Development and validation of a multimodal model integrating gut microbiota and metabolite for identifying sarcopenia in patients with MASLD: a study from two centers in China

Abstract

Background and aims

Metabolic dysfunction-associated steatotic liver disease (MASLD) is a common chronic liver disease worldwide, and identifying sarcopenia is critical since it is correlated with poor prognosis. Little is known about mechanistic alterations in the pathogenesis of this condition. This study aimed to explore the alterations in the gut microbiome and metabolome in patients with sarcopenia and develop a predictive model.

Methods

We performed shotgun metagenomic sequencing and untargeted liquid chromatography-mass spectrometry (LC-MS) metabolomic profiling of fecal samples from the discovery cohort (70 patients without sarcopenia, 30 with sarcopenia). A microbiota-metabolite score (MM score) was developed using LASSO regression to identify key microbiome and metabolite features associated with sarcopenia. A multimodal prediction model incorporating the MM score and clinical parameters was then developed and validated in an independent cohort of 50 patients.

Results

Patients with sarcopenia exhibited altered gut microbiota and metabolomic profiles, with significantly elevated Enterococcus faecium and Bacteroides vulgatus species, and elevated bile acids. Integration of the MM score with clinical variables (age, BMI, AST, presence of diabetes) resulted in a multimodal model with an AUC of 0.911, outperforming existing models including FIB-4 (AUC 0.765), NFS (AUC 0.724), and using only MM score alone (AUC 0.818). In a prospective validation cohort, the multimodal model demonstrated superior diagnostic performance (AUC 0.897), with significant improvements in clinical utility as evidenced by calibration curves and decision curve analysis.

Conclusions

This study developed a novel multimodal model combining gut microbiome, metabolomics, and clinical data for accurate prediction of sarcopenia, offering a promising approach for early identification of high-risk MASLD patients with sarcopenia.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12937-025-01198-2.

Introduction

Metabolic dysfunction-associated steatotic liver disease (MASLD), formerly termed non-alcoholic fatty liver disease (NAFLD), is the most prevalent chronic liver disease globally. While most patients follow an asymptomatic course, a subset progresses to advanced fibrosis, a critical prognostic determinant linked to cirrhosis and hepatocellular carcinoma [1, 2]. Beyond hepatic injury, MASLD intersects with systemic comorbidities, including sarcopenia, a progressive and generalized skeletal muscle disorder characterized by accelerated loss of muscle mass and strength [3–5]. Sarcopenia exacerbates MASLD progression through insulin resistance, chronic inflammation, and direct promotion of hepatic fibrosis [6–8], yet clinical approaches remain inadequate in addressing this complex relationship. Emerging evidence implicates gut microbiota dysbiosis and microbial metabolite perturbations in both sarcopenia pathogenesis and MASLD fibrosis via the gut-liver-muscle axis [9–15]. However, conflicting findings limit our understanding of whether gut microbiota alterations directly drive sarcopenia or are secondary to aging or metabolic dysfunction [16], highlighting the need to characterize the role of intestinal microbiota-derived metabolites in MASLD-associated muscle deterioration.

Existing diagnostic approaches for MASLD-associated sarcopenia rely on isolated clinical or imaging parameters (e.g., muscle mass quantification, fibrosis biomarkers), which fail to capture the complex pathophysiological changes of the gut-liver-muscle axis [9–12]. This limited perspective overlooks the significant advantages of integrating microbial functional signatures with host metabolic alterations—a critical gap hindering precision interventions. To address this, we developed a multimodal diagnostic framework combining gut microbiome profiling, metabolomic characterization, and clinical parameters. Unlike prior studies, our approach systematically integrates shotgun metagenomics and untargeted metabolomics to elucidate how microbial community dynamics interface with metabolic outputs in the intestinal microenvironment, thereby revealing the intricate interplay among gut microbiota, host metabolism, and sarcopenia progression. This framework, validated in an independent prospective cohort, identifies clinically relevant biomarkers that could facilitate risk stratification and guide personalized therapeutic interventions.

