Lactobacillus mucosae extracellular vesicles improve metabolic dysfunction-associated fatty liver disease by activating the AMPK pathway via the gut-liver axis

Abstract

Background and objective

Metabolic dysfunction-associated fatty liver disease (MAFLD) is the most prevalent chronic liver disease worldwide but lacks effective therapies. Gut microbiota play a key role in MAFLD via the gut-liver axis, yet targeted interventions remain unclear. This study investigates the therapeutic effects and mechanisms of Lactobacillus mucosae (L. mucosae) and its extracellular vesicles (EVs) in MAFLD.

Methods

At the clinical level, 16S rRNA sequencing was performed to analyze differences in gut microbiota composition between MAFLD patients and healthy controls, followed by the identification of a core probiotic strain, L. mucosae, using culture-based approaches. A high-fat diet-induced MAFLD mouse model was subsequently established to evaluate the therapeutic effects of L. mucosae, while free fatty acid-induced hepatocyte models were used to identify L. mucosae-EVs as the key active component. Mechanistic investigations and validation were further conducted in vivo and in vitro using RNA sequencing, qPCR, Western blot analyses, DIA-based EV proteomics, and Proteinase K-mediated EV protein functional assays.

Results

MAFLD patients showed reduced intestinal Lactobacillus abundance, negatively correlated with hepatic lipid parameters. L. mucosae administration alleviated steatosis, metabolic dysfunction, inflammation, and oxidative stress in mice. In vitro, L. mucosae-EVs were confirmed as the main active component, and Proteinase K-pretreated EVs showed attenuated protective effects on hepatocytes. Mechanistically, L. mucosae-EVs activated hepatic AMPK signaling via the gut-liver axis, remodeled gut microbiota, and restored intestinal barrier integrity.

Conclusion

This study systematically demonstrates for the first time that L. mucosae-EVs ameliorate MAFLD through multiple mechanisms, including gut microbiota remodeling, intestinal barrier restoration, and activation of hepatic AMPK signaling via the gut-liver axis. Proteinase K-mediated functional assays suggest that EV-associated protein cargo contributes, at least in part, to these protective effects. These findings provide a solid experimental foundation and theoretical basis for the development of precision microbiome-based therapeutic strategies for MAFLD based on L. mucosae and its EVs.

Graphical Abstract

L. mucosae-EVs improve MAFLD by activating the AMPK pathway through the gut-liver axis. Orally administered L. mucosae-EVs act through the gut-liver axis and activate hepatic AMPK signaling, resulting in synchronized metabolic improvements, including enhanced fatty acid oxidation, suppressed lipogenesis, improved insulin sensitivity, and attenuated oxidative stress, inflammation, and fibrosis. Restoration of gut microbiota homeostasis and reinforcement of intestinal barrier integrity further contribute to this systemic amelioration. EVs, extracellular vesicles; UC, ultracentrifugation

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12951-026-04843-y.

Introduction

Metabolic dysfunction-associated fatty liver disease (MAFLD), also referred to as metabolic dysfunction-associated steatotic liver disease (MASLD) and previously known as non-alcoholic fatty liver disease, is a chronic and progressive hepatic condition that arises due to excessive nutrition intake and insulin resistance [1–3]. The global prevalence of MAFLD exceeds 30%, and it is expected to become the leading cause of end-stage liver disease by 2030 [1–3]. The disease spectrum of MAFLD ranges from simple hepatic steatosis to steatohepatitis, which can further progress to liver fibrosis, cirrhosis, and even hepatocellular carcinoma [2]. Additionally, MAFLD is closely linked to extrahepatic metabolic disorders such as cardiovascular disease and type 2 diabetes, significantly increasing overall mortality and imposing a major public health burden. However, existing treatment strategies for MAFLD primarily rely on lifestyle interventions, which suffer from poor long-term adherence and limited efficacy [3]. Therefore, the development of novel and effective therapeutic measures has become a critical issue in this field that requires urgent attention.

Gut microbiota, through the gut-liver axis, contributes to MAFLD pathogenesis by dysbiosis-driven alterations in microbial metabolites (e.g., short-chain fatty acids and bile acids) that compromise intestinal barrier integrity, increase portal translocation of microbial products, and activate hepatic inflammatory pathways such as TLR4-NF-κB and NLRP3, thereby exacerbating lipotoxicity, chronic inflammation, and fibrogenesis [4–7]. Accordingly, targeted modulation of the gut microbiome, particularly through probiotics, has emerged as a promising strategy to ameliorate MAFLD and related metabolic disorders [8–10]. Lactobacillus mucosae (L. mucosae), a facultative anaerobe with strong mucosal colonization and mucus layer adaptation, has been proposed as a next-generation probiotic [11, 12]. Prior studies report that it lowers serum cholesterol via upregulation of hepatic cholesterol metabolism genes such as Srebp2 and Ldlr, partially restores microbiota homeostasis, alleviates metabolic endotoxemia, and remodels pathogenic microbiota, thereby improving high-fat diet-induced dyslipidemia and atherosclerosis in ApoE-/- mice [13, 14]. Collectively, these properties suggest that L. mucosae may also ameliorate MAFLD, although its direct impact on hepatic pathological progression and the underlying mechanisms have not been fully elucidated.

Extracellular vesicles (EVs) derived from probiotics are nanoscale lipid bilayer particles that encapsulate diverse bioactive molecules, including proteins, nucleic acids, and metabolites, and are capable of traversing the intestinal mucus barrier to enter the systemic circulation, thereby mediating inter-organ communication and regulating immune responses, intestinal homeostasis, and metabolic processes [15]. Accumulating evidence indicates that probiotic EVs enhance intestinal barrier integrity by upregulating tight junction proteins and suppressing TLR2/4-mediated inflammatory signaling, while simultaneously activating PPAR-α and PPAR-γ pathways to promote fatty acid oxidation and improve obesity-related metabolic dysfunction [16, 17]. Moreover, probiotic EVs have been shown to modulate gut microbiota composition, restore the FXR-FGF15 signaling axis, and activate intestinal aryl hydrocarbon receptor signaling, thereby inducing the expression of antimicrobial peptides such as Reg3 and antioxidant factors including NRF2, which in turn reduce bacterial translocation and endotoxin release and ultimately attenuate hepatic steatosis and inflammatory injury associated with alcoholic liver disease [18, 19]. Despite these findings underscoring the therapeutic potential of probiotic EVs in metabolic and liver diseases, the specific roles and underlying molecular mechanisms of probiotic EVs—particularly L. mucosae-EVs—in MAFLD remain limited.

In this study, we first performed a clinical-level analysis using 16S rRNA sequencing to compare the gut microbiota profiles of healthy controls and patients with MAFLD. A probiotic strain, L. mucosae, was then isolated and identified using culture-based approaches. Subsequently, MAFLD-related in vitro cellular and in vivo animal models were established to systematically evaluate the therapeutic efficacy and underlying mechanisms of L. mucosae and its EVs. This study aims to provide a solid experimental foundation and theoretical basis for the application of L. mucosae and its EVs in precision microbiome-based therapy for MAFLD.

