Application of gut microbiota metabolites in the treatment of knee osteoarthritis: a network pharmacology study

Abstract

Background

Knee osteoarthritis (KOA) is a prevalent degenerative joint disease affecting approximately 654 million people worldwide. The gut-joint axis theory suggests a intrinsic link between gut microbiota(GM) metabolites and KOA pathogenesis. This study employs network pharmacology to investigate the protective effects of GM metabolites against KOA and elucidate their underlying molecular mechanisms.

Methods

KOA-related targets and GM metabolite targets were retrieved from public databases. After deduplication, intersecting targets were identified and subjected to protein–protein interaction (PPI) network analysis, Gene Ontology (GO) enrichment, and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis to identify core targets and pathways. Functional association analysis was performed on core targets, followed by GO, KEGG, and functional clustering analyses. Results from both analytical rounds were compared. A “Gut Microbiota-Target-Metabolite” network was constructed to screen key metabolites and targets, which were subsequently validated using molecular docking, drug-like property assessment, and toxicity analysis.

Results

By integrating multi-source target prediction, network analysis, and molecular docking validation, this study first identified IL6, IL1B, and NFKB1 as core targets regulating KOA processes via GM metabolites. GO analysis revealed their functions primarily concentrate on immune response and inflammatory regulation. KEGG analysis highlighted the lipid and atherosclerosis pathway and TNF signaling pathway as key mechanisms. Butyrate, acetate, propionate, and trimethylamine oxide emerged as core metabolites. Molecular docking confirmed strong binding affinities with core targets. All four metabolites exhibited favorable bioavailability, acceptable Lipinski's rule violations, and no hepatotoxicity or carcinogenicity.

Conclusion

This study provides novel network pharmacology evidence supporting the gut-joint axis theory, revealing a potential mechanism whereby GM metabolites may synergistically intervene in KOA through multiple targets and pathways. It also identifies candidate targets and metabolites for gut microbiome-based prevention and treatment strategies for KOA.

Introduction

Knee osteoarthritis (KOA) is a chronic disease involving damage to the knee cartilage, bones, and surrounding tissues, primarily manifesting as joint pain, stiffness, and limited mobility[1]. Epidemiological studies indicate a global prevalence of approximately 16%, with rates influenced by gender, age, geographic region, and body mass index (BMI)[2]. Without active treatment, disability rates can reach up to 53%[3]. Due to its high disability and prevalence rates, KOA poses significant challenges to individual quality of life, healthcare systems, and socioeconomic development[4, 5].

Current clinical treatment primarily focuses on symptom relief, including nonsteroidal anti-inflammatory drugs(NSAIDs), intra-articular injections, and surgical interventions. These strategies can reduce patient pain in the short term, but long-term medication use may lead to gastrointestinal adverse reactions, while surgical treatment has limitations such as invasiveness and high costs[6]. The gut microbiota(GM) represents the largest microbial ecosystem in the human body and is increasingly recognized as a key factor regulating host health[7]. The concept of the “gut-joint axis” was first proposed by Professor Bradtzaeg of the University of Oslo, Norway[8], who posited that gut microbes and their metabolites influence joint health through immune and inflammatory responses[9]. Compared to healthy individuals, patients with osteoarthritis exhibit significant differences in gut microbiome function and composition, these being closely associated with disease progression[10, 11]. Recent studies indicate that short-chain fatty acids(SCFAs), as one type of microbiota metabolite, exert beneficial effects on osteoarthritis[12]. However, the mechanisms underlying the effects of GM metabolites remain incompletely understood. Therefore, elucidating the mechanisms by which the GM and its metabolites influence KOA may contribute to the development of novel and more effective therapeutic strategies.

As an emerging interdisciplinary field, network pharmacology has demonstrated significant value in investigating complex disease mechanisms. This approach explores drug-biological network interactions at the systems level, enabling efficient screening of disease-related active components and identification of their potential targets. In recent years, it has been widely applied to elucidate disease pathogenesis[13]. Therefore, we employed network pharmacology to systematically analyze the mechanisms of GM metabolites in KOA. The entire research workflow is illustrated in Fig. 1. We employed protein–protein interaction (PPI) networks, Gene Ontology (GO) analysis, Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis, GeneMANIA functional association analysis, and functional clustering analysis to identify key targets, key metabolites, and key signaling pathways[14]. Finally, we evaluated the affinity between key metabolites and key targets, and conducted drug-likeness properties and toxicity analysis for the key metabolites.

Fig. 1.

Research Flowchart

Methods and materials

Data sources and tools

All data sources and tools used are listed in Table 1.

Table 1.

