Butyric acid and valeric acid attenuate stress-induced ferroptosis and depressive-like behaviors by suppressing hippocampal neuroinflammation

Male SPF-grade C57BL/6 J mice (6–7 weeks old, 20–23 g) were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. The mice were housed under controlled conditions: a temperature of 20–22 °C, humidity of 45 ± 5%, and a 12-h light/dark cycle. Sterilized food and water were provided ad libitum. Following a 7-day acclimatization period in the new environment, the experiments were initiated. All procedures strictly adhered to the ethical guidelines of the Laboratory Animal Ethics Committee of Hebei Medical University. The experimental protocol was approved by the Animal Management and Ethics Committee (IACUC-Hebmu-2023011).

Fecal samples were collected in sterile EP tubes from each group of mice on the final day of restraint stress modelling, before and after antibiotic treatment, and at the conclusion of FMT. All the samples were immediately frozen in liquid nitrogen and stored at − 80 °C until further analysis.

The V3-V4 hypervariable region of the 16S rDNA gene was selected for amplification via the forward primer 5′-ACTCCTACGGGAGGCAGCA-3′ and reverse primer 5′-GGACTACHVGGGTWTCTAAT-3′. The PCR products were purified, quantified, pooled in equimolar concentrations, and sequenced via the paired-end 2 × 250 bp configuration on the Illumina NovaSeq platform (Shanghai Personal Biotechnology Co., Ltd., Shanghai, China). Microbiome bioinformatics analyses were conducted via QIIME 2 (version 2019.4) following official guidelines (https://docs.qiime2.org/2019.4/tutorials/↗) [24]. Amplicon sequence variants (ASVs) were generated via the DADA2 plugin, and the taxonomic classification of the ASVs was performed via the Greengenes database.

Gas chromatography-mass spectrometry (GC–MS) was utilized to measure the concentrations of seven common SCFAs: acetic acid, propionic acid, isobutyric acid, butyric acid, isovaleric acid, valeric acid, and caproic acid. Standard solutions were prepared as follows: acetic acid (64-19-7, Sigma, MO, USA), propionic acid (79-09-4, Sigma, MO, USA), butyric acid (107-92-6, Sigma, MO, USA), isobutyric acid (79-31-2, Sigma, MO, USA), valeric acid (109-52-4, Sigma, MO, USA), and isovaleric acid (503-74-2, Sigma, MO, USA) were mixed with water to create stock solutions at a concentration of 100 mg/mL. These stock solutions were then diluted with water to prepare a series of working solutions. Caproic acid (142-62-1, Sigma, MO, USA) was prepared in diethyl ether (60-29-7, Greagent, Shanghai, China) to produce a 100 mg/mL stock solution, which was further diluted with diethyl ether to generate a series of working solutions. Stock solutions were stored at − 20 °C, and all working solutions were freshly prepared prior to use.

For sample analysis, approximately 50 μL of 15% phosphoric acid, 10 μL of the internal standard solution (isohexanoic acid, 646–07-1, Sigma, MO, USA), and 140 μL of diethyl ether were added to a measured amount of sample in a 2 mL centrifuge tube. The mixture was homogenized for 1 min and centrifuged at 12,000 rpm for 10 min at 4 °C. The supernatant was collected and analysed via GC–MS at Shanghai Personal Biotechnology Co., Ltd. (Shanghai, China).

The mice were anaesthetized via intraperitoneal injection of 2% sodium pentobarbital (0.1 mL/20 g) and immediately euthanized. Brain and colon tissues were collected and fixed in 10% neutral formalin. The tissues were then dehydrated in graded ethanol, cleared in xylene, and embedded in paraffin. The tissue sections were baked in a 60 °C oven and stored at 4 °C until further use. For hematoxylin and eosin (HE) staining, the tissue sections were deparaffinized and rehydrated, followed by staining with hematoxylin solution, differentiation with hydrochloric acid alcohol, eosin staining, rinsing with tap water, dehydration, clearing, and mounting. The sections were then rinsed with running water, dehydrated, cleared, and mounted. Alcian blue staining was performed via an Alcian blue staining kit (G1560, Solarbio, Beijing, China) following the manufacturer's instructions. The stained sections were visualized and imaged using an Olympus DP80 microscope (Olympus, Tokyo, Japan). The microscopy images were analysed and quantified via ImageJ software (version 1.50b; National Institutes of Health, Maryland, USA).

