Antimicrobial Susceptibility and Molecular Characterisation of Escherichia coli Strains Isolated from Raw Milk and Raw Milk Products in Ireland

1. Introduction

Most E. coli strains are ubiquitous commensal organisms, existing harmlessly in humans, animals and their environment and seldomly causing disease [1]. However, some strains can be pathogenic due to the presence of specific virulence factors encoded by genes located on the chromosome or mobile genetic elements[1,2]. These pathogens cause a wide range of conditions, from common illnesses like cystitis, urinary tract infections (UTIs), and gastroenteritis to life-threatening extraintestinal infections, including pneumonia, neonatal meningitis, haemorrhagic colitis or haemolytic-uremic syndrome.[3]. Pathogenic E. coli are classified into pathotypes based on their virulence factors and their association with different diseases in humans[4,5,6]. Genes encoding virulence factors confer the potential for mechanisms to cause disease such as adhesion, invasion and toxin production amongst others[2,7,8]. In addition, efflux pump also play an important role in AMR, virulence and disinfectant resistance through their ability to transport and expel a wide range of unwanted material from the bacterial cell[9,10,11,12].

Illnesses caused by pathogenic E. coli require treatment using antibiotics, including last resort antibiotics on those specific cases where an infection is caused by multidrug resistant (MDR) organisms or where all other treatment options have failed[13,14]. Resistance to antimicrobials classified by the World Health Organisation (WHO) as critically important (CIA) for human medicine, is of major concern since infections caused by these organisms are associated with the most adverse health consequences[15]. Extended spectrum β-lactamases (ESBLs) such as SHV, TEM, CTX-M and AmpC β-lactamases, are enzymes that confer resistance to the CIA cephalosporins and have been commonly identified in E. coli from food and food producing animals in Ireland, Europe and worldwide[16,17,18,19]. The spread of MDR or ESBL-producing E. coli due to the misuse and/or overuse of antibiotics in animal and human health as well as the environment is a serious concern[20].

Food-producing animals act as potential reservoirs of AMR bacteria and antimicrobial resistant genes (ARGs), which can be transmitted to other animals and humans through direct contact, the food chain or the environment[21,22,23,24,25,26]. This poses a significant public health threat requiring a one health approach linking human health, animal health and the environment with a strong focus on the emergence and spread of AMR and ARGs[21,23,24]. AMR can be intrinsic, often due to the bacterial cell wall structure, permeability, and the presence of natural efflux pumps that prevent antimicrobial agents from reaching their targets within the cell, but it can also be acquired, due to natural selection, through mechanisms such as horizontal gene transfer, driven by plasmids or due to genomic point mutations[27,28]. Bacteria usually acquire ARGs due to selective pressure exerted through the use and misuse of antimicrobials, resulting in them to be ineffective[20,29]. The use of heavy metals (e.g. copper and zinc) and biocides (e.g. acetic acid and hydrogen peroxide) in food production settings is also known to lead to co-resistance (resistance to several drugs which is linked genetically) through co-selection and co-transfer mechanisms[30,31]. In recent years, especially since the COVID-19 pandemic, non-corrosive biocides such as Quaternary ammonium compounds (QACs) are the most frequently used disinfectant in the food industry with a significant increase in their use[32,33,34,35]. Benzalkonium Chloride (BAC) and Dialkyldimethylammonium Chloride (DDAC) are commonly used QACs in the dairy industry where they are used to disinfect the udder, milking equipment, milk storage tanks and dairy processing machines[36]. However, it has been demonstrated that their overuse can lead to mutation, toxicity, spread of ARGs and therefore lead to increased AMR in bacteria[32,37].

Dairy cows account for a large proportion of Ireland’s cattle herd, with over 1.4 million recorded in December 2024 (Central Statistics Office, 2025)[38]. Mastitis, which is the inflammation of the mammary gland, is the most prevalent dairy cow disease worldwide and the main cause of antimicrobial use in the dairy industry[39]. E. coli has been reported to be one of the main pathogens responsible for mastitis[39,40,41]. While CIA should not be used as a first line of treatment, if milk culture results indicate that there is no alternative treatment or in exceptional circumstances, fluoroquinolones, macrolides and 3^rd and 4^th generation cephalosporins, may be used[42]. In addition to that, and prior to the introduction of the European Veterinary Medicines Regulation (Regulation (EU) 2019/6)[43] in 2022, treatment of the entire herd with intramammary, long-acting antimicrobials at the end of the lactation and during the dry period was carried out in the majority of Irish herds as a preventative measure[44,45].

