Bioengineered ROS-tolerant probiotic reshapes gut microbiota-host axis to ameliorate type 2 diabetes in male mice

Abstract

The pandemic-scale progression of type 2 diabetes mellitus (T2DM) necessitates innovative interventions targeting the pathogenic triad of insulin resistance, dysregulation of lipid metabolism, and gut microbiome dysbiosis. Here, we report a synthetically bioengineered probiotic consortium (REcN-F/Ca) developed through directed metabolic adaptations of Escherichia coli Nissle 1917 (EcN) under iterative hydrogen peroxide selection, subsequently functionalized with fructooligosaccharide-calcium carbonate composites. REcN-F/Ca exhibits enhanced reactive oxygen species tolerance through upregulated antioxidant enzymes and hydrogen sulfide-mediated redox balancing, alongside improved gastrointestinal survivability. In high-fat diet-induced obese male mice, REcN-F/Ca restores gut microbiota diversity, enriches butyrogenic taxa (Lachnospiraceae and Blautia), and rescues short-chain fatty acids depletion. Transcriptomic profiling reveals PPAR signaling activation, driving lipid metabolism and suppressing adipose inflammation. These effects translate to systemic metabolic improvements with attenuated weight gain (−25.4%), restored glucose homeostasis, and reduced insulin resistance (HOMA-IR: −73.2%) in the obesity and T2DM murine model. Our findings establish REcN-F/Ca as a synthetically engineered probiotic that simultaneously corrects intestinal ecological perturbations and reverses host metabolic dysfunction, proposing a paradigm for metabolic syndrome management.

Introduction

The global prevalence of type 2 diabetes mellitus (T2DM) has reached epidemic proportions, characterized by with excessive circulating glucose and free fatty acids that drive dysregulated lipid metabolism, insulin resistance, systemic oxidative stress and gut microbiota dysbiosis^1,2,3,4 While pharmacological interventions (e.g., PPAR agonizts and GLP-1 analogs) offer symptomatic relief, their long-term efficacy is limited by an increased risk of side effect and failure to address gut microbiota dysbiosis—a hallmark of metabolic dysfunction^5,6,7,8,9 Probiotic has emerged as promising modulators of the gut-metabolic axis, yet conventional strains often lack the robustness to survive gastrointestinal stressors or the metabolic versatility to counteract host-pathological processes^10,11,12

Escherichia coli Nissle 1917 (EcN), a well-characterized probiotic, has shown potential in mitigating intestinal inflammation through competitive exclusion and biofilm formation¹³. However, its native antioxidant defenses are insufficient to neutralize reactive oxygen species (ROS) in obesogenic microenvironments, and its inability to persistently colonize the inflamed gut limits therapeutic utility¹⁴. In the diabetic gut microenvironment, elevated levels of ROS generated from mitochondrial dysfunction and NADPH oxidase activity exacerbate intestinal barrier damage and inflammation, creating a vicious cycle that perpetuates metabolic dysfunction^15,16. Notably, microbial communities in this milieu face dual challenges with nutrient abundance (e.g., glucose and lipids) and environmental stressors (e.g., ROS and bile acids). Analogous to how bacteria under antibiotic stress upregulate energy-dependent efflux pumps^17,18, we hypothesized that probiotic subjected to sublethal ROS pressure would evolve enhanced antioxidant defenses and metabolic flexibility to thrive in hostile niches—a strategy that could be harnessed therapeutically.

ROS, particularly hydrogen peroxide (H₂O₂), serve as both a stressor and a signaling molecule in microbial systems^19,20. While H₂O₂ is less reactive than hydroxyl radicals (•OH), its relative stability allows for controlled induction of oxidative stress, mimicking the chronic ROS exposure observed in diabetic intestines. Here, we engineer a next-generation probiotic by subjecting EcN to iterative H₂O₂-driven evolution to select for strains that couple ROS detoxification with metabolic adaptations favoring energy-efficient redox balancing, followed by sequential modification with fructooligosaccharides (FOS) and CaCO₃ mineralization. While other prebiotics (e.g., inulin) exist, FOS has a shorter chain length than inulin, making it potentially more readily fermentable by a broader range of beneficial bacteria²¹. Unlike some encapsulation polymeric materials (e.g., chitosan, alginate) that may require complex synthesis or raise biocompatibility concerns, CaCO₃ is a Food and Drug Administration–approved antacid²². Its mineralization process is simple, scalable, and results in a highly stable, protective shell²³. We hypothesized that this ROS-tolerant EcN (REcN-F/Ca) would resist luminal oxidative stress, reprogram host lipid metabolism, and rectify high-fat diet-induced microbiota dysbiosis through short-chain fatty acids-dependent mechanisms. Using a combination of transcriptomics, metabolomics, and gnotobiotic models, we demonstrate that REcN-F/Ca shows superior ability in restoring glucose homeostasis, attenuating adipose inflammation, and enriching butyrate-producing taxa (Blautia and Lachnospiraceae). Our work establishes a blueprint for engineering multifunctional probiotic that targets both microbial and host metabolic networks, bridging a critical gap in translational microbiome therapeutics.

Results

ROS-tolerant probiotic design and antioxidant mechanisms exploration

To induce reactive oxygen species (ROS) tolerance in Escherichia coli Nissle 1917 (EcN) probiotic, hydrogen peroxide (H₂O₂) was selected as the model ROS. The minimum inhibitory concentration (MIC) assay was implemented to quantify bacterial tolerance evolution. Serial passages (n = 10) of EcN were performed in 96-well plates containing H₂O₂ gradients, with concentrations decreasing twofold from column 1 to 12 (Supplementary Fig. 1). Each H₂O₂ concentration was tested in triplicate. Following 24 h co-culture, MIC values were defined as the lowest H₂O₂ concentration in clear wells adjacent to turbid growth-positive wells (Supplementary Fig. 2). Wild-type EcN (WT, passage zero, P0) surviving the highest H₂O₂ concentrations were iteratively passaged under identical selective pressure for ten cycles (P1-P10). The H₂O₂ MICs were determined after each passage. Progressive MIC elevation across passages demonstrated successful development of ROS-tolerant EcN (REcN) (Fig. 1a and Supplementary Fig. 2).

Fig. 1: Enhancement of metabolic functions and ROS tolerance.

a The fold increase in MIC of probiotic following iterative passages, indicating enhanced tolerance to oxidative stress. b GO annotations analysis highlighting the upregulation of genes associated with functions in REcN compared to EcN. c GSEA depicting the overrepresentation of genes involved in the oxidoreductase complex in REcN. d Heatmap illustrating the relative expression levels of antioxidant enzymes in REcN compared to EcN. Color scale from blue to red indicates low to high relative gene expression. e O₂^·- scavenging ability of probiotic measured by the Total Superoxide Dismutase Kit with WST-8. Data were obtained from independent samples (n = 3). Error bars represent the mean ± standard deviation, with significance level indicated as p < 0.0001 (****). A two-sample Student two-sided t test (e) was utilized for the statistical analysis. f H₂O₂ scavenging ability of probiotic measured by a dissolved oxygen meter. g GSEA depicting the overrepresentation of genes involved in the H₂S biosynthetic process in REcN. h Representative lead acetate-soaked filter papers exhibit a brown stain of lead sulfide, resulting from the reaction with gaseous H₂S. i The corresponding fluorescent images in the EcN and REcN strains treated with the WSP5 fluorescent H₂S probe as indicated by green fluorescence. Scale bars: 100 μm. (n  =  3 independent samples) Source data are provided as a Source Data file.

Subsequently, transcriptomic profiling via RNA sequencing elucidated molecular mechanisms underlying the adaptive phenotype of REcN. Comparative analysis revealed significant inter-strain expression heterogeneity through violin plots (Supplementary Fig. 3a), suggesting divergent transcriptional responses to sublethal H₂O₂ exposure. Differential expression analysis identified 912 upregulated (log₂FC > 1, p < 0.05) and 940 downregulated (log₂FC < − 1, p < 0.05) genes in REcN versus EcN (Supplementary Fig. 3b), potentially associated with enhanced antioxidant pathways. Hierarchical clustering of expression profiles demonstrated intra-group homogeneity, with REcN and EcN strains forming distinct clusters (Supplementary Fig. 3c), indicative of conserved strain-specific molecular signatures. Principal component analysis (PCA) further confirmed transcriptional divergence between REcN and EcN cohorts (Supplementary Fig. 3d), supporting the emergence of unique molecular features in engineered strains.

Gene Ontology (GO) annotation analysis was performed to systematically characterize the functional profiles of differentially expressed genes in REcN²⁴. The Comparative analysis of upregulated genes in REcN versus wild-type EcN revealed significant enrichment in three core functional categories: antioxidant activity (GO:0016209), molecular transducer activity (GO:0060089), and catalytic activity (GO:0003824) (Fig. 1b). This enrichment pattern suggests adaptive remodeling of stress response and metabolic regulatory networks in REcN. Gene set enrichment analysis (GSEA) further identified pronounced enrichment (normalized enrichment score [NES] = 1.15) of the oxidoreductase complex (GO:1990204) in REcN (Fig. 1c), implicating redox homeostasis as a key evolutionary trajectory under H₂O₂ selection pressure.

Subtype analysis of oxidoreductase-associated GO terms highlighted specific enrichment in oxidoreductase activities, especially those involving NAD(P)H and quinone or similar compounds (Supplementary Fig. 4a), suggested that the REcN undergone significant metabolic adaptations. NADH dehydrogenases are key components of the electron transport chain, playing a central role in energy metabolism and ROS management²⁵. The NADH dehydrogenase activity (GO:0003954, Supplementary Fig. 4b) and oxidoreductase activity, NAD or NADP as acceptor (GO:0016620, Supplementary Fig. 4c) were significantly enriched. To functionally validate the enhanced energy metabolism, we measured the NADH dehydrogenase activity in EcN and REcN. Consistent with the RNA-seq data, REcN exhibited a significantly greater decrease in NADH absorbance at 340 nm over a 30 min incubation compared to the wild-type EcN strain, confirming its superior NADH dehydrogenase activity (Supplementary Fig. 5). These metabolic adaptations likely enhance electron transport chain efficiency, thereby coupling energy metabolism with ROS mitigation. Notably, whole-genome sequencing (WGS) with single‑nucleotide polymorphisms (SNPs) and insertions/deletions (InDels) analysis revealed that 6 missense mutations were found in the nuoCD gene (Table 1 and Supplementary Data 1), which encodes NADH-quinone oxidoreductase subunit C/D (Supplementary Data 2). This offers a plausible genetic basis for the observed increase in NADH dehydrogenase activity, which we confirmed through the enzymatic assay. Directed H₂O₂ selection pressure fixed mutations in a key metabolic component, which led to stable transcriptional reprogramming, resulting in the optimized functional phenotype.

