Modulation of the microbiota-lipid-brain axis by modified Chaihu-Longgu-Muli Decoction ameliorates chronic stress-induced depression
1Department of Traditional Chinese Medicine, The First Affiliated Hospital, Fujian Medical University, Fuzhou, China
2National Regional Medical Center, Binhai Campus of the First Affiliated Hospital, Fujian Medical University, Fuzhou, China
*Correspondence: Xuanbin Huang, hxb2138@163.com Tianmin Wu, tianmwu@sina.comAbstract
Background and aims
Modified Chaihu-Longgu-Muli Decoction (mCLMD) has well-established clinical antidepressant efficacy, yet its precise systemic mechanisms of action remain incompletely characterized. Here, we employed an integrated multi-omics strategy to delineate the mechanisms by which mCLMD exerts antidepressant-like effects via modulation of the gut microbiota-lipid-brain axis.
Experimental procedures
Rats exposed to chronic unpredictable mild stress (CUMS) were administered graded doses of mCLMD. Depressive-like behaviors were assessed using the sucrose preference test, forced swim test, and open field test. To delineate the underlying mechanisms, we integrated network pharmacology analysis, 16S rRNA gene sequencing, serum metabolomics, and targeted validation of hippocampal signaling pathways.
Results
mCLMD administration dose-dependently reversed CUMS-induced depressive-like behaviors in rats. Consistent with network pharmacology predictions, 16S rRNA sequencing revealed that mCLMD ameliorated CUMS-induced gut dysbiosis, characterized by reduced relative abundance of pro-inflammatory genera (Colidextribacter, Oscillibacter) and enrichment of beneficial taxa (Romboutsia, Lactobacillus). Metabolomic profiling demonstrated concomitant restoration of dysregulated lipid and neurosteroid profiles in serum. Correlation analysis identified that reduced abundance of stress-associated pathobionts was tightly linked to decreased levels of peripherally derived neurosteroids with neurotoxic potential (e.g., pregnenolone), while enrichment of beneficial commensals correlated with elevated levels of neuroprotective endocannabinoid precursors (including 1-stearoyl-2-arachidonoylglycerol). These peripheral immunometabolic alterations were accompanied by the transcriptional upregulation of the hippocampal cAMP-BDNF–TrkB signaling pathway and restoration of monoaminergic neurotransmission.
Conclusion
The antidepressant effects of mCLMD are strongly associated with systemic remodeling of the gut microbiota-lipid-brain axis. These therapeutic effects potentially stem from both the direct pharmacological activity of mCLMD’s bioactive compounds and indirect modulation of the gut microbiota and host metabolism. Collectively, our findings provide robust preclinical evidence underpinning the clinical application of mCLMD in the management of major depressive disorder.
1Introduction
Major depressive disorder (MDD) is a chronic, relapsing psychiatric disorder affecting over 280 million people worldwide and the leading cause of years lived with disability globally (Rong et al., 2025). For more than half a century, first-line antidepressant therapies have been based on the monoamine deficiency hypothesis. However, this classical framework cannot fully explain the 2–4 week delay in therapeutic onset or the ~30% rate of treatment resistance seen in clinical practice (Cipriani et al., 2018; Ding et al., 2024). Mounting evidence now demonstrates that MDD pathogenesis is not limited to the central nervous system but involves systemic disturbances including neuroinflammation, hypothalamic–pituitary–adrenal (HPA) axis hyperactivity, and widespread metabolic dysregulation (Ding et al., 2024). These interconnected pathologies highlight the need for safer, multi-target therapeutic strategies.
Traditional Chinese medicine (TCM), with its holistic philosophy and multi-component regulatory effects, represents a valuable resource for developing novel antidepressants. Chaihu-Longgu-Muli Decoction (CLMD) is a classical herbal formula that has been used clinically for centuries to manage anxiety, insomnia, and depressive symptoms. Numerous preclinical studies and meta-analyses consistently show that CLMD exerts significant neuroprotective effects—including reducing neuroinflammation and upregulating hippocampal brain-derived neurotrophic factor (BDNF)—and enhances efficacy while reducing side effects when combined with selective serotonin reuptake inhibitors (Wang et al., 2019; Jia et al., 2023).
Building on this work, we developed a modified formulation of CLMD (mCLMD) by integrating modern pharmacological knowledge with clinical experience. This optimized formula incorporates botanicals with well-documented neuroprotective, anxiolytic, and prebiotic properties—including Bulbus Lilii, Radix Rehmanniae, Radix Polygalae, and Sclerotium Poriae Pararadicis—while omitting components with redundant or non-essential effects. These modifications were designed to synergistically enhance BDNF/TrkB signaling (Wang et al., 2024; Wang et al., 2025), modulate monoaminergic and GABAergic neurotransmission (Chen et al., 2023; Zhou et al., 2025), and maintain gut microbial homeostasis (Du et al., 2020; Liu et al., 2024; Yan et al., 2025).
Despite its promising clinical efficacy, the systemic mechanisms underlying mCLMD’s antidepressant effects remain poorly understood. While network pharmacology studies have provided a preliminary framework pointing to a “periphery-to-central” mode of action, rigorous in vivo validation using multi-omics approaches is needed to identify the specific biological pathways linking peripheral immunometabolic changes to central neuroplasticity (Zhao et al., 2024).
The microbiota-gut-brain axis has emerged as a key paradigm in understanding MDD pathogenesis. Gut dysbiosis triggers both systemic immune activation and metabolic dysregulation (Yang et al., 2020; Hao et al., 2024) Importantly, the gut microbiome is now recognized as a major regulator of systemic lipid mediators and neurosteroid biosynthesis. Microbiota-driven alterations in circulating endocannabinoid precursors and neurosteroids can directly influence CNS function by crossing the blood–brain barrier and modulating hippocampal neuroplasticity via the cAMP-BDNF–TrkB pathway, which is significantly suppressed during chronic stress (Casarotto et al., 2021; Markov and Novosadova, 2022).
