The mechanical barrier, located beneath the mucus layer, is the most important factor affecting the selective permeability of the intestinal mucosa (17, 18). This barrier is composed of specialized epithelial cells, including goblet, Paneth, and M cells. The apical junction complex (AJC) is a critical structural component that consists of TJs and adherens junctions (AJs), the two primary modes of cellular connection (19, 20). TJ, the core regulatory element of the mechanical barrier, are composed of transmembrane proteins (claudins, occludin, and junctional adhesion molecules) and cytoplasmic scaffold proteins (ZO-1/2/3), which maintain barrier integrity through dynamic anchoring to the cytoskeleton (21–23). ZO-1 plays a central regulatory role in maintaining the structural stability of AJC through its interaction with cytoskeletal proteins (24). Studies have demonstrated that TJ protein expression levels are significantly negatively correlated with intestinal permeability, and that ZO-1 is an established biomarker for assessing TJ function (25). Under pathological conditions, the expression of ZO-1 is downregulated, resulting in endotoxin (such as LPS) translocation and activation of the TLR4/NF-κB pathway in liver Kupffer cells, driving liver inflammation and insulin resistance (26, 27).

Bacteroidetes and Firmicutes are the dominant phyla in the human gut microbiota. Actinobacteria and Proteobacteria cooperate to form multidimensional biological barrier (28). These symbiotic bacteria prevent the colonization of pathogenic bacteria, such as Salmonella, by secreting bacteriocins, producing SCFAs, and competitively deprressing pathogenic bacteria of essential nutrients. Butyric acid not only decreases intestinal pH and inhibits pathogen proliferation, but also activates the peroxisome proliferator-activated receptor gamma (PPARγ) signaling pathway in the intestinal epithelium, upregulates TJ protein expression, and strengthens the mechanical barrier (29). The gut microbiota also maintains immune tolerance to symbiotic bacteria by inducing regulatory T cell (Treg) differentiation and secretory IgA (sIgA) production through low-intensity interactions of model molecules (such as LPS and peptidoglycan) with intestinal epithelial TLR. Notably, Kupffer cells clear portal vein microbiota antigens and block systemic inflammation, while hepatic-originated complement components, such as complement component 5a (C5a), exert feedback regulation on the secretion of intestinal sIgA, forming closed-loop immune homeostasis (30, 31). In contrast, hepatic primary BAs (CA and CDCA) are transformed into secondary BAs (DCA and LCA) by the gut microbiota (e.g., Bacteroides and Clostridium), which enhance barrier function through the following mechanisms: 1) direct antibacterial effect (hydrophobic BAs destroy pathogen membrane structure) (32) and 2) signal activation (conjugated BAs activate intestinal epithelium FXR, promote antibacterial peptide [e.g., ANG4] expression, and mucus secretion) (33). Additionally, microbiota metabolites, such as tryptophan derivatives, are transported to the liver via the portal vein, regulating the FXR/FGF19 signaling axis, inhibiting lipid synthesis, and maintaining BA homeostasis (34).

The intestinal chemical defense system forms a cascade of antibacterial barriers through the synergistic action of gastric acid, BAs, digestive enzymes (including proteases and lipases), lysozymes, and antimicrobial peptides (AMPs) (35). Among them, gastric acid, trypsin, and lysozyme play a non-specific bactericidal role by destroying microbial cell walls (such as the outer membrane of gram-negative bacteria) or degrading antigenic proteins, thus maintaining the integrity and function of the gut barrier (36). BAs play a dual regulatory role in this system (37). On one hand, they act as physiological detergents, with hydrophobic BAs (e.g., DCA) disrupting the membrane structures of gram-positive bacteria at critical concentrations (> 2 mM). On the other hand, they act as signaling molecules, with conjugated BAs activating FXR/TGR5 receptors to upregulate TJ protein (occludin) expression and drive microbiota-mediated secondary BA conversion. In addition, AMPs (e.g., defensins and cathelicidins) disrupt pathogen membrane integrity through charge interactions while exhibiting selective tolerance to symbiotic bacteria, thereby maintaining barrier homeostasis (38).

The gut immune barrier is primarily composed of gut-associated lymphoid tissue (GALT) and sIgA. Among them, GALT sustains mucosal immune homeostasis through the “antigen sampling-immune education-effect output” cascade. The core mechanism is that micropleated cells (M cells) transport intestinal lumen antigens to Pyle’s node, and integrin alpha E (CD103)⁺ dendritic cells induce Treg differentiation through retinoic acid signaling, thus establishing immune tolerance to symbiotic bacteria (39). Furthermore, the gut-specific chemokine chemokine (C-C motif) ligand 25 (CCL25) directs plasmablasts to migrate to the lamina propria and differentiate into sIgA plasma cells, which produce 80% of intestinal immunoglobulins (40). Subsequently, sIgAs block pathogen adhesion and invasion by binding to fragment antigen-binding (Fab) fragments (such as flagellin). In addition, sIgA can limit the excessive proliferation of symbiotic bacteria through Fc alpha receptor (FcαR)-mediated encapsulation and maintain the spatial colonization stability of the microbiota (41). Clinical studies have shown that reduced sIgA levels lead to abnormal amplification and translocation of proteobacteria, such as Escherichia coli, which activate the TLR4/NF -κB pathway in Kupffer cells via the portal vein, driving hepatic inflammation and metabolic disorders (42). In addition to sIgA, intestinal epithelial cells release IgE, IgG, and various other immunoglobulins that play key roles in maintaining intestinal humoral immunity. IgG activates neutrophil extracellular trap formation (NETosis) via FcγR in areas of gut barrier injury, thereby exacerbating tissue inflammation (43) (Table 1). In summary, the gut–liver axis is a bidirectional metabolic–immune circuit primarily mediated by portal circulation and enterohepatic bile acid cycling. Its physiological “gatekeeping” relies on multilayer intestinal barriers (mechanical, biological, chemical, and immune) that constrain microbial translocation while preserving host–microbiota mutualism. Disruption of these barriers and bile acid/SCFA signaling shifts the hepatic microenvironment toward steatosis-prone metabolism and inflammation, providing a structural basis for MAFLD initiation and progression (Figure 1). Building on this barrier–mucosal immune framework, the spleen can function as a key “gain-control” organ in systemic immunity. When gut-derived antigenic load and inflammatory cues increase, splenic immune programming (e.g., T-cell polarization and monocyte/macrophage activation) can elevate the basal inflammatory tone and cytokine output. Consequently, splenic immune cells may be recruited to the liver, amplifying intrahepatic inflammation and fibrogenic signaling. This cross-organ circuitry supports the view that immune dysregulation in MAFLD/MASLD is not purely liver-centric but can emerge from coordinated gut–spleen–liver crosstalk.