Materials and methods

Study design and population

This was a two-center study conducted at the Third Affiliated Hospital of Sun Yat-sen University (SYSUTH) and Shenzhen Hospital of Southern Medical University to develop and validate a multimodal diagnostic model for predicting sarcopenia. The study comprised two phases: a retrospective modeling phase and a prospective validation phase. These patients underwent Dual-energy X-Ray Absorptiometry (DXA) to assess sarcopenia and were willing to provide stool samples for further investigation during the period from January 2020 to December 2022. Inclusion criteria encompassed adults aged 18 to 75 years diagnosed with MASLD through imaging or biopsy, in conjunction with the presence of t least one cardiometabolic risk factor as outlined in the clinical guidelines [17]. Cardiometabolic criteria included the following: (1) Body mass index (BMI) ≥ 23 kg/m² or waist circumference ≥ 94 cm (males) and ≥ 80 cm (females), (2) Fasting plasma glucose ≥ 5.6 mmol/L (100 mg/dL) or glycated hemoglobin (HbA1c) ≥ 5.7% or a previous diagnosis of type 2 diabetes or treatment for type 2 diabetes, (3) Blood pressure (BP) ≥ 130/85 mmHg or treatment for hypertension, (4) Triglycerides ≥ 1.7 mmol/L or lipid lowering therapy, (5) High-density lipoprotein cholesterol (HDL-C) < 1.0 mmol/L (males) or < 1.3 mmol/L (females) or lipid lowering therapy. The exclusion criteria were as follows: significant alcohol consumption (> 20 g/d for women and > 30 g/d for men); treated with antibiotics/probiotics within the previous 3 months; presence of other chronic liver diseases, such as viral hepatitis or autoimmune hepatitis; hepatobiliary malignancies; Pregnancy (confirmed by urine pregnancy test or self-reported); Use of a radiographic contrast agent like barium within the past seven days; Body weight exceeding 204 kg or height greater than 196 cm (which exceeded DXA operational limits). The prospective cohort comprised 50 MASLD patients recruited from January 2023 to July 2024, meeting the same inclusion and exclusion criteria (Figure S1). In this phase, stool samples were also collected for shotgun metagenomic and untargeted metabolomic profiling to validate the multimodal diagnostic model developed from the retrospective cohort. The study adhered to the principles outlined in the Declaration of Helsinki. The study protocol was approved by the internal review boards and ethics committees of the SYSUTH and the Shenzhen Hospital of Southern Medical University. Written informed consent was obtained from all participating. Details of clinical and laboratory data, stool sample collection and processing, assessment of sarcopenia, and bioinformatic analysis are described in the supplement.

Statistical analysis

Wilcoxon rank-sum test (two-tailed) was performed to assess differences in α diversity, while the permutation multivariate analysis of variance (PERMANOVA) was used to evaluate differences in β diversity between the two groups. Spearman correlation analysis was used to examine relationships between differentially enriched features. AUROC comparisons were performed using the Delong test. A P value < 0.05 was considered statistical significance, and all analyses were conducted using R software (version 4.1.3).

Results

Patient characteristics

Of the 150 patients analyzed, 33.3% (n = 50) were diagnosed with sarcopenia. Patients with sarcopenia exhibited several characteristic clinical profiles: they were predominantly female, older, and had higher BMI values compared to non-sarcopenic individuals, as well as elevated aspartate aminotransferase (AST) and alanine aminotransferase (ALT) levels, increased diabetes prevalence, and reduced albumin and platelet counts (Table S1). The overall cohort had a median age of 50 years with 46.0% being male. The mean BMI was 28.2 kg/m², with 42.6% (64/150) having type 2 diabetes and 14.6% (22/150) diagnosed with hypertension.

We performed high-throughput shotgun metagenomic sequencing of all fecal samples to characterize the taxonomic composition of microbial communities. We also conducted to non-targeted LC-MS metabolomics for identification of both established and novel metabolites (Fig. 1A). Our analytical approach focused on three main objectives: (1) Identifying disease-specific alterations in individual microbial and metabolic profiles; (2) Exploring associations between disease-enriched microbiota and metabolites; and (3) Evaluating the potential diagnostic value of these features for sarcopenia. We validated the identified features in an independent cohort comprising 15 patients with sarcopenia and 35 MASLD patients without sarcopenia.

Fig. 1.

