The antibiotic susceptibility testing strips were purchased from Hangwei Company (Hangzhou, China). The E. coli ATCC 25922, was obtained from the American Type Culture Collection. The primary antibodies, including phosphorylated NF-κB p65, Phospho-NF-κB p65, NLRP6, TLR4, Occludin, ZO-1, and β-actin, were all sourced from Proteintech (Wuhan, China). Goat anti-rabbit or rabbit anti-mouse secondary antibodies were acquired from Beyotime (Shanghai, China). The cell viability assay kit was also procured from Beyotime (Shanghai, China).

All the experimental methods in this study were evaluated and approved by the Experimental Animal Ethics Committee of Yangzhou University (NO. 202503247). The environmental conditions of this facility meet the requirements for standard animal experimentation facilities outlined in the Chinese National Standard Experimental Animal Environment and Facilities (GB14925–2023). Animal husbandry management and experimental procedures are conducted in compliance with all applicable Chinese regulations.

From June to September 2024, this study collected milk samples from dairy cows with clinical mastitis across four intensive dairy farms (herd size > 1,000 cows each) in Xuzhou, Suqian, Huaian, and Yancheng cities, Jiangsu Province, China. A total of 96 samples were obtained based on the availability of eligible cases during the study period (convenience sampling). Clinical mastitis was identified and classified based on a modified version of the National Mastitis Council (NMC) scoring system. Only cases presenting with both local signs of udder inflammation (redness, swelling, heat, or pain) and visible abnormalities in the milk (e.g., flakes, clots, or watery secretion) were included (i.e., bovine individuals with a clinical severity score of 2). To objectively confirm the presence of mammary gland inflammation and to standardize the inclusion threshold across all participating farms, a positive California Mastitis Test (CMT) result ( ≥ ++) was required. The procedures for sample collection and processing are as follows.

Of the 96 initial milk samples, 24 E. coli isolates were recovered following the exclusion of contaminated samples and the subculture procedure described above. All E. coli isolates characterized in this study originated from samples in which E. coli was the dominant species and were selected and purified through this entire workflow.

Genomic DNA was extracted from the isolated E. coli strains using the Bacterial Genomic DNA Extraction Kit (DP302-02) (Tiangen, Beijing, China), following the manufacturer’s protocol. The procedure was briefly summarized as follows: Bacterial cultures grown overnight were collected by centrifugation, and the resulting cell pellets were resuspended in buffer. Cell lysis was achieved through lysozyme treatment, followed by protein digestion with Proteinase K. DNA was then purified using spin columns. The purified DNA was dissolved in elution buffer and stored at -20°C for subsequent PCR amplification and sequencing analysis. PCR amplification was performed using the extracted bacterial DNA as the template. The primer sequences for 16S rDNA identification are listed in Supplementary File 1. The PCR reaction mixture (total volume: 50 μL) consisted of 22.5 μL of 2 × Taq Master Mix, 2 μL each of forward and reverse primers, 1 μL of DNA template, and 22.5 μL of ultrapure water. The amplification program was set as follows: initial denaturation at 98°C for 2 min; followed by 30 cycles of denaturation at 98°C for 10 s, annealing at 58°C for 30 s, and extension at 72°C for 40 s; with a final extension at 72°C for 1 min. The PCR products were verified by 1% agarose gel electrophoresis and subsequently sent to Nanjing TsingKe Biotech Co., Ltd. for sequencing. The obtained sequences were subjected to BLAST analysis via the NCBI database¹ to determine the genus of the pathogenic bacteria. And the automated microbial identification system (BioReader5000, France) was employed to differentiate between E. coli and Shigella spp. All isolates identified as E. coli were stored at -80°C.

