The use, handling, storage, and laboratory investigation of E. coli isolates in this study were reviewed and approved by the Faculty of Veterinary Science Biosafety Committee (CU-VET-BC), Chulalongkorn University, Thailand, under Institutional Biosafety Committee (IBC) approval No. 2331030. This investigation used 333 previously collected E. coli isolates from the culture collection of the Department of Veterinary Public Health, Faculty of Veterinary Science, Chulalongkorn University, comprising isolates originating from fish (n = 251) and cultivation water (n = 82).

No animals were newly recruited, handled, sampled, or subjected to experimental procedures specifically for the present study. The bacterial isolates originated from samples collected in a previous study involving hybrid red tilapia raised in open-cage-culture systems. The original animal sampling protocol had been reviewed and approved by the Chulalongkorn University Animal Care and Use Committee (IACUC; Approval No. 2031048). Accordingly, the present study constituted laboratory-based analysis of archived bacterial isolates under the applicable institutional biosafety approval. All procedures involving bacterial cultures were conducted in accordance with the approved institutional biosafety requirements and laboratory procedures of Chulalongkorn University.

This study was conducted from January to October 2021 at tilapia farms along the Kwae Noi River in Kanchanaburi, Thailand,

A total of 333 E. coli isolates were obtained from the culture collection of the Department of Veterinary Public Health, Faculty of Veterinary Science, Chulalongkorn University. The isolates originated from two primary sources: fish samples (n = 251), comprising fish meat (n = 22), intestines (n = 42), liver and kidney (n = 99), and fish carcass rinses (n = 88), and cultivation water samples (n = 82). We collected samples between October 2019 and November 2020 from hybrid red tilapia raised in open-cage-culture systems. Farms were purposively selected based on production system, accessibility, and the availability of market-sized fish at the time of sampling. To minimize clonal duplication and ensure genomic diversity, we selected only one E. coli isolate from each sample, in accordance with the Food and Agriculture Organization guidelines for AMR surveillance in food animals [27]. The AMR profiles of these isolates have been reported previously using the agar dilution method [28].

All isolates (n = 333) were evaluated for susceptibility to disinfectants and heavy metals by determining MICs using the agar dilution method with two-fold serial dilutions [29]. Because no standardized Clinical and Laboratory Standards Institute (CLSI) or European Committee on Antimicrobial Susceptibility Testing (EUCAST) methods or interpretive breakpoints are available for biocides and heavy metals, we adapted the procedure from the general MIC testing principles described in CLSI VET01, and selected concentration ranges based on previously published MIC distributions [29, 30].

Heavy metal stock solutions were prepared in sterile deionized water, completely dissolved, and inspected for precipitation before use. Each isolate was tested in triplicate. Five disinfectants commonly used as active ingredients in household products were evaluated: benzalkonium chloride (BKC; 0.5–1,024 µg/mL) (Sigma, St. Louis, MO, USA), CHX (0.5–1,024 µg/mL) (Sigma), GLU (256–4,096 µg/mL) (Sigma), nitrilotriacetic acid (NTA; 256–4,096 µg/mL) (Sigma), and triclosan (TCS; 0.002–8 µg/mL) (Sigma). Three heavy metals were also evaluated: AgNO₃ (0.5–128 µg/mL), CuSO₄ (256–4,096 µg/mL), and ZnCl₂ (256–4,096 µg/mL) (Sigma).

Reference strains included Staphylococcus aureus American Type Culture Collection (ATCC) 29213, E. coli ATCC 25922, and Pseudomonas aeruginosa ATCC 27853. We compared the MICs obtained for the quality-control strains with previously published reference MIC ranges to verify assay validity and reproducibility [30].

ECOFFs were determined for agents lacking established values in aquatic E. coli using ECOFF Finder software (version 2010-v2.1) [31], developed by EUCAST. We defined the ECOFF for each agent as the 97.5th percentile of the MIC distribution. MIC frequency data generated from the two-fold dilution series were entered into the software, which iteratively fitted a log-normal distribution to the presumed wild-type population. We evaluated the reliability of the 97.5th-percentile estimate by comparing the observed and modeled MIC distributions, and we rounded the calculated ECOFF up to the next tested two-fold dilution.

All E. coli isolates were evaluated for biofilm formation using the tissue culture plate assay (TCPA) [32]. Pure cultures were revived on plate count agar (PCA; Difco), and two colonies from each isolate were inoculated into 3 mL of tryptic soy broth (TSB; Difco) and incubated at 37°C with shaking at 120 rpm for 16 ± 1 h. Cultures were adjusted to 1 × 10⁶ Colony-forming unit/mL, and 200 µL aliquots were transferred into sterile 96-well microplates and incubated statically at 37°C. Each isolate was tested in three technical replicates to ensure assay reproducibility. E. coli ATCC 25922 and sterile TSB served as the positive and negative controls, respectively.