Database, software and analysis platform

NO	Database, software and analysis platform	Website	Version
1	gutMGene	http://bio-computing.hrbmu.edu.cn/gutmgene/
2	Pubchem	https://pubchem.ncbi.nlm.nih.gov/
3	Swiss target prediction	https://swisstargetprediction.ch/predict.php
4	Similarity ensemble approach	https://sea.bkslab.org/
5	Bioinformatics	http://www.bioinformatics.com.cn/
6	GeneCards	https://www.genecards.org/
7	OMIM	https://www.omim.org/
8	Therapeutic target database	https://ttd.idrblab.cn/
9	Cytoscape	https://cytoscape.org/	V3.8.0
10	STRING	https://string-db.org/
11	Gene MANIA	https://genemania.org/
12	UniPort	https://www.uniprot.org/
13	RCSB PDB	https://www.rcsb.org/
14	PyMOL	https://pymol.org/	V3.1
15	AutoDock	https://autodock.scripps.edu/	V1.5.7
16	ChemBio3D	https://www.chemdraw.com.cn/ChemBio3D.html
17	SwissADME	https://swissadme.ch/
18	ADMETlab	https://admetmesh.scbdd.com/

Identify the functional targets of GM and GM metabolites

Acquire targets for GM and GM metabolites from the gutMGene database. To systematically identify potential target sites for GM metabolites, we employ a multi-algorithm cross-prediction strategy. Metabolite SMILES obtained from PubChem are used for target prediction via the Similarity Ensemble Approach (SEA) and Swiss Target Prediction (STP). The targets predicted by SEA and STP were intersected, and the resulting intersection targets were identified as the targets for GM metabolites.

Identifying core targets of the GM and its metabolites in the treatment of knee osteoarthritis

Target data for KOA were collected from GeneCards, OMIM, and TTD databases. The data were integrated and duplicates removed. Overlapping targets for this disease were then identified through VENN diagram analysis. The overlapping targets between GM metabolites, GM, and KOA constitute the core targets for KOA regulation.

Protein–protein interaction (PPI) networks construction and core target screening

Submit the overlapping targets among GM metabolites, GM, and KOA to the string platform for analysis. Subsequently, utilize Cytoscape 3.8.0 software to construct the PPI network diagram. Employ algorithms within the Cytohubba plugin to further identify core functional targets within the PPI network.

GO function and KEGG Enrichment analysis

To explore the biological functions of core targets, GO functional and KEGG enrichment analyses were performed using a bioinformatics platform. GO entries (including biological processes (BP), cellular components (CC), and molecular functions (MF)) and KEGG pathways with P-values < 0.05 and false discovery rates (FDR) < 0.05 were considered statistically significant. Results were visualized using bar charts and bubble charts.

GeneMANIA functional association network analysis

To further explore functional associations and potential interactions among core targets, this study employed the GeneMANIA platform for functional association network analysis. The core targets identified in the preliminary screening were submitted as the input gene list, with the species set to Homo sapiens and the default automatic weight selection mode applied. GeneMANIA constructs gene–gene functional interaction networks based on the guilt-by-association algorithm. This approach integrates multi-source functional association data, including protein–protein interactions, co-expression, colocalization, genetic interactions, pathway participation, and shared protein domains.

Molecular docking

We obtained target structure files from UniPort and the RCSB Protein Data Bank(RCSB PDB), which were optimized using PyMOL software. Metabolite structure files were downloaded from the PubChem database and optimized using ChemBio3D software. Using tools such as AutoDock and its Grid plugin, we generated pdpqt format files for target-metabolite pairs and metabolite active site structures. Molecular docking visualization was performed using PyMOL software. Binding energy(BE) was calculated using Vina. The BE below 0 kcal/mol indicates the formation of a stable structure between the target and the component.

The evaluation of drug-likeness properties and toxicity

To evaluate the drug-like properties of core metabolites and their feasibility and safety as potential therapeutic agents, this study employed the SwissADME platform to conduct drug-likeness properties on the screened core metabolites. The SMILES structures of the metabolites were submitted to the platform to calculate physicochemical parameters, primarily including molecular weight, lipophilicity coefficient, hydrogen bond donor count, and hydrogen bond acceptor count. The ADMETlab platform was employed for toxicity prediction analysis. The SMILES structures of the core metabolites were submitted to the platform to predict multidimensional toxicity endpoints.

