https://orcid.org/0000-0002-8479-3372
Figures
Abstract
Escherichia coli-induced bovine mastitis represents a major challenge in dairy production due to the prevalence of multidrug-resistant strains. This study repurposes metformin as a dual-function agent that simultaneously targets bacterial virulence and host inflammation. Epidemiological surveillance identified phylogroup B1 as the most prevalent (52.5%) and resistant E. coli lineage. Against a representative B1 strain, metformin potently inhibited and dispersed bacterial biofilms, and synergized with conventional β-lactam antibiotics. Bacterial transcriptomics revealed metformin downregulated genes critical for membrane integrity and metabolism. In parallel, metformin attenuated the inflammatory response in bovine mammary epithelial cells and in murine and ovine mastitis models. In vivo, it significantly reduced bacterial colonization in mammary tissue and suppressed key pro-inflammatory cytokines. Mechanistically, metformin activated the AMPK/SIRT1 axis, leading to deacetylation of NF-κB p65. In the ruminant model, this culminated in epigenetic regulation, with increased chromatin compaction at promoters of inflammatory genes, and a significant inverse correlation (r = -0.77) between NF-κB binding and chromatin accessibility. Collectively, metformin combats resistant E. coli mastitis through a dual mechanism: disrupting biofilm-dependent bacterial persistence and reprogramming host immunometabolism via AMPK/SIRT1-mediated epigenetic regulation. These findings provide a compelling non-antibiotic strategy for overcoming antimicrobial resistance.
Author summary
Bovine mastitis caused by drug-resistant E. coli is difficult to treat with conventional antibiotics. We explored repurposing metformin, a common diabetes drug, as a new strategy. We found that metformin fights mastitis in two ways at once. First, it directly attacks the bacteria by breaking apart their protective biofilms and damaging their cell membranes. Second, it calms the host’s harmful overreaction to infection by activating a metabolic pathway (AMPK/SIRT1) that turns off inflammatory genes through epigenetic changes. This dual-action approach—targeting both the pathogen and the host’s immune response—was effective in mouse and sheep models of mastitis. Our work suggests metformin could be a promising, resistance-proof therapy for this economically important animal disease.
Citation: Xu T, Cao W, Fan S, Liu R, Zhu H, Lu X, et al. (2026) Repurposing metformin as a dual-function agent to combat E. coli-induced mastitis: Mechanistic insights into biofilm dispersion and AMPK/SIRT1-mediated NF-κB inhibition. PLoS Pathog 22(3): e1014012. https://doi.org/10.1371/journal.ppat.1014012
Editor: Fengyi Wan, Johns Hopkins University Bloomberg School of Public Health, UNITED STATES OF AMERICA
Received: November 10, 2025; Accepted: February 19, 2026; Published: March 6, 2026
Copyright: © 2026 Xu et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All transcriptomic data have been deposited in a public repository (https://ngdc.cncb.ac.cn/gsa/) with GSA accession numbers: CRA036621, CRA036624, and CRA036969. Other data are provided within the manuscript or as supplementary files.
Funding: Z.Y. was supported by the National Key Research and Development Program of China (https://www.most.gov.cn/index.html) with grant number 2023YFF1000902. T.X. and K.Z. were supported by the National Natural Science Foundation of China (https://isisn.nsfc.gov.cn/) with grant number of 32102731 and 32202883, respectively. T.X. was supported by China Postdoctoral Science Foundation (https://www.chinapostdoctor.org.cn/home) with grant number 2023M732994, Natural Science Research of Jiangsu Higher Education Institutions (https://jyt.jiangsu.gov.cn/) with grant number 25KJA230003.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Bovine mastitis, primarily caused by bacterial infection, is diagnosed by pathogen isolation from milk. Escherichia coli is a predominant environmental pathogen responsible for a significant proportion of clinical cases [1,2]. Antibiotic therapy is often ineffective due to the rise of multidrug-resistant (MDR) pathogens, which also pose a public health risk. The World Health Organization has highlighted antimicrobial resistance as a critical global threat. To address this, there is an urgent need to develop safe and effective non-antibiotic alternatives for diseases like bovine mastitis.
Despite the growing resistance, β-lactams antibiotics are potent antimicrobial agents and widely used in livestock as compared to other antibiotics. Notably, over 90% of milk-isolated E. coli exhibit β-lactam resistance (BLR) [3]. This resistance is exacerbated by two key factors: biofilm formation, which protects bacterial communities [4–6], and the impermeable outer membrane of Gram-negative bacteria [7]. Therefore, strategies that disrupt biofilms and compromise membrane integrity are needed.
Metformin, a widely used agent for the first-line prevention of type 2 diabetes and has U.S. Food and Drug Administration (FDA)-approved compounds over the past 5 decades. Notably, the expansive effects of metformin have been evidenced in inflammation-inhibition in cells, animal models, patient records, and randomized clinical trials [8–12]. As an adjuvant to doxycycline, metformin restores susceptibility of tetracycline-resistant E. coli B2 and K. pneumoniae by potentiating intracellular accumulation of antibiotics in the more recent studies [13,14]. Importantly, the potential bactericidal and host-defending effects of metformin on raw milk-derived BLR E. coli B1 and the underlying mechanisms remain largely unknown.
E. coli mastitis is characterized by a damaging inflammatory response driven by the TLR4/NF-κB pathway [15,16]. However, non-antibiotic therapies that can concurrently inhibit this inflammatory response and target the pathogen itself are lacking, particularly against the growing challenge of antimicrobial resistance (AMR). In addition, we have previously reported that metformin exerts anti-inflammatory effects in mastitis by activating AMPK, which inhibits NF-κB nuclear translocation, suppressing pro-inflammatory responses in mammary epithelial cells, and restoring redox balance via the Nrf2 axis in LTA or LPS-challenge in vitro [17,18]. Moreover, metformin ameliorates murine colitis via blocking STAT3 acetylation by reducing acetyl-CoA production [19]. However, its precise antibacterial and anti-inflammatory mechanisms remain incompletely understood and require further investigation. Taken together, we clarified that metformin combats E. coli mastitis through a dual-targeting strategy: disrupting bacterial biofilm (pathogen-directed) and reprogramming host immunometabolism to suppress excessive inflammation (host-directed). Nevertheless, an integrated understanding of its antibacterial and immunomodulatory mechanisms is incomplete.
In this study, we first characterized the predominant MDR E. coli phylogroup from clinical bovine mastitis. We then systematically investigated metformin’s efficacy, focusing on its ability to disrupt biofilms and compromise bacterial membrane integrity. Furthermore, we elucidated its protective immunomodulatory role in vitro and validated its dual therapeutic action in both murine and lactating ruminant infection models. Our integrated approach reveals that metformin combats E. coli mastitis via a dual mechanism, providing novel mechanistic insights and a strong preclinical rationale for its repurposing as a dual-targeting therapeutic agent.
Materials and methods
Ethics statement
The experimental procedures in the current study used for cows (YZU-202502057), mice (YZU-202503173), and sheep (YZUDWLL-202103–312) were approved by the Animal Experiment Committee of Yangzhou University. All experimental protocols were performed in accordance with approved guidelines and regulations.
Experimental design
Bacteria isolation and reagents.
A total of 186 quarter-milk samples from mastitis-infected cows (with positive result using California Mastitis Test and a somatic cell count > 400,000 cells/mL of fresh milk) were randomly collected from four major dairy farms in Jiangsu province of China, including Northern Jiangsu (n = 98), Mid Jiangsu (n = 42), Southern Jiangsu (n = 46). The samples were continuously collected from Spring (March) to Autumn (September) over two seasons in 2023. All milk samples (50 mL for each quarter) were processed within 24 h for the experiment.
The isolation and identification of bacteria were conducted as recommended by the U.S. National Mastitis Council [20,21]. In brief, the milk samples were plated onto blood agar supplemented with 5% fresh sheep whole blood and incubated at 37°C aerobically for 24 h. Based on the morphology of colonies, a single identical colony of each sample was then sub-cultured by streaking on Luria Broth (LB) agar. Plate-cultured bacteria were expanded in nutrient broth at 37°C aerobically for 24 h. All the suspected isolates were further confirmed via 16S rDNA sequencing [22]. The confirmed isolates were kept in 15% glycerol at -80 °C as frozen stock. A PCR method was performed to rapidly determine the phylogenetic groups of E. coli strains into four main groups (A, B1, B2, and D) according to the positive expression of three reference genes (chuA, yjaA, TspE4.C2) as described [23].
Mammalian cell, mouse, and sheep infections.
Primary bMEC cultures were performed as described previously. 2 × 105 of cells were seeded in six-well plates with medium containing Roswell Park Memorial Institute (RPMI) with 10% heat-inactivated fetal bovine serum (Gibco, San Diego, CA, USA) and antibiotics (penicillin 100 IU/ml; streptomycin 100 mg/ml) (Thermo Fisher Scientific, Pleasanton, CA, USA). The incubation of cells was maintained at 37°C with a 5% CO2 condition. The E. coli-derived LPS-induced inflammatory cell model was established and described previously. The optimal doses used in the current study were selected according to the previous study. Cells challenged with LPS (2 μg/mL) for 6 h were selected as the positive control (LPS); Cells pretreated with Metformin at a concentration of 3 mM for 12 h, followed by the LPS (2 μg/mL) challenge for 6 h (LPS + Met). Cells treated with PBS for 18 h were selected as the negative control (NC).