According to the National Dairy Council, milk for sale in Ireland is generally pasteurised, however, it is legal to sell raw milk and there is a small sector of the population that consumes raw milk and raw milk products such as butter, cheese and cream[46]. Consumers could potentially acquire AMR and ARGs by consuming raw milk or raw milk dairy products[47,48]. Although there is limited information on the prevalence of AMR E. coli and ARG’s in raw milk and raw milk products in Ireland and Europe, AMR E. coli have been detected in milk from animals with mastitis and from healthy cows[49,50,51,52]. Therefore, the aim of this study was to investigate the prevalence of commensal E. coli and fluoroquinolone resistant, ESBL/AmpC β-lactamase and carbapenemase producing E. coli in raw milk and raw milk products in Ireland to help understand the risks of raw milk as a vehicle for AMR E. coli. AMR was determined using phenotypic methods and a subset of the strains were characterised using Whole Genome sequencing (WGS) to obtain a deeper understanding of the ARGs together with other related or significant genetic determinants, i.e. virulence factors, plasmid replicons, disinfectant resistance, efflux pumps genes, MLST and phylogenetic diversity and pathotype.

2. Materials and Methods

Sample Selection

A total of 139 samples received in our laboratories between March and September 2022 as part of different routine DAFM testing programs were selected for this study: A) 87 raw milk samples from individual dairy farms collected as part of the national residue control plan of Ireland (RM-NRCP). This milk is intended for pasteurisation or further processing. B) 38 raw milk samples from individual food business operators (RM-FBOs) approved for selling raw milk for human consumption. This milk is intended to be consumed raw. C) 14 raw dairy products (RDP) intended for human consumption: 5 butter, 7 cheeses and 2 cream samples.

Isolation of ESBL/AmpC β-Lactamase and Carbapenemase Producing E. coli

Samples were diluted 1:10 in buffered peptone water (BPW; Syntec Scientific, Dublin Ireland) and incubated for 18-22 h at 37^oC. ESBL, AmpC and carbapenemase producing E. coli were isolated using the method previously described and recommended by the European Union Reference Laboratory for Antimicrobial Resistance (EURL-AR)[53]. In brief, the enriched BPW was spread plated onto appropriate selective agars i.e., MacConkey plates supplemented with 1 mg/L cefotaxime (E&O laboratories Ltd., Bonnybridge, UK) for ESBL and AmpC producing E. coli and ChromID CARBA and ChromID OXA-48 plates (BioMérieux, Basingstoke, UK) for the selective isolation of carbapenemase producing E. coli. All selective plates were incubated at 37°C for 18-22 h. Presumptive colonies on the chromogenic agars and those that grew up close to the ciprofloxacin disc were purified on the selective media from which they were obtained and on cystine–lactose–electrolyte-deficient (CLED) agar (W12515, Fannin Ltd, Dublin Ireland) as previously described[54].

Isolation of Fluoroquinolone Resistant E. coli

Isolation of fluoroquinolone resistant E. coli was carried out according to the method previously described[54]. In brief, a sterile cotton swab was dipped into enriched BPW and spread over the entire surface of a MacConkey agar plate (E&O laboratories Ltd, Bonnybridge, Scotland, UK.) which was streaked with the same swab after rotating the plate approximately 60 degrees. Immediately after the plate was inoculated, a 5 µg ciprofloxacin paper disk (CT0425B, Fannin Ltd, Dublin Ireland) was placed aseptically onto the middle of the agar plate. Plates were incubated as described above and presumptive colonies were those that grew up close to the ciprofloxacin disc. Presumptive colonies were purified as previously described[54].

Isolation of Commensal E. coli

Commensal E. coli was isolated from raw milk and raw milk products following the standard ISO 16649-2:2001[55]. Milk, cream and butter samples were diluted in maximum recovery diluent (MRD: SKU: KM0040, E&O Laboratories, Scotland, UK) while cheese samples were diluted in 2% tri-sodium citrate solution (Scientific Laboratory Supplies, Rathcoole, Dublin) as previously described[56]. A 1 ml volume of the required dilutions (Milk: 10^o and 10^-1, cream and butter: 10^-1 and cheese: 10^-1 and 10^-2) was pipetted into the centre of a Petri dish followed by the addition of 15 ml Tryptone Bile X-Glucuronide (TBX) Medium (CM0945B, Fisher Scientific Ltd, Dublin, Ireland) which was tempered to 45 ± 2^oC followed by gentle mixing. Once solidified, plates were incubated for 4 h at 37^oC followed by 18 to 24 h at 44 ± 1^oC.

