Adenosine from high-fat-diet-tolerant monkey-derived Limsolactobacillus reuteri MacFasB02 modulates cholesterol metabolism to alleviate hyperlipidemia and inflammation

Abstract

Hyperlipidemia is a leading global health challenge, limited by the safety liabilities of current pharmacotherapies. Here, we isolated a novel Limosilactobacillus reuteri strain, MacFasB02, from fecal samples of cynomolgus monkeys tolerant to chronic high-fat diet (HFD). This study aimed to systematically evaluate its probiotic properties and therapeutic potential against hyperlipidemia. In vitro, MacFasB02 exhibited robust growth, acid production, and tolerance to acidic and bile environments. In HFD-fed mice, 13-week MacFasB02 administration reduced weight gain, serum triglycerides, low-density lipoprotein cholesterol and total cholesterol, while ameliorating hepatic steatosis and inflammation, as well as restoring intestinal barrier integrity by enhanced villus architecture, goblet cell function, and tight junction proteins expression. Metagenomic analysis revealed gut microbiota remodeling. Transcriptomic profiling coupled with in vivo validation demonstrated upregulation of Apoa1 and Pltp in cholesterol metabolism. Untargeted metabolomics integrated with whole-genome sequencing and supernatant metabolite profiling identified adenosine as a key MacFasB02-derived metabolite in purine metabolism. Consistently, In vitro experiments showed that adenosine reduced lipid accumulation and inflammation in hepatocytes by regulating Apoa1 and Pltp to modulate cholesterol metabolism. Collectively, MacFasB02 exerts dual lipid-lowering and anti-inflammatory effects probably via adenosine-mediated modulation of cholesterol metabolism, promising potential as a live biopharmaceutical agent for hyperlipidemia.

Introduction

As global dietary patterns shift toward high-calorie intake, hyperlipidemia (HLP) has emerged as a significant public health concern¹. Characterized by elevated serum total cholesterol (TC), triglycerides (TG), and low-density lipoprotein cholesterol (LDL-C), along with reduced high-density lipoprotein cholesterol (HDL-C)², HLP contributes to the development and progression of atherosclerosis and increases cardiovascular risk³. Moreover, HLP impairs intestinal barrier function, increases permeability, and enables translocation of gut-derived endotoxins, thereby driving chronic systemic inflammation⁴. Despite its prevalence and multisystem impact, current lipid-lowering therapies face significant limitations. For example, statins may cause myotoxicity and hepatotoxicity⁵, and cholesterol absorption inhibitors such as ezetimibe may cause upper respiratory tract infections, joint pain, and gastrointestinal side effects⁶. In addition, although PCSK9 inhibitors offer therapeutic benefits, their high cost and storage requirements limit accessibility⁷. Traditional Chinese medicines frequently lack mechanistic clarity and exhibit a slow onset of therapeutic efficacy⁸. Thus, safe, effective, and affordable therapeutic alternatives are needed.

Modulation of the gut microbiota, particularly via probiotics, represents a promising therapeutic strategy for HLP⁹. Probiotics, including Limosilactobacillus reuteri (L. reuteri), Enterococcus faecium, and Bifidobacterium animalis, have been reported to exert therapeutic effects on HLP¹⁰. Among them, L. reuteri, a probiotic in the human gut microbiota, has shown promising potential due to its capacity to colonize and modulate host’s metabolism¹¹. L. reuteri HI120 has been reported to exert lipid-lowering effects by activating the AMPK pathway to reduce lipid accumulation and downregulating the expression of PPARγ and SREBP-1c to inhibit lipid synthesis¹². Additionally, L. reuteri TISTR2736 ameliorates inflammation by activating the liver IRS1/PI3K/AKT signaling pathway¹³. However, HLP is a metabolic dysfunction disease concurrently linked to lipid metabolism and inflammation¹⁴, and most existing studies focus on either lipid-lowering or anti-inflammatory effects, without systematically exploring both effects and the underlying mechanisms of L. reuteri strains^15–17. Moreover, certain L. reuteri strains, such as L. reuteri A9 and L. reuteri HI120, are reported to modulate host lipid metabolism by increasing beneficial microbial metabolites, such as short-chain fatty acids propionate and butyrate^12,16, suggesting that their roles in HLP could be mediated by derived metabolites. Nevertheless, the underlying mechanisms of such L. reuteri strains and the derived metabolites in HLP remain to be further elucidated.

Here, we reported the isolation of a novel L. reuteri strain, MacFasB02 (Patent No.: ZL202411412756.7), from cynomolgus monkeys that exhibited tolerance to a long-term high-fat diet (HFD), whose fecal microbiota was shown to alleviate HFD-induced hyperlipidemia in rats¹⁸. This L. reuteri strain, MacFasB02, was found to exhibit dual lipid-lowering and anti-inflammatory activities In vitro and in vivo, by modulating lipid metabolism via adenosine signaling. Our current study thus provides critical insights and a candidate strain for the development of a live biopharmaceutical agent targeting dyslipidemia.

Results

Isolation and characterization of L. reuteri strain MacFasB02 from feces of HFD-tolerant cynomolgus monkeys

A novel L. reuteri strain, MacFasB02, was isolated from the feces of cynomolgus monkeys that exhibited tolerance to a 12-month long-term HFD. It formed white, circular colonies on MRS agar with smooth surfaces, defined edges, and a central papilla-like protrusion (Fig. 1A). It exhibited the typical purple-blue, short-rod morphology of L. reuteri species, with Gram-positive staining. Sanger sequencing confirmed that it had 99% identity to L. reuteri based on the 16S rRNA gene sequence (Fig. 1B). MacFasB02 exhibited robust growth in MRS broth, with the medium pH steadily declining, demonstrating sustained acid production (Fig. 1C). MacFasB02 displayed strong tolerance to bile salt concentrations ranging from 0.1 to 0.5% (Fig. 1D). Moreover, it exhibited excellent acid tolerance, with robust growth at pH 4.0–5.0 (proliferation rate > 100%) and maintained considerable viability even at pH 3.0 (proliferation rate > 72%) (Fig. 1E). These results collectively demonstrated that MacFasB02 isolated from the feces of HFD-tolerant cynomolgus monkeys possessed good acid and bile salt tolerance.