Fecal microbiome structure in patients with or without sarcopenia.A Fecal samples were collected from 100 MASLD patients, including 35 with sarcopenia and 65 without sarcopenia, for metagenomic and metabolomic profiling. B α-diversity analysis of gut microbiota. The Shannon index (t-test: t = 2.56, df = 35.8, P = 0.0149) and Chao1 index (Wilcoxon test: W = 1397, P = 0.00915) were significantly lower in the sarcopenia group compared to the non-sarcopenia group. C β-diversity analysis using principal coordinate analysis (PCoA). The gut microbiota composition showed significant separation between the non-sarcopenia and sarcopenia groups (PERMANOVA p < 0.05). Relative abundance of gut microbiota at the phylum level in individual patients. D Phylum-level comparison demonstrating higher Firmicutes and Bacteroidota with lower Proteobacteria, Verrucomicrobiota, and Actinobacteriota in sarcopenia patients compared to non-sarcopenia patients. E Relative abundance of gut microbiota at the species level in individual patients. F Comparison of gut microbiota composition at the species level. The sarcopenia group displayed higher Enterococcus_faecium and lower Klebsiella_pneumoniae and Collinsella_tanakaei proportions compared to the non-sarcopenia group. G Comparison of gut microbiota composition at the species level. The sarcopenia group displayed higher Enterococcus faecium proportion compared to the non-sarcopenia group

Microbial community structure specific to sarcopenia unveiled by shotgun metagenomic sequencing

We performed metagenomic sequencing and mapped high-quality reads to an updated gut microbiome gene catalog, generating microbiome profiles of the discovery cohort. To evaluate differences in bacterial community composition between sarcopenia and non-sarcopenia groups, we analyzed both α-diversity and β-diversity metrics. Alpha diversity analysis revealed significant differences between groups, with Shannon index significantly lower in sarcopenia patients compared to controls (t = 2.56, df = 35.8, P = 0.0149). Similarly, Chao1 richness was significantly reduced in the sarcopenia group (Wilcoxon test: W = 1397, P = 0.00915), indicating diminished microbial diversity. Beta diversity analysis using PERMANOVA revealed significant differences in gut microbial community structure between groups (R² = 0.344, F (1,98) = 51.377, P = 0.001). These findings indicate alterations in the gut bacterial community of sarcopenia patients. β-diversity analysis also demonstrated significant differences in gut microbiota composition between cases and controls (R² = 0.171, P < 0.05, Fig. 1C). PERMANOVA analysis further confirmed significant differences in gut microbial community structure between groups (R²= 0.344, F(1,98) = 51.377, P = 0.001). To account for potential site-specific technical variations in metabolomics data, we systematically evaluated confounding effects using PERMANOVA. The analysis demonstrated that disease status emerged as the most significant contributor to microbial β-diversity variation (R² = 0.20, P = 0.001), while sampling center accounted for only a small proportion of variance (R² = 0.06, P = 0.001). These results confirm that the observed community differences primarily reflect disease pathophysiology rather than technical variations between centers.

We then examined the overall gut microbial composition in sarcopenia patients at both phylum and species levels using Principal Coordinates Analysis (PCoA). At the phylum level, microbial relative abundances differed significantly between the two groups (Fig. 1D-E) (P < 0.05). The sarcopenia group displayed higher proportions of Firmicutes, Bacteroidota and lower proportions of Proteobacteria, Verrucomicrobiota and Actinobacteriota, compared to the control group. At the species level, microbial composition also showed significant differences between groups (Fig. 1F-G) (P < 0.05).

Functional characterization of the microbiome in patients with sarcopenia

We further characterized functional alterations in the gut microbiota associated with sarcopenia. Notably, Bacteroides vulgatus, Ruminococcus bromii, Streptococcus constellatus, and Klebsiella pneumoniae were more abundant in the group without sarcopenia, while Enterococcus faecium and Anaerostipes hadrus showed higher abundance in the sarcopenia group. These findings suggest distinct microbial signatures associated with sarcopenia pathophysiology. Correlation analysis among differentially abundant species revealed complex interspecies relationships (Fig. 2A-B).

Fig. 2.

Functional alterations of gut microbiota associated with sarcopenia. A Differentially abundant bacterial species between the non-sarcopenia and sarcopenia groups. B Correlation analysis of the significantly altered species

Clusters of Orthologous Groups (COG) functional annotation revealed significant enrichment of several categories in the sarcopenia group, including hemerythrin superfamily proteins, CRISPR-associated proteins, and archaeal ADP-dependent phosphofructokinase/glucokinase (Figure S2A). These enriched pathways may represent specific microbial functional adaptations in the sarcopenia microenvironment. Analysis using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database identified distinct functional differences between sarcopenia and control groups. At both level 1 and level 2 pathways, we observed significant enrichment of metabolic processes in the sarcopenia group, particularly those related to carbohydrate metabolism (Figure S2B-C), which could contribute to the pathogenesis of sarcopenia.