Results

Identification of GM Metabolites associated with knee osteoarthritis-related targets

Based on the GM metabolites and their corresponding targets identified from the SEA and STP databases, we obtained 696 overlapping targets associated with GM using VENN diagrams (Fig. 2A). After searching GeneCards, OMIM, and TTD databases, integrating results, and removing duplicates, we identified 2378 KOA-related targets (Fig. 2B). The 153 overlapping targets appearing in the VENN diagram represent key targets regulated by GM metabolites in KOA (Fig. 2B). Finally, after performing an intersection analysis between the 153 overlapping targets and GM targets, we identified 33 overlapping targets considered key targets involved in KOA development (Fig. 2C). We also constructed a gut-metabolite-target-disease network to elucidate their complex relationships (Fig. 2D).

Fig. 2.

The potential targets acquisition. A The overlapping targets of gut microbiota metabolites between SEA and STP. B The overlapping targets between disease targets and gut microbiota metabolites. C The overlapping targets between 153 targets and gut microbiota targets. D The Gut-Metabolites-KOA-Targets network

Identification of core targets in PPI networks

To identify core therapeutic targets for KOA treatment, we constructed a PPI network by submitting 33 overlapping targets to the String platform. Visualized using Cytosccape 3.8.0 software, this PPI network comprised 33 nodes and 236 edges (Fig. 3A). We further screened core targets from the PPI network using six algorithms (MCC, MNC, Degree, EPC, Radiality, and Stress) from the Cytohubba plugin in Cytoscape 3.8.0. As shown in Fig. 3C, we selected the top 10 targets from each algorithm and identified overlapping targets using a VENN diagram. The results revealed 6 genes identified as core therapeutic targets for knee osteoarthritis: AKT1, IL6, PPARG, JUN, IL1B, and NFKB1 (Fig. 3B).

Fig. 3.

The identification of core targets. A The PPI network of overlapping targets. B The core targets acquired from 6 algorithms using VENN diagram. C The identification of core targets using 6 algorithms from CytoHubba plug-in in Cytoscape

GO function and KEGG enrichment analysis

To further investigate the biological functions of GM metabolites regulating KOA, we submitted the 33 core targets identified to the bioinformatics website for GO functional and KEGG pathway enrichment analysis. In the GO enrichment analysis, based on the threshold value of FDR was set at < 0.05, the top 10 most significant entries were selected from each of the three major categories: BP, CC, and MF. As shown in Fig. 4A, within BP, genes were significantly enriched in pathways related to host immune and inflammatory responses to bacterial infection. Simultaneously, multiple entries were closely associated with the regulation of vascular endothelial cell migration, suggesting these genes may participate in regulating vascular barrier function and tissue repair processes during infection or inflammation. Notably, these genes are primarily localized to epigenetic modification complexes within the nucleus (particularly histone deacetylase complexes) and specific signaling microdomains on the cell membrane (membrane rafts and cisternae), aligning closely with their potential roles in transcriptional regulation and signaling responses. We also observed significant MF enrichment of these genes in histone modification enzyme activity and related binding functions. Figures 4B, D, and F reveal targets involved in BP, CC, and MF regulation. Regarding KEGG enrichment analysis, pathways associated with our input gene set include Lipid and atherosclerosis, Yersinia infection, AGE-RAGE signaling pathway in diabetic complications, Measles, Chagas disease, Toll-like receptor signaling pathway, and Non-alcoholic fatty liver disease (Fig. 4C). The target-pathway network is visualized with yellow circular nodes representing targets and blue triangular nodes indicating pathways (Fig. 4E). These findings suggest that GM metabolites exert therapeutic effects by regulating multiple biological functions and pathways.

Fig. 4.

Biological analysis of gut microbiota metabolites against KOA. A The top 10 entries in BP, CC and MF based on P value. B, D and F The genes involved in the regulation of BP, CC and MF. C The top 15 KEGG pathways based on P value. E The target-pathway network

GeneMANIA functional association (GMFA) network analysis of six core targets

GeneMANIA is a flexible and powerful retrieval tool for predicting genes and constructing gene interaction networks. By integrating data from multiple sources, it helps researchers identify other genes associated with the input gene. To further analyze the six core genes identified earlier, we expanded the associated gene sets for each core gene based on seven dimensions (Physical Interactions, Co-expression, Predicted, Co-localization, Genetic Interactions, Pathway, and Shared protein domains). We then selected the top 10 most relevant genes for each (Fig. 5). This process enables us to more precisely identify genes with therapeutic potential.

Fig. 5.