Nutrition and dietary composition are key factors regulating the structure and function of gastrointestinal microbiota communities. Early cross-regional studies revealed that African children following a traditional high-fiber diet exhibited a 38% Prevotella/Enterobacteriaceae ratio. In contrast, Italian children on a Westernized diet had a higher Firmicutes/Enterobacteriaceae ratio, which was closely associated with the conversion of excess choline into trimethylamine (TMA) catalyzed by choline trimethylamine-lyase (CutC) and its activator CutD in Western diets and the subsequent production of Trimethylamine N-oxide (TMAO), a proinflammatory mediator. A modern refined diet (fiber < 15 g/day) decreases Prevotella abundance and SCFA production, directly impairing PPARα-mediated fatty acid oxidation (61). Similarly, a low-fiber diet reduces microbiota BA hydrolase activity, leading to an accumulation of primary BAs(e.g., CA), which inhibits Cytochrome P450 family 7 subfamily A member 1 (CYP7A1) expression by activating the FXR/SHP pathway in hepatocytes and exacerbating cholesterol metabolism imbalance (62, 63). HFD has profound effects on the composition of the gut microbiota, and remodeling of the gut microbiota occurs independently of obesity. In mice fed HFD for 8 weeks, the abundance of Proteobacteria increased, and that of SCFA-producing Roseburia decreased (64, 65). In addition, HFD induced the upregulation of microbiota choline metabolic pathway activity, which was accompanied by a simultaneous increase in serum TMAO levels (66). Transplantation of gut microbiota from HFD-fed donor mice into germ-free recipients recapitulated the phenotypes of hepatic steatosis, inflammation, and fibrosis, confirming that microbiota disorder is the core medium of diet-induced liver damage (8). In a cohort study of patients with advanced liver disease, dietary patterns rich in coffee (≥ 3 cups/day), whole grains (> 30%), and fermented dairy products (200 g per day) significantly increased intestinal microbial alpha diversity and reduced the risk of hospitalization (67). In further animal studies, a chronic high-fat high-cholesterol (HFHC) diet (> 14 months) induced liver steatosis (14 weeks), inflammation (20 weeks), fibrosis (36 weeks), and liver cancer (56 weeks) progression. Concurrently, the bile salt hydrolase (BSH) activity of gut microbiota was reduced, resulting in the inhibition of intestinal FXR signaling and intensification of lipid accumulation in the liver. Thus, the evolution of liver steatosis and liver inflammation in MAFLD has evolved into a classic example of diet-induced inflammation, suggesting that excess nutrients can spread disease along the “gut-liver axis.”

The biological clock is the core regulatory system that enables organisms to adapt to day–night environments, and its interaction with the gut microbiota has been gradually clarified. Studies have demonstrated that approximately 20% of the gut microbiota exhibit circadian fluctuations in abundance and metabolic function, and experimental evidence supports that these oscillations are strongly shaped by host behavioral timing (especially feeding schedules), rather than implying a universal direct regulatory mechanism. The gut microbiota prioritizes energy metabolism and DNA replication during the daytime and shifts to detoxification pathways and SOS response-mediated DNA repair at night. Moreover, the feeding rhythms of the host not only drive diurnal reprogramming of microbiota functions but also shape dynamic fluctuations in microbiota composition. Together, these observations suggest that feeding cycles act as major zeitgebers (timing cues) for the microbial rhythmicity (68, 69).

It is worth noting that defects in core clock genes in animal models are associated with an imbalance in microbiota homeostasis. Mutations in the circadian locomotor output cycles kaput (Clock) have been reported to be accompanied by the proliferation of proinflammatory Enterobacteriaceae, and deletions of circadian rhythm target genes Per 1/2 or brain and muscle ARNT-like protein 1 (BMAL1) disrupt microbial rhythmicity and are linked to gut barrier dysfunction (70). These findings indicate that the biological clock system sustains symbiotic equilibrium between the microbiota and host through a dual mechanism involving genetic regulation and environmental cues, such as photoperiod and feeding time. Furthermore, its disruption may contribute to the pathogenesis of metabolic diseases by altering microbial rhythms, although its causal relevance in humans requires interventional confirmation.

In addition, in a germ-free mouse experiment, transplantation of gut microbiota from trans-time travelers was sufficient to induce obesity and impaired glucose tolerance in recipient animals (71, 72), supporting a causal contribution of altered microbiota configurations within this experimental setting. Furthermore, although HFD disrupts the circadian rhythm of the microbiota and induces obesity and metabolic syndrome, limiting the time window of HFD intake (time-restricted feeding in mice) can restore glucose metabolism homeostasis and suppress weight gain (73, 74). This protective effect has been associated with a partial restoration of gut microbiota rhythmicity, manifested by a decrease in the abundance of obesity-associated bacteria and an increase in the functional activity of beneficial taxa, such as Oscillibacter, Ruminococcus, and SCFA-producing strains (75). Further studies have revealed that HFD intake with circadian rhythm disruption in experimental models exacerbates gut microbiota dysregulation, resulting in impaired gut barrier function and endotoxin entry into the bloodstream, which is linked to systemic inflammation and may accelerate metabolic syndrome progression (76). These findings suggest that regulating microbial circadian rhythms is a mechanistically plausible intervention avenue supported by animal studies; however, translation to humans will require well-designed clinical trials with time-resolved microbiome and metabolic endpoints.

Smoking, a global public health problem, is closely associated with the onset and progression of MAFLD. Large-scale population-based studies have shown that smoking is an independent risk factor for MAFLD, and its harmful effect are dose-dependent (77, 78). A cross-sectional analysis of 6,852 patients with MAFLD showed that those who smoked ≥ 20 cigarettes per day had a significantly higher risk of liver fibrosis than light smokers (< 10 cigarettes per day), with an increased risk of fibrosis of 30% per 10 years of smoking history. A multicenter cohort study from 16 centers in the United States further confirmed that smokers had a significantly higher risk of liver steatosis, lobular inflammation, and cirrhosis progression over 5 years than nonsmokers (79). In addition, there are sex differences in the effects of smoking on MAFLD. A multicenter cohort study conducted in China demonstrated that female smokers have a 1.7-fold higher risk of developing MASH than male smokers, and that liver fibrosis progresses faster among females (80). Recent studies have revealed that gut microbiota disorders are central to the mechanism by which smoking exacerbates MAFLD. Nicotine and other tobacco smoke components can directly interact with the gut microbiota, remodeling its structure. For example, the abundance of proinflammatory bacteria (such as Prevotella and Veillonella) in the intestinal tract of smokers increases significantly, whereas the proportion of beneficial bacteria, such as Firmicutes, decreases. Inhibition of gut colonization by SCFA-producing bacteria (such as Bifidobacterium) decreases SCFA levels in the cecum and increases intestinal pH, thus destroying the integrity of the gut barrier (81). Smoking also significantly increases the proportion of Gram-negative bacteria such as Proteobacteria, resulting in increased LPS production (82). LPS is transferred to the liver through the portal vein, binds to TLR4 on Kupffer cell surfaces, activates the NF-κB pathway, drives the release of proinflammatory factors such as TNF-α and IL-6, and triggers insulin resistance and excessive lipid deposition in hepatocytes (83, 84). Animal models have provided direct evidence for this mechanism. SCFA levels in the cecum of rats exposed to cigarette smoke for 4 weeks were reduced, and antimicrobial peptide expression was suppressed. Additionally, in the intestines of mice treated with cigarette smoke condensate, the abundance of pathogenic bacteria, such as Erysipelothrix, is increased, accompanied by increased serum LPS levels and increased liver steatosis (85–88). In conclusion, smoking may affect the composition of the gut microbiota by changing the intestinal microenvironment (such as excessive proliferation of pathogenic bacteria), thereby regulating the host immune inflammatory response and SCFA production, ultimately leading to MAFLD progression.

Particulate matter with a diameter of less than 2.5 micrometers (PM2.5), a major air pollutant in China, has attracted considerable attention owing to its extensive exposure and multi-organ toxicity. Recent studies have found that PM2.5 not only directly damages the respiratory system but also becomes a new environmental driver of MAFLD progression by reshaping intestinal microecology, disrupting the intestinal mucosal barrier, and triggering systemic inflammatory responses. In animal experiments, short-term exposure (7 days) can cause significant spatial heterogeneity in the gastrointestinal microbiota of mice. β-diversity increases from the proximal (stomach) to the distal (colon) regions, suggesting that PM2.5 exerts a stronger selective pressure on the distal intestine (89). Notably, prolonged exposure (12 weeks) did not significantly reduce the α-diversity of the microbiota, but resulted in a decrease in the relative abundance of Firmicutes and a 2-fold increase in Proteobacteria, resulting in the development of a proinflammatory microbiota phenotype (90). Further studies have shown that PM2.5 reduces cecal butyrate levels by inhibiting the growth of SCFA-producing bacteria (such as Roseburia and Faecalibacterium) and impairs its regulatory effect on liver lipid metabolism (91). In addition to dysbiosis, PM2.5 impairs gut barrier integrity via a dual pathway. PM2.5 downregulates the mRNA expression of the TJ protein ZO-1 and occludin in intestinal epithelial cells and induces a burst of mitochondrial reactive oxygen species (ROS) burst (92). ROS-mediated apoptosis: Excessive ROS activate the JNK/c-Jun pathway, triggering apoptosis of colonic epithelial cells and causing a mucosal mechanical barrier breach. These changes cause metabolic endotoxemia-gut microbiota disorder, promote increased LPS production and translocation to the portal venous circulation through the damaged barrier, activate the TLR4/NF-κB pathway in Kupffer cells, and increase the release of proinflammatory factors (TNF-α and IL-6) (93).

