal pellets were retrieved using sterile forceps. Samples were immediately transferred into pre-weighed sterile tubes and stored at − 80 °C until further analysis.

DNA extraction and sequencing

Gut microbial community composition was compared between groups using a culture-independent 16S rDNA gene sequencing approach. Genomic DNA was extracted from weighed fecal samples using the ZymoBIOMICS DNA Miniprep Kit (Zymo Research, CA, USA), following the manufacturer’s protocol. DNA concentration and purity were assessed with a Nanodrop 2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA), and integrity was verified by 1.0% agarose gel electrophoresis. Purified DNA samples were stored at − 20 °C until further analysis. For 16S rRNA gene sequencing, DNA samples were sent to Hylabs (Hy Laboratories Ltd., Rehovot, Israel). The V4 region was amplified using the following primers: forward (CS1_515F) ACACTGACGACATGGTTCTACAGTGCCAGCMGCCGCGGT and reverse (CS2_806R) TACGGTAGCAGAGACTTGGTCTGGACTACHVGGGTWTCT.

Gut microbiota analysis

Raw FASTQ sequencing data were processed and analyzed using the QIIME2 pipeline (version 2024.10), as previously described^15,29,30. Briefly, paired-end reads were demultiplexed with the q2-demux plugin, followed by denoising and clustering using DADA2 via q2-dada2 to enhance taxonomic resolution. Sequence alignment and phylogenetic tree construction for all amplicon sequence variants (ASVs) were performed using MAFFT and fasttree2 through the q2-alignment and q2-phylogeny plugins, respectively. ASVs, representing unique DNA sequences differing by as little as one nucleotide, were used for high-resolution microbial community profiling. Taxonomic classification was performed using the q2-feature-classifier plugin in QIIME 2 with a Naive Bayes classifier trained on the SILVA 138.2 reference database, trimmed to the V4 region, with a minimum confidence threshold of 0.99 to minimize contamination. The feature table was filtered using q2-feature-table to exclude features annotated as mitochondria or chloroplasts, as well as those present in ≤ 20% of samples.

Multiple alpha and beta diversity metrics were employed to comprehensively characterize microbial community structure, using sequence counts. Alpha diversity, reflecting within-sample diversity, was assessed using observed richness, inverse Shannon index, and Faith’s phylogenetic diversity (Faith’s PD), capturing species richness, evenness, and phylogenetic relationships. Beta diversity, representing differences between samples, was visualized via Principal Coordinate Analysis (PCoA) to highlight group-level variation. This approach enabled robust evaluation of both intra- and inter-group diversity. Statistical significance in beta diversity was determined using PERMANOVA implemented in the vegan package in R (version 4.5.0; https://www.r-project.org/). All figures were generated using R.

To identify bacterial taxa associated with HICT versus SED, or HICT-FMT versus SED-FMT, statistical analyses were performed using zero-inflated count models. Specifically, the zero-inflated negative binomial (ZINB) model was applied using the zinbwave package in R, followed by differential abundance testing with DESeq2. Weights were first calculated with zinbwave, which provides a flexible framework for analyzing high-dimensional zero-inflated count data. Size factors were then estimated using DESeq2³¹. To account for multiple comparisons, p-values were adjusted using the Benjamini–Hochberg false discovery rate (FDR) correction, with statistical significance defined as an adjusted p-value (padj) ≤ 0.1. Alpha diversity metrics were calculated after rarefying the ASV count table to an even depth of 10,000 reads per sample. Beta diversity was assessed using Aitchison distance on CLR-transformed count data without rarefaction, and differential abundance testing was conducted using DESeq2 on raw counts with internal normalization.

Fecal microbiota transplantation (FMT)

FMT procedures were conducted in naïve recipient mice. To maximize gut microbiome depletion, recipient mice were administered daily with broad-spectrum of antibiotics (ciprofloxacin 0.1 mg/ml, metronidazole 0.5 mg/ml, vancomycin 0.25 mg/ml) in their drinking water for two wks prior to FMT, as previously described³². Antibiotic treatment was discontinued three days before FMT. FMT was performed once, using fecal material from individual donor mice as follows. Fresh fecal samples were collected from each HICT and SED donor mouse, and stool pellets were suspended in sterile PBS (1.0 ml per pellet) and homogenized. The homogenate was incubated briefly (5 min at 25 °C) to facilitate separation of solids. Each recipient mouse was administered 200 μl of fecal suspension from either a HICT or SED single donor mouse by oral gavage, using a sterile feeding tube (20 ga × 38 mm, Instech Laboratories, Inc.) attached to a 1.0 ml syringe (PIC, cat no. 1510022). FMT was performed once. Following FMT, mice were housed in small groups (2–3 per cage) to minimize cross-contamination. The impact of FMT was evaluated three wks post-transplantation, following assimilation of donor microbiota.

Targeted metabolomics for short chain fatty acid quantification

Fresh faeces and sera samples were collected from HICT and SED mice 48 h after the last exercise bout, and 3 wks following FMT as described above. Blood samples were drawn directly from the heart into EDTA-coated anticoagulant tubes, followed by centrifugation (2000 g, 14 min, 4 °C) to separate plasma. The serum fraction was collected and stored at -80 °C. The targeted metabolomics was performed at the Weizmann Institute, Israel, using LC–MS analysis to detect butyrate, propionate, and acetate SCFAs levels^29,33.

