Intermittent fasting after disease start reduces brain inflammation and nerve damage through a stress-response and cell cleanup pathway in a multiple sclerosis model

C57BL/6J mice were purchased from the Laboratory Animal Center of Shaanxi Normal University (Xi'an, China), and C57BL/6 N mice were obtained from Beijing Huafukang Biotechnology Co., Ltd. (Beijing, China). Trib3 heterozygous knockout (Trib3^⁺/⁻) mice were purchased from Cyagen Biosciences (Guangzhou, China) (strain number KOCMP-228775-Trib3-B6N-VA, NCBI ID: 228775, catalog # S-KO-06197). Homozygous Trib3^⁻/⁻ mice and corresponding wild-type (WT) littermates were obtained by breeding Trib3^⁺/⁻ heterozygotes with C57BL/6 N WT mice, followed by intercrossing of the heterozygous offspring. All mice were housed under specific-pathogen-free (SPF) conditions (24 ± 2 °C; 12 h light/dark cycle) and were group-housed, five per cage. Food was provided according to the experimental design (ad libitum, time-restricted, or calorie-restricted feeding), and water was available ad libitum. All procedures were approved by the Animal Research and Ethics Committee of Shaanxi Normal University and complied with institutional and national guidelines for animal welfare. Mice were acclimated for at least one week before experimentation and randomly assigned to experimental groups (n values provided in figure legends). All animal experiments were designed and reported in accordance with the ARRIVE 2.0 guidelines. Female mice aged 8 weeks were used for EAE induction to ensure consistency across experimental groups, as female C57BL/6 mice are more susceptible and display a more reproducible EAE phenotype than males [29, 30]. Active experimental autoimmune encephalomyelitis (EAE) was induced in C57BL/6 mice by immunization with the MOG_{35 − 55} peptide emulsified in complete Freund's adjuvant (CFA) containing Mycobacterium tuberculosis H37Ra, followed by pertussis toxin (PTX) injections. Briefly, MOG_35–55 powder was dissolved in sterile water to prepare a 50 mg/mL stock, which was then diluted in PBS to 2 mg/mL working solution. Equal volumes of MOG solution and CFA (containing ~ 1 mg/mL MT H37Ra) were emulsified by mechanical homogenization (e.g. 6 × 30 s bead-beating) until a stable emulsion formed (verified by drop test in water). Each mouse received a total of 200 µg MOG_35–55 (100 µL per injection site) subcutaneously at 2 flank/back sites. Immediately after immunization (day 0) and again 48 h later, 200 ng of pertussis toxin (in 200 µL PBS, IP) were administered to further enhance blood–brain barrier permeability and disease induction. This active immunization protocol (MOG/CFA + PTX) is a standard C57BL/6 EAE model.

Mice were examined daily from day 7 post-immunization onward for neurological deficits by two blinded observers. A 0–5 clinical scoring scale was used (no symptoms to moribund) with 0.5-point increments for intermediate signs. Criteria were: 0 = normal; 0.5 = tail tip atony; 1 = limp tail; 1.5 = partial hind limb weakness; 2 = hind limb paresis; 2.5 = severe paresis (one limb paralyzed, other weak); 3 = bilateral hind limb paralysis; 3.5 = hind limb paralysis plus partial forelimb weakness; 4 = complete paralysis (hind and forelimbs, mouse cannot right itself); 5 = moribund or dead. These scores reflect ascending paralysis in EAE. Body weight was recorded concurrently. Cumulative and average clinical scores were calculated for each group. Mice reaching a score of 4 or above were monitored closely and euthanized if humane endpoints were met, which were defined as moribund state or inability to access food and water (score = 5). In our cohort, no mice reached this endpoint; therefore, all animals, including those that transiently scored 4, were retained in the clinical scoring dataset and were available for downstream analyses. The scoring and endpoint criteria followed established institutional guidelines for EAE monitoring and humane care [31, 32].

Intermittent fasting (IF) was implemented as an alternate-day feeding schedule. In this regimen, mice underwent 24 h of food deprivation ("fasting day") followed by 24 h of ad libitum (AL) feeding ("feeding day"). Water was available at all times. Four feeding groups were established: (1) EAE_AL: EAE induction, fed ad libitum throughout; (2) EAE_postIF: Immunized on day 0, ad libitum feeding until day 10, then IF every other day from day 10 until endpoint; (3) EAE_IF: Immunized on day 0, IF every other day from day − 14 until endpoint; (4) EAE_preIF: Immunized on day 0, IF every other day from day − 14 to day 10, followed by ad libitum feeding thereafter.

