Rhein-induced free fatty acid generation cooperates with silver ions for biofilm-infected diabetic wound healing

Abstract

Diabetic foot infection (DFI) significantly increases the risks of hospitalization and amputation, with biofilm infections caused by methicillin-resistant Staphylococcus aureus (MRSA) and carbapenem-resistant Escherichia coli (CREC) being key drivers of treatment failure. Here, we analyzed MRSA/CREC biofilm-infected ulcer tissues from patients using proteomics and metabolomics. Compared with nonhealing tissues, healed tissues exhibit activation of low-density lipoprotein receptor (LDLR)-mediated low-density lipoprotein cholesterol (LDL-C) endocytosis and increased free fatty acid production; thus, we hypothesize a potential countermeasure against biofilms. However, the oxidative instability of free fatty acids limits their therapeutic use. Building on our findings demonstrating the potent anti-MRSA/CREC activity of rhein and its synergy with Ag⁺, we developed an injectable hydrogel for sustained codelivery of both agents. This formulation outperforms Ag⁺ monotherapy, accelerating wound closure in biofilm-infected diabetic wounds. Mechanistically, rhein binds to LDLR, upregulates its expression, and enhances the LDL-C endocytosis pathway in vascular endothelial cells to generate more free fatty acids. Among these, eicosapentaenoic acid (EPA) is identified as the most critical free fatty acid that promotes healing, as it inhibits biofilm formation and synergizes with Ag⁺ to exert antibacterial and pro-healing effects. In summary, this study identifies endogenous host lipid metabolites as “metabolic antibiotics” and establishes a synergistic therapeutic strategy that concurrently suppresses biofilm formation, inhibits planktonic bacteria, alleviates inflammation, and promotes tissue repair in drug-resistant biofilm-infected diabetic wounds.

Introduction

Diabetic foot ulcers (DFUs), which are severe late-stage complications of diabetes mellitus, represent a leading cause of diabetes-related hospitalizations.¹ The global burden of DFUs is staggering, with annual lower-limb amputations exceeding 1 million cases, accompanied by high mortality rates of 24.6% and 45.4% at 5 and 10 years post-diagnosis, respectively.^2,3 A critical factor underlying DFU pathogenesis and impaired healing is the prevalence of multidrug-resistant infections, particularly those mediated by methicillin-resistant Staphylococcus aureus (MRSA)⁴ and carbapenem-resistant Escherichia coli (CREC),⁵ which collectively account for >80% of diabetic foot infections (DFIs).⁶ Patients with DFIs face dramatically worse outcomes than their noninfected counterparts do, demonstrating 50-fold and 150-fold increases in hospitalization and amputation risks, respectively.⁷ While antibiotics remain the cornerstone of DFI treatment, the escalating crisis of antibacterial resistance severely compromises their therapeutic efficacy because of microbial biofilm formation. Within biofilms, compared with their planktonic counterparts, bacteria exhibit dramatically elevated (100–1000 fold) tolerance to antibiotics.⁸ Silver ions (Ag⁺), widely utilized as broad-spectrum antibacterial agents in wound care, also demonstrate limited activity against biofilm-embedded pathogens.⁹ Moreover, conventional silver-based formulations suffer from uncontrolled Ag⁺ release kinetics, often resulting in local cytotoxic concentrations that may impede wound healing.¹⁰ These limitations underscore the urgent need for novel antibiofilm strategies that combine enhanced safety profiles with robust efficacy while minimizing resistance development.^11,12

Diabetic ulcers are associated with metabolic dysregulation involving both glucose and lipid homeostasis.¹³ Chronic hyperglycemia accelerates the pathological deposition of low-density lipoprotein cholesterol (LDL-C) in vascular walls, resulting in progressive stenosis and subsequent lower limb ischemia.^14–17 Emerging research reveals that LDL-C undergoes receptor-mediated endocytosis to yield free fatty acids,¹⁸ among which certain species have demonstrated significant antibiofilm activity.¹⁹ However, the topical application of free fatty acids as antibacterial agents is limited by poor dose control, wound exudate incompatibility,²⁰ and oxidation-prone instability in unsaturated types, which reduces efficacy and generates cytotoxic metabolites.^21,22 In this context, targeted modulation of physiological LDL-C endocytosis to promote sustained endogenous production of antibacterial free fatty acids represents a promising therapeutic strategy to overcome current limitations in the management of biofilm-associated diabetic foot infections.

In clinical practice, the Chinese herb rhubarb modulates obesity-related adipogenesis, lipolysis, and inflammation.²³ Additionally, rhubarb powder is applied topically to treat infectious skin ulcers because of its broad-spectrum antibacterial activity,²⁴ which is attributed to its diverse anthraquinone compounds.²⁵ Among these compounds, rhein exhibits a broad spectrum of pharmacological activities relevant to wound healing, including the modulation of lipid metabolism, the promotion of adipose thermogenesis,²⁶ and the inhibition of bacterial growth^27,28 and biofilm formation.²⁹ However, the mechanism underlying the dual regulatory effects of rhein on lipid metabolism and anti-infection activity in the context of DFI-induced wound healing remains unclear.

In this study, we applied four-dimensional data-independent acquisition (4D-DIA) proteomic sequencing to skin biopsies from patients with healed DFIs, patients with nonhealed DFIs, and patients with nonulcer diabetes. Additionally, untargeted metabolomic sequencing was performed on ulcer surface biopsies from patients with healed DFIs and nonhealed DFIs. No significant differences were observed in key baseline characteristics, including age, sex, diabetes duration, HbA1c level, BMI, or initial ulcer features, between the DFI-healer and DFI-nonhealer subgroups selected for untargeted metabolomic analysis (P > 0.05). Sample details are provided in Data S2 for all enrolled patients, where the IDs highlighted in yellow correspond to the samples included in the metabolomic analysis. The results revealed enhanced activation of the LDL-C endocytosis pathway and increased production of free fatty acids in patients with healed DFIs. Inspired by the potent antibacterial activity of rhein against MRSA and CREC and its synergy with Ag⁺, we engineered an injectable hydrogel for sustained codelivery of both agents. This hydrogel was fabricated through a dynamic crosslinking gelation mechanism using natural keratin as the sole matrix material. In vitro assessments against planktonic bacteria and biofilms, alongside diabetic murine infection models, demonstrated the hydrogel’s ability to disrupt existing biofilms and prevent new biofilm formation. Integrated analysis, including network pharmacology targets, quantitative lipidomics from murine tissues, and patient-derived proteomic/metabolomic data, revealed coordinated enrichment of KEGG pathways in LDL-C endocytosis following hydrogel treatment. Mechanistically, rhein binds the low-density lipoprotein receptor (LDLR), potentiating LDL-C endocytosis to increase the biosynthesis of free fatty acids, particularly eicosapentaenoic acid (EPA). EPA inhibits biofilm reformation and acts synergistically with Ag⁺ to enhance antibacterial effects, collectively accelerating wound closure in diabetic infections. Overall, by leveraging the LDL-C-rich microenvironment of diabetic ulcers, this rhein/Ag⁺ combination strategy activates lipid metabolism pathways and exerts synergistic antibacterial effects, offering a promising therapeutic approach for chronic, drug-resistant wound infections.

Results

Activation of LDLR-mediated LDL-C endocytosis promotes free fatty acid generation in the healing of DFIs with drug-resistant biofilms

To analyze healing-associated differential proteins and metabolites, we collected debridement samples from 45 DFU patients with MRSA and CREC infections and recorded demographic, clinical, and laboratory data during a 3-month follow-up period with weekly wound healing documentation (Clinical Trial Registration Number: ChiCTR2400084803; Data S2). Patients demonstrating >50% wound area reduction after standard therapy were classified as DFI-healers, while those with <50% reduction were designated as DFI-nonhealers. Among the 36 completed follow-up cases, untargeted metabolomics was performed on 9 DFI-nonhealers and 10 DFI-healers, while 4D-DIA proteomics was used to analyze 5 DM-nonulcers and 19 DFI tissues (Fig. 1a).

Fig. 1.

Activation of LDLR-mediated LDL-C endocytosis promotes free fatty acid generation in the healing of DFIs with drug-resistant biofilms. a Schematic of the experimental design, including the study population, data collection, proteomics and metabolomics analysis, and validation. b Macroscopic wound appearance at the 3-month follow-up and biofilm assessment through Gram staining, SEM visualization, and CLSM live/dead staining of infected tissues. Scale bars: 200/20 μm (Gram staining), 2 μm (SEM), and 200 μm (CLSM 3D reconstruction). c KEGG coenrichment analysis of differentially abundant metabolites and proteins in wound tissues from DFI-healers and DFI-nonhealers. Pathways whose FDRs were < 0.1 according to either proteomic or metabolomic analysis are displayed. Bubble color indicates the raw P value, and bubble size represents the count of differential molecules in the pathway. d Correlation analyses between plasma free fatty acids and plasma LDL-C and between DMIST scores and free fatty acids in DFI tissues. The blue dots represent individual samples; the blue solid line indicates the regression line; and the dashed lines indicate 95% confidence intervals. e Fold change differences in free fatty acid, amino acid, and oxidized lipid profiles between DFI-healers (n = 10) and DFI-nonhealers (n = 9) by metabolomics sequencing. f Chemiluminescent quantification of free cholesterol, amino acids and free fatty acids in the DFI-healers and DFI-nonhealers (n = 9/group) and ELISA quantification of LDL-C levels in the same groups (n = 6/group). g Immunohistochemical staining of LDLR, LDLRAP1/ARH, and Clathrin in the wound tissues of the DFI-healers, DFI-nonhealers and DM-nonulcers (n = 4/group). Scale bars: 500/100 μm. The data are presented as the mean ± SD from three independent experiments. For d, the Pearson correlation coefficients and two-tailed P values are presented. For f, statistical significance was determined by an unpaired two-tailed Student’s t test. For g, statistical significance was determined by one-way ANOVA with Tukey’s multiple comparisons test. *P < 0.05, **P < 0.01, ***P < 0.001