Materials and methods

Clinical cohort and sample collection

This study recruited 56 volunteers from the Second Affiliated Hospital of Nanchang University, including 29 diagnosed MAFLD patients and 27 healthy controls. Before inclusion, all participants completed a detailed questionnaire regarding their medical history, lifestyle, and dietary habits, and provided fresh fecal samples. After collection, fecal samples were stored at −80 °C within 2 h for subsequent 16S rRNA gene high-throughput sequencing and microbiological culture analysis. The inclusion criteria for MAFLD patients were based on the "Guidelines for the Prevention and Treatment of Metabolic Dysfunction-Associated (Non-Alcoholic) Fatty Liver Disease (Version 2024)" [20]. All MAFLD patients were 18–65 years old and underwent liver ultrasound imaging and transient elastography, with fatty liver severity graded according to the ultrasound attenuation parameter. Exclusion criteria included: (1) coexisting liver diseases with clear causes (such as viral hepatitis, drug-induced liver injury, autoimmune liver diseases, alcoholic liver disease, etc.) or biliary tract diseases; (2) malignancies, or severe underlying diseases of important organs such as the heart, brain, lungs, and kidneys, which may affect the study outcomes; (3) pregnancy or breastfeeding; (4) use of hormones, probiotics, antibiotics, or other medications in the past 3 months.

All participants were permanent residents of Nanchang, China, and provided written informed consent. The study adhered to the Declaration of Helsinki and WHO’s ethical guidelines, with protocol approval from the Biomedical Research Ethics Committee of the Second Affiliated Hospital of Nanchang University (Clinical Ethics No. O-Medical Research Ethics-2025-019).

Probiotic strain properties and safety evaluation

L. mucosae was isolated, identified, and cultured from fecal samples obtained from healthy controls and designated as strain HG003 (CGMCC No.37136). The growth kinetics of the strain were evaluated by plotting growth curves based on optical density at 600 nm. Tolerance to different pH conditions and bile salt concentrations was assessed by incubating the strain in PBS and quantifying viable bacterial counts after 4 h. Antioxidant activity was evaluated by measuring the strain’s ability to scavenge superoxide anions, hydroxyl radicals, and DPPH radicals, as well as its iron ion chelation capacity and reducing power. In vitro adhesion to HT-29 intestinal epithelial cells was assessed by co-incubating the strain with cells, followed by Gram staining to visualize bacterial adherence. Safety was evaluated using hemolysis assays on blood agar and antibiotic susceptibility testing with the Kirby-Bauer disk diffusion method. Finally, the antimicrobial activity of the strain’s metabolic products was assessed using the Oxford cup method by measuring inhibition zone diameters against intestinal pathogens.

Preparation, characterization, and tracing of EVs

L. mucosae was anaerobically cultured in MRS medium at 37 °C to the stationary phase. Bacterial cells and debris were removed by centrifugation (10,000 × g, 30 min, 4 °C) and filtration through a 0.22 μm membrane. EVs were isolated from the supernatant by ultracentrifugation at 120,000 × g for 90 min at 4 °C, washed once with sterile PBS, and re-centrifuged under the same conditions. The final EV pellet was resuspended in PBS and stored at −80 °C. The supernatant after EV removal was collected as EV-depleted supernatant for control experiments. EV protein concentration was measured using a BCA Protein Assay Kit (Beyotime, China) after lysing EVs in PBS containing 0.1% Triton X-100, following the manufacturer’s instructions. EV morphology, size distribution, and particle concentration were characterized by TEM (HT7700, HITACHI, Japan) and NTA (ZetaView PMX-120, Particle Metrix, Germany) calibrated with 100 nm beads.

For tracing experiments, EVs were labeled with PKH26 (Sangon Biotech, China) for in vitro uptake by HT-29 and HepG2 cells or with DiR (Dolemi, China) for in vivo oral administration in mice; uptake and biodistribution were visualized by confocal microscopy and ex vivo fluorescence imaging, respectively. The protein cargo of EVs was analyzed by label-free DIA proteomics: EVs were lysed, trypsin-digested, and analyzed by LC–MS/MS (Orbitrap Astral, Thermo Fisher Scientific), and DIA data were searched against the Lactobacillus mucosae UniProt database (taxonomy ID: 97,478) using Diann software with standard parameters.

Experimental animals and treatment protocols

Healthy male C57BL/6 mice (aged 6–8 weeks) were obtained from Gempharmatech Co., Ltd. and housed in specific pathogen-free (SPF) conditions at the Biomedical Testing Center Animal Platform of Nanchang University (SYXK(Gan) 2021–0004). Environmental conditions were maintained at 50% ± 10% humidity, 22 °C ± 2 °C temperature, and a 12-h light/dark cycle. All mice had free access to food and water, and were fed standard chow. Prior to the experiment, the mice underwent a one-week acclimatization period.

MAFLD was induced by feeding mice a high-fat diet (Xiaoshuyoutai, China, D12492) for 12 weeks, with 60% of calories from fat. Pre-experiment: From week 12, MAFLD mice were orally gavaged daily for 4 weeks with 200 µL bacterial suspensions (1 × 10⁹ CFU/mL) of L. brevis, L. gasseri, or L. mucosae to screen strains with potential therapeutic effects. Experiment 1: From week 12, MAFLD mice received 200 µL L. mucosae (1 × 10⁹ CFU/mL) daily for 4 weeks to formally assess therapeutic efficacy. Experiment 2: From week 12, normal-diet mice received 50 µg L. mucosae-EVs or PBS daily for 4 weeks to assess safety. Experiment 3: From week 12, MAFLD mice received 50 µg L. mucosae-EVs daily for 4 weeks to investigate the mechanism of action. Experiment 4: From week 12, MAFLD mice received 50 µg L. mucosae-EVs daily, or 50 µg L. mucosae-EVs combined with intraperitoneal 20 mg/kg AMPK inhibitor Compound C, for 4 weeks to validate the mechanism of EV treatment.

At week 16, mice were euthanized under isoflurane anesthesia. Blood samples were collected, centrifuged to separate the serum, and stored at −80 °C. Liver and colon tissues were harvested, weighed, photographed, fixed in 4% paraformaldehyde, or snap-frozen in liquid nitrogen and stored at −80 °C for later use. All experimental procedures were approved and supervised by the Animal Ethics and Welfare Committee of Nanchang University (Animal Ethics No. NCULAE-20241023001).

Cell culture and treatment protocols

The HepG2, AML12, and HT-29 cell lines were obtained from the Cell Bank of the Chinese Academy of Sciences and maintained in Dulbecco’s Modified Eagle Medium (Servicebio, China, G4511) supplemented with 10% heat-inactivated fetal bovine serum (Sangon Biotech, China, BC-SE-FBS07C) and 1% penicillin–streptomycin (Servicebio, China, G4003). For AML12 cell culture, 1% dexamethasone (OriLeaf, China, S17003) was also added to the medium. All cells were maintained at 37 °C with 5% CO₂ and saturated humidity in a conventional incubator.

To establish the MAFLD cell model, HepG2 and AML12 cells were treated with 0.75 mM free fatty acids (FFA) (Kunchuang Biotechnology, China, KC006) for 24 h. Pre-experiment: The appropriate concentrations of various L. mucosae components for improving MAFLD were tested. In the FFA-induced HepG2 cell model, cells were treated with different concentrations of L. mucosae supernatant, heat-killed L. mucosae cells, and its derived EVs for 24 h. Experiment 1: To evaluate the effective components of L. mucosae for improving MAFLD, HepG2 cells were treated with 10% L. mucosae supernatant or 1 × 10⁷ CFU/mL heat-killed L. mucosae cells for 24 h. Experiment 2: To identify the active components in the L. mucosae supernatant, HepG2 cells were treated with 50 μg/mL L. mucosae-EVs or the EV-depleted supernatant (10%) for 24 h. Experiment 3: To further investigate the in vitro mechanism of L. mucosae-EVs, HepG2 and AML12 cells were treated with 50 μg/mL L. mucosae-EVs, with or without 5 μmol/L Compound C, for 24 h. Experiment 4: To validate the role of EV-associated proteins, HepG2 cells were treated with 50 μg/mL intact or Proteinase K-pretreated L. mucosae-EVs for 24 h.

