This study involved fecal sampling of farm animals and farm workers. The farm owner provided verbal permission for the collection of fecal samples. Ethical approval was not required for animal sampling because only fecal samples were collected after defecation from the ground, without any direct contact with the animals.

The protocol for human sampling was approved by the Naresuan University Institutional Review Board (COA No. 83/2013), and written informed consent was obtained from all farm workers prior to their participation in the study.

All procedures involving 3GC-R E. coli were conducted in a biosafety level 2 laboratory and were performed in accordance with standard biosafety guidelines and institutional regulations.

This study was conducted between March and April 2016 on a small-scale local farm in Phitsanulok province (16°49′N, 100°15′E), located approximately 377 km north of Bangkok, a region characterized by high livestock farming density. The farm, situated 30 km from the city center, raises three types of livestock, including laying hens, swine, and cattle, with approximate population sizes of 2,000, 100, and 80 animals, respectively.

Laying hens were reared in rows of battery cages, with three hens housed per cage. Swine were housed in small groups of four to five animals per pen with concrete flooring. Cattle were raised in a large outdoor pen with a soil surface. All animals were raised in close proximity to one another within a distance of approximately 200 m and were managed by three farm workers.

Standardized husbandry practices were applied to all animals throughout the rearing period. Commercial feed and groundwater were provided. Farm workers entered the animal areas twice daily to feed and monitor the animals. Biosecurity measures were not consistently or strictly applied; however, farm workers changed their boots before entering and after leaving the animal areas. Wastewater generated from the farm was collected in a sedimentation tank to remove large solids, after which the effluent was discharged into the environment without further treatment. Sludge and animal manure were directly applied as fertilizers.

Sample size estimation for laying hens was performed using the single-population proportion formula, assuming an expected prevalence of 27% based on a previous study [15], a margin of error of 5%, and a 95% confidence interval (CI). Calculations were conducted using Epitools (https://epitools.ausvet.com.au/).

One fecal sample was collected from each swine pen, whereas fecal samples from cattle were randomly collected. The final sample size varied depending on the availability of fecal material at the time of collection. Three farm workers agreed to provide fecal samples. In total, 265 fecal samples were collected, comprising samples from laying hens (n = 210), cattle (n = 33), swine (n = 19), and farm workers (n = 3) (Table 1).

Fresh fecal samples were randomly collected from animals using sterile spoons and sterile containers. For laying hens, voided feces were collected from beneath each cage, whereas fecal samples from swine and cattle were collected from the ground. The portion of feces that had not touched the ground was carefully selected and collected without disturbing the animals.

Fecal samples from the three farm workers were collected using sterile Amies swabs (Deltalab, Barcelona, Spain). All samples were transported to the laboratory on ice and processed within 6 h of collection.

A total of 265 fecal samples, including samples from laying hens (n = 210), cattle (n = 33), swine (n = 19), and farm workers (n = 3), were collected. For animal fecal samples, 1 g of fecal material was suspended in 9 mL of tryptic soy broth (Oxoid, Basingstoke, UK) and incubated overnight at 37°C. Following incubation, the enriched cultures were streaked onto eosin methylene blue (EMB) agar (Oxoid) supplemented with cefotaxime (2 µg/mL) (EMB-CTX). Fecal swabs obtained from farm workers were directly inoculated onto EMB-CTX agar. All inoculated plates were incubated at 37°C for 24 h. From each sample, one presumptive E. coli colony displaying a metallic green sheen was selected. Species identification was confirmed using the RapID™ ONE System (REMEL Inc., KS, USA).

Antibiotic susceptibility testing was performed against 12 antimicrobial agents using the disk diffusion method in accordance with Clinical and Laboratory Standards Institute (CLSI) guidelines [17]. Freshly grown colonies were suspended in normal saline, and the turbidity was adjusted to 0.5 McFarland standard (1.5 × 10⁸CFU/mL). The bacterial suspension was spread evenly onto Mueller–Hinton agar (Oxoid, UK) using a sterile cotton swab. Antibiotic disks were placed on the agar surface, and plates were incubated at 37°C for 16–18 h.

