Liubao tea (LPT) was purchased from China Tea Co., Ltd. (Wuzhou, Guangxi, China; Batch No.: S002/2022) and complied with the specifications of GB/T 32719.4-2016 (Brick-type Liubao Tea, Grade 2). Brequinar was acquired from MedChemExpress (MCE, Monmouth Junction, NJ, USA; Cat. No.: HY-108325).

The animal experiments were approved by the Animal Ethics Committee of Liuzhou Workers’ Hospital (Approval No.: KY202499). In the present study, the destabilization of the DMM technique was used to establish an OA model. To evaluate the therapeutic effects of Liubao tea, male C57BL/6 mice (6–8 weeks old) were randomly divided into four groups (n = 6 per group): Sham-operated control group, DMM group, DMM + low-dose Liubao Tea Extract (LPTE) group (100 mg/kg/day via intragastric gavage), and DMM + high-dose LPTE group (300 mg/kg/day via intragastric gavage). Samples were collected for analysis after 4 weeks of intervention.

For the DMM group: Mice were anesthetized, and hair over the surgical site was shaved off. The mice were placed in a supine position, and their left hind limbs were secured to maintain the knee joints flexed at 90°, followed by thorough disinfection of the surgical area. The skin was incised to expose the patellar ligament; a sharp scalpel was then used to dissect the joint capsule along the medial edge of the patellar ligament. The medial meniscus attaches to the tibial plateau via the medial meniscotibial collateral ligament (MMTL). The MMTL was transected, and the surgical wound was irrigated with sterile saline. The joint capsule was sutured with 7-0 surgical sutures, followed by skin closure. Topical amoxicillin ointment was applied to prevent wound infection.

For the Sham-operated control group, only the joint capsule was incised and then sutured in layers, with no additional interventions (i.e., no MMTL transection).

To investigate the role of pyrimidine synthesis inhibition in OA, a separate cohort of DMM-induced OA mice was treated with Brequinar (a specific de novo pyrimidine synthesis inhibitor). These mice were further subdivided into two groups: DMM + Vehicle group and DMM + Brequinar group (20 mg/kg Brequinar administered intragastrically [i.g.]).

Whole knee joints of mice were excised, and redundant tissues were carefully dissected away prior to fixation of the samples in paraformaldehyde. Subsequently, a NEMO-NMC200 micro-CT system was used to scan the samples at 50 kV and 800 μA. A 1.8 mm-thick region of the medullary cavity was selected, and 125 consecutive sections of the femoral epiphyseal plate were used for three-dimensional (3D) reconstruction. N-Recon software was employed for 3D image reconstruction, while Avatar software was used to quantify trabecular bone microstructural parameters, including trabecular thickness (Tb.Th), trabecular separation (Tb.Sp), trabecular number (Tb.N), and bone volume/tissue volume (BV/TV) ratio.

Mouse knee joints were first fixed in 4% neutral buffered paraformaldehyde (PFA; Servicebio, Wuhan, China) at room temperature for 48 h, followed by decalcification in 12% ethylenediaminetetraacetic acid (EDTA, pH 7.4; Servicebio, Wuhan, China) at 4°C for 1 month (with the EDTA solution refreshed every 3 days to ensure consistent decalcification efficiency).

Following decalcification, samples were dehydrated through a graded ethanol series (70%, 80%, 90%, 95%, and 100% ethanol, 1 h each), cleared in xylene (twice, 30 min each), and then embedded in paraffin wax. A microtome (Leica, Wetzlar, Germany) was used to cut serial 4-μm-thick sagittal sections of the knee joint (centered on the medial femoral condyle and tibial plateau), which were then mounted on poly-L-lysine-coated glass slides. Hematoxylin and eosin (H&E; Servicebio, Wuhan, China) staining was used to measure cartilage surface thickness, while safranin O-fast green staining (Servicebio, Wuhan, China) was employed to evaluate cartilage matrix integrity (reflecting glycosaminoglycan content) following standard protocols. A digital slide scanning system (Pannoramic MIDI, 3DHISTECH, Budapest, Hungary) was used to acquire histological images at 20× magnification.