Altered fecal metabolite profiles in patients with sarcopenia

We performed fecal metabolomic profiling to investigate metabolic alterations associated with sarcopenia in MASLD patients. Several significantly altered metabolites were identified between sarcopenia and non-sarcopenia groups. Volcano plot analysis revealed significantly elevated levels of ursodeoxycholic acid, O-desmethylangolensin, and cholic acid in the sarcopenia group, while metabolites such as acylcarnitine 13:0 were significantly decreased (Fig. 3A). Heatmap visualization highlighted the top 10 differentially abundant metabolites between groups (Fig. 3B). Notably, levels of bile acid-related metabolites (e.g., ursodeoxycholic acid and cholic acid) were markedly elevated in patients with sarcopenia, suggesting a potential link between bile acid metabolism and sarcopenia pathophysiology. Principal Coordinate Analysis (PCoA) of fecal metabolomic data demonstrated distinct clustering patterns between sarcopenia and non-sarcopenia groups (Fig. 3C). The two groups exhibited a clear separation along the first two principal components, reflecting significant compositional differences in their fecal metabolomes (P < 0.05). Correlation analysis (Fig. 3D) revealed strong positive associations between bile acid metabolites (e.g., ursodeoxycholic acid and O-desmethylangolensin), whereas inverse correlations were observed between these metabolites and compounds such as 2-piperidone and taurine, highlighting distinct metabolic signatures in sarcopenia patients. KEGG pathway enrichment analysis (Figure S3) demonstrated that differentially abundant metabolites were significantly enriched in pathways related to the citrate cycle and amino acid biosynthesis, suggesting these metabolic pathways may play crucial roles in sarcopenia development.

Fig. 3.

Fecal metabolomic profiles in non-sarcopenia and sarcopenia. A Volcano plot of differential fecal metabolites between non-sarcopenia and sarcopenia groups. B Heatmap showing the top 10 metabolites with the most significant differential abundance between the two groups. C Principal Coordinate Analysis (PCoA) demonstrating distinct clustering and dispersion of fecal metabolites between non-sarcopenia and sarcopenia groups. D Correlation analysis between fecal metabolites

Development and validation of a microbiota-metabolite-based predictive model for sarcopenia in MASLD patients

To identify potential microbiota and metabolites associated with sarcopenia in MASLD patients, we performed least absolute shrinkage and selection operator (LASSO) regression analysis using the most significantly differentially expressed species and metabolites (Fig. 4A). Five key variables were identified, including three bacterial species (Enterococcus faecium, Anaerostipes hadrus, and Bacteroides vulgatus) and two metabolites (cholic acid and ursodeoxycholic acid) (Fig. 4B). A microbiota-metabolite score (MM score) was then constructed based on the β coefficients of these five variables, achieving an area under the curve (AUC) of 0.818 (95%CI:0.729–0.903) in receiver operating characteristic (ROC) analysis (Fig. 4C). The precision-recall AUC (PR-AUC) of the MM score was 0.768 (Figure S4). Patients were stratified into high-score and low-score groups based on the median MM score. PCoA revealed significant dispersion between the two groups (P < 0.001), indicating that the MM score effectively captures distinct microbiota-metabolite profiles (Fig. 4D). To further evaluate the clinical utility of the MM score, we performed univariate and multivariate logistic regression analyses integrating clinical variables (age, gender, BMI, ALT, AST, albumin, platelet count, and diabetes status) with the MM score (Fig. 4E). Multivariate analysis identified age, BMI, AST, diabetes, and the MM score as independent predictors of sarcopenia in MASLD patients (P < 0.05). Multicollinearity diagnostics confirmed minimal correlation between predictors (highest correlation coefficient: 0.14; all VIF values < 1.06; maximum condition index: 1.31), demonstrating that the MM score captures unique biological information not represented by conventional clinical parameters. Concerning the risk of optimism bias in our multivariable model, we also applied the bootstrap optimism correction with nested cross-validation. The original and shrinkage-adjusted coefficients for all predictors were listed in the Table S2. The adjustment preserved the direction and relative importance of each feature while reducing their absolute magnitude to provide more realistic effect estimates. A nomogram model was developed using these five variables (Fig. 4F). Calibration curves demonstrated good agreement between the predicted and observed probabilities, and decision curve analysis (DCA) indicated high clinical utility and net benefit (Fig. 4G-H). The nomogram model demonstrated excellent discrimination (AUC = 0.911) with balanced diagnostic metrics at the optimal threshold of 0.544 (sensitivity = 0.743, specificity = 0.954, PPV = 0.897, NPV = 0.873), consistently maintaining high performance across clinically relevant decision thresholds. Compared to existing models such as FIB-4 and NFS, the nomogram achieved superior diagnostic performance with an AUC of 0.911 (95% CI: 0.849–0.973) [vs. FIB-4: AUC 0.765 (95% CI: 0.657–0.873), NFS: AUC 0.724 (95% CI: 0.625–0.824), DeLong tests P < 0.0001 both comparisons) (Fig. 4I). Finally, patients were stratified into high-risk and low-risk groups based on the median nomogram score. PCoA analysis showed significant dispersion between the two groups, further validating the model’s ability to distinguish patients with different sarcopenia risks (P < 0.001) (Fig. 4J).