The agar disk diffusion method was employed to assess the antibiotic susceptibility of E. coli in Mueller-Hinton agar, as well as to detect the production of ESBLs. The following 18 antibiotic discs (Hangwei, China) were used: cefalexin (CA, 30 μg); ampicillin (AM, 10 μg); cephalothin (CF, 30 μg); ceftazidime (CAZ, 30μg); cefotaxime (CTX, 30 μg); cefepime (FEP, 30 μg); amoxicillin (AMC, 20 μg); meropenem (MEM, 10 μg); penicillin (P, 10U); Amikacin (AK, 30 μg); streptomycin (S, 10 μg); kanamycin (KAN, 30 μg); gentamicin (GM, 10 μg); tetracycline (TE, 30 μg); ciprofloxacin (CIP, 5 μg); chloramphenicol (C, 30 μg); trimethoprim-sulfamethoxazole (SXT, 20 μg); polymyxin B (PB, 300IU). The determination of drug susceptibility was based on interpretive criteria (inhibition zone diameters and MIC breakpoints for Enterobacteriaceae) published in the Clinical and Laboratory Standards Institute (CLSI) document M100 (32nd Edition), which provides standards for testing bacteria of human origin (Schuetz et al., 2025). E. coli ATCC25922 was employed as a quality control strain. In the absence of explicit standards for detecting bovine-origin E. coli, the aforementioned identification criteria represent a common and accepted practice in veterinary microbiology research. Using a sterile bacteriological swab, evenly streak the actively growing bacterial suspension across 100 mm diameter LB agar plates. Place 4–5 antibiotic sensitivity discs on each plate, ensuring a minimum center-to-center distance of 24 mm between adjacent discs to prevent overlap of inhibition zones and thereby avoid compromising measurement accuracy. The specific criteria for assessing bacterial susceptibility to antibiotics are provided in Table 1. Multidrug resistance was defined as resistance to at least one agent in three or more antimicrobial categories. The production of ESBLs was confirmed by resistance to third-generation cephalosporins and a difference of ≥ 5 mm in the zone of inhibition in the combination disk diffusion test (Magiorakos et al., 2012).

MLST typing was conducted using Pasteur’s MLST scheme. We performed MLST typing utilizing a standardized protocol specific to E. coli, and eight housekeeping genes: dinB, icdA, pabB, polB, putB, trpA, trpB, and uidA. The primer information is listed in Supplementary File 1. Genomic DNA was extracted from E. coli isolates for PCR amplification under the following conditions: initial denaturation at 96°C for 1 min, followed by 30 cycles of denaturation (96°C for 1 min), annealing (55°C for 1 min), and extension (72°C for 1 min), with a final extension at 72°C for 10 min. The PCR products were sent to Nanjing Tsingke Biotechnology Co., Ltd. for sequencing. Upon completion of sequencing, the verified PCR product sequences were submitted to the E. coli MLST database² to obtain the allele numbers of each housekeeping gene.

To analyze potential evolutionary relationships among the strains, the Phyloviz 2.0 software was employed to generate and visualize the minimum spanning tree (MST) and hierarchical clustering analysis results of E. coli strains (Feil et al., 2004). The geographical distribution and origins of E. coli ST479 were analyzed using global public MLST data of E. coli obtained from PubMLST database³

In summary, bacterial genomic DNA was extracted using a bacterial genomic DNA extraction kit (DP302, TIANGEN, Beijing, China) following the manufacturer’s instructions. The quality of the DNA was assessed using a NanoDrop ND-1,000 spectrophotometer (ThermoFisher Scientific, Wilmington, United States). Qualified DNA samples were randomly sheared into fragments of approximately 350 bp using an ultrasonicator. The processed DNA fragments were subjected to the entire library preparation procedure using the NEBNext^® Ultra™ DNA Library Prep Kit for Illumina (NEB, United States), which included steps such as end repair, A-tailing, adapter ligation, purification, and PCR amplification. After library construction, the library was initially quantified using Qubit 2.0 and the insert size was verified using the Agilent 2100. Once the insert size met the expected criteria, the effective concentration of the library was accurately quantified using RT-qPCR to ensure library quality. Sequencing was subsequently performed using Illumina PE150 (NEB, United States).