Inhibition zones were measured manually using standardized rulers and interpreted as susceptible, intermediate, or resistant according to CLSI criteria [17]. Intermediate results were considered indicative of susceptibility. Isolates exhibiting resistance to at least three antimicrobial classes were classified as multidrug-resistant (MDR) [18].

The tested antibiotics included seven β-lactams; amoxicillin/clavulanate (20/10 µg), ceftazidime (30 µg), cefpodoxime (10 µg), cefotaxime (30 µg), cefepime (30 µg), aztreonam (30 µg), and imipenem (10 µg), and five non-β-lactams; chloramphenicol (30 µg), gentamicin (10 µg), ciprofloxacin (5 µg), doxycycline (30 µg), and trimethoprim–sulfamethoxazole (1.25/23.75 µg) (Oxoid). E. coli ATCC (American Type Culture Collection) 25922 was used as the quality control strain. All antibiotic disks were stored in sealed containers in the dark at 2°C–4°C prior to use, according to the manufacturer’s instructions.

MICs for transconjugants were determined using the broth microdilution method following CLSI guidelines [17]. Two-fold serial dilutions of each antibiotic were prepared in 200 µL of cation-adjusted Mueller–Hinton broth (Oxoid) in 96-well microtiter plates. Bacterial suspensions were inoculated into each well at a final concentration of 5 × 10⁵ CFU/mL. Plates were incubated at 37°C for 18–20 h. E. coli ATCC 25922 was included as a control. MICs were defined as the lowest concentration of an antibiotic that inhibited visible bacterial growth.

Phenotypic detection of ESBL production was performed using the combination disk method according to CLSI guidelines [17]. Cefotaxime (30 µg) and ceftazidime (30 µg), alone and in combination with clavulanic acid (30/10 µg), were used (Becton, Dickinson and Company, MD, USA). ESBL production was defined as an increase of ≥5 mm in the inhibition zone diameter for antibiotic–clavulanate disks compared with the corresponding antibiotic disks without clavulanic acid. E. coli ATCC 25922 and a known ESBL-producing E. coli strain were used as negative and positive controls, respectively [15].

Detection of genes encoding β-lactamases (bla_CTX-M and bla_CMY-2) was performed by polymerase chain reaction (PCR) using previously published primers and conditions [19, 20]. The bla_CTX-M alleles were identified using group-specific primers for group 1 and group 9. Full-length amplification of bla_CMY-2-like genes was performed using previously described primers.

Individual colonies were suspended in 300 µL of distilled water and used as PCR templates. Amplification reactions were performed in a total volume of 20 µL containing 1 µL of template DNA, 1× PCR buffer, 1.5 mM MgCl₂, 0.2 mM dNTPs, 0.5 µM of each primer, and 1 U of Taq polymerase (Vivantis Technologies, Selangor, Malaysia). PCR amplification was carried out using a Veriti™ 96-well thermal cycler (Applied Biosystems, CA, USA) under the specified cycling conditions. Distilled water and previously confirmed bla_CTX-M/bla_CMY-2-positive E. coli strains were included as negative and positive controls, respectively [14].

PCR products were analyzed by 1% agarose gel electrophoresis. Selected PCR products were purified using a DNA purification kit (RBC Bioscience, New Taipei City, Taiwan) and submitted for sequencing (First BASE Laboratories, Selangor, Malaysia). Nucleotide sequences were compared with reference sequences in the GenBank database using the Basic Local Alignment Search Tool algorithm available through the National Center for Biotechnology Information.