Experimental autoimmune encephalomyelitis (EAE)

The PLP_139–151- induce EAE model in female SJL/JCrHsd mice was employed as previously described^3–8. EAE was induced in recipient mice three wks after FMT from either HICT or SED donor groups, referred to as HICT-FMT-EAE and SED-FMT-EAE, respectively. A control group (Control EAE) received saline via oral gavage. Neurological symptoms were monitored daily for up to 30 days post EAE induction using a standardized clinical scoring system: 0—no symptoms; 1—partial tail tonicity loss; 2—complete tail atony; 3—hind limb weakness and/or impaired righting reflex; 4—hind limb paralysis; 5—quadriplegia; 6—death due to EAE. Disease severity was evaluated using three parameters: onset index (OI), maximal clinical score (MCS), and burden of disease (BOD). Measurements were performed for each mouse at 20 days post-EAE induction, corresponding to the acute phase. The cumulative BOD score was calculated by summing daily clinical scores across the entire follow-up period (area under the curve). The OI was determined based on the day of clinical onset according to the following scale: 5 = days 10–11; 4 = 12–13; 3 = 14–15; 2 = 16–17; 1 = 18–19; 0 =  ≥ 20. Body weights were monitored throughout the experiment to ensure animal welfare and ethical compliance.

Histopathology analyses

Spinal cord histopathology analyses were performed as previously described^3–8. At 15 days post-EAE induction, corresponding to the acute phase of disease, subsets of SED-FMT-EAE and HICT-FMT-EAE mice were euthanized for histopathological evaluation. Mice were deeply anesthetized with a lethal dose of sodium pentobarbital and transcardially perfused via the ascending aorta with ice-cold phosphate-buffered saline (PBS), followed by 4% paraformaldehyde (PFA). Dissected tissues were post-fixed in 4% PFA for 24 h, then processed for paraffin embedding. Serial transverse Sects. (6 μm) were obtained from mid-cervical, mid-thoracic, and mid-lumbar regions of the spinal cord and analyzed for inflammation, demyelination, and axonal injury.

Immunohistochemical staining was performed on adjacent sections using monoclonal rabbit anti-CD3 (RM-9107-SO; 1:800, Thermo Scientific) to identify T cells, and polyclonal rabbit anti- ionized calcium-binding adapter molecule 1 (Iba1 ,019-19741, Wako) to detect microglia. Myelin integrity was assessed using Luxol Fast Blue (LFB; Sigma, S3382-25G) counterstained with nuclear fast red (Sigma, N8002), while axonal preservation was evaluated via Bielschowsky silver staining using silver nitrate (Chem-lab CL00.2614.0250) and ammonia solution (32%, Merck; 1.05426.1000). For each staining protocol, three sections per mouse (one per spinal cord level) were quantified in a blinded manner.

Immune cell infiltration was assessed in H&E-stained sections by counting perivascular immune cells and reporting the average number per mm². CD3⁺ T cells and Iba1⁺ microglia were quantified in both perivascular and parenchymal regions and reported as the total average number per mm². Demyelination was quantified by measuring the area of LFB signal loss. Axonal damage was scored based on silver stain density: 0—uniform staining; 1—sporadic loss; 2—frequent small areas of loss; 3—extensive loss throughout white matter. All image analyses were performed using ImageJ software (version 1.51H, NIH, USA).

In vitro analyses of encephalitogenic lymph node cells (LNCs)

Proliferation assay for autoimmune LNCs was performed as previously described^3–8. At 10 days post-PLP immunization, inguinal lymph nodes were collected from HICT-FMT and SED-FMT mice. LNCs were prepared as single-cell suspensions and cultured for 72 h in the presence of either 10 μg/ml PLP peptide, 2.5 μg/ml concanavalin A (ConA), or without stimulation. Cellular proliferation was assessed by flow cytometric analysis of bromodeoxyuridine (BrdU) incorporation, as previously described. Proliferative responses were quantified as the fraction of BrdU-positive cells among total LNCs or among CD3 + cells. All samples were analyzed using a Cytomics FC 500 flow cytometer (Beckman Coulter, Life Science) and CXP analysis software (version 2.1; Informer Technologies, Inc).

Statistical analysis

The normality of variable distributions was assessed using the Shapiro–Wilk test, followed by appropriate statistical comparisons. For two-group comparisons, either the unpaired Student’s t-test or the two-tailed Mann–Whitney test was applied, depending on the outcome of the normality test. Comparisons involving more than two groups were performed using one-way analysis of variance (ANOVA), and, when significant, the Newman-Keuls multiple comparison test was used to identify specific differences. Pairwise statistical comparisons were performed only between groups relevant to the primary hypotheses (e.g., HICT-FMT vs. SED-FMT) to maintain interpretability and minimize the risk of Type I error. Comparisons not aligned with these hypotheses, such as those involving the Control-EAE (saline) group, were not included. Each experiment was independently repeated two to three times to ensure consistency. Data analysis was conducted using GraphPad Prism software (version 5). Statistical significance was defined as p < 0.05. All results are presented as mean ± standard error of the mean (SEM).