Positive and Negative Controls

For each batch of samples tested, positive and negative controls were used i.e., E. coli NCTC 13353 (ESBL producer), E. coli NCTC 13476 (carbapenemase resistant) and E. cloacae 03-577qnr-AI (fluoroquinolone resistant). E. coli ATCC 25922 was used as the commensal positive control and as a negative control for the ESBL/AmpC β-lactamase producers and carbapenemase producing E. coli. These control isolates were obtained from the EURL-AR.

Bacterial Species Confirmation

Matrix Assisted Laser Desorption Ionization-Time of Flight (MALDI-TOF) Mass Spectrometry (Bruker Daltronics GmbH, Bremen, Germany) was used to confirm all E. coli isolates as previously described[57]. Confirmed E. coli isolates were maintained at -80^oC pending further analyses.

Antimicrobial Susceptibility Testing

Antimicrobial susceptibility testing (AST) was carried out by determining the Minimum inhibitory concentration (MIC) using the Sensititre™ EU surveillance Salmonella/E. coli EUVSEC3 AST Plates (Thermo Scientific, MA, USA). A control strain of E. coli ATCC 25922 was included alongside each batch of strains that were tested. In addition, the negative control well was streaked onto blood agar for confirmation of no contamination. The interpretive breakpoints used in this study were those recommended by the EURL-AR and according to the Commission Implementing (EU) Decision 2020/1729[58,59,60]. Presumptive ESBL, AmpC or carbapenemase producing E. coli isolates were tested further with the second panel of antimicrobials, the Sensititre™ EU Surveillance ESBL EUVSEC2 AST plates, which included cefoxitin, cefepime, and clavulanate in combination with cefotaxime and ceftazidime, for the detection of ESBL and AmpC production. EUVSEC2 plates also included imipenem, meropenem and ertapenem to phenotypically verify presumptive carbapenemase producing E. coli (EUCAST 2013)[59]. Depending on the results obtained from the susceptibility testing performed with the second panel, the isolates were subsequently classified as presumptive ESBL, AmpC, ESBL/AmpC or carbapenemase producers according to EUCAST 2013 criteria[59,60]. Classification of presumptive chromosomal or plasmid mediated fluoroquinolone resistance was carried out as previously described[54].

Whole Genome Sequencing Analysis

Selection of Isolates

In total, 33 isolates were selected for WGS, and this included all phenotypically AMR resistant isolates (13) and 20 isolates that were fully susceptible to all antimicrobials.

Culture Preparation and DNA Extraction

E. coli isolates were recovered from frozen stocks on Columbia Agar and Horse blood (E & O Laboratories LTD) and incubated for 24 h at 37°C. A 1 μl loopful of pure culture was taken from the plate and re-suspended in 100 μl nuclease free water. DNA was extracted using the Roche MagNA pure 96 system and quality checked as previously described[17].

Library Preparation, Amplification, Denaturing and Sequencing of Libraries

Sample libraries for all isolates were prepared using the Illumina DNA prep kit (Illumina, Inc., Eindhoven, Netherlands) as per manufacturer’s instructions and using 30 μl of stock DNA to give an input range of 100–500 ng. The library was amplified by 5 PCR cycles with normalisation and denaturing performed as previously described[17]. The molarity of each sample was normalised to 2 nM and the same quantity of each normalised library was pooled to produce a pooled amplified library which was then loaded with 2% PhiX (control library). The isolates included in this study were distributed over four sequencing runs on the NextSeq 2000 (Illumina) using a NextSeq™ 1000/2000 P1 Reagent Cartridge (300 cycles) and NextSeq™ 1000/2000 P1 Flow Cell.

Data Analysis

Statistical Analysis

Using Epitools[61], a Chi-square test from cross tabulation of the raw data was employed to determine whether there was an association between RM-NRCP or RM-FBO recovery of commensal E. coli. Contingency tables were analysed, and results were considered statistically significant when p < 0.05. Due to the low sample numbers of butters, cheeses and creams that were received, statistical analysis was not performed on RDP.

Analysis of WGS Data

The following run metrics were used to check that each WGS run met the basic quality metrics requirements for raw sequence data: >70% bases higher than Q30 at 2 × 300 bp and cluster density of 800 -1200 k/mm². In addition, the percentage of reads that aligned to the PhiX was also checked to ensure that the starting concentration of the libraries were not over or under-estimated.