Hippocampal and colon tissues were homogenized on ice in RIPA lysis buffer supplemented with PMSF (R0010, Solarbio, Beijing, China). The homogenates were subsequently centrifuged at 12,000 rpm for 10 min at 4 °C, after which the supernatants were collected. Equal amounts of protein were separated by SDS-PAGE and transferred onto PVDF membranes. The membranes were blocked with Y-Tec ready-to-use blocking buffer (YWB0501S, Yoche, Shanghai, China) for 5 min and incubated with the following primary antibodies: transferrin (TF, 1:1000, A1448, ABclonal, Wuhan, China), transferrin receptor (TFR, 1:1000, 13–6800, Invitrogen, CA, USA), Claudin-5 (1:2000, AF5216, Affinity Biosciences, Jiangsu, China), Occludin (1:2000, ET1701-76, Huabio, Hangzhou, China), G protein-coupled receptor 41 (GPR41, 1:1000, Huabio, Hangzhou, China), Ras homologue family member A (RhoA, 1:1000, 2117S, CST, MA, US), Rho-associated coiled-coil containing protein kinase 1 (Rock1, 1:1000, ET1609-53, Huabio, Hangzhou, China) and β-actin (1:2000, 66,009–1-Ig, Proteintech, Wuhan, China). The secondary antibodies used included goat anti-rabbit IgG (1:2000, A23920, Abbkine, Beijing, China) and goat anti-mouse IgG (1:2000, A23710, Abbkine, Beijing, China). The membranes were visualized via an Odyssey imaging system (Odyssey V3.0, LI-COR Biosciences, Nebraska, USA), and the data were analysed and quantified via ImageJ software.

Total RNA was extracted from hippocampal tissue via TRIzol reagent (15596018CN; Thermo Fisher Scientific, MA, USA) according to the manufacturer's protocol. Complementary DNA (cDNA) was synthesized via a cDNA synthesis kit (A2791; Promega, Madison, WI, USA). RT-qPCR was performed via a qPCR kit (A6001, Promega, Madison, WI, USA). Each sample was analysed in triplicate, with GAPDH used as the internal reference gene for normalization. The sequences of primers used for RT-qPCR are provided in Table 1.

ELISA kits were used to measure the levels of corticosterone (E-OSEL-M0001, Elabscience, Wuhan, China), IL-1β (SEA563Mu, CLOUD-CLONE, Wuhan, China), IL-6 (SEA079Mu, CLOUD-CLONE, Wuhan, China), IL-4 (SEA077Mu, CLOUD-CLONE, Wuhan, China), and IL-10 (SEA056Mu, CLOUD-CLONE, Wuhan, China) in mouse serum, as well as butyric acid (CEO777Ge, CLOUD-CLONE, Wuhan, China) in hippocampal tissue. All the assays were conducted in accordance with the manufacturer's instructions.

The levels of GSH and oxidized glutathione (GSSG) were measured via a GSH and GSSG detection kit (S0053, Beyotime, Shanghai, China). The MDA content was quantified via an MDA detection kit (S0131S; Beyotime, Shanghai, China), and the Fe²⁺ level was assessed via a tissue iron content detection kit (AKIC001M; Boxbio, Beijing, China). All procedures followed the manufacturer's instructions.

