Gut Bacteria Produce Propionic Acid That Helps ApoA-I Improve Fatty Liver by Activating Liver Fat-Breakdown Pathways

To assess the impact of apoA-I on MASLD development, we established a mouse model of MASLD using a high-fat, high-cholesterol, high-fructose diet (HFCFD) (Figure 1A and S1A,B). Compared with those in WT mice fed HFCFD, serum concentrations of alanine transaminase (ALT) and aspartate transferase (AST) were significantly lower in apoA-I Tg-HFCFD mice, indicating that elevated apoA-I mitigated hepatic injury in this model (Figure 1B,C; Table A1). The examination of liver sections stained with H&E and Oil Red O indicated that apoA-I significantly diminished HFCFD-induced hepatic steatosis and inflammatory infiltration (Figure 1D; Figure S2B–E). Furthermore, apoA-I significantly reduced hepatic triglyceride (TG) and cholesterol (Chol) levels in MASLD mice (Figure 1E,F). These results collectively suggest that apoA-I has therapeutic effects on MASLD.

Considering the well-documented involvement of the gut microbiota in MASLD pathogenesis [22,23], we investigated whether the protective effects of apoA-I on MASLD rely on these microbial communities. We conducted faecal microbiota transplantation (FMT) from HFCFD-fed apoA-I Tg mice (Tg-HFCFD) to HFCFD-fed WT recipients (designated WT-HFCFD-FMT). Additionally, we administered an antibiotic cocktail (ABX) to HFCFD-fed apoA-I Tg mice to deplete the gut microbiota (designated Tg-HFCFD-ABX) (Figure 2A; Figure S1C,D).

Compared with WT-HFCFD mice, WT-HFCFD-FMT mice presented significantly lower serum ALT and AST levels, hepatic steatosis, hepatocellular ballooning, and hepatic TG and Chol contents (Figure 2B–F; Figure S2G–J; Table A2). These findings indicate that FMT from apoA-I Tg mice effectively mitigates MASLD. Conversely, compared with Tg-HFCFD mice, Tg-HFCFD-ABX mice exhibited exacerbated hepatic injury, steatosis, and ballooning (Figure 2B–F; Figure S2G–J; Table A2). These findings indicate that the depletion of intestinal flora eliminates the hepatoprotective benefits of apoA-I in the context of MASLD. Collectively, our observations strongly support the conclusion that apoA-I alleviates MASLD in a gut microbiota-dependent manner.

To characterise alterations in the intestinal flora induced by apoA-I, we conducted 16S rRNA gene sequencing on faecal samples obtained from WT-CHOW, WT-HFCFD, and Tg-HFCFD mice. Principal coordinate analysis clearly revealed segregation in the composition of the microbial communities across the three groups (Figure 3C), indicating that both diet and genotype significantly shaped the structure of the gut microbiota. Compared with WT-CHOW, HFCFD feeding significantly reduced alpha diversity, as assessed by the Chao index and observed features; however, compared with WT-HFCFD mice, apoA-I overexpression did not significantly alter microbial richness (Figure 3A,B).

We next examined specific taxonomic shifts. Analysis at the phylum level revealed that HFCFD induced a typical dysbiotic pattern characterised by a marked reduction in the abundance of Bacteroidota and a rise in the abundance of Firmicutes, thereby increasing the Firmicutes/Bacteroidota ratio, a hallmark of MASLD-associated dysbiosis [24]. This shift was notably reversed in the Tg-HFCFD mice (Figure 3D–G). Furthermore, at the family level, HFCFD feeding led to a reduction in Muribaculaceae, which is known as a key SCFA-producing family [25]. Concurrently, we observed an increase in Erysipelotrichaceae, a bacterial family that has been associated with steatosis and metabolic syndrome [26]. ApoA-I overexpression significantly counteracted these changes (Figure 3H,I).