MBioSEQ Ridom Typer software (formerly known as SeqSphere), version 10.04, 2024-07 (Ridom GmbH, Münster, Germany) was used to process and analyse WGS data, i.e., quality trimming, de novo assembly, quality assessment and analysis of WGS data. Raw reads (FASTQ) were automatically trimmed prior to assembly using the default settings for trimming i.e., trimmed at both ends until the average base quality was >30 in a window of bases. Paired-end raw read data were assembled using the SPAdes assembler’s algorithm (V3.15.4) set at the automatic mode to determine the coverage cut-off and minimum length of contigs. De novo assemblies with >30x coverage, 5.0 Mb ± 10% bp in size, an N50 ≥ 15,000 and core percentage of ≥95% were included in the downstream WGS analysis. Species confirmation, MLST using the Warwick scheme and core genome MLST (cgMLST) from assembled E. coli genomes was conducted using SeqSphere+. Additionally, the integrated NCBI AMRFinderPlus (AMR Gene Finder plus) software[62,63,64] and virulence factor database[65] (VFDB) were used to identify AMR and virulence, respectively. Furthermore, genes related to disinfectants and general efflux were also identified within AMRFinderPlus. In silico phenotyping was performed using the freely available online ClermonTyper, based on an In silico quadruplex PCR targeting the genes arpA, chuA, yjaA and TspE4.C2 and is accessible at http://clermontyping.iame-research.center/. Each FASTA file was uploaded, as input and the species/phylogroup was determined in two distinct ways: by in silico PCR and Mash. Mash is the species phylogroup of the nearest genome[66]. The presence or absence of the different amplicons constituted a profile that allowed for phylogroup assignment.

CGE tools (Center for Genomic Epidemiology) available at http://cge.cbs.dtu.dk/services, were used for further analysis of assembled data, i.e., in silico detection of plasmid replicons using PlasmidFinder 2.1[67,68] while PathogenFinder 1.1[69] was used to predict the potential of each E. coli isolate to be a human pathogen by uploading the FASTA file and choosing the proteobacteria as the phylum and gamma as the class. The number and type (pathogen specific or non-pathogen specific) of query strain orthologues in the protein family database are used to assign a human pathogen probability score (range 0.0 – 1.0), where a probability score > 0.5 is considered as the threshold for consideration as a human pathogen.

3. Results

Bacterial Culture

Forty-seven E. coli isolates were recovered from the 139 milk and dairy product samples tested (33.8% isolation rate) and this is summarised in Table 1. The percentage of commensal E. coli positive samples amongst the sample types was 29.9%, 39.5% and 42.9% from RM-NRCP, RM-FBO and RDP respectively. Although a higher proportion of commensal E. coli were recovered from RM-FBO samples compared to RM-NRCP, Chi-Squared analysis indicated that the difference was not significant (P > 0.05). In addition, four isolates were recovered from the MacConkey agar plates with the ciprofloxacin paper disk (3 RM-NRCP and one RDP from cream), and two isolates (RM-FBO and RM-NRCP) were isolated from MacConkey plates with cefotaxime. No presumptive carbapenemase producing E. coli were isolated from any of the samples tested.

Table 1. E. coli isolated from raw milk and raw milk dairy products.

Table 1. E. coli isolated from raw milk and raw milk dairy products.

Antimicrobial Susceptibility Testing

AST resulted in 34 isolates found fully susceptible to all antimicrobials tested and 13 isolates resistant to between two and nine antimicrobials (Table 2). The majority of resistant isolates were multidrug resistant (MDR) with ten of the 13 resistant isolates displaying resistance to at least one antimicrobial in three or more antimicrobial classes (Table 2). Ampicillin resistance was the highest (25.3%), followed by sulfamethoxazole (23.4%), trimethoprim (21.3%) and tetracycline (14.9%). Resistance levels were moderate for ciprofloxacin and nalidixic acid (8.5%) and chloramphenicol (8.5%) and only one isolate was resistant to gentamicin (2.1%). None of the isolates were resistant to colistin (polymyxin) or meropenem (carbapenem). Amongst the two presumptive ESBL/AmpC β-lactamase producers, the E. coli isolated from RM-NRCP was found to be a presumptive AmpC while the isolate from RM-FBO was found to be a presumptive ESBL producer when the EU Surveillance ESBL EUVSEC2 AST plate was used. The four presumptive fluoroquinolone isolates were indicative of phenotypic AMR chromosomal fluroquinolone resistance i.e., resistant to both ciprofloxacin and nalidixic acid[27].

Whole Genome Sequencing

Following WGS, one of the 33 sequenced genomes failed the quality metrics from the raw sequencing data. This was an isolate which was fully susceptible and was subsequently excluded from the WGS analysis. Over each run on the Miseq, the index reads were evenly distributed.