Statistical analyses were performed via GraphPad Prism 8 software (version 8.0.2; GraphPad Software, San Diego, USA). The data are presented as the means ± standard deviations (SDs). For normally distributed data, Student's t test was used for comparisons between two groups, whereas one-way analysis of variance (ANOVA) was employed for multiple group comparisons. Nonnormally distributed data were assessed via the Mann–Whitney U test for two-group comparisons and the Kruskal–Wallis test for multiple-group comparisons. Two-way repeated-measures ANOVA was used to assess the statistical significance of the differences in body weight across the groups. P < 0.05 was considered statistically significant. For the omics data in the article, particularly the metabolomics data identifying core metabolites, we used PASS 15 software to calculate the statistical power for the expression differences of butyric acid and valeric acid between the control group and the acute/chronic stress groups. The effect size was determined based on the mean and standard deviation of each group. With the parameters set at α = 0.05 and a sample size of 6, the calculated statistical power was 1-β > 0.90, indicating that the selected sample size was statistically meaningful. Furthermore, for SCFA metabolomics analysis, P values were calculated via either Student's t test or the Mann–Whitney U test. Metabolites with P < 0.05 and VIP > 1 were considered statistically significant. Spearman correlation analysis was performed to explore the relationships between the gut microbiota and SCFAs. Correlations with |rho|> 0.5 and P < 0.05 were considered significant and are presented accordingly.

To ascertain whether stress induces ferroptosis in hippocampal neurons and fosters depression-like behaviors, we used acute (3-day) and chronic (7-day) restraint stress models in mice [14 –17] and administered the ferroptosis inhibitors Fer-1 and DFO (Fig. 1A). Successful stress induction was confirmed by elevated serum corticosterone (Fig. 1B) and reduced body weight (Fig. 1C) in both stress paradigms. Stress exposure significantly upregulated the expression of hippocampal ferroptosis markers, including TF and TFR (Fig. 1D-F), Fe²⁺ accumulation (Fig. 1G), and MDA levels (Fig. 1H), while depleting GSH reserves (Fig. 1I). Histopathological analysis revealed increased red-stained neurons with disrupted nuclear structures in the stressed hippocampus (Fig. S1A-B), paralleled by depressive-like behaviors: reduced central locomotion and central dwelling time in the open field test (Fig. 1J–L), along with prolonged immobility in the tail suspension test (Fig. 1M). Crucially, the administration of Fer-1 and DFO significantly reversed these ferroptotic alterations (Fig. 1D–I), attenuated neuronal damage (Fig. S1A-B), and normalized behavioral deficits (Fig. 1J–M). Collectively, these findings demonstrate that restraint stress triggers hippocampal neuronal ferroptosis, which mechanistically drives the emergence of depressive-like phenotypes in mice.

To determine whether alterations in the gut microbiota contribute to stress-induced ferroptosis in hippocampal neurons, we conducted FMT experiments (Fig. 2A). Antibiotic cocktail pretreatment effectively depleted gut microbial diversity (Fig. 2B–C) and restructured bacterial communities (Fig. 2D–E), establishing an antibiotic-treated mouse model. Subsequent FMT from 3-day and 7-day stress-exposed donors to antibiotic-treated recipients recapitulated hippocampal ferroptotic signatures in recipient mice (termed the FRS-3 d and FRS-7 d groups), as evidenced by upregulated TF and TFR expression (Fig. 2F–H), elevated Fe²⁺ and MDA levels (Fig. 2I–J) alongside depleted GSH (Fig. 2K), accompanied by increased red-stained neurons (Fig. S2A–B) and depressive-like behavioral manifestations (Fig. 2L–O). These findings collectively demonstrate that gut microbial remodelling drives stress-induced ferroptosis and associated neurobehavioral pathologies.

To identify the gut microbial taxa critically involved in stress-induced hippocampal ferroptosis, we performed 16S rDNA sequencing on fecal samples from time-course stress models and FMT recipients. Both 3-day and 7-day stress exposures significantly reduced the gut microbial α diversity (Fig. 3A–B) and restructured the community composition (Fig. 3C). At the phylum level, stressed mice exhibited dominance of p_Firmicutes and p_Bacteroidetes, with elevated Firmicutes/Bacteroidetes (F/B) ratios (Fig. 3D–E). FMT recipients displayed distinct microbial profiles (Fig. 3J), showing a restored predominance of p_Firmicutes and p_Bacteroidetes (Fig. 3K). However, the α diversity indices (Fig. 3H–I) and F/B ratios (Fig. 3L) in the FMT groups showed nonsignificant intergroup variations.

