Male C57BL/6 mice, 7 weeks old and free of specific pathogens, were purchased from Chengdu Dashuo Laboratory Animals (Chengdu, China), housed individually and assigned unique numbers. The mice were acclimated for a week before the experiment. The mice were housed individually under standard condition (23°C ± 2°C, 50% ~ 70% humidity, on a 12–12 h light–dark cycle with lights on at 07,00 and lights off at 19,00, with access to food and water ad libitum) and were allowed to habituate to laboratory conditions for 1 week prior to the experiments. Weight-matched animals were randomly divided to control group (Ctrl, n = 12 mice) and chronic mild stress group (CMS, n = 30 mice). Mice from CMS group were exposed to chronic mild stress as described in 2.3 section. Control mice were not subjected to any stress and housed in their cages similar to the habituation period.

All animal experiments were approved by the Institutional Animal Care and Use Committee at the Guizhou University of Traditional Chinese Medicine (20,231,215,001, Guiyang, China). All experiments were performed in accordance with ARRIVE guidelines 2.0 (2020) (Percie et al., 2020) and with the “Guidelines for the Care and Use of Laboratory Animals” from the United States National Institutes of Health.

Mice from CMS group were exposed to chronic mild stress for 4 weeks as described (Agudelo et al., 2014). Every day, mice in the stressed group were exposed to 2–3 of the following stressors in random order: tilting of the cage (45°, 24 h), clamping of the tail (10 min), shaking of the cage (30 min), continuous light (24 h), continuous darkness (24 h), restraint (2 h), strobe light (12 h), empty bottle (6 h), wet bedding (12 h), empty cage (24 h), and scent stimuli (mice were placed in a pepper-smelled chamber for 15 min).

During the 4 weeks, all animals were weighed every Wednesday and Sunday at 15:00. Once weekly, they were deprived of food and water for 12 h and assessed in the sucrose preference test (see section 2.3.1 below). Also once weekly, their physical state was assessed in terms of the coat score assay as described (Cao et al., 2013). The fur in the following seven body areas was assigned 0 point if it was unkempt (un-groomed) or 1 point if it was well-groomed: head, neck, forepaws, dorsal coat, ventral coat, hind paws and tail. The seven scores were summed to obtain the total coat score.

After 4 weeks, animals whose body weight was at least one standard deviation smaller or larger than that of unstressed control animals were classified as showing stress-induced weight loss (WL) or gain (WG), respectively. These two groups were confirmed by body mass index and serum lipid level, and compared in several ways such as based on the fecal microbiome, systemic behavior, hormone levels, peripheral and neural cytokines and hippocampal neurogenesis (see below).

All fecal samples were collected at the same time (19:00) to avoid circadian influences on the microbiome. Fecal samples were examined instead of cecum contents to assess whether differences WL, WG or control. Fecal samples from 7 control animals, 7 WL animals and 6 WG animals were subjected to microbiome analysis. The composition of the fecal microbiome was analyzed using 16S rRNA sequencing, which was performed by Majorbio Bio-Pharm Technology (Shanghai, China). Sequencing experiments involved the following steps: DNA extraction, PCR amplification and product purification, MiSeq library construction, and MiSeq sequencing. PCR amplification was carried out using TransStart FastPfu DNA Polymerase (TransGen Biotech, Beijing, China) and an ABI GeneAmp® 9,700 system (Thermo Fisher Scientific, Wilmington, DE, United States). Bacterial DNA fragments were amplified using forward primer-338 (5′-ACTCCTACGGGAGGAGCAG-3′) and reverse primer-806 (5′-GCACTACHVGGGTWTCTAAT-3′). The DNA was quantified using a Nanodrop spectrophotometer (Thermo Fisher Scientific), and paired end sequencing of the V3-V4 region of 16S rRNA was performed as previously described (Pearson-Leary et al., 2020) using an Illumina MiSeq PE300 platform (Illumina, San Diego, United States). The raw sequencing reads were deposited into the NCBI Sequence Read Archive (BioProject ID PRJNA1047189).