Based on these findings, we hypothesized that mCLMD exerts antidepressant-like effects by restructuring the gut microbiota and correcting systemic metabolic abnormalities, particularly in lipid and neurosteroid pathways. To test this hypothesis, we used an integrated approach combining network pharmacology, 16S rRNA gene sequencing, UPLC-MS/MS untargeted metabolomics, and targeted molecular validation of hippocampal signaling pathways in a rat model of chronic unpredictable mild stress (CUMS). Our findings offer insights into the systemic mechanisms of mCLMD, supporting its clinical application in the treatment of major depressive disorder.
2Materials and methods
2.1Preparation of modified Chaihu-Longgu-Muli Decoction (mCLMD)
Modified Chaihu-Longgu-Muli Decoction (mCLMD) is derived from the classical formula recorded in the Treatise on Cold Damage Disorders. To optimize its antidepressant efficacy, we made three key modifications: addition of six herbs (Bulbus Lilii, Radix Rehmanniae, Radix Polygalae, Semen Ziziphi Spinosae, Fructus Tritici Levis, and Rhizoma Chuanxiong), replacement of Poria with Sclerotium Poriae Pararadicis, and substitution of Ginseng Radix with Radix Codonopsis. This formulation follows the core TCM therapeutic principle of “soothing the liver, relieving depression, nourishing the heart and calming the mind”, which addresses the core TCM pathogenesis of depression: liver qi stagnation with heart-spleen deficiency. The optimized formula was specifically designed to regulate neuroinflammation, promote hippocampal neuroplasticity, and exert antidepressant effects. It comprises 15 herbal and mineral ingredients, with full details of composition and dosage listed in Table 1.
| Chinese name | Latin botanical name | Medicinal name (Pharmacopoeia) | Weight (g) |
|---|---|---|---|
| Chai Hu | Bupleurum chinense DC. | Radix Bupleuri | 10 |
| Gui Zhi | Cinnamomum cassia Presl | Ramulus Cinnamomi | 10 |
| Long Gu | Fossilia Ossis Mastodi | Os Draconis (Fossilized bones) | 20 |
| Mu Li | Ostrea gigas Thunberg | Concha Ostreae | 20 |
| Bai He | Lilium lancifolium Thunb. | Bulbus Lilii | 20 |
| Sheng Di Huang | Rehmannia glutinosa Libosch. | Radix Rehmanniae | 15 |
| Huang Qin | Scutellaria baicalensis Georgi | Radix Scutellariae | 10 |
| Ban Xia | Pinellia ternata (Thunb.) Breit. | Rhizoma Pinelliae Praeparatum | 10 |
| Suan Zao Ren | Ziziphus jujuba Mill. var. spinosa | Semen Ziziphi Spinosae | 15 |
| Yuan Zhi | Polygala tenuifolia Willd. | Radix Polygalae | 8 |
| Fu Xiao Mai | Triticum aestivum L. | Fructus Tritici Levis | 30 |
| Gan Cao | Glycyrrhiza uralensis Fisch. | Radix Glycyrrhizae | 5 |
| Dang Shen | Codonopsis pilosula (Franch.) Nannf. | Radix Codonopsis | 15 |
| Fu Shen | Poria cocos (Schw.) Wolf | Sclerotium Poriae Pararadicis | 15 |
| Chuan Xiong | Ligusticum chuanxiong Hort. | Rhizoma Chuanxiong | 10 |
All crude medicinal materials were authenticated by a licensed TCM pharmacist and obtained from the Department of Traditional Chinese Medicine Pharmacy, First Affiliated Hospital of Fujian Medical University. Mineral ingredients (Os Draconis and Concha Ostreae) were crushed and decocted alone for 30 min prior to adding the remaining herbs. The decoction was prepared according to the water extraction guidelines for TCM formulas in the Pharmacopoeia of the People’s Republic of China (2020 Edition), with minor adjustments based on classical literature for the original Chaihu-Longgu-Muli Decoction. Briefly, the combined herbs were soaked in 10 volumes of distilled water (v/w) for 30 min and extracted twice under reflux for 1 h each time. The combined filtrates were concentrated under reduced pressure at 60 °C using a rotary evaporator to a final concentration of 2.22 g crude drug/mL.
2.1.1Quality control
All raw materials complied with the quality and safety standards of the Pharmacopoeia of the People’s Republic of China (2020 Edition). The extraction process was strictly standardized across all batches to ensure consistent extraction efficiency. All batches were tested for heavy metals and arsenic, with results within acceptable safety limits. Stock extracts were stored at 4 °C and diluted to the desired concentrations with distilled water immediately prior to intragastric administration.
2.2Network pharmacology analysis
2.2.1Active ingredients and target acquisition
We retrieved active compounds for each herb in mCLMD from the TCMSP and HERB databases (data retrieved January 2026). For ingredients not covered in these databases, we supplemented 14 additional antidepressant-active components identified through a systematic literature review. Compounds were screened using standard thresholds of oral bioavailability (OB) ≥ 30% and drug-likeness (DL) ≥ 0.18, the standard criteria for bioactive component identification in TCM network pharmacology studies (Ru et al., 2014). All compounds with confirmed toxicity were excluded from further analysis. We collected corresponding targets for the screened active compounds and converted them to official human gene symbols using the UniProtKB database (release 2026_01). We used human targets because core pathogenic pathways in depression and their gene orthologs are highly evolutionarily conserved between Homo sapiens and Rattus norvegicus (average sequence identity > 85% for the neurotrophic signaling and monoaminergic neurotransmission pathways investigated in this study).
2.2.3Network construction and core target screening
We identified overlapping targets between mCLMD active compounds and depression-related targets using a Venn diagram. We constructed a protein–protein interaction (PPI) network of the overlapping targets using the STRING database (confidence score > 0.90, species: Homo sapiens, disconnected nodes removed). We selected this high-confidence threshold to balance network comprehensiveness and biological reliability. The PPI network was visualized in Cytoscape (v3.10). We screened core hub targets using the CytoNCA plugin via two rounds of median filtering based on four centrality metrics: degree, betweenness, closeness, and eigenvector centrality.