Intracellular lipid accumulation was evaluated using Oil Red O and Nile Red staining. Cells were washed with PBS, fixed with 4% paraformaldehyde, and stained accordingly. For Oil Red O, cells were incubated with the solution (Servicebio, China, G1015) for 30 min, differentiated with 60% isopropanol, and counterstained with hematoxylin. For Nile Red, cells were treated with 1 μg/mL Nile Red (Merck, Germany, N3013) at 37 °C for 20 min and counterstained with DAPI. Stained cells were examined using optical and fluorescence microscopy (Nikon, Japan; Perkin Elmer & Olympus, Germany).

Biochemical analysis

Biochemical parameters, including total cholesterol (TC), triglycerides (TG), HDL, LDL, alanine aminotransferase (ALT), and aspartate aminotransferase (AST), were measured in serum and tissues using commercial kits (Nanchang Jianchen, China) to assess lipid metabolism and liver function. Myocardial injury biomarkers (CK-MB) and renal function markers (BUN, CREA, UA) were also analyzed for biological safety. Oxidative stress was evaluated by measuring the activities of superoxide dismutase (SOD) and catalase (CAT), as well as the levels of glutathione (GSH) and malondialdehyde (MDA) in liver tissue. The levels of inflammatory cytokines (IL-1β, IL-6, TNF-α) and endotoxin (LPS) in serum and tissue homogenates were quantified using ELISA kits (Jianglai Biotech, China). Fasting blood glucose (FBG) and fasting insulin (FINS) levels were measured to evaluate glucose homeostasis, and HOMA-IR and HOMA-IS were calculated. Glucose clearance and insulin sensitivity were assessed through IPGTT and IPITT. In cell experiments, TC and TG levels in cell pellets, and ALT and AST activities in culture supernatants were measured. All procedures followed kit instructions to ensure consistency and comparability of results.

Histopathological and immunostaining analysis

Liver tissue was fixed in 4% paraformaldehyde, embedded in paraffin, and sectioned for various staining procedures. Hematoxylin and eosin (HE) staining (Servicebio, China, G1005) was used to observe tissue structure and inflammatory cell infiltration; Oil Red O staining (Servicebio, China, G1015) detected neutral lipid accumulation; Masson’s trichrome staining (Servicebio, China, G1006) assessed liver fibrosis by visualizing collagen fibers. Liver reactive oxygen species (ROS) levels were measured using the dihydroethidium (DHE) fluorescent probe (Servicebio, China, G1746), with red fluorescence indicating superoxide anion content. Immunohistochemistry detected α-SMA (Servicebio, China, GB13044) and Collagen I (Servicebio, China, GB154197) expression. For intestinal tissue, colon paraffin sections were stained with HE (Servicebio, China, G1005) to evaluate mucosal structure and inflammation. Dual immunofluorescence staining for Occludin (Servicebio, China, GB151401) and ZO-1 (Servicebio, China, GB121981) was performed to analyze tight junction protein distribution and integrity in the intestinal barrier.

Quantitative real-time PCR analysis

Total RNA was extracted from mouse liver tissues using TRIzol reagent (Life Technologies, USA), and RNA concentration and purity were determined with a NanoDrop 2000 spectrophotometer (Thermo Scientific, USA). Subsequently, 1 μg of total RNA was reverse-transcribed into cDNA using Hifair® AdvanceFast gDNA Removal RT Premix (Yeasen, China, 11151ES60). Quantitative real-time PCR (qPCR) was performed on an Applied Biosystems platform using Hieff UNICON® SYBR Green Premix (Yeasen, China, 11185ES08). The amplification conditions were as follows: initial denaturation at 95 °C for 30 s, followed by 40 cycles of denaturation at 95 °C for 10 s and annealing/extension at 60 °C for 30 s. β-actin was used as the internal reference gene, and relative gene expression levels were calculated using the 2^ − ΔΔCt method. The primer sequences are listed in Table S2.

Western blotting analysis

Tissue and cell samples were lysed with RIPA buffer containing protease inhibitors (Solarbio, China) on ice, and lysates were collected by centrifugation. Protein concentrations were measured using the BCA assay. Equal amounts of protein were separated by SDS-PAGE, transferred to PVDF membranes (Millipore, USA), and blocked with 5% non-fat milk for 2 h. Membranes were incubated with primary antibodies overnight at 4 °C, then with HRP-conjugated secondary antibodies for 2 h at room temperature. Protein bands were visualized using a chemiluminescent substrate (Invitrogen, USA) and imaged with an automated gel system (Tanon, China). Band intensities were quantified using ImageJ software. Antibody details are in Table S3.

Flow cytometry analysis

Cells were loaded with the DCFH-DA fluorescent probe (Beyotime, China) and incubated at 37 °C in the dark for 20 min. Afterward, the cells were washed three times with PBS to remove any residual probe that had not entered the cells. The intensity of DCF fluorescence signals was measured using a flow cytometer (Sony ID7000, Japan) to assess intracellular reactive oxygen species (ROS) levels.

Fecal 16S rRNA sequencing

Fecal samples were stored at −80 °C until analysis. Microbial DNA was extracted using a commercial kit (Personalbio, China), and its quality and concentration were assessed with a NanoDrop system (Thermo Scientific, USA). The V3–V4 regions of the 16S rRNA gene were amplified using barcoded primers and subjected to paired-end sequencing on the Illumina NovaSeq platform. Raw data were processed with the QIIME2 pipeline, using DADA2 for quality control and generating amplicon sequence variants (ASVs). Taxonomic assignment was performed using the classify-sklearn module and the SILVA database. Alpha (Chao1 and Shannon) and beta (Bray–Curtis) diversity indices were calculated, followed by principal coordinate and taxonomic composition analyses. All raw sequencing data produced in this study were deposited in the NCBI SRA database in FASTQ format, with the human fecal 16S rRNA sequencing data registered under accession number PRJNA1365503 and the mouse fecal 16S rRNA sequencing data under accession number PRJNA1365674.

Liver RNA sequencing

Total RNA extraction, quality assessment, library construction, and sequencing of liver tissue were outsourced to Shanghai Weihuan Biotechnology Co., Ltd. (APExBIO, China, https://www.apexbio.cn/). The criteria for screening differentially expressed genes were adjusted p-value < 0.05 and absolute fold change ≥ 1.5. The differentially expressed gene set was analyzed using Gene Ontology (GO) functional analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment, with statistical significance of enrichment determined by the hypergeometric distribution test. Additionally, Gene Set Enrichment Analysis (GSEA, http://www.broadinstitute.org/gsea) was used to systematically evaluate the distribution patterns of other predefined gene sets in the phenotype-related ranked lists at the whole transcriptome level. All raw sequencing datasets generated in the present study were deposited in the NCBI SRA database as FASTQ files under the accession number PRJNA1365667.