Table 1 SNPs/InDels analysis of nuoCD gene in REcN compared to EcN

Targeted quantification of antioxidant enzymes demonstrated marked upregulation of sodA, sodB, sodC, and katG in REcN (Fig. 1d), corresponding to functional validation through enzymatic assays. Distinctly, REcN exhibited 1.8-fold higher superoxide dismutase (SOD) activity than EcN (Fig. 1e), catalyzing the disproportionation of superoxide radical (O₂·^-) to produce H₂O₂ and O₂. The generated H₂O₂ could be further scavenged by the catalase (CAT) in REcN, which was confirmed by the consumption of H₂O₂ to O₂ (Fig. 1f), collectively confirming systemic ROS-neutralizing capabilities. The transcriptomic data also indicted that the expression of soxR was downregulated, while the expression of oxyR was markedly upregulated in REcN (Supplementary Fig. 6). This shift in regulatory network activity aligns with the observed enhancement in peroxide-scavenging capabilities and suggests a prioritized adaptation to H₂O₂ stress.

GO analysis at the biological process level revealed significant enrichment in hydrogen sulfide (H₂S) biosynthetic process (GO:0070814) and metabolic process (GO:0070813) in REcN (Supplementary Fig. 7a). Further GSEA corroborated these findings, demonstrating strong enrichment signals for H₂S biosynthesis (NES = 1.15) and sulfur metabolism (NES = 1.11) in REcN (Fig. 1g and Supplementary Fig. 7b). These pathways may synergistically enhance antioxidant defenses, as H₂S functions as a potent reducing agent capable of directly scavenging ROS^26,27, while sulfur metabolism supports cysteine biosynthesis—a critical precursor for glutathione synthesis²⁸.

Transcriptomic profiling identified coordinated upregulation of H₂S biosynthesis genes cysH, cysC, cysJ, cysI, cysN, and cysD in REcN (Supplementary Fig. 7c). Notably, sseA—encoding 3-mercaptopyruvate sulfurtransferase (3-MST), a key enzyme in the non-canonical H₂S production pathway—showed 1.9-fold induction (Supplementary Fig. 7d)²⁹. Therefore, we further employed the classic lead acetate reactivity test to detect H₂S (Supplementary Fig. 8)³⁰. In this method, a filter paper saturated with a 2% lead acetate solution was positioned above a well plate containing the bacterial suspension. H₂S produced by the probiotic subsequently reacted with lead acetate to form a brown lead sulfide precipitate. The results indicated that the REcN strain exhibited a higher H₂S production after 24 h of incubation (Fig. 1h). This observation was further corroborated by the H₂S–nucleophilic substitution–cyclization–based fluorescent probe analysis, which demonstrated a significant increase in H₂S production in the REcN strain (Fig. 1i). To establish a direct causal link between H₂S and ROS tolerance, we supplemented the wild-type EcN with the H₂S donor NaHS (0.2 mM). This intervention dramatically increased its H₂O₂ MIC by 16-fold (Supplementary Fig. 9), demonstrating that H₂S confers a robust antioxidant phenotype.

To explore additional mechanisms of ROS tolerance, we first assessed whether evolutionary adaptation altered outer membrane permeability. The ANS binding assay revealed no significant difference between EcN and REcN (Supplementary Fig. 10), indicating that the ROS tolerance was not attributable to a reduced uptake of oxidative compounds. The growth curve analysis revealed that REcN exhibited a faster growth rate during the exponential phase compared to EcN, although both reach similar final cell densities (Supplementary Fig. 11). This enhanced growth kinetics likely contributes to its competitive advantage in the gut environment. These multilayered adaptations—spanning enzymatic ROS neutralization (SOD/CAT), redox cofactor recycling (NAD(P)H oxidoreductases), and H₂S-mediated antioxidant mechanisms—suggest potential applicability of REcN in managing oxidative stress-associated pathologies. The systemic enhancement of interconnected antioxidant networks positions this bioengineered probiotic as a candidate for further investigation in metabolic disorders linked to redox imbalance, including diabetes and obesity³¹.

REcN-driven metabolic reprogramming and therapeutic implications

Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis was employed to characterize metabolic adaptations in REcN³². Upregulated genes in REcN exhibited significant enrichment in lipid metabolism, energy metabolism, and carbohydrate metabolism (Fig. 2a). Specifically, scatterplot analysis highlighted pronounced activation of fatty acid degradation pathways (ko00071, Fig. 2b), with GSEA confirming enhanced lipid catabolic processes (NES = 1.48), fatty acid catabolic processes (NES = 1.29), and fatty acid degradation (NES = 1.15) in REcN (Fig. 2c–e).

Fig. 2: Metabolic pathway activation and gene expression analysis.

a The KEGG annotations of differentially expressed genes (upregulated in REcN compared to ECN), highlighting significant enrichment in pathways related to lipid metabolism, energy metabolism, and carbohydrate metabolism. b KEGG enrichment analysis focusing on metabolism, with significant enrichment observed in pathways such as oxidative phosphorylation and fatty acid degradation. Pathways are shown by p.adjusted. c–e GSEA depicting the overrepresentation of genes involved in the lipid catabolic process (c), fatty acid catabolic process (d), and fatty acid degradation (e) in REcN. f Oil red O staining in 3T3L1 adipocytes treated without or with the culture supernatant of probiotics. Red indicatesa lipid droplet. Scale bars: 50 μm. (n  =  3 independent samples). g Schematic diagram of glycolysis and the TCA cycle, illustrating key enzymes and intermediates. h,i The α-GC (h) and α-AMS (i) inhibition percentage between ECN and REcN strains. Data were obtained from independent samples (n = 3). Error bars represent the mean ± standard deviation, with exact p-values. A two-sample Student two-sided t test (h, i) was utilized for the statistical analysis. Source data are provided as a Source Data file.

Functional validation using Oil Red O staining demonstrated a significant reduction in cytoplasmic lipid droplet accumulation in 3T3-L1 adipocytes cultured with REcN-conditioned medium (Fig. 2f and Supplementary Fig. 12), quantitatively corroborating transcriptome-predicted lipid catabolic enhancement. Given the established link between T2DM and dysregulated lipid/carbohydrate metabolism^33,34, these findings suggest REcN may modulate host metabolic homeostasis through microbiome-lipid axis interactions.

Concurrently, KEGG analysis revealed marked upregulation of energy transduction pathways in REcN, including oxidative phosphorylation (ko00190) and the electron respiratory chain (Fig. 2b and Supplementary Fig. 13a–c). Notably, the tricarboxylic acid (TCA) cycle pathway (ko00020) was significantly enhanced in REcN (Fig. 2b and Supplementary Fig. 13d), with a large number of associated genes being significantly upregulated (Supplementary Fig. 14). The upregulation of key pathway, such as pyruvate metabolism, in the TCA cycle was observed (Fig. 2b), suggesting that the REcN strain possesses enhanced energy production capabilities. The TCA cycle is a central metabolic pathway that plays a crucial role in energy production by generating ATP and NADH³⁵. The enhanced expression of enzymes such as isocitrate dehydrogenase (icd, 1.9-fold), succinate dehydrogenase (sdhB, 2.7-fold), and malate dehydrogenase (mdh, 1.3-fold) indicated that the REcN strain was able to utilize carbon sources and produce energy (Fig. 2g and Supplementary Fig. 15)³⁶. Efficient energy production and utilization are essential for maintaining cellular homeostasis and coping with environmental challenges³⁷. This metabolic flexibility may endow REcN with superior capacity to withstand oxidative stress.

Moreover, both EcN and REcN probiotics exhibited inhibitory effects on α-glucosidase (α-GC) (Fig. 2h), which can decelerate glucose absorption and thereby assist in reducing blood glucose levels³⁸. In addition, the probiotic demonstrated a pronounced inhibitory impact on α-amylase (α-AMS) (Fig. 2i), thereby diminishing starch degradation and subsequent glucose release³⁹. Collectively, superior metabolic plasticity of REcN—integrating enhanced lipid catabolism, TCA cycle flux, and antioxidant networks—positions it as a multifaceted candidate for mitigating metabolic syndrome-associated dysregulation.

Surface engineering and gastrointestinal stability of REcN

To augment the gastrointestinal (GI) tract survivability and bioactivity of REcN, sequential surface modifications were implemented using fructooligosaccharide (FOS) prebiotics and calcium carbonate (CaCO₃) mineralization. Transmission electron microscopy (TEM) imaging revealed a homogeneous electron-lucent layer (a thickness of approximately 60 nm) encapsulating the REcN cell (Fig. 3a, b and Supplementary Fig. 16, red arrow), confirming successful FOS polymerization (designated REcN-FOS). Subsequent CaCO₃ mineralization generated a dense inorganic shell on REcN-FOS surfaces (Fig. 3c), with energy-dispersive X-ray spectroscopy (EDS) mapping demonstrating uniform Ca, C, and O distributions (Fig. 3c). Detection of nitrogen (N), a microbial biomass signature, confirmed bacterial structural integrity post-modification (designated REcN-FOS/CaCO₃). X-ray diffraction (XRD) analysis confirmed the successful mineralization with CaCO₃, identifying the crystalline phase as calcite (Supplementary Fig. 17).