Fig. 1. Isolation and characterization of L. reuteri strain, MacFasB02, from feces of HFD-tolerant cynomolgus monkeys.

A Colony morphology and Gram staining of MacFasB02. B Phylogenetic analysis of MacFasB02 based on 16S rRNA gene sequences. C Growth profile and acid production kinetics of MacFasB02. D Bile salt tolerance of MacFasB02. E Acid tolerance of MacFasB02. B was generated using Majorbio Cloud Platform (https://www.majorbio.com). C–E was generated using GraphPad Prism.

MacFasB02 with dual lipid-lowering and anti-inflammatory effects on HFD mice

An HFD-induced HLP mouse model was used to evaluate the effects of MacFasB02 in vivo (Fig. 2A). After 10 weeks of HFD, mice showed significantly elevated LDL-C, TG, and TC levels and reduced HDL-C compared to NCD controls (Fig. 2B), along with significantly accelerated body weight gain (Fig. 2C), confirming the successful establishment of HLP mice. Following 1 week of ABX treatment, HFD mice received 13 weeks of daily MacFasB02 intervention. Food intake showed no significant difference between HFD + L. r and HFD + PBS mice, yet body weight gain was markedly lower in HFD + L. r mice (Fig. 2D). Lipid profiling revealed significantly lower LDL-C, TG, and TC and higher HDL-C in both HFD + L. r and NCD + PBS mice compared to HFD + PBS mice (Fig. 2E). In the mouse liver, MacFasB02 alleviated HFD-induced hepatic steatosis and inflammatory infiltration, as assessed by HE staining (Fig. 2F), and intracellular lipid deposition, as assessed by Oil Red O staining, in both HFD + L. r and NCD + PBS (Fig. 2G). In the intestine, HE staining showed that MacFasB02 restored the ileum villus height and crypt depth, with significantly lower Chiu scores in HFD + PBS mice (Fig. 2H). AB-PAS staining revealed that MacFasB02 increased goblet cell density and acidic mucus secretion in the ileum and colon (Fig. 2I). Immunofluorescence showed that MacFasB02 improved the expression of Occludin, ZO-1, and MUC2 in HFD + PBS mice (Fig. 2J). Collectively, these findings indicated that MacFasB02 not only effectively improved lipid disorders but also repaired HFD-induced intestinal barrier damage by improving mucosal structure, reducing inflammation, and promoting mucus secretion.

Fig. 2. MacFasB02 improved lipid disorder and repaired the intestinal barrier damage in mice.

A Schematic of MacFasB02 intervention in HLP mice. B Serum lipid profiles of NCD and HLP mice at Week 10. C Weekly body weight changes during HFD induction. D, E Effects of MacFasB02 intervention on food intake, body weight, blood lipid levels. F, G HE staining and Oil red O staining of the mouse liver after MacFasB02 intervention. H HE staining of the mouse ileum and colon after MacFasB02 intervention. I Alcian blue-PAS staining of the mouse colon after MacFasB02 intervention. J Immunofluorescence staining of the mouse ileum and colon after MacFasB02 intervention. p values are calculated using one-way ANOVA for multiple-group comparisons and Student’s t test for two-group comparisons. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. NCD + PBS: mice fed a normal chow diet and treated with phosphate-buffered saline; HFD + PBS: mice fed a high-fat diet and treated with phosphate-buffered saline; HFD + L. r: mice fed a high-fat diet and treated with 10¹⁰ CFU/mL MacFasB02; TC Total cholesterol, TG triglycerides, LDL-C low-density lipoprotein cholesterol, HDL-C high-density lipoprotein cholesterol, ZO-1 zonula occludens-1, MUC2 mucin 2. A was created in BioRender (https://BioRender.com). B was generated using Majorbio Cloud Platform (https://www.majorbio.com). B–E, H, J was generated using GraphPad Prism.

Metagenomics revealed that MacFasB02 reshaped gut microbiota structure

MacFasB02 exerted lipid-lowering and anti-inflammatory effects in vivo. To determine whether these are mediated by modulation of the gut microbiota, we performed metagenomic sequencing of mouse feces. MacFasB02 restored the reduction in microbial richness indicated by the Chao1 index in HFD mice (Fig. 3A). The Shannon index showed a recovery trend in the HFD + L. r (Fig. 3B). PCA demonstrated that HFD + L. r mice clustered closer to NCD + PBS and distinctly separated from HFD + PBS mice, with PERMANOVA confirming significant compositional differences (R² > 0.5, p < 0.05; Fig. 3C, D), indicating that MacFasB02 partially reversed HFD-induced dysbiosis. NMDS (stress = 0.093) further indicated that MacFasB02 shifted microbial structure toward NCD + PBS (Fig. 3D), demonstrating reversal of HFD-induced dysbiosis. At the species level, heatmaps confirmed robust colonization of L. reuteri in HFD + L. r mice (Fig. 3E, H, K). At the phylum level, Bacteroidetes increased from 26.29 to 30.61%, while Bacillota decreased from 60.85 to 57.32%, resulting in a reduced Bacillota/Bacteroidetes ratio (2.31 to 1.87; Fig. 3G). LEfSe analysis revealed a characteristic microbiota signature in HFD + L. r mice at both the genus and species levels (Fig. 3I, L). In such a characteristic microbiota signature, HFD + L. r mice showed enriched Limosilactobacillus, Paramuribaculum, and Dorea compared to HFD + PBS mice at the genus level (Fig. 3J). Correspondingly, the abundances of Limosilactobacillus reuteri, Parabacteroides intestinalis, Schaealia arabinosiphila, as well as Dorea sp. and Adlercreutzia caecimuris were significantly increased in HFD + L. r mice compared to HFD + PBS mice at the species level (Fig. 3M). Collectively, MacFasB02 promoted restoration of a beneficial gut microbiota composition in HFD mice.

Fig. 3. Metagenomics of cecal contents revealed that MacFasB02 remodeled gut microbiota.