2.2.4GO and KEGG enrichment analysis
We performed Gene Ontology (GO) functional annotation and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis on the core hub targets using the clusterProfiler package in R (v4.5.3). An adjusted p-value < 0.05 (Benjamini-Hochberg correction) was set as the statistical significance threshold.
2.3Animals, ethical statement, and welfare monitoring
Male Sprague–Dawley rats (6–8 weeks, 200 ± 20 g) were housed under standard SPF conditions. Procedures were approved by the IACUC of Jiangxi Zhonghong Boyuan Biological Technology Co., Ltd. (K-2024-0740-1) and adhered to NIH guidelines and the 3Rs principles. In vivo models were irreplaceable for studying systemic microbiota-brain interactions (Replacement). Sample sizes (n = 6/group) were statistically minimized for exploratory multi-omics screening (Reduction). Trained staff conducted daily welfare monitoring of body weight and behavior (Refinement). Humane endpoints (>20% weight loss, self-mutilation, or resource inaccessibility) were established; no animals reached these endpoints or died prematurely. Rats failing initial modeling criteria (sucrose preference <60%) were excluded pre-treatment. As terminal multi-organ collection precluded rehoming, rats were deeply anesthetized (2% sodium pentobarbital, 40 mg/kg, i.p.) and humanely euthanized via cervical dislocation per AVMA guidelines.
2.4Experimental design and drug administration
After a 1-week acclimation period, we randomly assigned rats to 6 groups (n = 6 per group) using a random number table:
- Control group (CON): normal housing + intragastric administration of distilled water.
- Model group (MOD): CUMS exposure + distilled water
- Positive control group (VEN): CUMS exposure + venlafaxine hydrochloride at 15.63 mg/kg/day.
- Low-dose mCLMD group (CL): CUMS exposure + mCLMD at 11.09 g/kg/day.
- Medium-dose mCLMD group (CM): CUMS exposure + mCLMD at 22.19 g/kg/day.
- High-dose mCLMD group (CH): CUMS exposure + mCLMD at 44.38 g/kg/day.
The medium mCLMD dose was converted from the standard clinical adult dose (213 g crude drug per 60 kg body weight) using body surface area normalization; low and high doses were set at 0.5 × and 2 × the medium dose, respectively.
All rats first underwent 4 weeks of CUMS exposure to establish the depressive-like model. We confirmed successful model establishment using the sucrose preference test (SPT) at the end of the 4-week CUMS period; only rats with sucrose preference < 60% relative to the CON group mean were included in the subsequent 4-week treatment phase.
All treatments were administered once daily via intragastric gavage at a constant volume of 10 mL/kg body weight for 4 consecutive weeks, between 9:00 and 10:00 a.m. The CUMS procedure continued for a total of 8 weeks: 4 weeks for model establishment, followed by 4 weeks of concurrent treatment and stress exposure (Markov and Novosadova, 2022). We applied 7 unpredictable mild stressors in random order (1–2 stressors per day, with no identical stressor repeated within 72 h to maintain unpredictability): 24-h food deprivation, 24-h water deprivation, 12-h reversed light/dark cycle, 24-h wet cage bedding, 45° cage tilt for 12 h, 30-min white noise exposure (85 dB), and 1-min tail clamping (1 cm from the tail tip). All stressors were applied between 8:00 a.m. and 6:00 p.m.
After the 4-week treatment period, we performed behavioral tests in the following order: SPT, open-field test (OFT), and forced swim test (FST). A 24-h recovery interval was allowed between tests to minimize carry-over effects. To balance mechanistic depth with sequencing and metabolomics costs, we prioritized the high-dose group because it exhibited the most robust and consistent behavioral efficacy across all behavioral paradigms. The present study was therefore designed primarily to identify core therapeutic mechanisms associated with maximal pharmacodynamic response rather than to establish dose–response omics relationships.
At the end of the experiment, we deeply anesthetized rats with 2% sodium pentobarbital (40 mg/kg, intraperitoneal injection). We collected blood from the abdominal aorta within 5 min of anesthesia induction, separated serum by centrifugation at 3000 × g for 15 min at 4 °C, and stored samples at −80 °C for untargeted metabolomics analysis. Rats were then euthanized by cervical dislocation in accordance with AVMA guidelines. We rapidly isolated whole brains on ice within 3 min of euthanasia and dissected bilateral hippocampi for molecular assays. We aseptically collected colonic fecal contents from the distal colon for 16S rRNA gene sequencing. All samples were snap-frozen in liquid nitrogen immediately after collection and stored at −80 °C until analysis.
Blinding and randomization: Allocation concealment was achieved using sequentially numbered, opaque sealed envelopes. Behavioral testing, sample processing, and all omics data analyses were performed by investigators blinded to group assignments. Furthermore, the order of sample processing and instrument injection for all molecular assays and omics analyses was fully randomized to minimize potential batch effects.
2.5Behavioral tests
All behavioral tests were performed in a sound-attenuated, temperature-controlled room (22 ± 2 °C). All experimenters were blinded to group assignments throughout testing, and all apparatuses were thoroughly cleaned with 75% ethanol between animals to eliminate residual olfactory cues.
2.5.1Sucrose preference test (SPT)
The SPT was used to assess anhedonia, a core depressive-like phenotype. Rats were first habituated to two pre-weighed water bottles for 48 h, with bottle positions exchanged every 24 h to prevent position bias. Following 16 h of food and water deprivation, each rat was given simultaneous free access to two pre-weighed bottles: one containing 1% (w/v) sucrose solution and the other containing distilled water, for a 1-h test period. Sucrose preference was calculated as: (sucrose intake / (sucrose intake + water intake)) × 100%.
2.5.2Open-field test (OFT)
The OFT was used to assess locomotor activity and anxiety-like behavior. We gently placed each rat in the center of a black square arena (100 × 100 × 40 cm) and allowed it to explore freely for 5 min. We recorded the total distance traveled and the time spent in the central zone (50 × 50 cm, defined as the area more than 25 cm away from any wall) using an automated video tracking system (EthoVision XT 15, Noldus Information Technology, Wageningen, Netherlands).