The murine model was employed for rapid, controlled validation of systemic protection and mechanistic pathways, as well as for specifically model mastitis pathology and assess local therapeutic efficacy. 8-week-old female SPF-grade ICR mice (n = 8 in each group) and 1-week lactating female ICR mice (n = 8 in each group) were purchased from the Center of Comparative Medicine at Yangzhou University. Allow adaptation to the housing environment (23°C ± 2°C; 55% ± 10% of humidity) for 1 week in advance. Before intraperitoneal injection, fast the mice for 4 hours to reduce interference from gastrointestinal contents. 0.5 mL of 1 × 108 CFU/mL of E. coli B1 suspension was intraperitoneally or intramammary injected for 6 h as the infected group (EC) (the control groups are injected with an equal volume of sterile saline as Saline or no treatment as NC). A 6-hour E. coli infection was followed by continuous injection of metformin for 16 hours at 6-hour intervals; this group was designated as the Post-infection Metformin (EM) group. Collection of the liver, spleen, and mammary tissues of the mice was accomplished after euthanization for subsequent analysis.
Sheep experiment.
Use 15 local Hu sheep of similar age, similar calving time, weighing approximately 30 kg, and in the early lactation period. The trial may only commence after clinical and laboratory examinations confirm the sheep are healthy. During the experimental period, the animals are fed a total mixed ration (TMR) and provided water ad libitum. The lactating Hu sheep model provides a physiologically relevant ruminant system to evaluate therapeutic efficacy in a target species with mammary gland biology comparable to dairy cows, enhancing translational relevance.
The experimental animals are randomly divided into three groups (n = 5): NC group, EC group, and EM group. The EC group and the EM group receive an intramammary infusion of 3 mL of an E. coli B1 suspension (10⁹ CFU/mL). After 6 hours of the infection, the EM group began receiving intramammary infusions of metformin (at a dose of 100 mg/kg body weight) every 6-hour interval twice. Physiological parameters such as body temperature and respiratory rate are monitored. Punch biopsy was applied for sampling of mammary tissue after treatment with local anesthesia.
Antimicrobial susceptibility
Antimicrobial susceptibility testing was performed on Mueller-Hinton broth (MHB) by the Kirby-Bauer disk diffusion method [24]. The 13 commercial antibiotic disks were conducted in the experiment included tetracycline (TCY, 30 μg), streptomycin (STR, 10 μg), ciprofloxacin (CIP, 5 μg), gentamycin (CN, 10 μg), imipenem (IPM, 10 μg), Meropenem (MEM, 10 μg), Piperacillin (PRL, 100 μg), lincomycin (MY, 2 μg), Cephalothin (KF, 30 μg), penicillin (PCN, 10 μg), Amoxicillin (AML, 20 μg), Ceftazidime (CAZ, 30 μg), Cefotaxime (CTX, 30 μg). The antimicrobial disks were obtained from Tianhe (Tianhe Microbial Reagent Co., Hangzhou, China). The E. coli ATCC 25922 strain was used as the control strain, and the test results were then recorded as susceptible or resistant based on the zone diameter interpretative standards of the Clinical and Laboratory Standards Institute [25].
MIC determination
The determination of MIC for metformin on each bacterial strain was performed using the standard microdilution approach recommended by the CLSI (2017). Briefly, a two-fold dilution of metformin with an equal volume of bacterial cultures containing 1.5 × 106 CFU/ mL in a 96-well microplate is available. The MIC values were recognized when the lowest concentrations of the metformin showed visible bacterial growth.
Bacteria growth curve determination
The overnight culture at 37°CofE.coli B1, MRSA SH01, and K. Peneumoniae H06 from a single colony was prepared and diluted 1:10000 into MHB medium. Then, the diluted cultures of E. coli B1 were incubated for 0 h or the exponential phase (4 h or 8 h) at 37 °C with 200 rpm shaking. Metformin at a concentration of 0, 1 × MIC, 2 × MIC, 4 × MIC, and 8 × MIC was added to the cultures, respectively. The optical density at OD600nm was measured by a spectrophotometer (Tecan, Spark, Switzerland) every 3 h throughout incubation. Meanwhile, the cultures of exponential-phase bacteria were sampled with 50 μL at time points 0, 4, 6, 8, 12, and 24 h for spotting on MHB agar plates and counting colonies after overnight incubation at 37 °C. Experiments were performed with three replicates.
Screening of resistance development in serial passage
Drug resistance development was performed with serial-passage experiments in 96-well microliter plates. Two-fold serial dilution of 100 μL metformin in MHB was added to the wells containing 100 μL of E. coli suspension (1.5 × 106 CFU/ mL). The highest concentration of metformin was selected as 8 times that of MIC. After cultivation of the plates at 37 °C for 24 h, the MIC of the culture was defined as OD600nm < 0.1, and 0.5 × MIC was diluted 1,000-fold in MH and taken as inoculum for the next passage. The repetition was processed for 26 days with triplicates, and the increase in MIC of metformin was calculated.
Checkboard assay
The synergistic effect of metformin together with antibiotics and the fractional inhibitory concentration index (FICI) was measured by checkboard assay. In brief, the MIC of the antibiotics to be tested were firstly determined separately. Antibiotics were diluted as abscissa, while dilutions of metformin were presented as ordinate. Overnight bacterial culture 0.5 Mc-Farland turbidity standardized, followed by the dilution at 1: 100 in MHB. Cultures were incubated at 37 °C for 24 h. 600 nm was taken to determine the optical density using a spectrophotometer (Tecan, Spark, Switzerland). The FICI was assessed according to the formula as recommended [26]:
Criteria: FICI ≤ 0.5 is defined as synergy.
Assessment of biofilm biomass and viable cells in biofilm
The metformin was diluted to 0, 0.5, 1, 2, 4, 8, 16, and 32 mg/mL, and biofilm culture was performed in a 96-well plate as described above. The cultures were washed once with PBS to remove planktonic cells and used for either CV staining or viable cells counting. After 1.5% CV incubation at RT for 30 min, the plate was washed three times with PBS to remove excess dye and dried. Then, the bound dye in each well was dissolved by adding 95% ethanol for 20 min. The optical density of the dissolved dye was detected at 595nm. The viable cells from mature biofilm were collected and plated on agar plates with serial 10-fold dilutions.
Biofilm morphology assay
The morphology of biofilm was determined by using microscopy observation. Metformin was diluted to 0.5 × MIC, 1 × MIC, and 2 × MIC in MHB medium. Overnight culture was added to a 6-well plate for 24 h incubation at 37°Cwithout shaking, and an 18 mm × 18 mm glass cover slide was placed. For live or dead cell determination in mature biofilm, the 24 h-incubated biofilm was treated with or without metformin for another 24 h. The slides were washed with PBS to remove planktonic cells and stained with SYTO9 (Green) and PI (Red) pre-mixture using LIVE/DEAD BacLight Bacterial Viability Kit (L7007, Thermo Fisher Scientific, MA). The images were captured using DMi8 Microsystems BmbH (Leica, Wetzlar, Germany). For the biofilm formation inhibition experiment, 40 μL of overnight culture was added to a 6-well plate containing 1960 μL of diluted metformin. Cells were stained with either 1.5% CV and visualized using DMi8 Microsystems BmbH (Leica, Wetzlar, Germany) or processed for scanning electron microscope using Hitachi Model SU-8010 SEM (Hitachi, Tokyo, Japan).
Cell viability
The CCK-8 Cell Counting Kit (Vazyme, Nanjing, China) was used to determine bMEC viability according to the manufacturer’s protocols. 1 × 105 cells/mL were seeded into 96-well plates. Cells with specific treatment were incubated with 10 mL of CCK-8 at 37 °C for 2 h, followed by measurement of the spectrum value at OD450 using a microplate reader (SPARK, TECAN, Switzerland).
Detection of biochemical parameters in plasma and mammary tissue
SAA and CRP levels in sheep plasma were measured using Enzyme-Linked Immunosorbent Assay (ELISA) kits according to the manufacturer’s instructions. The ELISA kits were commercially purchased from CUSABIO (CSB-E14973Sh and CSB-E15081Sh).Parameters of MDA in sheep plasma and NAD+ in mammary tissue or cells were detected using a commercial kit and performed according to the instructions of the manufacturer. The levels of MDA and NAD+ were measured using the following kits (Beyotime, China): Total Superoxide Dismutase Assay Kit with WST-8 (S0101S), Lipid Peroxidation MDA Assay Kit (S0131S), and NAD+/NADH Assay Kit with WST-8 (S0176S).
RNA isolation and quantitative real-time PCR (RT-qPCR)
The treated bacterial cells, bovine mammary cells, mice, or sheep mammary tissue were taken for total RNA extraction using the Bacteria RNA Extraction Kit or RNA-easy Isolation Reagent (Vazyme, China). 1 μg of RNA was reverse transcribed using HiScript Reverse Transcriptase (Vazyme, China). The RT-qPCR for synthesized cDNA was performed with SYBR AceQ qPCR SYBR Green Master Mix (Vazyme, China). Primers used in the PCR system are provided in S1 Table. The 2-ΔΔCt method was used to present the relative expression of each target gene, as elucidated previously [27]. Internal references, including Gapdh, Uxt, and Rps9 (eukaryotes) and 16s rRNA (prokaryotes) were selected for normalization of each target gene’s expression.
Western-blotting
Protein extraction was conducted by RIPA lysis buffer (Beyotime, Shanghai, China) and was quantified with BCA measurement (Pierce, Rockford, IL, USA). The SDS-PAGE gel with a gradient proportion of 4%-20% was used for separating proteins with different molecular sizes. The primary antibody for target proteins measured was purchased from Cell Signaling Technology (CST, Danvers, MA, USA), followed by the HRP-labeled secondary antibodies. β-actin was taken as an internal reference for normalization of each blot’s quantification.
Immunofluorescence detection
For cellular immunofluorescence, cells seeded on a 12-well plate were fixed with 4% paraformaldehyde (15 min, RT) and washed with PBS. Bovine serum albumin (BSA) (5%) was used for blocking slides and subsequently incubated with primary antibody (Ac-p65 and SIRT1) overnight. A Cy3-labeled secondary antibody was used for staining the cells. DAPI (1 μg/mL, D8417, Sigma-Aldrich) was used for nuclear detection.