The 300 bp reads were truncated at any site receiving an average quality score of <20 over a 50 bp sliding window, and the truncated reads shorter than 50 bp were discarded, reads containing ambiguous characters were also discarded. Only overlapping sequences longer than 10 bp were assembled according to their overlapped sequence. Barcode accurately matched and separated samples, primers adjusted the sequence direction, and selected sequences with mass greater than 20 (that is, the accuracy of 99%) for downstream analysis. In order to remove the primer cleanly, set the primer mismatch rate to 0.15. Unexcised sequences of primers are discarded directly to prevent false operational taxonomic units (OUTs). Then the optimized sequences were clustered into OTUs using UPARSE 7.1 with 97% sequence similarity level. The most abundant sequence for each OTU was selected as a representative sequence.

Bioinformatic analysis of the gut microbiota was carried out using the Majorbio Cloud platform.¹ Based on the OTUs information, rarefaction curves and alpha diversity indices including observed OTUs, Ace, Chao richness, Simpson and Shannon index were calculated with Mothur v1.30.1. The similarity among the microbial communities in different samples was determined by principal coordinate analysis (PCoA) based on Bray–curtis dissimilarity using Veganv2.5–3 package. The PERMANOVA test was used to assess the percent-age of variation explained by the treatment along with its statistical significance using Vegan v2.5–3 package.

Five mice were randomly selected from each group. These mice were anesthetized (isoflurane 4% induced anesthesia and 2% maintained anesthesia in mice), cardiac puncture was performed to collect the whole blood, and the serum was collected by centrifuging at 4°C 1500 r / min for 15 min. The serum was assayed for low-density lipoprotein cholesterol (LDL), high-density lipoprotein cholesterol (HDL), total cholesterol (TC), triglyceride (TG), corticotropin-releasing hormone (CRH), adrenocorticotropin (ACTH) and corticosterone (CORT), interleukin-6 (IL-6), tumor necrosis factor α (TNF-α) and interleukin-10 (IL-10) (Beyotime. Shanghai, China) in strict accordance with the manufacturer’s instructions. These kits obtained from the Guizhou Lvmeng Weiye Biotechnology Co., LTD (Guizhou, China).

After blood collection, all mice were perfused with 0.9% NaCl. Brains were removed and divided into two hemispheres. The intact hippocampus was removed from left hemisphere and used for transcriptome sequencing and cytokine detection while the right hemisphere was used for immunohistochemical staining.

Hippocampi were dissociated and lysed in RIPA lysis buffer (catalog no. R0010, Solarbio, Beijing, United States) containing phenylmethanesulfonyl fluoride (catalog no. IP0280, Solarbio). Lysates were centrifuged for 15 min at 1000 g at 4°C, the concentration of total protein in the supernatant was determined using a commercial bicinchoninic acid kit (Boster, Wuhan, China), then samples were diluted to a final total protein concentration of 1 g/mL. The pro-inflammatory factor IL-1β was quantitatively analyzed by the double antibody sandwich method (Piao et al., 2023). The diluted samples were added to 96-well plates coated with anti-mouse IL-1β antibody, labeled with biotin, and incubated with avidin peroxidase complex labeled antibody. OD values were detected at 450 nm and protein expression was calculated. The manufacturer-specified detection limit was 2 pg. / mL.

The hippocampal transcriptome was analyzed as described (Zhang et al., 2021). Total RNA of hippocampus was extracted using TRIzol® (Invitrogen, Carlsbad, CA, United States) according to the manufacturer’s instructions. RNA was sequenced by Majorbio Bio-Pharm Technology, and all sequences were uploaded to the NCBI Sequence Read Archive under BioProject ID PRJNA1046681.

The right hemispheres from each group mice were fixed with 4% paraformaldehyde for 48 h, dehydrated, frozen, sliced, thoroughly washed with 0.5% Triton X-100 for 15 min, blocked with 10% donkey serum for 1 h, and incubated at 4°C overnight with monoclonal antibody against Iba1 (1:400; Abcam, Cambridge, United Kingdom) to label microglia or against glial fibrillary acidic protein (1,400; Cell Signaling Technology, Danfoss, MA, United StAtes) to label astrocytes. Sections were washed three times with phosphate-buffered saline (PBS), incubated with secondary antibody in the dark for 2 h, and counterstained with DAPI (1,10,000; Roche, Switzerland).