A, B Box plots showing Chao and Shannon indices in metagenomics of mouse cecal contents. C PCA D NMDS analysis. E Heatmaps of microbial composition at phylum level. F LEfSe analysis at the phylum level. J Comparison of relative abundance of phyla Bacteroidota and Bacillota. H Heatmaps of microbial composition at the genus level. I LEfSe analysis at the genus level. J Comparison of the relative abundance of MacFasB02-enriched genera in (I). K Heatmaps of microbial composition at the species level. L LEfSe analysis at species level. M Comparison of the relative abundance of MacFasB02-enriched species in (L). p values are calculated using Kruskal-Wallis test for α-diversity and PERMANOVA for β-diversity, as well as LEfSe analysis for differentially abundant microbial communities. *p < 0.05; **p < 0.01; ***p < 0.001. NCD + PBS: mice fed a normal chow diet and treated with phosphate-buffered saline; HFD + PBS: mice fed a high-fat diet and treated with phosphate-buffered saline; HFD + L. r: mice fed a high-fat diet and treated with MacFasB02. A–M was generated using Majorbio Cloud Platform (https://www.majorbio.com).

Non-targeted metabolomics identified adenosine as a key metabolite derived from MacFasB02 in alleviating HFD-induced HLP

Given that MacFasB02 modulates the gut microbiota, we performed non-targeted metabolomic analysis of mouse cecal contents to investigate the metabolic basis of its probiotic effects. PLS-DA revealed the metabolic profiles of HFD + L. r clustered separately from HFD + PBS and closer to NCD + PBS (Fig. 4A) with permutation tests confirming model validity (Fig. 4B), indicating that MacFasB02 ameliorated lipid disorder. 691 DAMs were identified between HFD + PBS and NCD + PBS mice (274 down, 417 up), and 424 DAMs between HFD + PBS and HFD + L. r mice (237 down and 187 up) (Fig. 4C). Venn analysis revealed 51 metabolites that were upregulated in the HFD + L. r towards the level of the NCD + PBS compared to HFD + PBS group, and 59 metabolites that were downregulated accordingly. This core set of 110 DAMs is associated with the ameliorative effect of MacFasB02 on HLP (Fig. 4D, E). KEGG enrichment analysis revealed that the core set of 110 DAMs was primarily involved in pathways related to nucleotide and purine metabolism (Fig. 4F), especially adenosine, a key metabolite in purine metabolism (Fig. 4G). Adenosine decreased in HFD + PBS mice but was rescued by MacFasB02 (Fig. 4D). These results suggest adenosine serves as a core MacFasB02-derived metabolite in purine metabolism.

Fig. 4. Non-targeted metabolomics of mouse cecal contents revealed adenosine as a key metabolite in purine metabolism.

A PLS-DA score plot of mouse cecal metabolites. B Permutation test validating model reliability of metabolomics. C Comparative analysis of differential metabolites across groups. D Venn diagrams and heatmaps of downregulated metabolites in HFD + PBS mice versus the other two groups. E Venn diagrams and heatmaps of upregulated metabolites in HFD + PBS mice versus the other two groups. F, G KEGG pathway enrichment analysis and constructed network of the core 110 DAM identified from (D, E). p values are calculated using univariate tests for screening DAMs. NCD + PBS: mice fed a normal chow diet and treated with phosphate-buffered saline; HFD + PBS: mice fed a high-fat diet and treated with phosphate-buffered saline; HFD + L. r: mice fed a high-fat diet and treated with MacFasB02. A–G was generated using Majorbio Cloud Platform (https://www.majorbio.com).

Liver transcriptomics highlighted the cholesterol metabolism pathway modulated by MacFasB02 and its correlation with adenosine

Given that adenosine is identified as a core MacFasB02-derived metabolite in HLP mice, we performed liver transcriptomic sequencing to identify the underlying molecular pathways. PCA revealed differential gene expression profiles among groups, with clear separation following MacFasB02 intervention (Fig. 5A). Differential expression analysis indicated that the HFD + L. r mice exhibited 381 upregulated differentially expressed genes (DEGs) and 283 significantly downregulated DEGs compared to HFD + PBS mice, whereas HFD + PBS mice exhibited 210 upregulated and 396 downregulated DEGs compared to NCD + PBS mice (Fig. 5B). Venn analysis identified 71 co-downregulated and 70 co-upregulated DEGs in HFD + PBS mice compared to NCD + PBS, which were reversed in HFD + L. r mice (Fig. 5C, D), resulting in a co-regulated gene set of 141 DEGs implicating in the beneficial therapeutic effect of MacFasB02 on HLP. KEGG enrichment analysis showed significant enrichment in cholesterol metabolism (Fig. 5E). Hmgcr, a core marker in the cholesterol metabolism pathway, was selected for further validation in the mRNA and protein expression levels using qPCR and WB (Fig. 5F). To determine whether these effects are mediated by adenosine, Pearson correlation was performed to reveal significant associations between adenosine levels and the co-regulated gene set (Fig. 5G). KEGG enriched string diagram of the adenosine-associated gene dataset further identified cholesterol metabolism as the central pathway, while the PPAR signaling pathway was also significantly enriched. This diagram also confirmed that Pltp and Apoa1 were co-enriched in both the cholesterol metabolism and PPAR signaling pathways (Fig. 5H). qPCR and Western blot validation showed that Apoa1 and Pltp expression were significantly downregulated at both mRNA and protein levels (Fig. 5I). Taken together, these results suggested that MacFasB02 potentially modulates cholesterol metabolism by adenosine-dependent regulation of the Apoa1 and Pltp genes.

Fig. 5. Liver transcriptomics revealed that MacFasB02 rescued HLP in HFD mice by modulating the cholesterol metabolism pathway, which was correlated with adenosine.