Results

HICT- derived gut microbiota attenuates clinical severity, neuroinflammation, and CNS damage in EAE mice

To examine the impact of the gut microbiota after ET on the development of EAE, feces were collected from donor HICT and SED mice and transplanted into recipient mice, followed by PLP immunization (EAE-induced HICT fecal microbiome transplanted—HICT-FMT-EAE, and EAE-induced SED fecal microbiome transplanted—SED-FMT-EAE, n = 10/group; Fig. 2). Saline-treated PLP-immunized mice served as controls (Control-EAE, n = 5; Fig. 2). The clinical course of EAE was monitored, and the average onset of disease, maximal clinical score at the acute phase, and overall burden of disease were recorded. Body weights were monitored throughout the study for ethical compliance and animal welfare; however, these data are not presented as they were not predefined outcome measures.

Fig. 2.

Fecal microbiota transplantation (FMT) from high-intensity continuous training (HICT) donors induces attenuation of experimental autoimmune encephalomyelitis (EAE) in recipient mice. Healthy donor mice were subjected to a HICT treadmill-running program, or sedentary (SED) period, and their feces were transplanted to naïve recipients (HICT-FMT and SED-FMT groups, respectively), followed by proteolipid (PLP)_139–151 peptide immunization to induce EAE. Clinical course (A) and clinical parameters (B–D) of EAE following FMT from HICT (HICT-FMT-EAE, n = 10) and sedentary (SED-FMT-EAE, n = 10) mice and saline treated control mice (Control-EAE, n = 5). The severity of EAE was scored according to a 0–6 scale. Mice that were transplanted with gut microbiome after HICT developed a significantly milder EAE course (A). HICT-FMT-EAE mice exhibited delayed onset (B), reduced maximal clinical scores (MCS, C), and a significantly reduced overall burden of disease (BOD, D). Groups of HICT-FMT-EAE and SED-FMT-EAE mice were sacrificed for histopathology analysis to evaluate CNS inflammation and tissue damage at 15 days post EAE induction, at the acute phase of disease (E–S, n = 6/group). Evaluation of inflammatory infiltrates (E–G), CD3 + T cells (H–J) and Iba-1 + microglia (K–M), demyelination (N–P) and axonal injury (Q–S) on cross sections of the spinal cords of SED-FMT-EAE (E,H,K,N,Q) or HICT-FMT-EAE (F,I,J,O,R) mice. Quantification of inflammatory infiltrations (G,J,M), demyelination and axonal damage (P,S) in spinal cord white matter. In HICT-FMT-EAE there were less inflammatory infiltrates (G), CD3 + T cells (J), Iba-1 + microglia (M) infiltrations and reduced demyelination (P) and axonal injury (S) within the CNS than in SED-FMT-EAE. Arrows indicate perivascular infiltrates (E,F), perivascular CD3 + T cells (H,I) and Iba1 + microglia (K,L). Dashed line indicate areas of demyelination (N,O) and axonal damage (Q,R). Scale bars = 100 μm; Data are presented as mean ± SEM. Each dot represents a single sample. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

All mice remained in good health throughout the experiment, and no mortality occurred in any experimental group. PLP immunization induced a significantly milder clinical course of EAE in HICT-FMT-EAE mice compared to both control and SED-FMT-EAE groups (Fig. 2A), with delayed onset (> 2 days, Fig. 2B), ~ 30% lower peak clinical scores (Fig. 2C), and ~ 50% reduced overall disease burden (Fig. 2D). The control EAE (saline) group exhibited slightly higher, non-significant, clinical severity compared to the SED-FMT-EAE group (Fig. 2B–D), which may reflect natural variability in disease onset and progression commonly observed in EAE models.

To determine whether the attenuated disease in recipient mice induced by FMT from HICT mice was associated with decreased neuroinflammation and tissue damage, histological analysis was performed in a separate experiment at the peak of disease (day 15 post-induction), to compare HICT-FMT-EAE and SED-FMT-EAE groups (Fig. 2E–S; n = 6/group). Histological analysis revealed significantly reduced neuroinflammation in HICT-FMT-EAE mice, including ~ 25% fewer perivascular infiltrates (Fig. 2E–G), ~ 50% fewer CD3⁺ T cells (Fig. 2H–J), and ~ 40% fewer Iba-1⁺ microglia (Fig. 2K–M). Additionally, HICT-FMT-EAE mice showed > 30% less demyelination (Fig. 2N–P) and ~ 15% greater axonal preservation, with minimal severe axonal loss (Fig. 2Q–S). These findings indicate that FMT from HICT donors effectively attenuates EAE development and CNS pathology in recipient mice.

HICT induces gut microbiota alterations

To identify bacterial taxa potentially mediating HICT’s protective effects, we analyzed fecal microbiota from healthy HICT and SED mice (Fig. 3; n = 5/group). Sequencing the V4 regions of the 16S rRNA gene generated approximately 287,622 sequences reads after filtration. The mean was 18,598 sequence reads per mice, and the filtered dataset contained 544 features covering 8 phyla, 10 classes, 23 orders, 31 families and 59 genera (Supplementary Table 1).

Fig. 3.