The transferability of bla_CTX-M and bla_CMY-2 was assessed using the broth mating method with sodium azide-resistant E. coli J53 as the recipient strain, as previously described [15]. Donor and recipient cultures were mixed at a 1:1 ratio and incubated overnight at 37°C without shaking. Serial dilutions of the mating mixture were plated on tryptic soy agar containing cefotaxime alone, sodium azide alone, or a combination of cefotaxime and sodium azide for selection of donor, recipient, and transconjugant colonies, respectively. Positive control strains carrying bla_CTX-M or bla_CMY-2 from previous studies were included [14]. All plates were incubated at 37°C overnight. Conjugation frequency was calculated as the number of transconjugants (CFU/mL) divided by the number of donor cells (CFU/mL). Transfer of resistance genes was confirmed by PCR. MICs of transconjugants were determined using the broth microdilution method according to CLSI guidelines [17]. Plasmid replicon types were identified by PCR-based replicon typing (PBRT) [21], using primers and conditions provided in Supplementary Table S1.

The contribution of efflux pumps to ceftazidime resistance was evaluated by determining MICs with and without the EPI PAβN (25 µg/mL) (Sigma-Aldrich Inc., MO, USA). A previous study conducted in our laboratory demonstrated that PAβN at this concentration did not substantially affect E. coli growth. An MDR P. aeruginosa strain with known efflux pump overexpression was included as a positive control. All assays were incubated at 37°C for 18 h. A reduction of at least four-fold in ceftazidime MIC in the presence of PAβN was considered indicative of efflux pump involvement in antibiotic resistance [10].

Genetic relationships among 3GC-R E. coli isolates were examined by PFGE following established protocols [15]. Bacterial cultures were adjusted to a turbidity of 1.3–1.4 at an optical density of 625 nm, embedded in agarose plugs, digested with XbaI, and separated using a CHEF Mapper® XA system (Bio-Rad Laboratories, USA). Chromosome DNA from Saccharomyces cerevisiae was used as a molecular size marker. PFGE patterns were analyzed by visual inspection and interpreted according to previously described criteria [22].

Descriptive statistics, including prevalence and AMR rates, were analyzed using Microsoft Excel. Comparisons of 3GC-R E. coli prevalence and antibiotic resistance rates among animal species were performed using Fisher’s exact test. Statistical analyses were conducted using Minitab version 15, and differences were considered statistically significant at p < 0.05.

In total, 265 fecal samples were collected from laying hens (n = 210), cattle (n = 33), swine (n = 19), and farm workers (n = 3), among which 42 samples (15.8%) yielded 3GC-R E. coli (Tables 1 and S2). The proportions of samples positive for 3GC-R E. coli were highest in swine (47.4%), followed by cattle (27.3%) and laying hens (10.5%). The prevalence of 3GC-R E. coli was significantly higher in swine than in laying hens (p < 0.001) (Figure 1). A similar significant difference was observed when comparing cattle and laying hens (p = 0.02). In addition, fecal samples from two of the three farm workers tested positive for 3GC-R E. coli.

Assessment of the antibiotic susceptibility of 3GC-R E. coli against 12 antibiotics showed high resistance rates to β-lactam agents. Resistance to cefotaxime and cefpodoxime was detected in 100% of isolates, whereas resistance to ceftazidime and aztreonam was observed in 73.8% and 69.0% of isolates, respectively. Resistance to amoxicillin–clavulanate and cefepime ranged from 35.7% to 38.1% (Table 2).

Among non-β-lactam antibiotics, resistance to chloramphenicol, gentamicin, doxycycline, and trimethoprim–sulfamethoxazole ranged from 40.5% to 52.4%. Although resistance to ciprofloxacin was detected in 11.9% of isolates, all isolates remained susceptible to imipenem. Notably, isolates recovered from swine and cattle exhibited significantly higher resistance rates to chloramphenicol, gentamicin, and doxycycline compared with those from laying hens (Figure 2).