Proton pump inhibitors (PPIs), the most widely used antacids worldwide, have been associated with gut microbiota disorders in multi-omics studies. A large population cohort study in the Netherlands revealed reduced α-diversities in the gut microbiota of PPI users and an abundance shift (94) in nearly one-fifth of the microbiota taxa. Metagenomic analysis further confirmed that PPIs significantly altered the activity of 133 metabolic pathways involved in the functional remodeling of lipid synthesis, Nicotinamide Adenine Dinucleotide (NAD +) fermentation, and purine degradation. This phenomenon of “oral-intestinal microbiota homology” results in the loss of integrity of the intestinal mechanical barrier and increased translocation of endotoxins (e.g., LPS) (95). The release of LPS into the blood drives the secretion of hepatic inflammatory factors (TNF-α and IL-6) by activating the TLR4/NF-κB pathway in Kupffer cells and inhibiting insulin receptor substrate phosphorylation, thereby exacerbating hepatocyte steatosis and MAFLD progression (96). In addition, PPIs induce microbiota imbalance and barrier damage to form a vicious circle, eventually promoting MAFLD progression to the fibrosis stage. This finding suggests that the risks associated with long-term PPI use should be considered in clinical practice.

Metformin is a first-line Type 2 diabetes mellitus (T2DM) drug, and its hypoglycemic effect is partially achieved through microbiota-host interaction. Intervention studies have shown that metformin specifically increases the abundance of SCFA-producing bacteria while inhibiting inflammation-associated enterobacteriaceae (97). Transplantation experiments conducted in germ-free mice demonstrated that microbiota reshaped by metformin treatment led to a reduction in blood glucose levels and an enhancement in glucagon-like peptide-1(GLP-1) secretion (98). However, this repopulation also resulted in adverse reactions, such as diarrhea/abdominal distension, which may be related to the increased intestinal gas production caused by the hyperproliferation of Methanobrevibacterium (99). Studies suggest that metformin’s modulation of the gut microbiota may have potential benefits for patients with MAFLD by improving gut microbiota structure and reducing inflammation levels, thereby reducing liver steatosis and inflammatory response (100).

The traditional “sterile uterus” hypothesis has been overturned, and the latest metagenomic studies have confirmed that human fetal intestinal colonization begins in utero. Low biomass but highly specific microbiota communities (such as Lactobacillus and Bifidobacterium) were detected in the placenta and amniotic fluid, with 16S rRNA gene abundance of 10^{^2}–10^{^3} copies/g tissue (101). In mouse models, orally administered fluorescently labeled Enterococcus faecium can cross the placental barrier via maternal circulation and colonize the fetal intestine, and the detection rate of labeled bacteria in the meconium microbiota is approximately 70% (102). This finding indicates a potential pathway for vertical maternal-fetal microbial transmission. The mode of delivery is a decisive factor for the initial microbiota of newborns. Lactobacillus crispus and Prevotella bivia from the maternal vagina are the main gut microbiota of newborns, and they are metabolically characterized by SCFAs synthesis pathway enrichment (103). The microbiota may also be derived from maternal skin bacteria (such as Streptococcus and Cutibacteria), which is accompanied by an increase in LPS biosynthesis pathway activity (10). These primary microbiota evolve to become more diverse and relatively stable. At 3 years of age, the gut microbiota of children becomes similar to that of adults. Early colonization and composition of the gut microbiota have profound effects on the host’s metabolic health. Studies have shown that the composition of the early gut microbiota is closely related to the metabolic state of the host, especially with the onset and development of MAFLD (104). For instance, microbiota disruption leads to the downregulation of occludin expression, an increase in intestinal permeability, enhanced endotoxin (LPS) translocation, and activation of the TLR4/NF-κB pathway in hepatic Kupffer cells, thereby promoting TNF-α secretion and hepatic fat deposition (6). In addition, certain bacterial species (e.g., Akkermansia muciniphila) regulate BA metabolism by inducing FXR signaling and inhibiting hepatocyte lipid synthesis, whereas Clostridium-derived secondary BAs enhance white fat browning (105) via the TGR5-cAMP-PKA pathway (106, 107). Thus, colonization and composition of the early gut microbiota may provide new insights into the prevention and treatment of MAFLD.

The global aging population is accelerating, and the population aged 65 and over is projected to exceed 20% (approximately 1.5 billion) by 2050. Among these, 80% will be in low- and middle-income countries, and the prevalence of MAFLD in this population is expected to increase significantly (108). Aging drives gut-liver axis imbalance by remodeling the structure and function of the gut microbiota. The α-diversity of the gut microbiota in older adults is significantly lower than that in younger individuals and is imbalanced, characterized by the expansion of pathogens and the depletion of symbiotic bacteria (109, 110). This disorder results in a decrease in SCFA synthesis and a concomitant decrease in 7α-dehydroxylase activity, leading to the accumulation of unconjugated BAs (DCA and LCA), which, in turn, inhibits ileal FXR signaling (111). Gut barrier injury increases LPS levels in the portal vein, inhibits insulin receptor phosphorylation through activation of the hepatic TLR4/MyD88 pathway, and stimulates TGF-β1 secretion by hepatic stellate cells. Additionally, accumulated hydrophobic BAs induce mitochondrial ROS bursts and inhibit CPT1 α-mediated fatty acid oxidation (26). These cascades ultimately lead to hepatocellular steatosis, inflammation, and fibrosis, which constitute the core pathological mechanisms of aging-related MAFLD.

Host epigenetic programs dynamically interact with gut microflora through multi-level regulation. At the histone modification level, histone deacetylase 3 (HDAC3) maintains microbiota homeostasis through spatiotemporal regulation of intestinal epithelial antibacterial programs. Gut-specific HDAC3 knockout mice have decreased ileal α-defensin mRNA expression, resulting in increased proteobacteria abundance, while the microbiota metabolite butyric acid promotes Treg differentiation by inhibiting HDAC3 activity (112, 113). Silent mating type Information Regulation2 homolog 1(SIRT1) deficiency exacerbates fatty diet-induced hepatic lipid deposition (114, 115) by decreasing the ratio of secondary to primary BAs and inhibiting ileal FXR signaling. Chromatin remodeling mechanisms further drive the coevolution of the gut barrier and microbiota. Chromodomain Helicase DNA Binding protein 1(CHD1) deletion leads to a reduced expression of the TJ protein Claudin-3, and intestinal permeability is positively correlated with γ-proteobacteria abundance (116). Lysine-Specific Demethylase 5(KDM5/JARID1) regulates ZO-1 transcription through histone H3 lysine 4 trimethylation(H3K4me3) modification, and its intestinal epithelial-specific knockout increases LPS levels in the portal vein and activates the TLR4/MyD88 pathway in the liver (117). Non-coding RNA mediate precise regulation across species. Host miR-122-5p targets E. coli through exosomes, inhibits flagellin synthesis, and reduces bacterial biofilm formation capacity (118). Clinical cohorts have shown a significant negative correlation between fecal miR-21-5p levels and Enterobacteriaceae abundance (119) in patients with MAFLD. The intergenerational effects of environmental stress highlight the importance of epigenetic memories. Dexamethasone exposure during pregnancy induces Cytosine-phosphate-Guanine (CpG) island hypermethylation in the promoter region of the solute carrier family 5 member 8 (SLC5A8/SMCT1) gene in offspring mice, resulting in decreased SCFAs transporter expression, accompanied by an imbalance in the Bacteroides/Firmicutes ratio and an increased risk of liver steatosis in adulthood (120). These findings reveal the pivotal role of epigenetic regulation in the gut-liver axis, providing new targets for intervention in metabolic liver disease.