Modified Mankin Score: This score evaluates cartilage degeneration based on four parameters, with a total score ranging from 0 (normal cartilage) to 14 (severe degeneration): Cartilage structural integrity (0–4 points): 0 = intact cartilage surface and structure; 1 = superficial fibrillation without loss of cartilage thickness; 2 = fissures extending to the middle zone of cartilage; 3 = fissures reaching the deep zone of cartilage; 4 = complete loss of cartilage (exposing subchondral bone). Chondrocyte morphology (0–3 points): 0 = normal chondrocyte distribution (columnar arrangement in the deep zone); 1 = mild chondrocyte clustering (≤3 cells per cluster); 2 = moderate clustering (4–6 cells per cluster); 3 = severe clustering (>6 cells per cluster) or extensive chondrocyte loss. Safranin O staining intensity (0–4 points): 0 = intense, uniform staining (abundant glycosaminoglycans); 1 = mild reduction in staining intensity; 2 = moderate reduction (focal loss of staining); 3 = severe reduction (diffuse loss of staining); 4 = no detectable staining (complete glycosaminoglycan depletion). Tidemark integrity (0–3 points): 0 = intact, continuous tidemark; 1 = mild irregularity of the tidemark; 2 = partial disruption (focal breaks) of the tidemark; 3 = complete disruption (diffuse breaks) or disappearance of the tidemark. OARSI Score: This score quantifies the extent (Grade, G) and depth (Stage, S) of cartilage lesions: Grade (G, 0–4): Defined by the percentage of the cartilage surface involved: 0 = no visible damage; 1 = focal superficial lesions (<10% of the surface); 2 = multifocal superficial lesions (10%–50% of the surface); 3 = lesions involving ≥50% of the cartilage surface; 4 = lesions involving the entire cartilage surface. Stage (S, 0–3): Defined by the depth of cartilage damage: 0 = no damage; 1 = damage limited to the superficial zone (≤1/3 of cartilage thickness); 2 = damage extending to the middle zone (>1/3 to ≤2/3 of cartilage thickness); 3 = damage reaching the deep zone (>2/3 of cartilage thickness) or subchondral bone. The intraclass correlation coefficient (ICC) was used to calculate inter-observer agreement, with an ICC > 0.8 considered indicative of good consistency.

The serum levels of IL-6, IL-1β, and TNF-α were quantified using commercially available enzyme-linked immunosorbent assay (ELISA) kits, in accordance with the manufacturers’ instructions. The kits were purchased from Jianglai Biotechnology Co., Ltd. (China), with the following catalog numbers: IL-6 (Cat. No. JL20268), IL-1β (Cat. No. JL18442), and TNF-α (Cat. No. JL10484).

PCR amplification and sequencing were performed according to the standard protocol provided by Shanghai OE Biotech Co., Ltd. Genomic DNA was used as the PCR template, with BARCODE-specific primers and Agencourt AMPure XP beads employed for the reaction. Primers targeting the V3-V4 region of the 16S rRNA gene were selected to ensure high amplification efficiency and accuracy, with the following sequences: forward primer (343F): TACGGRAGGCAGCAG; reverse primer (798R): AGGGTATCTAATCCT. PCR products of equal volume were pooled based on their concentrations and sequenced on the Illumina MiSeq PE300 platform. The resulting data were subsequently analyzed using the OE Biotech Cloud Platform (https://cloud.oebiotech.com/).