Statistical analysis

Statistical analysis and data visualization were conducted using GraphPad Prism 9.0, SPSS 26.0, and R 4.3.2, based on independent biological replicates with sample sizes indicated in the figure legends. Normality was tested using the Shapiro–Wilk method. For normally distributed data, mean ± standard deviation are reported and analyzed with an unpaired t-test or one-way ANOVA followed by Bonferroni post hoc analysis. Welch’s t-test or Welch’s ANOVA with Dunnett’s T3 correction was used for data with unequal variances. Non-normally distributed data are presented as median (interquartile range) and analyzed using the Mann–Whitney U test or Kruskal–Wallis test with Dunn’s post hoc test. Categorical data are expressed as counts (percentages) and compared using the chi-square or Fisher’s exact test. Correlation analysis was performed using Spearman’s rank correlation. All tests were two-sided, and p < 0.05 was considered statistically significant.

Statement on the use of AI in manuscript preparation

During the writing process, DeepSeek was used under the authors’ supervision to assist in polishing the language. The final manuscript was thoroughly reviewed and approved by the authors, who are solely responsible for its content.

Results

MAFLD patients show reduced Lactobacillus in the gut microbiota, correlating negatively with liver lipid markers

A total of 27 healthy controls (HC) and 29 MAFLD patients (MF) were recruited to investigate the role of gut microbiota in MAFLD. There were no significant differences between the groups in age and gender (p > 0.05), but the MF group had significantly higher levels of BMI, ALT, AST, TC, TG, LDL, and GLU, and significantly lower HDL levels compared with the HC group (p < 0.05, Table S1). 16S rRNA gene V3–V4 region sequencing of fecal samples revealed lower alpha diversity indices (Chao1, Faith_pd, Observed_species) in the MF group (p < 0.05), while the Simpson index showed no significant difference (p > 0.05) (Fig. S1A). Beta diversity analysis indicated significant microbial composition differences (Fig. S1B). At the phylum level (Fig. 1A), Firmicutes and Proteobacteria showed an increasing trend in the MF group (p > 0.05), while Actinobacteria was significantly reduced (p < 0.05), and Bacteroidetes decreased but not significantly (p > 0.05) (Fig. S1C). At the genus level (Fig. 1B), the MF group had significantly lower Lactobacillus abundance (p < 0.05), while Enterococcus increased significantly (p < 0.05) (Fig. 1D). LEfSe, random forest models, and functional prediction analysis further highlighted Lactobacillus as a key genus distinguishing the groups (Figs. 1C, F and S1D, E). Correlation analysis (Fig. 1E) showed that Lactobacillus was negatively correlated with BMI, ALT, and TG, suggesting its potential as a probiotic target for MAFLD treatment.

Fig. 1.

Decreased relative abundance of Lactobacillus in the gut microbiota of MAFLD patients negatively correlates with hepatic lipid deposition. A Bar plot of microbial composition at the phylum level. B Bar plot of microbial composition at the genus level. C Cladogram from LEfSe (LDA Effect Size) analysis. D Relative abundances of Bifidobacterium, Lactobacillus, Enterococcus, and Streptococcus. E Heatmap of correlation analysis between gut microbial genera and indicators of hepatic lipid deposition. F Random forest analysis of species differences and the corresponding ROC curve. Data are presented as mean ± SD or median (IQR) and were compared using the independent samples t-test or Mann–Whitney U test. **p < 0.01. HC, healthy controls (n = 27); MF, MAFLD patients (n = 29). BMI, body mass index; ALT, alanine aminotransferase; TG, triglycerides

Screening and characterization of L. mucosae as a promising probiotic for MAFLD treatment

To screen beneficial Lactobacillus for improving MAFLD (Fig. S2A), we used culture-based techniques, high-throughput sequencing, and acid and bile salt tolerance tests (Fig. S2B). Candidate strains, including L. brevis, L. gasseri, and L. mucosae, were isolated. A pre-experiment established an MAFLD animal model and evaluated the intervention effects of each strain (Fig. S2C). Among the strains, L. mucosae showed the most significant improvement in MAFLD (Fig. S2D–I). In vitro evaluations of L. mucosae revealed it entered the exponential growth phase after ~ 8 h and reached a stable phase at ~ 20 h (Fig. 2A). The strain exhibited high tolerance to various pH conditions (2.0–7.0) and bile salt concentrations (0.0–0.3%), with high survival rates after 4 h of exposure (Fig. 2B–C). Antioxidant tests showed scavenging rates of 44.61%, 53.36%, and 74.63% for DPPH, hydroxyl radicals, and superoxide anions, respectively, with an Fe²⁺ chelation ability of 36.12% and a total reducing power of 0.78 (Fig. 2D). L. mucosae also exhibited strong adhesion to HT-29 cells (Fig. 2E), no hemolytic activity (Fig. 2F), inhibitory effects against intestinal pathogens (Fig. 2G), and some resistance to antibiotics (Fig. 2H). These results suggest that L. mucosae has excellent probiotic characteristics and potential for MAFLD treatment.

Fig. 2.

Evaluation of the probiotic properties of L. mucosae. A Growth curve. B Acid tolerance assay. C Bile salt tolerance assay. D Antioxidant activity assay. E Cell adhesion assay (scale bar: 20 μm and 10 μm). F Hemolytic activity assay. G Antimicrobial activity assay. H Antibiotic susceptibility test. n = 3. c, culture medium control; m, L. mucosae suspension; P. aeruginosa, Pseudomonas aeruginosa; BHS, β-hemolytic Streptococcus; C. albicans, Candida albicans; S. typhi, Salmonella typhi; S. aureus, Staphylococcus aureus; E. coli, Escherichia coli; L. monocytogenes, Listeria monocytogenes; S. flexneri, Shigella flexneri; PEN, penicillin; CHL, chloramphenicol; ERY, erythromycin; AMP, ampicillin; LIN, lincomycin; CRO, ceftriaxone; TET, tetracycline; SXT, trimethoprim-sulfamethoxazole; GEN, gentamicin; CIP, ciprofloxacin

L. mucosae reduces hepatic lipid deposition in high-fat diet-induced MAFLD mice

Building on the probiotic characteristics of L. mucosae observed in preliminary and in vitro studies, we evaluated its therapeutic effects on MAFLD in vivo (Fig. 3A). Compared with the MAFLD model group established by a 60% high-fat diet, mice receiving the L. mucosae intervention exhibited significant decreases in both body weight and liver weight (Figs. 3B, F and S3A–C). The MP group also exhibited significant improvements in serum TC, TG, HDL, LDL, and liver TC and TG content (Figs. 3C and S3D, E). Regarding glucose metabolism, the MP group showed better glucose homeostasis, lower insulin resistance, and improved pancreatic β-cell function (Figs. 3D–E and S3F–H). HE staining revealed reduced steatosis in the MP group compared to the M group, with L. mucosae alleviating hepatic steatosis (Fig. 3G). Oil Red O and Masson’s trichrome staining confirmed reduced lipid accumulation and fibrosis in the MP group (Figs. 3I and S3I). The MP group also showed reduced serum levels of IL-1β, IL-6, TNF-α, LPS (Fig. 3L), and improved liver oxidative stress markers (MDA, CAT, SOD, GSH) (Fig. 3J, K). Additionally, the MP group exhibited reduced serum ALT and AST activities, reflecting an overall improvement in hepatic function (Fig. 3H). In summary, L. mucosae improved lipid metabolism, glucose homeostasis, liver function, inflammation, and oxidative stress in the MAFLD mouse model, thereby supporting its potential utility as a probiotic intervention for MAFLD.

Fig. 3.