2.5.3Forced swim test (FST)
The FST was used to measure behavioral despair, another key depressive-like phenotype. Rats underwent a 15-min pre-swim session 24 h prior to the formal test to induce behavioral despair. For the test session, we placed each rat individually in a transparent cylindrical tank (50 cm height, 20 cm diameter) filled with water (24 ± 1 °C) to a depth of 30 cm, such that the rat could neither touch the tank bottom nor escape. The total test duration was 6 min, with the first 2 min serving as an acclimation period. We recorded the cumulative immobility time during the final 4 min, defined as floating passively in the water with only minimal movements necessary to keep the head above the surface.
2.6Molecular biological assays in the hippocampus
2.6.1Enzyme-linked immunosorbent assay (ELISA)
Hippocampal concentrations of cyclic adenosine monophosphate (cAMP, Cat. No. E-EL-0056) were quantified using commercial ELISA kits from Elabscience Biotechnology Co., Ltd. (Wuhan, China). Levels of 5-hydroxytryptamine (5-HT, Cat. No. MM-0442R1), dopamine (DA, Cat. No. MM-0355R1), and norepinephrine (NE, Cat. No. MM-0556R1) were measured with kits from Jiangsu Meimian Industrial Co., Ltd. (Jiangsu, China). All assays were performed following the manufacturers’ protocols without modification. All kits were validated for rat tissue samples, with reported intra-assay coefficients of variation (CV) < 10% and inter-assay CV < 15%. Absorbance was read at 450 nm on a Tecan Infinite M200 microplate reader (Tecan Group Ltd., Männedorf, Switzerland), and analyte concentrations were calculated from standard curves generated in parallel for each assay plate.
2.6.2Quantitative real-time polymerase chain reaction (qRT-PCR)
Total RNA was isolated from frozen hippocampal tissues using Trizon Reagent combined with an RNA extraction kit (CWBIO, Beijing, China) per the manufacturer’s instructions. RNA purity and concentration were determined spectrophotometrically with a NanoPhotometer (Implen, Munich, Germany); samples with A260/A280 ratios between 1.8 and 2.2 were deemed acceptable for downstream analysis. Complementary DNA (cDNA) was synthesized from 1 μg of total RNA using the HiScript II Q RT SuperMix for qPCR (Vazyme, Nanjing, China). Quantitative PCR was performed in triplicate using SuperStar Universal SYBR Master Mix (CWBIO, Beijing, China) on a CFX Connect™ Real-Time PCR Detection System (Bio-Rad, Hercules, CA, USA). All primer pairs were designed to span exon-exon junctions to prevent genomic DNA amplification, and amplification efficiencies (90–110%) were experimentally validated for each target. Actb (encoding β-actin) served as the endogenous reference gene, and its stable expression across all experimental groups was confirmed (one-way ANOVA, p > 0.05). Relative mRNA expression levels of Bdnf and Ntrk2 (encoding TrkB) were calculated using the 2 < sup > −ΔΔCt</sup > method. Primer sequences are listed below: Bdnf (forward: 5’-CCTGGCAGGCTTTGATGAGA-3′, reverse: 5’-ACCTGGTGGAACTCAGGGT-3′); Ntrk2 (forward: 5’-GGCCGTGAAGACGCTGA-3′, reverse: 5’-ATTTGCTGAGCGATGTGCAG-3′); and Actb (forward: 5’-GCCATGTACGTAGCCATCCA-3′, reverse: 5’-GAACCGCTCATTGCCGATAG-3′).
2.716S rRNA gene sequencing analysis
2.7.1DNA extraction and PCR amplification
Total bacterial genomic DNA was extracted from frozen colonic fecal pellets using the QIAamp Fast DNA Stool Mini Kit (Qiagen, Hilden, Germany) following the manufacturer’s standard protocol. DNA quality and concentration were assessed using a Qubit 2.0 Fluorometer (Thermo Fisher Scientific, Waltham, MA, USA) and 1% agarose gel electrophoresis. The V3–V4 hypervariable regions of the 16S rRNA gene were amplified using barcoded universal primers (341F, 5’-CCTAYGGGRBGCASCAG-3′, 806R, 5’-GGACTACNNGGGTATCTAAT-3′). PCR reactions were carried out with Phusion High-Fidelity PCR Master Mix (New England Biolabs, Ipswich, MA, USA) under the following cycling conditions: initial denaturation at 95 °C for 3 min, followed by 30 cycles of 95 °C for 30 s, 55 °C for 30 s, and 72 °C for 45 s, with a final extension at 72 °C for 5 min.
2.7.2Library construction and sequencing
PCR amplicons were purified using AMPure XP magnetic beads (Beckman Coulter, Brea, CA, USA) to remove primer dimers and non-specific products. Sequencing libraries were prepared with the NEBNext Ultra II DNA Library Prep Kit (New England Biolabs, Ipswich, MA, USA) according to the manufacturer’s recommendations. Paired-end sequencing (2 × 250 bp) was performed on an Illumina NovaSeq 6,000 platform (Illumina Inc., San Diego, CA, USA), generating an average of 50,000 high-quality reads per sample.
2.7.3Sequencing data preprocessing and ASV analysis
Raw paired-end sequencing reads were demultiplexed and processed using QIIME 2 (version 2023.2) (Bolyen et al., 2019). The Deblur algorithm was employed for sequence denoising, quality filtering, and chimera removal, producing amplicon sequence variants (ASVs) at single-nucleotide resolution (Amir et al., 2017). Deblur was chosen over alternative denoising methods due to its superior computational efficiency, significant parallel processing capabilities, and consistent sequence identification across independent datasets. All samples met stringent quality control criteria: Q30 quality score > 90% (corresponding to a sequencing error rate < 0.1%), effective sequence retention ratio > 70%, and Goods_coverage index > 0.997, confirming sufficient sequencing depth to capture the full microbial community diversity present in the samples.