Biofilm formation in DFUs was assessed through Gram staining (GS), scanning electron microscopy (SEM), and confocal laser scanning microscopy (CLSM) 3D reconstruction (Fig. 1b). GS revealed abundant Gram-positive (green arrows) and Gram-negative bacteria (blue arrows), and SEM revealed uneven biofilm surfaces with extracellular polymeric substances in both the DFI-nonhealers and the DFI-healers. A qualitative analysis using CLSM with SYTO-9/PI staining revealed a homogeneous biofilm distribution in the skin tissue, with SYTO-9 marking live bacteria and PI marking dead bacteria. Supplementary Fig. 1a shows three-dimensional principal component analysis (PCA) plots derived from proteomic profiling. The DFI-nonhealer and DFI-healer groups exhibited significant separation along the first principal component (PC1), whereas the DM-nonulcer group (DM) showed partial segregation along PC2. Supplementary Fig. 1b, c present the PCA plot of metabolomics and the heatmap of differentially abundant metabolites, respectively. The top 20 differentially expressed proteins and lipases in the fat digestion and absorption pathway are shown in Supplementary Fig. 2d. KEGG coenrichment analysis revealed activated lipid metabolism pathways, including fat digestion and absorption, α-linolenic acid and linoleic acid metabolism, and fatty acid degradation (Fig. 1c).

To elucidate the relationships among lipid metabolism, inflammatory markers, and the DFI, we performed comprehensive correlation analyses between ulcer depth, maceration, inflammation/infection, size, tissue type of the wound bed, type of wound edge, and tunneling/undermining (DMIST) as previously described.³⁰ Our analysis revealed no significant associations between DMIST scores and triglyceride (TG), total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), white blood cell (WBC), or neutrophil (NEUT) levels in plasma, whereas there was a positive correlation between DMIST scores and plasma free fatty acid levels (Supplementary Fig. 2a–f). Notably, an inverse relationship was found between plasma LDL-C levels and plasma free fatty acid levels, as well as a strong negative correlation between DMIST scores and free fatty acid levels in the DFI-wounded tissue samples (Fig. 1d). The free fatty acids in DFI tissues originate primarily from triglyceride catabolism, with minor contributions from cholesterol and phospholipid metabolism.^31,32 Further analysis of differential lipids in DFI-affected tissues revealed elevated triglyceride and phospholipid levels in the DFI-healer group (Supplementary Fig. 3), along with increased abundance of total amino acids and multiple free fatty acids, including tetradecanedioic acid, lyngbic acid, dodecanedioic acid, tetradecanoic acid, 16-hydroxyhexadecanoic acid, hexadecanedioic acid, 3-hydroxyhexadecanoic acid, and DL-2-hydroxyoctadecanoic acid, and decreased abundance of the oxidized lipid 16-oxohexadecanoic acid (Fig. 1e).

LDL-C is well known to undergo hydrolysis by lysosomal hydrolases into free cholesterol, amino acids, and free fatty acids.¹⁸ These liberated free fatty acids, such as arachidonic acid, linoleic acid, α-linolenic acid, docosahexaenoic acid, and EPA, serve as precursors for oxylipin synthesis, which subsequently promotes inflammatory responses and oxidative stress.²² Our findings revealed an inverse correlation between plasma LDL-C levels and free fatty acid levels, combined with significant intergroup differences in low-density lipoprotein receptor-associated protein 1 (LDLRAP1) expression (Supplementary Fig. 4), led us to focus on the LDLR-mediated LDL-C endocytic pathway. LDLR facilitates the absorption of LDL-C from blood into cells, which is crucial in humans, as low-density lipoprotein is the primary cholesterol transporter.³³ To elucidate the potential association between the LDL-C endocytosis-mediated free fatty acid generation pathway and wound healing in the DFI, we performed quantitative analyses using ELISA and chemiluminescence assays to measure LDL-C, free cholesterol, total amino acids, and free fatty acids in wound tissues. Our findings revealed that compared with the DFI-nonhealers, the DFI-healers presented significantly lower LDL-C levels and higher concentrations of free cholesterol, total amino acids, and free fatty acids in the wound tissue (Fig. 1f). Immunohistochemical staining further confirmed the increased expression of LDLR, its adaptor protein LDLRAP1/ARH, and Clathrin at the wound margins in the DFI-healers (Fig. 1g). Taken together, these results provide comprehensive evidence characterizing the dynamics of LDL-C endocytosis and subsequent free fatty acid generation in DFI-healer wounds, highlighting enhanced LDL-C uptake and processing, the consequent accumulation of metabolic byproducts, and the coordinated upregulation of endocytic machinery components during the healing process.

In vitro antibacterial and antibiofilm activity of the Ker/Ag⁺/Rh hydrogel against MRSA and CREC

In our preliminary experiments, we compared the antibacterial activities of six anthraquinone compounds derived from rhubarb. The results demonstrated that rhein exhibited a potent antibacterial effect against both MRSA and CREC (Supplementary Fig. 5). The antibacterial efficacy of rhein and its synergistic interaction with Ag⁺ were further quantitatively evaluated using the standard minimum inhibitory concentration (MIC) method, with comparisons made against clinically established antibacterial regimens, including cefixime, vancomycin, ampicillin, and ofloxacin (Fig. 2a). MRSA (USA300) and CREC (RJ112) strains isolated from chronic wound biofilms were chosen as model pathogens. MIC determinations were performed in strict accordance with the 2022 Clinical and Laboratory Standards Institute (CLSI) guidelines. Rh demonstrated antibacterial activity against both strains, as evidenced by MICs. The combination of Rh with silver ions (Ag⁺) resulted in particularly potent synergistic effects, with a fractional inhibitory concentration (FIC) index below the 0.5 threshold.

Fig. 2.

The Ker/Ag⁺/Rh hydrogel inhibits the planktonic bacterial growth of MRSA and CREC and disrupts their biofilms in vitro. a MIC and FIC index of Rh and Ag⁺ against MRSA and CREC, with clinical antibiotics as controls. Rhein (+Ag⁺) means rhein combined with Ag⁺; Ag⁺ (+Rhein) means Ag⁺ combined with rhein. b Schematic illustration of the formation mechanism of the Ker/Ag⁺/Rh hydrogel. c Macroscopic appearance and SEM images of the Ker/Ag⁺ hydrogels at various Ag⁺ concentrations. Scale bar: 200 μm. d Rheological properties of the Ker/Ag⁺ hydrogels at various Ag⁺ concentrations. e Shear-thinning behavior and injectability of the Ker/Ag⁺ hydrogel. f, g Cumulative release profiles of Ag⁺ and Rh from the Ker/Ag⁺/Rh hydrogel (with a Ag⁺ concentration of 1 mM and a Rh concentration of 1.5 mg/mL). h Schematic of the in vitro anti-planktonic bacteria assay against MRSA and CREC. i Colony-forming unit (CFU) counts and SEM images (scale bar: 0.5 μm) of MRSA after treatment. j Quantitative analysis of bactericidal rates against MRSA. k CFU counts and SEM images (scale bar: 2 μm) of CREC after treatment. l Quantitative analysis of bactericidal rates against CREC. m Schematic of the in vitro anti-biofilm assay against MRSA and CREC. n, q The inhibition of biofilm formation was assessed by crystal violet staining and quantified by UV absorbance. o, r CLSM 3D reconstructions via SYTO-9/PI staining and SEM images of biofilms. Scale bars: 100/10 μm (o); 100/20 μm (r). p, s Quantitative viability analysis of SYTO-9/PI-stained biofilms. The data are presented as the mean ± SD from three independent experiments. Statistical significance was determined by one-way ANOVA (Tukey’s multiple comparisons test) for j, l, n, and q or two-way ANOVA (multiple comparisons) for p and s. *P < 0.05, **P < 0.01, ***P < 0.001 vs control; ^#P < 0.05, ^##P < 0.01, ^###P < 0.001 vs VAN/CFM; ^&P < 0.05, ^&&&P < 0.001 vs Ker/Ag⁺

Building upon these promising results, we engineered an innovative injectable hydrogel for the codelivery of Rh and Ag⁺ in slow release mode (Fig. 2b). This hydrogel utilized keratin, a natural epidermal structural protein, as the matrix material through dynamic Ag‒S crosslinking between thiol groups in keratin and Ag⁺ ions (denoted as Ker/Ag⁺). Rh incorporation was achieved through coordination between its carboxyl groups and Ag⁺ (denoted as Ker/Ag⁺/Rh).³⁴ Notably, the gelation process occurred spontaneously at room temperature through simple physical mixing, eliminating the need for external crosslinkers or chemical modification of keratin. Dynamic Ag–S crosslinking was confirmed by modulating the Ag⁺ concentration (Fig. 2c). At 0 mM Ag⁺, the hydrogel was formed via covalent disulfide bonds between keratin molecules (denoted as Ker), resulting in poor injectability. With increasing Ag⁺ concentration, the hydrogel exhibited enhanced injectability, as evidenced by smooth extrusion through a syringe. SEM revealed interconnected porous structures in all the hydrogels, with larger pore sizes and higher swelling ratios at elevated Ag⁺ concentrations (Fig. 2c; Supplementary Fig. 6a, b), corroborating the formation of more Ag‒S bonds. Rheological analysis revealed a reduced storage modulus (G’) at higher Ag⁺ concentrations, indicating the formation of softer hydrogels, whereas G’ consistently exceeded the loss modulus (G”), reflecting elastic behavior (Fig. 2d; Supplementary Fig. 7a, b). The dynamic cross-linked network endowed the self-healing properties of the hydrogel, as verified by macroscopic recovery and cyclic strain tests (Supplementary Fig. 8a–c). Shear-thinning tests further confirmed the injectability of the hydrogel at 1 mM Ag⁺ (Fig. 2e). All these performances confer excellent practical applicability in wound management, such as seamless filling of irregular wounds while matching skin mechanics. We also determined the in vitro release profiles of Ag⁺ and Rh. Both Ag⁺ and Rh exhibited slow release profiles (Fig. 2f, g). Owing to its simple formulation design, this hydrogel showed excellent hemocompatibility (hemolysis rate <5%; Supplementary Fig. 9a) and cytocompatibility (Supplementary Fig. 9b, c).