On "feeding days," food pellets were placed in the cage at 17:00; on "fasting days," pellets were removed at 17:00. Food intake and body weight were monitored daily. Alternate-day fasting has been shown to improve metabolic and inflammatory parameters in rodents.

EAE_postTR: Mice were immunized on day 0 and maintained on ad libitum feeding until day 10. From day 10 until the experimental endpoint, they were subjected to a time-restricted feeding (TR) regimen consisting of a 16 h fasting and 8 h feeding cycle. The feeding window was set from 8:00 p.m. to 4:00 a.m., corresponding to the dark (active) phase of the circadian cycle. During the feeding period, mice had free access to standard chow, and all mice had ad libitum access to water throughout the experiment.

EAE_postCR: Mice were immunized on day 0 and maintained on ad libitum feeding until day 10. From day 10 until the endpoint, mice received a moderate caloric restriction (CR) regimen, in which the amount of chow provided was adjusted to 70% of the average daily food intake of ad libitum-fed EAE_AL mice recorded over the corresponding period. The baseline consumption data were derived from previously established EAE_AL cohorts with continuous food intake monitoring from day 0 to the endpoint, allowing dynamic adjustment of food quantity. All mice had ad libitum access to water throughout the experiment. Mice were group-housed (five per cage) under controlled temperature and light/dark conditions. Daily food and water intake were recorded per cage, and the mean consumption per mouse was calculated by dividing the total cage intake by the number of mice.

At the experiment endpoint, mice were deeply anesthetized (100 µL 1% pentobarbital IP) and transcardially perfused with cold PBS to clear blood, followed by 4% paraformaldehyde (PFA) in PBS. Spinal cords were dissected. For histology/immunostaining, tissues were post-fixed in 4% PFA overnight at 4 °C, then either cryoprotected in sucrose and embedded in OCT for frozen sectioning or processed for paraffin embedding. For molecular analyses (flow cytometry or Western blot), fresh tissues (spinal cord and spleen) were collected on ice without fixation. Spleens were mashed through a 70 μm strainer to obtain single-cell suspensions. Spinal cords were collected from perfused mice and processed immediately for mononuclear cell isolation. Tissues were minced into approximately 1 mm fragments and centrifuged at 400 × g for 3 min to remove debris. The supernatant was discarded, and enzymatic dissociation was performed using the Neural Tissue Dissociation Kit (P) (Miltenyi Biotec, Cat# 130-107-677) according to the manufacturer's protocol. Briefly, for each spinal cord sample, 50 µL of Enzyme P, 10 µL of Enzyme A, 1900 µL of Buffer Z, and 20 µL of Buffer A were added and incubated at 37 °C, 90 rpm for 30 min with gentle trituration every 10 min. After digestion, 35 mL of PBS was added to stop the reaction, and the suspension was filtered through a 70 μm strainer and centrifuged at 400 × g for 5 min. The resulting pellet was resuspended in 5 mL of 70% Percoll (Cytiva, Cat# 17089109), overlaid with 5 mL of 30% Percoll, and centrifuged at 800 × g for 20 min (acceleration = 9, deceleration = 1) without brake. Mononuclear cells were collected from the interphase between the 30% and 70% layers, washed with PBS, and resuspended for downstream analyses. Red blood cells in splenocyte preparations were lysed with RBC lysis buffer (ACK). Cell counts were determined by hemocytometer, and viability by trypan blue. Typically, 1–2 × 10⁵ cells per sample were used for flow cytometry.

Spinal cord segments were fixed, then cut into 7 μm sections. For inflammation assessment, sections were stained with Hematoxylin and eosin (H&E) using a standard kit (Servicebio, Cat# G1005). For myelin visualization, adjacent sections were stained with Luxol Fast Blue (LFB) (Solarbio, Cat# G1460) followed by cresyl violet or eosin counterstain. Demyelination was evaluated by the degree of LFB loss and vacuolation. Inflammatory infiltrates and lesion scores were graded by a blinded observer on a 0–3 scale according to published criteria. For H&E staining, inflammation was scored as follows: 0 = no inflammatory cells; 1 = few scattered cells; 2 = perivascular infiltrates; 3 = dense perivascular cuffing extending into the parenchyma. For LFB staining, demyelination was graded as 0 = no demyelination; 1 = focal demyelination; 2 = multiple demyelinated areas; 3 = extensive or confluent demyelination [33, 34].