High-intensity continuous training (HICT) alters gut microbiota composition. Healthy mice were subjected to a HICT treadmill-running program, or SED period. Their feces and blood were collected and analyzed for gut microbiota profile (n = 5/group). A: Box plots illustrate alpha diversity indices of observed amplicon sequence variants (ASVs), Shannon-Weiner index and Faith’s phylogenetic diversity (PD) in bacterial microbiota. B: Principal-coordinate analysis (PCoA) of the SED and HICT mice. Taxa without genus-level annotation were retained for PCoA at their lowest taxonomic rank, increasing the number of plotted features relative to the main text. Relative abundance of the gut microbiome composition at the Order’s level (C), Genus (D), and Phyla (E) level. Differential abundance analysis plot associated with SED and HICT at the family (F) and genus (G) levels. H–O: Log2 fold-change plots display all differentially tested taxa with family- or genus-level annotation. For clarity, ASV-level boxplots show representative statistically significant ASVs selected from the DESeq2 analysis. Statistical analyses were performed by PERMANOVA and DEseq2. Each dot represents a single sample. Data are presented as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001.

Alpha diversity metrics (observed taxa, inverse Shannon index, Faith’s PD) showed a significant decrease in the PD and Shannon indexes diversity in the HICT compared to SED group (p < 0.05, Fig. 3A). However, PCoA (Fig. 3B) and PERMANOVA (Bray–Curtis p = 0.016, Fig. 3C,D) revealed distinct microbial community structures, indicating HICT-induced compositional shifts. The dominant phyla in the gut microbiota across all groups were Bacteroidetes, Firmicutes, Tenericutes and Proteobacteria (Fig. 3E). Most samples had low read counts (< X 100 reads) for other phyla, with some outlying samples (Fig. 3C–E).

We further identified associations between the groups in the individual amplicon sequence variants (ASVs). Due to the large number of zero counts of most ASVs, statistical analysis using zero-inflated count models was performed. Then, differential analysis testing using DESeq2, a method for differential analysis of count data was performed, identified several taxa associated with HICT (Fig. 3F–O, Supplementary Table 1, marked in bold). HICT was significantly associated with five features, including the genera Escherichia and Faecalimonas (Fig. 3J,H, respectively) and specific ASVs from the Lachnospiraceae family (Fig. 3H–O; corrected p-values ranging from 0.0045 to 0.077). In contrast, six features were associated with SED, including several Lachnospiraceae ASVs lacking genus-level classification (Fig. 3H–O); log₂ fold change range: 6.5 to − 4.07).

These findings highlight the heterogeneous response of this family and suggest that HICT selectively modulates specific microbial populations.

FMT from HICT donor mice induces distinct gut microbiota composition

To assess whether HICT-induced microbiota changes contribute to EAE protection, we analyzed gut microbiota composition in recipient mice after FMT from HICT or SED donors (HICT fecal microbiome transplanted—HICT-FMT and SED fecal microbiome transplanted—SED-FMT; Fig. 4; n = 5/group). 16S rRNA sequencing yielded 645,615 high-quality reads. The mean was 34,405.5 sequence reads per mice, and the filtered dataset contained 580 features covering 8 phyla, 10 classes, 23 orders, 31 families and 59 genera (Supplementary Table 2).

Fig. 4.

Fecal microbiota transplantation (FMT) from high-intensity continuous training (HICT) donors induces alterations in gut microbiota composition. Healthy donor mice were subjected to a HICT treadmill-running program, or SED period. Their feces were collected and transplanted to naïve recipients (HICT-FMT and SED-FMT, n = 5/group). Feces and blood samples from HICT-FMT and SED-FMT were collected to analyze the gut microbiome profile and serum SCFA levels. A: Box plots illustrate alpha diversity indices of Observed amplicon sequence variants (ASVs), Shannon-Weiner index and Faith’s phylogenetic diversity (PD) in bacterial microbiotas. B: Principal-coordinate analysis (PCoA) of the SED-FMT and HICT-FMT mice. Taxa without genus-level annotation were retained for PCoA at their lowest taxonomic rank, increasing the number of plotted features relative to the main text. Relative abundance of the gut microbiota composition at the Order’s level (C) and Genus’s level (D). Differential abundance analysis plot associated with SED-FMT and HICT-FMT at the family (E) and genus (F) levels G-N: Count plots of ASVs that were found significantly different in the DESeq2 analysis between SED-FMT and HICT-FMT mice. Log2 fold-change plots display all differentially tested taxa with family- or genus-level annotation. For clarity, ASV-level boxplots show representative statistically significant ASVs selected from the DESeq2 analysis. Statistical analyses were performed by PERMANOVA and DEseq2. Each dot represents a single sample. Data are presented as mean ± SEM. *p < 0.05, **p < 0.01.

The number of observed taxa, the inverse Shannon diversity indices, and Faith’s PD were used to analyze the bacterial diversity in HICT-FMT and SED-FMT groups. Alpha diversity was similar between groups (p > 0.05, Fig. 4A).

PCoA (Fig. 4B) and PERMANOVA (Bray–Curtis p = 0.01, Fig. 4C,D) confirmed distinct microbial profiles between HICT-FMT and SED-FMT mice. The dominant bacterial phyla in the gut microbiota across groups were Bacteroidetes, Firmicutes, and Proteobacteria. Most samples showed low reads counts for other phyla, with some outlying samples.

Differential abundance analysis at the ASV level was performed using DESeq2 to identify taxa associated with SED-FMT and HICT-FMT (Fig. 4E–N). SED-FMT was associated with a significant enrichment of specific ASVs within the Ruminococcaceae and Lachnospiraceae families, including ASVs assigned to Oscillospira, Roseburia, and Parabacteroides genera (Fig. 4E–N; adjusted p < 0.05; positive log₂ fold change). In contrast, HICT-FMT was associated with an increase of multiple ASVs primarily belonging to the Faecalimonas genus (Fig. 4E–N; adjusted p < 0.05; negative log₂ fold change). Notably, the ASVs differentially associated with SED-FMT and HICT-FMT largely belonged to the same microbial families identified in the original HICT versus SED comparison, indicating consistency in the taxonomic groups affected across both experiments (Supplementary Table 2, marked in bold).