2.7.4Taxonomic annotation and diversity analysis
Taxonomic classification of ASVs was performed against the SILVA 16S rRNA gene database (version 138.1) using a naive Bayes classifier trained specifically on the V3–V4 hypervariable region, with a 97% sequence identity confidence threshold. Alpha diversity indices (Observed_ASV, Shannon, Simpson, Chao1, ACE, Goods_coverage, and PD_whole_tree) were calculated within QIIME 2. Between-group differences in alpha diversity were tested using the Wilcoxon rank-sum test with Benjamini-Hochberg false discovery rate (FDR) correction for multiple comparisons. Beta diversity was assessed using Bray–Curtis dissimilarity, weighted UniFrac, and unweighted UniFrac distances. Principal Coordinate Analysis (PCoA) was used to visualize global differences in microbial community structure between experimental groups. Statistical significance of between-group compositional differences was determined using permutational multivariate analysis of variance (PERMANOVA) with 999 permutations for each distance matrix. Unweighted Pair-group Method with Arithmetic Mean (UPGMA) clustering was performed to evaluate hierarchical similarity between samples based on phylogenetic distances.
2.7.5Differential taxa analysis and functional prediction
Linear discriminant analysis Effect Size (LEfSe) was used to identify differentially abundant taxa at all taxonomic levels between groups, with a significance threshold set at LDA score > 3.0. Predictive functional profiling of the gut microbial communities was conducted using Tax4Fun2 (version 1.1.5). Representative ASV sequences were aligned to the Ref99NR database at 97% sequence identity, and functional predictions were generated after normalization by 16S rRNA gene copy number. The average fraction of taxa unused (FTU) was 0.12, indicating high prediction reliability. KEGG Orthology (KO) annotations at Level 3 were compared between groups using the Wilcoxon rank-sum test with Benjamini-Hochberg FDR correction. Pathways with an adjusted q-value < 0.05 were considered significantly altered.
2.9Statistical analysis
All data are presented as mean ± standard deviation (SD) unless otherwise specified. Statistical analyses were performed using R (version 4.5.3, R Foundation for Statistical Computing, Vienna, Austria) and GraphPad Prism (version 9.5.1, GraphPad Software Inc., San Diego, CA, USA). Data normality was assessed using the Shapiro–Wilk test, and homogeneity of variance was evaluated using Levene’s test. Normally distributed data with equal variance were analyzed using one-way analysis of variance (ANOVA), followed by Dunnett’s post-hoc test for pairwise comparisons against the model (MOD) group. For non-normally distributed data or data with unequal variance, the Kruskal–Wallis test was used, followed by Dunn’s post-hoc test for comparisons against the MOD group. Correlation analyses were performed using the Spearman rank correlation test. Outliers were detected using Grubbs’ test and excluded from analysis if p < 0.05. All statistical tests were two-tailed, and a p-value < 0.05 was considered statistically significant.
3Results
3.1mCLMD alleviates CUMS-induced depressive-like behaviors
To evaluate the antidepressant efficacy of mCLMD, a comprehensive battery of behavioral tests was conducted in a chronic unpredictable mild stress (CUMS)-induced rat model of depression. Chronic CUMS exposure induced anhedonia, the core depressive phenotype, as evidenced by a marked reduction in sucrose preference in the Model (MOD) group compared with the Control (CON) group (***p < 0.001). mCLMD treatment dose-dependently reversed this anhedonic deficit. All three mCLMD-treated groups (low-dose [CL], medium-dose [CM], and high-dose [CH]) exhibited significantly higher sucrose preference than the MOD group (all ***p < 0.001, Figure 1A), with efficacy comparable to that of the venlafaxine (VEN) positive control group.
Consistent with the amelioration of anhedonia, mCLMD also attenuated behavioral despair in the forced swim test (FST). CUMS exposure markedly prolonged immobility time, whereas mCLMD treatment reduced this parameter in a dose-dependent manner. Specifically, the CH and CM groups showed significantly shorter immobility times relative to the MOD group (**p < 0.01 and *p < 0.05, respectively), while no statistically significant difference was observed between the CL and MOD groups (Figure 1B).
Locomotor activity and anxiety-like behavior were next assessed using the open-field test (OFT). MOD rats exhibited a significant reduction in total distance traveled (**p < 0.01) and spent significantly less time in the central zone (***p < 0.001) compared with CON rats. mCLMD treatment restored locomotor activity across all tested doses (*p < 0.05 vs. MOD, Figure 1C). With respect to time spent in the central zone, both the CH (**p < 0.01) and CL (*p < 0.05) groups showed significant increases relative to the MOD group (Figure 1D). These behavioral changes were corroborated by representative movement trajectories: MOD rats displayed typical thigmotaxis, remaining predominantly along the arena walls, whereas CON and CH rats frequently entered and extensively explored the central region (Figure 1E). Taken together, these results demonstrate that mCLMD effectively ameliorates CUMS-induced anhedonia, behavioral despair, and anxiety-like behavior in rats.
3.2Network pharmacology analysis uncovers multi-target synaptic and neurotrophic mechanisms of mCLMD
To elucidate the molecular mechanisms underlying the antidepressant effects of mCLMD, an integrated network pharmacology approach was employed. Putative active compounds of mCLMD and their corresponding targets were retrieved from the TCMSP and HERB databases, while depression-associated targets were compiled from multiple well-curated disease databases. Venn diagram analysis identified 346 overlapping genes, which were designated as potential therapeutic targets of mCLMD for depression (Figure 2A).
An herb-compound-target network was then constructed to visualize the multi-component synergistic effects of the formula (Figure 2B). A protein–protein interaction (PPI) network was subsequently generated using the STRING database with a high confidence threshold of 0.90 (Figure 2C). Topological analysis of the PPI network identified a core subnetwork consisting of 22 hub genes, among which BDNF and NTRK2 (encoding TrkB) are well-established key regulators of depression pathogenesis (Figure 2D).