Genome-wide association studies (GWAS) have revealed that carriers of the PNPLA3 rs738409 risk allele have an increased abundance of Prevotella in their colonic contents and an encoded alcohol dehydrogenase that catalyzes endogenous ethanol production and promotes lipid droplet deposition by inhibiting hepatocyte mitochondrial complex I activity (121, 122). Similarly, the TM6SF2 rs58542926 T allele results in a decreased abundance of the Bacteroides phylum, decreased synthesis of the secondary BA ursodeoxycholic acid (UDCA), impaired FXR-mediated carnitine palmitoyltransferase 1(CPT1) α expression, and activation of the NACHT, LRR, and PYD domains-containing protein 3 (NLRP3) inflammasome (123).

Gut-derived Tregs form immune homeostasis loops in response to microbiota metabolites. Butyric acid (≥ 50 μM) activates the G protein-coupled receptor 109A (GPR109A/HM74A) receptor on Tregs, induces C-C chemokine receptor type 9 (CCR9) expression, drives Treg migration to the lamina propria of the colon, and promotes IL-10 secretion (124–126) through histone H3 lysine 27 acetylation(H3K27ac) modification. This process inhibits hepatic CD8+ T cell proliferation and TNF-α release. Simultaneous mechanistic studies suggest that butyric acid upregulates forkhead box protein P3(Foxp3) expression by inhibiting HDAC3 activity in liver macrophages and blocks TLR4/NF-κB pathway overactivation (127) through suppressor of cytokine signaling 3(SOCS3)-mediated ubiquitination degradation. This regulatory axis is impaired in patients with MAFLD. Serum butyrate levels < 15 μM are associated with a decreased hepatic Treg/Teff ratio and are significantly negatively correlated with the lobular inflammation score and fibrosis stage (128, 129)(Figure 3).In summary, MAFLD-relevant dysbiosis is shaped by a convergence of exogenous exposures (dietary patterns, circadian disruption, smoking, air pollution, and medications) and endogenous host determinants (early life microbial seeding, aging, epigenetic programs, and genetic–immune context). Although these factors differ in timing and intensity, they tend to converge on common downstream consequences, including reduced beneficial metabolite output, impaired barrier integrity, and heightened portal inflammatory signaling. Conceptualizing risk factors through shared microbiota-barrier-liver nodes helps explain clinical heterogeneity and identifies modifiable entry points for prevention.

Serum LPS levels are significantly elevated in patients with MAFLD, driving disease progression through a dual mechanism. First, LPS directly activates intrahepatic inflammation in the liver. It binds with high affinity to the toll-like receptor 4–myeloid differentiation factor 2 (TLR4-MD2) complex on the surface of hepatocytes and Kupffer cells, triggers myeloid differentiation primary response 88(MyD88) -dependent signaling pathways, induces IL-6 and IL-1β secretion, and directly leads to hepatocyte apoptosis and MASH deterioration (130). Serum co-receptor lipopolysaccharide-binding protein (LBP) levels are abnormally elevated in patients with MAFLD, and LBP gene knockout significantly improves liver lipid deposition and insulin resistance (131, 132). Additionally, LPS activates the NLRP3 inflammatory body through the TLR4/TRIF pathway, promotes caspase-1 self-cleavage and release of the gasdermin D N-terminal domain (GSDMD-N), drives mature IL-1β production, and intensifies the intrahepatic inflammatory cascade (133). Second, LPS indirectly aggravates liver injury by disrupting gut-liver axis homeostasis, activating the intestinal TLR4/MyD88 signaling axis, and upregulating myosin light chain kinase, resulting in decreased expression of the TJ protein occludin and significantly increasing intestinal permeability (134). Increased LPS levels in the portal vein aggravate the disease through a dual pathway: 1) activation of the NF-κB pathway in Kupffer cells, promotion of TGF-β1 secretion, induction of hepatic stellate cell activation, and collagen deposition (135); and 2) inhibition of Peroxisome Proliferator-Activated Receptor Alpha(PPARα) signaling in hepatocytes, reduction of fatty acid β-oxidation, and upregulation of sterol regulatory element-binding protein 1c(SREBP-1c143), a key factor in lipid generation. Eventually, a self-reinforcing vicious cycle develops between the gut and liver. Intrahepatic inflammatory factors (e.g., TNF-α) inhibit intestinal epithelial repair through portal feedback, resulting in a significant positive correlation between the abundance of LPS-producing proteobacteria in the gut and liver fibrosis stage (136). This cycle of “intestinal leakage-LPS into the liver-liver damage” is the core driving force for MAFLD progression to end-stage liver disease, providing a theoretical basis for precise treatment targeting the LPS/TLR4 pathway.

Peptidoglycans (PGN) possess immune recognition properties owing to their structural heterogeneity. The cell walls of gram-positive bacteria are approximately 50–100 nm thick, and the lipoteichoic acid in their cell walls is crosslinked with PGN to form a highly crosslinked three-dimensional network structure. In contrast, the cell walls of gram-negative bacteria (3–8 nm thin layer) are loosely anchored to the outer membrane (137, 138) by the Braun lipoprotein. This structural difference results in a higher PGN release efficiency in gram-positive bacteria than in gram-negative bacteria and significantly affects the activation threshold of the host pattern recognition receptor (139). In the HFD model, decreased fasting glucose and triglyceride (TG) deposition in TLR2^-/-mice were directly correlated with a decreased abundance of PGN-producing Lactobacillus (140). However, in the methionine- and choline-deficiency (MCD) model, isogenic defects resulted in increased liver inflammation scores, associated with increased PGN release and decreased levels of short-chain fatty acids (SCFAs) in Enterobacteriaceae, suggesting that dietary intervention dynamically modulates TLR2 signaling sensitivity (141) by reshaping the metabolic profile of the microbiota. PGN activates nucleotide-binding oligomerization domain-containing protein 1/2(NOD1/2) receptors with high affinity, synergistically triggering the NF-κB and MAPK pathways to drive a proinflammatory factor storm. NOD1/2 double-knockout mice have significantly decreased TG content and inflammatory cell infiltration area, confirming the key role of this pathway in lipid metabolism disorders (142, 143).

When microbial (bacterial) DNA is internalized by immune cells and trafficked to endosomal compartments, its unmethylated CpG motifs (5′-CpG-3′) can be recognized by toll-like receptor 9 (TLR9, an endosomal DNA-sensing PRR), initiating downstream signaling. The continued activation of extracellular signal-regulated kinase 1/2 (ERK1/2) and c-Jun N-terminal kinase (JNK) in the MAPK pathway has been reported in experimental systems to promote macrophage secretion of TNF-α and IL-12, while the NF-κB pathway can synergistically facilitate NLRP3 inflammasome assembly by accelerating inhibitor of kappa B alpha (IκBα) degradation and enhancing nuclear factor kappa B subunit p65 (p65) nuclear translocation, thereby providing mechanistic support for inflammation that may contribute to progression from MAFLD to MASH in animal models (144, 145). Notably, in preclinical studies, TLR9 inhibits interferon regulatory factor 7 (IRF7) phosphorylation via the IκB kinase α (IKKα)–microtubule-associated protein 1A/1B-light chain 3 (LC3) complex, reducing type I interferon synthesis and forming a “proinflammatoryanti-interferon” imbalance that may exacerbate metabolic inflammation by blunting antiviral responses (146).