Following microbial community profiling, random forest analysis was employed to prioritize the ferroptosis-related gut microbiota (Fig. 3F-G, M–N). Our focus was on genera that were consistently altered in both stress-exposed and FMT recipient mice. This analysis revealed shared microbial perturbations through comparative analyses between stress-exposed groups and their FMT recipients: g_Butyricimonas, g_Prevotella, g_Akkermansia, g_Candidatus Arthromitus, and g_[Prevotella] in 3 d models (Fig. 4A–E) and g_Butyricimonas, g_Allobaculum, g_Lactobacillus, g_Mucispirillum, g_Bilophila, g_Streptococcus, and g_Anaerotruncus in 7 d models (Fig. 4F–L). Notably, all genera except g_Candidatus Arthromitus and g_Bilophila are established SCFA producers [25 –30], defined as organic fatty acids with fewer than six carbon atoms derived from microbial fermentation of undigested dietary fibres. Quantitative analysis confirmed significant alterations in g_Butyricimonas (3 d: Fig. 4A; 7 d: Fig. 4F), g_Akkermansia (Fig. 4C), and g_Allobaculum (Fig. 4G) across both the stress and FMT groups. Moreover, g_Prevotella (Fig. 4B), g_Candidatus Arthromitus (Fig. 4D), g_Mucispirillum (Fig. 4I), g_Streptococcus (Fig. 4K), and g_Anaerotruncus (Fig. 4L) presented marked differences in stress models, with less pronounced variations between FMT subgroups. These findings highlight stress-induced dysregulation of SCFA-producing taxa as a potential driver of hippocampal ferroptosis.

The stress-induced disruption of SCFA-producing gut bacteria prompted us to delve into the specific SCFAs implicated in hippocampal neuron ferroptosis in mice under stress at various time intervals. By employing SCFA metabolomics, we quantified the serum concentrations of seven prevalent SCFAs and revealed significant disparities between the stressed and control mice at both the 3-day and 7-day time points (Fig. 5A, C). In the 3-day stress group, a marked reduction in the serum concentration of butyric acid was observed (Fig. 5B, E). In the 7-day stress group, the levels of isobutyric acid, valeric acid, and caproic acid significantly decreased (Fig. 5D, F). To further elucidate the interplay between the gut microbiota and SCFAs, we conducted a correlation analysis between the top 20 gut microbial taxa identified through random forest analysis (Fig. 3F–G) and the serum concentrations of the seven SCFAs (Fig. 5E–F). In the 3-day stress group, butyric acid emerged as the metabolite most robustly linked to the gut microbiota, demonstrating significant positive correlations with the abundances of g_Butyricimonas, g_Akkermansia, and g_Candidatus Arthromitus (Fig. 5G). In the 7-day stress group, valeric acid was the pivotal SCFA, with a positive correlation with the abundances of g_Butyricimonas, g_Allobaculum, and g_Streptococcus but a negative correlation with g_Mucispirillum (Fig. 5H). These findings underscore butyric acid and valeric acid as pivotal SCFAs that may be instrumental in driving stress-induced ferroptosis of hippocampal neurons in mice.

To delineate the temporal-specific roles of butyric acid and valeric acid in stress-induced hippocampal ferroptosis, we conducted targeted SCFA supplementation experiments (Fig. 6A). Quantitative analysis revealed significantly reduced hippocampal butyric acid levels in the 3-day stress models (RS3 + NS vs. C3 + NS), which were effectively normalized by butyric acid supplementation (Fig. 6B). Similarly, valeric acid administration restored hippocampal valeric acid depletion in 7-day stress models (Table 2). Stress exposure markedly upregulated ferroptotic markers across both stress durations, including elevated TF and TFR expression (Fig. 6C–F), Fe²⁺ accumulation (Fig. 6G), and MDA levels (Fig. 6H), coupled with GSH depletion (Fig. 6I). Critically, butyric acid or valeric acid intervention reversed these pathological signatures (Fig. 6C–I) while attenuating stress-induced hippocampal neurodegeneration (Fig. S3A–B) and depressive-like behaviors (Fig. 6J–N). These findings demonstrate that gut microbiota-derived butyric acid and valeric acid differentially regulate hippocampal ferroptosis in a time-dependent manner, suggesting their therapeutic potential against stress-related neural damage.