Fresh fecal samples were collected from DMM-induced mice, either treated with Liubao tea or left untreated. Samples from each group were pooled to a total weight of 1 g and suspended in sterile PBS at a concentration of 0.125 g/mL. Recipient mice received daily intragastric administration of an antibiotic cocktail—consisting of vancomycin (50 mg/kg), neomycin sulfate (100 mg/kg), metronidazole (100 mg/kg), and ampicillin (100 mg/kg)—for 5 days prior to fecal microbiota transplantation (FMT). Following the antibiotic regimen, the mice were randomly assigned to two groups: the FMT-DMM group, which received fecal suspension from DMM-induced mice, and the FMT-LPTE group, which received fecal suspension from DMM-induced mice treated with Liubao tea. Each group was then administered 200 μL of the corresponding fecal suspension via intragastric gavage three times per week.

Mouse serum samples (100 μL) were mixed with 400 μL of precooled methanol (-20°C) for protein precipitation. The mixture was vortexed vigorously for 3 min and then centrifuged at 14,000 × g for 15 min at 4°C. The resulting supernatant was transferred to a new 1.5 mL microcentrifuge tube, dried under a nitrogen gas stream at room temperature, and reconstituted in 100 μL of methanol. After vortexing and centrifugation (14,000 × g, 10 min, 4°C), 10 μL of the supernatant was injected into an Ultra Performance Liquid Chromatography-Quadrupole Time-of-Flight Mass Spectrometry (UPLC-Q-TOF MS) system for analysis. Chromatographic separation was achieved using a Waters ACQUITY UPLC BEH C18 column (2.1 × 50 mm, 1.7 μm) maintained at 40°C. The mobile phase consisted of 0.1% formic acid in water (solvent A) and acetonitrile (solvent B), with the following gradient program: 0–1.5 min: 95% A/5% B; 1.5–4.5 min: 75% A/25% B; 4.5–7 min: 70% A/30% B; 7–11 min: 45% A/55% B; 11–12 min: 15% A/85% B; 12–13.5 min: 5% A/95% B; 13.5–14 min: 95% A/5% B. The flow rate was 0.3 mL/min, and the injection volume was 1 μL. Mass spectrometry was performed with an electrospray ionization (ESI) source operating in both positive and negative ion modes. Key MS parameters were configured as follows: mass range, m/z 100–1000; spray voltage, 3.5 kV (positive) and 3.2 kV (negative); sheath gas flow rate, 40 arbitrary units (arb); auxiliary gas flow rate, 5 arbitrary units (arb); ion transfer tube temperature, 320°C; auxiliary gas heater temperature, 350°C; collision energy, 20–40 eV (for MS/MS). Raw data were processed using Progenesis QI software (Waters Corporation) for peak alignment, denoising, and normalization. Metabolites were identified by matching accurate mass, retention time, and MS/MS fragmentation patterns with the Human Metabolome Database (HMDB), MetLin, and an in-house standard library. Multivariate statistical analyses, including principal component analysis (PCA) and orthogonal partial least squares-discriminant analysis (OPLS-DA), were performed using SIMCA-P 14.1 software to screen for differential metabolites (variable importance in projection (VIP) > 1 and p < 0.05 by Student’s t-test).

Liubao tea samples were freeze-dried using a Scientz-100F lyophilizer (Scientz Biotechnology Co., Ltd., Ningbo, China) and ground into a fine powder with an MM 400 grinder (Retsch GmbH, Haan, Germany) at 30 Hz for 1.5 min. Approximately 50 mg of sample powder was accurately weighed and mixed with 1200 μL of pre-cooled 70% (v/v) methanolic aqueous solution (pre-cooled to -20°C) containing internal standards (with the volume scaled proportionally for samples < 50 mg). The mixture was vortexed for 30 s every 30 min (total of 6 times), centrifuged at 12,000 rpm (13,800 × g) at 4°C for 3 min, and the supernatant was filtered through a 0.22 μm hydrophilic polyvinylidene fluoride (PVDF) microporous membrane before being stored in amber injection vials for UPLC-MS/MS analysis. UPLC analysis was performed on an ExionLC™ AD system (Sciex LLC, Framingham, MA, USA) equipped with an Agilent ZORBAX SB-C18 column (1.8 µm, 2.1 mm × 100 mm; Agilent Technologies, Santa Clara, CA, USA). The mobile phase consisted of solvent A (0.1% (v/v) formic acid in ultrapure water) and solvent B (0.1% (v/v) formic acid in acetonitrile), with a gradient program as follows: 95% A/5% B (0 min) → 5% A/95% B (9 min, held for 1 min) → 95% A/5% B (10.1 min, held for 2.9 min).

