Intermittent fasting protects against stress-induced depression and demyelination via the gut microbiota–brain axis Authors: Xin Ding, Rumi Murayama, Yi Cai, Yong Yue, Jian-Jun Yang, Kenji Hashimoto Published: 2026 ABSTRACT Chronic stress is a major risk factor for depression and disrupts myelin integrity in brain regions involved in emotional regulation. Although intermittent fasting (IF) improves metabolic and inflammatory states, its effects on stress-induced depression and demyelination remain unclear. Here, we investigated whether IF alleviates depression-like behaviors and myelin deficits in mice exposed to chronic restraint stress (CRS) and whether these effects involve modulation of the gut microbiota. Adult male C57BL/6 J mice underwent 14 days of CRS while maintained on either an ad libitum (AL) diet or an IF regimen. CRS induced robust depression-like phenotypes—characterized by increased immobility in the forced swimming test and reduced sucrose preference—without affecting locomotor activity, whereas IF significantly attenuated these behavioral abnormalities. Black-Gold II staining and myelin basic protein (MBP) immunofluorescence revealed marked demyelination in the corpus callosum, medial prefrontal cortex, and hippocampus of CRS mice, which was substantially reversed by IF. 16S rRNA sequencing demonstrated that IF reshaped gut microbial diversity and community composition under stress. Species-level analyses identified Prevotellamassilia timonensis and Muricoprocola aceti as positively associated with myelin integrity and behavioral improvement, whereas Anaeroplasma abactoclasticum showed negative associations. Functional pathway prediction further indicated that IF partially normalized stress-induced alterations in microbial metabolic functions. Collectively, these findings demonstrate that IF mitigates depression-like behaviors and preserves myelin integrity in CRS-exposed mice, potentially through gut microbiota–mediated mechanisms. IF may therefore represent a promising non-pharmacological strategy for alleviating stress-related neurobiological dysfunction. Introduction Chronic stress is a major contributor to the development of depression. Prolonged exposure to stress disrupts neural homeostasis and triggers pathological processes—including persistent neuroinflammation, oxidative injury, and impaired myelin integrity—that collectively increase vulnerability to depressive symptoms and hinder recovery [1–6]. Although antidepressant medications remain a cornerstone of treatment, their clinical utility is often limited by delayed therapeutic onset, incomplete symptom remission, and adverse effects that reduce adherence [7–12]. These limitations underscore the urgent need for novel, mechanism-based strategies that enhance resilience to chronic stress and target underlying neurobiological abnormalities, including white-matter dysfunction [6]. Intermittent fasting (IF), a dietary regimen characterized by alternating periods of energy restriction and normal feeding, has emerged as a promising non-pharmacological intervention with broad metabolic and cellular benefits. IF improves insulin sensitivity, reprograms energy metabolism, and activates cytoprotective signaling pathways that enhance mitochondrial function and neurotrophic factor expression [13–18]. Increasing evidence further indicates that IF exerts neuroprotective effects by reducing oxidative stress, suppressing neuroinflammation, and conferring protection against neurodegenerative and psychiatric disorders [19, 20]. In addition, IF reshapes gut microbial communities and modulates peripheral immune responses, thereby influencing systemic and central inflammatory states [21, 22]. Despite these advances, it remains unclear whether IF can counteract the behavioral and white-matter abnormalities induced by chronic stress. The gut–brain axis—encompassing neural, endocrine, metabolic, and immune pathways that connect the gastrointestinal system with the central nervous system—is increasingly recognized as a key regulator of emotional and stress-related processes [23–35]. Chronic stress–associated gut dysbiosis can impair intestinal barrier integrity, elevate systemic inflammation, and alter microbial metabolite production, ultimately influencing neuroinflammatory signaling, neurotransmission, and mood regulation [36–39]. These observations raise the possibility that IF may act, at least in part, through microbiota-dependent mechanisms to preserve neural integrity under stress. However, whether IF modulates the gut–brain axis to protect against chronic stress–induced demyelination and depressive-like behaviors has not been directly examined. Growing evidence indicates that gut microbiota dysregulation can influence white-matter integrity through convergent immune, metabolic, and neural pathways. Chronic stress–induced dysbiosis promotes peripheral inflammation, elevates pro-inflammatory cytokines, and disrupts blood–brain barrier function, thereby impairing oligodendrocyte maturation and myelin maintenance in stress-sensitive brain regions [6, 23, 24, 40, 41]. Microbiota-derived metabolites—including short-chain fatty acids, bile acids, and tryptophan-related molecules—also regulate microglial activation and oligodendrocyte differentiation, linking intestinal ecology to central myelin plasticity [6, 23, 24, 42]. In addition, vagus nerve signaling integrates peripheral microbial signals with central stress circuitry, suggesting that disrupted gut–brain communication may contribute to demyelination in depression-related conditions [24, 32, 33]. Collectively, these findings support the