Biofilm formation was quantified using the crystal violet (CV) assay [32]. After incubation, the wells were washed, air-dried for 4–6 h, fixed with 2% sodium acetate for 15 min, stained with 1% CV for 10 min, washed again, and air-dried. The bound stain was solubilized with 30% acetic acid, and the optical density (OD) was measured at 550 nm.

Biofilm production was classified according to the method described by Stepanović et al. [33] by comparing the OD of each isolate with the negative-control cut-off value (ODc):

1. OD ≤ ODc = No biofilm production

2. ODc < OD ≤ 2 × ODc = Weak biofilm production

3. 2 × ODc < OD ≤ 4 × ODc = Moderate biofilm production

4. OD > 4 × ODc = Strong biofilm production

This study targeted disinfectant resistance genes associated with efflux pump mechanisms in E. coli, including chromosomally encoded genes (sugE(c), emrE, mdfA, and ydgE) and the plasmid-encoded gene sugE(p). In addition, we screened for small multidrug resistance family efflux pump genes (qacE, qacEΔ1, qacF, and qacG), which are frequently detected in Gram-negative bacteria and are associated with resistance to commonly used disinfectants. We also investigated five heavy metal resistance genes corresponding to the evaluated phenotypes, namely silA, silE, copA, pcoA, and zntB (Supplementary Table S1).

Genomic DNA was extracted using the whole-cell boiling method [34]. Polymerase chain reaction (PCR) was performed in a final reaction volume of 25 µL containing AllTaq DNA PCR Master Mix (Thermo Fisher Scientific, MA, USA), 5 µL of DNA template, 2 µL of each primer, and 16 µL of distilled water. PCR amplification consisted of an initial denaturation at 94°C for 5 min, followed by 30 amplification cycles, each comprising denaturation at 94°C for 30 s, annealing at the gene-specific temperature (Supplementary Table S1), and extension at 72°C for 30 s, followed by a final extension at 72°C for 7 min. PCR products were separated on 1.5% agarose gels stained with RedSafe™, visualized using an Omega Fluor™ imaging system (Aplegen), and compared with a 100-bp DNA ladder.

Descriptive statistics were used to summarize biofilm production and the phenotypic and genotypic resistance profiles of the 333 E. coli isolates recovered from paired fish and water samples. Pearson's chi-square test was used to compare biofilm production between isolates recovered from fish and cultivation water.

Principal component analysis was performed to cluster isolates by resistance gene profiles. Multivariable logistic regression analysis was conducted to identify factors associated with the presence of int1, a marker of classical and non-classical integron-mediated multidrug resistance. Predictor variables included AMR, disinfectant resistance, and biofilm formation level. We selected candidate variables based on biological relevance and significant univariable associations. We evaluated multicollinearity using variance inflation factors and assessed model performance using the Hosmer-Lemeshow goodness-of-fit test. Statistical significance was defined as p < 0.05. All statistical analyses were performed using Stata version 18.0 (StataCorp, College Station, TX, USA).

The MIC distributions and proposed ECOFFs are presented in Table 1. A bimodal TCS MIC distribution was observed among tilapia-derived E. coli isolates, particularly those recovered from cultivation water, enabling the proposal of an ECOFF of 0.004 µg/mL. In contrast, the other disinfectants exhibited unimodal MIC distributions (Supplementary Figure S1). BKC MICs ranged from 4 to 256 µg/mL in isolates from both sources, whereas CHX MICs were predominantly 2–32 µg/mL in fish isolates and 4–16 µg/mL in water isolates. GLU and nitrilotriacetic acid (NTA) exhibited high MICs ranging from 1,024 µg/mL to >4,096 µg/mL.

All tested heavy metals exhibited narrow MIC distributions. The lowest MIC₅₀ was observed for AgNO₃ (16 µg/mL), whereas higher MIC₅₀ values were recorded for CuSO₄ (2,048 µg/mL) and ZnCl₂ (1,024 µg/mL). Overall, the MIC distributions were similar between fish and water isolates, with a significant difference observed only for AgNO₃.