The expanded target dataset of the 8 core targets with GeneMANIA functional association network analysis (GMFAN)

The GO function and KEGG enrichment analysis of GMFA

We integrated the gene set derived from GMFA expansion with the previous 33 core genes and removed duplicates, ultimately obtaining 90 gene targets. To thoroughly analyze the functions of these 90 targets derived from the GMFA dataset, we conducted comprehensive GO and KEGG enrichment analyses. As shown in Fig. 6B, the most significantly enriched (highest enrichment score) entry in the BP category for the gene set is “response to molecule of bacterial origin,” indicating that the target gene set plays a central role in biological processes responding to bacterial molecules (such as lipopolysaccharide). Other enriched entries involve innate immune responses and transcriptional program regulation. The most significantly enriched entry in the CC category is “transcription regulator complex,” which aligns strongly with the BP enrichment for transcription regulation. This suggests that the products of these genes are primarily localized within transcription regulatory complexes to execute their regulatory functions. Other significantly enriched entries imply subcellular localization diversity for this gene set, involving: epigenetic regulatory complexes (e.g., histone deacetylase complexes), signaling platforms (e.g., membrane microdomains), and immune effector sites (e.g., early phagosomes). Within the MF category, enriched entries clearly divide into two major functional modules: the first is binding functions, and the second is enzymatic activity functions. As shown in Fig. 6A, we generated a bubble plot based on the top 15 pathways selected by P-value screening. Pathways associated with the input gene set include TNF signaling pathway, Lipid and atherosclerosis, Non-alcoholic fatty liver disease, Fluid shear stress and atherosclerosis, Human cytomegalovirus infection, Chagas disease, and Toll-like receptor signaling pathway. Additionally, two key pathways—Lipid and atherosclerosis and TNF signaling pathway—emerged from two KEGG enrichment analyses (Fig. 7). Visual comparisons of GO and KEGG enrichment results for the 90-target and 33-target sets are presented (Fig. 6C-F). Overall, the GO and KEGG enrichment analyses for the 90-target set provide a more comprehensive and in-depth understanding of KOA-related GO and KEGG enrichment, revealing the complex molecular mechanisms underlying the action of GM metabolites.

Fig. 6.

GO and KEGG enrichment analysis of the expanded core target dataset. A The top 15 KEGG pathways based on P value. B The top 10 entries in BP, CC and MF based on P value. C The comparison of KEGG enrichment analysis of 33 hub targets and expanded target dataset. D, E and F The comparison of BP, CC and MF of 33 hub targets and expanded target dataset

Fig. 7.

[14]. Two key pathways identified. A Results of the First KEGG Enrichment Analysis: The lipid and atherosclerosis. B Results of the Second KEGG Enrichment Analysis: TNF signaling pathway

Functional cluster analysis

To further investigate the mechanism by which GM metabolites target KOA, we performed a modular network analysis by uploading 90 targets to the STRING database and identifying core clusters using the MCODE algorithm. As shown in Fig. 8A, after removing disconnected nodes, the GMFA PPI network comprised 88 nodes and 843 edges. Cluster 1 comprised 26 nodes and 302 edges with a score of 24.160 (Fig. 8B); Cluster 2 comprised 8 nodes and 13 edges with a score of 3.741 (Fig. 8B); Cluster 3 comprised 6 nodes and 7 edges with a score of 2.800 (Fig. 8B).

Fig. 8.

The cluster analysis of GMFAN. A The PPI network of expanded core target dataset. B The top 3 clusters

We performed GO enrichment analysis on Clusters 1, 2, and 3, as shown in Fig. 9. Genes in Cluster 1 were significantly enriched in biological processes such as responses to external stimuli (e.g., lipopolysaccharides, bacterial-derived molecules), reactive oxygen species responses, and the extrinsic apoptosis signaling pathway. This suggests that its core function is closely related to innate immune responses and cellular stress responses. Cluster 2 specifically enriched biological processes including regulation and production of interferon-γ and regulation of interleukin-10 production. It exhibited localization preferences toward cellular components such as the extracellular membrane and the endocytosis-lysosomal system, indicating this cluster primarily participates in adaptive immune regulation and specific cellular content processing. Cluster 3 exhibits highly concentrated functions in the field of epigenetic regulation, showing significant enrichment in biological processes such as histone deacetylation and histone modification, as well as cellular components including histone deacetylase complexes and transcription regulatory complexes. Its molecular functions explicitly point to NAD-dependent histone/ protein deacetylase activity, suggesting this cluster serves as a key regulatory module for epigenetic reprogramming. Collectively, the three gene clusters correspond to three core biological layers: innate immunity and stress, adaptive immune regulation, and epigenetic control.

Fig. 9.