Given that Muribaculaceae is a proficient producer of SCFAs, and considering the established importance of these metabolites in regulating glucose and lipid metabolism [12,13,16], we hypothesised that apoA-I may improve MASLD by increasing SCFA production. Consequently, we quantified the faecal SCFA levels. As expected, compared with CHOW-fed WT mice, mice on the HFCFD regimen exhibited substantial decreases in acetic, propionic, and butyric acid concentrations. Notably, the overexpression of apoA-I led to a specific and marked increase in PPA levels among HFCFD-fed mice (Figure 3J–L).

Collectively, these findings demonstrate that apoA-I remodels the gut microbiota in MASLD mice, attenuating diet-induced dysbiosis and specifically increasing the production of PPA, which may underlie its hepatoprotective mechanism.

To clarify how PPA influences hepatic lipid metabolism and explore its underlying mechanism, we investigated its effect on OA-induced lipid accumulation in HepG2 cells. Staining with Oil Red O revealed that at concentrations of 0.5 mM and greater, PPA effectively reduced lipid accumulation without compromising cell viability (Figure 4A–C). On the basis of these results, 0.5 mM and 1 mM PPA were selected for all subsequent experiments. Moreover, this concentration range effectively elicits biological activity in vitro and is well-established for investigating the functional roles and molecular mechanisms of PPA [27]. Consistently, BODIPY 493/503 staining further confirmed a marked decrease in both the quantity and size of lipid droplets in OA-loaded HepG2 cells following PPA treatment (Figure 4D,E). Given that lipid droplet catabolism is primarily regulated by lipolysis, we hypothesised that PPA may stimulate this catabolic pathway. To test this hypothesis, we evaluated the activity of adipose triglyceride lipase (ATGL), the primary regulatory enzyme that catalyses the first stage of triglyceride hydrolysis [28]. Western blot analysis demonstrated that PPA treatment significantly increased ATGL phosphorylation in OA-treated cells (Figure 4F,G), indicating the activation of intracellular lipolysis.

We next examined whether the fatty acids released via lipolysis were subsequently involved in oxidative metabolism. Accordingly, we examined the expression of carnitine palmitoyltransferase 1A (CPT1A), a mitochondrial enzyme essential for fatty acid β-oxidation [29]. Notably, PPA treatment robustly upregulated CPT1A expression (Figure 4H,I), suggesting a concomitant increase in fatty acid oxidation.

In summary, these results demonstrate that PPA mitigates hepatocellular lipid accumulation. This effect is associated with enhanced ATGL-mediated lipolysis and the promotion of fatty acid oxidation.

Previous studies have shown that Ca²⁺–CaMKII signalling mediates ATGL phosphorylation and hepatic lipolysis [30,31]. Therefore, we evaluated alterations in intracellular Ca²⁺ levels and CAMKII phosphorylation in OA-overloaded HepG2 cells exposed to PPA. We observed a marked increase in intracellular Ca²⁺ levels and phosphorylated CAMKII in OA-overloaded HepG2 cells stimulated with 0.5 mM and 1 mM PPA (Figure 5A–C). To further investigate the role of Ca²⁺ in CAMKII and ATGL activation, we applied BAPTA, a calcium chelator, to sequester intracellular Ca²⁺. BAPTA treatment in OA-exposed HepG2 cells abolished the PPA-induced responses, which was supported by decreases in CAMKII and ATGL activity (Figure 5D–F). These results indicate that PPA promotes CAMKII and ATGL activation in a Ca²⁺-dependent manner in OA-overloaded HepG2 cells.

G-protein-coupled receptors 41 and 43 (GPR41, GPR43) function as the primary receptors for SCFAs and are involved in regulating lipid metabolism in the liver [14,16,32]. Upon activation by SCFAs, GPR43, which is expressed in the liver, triggers the Gαq-phospholipase C (PLC)-inositol trisphosphate (IP₃) pathway, leading to increased intracellular Ca²⁺ release [33]. We therefore hypothesised that GPR43 participates in PPA-induced activation of the Ca²⁺–CAMKII–ATGL cascade. To test this hypothesis, we inhibited GPR43 using the specific antagonist GLPG0974. The results showed that GLPG0974 significantly attenuated the PPA-induced increase in the level of intracellular Ca²⁺ (Figure 6A), as well as the activation of CAMKII and ATGL (Figure 6B–D). Collectively, these data support the potential involvement of GPR43 in PPA-induced activation of the Ca²⁺–CAMKII–ATGL signalling axis in OA-overloaded HepG2 cells.

