Phlorizin reduces depression-like behaviors in mice by changing gut bacteria and methionine to lower inflammation in the memory area of the brain

Mouse body weights were determined per week over a 5-week period to appraise the influence of PHZ (Figure 1A). As to the CORT-treated mice, the body weight gradually decreased relative to that of controls, indicating that CORT may affect body weight in growing mice. Throughout the medication administration period, the mice in all treatment groups experienced a significantly greater rate of weight gain than the model group. It is noteworthy that our findings indicate that PHZ exerts a similar effect to sertraline on counteracting CORT-induced weight variations. Depressive-like behaviourbehaviors were assessed using the sucrose preference test (SPT), immobility time in the forced swim test (FST), immobility time in the tail suspension test (TST), and open field test (OFT). CORT-induced depression caused anhedonia and despair, as reflected by a decrease in SPT and a rise in immobility time relative to controls (Figure 1B–D). An obvious decrease in immobility duration in both the TST and FST after the administration of PHZ indicates the PHZ's therapeutic potential. In the OFT (Figure 1E–G), mice with CORT-induced depression exhibited reduced locomotor activity, and central area retention time than controls. In contrast, PHZ increased the total distance moved and time spent in the centrecenter after 5 weeks of administration compared with the CORT group, suggesting that PHZ alleviated CORT-induced depression.

To appraise the influence of PHZ on neuroinflammation among CORT-treated mice, we first measured the levels of anti-inflammatory cytokines (IL-4 and IL-10) and pro-inflammatory cytokines (IL-1β and TNF-α) within the hippocampus via ELISA. CORT treatment remarkably raised the expressions of pro-inflammatory cytokines and decreased the levels of anti-inflammatory cytokines, relative to controls. By contrast, PHZ treatment decreased the expressions of IL-1β and TNF-α, while increasing IL-10 and IL-4 levels in a dose-dependent manner (Figure 2A). The above finding indicates that PHZ effectively modulates the inflammatory response in the hippocampus of mice suffering from CORT-induced depression. Hence, the anti-inflammatory effects of PHZ prompted us to further investigate its role in microglial activation and polarization. To investigate PHZ's potential role in microglial activation, immunofluorescence staining for the microglial marker Iba1 was performed (Figure 2B). CORT treatment brought about the obviously heightened quantity of activated Iba-1⁺ microglia within the hippocampus. However, PHZ treatment maintained Iba-1⁺ microglia numbers comparable to controls, indicating that PHZ attenuated CORT-induced microglial activation (Figure 2C). We further investigated the effects of PHZ on microglial polarization. Western blot analysis was used to measure NOS2 (M1 marker) and Arg1 (M2 marker) levels in the hippocampus (Figure 2D–F). In the CORT-treated group, a marked increase in NOS2 levels and reduction in Arg1 levels were observed, indicating a shift towards M1 polarization. In contrast, PHZ treatment significantly decreased NOS2 levels and increased Arg1 levels, suggesting that PHZ promoted M2 polarization and reduced M1 activation in the hippocampus. These results demonstrate that PHZ effectively suppresses abnormal microglial activation, modulates the polarization, and reduces neuroinflammation.