Fig. 4.

Construction of a diagnostic model for identifying sarcopenia based on microbial and metabolomic signatures. A LASSO regression analysis of the most significantly differentially abundant species and metabolites. Identification of key species and metabolites contributing to the model. B Selection of the top 3 species (Enterococcus faecium, Anaerostipes hadrus, Bacteroides vulgatus) and 2 metabolites (Cholic acid, ursodeoxycholic acid) with the greatest discriminatory power for sarcopenia diagnosis. C Construction of the microbiota-metabolite score (MM score) by fitting these 5 variables using β coefficients. D Patients were classified into high-risk and low-risk groups based on the median MM score. PCoA shows substantial separation between the two groups. E Univariate and multivariate regression analyses incorporating clinical variables (Age, Gender, BMI, ALT, AST, ALB, PLT, Diabetes) alongside MM score. Multivariate analysis reveals that Age, BMI, AST, Diabetes, and MM score significantly contribute to the model. F A multimodal model constructed using Age, BMI, AST, Diabetes, and MM score to predict sarcopenia in MASLD patients. G-H Calibration curve and Decision Curve Analysis (DCA) demonstrating that the multimodal model has excellent accuracy and predictive performance. I ROC analysis of different models in identifying sarcopenia. The multimodal model displayed the highest AUC, compared to clinical parameter-only models (P < 0.05). J Based on the multimodal model, patients are divided into high risk and low risk groups based on the median score. PCoA shows a clear separation between the two groups, indicating distinct microbial and metabolic profiles

Validation of the nomogram in a prospective dataset

In a prospective study including 50 patients with DXA examinations (30 MASLD patients without sarcopenia and 20 MASLD patients with sarcopenia), the predictive performance of the nomogram was evaluated and compared to the MM score, FIB-4, and NFS. The AUC for the nomogram was 0.897 (95% CI: 0.784–0.985), demonstrating superior diagnostic accuracy compared to the FIB-4 [AUC 0.743 (95% CI: 0.559–0.926), Delong test, P = 0.014), and NFS [AUC 0.737 (95% CI: 0.567–0.907), Delong test, P = 0.006) (Fig. 5A). Calibration curves and decision curve analysis (DCA) confirmed the predictive precision and clinical utility of the nomogram and MM score (Fig. 5B-C). Both tools provided reliable predictions, with the nomogram showing superior performance. We performed subgroup analyses to assess the stability of the nomogram’s diagnostic performance across various clinical parameters. The AUC values for the nomogram remained stable across subgroups stratified by age (≤ 55 or > 55 years), gender, BMI (< 28 or ≥ 28), diabetes status (yes or no), FIB-4 values (< 1.3 or ≥ 1.3), and NFS values (≤ −1.455 or > −1.455), with no significant differences observed in any subgroup (P > 0.05) (Fig. 5D).

Fig. 5.