concept that restoration of gut microbial homeostasis may preserve white-matter integrity and enhance resilience to chronic stress. In the present study, we investigated whether intermittent fasting alleviates depression-like behaviors and prevents demyelination in mice exposed to chronic restraint stress (CRS) [43, 44]. To determine whether gut microbiota alterations contribute to these effects, we performed 16S rRNA sequencing of fecal samples collected throughout the experimental period. This study aims to clarify whether IF represents a viable strategy to promote stress resilience and preserve white-matter integrity through modulation of the gut–brain axis. Materials and methods Animals Adult male C57BL/6 J mice (8 weeks old, 20–25 g) were obtained from Japan SLC (Hamamatsu, Shizuoka, Japan). Animals were maintained under controlled temperature and a 12-h light/dark schedule (lights on 07:00–19:00) with unrestricted access to standard chow (CE-2; CLEA Japan, Tokyo, Japan) and water. After a 7-day acclimation period, mice were randomly assigned to one of four experimental groups: Control + ad libitum (AL) feeding, chronic restraint stress (CRS) + AL, Control + intermittent fasting (IF), and CRS + IF (n = 8 per group). Randomization was performed using a computerized method to balance baseline body weights across groups. Behavioral assessments were conducted by an experimenter blinded to treatment assignments. Ethical approval All procedures were approved by the Chiba University Animal Care and Use Committee (Approval No. 7-345), in accordance with the Japanese national guidelines for animal experimentation. Every effort was taken to minimize discomfort to the animals. Chronic restraint stress (CRS) model The CRS paradigm was conducted with minor adjustments from previously described methods [43]. Briefly, mice were individually placed in ventilated 50-mL plastic restraint tubes for 4 h per day (17:00–21:00) over a 14-day period. The tubes restricted major body movements and prevented turning but did not apply pressure that could cause physical harm. Food and water were withheld during the restraint period. After each daily session, animals were returned to their home cages. Control mice were handled only for routine husbandry and were otherwise left undisturbed during the corresponding time window. Intermittent fasting (IF) Mice were allocated to either the IF regimen or an AL feeding control group. The IF protocol was adapted from Serger et al. [45]. Animals in the IF group underwent an alternate-day feeding schedule: food was removed or provided at 21:00 each day to maintain consistent timing. Mice experienced a 24 h fasting period at the beginning of the protocol and then continued with fasting every other day (e.g., fasting during 0–24 h, 48–72 h, and 96–120 h), with unrestricted food access during the intervening 24 h refeeding periods (24–48 h, 72–96 h, 120–144 h). Water was available at all times. Mice in the AL group had continuous access to both food and water throughout the experiment. Behavioral tests Behavioral assessments were conducted according to established procedures [43, 46, 47]. The locomotion test (LMT), forced swimming test (FST), and sucrose preference test (SPT) were performed on days 15, 16, and 17, respectively (Fig. 1A).Fig. 1IF improves depression-like behaviors without affecting locomotion in CRS-exposed mice.A Schematic overview of the experimental design. Mice underwent 14 days of chronic restraint stress (CRS) with either ad libitum (AL) feeding or intermittent fasting (IF). Behavioral testing—including the locomotion test (LMT), forced swimming test (FST), and sucrose preference test (SPT)—was conducted on days 15–17, followed by brain tissue collection. B Body weight gain on day 14 relative to baseline (one-way ANOVA with Tukey’s post hoc test; F = 8.246, p < 0.01). C Total distance traveled in the LMT (one-way ANOVA with Tukey’s post hoc test; F = 1.498, p = 0.24). D Immobility time in the FST (one-way ANOVA with Tukey’s post hoc test; F = 6.509, p < 0.01). E Sucrose preference in the SPT (one-way ANOVA with Tukey’s post hoc test; F = 5.121, p < 0.01). Data are presented as mean ± SEM (n = 8 per group). *p < 0.05, **p < 0.01, ***p < 0.001. IF improves depression-like behaviors without affecting locomotion in CRS-exposed mice. A Schematic overview of the experimental design. Mice underwent 14 days of chronic restraint stress (CRS) with either ad libitum (AL) feeding or intermittent fasting (IF). Behavioral testing—including the locomotion test (LMT), forced swimming test (FST), and sucrose preference test (SPT)—was conducted on days 15–17, followed by brain tissue collection. B Body weight gain on day 14 relative to baseline (one-way ANOVA with Tukey’s post hoc test; F = 8.246, p < 0.01). C Total distance traveled in the LMT (one-way ANOVA with Tukey’s post hoc test; F = 1.498, p = 0.24). D Immobility time in the FST (one-way ANOVA with Tukey’s post hoc test; F = 6.509, p < 0.01). E Sucrose preference in the SPT (one-way ANOVA with Tukey’s post hoc test; F = 5.121, p < 0.01). Data are presented as mean ± SEM (n = 8 per group). *p < 0.05, **p < 0.01, ***p < 0.001. On day 15, spontaneous activity was evaluated using the SCANET MV-40 system (MELQUEST Co., Ltd., Toyama, Japan). Each mouse was placed in an acrylic testing chamber (560 × 560 × 330 mm), and total movement was automatically recorded for 60 min. Chambers were cleaned thoroughly between animals to avoid olfactory interference. On day 16, depression-like behavior was assessed using an automated FST apparatus (SCANET MV-40). Mice were individually placed in a