To explore whether the efficacy of PHZ on depression is related to the gut microflora, the microbial composition of fecal samples from each experimental group was analyzed by 16S rRNA sequencing. As indicated by the α diversity indices (Figure 3A), CORT treatment resulted in significantly lower Chao1, Shannon, and Simpson PD indices, reflecting reduced microbial richness and diversity, which corresponds to former reports on gut dysbiosis in depression models [21,22]. PHZ reversed these indicators, increasing microbial richness and diversity. Beta diversity further showed remarkable disparities in microbial communities between CORT and control mice (Figure 3B), suggesting a shift in the microbiota composition due to CORT treatment. PHZ-50 intervention affected the largest number of bacterial genera and modulated the host gut microbiota more efficiently (Figure 3C), restoring microbial balance in the CORT mice. The relative abundances (RAs) of the gut microflora were enumerated at the phylum level (Figure 4A). CORT significantly increased Firmicutes, Verrucomicrobiota, and Proteobacteria relative to controls, whereas PHZ recovered these levels and prominently lessened the Firmicutes/Bacteroidetes ratio, in a dose-dependent manner (Figure 4B). At the family level, Muribaculaceae, Bacteroidaceae, Prevotellaceae, Lachnospiraceae, Akkermansiaceae, and Lactobacillaceae were the dominant families (Figure 4C). Compared with the CORT group, PHZ treatment significantly lessened the RAs of Akkermansiacea, Bacteroidaceae, and Lachnospiraceae, whereas it markedly raised the abundance of Muribaculaceae (Figure 4D). At the genus level, the top 20 genera by RAs (Figure 5A) encompassed 13 genera whose relative abundance trends were reversed by PHZ treatment (Figure 5B). In detail, Alloprevotella and Lachnospiraceae NK4A136 group significantly increased, whereas Bacteroides, Lachnoclostridium, Akkermansia, Hungatella, and Blautia decreased. Finally, the PICRUSt2 analysis for forecasting the functional profiles of the gut microbial community exhibited several remarkable disparities between groups at KEGG levels. Specifically, 20 mg/kg was chosen as the low dose for PICRUSt2 experiments, as it represents the minimum effective dose with anti-inflammatory and antidepressant effects, while avoiding high-dose confounding factors. The results of the PICRUSt2 analysis indicate that PHZ may influence metabolic pathways by affecting the microbiota, such as the pentose phosphate pathway, pentose metabolism, glucuronidation, cysteine and methionine metabolism, and fatty acid biosynthesis (Figure 6).

Given the significant role of metabolites in host–microbiota communication, we analyzed serum metabolites. Orthogonal Projections to Latent Structures Discriminant Analysis (OPLS-DA) revealed clear separation between the experimental groups and CORT-treated mice (Figure 7A–C). The score plots of serum metabolites, shown in Figure S1, confirmed the model's predictive ability for subsequent analysis. Subsequently, 38 differential metabolites were identified (Table S1). CORT reduced the levels of 23 metabolites, including prostaglandin 2, L-Met, and L-palmitoylcarnitine, while significantly increasing the levels of 15 metabolites, such as N-arachidonoyl glycine, N-palmitoyl leucine, trigonelline, and cortisone acetate. PHZ significantly modulated the abundance of these metabolites (Figure 7D). Further metabolic pathway analysis indicated that PHZ affected nine key metabolic pathways, including cysteine and methionine metabolism, phenylalanine metabolism, steroid hormone biosynthesis, pentose and glucuronate interconversions, arachidonic acid metabolism, fatty acid degradation, aminoacyl-tRNA biosynthesis, and drug metabolism-cytochrome P450 (Figure 7E). Notably, cysteine and methionine metabolism exhibited the highest pathway impact value and was regarded as the most critical pathway.

We further explored whether PHZ can modulate the Met metabolism in brain tissue. To this end, we conducted targeted metabolomics analysis on hippocampal tissue, as the hippocampus is highly susceptible to inflammatory damage and is crucial for emotional regulation. This allowed us to investigate the effects of PHZ on Met metabolism and the potential role of microbial depletion in this process. Figure 8 shows that PHZ-20 prominently raised the levels of SAMe and L-MET in the hippocampus. These two substances are key regulatory factors of Met metabolism and are closely related to neuroinflammation [23]. ABx treatment significantly weakened this effect. These results suggest that PHZ-mediated regulation of metabolites may contribute to its anti-inflammatory effects, and that the gut microbiota plays a pivotal role in this metabolic regulation.