Fig. 3: Survival and stability of engineered probiotic in simulated GI environments.

a–c The TEM images of REcN (a), REcN-FOS (b), and REcN-FOS/CaCO₃ (c), the red arrow indicates the FOS modification, the corresponding elemental maps showing the distribution of Ca (yellow), C (red), O (green), and N (blue). Scale bars: 500 nm. (n  =  3 independent samples). d–f Probiotic viability in SGF (d), SGF + SIF (e) and SGF + SIF + SICF (f). g •OH-scavenging ability of probiotic treated by SGF, SGF + SIF, and SGF + SIF + SICF, respectively, detected by DMPO probe. h–j Gene expression of IL-10 (h), IL-1β (i), and Tnf-α (j). k, l Representative IVIS images (k) and the corresponding intensity (l) of luminescence signals of the intestinal tracts after oral gavage of probiotics for 6 h. Data were obtained from independent samples (n = 3). Error bars represent the mean ± standard deviation, with exact p-values or significance level indicated as p < 0.0001 (****). A one-way analysis of variance with the Tukey multiple-comparisons test (d–f, l) or two-sample Student two-sided t test (h–j) was utilized for the statistical analysis. Source data are provided as a Source Data file.

Subsequently, to rigorously evaluate the protective efficacy of the CaCO₃ shell during sequential GI transit, we subjected the probiotic to a simulated journey from the stomach to the inflamed colon. Given the GI transit times, with rapid gastric emptying occurring within approximately 20 min, followed by small intestinal transit for about 2 h, and ultimately reaching the inflamed colon⁴⁰, the viability and ROS-scavenging capacity of EcN, REcN, and REcN-FOS/CaCO₃ were assessed after first incubation for 20 min in simulated gastric fluid (SGF), then incubation for 2 h in simulated intestinal fluid (SIF), and finally incubation for 2 h in simulated inflammatory colon fluid (SICF). After each sequential challenge, REcN demonstrated greater activity than EcN (Fig. 3d–f), likely due to the superior energy metabolism that enables it to better withstand GI environmental assaults. Notably, the modification with CaCO₃ further enhanced the viability of the REcN probiotic, confirming the robust protective role of CaCO₃ in the digestive tract. Importantly, after each sequential challenge, REcN-FOS/CaCO₃ demonstrated the strongest •OH scavenging capacity (Fig. 3g), suggesting preserved bioactivity under GI stress. To determine the origin of the antioxidant activity, we compared •OH scavenging across different components. Live REcN demonstrated the highest capacity, significantly exceeding that of live EcN (Supplementary Fig. 18). The lysates of both strains also showed substantial activity, indicating that stable bacterial components contribute significantly. The FOS/CaCO₃ material itself possessed a baseline scavenging ability, confirming that the overall effect of REcN-FOS/CaCO₃ is a composite of the evolved superior intrinsic ability, its released components, and the coating matrix. To functionally validate the pH-responsive behavior, we quantified the dissolution kinetics of the CaCO₃ shell. Inductively coupled plasma optical emission spectrometry (ICP-OES) analysis revealed a rapid release of Ca²⁺ ions in SGF (pH 1.2, ~ 50 mg L⁻¹ after 20 min), contrasting with a slow, sustained release in SIF (pH 6.8, ~ 18 mg/L after 2 h) and SICF (pH 7.8, ~ 12 mg/L after 2 h) (Supplementary Fig. 19). This controlled dissolution in the upper GI tract ensures that a maximal payload of viable REcN is delivered to the target colonic environment.

Quantitative real-time PCR (qRT-PCR) analysis revealed a significant 2.7-fold upregulation of the anti-inflammatory cytokine IL-10 transcript concomitant with a suppression of pro-inflammatory IL-1β and Tnf-α expression in macrophage cultures exposed to engineered probiotic (Fig. 3h–j). Furthermore, the modified probiotic exhibited superior cytocompatibility and hemocompatibility (Supplementary Fig. 20).

To directly evaluate the gastrointestinal colonization capacity of the engineered probiotic, an ex vivo tracking experiment using Cy5.5-labeled strains was performed. Quantitative fluorescence imaging of intestinal tissues revealed that REcN-FOS/CaCO₃ achieved the highest retention level among all groups (Fig. 3k, l). The FOS prebiotics improved the retention of both strains compared to their unencapsulated counterparts. Notably, the fluorescent signal of REcN-FOS/CaCO₃ was significantly greater than that of EcN-FOS/CaCO₃, indicating that the ROS-tolerant phenotype of REcN also enhanced gut colonization.

Amelioration of phenotypes in T2DM and obesity murine model

To establish causal links between REcN-FOS/CaCO₃ (REcN-F/Ca) supplementation and metabolic improvement, a high-fat diet (HFD)-induced obesity murine model was employed. Compared to normolipidic diet (ND)-fed controls, HFD-fed mice developed pronounced obesity phenotypes, evidenced by elevated body weight gain (Δ + 71.9% versus Δ+43.5%), peritesticular fat accumulation, and higher epididymal adipose tissue (eAT) weight (Fig. 4a–d). In addition, the wild-type EcN probiotic encapsulated within the protective shell demonstrated a mitigating effect on obesity, which can be attributed to the action of the EcN probiotic. The F/Ca shell showed a negligible inhibitory effect on obesity. However, oral administration of REcN-F/Ca during HFD feeding mitigated obesity development, as evidenced by reduced body weight gain and fat accumulation compared to HFD-fed mice that did not receive supplementation (Fig. 4a–d).

Fig. 4: Systemic metabolic remediation by REcN-F/Ca in HFD-induced obesity and T2DM.

a Photographs of mice at the end of the study period, illustrating body weight reduction after REcN-F/Ca treatment. Scale bars: 2 cm. b Photographs of dissected mice highlighting reduced epididymal fat accumulation in the HFD + REcN-F/Ca group. Scale bars: 2 cm. c Body weight progression over 10 weeks in mice fed with an ND, HFD, HFD supplemented with EcN-F/Ca, HFD supplemented with F/Ca, and HFD supplemented with REcN-F/Ca, showing significant weight reduction with REcN-F/Ca treatment. d The weight of eAT in different groups. e The homeostatic model assessment of insulin resistance (HOMA-IR = [glucose (mg dL⁻¹) × insulin (μU mL⁻¹)]/405). f, g Dynamic metabolic profiling during intraperitoneal glucose tolerance test (IPGTT): glucose-stimulated insulin secretion kinetics (f) and blood glucose (g). h Multiplex immunofluorescence of eAT sections: macrophages (F4/80, red); M1 macrophages (CD86, green); nuclei (DAPI, blue), Scale bars: 100 μm. i Immunofluorescence of eAT sections: lipid droplets (BODIPY, green); nuclei (DAPI, blue), Scale bars: 200 μm. j Quantitation of lipid droplets infiltration. k, l Transcriptional regulation in mature adipocyte fraction: a chemokine for recruiting macrophages S100a8 (k) and a proinflammatory cytokine Tnfα (l). m Gene expression of Pparα in colon tissues. n Immunofluorescence of colon tissues: 8-OHdG (green); nuclei (DAPI, blue), Scale bars: 100 μm. o Quantitation of oxidative stress infiltration. Data were obtained from independent samples (n = 6). Error bars represent the mean ± standard deviation, with exact p-values or significance level indicated as p < 0.0001 (****). A one-way analysis of variance with the Tukey multiple-comparisons test was utilized for the statistical analysis. Source data are provided as a Source Data file.

Metabolic profiling revealed HFD-driven hyperglycemia and insulin resistance, both ameliorated by REcN-F/Ca (Fig. 4e). During intraperitoneal glucose tolerance testing (IPGTT), ND-fed mice exhibited transient increases in blood insulin levels following intraperitoneal glucose injection, with a rapid return to baseline (Fig. 4f). In contrast, HFD-fed mice exhibited a significant decline in blood insulin levels following the glucose challenge (Fig. 4f). This pattern is indicative of failure to mount an appropriate secretory response, a hallmark of T2DM pathology, rather than mere insulin resistance⁴¹. On the contrary, in mice fed with F/Ca, the blood insulin levels continued to rise after glucose injection, which is a typical characteristic of insulin resistance. This minor improvement was metabolically insufficient, as evidenced by the poor overall glucose tolerance and high HOMA-IR, which remained comparable to the HFD group. However, HFD-fed mice supplemented with REcN-F/Ca demonstrated insulin responses comparable to those of ND-fed mice (Fig. 4f). Consistent with these observations, the increases in blood glucose levels following intraperitoneal glucose injection in HFD-fed mice were normalized to levels similar to those of ND-fed mice when supplemented with REcN-F/Ca (Fig. 4g).

Inflammation within the eAT is known to contribute to the development of obesity-induced diabetes⁴¹. Histopathological evaluation of eAT in HFD and HFD + F/Ca groups revealed M1 macrophage infiltration (Fig. 4h), a hallmark of obesity-associated adipose inflammation. In line with the inhibitory effects of REcN on diabetes pathogenesis, the eAT of HFD-fed mice supplemented with REcN-F/Ca exhibited minimal macrophage accumulation (Fig. 4h). The administration of REcN significantly also reduced the number of lipid droplets infiltrating the eAT of HFD-fed mice (Fig. 4i,j). In addition, RT-qPCR analysis indicated that the expression of S100a8 and Tnfα (Fig. 4k, l), which encode a chemokine that recruits macrophages and an inflammatory cytokine, respectively, was downregulated in the mature adipocyte fraction of the eAT in HFD-fed mice supplemented with REcN-F/Ca. Therefore, the administration of REcN-F/Ca not only inhibited body weight gain but also attenuated inflammatory responses in the eAT of HFD-fed mice.

Notably, colonic Pparα expression was upregulated in REcN-F/Ca-treated HFD mice versus ND mice (Fig. 4m). As a nuclear receptor regulating lipid oxidation and anti-inflammatory pathways, PPARα activation may mechanistically link the metabolic benefits of REcN to transcriptional reprogramming of host energy utilization and inflammatory responses. Owing to the superior antioxidant capacity of REcN, the colon tissues after therapy showed the least oxidative stress (Fig. 4n, o). Furthermore, histopathological analysis via hematoxylin and eosin (H&E) staining demonstrated distinct hepatic morphological alterations across experimental cohorts. The ND group maintained intact hepatic lobular architecture with characteristic radial hepatocyte arrangement around central veins (Supplementary Fig. 21). Conversely, the HFD group exhibited pronounced microvesicular and macrovesicular steatosis, accompanied by hepatocyte ballooning degeneration. Notably, REcN-F/Ca intervention substantially attenuated HFD-induced hepatosteatosis. Critical extrahepatic organs (heart, spleen, lung, and kidney) displayed preserved histoarchitectural integrity across all groups, suggesting the absence of multiorgan toxicity after REcN-F/Ca treatment.