The antibacterial effects of Ker/Ag⁺/Rh against planktonic MRSA and CREC were assessed using the plate-counting method, with Ker/Rh, Ker/Ag⁺, Ker, vancomycin (VAN) and cefixime (CFM) used for comparison (Fig. 2h). Quantitative CFU analysis revealed that compared with Ker/Rh and Ker/Ag⁺, Ker/Ag⁺/Rh reduced the MRSA burden by 105-fold and 55-fold, respectively, and reduced the CREC burden by 352-fold and 242-fold, respectively (Supplementary Fig. 16 a, b). The results of the inhibition rate calculations revealed that Ker had no inhibitory effect on MRSA, whereas VAN, Ker/Ag⁺, and Ker/Ag⁺/Rh exhibited inhibition rates of 78.05%, 59.75%, and 98.79%, respectively. Similarly, Ker showed no significant activity against CREC, whereas CFM, Ker/Ag⁺, and Ker/Ag⁺/Rh achieved inhibition rates of 72.65%, 50.61%, and 97.73%, respectively (Fig. 2j, l). SEM revealed distinct morphological changes in MRSA and CREC following treatment (Fig. 2i, k). The untreated control and Ker groups displayed smooth, intact bacterial membranes with well-defined rod-shaped morphologies. In contrast, surface wrinkling and deformation were induced in both the VAN/CFM and Ker/Ag⁺ groups, whereas Ker/Ag⁺/Rh caused severe structural damage, rupturing the cell walls and membranes and leading to cytoplasmic leakage. The inhibition of biofilm formation was further investigated using a crystal violet staining assay (Fig. 2m). Among all the groups, the Ker/Ag⁺/Rh group had the lowest crystal violet absorbance, indicating that the strongest inhibition of biofilm formation occurred in both the MRSA and the CREC biofilms (Fig. 2n, q; Supplementary Fig. 16c, d). CLSM and SEM further confirmed the effective inhibition of biofilm development in the Ker/Ag⁺/Rh group, which was more pronounced than that achieved with VAN or CFM. Notably, compared with Ker/Ag⁺/Rh, Ker/Ag⁺ exhibited significantly weaker biofilm formation inhibition activity against both MRSA and CREC biofilms (Fig. 2o, p, r, s). Collectively, while VAN/CFM and Ker/Ag⁺ exhibited measurable antibacterial and biofilm-inhibitory effects, the Ker/Ag⁺/Rh hydrogel demonstrated superior efficacy against both MRSA and CREC in vitro.

Ker/Ag⁺/Rh hydrogels promote the healing of MRSA/CREC biofilm-infected ulcers in T2DM animal models

A T2DM mouse model was successfully established using a high-fat diet combined with low-dose streptozotocin (STZ) (30 mg/kg) (Supplementary Fig. 10a). Initial evaluation of the therapeutic effects of the Ker/Ag⁺/Rh hydrogels containing different Rh concentrations (0.5–1.5 mg/mL) on MRSA and CREC biofilm-infected wounds was performed. Infected wounds treated with blank urea ointment and uninfected wounds treated with blank urea ointment were defined as the Vehicle and Uninfected groups, respectively. Compared with the vehicle and uninfected groups, the Ker/Ag⁺/Rh groups exhibited accelerated healing in a dose-dependent manner, with the 1.5 mg/mL Rh group showing optimal efficacy (Supplementary Fig. 10b, c). We further compared the therapeutic effects of Ker/Ag⁺/Rh (1.5 mg/mL Rh) with those of Ker/Ag⁺, Ker and Mupirocin Ointment (MO) on MRSA and CREC biofilm-infected wounds (Fig. 3a). All the wounds displayed redness and exudation on day 1 because of biofilm infection. Compared with the Uninfected group, the Ker/Ag⁺/Rh group achieved 50% wound closure by day 3, surpassing the other treatment groups, and reached healing levels comparable to those of the Uninfected group by day 12 (Fig. 3b, c).

Fig. 3.

Ker/Ag⁺/Rh hydrogel promotes wound healing in T2DM mice with MRSA/CREC biofilm infection. a Experimental timeline for the wound healing test. b, c Representative images of wounds and quantitative data of relative wound sizes at different time points (n = 6/group). d Quantification of the reduction in bacterial viability in the wounds on day 3 (n = 6/group). e CLSM imaging of biofilms and quantitative viability assessment based on SYTO-9/PI staining on day 3 (n = 4/group). Scale bar: 200 μm. f, g Relative expression of icaA, icaB, icaC, icaD, ompa, fnbA and fnbB in MRSA and CREC biofilms in the vehicle group (n = 3) and treatment group (n = 6) determined by RT‒qPCR analysis. h, i Representative immunofluorescence staining images and quantification of NETs in wounds on day 3 (n = 6/group); scale bar: 500/100 μm. Data are represented as the mean ± SD from two independent experiments unless otherwise specified. Statistical significance was determined by one-way ANOVA (Tukey’s multiple comparisons test) for c, d, g, and i or by two-way ANOVA (multiple comparisons) for e. ns: not significant; **P < 0.01, ***P < 0.001 vs vehicle; ^#P < 0.05, ^##P < 0.01, ^###P < 0.001 vs MO; ^&P < 0.05, ^&&P < 0.01, ^&&&P < 0.001 vs Ker/Ag⁺

The synergistic efficacy of rhein and Ag⁺ was further validated in a rat model of diabetic ulcer infection with MRSA and CREC mixed biofilms. Animals were randomly assigned to the Vehicle, Ker/Rh, Ker/Ag⁺, and Ker/Ag⁺/Rh groups. Consistent with the findings in mice, compared with the Ker/Rh and Ker/Ag⁺ groups, the Ker/Ag⁺/Rh group exhibited the most pronounced acceleration of wound closure throughout the 12-day observation period, with significantly reduced wound areas on days 7 and 12 (P < 0.01 and P < 0.001, respectively; Supplementary Fig. 17a). The bacterial burden in the wound tissue was assessed on days 3 and 7 posttreatment. Compared with the Ker/Rh and Ker/Ag⁺ groups, the Ker/Ag⁺/Rh group consistently presented the lowest bacterial loads, with reductions of 2–3 log₁₀ CFU at both time points (P < 0.001; Supplementary Fig. 17b). These results confirm that the synergistic antibacterial and pro-healing effects of rhein and Ag⁺ are reproducible across species, supporting the therapeutic advantage of their combination.

Histological analysis on day 12 revealed that compared with the other groups, the Ker/Ag⁺/Rh group exhibited the most prominent epithelial regeneration (blue dashed line), with superior epithelial tongue length, thickness and coverage (Supplementary Fig. 11). Immunohistochemical staining revealed the lowest HIF-1α expression in the Ker/Ag⁺/Rh group on day 12, with VEGF-A expression matching that in the uninfected group (Supplementary Fig. 12a, b). Masson’s trichrome staining revealed that compared with the other groups, the Ker/Ag⁺/Rh group had regularly arranged, densely packed collagen fibers with a significantly greater collagen volume fraction (Supplementary Fig. 12c, d).

On day 3 post-treatment, the bacterial load in the wounds was quantified via swab collection, followed by serial dilution and colony counting (Fig. 3d). The Ker/Ag⁺/Rh group achieved a high sterilization efficacy of >98.79%, significantly outperforming both the MO and Ker/Ag⁺ treatments. Live/dead staining combined with confocal laser scanning microscopy (CLSM) 3D reconstruction was performed to qualitatively analyze biofilm reformation across groups. The results revealed that dense, block-like biofilms persisted in the vehicle group, whereas the Ker/Rh-treated group exhibited notably dispersed green fluorescence, indicating effective suppression of biofilm reformation (Fig. 3e, Supplementary Fig. 18a). Compared with the Ker/Rh group, the Ker/Ag⁺-treated group displayed greater biofilm density. Notably, the Ker/Ag⁺/Rh group exhibited further dispersal of green fluorescence along with the strongest red fluorescence signal, suggesting a synergistic effect between Rh and Ag⁺ in inhibiting biofilm reformation.

SEM observations further confirmed the biofilm reformation results across groups (Supplementary Fig. 19). In the vehicle group, mature and structurally stable biofilms were formed, which displayed dense sheet-like structures with abundant mucus extracellular skeletons. Grape cluster-like MRSA (0.5–1.0 μm) and rod-shaped CREC (1.5–2.5 μm) were tightly embedded within the extracellular polymeric substances. Compared with the vehicle group, the Ker/Rh group showed disrupted biofilm formation, presenting an irregular hollow structure with weakened bacterial adhesion and aggregation, reduced bacterial stacking, and partial bacterial shrinkage. In the Ker/Ag⁺ group, partial loss of extracellular matrix and sporadic collapse of biofilm architecture were observed, although the bacteria remained wrapped within a relatively compact biofilm. The Ker/Ag⁺/Rh group exhibited fragmented biofilms with loose reticular structures, and almost no intact bacterial cells could be observed.