To comprehensively explore the anti-inflammatory properties of L-Met, we treated LPS-induced BV2 microglial cells with L-Met and assessed its effect on microglial polarization and production of proinflammatory mediators. BV2 cell viability remained unaffected after treatment with L-Met (Figure S2). LPS stimulation significantly raised the yield of proinflammatory cytokines, like IL-1β, TNF-α, and NOS2, indicating an M1 (pro-inflammatory) polarization of microglia. In contrast, levels of M2 anti-inflammatory cytokines, like IL-4, IL-10, and Arg1, were remarkably reduced following LPS exposure. Importantly, L-Met treatment effectively inhibited the synthesis of LPS-induced M1 pro-inflammatory cytokines (IL-1β, TNF-α, and NOS2) while facilitating the production of M2 anti-inflammatory cytokines (IL-4, IL-10, and Arg1) (Figure 9A). Results suggest that L-Met reduces the inflammatory response and shifts microglial polarization towards the anti-inflammatory M2 phenotype. Flow cytometry deeply confirmed its regulatory effect on microglial phenotype (Figure 9B). These findings provide cellular-level evidence supporting the notion that L-Met, a key metabolite in the Met pathway, regulates microglial polarization and inflammation. This mechanism accords with our previous animal findings that PHZ regulates inflammatory markers and microglial activation, indicating that L-Met is a central mediator of the MGB in its antidepressant action.

To systematically construe the correlations among inflammatory factors, serum metabolites, and intestinal microbial community, a correlation matrix was produced via computing the Spearman's correlation coefficient (Figure 10). Notably, at the genus level, Parabacteroides, Lachnoclostridium, Parasutterella, Blautia, Hungatella, and Bacteroides proactively relate to pro-inflammatory factors, while Muribaculum is negatively linked to pro-inflammatory factors. Regarding correlations between serum metabolites and gut microbiota, L-Met was associated with all fungal genera. Overall, these findings support the key role of PHZ in modulating the gut–metabolite axis, thereby contributing to its antidepressant effects.

Shanghai Yuanye Biotechnology Co. (Shanghai, China) supplied the PHZ. Sertraline was purchased from Guang'anmen Hospital, China Academy of Chinese Medical Sciences (Beijing, China) (Zhejiang Jingxin Pharmaceutical Co., Ltd., Shaoxing, China, National Medicine Standard H20051076; specification: 50 mg). SAME (S5109) and L-Met (S5633) were purchased from Selleck (Selleck Chemicals, Houston, TX, USA). CORT (S30265) was purchased from Shanghai Yuanye Bio-Technology Co., Ltd (Shanghai, China). Neomycin sulfate (B540734), Ampicillin (A430258), Vancomycin hydrochloride (A430243), and Metronidazole (A429689) were purchased from Sangong Biotechnology (Shanghai, China). DAPI (S2110) was purchased from Solarbio (Beijing, China). Iba-1 (ab178846), NOS2 (ab178945), antibodies against alpha tubulin (ab176560), and FITC-labeled secondary antibodies (ab6717) were purchased from Abcam (Cambridge, UK). Antibodies against caspase 1 (sc-56036) were obtained from Santa Cruz Biotechnology, Inc. (Dallas, TX, USA); antibodies against Arg1 (66129-1-l) were obtained from Proteintech Group, Inc. (Wuhan, China); CD86 (12-0862-82), CD206 (17-2061-82), CD11b (17-0112-83), and ELISA kits of IL-10 (BMS607-3), IL-4 (88-7044-88), IL-1β (BMS6002), and TNF-α (BMS607-3) were obtained from Thermo Fisher Scientific (Waltham, MA, USA); Antibodies against GAPDH (G0100) were obtained from Lablead Biotech Co., Ltd. (Beijing, China); CCK-8 (C0037; Beyotime, Shanghai, China), EZ-press RNA Purification Kit (B0004D; EZBioscience, Roseville, CA, USA, USA), Color Reverse Transcription Kit (A0010CG; EZBioscience, Roseville, CA, USA), Color SYBR Green qPCR Master Mix (A0012-R2; EZBioscience, Roseville, CA, USA). BV2 mouse microglial line, fetal bovine serum (FBS), and 1% penicillin-streptomycin were supplied by ScienCell (Roseville, CA, USA). Cell Counting Kit-8 assay (CCK-8) was provided by Dojindo (Mashiki, Japan).