PPAR signaling pathway mediated lipid metabolism

To delineate the molecular mechanisms underlying the metabolic improvements conferred by REcN in obesity and T2DM, transcriptomic profiling of colonic tissues was conducted post-treatment. GO enrichment analysis revealed that REcN probiotic treatment significantly enhanced the capacity of colonic tissue for lipid and fatty acid metabolism and catabolism (Fig. 5a and Supplementary Fig. 22). The KEGG enrichment chord diagram further validated enhanced fatty acid degradation (ko00071) and fat digestion/absorption (ko04975), alongside upregulated steroid hormone biosynthesis (ko00140), pentose-glucuronate interconversions (ko00040), and glycerolipid metabolism (ko00561) (Fig. 5b). Glucocorticoids, which are part of the steroid hormone biosynthesis pathway, can maintain blood glucose stability by inhibiting glucose utilization and promoting lipolysis. They also possess potent anti-inflammatory and immunosuppressive properties, commonly utilized in the treatment of autoimmune diseases⁴². The pentose and glucuronate interconversions pathway is not only involved in energy production but also in the synthesis of NADPH⁴³, which is crucial for the enhancing antioxidant capacity of the colonic tissue. Modulation of glycerolipid metabolism can improve insulin sensitivity and lower blood glucose levels⁴⁴, thereby aiding in diabetes management. Transcriptomic profiling revealed significant downregulation of key pro-inflammatory signaling cascades following REcN-F/Ca intervention, including IL-17 signaling pathway, TNF signaling pathway, and NF-kappa B signaling pathway (Supplementary Fig. 23). These findings position REcN-F/Ca as a promising immunomodulatory biotherapeutic for inflammation-driven T2DM, where chronic low-grade inflammation perpetuates insulin resistance and ectopic lipid accumulation.

Fig. 5: Molecular pathway analysis and gene expression in response to REcN-F/Ca treatment.

a, b Circular chord diagram depicting the relationship between upregulated genes and associated GO terms (a) and KEGG pathways (b) in colon tissues after REcN-F/Ca treatment, highlighting the impact of REcN treatment on metabolic processes. c–l Relative gene expression levels of key enzymes in lipid metabolism, including Slc27a4 (c), Fabp1 (d), Fabp6 (e), Apoa1 (f), Apoa5 (g), Lpl (h), Ehhadh (i), Acaa1b (j), Acox2 (k), and Cpt1b (l). Data were obtained from independent samples (n = 6). Error bars represent the mean ± standard deviation, with exact p-values or significance level indicated as p < 0.0001 (****). A two-sample Student two-sided t test was utilized for the statistical analysis. m Schematic diagram of the PPAR signaling pathway, illustrating the potential role of REcN in modulating this pathway to influence lipid metabolism. Red indicates upregulated genes, proteins or pathways. Source data are provided as a Source Data file.

Notably, the PPAR signaling pathway (ko03320) was also significantly enriched and upregulated (Fig. 5b). The PPAR signaling pathway emerged as a central regulatory node and coordinated with the adipocytokine signaling pathway in energy metabolism and inflammatory responses. Moreover, several key genes, including Slc27a4, Fabp1, Fabp6, Apoa1, Apoa5, Lpl, Ehhadh, Acaa1b, Acox2, and Cpt1b, were significantly upregulated (Fig. 5c–l), thereby enhancing lipid transport, fatty acid transport, fatty acid oxidation, and ultimately lipid metabolism in colonic tissues following probiotic treatment (Fig. 5m and Supplementary Fig. 24). This transcriptional reprogramming establishes a metabolic network favoring lipid mobilization over storage (Fig. 5m), potentially mediated by PPARα-driven crosstalk with adipocytokine signaling pathways to enhance insulin sensitivity and mitigate inflammation.

To establish a causal role for the PPARα pathway, we pharmacologically inhibited PPARα signaling in REcN-F/Ca-treated mice using the antagonist GW6471. The intervention largely abolished the metabolic benefits conferred by REcN-F/Ca. Specifically, we observed a significant rebound in body weight, eAT weight, HOMA-IR index, and a marked deterioration in both insulin secretion and blood glucose levels during tolerance tests (Supplementary Fig. 25). The metabolic profile of this group reverted to a state closely resembling that of the HFD control group, demonstrating that the host PPARα pathway is indispensable for the metabolic remediation by REcN-F/Ca.

Modulation of gut microbiome

Given the well-documented ability of probiotic to modulate the gut microbiota and confer beneficial effects on the host⁴⁰, fecal samples from each group were subjected to 16S rRNA gene sequencing after 10 weeks of intervention. Alpha diversity analysis revealed significant reductions in microbial richness (Observed species [Sobs] index: HFD 94 ± 34 versus ND 350 ± 99; Abundance-based coverage estimator [Ace] index: HFD 111 ± 41 versus ND 383 ± 96) and diversity (Shannon index: HFD 1.8 ± 0.3 versus ND 3.1 ± 0.9) in HFD-fed mice, which were partially restored by REcN-F/Ca treatment (Sobs index: 212 ± 22; Ace index: 238 ± 23; Shannon index: 2.8 ± 0.2) (Fig. 6a–c). The Chao index, Simpson index, and Shannon index all indicated that HFD-fed reduced the richness and diversity of the gut microbiota, while REcN probiotic treatment can restore the disrupted intestinal microbiota (Supplementary Fig. 26), suggesting that this bioengineered probiotic can help restore a healthier gut microbiota composition¹⁴. Furthermore, Beta diversity analysis via principal component analysis (PCA) demonstrated distinct clustering of gut microbiota communities among ND, HFD, and HFD+REcN-F/Ca groups (ANOSIM R = 0.2569, p = 0.014; Fig. 6d), with hierarchical clustering and NMDS further confirming compositional divergence (Supplementary Fig. 27). This change is likely due to the beneficial effects of the engineered probiotic, which can modulate the gut microbiota to promote health.

Fig. 6: Gut microbiota restructuring mediated byREcN-F/Ca intervention.

a–c Alpha diversity metrics of faecal microbiome quantified via Sobs (a), Ace (b), and Shannon (c) index on OUT level after different treatment. d Beta diversity profiling through PCA after different treatments. e The relative abundance of microbiota composition at the family level across ND, HFD, and HFD+REcN-F/Ca groups. f LDA score plot identifying differentially abundant bacterial taxa between groups using the LEfSe method, with significant changes observed in Blautia and Lachnospiraceae. The prefix of “c_”, “o_”, “f_”, and “g_” refer to class, order, family, and genus levels respectively. Unclassified genera denoted as “norank_f_” and “norank_o_”. g, h Relative abundance of Lachnospiraceae at the family level (g) and Blautia at the genus level (h) in the gut microbiome after different treatment. i–l The concentrations of SCFAs, including acetic acid (i), propionic acid (j), butyric acid (k), and isobutyric acid (l), significantly altered by HFD and modulated by REcN-F/Ca treatment. Data were obtained from independent samples (n = 4) (a–c, g, and h), (n = 6) (i–l). Error bars represent the mean ± standard deviation, with exact p-values or significance level indicated as p < 0.0001 (****). The Mann-Whitney test was employed to analyze inter-group differences in alpha diversity (a–c). A one-way analysis of variance with the Tukey multiple-comparisons test was utilized for the statistical analysis in (g–l). Source data are provided as a Source Data file.

Taxonomic profiling at the family and genus levels identified REcN-F/Ca-driven enrichment of Lachnospiraceae (ND: HFD: HFD+REcN-F/Ca = 13.14%: 4.26%: 32.15%) (Fig. 6e and Supplementary Fig. 28), negatively correlated with systemic inflammation in T2DM^45,46. Lachnospiraceae is associated with butyric acid production, promoting the growth of host epithelial cells. Conversely, pro-inflammatory Streptococcaceae abundance was reduced after REcN-F/Ca treatment (HFD 3.68% versus HFD+REcN-F/Ca 0.04%). Linear discriminant analysis (LDA) effect size (LEfSe) analysis highlighted Lachnospiraceae (LDA score = 4.88) and Blautia (LDA = 4.18) as signature taxa enriched in REcN-F/Ca-treated mice (Fig. 6f–h and Supplementary Fig. 29). The recent study identified that the reduction of Blautia luti and Blautia wexlerae is associated with insulin resistance in obese individuals^41,47. Blautia prevents pathogen colonization through the production of bacteriocins and exhibits anti-inflammatory properties and glucose homeostasis maintenance by upregulating the production of short-chain fatty acids (SCFAs)⁴¹. We also observed the specific enrichment of the [Eubacterium]_coprostanoligenes_group in the REcN-F/Ca group (Supplementary Fig. 30). These taxa are associated with butyrate production, pathogen exclusion via bacteriocins, and immunomodulation^{41,45,46,47,48}.

Given the critical role of SCFAs in maintaining gut barrier integrity, systemic metabolic homeostasis, and neuroendocrine signaling, we quantified fecal SCFAs levels across experimental groups. HFD-fed mice exhibited significant reductions in acetic acid (ND: 1.46 ± 0.51 versus HFD: 0.21 ± 0.13 μg mg^-1), propionic acid (ND: 0.29 ± 0.04 versus HFD: 0.04 ± 0.03 μg mg^-1), butanoic acid (ND: 0.15 ± 0.07 versus HFD: 0.011 ± 0.004 μg mg^-1), and isobutyric acid (ND: 0.023 ± 0.008 versus HFD: 0.007 ± 0.006 μg mg^-1) compared to ND control (Fig. 6i–l). REcN-F/Ca intervention restored SCFAs concentrations to the ND-equivalent levels (acetic acid: 1.45 ± 0.26 μg mg⁻¹; propionic acid: 0.19 ± 0.07 μg mg⁻¹; butanoic acid: 0.21 ± 0.05 μg mg⁻¹; isobutyric acid: 0.027 ± 0.008 μg mg⁻¹), concomitant with the observed enrichment of SCFAs (especially for butyrate-producing taxa (Lachnospiraceae, Blautia, and [Eubacterium]_coprostanoligenes_group).