transparent cylinder (23 cm diameter, 31 cm height) filled with water to a depth of 15 cm at 23 ± 1 °C. The system quantified immobility using its integrated analysis software, calculating immobility time as total time minus active movement time. Behavioral data from the 6 min test session were used for analysis. The predefined exclusion criterion was an immobility time of less than 5 s; however, no mice met this criterion, and therefore no animals were excluded from the analysis. To habituate mice to the sucrose solution, both water and 1% sucrose were available for 48 h prior to testing (days 15–16). On day 17, following a 4 h period without food or water, mice were given simultaneous access to two identical bottles—one containing water and the other 1% sucrose (50 mL). The bottles containing water and sucrose were weighed at the end of this period. The sucrose preference was calculated as a percentage of sucrose solution consumption to the total liquid consumption. Collection of intestinal contents Mice were anesthetized with a combination of medetomidine (0.75 mg/kg; Nippon Zenyaku Kogyo, Fukushima, Japan), midazolam (4 mg/kg; Sandoz K.K., Tokyo, Japan), and butorphanol (5 mg/kg; Meiji Seika Pharma, Tokyo, Japan). Under sterile conditions, the intestinal tract was exposed, and luminal contents were collected from the colon as described previously [48, 49]. Samples were immediately snap-frozen in liquid nitrogen and stored at −80 °C until analysis. 16S rRNA sequencing analysis Genomic DNA was isolated from approximately 250 mg of colon contents using the QIAamp Fast DNA Stool Mini Kit (Qiagen; cat. no. 51604; Lot 181023747) following the manufacturer’s instructions [43, 49, 50]. DNA concentrations were adjusted to 10 ng/µL, and samples were submitted to BGI Japan (Kobe, Japan) for 16S rRNA gene profiling. The bacterial V3–V4 region was amplified using dual-indexed primers (341 F/806 R), and paired-end sequencing was performed. Bioinformatic processing followed a standardized pipeline. Paired-end reads were merged using fastq-join, and quality filtering included chimera removal with Usearch 6.1, retaining reads with ≥99% of bases above a Q-score of 20. The sequencing dataset is publicly available in the NCBI Sequence Read Archive under accession number PRJNA1268743. For taxonomic classification, non-chimeric sequences were assigned to bacterial lineages using the RDP Naïve Bayesian Classifier with an 80% confidence cutoff. Alpha-diversity metrics and taxonomic relative abundances were analyzed using the Kruskal–Wallis test followed by Dunn’s post hoc comparisons. Beta-diversity was assessed through principal component analysis (PCA), principal coordinates analysis (PCoA), and unweighted UniFrac distance calculations; statistical significance was evaluated using ANOSIM implemented in the vegan package in R (v4.2.2). To identify discriminatory taxa among experimental groups, Linear Discriminant Analysis Effect Size (LEfSe) was applied. Only taxa with LDA scores >3.0 and false discovery rate (FDR) < 0.05 were considered significant. Visualization of microbial community differences was generated using taxonomic bar plots, cladograms, and related tools available through Metware Cloud and OmicStudio (https://cloud.metware.cn/; https://www.omicstudio.cn/) [49, 51]. Immunofluorescence Mice were deeply anesthetized and perfused transcardially with 0.9% saline followed by 4% paraformaldehyde (PFA). Brains were removed and post-fixed in 4% PFA overnight at 4 °C. Tissue was then sectioned coronally at 30 μm thickness using a vibratome (VT1000S, Leica Microsystems, Wetzlar, Germany). Sections were rinsed in phosphate-buffered saline (PBS) and incubated for 1 h at room temperature in a blocking solution consisting of 0.3% Triton X-100 and 5% normal goat serum. Overnight incubation at 4 °C with a primary antibody against myelin basic protein (MBP; 1:500, ab218011, Abcam) followed. After PBS washes, sections were exposed for 2 h at room temperature to Alexa Fluor 594–conjugated goat anti-rabbit IgG (1:1000, A-11012, Invitrogen). Slides were coverslipped using DAPI-containing mounting medium (H1200, Vector Laboratories), and images were acquired with a Keyence BZ-900 fluorescence microscope (Tokyo, Japan). Immunofluorescence procedures followed previously described approaches [43, 50, 52, 53]. Quantification of MBP signal intensity in designated brain regions was performed using Fiji-ImageJ software. Black-Gold II (BG II) staining Black-Gold II staining was used to assess myelin integrity, following standard procedures [43, 50, 52–54]. Coronal brain sections (30 μm) were mounted on gelatin-coated slides and air-dried for approximately 30 min. After a brief rehydration in distilled water (2 min), sections were incubated in BG II Solution A (TR-100-BG, Biosensis, Australia) at 65 °C for ~20 min or until optimal myelin staining developed. Slides were then rinsed in distilled water and treated with Solution B for 3 min, followed by an additional rinse. Sections were dehydrated in xylene for 1 min and coverslipped using a non-aqueous mounting medium. Images were captured using a Keyence BZ-900 microscope (Tokyo, Japan). Quantification of myelin-positive regions was performed with Fiji-ImageJ. The percentage of demyelination area was calculated as (corpus callosum area - myelin staining area)/corpus callosum area x 100%. Statistical analyses Data are presented as mean ± standard error of the mean (SEM) or median with interquartile range (IQR), as appropriate. Statistical analyses were performed using GraphPad Prism 9.4.1 and R version 4.2.2. Normality of data distribution was