For the mammary tissue of mice and sheep, 3 μm sections of the tissue were permeabilized with 0.3% Triton for 10 min at RT (Beyotime, China). Primary antibody (ZO-1) was incubated following 4% BSA blocking. Cy3-labeled secondary antibody was used for staining of the target antibodies. Slides were sealed with DAPI-containing anti-fluorescence quenching mounting medium. The observation of the slides for both cells and mammary tissues was imaged using a DMi8 Microsystems BmbH (Leica, Wetzlar, Germany) microscope.
Histological analysis
Pieces of tissue from mice and sheep in all experimental groups were fixed in buffered formalin phosphate (10%), followed by paraffin-embedding, sectioned, and stained with hematoxylin and eosin, and a phase-contrast microscope was used for capturing (Nikon, Tokyo, Japan).
Mice lethality evaluation
The rates of mortality were evaluated over a duration of 24 h in all groups defined in the present study. The time-points were selected at 8 h, 12 h, 18 h, and 24 h post-injection of E.coli.
Bacterial load quantification in mammary tissue
Approximately 100 mg of mammary tissue was homogenized in sterile PBS, serially diluted, and plated on MHB agar. Colonies were enumerated after 24 h incubation at 37°C and expressed as CFU/g of tissue.
Terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling (TUNEL) staining
Apoptotic cells in sections of sheep mammary tissue were detected using the TUNEL In Situ Apoptosis Kit (HRP-DAB, E-CK-A331, Elabscience, China). Following deparaffinization and proteinase K treatment, the sections were incubated with the TUNEL reaction mixture at 37 °C for 2 hours. Streptavidin-HRP was applied to label the fragmented DNA ends. DAB chromogenic development was performed, and the sections were observed under a light microscope. The cell nuclei were counterstained with hematoxylin, followed by dehydration, transparency treatment, and mounting. Apoptotic cells were identified by the presence of brown signals (DAB) in the nuclei, and the extent of apoptosis was assessed in combination with morphological analysis.
Plasmid construction, lentivirus production, and transduction
For silencing of Prkaa1 expression, pGLVH1/GFP/Puro vector containing shRNA targeting Prkaa1 (PRKAA1-Bos-1071, PRKAA1-Bos-1275, PRKAA1-Bos-1460, and PRKAA1-Bos-1536) and negative control (sh-NC) were commercially obtained from GenePharma. The relative mRNA expression of Prkaa1 in MAC-T cells was assessed to evaluate the silencing efficiency of the constructed plasmids by transfection using Lipofectamine 2000 (Invitrogen) (S5 Fig). Sequences of shRNAs and primers used for the efficiency test are listed in S1 Table.
HEK293T cells were co-transfected with pGLVH1-PRKAA1-GFP-Puro, pCMV-VSV-G, and pCAG-deltaR8.9 packaging plasmids using Lipofectamine 2000 (Invitrogen). Supernatants were collected after 48 h of transfection for the infection of cells in the presence of 5 μg/mL polybrene (H8761, Solarbio). After 24 hours, cells were treated with puromycin (2 μg/mL, P8230, Solarbio) and selected as positive accordingly.
Transcriptomic analysis
After RNA extraction from treated bacteria, bMEC, and sheep mammary tissue, the RNA was prepared and sent to Majorbio Biotechnology Co., Ltd. for transcriptomic sequencing using Illumina HiSeq × TEN (2 × 150 bp read length). The data generated from the Illumina platform were used for bioinformatics analysis. All of the analyses were performed using the free online platform of Majorbio Cloud Platform (www.majorbio.com) from Shanghai Majorbio Bio-pharm Technology Co., Ltd.
Chromatin immunoprecipitation analysis
Experiment was performed according to the protocols described previously [28]. Briefly, mammary tissues were crushed in liquid nitrogen. Approximately 200 μg of powdered tissue was homogenized in PBS supplemented with protease inhibitor cocktail (Roche, #11697498001). Proteins were cross-linked with 1% formaldehyde for 10 min at room temperature, and the reaction was quenched with 2.5 M glycine. After centrifugation (4°C, 4000 × g, 5 min), the pellets were washed with PBS and resuspended in SDS lysis buffer with protease inhibitors. Chromatin was sonicated on ice to obtain fragments ranging from 200 to 500 bp and precleared with protein G agarose beads (Santa Cruz Biotechnology, sc-2003). Precleared chromatin was immunoprecipitated using 4 μg of anti-P65 antibody (Abcam, ab16502) at 4°C overnight, with normal rabbit IgG used as a negative control. Immunocomplexes were captured with protein A/G agarose beads. After reversal of cross-links in elution buffer with 1% SDS at 65°C for 1 h, immunoprecipitated DNA was quantified by qPCR using primers specific to promoter regions of Tnf, Il6, and Tlr4 (S1 Table).
Chromatin compaction assay
Chromatin accessibility was assessed using a real-time PCR-based assay following established methods with minor adjustments [29]. Briefly, nuclei isolated from mammary tissue were digested with 20 U of SacI, SspI, and MspI at 37°C for 2 h to target the Tnf, Il6, and Tlr4 promoter, respectively (Fig 9C). DNA was subsequently purified and quantified. Undigested samples served as input controls. The relative amount of uncut DNA was determined by qPCR using 75 ng of template, SYBR Premix EX Taq (Takara, DRR420A), and an ABI system. Primer sequences are provided in S1 Table.
Statistics
Statistical analyses were performed with GraphPad Prism 8 (GraphPad, USA). All statistical analyses were based on independent biological replicates, with group size (n) defined accordingly. All collected data points were included without exclusion. Bacterial loads in tissues and bacterial exponential assay were Log10-transformed prior to statistical analysis. Normality was assessed using the Shapiro-Wilk test, and homogeneity of variances was evaluated with Bartlett’s test. For data that did not meet parametric assumptions, equivalent non-parametric tests (Kruskal-Wallis with Dunn’s post hoc) were applied. Data are presented as mean ± standard error of the mean (SEM). Differences among multiple groups were analyzed by one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test for pairwise comparisons. For transcriptomic data, differential expression analysis was performed using DESeq2. Significantly differentially expressed genes were identified using an adjusted P-value (Benjamini-Hochberg false discovery rate, FDR) threshold of < 0.05. A P value of less than 0.05 was considered statistically significant. In figures, bars labeled with different lowercase letters (e.g., a, b, c) indicate statistically significant differences (P < 0.05) between groups. Bars sharing the same letter are not significantly different.
Results
Identification of a predominant, multidrug-resistant E. coli phylogroup B1 as the primary target
To establish a clinically relevant model, we first characterized the epidemiological landscape of mastitis pathogens in our region. Bacterial isolation from 186 clinical mastitis milk samples yielded 306 isolates, with a detection rate of 88.70%. Among 256 isolates of major mastitis pathogens, Escherichia coli (26.17%) and Klebsiella spp. (20.81%) were predominant (Fig 1A). Phylogenetic grouping of the 78 E. coli isolates from bovine mastitis in Jiangsu was conducted using the phylogroup identification proposed by Clermont et al. [30]. Phylogenetic analysis of the 78 E. coli isolates revealed that phylogroups B1 (52.5%) and A (37.2%) were most prevalent, while B2 (1.3%) and D (9.0%) were less common (Fig 1B). Antibiotic resistance profiling showed that resistance to tetracycline and gentamicin was mainly associated with phylogroup D. Resistance to streptomycin was predominant in phylogroup B1, and resistance to ciprofloxacin was primarily found in groups B2 and D. Resistance to β-lactams (lincomycin, penicillin, amoxicillin, cefalotin) was widespread across groups B1, B2, and D (Fig 1C). Consistent with phenotypic resistance, analysis of 10 major resistance genes revealed that phylogroup B1 carried the strongest and most prevalent resistance gene profile, including the highest frequencies of blaCTX-M, blaSHV, tetC, qnrS, and oqxA. In contrast, groups A and D exhibited generally lower resistance gene carriage, and group B2 carried the fewest determinants overall (Fig 1D).
(A) Distribution of bacterial species isolated from clinical mastitis milk samples. (B) The phylogenetic group distribution (A, B1, B2, D) among the 78 isolated E. coli strains, with group B1 being the most prevalent. (C) Antibiotic resistance profiles for the E. coli phylogroups against eight different antibiotics. (D) The detection rates of major antibiotic resistance genes (e.g., blaTEM, tetA, oqxA) across the E. coli phylogroups.
These findings indicated that phylogroup B1 was not only the most prevalent but also carried the most concerning resistance profile among clinical isolates. Therefore, we selected a representative B1 strain as our primary model to investigate therapeutic strategies against this dominant, resistant pathogen.
Metformin directly impairs E. coli B1 virulence by inhibiting growth and disrupting biofilm formation
We next assessed whether metformin could directly compromise key virulence traits of the predominant, MDR E. coli B1, focusing on traits critical for persistence and antibiotic resistance. We first determined the basic pharmacodynamic profile of metformin against E. coli B1. The minimal inhibitory concentration (MIC) and minimal bactericidal concentration (MBC) were 8 mg/mL and 16 mg/mL, respectively. Time-kill assays demonstrated that metformin exerted a concentration-dependent bactericidal effect, particularly during the exponential growth phase (Fig 2A and 2B). To evaluate the risk of resistance development, a critical consideration for any potential therapeutic, we performed a serial passage experiment. Notably, a discernible increase in MIC (2 × MIC) emerged only after 15 days of continuous exposure (Fig 2C), indicating a low propensity for rapid resistance selection under monotherapy. Given the high β-lactam resistance in our isolates, we investigated whether metformin could act as an antibiotic adjuvant to restore efficacy. Checkerboard assays revealed potent synergistic effects (FICI ≤ 0.5) between metformin (at a sub-inhibitory concentration of 0.5 × MIC) and several antibiotics, including penicillin, cefoxitin, and lincomycin, reducing their MICs by 16- to 32-fold (S2A Fig). This finding supports the potential of metformin as a combination therapy enhancer against resistant infections.