The four presumptive chromosomal fluoroquinolones were confirmed genotypically and three of the four displayed double mutations in gyrA and two in parC/parE (Table 3). The most frequent mutations were observed in codon 83 of gyrA and codon 80 of parC gene. The presumptive fluoroquinolone E. coli recovered from RDP, harboured a single mutation in codon 83 of gryA (Table 3). The presumptive AmpC β-lactamase producer encoded the upregulated AmpC promotor at position 42 (C-42T mutation), and harboured the blaTEM-1 gene, while the presumptive ESBL producer harboured the gene blaCTX-M-3 and was also phenotypically resistant to Gentamicin which was confirmed genotypically through identification of the ant(2'')-Ia gene (Table 3). Other aminoglycoside resistant genes responsible for encoding resistance to kanamycin and streptomycin were identified amongst the isolates, although none of them were phenotypically resistant to either kanamycin or streptomycin.

Ampicillin resistance was observed in 12 isolates which included the ESBL and AmpC isolates. BlaTEM-1 was identified in eight of the ampicillin-resistant commensal E. coli isolates, while blaTEM-30, blaTEM-39 in combination with blaOXA-1 and blaOXA-1, were identified in the other three ampicillin-resistant commensal E. coli isolates. Finally, four isolates resistant to chloramphenicol harboured the floR gene, seven isolates resistant to tetracyclines harboured the tet(A) or tet(B) genes, while 10 ten isolates resistant to both trimethoprim and sulfamethoxazole harboured the dfrA and sul1/sul2 genes (Table 3).

Efflux pump genes were present in all 32 E. coli genomes i.e., both fully susceptible and r isolates. Three efflux pump genes were identified i.e., the acrF, mdtM and emrE genes, either alone or in combination, in 29, 23 and 17 genomes, respectively (Table 3).

The resistance gene qacEΔ1, which confers resistance to disinfectants was identified in four of the sequenced E. coli isolates (Table 3). Three of these isolates were from RM-NRCP and two of these were fluoroquinolone resistant. The ESBL isolate from RM-FBO also harboured the qacEΔ1 gene.

Fifteen plasmid replicons were identified in 30 E. coli isolates i.e., IncFIB(AP001918), IncFII, IncFIA, IncFIC(FII), IncFIB(pB171), IncI1-I(Alpha), IncFII(pCoo), IncFII(29), IncX1, IncFII(pSE11), IncFII(pHN7A8), IncY, IncI2(Delta), ColpVC and IncX4, present respectively in 21, nine, seven, seven, four, four, four, two, and one of each of the rest of isolates. Twelve of the 13 AMR isolates harboured at least two plasmid replicons and the total number of plasmid replicons identified amongst all 12 isolates was six with IncFIB(AP001918) and IncFII, identified in 85% and 69% of resistant E. coli. The only resistant isolate without detectable plasmid replicons was the fluoroquinolone-resistant E. coli isolate recovered from RMD. Similarly, 18 of the 19 isolates that were fully susceptible harboured a total of 14 plasmid replicons, including major plasmid replicons such as IncFIB(AP001918), IncFIA, and IncFII. IncFIB(AP001918) was present in 53% of the fully susceptible isolates.

All 32 sequenced E. coli isolates were predicted to be human pathogens with an average probability score of 0.91 (range 0.84 to 0.96) as shown in Table 3. The strains with the highest average probability scores, i.e., 0.96, were the ESBL isolated from RM-FBO and the fluoroquinolone resistant isolate from RMD. The least probability score (0.84) was identified in an isolate from RM-NRCP which was fully susceptible. The genome sequences of the E. coli matched pathogenic families with a range of 53 to 755 per isolate i.e., 53 pathogenic families matched a fluoroquinolone resistant isolate from RM-NRCP, and 755 pathogenic families matched a fully susceptible isolate from RM-FBO (Table 3). In silico phylogrouping permitted the classification of E. coli isolates into four of the seven main phylogroups i.e., A, B1, C and E, in ten, seventeen, four and one isolate, respectively. Two of the fluoroquinolone resistant isolates were assigned to phylogroup A, whereas the other two were assigned into phylogroups B1 and C. The ESBL and AmpC producing isolates were assigned to phylogroups B1 and C, respectively (Table 3). There was 100% concordance between the in-silico PCR assay and the Mash genome-clustering tool.