evaluated with the Shapiro–Wilk test, and variance homogeneity with Bartlett’s test. For body weight measurements across the four groups, nonparametric repeated-measures analysis was conducted using the Friedman test followed by Dunn’s multiple-comparisons test. Locomotor activity, immobility time in the FST, and sucrose preference in the SPT were analyzed by two-way ANOVA with Tukey’s post hoc correction. Statistical significance was set at p < 0.05. For microbiota analyses, FDR values were computed in R. Volcano plots were generated with the ggplot2 package. Differentially altered taxa were defined as follows: FDR < 0.05 (or p < 0.05 where noted) and log₂ fold change (log₂FC) > 0 for upregulated taxa, and FDR < 0.05 (or p < 0.05) with log₂FC < 0 for downregulated taxa when comparing Control + AL vs. CRS + AL and CRS + AL vs. CRS + IF groups. Spearman correlation analyses were performed and visualized using GraphPad Prism 9.4.1. Results IF alleviated depression-like behaviors in CRS-exposed mice without affecting locomotion To assess whether IF confers behavioral benefits under chronic stress, mice were subjected to 14 days of CRS while maintained on either an IF or AL feeding regimen (Fig. 1A). Behavioral testing (LMT, FST, and SPT) was conducted from days 15–17. As expected, body weight gain was reduced in the CRS + AL, Control + IF, and CRS + IF groups compared with the Control + AL group (Fig. 1B). However, locomotor activity did not differ among the four groups (Fig. 1C), indicating that subsequent behavioral outcomes were not confounded by changes in general mobility. CRS exposure induced marked depression-like behaviors, as evidenced by increased immobility time in the FST and reduced sucrose preference in the SPT (Fig. 1D, E). Notably, mice in the CRS + IF group exhibited significantly reduced immobility time and higher sucrose preference relative to the CRS + AL group, demonstrating that IF effectively mitigated CRS-induced behavioral deficits (Fig. 1D, E). Together, these results show that IF attenuates depression-like behaviors in CRS-exposed mice while leaving motor function unchanged. IF attenuated demyelination in the corpus callosum (CC), medial prefrontal cortex (mPFC) and hippocampus in CRS-exposed mice Myelin disruption has been implicated in the development of depression-like behaviors induced by CRS [43]. To examine whether IF modulates CRS-associated myelin abnormalities, we assessed myelin integrity using Black-Gold II (BG II) staining and MBP immunofluorescence in brain regions relevant to mood regulation [55, 56]. BG II staining revealed robust demyelination in the CC of CRS + AL mice compared with the Control + AL group (Fig. 2A, C). Importantly, IF markedly reduced this myelin loss, as indicated by a significantly lower demyelination ratio in CRS + IF mice relative to CRS + AL mice (Fig. 2C). Consistent with these findings, MBP fluorescence intensity in the CC was substantially decreased in CRS + AL mice but restored by IF to levels comparable to Control + AL mice (Fig. 2B, D).Fig. 2IF attenuates CRS-induced myelin loss and correlates with depression-like behaviors.A, B Representative images of Black-Gold II (BG II) staining and MBP immunofluorescence in the corpus callosum (CC). C Quantitative analysis of demyelination area in the CC area of Black gold (%): Two-way ANOVA: Interaction, F(1,28) = 8.783, p = 0.006; IF, F(1,28) = 1.129, p = 0.297; CRS, F(1,28) = 1.947, p = 0.174. D MBP fluorescence intensity in the CC: Two-way ANOVA: Interaction, F(1,28) = 2.701, p = 0.111; IF, F(1,28) = 51.83, p < 0.001; CRS, F(1,28) = 30.62, p < 0.001). E, F MBP fluorescence intensity in the medial prefrontal cortex (mPFC) and hippocampus (mPFC-MBP: Two-way ANOVA: Interaction, F(1,28) = 1.966, p = 0.172; IF, F(1,28) = 10.58, p = 0.003; CRS, F(1,28) = 9.867, p = 0.004; Hippocampus-MBP: Two-way ANOVA: Interaction, F(1,28) = 12.68, p = 0.001; IF, F(1,28) = 2.084, p = 0.16; CRS, F(1,28) = 1.922, p = 0.177). G Representative MBP immunofluorescence images in the mPFC. Scale bars: 500 μm (overview) and 200 μm (inset). H-J: Scatter plots show Spearman’s rank correlations for (H) Sucrose preference (%) vs. demyelinating ratio of BG II in the CC, (I) Immobility time in FST (s) vs. MBP fluorescence intensity in the CC, and (J) Immobility time vs. MBP fluorescence intensity in the mPFC. Correlation coefficients (r) and p values are shown on each plot; dashed lines indicate 95% confidence intervals. Scale bars: 500 μm (overview) and 200 μm (inset). Data represent n = 8 mice per group. *p < 0.05, **p < 0.01, ***p < 0.001. IF attenuates CRS-induced myelin loss and correlates with depression-like behaviors. A, B Representative images of Black-Gold II (BG II) staining and MBP immunofluorescence in the corpus callosum (CC). C Quantitative analysis of demyelination area in the CC area of Black gold (%): Two-way ANOVA: Interaction, F(1,28) = 8.783, p = 0.006; IF, F(1,28) = 1.129, p = 0.297; CRS, F(1,28) = 1.947, p = 0.174. D MBP fluorescence intensity in the CC: Two-way ANOVA: Interaction, F(1,28) = 2.701, p = 0.111; IF, F(1,28) = 51.83, p < 0.001; CRS, F(1,28) = 30.62, p < 0.001). E, F MBP fluorescence intensity in the medial prefrontal cortex (mPFC) and hippocampus (mPFC-MBP: Two-way ANOVA: Interaction, F(1,28) = 1.966, p = 0.172; IF, F(1,28) = 10.58, p = 0.003; CRS, F(1,28) = 9.867, p = 0.004; Hippocampus-MBP: Two-way ANOVA: Interaction, F(1,28) = 12.68, p = 0.001; IF, F(1,28) = 2.084, p = 0.16; CRS, F(1,28) = 1.922, p = 0.177). G Representative MBP immunofluorescence images in the mPFC. Scale bars: 500 μm (overview) and 200 μm (inset). H-J: Scatter plots show