In the methionine–choline deficient diet (MCD) feeding model, activation of the TLR9–MyD88 axis in Kupffer cells is associated with increased IL-1β secretion and has been linked to hepatocyte injury markers such as LDH release through caspase-1–dependent inflammatory cell death signaling in this model (147). Consistent with a functional role in mice, TLR9^-/- mice show reduced steatosis area, portal vein fibrosis score, and hepatic hydroxyproline content, supporting that TLR9–MyD88 signaling is an important contributor to MASH-like disease severity in this experimental setting, rather than establishing it as the sole driver in humans (148).

Bacterial DNA may also contribute to liver injury through broader immune reprogramming. Abnormal B cell activation promotes immunoglobulin G subclass 2c (IgG2c) secretion, leading to immune complex deposition and activation of the complement C3a–C5a axis. NK cells can induce hepatocyte apoptosis through perforin and Fas ligand/Fas receptor (Fas/FasL) signaling and, together with interferon-gamma (IFN-γ) secretion, may amplify inflammatory cascades (149, 150). In an early translational study, the TLR9 antagonist oligodeoxynucleotide (ODN) TTAGGG (5 mg/kg) reduced serum alanine aminotransferase (ALT) levels and improved NAS scores after four weeks of treatment; however, in patients, plasma bacterial DNA levels were reported to be strongly correlated with fibrosis stage, which supports clinical association but does not establish causality (151).

Outer membrane vesicles (OMVs) are nanosized bilayer membrane vesicles with a diameter of 20–300 nm secreted by gram-negative (such as Escherichia coli) and some gram-positive bacteria. OMVs completely encapsulate LPS, virulent proteins, and circular DNA. OMVs deliver LPS directly to the cytoplasm of host cells through a membrane fusion-mediated Trojan horse mechanism, activating caspase-11-dependent non-classical inflammatory bodies and driving increased IL-1β secretion (152). In the gut microenvironment, OMVs are activated via a dual receptor coactivation pattern: MyD88-dependent signaling triggered by the TLR4-MD2 complex in intestinal epithelial cells and activation of NLRP3 inflammatory bodies in macrophages, resulting in the spatiotemporal co-release of TNF-α and IL-6 and an “inflammatory storm” (153, 154). This multi-target effect makes OMVs critical “inflammation amplifiers” for disrupting the enterohepatic axis in MAFLD. Clinical cohort studies have demonstrated a positive correlation between fecal OMV concentrations and serum ALT levels in patients with MAFLD, and liver tissue OMVs are highly localized within fibrotic regions (155, 156). At present, the molecular pathway by which OMVs affect host-microbiota interactions by regulating gut barrier TJ proteins and BA metabolism requires systematic analysis. Nano-intervention strategies based on OMV component editing may provide new ideas for remodeling the gut-liver immune homeostasis (Figure 4).

SCFAs (acetate, propionate, and butyrate) are the core metabolites of dietary fiber that are fermented by the gut microbiota. Approximately 95% of SCFAs enter the liver metabolic pool after absorption through the portal vein, 30–40% of which are converted into glycogen precursors, 10% are catalyzed by acyl-coA synthetase short-chain family member 2(ACSS2) to generate acetyl coenzyme A to participate in lipid neogenesis, and the remaining act on cells via G protein-coupled receptor 41/43(GPR41/43). Gut-brain-liver tripartite dialogue: the activation of intestinal L cells increases GLP-1 secretion and prolongs peptide YY(PYY) half-life, regulating food intake by inhibiting ghrelin (164). However, an abnormally high expression of GLP-1/PYY in the HFD model increases the risk of hepatic lipid accumulation via the activation of agouti-related protein/neuropeptide Y(AgRP/NPY) neurons in the hypothalamus (165). SCFAs exhibit concentration-dependent metabolic reprogramming in the liver. Physiological concentrations (50–200 μM) of propionate inhibit acetyl-CoA carboxylase (ACC) activity through AMP-activated protein kinase α subunit threonine 172(AMPKα Thr172) phosphorylation, decrease malonyl-CoA levels, and block fatty acid elongation. Supraphysiological concentrations (> 350 μM) of acetate drive the upregulation of fatty acid synthase (FASN) expression and promote lipid droplet formation (166, 167). Butyrate acts as an HDAC inhibitor and increases the acetylation of the adiponectin gene H3K27. Simultaneously, apparent silencing is achieved by methylation of the CpG island in the resistin promoter region, synergistically improving insulin sensitivity (168). Clinical cohorts have shown a decrease in total SCFAs in the portal vein of patients with MAFLD and a negative correlation between the butyrate ratio and the degree of hepatic steatosis, suggesting the need to develop strategies to precisely regulate SCFA composition/concentration based on individualized metabolic phenotypes (169).

BAs convert cholesterol to primary BAs (CA/CDCA) via classical (CYP7A1) and alternative (CYP27A1/CYP7B1) pathways, which are secreted into the gut via glycine and taurine conjugation reactions. The gut microbiota converts primary BAs to secondary BAs (DCA/LCA) using BSH and 7α-dehydroxylase and maintains enterohepatic circulation 8–10 times a day via the apical sodium-dependent bile acid transporter (ASBT) (170). Patients with MAFLD have increased total serum BA concentrations, but the proportion of secondary BAs decreases significantly, with toxic taurine-bound CDCA accumulation (171). Secondary BAs drive pathological processes by abnormally regulating the FXR signaling axis. A reduced ability of hepatocyte CDCA to activate FXR leads to a failure of CYP7A1 negative feedback inhibition, triggering abnormal CA accumulation and inducing apoptosis through a mitochondrial cytochrome P450 family 2 subfamily E member 1(CYP2E1)-dependent ROS burst (172). INT-777 (10 μM), an activator of TGR5, inhibits the assembly of NLRP3 inflammatory bodies, decreases the polarization ratio of classically activated macrophages(M1)-type macrophages, and enhances the expression of bile salt export pump (BSEP), a key protein in biliary secretion, thus alleviating cholestatic injury (173, 174). Clinical cohorts showed a negative correlation between DCA/CA ratios and liver controlled attenuation parameter (CAP) values, while treatment with the FXR agonist obeticholic acid (OCA 25 mg/d) for 24 weeks reduced NAS scores, suggesting a new intervention strategy targeting the metabolic remodeling of BA or MAFLD (175).

TMAO is converted from dietary choline/L-carnitine to TMA by gut microbiota (such as Anaerococcus and Clostridium) through the CutC/D enzyme, which is oxidized by flavin-containing monooxygenase 3(FMO3) after entering the liver through the portal vein. Clinical studies have shown that patients with MAFLD have elevated serum TMAO levels compared to healthy individuals and a geometrically increased risk of advanced liver fibrosis (176). Animal experiments further confirmed that HFD-enriched CutC-positive microbiota increased intestinal choline-TMA conversion efficiency, decreased liver phosphatidylcholine bioavailability, aggravated very low-density lipoprotein (VLDL) secretion disorders, and increased hepatic lipid accumulation (177). However, the mechanism underlying the relationship between TMAO and the onset and development of MAFLD remains unclear. TMAO accelerated the degradation of insulin-induced gene 2(Insig-2) and increased the nuclear translocation efficiency of SREBP-1c through the ubiquitin-proteasome system. Concurrently, TMAO induced enhancer of zeste homolog 2(EZH2) to mediate H3K27me3 modification of the PPARα promoter region, inhibit CPT1A transcription, and lead to a decrease in the mitochondrial fatty acid oxidation rate (178). Further studies have shown that adipose tissue macrophages are the key targets of TMAO’s proinflammatory effect of TMAO. TMAO promotes the conversion of adipose tissue macrophages to the proinflammatory M1 phenotype through the STAT3 Tyr705 site and inhibits the expression of the alternatively activated macrophage (M2) marker arginase 1(Arg1), resulting in the formation of a chronic inflammatory microenvironment (179, 180). Furthermore, TMAO disrupts BA homeostasis by antagonizing FXR signaling. In vitro experiments have shown that 50 μM TMAO reduces FXR binding to the BSEP promoter of the BA efflux pump, resulting in cholestasis in the liver (181). The corresponding DCA further inhibits FXR signaling and promotes lipid regeneration by activating TGR5 receptors on the hepatocyte membrane, forming a self-reinforcing loop of “TMAO→FXR inhibition →DCA accumulation → lipid toxicity exacerbation” (182). In hepatic stellate cells, TMAO promotes IL-1β secretion by activating NLRP3 inflammatory bodies and indirectly stimulates collagen deposition in the liver. TMAO in hepatic sinusoidal endothelial cells downregulates the TJ protein Claudin-5 expression, increases vascular permeability, and promotes lipid particle transendothelial transport (183). This cell-type-specific regulation suggests that TMAO may drive both liver inflammation and abnormal lipid distribution in the liver. However, the specific molecular targets of this compound require further investigation.