The raw data were filtered and processed using the fastp software. After quality control of the original sequencing data, genome assembly was carried out by Unicycler software, and the assembly results were evaluated. Bacterial antibiotic-resistant genes and virulence genes were annotated through the Comprehensive Antibiotic Research Database 3.2.9⁴ and Virulence Factors of Pathogenic Bacteria Database,⁵ and statistics and analysis were carried out according to the information in the databases. Plasmid replicons were determined using Plasmid Finder 3.1 Database⁶ and plasmid replicons were typed (if applicable) using Plasmid MLST 2.0 Database⁷. Bacterial genome sequence was annotated with online tool Bakta 1.8.2 and visualized using Proksee 1.1.0.⁸

BMECs were cultured in DMEM/F12 (Thermo Fisher, Waltham, MA, United States) supplemented with 10% fetal bovine serum and 1% ampicillin and streptomycin at 37°C in an atmosphere containing 5% carbon dioxide. Upon reaching approximately 80% confluence (approximately 10^⁶ cells/mL), ampicillin and streptomycin were removed from the culture, and the cells were incubated in six-well plates for 24 h. Subsequently, the cells were treated with E. coli (MOI 100:1), while the control group was treated with sterile PBS. Following an appropriate incubation period, cells were collected for subsequent analysis. Each treatment group consisted of three biological replicates.

All experiments were conducted using lactating BALB/c female mice aged 9–12 weeks, with one mouse per cage. These mice had free access to food and water. Between postpartum days 5 and 7, mice with similar body weights were selected and randomly divided into three groups for treatment. In brief, overnight cultured bacteria were subcultured and propagated, with bacterial concentration determined through serial dilution and colony counting. A bacterial suspension was collected and centrifuged at 5,000 rpm for 5 min, followed by two washes and resuspensions in PBS to adjust the concentration to 5 × 10^⁹ CFU/mL. Subsequently, a 50 μL bacterial solution was directly injected into the fourth pair of mammary glands of the mice via the mammary duct. Control group mice received an equivalent injection of 50 μL sterile PBS solution. Model validation and criteria for success are as follows: The model was confirmed 24 h after injection by a combined assessment of mammary morphology, histopathology, and biochemical indicators. Only animals meeting all of the following predefined criteria were considered to have successful mastitis induction and were included in subsequent analyses: Clinical assessment: visual inspection of the injected fourth pair of mammary glands showing pronounced congestion (erythema) and swelling. Histopathological evaluation: excised mammary tissues fixed, sectioned, and stained with hematoxylin and eosin (H&E) to assess the degree of inflammatory cell infiltration in the mammary parenchyma. Increased relative expression of inflammatory genes in tissue: elevated expression of inflammatory genes such as TNF-α, IL-1β, and IL-6. Criteria for judgment based on prior research (Lee et al., 2003). There are three mice in each group, with a total of nine mice. The mice were monitored for 24 h, with body temperature measured hourly, followed by dissection to harvest mammary tissue. Each treatment group consisted of three biological replicates.

Total RNA was extracted from cultured cells using Trizol reagent (Vazyme, Nanjing, China) according to the manufacturer’s instructions. Complementary DNA was synthesized using a reverse transcription kit (ThermoFisher Scientific, Wilmington, DE, United States). RT-qPCR experiments were conducted using SYBR Green PCR Master Mix (ThermoFisher Scientific, Wilmington, DE, United States) on the CFX Connect real-time PCR detection system (Bio-Rad, RI, United States). GAPDH was utilized as the house-keeping gene. The relative mRNA expression levels of the target genes were calculated using the 2^{–Δ Δ Ct} method. All primer sequences are listed in Supplementary File 2.