These results indicate that FMT from HICT donors induces distinct microbial signatures in recipients, potentially contributing to reduced EAE severity.

HICT and FMT from HICT donor induce decreased serum levels of SCFA mice

To assess microbial activity, we measured serum and fecal levels of SCFAs (n = 5/group; Fig. 5A–I). HICT mice exhibited significantly lower serum levels of acetate, propionate, and butyrate (Fig. 5A–C), whereas fecal SCFA levels remained unchanged (p > 0.05, Fig. 5D–F).

Fig. 5.

High-intensity continuous training (HICT) and fecal microbiota transplantation (FMT) from HICT donors induce reduction of serum levels of short- chain fatty acids (SCFAs). Healthy donor mice were subjected to a HICT treadmill-running program, or SED period. Their feces were collected and transplanted to naïve recipients (HICT-FMT and SED-FMT). Feces and blood samples from HICT, SED, HICT-FMT and SED-FMT were collected to analyze serum SCFA levels (n = 5/group). Quantitation of serum (A–C) and fecal (D–F) SCFA levels in HICT and SED mice. Serum concentrations of butyrate, propionate, and acetate were significantly reduced in HICT mice compared to SED mice. No significant differences in fecal butyrate, propionate, and acetate concentrations were observed between the two groups. G-I: Quantitation of serum SCFA levels in HICT-FMT and SED-FMT mice. Serum concentrations of butyrate, propionate, and acetate were significantly reduced in HICT-FMT mice compared to SED-FMT mice. Each dot represents a single sample. Data are presented as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001.

Consistent with donor profiles, HICT-FMT mice also exhibited significantly lower serum levels of propionate and acetate (Fig. 5G–I), while fecal SCFA levels remained unchanged. These findings indicate that FMT from HICT donors induces distinct metabolic signatures in recipient mice and preserves key serum metabolic alterations observed in the donors, potentially contributing to reduced EAE severity.

FMT from HICT donor mice modulates PLP- reactive LNC proliferation in recipient mice

The in vivo experiments showed that FMT from HICT donor mice reduced the potency of PLP-reactive LNCs to induce brain inflammation in transplanted mice. We therefore hypothesized that FMT from HICT mice may actively interfere with the generation of effector T cells. To test whether HICT-FMT modulates autoreactive T cell responses, we assessed in additional independent cohorts the proliferation of PLP-reactive LNCs from HICT-FMT and SED-FMT mice using BrdU incorporation and flow cytometry (Fig. 6, n = 8/group). While overall LNC proliferation in response to PLP stimulation in vitro was similar between groups (p > 0.05, Fig. 6A–C), HICT-FMT significantly reduced the proportion of proliferating CD3⁺BrdU⁺ T cells by ~ 50% compared to SED-FMT (p < 0.05, Fig. 6D–F), indicating suppression of PLP-specific T cell responses.

Fig. 6.

Fecal microbiota transplantation (FMT) from high-intensity continuous training (HICT) donors induces reduction in autoimmune T cell proliferation in recipient mice. Healthy mice were subjected to a HICT treadmill-running program, or sedentary (SED) period. Their feces were collected and transplanted to naïve recipients (HICT-FMT and SED-FMT), followed by proteolipid protein (PLP) immunization. Lymph node cells (LNCs) were excised from HICT-FMT and SED-FMT mice at 10 days post PLP immunization, stimulated in vitro for 72 h with proteolipid (PLP) peptide or Concanavalin A (ConA), and analyzed for bromedeoxyuridine (BrdU) incorporation by flow cytometry (n = 8/group). Quantitation of BrdU + cells out of total LNCs in response to PLP (A) or ConA (G) stimulation, and BrdU + cells in CD3 + T cells in response to PLP (D) or ConA (J) stimulation in SED-FMT and HICT-FMT mice. Representative flow cytometry plots of BrdU + cells out of total LNCs in response to PLP (B,C) or ConA (H,I), and in CD3 + T cells in response to PLP (E,F) or ConA (K,L) stimulation in SED-FMT (B,E,H,K) and HICT-FMT (C,F,I,L) mice. A significant reduction in BrdU incorporation was observed in the fraction of BrdU + CD3 + T cells in response to PLP (D–F). No differences were noted in the fraction of BrdU + LNCs in response PLP (A–C), nor in BrdU + LNCs (G–I) or BrdU + CD3 + T cells (J–L) in response to ConA. Data are presented as mean ± SEM. *p < 0.05.

To determine if this effect was antigen-specific, we stimulated LNCs with the non-specific mitogen ConA. Both groups showed comparable overall proliferation (p > 0.05, Fig. 6G–I), and comparable fractions of CD3 + T cells entering the cell cycle following ConA stimulation (p > 0.05, Fig. 6J–L), suggesting that the inhibitory effect of HICT-FMT was specific to the PLP autoantigen. These results indicate that FMT from HICT donors induces immune tolerance by selectively dampening encephalitogenic T cell responses.