Mouse TNF-α, IL-1β, IL-4, IL-6, and IL-10 levels in the hippocampal supernatant were quantified through ELISA kits complying with the manufacturer's protocols.

During general anesthesia with 1% pentobarbital sodium (50 mg/kg; Sigma-Aldrich, St. Louis, MO, USA), mouse chest was opened, the heart exposed, and normal saline solution was injected into the heart tip. Following the removal of fluids and the absence of any blood color, whole brain tissue was extracted from the decapitated head and then immersed in 4% paraformaldehyde (Solarbio, Beijing, China) for extra 24 h. The brain specimens were dried with various alcohol concentrations, treated with xylene, and then immobilized in paraffin. Subsequently, the brain samples were cut into consecutive coronal slides, each measuring 5 μm in thickness. Tissue slices were subjected to antigen retrieval by placing them in citric acid antigen repair solution (pH 6.0; Solarbio, Beijing, China). Subsequently, they were treated with the primary antibody Iba-1 (1:2000) in a humid compartment throughout the night at 4 °C. The subsequent day, the slices were cleaned using PBS, subsequently processed with drops of an FITC-conjugated secondary antibody (1:3000), before cultivation for 50 min and away from light. After restaining through DAPI (1:100) for 10 min and washing through PBS, an antifluorescence quenching sealing agent (Solarbio, Beijing, China) was utilized. Finally, slides were observed under a fluorescence microscope (Olympus, Tokyo, Japan) [44,45].

Total proteins from the hippocampus were obtained by a RIPA lysis buffer that contained a 1× protease inhibitor cocktail. To determine the concentration of proteins, a BCA assay test kit (Thermo Fisher Scientific, Waltham, MA, USA) was employed as per the manufacturer's guidelines. To sum up, the proteins underwent separation via a 10% SDS-PAGE gel and were subsequently delivered onto a 0.45 µm PVDF membrane (Millipore, Billerica, MA, USA). Such membranes were then cultivated using 5% non-fat dried milk for 1 h at 23 °C, before an all-night incubation at 4 °C using primary antibodies targeting NOS2 (1:500), Arg1 (1:5000), and alpha tubulin (1:5000). After performing three washes with TBST, the membranes were nurtured using the appropriate goat antirabbit IgG-HRP secondary antibody (1:10,000) for 1 h at 25 °C. Another wash with TBST was performed, after which the ECL reagent was used for observing immunoreactivity. Afterward, such membranes were scanned with ChemiDoc Imagers (BioRad, Hercules, CA, USA). Concurrently, band intensity was assessed via ImageJ software (version 1.52a, National Institutes of Health, Bethesda, MD, USA) [46].

To assess variations of gut microflora, stool samples from five mice per group underwent 16S rRNA sequencing. After DNA extraction, the DNA was detected through 1% agarose gel electrophoresis. Pure DNA underwent polymerase chain reaction amplification as per the manufacturer's guidelines. Sequences that had been optimized through sequence splicing and filtration were obtained, and chimeric reads were removed. Libraries were sequenced through the MiSeq PE300 platform (Illumina, San Diego, CA, USA).

Bacterial diversity was appraised by means of alpha diversity indicators, like Chao1, PD, Shannon, and observed species, as well as principal coordinate analysis (PCoA) on the grounds of Bray-Curtis distances. Enriched bacteria were recognized with linear discriminant analysis (LDA) through LEfSe. Additionally, differential bacterial species were gauged using p ≤ 0.05. The prediction spectrum for gene function within the bacterial domain was constructed using PICRUSt2 (https://github.com/picrust/picrust2↗, accessed on 5 January 2025) [47].