This HFD-induced SCFAs depletion aligns with gut dysbiosis characterized by reduced microbial diversity and diminished beneficial taxa. REcN-F/Ca treatment partially restores microbiota diversity and abundance of beneficial bacteria, suggesting a potential therapeutic role in modulating gut health. The increase in Blautia and Lachnospiraceae, along with the restoration in SCFAs levels, suggests improved gut function and metabolic health.

Discussion

This study presents a multifaceted strategy for engineering next-generation probiotics with enhanced ROS tolerance and systemic antioxidant capacity, demonstrating their therapeutic efficacy in ameliorating metabolic syndrome. By subjecting EcN to iterative H₂O₂ selection pressure, we evolved a robust variant (REcN) exhibiting significantly elevated ROS tolerance—quantified by a progressive 32-fold increase in H₂O₂ MIC over ten passages. Transcriptomic profiling revealed that this adaptive phenotype stems from a coordinated rewiring of redox homeostasis networks, including the pronounced enrichment of oxidoreductase complexes and the upregulation of core antioxidant enzymes. Functional validation confirmed enhanced SOD and CAT activities of REcN, enabling efficient O₂·^- disproportionation and H₂O₂ detoxification—a critical advantage for probiotics navigating inflammatory microenvironments rich in phagocyte-derived ROS^31,49,50,51. The prior study primarily focused on protecting probiotics from ROS using externally applied nanomaterials or coatings that act as scavengers^12,40. We used directed evolutionary pressure (iterative H₂O₂ selection) to intrinsically enhance the ROS tolerance and metabolic resilience. This creates a robust chassis that does not solely rely on external protection.

Our analysis further uncovered a previously underappreciated role of microbial H₂S biosynthesis in probiotic ROS defense. REcN exhibited significant enrichment in H₂S metabolic pathways, accompanied by the upregulation of key enzymes and an increase in H₂S production. As H₂S directly scavenges •OH and peroxynitrite while sustaining cellular glutathione pools^28,52, this gaseous mediator constitutes a pivotal non-enzymatic antioxidant mechanism in engineered probiotics. The concurrent enhancement of NAD(P)H-dependent oxidoreductase activities further optimizes redox cofactor recycling—synergizing with enzymatic and H₂S-mediated defenses to create a multi-tiered ROS-neutralizing system. These adaptations collectively enable REcN to thrive in oxidative niches while delivering localized antioxidant effects.

Beyond antioxidant capabilities, REcN underwent profound metabolic reprogramming, favoring energy efficiency and substrate flexibility. KEGG analysis revealed marked upregulation of oxidative phosphorylation, TCA cycle, and fatty acid degradation pathways. This metabolic rewiring—characterized by the upregulation of Icd, SdhB, and Mdh—potentiates ATP synthesis while minimizing electron leakage, thereby reducing intrinsic ROS generation^53,54. Crucially, this metabolic plasticity translated into functional benefits: REcN-conditioned media reduced lipid droplet accumulation in adipocytes and inhibited α-glucosidase/α-amylase—enzymes whose overactivity exacerbates postprandial hyperglycemia in diabetes^38,39,55,56. Notably, the concurrent enhancement of the TCA cycle and sulfur metabolism suggests an evolutionary trade-off where energy production is coupled to ROS detoxification—a metabolic rewiring previously observed in extremophiles but rarely engineered into probiotics.

To ensure targeted delivery of REcN to inflammatory loci, we engineered a sequential FOS/CaCO₃ encapsulation system (REcN-F/Ca). The mineralized shell provided exceptional GI protection, maintaining viability and ROS scavenging capacity after exposure to SGF, SIF, and SICF. This stability is paramount for probiotics intended to modulate colonic inflammation, where luminal ROS concentrations can reach cytotoxic levels^57,58,59. In vivo assessments in HFD-fed mice demonstrated the ability of REcN-F/Ca to counteract core metabolic perturbations: it attenuated weight gain, restored glucose homeostasis, and suppressed adipose inflammation—reducing macrophage infiltration and downregulating Tnfα/S100a8 in epididymal fat. Mechanistically, these benefits were linked to probiotic-induced activation of the PPARα signaling axis in colonic tissue. REcN-F/Ca treatment significantly upregulated Pparα targets governing fatty acid oxidation and lipid transport, establishing a transcriptional framework favoring lipid mobilization over storage. Most importantly, the pharmacological inhibition of PPARα signaling almost abrogated the therapeutic effects of REcN-F/Ca, unequivocally demonstrating that the activation of this host signaling pathway is not merely associated with, but is strictly necessary for, the observed metabolic improvements. Given the dual role of PPARα in enhancing insulin sensitivity and repressing inflammation^60,61, this pathway provides a compelling molecular bridge between probiotic action and systemic metabolic improvement.

Concomitantly, REcN-F/Ca remodeled the dysbiotic microbiota in HFD mice, restoring α-diversity and enriching butyrogenic taxa (Lachnospiraceae, Blautia, and [Eubacterium]_coprostanoligenes_group). This mechanistic framework aligns with clinical observations linking Blautia abundance to improved glycemic control in prediabetic cohorts^41,47,62. This shift correlated with normalization of fecal SCFAs levels—particularly butyrate, which improves insulin sensitivity, enhances gut barrier integrity, and suppresses endotoxemia-driven inflammation^63,64. The depletion of pro-inflammatory Streptococcaceae further underscores the ability of REcN-F/Ca to rectify microbiome-immune crosstalk in metabolic disease.

While our study establishes proof-of-concept for bioengineered probiotics in metabolic disease, long-term safety assessments are warranted, given concerns that chronic ROS tolerance could inadvertently promote plasmid-borne antibiotic resistance⁶⁵. Crucially, EcN is one of the few well-characterized, common probiotic strains that is Gram-negative. We have found that Gram-negative bacteria possess inherent structural and metabolic features that make them particularly effective at purifying and tolerating ROS-rich environments⁶⁶. The presence of an outer membrane and specific periplasmic redox systems provides a foundational advantage in managing oxidative stress. Therefore, Gram-negative strains like EcN represent an ideal starting chassis for ROS-targeted engineering. Future work will indeed explore applying this biomimetic engineering paradigm to other promising Gram-negative candidates, as well as to more robust Gram-positive probiotics, to create a suite of tailored therapeutics for various diseases. In addition, future investigations should systematically determine the optimal therapeutic dose and the minimum effective dose. A comprehensive dose-response study represents a critical and logical next step in the translational development pathway, essential for maximizing therapeutic outcomes while ensuring safety and cost-effectiveness in potential future clinical applications. Furthermore, the therapeutic efficacy was demonstrated in a prevention model, wherein probiotic administration was initiated concurrently with the high-fat diet. Although this design effectively shows the capacity of REcN-F/Ca to counteract metabolic dysfunction, future studies employing an intervention model—where treatment begins after the establishment of obesity and insulin resistance—are essential to fully assess its clinical translational potential for treating existing disease. Finally, this investigation was conducted exclusively in male mice. Given the well-documented sexual dimorphism in immune responses, gut microbiota composition, and metabolic disease progression, evaluating the efficacy and mechanism of action of REcN-F/Ca in female subjects constitutes a critical and immediate future direction. Addressing these aspects will be vital for comprehensively understanding the therapeutic scope of our bioengineered probiotic.

This study demonstrates that iterative H₂O₂-driven adaptive evolution and surface engineering synergistically transform EcN into a robust therapeutic agent, REcN-F/Ca, capable of targeting the oxidative, metabolic, and dysbiotic hallmarks of T2DM and obesity. By simulating the ROS-enriched diabetic gut microenvironment through sublethal H₂O₂ exposure, we selected for EcN variants with enhanced energy metabolism and antioxidant defenses, enabling efficient ROS neutralization via enzymatic scavenging and H₂S-mediated redox balancing. The dual FOS-CaCO₃ coating further augmented gastrointestinal survivability and sustained delivery of therapeutic effects, driving PPAR signaling activation to restore lipid metabolism and enriching butyrate-producing taxa (Blautia and Lachnospiraceae) that rescued SCFAs depletion. These adaptations collectively attenuated HFD-induced metabolic dysfunction, reducing adiposity, insulin resistance, and adipose inflammation. Our findings validate ROS preconditioning as a biomimetic strategy to engineer probiotics that thrive in pathological niches, bridging microbial resilience with host metabolic regulation—a paradigm with broad applicability for microbiome-targeted therapies.

Methods

Research ethics statement

The animal experiments investigating the therapeutic effects on high-fat diet-induced type 2 diabetes were conducted with approval from the Institutional Animal Care and Use Committee of the Beatcell Model Biological Center (Approval No. 2025-07-31 A). All animal procedures were performed in strict compliance with the “Animal Management Regulations of the Ministry of Health of the People’s Republic of China” and the “Guidelines for the Care and Use of Laboratory Animals (Chinese version)”.

Chemicals

All chemical reagents were of analytical grade or higher purity. Hydrogen peroxide, glycerol, norepinephrine, fructooligosaccharides, calcium chloride, sodium carbonate, 3-isobutyl-1-methylxanthine, dexamethasone, insulin, paraformaldehyde, lead(II) acetate, and dimethylformamide were procured from Aladdin Regents. Acetonitrile, methanol, isopropanol, and formic acid were obtained from Fisher Scientific. Molecular biology reagents comprised PrimeScript RT Master Mix (Takara Bio, #RR036A), TB Green Premix Ex Taq II (Takara Bio, #RR820A), TransStart FastPfu DNA Polymerase (TransGen Biotech, #AP221-01), and dNTP mixtures (Beyotime, #D7371). Commercial kits included: Total Superoxide Dismutase Assay Kit (Beyotime, #S0101S), α-Amylase Assay Kit (Shanghaiyuanye, #S31302), α-Glucosidase Activity Assay Kit (Solarbio, # BC2550), Total RNA Kit (Omega Bio-tek, #R6834), FastPure Microbiome DNA Isolation Kit (Vazyme, #DC502-01), NEXTFLEX Rapid DNA-Seq Kit (Bioo Scientific, #5144-03).