Spearman’s rank correlations for (H) Sucrose preference (%) vs. demyelinating ratio of BG II in the CC, (I) Immobility time in FST (s) vs. MBP fluorescence intensity in the CC, and (J) Immobility time vs. MBP fluorescence intensity in the mPFC. Correlation coefficients (r) and p values are shown on each plot; dashed lines indicate 95% confidence intervals. Scale bars: 500 μm (overview) and 200 μm (inset). Data represent n = 8 mice per group. *p < 0.05, **p < 0.01, ***p < 0.001. Similar patterns were observed in the mPFC and hippocampus, where CRS exposure led to reduced MBP immunoreactivity, indicating myelin loss. IF intervention effectively reversed these deficits in both regions (Fig. 2E, F). Analysis of hippocampal subfields showed parallel reductions in MBP signal across CA1, CA3, and DG in CRS + AL mice, all of which were ameliorated by IF (Supplemental Fig. 1A–D). We next explored the relationship between demyelination and behavioral outcomes. Spearman correlation analyses showed significant associations between sucrose preference and CC demyelination ratio, as well as between FST immobility and MBP intensity in the CC and mPFC (Fig. 2H–J). No significant correlations were detected between behavioral measures and other myelin indices (Supplemental Fig. 1E–I). Taken together, these findings indicate that CRS induces widespread myelin deficits in the CC, mPFC, and hippocampus, and that IF effectively mitigates these structural abnormalities. These results suggest that preservation of myelin integrity may contribute to the antidepressant-like effects of IF. IF altered gut microbiota diversity (α-diversity, β-diversity) and community structure in CRS-exposed mice We first evaluated gut microbial α-diversity using the Shannon, Chao1, and Simpson indices across the four experimental groups. All three metrics showed significant overall group differences by Kruskal–Wallis test (p = 0.006, 0.047, and 0.01, respectively) (Fig. 3A–C). To assess β-diversity, Bray–Curtis dissimilarity was calculated, revealing significant shifts in community structure among groups (p < 0.01) (Fig. 3D).Fig. 3IF modulates gut microbiota diversity and community structure in CRS-exposed mice.A–C α-diversity indices across groups: Shannon (Kruskal–Wallis: H = 5.92, p = 0.006), Chao1 (H = 6.84, p = 0.047), and Simpson (H = 7.57, p = 0.01). D β-diversity assessed using Bray–Curtis distances (Kruskal–Wallis: H = 47.55, p < 0.01). E, F Principal component analysis (PCA) and principal coordinates analysis (PCoA), with R² and p values indicated on each panel. G Venn diagrams showing the number of differentially abundant species among the four groups. H, I Similarity analysis and partial least squares discriminant analysis (PLS-DA), with R, Q², and statistical significance displayed on each panel. J LEfSe cladogram illustrating taxa enriched in each group (LDA > 3, FDR < 0.05). K Top discriminatory taxa for each group identified by LEfSe analysis. Data represent n = 8 mice per group. *p < 0.05; **p < 0.01; ***p < 0.001. IF modulates gut microbiota diversity and community structure in CRS-exposed mice. A–C α-diversity indices across groups: Shannon (Kruskal–Wallis: H = 5.92, p = 0.006), Chao1 (H = 6.84, p = 0.047), and Simpson (H = 7.57, p = 0.01). D β-diversity assessed using Bray–Curtis distances (Kruskal–Wallis: H = 47.55, p < 0.01). E, F Principal component analysis (PCA) and principal coordinates analysis (PCoA), with R² and p values indicated on each panel. G Venn diagrams showing the number of differentially abundant species among the four groups. H, I Similarity analysis and partial least squares discriminant analysis (PLS-DA), with R, Q², and statistical significance displayed on each panel. J LEfSe cladogram illustrating taxa enriched in each group (LDA > 3, FDR < 0.05). K Top discriminatory taxa for each group identified by LEfSe analysis. Data represent n = 8 mice per group. *p < 0.05; **p < 0.01; ***p < 0.001. Ordination analyses further supported these observations. Both PCA and PCoA (based on OTU-level Bray–Curtis distances) demonstrated distinct clustering among groups, confirmed by significant ANOSIM values (PCA: R = 0.39, p = 1 × 10⁻⁴; PCoA: R = 0.48, p = 1 × 10⁻³) (Fig. 3E, F). Venn diagrams illustrated group-specific and shared taxa (Fig. 3G). Similarity analysis and PLS-DA also showed clear separation of microbial communities across the four groups (similarity: r = 0.71, p < 0.01; PLS-DA: Q² = 0.41, p < 0.01) (Fig. 3H, I). To identify specific microbial signatures associated with each condition, LEfSe analysis (LDA > 3, FDR < 0.05) was performed. Cladograms illustrate taxa enriched in each group, with concentric rings indicating hierarchical taxonomic levels from phylum to species (Fig. 3J). The top discriminatory taxa for each group are shown in Fig. 3K. In the Control + AL group, potential microbial markers included f_Prevotellaceae, g_Prevotellamassilia, g_Oscillibacter, g_Flintibacter and f_Porphyromonadaceae. Control + IF group potential microbial markers included f_Bacteroidaceae, g_Bacteroides, p_Bacteroides, o_Bacteroidales and c_Bacteroidia. CRS + AL group potential microbial markers included g_Alistipes, f_Rikenellaceae and g_Harryflintia. CRS + IF group potential microbial markers included f_Lachnospiraceae, g_Kineothrix, p_Mycoplasmatota, f_Anaeroplasmataceae and c_Mollicutes. Taken together, these findings demonstrate that IF profoundly reshapes gut microbial diversity and composition in CRS-exposed mice. IF modulated specific bacterial species associated with demyelination We next examined compositional changes in the gut microbiota across groups. At