(A) Bacterial growth curves of E. coli B1, MRSA, and K. pneumoniae treated with a range of metformin concentrations (0, 0.5, 1, 2, 4 × MIC) (n = 3). (B) The bactericidal effect of metformin on E. coli B1 viability during the exponential growth phase (4h and 8h), determined by colony counting (n = 3). (C) The development of metformin resistance in E. coli B1 over a 26-day serial passage experiment. (D) Quantification of biofilm biomass by crystal violet staining (n = 3). (E) The number of viable bacteria (CFU) within pre-formed biofilms after treatment with metformin. (F) Microscopic analysis of biofilms: (a) Light microscopy images; (b) Visual appearance of crystal violet-stained biofilms; (c) LIVE/DEAD BacLight staining (green: live bacteria, red: dead bacteria) for bacteria in pre-formed biofilm; (d) Scanning Electron Microscope (SEM) images showing the structural disintegration of E. coli in biofilms following metformin treatment. Results are expressed as means ± SEM (A, B, D). Bars with different lowercase letters indicate statistically significant differences (p < 0.05, one-way ANOVA with Tukey’s test).
Since biofilm formation is a major mechanism of antibiotic tolerance and chronic infection in mastitis, we focused on metformin’s impact on this critical virulence phenotype. Crystal violet quantification showed that metformin inhibited E. coli B1 biofilm formation in a dose-dependent manner, with ~70% inhibition achieved at 2 × MIC (Fig 2D). More importantly, metformin not only prevented biofilm formation but also significantly reduced the viability of bacteria embedded within pre-established mature biofilms (Fig 2E). To visualize these effects, we employed complementary microscopy techniques. Light and scanning electron microscopy revealed that metformin treatment fragmented the interconnected biofilm architecture and reduced bacterial adhesion (Fig 2F-a, 2F-b, and 2F-d). Live/Dead staining further confirmed a shift from predominantly live (green) to dead (red) bacteria within the biofilm matrix upon treatment, with near-complete abolition of biofilm at 2 × MIC (Fig 2F-c). Collectively, these results demonstrate that metformin possesses direct antibacterial activity against E. coli B1, synergizes with conventional antibiotics, and potently disrupts the biofilm lifestyle that underpins treatment failure. To elucidate the molecular basis for these phenotypic effects, we next conducted a transcriptomic analysis.
Transcriptomic analysis reveals that metformin compromises bacterial membrane integrity and metabolic pathways
To elucidate the molecular basis for metformin’s anti-biofilm and antibacterial effects, we performed transcriptomic analysis on metformin-treated E. coli B1. Treatment with metformin resulted in the upregulation of 1,095 differentially expressed genes (DEGs) and the downregulation of 1,148 DEGs (Fig 3A). GO annotation analysis revealed that the DEGs were primarily associated with transmembrane transporter activity, transferase activity, and ATP binding in the molecular function category, as well as with the integral component of membrane and integral component of plasma membrane in the cellular component category (Fig 3B). KEGG pathway analysis indicated that among the DEGs, 21 were related to drug resistance and 109 were associated with membrane transport, both categories showing predominant downregulation. In addition, 141 genes involved in carbohydrate metabolism and 75 genes involved in energy metabolism were downregulated (Fig 3C). Notably, biofilm formation in bacteria relies on the biosynthesis and secretion of glycoproteins and phospholipids. We observed significantly more downregulated than upregulated genes in pathways related to amino acid metabolism, glycan biosynthesis, and lipid metabolism. This pattern correlates with the marked reduction in bacterial biofilm formation capacity induced by metformin. Finally, we validated the expression of relevant genes following metformin treatment using qPCR, and the results were consistent with the transcriptome data (Fig 3D). These data indicate that metformin disrupts E. coli B1 by dysregulating genes critical for outer membrane integrity, transport, and energy metabolism, providing a mechanistic basis for its observed ability to inhibit growth and biofilm formation. Having established metformin’s direct antibacterial actions, we then wondered whether it could also protect the host by modulating the damaging inflammatory response to infection.
E.coli B1 was grown to the exponential phase (6h), then exposed to 8 mg/mL metformin for 4 h. (A) volcano plot of differentially expressed genes (DEGs), with a number of 1095 genes upregulated and 1148 genes downregulated. (B) GO enrichment analysis of DEGs, with significant terms related to membrane components and transporter activity. (C) KEGG pathway annotations analysis, showing significant downregulation in pathways including membrane transport and carbohydrate metabolism. (D) qPCR validation of selected genes related to membrane integrity, metabolism, and stress responses (n = 3). The colors from dark to light indicate the downregulation or upregulation of relative gene expression levels.
Metformin suppresses LPS-induced inflammation in bMECs via the AMPK/SIRT1/NF-κB axis
We employed an in vitro model of LPS-challenged bovine mammary epithelial cells (bMECs) to dissect metformin’s impact on the host inflammatory response. We employed transcriptome sequencing to analyze the gene expression profiles associated with metformin’s intervention in LPS-induced inflammatory responses and identified DEGs (Fig 4A). As shown in Fig 4B, overlapped DEGs (108 genes) from the LPS vs. NC group and the LM vs. LPS group were considered as co-regulated DEGs and subjected to GO annotation and KEGG enrichment analysis. GO annotation results revealed (Fig 4C) that the three groups co-regulated biological processes such as inflammatory response, cellular response to lipopolysaccharide, and defense response to symbiont; cellular components including CXCR chemokine receptor binding; and molecular functions related to ribosome-associated biological activities. KEGG enrichment analysis indicated that, in addition to ribosome-related functions being affected, the major signaling pathways involved were the TNF signaling pathway and the NF-κB signaling pathway (Fig 4D). We further validated selected pro-inflammatory factors using qPCR, and the results were consistent with the RNA-seq findings (S3A and S3B Fig). Since metformin is a known activator of AMP-activated protein kinase (AMPK), a key metabolic regulator with anti-inflammatory roles, we investigated whether this pathway mediated the observed effects. Metformin is a classic activator of the AMPK signaling pathway. Consistent with our previous findings in bMECs, we have demonstrated that metformin effectively activates the AMPK pathway by enhancing its phosphorylation level [17]. Therefore, we further investigated the role of the AMPK signaling pathway in mediating downstream inflammatory responses by establishing an AMPK knockdown system (S4A–S4D Fig). The results showed that knockdown of AMPK using shPRKAA1 suppressed p-AMPK levels, while leading to upregulation of both p- and Ac-p65. Similarly, AMPK knockdown exacerbated the LPS-induced expression of p- and ac-p65, and under these conditions, metformin failed to inhibit p65 activation. Furthermore, AMPK knockdown reduced the activity of the deacetylase SIRT1, suggesting that SIRT1 may function downstream of AMPK and regulate protein or gene expression through deacetylation (Fig 5A and 5B). Since SIRT1 activity is regulated by NAD⁺ levels, we measured intracellular NAD⁺ content and the NAD ⁺ /NADH ratio. As expected, AMPK knockdown significantly reduced both parameters. Interestingly, metformin treatment partially restored the NAD⁺ content and NAD ⁺ /NADH ratio even in AMPK-knockdown cells (Fig 5C).
(A) PCA analysis. (B) Heatmap of DEGs. (C) GO term enrichment analysis of the co-regulated DEGs among NC, LPS, and LPS + Met groups. (D) Co-regulated KEGG pathway analysis.
(A-B) Western blot of p-AMPK, SIRT1, p-p65, and Ac-p65 after AMPK knockdown by GFP-labeled pGLVH1-PRKAA1-GFP-Puro construct (n = 4). (C) The intracellular NAD⁺ levels and NAD ⁺ /NADH ratio (n = 3). (D) The mRNA expression levels of inflammatory genes after pharmacological activation (SRT1720) or inhibition (EX527) of SIRT1 (n = 3). (E-F) Protein levels of p-AMPK, SIRT1, and p-p65 after SIRT1 agonist/antagonist treatment (n = 4). (G-H) Acetylation levels of p65 (K310) and global Pan-acetylation (n = 4). Results are expressed as means ± SEM (B-D and F-H). Bars with different lowercase letters indicate statistically significant differences (p < 0.05, one-way ANOVA with Tukey’s test).
To directly test SIRT1’s role, we used pharmacological tools. The SIRT1 agonist SRT1720 (2.5 μM) mimicked metformin’s effect, potently suppressing mRNA expression of Il6, Tnf, and Cxcl8 induced by LPS. Conversely, the inhibitor EX527 (2.5 μM) significantly attenuated metformin’s anti-inflammatory efficacy (Fig 5D). At the protein level, SIRT1 activation decreased, while its inhibition increased, the acetylation of p65 at K310 and global pan-acetylation (Fig 5G and 5H). Notably, EX527 also diminished metformin-induced AMPK phosphorylation (Fig 5E and 5F), hinting at a potential positive feedback loop. Furthermore, SIRT1 modulation directly influenced cellular redox state, as evidenced by corresponding changes in ROS levels (S6A Fig). Immunofluorescence confirmed that SIRT1 activation correlated with reduced nuclear Ac-p65 signal (S6B and S6C Fig). Collectively, these data delineate a defined signaling cascade in bMECs: metformin activates AMPK, which enhances SIRT1 expression and activity, leading to the deacetylation and functional inhibition of NF-κB p65, thereby resolving the inflammatory response.