Functional enrichment analysis was performed to explore the systemic mechanisms of mCLMD and guide experimental validation. Gene Ontology (GO) enrichment analysis (Figure 2F) revealed that the target genes were significantly enriched in biological processes and cellular components critical for central nervous system function, including regulation of membrane potential, synaptic transmission, and postsynaptic specialization. These findings strongly support the hypothesis that mCLMD exerts its effects by modulating neuroplasticity.
Importantly, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis (Figure 2E) further revealed a highly coordinated “peripheral-central” dual-targeting profile. Centrally, target genes were significantly enriched in pathways directly related to neurotrophic regulation, including the cAMP signaling pathway, neuroactive ligand-receptor interaction, and serotonergic synapse. Peripherally, a prominent cluster of pathways involved in immunometabolic homeostasis was observed, encompassing systemic inflammatory cascades (TNF, IL-17, Toll-like receptor, and NOD-like receptor signaling pathways) and key metabolic networks (tryptophan metabolism, sphingolipid signaling, and arachidonic acid metabolism).
Taken together, these bioinformatic results suggest that mCLMD exerts its antidepressant effects not solely through direct binding to central nervous system receptors. Instead, it acts via a “dual-drive” mechanism that integrates central synaptic modulation with systemic regulation of the peripheral immunometabolic microenvironment. This mechanistic hypothesis provided the rationale for the comprehensive in vivo experimental design, encompassing 16S rRNA gut microbiome sequencing, untargeted serum metabolomics, and molecular validation of the hippocampal cAMP/BDNF/TrkB signaling pathway.
3.3mCLMD restores hippocampal monoamine neurotransmitters and upregulates cAMP/BDNF/TrkB pathway transcription
Building on the network pharmacology findings that identified synaptic transmission and the cAMP signaling pathway as core therapeutic modules, the effects of mCLMD on hippocampal neurochemistry were next examined. ELISA assays revealed that CUMS exposure caused a marked reduction in hippocampal levels of 5-hydroxytryptamine (5-HT), dopamine (DA), and norepinephrine (NE) (all ***p < 0.001 vs. CON). High-dose mCLMD (CH) treatment completely restored all three monoamine neurotransmitters (5-HT: **p < 0.01; DA and NE: ***p < 0.001 vs. MOD, Figures 3A–C). Medium-dose mCLMD (CM) also significantly increased DA levels (**p < 0.01) and induced moderate increases in 5-HT and NE concentrations (*p < 0.05 vs. MOD).
Key components of the cAMP/BDNF/TrkB signaling axis were then evaluated. CUMS significantly suppressed hippocampal cAMP levels (***p < 0.001 vs. CON). mCLMD treatment reversed this suppression in a dose-dependent manner, with all three doses producing significant elevations in cAMP concentrations (all ***p < 0.001 vs. MOD, Figure 3D). At the transcriptional level, qPCR analysis confirmed that CUMS significantly downregulated mRNA levels of both Bdnf and its receptor Ntrk2 (encoding TrkB). All three doses of mCLMD significantly increased Ntrk2 mRNA expression (all ***p < 0.001 vs. MOD, Figure 3F), while Bdnf mRNA levels were significantly upregulated in both the high-dose and medium-dose groups (***p < 0.001 vs. MOD, Figure 3E).
The magnitude of these restorative changes in the CH group was comparable to that observed in the venlafaxine (VEN) positive control group. Collectively, these data demonstrate that mCLMD normalizes hippocampal monoamine neurotransmitter levels and activates the cAMP/BDNF/TrkB signaling pathway at the transcriptional level.
3.4mCLMD restructures the gut microbial community disrupted by chronic stress
3.4.1mCLMD reverses CUMS-induced alterations in gut microbial diversity
Alpha diversity indices were compared between groups using the Wilcoxon rank-sum test. The Chao1 index, a measure of microbial richness, was significantly higher in the MOD group than in the CON group (*p < 0.05; Figure 4A)—a finding consistent with previous clinical and preclinical studies showing increased gut microbial richness in both patients with major depressive disorder (MDD) and chronic stress-induced depression animal models (Duan et al., 2021). High-dose mCLMD completely reversed this increase, restoring microbial richness to levels indistinguishable from the CON group (**p < 0.01 vs. MOD). The Shannon diversity index, which reflects both richness and evenness, showed no significant difference between the CON and MOD groups, but was significantly lower in the CH group than in the MOD group (*p < 0.05; Figure 4B), indicating that mCLMD reshaped the community structure by selectively reducing the abundance of potentially pathogenic taxa.
A Venn diagram showed the distribution of shared and unique amplicon sequence variants (ASVs) across the three groups: 1086 ASVs were common to all groups, while 1,604, 1,606, and 1,125 unique ASVs were identified in the CON, MOD, and CH groups, respectively (Figure 4C). Beta diversity analysis using Principal Coordinate Analysis (PCoA) based on Bray–Curtis distances revealed distinct separation of microbial community structures across the three groups (PERMANOVA, **p < 0.01), with PCoA1 and PCoA2 explaining 14.7 and 11.78% of the total variance, respectively (Figure 4D). The CH group clustered significantly closer to the CON group than to the MOD group, confirming that mCLMD partially restored the overall microbial community structure disrupted by chronic stress.
3.4.2mCLMD normalizes depression-associated microbial Dysbiosis and restores predictive metabolic functions
At the phylum level, Bacillota and Bacteroidota were the dominant taxa across all groups, consistent with the normal gut microbial composition of rats. At the genus level, CUMS exposure caused pronounced shifts in the abundance of multiple taxa closely associated with depression pathogenesis, and these changes were largely reversed by mCLMD treatment (Figure 4E).
Linear discriminant analysis Effect Size (LEfSe) with an LDA score threshold of 3.0 was used to identify key differential taxa driving community separation. In the CON vs. MOD comparison (Figure 4F), the MOD group was characterized by significant enrichment of pro-inflammatory genera associated with depression, including Colidextribacter (LDA = 3.93), Oscillibacter (LDA = 3.88), and Desulfovibrio (LDA = 3.60). In contrast, Blautia (LDA = 4.46)—a genus with well-documented anti-inflammatory and antidepressant-like effects in preclinical models—was significantly depleted in the MOD group. In the MOD vs. CH comparison (Figure 4G), mCLMD treatment led to a sharp decline in these pro-inflammatory genera, while significantly enriching beneficial taxa such as Lactobacillus (LDA = 4.75) and Romboutsia (LDA = 4.51).