The flow rate was 0.35 mL/min, the column temperature was 40°C, and the injection volume was 2 μL (with the needle washed using 50% methanol between injections). MS/MS detection was conducted on an ESI-QTRAP mass spectrometer (Sciex LLC, Framingham, MA, USA) with the following parameters: source temperature, 500°C; ion spray voltage, 5500 V (positive mode)/-4500 V (negative mode); gas curtain (CUR), gas 1 (GSI), and gas 2 (GSII) set to 25, 50, and 60 psi, respectively; collision-activated dissociation (CAD) at high level. Targeted multiple reaction monitoring (MRM) scans were performed with medium-pressure collision gas (nitrogen), and the declustering potential (DP) and collision energy (CE) were optimized individually for each target metabolite.

To explore the active components, potential targets, and key pathways of Liubao tea in the treatment of OA, a network pharmacology analysis was performed as follows. Briefly, one Liubao tea sample was subjected to metabolomic analysis using a UPLC-MS/MS platform. The Simplified Molecular-Input Line-Entry System (SMILES) strings of the identified metabolites were retrieved from the PubChem database (https://pubchem.ncbi.nlm.nih.gov/). The chemical constituents of Liubao tea were screened in silico for drug-likeness and pharmacokinetic properties using the ADMETlab 2.0 platform, consistent with previous reports (20, 21). Briefly, potential bioactive compounds were identified by applying a stringent multi-criteria filter. They were required to comply with Lipinski’s Rule of Five: molecular weight ≤ 500, hydrogen-bond donors ≤ 5, hydrogen-bond acceptors ≤ 10, logP ≤ 5, and rotatable bonds ≤ 10. In addition, they had to meet key ADMET thresholds, including predicted human oral bioavailability (F20%) ≥ 0.7, low blood-brain barrier permeability (≤ 0.7), low hERG blockade risk (≤ 0.7), low hepatotoxicity potential (≤ 0.7), and a topological polar surface area (TPSA) ≤ 140 Ų. OA-pathogenic genes were retrieved from the DisGeNet database (https://www.disgenet.org/). We searched for all OA-related diseases and their corresponding genes, and genes supported by literature reports were defined as OA-pathogenic genes. The overlapping targets between the active components of Liubao tea and OA were identified using Venn diagram analysis. A “Liubao tea-OA” network was constructed with these intersection targets using Cytoscape 3.9.1.

Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed to annotate the functional roles and potential signaling pathways of the overlapping genes. For GO enrichment analysis, the DAVID database (https://davidbioinformatics.nih.gov/) was utilized with the following parameters: species was restricted to Homo sapiens, gene identifier was set as official gene symbol, and enrichment categories included Biological Process (BP), Cellular Component (CC), and Molecular Function (MF) (GOTERM_BP_DIRECT, GOTERM_CC_DIRECT, GOTERM_MF_DIRECT). Terms with a false discovery rate (FDR) < 0.01 were considered significantly enriched, and the top enriched terms in each category were selected for visualization. KEGG pathway enrichment analysis was conducted using the KEGG database (https://www.genome.jp/kegg/) or DAVID database. The same gene list and species setting as GO analysis were applied. Significantly enriched pathways were filtered with FDR < 0.01, and key pathways were summarized for subsequent analysis. All visualization plots (e.g., bar charts for GO terms, bubble charts for KEGG pathways) were generated using R software (version 4.0.2).