L. mucosae administration alleviates hepatic lipid deposition in MAFLD mice. A Schematic of the animal experimental design. B Body weight and liver weight. C Serum levels of TC and TG. D Blood glucose curves during the IPGTT. E Blood glucose curves during the IPITT. F Macroscopic appearance of livers. G Representative photomicrographs of liver sections stained with HE (scale bar: 200 μm and 50 μm). H Serum levels of ALT and AST. I Representative photomicrographs of liver sections stained with Oil Red O (scale bar: 200 μm and 50 μm). J, K Hepatic levels of MDA, SOD, GSH, and CAT. L Serum levels of IL-1β, IL-6, TNF-α, and LPS. Data are expressed as mean ± SD and were analyzed by one-way ANOVA or Welch’s ANOVA, with Bonferroni or Dunnett’s T3 post hoc test. ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001. C, control group; M, model group; MP, L. mucosae treatment group. n = 6 mice per group.TC, total cholesterol; TG, triglycerides; FBG, fasting blood glucose; FINS, fasting insulin; IPGTT, intraperitoneal glucose tolerance test; IPITT, intraperitoneal insulin tolerance test; ALT, alanine aminotransferase; AST, aspartate aminotransferase; MDA, malondialdehyde; SOD, superoxide dismutase; GSH, glutathione; CAT, catalase; IL-1β, interleukin-1β; IL-6, interleukin-6; TNF-α, tumor necrosis factor-α; LPS, lipopolysaccharide

L. mucosae-EVs alleviate FFA-induced lipid accumulation in HepG2 cells

To identify the active components of L. mucosae responsible for its beneficial effects, we first established an in vitro MAFLD model by treating HepG2 cells with 0.75 mM FFA for 24 h (Fig. S4A). After separating the supernatant from L. mucosae, we tested the effects of heat-killed L. mucosae cells (1 × 10⁷ CFU/mL) and supernatants (10%) on ALT, AST, and intracellular lipid levels. The supernatant significantly reduced ALT and AST levels and alleviated lipid accumulation, while the heat-killed L. mucosae cells had no effect (Fig. 4A, C). Oil Red O staining confirmed the supernatant’s role in reducing lipid deposition (Fig. 4B). Based on recent studies suggesting the therapeutic potential of EVs in metabolic diseases, we hypothesized that L. mucosae-EVs may play a key role [15–19]. EVs were isolated by ultracentrifugation (Fig. S4B) and characterized by NTA and TEM, showing a particle concentration of 3.4 × 10¹¹ particles/mL and a median size of 129.1 nm (Fig. 4D–E). Confocal microscopy confirmed effective cellular uptake of EVs by HepG2 and HT-29 cells (Fig. 4F). The optimal dose of 50 μg/mL EVs significantly reduced TG content, with no further improvement at higher concentrations (Fig. S4E). Comparison of 50 μg/mL EVs and EV-depleted supernatant revealed that only EVs significantly improved ALT, AST, TC, TG levels, and lipid deposition (Fig. 4G–I). These findings suggest that L. mucosae-EVs are the key components responsible for its lipid-improving effects in MAFLD.

Fig. 4.

L. mucosae-EVs alleviate FFA-induced lipid accumulation in HepG2 cells. A Schematic diagram of the cellular experimental design. B Representative images of Oil Red O staining (scale bar: 100 μm and 25 μm). C ALT and AST levels in cell culture supernatants and intracellular TC and TG contents. D Representative transmission electron microscopy (TEM) image of L. mucosae-EVs (scale bar: 100 nm). E Nanoparticle tracking analysis (NTA) of L. mucosae-EVs. F Cellular uptake of L. mucosae-EVs. Red arrows indicate representative intracellular fluorescence signals of PKH26-labeled L. mucosae-EVs (scale bar: 10 μm). G Schematic diagram of the cellular experimental design. H Representative images of Oil Red O staining (scale bar: 100 μm and 25 μm). I ALT and AST levels in cell culture supernatants and intracellular TC and TG contents. Data are expressed as mean ± SD and were analyzed by one-way ANOVA or Welch’s ANOVA, with Bonferroni or Dunnett’s T3 post hoc test. ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001. C, control group; M, model group; MC, heat-killed L. mucosae cells group; MS, L. mucosae supernatant group; ME, L. mucosae-EVs group; MES, EVs-depleted supernatant group. n = 3. Sup, supernatant. Sup (-EVs), EVs-depleted supernatant. EVs, extracellular vesicles; FFA, free fatty acids; ALT, alanine aminotransferase; AST, aspartate aminotransferase; TC, total cholesterol; TG, triglycerides; TEM, transmission electron microscopy; NTA, nanoparticle tracking analysis

In Vivo distribution and safety of L. mucosae-EVs

To assess the in vivo bioavailability of L. mucosae-EVs, we used in vivo imaging to track their biodistribution in mice (Fig. S4F). The results showed that after oral gavage, fluorescently labeled EVs primarily accumulated in the liver, spleen, and gastrointestinal regions (Fig. S4G–H), suggesting intestinal exposure and gastrointestinal–hepatic tissue distribution along the gut-liver axis. A systematic safety evaluation revealed no significant changes in serum ALT, AST, CK-MB, or kidney function markers (CREA, UA, BUN) after EV administration (Fig. S4I). HE staining of major organs, including the brain, heart, liver, spleen, lungs, kidneys, and colon, revealed no evident pathological alterations in the EV-treated group (Fig. S4J–K). These findings suggest that orally administered L. mucosae-EVs can be detected in gastrointestinal and hepatic tissues without evident systemic toxicity, supporting their potential clinical application in MAFLD treatment.

L. mucosae-EVs ameliorate MAFLD by activating the AMPK signaling pathway

To explore the mechanism of L. mucosae-EVs, we performed RNA sequencing on liver tissues from control (C) and MAFLD model (M) mice (Fig. S5A). Principal coordinate analysis (PCoA) and heatmap analysis revealed significant gene expression differences between the two groups (Fig. S5B–C). Among the top 20 upregulated and downregulated genes, many were involved in liver lipid metabolism (Figs. 5A–B and S5D). Gene Ontology (GO) analysis highlighted lipid metabolism as key in MAFLD pathogenesis (Fig. 5D). KEGG pathway analysis identified significant changes in PI3K-Akt, JAK-STAT, MAPK, and AMPK signaling pathways (Fig. 5C). qPCR validation showed that EV treatment activated AMPK pathway genes, including Acaca, Sirt1, and Ppargc1a (Fig. 5H), while other pathway genes remained unchanged (Fig. S5H). Western blotting confirmed that EVs promoted AMPK phosphorylation (Fig. 5I). Additionally, Gene Set Enrichment Analysis (GSEA) identified pronounced enrichment of pathways associated with oxidative stress, fibrosis, and inflammation, indicating their potential involvement in the progression of MAFLD (Figs. 5E–G and S5E–G).

Fig. 5.