The cluster analysis of GMFAN. A The GO and KEGG enrichment analysis of Cluster 1. B The GO and KEGG enrichment analysis of Cluster 2. C The GO and KEGG enrichment analysis of Cluster 3

KEGG enrichment analysis results for the three clusters can be categorized into multiple distinct groups (Fig. 9). These classifications encompass Environmental Information Processing, Organismal Systems, Metabolism, Cellular Processes, and disease-related aspects. Cluster 1 showed significant enrichment in pathways related to immune and stress responses, such as the TNF signaling pathway, IL-17 signaling pathway, and Toll-like receptor signaling pathway. It also exhibited high enrichment in pathways closely associated with human diseases, including “lipids and atherosclerosis,” indicating strong correlations with inflammatory responses, innate immunity, and cardiovascular metabolic diseases. Cluster 2 showed enrichment in immune signaling (e.g., cytokine-cytokine receptor interactions) and metabolic pathways (e.g., tryptophan metabolism), but overall enrichment was weaker with generally lower gene counts, suggesting more dispersed functions. Cluster 3 primarily enriched pathways related to cellular processes, notably “neutrophil extracellular trap formation,” and also featured in metabolic regulation pathways such as the “PPAR signaling pathway.”This comprehensive clustering analysis reveals intricate mechanisms linking metabolites, fundamental biological processes, molecular functions, and potentially therapeutic cellular signaling pathways. By unraveling these complex interactions, the study not only deepens our understanding of biological systems but also provides crucial reference points for future research.

Identification and molecular docking of key GM metabolites

To investigate the potential mechanisms of core GM metabolites in KOA treatment, this study constructed a “gut microbiota–target–metabolite” (G-T-M) interaction network to systematically analyze the complex relationships among these three components (Fig. 10). The network comprises 108 gut microbiota species, 8 GM metabolites, and 6 core targets. Red arrows represent targets, orange circles denote metabolites, and green squares indicate GM. Notably, IL6 and IL1B were the targets most frequently associated with metabolites, while Butyrate, Acetate, Propionate, and Trimethylamine oxide were the metabolites most extensively linked to gut microbiota, suggesting their potential key roles in KOA regulation. Using Cytoscape's network analysis tools, key nodes within the network were further identified. Based on degree centrality scores, Butyrate, Acetate, Propionate, and Trimethylamine oxide were identified as core metabolites, while IL6, IL1B, and NFKB1 were identified as core targets. To validate interactions between these core metabolites and targets, molecular docking analysis was subsequently performed. Generally, a Binding Energy < 0 kcal·mol⁻¹ indicates that ligands and receptors can spontaneously form stable conformations[14]. As shown in Table 2, the binding energies of all metabolites and targets were below this threshold, indicating strong binding affinity. Most notably, butyrate and propionate exhibited the highest binding affinities toward the targets. Visual representations of the molecular docking results are presented in Fig. 11.

Fig. 10.

The gut microbiota-targets-metabolites network

Table 2.

The binding affinity between metabolites and targets

Targets	Metabolites	Binding Affinity(kcal/mol)	Targets	Metabolites	Binding affinity(kcal/mol)
IL6	Butyrate	−3.7	IL1B	Butyrate	−3.8
	Acetate	−3.0		Acetate	−3.2
	Propionate	−3.5		Propionate	−3.6
	Trimethylamine oxide	−2.7		Trimethylamine oxide	−2.9
NFKB1	Butyrate	−3.7
	Acetate	−3.1
	Propionate	−3.4
	Trimethylamine oxide	−2.9

Fig. 11.

Molecular docking results. A Molecular docking of Acetate with IL1B, IL6 and NFKB1 respectively. B Molecular docking of Butyrate with IL1B, IL6 and NFKB1 respectively. C Molecular docking of Propionate with IL1B, IL6 and NFKB1 respectivel. D Molecular docking of Trimethylamine oxide with IL1B, IL6 and NFKB1 respectively

The evaluation of drug-likeness properties and toxicity

To evaluate the feasibility and safety of core metabolites as potential therapeutic agents, we conducted their drug-like property assessment and toxicity analysis. Drug-like property evaluation was performed on the ADMETlab website, adhering to the classic Lipinski's five rules while incorporating Veber's rules [14, 15], results are presented in Table 3, showing that all four core metabolites exhibit favorable drug-like characteristics. Toxicity prediction analysis was performed on the ADMETlab platform [13], results are shown in Table 4.The toxicity prediction results for the four core metabolites indicate that none exhibit cardiac toxicity risk, drug-induced liver injury risk, or carcinogenicity. Notably, all four metabolites yielded positive H-HT (Human Hepatotoxicity) predictions, suggesting potential hepatotoxic effects at the hepatocyte level. However, in conjunction with negative DILI results, this hepatocyte-level action may be insufficient to induce clinically significant liver injury events.

Table 3.