Metformin confers dual therapeutic benefits in murine models of systemic and mammary gland infection
To evaluate therapeutic potential in vivo, we employed a tiered strategy. First, a systemic intraperitoneal infection model demonstrated that metformin improved survival and activated the AMPK/NF-κB axis in the liver (S7 Fig), confirming the operation of this core anti-inflammatory axis in vivo despite tissue-specific nuances in SIRT1 regulation. To directly evaluate therapeutic efficacy in the target organ, we next employed a lactating mouse mammary gland infection model.
To directly model mastitis and assess local efficacy, we then used a lactating mouse mammary gland infection model. Metformin treatment post-infection significantly reduced the bacterial load within mammary tissue (Fig 6B), demonstrating direct antibacterial activity in vivo. This direct antibacterial effect in vivo was accompanied by a marked alleviation of histopathological damage, a reduction in infection-driven systemic leukocytosis and neutrophilia (Fig 6A and 6C), and suppression of local pro-inflammatory gene expression (e.g., Il1b, Tnf, Cxcl1) (Fig 6D). At the molecular level, analysis of the mammary NAD⁺ level revealed that while the absolute NAD⁺ content showed a decreasing trend in the EC group (p = 0.07), the functional NAD ⁺ /NADH ratio was significantly reduced. Metformin treatment restored both parameters (Fig 6E) and concurrently enhanced AMPK phosphorylation while decreasing the acetylation of p65 (Fig 6F). Thus, in the target organ, metformin concurrently executes pathogen-directed (bactericidal) and host-directed (immunomodulatory via AMPK/SIRT1) actions, leading to improved outcomes.
(A) Representative H&E staining and immunofluorescence for tight-junction proteins (ZO-1 and Occludin) of mammary gland tissue from lactating mice, showing reduced inflammatory damage after metformin treatment (n = 4). (B) Bacterial load quantification in murine mammary tissues (n = 8). (C) Complete blood count (CBC) parameters, including white blood cell (WBC), lymphocyte (Lym), monocytes (Mon), and neutrophil (Neu) counts (n = 8). (D) The relative mRNA expression of pro-inflammatory genes in mammary tissue, inhibited by metformin treatment (n = 4). (E) The NAD⁺ levels and NAD⁺/NADH ratio in murine mammary tissues (n = 8). (F) Protein levels of SIRT1, p-AMPK, and Ac-p65 in murine mammary tissues (n = 4). Results are expressed as means ± SEM (B-F). Bars with different lowercase letters indicate statistically significant differences (p < 0.05, one-way ANOVA with Tukey’s test).
Metformin confers therapeutic efficacy and epigenetically silences inflammation in a ruminant mastitis model
To definitively evaluate the translational potential of metformin, we established an E. coli mastitis model in lactating Hu sheep—a physiologically relevant dairy species. Intramammary challenge induced acute clinical mastitis, characterized by elevated body temperature and leukocytosis. Post-infection treatment with metformin (100 mg/kg, intramammary) significantly alleviated these clinical signs (S8A and S8B Fig).
Crucially, metformin exerted a direct antibacterial effect in vivo, significantly reducing the bacterial load in mammary tissue compared to the infected-only group (Fig 7B). This reduction in pathogen burden was accompanied by a mitigation of the systemic inflammatory response, as evidenced by decreased plasma levels of the acute-phase protein SAA and the oxidative stress marker MDA (Fig 7C). Histopathological analysis demonstrated that metformin preserved mammary alveolar architecture, reduced inflammatory cell infiltration, and diminished epithelial apoptosis (TUNEL staining) (Fig 7D). Furthermore, immunofluorescence revealed that metformin treatment restored the expression and continuity of the tight junction proteins ZO-1 and Occludin, indicating enhanced epithelial barrier integrity (Fig 7D). At the molecular level, metformin activated the core immunometabolic axis identified in vitro: Western blot analysis of mammary tissue showed that metformin enhanced AMPK phosphorylation, increased SIRT1 expression, and concurrently suppressed the activation of NF-κB (p65) and its downstream target IL-6 (Fig 7E).
(A) The experimental timeline for intramammary infection and metformin treatment in sheep. Icons were obtained from the open-source repositories (https://openclipart.org/). (B) The concentration of E. coli in the lactating sheep mammary gland was determined by plate coating (n = 5). (C) Plasma levels of SAA, CRP, and MDA (n = 5). (D) Histopathological and cellular analysis of mammary tissue: H&E staining, TUNEL staining, and immunofluorescence (tight junction integrity) (n = 5). (E) Comparison of AMPK, p65, and IL-6 protein expression in mammary tissue (n = 5). Results are expressed as means ± SEM. Bars with different lowercase letters indicate statistically significant differences (p < 0.05, one-way ANOVA with Tukey’s test).
To obtain a comprehensive view of metformin’s impact on the host response, we performed transcriptomic analysis. RNA-seq of mammary tissue from metformin-treated sheep revealed a broad repression of infection-induced inflammatory pathways. Differentially expressed genes were significantly enriched in “Cytokine-cytokine receptor interaction,” “IL-17 signaling pathway,” and “NF-κB signaling pathway” (Fig 8B–8D). qPCR validation confirmed the downregulation of key pro-inflammatory mediators, including Il1b, Cxcl8, Cxcl6, Tlr4, Nfkb1, Tnf, Nlrp3, and Tlr2 (Fig 8E). This confirms that metformin reprograms the host transcriptional landscape toward resolution.
(A) PCA analysis. (B) Volcano plot of differentially expressed genes (DEGs) between the E. coli-challenged (EC) and metformin-treated (EM) groups. (C) GO annotations analysis of DEGs, with significant terms in biological processes such as response to stimulus. (D) KEGG pathway analysis showing significant enrichment in immune and inflammatory pathways, including cytokine-cytokine receptor interaction and IL-17 signaling pathway. (E) qPCR validation of key DEGs involved in inflammation (Il1b, Tnf), chemotaxis (Cxcl8), and immune signaling (Tlr4, Nfkb1, Nlrp3). Results are expressed as means ± SEM (n = 3). Bars with different lowercase letters indicate statistically significant differences (p < 0.05, one-way ANOVA with Tukey’s test).
We next dissected the epigenetic mechanism by which the AMPK/SIRT1 axis enforces this sustained anti-inflammatory state. Metformin restored the infection-depleted NAD⁺ pool in mammary tissue (Fig 9A), which correlated with enhanced SIRT1 deacetylase activity. This was evidenced by a global reduction in protein acetylation and specific deacetylation of NF-κB p65 at Lys310 and histone H3 at Lys14 (Fig 9B). To link deacetylation to transcriptional silencing, we examined chromatin dynamics at specific gene promoters. Putative binding sites for NF-κB were searched for with the JASPAR Program (https://jaspar.elixir.no/search?q=&collection=CORE&tax_group=vertebrates). Filters were set ≥ 0.90 as a threshold for similarity of the core sequence (Fig 9C). Chromatin immunoprecipitation (ChIP) assays showed that metformin significantly reduced NF-κB p65 binding to the promoters of Tnf, Il6, and Tlr4 (Fig 9D). This decrease in transcription factor occupancy was mechanistically explained by increased chromatin compaction at these loci, as measured by a nuclease accessibility assay (Fig 9E). A strong inverse correlation (r = -0.77, p < 0.01) was observed between NF-κB binding abundance and chromatin compaction across all samples (Fig 9F). Collectively, these results demonstrate that in a translational ruminant model, metformin alleviates E. coli mastitis through a dual mechanism: directly reducing bacterial burden and reprogramming the host immunometabolic response. The host-directed action is mediated by the AMPK/SIRT1 axis, which enforces resolution of inflammation through epigenetic chromatin remodeling and silencing of NF-κB-driven gene transcription.
(A) NAD⁺ levels in mammary tissue, which were decreased by E. coli infection and restored by metformin treatment. (B) Western blot analysis showing that metformin reduces the infection-induced acetylation of p65, histone H3 (H3K14), and pan-acetylated proteins. (C) Schematic diagram of the promoter regions of the Tnf, Il6, and Tlr4 genes, indicating NF-κB binding sites and restriction enzyme sites used for chromatin accessibility assay. (D) Chromatin Immunoprecipitation (ChIP)-qPCR analysis of NF-κB p65 binding to the promoters of pro-inflammatory genes (Tnf, Il6, and Tlr4). (E) chromatin compaction assay measuring the accessibility of promoter regions. (F) Correlation analysis reveals a significant negative correlation between NF-κB binding abundance and chromatin compaction levels. Coefficients of correlation between the degree of chromatin compaction and percentage of NF-κB binding were analyzed using Pearson Correlations. Results are expressed as means ± SEM (B, D and E) (n = 5). Bars with different lowercase letters indicate statistically significant differences (p < 0.05, one-way ANOVA with Tukey’s test).