the genus level, the ten genera showing the greatest differences in relative abundance included Mucispirillum, Kineothrix, Lawsonibacter, Oscillibacter, Alistipes, Prevotellamassilia, Bacteroides, Duncaniella, Lacrimispora, and Anaeroplasma (Fig. 4A). At the species level, notable differences were observed for Lawsonibacter asaccharolyticus, Alistipes putredinis, Bacteroides acidifaciens, Muricoprocola aceti, Duncaniella dubosii, Prevotellamassilia timonensis, Mucispirillum schaedleri, Oscillibacter valericigenes, Fretibacterium fastidiosum, and Anaeroplasma abactoclasticum (Fig. 4B). These findings indicate that IF produces marked shifts in microbial composition in CRS-exposed mice.Fig. 4IF alters specific bacterial species and their associations with demyelination metrics.A, B Relative abundance of the top significantly altered bacterial genera (A) and species (B) across the four groups. C, D Volcano plots showing differentially abundant species between (C) Control + AL vs. CRS + AL and (D) CRS + AL vs. CRS + IF. The red dashed line indicates FDR = 0.05; the black dashed line indicates p = 0.05. E, F Lists of species significantly altered between (E) Control + AL vs. CRS + AL and (F) CRS + AL vs. CRS + IF (Mann–Whitney U test, p < 0.05). G–J Violin plots showing the relative abundance of four species significantly altered in both comparisons: Prevotellamassilia timonensis, Muricoprocola aceti, Fretibacterium fastidiosum, and Anaeroplasma abactoclasticum. K Heatmap illustrating the abundance of these four species across individual samples. L Correlation heatmap showing Spearman’s associations between the four species and myelin-related metrics. Significant correlations (p < 0.05) are indicated in each cell. Data represent n = 8 mice per group. *p < 0.05; **p < 0.01; ***p < 0.001. IF alters specific bacterial species and their associations with demyelination metrics. A, B Relative abundance of the top significantly altered bacterial genera (A) and species (B) across the four groups. C, D Volcano plots showing differentially abundant species between (C) Control + AL vs. CRS + AL and (D) CRS + AL vs. CRS + IF. The red dashed line indicates FDR = 0.05; the black dashed line indicates p = 0.05. E, F Lists of species significantly altered between (E) Control + AL vs. CRS + AL and (F) CRS + AL vs. CRS + IF (Mann–Whitney U test, p < 0.05). G–J Violin plots showing the relative abundance of four species significantly altered in both comparisons: Prevotellamassilia timonensis, Muricoprocola aceti, Fretibacterium fastidiosum, and Anaeroplasma abactoclasticum. K Heatmap illustrating the abundance of these four species across individual samples. L Correlation heatmap showing Spearman’s associations between the four species and myelin-related metrics. Significant correlations (p < 0.05) are indicated in each cell. Data represent n = 8 mice per group. *p < 0.05; **p < 0.01; ***p < 0.001. To identify taxa driving these differences, volcano plots were generated for the top 200 species based on total abundance. In the CRS + AL versus Control + AL comparison, species with p < 0.05 and log₂FC > 0 were considered upregulated, while those with p < 0.05 and log₂FC < 0 were downregulated (Fig. 4C). An analogous classification was applied to the CRS + IF versus CRS + AL comparison (Fig. 4D). Differentially abundant species for both comparisons are summarized in Fig. 4E, F. Four species displayed notable bidirectional regulation across conditions: Fretibacterium fastidiosum, Anaeroplasma abactoclasticum, Prevotellamassilia timonensis, and Muricoprocola aceti. Fretibacterium fastidiosum and Anaeroplasma abactoclasticum decreased further in CRS + IF mice, whereas Prevotellamassilia timonensis and Muricoprocola aceti increased relative to the CRS + AL group. Violin plots and a heatmap illustrate these species-level differences across samples (Fig. 4G–K). We then assessed correlations between these taxa and myelin-related outcomes. Significant associations were observed (Fig. 4L): Prevotellamassilia_timonensis and Muricoprocola_aceti were positively correlated with MBP fluorescence intensity in the CC or hippocampus. Anaeroplasma_abactoclasticum showed a negative correlation with MBP intensity in the hippocampus. Together, these findings suggest that IF enriches bacterial species positively associated with myelin integrity while reducing species negatively associated with demyelination. These species may contribute to the protective effects of IF on white matter in CRS-exposed mice. Differential species functional annotation enrichment and correlations with depression-like behaviors We next performed functional prediction and pathway enrichment analyses (KEGG, COG, MetaCyc) for the CRS + AL vs. Control + AL and CRS + IF vs. CRS + AL comparisons (Fig. 5A, B). Several pathways showed opposite regulation patterns across conditions. For example, “Endocrine system nucleotide metabolism” and “Photosynthesis” were reduced in the CRS + AL group but significantly increased following IF intervention. Conversely, “Energy production and conversion” was elevated in CRS + AL mice and suppressed in the CRS + IF group. These findings indicate that IF partially reverses stress-induced alterations in microbial functional pathways.Fig. 5Functional pathway enrichment of differential species and correlations with depression-like behaviors.A Differentially enriched microbial metabolic pathways between Control + AL and CRS + AL groups, summarized across KEGG Level 2 categories, COG functional categories, and MetaCyc pathways. B Differentially enriched pathways between CRS + AL and CRS + IF groups, displayed