Most isolates (52.4%) exhibited resistance to 4–6 antibiotics, while 40.5% of isolates were resistant to 7–9 antibiotics (Table 2). A total of 24 distinct resistance profiles were identified. The most frequent resistance pattern was amoxicillin/clavulanate–ceftazidime–cefpodoxime–cefotaxime, which was detected in 14.3% of isolates (Table S3). Overall, 24 isolates (57.1%) were resistant to at least three antibiotic classes and were therefore classified as MDR. The largest proportion of isolates (21.4%) showed resistance to antibiotics belonging to four antimicrobial classes (Table 2).

Screening for β-lactamase genes demonstrated that all 3GC-R E. coli isolates harbored bla_CTX-M, bla_CMY-2, or both (Tables 1 and S2). Twenty-nine isolates (69.0%) carried bla_CTX-M alone, including bla_CTX-M-55 (n = 21), bla_CTX-M-14 (n = 5), and bla_CTX-M-15 (n = 3). Nine isolates (21.4%) simultaneously carried bla_CTX-M and bla_CMY-2. Among the 38 bla_CTX-M-positive isolates, 28 (66.7%) were confirmed as ESBL producers (Tables 1 and S2). In addition, four isolates (9.5%) carried bla_CMY-2 alone.

Conjugation experiments were performed using 23 3GC-R E. coli isolates as donors, including 19 bla_CTX-M positive and four bla_CMY-2 positive isolates. Of these, 22 isolates—comprising 11 bla_CTX-M-55, five bla_CTX-M-14, two bla_CTX-M-15, and four bla_CMY-2, successfully transferred their resistance genes to the recipient E. coli J53. Conjugation frequencies ranged from 4.1 × 10^-8 to 7.2 × 10^-2 (Tables 3 and S4).

For transconjugants carrying either bla_CTX-M or bla_CMY-2, cefotaxime MIC values were increased by up to >32-fold compared with the recipient strain (E. coli J53; cefotaxime MIC = 0.25 µg/mL). Among these transconjugants, 13 were successfully characterized by PBRT. Detected plasmid replicon types included IncF (n = 5), IncFIA (n = 1), IncFIB (n = 3), and IncI1-Iγ (n = 4) in transconjugants carrying bla_CTX-M and bla_CMY-2 (Table 3).

To assess the role of efflux pumps in ceftazidime resistance, 31 3GC-R E. coli isolates resistant to ceftazidime by disk diffusion were further examined. According to CLSI guidelines [17], E. coli isolates with ceftazidime MICs of 8 µg/mL and ≥16 µg/mL are categorized as intermediate and fully resistant, respectively. Accordingly, ceftazidime MIC values ranged from 16 to 128 µg/mL in 29 isolates, while two isolates exhibited MICs of 8 µg/mL (Table S2).

In the presence of PAβN, four-fold and eight-fold reductions in ceftazidime MICs were observed in 11 and one isolates, respectively (Table 4). Inhibition of efflux pump activity by PAβN in a single isolate from cattle reduced the MIC from 8 µg/mL to 2 µg/mL (Table S2), restoring susceptibility to ceftazidime.

Genotypic analysis of 3GC-R E. coli isolates identified 23 distinct PFGE profiles (I–XXIII) among 41 isolates (Figure 3 and Table S2). One isolate from a farm worker could not be typed. Identical PFGE profiles were detected among isolates from laying hens (profiles I and II), swine (profile V), and cattle (profile VI). In addition, similarities were observed among isolates recovered from different animal species, with E. coli isolates from laying hens and swine sharing profiles III and IV, and isolates from swine and cattle sharing profile VI.

The inappropriate use of antibiotics is a major driver of the emergence of ARB in livestock production systems. Previous studies conducted in low- and middle-income countries, particularly those focusing on small-scale farms, have reported a high prevalence of 3GC-R E. coli in farm animals, including chickens in Indonesia, Ecuador, and Zambia (66.1%–77.8%) [23–25], swine in Vietnam and Argentina (47.8%–89.0%) [26, 27], and cattle in Tanzania and Argentina (34.8%–95.0%) [26, 28]. Similarly, a recent study from small- and middle-scale farms in Peru reported ESBL-producing E. coli prevalences of 75.4% in dairy cows, 61.3% in pigs, and 34.1% in laying hens [29].