To further evaluate the antibiofilm capacity of Ker/Ag⁺/Rh, the expression of several virulence factors (day 3) was quantified by reverse transcription quantitative PCR (RT‒qPCR). In terms of CREC biofilm formation, ompA, papC, and KpsM are critical regulatory genes. The ompA gene encodes outer membrane protein A, a critical virulence factor that mediates the adhesion of Escherichia coli to host cells through fibronectin binding.³⁵ PapC facilitates the surface localization and polymerization of pilus-adhesin subunits,³⁶ and KpsM, together with KpsT, mediates the transmembrane export of capsular polysaccharides to form a capsular layer, thereby stabilizing microcolonies and promoting biofilm maturation and persistence.³⁷ The ica operon (icaA, icaB, icaC, and icaD) in MRSA regulates polysaccharide intercellular adhesin production, which is essential for biofilm formation and stability.^38,39 Fibronectin-binding proteins (fnbA and fnbB) facilitate bacterial aggregation and extracellular matrix interactions, further promoting biofilm development.⁴⁰ RT‒qPCR analysis revealed that Ker/Ag⁺/Rh treatment dramatically suppressed the expression of key biofilm-related genes (ompA, icaA, icaB, icaC, icaD, and fnbA), although fnbB expression remained unaffected (Fig. 3f, g). To distinguish the contributions of rhein and Ag⁺, we compared the effects of Ker/Rh, Ker/Ag⁺, and Ker/Ag⁺/Rh. For MRSA, Ker/Rh inhibited icaA but weakly suppressed icaB/icaC, and Ker/Ag⁺ showed limited inhibition of the ica operon. For CREC, both single agents reduced ompA and papC, whereas only Ker/Rh moderately inhibited KpsM. In all the cases, Ker/Ag⁺/Rh exhibited the strongest suppression of icaA/B/C, ompA, papC, and KpsM, significantly outperforming both monotherapies (Supplementary Fig. 18 b). It is well known that bacterial infections in diabetic wounds can activate host phagocytes and endothelial cells, leading to the excessive production of cytokines and chemokines. This inflammatory cascade promotes neutrophil recruitment and the formation of neutrophil extracellular traps (NETs), which in turn exacerbates the immune response and delays wound healing.⁴¹ Immunofluorescence analysis revealed that on day 3, compared with the vehicle, MO, Ker, and Ker/Ag⁺ treatments, treatment with Ker/Ag⁺/Rh significantly reduced neutrophil infiltration (as indicated by MPO) and NET formation (as indicated by Cit-H3) (Fig. 3h, i). Furthermore, by day 7, a marked decrease in the levels of proinflammatory cytokines, including IL-1β, IL-6, and TNF-α, was observed (Supplementary Fig. 13a, b). These results demonstrated that Ker/Ag⁺/Rh effectively disrupted MRSA/CREC biofilm adhesion and regeneration through a multifaceted mechanism. This synergistic action integrated antibiofilm, antibacterial, and anti-inflammatory effects, thereby facilitating the transition of wound healing from the inflammatory phase to the proliferative phase.

The Ker/Ag⁺/Rh hydrogel activates LDL-C endocytosis to generate EPA

Network pharmacology analysis revealed that Ker/Ag⁺/Rh treatment primarily modulated signaling pathways, including those related to PI3K-AKT, lipid metabolism and atherosclerosis, and focal adhesion (Fig. 4a). To identify differentially expressed lipids in the skin tissue between the vehicle- and the Ker/Ag⁺/Rh-treated model mice, a t test was conducted on the basis of quantitative lipidomics data, using the relative lipid content from both groups. Lipids meeting the criteria of |FC | ≥ 1.5 and P < 0.05 were considered significantly differentially expressed. This analysis revealed 399 significantly altered lipids (|FC | ≥ 1.5, P < 0.05), among which 357 were upregulated and 42 were downregulated (Fig. 4b, c). The bar graph of the fold changes for the top 25 differentially expressed lipids is shown in Fig. 4d. Notably, free fatty acid (20:5), which is typically known as EPA, is a reported potent antibacterial agent owing to its unsaturated structure. KEGG annotation further revealed that these differential lipids were predominantly enriched in pathways related to lipids and atherosclerosis (Fig. 4e). An integrated KEGG enrichment analysis combining network pharmacology results from rhein treatment in DFU, proteomic and metabolomic sequencing data from skin tissues of clinical DFI patients, and lipidomic sequencing data from model mouse skin consistently highlighted the LDL-C endocytosis pathway as a commonly enriched pathway (Fig. 4f). On the basis of these convergent findings, we hypothesized that Ker/Ag⁺/Rh promoted wound healing by modulating LDL-C endocytosis. This hypothesis was subsequently validated through targeted analysis of key molecules within this pathway.

Fig. 4.

The Ker/Ag⁺/Rh hydrogel enhances LDL-C endocytosis to produce the unsaturated free fatty acid EPA. a Network pharmacology analysis of the enrichment of genes related to the Ker/Ag⁺/Rh-targeted pathway in response to treatment of mice with DFI. b PCA score plot showing the separation of lipidomic profiles between the Ker/Ag⁺/Rh and vehicle groups. c Volcano plot of differential lipids (VIP > 0.5/1, |FC | ≥ 1.2, P < 0.05) between the Ker/Ag⁺/Rh and vehicle groups, with point size reflecting variable importance. d Top 25 lipid species ranked by fold change. e KEGG enrichment analysis of differentially abundant metabolites in Ker/Ag⁺/Rh- and vehicle-treated wound tissues. f Schematic of the coenrichment analysis highlighting the convergent regulation of the LDL-C endocytosis pathway. g Quantification of EPA in wound tissues via lipidomic sequencing following treatment with Ker/Ag⁺/Rh or vehicle. h Immunohistochemical staining of LDLR and LDLRAP1/ARH expression on day 7 posttreatment with Ker/Ag⁺/Rh, Ker/Ag⁺ or vehicle and quantitative analysis using ImageJ software (n = 6/group). i CLSM visualization of Clathrin-CD29 colocalization in wound-edge tissues (nuclei, DAPI/blue; Clathrin/red; CD29/green; merge/yellow; a, wound edge; b, wound bed) and quantitative analysis of fluorescence intensity using ImageJ software (n = 6/group). j Biochemical assays of free cholesterol, amino acids, and free fatty acids on day 7 (n = 6/group). Data are represented as the mean ± SD from two independent experiments unless otherwise specified. Statistical significance was determined by one-way ANOVA with Tukey’s multiple comparisons test. ns: not significant; *P < 0.05, **P < 0.01, ***P < 0.001. Scale bars: 500/200 μm (h), 50/5 μm (i)

Quantitative lipidomic analysis revealed that among the differential free fatty acids, the level of the unsaturated free fatty acid EPA was significantly elevated in the Ker/Ag⁺/Rh-treated group (Fig. 4g). Immunohistochemical analysis revealed a significant increase in the mean optical density of LDLR and LDLRAP1/ARH in the wound tissues of the Ker/Ag⁺/Rh group on day 7 compared with both the Vehicle and Ker/Ag⁺ groups (Fig. 4h). These proteins were predominantly expressed in vascular endothelial cells within the dermal layer. No significant difference was observed between the vehicle and Ker/Ag⁺ groups. Compared with the other groups, the Ker/Ag⁺/Rh group exhibited increased colocalization of Clathrin (red) with CD29-positive mesenchymal cells (green) among keratinocytes, fibroblasts, and vascular endothelial cells, as demonstrated by immunofluorescence staining (Fig. 4i). Biochemical quantification of lipid metabolites confirmed that Ker/Ag⁺/Rh treatment increased the levels of free cholesterol, amino acids and free fatty acids, whereas Ker/Ag⁺ treatment increased the levels of only amino acids (Fig. 4j). These findings collectively confirmed that Ker/Ag⁺/Rh, particularly the rhein component, induced the activation of LDL-C endocytosis to increase the production of free cholesterol, amino acids, and free fatty acids, among which EPA may be the free fatty acid that inhibits biofilm formation.^42,43

Antibacterial activity of EPA against planktonic and biofilm-embedded MRSA and CREC

Antibacterial assessments of planktonic MRSA and CREC revealed that EPA exhibited dose-dependent bactericidal activity at concentrations ranging from 0.1 to 0.8 mM (Fig. 5a, b). SEM imaging further confirmed the membrane-disruptive action of EPA (0.4 and 0.8 mM), where the untreated controls maintained smooth, intact cellular morphologies (Fig. 5c). These morphological alterations aligned with established mechanisms of unsaturated free fatty acids.⁴⁴ Crystal violet staining and live/dead fluorescence staining (SYTO-9/PI) tests were performed to determine the biofilm disruption ability of EPA (Fig. 5d, e). EPA also exhibited dose-dependent biofilm disruption activity against both MRSA and CREC at concentrations of 0.1–0.8 mM. The data in Fig. 5f, g revealed significantly reduced bacterial viability in both MRSA and CREC biofilms treated with 0.4 mM and 0.8 mM EPA, whereas the control groups maintained dense, viable biofilms (predominantly green fluorescence). To evaluate the therapeutic potential of EPA under physiologically relevant conditions, we established a HUVEC–bacteria coculture model at an optimal multiplicity of infection (MOI) of 1 (optimized for high infection efficiency and minimal cytotoxicity).⁴⁵ A scratch assay of HUVECs cocultured with MRSA and CREC was performed with 0.2 and 0.4 mM EPA (Fig. 5h). HUVECs cocultured with MRSA and CREC without EPA treatment were used as the vehicle group, and those not cocultured with either MRSA or CREC were used as controls. Compared with vehicle treatment, EPA treatment significantly increased endothelial wound closure rates over 48 h (Fig. 5i). These findings collectively demonstrated that EPA, produced through Ker/Ag⁺/Rh-activated LDL-C endocytosis, showed potent antibacterial activity against planktonic and biofilm-embedded MRSA and CREC.