Choline, an essential nutrient, is exogenously derived from the diet and endogenously synthesized by phosphatidylethanolamine N-methyltransferase (PEMT). It completes the lipidation of apolipoprotein B-100 by generating phosphatidylcholine (PC), which drives the assembly of VLDL particles (184). A choline-deficient diet reduced hepatic PC synthesis to approximately one-third of that in the control group, resulting in a decrease in the VLDL secretion rate and accumulation of TG in the liver. This model has been widely used for MASH mechanism analysis (185). In patients with MAFLD, the abundance of Bacteroides increases, CutC enzyme activity increases the conversion efficiency of dietary choline to TMA, and TMA concentration in the portal vein significantly increases. TMAO reduces endogenous choline synthesis by half in the healthy control group by inhibiting H3K4me3 modification of the PEMT promoter region, with a double effect of “exogenous entrapment and endogenous inhibition” (186). Clinical cohort studies have demonstrated that patients with MAFLD exhibit decreased free plasma choline levels, elevated serum TMAO concentrations, and abnormal metabolic axes that drive lipid neogenesis and inflammatory cascades (187) through the activation of SREBP-1c and NLRP3 inflammatory bodies. Glycerophosphocholine supplementation (2 g/d) for 12 weeks reduced liver fat content and improved NAS scores, suggesting that the precise regulation of choline metabolism is a potential therapeutic approach or a novel intervention strategy for MAFLD (188).

Endogenous ethanol is mainly produced by the gut microbiota (such as Escherichia coli and Klebsiella pneumoniae) through pentose phosphate metabolism. Clinical studies have shown that ethanol-producing bacteria are more abundant in patients with MASH than in healthy individuals, and the concentration of ethanol in peripheral blood is significantly positively correlated with the liver fat fraction. Single-cell transcriptome analysis further confirmed that alcohol dehydrogenase 1B(ADH1B) and aldehyde dehydrogenase 2(ALDH2), key enzymes of ethanol metabolism, were abnormally upregulated in MASH livers, suggesting a surge in the liver metabolic load (189, 190). In addition, the signaling pathway related to ethanol metabolism in the liver showed abnormal activation. From the perspective of liver metabolism, ethanol exposure can aggravate lipid deposition via dual mechanisms. Ethanol promotes lipid neogenesis by activating liver SREBP-1c and its downstream target genes, ACCα and FASN. Conversely, an increased NADH/NAD+ ratio inhibits PPARα transcription activity, resulting in a decreased mitochondrial β-oxidation rate (191, 192). Furthermore, CYP2E1 catalyzes the oxidation of ethanol to produce large amounts of acetaldehyde and ROS, directly damaging hepatocyte DNA. In addition, ROS promote the secretion of IL-1β and IL-18 by activating NLRP3 inflammatory bodies, initiating neutrophil infiltration and the activation of fibrosis precursor cells, and further aggravating liver inflammatory infiltration and fibrosis (86). At the gut microenvironment level, ethanol and its metabolites (especially acetaldehyde) can destroy intestinal epithelial TJ, resulting in impaired gut barrier function. For example, acetaldehyde (50 μM exposure for 24 hours) induced ubiquitination degradation of intestinal epithelial TJ proteins (occludin and claudin-1) and increased intestinal permeability.LPS enters the liver via the portal vein, activates the TLR4/MyD88 pathway in Kupffer cells, and upregulates TNF-α and IL-6 expression. These inflammatory factors can further aggravate cholestasis by inhibiting FXR signaling in hepatocytes and accelerating MAFLD progression.

Kupffer cells inhibit FXR signaling, leading to cholestasis (193). A clinical probiotic (Vivomixx^®/Visbiome^® Strains Liquid [VSL #3]) intervention for 12 weeks reduced the abundance of ethanol-producing bacteria and the NAS score of the liver, proposing a novel direction for targeted microbiome-metabolic axis or MASH treatment (194, 195).

The gut barrier maintains host-microbiota interface homeostasis through TJ proteins, ZO-1 and occludin. Fecal zonulin concentrations are elevated in patients with MAFLD, suggesting early intestinal leakage (196). Animal models have demonstrated that gut barrier damage results in portal endotoxin and bacterial DNA translocation, thereby activating Kupffer cells. The TLR4/MyD88 pathway drives NF-κB nuclear translocation and NLRP3 inflammasome assembly, resulting in increased IL-1β secretion (197, 198).

Concentration-dependent bidirectional regulation of microbiota metabolites: secondary BAs (DCA, 10 μM) inhibit SREBP-1c expression by activating farnesoid X receptor (FXR). SCFAs enhance AMPK phosphorylation via G-protein coupled receptor 43(GPR43) at physiological concentrations (< 200 μM), but acetate promotes lipid neogenesis (199) by ACSS2-driven FASN expression at supraphysiological concentrations (> 350 μM). HDL3 neutralizes portal endotoxins via the apolipoprotein A-I binding domain and inhibits Kupffer cell activation in healthy livers. However, HDL3 depletion exacerbated hepatocyte apoptosis in an alcoholic liver injury model (200, 201). A clinical study showed that an oral occludin mimetic peptide (OM-101,50 mg/d) intervention for 8 weeks resulted in decreased fecal protein levels and improved NAS scores, whereas recombinant HDL3 infusion (80 mg/kg) significantly reduced serum LPS levels (202).

Gut barrier damage significantly increases the concentration of gram-negative bacterial endotoxin (LPS) in the portal vein, leading to metabolic endotoxemia (203). In experimental models, LPS triggers the MyD88-dependent signaling axis by binding to TLR4 (an innate receptor sensing bacterial LPS) on Kupffer cell surfaces, increases the nuclear translocation efficiency of NF-κB p65, promotes the release of TNF-α and IL-6, and is linked to increased serine 307 phosphorylation of insulin receptor substrate 1 (IRS-1) in hepatocytes, thereby contributing to insulin resistance in this setting (204–208).

A single-cell transcriptome study revealed that TLR4^+ macrophages expanded in the human MASH liver, secreted IL-6, and upregulated stearoyl-coA desaturase 1 (SCD1) expression through STAT3 Tyr705 phosphorylation, which is consistent with an association between this macrophage state and lipogenic signaling rather than establishing causality (209). The TRIF pathway downstream of TLR4 (a MyD88-independent adaptor route) coactivates the NLRP3 inflammasome, increasing caspase-1 activity and leading to GSDMD-mediated pyroptotic cell death signaling (210). A clinical cohort study revealed elevated serum soluble cluster of differentiation 14 (CD14) levels in patients with MASH, and a positive correlation was found between CD14 levels and liver fibrosis stage (211).From an “axis-to-cytokine” perspective, the gut–liver axis most directly reflects increased portal PAMP burden and consequent hepatic innate immune activation, which drives the local production of proinflammatory mediators such as TNF-α, IL-6, and IL-1β and couples hepatocellular injury to stellate-cell–mediated profibrotic responses. In contrast, the gut–spleen–liver axis highlights splenic immune amplification and the establishment of systemic cytokine tone, followed by immune cell trafficking and circulating inflammatory mediators that collectively shape the hepatic inflammatory microenvironment. This distinction helps reconcile locally triggered hepatic cytokine induction with broader, systemic immune dysregulation observed during disease progression.