Bacterial strains and cultivation

The wild-type Escherichia coli Nissle 1917 (EcN) was obtained from BeNa Culture Collection Co., Ltd (Beijing, China). The strain was routinely cultured in Luria-Bertani (LB) broth at 37 °C with shaking. For all experiments, after overnight growth, the bacterial strain was transferred to fresh media and permitted to propagate until they reached the exponential phase. A 1000-fold dilution was subsequently conducted using the LB broth, resulting in a suspension with approximately 10⁶ colony-forming units (CFU) per milliliter for subsequent experimental procedures. Stock solutions of each microbial passage were prepared in the LB broth, supplemented with 25% glycerol, and stored at -80 °C for long-term preservation.

Selection of ROS-resistant strain

The initial antimicrobial efficacy of ROS, specifically H₂O₂, was evaluated through standard planktonic minimum inhibitory concentration (MIC) assays. These experiments were performed in triplicate utilizing 96-well microplates. A volume of 100 µL of H₂O₂ solution was serially diluted in bacterial broth and inoculated with 100 µL of bacterial suspensions (1 × 10⁶ CFU mL⁻¹), resulting in a total well volume of 200 µL. The final tested concentrations of H₂O₂ ranged from 15% to the twenty-third half-dilution. The mixtures were incubated at 37 °C for 24 h in a shaking incubator.

To investigate the evolution of resistance, bacteria were subjected to repeated exposure to ROS over 10 successive passages. A wild-type (WT) bacterial parent strain was prepared and inoculated into a serially diluted series of H₂O₂. The MIC was defined as the lowest concentration of the antimicrobial agent that inhibited visible microbial growth after 24 h. Sub-MIC cultures (the first three wells below the MIC) containing surviving bacteria were collected and diluted 100,000-fold. A volume of 20 µL of the diluted bacterial solution was then subcultured on agar at 37 °C for 24 h. A single colony was selected for inoculum preparation at a density of 1 × 10⁶ CFU mL⁻¹ for the subsequent passage, where the bacteria were exposed to a new series of diluted H₂O₂. The entire procedure, from initial inoculation to the preparation of the new inoculum, was repeated for up to 10 passages, and the MICs at each passage were recorded. Finally, the tenth passage of probiotics was selected and named ROS-resistant EcN (REcN).

The fold increase in the MIC was calculated using the following Eq. (1):

$${{{\rm{MIC}}}\,{{\rm{fold}}}\,{{\rm{change}}}}_{{{\rm{experimental}}}}=\frac{{{{\rm{MIC}}}}_{{{\rm{experimental}}}\,{{\rm{passage}}}\,N}}{{{{\rm{MIC}}}}_{{{\rm{WT}}}}}$$

(1)

where “experimental” denotes each passage with antimicrobial exposure (e.g., passage 5 represents 5 passages of antimicrobial exposure).

RNA sequencing analysis

The experimental protocol utilized the TruSeq^TM Stranded Total RNA Library Prep Kit for library construction. During the synthesis of the second cDNA strand, dTTP in the dNTP mixture was substituted with dUTP, leading to the incorporation of A/U/C/G bases into the second cDNA strand. Subsequently, the second cDNA strand was enzymatically degraded using uracil-DNA glycosylase (UNG) prior to PCR amplification, ensuring that the library exclusively contained the first cDNA strand. The detailed procedures are as follows:

(a) Total RNA extraction: Total RNA was isolated from EcN and REcN samples. The concentration and purity of the extracted RNA were determined using a Nanodrop2000 spectrophotometer. RNA integrity was assessed via agarose gel electrophoresis, and the RNA Integrity Number (RIN) was measured using an Agilent 2100 Bioanalyzer. For library construction, each sample required a total RNA amount of 2 µg, with a concentration of ≥ 100 ng µL⁻¹ and an OD_260/280 ratio ranging from 1.8 to 2.2.

(b) rRNA removal: Unlike eukaryotic mRNA, prokaryotic mRNA lacks a polyA tail at the 3’ end, precluding the separation of mRNA from total RNA through A-T base pairing with Oligo dT. Therefore, rRNA removal was employed for transcriptome analysis.

(c) mRNA fragmentation: The Illumina platform sequences short fragments, whereas the enriched mRNA consists of full-length RNA sequences with an average length of several kilobases. Thus, mRNA must be randomly fragmented. The addition of a fragmentation buffer enables the random breakage of mRNA into approximately 200 bp fragments.

(d) cDNA synthesis via reverse transcription: Under the action of reverse transcriptase and using random primers, the mRNA template was used to synthesize the first cDNA strand. During the second-strand synthesis, dUTP was used in place of dTTP in the dNTP mixture, resulting in the incorporation of A/U/C/G bases into the second cDNA strand.

(e) Adapter ligation: The double-stranded cDNA structure has sticky ends. The End Repair Mix was added to blunt the ends, after which an A base was added to the 3’ end to facilitate the connection of the Y-shaped adapter.

(f) Digestion of the second cDNA strand with UNG enzyme: Prior to PCR amplification, the second cDNA strand was digested using the UNG enzyme, ensuring that the library exclusively contained the first cDNA strand.

(g) Illumina HiSeq sequencing: (1) The library was enriched and amplified via PCR for 15 cycles; (2) The library was quantified using TBS380 (Picogreen), and samples were pooled according to data proportions for sequencing; (3) Bridge PCR amplification was performed on the cBot to generate clusters; (4) Sequencing was conducted on the Illumina HiSeq platform, generating 2 × 150 bp or 300 bp reads.

Whole-genome sequencing and SNPs/InDels analysis

(a) Genomic DNA extraction: The bacterial strains (EcN and REcN) were inoculated onto an LB agar plate and incubated overnight at 37 °C. A single colony was picked with an inoculation needle and transferred into 200 mL of LB liquid medium, followed by incubation at 37 °C with shaking until the logarithmic growth phase was reached. The bacterial pellet was collected by refrigerated centrifugation and washed with sterile water to remove residual medium components. Genomic DNA was extracted and quantified, and high-quality DNA was used for subsequent library construction.

(b) Library construction and genome sequencing: Sequencing library was prepared using the Illumina platform. Genomic DNA was fragmented to approximately 400 bp using a Covaris M220 Focused‑Acoustic Shearer. Library construction was performed with the NEXTFLEX Rapid DNA‑Seq Kit. Adapter‑ligated fragments were enriched by PCR. The final library was subjected to paired‑end sequencing (2 × 150 bp) on an Illumina NovaSeq 6000.

(c) Sequencing data quality control: Initially, adapter sequences were removed from the reads. The 5’ ends of reads were then trimmed to eliminate any bases that were not A, G, C, or T. Next, read ends with low sequencing quality (Phred score ≤ Q20) were trimmed. Furthermore, reads containing more than 10% undetermined bases (N) were filtered out. Finally, fragments shorter than 30 bp after adapter removal and quality trimming were discarded.

(d) Read mapping: Clean reads were aligned to a designated reference genome using the BWA‑MEM algorithm. PCR duplicates were identified and removed with Picard‑tools. Sequencing depth and genome coverage were calculated from the resulting BAM alignment files.

(e) SNP and InDel analysis: SNPs/InDels were detected using a standardized variant‑calling pipeline. Initially, local realignment around InDel regions was performed with GATK v4.3.0.0 to reduce false‑positive SNP calls. Variant calling was then carried out with Snippy v4.6.0, applying filters for minimum sequencing depth and mapping quality. Detected variants were functionally annotated using SnpEff to predict their potential effects on coding and regulatory regions.

H₂S detection

The classic lead acetate reactivity test has historically been employed to detect the presence of H₂S. For this test, an appropriate quantity of lead(II) acetate was dissolved in distilled water to prepare a 2% lead acetate solution. A piece of filter paper was cut and immersed in the lead acetate solution until fully saturated. The paper was then carefully removed from the solution and positioned over the surface of a 12-well plate containing bacterial suspensions, and the bacteria were incubated at 37 °C for 24 h. Following incubation, the color change of the filter paper was meticulously observed. Subsequently, the WSP5 H₂S-nucleophilic substitution-cyclization-based fluorescent probe was utilized to detect by an inverted fluorescence microscope (IFM, IX73, Olympus).

Superoxide dismutase and catalase activities tests of REcN

The SOD-like activity of REcN was evaluated using the Total Superoxide Dismutase Kit with WST-8. The optical density at 450 nm (OD₄₅₀) was measured using a microplate reader (SpectraMax i3, Molecular Devices, San Francisco, CA, USA). The CAT-like activity of REcN was assessed using a dissolved oxygen meter. REcN and H₂O₂ (10 mM) were added to 20 mL of phosphate-buffered saline (PBS) under magnetic stirring. The concentration of dissolved oxygen was measured at 1 min intervals for a total duration of 10 min.

α-amylase and α-glucosidase inhibition experiments

The α-amylase activity inhibition of the probiotic was detected using the α-Amylase Kit. The optical density at 540 nm (OD₅₄₀) was measured using a microplate reader. The inhibition of α-glucosidase was evaluated by monitoring the decomposition of p-nitrophenyl-α-D-glucopyranoside to p-nitrophenol using the α-Gucosidase Activity Assay Kit. The optical density at 400 nm (OD₄₀₀) was measured using a microplate reader.

NADH dehydrogenase activity assay

NADH dehydrogenase activity was assessed by monitoring the oxidation of NADH. Briefly, EcN and REcN (1 × 10⁸ CFU mL⁻¹) were incubated with 20 µM NADH solution. The reaction mixture was incubated at 37 ^oC for 30 min, and the absorbance at 340 nm was measured using a microplate reader at the start and end of the incubation period. The decrease in absorbance was used as an indicator of NADH dehydrogenase activity.

Outer membrane permeability assay

Outer membrane permeability was assessed using the fluorescent probe 8-anilino-1-naphthalenesulfonic acid (ANS). Briefly, EcN and REcN (1 × 10⁵ CFU mL⁻¹) cells were harvested, washed, and resuspended in PBS. ANS was added to a final concentration of 10 µM, and after incubation, fluorescence intensity was measured (excitation 370 nm, emission 450–600 nm) using a microplate reader (SpectraMax i3, MolecularDevices).