For cellular localization of inflammation, cryosections (7 μm) of PFA-fixed spinal cord were prepared. Sections were air-dried for 30 min, rinsed in PBS (4 × 5 min washes), and permeabilized in ice-cold acetone for 10 min. After drying, sections were encircled with a hydrophobic barrier (PAP pen), and incubated in blocking buffer (3% BSA, 0.2% Triton X-100 in PBS) for 30 min at room temperature. Primary antibodies (anti-IBA1, anti-GFAP, anti-MBP) diluted in blocking buffer were applied to sections and incubated overnight at 4 °C. After four washes (PBS, 5 min each), fluorescently labeled secondary antibodies (Alexa Fluor 488 or 594 conjugates) were applied (1 h at room temperature, dark). Sections were then washed (4 × 5 min in PBS), stained with DAPI (300 nM in PBS, 5 min), rinsed, and mounted with antifade medium. Stained sections were imaged by fluorescence microscopy. Cell counts and area quantification (percent IBA1⁺ or GFAP⁺ area) were performed using ImageJ (NIH). For each mouse, one representative spinal cord section at a comparable lumbar level was analyzed. In each section, five corresponding, non-overlapping regions of interest (ROIs) encompassing both gray and white matter were selected at identical anatomical locations across animals. The mean value of the five ROIs was used as the representative value for each animal (n = animals). All images were acquired and analyzed under identical parameters across groups. For MBP immunostaining, myelin integrity was semiquantitatively evaluated using a 0–3 pathological scoring system based on established EAE histopathological criteria [35], where 0 = normal myelination, 1 = focal myelin loss, 2 = multiple demyelinated areas, and 3 = extensive or confluent demyelination. Detailed information on primary antibodies for immunofluorescence is provided in Supplementary Table S1.

Mononuclear cells isolated from spinal cord and spleen were analyzed by flow cytometry to quantify T cell subsets. For surface staining, cells were resuspended in FACS buffer (PBS + 2% FBS) and incubated with fluorescently conjugated anti-CD4 (PerCP) antibody or isotype controls for 30 min at room temperature in the dark. For intracellular cytokine staining, cells were first stimulated with PMA (50 ng/mL) and ionomycin (500 ng/mL) in the presence of brefeldin A (GolgiStop) for 4 h at 37 °C. After stimulation, cells were washed (FACS buffer, 3×), surface-stained for CD4 as above, then fixed and permeabilized using a commercial kit (e.g. BD Cytofix/Cytoperm) following manufacturer instructions. Cells were then stained with anti-IL-17 A (APC) and anti-IFN-γ (PE) antibodies for 30 min (dark), washed (Perm/Wash buffer 2×), and resuspended in PBS for acquisition. In parallel, a fraction of cells (unstimulated) was fixed/permeabilized and stained for intracellular FoxP3 (PE) to identify regulatory T cells, following the same wash/incubation steps. Samples were acquired on a BD FACSAria flow cytometer. Compensation was set using single-stained controls. Data were analyzed with FlowJo v10 (BD), gating on lymphocytes (FSC/SSC), single cells, then CD4⁺ T cells, with further gating for IL-17 A⁺ (Th17), IFN-γ⁺(Th1), and FoxP3⁺ (Tregs) subsets. The percentages of cytokine-positive cells among CD4⁺ T cells were reported. This protocol follows established methods for CNS and splenocyte flow analysis in EAE. The antibodies used for flow cytometry are listed in Supplementary Table S2. Although we did not use a dedicated live/dead viability dye, viable cells were selected based on FSC/SSC characteristics and singlet discrimination. This approach is documented in the literature for CNS-derived single-cell suspensions [36, 37].

Spinal cord tissues were lysed in RIPA buffer containing protease and phosphatase inhibitors (Roche) on ice. Samples were sonicated (1 min) and centrifuged (12,000 ×g, 10 min, 4 °C) to pellet debris. Protein concentration in the supernatant was determined by BCA assay. Equal protein amounts (20 µg per lane) were mixed with 5× SDS loading buffer, boiled at 100 °C for 10 min, and resolved on 10% SDS-PAGE gels (5% stacking). Proteins were transferred to PVDF membranes (wet transfer, 300 mA, 1 h). Membranes were blocked in 5% non-fat milk (in PBST) for 1 h at room temperature, then incubated overnight at 4 °C with primary antibodies (e.g. anti-TRIB3, anti–phospho-PERK, anti-p62, β-actin; see Key Resources). After washing (PBST 3 × 5 min), HRP-conjugated secondary antibodies were applied for 1 h at room temperature. Blots were washed again and developed using ECL substrate. Chemiluminescence images were captured on a digital imager. Band intensities were quantified with ImageJ and normalized to the β-actin loading control. This procedure follows standard immunoblotting protocols. Detailed information on antibodies used for western blotting is provided in Supplementary Table S3.