Discussion
Infectious diseases triggered by multidrug-resistant (MDR) pathogens like Escherichia coli represent a critical challenge to both animal and human health, necessitating innovative therapeutic strategies that extend beyond direct bacterial killing. This study repurposes metformin, a well-known metabolic modulator, as a dual-function agent against E. coli-induced bovine mastitis. We demonstrate that metformin simultaneously disrupts key bacterial virulence programs and reprograms host immunometabolism to resolve excessive inflammation. This dual-targeting strategy is mechanistically grounded in metformin’s ability to disperse bacterial biofilms and, through the AMPK/SIRT1 axis, enforce epigenetic silencing of NF-κB-driven inflammatory genes. Our findings not only provide a novel non-antibiotic approach to combat MDR infections but also exemplify how modulating host immunometabolism can be harnessed for therapeutic benefit against infectious diseases. Our epidemiological surveillance, consistent with reports from other regions [31–34], identified phylogroup B1 as the predominant and most multidrug-resistant E. coli lineage in clinical bovine mastitis cases. The high prevalence of biofilm-associated virulence factors in this group underscores its persistence (S1 Fig). This clinical prevalence prompted us to investigate the mechanisms underlying both its persistence and its potential susceptibility to metformin. A key virulence factor associated with antibiotic resistance in Gram-negative bacteria is the outer membrane protein (OMP) [35]. In the current isolates, virulence gene profiling revealed a high prevalence of genes encoding OMPs such as OmpC and OmpF, particularly in phylogroup B1. Notably, while loss-of-function mutations in porins like OmpC and OmpF can confer resistance by limiting antibiotic influx [36,37], their presence in B1 isolates suggests an alternative resistance strategy, potentially linked to enhanced biofilm formation which is known to be influenced by OMPs [38]. Metformin directly targeted this resilience. It not only inhibited planktonic growth but, more importantly, potently disrupted biofilm formation and reduced viability within pre-formed biofilms. Transcriptomic analysis of metformin-treated E. coli B1 provided mechanistic insight, revealing a significant downregulation of genes involved in membrane transport, integrity, and central metabolism. The downregulation of OMP-related genes (e.g., ompC, ompW) and glycogen metabolism genes (e.g., glgS) is particularly relevant, as these pathways are critical for biofilm architecture and stability [39,40]. Furthermore, the upregulation of envelope stress response genes indicates that metformin inflicts substantial damage on the bacterial membrane, triggering a protective stress response [41,42]. This indicates that metformin’s primary antibacterial action against E. coli B1 is mediated by compromising membrane homeostasis, leading to biofilm dispersion and bacterial death. This mechanism of action—destabilizing the bacterial membrane and biofilms—also provides a rational basis for the observed synergistic effect between metformin and conventional antibiotics (such as β-lactams) against this resistant strain (S2A Fig). By disrupting these key defensive structures, metformin likely facilitates antibiotic penetration and efficacy. Thus, the primary anti-virulence effect of metformin, centered on biofilm disruption and membrane stress, not only stands as a direct therapeutic strategy but also enhances its potential utility as an adjuvant in combination therapies.
Beyond its antibacterial properties, metformin exhibits anti-inflammatory effects, necessitating further exploration of molecular interactions to elucidate underlying mechanisms and advance therapeutic strategies for inflammatory diseases, such as bovine mastitis (S10 Fig). This study elucidates that metformin’s potent anti-inflammatory action in bovine mastitis is orchestrated through the AMPK/SIRT1/NF-κB signaling axis. This pathway integrates metabolic sensing with immune regulation, a concept central to immunometabolism. In bovine mammary epithelial cells (bMECs), metformin’s activation of AMPK was indispensable for suppressing the LPS-induced inflammatory cascade, a finding consistent with its established role as an AMPK activator [43,44] and with our previous work [18,45]. Genetic ablation of AMPK not only abolished metformin’s ability to inhibit NF-κB activation but also revealed a critical link to SIRT1, as AMPK knockdown reduced SIRT1 expression and activity. The relationship between AMPK and the NAD+-dependent deacetylase SIRT1 revealed an intriguing complexity. An interesting divergence was observed in the regulation of NAD+ homeostasis. Although AMPK is a established upstream activator of SIRT1 [46], likely through modulating NAD+ levels, our data indicated that metformin could partially restore NAD+ even in AMPK-knockdown cells. This points to the existence of AMPK-independent mechanisms by which metformin influences NAD+ metabolism, possibly via its direct action on mitochondrial complex I and subsequent alterations in cellular redox state [47]. However, the complete abolition of metformin’s effect on p65 acetylation and inflammation upon AMPK loss underscores a non-redundant, critical role for AMPK. It appears that AMPK is not only a mediator of NAD+ flux but is essential for functionally linking the improved NAD+ landscape to the specific deacetylation of pro-inflammatory targets by SIRT1. This suggests a cooperative model where metformin primes the metabolic environment (increasing NAD+ via multiple routes) while AMPK acts as a required signal transducer to harness this potential for targeted anti-inflammatory action. This nuanced interplay suggests a cooperative model wherein metformin employs both AMPK-independent (e.g., via mitochondrial complex I inhibition) and AMPK-dependent pathways to optimize the cellular condition for SIRT1-mediated, target-specific deacetylation of inflammatory mediators like NF-κB p65. This mechanism aligns with broader observations that metformin exerts anti-inflammatory effects via AMPK and SIRT1 in various contexts [48–50], and specifically that SIRT1 deacetylates p65 at K310 to inhibit NF-κB transcriptional activity [51,52].
The physiological and translational relevance of this immunometabolic pathway was robustly confirmed through a tiered in vivo strategy. We first employed a systemic murine infection model, which established that metformin could improve survival and modulate the AMPK/NF-κB axis in a major metabolic organ (liver), providing proof-of-concept for its systemic bioactivity (S7 Fig). To directly model mastitis pathology and assess local therapeutic efficacy, we then utilized lactating mouse and Hu sheep mammary gland infection models. This study exemplifies how repurposing a metabolic modulator like metformin can recalibrate dysregulated host immunometabolism during bacterial infection, offering a novel host-directed therapeutic strategy beyond direct antimicrobial action. In both models, metformin treatment recapitulated the dual benefits observed in vitro: it significantly reduced bacterial loads in mammary tissue, confirming direct antibacterial activity in vivo, and concurrently attenuated infection-driven inflammation and tissue damage. Importantly, a physiologically relevant ruminant model of lactating Hu sheep uncovered the mechanism reached its definitive epigenetic endpoint. Metformin restored infection-depleted mammary NAD+ levels, enhanced SIRT1 activity, and reduced acetylation of NF-κB p65 and histone H3 (H3K14). Most notably, we demonstrated that this deacetylation led to increased chromatin compaction at the promoters of key pro-inflammatory genes (Tnf, Il6, Tlr4), thereby reducing NF-κB binding and silencing their transcription. This chromatin-mediated repression provides a stable molecular explanation for the sustained resolution of inflammation, extending the mechanism beyond cytoplasmic signaling inhibition to nuclear transcriptional reprogramming. This aligns with emerging concepts that metabolic interventions can influence inflammatory gene expression via epigenetic modifications in the context of infection [53–55].
Metformin’s dual pathogen- and host-directed actions define a synergistic therapeutic strategy. It directly reduces bacterial burden to lessen the immune stimulus while recalibrating the host response through the AMPK/SIRT1 axis, thereby preventing the excessive inflammation characteristic of severe mastitis. This two-pronged approach effectively combines direct anti-virulence effects (e.g., biofilm disruption) with host-directed immunomodulation—a strategy of particular value in the antimicrobial resistance era due to its reduced potential for driving pathogen resistance. Notably, key effects such as biofilm disruption and antibiotic synergy were achieved at sub-inhibitory in vitro concentrations (0.5 × MIC). Although the planktonic MIC is high (8 mg/mL), the standard clinical practice of local intramammary infusion enables effective local delivery. Future formulations (e.g., nanoparticles) could further optimize bioavailability.
Taken together, our work establishes the repurposing of metformin as a prototype for a dual-targeting therapeutic paradigm against MDR bacterial mastitis. Beyond elucidating a detailed mechanistic basis, our findings underscore the therapeutic potential of modulating host immunometabolism as a powerful complement to direct antimicrobials. Refining this strategy and exploring analogous immunometabolic pathways may yield novel, resistance-evading interventions in the ongoing fight against AMR.
Supporting information
S1 Table. Primers and relevant sequences used in this study.
https://doi.org/10.1371/journal.ppat.1014012.s001
(XLSX)
S1 Fig. The prevalence of virulence genes in the E. coli isolates.
Among the 78 mastitis-derived E. coli isolates, a total of 8 virulence genes were detected. Four of these genes—ompC, fimH, ECs3703, and ompF—exhibited high detection rates, accounting for 98.7%, 94.9%, 92.3%, and 82.1% of the isolates, respectively. The detection rates for the remaining genes were as follows: irp2 (32.1%), fyuA (32.1%), iucD (24.4%), and F17A (1.3%). Fourteen other virulence genes—papC, cnf1, cnf2, afaD-8, afaE-8, stx1, stx2, sfaD, LT1, STb, colV, HlyA, eaeA, and ler—were not detected.
https://doi.org/10.1371/journal.ppat.1014012.s002
(TIF)
S2 Fig. The combination effect of metformin with antibiotics and the RNA-Seq correlations.
(A) Checkerboard assay demonstrating the synergistic effect between metformin and antibiotics (Cefoxitin, Lincomycin, Penicillin), quantified by the Fractional Inhibitory Concentration Index (FICI). (B) Correlations of differentially expressed genes (DEGs) following metformin treatment. (C) Heatmap of DEGs demonstrating distinct clustering between control and metformin-treated groups.
https://doi.org/10.1371/journal.ppat.1014012.s003
(TIF)
S3 Fig. qPCR validation of pro-inflammatory genes in bovine mammary epithelial cells.
(A) Genes expression obtained from RNA-Seq dataset. (B) Validation of RNA-Seq by qPCR analysis. The consistent trend between qPCR and RNA-Seq data indicates that the RNA-Seq results are reliable.Results are expressed as means ± SEM (n = 3). Bars with different lowercase letters indicate statistically significant differences (p < 0.05, one-way ANOVA with Tukey’s test).
https://doi.org/10.1371/journal.ppat.1014012.s004
(TIF)
S4 Fig. The construction of the gene knockout system for Prkaa1.