To evaluate whether PPA has therapeutic value for MASLD, 6-week-old male WT mice were assigned to three groups. Each group received one of three dietary regimens, namely, a chow diet, a HFCFD or a HFCFD supplemented with 2 g/kg PPA for 16 weeks (Figure 7A). Compared with the HFCFD-fed MASLD mice, the mice receiving PPA supplementation had notably lower serum concentrations of ALT and AST (Figure 7B,C; Table A3), indicating that PPA improved liver function. Additionally, examination of liver sections using H&E and Oil Red O staining confirmed that PPA supplementation mitigated HFCFD-induced hepatic steatosis and inflammatory infiltration (Figure 7D; Figure S2L–O). Consistent with these findings, hepatic TG and Chol levels, which were elevated in the HFCFD-fed mice, markedly decreased in the HFCFD-PPA-fed mice (Figure 7E,F). These findings collectively indicate that PPA exerts protective effects against MASLD progression by ameliorating hepatic lipid accumulation.

Furthermore, we investigated whether the Ca²⁺–CAMKII–ATGL signalling pathway contributes to the PPA-mediated amelioration of MASLD. Compared with mice in the HFCFD group, mice in the PPA supplementation group exhibited increased phosphorylation of CAMKII and ATGL (Figure 7G–I). Concurrently, we observed upregulation of CPT1A in the HFCFD-PPA-fed mice (Figure 7G,J). Taken together, these results demonstrate that PPA attenuates MASLD by promoting hepatic lipolysis through the activation of CAMKII and ATGL, and by enhancing hepatic mitochondrial β-oxidation.

All experimental protocols involving animals complied with the ethical standards authorised by the Animal Ethics Committee of Capital Medical University (Approval No. AEEI-2025-1105). Male wild-type (WT) mice on a C57BL/6J background were provided by Vital River Laboratory (Beijing, China). Transgenic mice expressing human apolipoprotein A-I (apoA-I Tg; Strain: 001927) were obtained from The Jackson Laboratory. In this study, male mice of both genotypes, aged six weeks, were housed under specific pathogen-free (SPF) conditions with unrestricted access to food and drinking water.

At the end of Animal Study 1, freshly obtained faeces samples from WT-CHOW, WT-HFCFD, and Tg-HFCFD mice were immediately flash-frozen in liquid nitrogen and preserved at −80 °C.

DNA extraction was performed, followed by PCR amplification of the V3–V4 hypervariable region of the bacterial 16S rRNA gene with the primer pair 341F/806R. The resulting amplicon libraries were subjected to paired-end sequencing on an Illumina NovaSeq 6000 system (Novogene, Beijing, China). Sequence data processing was conducted in QIIME2 (v2020.2): primers were removed with cutadapt (v2.8), and sequences were denoised and filtered for chimaeras using DADA2 to generate amplicon sequence variants (ASVs). Libraries containing fewer than 8000 quality-filtered sequences were discarded. The retained samples were rarefied to an even depth of 8500 sequences per sample. Taxonomic classification was carried out with a pretrained naïve Bayes classifier (silva-132-99-nb-classifier.qza) referencing the SILVA 132 database at 99% identity. Alpha diversity was calculated using the Chao1 and observed features metrics. Moreover, beta diversity was quantified with the Bray–Curtis distance, displayed via principal coordinate analysis (PCoA), and group differences were statistically tested using ANOSIM. The differential abundance of bacterial taxa across groups was examined with the Mann–Whitney U test. Raw sequencing reads have been deposited in the NCBI SRA under BioProject accession number PRJNA1377218.