Validation of the multimodal model in a prospective cohort. A Prospective inclusion of 50 MASLD patients with DXA results. ROC analysis shows that the multimodal model’s AUC is 0.936, which is superior to alternative models (P < 0.05). B-C Calibration curve and Decision Curve Analysis (DCA) demonstrate that the multimodal model and MM score have excellent accuracy and predictive performance in the prospective cohort. D AUC results indicate that the nomogram performs consistently robustly across different clinical subgroups, including different age strata, gender, BMI strata, Diabetes status, FIB-4 strata, and NFS strata

Discussion

Our study reveals significant alterations in gut microbiome and metabolome of sarcopenia patients, suggesting these changes contribute to sarcopenia development. We developed and validated a multimodal prediction model incorporating microbiome, metabolomic, and clinical data that effectively predicts sarcopenia in MASLD patients.

Previous studies have established connections between gut microbiota and MASLD, with most research showing reduced bacterial diversity in MASLD patients [18–20]. Diminished microbial richness has also been associated with obesity [21], inflammatory bowel disease [22], and recurrent Clostridium difficile-associated diarrhea [23]. Le Chatelier et al. observed that individuals with lower bacterial richness exhibited higher total adiposity, insulin resistance, dyslipidemia, and a more pronounced inflammatory phenotype compared to those with greater microbial diversity. Low bacterial richness was characterized by: (1) decreased butyrate-producing bacteria; (2) increased mucolytic potential; (3) reduced hydrogen and methane production with increased hydrogen sulfide formation capacity; and (4) elevated oxidative stress potential [24]. These findings suggest that reduced microbial diversity may trigger responses that contribute to the pathogenesis of MASLD. In our study, both α- and β-diversity were lower in sarcopenia patients. Sarcopenia in MASLD patients typically indicates a more severe inflammatory phenotype and metabolic dysfunction [25], suggesting reduced gut microbiota diversity may contribute to sarcopenia development through these mechanisms.

In current study, several differentially enriched COG categories, including CRISPR-Cas system components and iron transporters, may reflect microbial adaptive responses to host immune pressure and nutrient availability in the altered gut environment associated with sarcopenia. Additionally, KEGG pathway analysis revealed enrichment of disease-related pathways (e.g., p53 signaling, colorectal cancer, and viral infections) and glycan biosynthesis pathways in the MASLD with sarcopenia group, which may indicate increased microbial involvement in host-microbe interactions and inflammation. In contrast, pathways related to endocytosis and Fc gamma R–mediated phagocytosis were reduced, potentially suggesting impaired immune-related microbial functions.

In the context of MASLD, we identified three bacterial species - Enterococcus faecium, Anaerostipes hadrus, and Bacteroides vulgatus- along with two metabolites, cholic acid and ursodeoxycholic acid, which may play a significant role in the diagnosis of sarcopenia. Enterococcus faecium, a common inhabitant of the human gut, was found to be associated with sarcopenia. Previous studies [26–29] have emphasized the role of Enterococcus species in gut dysbiosis, which is often seen in metabolic disorders like MASLD and cirrhosis. Notably, these previous studies [26–29] reported varying outcomes. Such discrepancies may be attributed in part to factors such as age, sex, geography, race/ethnicity, genetics, regional dietary patterns, and lifestyle [30–32].

In our current study, we established a positive correlation between Enterococcus faecium and sarcopenia. Enterococcus faecium is thought to contribute to the pathogenesis of liver fibrosis by disrupting the gut - liver axis. This disruption, in turn, exerts an influence on liver inflammation and immune responses. The potential underlying mechanism may involve Enterococcus faecium generating a key bioactive metabolite, tyramine. It is likely that this tyramine activates PPAR-γ, which ultimately leads to lipid accumulation, inflammation, and fibrosis in the liver [33]. Previous reports have indicated that sarcopenia is associated with the progression of liver fibrosis in patients with MASLD. Specifically, in the context of MASLD, the severity of liver fibrosis is directly proportional to the likelihood of developing sarcopenia [8]. Since Enterococcus faecium promotes severe fibrosis development, its positive correlation with sarcopenia observed in our study is biologically plausible.