Discussion

This study is the first to investigate the impact of ET-modulated gut microbiota on EAE severity and associated metabolic profiles, integrating microbiota transfer with metabolic analysis in the context of neuroinflammation. Our results provide compelling evidence that HICT induces gut microbiota alterations that are associated with protective effects against EAE, a widely used murine model of MS. By employing FMT, we demonstrate that the gut microbiota from HICT-trained mice confers neuroprotection in EAE recipients, associated with delayed disease onset, reduced clinical severity, and attenuated neuropathological hallmarks of EAE. These findings support the hypothesis that ET-induced microbial reprogramming may contribute to modulation of autoimmune neuroinflammation and suggest that the gut microbiome is a critical intermediary in the ET-neuroimmunity axis. However, while our data reveal correlations between microbial composition, metabolite profiles, and disease outcomes, the study design does not allow determination of causal mechanisms.

FMT, initially established as a highly effective treatment for recurrent Clostridioides difficile infections, has demonstrated expanding utility in modulating immune responses and ameliorating inflammation in chronic gastrointestinal and systemic autoimmune diseases^26,34. In ulcerative colitis, randomized controlled trials and meta-analyses have consistently shown improved rates of clinical and endoscopic remission following FMT compared to placebo, with a safety profile comparable to standard interventions^26,34. Beyond gastrointestinal disorders, preclinical and clinical studies suggest a therapeutic potential for FMT in extraintestinal autoimmune diseases such as MS. Indeed, several studies have consistently demonstrated that patients with MS harbor a distinct gut microbiota composition compared to healthy individuals, characterized by altered abundances of specific taxa such as Akkermansia^25,28. Particularly, FMT from MS patients into germ-free mice has been shown to enhance CNS-specific autoimmunity, whereas targeted modulation of the gut microbiota, including via FMT, has been associated with reduced pro-inflammatory profiles and attenuation of disease severity^25,28. Furthermore, MS-derived ileal microbiota transplantation into germfree transgenic mice prone to develop spontaneous EAE induced substantially higher rates of disease than analogous material from healthy twin donors³⁵. It was suggested that the active organisms were Eisenbergiella tayi and Lachnoclostridium, members of the Lachnospiraceae family. Our findings are consistent with and extend these observations and support the hypothesis that ET-induced gut microbiome reprogramming can elicit systemic protective effects that curb the onset and progression of autoimmune responses within the CNS. This is consistent with the established notion that commensal microbiota and their metabolites influence CNS immune homeostasis via the gut-brain and gut-immune axes^36,37.

Furthermore, our findings demonstrate that mice receiving FMT from HICT-trained donors exhibited significantly milder EAE phenotypes compared to those receiving FMT from sedentary donors. Histopathological analyses revealed reduced perivascular immune cell infiltration, decreased T cell and microglial accumulation, and diminished demyelination and axonal injury in the CNS of HICT-FMT recipients. These results mirror our previous work demonstrating that HICT reduces the invasiveness of encephalitogenic T cells and preserves CNS integrity⁶. Importantly, the current study supports the gut microbiota as a sufficient contributor to these effects, independent of direct ET or central adaptations induced by ET.

High-resolution 16S rRNA sequencing revealed distinct microbial community structures between HICT and SED mice, accompanied by a significant reduction in phylogenetic diversity and Shannon diversity in HICT mice. This contrasts with some animal studies reporting increased diversity post-ET^38,39, while others, particularly in humans, found no significant changes, highlighting variability based on host species, ET type, duration, and methodology^40–42.

Notably, HICT was associated with alterations in the abundance of several gut microbial taxa, including a heterogeneous response within the Lachnospiraceae family, with certain ASVs enriched in HICT while others were more abundant in SED, alongside increased Faecalimonas and Escherichia and decreased Mucispirillum. This heterogeneity within Lachnospiraceae reflects the well-documented functional diversity of this family, as some members are key SCFA producers with anti-inflammatory properties, whereas others have been implicated in pro-inflammatory processes and autoimmune susceptibility^43,44. The enrichment of specific Lachnospiraceae taxa in HICT mice is consistent with previous reports from both animal and human studies, where endurance and moderate-to-vigorous training increased butyrate-producing genera^45–47. At the same time, the higher abundance of unidentified Lachnospiraceae ASVs in SED mice underscores that not all members of this bacterial family respond uniformly to ET, cautioning against generalized interpretations at the gut microbiota family level alterations.

Similarly, the increased abundance of Faecalimonas in HICT mice may reflect shifts toward carbohydrate-degrading taxa that promote cross-feeding and SCFA production⁴⁸, or alternatively expansion of taxa linked to inflammatory states⁴⁴. However, these possibilities cannot be distinguished using 16S rRNA sequencing alone and therefore remain speculative. The rise in Escherichia observed in HICT contrasts with large human cohorts, where higher fitness levels correlated with lower Enterobacteriales and greater enrichment of SCFA producers⁴⁹, yet rodent studies indicate that intensive ET can disrupt gut barrier integrity and transiently favor Proteobacteria expansion^46,50. Since Escherichia coli is a known pathobiont capable of triggering LPS-mediated inflammation⁵¹, its enrichment in our model may reflect HICT- associated stressors. Thus, the observed increase in Escherichia should be interpreted cautiously and does not imply a direct pro-inflammatory role in this model.