The relative abundances of these key taxa were further quantified to confirm the LEfSe findings (Figure 4H). Colidextribacter, Oscillibacter, and Desulfovibrio were all significantly enriched in the MOD group and reduced to levels comparable to the CON group by mCLMD treatment (all *p < 0.05). Among beneficial genera, Lactobacillus and Romboutsia were significantly increased in the CH group compared with the MOD group (both *p < 0.05), while Dubosiella—a taxon markedly depleted by chronic stress—showed a clear trend toward recovery after mCLMD treatment. Notably, the restorative effects of mCLMD on gut microbial dysbiosis were more pronounced than those of venlafaxine, which only exhibited mild trends toward normalization of gut microbial profiles.
To explore the potential functional consequences of these mCLMD-induced microbial shifts, predictive functional profiling was performed using Tax4Fun2. Comparative analysis of KEGG Level 3 pathways between the MOD and CH groups revealed significant inferred functional reprogramming of the gut microbiome (Figure 4I). Specifically, the CH group showed significant depletion of stress-associated pathways enriched in the MOD group, including pro-inflammatory pathways such as bacterial chemotaxis (q = 0.018) and flagellar assembly (q = 0.016), as well as phenylalanine, tyrosine, and tryptophan biosynthesis (q = 0.018). Conversely, mCLMD treatment resulted in a predicted enrichment of functional modules critical for neuroprotection and metabolic homeostasis that were suppressed by CUMS, including the cAMP signaling pathway (q = 0.041), GABAergic synapse (q = 0.014), fatty acid biosynthesis (q = 0.044), and glycerolipid metabolism (q = 0.018). It should be emphasized that these are computational inferences derived from 16S rRNA gene profiles and do not represent direct metagenomic or transcriptomic evidence.
These predictive functional changes are highly consistent with both the serum metabolomics findings and the targeted molecular validation of hippocampal neurotrophic signaling, further supporting the hypothesis that mCLMD exerts its antidepressant effects, at least in part, through modulation of the microbiota-lipid-brain axis.
3.5mCLMD reprograms the serum metabolome and normalizes CUMS-induced metabolic disturbances
3.5.1mCLMD reverses global metabolic dysregulation induced by chronic stress
Orthogonal partial least squares discriminant analysis (OPLS-DA) revealed distinct separation of serum metabolic profiles across the CON, MOD, and CH groups (Figure 5A), indicating that CUMS exposure induced profound systemic metabolic disturbances that were largely reversed by high-dose mCLMD treatment. Model validity was verified via 200-fold permutation tests, showing no signs of overfitting (CON vs. MOD: Q2 = 0.58; MOD vs. CH: Q2 = 0.542).
Volcano plot analysis identified 238 significantly differential metabolites in the CON vs. MOD comparison and 238 differential metabolites in the MOD vs. CH comparison (Figure 5B). Among the metabolites significantly altered by CUMS, 78.3% were normalized by high-dose mCLMD (defined as metabolites showing a significant difference between the MOD and CON groups, but no significant difference between the CH and CON groups post-treatment). Hierarchical clustering analysis of the top 50 biologically relevant differential metabolites (variable importance in projection [VIP] > 1, adjusted p < 0.05) confirmed that the stress-induced metabolic dysregulation pattern was effectively reversed in the CH group (Figure 5C). These differential metabolites were predominantly involved in three core biological processes: lipid homeostasis, amino acid metabolism, and bile acid biosynthesis—all closely linked to the microbiota-gut-brain axis.
3.6Integrated multi-omics analysis reveals key microbiota-metabolite associations correlated with mCLMD’s antidepressant-like effects
3.6.1Global omics concordance
To define the functional link between mCLMD-induced gut microbial remodeling and systemic metabolic normalization, integrated multi-omics correlation analyses were performed. Procrustes analysis using Bray–Curtis distances identified a highly significant global concordance between gut microbial community structure and serum metabolome profiles (M2 = 0.65, p = 0.003; Figure 6A). This strong bidirectional coupling confirms that mCLMD-induced systemic metabolic reprogramming is fundamentally driven by changes in the gut microbiome.
4Discussion
Traditional Chinese medicine (TCM) formulations exert their therapeutic effects through synergistic interactions between multiple components acting on diverse biological targets. Here, network pharmacology predicted that modified Chaihu Longgu Muli Decoction (mCLMD) produces antidepressant-like effects via both direct central receptor interactions and a systemic “peripheral-to-central” regulatory mechanism. A key challenge is identifying the specific biological pathways linking peripheral immunometabolic changes to central neuroplasticity. Our integrated multi-omics (Zhao et al., 2024) and targeted hippocampal analyses address this critical gap by suggesting that the microbiota-lipid-brain axis may play a crucial role as a potential mediator of mCLMD’s therapeutic action—a concept supported by recent pharmacological reviews of TCM for depression (Li et al., 2023).
Consistent with our a priori hypothesis, 16S rRNA gene sequencing identified the gut microbiota as a key peripheral target of mCLMD. At the taxonomic level, chronic unpredictable mild stress (CUMS) exposure significantly enriched pro-inflammatory, depression-associated genera including Colidextribacter and Oscillibacter. Both genera are consistently elevated across depressive-like animal models and clinical major depressive disorder (MDD) cohorts (Zhao et al., 2022; Gao et al., 2023), where they compromise intestinal barrier integrity and propagate depressive-like phenotypes through immune signaling and aberrant vagus nerve activation (Liu et al., 2023). High-dose mCLMD treatment significantly reduced the abundance of these pathobionts. Conversely, mCLMD significantly enriched beneficial genera such as Romboutsia, Lactobacillus, and Blautia. Lactobacillus species are well-characterized psychobiotics that alleviate depressive-like behaviors by reinforcing intestinal tight junctions and attenuating neuroinflammation (Dziedzic et al., 2024). These shifts indicate that mCLMD promotes an anti-inflammatory gut microbial environment, which may mediate its downstream systemic effects.