Fig. 5.

Antibacterial activity of EPA against planktonic and biofilm-embedded MRSA and CREC. a Schematic of the antibacterial test against planktonic bacteria. b CFU assay of EPA at varying concentrations against planktonic bacteria and corresponding quantitative inhibition rates. c SEM image of EPA-induced morphological damage to bacterial membranes; scale bars: 1/2 μm. d Schematic of the antibacterial test against MRSA/CREC mixed biofilms. e Crystal violet staining of biofilms treated with varying EPA concentrations and corresponding quantification analysis. No EPA treatment was used as a control. f, g CLSM images of live/dead staining in biofilms and corresponding quantification analysis of bacterial viability; scale bars: 100 μm. h Schematic of the scratch assay using a HUVEC–MRSA/CREC mixed bacterial coculture infection model. i Cell migration behaviors and wound closure rates of infected HUVECs following EPA treatment; scale bar: 100 μm. The data are presented as the mean ± SD from three independent experiments. Statistical significance was determined by one-way ANOVA (Tukey’s multiple comparisons test) for b, e, and i or two-way ANOVA (multiple comparisons) for g. ns: not significant; *P < 0.05, **P < 0.01, ***P < 0.001 vs control, ^###P < 0.001 vs 0.4 mM EPA

Rhein enhances LDL-C endocytosis to potentiate EPA accumulation and anti-biofilm capacity

To verify EPA production through the rhein-activated LDL-C endocytosis pathway, we performed systematic validation experiments. We selected a rhein loading of 0.05–0.4 mg/mL in the Ker/Ag⁺/Rh hydrogel for testing on its negligible effect on cell viability (Supplementary Fig. 14a). Using the MRSA/CREC-infected HUVEC model, western blot analysis revealed that compared with the vehicle and Ker/Ag⁺ treatment groups, the Ker/Ag⁺/Rh group upregulated the expression of proteins associated with the LDL-C endocytosis pathway, including LDLR, LDLRAP1/ARH, and Clathrin (Supplementary Fig. 14b, c). Molecular docking simulations revealed LDLR as the primary biological target of Rh, demonstrating strong binding affinity (–8.702 kcal/mol) through hydrogen bonding and π-anion interactions at the protein’s active site (Fig. 6a, b). RT‒qPCR analysis confirmed that this protein-level activation occurred without transcriptional regulation, as Ldlr mRNA expression remained unchanged between the Ker/Ag⁺/Rh and Ker/Ag⁺ groups (Supplementary Fig. 15a).

Fig. 6.

Rhein enhances LDL-C endocytosis to potentiate EPA accumulation and anti-biofilm capacity. a Molecular docking simulation of Rh-LDLR interactions, demonstrating hydrogen bonds and π-anion interactions at the LDLR active site. b Analysis of the binding affinity of rhein for proteins related to LDL-C endocytosis, including LDLR, Clathrin and LDLRAP1/ARH. c Schematic of a coculture infection model of HUVECs with MRSA and CREC-mixed bacteria. d Scratch assay of infected HUVECs treated with Ker/Ag⁺, Ker/Ag⁺/Rh, and Ker/Ag⁺/Rh with PCSK9 intervention (n = 3/group). e mRNA expression levels of IL-1β, IL-6, and TNF-α in HUVECs determined by RT‒qPCR analysis (n = 3/group). f Protein levels of LDLRAP1/ARH and Clathrin in HUVECs determined by western blot analysis (n = 3/group). g Biochemical analysis of free cholesterol, amino acids, and free fatty acids in HUVECs (n = 3/group). h ELISA of LDL-C and EPA in HUVECs (n = 4/group). i Schematic of the wound healing test using MRSA/CREC mixed biofilm-infected T2DM mice. j Representative wound images and quantified closure rates after treatment with Ker/Ag⁺/Rh, Ker/Ag⁺/Rh + PCZ, and Ker/Ag⁺/Rh + PCZ + EPA (n = 3/group). k CLSM images of live/dead staining in biofilms (green for live bacteria and red for dead bacteria) with quantification of bacterial viability in biofilms (n = 3/group). Data are represented as the mean ± SD from two independent experiments unless otherwise specified. Statistical significance was analyzed via one-way ANOVA with Tukey’s multiple comparisons test for d–h, and j or two-way ANOVA with multiple comparisons for k. Scale bars: 200 μm (d) and 150 μm (k). * vs Ker/Ag⁺/Rh, ^# vs Ker/Ag⁺/Rh + PCZ. ns: not significant; *P < 0.05, **P < 0.01, ***^/###P < 0.001

LDLR plays a pivotal role in LDL-C endocytosis through a well-characterized mechanism in which its extracellular segment binds LDL-C while the intracellular segment recruits LDLRAP1/ARH, facilitating clathrin-mediated vesicle formation for lysosomal degradation and LDLR recycling.¹⁸ LDLR degradation is regulated through two primary mechanisms: 1) Proprotein Convertase Subtilisin/Kexin Type 9 (PCSK9) binds to LDLR and promotes its internalization and lysosomal degradation, thereby preventing receptor recycling to the cell surface; and 2) the intracellular ubiquitin ligase inducible degrader of the LDLR (IDOL; also known as MYLIP) mediates LDLR degradation via the ubiquitin‒proteasome pathway.^18,46,47 We treated infected HUVECs with recombinant PCSK9 and IDOL protein and subsequently assessed LDLR protein expression levels. PCSK9 was selected on the basis of its complete antagonistic effect on LDLR at a concentration of 0.8 μg/mL (Supplementary Fig. 15b, c). To determine whether the antibacterial ability of Ker/Ag⁺/Rh is associated with the LDL-C endocytosis pathway, we conducted rescue experiments using PCSK9. As shown in Fig. 6c, d, MRSA/CREC-infected HUVECs treated with Ker/Ag⁺/Rh (0.05 mg/mL rhein) displayed enhanced migration ability, which was attenuated by PCSK9 cotreatment. We collected cell samples from different treatment groups for RT‒qPCR analysis. The results revealed that compared with the vehicle, both the Ker/Ag⁺ and the Ker/Ag⁺/Rh hydrogels significantly downregulated the expression of proinflammatory cytokines (IL-1β, IL-6, and TNF-α). However, the addition of PCSK9 partially restored the expression of proinflammatory factors in the Ker/Ag⁺/Rh group (Fig. 6e). Western blot analysis confirmed that PCSK9 reversed the Ker/Ag⁺/Rh-induced upregulation of LDLRAP1/ARH and Clathrin protein expression (Fig. 6f). Biochemical assays and ELISA demonstrated that Ker/Ag⁺/Rh treatment increased the levels of key metabolic intermediates, including free cholesterol, amino acids, free fatty acids, and, specifically, EPA, whereas these effects were counteracted by PCSK9 intervention (Fig. 6g, h). Notably, while both Ker/Ag⁺ and Ker/Ag⁺/Rh reduced LDL-C levels relative to those in the vehicle group, the addition of PCSK9 to the Ker/Ag⁺/Rh treatment selectively reversed the effect of rhein, restoring LDL‑C levels to those observed in the vehicle group, thereby confirming the pathway-specific modulation of rhein treatment.

We further employed prochlorperazine dimaleate (PCZ), a specific inhibitor of Clathrin-mediated endocytosis,^48,49 to investigate the relationship between EPA accumulation mediated by LDL-C endocytosis and biofilm-infected diabetic wound healing (Fig. 6i). Pharmacological inhibition of LDL-C endocytosis with PCZ significantly impaired wound healing kinetics, whereas exogenous EPA administration reversed this deficit (Fig. 6j), indicating that LDL-C endocytosis promoted wound healing in biofilm-infected diabetic wounds through EPA production. Live/dead biofilm imaging (SYTO-9/PI) revealed that the antibacterial efficacy of the PCZ-treated wounds reduced (green fluorescence), whereas EPA supplementation restored antibacterial efficacy, as evidenced by increased bacterial killing (red fluorescence), confirming that EPA production by the LDL-C endocytosis pathway contributed to the inhibition of biofilm infection in diabetic wounds (Fig. 6k). These findings establish a complete mechanistic cascade in which rhein binds to LDLR, enhances LDL-C endocytosis, and promotes the production of EPA metabolites that confer antibiofilm activity. Rhein treatment provides a novel therapeutic framework for developing targeted wound healing interventions through the modulation of lipid endocytosis metabolism.

Discussion

Biofilms are major contributors to nonhealing diabetic DFIs. While debridement remains the clinical standard for biofilm removal, its efficacy is limited by rapid biofilm reformation.⁵⁰ Clinical guidelines recommend combining debridement with topical antibacterial agents to treat DFIs,⁵¹ yet biofilms impede drug penetration (e.g., Ag⁺) and create a hypoxic, nutrient-deprived microenvironment that reduces bacterial metabolic activity and antibiotic susceptibility.^6,52–55 To identify novel therapeutic targets. We prospectively collected wound samples at initial presentation from patients with MRSA and CREC biofilm-infected DFU. These patients were then followed for three months and stratified into DFI-healer and DFI-nonhealer groups on the basis of their clinical outcomes at the end of this period. A comparative analysis of protein and metabolite profiles was then conducted between these groups. Unlike prior studies focusing solely on microbial or host mechanisms, we integrated host lipid metabolism (free fatty acids), drug-resistant pathogens, vascular endothelial cells, and LDL-C endocytosis into a unified framework to evaluate wound healing.