A 6-week intervention with TAK-242 (1 mg/kg), a TLR4 inhibitor, in animal studies reduced portal LPS and improved NAS scores, supporting a causal contribution of gut-derived PAMPs–TLR4 signaling within the experimental model, while clinical causality in humans remains to be established (212, 213). Collectively, mechanistic evidence supports a model in which dysbiosis and barrier failure increase hepatic exposure to microbe-derived metabolites and molecular patterns, thereby reprogramming lipid handling and immune tone in the liver. Key effectors include bile acid signaling imbalance, SCFA depletion or context-dependent excess, and endotoxin-driven inflammatory cascades (e.g., TLR4–NF-κB/NLRP3), which together promote insulin resistance, steatosis, and fibrogenic progression. Framing these pathways as “metabolite–barrier–immune” modules provides a coherent bridge from microbial ecology to measurable hepatic phenotypes and therapeutic targets (Figure 5).

Exercise improves MAFLD by regulating gut microbiota composition and function. Clinical data show that 12 weeks of moderate-intensity aerobic exercise (5 times a week, 60% VO₂max) increases Akkermansia abundance, elevates serum butyrate concentrations, and reduces liver fat content in patients with MAFLD. Animal experiments confirmed that 5 weeks of aerobic resistance training (treadmill slope 8°, 60 min/day) increased the relative abundance of Bacteroidetes and inhibited the increase in the Firmicutes/Bacteroidetes ratio in HFD-fed rats. These changes were accompanied by a decrease in TG levels in the liver and an improvement in the insulin sensitivity index (214). Metagenomic analysis revealed that exercise upregulates the SCFA synthesis pathway in the microbiota and enhances the BA 7α-dehydroxylase gene cluster, promoting the production of secondary BAs (DCA) and activating hepatocyte FXR signaling (215). A randomized controlled trial (RCT) showed that 8 months of high-intensity interval training (HIIT) increased Ruminococcaceae abundance and decreased liver fibrosis scores in patients with MAFLD, and was significantly negatively correlated with fecal propionate concentration (216). Although further studies on the association between metabolic groups and host phenotypes of microbiota are needed, the available evidence supports the hypothesis that moderate-intensity exercise (e.g., brisk walking with an oxygen uptake of 12 mL/kg/min) for ≥ 150 min per week can maintain dominant colonization. Monitoring exercise intensity using wearable devices to optimize metabolic benefits is recommended (217, 218).

Time-restricted feeding (TRF, 16:8 cycles) improves MAFLD by reconstructing spatiotemporal dynamics of the gut microbiota. Clinical data show that 8 weeks of TRF intervention increases Akkermansia muciniphila abundance in patients with MAFLD, significantly increases the relative abundance of Faecalibacterium prausnitzii, and restores microbiota α diversity (219). In an HFD model, TRF enhanced the amplitude of abundance fluctuations of Bacteroides during the active host phase, driving rhythmic butyrate synthesis and BA 7α-dehydroxylase activity (220).

Metagenomic identification of Ruminococcus sp. AM28–29LB is a core effector (abundance is negatively correlated with liver steatosis) that upregulates aldehyde oxidase 1(AOX1) expression in hepatocytes by synthesizing vitamin B3 and inhibiting nicotinamide (NAM)-dependent lipid accumulation (221). This bacterium also activates PPARγ in the intestinal epithelium by secreting acetic acid/butyric acid, enhances the expression of TJ ZO-1, and reduces intestinal permeability (29). ^A fecal bacterial transplantation (FMT) experiment confirmed that the TRF donor microbiota could reduce the total cholesterol in the livers of recipient sterile mice and reproduce the inhibitory effect of AOX1 expression, clarifying the causal relationship between the microbiota and host (67). In the future, synergistic interventions involving TRF, dietary fiber (such as inulin 10 g/d), and metabolic benefit optimization by targeting Ruminococcus functional gene cluster enhancement should be explored (222).

Treating microbiota disorders presents multiple challenges. However, probiotic interventions have unique advantages. A multicenter RCT showed that a 12-week probiotic combination significantly reduced ALT and serum endotoxin (LPS) levels and increased Akkermansia muciniphila abundance (223). A liver biopsy confirmed that the proportion of hepatocyte ballooning in the probiotic group decreased significantly, and the NAS score improved significantly by ≥ 2 points (224). Mechanistic studies have revealed that Bifidobacterium longum BB536 activates AMPKα phosphorylation in hepatocytes by secreting acetic acid, inhibiting SREBP-1c-mediated lipid synthesis, and upregulating PPARα-driven fatty acid oxidation genes (225, 226). Animal experiments have shown strain-specific effects. BB536 decreased TG by approximately 40% in rats fed an HFD for 12 weeks, whereas Lactobacillus acidophilus NCFM decreased TG by only 20% (227). There is an urgent need to establish a strain-dose-treatment standardization protocol (recommended minimum effective dose 1 × 10⁹ CFU/day for ≥ 8 weeks) and analyze the interaction network of probiotics, bacteriophages, and dietary components using multi-component technology to promote personalized microecological treatment of MAFLD (228, 229).

Prebiotics (e.g., inulin and fructooligosaccharides) modify metabolic homeostasis by targeting the beneficial gut microbiota. A mouse model on an HFD showed that a 10% inulin intervention for 8 weeks significantly increased serum glucagon-like peptide-2 (GLP-2) concentrations and enhanced intestinal epithelial TJ protein (occludin) expression, resulting in a reduction in portal vein endotoxin (LPS⁾ (230). A metabolomic analysis confirmed that a fructooligosaccharide (5% w/w) intervention increased the proportion of the secondary BA decCA and inhibited SREBP-1c-mediated lipid synthesis (231) by activating the hepatic FXR signaling. Clinical studies have shown that 24 weeks of galactooligosaccharide (GOS, 5 g/d) intervention significantly reduced ALT levels, improved the homeostatic model assessment of insulin resistance(HOMA-IR) index, and markedly increased fecal butyrate concentrations (232) in patients with MAFLD. However, the available evidence is mostly based on surrogate endpoints, and there is an urgent need to verify the benefits histologically by performing liver biopsy and quantifying the liver fat fraction (233) using magnetic resonance imaging-proton density fat fraction (MRI-PDFF). In the future, metagenomics and spatial metabolomics technologies should be combined to analyze the spatiotemporal interactions between prebiotics and microbiota metabolites, and multicenter RCTs should be performed to verify the synergistic effects of GLP-1 receptor agonists.

Antibiotics, important tools for regulating gut microbiota, have a dose-dependent bidirectional effect in the treatment of MAFLD. Non-absorbable antibiotics, such as rifaximin, exert therapeutic effects through local regulation of microbial metabolism. In an HFD-induced MASH mouse model, rifaximin (50 mg/kg/d) decreased portal vein endotoxin levels, liver TG content, and collagen deposition area after 6 weeks of intervention. This mechanism involves FXR signal inhibition and secondary BA metabolic reprogramming (234–236). A RCT conducted by Gangarapu et al. showed that rifaximin (1200 mg/day for 4 weeks) reduced serum ALT levels and NAS scores in patients with MAFLD, which were associated with improved gut microbiota beta diversity index and CDCA/FXR pathway inhibition (237). However, Cobbold et al. found that a 400 mg bid for 6 weeks regimen increased the HOMA-IR index and decreased Bacteroides abundance, suggesting that long-term use of high doses may disrupt metabolic homeostasis (238). Triclosan (0.3 mg/kg/day) intervention specifically enriched Lactobacillus and Chaetospiraceae, increased the Bacteroides/Firmicutes ratio in MAFLD mice, and reduced the fatty degeneration area in the liver. Its efficacy is related to the upregulation of butyric acid synthesis gene expression, suggesting that triclosan can regulate the gut microbiota to treat MAFLD (239). However, antibiotics may have different effects, with mice treated with antibiotics exhibiting severe insulin resistance and an increased incidence and severity of histological lesions in MAFLD. These effects are associated with antibiotic-induced disturbances in the gut microbiota, resulting in changes in bifidobacteria and prevobacteria (240, 241). Therefore, antibiotics are a double-edged sword. On the one hand, antibiotics can regulate gut microbiota disorders by inactivating harmful bacteria to alleviate or treat the condition. However, long-term use of antibiotics may lead to bacterial drug resistance, affect the growth of beneficial bacteria in the body, and result in a new microbiota disorder.