To measure hippocampal neurogenesis, all mice received intraperitoneal injections of 5-bromo-2-deoxyuridine (5 mg/mL BrdU in 0.9% saline; Sigma, St. Louis, MO, USA) at a dose of 50 mg/kg once every 12 h for 7 days. At 48 h after the last injection, mice were perfused and brains were taken, and one of every six sequential tissue slices containing hippocampus was permeabilized with 0.5% Triton X-100 in PBS for 15 min, incubated in 2 N HCl for 28 min at 37°C, washed twice with 0.1 M borate buffer (pH 8.5) for 10 min each, blocked in 10% donkey serum for 2 h, and incubated with primary antibody against doublecortin (1:300, Cell Signaling Technology) to label immature neurons, BrdU (1:400, Cell Signaling Technology) to label proliferating neural stem/progenitor cells, or nuclear protein NeuN (1:400, Cell Signaling Technology) to label neurons. Then sections were incubated for 2 h at room temperature with Alexa-conjugated secondary antibody (1:300; Invitrogen, Carlsbad, CA, United States), followed by counterstaining with DAPI for 5 min.

In both types of staining, sections were imaged under a fluorescence microscope (Olympus IX 73, Tokyo, Japan), and images were analyzed using Image J 1.45 (United States National Institutes of Health, Bethesda, MD, United States). Percentages of total area that were positively stained or numbers of cells were averaged across five micrographs at 40× magnification per mouse.

Data were analyzed for normal distribution using the Shapiro–Wilk test, and reported as mean ± standard error. Differences among unstressed control mice and those showing stress-induced weight loss or gain were assessed for significance using one-way analysis of variance, followed by Tukey’s post hoc test for multiple comparisons. All statistical analyses were performed using GraphPad Prism 8.0 (GraphPad, San Diego, CA, United States). Results associated with p < 0.05 were considered significant.

Mice were exposed to CMS for 4 weeks and divided into weight gain (WG), weight loss (WL) and weight insusceptible sub-group based on their body weight changes relative control mice (Figures 1A,B). We traced the changes in body mass index and food consumption of each sub-group of animals during the CMS period, and found that WG mice showed increase and WL mice showed decrease in their body weight from week-2 to week-4 when compared with control mice (Figure 1C). However, there was no significant difference in food consumption among control, WG, WI, and WL mice during 4 weeks of CMS exposure (Figure 1D).

Chronic mild stress led to significant weight gain or loss in subsets of animals starting halfway through the four-week stress paradigm, even though all animals consumed similar amounts of normal diet throughout the paradigm (Figures 1A–D). Animals showing stress-induced weight gain also showed up-regulation of total cholesterol and triglyceride in serum, consistent with obesity (Du et al., 2019; Figures 1E–H). There was no significant difference in the above indexes between weight insusceptible sub-group and the control group.

Chronic mild stress induced depression- and anxiety-like symptoms in mice, regardless of whether they showed significant weight gain or loss (Figure 2; Supplementary Figure S2). However, the detailed nature of the symptoms differed between the two groups: animals that gained weight performed worse in the forced swimming and tail suspension tests than animals that lost weight, indicating greater behavioral despair (Porsolt et al., 2001); conversely, animals that lost weight showed lower sucrose preference, indicating greater anhedonia (Liu et al., 2018).

Chronic mild stress did not alter the animals’ natural lack of location preference (Figures 3A–C), but it did significantly reduce the time that animals with weight gain spent exploring objects in novel locations (Figures 3D–F) and exploring novel locations (Figures 3G–I). These results link stress-induced obesity to impaired memory of objects and locations. These behavioral changes were not observed in the weight insusceptible sub-group.

Taken together, these behavioral experiments suggest that chronic mild stress can induce depression- and anxiety-like behaviors in animals regardless of whether it also leads to significant weight gain, whereas such stress impairs cognitive function only when it leads to significant weight gain.