In silico MLST typing yielded 22 different STs amongst the 32 E. coli, with ST10 the most common followed by ST58 identified in five and three isolates respectively. One isolate of unknown ST was also identified. The distribution of virulence genes amongst the 22 ST’s of E. coli isolates is shown in Figure 1. Major virulence determinants were identified in the sequenced E. coli isolates, including shiga-toxin producing (stx2A and stx2B), the cytolethal distending toxin operon (cdtABC), α-hemolysin (hlyA, hlyB, hlyD), the gene encoding arylsulfatase A enzyme (aslA) Cytotoxic Necrotizing Factor 1 (cnf1), EAST1 (astA/east1), high-affinity iron-acquisition systems (entB–E, entS, fepA–D/G, fes), iron uptake including the salmochelin siderophore system (iroBCDEN), the 8-gene cluster (chu genes) involved in heme transport and processing (chuA, chuS-Y), the type III secretion system (T3SS) effector genes (espL1, espL4, espP, espR1, espX1, espX4, espX5, espY1, espY2, espY3, espY4). Multiple fep genes encoding ferric enterobactin transport functions (fepA, fepB, fepC, fepD, fepG) were identified in all but one of the E. coli isolates, as was the ompA gene which affects adhesion and plays a key role in the resistance of bacteria to antibiotics. Genes involved in aerobactin biosynthesis (iucA–D, iutA), the outer membrane hemin receptor for heme uptake (chuA–Y) and key adherence factors such as fdeC, type 1 fimbriae (fimA–H), P fimbriae (papA–K, papG), sfaX and associated regulators were also identified. In general, all isolates within each ST displayed the same virulence profile. Amongst all 32 isolates, one isolate of ST10 (identified as fluoroquinolone resistant), harboured two stx2 gene toxins (stx2A and stx2B). The aggregative heat-stable enterotoxin 1 (EAST1) was identified in six isolates of ST10, ST58, ST88, ST410, ST8842 and ST17585 (Figure 1).

The neighbor joining phylogenetic tree (NJT) showed a wide range of diversity with 30 different MLST’s observed amongst the 32 E. coli isolates ( Figure 2 ). Two clusters were identified, one of ST101 which included two isolates that were one allele different through cgMLST, and both were isolated from RM-FBO and both fully susceptible. The second cluster included two isolates of ST2539 isolated from RMD, both were indistinguishable to each other, and one was isolated from butter and the other from cream ( Figure 2 ).

4. Discussion

Although raw milk and raw milk products are appealing to some consumers as a nutritious alternative to heat treated products, the lack of a thermal treatment poses a risk. Raw milk is not included in the AMR EU surveillance programme under Commission Implementing Decision (EU) 2020/1729[70] and to the authors’ knowledge, no previous work has examined AMR, virulence factors, plasmid replicons or disinfectant-resistance genes in E. coli from Irish raw milk. Only limited food safety data exist from the past two decades for raw milk in Ireland; notably, our group reported shiga toxin-producing E. coli (STEC) in 14% of raw milk filters and 1% of raw milk in 2017-2019[71]. In addition, the Food Safety Authority of Ireland (FSAI) reported STEC in 6% of raw milk filters in 2012–2013[72], while a study conducted between 2001 and 2003 detected E. coli O157 in 12.5% of milk filters sampled[73].

The consumption of raw milk and raw-milk cheeses in Europe represents a recognised public-health concern due to frequent contamination with E. coli, including STEC and AMR strains[74,75,76,77]. European studies consistently report E. coli in raw milk at variable but often significant levels (approximately 2–22%), while raw milk cheeses, particularly artisanal products, may show even higher levels[74,75,77,78]. Importantly, a substantial proportion of isolates display AMR, with MDR phenotypes reported in 27-40% of strains and commonly resistant to β-lactams, tetracyclines, aminoglycosides, and fluoroquinolones[76,77,79,80]. The repeated detection of virulent and resistant E. coli in unpasteurised dairy products, together with documented STEC outbreaks linked to raw milk cheeses in several European countries, underscores the dual risk of foodborne infection and dissemination of AMR via the raw milk food chain[81,82].

Our study identified E. coli that were MDR, and contained virulence, pathogenic, and disinfectant-resistance genes in raw milk and raw milk products. This reinforces the need for maintaining high farm hygiene measures, as E. coli levels in milk are known to reflect cleaning efficacy of milking equipment, and bulk milk storage temperatures[63,83,84]. Approximately one fourth of the E. coli isolates in our study (27.7%) exhibited AMR and 21% of these were classified as MDR. To the best of the authors knowledge, there is no previous research on AMR in raw milk and their products in Ireland. These results could therefore serve as a baseline for future research in Ireland. Encouragingly no carbapenemase producing E. coli were detected. The lower proportion of resistant isolates in raw milk destined for direct consumption, compared with milk intended for pasteurisation, may reflect greater compliance with the relevant hygiene requirements laid down in Annexes I and II of Regulation (EC) 852/2004[85] as well as those set out in Annex III (Section IX) of Regulation (EC) No 853/2004[86].