The evaluation of drug-likeness properties on key metabolites

Compound	MW	HBA	HBD	MLogP	Lipinski’ sviolations	Bioavailability Score	TPSA
Butyrate	87.1	2	0	0.49	0	0.85	40.13
Acetate	59.04	2	0	-0.49	0	0.85	40.13
Propionate	73.07	2	0	0.03	0	0.85	40.13
Trimethylamine oxide	75.11	1	0	-1.66	0	0.55	29.43

Note[14, 15]: MW, molecular weight < 500; HBA, hydrogen bond acceptor < 10; HBD, hydrogen bond donor ≤ 5; MLog P, Moriguchi octanol–water partition coefficient ≤ 4.15; Lipinski’sviolations ≤ 1; Bioavailability score > 0.1; TPSA: topological polar surface area < 140

Table 4.

The evaluation of toxicity on key metabolites

Compound	hERG Blockers	H-HT	DILI	Carcinogencity	LD50
Butyrate	Non-blocker	positive	negative	negative	3.301 mg/kg
Acetate	Non-blocker	positive	negative	negative	2.579 mg/kg
Propionate	Non-blocker	positive	negative	negative	3.158 mg/kg
Trimethylamine oxide	Non-blocker	positive	negative	negative	1.877 mg/kg

Note[13]: H-HT, Human Hepatotoxicity; DILI, Drug Induced Liver Injury

Discussion

The prevalence of KOA has been on an upward trend, driven not only by population aging and rising obesity rates but also linked to modern dietary patterns and physical activity levels [16]. As understanding deepens regarding the GM and its metabolites' critical role in maintaining structural and functional homeostasis, increasing attention is focused on the microbiota as an inducer or modulator of skeletal health and disease [17]. Multiple data sets support the association between dysbiosis and low-grade inflammation, a key component in the pathogenesis of musculoskeletal disorders [18]. Beneficial metabolites produced by the GM act as positive regulators to maintain homeostasis in the host system.

Network pharmacology is an interdisciplinary research field integrating systems biology, genomics, and proteomics. By synthesizing large-scale data, it systematically elucidates therapeutic mechanisms for complex diseases, offering novel strategies for drug development and disease treatment [19]. This study demonstrates that network pharmacology significantly contributes to understanding the molecular mechanisms of GM metabolites in KOA and identifying potential therapeutic metabolites. Our findings indicate that IL6, IL1B, and NFKB1 are three core targets regulated by GM metabolites in KOA. Based on G-T-M network analysis, we identified butyrate, acetate, propionate, and trimethylamine oxide as metabolites with promising therapeutic potential for KOA. KEGG enrichment analysis indicated that lipid metabolism and atherosclerosis, along with the TNF signaling pathway, represent key intervention pathways for KOA treatment. The biological functions of GM metabolites primarily focus on immune regulation and inflammatory responses.

KOA, as a common chronic degenerative disease of the osteoarticular system, exhibits pathological features including synovial inflammation, cartilage degeneration, osteophyte formation, and subchondral sclerosis [20]. KOA is not merely a degenerative disease resulting from localized mechanical damage but rather a systemic immune-metabolic disorder characterized by chronic low-grade inflammation [21]. Pro-inflammatory cytokines serve as key mediators in the pathophysiological disruption of osteoarthritis. Currently identified pro-inflammatory cytokines in KOA primarily include IL-1β, TNF-α, IL-6, IL-15, IL-17, IL-18 and IL-21. Among these, IL-1β, TNF-α and IL-6 are the principal pro-inflammatory cytokines involved in the pathogenesis of osteoarthritis [22]. IL-1β, a representative member of the IL-1 family, inhibits chondrocyte synthesis of extracellular matrix (ECM) components and disrupts the production of key structural proteins such as type II collagen and aggrecan [23, 24]. It also impairs chondrocyte function in synthesizing matrix metalloproteinases (MMPs), enzymes that destructively affect cartilage components [25]. TNF-α exhibits marked synergistic effects with IL-1β, often amplifying IL-1β's actions [26, 27]. IL-6 production in affected joint tissues typically responds to IL-1β and TNF-α, primarily promoting osteoclast formation to enhance bone resorption while synergizing with IL-1β and TNF-α [28]. Osteoblasts stimulated by all three cytokines become sources of these factors and may adversely affect adjacent cartilage by producing MMPs, further driving disease progression [23].