A PCA of the mouse live transcriptomic profiles. B DEGs obtained by intergroup comparison of NCD + PBS vs. HFD + PBS and HFD + PBS vs. HFD + L. r. C, D Venn diagrams and heatmaps of downregulated and upregulated genes in the HFD + PBS group compared with the other two groups. E KEGG enrichment analysis of the 110 co-regulated liver DEG set. F Validation of mRNA and protein expression of Hmgcr in the mouse liver, a key gene in the cholesterol metabolism pathway. G Pearson correlation between cecal adenosine levels and co-regulated liver DEG set. H KEGG enriched string diagram of adenosine-correlated DEGs. I Validation of mRNA and protein expression of Apoa1and Pltp. p values are calculated using one-way ANOVA for multiple-group comparisons, as well as DESeq2 for identifying DEGs. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. NCD + PBS: mice fed a normal chow diet and treated with phosphate-buffered saline; HFD + PBS: mice fed a high-fat diet and treated with phosphate-buffered saline; HFD + L. r: mice fed a high-fat diet and treated with MacFasB02; Hmgcr: 3-Hydroxy-3-methylglutaryl-CoA reductase; Apoa1 Apolipoprotein A1, Pltp Phospholipid transfer protein. A–E, G, H was generated using Majorbio Cloud Platform (https://www.majorbio.com). F, I was generated using GraphPad Prism.

Given that Apoa1 and Pltp in cholesterol metabolism reduce lipids via reverse transport and suppress inflammation (Barter et al., 2016; Goldstein et al., 2018), MacFasB02 exerts dual lipid-lowering and anti-inflammatory effects probably through adenosine-mediated regulation of Apoa1and Pltp in cholesterol metabolism pathways.

Validation of the capacity of adenosine production in MacFasB02

To validate the capacity for adenosine production in MacFasB02, we analyzed its whole-genome sequence, obtained using a hybrid approach integrating Illumina next-generation sequencing and PacBio long-read sequencing, followed by non-targeted metabolomics. Our results showed that MacFasB02 has a genome of 2.02 megabase pairs (Fig. 6A) (Table S2). KEGG annotation analysis revealed that the genome of MacFasB02 contained genes involved in the purine metabolism pathway (Fig. 6B). The genes annotated by KEGG in the MacFasB02 genome were mapped to the standard purine metabolism pathway (map00230, highlighted in red), particularly the nagD and deoD genes, which play a crucial role in adenosine synthesis (Fig. 6C), indicating that MacFasB02 has the genetic basis to synthesize adenosine through the purine metabolism pathway.

Fig. 6. MacFasB02 produces adenosine during its purine metabolism.

A, B Circos plot in whole-genome analysis and its KEGG functional enrichment of MacFasB02. C Gene and metabolite network of purine metabolism based on KEGG ortholog analysis of the MacFasB02 genome annotation. D, E Volcano plot and PLS-DA scores plot of DAM in non-targeted metabolomics of MacFasB02 culture medium supernatants. F Permutation test for validation of the PLS-DA model in (E). G, H KEGG pathway enrichment and the top 30 DAMs in non-targeted metabolomics of MacFasB02 culture medium supernatants. nagD encodes a 5’-nucleotidase (EC number 3.1.3.5). deoD encodes a purine-nucleoside phosphorylase (EC number 2.4.2.1). A, B, D–H was generated using Majorbio Cloud Platform (https://www.majorbio.com). C was generated using KEGG Mapper (https://www.genome.jp/kegg/mapper).

Non-targeted metabolomics of MacFasB02 supernatants identified DAMs between MacFasB02 CMS and controls (Fig. 6D). PLS-DA confirmed significant separation between the two groups of samples, which was supported by permutation tests (n = 200, all permuted R₂Y and Q₂ values were lower than those of the original model, p < 0.05) (Fig. 6E, F). KEGG pathway enrichment analysis indicated that DAMs were enriched in purine metabolism (Fig. 6G), and adenosine was confirmed as a key component of this pathway among the top 30 DAMs (Fig. 6H), identifying it as a core MacFasB02-derived metabolite. Taken together, these results provided convergent evidence that MacFasB02 produces adenosine in its purine metabolism.

Adenosine exerts lipid-lowering and anti-inflammatory effects on hepatocytes

The role of adenosine was then validated in a hepatocyte model. Oil Red O staining showed that adenosine treatment significantly reduced intracellular lipid droplet amount and volume in mouse AML12 hepatocytes (Fig. 7A), indicating inhibition of lipid accumulation. ELISA revealed decreased secretion of TNF-α and IL-6 on hepatocytes treated with adenosine (Fig. 7B), demonstrating suppression of pro-inflammatory cytokine release. qPCR and Western blot analyses confirmed downregulation of the core cholesterol-producing enzyme Hmgcr and upregulation of HDL components Apoa1 and Pltp at both the mRNA and protein levels in hepatocytes treated with adenosine (Fig. 7C, D). This result was consistent with the in vivo findings in MacFasB02-treated HFD-induced HLP mice, implicating a consistent role for adenosine in reducing cholesterol production and enhancing HDL lipid transport. Mechanistically, adenosine exerted lipid-lowering and anti-inflammatory effects by regulating the expression of key genes in cholesterol metabolism and reducing pro-inflammatory cytokine release.

Fig. 7. Adenosine exerts lipid-lowering and anti-inflammatory effects on hepatocytes by modulating cholesterol metabolism and its pathway.

A Oil Red O staining on mouse AML hepatocytes treated with adenosine. B ELISA quantification of TNF-α and IL-6 in hepatocytes supernatants treated with adenosine. C qPCR analysis of Hmgcr, Apoa1, and Pltp mRNA levels in hepatocytes treated with adenosine. D Western blot detection of corresponding protein expression in hepatocytes treated with adenosine. p values are calculated using Student’s t test for two-group comparisons. *p < 0.05; **p < 0.01; ***p < 0.001. Hepatocytes were treated with 0.1% DMSO and 1 mmol/L adenosine for the control and adenosine groups, respectively; Hmgcr 3-Hydroxy-3-methylglutaryl-CoA reductase, Apoa1 Apolipoprotein A1, Pltp Phospholipid transfer protein. B–D was generated using GraphPad Prism.

Discussion

We reported the isolation of a novel L. reuteri strain, MacFasB02, from fecal samples of cynomolgus monkeys that exhibited tolerance to a long-term HFD, with robust acid and bile tolerance. MacFasB02 remodeled the gut microbiota and exhibited dual lipid-lowering and anti-inflammatory effects in HFD-induced HLP mice through upregulation of Apoa1 and Pltp in cholesterol transport and downregulation of Hmgcr in cholesterol production, as well as suppression of pro-inflammatory cytokine secretion via its adenosine production. The schematic diagram of the proposed underlying mechanism is presented in Fig. S1.