Animals showing stress-induced weight gain also showed upregulation of corticotropin-releasing hormone, adrenocorticotropin and corticosterone in serum, consistent with HPA activation (Stout et al., 2002; Figures 4A–C). Animals showing stress-induced weight loss showed upregulation of corticosterone relative to unstressed controls, but the levels were not as high as in mice with stress-induced weight gain. Consistent with the link between HPA activation and upregulation of pro-inflammatory cytokines (Ensminger et al., 2021), we found that animals showing stress-induced weight gain also showed upregulation of tumor necrosis factor-α and interleukin-6 in serum (Figures 4D–F). This upregulation was not observed in animals showing stress-induced weight loss and weight insusceptible. Both WG and WI animals showed downregulation of the anti-inflammatory cytokine interleukin-10. There was no significant difference in the above indexes between weight insusceptible sub-group and the control group.

Since the WI group showed no significant difference from the control group in behavioral test results such as anxiety, depression and cognitive function, and the expression levels of obesity-related indicators and pro-inflammatory factors. Therefore, in the subsequent analysis, we focused on WG and WL groups.

Fecal bacteria from animals showing stress-induced weight gain or loss showed significantly higher Ace and Chao indices than those from unstressed controls (Figure 5A), suggesting that stress increased the richness of gut microbiota. In contrast, stress did not significantly affect Shannon or Simpson indices of fecal bacteria, suggesting minimal effects on the alpha diversity of gut microbiota. Principal component analysis of variation in non-phylogenetic Bray–Curtis metrics showed that the structure of the bacterial community differed significantly between animals showing stress-induced weight loss or gain, and between each of those groups and unstressed controls (Figure 5B).

At the phylum level, animals with stress-induced weight gain showed a significant increase in abundance of Firmicutes bacteria but significant decrease in Bacteroidota bacteria, while the opposite was observed in animals with stress-induced weight loss (Figure 5C). At the family level, animals with stress-induced weight gain showed a significant increase in abundance of Erysipelotrichaceae, Lactobacillaceae and Prevotellaceae, but significant decrease in Lactobacillaceae bacteria; the opposite was observed in animals with stress-induced weight loss (Figure 5D).

At the genus level, animals with stress-induced weight gain showed a significant increase in Dubosiella, while the opposite was observed in animals with stress-induced weight loss (Figure 5E). Those with weight gain showed a significant increase in abundance of Turicibacter bacteria but a significant decrease in abundance of Lactobacillus and Alistipes bacteria. Animals with weight loss showed a significant increase in abundance of Enterorhabdus bacteria. In both groups of mice, stress induced similar changes in abundances of Alloprevotella, bifidobacterium, Desulfovibrio, Prevotellaceae UG-001 and Lachnospiraceae NK4A136 bacteria.

Mice with stress-induced weight gain showed upregulation of 1,431 genes and downregulation of 788 genes in the hippocampus relative to mice with stress-induced weight loss (Figures 6A–C). Most of these differentially expressed genes were involved in processes related to learning or memory, eating behavior, stress response, cytokine production, hormone secretion regulation, oxidative stress response and corticosteroid response; and these processes involved apoptosis, fatty acid biosynthesis, and signaling mediated by PI3K-Akt, NF-κB, PPARs, Toll-like receptor and NOD-like receptor (Figures 6D–F). We found that oxidative stress response in the hippocampus was implicated in stress-induced body weight change (Supplementary Figure S3). Gene set enrichment analysis confirmed that genes involved in responding to oxidative stress were upregulated in the hippocampus of animals showing stress-induced weight gain relative to those showing weight loss (Figures 7A,B; Supplementary Figure S2).

Both groups of mice showed significantly larger numbers of hippocampal microglia (detected based on Iba1 expression) than unstressed controls, and a significantly larger proportion of those microglia had enlarged somata, reduced branching and thicker, shorter processes than in controls (Figures 7C–H). These results indicate stress-induced hyperactivation of hippocampal microglia, which was stronger in animals with weight gain than in animals with weight loss. The greater activation in mice with weight gain was associated with higher levels of the pro-inflammatory cytokine interleukin-1β in hippocampus (Figure 7I). Both groups of mice showed significantly larger numbers of hippocampal astrocytes (detected based on expression of glial fibrillary acid protein; Figures 7J,K).