(A) 48 hours after co-transfection with the three-plasmid system (pGLVH1-PRKAA1-GFP-Puro, pCMV-VSV-G, and pCAG-deltaR8.9), observation under a fluorescence microscope revealed that nearly all 293T cells emitted green fluorescence. (B) Fluorescence intensity of plasmids. (C) Fluorescence intensity of different MOI values of lentivirus-infected target cells. (D) Selection of knockout target sequences from four designed candidates.
https://doi.org/10.1371/journal.ppat.1014012.s005
(TIF)
S5 Fig. The effect of SIRT1 agonist or antagonist on bovine mammary epithelial cell viability.
(A) Viability of cells treated with SIRT1 agonist (SRT1720) tested by CCK-8 assay. (B) Viability of cells treated with SIRT1 antagonist (EX527) tested by CCK-8 assay. Results are expressed as means ± SEM (n = 3). Bars with different lowercase letters indicate statistically significant differences (p < 0.05, one-way ANOVA with Tukey’s test).
https://doi.org/10.1371/journal.ppat.1014012.s006
(TIF)
S6 Fig. ROS content and SIRT1 expression during SIRT1 activation or inhibition.
(A) ROS levels in cells during SIRT1 agonist or inhibition. LPS stimulation significantly increased cellular ROS levels, which were then reduced by metformin treatment. Subsequent treatment with a SIRT1 agonist further decreased ROS levels compared to metformin alone, whereas a SIRT1 inhibitor significantly increased them. Results are expressed as means ± SEM (n = 3). Bars with different lowercase letters indicate statistically significant differences (p < 0.05, one-way ANOVA with Tukey’s test). (B-C) SIRT1 and Ac-p65 expression by immunofluorescence during SIRT1 agonist or inhibition. Compared to the metformin treatment group, SIRT1 fluorescence intensity was significantly stronger following treatment with a SIRT1 agonist and significantly weaker with a SIRT1 inhibitor. On the contrary, Ac-p65 fluorescence intensity was stronger in LPS challenged cells as compare to control group, while the intensity induced by LPS were both downregulated by Metformin or Metformin with SRT treated cells. The activation of Ac-p65 was recovered by EX557 in condition of the metformin addition.
https://doi.org/10.1371/journal.ppat.1014012.s007
(TIF)
S7 Fig. Metformin confers therapeutic protection against E. coli B1 infection in murine hepatic models.
(A) Survival curve of mice challenged with E. coli and subsequently interfered by metformin at doses of 25, 50, and 75 mg-kg, respectively (n = 10). (B) Liver or spleen tissue images of the mice treated with E. coli B1 or followed by metformin injection. (C) Representative H&E-stained sections of liver and spleen tissues, illustrating the attenuation of infection-induced hemorrhagic necrosis and inflammatory infiltration by metformin. (D) Western blot analysis of liver tissues confirming metformin-mediated activation of AMPK pathway and inhibition of NF-κB (p-p65) (n = 4). Results are expressed as means ± SEM. Bars with different lowercase letters indicate statistically significant differences (p < 0.05, one-way ANOVA with Tukey’s test).
https://doi.org/10.1371/journal.ppat.1014012.s008
(TIF)
S8 Fig. Metformin attenuates E. coli B1-induced alterations of body temperature and blood parameters in lactating Hu sheep.
(A) Changes in body temperature following infection and treatment. (B) Hematological parameters (WBC, Neu count), showing that metformin counteracts infection-induced leukocytosis. Results are expressed as means ± SEM (n = 5). Bars with different lowercase letters indicate statistically significant differences (p < 0.05, one-way ANOVA with Tukey’s test).
https://doi.org/10.1371/journal.ppat.1014012.s009
(TIF)
S9 Fig. Analysis of transcriptional profile in the mammary tissue of lactating sheep challenged by E. coli B1.
(A) Volcano plot of differentially expressed genes (DEGs) between the E. coli-challenged (EC) and control (NC) groups. (B) GO enrichment analysis of DEGs, with significant terms in biological processes such as response to stimulus. (C) KEGG pathway analysis showing significant enrichment in immune and inflammatory pathways. (D) Gene Enrichment Chord Diagram of DEGs.
https://doi.org/10.1371/journal.ppat.1014012.s010
(TIF)
S10 Fig. The observation of therapeutic effect of metformin on E. coli-induced mastitis.
(A) The experimental timeline for intramammary metformin treatment in mastitis-infected cows. Icons were obtained from the open-source repositories (https://openclipart.org/). (B) Clinical observation and evaluation of mammary appearance after treatment with metformin infusion and the clots changing during 3-day therapeutic. The redness of the mammary surface was recovered to health and the clots in the milk completely disappeared after 3-day treatment with metformin.
https://doi.org/10.1371/journal.ppat.1014012.s011
(TIF)
S11 Fig. Graphical abstract.
Created with the open-source repositories: https://bioart.niaid.nih.gov/ or https://bioicons.com/.
https://doi.org/10.1371/journal.ppat.1014012.s012
(TIF)
S1 Data. Excel spreadsheet containing, in separate sheets, the data points presented in Figs 1–9.
https://doi.org/10.1371/journal.ppat.1014012.s013
(XLSX)
References
- 1. Esener N, Green MJ, Emes RD, Jowett B, Davies PL, Bradley AJ, et al. Discrimination of contagious and environmental strains of Streptococcus uberis in dairy herds by means of mass spectrometry and machine-learning. Sci Rep. 2018;8(1):17517. pmid:30504894
- 2.
Yang Y, Peng Y, Jiang J, Gong Z, Shang S. Isolation and characterization of multidrug‐resistant Klebsiella pneumoniae from raw cow milk in Jiangsu and Shandong provinces, China. Transboundary and Emerging Diseases. 2020.
- 3. Yu ZN, Wang J, Ho H, Wang YT, Huang SN, Han RW. Prevalence and antimicrobial-resistance phenotypes and genotypes of Escherichia coli isolated from raw milk samples from mastitis cases in four regions of China. J Glob Antimicrob Resist. 2020;22:94–101. pmid:31887413
- 4. Sanchez-Vizuete P, Orgaz B, Aymerich S, Le Coq D, Briandet R. Pathogens protection against the action of disinfectants in multispecies biofilms. Front Microbiol. 2015;6:705. pmid:26236291
- 5.
Gędas A, Olszewska MA. Chapter 1 - Biofilm formation and resistance. In: Simoes M, Borges A, Chaves Simoes L, editors. Recent Trends in Biofilm Science and Technology. Academic Press; 2020. pp. 1–21.
- 6. Tyerman JG, Ponciano JM, Joyce P, Forney LJ, Harmon LJ. The evolution of antibiotic susceptibility and resistance during the formation of Escherichia coli biofilms in the absence of antibiotics. BMC Evol Biol. 2013;13:22. pmid:23356665
- 7. Lam SJ, O’Brien-Simpson NM, Pantarat N, Sulistio A, Wong EHH, Chen Y-Y, et al. Combating multidrug-resistant Gram-negative bacteria with structurally nanoengineered antimicrobial peptide polymers. Nat Microbiol. 2016;1(11):16162. pmid:27617798
- 8. Soberanes S, Misharin AV, Jairaman A, Morales-Nebreda L, McQuattie-Pimentel AC, Cho T, et al. Metformin targets mitochondrial electron transport to reduce air-pollution-induced thrombosis. Cell Metab. 2019;29(2):335–347.e5. pmid:30318339
- 9. Horiuchi T, Sakata N, Narumi Y, Kimura T, Hayashi T, Nagano K, et al. Metformin directly binds the alarmin HMGB1 and inhibits its proinflammatory activity. J Biol Chem. 2017;292(20):8436–46. pmid:28373282
- 10. Xiong W, Sun K-Y, Zhu Y, Zhang X, Zhou Y-H, Zou X. Metformin alleviates inflammation through suppressing FASN-dependent palmitoylation of Akt. Cell Death Dis. 2021;12(10):934. pmid:34642298
- 11. Bharath LP, Agrawal M, McCambridge G, Nicholas DA, Hasturk H, Liu J, et al. Metformin enhances autophagy and normalizes mitochondrial function to alleviate aging-associated inflammation. Cell Metab. 2020;32(1):44–55.e6. pmid:32402267
- 12. Pryor R, Norvaisas P, Marinos G, Best L, Thingholm LB, Quintaneiro LM, et al. Host-microbe-drug-nutrient screen identifies bacterial effectors of metformin therapy. Cell. 2019;178(6):1299–1312.e29. pmid:31474368
- 13. Xiao X, Huan Q, Huang Y, Liu Y, Li R, Xu X, et al. Metformin Reverses tmexCD1-toprJ1- and tet(A)-Mediated High-Level Tigecycline Resistance in K. pneumoniae. Antibiotics (Basel). 2022;11(2):162. pmid:35203765
- 14. Liu Y, Jia Y, Yang K, Li R, Xiao X, Zhu K, et al. Metformin restores tetracyclines susceptibility against multidrug resistant bacteria. Adv Sci (Weinh). 2020;7(12):1902227. pmid:32596101
- 15. Ibeagha-Awemu EM, Wang M, Zhao X. Invited review: multiomics insights into the molecular and regulatory mechanisms underlying bovine mastitis. J Dairy Sci. 2025;108(9):9004–33. pmid:40675471
- 16. Wang M, Yang N, Laterrière M, Gagné D, Omonijo F, Ibeagha-Awemu EM. Multi-omics integration identifies regulatory factors underlying bovine subclinical mastitis. J Anim Sci Biotechnol. 2024;15(1):46. pmid:38481273
- 17. Xu T, Wu X, Lu X, Liang Y, Mao Y, Loor JJ, et al. Metformin activated AMPK signaling contributes to the alleviation of LPS-induced inflammatory responses in bovine mammary epithelial cells. BMC Vet Res. 2021;17(1):97. pmid:33648513
- 18. Arbab AAI, Lu X, Abdalla IM, Idris AA, Chen Z, Li M, et al. Metformin inhibits lipoteichoic acid-induced oxidative stress and inflammation through AMPK/NRF2/NF-κB signaling pathway in bovine mammary epithelial cells. Front Vet Sci. 2021;8:661380. pmid:34262962
- 19. Li X, Xiang Z, Wang X, He H, Xu M, Tan C, et al. Metformin attenuates colitis via blocking STAT3 acetylation by reducing acetyl-CoA production. J Adv Res. 2026;79:393–407. pmid:40174640
- 20.