The current study described the isolation and identification of MacFasB02, a novel L. reuteri strain from fecal samples of a cynomolgus monkey that exhibited tolerance to a long-term HFD. In vitro characterization demonstrated stable growth kinetics, sustained acid production, and robust tolerance to low pH and high bile salt concentrations, indicating traits essential for survival and colonization in the upper gastrointestinal tract^19,20. MacFasB02 exhibited a shorter lag phase and higher acid production rate than rodent-derived strains under identical conditions, suggesting enhanced host adaptation²¹. Its bile salt resistance enables duodenal persistence²². Notably, A pH range of 3.0–5.0 represents the typical acidic gastric environment postprandially²³, and the stable proliferation of MacFasB02 at this pH value suggested its capacity to survive passage through the upper gastrointestinal tract. In contrast to other Limosilactobacillus species, which show strong acid resistance but limited bile tolerance and poor in vivo colonization^24,25, MacFasB02 exhibited dual beneficial effects, ensuring effective transit through the upper digestive tract with maintained viability. The extraordinary bile salt tolerance of MacFasB02 may be attributable to its adaptation to the long-term exposure of bile salts as emulsifiers in the HFD feeding of its monkey host. These findings establish MacFasB02 as a physiologically robust strain with favorable traits for intestinal colonization even under HFD conditions, underscoring its potential translational value in HLP therapy.

Consistent with the HFD-tolerant characteristics of the host from which MacFasB02 was isolated, MacFasB02 exhibited dramatic probiotic effects, including both lipid-lowering and anti-inflammatory effects on HLP in vivo. In HFD-induced mice, MacFasB02 intervention suppressed weight gain, decreased serum TC, TG, and LDL-C levels, and alleviated hepatic steatosis, demonstrating its lipid-lowering activity²⁶. Concurrently, MacFasB02 ameliorated intestinal inflammation and lesions by upregulating junction proteins, the core components of the intestinal barrier²⁷. Thus, MacFasB02 exerts core anti-inflammatory effects via multi-pathway intestinal barrier repair. Most of the reported L. reuteri strains, such as L. reuteri NCIMB30242²⁸, L. reuteri A9¹⁶, L. reuteri LR¹⁵, and L. reuteri HI120¹², have been reported to lower lipids, with their anti-inflammatory effects unconfirmed. Thus, MacFasB02 isolated from HFD-tolerant monkeys exhibited unique advantages for HLP therapy.

MacFasB02 reshaped the gut microbiota structure in HLP mice, especially reversing the increased Bacillota/Bacteroidota ratio induced by HFD. Elevated Bacillota/Bacteroidota ratios are often associated with dysbiosis in metabolic disorders²⁹, suggesting that MacFasB02 counteracts HFD-induced imbalance and restores microbial homeostasis. Notably, MacFasB02 significantly increased Paramuribaculum, Limosilactobacillus, Schaedlerella, and Dorea at the genus and species levels. The increased abundance of Limosilactobacillus indicated successful colonization by MacFasB02. Limosilactobacillus has been reported to promote gut health by producing lactic acid and antimicrobials³⁰. Paramuribaculum, a key butyrate producer, enhances intestinal barrier function and modulates lipid metabolism through elevated butyrate levels³¹. Dorea, capable of producing SCFAs, has been linked to metabolic phenotypes in humans, with context-dependent roles³². Schaedlerella, as a beneficial gut bacterium, further supports a favorable community shift³³. These results suggested that MacFasB02 modulated the gut microbiota towards a beneficial pattern. Despite belonging to the same genus, L. reuteri strains CCFM1190 and MacFasB02 exhibited distinct modulatory effects on gut microbiota. While L. reuteri CCFM1190 specifically enhances the abundance of Faecalibaculum, a genus associated with immune regulation³⁴, MacFasB02 predominantly increased short-chain fatty acid (SCFA)-producing taxa such as Paramuribaculum and Schaedlerella, indicating that taxonomic specificity and functional niches of probiotics vary significantly. Collectively, MacFasB02 demonstrated stable colonization and reshaped the gut microbiota toward a metabolically beneficial state, consistent with its lipid-lowering effects.

Adenosine, a key metabolite in the purine metabolism pathway, was significantly elevated in both MacFasB02 culture supernatants and mouse cecal contents, suggesting its role as a primary MacFasB02-derived metabolite. Genomic analysis and non-target metabolomics confirmed that adenosine was produced by MacFasB02 in its purine metabolism³⁵, which was found to be a key mediator of microbe-host crosstalk via the gut-liver axis. Adenosine is reported to regulate lipid metabolism through A₂A receptors, thereby alleviating hypercholesterolemia³⁶. Concurrently, adenosine exerts its primary anti-inflammatory effects by activating the A₂A receptor, which plays a critical dual regulatory role in suppressing neuroinflammation³⁷. Our results suggested that MacFasB02-derived adenosine represents a key mediator in lipid-lowering and anti-inflammation. In line with our findings, the engineered probiotic Escherichia coli Nissle 1917, designed to produce high levels of adenosine, has been reported to ameliorate inflammation³⁸. It should be noted that our natural probiotic, MacFasB02, isolated from HFD-tolerant monkeys, achieved dual effects along the gut-liver axis in HLP mice, underscoring the comparable importance of screening for beneficial probiotics in hosts with disease-resistant phenotypes to engineered probiotics.