The sequencing results of this study identified a variety of resistance determinants in a subset of the isolates, while the efflux pump genes acrF, mdtM, and emrE were detected either alone or in combination in all 32 isolates. AcrF, for which encoding genes were found in all but three of the sequenced isolates, is a member of the Resistance Nodulation and cell Division (RND) family and represents one of the seven known RND transporters in E. coli. Through the tripartite efflux AcrEF-TolC system[87], it contributes to the extrusion of a broad array of hydrophobic and amphiphilic compounds, including a wide range of antimicrobials and biocides, although non-functional variants have also been described[88]. Genes encoding EmrE, which extrudes hydrophobic compounds[89,90] and belongs to the Small Multidrug Resistant (SMR) family, and MdtM, a Major Facilitator Superfamily (MFS) antiporter that actively expels neutral and cationic compounds including QACs[91,92,93], were also found in more than half of the isolates. This further highlights the potential of raw milk isolates to expel structurally diverse toxic agents, including polyaromatic and cationic compounds commonly encountered in clinical and environmental contexts[94]. The fact that all sequenced isolates harboured efflux pump genes, regardless of their antimicrobial susceptibility profiles, seems to indicate that efflux pump genes presence alone could not explain variations in antibiotic susceptibility. The coexistence of qacEΔ1 with efflux systems identified in four isolates is expected to likely enhance survival under disinfectant pressure and raises concerns regarding co-selection between biocides and antimicrobials[30,95,96,97].

Previously reported research showed that efflux pump mechanisms are widespread in E. coli but do not necessarily confer QAC tolerance unless induced under specific conditions[98,99]. Moreover, recent experimental work demonstrated that E. coli may develop increased tolerance to benzalkonium chloride through adaptive mechanisms independent of integron encoded qac genes[100]. Our findings highlight the need for follow-up phenotypic assays, for example QAC specific MIC testing, time-kill studies, and biofilm tolerance assessments investigations on the E. coli isolates carrying qacEΔ1, to determine whether this gene contributes to measurable reductions in QAC susceptibility. This is because although qacEΔ1 is frequently embedded within class 1 integrons, which are mobile genetic elements (MGE), its association with phenotypic QAC tolerance has often been reported to be weak and inconsistent [101,102,103]. Genetic detection alone cannot reliably predict disinfectant resistance, underscoring the importance of phenotypic validation[104,105]. Genes conferring resistance to other biocides, such as chlorhexidine, hydrogen peroxide, ethanol, isopropanol, peracetic acid, povidone-iodine and sodium hypochlorite[106] were not identified in this present study.

The widespread detection of plasmid replicons in both resistant and fully susceptible E. coli indicates that plasmid carriage is common in dairy-associated commensal populations and is not solely driven by AMR[107]. The high prevalence of plasmid replicons in both resistant and susceptible isolates, particularly IncF plasmids such as IncFIB(AP001918) and IncFII, seems to indicate that these lineages may serve as stable hosts for MGEs capable of acquiring resistance genes under selective pressure, even in the absence of phenotypic resistance[107,108,109]. Notably, IncFIB(AP001918), which is frequently associated with ESBL-producing E. coli in human infections, was also detected in a substantial proportion of fully susceptible isolates, suggesting that clinically relevant plasmids can persist in the raw milk microbiota in the absence of detectable resistance phenotypes[68,110]. The predominance of IncF plasmid replicons, particularly IncFIB(AP001918) and IncFII, among AMR isolates supports their established role as major vectors for resistance gene dissemination in E. coli[109]. Collectively, these findings underscore the importance of including plasmid surveillance in microbiological assessments of raw milk and raw milk products, as such foods may contribute to the silent maintenance and spread of resistance associated plasmids along the food chain[108]. IncL/M plasmid replicons, common in European outbreaks of ESBL [111] and NDM-1 carbapenemase producers [112,113], were not detected. However, the reference based approach using PlasmidFinder which depends on a precompiled database struggles to capture large plasmids greater than 50 kb[114] and short-read sequencing in this present study limited full plasmid reconstruction and characterisation of plasmid structures[115,116]. Chromosomal fluoroquinolone resistance was confirmed through mutations in gyrA and parC, which reduce susceptibility and can increase efflux pump expression[117,118,119]. Among the four fluoroquinolone-resistant isolates in this study, three encoded all three efflux pumps identified, while one encoded AcrF alone.