The predicted targets obtained from the intersection were imported into the STRING online database (https://string-db.org), with the species set to Homo sapiens. A protein-protein interaction (PPI) network of OA-related targets was constructed with a confidence score cutoff of > 0.7. Subsequently, the network data were imported into Cytoscape 3.7.1, and these intersecting drug and disease targets were analyzed using the BisoGenet and CytoNCA plugins. Hub targets of the network were identified through topological analysis based on three key topological parameters: Degree, Betweenness Centrality, and Closeness Centrality. The size of the network nodes was set based on the Degree value.

The construction of the C-T-P network was performed as described in previous studies (22, 23). The potential targets of the active compounds of Liubao tea were predicted by using the following web servers: Similarity Ensemble Approach (SEA) Search Server (https://sea.bkslab.org/), CDD Vault HitFinder (https://www.collaborativedrug.com/), and SwissTargetPrediction (http://www.swisstargetprediction.ch/index.php). The union set of the prediction results is defined as the potential targets. Pathological genes for osteoarthritis (OA) were searched and downloaded from the DisGeNet database (https://disgenet.com/). We searched all OA-related diseases and their corresponding genes. Genes with literature reports were defined as pathological genes for OA. The intersection of target genes and OA pathological genes is defined as essential common genes. We retrieved the interactions between target genes and OA pathological genes and extracted protein-protein interactions that were reported in the literature. Combining the previous Liubao tea compound–target prediction results, the compound–target–pathological gene (C-T-P) network was constructed and visualized using Gephi software (https://gephi.org/).

High-resolution crystal structures of the hub protein targets were retrieved from the RCSB Protein Data Bank (RCSB PDB; https://www.rcsb.org/) and selected as molecular docking receptors. PyMOL software was used to process the proteins, including the removal of water molecules and phosphate groups, and the processed structures were saved in PDB format. AutoDock Vina 1.5.6 was employed for molecular docking to investigate protein-ligand interactions. Prior to docking, structural preprocessing of the proteins and small-molecule ligands was performed using AutoDock Tools: hydrogen atoms were added to the proteins and water molecules were removed; for the small-molecule ligands, hydrogenation was conducted and torsion angles were defined. Subsequently, the coordinates of the docking box were determined. The optimal conformations of the molecular simulations were ultimately obtained by comparing the docking scores. Discovery Studio 2019 and PyMOL software were utilized to visualize the 2D interaction diagrams between the test compounds and the key amino acid residues of the target proteins.

Data are presented as the mean ± standard error of the mean (SEM). Differences between two groups were assessed using Student’s t-test, while those among multiple groups were analyzed using one-way analysis of variance (ANOVA). A p-value < 0.05 was considered statistically significant. All statistical analyses were performed using GraphPad Prism 8.0 and SPSS 29.0.

To evaluate the therapeutic potential of Liubao tea against OA, a mouse model was established via destabilization of the DMM surgery (Figure 1A). Following Liubao tea intervention, micro-CT imaging revealed notable improvements in the knee joints of OA mice, including reduced joint space narrowing, smoother articular surfaces, and diminished osteophyte formation (Figure 1B). Quantitative assessment of subchondral bone microstructure indicated that Liubao tea significantly increased BV/TV, Tb.N, and Tb.Th, while decreasing Tb.Sp (Figure 1C). Histopathological analysis of knee joint sections stained with H&E and SOFG was performed to examine cartilage morphology across groups (Figures 2A, B). In sham-operated control mice, the cartilage structure and tidemark remained intact, with chondrocytes arranged uniformly. By contrast, the DMM group exhibited uneven and severely eroded cartilage margins, chondrocyte loss, cartilage layer thinning, and reduced or absent Safranin O staining. Liubao tea treatment mitigated these pathological changes, alleviating cartilage damage, restoring chondrocyte numbers, and improving cartilage architecture. Consistently, both the Modified Mankin scores and the OARSI scores were significantly reduced following Liubao tea administration (Figures 2C, D). Furthermore, serum analysis demonstrated markedly elevated levels of the pro-inflammatory cytokines IL-6, IL-1β, and TNF-α in DMM-induced mice, all of which were significantly suppressed after Liubao tea treatment (Figure 2E). Taken together, these findings indicate that Liubao tea alleviates joint structural deterioration and attenuates inflammatory responses in a mouse model of DMM-induced OA.