Our study also found Anaerostipes hadrus to be significantly associated with sarcopenia. Previous research has linked A. hadrus with severe liver fibrosis development. In patients with cirrhosis, fatty acids were the primary metabolites affected in the liver, portal vein, and gut microbiome. The synthesis of fatty acids by Anaerostipes hadrus affects the free fatty acids in the portal vein. Both the microbial production of fatty acids and the portal FFAs were linked to increased liver fibrosis [34]. Since sarcopenia is positively correlated with liver fibrosis [8], and Anaerostipes hadrus can trigger severe hepatic fibrosis, it is reasonable that Anaerostipes hadrus shows a positive correlation with sarcopenia. On the other hand, Bacteroides vulgatus displays a negative correlation with sarcopenia in the current study. Bacteroides vulgatus is a member of the Bacteroides genus. A recent study indicated that it was highly abundant in the feces of non- obese mice. Mice colonized with Bacteroides vulgatus showed improved glucose tolerance and insulin sensitivity. Bacteroides vulgatus appears to ameliorate high-fat diet-induced obesity by modulating intestinal serotonin synthesis and lipid absorption [35]. The negative correlation between Bacteroides vulgatus and sarcopenia may suggest it has protective effects against liver fibrosis, though this hypothesis requires further investigation.

Beyond microbial alterations, we identified two sarcopenia-associated metabolites—cholic acid and ursodeoxycholic acid—as potential biomarkers. A study from Japan demonstrated that the total fecal bile acid concentrations in patients with severe hepatic steatosis were significantly elevated compared to those in healthy controls [36]. Notably, MASLD patients exhibited marked increases in secondary unconjugated bile acids including CA, deoxycholic acid, chenodeoxycholic acid, UDCA, and lithocholic acid [36]. The gut microbiota plays a pivotal role in maintaining bile acid homeostasis through biotransformation processes. As bile acids regulate nutrient metabolism and influence BMI/insulin resistance [36, 37], their dysregulation may represent a critical link between MASLD and sarcopenia pathogenesis. Specifically, secondary bile acids like CA exhibit proinflammatory properties in the colonic environment [38], potentially exacerbating intestinal barrier dysfunction in MASLD patients. We propose that bile acid-mediated impairment of intestinal permeability promotes metabolic endotoxinemia via bacterial translocation, triggering insulin resistance and proinflammatory cytokine production—hallmarks of MASLD progression [39, 40]. Importantly, the observed positive correlation between blood endotoxin levels and liver fibrosis severity suggests a vicious cycle where CA-driven hyperendotoxemia may simultaneously promote both hepatic and muscular deterioration.

Despite providing valuable insights, our study has several limitations. First, the relatively small sample size may limit result generalizability. Second, while we identified bacterial species and metabolites associated with sarcopenia, causal relationships remain unclear; longitudinal studies and mechanistic experiments are needed to determine directionality and underlying biological processes. Third, population-specific factors may influence our findings. As presented in Fig. 1B, inherent heterogeneity in the gut microbiome composition in the same clinical subgroup was observed. This heterogeneity may stem from variations in diet, disease severity, comorbidities, or host genetics—all factors that influence gut microbiota composition and metabolic profiles. Expanding the study to include diverse populations would help address these potential confounding factors.

In conclusion, our study highlights the crucial role of gut microbiome and metabolomic alterations in sarcopenia progression and introduces a novel predictive model for MASLD risk stratification. With further validation, this model could significantly impact clinical practice and facilitate development of personalized treatment strategies for MASLD patients.

Authors’ contributions

SW, ML, QL and WG were responsible for the study concept and design. SW, ML, WL, YR and YW conducted the acquisition and cleaning of data. QL and WG performed the critical revision. SW, ML and WL were responsible for the the statistical analysis. QL and WG supervised the research. All authors contributed to the manuscript for important intellectual content and approved the submission.

Funding

This work was supported by the National Natural Science Foundation of China (82401448), Guangdong Basic and Applied Basic Research Foundation (2021A1515110799 and 2023A1515010144), Shenzhen Science and Technology Program (JCYJ20240813145414019 and JCYJ20220530142000001) and Shenzhen Nanshan District Science and Technology Plan Funding Program (NSZD2024015).

Data availability

The datasets used and analyzed in the current study are available from the corresponding author on reasonable request. The datasets (PRJNA1267271) generated for this study can be found in the SRA of NCBI: https://www.ncbi.nlm.nih.gov.

Declarations

Ethics approval and consent to participate

The Institutional Review Board of the Third Affiliated Hospital of Sun Yat-sen University (Approval Number: [2019]02 − 530 − 01) and the Shenzhen Hospital of Southern Medical University (Approval Number: NYSZYYEC20190017).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.