The concentrations of faecal short-chain fatty acids (SCFAs) were determined via an established LC–MS/MS method. Sample analysis was carried out on a TSQ Altis™ mass spectrometer coupled to a Vanquish™ Flex UHPLC system (Thermo Fisher Scientific, Waltham, MA, USA). Separation was achieved with a Waters ACQUITY UPLC BEH C18 column (2.1 × 100 mm, 1.7 μm; Waters Corporation, Milford, MA, USA) maintained at 40 °C, employing a gradient elution of 10 mM aqueous ammonium acetate (Sigma-Aldrich, St. Louis, MO, USA) and a mixture of acetonitrile and isopropanol (1:1; Sigma-Aldrich, St. Louis, MO, USA). Detection was performed in negative ionisation mode using multiple reaction monitoring (MRM). Quantitation was based on external calibration curves corrected by isotopically labelled internal standards specific to each SCFA. Method validation confirmed acceptable precision (intra- and interday RSD ≤ 15%), accuracy (85–115% recovery), and stability (RSD ≤ 15% over 24 h at 4 °C), supporting the reliable quantification of all the analytes.

To eliminate the gut microbiota, the mice were subjected to a two-stage antibiotic treatment. During the initial intensive phase (the first week), each mouse received a daily oral dose of a 200 μL antibiotic mixture containing neomycin sulphate (200 mg/kg; #N6386; Sigma-Aldrich, St. Louis, MO, USA), metronidazole (200 mg/kg; #A114755; Ambeed, Buffalo Grove, IL, USA), vancomycin (100 mg/kg; #A810922; Ambeed, Buffalo Grove, IL, USA), and ampicillin (200 mg/kg; #A307483; Ambeed, Buffalo Grove, IL, USA). Subsequently, during the maintenance period spanning 8 weeks, the animals’ drinking water was supplemented with neomycin sulphate (1 g/L), metronidazole (1 g/L), ampicillin (1 g/L), and vancomycin (500 mg/L). This antibiotic-supplemented water was freshly made and changed every 48 h to ensure its effectiveness.

Recipient mice were pretreated with antibiotics for three days to eliminate their endogenous intestinal microbiota prior to FMT. Fresh faecal pellets from Tg-HFCFD donor mice were suspended in sterile PBS, and then subjected to centrifugation at 500× g for 5 min. The supernatant was collected under anaerobic conditions. Recipient mice (WT-HFCFD-FMT) received 200 μL of this faecal supernatant by oral gavage four times weekly for 8 weeks.

After a 10 h fasting period with water provided ad libitum, the mice were euthanised following approved protocols. Livers were promptly excised. A consistent 1 cm × 1 cm section from each mouse was fixed in 4% paraformaldehyde (#G1101; Servicebio, Wuhan, China), processed for paraffin embedding, sectioned, and subjected to hematoxylin and eosin (H&E) staining for histological evaluation. For lipid analysis, fresh liver tissue was embedded in optimal cutting temperature (OCT) medium, rapidly frozen, and cryosectioned for Oil Red O staining. The residual hepatic tissue was snap-frozen in liquid nitrogen and maintained at −80 °C for subsequent analysis.

Hepatic steatosis, ballooning, and inflammation were graded on H&E-stained sections on the basis of established criteria (). For each mouse, 5 random nonoverlapping fields (200×) were examined. Scoring was conducted independently by two pathologists who were blinded to group allocation, and the final scores represent the average of both assessments. To quantify steatosis, Oil Red O-stained liver sections were analysed using ImageJ (version 1.53; National Institutes of Health, Bethesda, MD, USA). The software was used to determine both the lipid-positive area and the total tissue area, and the proportion of lipid area was calculated as (lipid-positive area/total tissue area) × 100%. Supplementary Tables S1–S3

Liver triglyceride (TG) and total cholesterol (TC) levels were quantified according to kit protocols (#E1025-105 and #E1026-105; Applygen Technologies, Beijing, China). Approximately 25 mg of frozen liver was homogenised in 500 µL of lysis buffer and clarified by centrifugation at 2000× g for 5 min. The resulting supernatant was split for protein quantification (BCA assay; #23227; Thermo Fisher Scientific, Waltham, MA, USA) and TG/TC assays. Reagent solutions (R1:R2 = 4:1) and standard curves (glycerol for TG, cholesterol for TC) were prepared. Ten microlitres from each sample was combined with 190 µL of working solution in a 96-well plate, followed by incubation at 37 °C for 15 min (TG assay) or 20 min (TC assay). The absorbance was recorded at 550 nm, and the TG/TC levels were normalised to the total protein content.