In healing wounds, we observed activation of the LDLR-mediated LDL-C endocytosis pathway, leading to elevated free fatty acid levels. Free fatty acids are derived primarily from triglyceride lipolysis and cholesterol metabolism. In triglyceride metabolism, adipocytes undergo lipolysis, breaking down triglycerides into free fatty acids and glycerol,⁵⁶ which becomes particularly vital during periods of increased ATP demand.⁵⁷ Cholesterol metabolism, especially the LDL-C endocytosis pathway, plays a pivotal role in atherosclerosis development,^58,59 with LDLR-mediated LDL-C endocytosis serving as the central mechanism for maintaining cholesterol homeostasis.⁶⁰ Diabetic ulcers, often associated with abnormal glucose and lipid metabolism,¹³ exhibit enhanced vascular deposition of LDL-C, which undergoes endocytosis and is metabolized into free fatty acids.¹⁸ Emerging evidence highlights free fatty acids as potent biofilm disruptors.⁴⁴ They inhibit bacterial adhesion, virulence, and flagellar motility while activating stress regulators that induce biofilm dispersion and antibiotic sensitization.^19,61 Additionally, free fatty acids restore the acidic skin pH, which is disrupted in chronic wounds. Our findings reveal a novel host defense strategy in which LDL-C is repurposed to generate antibacterial free fatty acids.

Our previous studies revealed that rhein had high antibacterial activity against MRSA and CREC, especially when it was combined with Ag⁺, which resulted in significant synergistic effects. Inspired by this, we designed a hydrogel dressing for codelivery of rhein and Ag⁺ in a slow release mode. Extensive research has demonstrated that hydrogel-based wound dressings can accelerate DFU healing and improve patient comfort by maintaining a moist wound environment and softening necrotic tissues and scabs.⁶² Notably, the design of this hydrogel incorporates several distinct advantages: (1) The use of skin-derived natural proteins as the sole matrix material. Keratin, a key intermediate filament protein expressed in epithelial cells, plays crucial roles in maintaining keratinocyte proliferation and epithelial integrity.^63,64 Notably, keratin exhibits excellent biocompatibility and wound-healing capacity, making it an attractive biomaterial for therapeutic applications.⁶⁵ (2) High clinical translation potential: This hydrogel was formed via spontaneous crosslinking through a simple physical mixing process. This crosslinker-free strategy not only preserves the inherent biocompatibility of the hydrogel but also facilitates large-scale production and future clinical translation. (3) Excellent clinical applicability: The unique dynamic cross-linking mechanism enables the injectability and shape adaptability of this hydrogel, fully meeting the clinical needs for filling irregular wounds and convenient use. (4) Codelivery of rhein and Ag⁺ in a slow-release mode achieves synergistic antibacterial effects and prevents potential toxicity risks from burst drug release.

In vivo studies using MRSA/CREC biofilm-infected T2DM models demonstrate that Ker/Ag⁺/Rh promotes infected wound healing via a coordinated mechanism involving inhibition of biofilm reformation, inhibition of bacterial growth, and anti-inflammatory effects. This action is driven by the rhein-induced upregulation of LDLR, which enhances LDL-C endocytosis and subsequent EPA accumulation. Additionally, EPA synergizes with Ag⁺, further enhancing antibacterial efficacy. Our results indicate that the antibacterial mechanism of Ker/Ag⁺/Rh is associated with the downregulation of key virulence genes involved in biofilm formation and adhesion, namely, the ica operon and fnbA in MRSA and ompA, papC and KpsM in CREC. These findings imply that EPA may interfere with extracellular polymeric substances by binding to polysaccharides/proteins, disrupting their cross-linking to reduce biofilm adhesion and mechanical strength, thereby facilitating deeper penetration of antibacterial agents.⁶⁶ Additionally, EPA can destabilize the biofilm architecture through hydrophobic interactions with lipid components in extracellular polymeric substances.⁶⁷ Owing to its lipophilic nature and similarities in polarity and solubility, EPA can also be incorporated into the bacterial cell wall. It acts synergistically with Ag⁺ to disrupt cell wall integrity, leading to increased membrane permeability, leakage of intracellular components, and ultimately bacterial death. This direct antibacterial mechanism effectively reduces viable bacterial counts within the biofilm.¹⁹

At the host-cell level, our multiomics investigation integrating clinical sample proteomics, untargeted metabolomics, network pharmacology predictions, and quantitative lipidomics of Ker/Ag⁺/Rh-treated mouse tissues revealed that rhein exerts its therapeutic effects on DFUs through the selective modulation of LDLR-mediated LDL-C endocytosis, a specific branch of lipid metabolism distinct from triglyceride-related pathways such as PPAR signaling.⁶⁸ In this study, we demonstrate a mechanism whereby EPA effectively inhibits biofilm formation and synergizes with Ag⁺ to exert a cascade of antibacterial effects. Interestingly, the metabolic products of this pathway, particularly free cholesterol and amino acids, play positive roles in wound repair by serving as essential building blocks for cell membrane regeneration and supporting tissue reconstruction.⁶⁹ This finding is especially significant in the context of ulcerative wounds where epidermal/dermal deficiency creates an increased demand for cholesterol during rapid repair phases, which cannot be met solely by endogenous synthesis, making the LDL-C endocytosis pathway a critical compensatory mechanism for the supply of membrane components during active healing processes.⁷⁰ Notably, EPA enhances the migration of HUVECs under infection conditions, a process critical for angiogenesis and re-epithelialization during wound healing. In summary, under infectious stress, rhein embedded in the Ker/Ag⁺/Rh hydrogel enhances the endocytosis of LDL-C, thereby converting the conventionally “harmful” lipid LDL-C into the unsaturated free fatty acid EPA. EPA effectively inhibits bacterial biofilm reformation and suppresses planktonic growth, thus functioning as an endogenous antibiotic. The generation of EPA helps to enhance the antibacterial efficacy of Ag⁺, collectively contributing to accelerated wound healing. By innovatively reprogramming detrimental lipids into beneficial antibacterial molecules through host metabolic pathways, this strategy not only circumvents the inherent limitations of using unsaturated fatty acids as topical agents, including oxidation, instability, and difficulties in dosage control but also offers a novel metabolic intervention for the treatment of drug-resistant bacterial wound infections (Fig. 7). Importantly, the host-directed metabolic reprogramming strategy described herein has therapeutic potential against biofilm infections caused by a wide range of pathogens beyond MRSA and CREC.^42,71 Future studies are warranted to test this hypothesis and explore whether targeting the host LDL-C endocytosis pathway can be generalized as a broadly applicable anti-infective strategy.

Fig. 7.

Schematic diagram summarizing the key findings of this study. Multiomics analysis of diabetic foot ulcer patient tissues revealed that activation of the LDLR-mediated LDL-C endocytosis pathway and subsequent free fatty acid production serve as critical biomarkers of healing in biofilm-infected wounds (a). The Ker/Ag⁺/Rh hydrogel effectively suppressed biofilm reformation and promoted wound healing in diabetic mice. The key hydrogel component, rhein, binds and enhances LDLR expression to facilitate LDL-C endocytosis. This metabolic reprogramming leads to the intracellular generation of EPA, an endogenous antibacterial fatty acid, which inhibits biofilm formation. Rhein- and Ag⁺-loaded hydrogels provide a groundbreaking combinational therapeutic strategy (b). This figure was created using Figdraw 2.0 (Home of Researchers)

This study has several limitations. First, our study did not systematically collect detailed data on patients’ statin use, which is a potential confounder of lipid metabolism. Although statin exposure appeared comparable between groups, the absence of comprehensive data precludes precise assessment of its potential effects. Second, the in vivo experiments were performed using two types of mixed bacterial infection animal models (MRSA and CREC), which may not fully reflect the complexity of polymicrobial wound environments. Future studies employing polymicrobial infection models are warranted to address this issue. Additionally, we do not propose that this regimen is universally applicable to all patients. Rather, we emphasize that for individuals with deficient endogenous EPA synthesis, alternative non-EPA-dependent antibacterial and wound-healing strategies would be more appropriate.

Materials and methods

Human subjects

Human subjects were recruited following approval by the Ethics Committee of Yueyang Hospital of Integrated Traditional Chinese and Western Medicine, Shanghai University of Traditional Chinese Medicine. All participants provided written informed consent. Specimens were collected in compliance with Good Clinical Practice (GCP), the National Medical Products Administration (NMPA) Medical Device Clinical Trial Quality Management Practices, the Declaration of Helsinki, and the International Ethical Guidelines for Biomedical Research Involving Human Subjects. The clinical trial was registered at ChiCTR (ChiCTR2400084803).

On the basis of sample size estimates from prior tissue omics studies in diabetic foot ulcers (DFU),^72,73 we enrolled 45 inpatients with biofilm-infected DFU caused by MRSA and CREC from the Department of Vascular Surgery at Yueyang Hospital between May 2023 and September 2024. Ulcer tissues were collected postdebridement. Patients were stratified into healed (DFI-healers) and nonhealers (DFI-nonhealers) groups after 3 months of follow-up; 36 participants completed the study. Additionally, discarded foot skin tissue was collected from 10 diabetic outpatients without ulcers (DM-nonulcers) after foot tumor resection or orthopedic surgery.

Inclusion criteria for patients with DFIs

Patients who met the following criteria were enrolled: aged ≥ 40 and ≤ 90 years; diabetes duration > 10 years; foot ulcer duration ≥ 4 weeks; ulcer area 6–20 cm²; Wagner grade 2–4; MRSA and CREC biofilm-confirmed infections by bacterial culture and antibacterial susceptibility testing; and scheduled for surgical debridement at our institution.

Inclusion criteria for DM-nonulcers

DM-nonulcers were defined as those who were aged ≥ 40 and ≤ 90 years; had type 2 diabetes without a history of foot ulcers; who did not have chronic comorbidities or prior foot surgeries; and who were scheduled for resection of benign skin lesions (e.g., lipomas, nevi) or orthopedic foot procedures.