Hippocampal tissues were homogenized in pre-cooled methanol (4:1, tissue to methanol, v/w) with centrifugation at 13,000 rpm for 15 min at 4 °C. The supernatant was filtrated via a 0.22 μm membrane and stored at −80 °C until analysis. Quantification of SAMe (399.4 g/mol, C15H22N6O5S), and L-Met (149.21 g/mol, C5H11NO2S) was performed using UPLC-MS/MS in multiple reaction monitoring mode. For each metabolite, calibration curves were formulated using standard solutions to ensure specificity and sensitivity. The mobile phase comprised acetonitrile with 0.1% formic acid and water, and the metabolites were separated on a reverse-phase column (Waters ACQUITY UPLC BEH C18, 2.1 × 100 mm, 1.7 μm; Waters Corp., Milford, MA, USA) using a gradient elution program. Metabolite concentrations were normalized to the tissue weight for comparative analysis.

Cell activity was appraised through the CCK-8 assay, performed on a 96-well plate (Corning Inc., Corning, NY, USA) [48]. Briefly, BV2 cells were plated and handled using discrepant concentrations of L-Met, both with and without LPS treatment, for a 24 h period. Afterwards, 10 μL CCK-8 reagent was supplemented to each well. The absorbance at 450 nm was gauged through a microplate reader (Thermo Fisher Scientific, Waltham, MA, USA) after a 2 h incubation at 37 °C.

The immortalized BV2 microglial cell line derived from mice was nurtured within MEM medium comprising 10% FBS and 1% penicillin/streptomycin. Such cells were preserved within a humid brooder (Thermo Fisher Scientific, Waltham, MA, USA) under 95% air and 5% CO₂ at 37 °C. The medium was displaced every other day. After reaching 80% confluence, the cells were isolated through 0.25% trypsin and passaged for continued propagation. For the experimental procedures, cells were handled with 10 mM Met for 16 h, before exposed to LPS (500 ng/mL) for 4 h.

According to the manufacturer's specifications, total RNA was obtained from each specimen through the EZ-press RNA Purification Kit (EZBioscience, Roseville, CA, USA). Complementary DNA synthesis was performed through the Color Reverse Transcription Kit, with the final reaction volume adjusted to 20 μL. Quantitative real-time PCR (qPCR) was executed thrice through the 2× Color SYBR Green qPCR Master Mix. As for all genes, the mRNA levels were standardized to the β-actin mRNA levels as a reference. Used primers in are listed in Table 1.

Cell surface markers CD11b, CD86, and CD206 were detected via flow cytometry. Briefly, after BV2 cells were treated as described in Section 4.10, cells were harvested through EDTA-free trypsin, underwent centrifugation at 1000 rpm for 5 min, and were rinsed twice using cold PBS to collect cell deposits. Subsequently, such cells were suspended in 100 μL PBS again and stained with anti-CD11b (1:80), anti-CD206 (1:160), and anti-CD86 (1:300) antibodies at 4 °C for 20 min. All specimens were construed via imaging flow cytometry (ImageStream^®X Mark II; Luminex Corp., Austin, TX, USA). In addition, data analytics was performed using IDEAS software (version 6.2; Luminex Corp., Austin, TX, USA).

All data are denoted as mean ± standard error. The data was visualized through GraphPad Prism (v 6.01) (GraphPad, San Diego, CA, USA). One-way analysis of variance was used for appraising data significance, before a post-hoc Student–Newman–Keuls test and an unpaired two-tailed Student's t-test. In this study, p < 0.05 held statistical significance. The significance levels were denoted as * p < 0.05, ** p < 0.01, and *** p < 0.001. The differences in effect size ratios of the linear discriminant analysis on core flora between the groups were rated through the non-parametric Kruskal–Wallis rank sum test. Spearman's correlation was adopted to evaluate the interrelationships among parameters.