The current study also revealed the underlying mechanism by which MacFasB02-derived adenosine alleviates HFD-induced HLP via the gut-liver axis. In vivo experiments demonstrated that MacFasB02 modulated the cholesterol metabolism pathway in HLP mice by downregulating Hmgcr and upregulating Apoa1 and Pltp. Hmgcr downregulation suppresses endogenous cholesterol synthesis³⁹ and increased expression of Apoa1 and Pltp promotes nascent HDL formation and maturation, enhancing cholesterol reverse transport and clearance⁴⁰. Besides, the KEGG enrichment also identified the PPAR signaling pathway as an enriched pathway, with Apoa1 and Pltp co-enriched in both pathways. This overlap suggests that MacFasB02’s regulation of cholesterol metabolism pathway may be synergistically reinforced by the PPAR signaling pathway, as PPAR α/γ are well-documented to regulate cholesterol transport and HDL maturation⁴¹. Notably, this cholesterol metabolic regulation may also synergize with the robust bile salt tolerance of MacFasB02, which enables the strain to interact with host bile acid metabolism. As reported, bile salt remodels bile acid profiles to modulate cholesterol absorption and catabolism, thereby enhancing lipid-lowering effects in HFD-induced metabolic disorders⁴². Furthermore, modulated cholesterol metabolism not only reduces circulating lipids but also attenuates tissue inflammation by inhibiting macrophage foam cell formation⁴³, providing a mechanistic basis for the strain’s dual effects. Crucially, In vitro treatment with adenosine recapitulated the same gene expression changes in Hmgcr, Apoa1, and Pltp, as observed in the HLP mouse liver, confirming adenosine as the metabolite mediator in lipid-lowering and anti-inflammatory effects. Taken together, these findings demonstrate that MacFasB02, through intrinsic metabolic programming and adenosine production, precisely modulates host cholesterol metabolism to achieve synergistic lipid-lowering and anti-inflammatory outcomes.

Compared with other known probiotic strains, especially L. reuteri strains, MacFasB02 has four unique advantages. First, it was isolated from HFD-tolerant cynomolgus monkeys, which share high genetic homology and similar metabolic phenotypes with humans, making their gut microbiota a valuable reservoir for screening probiotic strains with translational potential in HLP¹⁸. Second, MacFasB02 displayed extraordinary dual tolerance to both acid and bile salts, ensuring efficient gut colonization even in the adverse gastrointestinal environment induced by HFD. Third, while most reported L. reuteri strains exert a single therapeutic effect, either lipid-lowering or anti-inflammatory^15,28. MacFasB02 was confirmed to exhibit dual effects on HLP. Fourth, MacFasB02 featured in adenosine production, which specifically targets the cholesterol metabolism pathway, thereby establishing a new mechanistic paradigm for probiotic-based therapy.

While this study provided a comprehensive mechanistic framework for the action of MacFasB02, certain limitations remain. The findings were primarily derived from murine models, and their translational relevance to human physiology requires further validation. Subsequent efficacy and safety evaluations in non-human primate models represent critical steps toward clinical translation. Besides, although adenosine’s In vitro activity is well established, in vivo evidence linking therapeutic benefits directly to MacFasB02-derived adenosine and its causality requires further studies with an isogenic adenosine-deficient mutant. Furthermore, while adenosine is shown to mediate these effects, future research is warranted to confirm specific downstream A₂A receptors and signaling pathways involved in the observed beneficial effects of adenosine. This could be achieved through pharmacological blockade with selective receptor antagonists or through tissue-specific receptor-knockout animal models. An in-depth investigation will be needed to refine the mechanistic understanding of MacFasB02 and accelerate its development as a live biopharmaceutical agent.

In conclusion, we reported an L. reuteri strain, MacFasB02, isolated from cynomolgus monkeys exhibiting tolerance to a long-term HFD, which exerts dual lipid-lowering and anti-inflammatory effects through adenosine production and gut microbiota remodeling, thereby modulating cholesterol metabolism. In summary, our findings highlight probiotic MacFasB02 as a potential biopharmaceutical agent for HLP and underline the mechanistic and therapeutic importance of adenosine in the disease.

Methods

Isolation and characterization of MacFasB02

Fresh fecal samples were collected from cynomolgus monkeys exhibiting tolerance to long-term HFD and then subjected to microbial isolation. Homogenized fecal samples were centrifuged, and supernatants were serially diluted to 10⁶ CFU/mL. Aliquots were plated on Man-Rogosa-Sharpe (MRS) agar and incubated at 37 °C under facultative anaerobic conditions (85% N₂, 10% H₂, 5% CO₂) for 24 h. Colonies with white-to-cream color, circular shape, and well-defined edges were selected, purified via subculturing in MRS broth, and then cultured overnight under identical conditions, followed by Gram staining. Isolated bacteria were identified by Sanger sequencing of the 16S rRNA gene amplified from genomic DNA using universal primers 27 F and 1492 R (Table S1). Sequence homology was assessed by BLAST analysis against the NCBI nucleotide database.

HLP mouse modeling and MacFasB02 administration

Thirty male specific pathogen-free (SPF) C57BL/6 J mice were housed under controlled conditions (20–26 °C, 30–70% humidity, 12-h light/dark cycle), with free access to food and water provided ad libitum. The sample size of 10 mice per group was determined in accordance with the FDA Redbook and OECD Test Guideline^44–46. After one week of acclimatization, mice were randomly assigned to a normal chow diet (NCD) (10% kcal from fat, n = 10) or a high-fat diet (HFD) (45% kcal from fat, n = 20) for 10 weeks. The HLP status was confirmed by serum lipid test at Week 10. All mice then received a 1-week antibiotic cocktail to deplete gut microbiota for pseudo-germ-free treatment. Subsequently, the HFD group was split into HFD + PBS (n = 10) and HFD + L. r (n = 10), receiving daily oral gavage of PBS or MacFasB02 for 13 weeks. The concentration of MacFasB02 was set at 10¹⁰ CFU/mL according to Zhang et al.’s report⁴⁷. The NCD group continued the standard diet with PBS gavage (NCD + PBS, n = 10). At the end of this study, mice were anesthetized with 3–5% isoflurane inhalation for cardiac blood collection, with unconsciousness confirmed by loss of pedal withdrawal reflex. After blood collection, mice were immediately euthanized via cervical dislocation (compliant with AVMA Guidelines), and death was confirmed by cessation of respiration and heartbeat before collecting liver, intestinal segments, feces, and cecal content. The protocol was approved by the Animal Welfare and Ethics Committee of the Institute of Zoology, Guangdong Academy of Sciences (GIZ20231016), in accordance with ARRIVE guidelines.

Cell culture and adenosine treatment

AML12 hepatocytes were cultured in DMEM supplemented with 10% FBS and maintained at 37 °C in a humidified atmosphere with 5% CO₂. After 24 h, cells were treated with 0.1% DMSO (vehicle control) or 1 mmol/L adenosine for 48 h. Subsequently, cells and supernatants were collected separately for downstream analysis.