According to the manufacturer’s instructions, the Cell Counting Kit-8 (Beyotime, Shanghai, China) was used to detect cell viability. Briefly, BMECs were seeded in 96-well plates at a density of 2.5 × 10³ cells per well. Subsequently, E. coli was used for the attack. At the designated time points, the medium was replaced with 100 μL DMEM/F12 medium (supplemented with 10 μL CCK-8) and then incubated at 37 °C for 1.5 h. The absorbance at 450 nm was measured using a microplate reader (TECAN, Männedorf, Switzerland). All experiments were repeated three times.

Total protein from BMEC cells was harvested using the protein extraction kit (Beyotime, Shanghai, China). Proteins were separated using 10 or 15% SDS-PAGE and subsequently transferred to a 0.45 μm PVDF membrane. After blocking with 5% non-fat milk, the prepared PVDF membrane was incubated overnight at 4°C with specific primary antibodies, including NLRP6 (Cat No. 30973-1-AP), β-actin (Cat No. 20536-1-AP), TLR4 (Cat No. 19811-1-AP), NF-κB p65 (Cat No. 10745-1-AP), and Phospho-NF-κB p65 (Cat No. 80379-2-RR) sourced from Proteintech (Wuhan, China). Following three washes with TBST, the PVDF membrane was treated at room temperature for 2 h with goat anti-rabbit IgG (Beyotime, Shanghai, China). The membrane was washed three times with TBST. Finally, following the addition of enhanced chemiluminescence, the protein blots were imaged using the Western blotting detection system (Tanon, Shanghai, China).

Mammary tissue sections were dewaxed in xylene, and the slides were then hydrated in 95% ethanol for 5 min, 85% ethanol for 5 min, 75% ethanol for 5 min, and distilled water. Hematoxylin staining was then performed for 3 min, and then slides were rinsed with water for 2 min. After the runoff was transparent, the slides were sealed by the addition of a drop of neutral gum to the center of the paraffin section and adding a cover slip. The slides were then observed and photographed using a microscope (Leica, Germany).

The slides were incubated at 60°C for 30 min, dewaxed and hydrated. First, the sections were placed in xylene twice for 10 min each, incubated with 100, 95, 85, and 75% ethanol for 5–10 min and soaked in distilled water for 5 min. For antigen retrieval, the sections were incubated in citrate buffer (pH 6.0), heated in a microwave at high heat for 8 min, removed and cooled to room temperature. The slides were then incubated with the primary antibody, with the secondary antibodies in a box at 37°C for 1.5 h and then with a streptavidin-HRP antibody at 37°C for 20 min. After the runoff was transparent, the slides were incubated twice in xylene for 10 min each. Each slide was sealed with neutral gum and observed under the microscope.

All data were analyzed using GraphPad Prism 9.5.1 (San Diego, CA, United States) and expressed as mean ± standard deviation. Comparisons between two groups were conducted using the t-test, while comparisons involving more than two groups were performed using one-way analysis of variance to assess variance and test for significance. A significance level of P < 0.05 was considered statistically significant, with additional specific analyses detailed in the figure legends.

The 16S rRNA sequencing results indicated the detection of 24 strains of E. coli among 96 milk samples. Through antibiotic susceptibility testing, 24 strains (100%), 20 strains (91.7%), and 17 strains (70.8%) of E. coli exhibited resistance to penicillin, cefotaxime, and cephalexin, respectively. Additionally, 23 strains (95.8%) and 20 strains (83.3%) displayed resistance or intermediate susceptibility to cephalothin and ampicillin, respectively. Furthermore, 15 strains (62.5%) and 11 strains (45.8%) of E. coli showed intermediate susceptibility to cefepime and ceftazidime, respectively. All isolated strains demonstrated no resistance to amikacin, chloramphenicol, meropenem, and polymyxin B (Figure 1). Notably, we observed a striking phenomenon where approximately 45% (n = 11) of E. coli exhibited remarkable consistency in their resistance profiles. Concurrently, these strains displayed resistance to a greater variety of antibiotics, resulting in 9 distinct resistance changes (7 antibiotic resistances, 2 intermediate susceptibilities) when compared to the laboratory strain ATCC25922. Subsequently, we conducted further typing analyses to determine whether these populations belonged to the same clonal group.