The co-occurrence of chromosomal AMR determinants, efflux pumps, plasmid replicons, virulence factors, and biocide-resistance genes in raw milk isolates reflects strong and overlapping selective pressures within dairy production environments[120,121]. Antimicrobial use in dairy cattle, particularly for the treatment of mastitis, can enrich intrinsic resistance and efflux-mediated tolerance, contributing to the selection of resistant bacterial populations[122]. Manure and slurry can further disseminate resistant bacteria and antimicrobial residues sustaining environmental selection and contaminating soil and water[123,124,125]. Routine exposure to biocides, heavy metals, and other environmental contaminants can promote co-selection through efflux activation and membrane adaptation[30]. Environmental pathways, including contaminated water, biofilms on milking surfaces, and movement of animals or personnel, facilitate ongoing circulation and persistence of resistant strains[126]. Collectively, these pressures allow chromosomal resistance traits to persist and spread independently of plasmid mediated exchange, highlighting the need for an integrated approach towards antimicrobial use, rigid hygiene measures at both farm and milk processing levels, and environmental management practices given that soil antibiotic resistomes are widespread and connected to human and environmental AMR reservoirs[127,128,129].

This present study revealed ST10 to be the most prevalent lineage, with all ST10 isolates belonging to phylogroup A, a human-associated lineage that has been linked to the dissemination of AMR via plasmids and other mobile genetic elements[107]. ST58 was the second most common and belonged to phylogroup B1, an emerging zoonotic lineage increasingly associated with MDR, ESBL production, and extraintestinal infections[68,109,110,130]. Among the isolates, one ST10 was MDR and fluoroquinolone resistant, while all three ST58 isolates were resistant to at least three antimicrobial classes. Apart from one fluoroquinolone isolate (ST10) that was recovered from cream, all remaining ST10 and ST58 isolates were obtained from raw milk, indicating that these lineages may serve as vehicles for resistance determinants along the dairy supply chain[68,109,130]. Additional resistant lineages included an AmpC β-lactamase-producing ST88 isolate of phylogroup C, an ESBL-producing ST1126 of phylogroup B1, and three more fluoroquinolone resistant isolates within phylogroups C (two isolates), and B1, demonstrating that AMR in raw milk-associated E. coli occurs across multiple phylogenetic backgrounds[68,109,110,130]. While no direct evidence of transmission to humans was assessed in this study, the presence of potentially high-risk lineages highlights the importance of continued surveillance of the dairy food chain within a One Health framework, encompassing human, animal, and environmental reservoirs to better understand and mitigate dissemination of strains of public health relevance.

Amongst the virulence genes identified, some were associated with functions including stress, survival, regulation, iron uptake, type-three secretion systems, invasion, adherence, colonisation and toxin production. The identification of major virulence determinants in the sequenced E. coli isolates highlights virulence profiles with public health relevance. Toxins such as Shiga-toxin are strongly linked to haemorrhagic colitis and haemolytic uremic syndrome, representing one of the most critical food and waterborne threats[131,132]. Likewise, cdtABC, hlyA, and cnf1 can contribute to epithelial damage, systemic inflammation, and severe extraintestinal infections, including urosepsis and neonatal meningitis[133,134,135] while it has also recently been reported that CNF1 promotes colorectal carcinogenesis[136]. Iron-acquisition systems including enterobactin, salmochelin, aerobactin, and heme-uptake pathways are key fitness determinants in host environments and are strongly associated with high-risk extraintestinal pathogenic E. coli lineages[137]. Adherence factors type 1 and P fimbriae enhance colonisation of the urinary and gastrointestinal tracts and facilitate progression to invasive disease[138,139]. In this study 30 isolates carried the fimA to fimH type 1 fimbrial genes, which encode adhesin proteins that are key contributors to biofilm formation and bacterial adherence[140]. The coexistence of these elements may enhance the ability of strains to persist in the environment, adapt to different hosts and contribute to the spread of AMR, highlighting the interconnected risks across human, animal and environmental health. This co-existence could not only limit therapeutic options but also increase the likelihood of horizontal gene transfer, environmental persistence, and foodborne transmission, underscoring the need for a comprehensive, up to date surveillance across the Irish dairy sector[141].

In conclusion, to the best of our knowledge, this is the first study to characterize AMR, STs, virulence, plasmid replicons, and disinfectant-resistance determinants in E. coli from raw milk and raw milk products in the Republic of Ireland. Although the overall prevalence of AMR was moderate, the detection of MDR isolates, ESBL/AmpC producers, high-risk sequence types (ST10, ST58), and Shiga-toxin genes indicate that raw milk can act as a vehicle for transmission of clinically significant E. coli through the food chain. The identification of isolates harbouring genes encoding for QAC-resistance highlights the potential selective effects of biocide use, which is not routinely monitored. Strengthening surveillance efforts, optimising hygiene practices and ensuring prudent use of antibiotics and biocides are therefore recommended to mitigate the emergence and spread of resistant and pathogenic E. coli in the dairy processing environment.