Exclusion criteria for the DFI cohort

The exclusion criteria included comorbidities that impaired DFU healing (e.g., malignancies, vasculitis, and immune disorders), severe hepatic dysfunction, end-stage renal disease, long-term radiotherapy/chemotherapy/growth factor therapy, inadequate blood supply,⁷⁴ infection signs per IDSA/IWGDF 2023 criteria,⁷⁵ osteomyelitis, topical/systemic antibiotic use within 2 weeks preenrollment (to exclude plankton-driven wound pathology),⁷⁶ nondiabetic foot ulcers, uncontrolled hypertension (hypertensive crisis/malignant hypertension), severe cardiovascular/cerebrovascular diseases (e.g., class IV cardiac dysfunction), and acute infections.

Exclusion criteria for DM-nonulcers

Exclusions included medication use of systemic corticosteroids, immunosuppressants, or biological agents; systemic antimetabolite/cytotoxic chemotherapy; lipid‑modifying agents (e.g., fibrates, sustained‑release niacin); any investigational new drugs/biological agents within one month prior to study enrollment; prior foot surgeries; and diagnoses of diabetic foot, chronic hypertension, nephropathy, or cardiopulmonary diseases.

Mice

All animal experiments were approved by the Ethics Committee of Yueyang Hospital (Approval No. 231932 for mice and YYLAC‑2025‑287‑3 for rats). C57BL/6 J mice (10 weeks old, equal sex distribution) and rats (250–300 g, both sexes) were obtained from Shanghai SLAC Laboratory Animal Co., Ltd. (License, SCXK [Shanghai] 2022-0004; Certification, 20220004024199). Mice were housed under barrier-maintained, specific pathogen-free conditions (temperature, 23 ± 2 °C; humidity, 40% ± 10%; 12 h light/dark cycle) with ad libitum access to a high-fat/high-glucose diet and water. Health status was monitored daily.

Bacteria

The biofilm-forming strains methicillin-resistant Staphylococcus aureus (MRSA) USA300 (ATCC^® BAA-1717™) and carbapenem-resistant Escherichia coli (CREC) clinical isolate RJ112 were utilized. The quality control strains included MSSA ATCC^® 29213™ (VAN MIC, 1–2 μg/mL) and Escherichia coli ATCC^® 25922™ (ampicillin MIC 2–8 μg/mL). Clinical isolates were provided by Prof. Min Li’s laboratory (Renji Hospital, Shanghai Jiao Tong University).

Cell lines

The SV40T-transformed human umbilical vein endothelial cell line PUMC-HUVEC-T1 (Resource ID, 1101HUM-PUMC000437) was obtained from the National Biomedical Experimental Cell Resource Bank. The L929 murine fibroblast line (Catalog No. 1101MOU-PUMC000403) was obtained from the Chinese Academy of Sciences Cell Bank.

Main materials, reagents and drugs

The main chemical reagents, all antibodies, and related experimental materials are listed in the Supplementary Materials.

Preparation of hydrogels

All the hydrogels were fabricated using regenerated keratin as the matrix material, which was extracted following the method described in our previous work.⁷⁷ The Ker/Ag⁺ hydrogel was synthesized by blending a keratin solution with a silver nitrate (AgNO₃) solution at room temperature under stirring for 60 s. The final concentration of keratin in the mixture was fixed at 18% (w/v), while the Ag⁺ concentrations were adjusted to 0.5, 1, or 2 mmol/L. The mixture was subsequently allowed to stand statically at room temperature to facilitate hydrogel formation. For the rhein-loaded Ker/Ag⁺ hydrogel, rhein was first dissolved in a sodium bicarbonate solution (0.2 M NaHCO₃, pH 8.3) and heated to ensure complete dissolution. This solution was then mixed with the freshly prepared Ker/Ag⁺ hydrogel precursor under gentle mixing to achieve a certain rhein concentration, such as 1.5 mg/mL. The resulting mixture was allowed to form a hydrogel under static conditions at room temperature (recorded as Ker/Ag⁺/Rh). The keratin hydrogel (Ker) was prepared using the same procedure as the Ker/Ag⁺ hydrogel, except that silver nitrate was replaced with an equal volume of phosphate-buffered saline (0.01 M, pH 7.4).

In vitro release of rhein and silver ions

The release profile of rhein and silver ions from the Ker/Ag⁺/Rh hydrogel was measured by incubating a cylindrical sample (10 mm diameter × 12 mm height) in 5 mL of a PBS solution (0.01 M, pH 7.4) at 37 °C. The release medium was collected and replaced with fresh PBS at predetermined intervals. The rhein concentration was quantified by high-performance liquid chromatography (HPLC) with UV detection at 254 nm, while Ag⁺ release was measured using inductively coupled plasma‒mass spectrometry.

Confocal laser scanning microscopy (CLSM)

Bacterial viability within the biofilms was assessed via live/dead staining. The samples were stained with SYTO-9/PI (5 μM/2 μg/mL; 37 °C, 15 min, dark), rinsed in saline, and subjected to CLSM (Olympus FV1200) for 3D reconstruction. SYTO-9 and PI were excited at 488 nm and 559 nm, with emission collected at 500–540 nm and 600–680 nm, respectively. For tissue samples, Z-stack acquisition was restricted to 50 μm from the wound surface to exclude host cells. Z-stacks were acquired at 1 μm intervals and reconstructed into 3D renderings using FV10-ASW software (Olympus).

Proteomics, metabolomics and targeted lipidomics sequencing

Proteomic profiling was conducted using 4D-DIA mass spectrometry. MS raw data were analyzed using DIA-NN (v1.8.1) with a library-free method. the uniprot_proteomeUP000005640_hu database (a total of 82492 sequences) was used to construct a spectral library with deep learning algorithms of neural networks. The option of MBR was employed to create a spectral library from DIA data and then reanalyse using this library. The FDRs of the search results were adjusted to < 1% at both the protein and the precursor ion levels, and the remaining identifications were used for further quantification analysis.

The differentially abundant metabolites in ulcer tissues were obtained through untargeted metabolomics sequencing. The sample stored at −80 °C was thawed on ice. The sample was ground with liquid nitrogen. A 400 μL solution (methanol:water = 7:3, V/V) containing an internal standard was added to 20 mg of ground sample and shaken at 1500 rpm for 5 min. After being placed on ice for 15 min, the sample was centrifuged at 12,000 rpm for 10 min (4 °C). A 300 μL aliquot of the supernatant was collected and stored at −20 °C for 30 min. The sample was then centrifuged at 12,000 rpm for 3 min (4 °C). Two hundred microliter aliquots of the supernatant were transferred for LC‒MS analysis. The original data file acquired by LC‒MS was converted into mzXML format by ProteoWizard software. Peak extraction, peak alignment and retention time correction were performed by the XCMS program. The “SVR” method was used to correct the peak area. The peaks whose detection rates were lower than 50% in each group of samples were discarded. After that, metabolic identification information was obtained by searching the laboratory’s self-built database, integrated public database, AI database and metDNA.

The relative quantification of lipids was performed using a targeted lipidomics method via liquid chromatography–tandem mass spectrometry with multiple reaction monitoring. The analysis was carried out using an ExionLC UHPLC system coupled with a QTRAP 6500+ mass spectrometer (AB SCIEX) controlled by Analyst (v.1.6.3) software (AB SCIEX).

The extraction, detection, and quantitative analysis were performed by Wuhan MetWare Biotechnology Co., Ltd. (www.metware.cn).

Immunohistochemistry

Protein spatial distribution was evaluated by immunohistochemistry. Deparaffinized sections were subjected to heat-induced epitope retrieval (citrate buffer, pH 6.0, 95 °C), BSA blocking, and primary antibody incubation (4 °C, overnight). HRP-conjugated secondary antibodies and DAB chromogen were applied before hematoxylin counterstaining. Quantification analysis was performed by ImageJ software (mean optical density, 200 × fields).

ELISA Quantification of lipid metabolites was performed using commercial ELISA kits according to the manufacturer’s specific protocols. Samples and standards were added to antibody-coated wells, followed by incubation with HRP-conjugated detection antibodies. After thorough washing, the enzymatic reactions were developed with the TMB substrate. The absorbance was measured at 450 nm, and the metabolite concentrations were calculated against standard curves.

Microplate colorimetric assays

Free fatty acid, amino acid, and cholesterol levels were determined using colorimetric microplate assays. Tissue homogenates (0.1 g in isopropanol) or cell lysates (5 × 10⁶ cells) were centrifuged (8000 × g, 10 min). The supernatants were analyzed using commercial kits per the manufacturer’s protocols (absorbance, 490–570 nm).