Tail snips (~ 2 mm) from weanling mice were used for genomic DNA extraction (commercial kit; Biolinked, WE0186). PCR genotyping was performed using two primer pairs (F1/R1 and F2/R1 as listed in Supplementary Table S4) to distinguish the wild-type and mutant alleles. PCR products were separated on 1% agarose gels (150 V, 30 min). Wild-type alleles produce a 1602 bp band (PCR-I), whereas knockout alleles lack this band but yield a 546 bp product with the second primer set (PCR-II). All genotypes were confirmed by the expected band patterns.

Quantitative data are presented as mean ± standard error of the mean (SEM). Sample sizes (n) and numbers of independent experiments are indicated in figure legends. Statistical analyses were performed using GraphPad Prism 9.0. For comparisons among multiple groups, one-way ANOVA followed by Tukey's post hoc test was used to determine significance. For clinical score analysis, repeated-measures two-way ANOVA was used, with time as the within-subject (repeated) factor and treatment group as the between-subject factor. This approach has been widely applied for longitudinal EAE clinical score analysis [38]. A p value < 0.05 was considered statistically significant. Post hoc comparisons were performed as indicated in the figures (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). All statistical tests were two-tailed. Exact p values and details of tests are reported in figure legends. All key experiments were reproduced at least three times with consistent results.

Histological analysis of spinal cords at day 25 revealed marked differences consistent with the clinical scores. Hematoxylin/eosin (H&E) staining showed extensive inflammatory infiltrates in EAE_AL and EAE_preIF mice, whereas EAE_postIF and EAE_IF cords had markedly less parenchymal inflammation (Fig. 1 C and D). Likewise, Luxol Fast Blue (LFB) staining and MBP immunofluorescence demonstrated severe demyelination in EAE_AL and EAE_preIF mice, whereas EAE_postIF and EAE_IF groups preserved much more myelin (Fig.1C, F, S1C and D). Quantification of inflammation and demyelination (based on standard scoring) confirmed significantly lower pathology in EAE_postIF and EAE_IF versus EAE_AL (Fig. 1D and G). In parallel, markers of glial activation were reduced by EAE_postIF: immunofluorescent IBA1 (microglia) and GFAP (astrocytes) signals were significantly lower in the EAE_postIF and EAE_IF mice than in EAE_AL (Fig. 1H and J). Notably, GFAP was significantly lower in EAE_postIF than in EAE_preIF, suggesting that late IF was more effective at curbing astrocyte activation. Although the EAE_preIF regimen provided a transient improvement during early disease stages, its protective effects were not sustained once normal feeding resumed, indicating that intervention timing critically influences efficacy. Together, these data indicate that IF initiated at early clinical onset limits neuroinflammation and demyelination, consistent with reports that fasting regimens can promote remyelination and limit autoimmunity in EAE models.

In spinal cords, PBS_postIF mice exhibited lower PERK pathway activation than EAE_postIF mice: p-eIF2α, CHOP, and TRIB3 were all significantly reduced in PBS_postIF compared with EAE_postIF. This indicates that ER stress signaling was not activated under fasting conditions in non-EAE animals relative to diseased mice. Correspondingly, p62 expression was lowest in EAE_postIF, suggesting that autophagy was specifically engaged in the disease context. Together, these findings suggest that activation of the PERK–autophagy axis by intermittent fasting depends on the presence of autoimmune stress.

Likewise, histological analyses showed no reduction in spinal cord inflammation or demyelination: H&E, LFB, and MBP staining patterns were comparable across EAE_AL, EAE_postTR, and EAE_postCR (Fig. 7B–F). Microglial (IBA1) and astrocyte (GFAP) activation markers also showed no significant differences (Fig. 7G–I, S8C and D). In the spleen, EAE_postTR modestly increased total CD4⁺ counts (as expected under feeding restriction) and slightly reduced the percentage of CD4⁺IFN-γ⁺ Th1 cells, but had no consistent effect on Th17 or Treg fractions (Fig. 7J–O). EAE_postCR produced even smaller immunological changes.

Thus, unlike prolonged alternate-day fasting regimens, these milder interventions failed to confer protection in the EAE model. This indicates that the magnitude and temporal parameters of energy restriction constitute critical determinants for engaging neuroprotective mechanisms.