3T3L1 adipocytes and oil red O staining

3T3-L1 preadipocytes (ATCC CL-173) were commercially procured from Wuhan Pricella Biotechnology Co., Ltd. (Wuhan, China). Adipocyte differentiation was initiating performed with the seeding of 4 × 10⁴ cells per well in 12-well plates. Cultures were maintained overnight in DMEM supplemented with 10% (v/v) neonatal bovine serum under standardized conditions (37 °C, 5% CO₂). The medium was replaced with DMEM containing 10% (v/v) fetal bovine serum (FBS) for a 72-h growth-arrest phase. Subsequently, cells underwent terminal differentiation through sequential exposure to induction cocktail (0.5 mM 3-isobutyl-1-methylxanthine, 1 μM dexamethasone, and 10 μg mL⁻¹ insulin in 10% FBS/DMEM) for 48 h, followed by 144 h maturation in maintenance medium (10% FBS/DMEM) with medium renewal at 72 h intervals. An experimental intervention consisting of 10% (v/v) probiotic-conditioned medium was administered throughout the differentiation regimen.

Quantification of intracellular lipid accumulation employed a validated spectrophotometric protocol. Differentiated adipocyte monolayers underwent PBS rinsing prior to fixation with 4% paraformaldehyde/PBS (37 °C, 30 min). Fixed cells were were stained with Oil Red O for 30 min at ambient temperature, protected from light. After staining, the samples were destained with PBS until a clear background was achieved. Bright-field images were then acquired to visualize lipid deposition by an inverted fluorescence microscope (IFM, IX73, Olympus). For quantitative analysis, the lipid-bound dye was solubilized by adding 500 µL of isopropanol per well. An aliquot of 200 µL from each sample was transferred to a 96-well plate, and absorbance was measured at 490 nm using a microplate reader (SpectraMax i3, MolecularDevices).

Surface modification with fructooligosaccharide

A 20 µL volume of REcN was inoculated into 30 mL of LB broth for 24 h. Following centrifugation at 2400 × g for 5 min at 4 °C, REcN was isolated from LB broth. Subsequently, REcN was resuspended in a norepinephrine (NE, 2 mL, 1 mg mL⁻¹) solution and incubated on a shaker for 1 h. After another round of centrifugation at 2400 × g for 5 min at 4 °C, the fructooligosaccharide (FOS, 2 mL, 10 mg mL⁻¹) solution was added, and the mixture was incubated on a shaker for an additional hour to obtain REcN-FOS.

Calcium carbonate encapsulation

A solution of calcium chloride (CaCl₂, 0.2 mL, 0.33 mM) was then to the probiotic suspension, and the mixture was incubated on a shaker for 20 min. An equal volume of sodium carbonate (Na₂CO₃, 0.33 mM) solution was subsequently added to the homogeneous mixture, followed by further incubation on a shaker for 1 h. The probiotic precipitate was collected by centrifugation at 2400 × g for 5 min at 4 °C. Finally, the probiotic precipitate was washed with PBS to obtain REcN-FOS/CaCO₃ (REcN-F/Ca). The morphologies of probiotics were observed by the transmission electron microscopy (Fisher Talos F200X G2, Thermo Scientific, Waltham, MA, USA).

Viability of REcN-FOS/CaCO₃ in the inflamed colon

Probiotic (1 × 10⁷ CFU mL⁻¹) was first incubated in SGF (containing pepsin and HCl, pH 1.2) for 20 min, then the probiotic was collected, resuspended, and transferred to SIF (containing pancreatin in phosphate buffer, pH 6.8) for 2 h, followed by a final transfer to SICF (containing ammonium hydroxide, 200 μM H₂O₂ in phosphate buffer) for 2 h. incubated in 1 mL of medium supplemented with either SGF, SIF, or SICF. Viability (CFU count) was determined after each step.

In vitro •OH-scavenging activities of REcN-FOS/CaCO₃

Probiotic (1 × 10⁷ CFU mL⁻¹) was incubated in 1 mL of medium supplemented with either SGF, SIF, or SICF. The incubation was conducted at 37 °C (SGF for 20 min, SIF for 2 h, and SICF for 2 h) with gentle shaking to evaluate their stabilities in the GI environment. •OH was generated by co-incubating FeSO₄ (0.4 mM) and H₂O₂ (10 mM) for 20 mins. A 100 μL aliquot of the treated probiotic was added to 100 μL of the above mixture for 10 min. DMPO (2 μL) was then added to the mixture for 2 min. The mixture was drawn into a capillary tube, and •OH scavenging was detected using electron spin resonance (ESR) (Magnettech ESR5000, Bruker, Billerica, MA, USA).

Degradation of CaCO₃ shell

REcN-FOS/CaCO₃ (1 × 10⁸ CFU mL⁻¹) was incubated in 2 mL of medium supplemented with either SGF, SIF, or SICF. The incubation was conducted at 37 °C (SGF for 20 min, SIF for 2 h, and SICF for 2 h) with gentle shaking to evaluate their stabilities in the GI environment. The dissolution profile of the CaCO₃ shell was characterized by measuring the concentration of released Ca ions in SGF, SIF, and SICF over time using inductively coupled plasma optical emission spectrometry (ICP-OES).

Intestinal colonization test

Freshly prepared Cy5.5 dye (10 mg mL⁻¹) was slowly added to probiotic suspensions (10⁸ CFU mL^-1). The mixtures were incubated at 37 ^oC for 2 h in the dark with gentle shaking, followed by repeated PBS washes until the supernatant was clear. The labeled probiotic was then collected, and the FOS/CaCO₃ encapsulation was performed on Cy5.5-labeled EcN and REcN to generate EcN-Cy5.5-FOS/CaCO₃ and REcN-Cy5.5-FOS/CaCO₃, respectively. Mice were orally gavaged with the preparations, and intestinal tissues were collected 6 h post-administration for ex vivo fluorescence imaging using an IVIS Lumina XRMS Series III system (PerkinElmer).

Hemolytic activity test

Hemocompatibility assessment was conducted through quantitative erythrocyte lysis analysis. Murine erythrocytes were isolated via refrigerated centrifugation (500 × g, 10 min, 4 °C) and reconstituted as 5% (v/v) isotonic suspension. Probiotic medium (1 × 10⁸ CFU mL⁻¹) was incubated with erythrocyte suspensions (1:1 v/v) under physiological conditions (37 °C, 4 h). Following secondary centrifugation (500 × g, 10 min), supernatant hemoglobin release was spectrophotometrically quantified at 540 nm using a microplate reader. Hemolytic indices were calculated relative to positive (deionized water) and negative (saline) controls, with < 5% hemolysis established as a biocompatibility threshold.

Cell viability

Cytotoxicity profiling employed mitochondrial dehydrogenase activity quantification (MTT assay) in accordance with ISO 10993-5. RAW 264.7 macrophages were seeded at optimized densities (1 × 10⁴ cells/well) in 96-well plates. After 24 h adherence, cultures were exposed to probiotic (1 × 10⁸ CFU mL⁻¹) for 1 day. MTT solution (0.5 mg mL⁻¹ in PBS) was introduced for 4 h, followed by formazan crystal solubilization in dimethylformamide (200 μL, orbital shaking). Absorbance was measured at 490 nm using a microplate reader.

Quantitative PCR (RT-qPCR) analysis

RAW 264.7 (1 × 10⁴ cells/well) was seeded in 6-well plates and received probiotic (1 × 10⁸ CFU mL⁻¹) exposure for 1 day. Total RNA was then extracted using a Total RNA Kit and quantify by a NanoDrop 2000 spectrophotometer (Thermo Scientific, Waltham, MA, USA). Reverse transcription of 500 ng RNA employed PrimeScript RT Master Mix. Amplification reactions utilized TB Green Premix Ex Taq II on CFX Connect Real-Time System (Bio-Rad, Hercules, CA, USA) under standardized thermocycling: 95 °C/30 s, 40 cycles of 95 °C/5 s and 60 °C/30 s. Relative transcript quantification of IL-1β (F: 5’-TGTGACTCGTGGGATGATGAC-3’, R: 5’-CCACTTGTTGGCTTATGTTCTGTC-3’), IL-10 (F: 5’-TGCTGCCTGCTCTTACTGACTG-3’, R: 5’-ACTCCTGCATTAAGGAGTCGGT-3’), and Tnf-α (F: 5’-CGTCAGCCGATTTGCTATCT-3’, R: 5’-CGTCAGCCGATTTGCTATCT-3’) employed the comparative C_t method (2^−ΔΔCt) normalized to β-actin reference gene (F: 5’-GTGACGTTGACATCCGTAAAGA-3’, R: 5’-GTAACAGTCCGCCTAGAAGCAC -3’). For each gene target, a standard curve was generated using serial dilutions of cDNA to confirm PCR amplification efficiency. All qPCR primers were designed using established sequences from the NCBI database and purchased from a commercial synthesis service (Pinuofei Biotechnology). All primer sets used had efficiencies between 90% and 110%, with correlation coefficients (R²) greater than 0.98, ensuring accurate quantification.

Longitudinal metabolic phenotyping in high-fat diet-induced obese murine model

Male C57BL/6 J mice (n = 6, 6-week-old) were acclimatized in IVC cages under SPF conditions (22 ± 1 °C, 55% humidity, 12 h light-dark cycle) with autoclaved water and a standard feed. Following a 2-week standard chow, mice underwent a metabolic challenge via HFD for a 10-week intervention, establishing a validated insulin-resistant prediabetic model. The animal study included six groups: (1) ND group, (2) HFD group, (3) HFD + F/Ca vehicle group, (4) HFD + EcN-F/Ca group, (5) HFD + EcN-F/Ca group, and HFD + EcN-F/Ca + GW6471 (PPARα signaling inhibitor) group. Pharmacokinetic-informed intervention consisting of 0.5 mL probiotic suspension (5 × 10⁸ CFU mL⁻¹) was administered via orogastric gavage thrice weekly (at 9:00 AM on Monday, Wednesday, and Friday). For the HFD + EcN-F/Ca + GW6471 group, mice were treated with GW6471 (20 mg kg⁻¹ in DMSO) by intraperitoneal injection after orogastric gavage. Weekly body weight measurements was recorded. At week 10, fasting blood samples acquired via tail venipuncture were processed for HOMA-IR calculation, while fecal pellets underwent immediate flash-freezing for SCFAs quantification and gut microbiome analysis. IPGTT was conducted involving intraperitoneal D-glucose (2 g kg⁻¹, 20% w/v solution) with glycemic profiling at 0/30/60/90/120 min. Concurrent serum insulin dynamics were assessed using ELISA. Terminal procedures at week 10 employed cervical dislocation under isoflurane anesthesia (3% induction), with subsequent tissue harvest utilizing aseptic microdissection techniques. The F4/80, CD86, and BODIPY in eAT and the 8-OhdG in colon tissue were detected by immunofluorescence staining. And the main organs were stained by H&E.