Hogan JS, Gonzalez RN, Harmon RJ, Nickerson SC, Smith KL. Laboratory handbook on bovine mastitis. Madison, WI: National Mastitis Council; 1999.
- 21. Blum S, Heller ED, Krifucks O, Sela S, Hammer-Muntz O, Leitner G. Identification of a bovine mastitis Escherichia coli subset. Vet Microbiol. 2008;132(1–2):135–48. pmid:18571344
- 22. Frank JA, Reich CI, Sharma S, Weisbaum JS, Wilson BA, Olsen GJ. Critical evaluation of two primers commonly used for amplification of bacterial 16S rRNA genes. Appl Environ Microbiol. 2008;74(8):2461–70. pmid:18296538
- 23. Clermont O, Bonacorsi S, Bingen E. Rapid and simple determination of the Escherichia coli phylogenetic group. Appl Environ Microbiol. 2000;66(10):4555–8. pmid:11010916
- 24.
Ceriotti F, Zakowski J, Sine H, Altaie S, Horowitz G, Pesce AJ, et al. Clinical and Laboratory Standards Institute (CLSI). 2012.
- 25.
Wayne PA. Clinical and laboratory standards institute. Performance standards for antimicrobial susceptibility testing. 2011.
- 26. Drlica K. The mutant selection window and antimicrobial resistance. J Antimicrob Chemother. 2003;52(1):11–7. pmid:12805267
- 27. Pfaffl MW. A new mathematical model for relative quantification in real-time RT-PCR. Nucleic Acids Res. 2001;29(9):e45. pmid:11328886
- 28. Xu TL, Seyfert HM, Shen XZ. Epigenetic mechanisms contribute to decrease stearoyl-CoA desaturase 1 expression in the liver of dairy cows after prolonged feeding of high-concentrate diet. J Dairy Sci. 2018;101(3):2506–18. pmid:29274956
- 29. Vanselow J, Yang W, Herrmann J, Zerbe H, Schuberth H-J, Petzl W, et al. DNA-remethylation around a STAT5-binding enhancer in the alphaS1-casein promoter is associated with abrupt shutdown of alphaS1-casein synthesis during acute mastitis. J Mol Endocrinol. 2006;37(3):463–77. pmid:17170087
- 30. Clermont O, Christenson JK, Denamur E, Gordon DM. The Clermont Escherichia coli phylo-typing method revisited: improvement of specificity and detection of new phylo-groups. Environ Microbiol Rep. 2013;5(1):58–65. pmid:23757131
- 31. Suojala L, Pohjanvirta T, Simojoki H, Myllyniemi A-L, Pitkälä A, Pelkonen S, et al. Phylogeny, virulence factors and antimicrobial susceptibility of Escherichia coli isolated in clinical bovine mastitis. Vet Microbiol. 2011;147(3–4):383–8. pmid:20729012
- 32. Ghanbarpour R, Oswald E. Phylogenetic distribution of virulence genes in Escherichia coli isolated from bovine mastitis in Iran. Res Vet Sci. 2010;88(1):6–10. pmid:19580980
- 33. Blum SE, Leitner G. Genotyping and virulence factors assessment of bovine mastitis Escherichia coli. Vet Microbiol. 2013;163(3–4):305–12. pmid:23374653
- 34. Tomazi T, Coura FM, Gonçalves JL, Heinemann MB, Santos MV. Antimicrobial susceptibility patterns of Escherichia coli phylogenetic groups isolated from bovine clinical mastitis. J Dairy Sci. 2018;101(10):9406–18. pmid:30031577
- 35. Zhou G, Wang Q, Wang Y, Wen X, Peng H, Peng R, et al. Outer membrane porins contribute to antimicrobial resistance in Gram-negative bacteria. Microorganisms. 2023;11(7):1690. pmid:37512863
- 36. Choi U, Lee C-R. Distinct roles of outer membrane porins in antibiotic resistance and membrane integrity in Escherichia coli. Front Microbiol. 2019;10:953. pmid:31114568
- 37. Liu Y-F, Yan J-J, Lei H-Y, Teng C-H, Wang M-C, Tseng C-C, et al. Loss of outer membrane protein C in Escherichia coli contributes to both antibiotic resistance and escaping antibody-dependent bactericidal activity. Infect Immun. 2012;80(5):1815–22. pmid:22354022
- 38. Zhou G, Wang Y-S, Peng H, Liu H-Z, Feng J, Li S-J, et al. Outer membrane protein of OmpF contributes to swimming motility, biofilm formation, osmotic response as well as the transcription of maltose metabolic genes in Citrobacter werkmanii. World J Microbiol Biotechnol. 2022;39(1):15. pmid:36401137
- 39. Jackson DW, Suzuki K, Oakford L, Simecka JW, Hart ME, Romeo T. Biofilm formation and dispersal under the influence of the global regulator CsrA of Escherichia coli. J Bacteriol. 2002;184(1):290–301. pmid:11741870
- 40. Liu X, Li J, Wu R, Bai L. Characterization of glycogen-related glycoside hydrolase glgX and glgB from Klebsiella pneumoniae and their roles in biofilm formation and virulence. Front Cell Infect Microbiol. 2024;14:1507332. pmid:39744154
- 41. Mitchell AM, Silhavy TJ. Envelope stress responses: balancing damage repair and toxicity. Nat Rev Microbiol. 2019;17(7):417–28. pmid:31150012
- 42. Rousseau CJ, Fraikin N, Zedek S, Van Melderen L. Are envelope stress responses essential for persistence to β-lactams in Escherichia coli? Antimicrob Agents Chemother. 2023;67(10):e0032923. pmid:37787525
- 43. Agius L, Ford BE, Chachra SS. The metformin mechanism on gluconeogenesis and AMPK activation: the metabolite perspective. Int J Mol Sci. 2020;21(9):3240. pmid:32375255
- 44. Zhang Y, Wang Y, Bao C, Xu Y, Shen H, Chen J, et al. Metformin interacts with AMPK through binding to γ subunit. Mol Cell Biochem. 2012;368(1–2):69–76. pmid:22644486
- 45. Xu T, Wu X, Lu X, Liang Y, Mao Y, Loor JJ, et al. Metformin activated AMPK signaling contributes to the alleviation of LPS-induced inflammatory responses in bovine mammary epithelial cells. BMC Vet Res. 2021;17(1):97. pmid:33648513
- 46. Cantó C, Gerhart-Hines Z, Feige JN, Lagouge M, Noriega L, Milne JC, et al. AMPK regulates energy expenditure by modulating NAD+ metabolism and SIRT1 activity. Nature. 2009;458(7241):1056–60. pmid:19262508
- 47. Hou W-L, Yin J, Alimujiang M, Yu X-Y, Ai L-G, Bao Y-Q, et al. Inhibition of mitochondrial complex I improves glucose metabolism independently of AMPK activation. J Cell Mol Med. 2018;22(2):1316–28. pmid:29106036
- 48. Zhou X, Chen J, Chen L, Feng X, Liu Z, Hu L, et al. Negative regulation of Sirtuin 1 by AMP-activated protein kinase promotes metformin-induced senescence in hepatocellular carcinoma xenografts. Cancer Lett. 2017;411:1–11. pmid:28964787
- 49. Wang X-D, Yu W-L, Sun Y. Activation of AMPK restored impaired autophagy and inhibited inflammation reaction by up-regulating SIRT1 in acute pancreatitis. Life Sci. 2021;277:119435. pmid:33781829
- 50. Tian R, Li R, Liu Y, Liu J, Pan T, Zhang R, et al. Metformin ameliorates endotoxemia-induced endothelial pro-inflammatory responses via AMPK-dependent mediation of HDAC5 and KLF2. Biochim Biophys Acta Mol Basis Dis. 2019;1865(6):1701–12. pmid:31002870
- 51. Yang S-R, Wright J, Bauter M, Seweryniak K, Kode A, Rahman I. Sirtuin regulates cigarette smoke-induced proinflammatory mediator release via RelA/p65 NF-kappaB in macrophages in vitro and in rat lungs in vivo: implications for chronic inflammation and aging. Am J Physiol Lung Cell Mol Physiol. 2007;292(2):L567-76. pmid:17041012
- 52. Lawless MW, Norris S, O’Byrne KJ, Gray SG. Targeting histone deacetylases for the treatment of disease. J Cell Mol Med. 2009;13(5):826–52. pmid:19175682
- 53. Salminen A, Hyttinen JMT, Kaarniranta K. AMP-activated protein kinase inhibits NF-κB signaling and inflammation: impact on healthspan and lifespan. J Mol Med (Berl). 2011;89(7):667–76. pmid:21431325
- 54. Silva LG, Ferguson BS, Avila AS, Faciola AP. Sodium propionate and sodium butyrate effects on histone deacetylase (HDAC) activity, histone acetylation, and inflammatory gene expression in bovine mammary epithelial cells. J Anim Sci. 2018;96(12):5244–52. pmid:30252114
- 55. Dai H, Wei G, Wang Y, Ma N, Chang G, Shen X. Sodium butyrate promotes lipopolysaccharide-induced innate immune responses by enhancing mitogen-activated protein kinase activation and histone acetylation in bovine mammary epithelial cells. J Dairy Sci. 2020;103(12):11636–52. pmid:33010913