Liver RNA sequencing reveals potential mechanisms by which L. mucosae-EVs ameliorate MAFLD. A Volcano plot of differentially expressed genes. B Heatmap of differentially expressed genes. C Bubble plot of KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway enrichment analysis. D Bubble plot of GO (Gene Ontology) enrichment analysis. E–G Gene set enrichment analysis (GSEA). H Hepatic gene expression. I Western blot analysis of hepatic p-AMPK and AMPK protein expression. Data are expressed as mean ± SD and were analyzed by one-way ANOVA or Welch’s ANOVA, with Bonferroni or Dunnett’s T3 post hoc test. *p < 0.05, **p < 0.01. C, control group; M, model group; ME, L. mucosae-EVs group. n = 3 mice per group. EVs, extracellular vesicles

Compound C blocks the therapeutic effects of L. mucosae-EVs on MAFLD

To confirm the involvement of the AMPK pathway in the EV-mediated amelioration of MAFLD, HepG2 and AML12 cells were exposed to 0.75 mM FFA for 24 h to establish in vitro models of MAFLD (Figs. 6A and S6A).EV treatment significantly reduced ALT, AST, intracellular TC, and TG levels (Figs. 6E and S6E) and decreased lipid accumulation, as shown by Oil Red O and Nile Red staining (Figs. 6B, D and S6B, D). Flow cytometry confirmed a reduction in intracellular ROS levels (Figs. 6C and S6C). These effects were reversed by the AMPK inhibitor Compound C, which also antagonized AMPK activation (Figs. 6F and S6F).

Fig. 6.

The AMPK inhibitor abrogates the ameliorative effects of L. mucosae-EVs on FFA-induced lipid accumulation in HepG2 cells. A Schematic diagram of the cellular experimental design. B Oil Red O staining of HepG2 cells (scale bar: 100 μm and 25 μm). C Flow cytometric analysis of ROS in HepG2 cells. D Nile Red staining of HepG2 cells (scale bar: 100 μm). E ALT and AST levels in HepG2 cell culture supernatants and intracellular TC and TG contents. F Western blot analysis of p-AMPK, AMPK, SREBP-1c, CPT1A, and PPAR-α protein expression in HepG2 cells. Data are expressed as mean ± SD and were analyzed by one-way ANOVA or Welch’s ANOVA, with Bonferroni or Dunnett’s T3 post hoc test. ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001. C, control group; M, model group; ME, L. mucosae-EVs group; MEI, L. mucosae-EVs and AMPK inhibitor (Compound C) group. n = 3. EVs, extracellular vesicles; FFA, free fatty acids; AMPK Inh, AMPK inhibitor; ALT, alanine aminotransferase; AST, aspartate aminotransferase; TC, total cholesterol; TG, triglycerides; ROS, reactive oxygen species

In vivo, 4 weeks of L. mucosae-EVs gavage significantly reduced body and liver weight (Figs. S7A–C and 7B) and improved lipid metabolism, liver function, and glucose homeostasis (Fig. 7C, D, F, J). Western blot confirmed AMPK pathway activation (Fig. 7I), and hepatic HE and Oil Red O staining showed alleviation of steatosis (Fig. 7E, H). Co-treatment with Compound C blocked these improvements. Additionally, L. mucosae-EVs alleviated liver fibrosis (Figs. 8D–F and S7F) and reduced liver inflammation (Figs. 8G–H and S7D–E) and oxidative stress (Fig. 8A–C), but these effects were diminished by Compound C. EVs also reduced colonic inflammation and upregulated tight junction proteins (ZO-1, Occludin, Claudin-1), with these protective effects lost upon Compound C co-treatment (Figs. 8I and S7G–I). In conclusion, AMPK activation represents a major downstream pathway involved in the protective effects of L. mucosae-EVs both in vitro and in vivo.

Fig. 7.

L. mucosae-EVs ameliorate hepatic lipid deposition and glucose metabolism in MAFLD mice via the AMPK signaling pathway. A Schematic of the animal experimental design. B Macroscopic appearance of livers. C Serum levels of TC, TG, HDL, LDL and hepatic TC, TG contents. D Fasting blood glucose (FBG) and fasting insulin (FINS) levels. E Representative photomicrographs of liver sections stained with HE (scale bar: 200 μm and 50 μm). F Blood glucose curves during IPGTT and IPITT, along with HOMA-IR and HOMA-IS indices. G Serum ALT and AST levels. H Representative photomicrographs of liver sections stained with Oil Red O (scale bar: 200 μm and 50 μm). I Western blot analysis of hepatic p-AMPK, AMPK, SREBP-1c, CPT1A, and PPAR-α protein expression. J Western blot analysis of hepatic PEPCK and G6PC protein expression. Data are expressed as mean ± SD and were analyzed by one-way ANOVA or Welch’s ANOVA, with Bonferroni or Dunnett’s T3 post hoc test. ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001. C, control group; M, model group; ME, L. mucosae-EVs group; MEI, L. mucosae-EVs and AMPK inhibitor (Compound C) group. n = 6 mice per group. EVs, extracellular vesicles; TC, total cholesterol; TG, triglycerides; HDL, high-density lipoprotein; LDL, low-density lipoprotein; IPGTT, intraperitoneal glucose tolerance test; IPITT, intraperitoneal insulin tolerance test; HOMA-IR, homeostasis model assessment of insulin resistance; HOMA-IS, homeostasis model assessment of insulin sensitivity; ALT, alanine aminotransferase; AST, aspartate aminotransferase

Fig. 8.

L. mucosae-EVs alleviate hepatic oxidative stress, fibrosis, and inflammation and enhance intestinal barrier function in MAFLD mice. A Representative fluorescence images of ROS staining in liver sections (scale bar: 100 μm). B Hepatic levels of MDA, SOD, GSH, and CAT. C Western blot analysis of hepatic NRF2 and HO-1 protein expression. D Immunohistochemical staining of α-SMA in liver sections. E Immunohistochemical staining of Collagen I in liver sections (scale bar: 200 μm and 50 μm). F Western blot analysis of hepatic Collagen I and α-SMA protein expression. G Western blot analysis of hepatic p-NF-κB and NF-κB protein expression. H Serum LPS levels. I Representative immunofluorescence images (scale bar: 100 μm) and Western blot analysis of colonic barrier proteins. Data are expressed as mean ± SD and were analyzed by one-way ANOVA or Welch’s ANOVA, with Bonferroni or Dunnett’s T3 post hoc test. ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001. C, control group; M, model group; ME, L. mucosae-EVs group; MEI, L. mucosae-EVs and AMPK inhibitor (Compound C) group. n = 6 mice per group. EVs, extracellular vesicles; ROS, reactive oxygen species; MDA, malondialdehyde; SOD, superoxide dismutase; GSH, glutathione; CAT, catalase; LPS, lipopolysaccharide

L. mucosae-EVs improve the gut microbiota in MAFLD mice

To investigate the modulatory role of L. mucosae-EVs on the gut microbiota in MAFLD mice, 16S rRNA sequencing analyses demonstrated that mice with high-fat diet-induced MAFLD displayed gut microbial dysbiosis comparable to that observed in human patients. EV treatment significantly improved both α-diversity (Fig. S8A–G) and β-diversity, with beneficial changes in species composition (Fig. S9K). At the phylum level, EV administration was associated with a decrease in the relative abundance of Proteobacteria (p < 0.05) and Firmicutes (p > 0.05), accompanied by an increase in Bacteroidetes (p > 0.05) and Verrucomicrobia (p < 0.05) (Fig. S9A–E). At the genus level, EVs significantly inhibited the proliferation of Allobaculum (p < 0.05) and showed a trend toward inhibiting Dorea (p > 0.05). Furthermore, EV treatment resulted in a significant elevation in the relative abundance of Akkermansia (p < 0.05) and showed a promotive trend toward the enrichment of Lactobacillus (p > 0.05) (Fig. S9F–J). Species contribution analysis suggested that Lactobacillus may be a key genus mediating functional differences in the microbiota (Fig. S8H).