Finally, the reduced abundance of Mucispirillum, which has been reported to expand in chemically induced colitis models and has been associated with intestinal inflammation, suggests a potential role as a mucus-associated pathobiont under inflammatory conditions⁵². These compositional changes underscore the nuanced and context-dependent nature of ET-induced microbiota modulation and caution against broad generalizations at the family or genus level. Because our microbiome profiling is based on 16S rRNA sequencing, functional interpretation is necessarily indirect. Although several ASVs enriched in the HICT group map to taxa previously associated with immune regulation or metabolic activity, these associations should be viewed as hypothesis-generating rather than mechanistic. Direct functional approaches (metagenomics, metabolomics, or targeted immune assays) will be required to establish causal links between specific microbial functions and EAE modulation.

SCFAs, mainly acetate, propionate, and butyrate, are key metabolites produced by the gut microbiota during the fermentation of dietary fibers and resistant starches in the colon. These molecules play crucial roles in maintaining gut health, modulating the immune system and affecting metabolic regulation and the brain via the gut-brain axis⁵³. Interestingly, despite the enrichment of SCFA-producing taxa, serum levels of acetate, propionate, and butyrate were reduced in HICT mice, while fecal SCFA levels remained unchanged. This finding challenges the prevailing assumption that increased SCFA production is a primary mechanism of ET-induced immunomodulation. Several explanations are plausible. First, the reduced serum SCFA levels may reflect enhanced colonic absorption and utilization rather than decreased production. ET has been shown to upregulate SCFA transporters and increase SCFA oxidation in peripheral tissues^54,55. Second, shifts in microbial metabolism under HICT may favor alternative metabolic pathways, such as amino acid fermentation or neurotransmitter synthesis, at the expense of SCFA output under certain conditions⁵⁶. Third, recent studies suggest that the immunomodulatory effects of SCFAs depend not only on their abundance but also on the context of immune activation and host metabolic state⁴³. These data suggest that the protective effects of HICT may not be mediated solely by elevated SCFA production but may involve other microbial-derived metabolites, such as tryptophan catabolites, bile acids, or microbial-associated molecular patterns. Similar findings have emerged in studies of microbiome-driven modulation of neuroinflammation, where specific microbial taxa or metabolites, rather than overall SCFA concentrations, drive immune phenotypes⁵⁷. Differences between fecal and serum SCFA levels likely reflect altered host absorption and utilization rather than changes in microbial SCFA production. Accordingly, our data do not support a model in which elevated SCFA levels are the sole or primary mediators of protection, but rather suggest that SCFAs represent one component of a broader, context-dependent microbial–host interaction.

Following antibiotic-mediated depletion of the resident microbiota, FMT recipient mice developed distinct microbial communities depending on donor source. Differential abundance analysis revealed pronounced compositional differences between SED-FMT and HICT-FMT mice, with the majority of differentially abundant ASVs mapping to taxa within the order Clostridiales, including members of the Lachnospiraceae and Ruminococcaceae families. In particular, SED-FMT mice exhibited higher abundances of multiple Clostridiales-associated ASVs, which may reflect secondary engraftment of low-abundance donor taxa or selective expansion under sedentary-associated gut conditions. This pattern is consistent with reports linking specific Clostridiales clusters to pro-inflammatory states and unfavorable metabolic profiles, as summarized in recent reviews integrating evidence from both human and rodent studies^50,58. Conversely, HICT-FMT mice were characterized by enrichment of distinct ASVs within the same taxonomic groups, underscoring the functional and ecological heterogeneity of these bacteria. Several differentially abundant ASVs mapped to genera such as Faecalimonas and Oscillospira, which have been linked to carbohydrate fermentation, SCFA production, and immune modulation⁴⁸. As observed in the donor microbiome analysis, the presence of both positively and negatively enriched ASVs within shared taxonomic families cautions against broad functional interpretations at higher taxonomic levels. Together, these findings indicate that following microbiota depletion, FMT establishes donor-dependent microbial signatures while allowing recipient-specific ecological factors to shape the resulting community. Given the use of 16S rRNA sequencing, functional interpretations remain indirect and should be considered hypothesis-generating.

Serum SCFA profiles in FMT recipients mirrored those of their respective donors, with HICT-FMT mice showing reduced acetate and propionate levels. These findings indicate that while not all donor-associated features were equally retained, core compositional and metabolic traits were successfully transferred through FMT, confirming successful microbial engraftment, and supporting the concept that ET-trained microbiota can transfer functional immunomodulatory properties.

Functionally, HICT-FMT recipients exhibited reduced proliferation of LN-derived T cells in response to PLP stimulation, indicating systemic modulation of autoreactive T cell responses. This complements our previous findings where HICT downregulates the expression of the migration-associated integrins very late antigen (VLA)-4 and leukocyte function-associated antigen (LFA)-1, and chemokine receptor CCR5 in encephalitogenic T cells⁶. The current data extends this paradigm by demonstrating that microbial reprogramming may be consistent with immune modulation, though not tested directly. These may include modulation of dendritic cell function, expansion of regulatory T cells, or alterations in cytokine signaling cascades, all of which warrant further investigation.