The gut microbiota influences host physiology primarily by regulating metabolism and generating bioactive mediators. Our UPLC-MS/MS-based untargeted metabolomics revealed that mCLMD largely reverses CUMS-induced systemic metabolic dysregulation, particularly within lipid and neurosteroid networks. One key observation was the normalization of pregnenolone, a core neurosteroid whose abnormal accumulation reflects severe hypothalamic–pituitary–adrenal (HPA) axis dysregulation and systemic stress (Cavaleri et al., 2026). By correcting pregnenolone levels, mCLMD alleviates peripheral neurosteroid toxicity (Cao et al., 2025; Cavaleri et al., 2026). mCLMD also markedly reversed the stress-induced depletion of 1-stearoyl-2-arachidonoylglycerol (SAG), a critical diacylglycerol and direct precursor to 2-arachidonoylglycerol (2-AG) (Zhang et al., 2026). 2-AG is the most abundant endogenous ligand of the endocannabinoid system, which plays a well-established role in buffering stress, resolving neuroinflammation, and supporting synaptic plasticity (Zhang et al., 2026). The restoration of peripheral SAG suggests a potential repair of the endocannabinoid system. However, it is important to note that this remains speculative based on circulating metabolites. Future studies should quantitatively target 2-AG and cannabinoid receptor type 1 (CB1) in brain tissues to definitively confirm this peripheral-to-central lipid signaling axis. Additionally, restoration of short-chain acylcarnitine C3:0 indicates improved mitochondrial fatty acid β-oxidation, further reducing peripheral metabolic burden (Hamilton et al., 2020; Liu et al., 2026).
Integrated correlation analysis further linked these taxonomic changes to the observed metabolic phenotypes. The pathobiont Colidextribacter is strongly correlated with pregnenolone, suggesting that its overgrowth is strongly correlated with stress-induced neurosteroid dysregulation. Conversely, Romboutsia, a beneficial genus enriched by mCLMD, correlated strongly positively with SAG, indicating its key role in restoring the systemic endocannabinoid precursor pool. Functional pathway analysis also highlighted Lactobacillus as a core driver of restored primary and secondary bile acid biosynthesis, which is essential for lipid absorption and gut-liver-brain signaling (MahmoudianDehkordi et al., 2022).
These coordinated changes in circulating metabolites systemically support central neuroplasticity. Peripheral lipid mediators, including endocannabinoid precursors and bile acids, can cross the blood–brain barrier (BBB) or signal through vagal afferents to directly modulate CNS function. Restored endocannabinoid tone and resolved neurosteroid toxicity are known to reduce hippocampal neuroinflammation, thereby relieving suppression of the cAMP-BDNF–TrkB neurotrophic pathway. Recovered BDNF signaling in turn enhances neuronal survival and synaptic plasticity, which likely underlies the observed normalization of hippocampal monoamine neurotransmitter levels (5-HT, DA, NE). The detection of intact bioactive constituents, such as spinoside A and tamarixetin, in the systemic circulation highlights additional direct pharmacological effects of mCLMD. While we did not directly validate the brain distribution of these compounds, their documented ability to penetrate the BBB suggests they may exert direct neuroprotective effects (Dajas et al., 2015; Qiao et al., 2016), acting synergistically with microbiota-dependent indirect pathways to stabilize the hippocampal microenvironment (Zhao et al., 2023).
Thus, while our study provides a systemic framework for mCLMD’s antidepressant effects, several limitations warrant consideration. First, our multi-omics approach highlights systemic associations rather than strict causality. Future mechanistic validation using fecal microbiota transplantation (FMT) or antibiotic-mediated depletion is essential. Second, due to the limited volume of dissected hippocampal tissue—which was prioritized for ELISA and qRT-PCR assays—we could not perform protein-level validations (e.g., Western blot for BDNF, p-TrkB, and p-CREB). Since mRNA changes do not necessarily reflect functional protein activation, our conclusions regarding neuroplasticity signaling are preliminary and require future protein-level confirmation. Third, the sample size (n = 6 per group) was determined based on previous exploratory multi-omics studies in CUMS rodent models, which have consistently reported significant behavioral and omics-level differences with comparable group sizes (Duan et al., 2021; Lv et al., 2021). While standard for this context, it remains relatively underpowered for high-dimensional omics, carrying a theoretical risk of overfitting. To mitigate this, we employed strict statistical penalization (e.g., FDR corrections, permutation testing). Additionally, omics profiling was restricted to the high-dose group to establish maximal molecular boundaries; subsequent studies should include dose–response multi-omics assessments in larger cohorts. Fourth, the microbial functional pathways discussed were computationally inferred (Tax4Fun2) from 16S rRNA profiles. Shotgun metagenomics is required to confirm these predictive functional estimations. Finally, reduced center time in the open-field test was interpreted as anxiety-like behavior; however, concurrent decreases in total locomotion may confound this interpretation by reflecting general motor fatigue. Future studies using specific anxiety paradigms (e.g., elevated plus maze) are needed to disentangle true anxiety from widespread locomotor deficits.
Data availability statement
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary material.
Ethics statement
The animal study was approved by Institutional Animal Care and Use Committee of Jiangxi Zhonghong Boyuan Biological Technology Co., Ltd. The study was conducted in accordance with the local legislation and institutional requirements.
Conflict of interest
The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Generative AI statement
The author(s) declared that Generative AI was used in the creation of this manuscript. During the preparation of this work, the authors used Gemini (Model: Gemini 1.5 Pro, developed by Google) in order to improve academic language clarity, refine the stylistic expression, and ensure consistency in the formatting of the reference list. After using this tool, the authors have thoroughly reviewed and edited the generated content as needed and take full responsibility for the integrity and accuracy of the published article.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
Supplementary material
The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnmol.2026.1865663/full#supplementary-material