Proteomic and functional analysis of L. mucosae-EVs

To further characterize the functional components of L. mucosae-EVs, DIA-based proteomic analysis was performed. The top abundant EV proteins included peptidoglycan-binding protein LysM, ABC transporter-related proteins, lipoprotein, phosphoglycerate kinase, enolase, peptidoglycan hydrolase, and MucBP domain-containing protein (Table S4). GO and KEGG analyses suggested that these proteins were mainly associated with metabolic regulation, transport, and cellular functional processes (Fig. S10A).

To further verify the role of EV-associated proteins, Proteinase K-pretreated EVs were used in FFA-induced HepG2 cells (Fig. S10B). Compared with intact L. mucosae-EVs, Proteinase K-pretreated EVs showed weakened effects on reducing lipid accumulation, ALT and AST release, and intracellular TC and TG levels (Fig. S10C–D). These results suggest that EV-associated proteins contribute, at least in part, to the therapeutic activity of L. mucosae-EVs in MAFLD, although the precise core functional proteins or peptides and their direct host targets remain to be identified.

Discussion

MAFLD is now recognized as the most common chronic liver disease globally, with complex pathogenesis involving metabolic disturbances and gut microbial dysbiosis [1]. Lifestyle modification remains the first-line therapeutic approach, but its long-term adherence is poor and clinical efficacy is limited [1–3]. In recent years, the gut-liver axis has emerged as an important therapeutic target, with probiotic-based interventions showing considerable promise [4, 5, 8–10, 21]. This study provides, for the first time, an integrated evidence chain supporting the therapeutic potential and major downstream mechanisms by which L. mucosae-EVs ameliorate MAFLD, including AMPK-associated metabolic regulation, gut microbiota remodeling, and restoration of intestinal mucosal barrier integrity. These findings not only offer novel and original insights into the therapeutic potential of L. mucosae and its EVs in metabolic diseases but also establish a solid experimental foundation and theoretical basis for the development of precision microbiome-based therapeutic strategies targeting the gut-liver axis in MAFLD.

First, our clinical cohort study confirmed that the gut microbiota structure of MAFLD patients significantly differs from that of healthy controls, characterized by decreased diversity, reduced abundance of beneficial bacteria such as Lactobacillus, and an increase in potential pathogenic bacteria such as Enterococcus. This finding is highly consistent with the dysbiosis patterns of MAFLD previously reported [22, 23]. Previous studies have confirmed that Lactobacillus can improve metabolic diseases through mechanisms such as the production of short-chain fatty acids (SCFAs), bile acid metabolism modulation, and immune regulation [10, 24, 25]. Our correlation analysis also suggested a significant negative correlation between Lactobacillus and key hepatic lipid markers such as BMI, ALT, and TG. Additionally, multiple bioinformatics analyses, including LEfSe, random forest, and functional contribution analysis, indicated that the reduction of Lactobacillus may be associated with the severity of MAFLD, making it a promising microbial target for therapy. Based on this, we successfully isolated and screened a strain with notable potential for improving MAFLD, L. mucosae. In vitro probiotic property evaluations confirmed that L. mucosae possesses the core characteristics required for a probiotic candidate, including good gastrointestinal environment tolerance, significant antioxidant capacity, strong adhesion to intestinal epithelial cells, and inhibitory activity against pathogens. Subsequent in vivo animal experiments further demonstrated that L. mucosae significantly alleviated hepatic steatosis induced by a high-fat diet, improved dysregulated lipid and glucose metabolism, and inhibited systemic inflammation and hepatic oxidative stress. These cumulative lines of evidence establish L. mucosae as a potential probiotic for improving MAFLD.

The traditional view attributes the beneficial effects of probiotics to their structural components, such as cell walls and surface proteins, or their secreted soluble metabolites, including short-chain fatty acids and bacteriocins [24, 26, 27]. However, in recent years, probiotic-derived EVs, as "non-living probiotics," have gained attention due to their ability to carry a variety of active molecules, such as proteins, lipids, metabolites, and nucleic acids. These EVs can traverse mucosal barriers and act on distant target organs such as the liver. They offer advantages such as structural stability, ease of standardization, and low immunogenicity, making them a research hotspot in the field of microbiome-based therapies [15, 18, 19]. Several studies have pointed out that EVs from probiotics such as Lactobacillus and Bifidobacterium can act as the next-generation postbiotics, exhibiting unique advantages in models of metabolic diseases, gut injury, and liver damage [18, 19, 28, 29]. Our study also provides new insights into this topic. This study innovatively identifies that the key active component responsible for the beneficial effects of L. mucosae is not the bacterial cells themselves, but rather the EVs they release. This difference may be partly explained by the superior stability and tissue penetration capacity of EVs, which can cross the intestinal barrier and deliver bioactive cargo to distant organs, whereas heat-killed bacteria lack active secretion and colonization capacity [30]. These features suggest that L. mucosae-EVs may have translational advantages as non-living postbiotics for gut-liver axis-based intervention. In the FFA-induced HepG2 cell model, the therapeutic effect of the bacterial supernatant disappeared after EVs removal, whereas purified EVs significantly reversed lipid accumulation and damage, supporting the central role of EVs as mediators. Additionally, in vivo distribution experiments showed that DiR-labeled EVs were mainly detected in gastrointestinal and hepatic tissues after oral administration, supporting a gut–liver axis-associated biodistribution route. Notably, DiR-labeled EVs were also detected in the spleen, suggesting a potential involvement of systemic immune regulation in addition to direct gut-liver axis-mediated hepatic effects [31]. Although the present study did not specifically investigate splenic immune responses, this observation provides an important direction for future studies. These findings provide new mechanistic insights into how probiotics exert their biological functions.

In terms of mechanistic exploration, RNA sequencing (RNA-seq) guided our focus to metabolic and inflammation-related pathways such as PI3K-Akt, JAK-STAT, MAPK, and AMPK. Following treatment with L. mucosae-EVs, the expression levels of core AMPK pathway-associated genes, including Acaca, Sirt1, and Ppargc1a, were markedly increased, whereas alterations in genes belonging to other signaling pathways remained limited. Western blotting further confirmed that EVs specifically promote the phosphorylation of AMPK. AMPK, as a central molecular hub regulating energy metabolism and lipid homeostasis, plays an important role in the prevention and treatment of MAFLD by inhibiting lipogenesis, promoting fatty acid β-oxidation, and alleviating inflammation and oxidative stress, which is highly consistent with the phenomena observed in our study [32, 33]. In the FFA-induced HepG2 and AML12 cell models, the beneficial effects of L. mucosae-EVs on lipid deposition, oxidative stress, and cellular damage were largely reversed by the AMPK-specific inhibitor Compound C. Similarly, in the high-fat diet-induced MAFLD mouse model, Compound C nearly blocked the beneficial effects of EVs on liver steatosis, glucose metabolism, fibrosis, inflammation, and oxidative stress. This body of evidence suggests that AMPK activation represents a major downstream mechanism through which L. mucosae-EVs exert their metabolic and hepatoprotective effects. Furthermore, this study links AMPK-associated metabolic regulation with the gut–liver axis-associated biodistribution of L. mucosae-EVs, offering both exploratory and novel insights.