Previous studies have indicated that Liubao tea can modulate gut microbiota and ameliorate multiple diseases (15–17). Therefore, this study further investigated whether Liubao tea regulates the gut microbiota in a mouse model of DMM-induced osteoarthritis. The results demonstrated that Liubao tea treatment increased the α-diversity of the gut microbiota in OA mice, as reflected by higher values of the Observed Species, Shannon, and Chao1 indices (Figure 3A). β-diversity analysis using non-metric multidimensional scaling (NMDS) and principal coordinate analysis (PCoA) revealed a distinct separation in microbial community structure between the DMM-induced OA group and the Liubao tea-treated group, indicating substantial shifts in gut microbiota composition (Figures 3B, C). In addition, Linear Discriminant Analysis Effect Size (LEfSe) and phylogenetic tree-based analysis were conducted to identify differentially abundant bacterial taxa among the groups (Figures 4A, B). Collectively, these findings indicate that Liubao tea treatment induces substantial alterations in the gut microbiota of mice with DMM-induced OA.

Previous studies have established that FMT can ameliorate osteoarthritis in mice (24). Building on the observed modulatory effect of Liubao tea on the gut microbiota of DMM-induced OA mice, we further examined whether FMT from Liubao tea-treated donors confers therapeutic benefits in DMM-induced OA mice. Recipient DMM-induced OA mice received fecal microbiota derived from Liubao tea-treated donors (Figure 5A). Micro-CT analysis of knee joints showed that FMT resulted in a wider joint space, a smoother articular surface, and reduced osteophyte formation (Figure 5B). Consistent with these observations, FMT significantly increased BV/TV, Tb.N, and Tb.Th, while significantly reducing Tb.Sp (Figure 5C). Histological evaluation using H&E and SOFG staining revealed that FMT alleviated cartilage damage and improved cartilage architecture (Figures 5D, E). Accordingly, both the Modified Mankin scores and OARSI scores were significantly reduced in the FMT group (Figures 5F, G). Moreover, serum levels of the pro-inflammatory cytokines IL-6, IL-1β, and TNF-α were markedly reduced following FMT (Figure 5H). Collectively, these findings demonstrate that transplantation of fecal microbiota from Liubao tea-treated mice alleviates DMM-induced OA in recipient mice.

To further investigate the effects of Liubao tea on serum metabolites in DMM-induced OA mice, we performed untargeted metabolomic analysis on serum samples. The metabolomic data were subjected to OPLS-DA. Score plots revealed clear separation among the groups, confirming distinct metabolic profiles (Figure 6A). Permutation tests confirmed the validity of the OPLS-DA models, with R² and Q² values showing the expected trends, indicating robust stability and reliable predictability of the models (Figure 6B). Volcano plots were used to identify differential metabolites based on the following criteria: VIP > 1.0, FC > 2 or < 0.5, and p < 0.05. Compared with the DMM-induced OA group, 40 metabolites were upregulated and 34 were downregulated in the Liubao tea-treated DMM-induced OA mice (Figure 6C). KEGG pathway enrichment analysis was visualized using a differential abundance (DA) score plot, which highlighted pyrimidine metabolism as the most significantly downregulated pathway (Figure 6D). Collectively, these results indicate that Liubao tea intervention markedly alters the serum metabolome in DMM-induced OA mice.