A total of 12 distinct sequence types (ST) were identified, with ST479 being the most prevalent (n = 11). The remaining STs included ST303 (n = 2), ST87 (n = 2), ST368 (n = 1), ST399 (n = 1), ST21 (n = 1), ST414 (n = 1), 1,081 (n = 1), 360 (n = 1), and three STs that were not assigned in the database (designated as N1, N2, and N3) (Figure 2A and Supplementary File 3). Consistent with our previous hypothesis, isolates exhibiting high levels of antibiotic resistance were all of the same ST, namely ST479. ST479 was derived from mastitis milk samples collected from four dairy farms in Jiangsu, China (Figure 2B), indicating that ST479 not only possesses extensive antibiotic resistance but also exhibits regional transmission advantages. Phylogenetic analysis revealed high genetic diversity among the E. coli isolates. Notably, the ST479 isolates formed a distinct, tightly clustered group, which was clearly separated from the other, more diverse isolates (Figures 2C,D). To investigate the origin and transmission routes of this clone, we conducted further analysis on ST479.

The analysis of PubMLST public data indicates that ST479 is widely distributed globally, with Europe and Asia, particularly the United Kingdom and China, being the primary regions of distribution (Figure 3A). In the Pasteur MLST scheme, ST479 corresponds to ST224 in the Achtman MLST scheme. Our results show that ST479/224 was first detected and recorded by researchers in France in 2002. Since then, it has been continuously identified across various regions worldwide from 2002 to 2022, with a notable increase in detection rates occurring in 2014 (Figure 3B). Overall, E. coli ST479/224 has spread to varying extents in multiple countries, with the highest prevalence observed in the United Kingdom (25.4%), followed by China (16.3%), Spain (8.2%), Ireland (8.2%), and the United States (8.2%) (Figures 3C,D). Interestingly, 51% of ST479 was found in animal hosts, predominantly in pigs (20.4%) and cattle (12.2%) (Figures 3E,F).

The bacterial genome annotation reveals the 4987430-base pair sequence of E. coli ST479 (Figure 4A). The genomic annotation results indicate that ST479 contains 153 virulence genes, primarily focused on bacterial Adherence (n = 50), Effector delivery system (n = 29), and Motility (n = 48) according to the categories in The Virulence Factor Database (Supplementary File 4). More than 40 genes are associated with the bacterial pilus system, such as the stg ADCD gene cluster and the flgABCDEFGHIJKL gene cluster. These genes are primarily related to the rapid colonization and invasion of bacteria within the host. Additionally, several gene clusters associated with enterobactin synthesis and transport have been detected, including entABCDEF and fepABCD. Similar to most other MPEC isolates, ST479 possesses the type II secretion system (T2SS), type III secretion system (T3SS), and type VI secretion system (T6SS), as well as genes related to long polar fimbriae and iron capture (Table 2). The functional classification of 4,574 core genes in E. coli ST479 was explored using the COG database, revealing that most of these genes are involved in fundamental bacterial structure and metabolism (Figure 4B). The results of the KEGG enrichment analysis indicate that E. coli ST479 possesses a substantial number of genes associated with the Signal Transduction and Membrane Transport pathways, which are critically important for MPEC during infection of the bovine mammary gland (Figure 4C).