Consistent with the link between hyperactivation of hippocampal microglia and impaired neurogenesis in the adult hippocampus (He et al., 2022; Su et al., 2023; Zhang et al., 2020), regulating neurogenesis in the hippocampus was implicated in stress-induced body weight change (Supplementary Figure S4). Genes involved in regulating neurogenesis made up a significantly greater proportion of the genes differentially regulated by stress in animals showing weight gain than in those showing weight loss (Figures 8A,B). Immunostaining of thin sections from the subgranular zone of the hippocampus showed that both groups of mice contained significantly fewer immature neurons (detected based on doublecortin expression), proliferating neural stem cells/precursor cells (detected based on BrdU labeling) and newborn neurons (detected based on simultaneous doublecortin staining and BrdU labeling) than unstressed controls (Figures 8C–F). The reductions were significantly more severe in animals showing stress-induced weight gain.

Animals with weight gain, but not those with weight loss, showed significantly less differentiation of neural stem cells/precursor cells into neurons than unstressed controls (Figure 8G). In the subgranular zone, animals showing weight gain contained significantly fewer new mature neurons (detected based on simultaneous NeuN staining and BrdU labeling) than animals showing weight loss (Figures 8H–J).

Given our observation that stress-induced obesity in mice causes hyperactivation of the HPA axis and hippocampal glia, inhibition of hippocampal neurogenesis and cognitive decline, we next wanted to know whether these changes are related to the disturbance of gut microbiome. Our results showed that the abundance of Firmicutes, Dubosiella and Turicibacter correlated positively with stress-induced obesity, activation of HPA axis and hippocampal glia, impairment of hippocampal neurogenesis and cognitive decline. The abundance of Bacteroidota, Lactobacillus, Enterorhabdus, Prevotellaceae UCG-001, Alistipes, Lachnospiraceae NK4A136, Alloprevotella, Bifidobacterium and Desulfovibrio correlated negatively with stress-induced obesity, activation of HPA axis and hippocampal glia, impairment of hippocampal neurogenesis and cognitive decline (Figure 9A). The stress-induced obesity correlated positively with hyperactivation of the HPA axis and hippocampal glia, impairment of hippocampal neurogenesis and cognitive decline. The hyperactivation of the HPA axis correlated positively with activation of hippocampal glia, and correlated negatively with hippocampal neurogenesis. The hyperactivation of hippocampal glia correlated negatively with hippocampal neurogenesis. The hippocampal neurogenesis correlated negatively with cognitive decline (Figure 9B). These results suggested that stress-induced obesity in mice causes cognitive decline associated with inhibition of hippocampal neurogenesis and dysfunctional gut microbiota.

To further explore the mechanism of overactivation of microglia and neurogenesis impairment in hippocampus of WG mice, gene set enrichment analysis of KEGG pathway based on differentially expressed genes among WG, WL and Ctrl mice were performed. Multichannel gene set enrichment analysis, indicating the differentially expressed genes that positively regulated oxidative phosphorylation, fatty acid biosynthesis, toll-like receptor signaling pathway, adipocytokine signaling pathway, NF-kappa B signaling pathway, NOD-like receptor signaling pathway, TNF signaling pathway and ferroptosis were significantly enriched in hippocampus of WG mice compared to WL or Ctrl animals (Figures 10A–D), suggesting that the hyperactivation of microglia and neuroinflammation in hippocampus induced by stress-mediated obesity are associated with the activation of these signaling pathways. In addition, we found that PPAR signaling pathway, a signaling pathway involved in lipid metabolism and anti-inflammation, was inhibited in hippocampus of WG mice compared to WL animals (Figure 10E). The signaling pathways (PI3K-AKT, cGMP-PKG, ErbB, neurotrophin, Wnt and Notch), which are involved in the proliferation, differentiation and survival of neural stem/precursor cells, were inhibited in hippocampus of WG mice compared to WL animals (Figures 10E,F), suggesting that the impairment in hippocampal neurogenesis of WG mice are associated with the activation of these signaling pathways.