Bacterial growth and acid production kinetics analysis

Bacterial growth and acid production kinetics of MacFasB02 were assessed (n = 3) as previously described with modifications⁴⁸. A 1% (v/v) inoculum from glycerol stock was cultured in MRS broth at 37°C under static conditions. OD₆₀₀ and pH were measured using a UV-visible spectrophotometer and a calibrated pH meter, respectively, to profile growth and acidification.

Acid and bile salt tolerance detection

Acid and bile salt tolerances of MacFasB02 were assessed (n = 3) as previously described with modifications⁴⁸. Overnight cultures were inoculated into MRS broth under the condition of pH 3.0–7.0 or 0–0.5% bile salts and incubated statically at 37 °C for 8 h. Cultures were serially diluted (10⁷), plated on MRS agar, and incubated at 37 °C for 24 h. Viable colonies were counted to calculate survival rates.

Whole-genome sequencing

Bacterial whole-genome sequencing was performed as previously described with modifications⁴⁹. High-quality genomic DNA was sequenced using a hybrid sequencing strategy combining PacBio long-read sequencing and Illumina NGS, yielding a complete genome (average sequencing depth ≥ 100×). Raw reads were quality-trimmed using Fastp (v0.23.4), and assembly was performed with Unicycler (v0.5.0), which achieved full chromosomal and plasmid closure. Protein-coding sequences were predicted by Prodigal (v0.5.0). tRNAs and rRNAs were identified using tRNAscan-SE (v2.0.9) and Barrnap (v0.9). Genome circular plots were generated using Circos to visualize features, including GC content, GC skew, and the distribution of coding versus non-coding RNA. Gene annotation was performed using Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis.

Non-targeted metabolomics

Non-targeted metabolomics analyses were performed on MacFasB02 culture medium supernatant and controls (n = 3 per group), as well as on mouse cecal content from HFD + PBS (n = 8), HFD + L.r (n = 8), and NCD + PBS mice (n = 7) as previously described with modifications⁵⁰. MacFasB02 culture medium supernatants were collected 2 h after inoculation during logarithmic growth and filtered (0.22 μm) with protein removed using methanol precipitation. Cecal content metabolites were extracted using methanol-acetonitrile-water with L-2-chlorophenylalanine as internal standard, followed by homogenization, ultrasonication, centrifugation, and filtration (0.22 μm). Metabolomics analysis was performed using an ultra-high performance liquid chromatography-tandem high-resolution mass spectrometry system (UHPLC-Q Exactive HF-X, Thermo Scientific, USA). Chromatographic separation was performed on a Waters ACQUITY HSS T3 column (2.1 × 100 mm, 1.8 μm), while mass spectrometric detection was carried out in positive/negative ESI modes at 70,000 FWHM resolution (m/z 200). The raw data were processed using Progenesis QI (v3.0) for peak detection, alignment, and normalization. Metabolites were identified via MS/MS matching to HMDB, Metlin, and KEGG (mass error < 10 ppm), and only those with RSD < 30% in QC samples were retained. Multivariate analysis employed OPLS-DA using Progenesis QI (v3.0) with 200 permutation tests for validation. Differentially abundant metabolites (DAM) in the MacFasB02 culture medium supernatant were defined using a cutoff of VIP > 1.0 and p < 0.05. DAMs in the cecal contents were defined using a cutoff of |log₂FC | > 1 and p < 0.05 in the HFD + PBS group relative to other groups. KEGG pathway enrichment of DAMs was conducted using clusterProfiler (v4.10.0), and the resulting networks were visualized using igraph (v2.0.3).

Blood lipid tests

Serum samples of mice were collected after 10-week HFD feeding (n = 3 per group) and after 13 weeks of MacFasB02 administration (n = 6 per group). Lipid profiling was performed as previously described with modifications⁵¹. Serum TC, TG, HDL-C, and LDL-C were measured on an automated biochemical analyzer (Roche Cobas c501/c502, Roche Diagnostics, Basel, Switzerland).

Oil red O staining

Oil Red O staining was performed on liver sections (n = 3 per group) as previously described with modifications⁵². Fresh tissues were embedded in OCT, sectioned at 8 μm, fixed in 4% paraformaldehyde, stained with Oil Red O solution (9 min), differentiated in 60% isopropanol, and counterstained with hematoxylin before being photographed using microscopy.

Hematoxylin-Eosin (HE) and Alcian Blue-Periodic Acid-Schiff (AB-PAS) staining

Paraffin sections and HE (n = 3 per group) and AB-PAS staining (n = 3 per group) from the mouse liver, ileum, and colon were prepared as previously described with modifications⁵¹.

Immunofluorescence

Immunofluorescence was performed on mouse intestinal segments (n = 3 per group) as previously described with modifications⁵². Following antigen retrieval and fluorescence quenching, sections were blocked and incubated overnight at 4 °C with primary antibodies against mouse Occludin (OCLN) (Proteintech, Cat# 27260-1-AP), Zonula Occludens-1 (ZO-1) (Proteintech, Cat# 21773-1-AP), or Mucin 2 (MUC2) (Proteintech, Cat# 66378-1-Ig), then incubated with a fluorescein-conjugated goat anti-rabbit IgG secondary antibody (Proteintech, Cat# SA00013-2), followed by DAPI counterstaining, and imaged via confocal microscopy (STELLARIS 5, Leica Microsystems, Wetzlar, Germany), and fluorescence intensity was quantified using ImageJ.