Preclinical studies have demonstrated that neuroinflammation triggers oxidative stress to fuel lipid peroxidation while modulating iron metabolism regulators such as TFR, thereby increasing intracellular labile iron pools [31, 32]. To investigate whether butyric acid and valeric acid alleviate stress-induced hippocampal ferroptosis through anti-inflammatory mechanisms, we first evaluated intestinal barrier integrity. Alcian blue staining revealed significant reductions in the number of colonic goblet cells across the stress models (Fig. 7A–C), accompanied by decreased expression of the tight junction proteins Occludin and Claudin-5 (Fig. 7D–G), confirming stress-induced disruption of the gut barrier. Such barrier disruption enables the translocation of harmful microbial metabolites into systemic circulation, triggering proinflammatory cascades [33]. Consistent with this mechanism, stressed mice presented significantly elevated serum levels of the proinflammatory cytokines IL-1β and IL-6 (Fig. 7H–I) and reduced levels of the anti-inflammatory mediators IL-4 and IL-10 (Fig. 7J–K). Concurrently, increased blood–brain barrier permeability (Fig. 8A–D) and hippocampal cytokine dysregulation, characterized by upregulated IL-1β and IL-6 mRNA and downregulated IL-4 and IL-10 mRNA (Fig. 8E–H), were observed. Crucially, butyric acid and valeric acid supplementation restored intestinal and blood–brain barrier integrity (Fig. 7A–G, 8A–D), normalized both systemic and hippocampal cytokine profiles (Fig. 7H–K, 8E–H), and concurrently ameliorated ferroptotic molecular signatures (Fig. 6C–I) and depressive-like behaviors (Fig. 6J–N). These findings collectively demonstrate that butyric acid and valeric acid confer neuroprotection by repairing the intestinal barrier and blood–brain barrier, thereby alleviating neuroinflammation, which in turn mitigates hippocampal neuronal ferroptosis.

In addition to peripheral inflammatory infiltration through the compromised blood–brain barrier [34], SCFAs such as butyric acid and valeric acid traverse the blood–brain barrier via monocarboxylate transporters to directly modulate neuroinflammatory responses by engaging neuronal receptors such as GPR41, which regulates downstream signalling pathways [23, 35, 36]. RhoA, a member of the Rho GTPase family, regulates cytoskeletal dynamics and cellular migration through its downstream effector Rock. Activation of the RhoA/Rock signalling pathway promotes proinflammatory cytokine production, driving neuroinflammatory cascades [37, 38]. Given that GPR41 activation suppresses this pathway [23], to investigate the molecular mechanisms by which butyric acid and valeric acid improve neuroinflammation and alleviate hippocampal neuronal ferroptosis, we measured the protein expression of GPR41, RhoA, and Rock1 in the hippocampus of mice. The results revealed that 7 days of chronic stress significantly downregulated GPR41 expression while activating the RhoA/Rock1 pathway. Valeric acid supplementation significantly improved this effect (Fig. 8K–L). Notably, acute 3-day stress caused no significant alterations in this pathway (Fig. 8I–J), indicating that the therapeutic effects of valeric acid are specific to chronic stress. These findings demonstrate that valeric acid alleviates chronic stress-induced ferroptosis potentially by enhancing GPR41-mediated inhibition of RhoA/Rock1 signalling, thereby reducing neuroinflammation. Interestingly, the anti-ferroptotic effects of butyric acid appear to operate through mechanisms distinct from this pathway.