Metabolomic findings revealed that Liubao tea suppressed the pyrimidine metabolism pathway. To determine whether Liubao tea alleviates DMM-induced OA by regulating pyrimidine metabolism, we treated DMM-induced OA mice with Brequinar—a specific inhibitor of de novo pyrimidine synthesis (Figure 7A). Micro-CT analysis indicated that Brequinar administration markedly improved knee joint pathology in DMM-induced OA mice (Figure 7B). Quantitative evaluation showed that Brequinar significantly increased BV/TV, Tb.N, and Tb.Th, while significantly reducing Tb.Sp (Figure 7C). Correspondingly, H&E and SOFG staining revealed that Brequinar mitigated inflammatory infiltration and articular cartilage degradation (Figures 7D, E). Both the Modified Mankin scores and OARSI scores were significantly lower in the Brequinar-treated group compared with the DMM-induced OA group (Figures 7F, G). Additionally, Brequinar treatment significantly decreased serum levels of the pro-inflammatory cytokines IL-6, IL-1β, and TNF-α (Figure 7H). Collectively, these data demonstrate that inhibition of pyrimidine synthesis ameliorates DMM-induced OA, implying that the therapeutic effect of Liubao tea on DMM-induced OA may be mediated, at least in part, through suppression of pyrimidine metabolism.

The chemical components of Liubao tea were initially characterized using ultra-performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS), with the resulting positive ion flow chromatograms shown in Figure 8A. Using the UPLC-MS/MS platform and an in-house database, a total of 1989 metabolites were identified. These metabolites were primarily categorized into flavonoids (571 species), phenolic acids (290 species), quinones (30 species), and lignans and coumarins (123 species), among other classes (see Supplementary Table 1). In silico screening of the chemical constituents of Liubao tea for drug-likeness and pharmacokinetic properties was conducted using the ADMETlab 2.0 platform. Potential bioactive compounds were identified through a stringent multi-criteria filter that required compliance with Lipinski’s Rule of Five, yielding a total of 1186 targets corresponding to these bioactive components.

Subsequently, Venn diagram analysis was performed to identify overlapping targets between the bioactive component targets of Liubao tea and OA-related genes, resulting in 324 intersection genes (Figure 8B). Based on the screened bioactive components and their corresponding targets, a component-target network was constructed; compounds with a Degree value > 30 and their matched targets were visualized (Figure 8C). The GO and KEGG analyses of overlapping genes are presented in Figure 9. Furthermore, a PPI network of the 324 intersection targets was constructed using the STRING online database, containing 324 nodes and 810 interaction edges. Thereafter, the drug targets and network targets were imported into Cytoscape 3.7.1 for mapping and screening; three rounds of topological analyses based on Degree, Betweenness Centrality, and Closeness Centrality were performed to construct the hub target protein network (Figure 10A). This analysis ultimately identified TP53, IL6, and TNF as the top 3 core hub targets of Liubao tea for OA treatment (Figure 10B).

Subsequently, we integrated the compound-target network with the PPI network to construct a compound-target-protein (C-T-P) network. The original C-T-P network contained 10,146 compound-target pairs (Figure 11A). Within this network, 324 proteins functioned as both target proteins and pathogenic genes, which were defined as core shared proteins. Subsequently, Cytoscape software was employed for network optimization and screening of bioactive compounds, generating an optimized C-T-P network (Figure 11B). This optimized network comprised 7,823 compound-target pairs, involving 169 compounds, 1,186 predicted targets and 2,060 disease-related genes. All core shared proteins were retained in the optimized network. In the interaction network, compounds including eupatilin, 5,6,7,8-tetramethoxyflavone, 5-hydroxy-6,7,3’,4’,5’-pentamethoxyflavone, demethoxysudachitin and tamaridone exhibited relatively high connectivity degrees.

Finally, molecular docking analysis was performed on the top 3 bioactive components of Liubao tea and their corresponding core hub targets identified from the network pharmacology analysis. Results demonstrated that these chemical components, namely eupatilin, 5,6,7,8-tetramethoxyflavone, and 5-hydroxy-6,7,3’,4’,5’-pentamethoxyflavone, exhibited potential binding capabilities with the key targets TP53, IL6, and TNF (Figures 12A–C). This finding suggests that Liubao tea may exert its therapeutic effects by binding to these core hub targets.