using the same annotation frameworks. Bars indicate mean pathway proportions (%), effect sizes, 95% confidence intervals, and corresponding p values. Statistical comparisons were performed using Welch’s t-tests (significance threshold: p < 0.01, no multiple-testing correction). Red and blue boxes highlight pathways significantly upregulated or downregulated within each comparison, respectively. C Heatmap showing Spearman’s correlations between significantly altered species (as identified in Fig. 4E–F) and behavioral performance metrics (FST immobility and SPT sucrose preference). Significant correlations (p < 0.05) are marked on the heatmap. D Microbial co-occurrence network illustrating significant inter-species correlations (Spearman’s |r | > 0.4, p < 0.05). Positive correlations are shown in pink and negative correlations in blue. Node size reflects degree centrality; Muricoprocola aceti appears as a putative keystone species within the network. Data represent n = 8 mice per group. *p < 0.05; **p < 0.01; ***p < 0.001. Functional pathway enrichment of differential species and correlations with depression-like behaviors. A Differentially enriched microbial metabolic pathways between Control + AL and CRS + AL groups, summarized across KEGG Level 2 categories, COG functional categories, and MetaCyc pathways. B Differentially enriched pathways between CRS + AL and CRS + IF groups, displayed using the same annotation frameworks. Bars indicate mean pathway proportions (%), effect sizes, 95% confidence intervals, and corresponding p values. Statistical comparisons were performed using Welch’s t-tests (significance threshold: p < 0.01, no multiple-testing correction). Red and blue boxes highlight pathways significantly upregulated or downregulated within each comparison, respectively. C Heatmap showing Spearman’s correlations between significantly altered species (as identified in Fig. 4E–F) and behavioral performance metrics (FST immobility and SPT sucrose preference). Significant correlations (p < 0.05) are marked on the heatmap. D Microbial co-occurrence network illustrating significant inter-species correlations (Spearman’s |r | > 0.4, p < 0.05). Positive correlations are shown in pink and negative correlations in blue. Node size reflects degree centrality; Muricoprocola aceti appears as a putative keystone species within the network. Data represent n = 8 mice per group. *p < 0.05; **p < 0.01; ***p < 0.001. We then examined correlations between all significantly altered species (p < 0.05; Fig. 4E, F) and behavioral performance. As shown in Fig. 5C, Prevotellamassilia timonensis and Muricoprocola aceti were negatively correlated with FST immobility and positively correlated with sucrose preference, suggesting that higher abundance of these taxa is associated with improved behavioral outcomes. These results support the idea that IF-associated enrichment of specific species may contribute to the antidepressant-like effects observed in CRS-exposed mice. Finally, we constructed a microbial co-occurrence network to explore interspecies relationships (Fig. 5D). Species identified in Fig. 4E, F were included, and significant correlations (Spearman’s |r | > 0.4, p < 0.05) were visualized as positive (pink) or negative (blue) edges. Notably, Muricoprocola aceti displayed high degree and betweenness centrality, identifying it as a potential keystone taxon. Its central network position suggests that it may play an important role in mediating the beneficial effects of IF on depression-like behavior in CRS-exposed mice. Discussion In this study, we demonstrated that IF exerts robust protective effects against depression-like behaviors and demyelination in mice exposed to CRS. Consistent with prior evidence that chronic stress disrupts white-matter integrity and impairs mood-related neural circuits [43], CRS induced pronounced myelin loss in the CC, mPFC, and hippocampus. Importantly, IF attenuated these structural abnormalities and improved behavioral performance in both the FST and SPT. Together, these findings suggest that metabolic interventions such as IF may represent promising non-pharmacological strategies for reducing stress-induced neurobiological vulnerability. The partial inconsistency among behavioral–myelin correlations likely reflects methodological differences between Black-Gold II staining, which primarily indexes structural myelin integrity, and MBP immunofluorescence, which reflects regional protein expression. These complementary but nonidentical measures capture distinct aspects of myelin biology and may therefore yield divergent correlation patterns. Accordingly, the behavioral improvements observed in the present study should not be attributed solely to myelin preservation, and additional mechanisms—including inflammatory, metabolic, and synaptic processes—are likely involved. A central finding of this study is that IF markedly reshaped the gut microbiota in CRS-exposed mice, influencing both microbial diversity and taxonomic composition. CRS-induced dysbiosis was characterized by reduced abundance of beneficial taxa and enrichment of bacteria associated with inflammation and metabolic disturbance [57, 58], whereas IF partially restored microbial diversity and produced a distinct compositional profile. Functional enrichment analyses further suggested that IF reversed CRS-associated alterations in microbial metabolic pathways, particularly those related to nucleotide metabolism and energy conversion. Collectively, these observations support the idea that IF may improve emotional behavior and myelin integrity, at least in part, by normalizing