ST479 contains 60 resistance genes, predominantly comprising tetracycline antibiotics (n = 3), peptide antibiotics (n = 5), nucleoside antibiotics (n = 3), nitroimidazole antibiotics (n = 1), macrolide antibiotics (n = 11), fosfomycin (n = 4), fluoroquinolone antibiotics (n = 18), cephalosporins (n = 6), and aminocoumarin antibiotics (n = 9) (Figure 4D and Supplementary File 5). This includes the CTX-M-55 and aph(3’)-III genes, with the CTX-M-55 enzyme belonging to the CTX-M-1 group, first identified in E. coli producing CTX-M-55 type ESBLs in Thailand in 2004 (Kiratisin et al., 2007). Since then, it has rapidly spread to dozens of countries, including China, Japan, the USA, the UK, and Russia (Yang et al., 2023). Consequently, we conducted susceptibility tests to verify the production of CTX-M-type ESBLs. The results show that ST479 is a bla_CTX–M–55-positive E. coli capable of producing ESBLs (Figure 4E).

In the genomic annotation of ST479, we identified two types of plasmids, pCoo and pECOE, belonging to the Inc.FII and Inc.B/O/K/Z types. The annotation results reveal that the plasmid pCoo contains the aph(3’)-III resistance gene and seven virulence genes: EspC, tssM, espP, pilV, Pic, pilN, and set1A. The pECOED plasmid did not annotate any resistance genes but contains two virulence genes, SenB and tssM.

To compare the pathogenic effects of ST479 with those of the reference strain ATCC 25922, we developed inflammation models using BMECs and the mouse mastitis model (Figure 5A). After challenge with ST479, the morphology of BMECs displayed significant changes at 3 h post-infection, and by 12 h, there was complete cell death (Figure 5B). RT-qPCR results indicated a marked increase in the mRNA levels of inflammatory cytokines IL-1β, IL-6, IL-8, and TNF-α in BMECs at 3 h post-infection (Figures 5C–F). In comparison to the laboratory strain ATCC29522, the upregulation of inflammatory cytokine mRNA was lower in cells infected with ST479 (Figures 5G–J). However, there was a more significant decline in cell viability (Figure 5K). This suggests that alterations in the virulence genes and antibiotic resistance of ST479 may have modified its pathogenic capabilities and mechanisms relative to ATCC25922. The reduced inflammatory response likely enhances the host’s chances of survival, thereby facilitating the growth and dissemination of ST479. Furthermore, Western blotting results demonstrated that, similar to ATCC25922, ST479 activates the classical inflammatory signaling pathways NLRP6/TLR4/NF κB following the induction of inflammation (Figure 5L).

A murine mastitis model was established to comparatively assess the pathogenic potential and host inflammatory response induced by E. coli ST479 against the reference strain ATCC 25922. The mouse model of mastitis was successfully established, similar to the model of BMECs. Results indicated that 24 h after induction with ST479, the mammary tissues of the mice showed diffuse hemorrhage, tissue pallor, and other inflammatory symptoms (Figure 6B). Additionally, the body temperature increased rapidly and was significantly higher than that of the control group (Figure 6G). RT-qPCR analysis of mammary tissues revealed that the mRNA levels of inflammatory cytokines IL–1β, IL–6, IL–8, and TNF-α were significantly elevated (P < 0.05) (Figures 6C–F). H&E staining demonstrated that the morphology of the mouse mammary tissue was compromised, showing notable damage to the alveolar lumen structure and extensive infiltration of inflammatory cells (Figure 6H). Immunofluorescence imaging revealed a reduction in tight junction proteins ZO-1 and Occludin in the mammary glands of treated mice compared to the control group (Figures 6I,J). Complete blood count results indicated that mice with mastitis exhibited decreased numbers of white blood cells, platelets, lymphocytes, and platelet volume fraction (P < 0.05) (Figures 6K–N). Interestingly, although the percentage of neutrophils showed an upward trend, the absolute number of neutrophils significantly decreased (P < 0.05) (Figures 6O,P). Furthermore, based on clinical symptoms, inflammatory cytokine mRNA levels, and routine blood test parameters, mastitis induced by ST479 in mice appeared to be less severe than that induced by the laboratory strain ATCC25922.