Overall, 3GC-R E. coli was detected in 15.8% of fecal samples collected from farm animals and workers (Table 1). This prevalence is lower than that reported in previous studies of food-producing animals raised on small-scale farms in Thailand (27.1%–55.6%) [15,16]. Differences in detection rates may be attributed to variations in study location, sample collection strategies, and isolation methodologies, all of which can influence prevalence estimates.

The highest prevalence of 3GC-R E. coli was observed in swine (47.4%), followed by cattle (27.3%) and laying hens (10.5%) (Table 1). This distribution is consistent with reports from other countries, including Laos, Peru, and Argentina [26, 29, 30]. The higher prevalence observed in swine may be explained by the fact that pig production systems worldwide are associated with greater antimicrobial use than other livestock sectors, leading to increased antibiotic exposure and selective pressure for resistant strains [31].

In addition to livestock, 3GC-R E. coli was detected in the intestinal tracts of two of the three farm workers. This finding is consistent with previous reports demonstrating a high rate of 3GC-R E. coli carriage among apparently healthy individuals in this region [15], underscoring the potential occupational exposure risk in small-scale farming environments.

Antimicrobial susceptibility testing revealed concerning resistance patterns. At least half of the isolates were resistant to third-generation cephalosporins (cefpodoxime and ceftazidime), aztreonam, doxycycline, gentamicin, and trimethoprim–sulfamethoxazole (Table 2), all of which are considered critically important antibiotics for the treatment of bacterial infections. Moreover, 57.1% of isolates were classified as MDR, indicating a substantial risk of treatment failure should infections occur.

The presence of ESBL-producing E. coli in livestock represents a significant public health concern, particularly in Africa and Asia [32]. Among the 42 3GC-R E. coli isolates examined in this study, bla_CTX-M was predominantly detected either alone (n = 29) or in combination with bla_CMY-2 (n = 9), while ESBL production was phenotypically confirmed in 28 isolates (Table 1). The absence of ESBL activity in ten bla_CTX-M positive isolates may be attributed to low-level gene expression [33]. In addition, four isolates carried bla_CMY-2 alone.

The detection of bla_CTX-M-14, bla_CTX-M-15, bla_CTX-M-55, and bla_CMY-2 in farm animals aligns with findings from multiple regions worldwide [34–36], as these enzymes represent the most prevalent ESBL and plasmid-mediated AmpC β-lactamases globally [4, 37].

In Thailand, pAmpC-producing E. coli have been frequently reported in human clinical samples [37,38], whereas reports from farm animals remain limited. Only a small number of bla_CMY-2 positive E. coli isolates have previously been identified in domestic and wild animals, including pigs [38]. The presence of bla_CMY-2 alone or in combination with bla_CTX-M in farm animals in the present study (Table 1) suggests that livestock may serve as an underrecognized reservoir for pAmpC-positive 3GC-R E. coli in this region.

Conjugation experiments demonstrated that bla_CTX-M and bla_CMY-2 genes carried by 3GC-R E. coli isolates were transferable, resulting in cefotaxime MIC increases of up to >32-fold in transconjugants. These genes were associated with IncF and IncI1-Iγ plasmids, respectively (Table 2), both of which are well known for their capacity to disseminate among Enterobacterales.

IncF plasmids are the most frequently reported plasmids in E. coli and commonly harbor multiple bla_CTX-M variants. Previous studies have shown that conjugative transfer of bla_CTX-M is predominantly mediated by IncF plasmids [39], consistent with findings from backyard poultry in this region [15]. Similarly, two bla_CMY-2 genes in this study were located on IncI1-Iγ plasmids, which are narrow-host-range conjugative plasmids frequently detected in food animal–associated Enterobacterales [40]. On these plasmids, bla_CMY-2 is often embedded within insertion sequences, particularly ISEcp1, facilitating gene mobilization.