There has recently been a shift in the understanding of the pathogenesis of osteoarthritis (OA), with the emergence of a framework described as “mechanically initiated, metabolically and inflammatory driven, and mechanically sensitized.” This framework posits that abnormal mechanical loading triggers initial joint damage, while systemic metabolic dysfunction and persistent low-grade inflammation are key factors driving disease progression [29]. According to this framework, metabolic syndrome (MetS)—which includes obesity, diabetes, dyslipidemia, and hypertension—is increasingly recognized as playing a causal role in the pathogenesis of osteoarthritis (OA). The concept of “metabolic osteoarthritis” (MetS-OA) has emerged, in which metabolic disturbances and low-grade systemic inflammation exert deleterious effects on multiple joint tissues, including cartilage, bone, and the synovium, ultimately leading to complete loss of joint function [30]. Importantly, gut microbiota dysbiosis has been identified as a key upstream regulator of this process: metabolic dysfunction and altered microbial metabolites jointly drive systemic inflammation, oxidative stress, and immune dysregulation, thereby accelerating the progression of osteoarthritis (OA). The concept of “leaky gut syndrome” further elucidates how impaired intestinal barrier function allows microbial products to translocate into the systemic circulation, thereby exacerbating both systemic and local joint inflammation [31]. Collectively, this evidence supports the core finding of this study: that gut microbiota metabolites—particularly short-chain fatty acids (SCFAs)—exert therapeutic effects on KOA by targeting the NF-κB and tumor necrosis factor (TNF) signaling pathways, thereby disrupting the metabolic-inflammatory axis that drives disease progression.

Butyrate, acetate, and propionate all belong to SCFAs. SCFAs are crucial metabolites of the GM, serving as mediators in communication between the microbiome and the immune system. They suppress pro-inflammatory mediators, activate immune cells, and regulate the body's immune responses. Through SCFAs, GM participate in regulating numerous physiological processes between host and microbiota, including maintaining intestinal barrier function, immune modulation, and anti-inflammatory effects [32]. Extensive research indicates that most reduced GM in KOA patients are next-generation probiotics capable of managing metabolic diseases and producing SCFAs [11]. Therefore, restoring SCFA levels has become one of the most relevant therapeutic approaches for microbiota regulation [33].

It is worth noting that, among the four identified metabolites, trimethylamine N-oxide (TMAO) warrants closer examination due to its well-documented dual role in human health. Although our network pharmacology analysis identified TMAO as a potential therapeutic metabolite for KOA, a substantial body of literature has confirmed a close association between elevated TMAO levels and systemic inflammation, adverse cardiovascular events, and all-cause mortality [34, 35]. Mechanistically, TMAO has been shown to promote inflammation by activating the NLRP3 inflammasome, induce endothelial dysfunction through mitochondrial reactive oxygen species (ROS) production and endoplasmic reticulum stress, and disrupt cholesterol and bile acid metabolism [36]. Of particular relevance to this study, recent evidence directly indicates that TMAO is involved in the pathogenesis of osteoarthritis: TMAO upregulates the mechanosensitive ion channel Piezo1 in chondrocytes, making these cells more sensitive to mechanical loading and thereby exacerbating cartilage degradation [37]. Furthermore, TMAO promotes osteoclast differentiation and oxidative stress by activating the NF-κB pathway [38], which is a core target identified in our PPI network analysis. These findings suggest that TMAO may act as a pro-pathogenic factor in KOA, rather than merely a beneficial metabolite. However, the role of TMAO in health and disease remains controversial, with some studies questioning whether TMAO is a direct mediator or merely a bystander associated with underlying pathology. There is a marked discrepancy between our identification of TMAO as a potential therapeutic candidate and descriptions of its pathogenic role in the existing literature; this may reflect dose-dependent effects, context-specific actions, or the inherent limitations of network pharmacology approaches that predict interactions based on database associations rather than functional outcomes.

Obesity is a recognized risk factor for KOA. SCFAs regulate energy metabolism through complementary mechanisms, thereby aiding in weight management. SCFAs, as key representatives of GM metabolites, demonstrate multiple potentials for controlling body weight by regulating energy metabolism. Acetate activates the free fatty acid receptors FFAR2/GPR43 and FFAR3/GPR41 expressed on intestinal epithelial cells, stimulating the secretion of glucagon-like peptide-1 (GLP-1) and peptide YY (PYY) by enteroendocrine L cells. This effectively suppresses appetite and reduces energy intake, providing a crucial physiological basis for weight management [39, 40]. Butyrate reduces fat production by inhibiting histone deacetylase activity and downregulating genes associated with lipid synthesis [41]. Simultaneously, it promotes the release of leptin from white adipose tissue, further enhancing central appetite suppression and promoting energy expenditure, forming a positive feedback regulation of energy balance [42]. Propionate inhibits hepatic gluconeogenesis and suppresses hepatic lipid synthesis while promoting fatty acid oxidation, synergistically regulating glucose and lipid metabolism [43, 44]. Overall, these SCFAs influence appetite regulation, lipid metabolism, and liver function, offering potential targets for metabolic interventions in obesity-related knee osteoarthritis. Further research is needed to translate these findings into clinical strategies.