The specificity of CD3⁺ T-cell responses to PLP raises the possibility that HICT-associated microbiota may differentially modulate antigen-specific versus global T-cell regulation. Accumulating evidence indicates that gut microbiota can shape T-cell phenotypes by influencing the balance between pro-inflammatory Th17 cells and regulatory T cells (Tregs), largely through microbial metabolites such as SCFAs and other immunomodulatory products^59–61. In particular, defined commensal taxa within the order Clostridiales, including members of the Lachnospiraceae and Ruminococcaceae families, have been shown to promote Foxp3⁺ Treg differentiation and attenuate EAE severity, whereas other microbial configurations favor Th17 polarization^59,61. In the present study, differential abundance was assessed at the ASV level using DESeq2, with taxonomic assignment used solely for biological interpretation. Several ASVs enriched in the HICT group were mapped to taxa previously linked to immune regulation, supporting a plausible association between HICT-driven microbiota remodeling and reduced EAE severity. However, whether these microbial changes selectively modulate PLP-reactive T cells or exert broader effects on T-cell regulation remains to be determined and will require targeted functional and antigen-specific assays in future studies.

Limitations

Although our study provides robust evidence for gut microbiota-mediated immunomodulation in EAE, several limitations merit consideration. First, while 16S rRNA sequencing offers valuable taxonomic insights, it lacks the resolution to identify strain-level differences or functional gene content. Future studies employing metagenomic, metatranscriptomic, and metabolomic approaches are needed to map the functional landscape of HICT-altered microbiomes and identify specific microbial taxa or metabolites responsible for the observed effects. Second, while our model isolates the microbiota’s role, it does not account for potential interactions between ET-induced neural adaptations and microbial signals. ET is known to influence neurotrophic factors, synaptic plasticity, and neurogenesis, which may synergize with microbial signals to modulate CNS inflammation. Third, the absence of longitudinal microbiome and metabolite measurements following EAE induction restrict our ability to determine whether transplanted microbiota or SCFA levels persist, fluctuate, or correlate with disease progression. While our design focused on pre-induction characterization to address the primary hypothesis, dynamic profiling during and after EAE would provide valuable mechanistic insights into the stability and functional impact of the transplanted microbiota. Future studies should incorporate serial sampling of fecal microbiota and metabolite levels at multiple time point post-induction, alongside clinical and immunological assessments, to establish temporal relationships and strengthen causal inference between microbiota changes and disease outcomes. Such an approach would help clarify whether microbiota-mediated protection is sustained throughout the course of disease or limited to early phases. Fourth, while our findings demonstrate that microbiota from HICT-trained mice can transfer protection against EAE, they do not establish that microbiota is the sole or necessary mediator of ET-induced benefits. The observed protection supports a contributory role of ET-conditioned microbiota, but additional mechanisms, such as direct neuroprotective effects or systemic immune modulation, may also be involved. Additionally, recipient mice were housed under conventional, non-barrier conditions, which may have permitted environmental microbial exposure and limited the fidelity of fecal microbiota engraftment following transplantation. Future studies incorporating microbiota disruption in trained mice or complementary mechanistic approaches will be essential to determine whether microbiota is required for ET-induced protection. Fifth,

given the design of the immune analyses, limited to LN at day 10 post PLP immunization, these findings should be interpreted as associative, not causal, and do not establish a mechanistic link between microbiota, metabolites, and T-cell suppression. Finally, the translatability of these findings to human MS remains to be validated. While pre-clinical models provide mechanistic insights, human studies are essential to determine whether similar microbial shifts occur in response to ET and whether these changes correlate with clinical outcomes. Although FMT recipients showed similar key compositional features of HICT donor microbiota, engraftment is influenced by host ecological limitations, and alpha diversity differences indicate that recipient communities represent a reshaped rather than identical donor microbiota. Therefore, while the data supports a microbiota-dependent contribution to EAE protection, identification of specific mechanistic taxa will require further functional validation.

Despite these limitations, our findings have important implications for the development of microbiota-targeted therapies in MS. The demonstration that HICT-induced microbial changes can be transferred via FMT to confer neuroprotection suggests that the gut microbiome may serve as a conduit for the beneficial effects of ET. This opens the possibility of leveraging microbial interventions, such as FMT, probiotics, prebiotics, or postbiotics, to mimic or enhance the effects of ET in individuals unable to engage in high-intensity ET. Moreover, identifying specific microbial taxa or metabolites responsible for these effects could inform the development of precision microbiome-based therapeutics tailored to modulate neuroinflammation.

In conclusion, this study provides evidence that gut microbiota is a key contributing factor of the protective effects of HICT in EAE. By demonstrating that FMT from trained donors can recapitulate the benefits of ET, we highlight a novel pathway linking ET, microbial ecology, and neuroimmune regulation. These findings underscore the therapeutic potential of microbiota-targeted strategies in MS and support the integration of ET as a modifiable environmental factor in the management of neuroinflammatory diseases. Future studies should aim to identify the specific microbial and metabolic mediators of these effects and explore their translational relevance in human MS.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

LH and OA contributed to the acquisition, analysis, and interpretation of data. YG, NA, NF, PT, IS contributed to the acquisition of data. NG, AK, TB-H, SNV contributed to the interpretation and drafting of the work. OE contributed to the conception, design, interpretation and drafting of the work. All authors read and approved the final manuscript.

Funding

This work was supported by the Ministry of Innovation, Science & Technology (no. – 0008147).

Data availability

The 16S rRNA gene sequencing data generated in this study have been deposited in the NCBI Sequence Read Archive (SRA) under the BioProject accession number PRJNA1356509. The data will be made publicly available upon publication.

Competing interests

The authors declare no competing interests.