Metagenomic sequencing

Metagenomic sequencing was performed on cecal samples (n = 10) from HFD + PBS (n = 8), HFD + L.r (n = 8), and NCD + PBS (n = 7) mice, as previously described with modifications⁵⁰. Total microbial DNA was extracted using the TIANamp Stool DNA Kit. 150-bp paired-end sequencing was performed on the Illumina NovaSeq 6000. Raw reads were quality-filtered using Fastp (v0.20.0) to remove low-quality sequences (Q < 20 for >50% bases), adapters, and host contaminants. De novo assembly was performed with MEGAHIT (v1.1.2), and open reading frames were predicted using Prodigal (v2.6.3). A non-redundant gene set was constructed via CD-HIT (v4.6.1) (sequence identity ≥95%, coverage ≥90%). Kruskal-Wallis tests (p < 0.05) were used to compare α-diversity indices between groups. Based on the Bray-Curtis distance matrix, principal coordinate analysis (PCoA) and non-metric multidimensional scaling (NMDS) were performed in R using the vegan package (v2.6-4) to assess β-diversity, with PERMANOVA testing for group differences (p < 0.05). Linear discriminant analysis effect size (LEfSe) identified differentially abundant microbial communities at phylum, genus, and species levels (LDA > 2.0, p < 0.05). Using the annotations obtained from the NR database via DIAMOND (v2.0.15), the relative abundances of characteristic phylum, genus, and species associated with MacFasB02 were calculated.

Transcriptome sequencing

Liver transcriptome sequencing (n = 3 per group) was performed as previously described with modifications⁵⁰. Total RNA was extracted and its quality assessed by spectrophotometry and electrophoresis. Libraries were constructed using the polyA method and sequenced on Illumina NovaSeq 6000 with a 150 bp paired-end configuration. The raw reads were first quality-controlled and adapter-trimmed using fastp (v0.23.2). The clean reads were then aligned to the mouse reference genome (GRCm39) using HISAT2 (v2.2.1). Gene expression was quantified as transcripts per million (TPM). PCA revealed global transcriptional variation. Differentially Expressed Genes (DEGs) were identified using DESeq2 (v1.40.2) (|log₂FC | > 1, p < 0.05).

Quantitative PCR (qPCR)

Total RNA was extracted from the mouse liver and AML12 hepatocytes (n = 3 per group), then reverse transcribed into cDNA using PrimeScript™ RT reagent Kit (Takara, Cat# RR047A). Primers for mouse apolipoprotein A1 (Apoa1), phospholipid transfer protein (Pltp), and 3-hydroxy-3-methyl glutaryl-coenzyme A reductase (Hmgcr) were listed in Table S1. qPCR was conducted using SYBR Premix Ex Taq II (Takara, Cat# RR820A) on a real-time PCR system (QuantStudio 5, Applied Biosystems, USA). Relative mRNA expression levels were normalized to β-actin and calculated using the 2^−ΔΔCt method.

Western blotting

Total protein was extracted from liver and AML12 hepatocytes (n = 3 per group), and concentrations were determined by bicinchoninic acid assay. Primary rabbit antibodies against mouse Apoa1 (Proteintech, Cat# 60068-1-Ig), Pltp (Proteintech, Cat# 66865-1-Ig), Hmgcr (Proteintech, Cat# 70459-1-Ig), β-actin (Proteintech, Cat# 81115-1-RR), and HRP-conjugated goat anti-rabbit IgG (Proteintech, Cat# SA00001-2) were used. Proteins were detected using enhanced chemiluminescence and quantified using ImageJ, with β-actin as an internal control.

Enzyme-Linked Immunosorbent Assay (ELISA)

IL-6 and TNF-α concentrations in cell culture supernatants (n = 3 per group) were measured with ELISA kits (IL-6: Solarbio, Cat# P00158; TNF-α: Solarbio, Cat# SEKM-0034) by following the manufacturer’s protocols. Supernatants were centrifuged to remove debris and then incubated with the TMB substrate, and the absorbance was measured at 450 nm on a microplate reader.

Statistical analysis

All data were analyzed using GraphPad Prism (v9.5.1) (GraphPad Software, San Diego, CA). Multiple-group and two-group comparisons were performed using one-way ANOVA and Student’s T test, respectively. Data are expressed as mean ± SD, with statistical significance set at p < 0.05. Representative results from tissue sections and microscopy were obtained from at least three independent fields and showed consistent patterns. All images were uniformly processed and quantitatively analyzed in ImageJ (v1.54). Pearson correlation coefficients were calculated to assess associations between adenosine levels and DEGs (|r | > 0.8, p < 0.05). KEGG pathway enrichment analysis of genes was performed using clusterProfiler (version 4.3.2) (p < 0.05, FDR < 0.1). Venn diagrams and hierarchical clustering were performed using VennDiagram (v1.7.3) and Pheatmap (v1.0.12), respectively.

Author contributions

Y.J., H.J.A., J.H.R., and J.H.C. conceived and designed the study. Y.J., H.J.A., T.T.Z., J.M.G., X.L.Z., B.H.L., Y.Y.L., and X.J.Z. performed the experiments. Y.J., H.J.A., and J.J.L. analyzed the data. Y.J. wrote the original draft of the manuscript. Y.J., J.H.R., and J.H.C. contributed to writing, reviewing, and editing the manuscript. J.H.R. and J.H.C. supervised the research. J.H.R. and J.H.C. administered the project.

Data availability

The multi-omics data generated in this study have been deposited in the OMIX database (https://ngdc.cncb.ac.cn/omix) and GSA database (https://ngdc.cncb.ac.cn/gsub) at the China National Center for Bioinformation. OMIX013581: Non-targeted metabolomics of MacFasB02 culture medium Supernatant. Access link: https://ngdc.cncb.ac.cn/omix/release/OMIX013581. OMIX013583: Non-targeted metabolomics of mouse cecal contents. Access link: https://ngdc.cncb.ac.cn/omix/release/OMIX013583. CRA034905: Whole-genome sequencingof MacFasB02. Access link: https://ngdc.cncb.ac.cn/gsa/s/qwRF2z5T. CRA034902: Transcriptome sequencing of mouse liver samples. Access link: https://ngdc.cncb.ac.cn/gsa/s/XjRXx8S. CRA035096: Metagenomic Sequencing of mouse fecal samples. Access link: https://ngdc.cncb.ac.cn/gsa/s/22XFWxI1. CRA035456: Sanger sequencing identification of MacFasB02. Access link: https://ngdc.cncb.ac.cn/gsa/s/NBo8148Z.

Competing interests

The authors declare no competing interests.

Consent for publication

All authors approved the final manuscript and the submission to this journal.

Supplementary information

The online version contains supplementary material available at 10.1038/s41538-026-00765-z.