RT-qPCR analysis for epididymal adipose and colon tissues

Fresh frozen tissue samples (~ 100 mg), stored at − 80 °C, were retrieved. 1 mL of Trizol reagent was added, and the tissue was homogenized into a slurry using a homogenizer. The homogenate was transferred to a 1.5 mL RNase-free microcentrifuge tube and lysed for 10 min. Total RNA was extracted using a Total RNA Kit and quantify by a NanoDrop 2000 spectrophotometer. Reverse transcription of 500 ng RNA employed PrimeScript RT Master Mix. Amplification reactions utilized TB Green Premix Ex Taq II on CFX Connect Real-Time System under standardized thermocycling: 95 °C/30 s, 40 cycles of 95 °C/10 s and 60 °C/30 s. Relative transcript quantification of Tnf-α, S100a8, Pparα, Slc27a4, Fabp1, Fabp6, Apoa1, Apoa5, Lpl, Ehhadh, Acaa1b, Acox2, and Cpt1b employed the comparative C_t method (2^−ΔΔCt) normalized to β-actin reference gene. For each gene target, a standard curve was generated using serial dilutions of cDNA to confirm PCR amplification efficiency. All qPCR primers were designed using established sequences from the NCBI database and purchased from a commercial synthesis service (Pinuofei Biotechnology) (Supplementary Table 1). All primer sets used had efficiencies between 90% and 110%, with correlation coefficients (R²) greater than 0.98, ensuring accurate quantification.

Quantification of short-chain fatty acids

(a) Experimental design and replicates: Fecal samples for short-chain fatty acids quantification were collected at the study endpoint (week 10). The total number of samples analyzed was n = 18, corresponding to n = 6 biologically independent mice per group for the three primary cohorts: ND, HFD, and HFD+REcN-F/Ca. No samples were excluded from the analysis. A set of external calibration standards, processed identically to the samples, served as controls for quantification. Analyses were performed as single technical measurements per biological sample.

(b) Preparation of SCFAs standard solutions: Accurately weigh 10 mg of each of the eight standard compounds and dissolve them in a 50% acetonitrile aqueous solution to a final volume of 1 mL. Vortex the solution to ensure homogeneity, yielding a stock solution for each standard. Subsequently, mix 100 µL of each individual standard solution and dilute to 1 mL to obtain mixed standard stock solution A. Dilute mixed standard stock solution A by a factor of 50 to achieve a mixed standard stock solution B with a concentration of 20 µg mL⁻¹. Pipette 40 µL of mixed standard stock solution B, add 20 µL of 200 mM 3-Nitrophenylhydrazine hydrochloride and 20 µL of 120 mM 1-Ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride solution (containing 6% pyridine). Both solvents are dissolved in a 50% acetonitrile aqueous solution. Incubate the mixture at 40 °C for 30 min, then dilute to 200 µL with 50% acetonitrile aqueous solution and further dilute stepwise to working solutions L1-L14, which are then transferred into 1.5 mL Eppendorf (EP) tubes.

(c) Sample preparation: Accurately weigh 20 mg of the fecal sample and add 500 µL of extraction solvent (methanol:water = 4:1). Homogenize the mixture in a cryogenic grinding mill for 6 min (− 10 °C, 50 Hz), followed by low-temperature ultrasonication for 30 min (5 °C, 40 kHz). Allow the sample to stand at − 20 °C for 30 min, then centrifuge at 13,000 × g for 15 min at 4 °C. Pipette 20 µL of the supernatant into a new tube and add 20 µL of 200 mM 3-Nitrophenylhydrazine hydrochloride and 20 µL of 120 mM 1-Ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride solution (containing 6% pyridine). Incubate the mixture at 40 °C for 30 min, then dilute to 1000 µL with 50% acetonitrile aqueous solution for subsequent LC-MS/MS analysis.

(d) LC-MS/MS analysis: The LC-MS/MS analysis was performed using an ExionLC AD liquid chromatography system coupled with a QTRAP^® 6500 + mass spectrometer (Shanghai Megi Biomedical Technology Co., Ltd.). A 2 µL sample was injected and separated on a BEH C18 column (150 × 2.1 mm, 1.7 µm) before entering the mass spectrometer for detection. The mobile phase A consisted of 0.1% formic acid in water, and mobile phase B consisted of 0.1% formic acid in acetonitrile. The gradient elution program was as follows: 0.0-2.0 min, mobile phase B held at 10%; 2.0–11.0 min, mobile phase B increased from 10% to 55%; 11.0-12.0 min, mobile phase B increased from 55% to 95%; 12.0-13.0 min, mobile phase B held at 95%; 13.0-13.1 min, mobile phase B decreased from 95% to 10%; 13.1-16.0 min, mobile phase B held at 10%. The flow rate was maintained at 0.35 mL min⁻¹, and the column temperature was set at 40 °C. The mass spectrometer was operated in negative ion mode. The curtain gas was set at 35 psi, collision gas at Medium, ion source spray gas at 40 psi, auxiliary gas at 40 psi, spray voltage at − 4500 V, and ion source temperature at 450 °C.

(e) Quality control samples: QC samples, which are either pooled samples or mixed standard solutions with appropriate concentrations, are primarily used to evaluate the stability of the analytical system. During the instrumental analysis, one QC sample is inserted every 5–10 samples to assess the reproducibility of the entire analytical process. The stability RSD (relative standard deviation) of each analyte should be less than 15%.

(f) Data processing: After the analysis, the raw LC-MS data were imported into the AB Sciex OS quantitative software. The software automatically identified and integrated each ion fragment using default parameters, with manual verification as needed. A linear regression calibration curve was constructed with the analyte peak area as the ordinate and the analyte concentration as the abscissa. The concentration of the analyte in the sample was calculated by substituting the analyte peak area into the linear equation.

Intestinal microbiome analysis

(a) DNA extraction from samples: Genomic DNA was extracted from fecal samples of individuals in the disease and healthy groups using the FastPure Microbiome DNA Isolation Kit according to the manufacturer’s instructions. The integrity of the extracted genomic DNA was assessed by 1% agarose gel electrophoresis. DNA concentration and purity were measured using a NanoDrop 2000 spectrophotometer.

(b) PCR amplification and sequencing library construction: Using the extracted DNA as a template, the V3-V4 hypervariable region of the 16S rRNA gene was amplified by PCR with barcoded primers 338 F (5’-ACTCCTACGGGAGGCAGCAG-3’) and 806 R (5’-GGACTACHVGGGTWTCTAAT-3’). The PCR reaction mixture consisted of 4 μL of 5 × TransStart FastPfu buffer, 2 μL of 2.5 mM dNTPs, 0.8 μL of each primer (5 μM), 0.4 μL of TransStart FastPfu DNA polymerase, 10 ng of template DNA, and water to a final volume of 20 μL. The amplification program was as follows: initial denaturation at 95 °C for 3 min, followed by 27 cycles of denaturation at 95 °C for 30 s, annealing at 55 °C for 30 s, and extension at 72 °C for 30 s, with a final extension at 72 °C for 10 min and storage at 4 °C (PCR instrument: T100 Thermal Cycler). Each sample was amplified in triplicate. The PCR products from the same sample were pooled and purified using 2% agarose gel electrophoresis to verify the fragment size. The purified PCR products were quantified using a Synergy HTX microplate reader (BioTek, USA). The NEXTFLEX Rapid DNA-Seq Kit was used to construct the sequencing library: (1) adapter ligation; (2) magnetic bead selection to remove adapter dimers; (3) PCR amplification to enrich the library template; and (4) magnetic bead purification of the PCR products to obtain the final library. Sequencing was performed on the Illumina NextSeq 2000 PE300 platform.

(c) High-throughput sequencing data analysis: The raw paired-end sequencing reads were quality-controlled using fastp (version 0.19.6) and merged using FLASH (version 1.2.11). The quality control process included: (1) filtering bases with a quality score below 20 at the tail of reads, using a 50 bp sliding window to trim bases if the average quality score within the window was less than 20; (2) discarding reads shorter than 50 bp after quality control and reads containing N bases; (3) merging paired-end reads based on their overlap, with a minimum overlap length of 10 bp; (4) allowing a maximum mismatch rate of 0.2 in the overlap region and filtering out non-conforming sequences; and (5) demultiplexing samples based on barcodes and primers at the ends of the sequences, with zero mismatches allowed for barcodes and up to two mismatches for primers. USEARCH (version 11) was used to cluster the quality-controlled and merged sequences into operational taxonomic units (OTUs) at a similarity threshold of 97% and to remove chimeric sequences. Chloroplast and mitochondrial sequences were removed from all samples. To minimize the impact of sequencing depth on subsequent alpha and beta diversity analyses, all samples were rarefied to 20,000 sequences. After rarefaction, the average sequence coverage (Good’s coverage) for each sample remained at 99.09%. The RDP Classifier (version 2.13) was used to annotate the taxonomic classification of OTUs against the Silva 16S rRNA gene database (v138), with a confidence threshold of 70%, and the community composition of each sample was analyzed at different taxonomic levels. PICRUSt2 (version 2.2.0) was used for 16S functional prediction analysis.

Statistics and reproducibility

The data from all experiments were presented as mean values ± standard deviations (SD), with a minimum of three independent replicates (n ≥ 3). Statistical analyses were performed using GraphPad Prism software. A one-way analysis of variance (ANOVA) was conducted, followed by Tukey’s multiple comparisons test for comparisons among multiple groups and a two-sample Student’s two-sided t test for comparisons between two groups. The Mann-Whitney test was employed to analyze inter-group differences in alpha diversity. ANOSIM non-parametric test was applied to determine significant differences in microbial community structures between sample groups. Exact p-values were shown or the levels of statistical significance was set as p < 0.0001 (****) with 95% confidence interval. LEfSe analysis (LDA > 3, p < 0.05) was used to identify bacterial taxa with significant abundance differences between groups at the phylum to genus levels.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.