A substantial body of evidence indicates that dysbiosis of the gut microbiota, together with impairment of the intestinal barrier, represents a pivotal driving force underlying both the initiation and progression of MAFLD [4–7]. Notably, the present study further identifies the intrinsic regulatory capability of L. mucosae-EVs in modulating the gut microbiota itself. Despite being derived from probiotics, EVs are capable of partially mimicking the functions of their parent strain and reshaping the gut microbiota in a reverse manner [34, 35]. This "bidirectional dialogue" phenomenon has only been gradually recognized in recent years. Our 16S rRNA sequencing data show that EVs intervention significantly improved the gut microbiota diversity in MAFLD mice and reversed the high-fat diet-induced dysbiosis. At the genus level, EVs specifically promoted the enrichment of Akkermansia. This genus is considered a representative next-generation probiotic and has been associated with improved intestinal barrier integrity, insulin sensitivity, and reduced liver fat accumulation [36, 37]. Additionally, the abundance of Lactobacillus itself showed an upward trend, suggesting that EVs may enhance the colonization advantage of the parent strain through a "self-feedback" mechanism. Notably, EVs significantly inhibited the overgrowth of pro-inflammatory genera such as Allobaculum, which correlates with the reduced serum LPS levels and the upregulation of tight junction proteins ZO-1, Occludin and Claudin-1 in the intestine [38, 39]. This finding expands the functional interpretation of EVs from direct host-cell regulation to a dual "microbiota-host" regulatory role, providing a new dimension for understanding the complexity of microbiome interventions. This synergistic model of gut-liver axis upstream barrier repair and downstream liver AMPK activation offers a rational mechanistic framework for explaining the multifaceted benefits of L. mucosae-EVs in the treatment of MAFLD.

Furthermore, our proteomic analysis of L. mucosae-EVs provided mechanistic insight into their bioactivity. DIA-based proteomics revealed that EVs are enriched in proteins involved in metabolic regulation, transport, and cellular functional processes, including peptidoglycan-binding protein LysM, ABC transporter-related proteins, lipoprotein, phosphoglycerate kinase, enolase, peptidoglycan hydrolase, and MucBP domain-containing proteins. These protein categories may contribute to EV–host cell interaction and metabolic regulation [40, 41]. Functional validation using Proteinase K-pretreated EVs in FFA-induced HepG2 cells demonstrated that disruption of EV-associated proteins attenuated their capacity to reduce intracellular TC and TG levels as well as ALT and AST release, suggesting that EV-associated proteins contribute, at least in part, to their protective effects. Together with AMPK inhibition and gut microbiota remodeling data, these findings support a refined model in which L. mucosae-EVs exert multifaceted therapeutic effects through protein-associated, multi-component regulation of lipid metabolism, hepatocyte function, and gut–liver axis communication.

Although this study provides systematic evidence supporting the therapeutic potential of L. mucosae-EVs in MAFLD, several limitations remain. First, the clinical sample size is limited and based on a single-center design, without stratified analysis according to different stages of MAFLD or the severity of common comorbidities such as diabetes and obesity. Second, we primarily confirmed the central role of the AMPK pathway using pharmacological inhibitors, but future studies using tissue-specific gene knockout animal models are needed for further validation. Additionally, while we observed that L. mucosae-EVs could reshape the gut microbiota and increase the abundance of beneficial bacteria such as Akkermansia, it remains unclear whether microbiota changes are an essential component of their protective effects. Further studies utilizing antibiotic depletion, fecal microbiota transplantation, or germ-free animal models are required to clarify this. Moreover, the contribution of additional metabolites, including short-chain fatty acids and bile acids, to the regulation of AMPK activation has yet to be comprehensively and systematically explored. Finally, although DIA-based proteomic analysis and Proteinase K pretreatment suggest that EV-associated proteins contribute to the biological activity of L. mucosae-EVs, the specific core proteins or peptides and their host targets remain unidentified. Current uptake and biodistribution data support gut–liver axis-associated tissue distribution but not receptor-mediated targeting. Therefore, future studies should validate candidate EV proteins and their host targets using binding assays, target-specific molecular simulations, and preclinical safety evaluation.

Conclusion

In conclusion, this study provides the first evidence elucidating both the therapeutic efficacy and the mechanistic basis by which L. mucosae alleviates MAFLD via its derived EVs. Specifically, L. mucosae-EVs improve lipid metabolic dysregulation, attenuate inflammation and oxidative stress, and inhibit hepatic fibrosis progression by reshaping the gut microbiota, restoring intestinal barrier integrity, and activating hepatic AMPK-associated signaling via the gut–liver axis, with EV-associated protein cargo contributing at least partly to the protective effects observed in hepatocytes. These results establish a solid experimental foundation and theoretical basis for the future development of precision microbiome-based therapeutic strategies for MAFLD based on L. mucosae and its EVs.

Abbreviations

ALT

Alanine aminotransferase

AMPK

AMP-activated protein kinase

AST

Aspartate aminotransferase

BMI

Body mass index

CAT

Catalase

CFU

Colony-forming units

DIA

Data-independent acquisition

EVs

Extracellular vesicles

FBG

Fasting blood glucose

FFA

Free fatty acids

FINS

Fasting insulin

GSH

Glutathione

HC

Healthy controls

HDL

High-density lipoprotein

HE

Hematoxylin and eosin

HOMA-IR

Homeostasis model assessment of insulin resistance

HOMA-IS

Homeostasis model assessment of insulin sensitivity

IL-1β

Interleukin-1β

IL-6

Interleukin-6

IPGTT

Intraperitoneal glucose tolerance test

IPITT

Intraperitoneal insulin tolerance test

KEGG

Kyoto encyclopedia of genes and genomes

LDL

Low-density lipoprotein

LPS

Lipopolysaccharide

MAFLD

Metabolic dysfunction-associated fatty liver disease

MDA

Malondialdehyde

MF

MAFLD patients

NTA

Nanoparticle tracking analysis

qPCR

Quantitative real-time PCR

RNA-seq

RNA sequencing

ROS

Reactive oxygen species

SOD

Superoxide dismutase

TC

Total cholesterol

TG

Triglycerides

TNF-α

Tumor necrosis factor-α

ZO-1

Zonula occludens-1

α-SMA

α-Smooth muscle actin

Author contributions

Kaige Zhang and Yuyu Zeng contributed equally as co-first authors. Tingtao Chen and Zhengyuan Xie are co-corresponding authors. Kaige Zhang: data curation and writing the original draft. Yuyu Zeng: Formal analysis and Visualization. Zhaohui Liao, Jianhua Wu and Xilong Zhang: Methodology and software. Tingtao Chen: Project administration and resources. Zhengyuan Xie: Conceptualization and supervision.

Funding

This work was supported by the National Natural Science Foundation of China Regional Program (Grants 82260131 and 82460128), the Key Program of Jiangxi Provincial Natural Science Foundation (Grant 20212ACB206017), and the Jiangxi Provincial Graduate Student Innovation Special Fund (Grant YC2025-B184).

Data availability

All raw sequencing data generated in this study, including human and mouse fecal 16S rRNA sequencing as well as liver RNA sequencing data, have been deposited in the NCBI SRA database in FASTQ format under the accession numbers PRJNA1365503, PRJNA1365674, and PRJNA1365667, respectively.

Declarations

Ethics approval and consent to participate

The study involving human participants was approved by the Biomedical Research Ethics Committee of the Second Affiliated Hospital of Nanchang University (Clinical Ethics No. O-Medical Research Ethics-2025–019), with written informed consent and adherence to the Declaration of Helsinki and WHO guidelines, while all animal procedures were approved and supervised by the Animal Ethics and Welfare Committee of Nanchang University (Animal Ethics No. NCULAE-20241023001).

Consent for publication

All authors have agreed to publish this manuscript.

Competing interests

The authors declare no competing interests.