The detection of transferable bla_CMY-2 on IncI1-Iγ plasmids in farm animal isolates provides new evidence of regional plasmid dissemination potential. These plasmids often carry additional resistance determinants, including those conferring resistance to fluoroquinolones and aminoglycosides, contributing to the multidrug-resistant phenotype commonly observed in 3GC-R E. coli [41]. The presence of such transferable resistance genes in rural farming environments represents a significant public health concern due to frequent human–animal–environment interactions.

Efflux pump overexpression is an important contributor to resistance against multiple antibiotic classes. In this study, the contribution of efflux pumps to ceftazidime resistance was assessed using PAβN, a well-characterized inhibitor of RND efflux pumps. PAβN has been shown to reduce MICs of third-generation cephalosporins and carbapenems in several Gram-negative bacteria, including E. coli [42], Klebsiella pneumoniae [11], and P. aeruginosa [10].

A ≥4-fold reduction in ceftazidime MICs in the presence of PAβN was observed in 38.7% of the isolates (Table 4), indicating a contributory role of RND efflux pumps in ceftazidime resistance. Complete restoration of susceptibility in one isolate further supports efflux-mediated resistance. However, among the 12 isolates showing efflux pump overexpression, only one exhibited an MDR phenotype (Table S2), consistent with previous reports suggesting that efflux pump overexpression alone does not necessarily confer MDR. Instead, efflux-mediated resistance may progress to MDR following additional mutational events [43]. A minor (two-fold) reduction in MICs observed in 41.9% of isolates further suggests partial efflux involvement.

Transmission of ARB between livestock and farm workers has been documented in Japan, Vietnam, and Thailand [15, 27, 36]. Although high genetic diversity was observed among 3GC-R E. coli isolates in the present study, identical PFGE profiles were detected within and between animal species (Figure 3, Table S1), indicating that clonal dissemination can occur among animals raised in close proximity. Cross-species sharing of identical PFGE patterns between laying hens and swine and between swine and cattle supports local micro-ecological transmission rather than purely vertical spread.

Despite frequent interactions between animals and workers, clonal transmission between livestock and farm workers was not observed. Given the lack of direct contact among animal species, environmental dissemination via drainage water or manure is a plausible transmission route [44]. The absence of direct animal-to-human clonal spread in this study contrasts with previous reports [15, 36] and suggests that horizontal gene transfer via highly transferable plasmids carrying bla_CTX-M and bla_CMY-2 (Table 3) may play a more dominant role in resistance dissemination.

The WHO and other international agencies have developed global strategies to mitigate AMR. Thailand has implemented two National Action Plans on AMR (NAP-AMR 2017–2021 and 2023–2027) [45, 46], which emphasize reducing AMR-related morbidity and mortality across human, animal, and environmental sectors. Surveillance of AMR bacteria, including 3GC-R E. coli, within a One Health framework is a core strategy of both plans.

Small-scale family-owned farms dominate livestock production in Thailand and are essential to the national economy. However, close human–animal contact in these systems increases zoonotic risk. The findings of this study strongly support the NAP-AMR strategy advocating expanded One Health surveillance of 3GC-R E. coli across humans, animals, and the environment in smallholder farming systems.

This study has several limitations. Sampling was limited to a single farm with a relatively small sample size. Environmental samples, including feed, manure, water, and surrounding soil, were not collected, which may have provided further insight into transmission pathways. Longitudinal sampling was not performed, preventing assessment of temporal trends in 3GC-R E. coli prevalence. In addition, data on antibiotic usage within the farm were unavailable, limiting the ability to identify selective pressures driving resistance emergence.

Future studies should incorporate whole-genome sequencing to characterize mobile genetic elements and integrons associated with bla genes. Expanded genomic surveillance across humans, animals, and environmental sources on multiple smallholder farms is warranted. Strengthening biosecurity practices and maintaining antibiotic usage records at the farm-level are critical steps toward controlling the emergence and dissemination of ARB.