FMT provides innovative intervention strategies for MAFLD treatment by transplanting functional microbial communities from healthy donor feces into the recipient’s intestines and reconstructing the microbiota ecology. Its mechanism involves many dimensions, such as microbiota remodeling, metabolite regulation, and signal transduction repair of the gut–liver axis. Evidence from animal models supports causality within the experimental setting. In animal experiments, FMT decreased the relative abundance of Bacteroides and increased the proportion of Firmicutes in an HFD-induced MAFLD mouse model. This microbiota remodeling was accompanied by improved gut barrier function, reduced portal endotoxin levels, decreased hepatic TNF-α concentrations, and a reduction in the hepatic steatosis area (242). While these findings support a causal contribution in rodents, translation to humans requires further clinical confirmation.

In a study of patients with metabolic syndrome, lean donor FMT increased peripheral insulin sensitivity and decreased hepatic TG levels in the recipients. Its efficacy is associated with an increased abundance of butyrate producers (243). Witjes et al. conducted a parallel-group study showing that the beta diversity of recipient microbiota approached that of the donor after 8 weeks of vegetarian donor FMT, with Bacteroides and Rothschella changing most distinctly (244). However, the clinical outcomes of MAFLD are modest and heterogeneous. Although the MAFLD activity scores did not differ statistically, histological analysis suggested a higher rate of improvement in lobular inflammation in the FMT group. Overall, the differences in MAFLD activity and fibrosis scores were not statistically significant, although the necroinflammation score (including lobular inflammation and hepatocellular ballooning) showed a trend toward improvement (245). Although previous clinical trials have demonstrated short-term improvements in intestinal permeability in patients with MAFLD, imaging assessments have shown that intrahepatic lipid content remained at baseline levels (246) during the 6-month follow-up period after FMT. Taken together, current human studies suggest associations and potential short-term physiological effects but are insufficient to establish causality or durable hepatic benefits without adequately powered randomized trials.

The current FMT technology has multiple potential risks. First, the lack of internationally unified standards for donor screening may lead to iatrogenic transmission of multidrug-resistant microorganisms (e.g., vancomycin-resistant enterococci) or latent pathogens (e.g., Clostridium difficile in the sporulation stage) (247). Second, drug-resistance genes carried by the microbiota can be integrated into the genome of host intestinal symbionts via conjugation transfer, and the β-lactamase gene expression level in the microbiota after intervention has been reported to be higher than that before transplantation (246). Third, metabolites of heterologous microorganisms may abnormally activate the Th17 pathway through pattern recognition receptors. In animal studies, IL-17A expression levels in the colon of recipient mice were increased compared those to in the control groups, and the inflammatory infiltration index in the portal vein area of the liver was also increased. Therefore, further studies are required to evaluate the long-term efficacy and safety of FMT in patients with MAFLD are required (246).

Progress has been made in the design of engineered bacteria that target the core pathological link in the gut-liver axis. Animal model studies have revealed multiple mechanisms of engineering bacteria to treat MAFLD/metabolic dysfunction-associated steatotic liver disease (MAFLD). Recombinant Lactobacillus reuteri_ LR-01 increased intestinal mucosa thickness, increased ZO-1 expression, reduced portal vein endotoxin (LPS) levels, and significantly improved liver steatosis (248, 249) by activating the IL-22/STAT3 signaling axis. In addition, some engineered bacteria, by integrating butyrate synthase clusters (e.g., but operons), geometrically increased butyric acid concentrations in the colon, promoted fatty acid beta-oxidation through the PPARγ/AMPK dual pathway, and reduced the collagen deposition area (250) in a carbon tetrachloride-induced liver fibrosis model. Preclinical trials have confirmed the safety and metabolic regulatory potential of these engineered bacteria. Phase I trials showed that the orally-administered engineered bacteria Escherichia coli_ Nissle 1917 was well tolerated, had no abnormal levels of antibiotic resistance genes (ARGs), significantly increased the proportion of fecal secondary BAs (LCA), and decreased serum ALT levels (251, 252). Metabolic flux analysis revealed that acetyl-CoA pools in the liver of the engineered bacteria intervention group were enlarged, and the metabolic phenotypes of Kupffer cells were reshaped through succinate-mediated activation of HIF-1α (252, 253). Currently, there is an urgent need to address technical bottlenecks, such as the efficiency of colonization and genetic stability of engineered bacteria, and to develop metabolism-responsive intelligent strains using the clustered regularly interspaced short palindromic repeats-deactivated Cas9(CRISPR-dCas9) system, thereby promoting a new era of precision therapy targeting the gut-liver axis.

Complementing the metabolic regulation strategy of engineered bacteria, phage therapy blocks the vicious cycle of inflammatory signaling along the gut-liver axis by precisely eliminating the pathogenic bacteria. Although direct evidence for MAFLD is currently limited, alcoholic liver disease (ALD) research provides important references for elucidating its mechanisms. The cytolytic Enterococcus faecalis-specific phage cocktail designed by Duan et al. not only reduced the fatty degeneration area of liver tissue and serum ALT levels in ethanol-induced mouse models but also maintained the α diversity (254) of the gut microbiota. This finding is highly consistent with the mechanism of pathogen-specific proliferation, which induces metabolic disorders in MAFLD. Notably, clinical cohort analysis showed that for every 1 CFU/g increase in intestinal cytolytic E. faecalis abundance in patients with ALD, the 90-day mortality increased 2.3-fold, confirming its clinical urgency as a therapeutic target (254). Further mechanistic studies have shown that phage therapy ameliorates liver injury through a dual mechanism. First, the targeted elimination of TMA-producing proteobacteria (such as Klebsiella pneumoniae) reduces serum TMAO levels and inhibits the activation of the liver FMO3/NF-κB pathway. Second, reducing the secretion of cytolysin by cytolytic Enterococcus faecalis significantly increases intestinal epithelial TJ ZO-1 expression and reduces portal vein endotoxin (255). Despite its promising future, two major bottlenecks must be resolved before phage therapy can be applied in MAFLD. First, there are striking individual differences in the virulence genes carried by cytolytic Enterococcus faecalis (such as cylLL and esp) in the intestinal tract of patients with MAFLD, and a database of virulence genotype-phage sensitivity associations needs to be established. Second, whether the improvement in liver steatosis observed in the ALD model applies to the insulin resistance background specific to MAFLD (such as PI3K/Akt signaling suppression) remains to be verified. Two Phase I clinical trials (NCT05277350 and NCT0559984) are currently evaluating the effects of phage cocktails on gut microbiota homeostasis and liver elasticity in MAFLD patients. In summary, microbiota-targeted management of MAFLD includes lifestyle-based approaches (exercise and dietary timing/quality), ecological rebuilding strategies (probiotics/prebiotics and related modalities), and emerging precision tools that aim to modulate microbial functions and metabolite outputs. The most translatable pathway is to align interventions with the dominant pathogenic node in a given patient: restoring barrier competence, rebalancing bile acid/SCFA signaling, and dampening portal inflammatory inputs. Future studies should prioritize standardized endpoints and responder stratification to ensure that microbiota modulation translates into reproducible clinical benefits.