stress-induced microbial dysfunction. Species-level analyses identified several bacterial taxa potentially contributing to these beneficial effects. In particular, Prevotellamassilia timonensis and Muricoprocola aceti were positively associated with myelin integrity and improved behavioral outcomes, whereas Anaeroplasma abactoclasticum showed the opposite pattern. Because P. timonensis has been isolated from the human gut [59], it may represent a clinically relevant taxon influencing both peripheral and central physiology. Network analysis further indicated that Muricoprocola aceti may function as a putative keystone species, exhibiting high degree and betweenness centrality—features commonly used to identify taxa with disproportionate ecological influence on community structure and host physiology [60]. Although these associations do not establish causality, they highlight tractable candidates for future mechanistic investigation of microbiota–myelin interactions. Our findings also support the emerging concept that myelin plasticity is regulated by gut-derived signals. Chronic stress impairs oligodendrocyte maturation and reduces myelin protein expression, thereby disrupting communication within mood-related neural circuits [2, 6]. IF-induced microbial remodeling may create a biochemical environment favorable for oligodendrocyte function through immune modulation, metabolite production, and neurotrophic signaling. Consistent with this interpretation, IF has been reported to increase myelin-related protein expression and improve motor performance in aged mice [61]. Although the precise molecular pathways remain to be defined, our data support a working model in which IF enhances brain resilience to chronic stress through microbiota-dependent preservation of white-matter integrity. Beyond microbiota-mediated effects, systemic inflammation and metabolic dysfunction are increasingly recognized as core contributors to depression pathophysiology, influencing neuroimmune signaling, synaptic plasticity, and white-matter stability [1–6]. IF improves metabolic homeostasis, reduces peripheral inflammatory cytokines, and enhances mitochondrial and neurotrophic signaling [17–22], processes that may collectively support oligodendrocyte function and stabilize mood-related neural circuits. Thus, the behavioral and myelin-protective effects observed here may also reflect systemic metabolic reprogramming and anti-inflammatory actions induced by IF. Despite these strengths, several limitations should be acknowledged. First, only male mice were studied; given known sex differences in stress responsivity, metabolism, immune function, and microbiota composition, inclusion of females will be essential for generalizability [62, 63]. Second, although specific microbial taxa correlated with behavioral and myelin outcomes, these associations do not establish causality. Future studies using targeted microbial manipulation, germ-free models, or fecal microbiota transplantation will be required to determine whether these taxa directly mediate IF’s protective effects. Third, metabolomic profiling was not performed. Because microbial metabolites are key mediators of gut–brain signaling, and functional pathway analyses were computationally inferred, these findings should be considered hypothesis-generating. Integrated microbiome–metabolome–transcriptome analyses will be necessary to define the biochemical mechanisms of IF-induced neuroprotection. Fourth, 16S rRNA V3–V4 sequencing provides limited species-level resolution; therefore, reported species-level alterations should be regarded as putative, and future shotgun metagenomic multi-omics approaches are required for validation. Fifth, IF was initiated concurrently with CRS, primarily assessing preventive rather than reversal effects. Post-stress intervention studies are needed to determine therapeutic efficacy in established pathology. Sixth, because the FST preceded the SPT by 24 h, potential acute stress carry-over cannot be fully excluded, although identical testing schedules likely minimized systematic bias. Counterbalanced behavioral designs will strengthen future studies. Seventh, the vagus nerve—a key component of gut–brain signaling in depression [24, 32, 64–66]—was not experimentally tested; vagotomy or neuromodulation studies are needed to determine its role in IF-mediated effects. Finally, this study focused primarily on myelin-related outcomes, whereas IF may also influence neuroinflammation, synaptic plasticity, microglial activation, and neurotrophic signaling. Future work examining these pathways will provide a more comprehensive understanding of IF-induced stress resilience. In conclusion, IF mitigates CRS-induced depression-like behaviors and protects against myelin loss in key mood-regulating brain regions while reshaping the gut microbiota toward taxa associated with white-matter integrity and behavioral improvement. Although causal mechanisms remain to be established, these findings support a model in which dietary metabolic interventions enhance resilience to chronic stress via the gut–brain axis. By identifying candidate microbial mediators of myelin preservation, this study provides a foundation for future mechanistic and translational investigations. Ultimately, integrated multi-omics, causal microbiota manipulation, vagal-circuit interrogation, and inclusion of both sexes will be essential to determine whether IF can be translated into a viable non-pharmacological strategy for preventing or treating stress-related neurobiological disorders. Supplementary information Supplemental material Supplemental material