Additionally, diverse GM originating from the “G-T-M” network play a crucial role in regulating interactions between metabolites and the host. Elucidating these interactions is essential for advancing novel therapeutic strategies targeting KOA. Acetate production pathways are widely distributed across the microbiota, while species such as Akkermansia municiphilla have been identified as key mucin-degrading microorganisms producing propionate. Butyrate production is primarily dominated by Faecalibacterium prausnitzii, Eubacterium and Ruminococcus bromii [45]. These communities are also revealed in our “G-T-M” network. The abundance of Eubacterium in the gut is highly correlated with beneficial SCFAs levels across multiple clinical conditions, including inflammatory bowel disease, metabolic syndrome, and colorectal cancer. For instance, butyrate is often significantly reduced in the intestines of IBD patients, a phenomenon closely linked to its lost anti-inflammatory function. This occurs because butyrate produced by Eubacterium binds to G protein-coupled receptors on intestinal epithelial cells. It suppresses inflammation by inhibiting the NF-κB pathway or histone deacetylase activity, thereby downregulating pro-inflammatory cytokines like IFN-γ, IL-1β, IL-6, IL-8 and TNF-α, while upregulating anti-inflammatory cytokines such as IL-10 and TGF-β. thereby alleviating intestinal inflammation [46].

This study preliminarily elucidates the potential value of GM metabolites in KOA intervention, but several limitations remain that require further exploration and refinement in future research: First, regarding data sources, this study primarily relied on bioinformatics databases such as gutMGene for predictive analysis. Although this database is relatively systematic, it still contains some microorganisms and metabolic targets with uncharacterized functions, which may affect the completeness of mechanism inference. Additionally, delays in database updates may limit the timeliness and reliability of results. Second, regarding metabolite safety assessment, while the four screened metabolites have reported good bioavailability, experimental evidence on their hepatotoxicity in human liver cells remains insufficient. Therefore, future studies should validate their preclinical safety through standardized in vitro hepatotoxicity testing. Finally, regarding validation, the conclusions of this study are primarily based on network pharmacology and bioinformatics predictive analyses and have not yet been substantiated by in vitro or in vivo experiments. Network pharmacology predictions are, by their nature, hypothesis-generating rather than confirmatory; they provide a systematic basis for prioritizing targets and pathways, but do not constitute direct evidence of biological activity or therapeutic efficacy. Subsequent studies should systematically validate the molecular targets, signaling pathways, and specific effects of these metabolites in osteoarthritis through cellular experiments, animal models, and clinical sample testing. Furthermore, while the molecular docking results support a potential binding interaction, the binding energy indicated is relatively weak, and these findings should be interpreted with caution. Computer-aided docking alone cannot account for conformational changes, allosteric effects, or the biological context of target binding; therefore, experimental validation—such as surface plasmon resonance (SPR) or cell-based thermal migration assay (CETSA)—is required to confirm the actual interactions and determine their functional implications.

Conclusion

This study systematically elucidated the protective mechanisms of GM metabolites in KOA by integrating network pharmacology and bioinformatics approaches. Results indicate that butyrate, acetate, propionate, and trimethylamine oxide exert anti-inflammatory and immunomodulatory effects by regulating lipid and atherosclerosis pathways, TNF signaling pathways, and other inflammation- and immunity-related pathways through key targets such as IL6, IL1B, and NFKB1, thereby slowing KOA progression. Among these, butyrate exhibited a prominent central role within the regulatory network. While this study deepens our understanding of the potential therapeutic value of GM metabolites in KOA treatment, the conclusions are primarily based on computational prediction analyses. Further validation of their mechanisms of action, pharmacodynamic properties, and preclinical safety through subsequent experiments remains necessary.

Abbreviations

KOA

Knee osteoarthritis

GM

Gut microbiota

PPI

Protein–Protein interaction

GO

Gene ontology

KEGG

Kyoto encyclopedia of genes and genomes

SEA

Similarity ensemble approach

STP

Swiss target prediction

BP

Biological process

CC

Cellular component

MF

Molecular function

SCFAs

Short-chain fatty acids

Authors' contributions

Fengjiao Chen and Yufeng Tao wrote the main manuscript; Jing Deng and Leyi Zhang collected database information; Lanlan Yu and Zhuoxi Yang conducted data analysis; Yixuan Zhang and Siru Chen created the illustrations; Chi Zhang reviewed the manuscript.

Data Availability

The data used in this study are available from the corresponding author upon request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent to Publish declaration

Not applicable.

Competing interests

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The authors declare no competing interests.