Determination of the minimum inhibitory concentration (MIC) and fractional inhibitory concentration (FIC) index

Antibiotic susceptibility was determined via 96-well plate checkerboard broth microdilution combined with the agar dilution method per CLSI M07-A10. Inocula (5 × 10⁵ CFU/mL in cation-adjusted Mueller-Hinton broth) were exposed to serially diluted antibiotics (512−0.5 μg/mL) and incubated for 18–24 h at 35 °C. MICs were recorded as the lowest concentration inhibiting ≥ 99.9% of bacterial growth (≤ 50 CFU per MH agar plate) instead of judging by visible growth. Drug interactions were evaluated using the FIC index and calculated according to the literature.⁷⁸ FIC index ≤ 0.5, between 0.5 and 2, and > 2 were defined as synergistic, indifferent, and antagonistic effects, respectively.⁷⁹

Plate-counting assays

Antibacterial efficacy against planktonic bacteria was determined according to a modified ISO 22196 protocol under diabetic wound-mimicking conditions. In brief, Mueller–Hinton broth supplemented with NaHCO (pH 7.5–8.5), glucose (20 mM), and 10% FBS was used to simulate hyperglycemia and wound exudate components. A 220 µL reaction system was prepared; bacterial suspensions (1 ×10⁵ CFU/mL in 200 μL) were mixed with 20 µL of test agent, including PBS (control), VAN (30 μg/mL), CFM (1 mg/mL), and the Ker, Ker/Ag⁺, or Ker/Ag⁺/Rh hydrogel leaching solutions at 37 °C. The hydrogel leaching solutions were obtained by immersing 20 mg of Ker/Ag⁺/Rh, Ker/Ag⁺, or Ker hydrogels in 10 mL of complete cell culture medium (DMEM/F12 with 10% FBS) at 37 °C for 24 h, followed by sterile filtration through 0.22 μm membranes. The final concentrations in the reaction system were as follows: VAN, 2.73 µg/mL (from a stock of 30 µg/mL, 11-fold diluted during plating); and CFM, 91 µg/mL (from a stock of 1 mg/mL, 11-fold diluted during plating). On the basis of the initial hydrogel composition, the concentrations of Ag⁺ and rhein in the leachate were 11.01 μM and 16.51 μg/mL, respectively. After 24 h of incubation, the cultures were stopped with an ice-cold neutralizer (0.1% Tween-80 in 0.9% NaCl). Serial 10⁻ dilutions were plated on Tryptic Soy Agar, and colonies were counted with an automated colony counter. Bactericidal rate (%) = (CFU of control − CFU of sample) /CFU of control × 100%.

Crystal violet staining

The inhibitory effects of different treatments on biofilm formation were assessed using crystal violet staining. MRSA and CREC were co-incubated during biofilm formation, which included PBS (control), 30 μg/mL VAN or 1 mg/mL CFM, and Ker/Ag⁺ or Ker/Ag⁺/Rh hydrogel leaching solutions. Briefly, an overnight bacterial culture was diluted in Mueller–Hinton broth to an optical density at 590 nm (OD₅₉₀) of 0.10 ± 0.02.⁸⁰ Then, 100 µL of this bacterial suspension was added to each well of a 96-well plate containing a sterile 6 mm circular silicone sheet together with 100 µL of the respective treatment. The final concentrations in the reaction system were as follows: VAN at 15 µg/mL; CFM at 0.5 mg/mL; Ker/Ag⁺ hydrogel leachate with a final Ag⁺ concentration of 5.5 µM; and Ker/Ag⁺/Rh hydrogel leachate with a final Ag⁺ concentration of 5.5 µM and a final Rh concentration of 8.3 µg/mL. The bacterial suspension was subsequently incubated statically at 37 °C for 72 h under constant humidity to allow biofilm formation. After 72 h of incubation, the soft rubber substrates were gently rinsed with sterile PBS (pH 7.4) to remove planktonic bacteria, treated with 2.5% glutaraldehyde for 2 h, and air-dried. Biofilms were stained with 0.5% crystal violet (1 mL/well, 25 min), rinsed until the effluent was colorless, and destained in 95% ethanol (1 mL, 15 min with agitation). Destaining solution was transferred to a fresh 96-well plate for OD₅₉₀ measurement.

In vivo assessment of diabetic wound healing

Diabetic wound closure was evaluated in MRSA and CREC biofilm-infected mice and rat models. C57BL/6 J mice were fed a high-fat/high-glucose diet for three months and subsequently administered STZ (30 mg/kg/day for 7 days, i.p.) to establish a T2DM model. The rats were fed a high‑fat/high‑glucose diet for four weeks, followed by STZ injection at 50 mg/kg/day for 7 days (i.p.). Successful induction of diabetes was confirmed by a fasting blood glucose concentration exceeding 11.1 mmol/L, accompanied by polyuria and polydipsia. Two full-thickness dermal wounds (1 cm in diameter) were created per mouse after isoflurane anesthesia and skin disinfection. For fixed wounds, silicone rings were sutured periwound. An infected wound model was established by applying a cotton ball saturated with a mixed bacterial suspension containing MRSA and CREC at a concentration of 1 × 10⁸ CFU/mL in a 1:1 ratio.⁸⁰ The wound was then covered with 3 M Tegaderm film for 72 h to allow the formation of a mature biofilm. The biofilm-infected wounds were randomly divided into five groups (n = 6/group) and treated with urea ointment (vehicle), mupirocin ointment (MO), 0.5 mL of Ker hydrogel, 0.5 mL of Ker/Ag⁺ hydrogel, or 0.5 mL of Ker/Ag⁺/Rh (1.5 mg/mL rhein) hydrogel. Diabetic wounds without biofilm infection composed the uninfected group. Treatment was initiated after biofilm establishment. Each treatment involved rinsing the wound with saline before topical application of the corresponding agent. The first day of treatment was designated day 1, and all treatments were applied once daily. Full-thickness ulcer tissue was collected on days 3, 7, and 12 posttreatment for downstream molecular and histopathological analyses (n = 6 biologically independent samples per time point).

For the experiments designated as rescue regimens, the biofilm-infected wounds were divided into four groups (n = 3/group) and treated with urea ointment (vehicle group), Ker/Ag⁺/Rh hydrogel, Ker/Ag⁺/Rh hydrogel plus PCZ, or Ker/Ag⁺/Rh hydrogel plus PCZ plus EPA. PCZ was administered once daily via tail vein injection at doses of 0.8 mg/kg and 50 mg/kg, followed by the application of the hydrogels. EPA was applied topically to the wound at a concentration of 0.4 mM prior to coverage with the hydrogels. Wound closure was photographically recorded daily. On day 4 post-treatment, full-thickness ulcer tissues were harvested for live/dead bacterial staining and CLSM analysis.

Determination of the wound bacterial load

On day 3 after treatment, the ulcer wounds in the mice were cleansed with sterile saline to remove necrotic tissue and moisten the wound surface. A Levine swab was firmly rotated over the entire wound area to ensure adequate collection of exudates and was immediately placed into a centrifuge tube containing 1 mL of sterile PBS. The tube was vortexed for 1 min to prepare a bacterial suspension. Tenfold serial dilutions were performed up to 10⁻⁴ and the diluted suspensions were evenly spread onto solid culture media. The bacteria were incubated at 37 °C and 60% humidity for 24 h. Finally, the bactericidal rates (%) were calculated as (CFU of the control − CFU of the treatment)/CFU of the control × 100%.⁸¹

Cell scratch assay

HUVECs were routinely resuscitated and cultured in high-glucose DMEM supplemented with 10% FBS at 37 °C under 5% CO₂. After the cells were seeded into 6-well plates at 1 × 10⁶ in number and incubated until they reached full adhesion and 90% confluence, the cells were infected with MRSA and CREC at an MOI of 1 to establish an infection model.⁸¹ A uniform wound was created in each well using a 200 μl pipette tip, followed by two PBS washes to remove debris and replacement with serum-free drug-containing medium to minimize proliferation effects. The experimental groups included the control (uninfected HUVECs), vehicle (infected untreated), 0.2 mM EPA, and 0.4 mM EPA (both infected and treated) in DMSO. For the PCSK9 rescue assay, infected HUVECs were treated with Ker/Ag⁺, Ker/Ag⁺/Rh, and Ker/Ag⁺/Rh + PCSK9. Images were taken at 0, 12, 24, 36, and 48 h at a fixed position under 4× magnification. Wound areas were measured using ImageJ, and the wound healing rate was calculated as (initial area – area at time t) / initial area × 100%.

Statistical analysis

Statistical analysis was performed by GraphPad Prism 8.0 and R 4.2.0 using appropriate packages. Statistical significance was determined by unpaired or paired Student’s t tests or nonparametric tests for comparisons of two groups, one-way ANOVA or nonparametric tests for comparisons of more than three groups, and two-way ANOVA or Pearson’s correlation for analyzing interactions or assessing correlations. All the results are presented as the mean ± SD. Statistical significance was defined as P < 0.05 after multiplicity correction. The number of replicates or animals used is given in the figure legends.

Author contributions

F.L., G.Y., and L.Z. conceived and supervised the study, designed the research, interpreted the results, and wrote the paper. T.D., J.S., M.C., S.W., F.S., Q.L., L.Z., Y.Z., P.W., and Y.D. performed the experiments and discussed and analyzed the data. T.D., J.S., S.W., and M.C. wrote the initial draft. S.C., J.S., T.D., and S.W. performed the supplementary experiments, processed the corresponding data, and contributed to the writing of the relevant results in response to the reviewers’ comments. F.S., Q.L., L.Z., and Y.D. collected clinical samples and provided clinical guidelines. P.W. and J.Z. provided the clinical guidelines and clinical analysis. Y.W., H.L., X.L., and Y.Z. assisted in providing reagents or providing technical support. F.L. is the primary contact for communication. All the authors have read and approved the article.

Data availability

The mass spectrometry data supporting the findings of this study were processed using Proteome Discoverer 3.0 (for proteomics) and Compound Discoverer 3.3 (for metabolomics). The proteomic dataset has been deposited in the iProX database (Project ID: IPX0011022000; ProteomeXchange ID: PXD060612), which contains two subprojects: IPX0011022001 and IPX0011022002. The dataset is fully publicly accessible without password restriction (https://www.iprox.cn//page/project.html?id=IPX0011022000). The metabolomics data have been deposited in the Metabolights repository under accession numbers MTBLS13611 (quantitative lipidomics data from 6 mouse samples) and MTBLS13603 (untargeted metabolomics data from human samples). All relevant data are also provided within the Supplementary materials or are available from the corresponding authors upon reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41392-026-02869-0.