Blood samples were collected and centrifuged at 3000 rpm for 15 min to obtain serum. Alanine aminotransferase (ALT) and aspartate aminotransferase (AST) levels were determined with an automatic biochemical analyser (Rayto Life and Analytical Sciences, Shenzhen, China). Serum non-esterified fatty acids (NEFA) were measured using a commercial kit (#A042-2-1; Nanjing Jiancheng Bioengineering Institute, Nanjing, China). Serum levels of tumour necrosis factor-α (TNF-α) and interleukin-1β (IL-1β) were evaluated by ELISA (#JL10484 and #JL18442, respectively; Jianglai Biotechnology, Shanghai, China).

A CCK-8 assay (#A311-01; Vazyme, Nanjing, China) was used to assess cell viability in accordance with the supplier’s guidelines.

The Oil Red O method was used to detect lipid accumulation. The cells were fixed for 15 min at room temperature using 4% paraformaldehyde (#G1101; Servicebio, Wuhan, China). Following the PBS washes, the samples were incubated with Oil Red O working solution (#O8010; Solarbio, Beijing, China) for 30 min in the dark. Stained lipid droplets were then examined under a microscope. For quantification, lipid droplets were extracted with 4% NP-40 solution (#N8030; Solarbio, Beijing, China), and the absorbance was read at 520 nm on a microplate reader.

Intracellular lipids were visualised using a Lipid Droplets Green Fluorescence Assay Kit (#C2053; Beyotime Biotechnology, Shanghai, China) as directed. Briefly, after being fixed with 4% paraformaldehyde and washed with PBS, the cells were exposed to staining solution (containing BODIPY 493/503 and Hoechst 33342) for 20 min in the dark. Following a final PBS wash, images were captured using a confocal microscope.

Cytosolic Ca²⁺ concentrations were determined using a calcium assay kit (#S1063S; Beyotime Biotechnology, Shanghai, China) in strict accordance with the provided protocol, and absorbance readings were taken at 575 nm.

Proteins from HepG2 cells or liver samples were extracted using lysis buffer supplemented with protease (#B14001↗; Selleck, Shanghai, China) and phosphatase (#B15001↗; Selleck, China) inhibitor cocktails. The protein concentration was quantified via a BCA assay (#23227; Thermo Fisher Scientific, Waltham, MA, USA).

Protein levels were evaluated by Western blotting using antibodies against phospho-CaMKII (#12716T; Cell Signaling Technology (CST), Danvers, MA, USA), CaMKII (#4436T; CST, Danvers, MA, USA), phospho-ATGL (#ab317611; Abcam, Cambridge, UK), ATGL (#55190-1-AP; Proteintech, Rosemont, IL, USA), β-actin (#60008-1-Ig; Proteintech, Rosemont, IL, USA), apoA-I (#14427-1-AP; Proteintech, Rosemont, IL, USA), GPR43 (#84544-1-RR; Proteintech, Rosemont, IL, USA) and vinculin (#26520-1-AP; Proteintech, Rosemont, IL, USA). The band intensity was analysed using ImageJ software (version 1.53; National Institutes of Health, Bethesda, MD, USA).

The colon was excised, and its length was recorded. Colon segments were histologically evaluated after fixation in 4% paraformaldehyde, paraffin embedding, and sectioning at a thickness of 5 μm. The sections were subsequently stained with H&E for microscopic observation.

The results are expressed as the mean ± SEM. GraphPad Prism (version 9.0) software was used for all the statistical analyses, with p < 0.05 considered to indicate statistical significance. Differences between two groups were assessed using two-tailed Student’s t tests. Comparisons involving three or more groups were performed by one-way ANOVA, followed by Tukey’s post hoc test.

The original contributions presented in this study are included in the article/. Further inquiries can be directed to the corresponding author(s). Raw 16S rRNA sequencing data are publicly accessible in the NCBI Sequence Read Archive under BioProject ID PRJNA1377218. Supplementary Materials