Most E. coli isolates (87.7%) exhibited weak biofilm production (Table 2). Strong biofilm production was uncommon (4.5%) and was detected predominantly among isolates recovered from cultivation water (2.4%), followed by intestinal samples (1.5%) and fish carcass rinses (0.6%). Biofilm production differed significantly between isolates recovered from fish and cultivation water (χ² = 5.9, p = 0.015). Isolates recovered from cultivation water had approximately twice the odds of biofilm formation compared with those recovered from fish (Odds ratio [OR] = 1.89; 95% CI = 1.13–3.16).

The most frequently detected disinfectant resistance genes were ydgE (97.9%), followed by mdfA (92.8%), sugE(c) (66.7%), and emrE (50.5%) (Figure 1). In contrast, qacF (8.1%), qacE (5.1%), and qacEΔ1 (3.9%) were detected at low frequencies, whereas qacG was not detected. None of the five tested heavy metal resistance genes were detected in the E. coli isolates. Disinfectant and heavy metal resistance genes previously reported in E. coli from aquaculture and terrestrial production systems are summarized in Supplementary Table S2.

Clustering based on the presence or absence of resistance genes identified four distinct clusters (Figure 2). Each cluster contained isolates from different sample types and farm origins but differed in its predominant resistance gene profile (Supplementary Table S3). Clusters 1 and 2 were predominantly characterized by emrE, whereas Clusters 3 and 4 were predominantly characterized by ydgE and sugE(c). The plasmid-encoded genes qacE, qacEΔ1, and qacF were detected across all clusters. Notably, sugE(p) was detected exclusively in Cluster 4.

Multivariable logistic regression analysis showed that the presence of int1 was significantly and positively associated with strA, tetA, and aadA2, whereas it was significantly and negatively associated with floR. The presence of int1 was also significantly associated with the disinfectant resistance genes qacF, emrE, and mdfA (Table 3).

This study provides a comprehensive assessment of AMR in E. coli isolated from tilapia aquaculture, focusing on susceptibility to commonly used disinfectants and heavy metals, together with biofilm-forming capacity. Five disinfectants commonly used in household and environmental settings were selected based on previous evidence [7]. A bimodal TCS MIC distribution was observed, indicating two distinct bacterial subpopulations: wild-type isolates with intrinsic susceptibility and non-wild-type isolates with acquired reduced susceptibility or tolerance to TCS. Notably, some isolates had MICs exceeding 1 µg/mL, higher than those previously reported for E. coli isolated from animal production systems in Thailand [30].

The bimodal MIC distribution, along with MIC values exceeding 1 µg/mL, supports the presence of a distinct subpopulation with reduced susceptibility to TCS and indicates a shift in the susceptibility profile of E. coli in tilapia aquaculture. TCS inhibits bacterial fatty acid biosynthesis by targeting the enoyl-acyl carrier protein reductase FabI, encoded by fabI, and mutations in fabI represent one of the principal mechanisms underlying TCS resistance [35]. In addition, increased activity of multidrug efflux pumps, particularly resistance-nodulation-division systems such as AcrAB-TolC and MexAB-OprM, may contribute to reduced intracellular accumulation of TCS and thereby promote reduced susceptibility [36]. The MIC₅₀ values of the remaining disinfectants observed in this study were slightly higher than those previously reported for E. coli isolated from animal food production systems (Table 4) [30, 37–41], suggesting reduced susceptibility and the potential emergence of tolerance to multiple disinfectants.

Compared with the only previous study conducted in Thailand using isolates from the swine production chain, the MIC₅₀ values for BKC, CHX, GLU, and TCS were at least one two-fold dilution higher [30]. This shift in susceptibility has important implications for farm biosecurity because disinfectant efficacy is influenced by water quality, temperature, disinfectant concentration, contact time, and organic matter [42]. By providing susceptibility data from Thai tilapia aquaculture, the present study offers region-specific evidence supporting thorough pre-cleaning procedures and the application of disinfectants at validated concentrations with adequate contact times to maximize bacterial inactivation while minimizing exposure to sublethal concentrations.

Differences between the proposed ECOFFs in the present study and those reported previously may reflect variations in the ecological origin of the isolates, local patterns of disinfectant use, and sample size. Nevertheless, these ECOFFs provide a valuable baseline for monitoring temporal changes in susceptibility and detecting the emergence of non-wild-type populations during aquaculture surveillance. The observed differences may also be associated with the prolonged environmental persistence of disinfectants in aquaculture systems. For example, BKC has been reported to persist in water for up to 38 days, thereby maintaining continuous selective pressure on microbial communities [43]. Collectively, these findings highlight the role of disinfectants in driving AMR in aquaculture through sustained environmental selective pressure, thereby facilitating the emergence and persistence of resistant bacterial populations.

All tested heavy metals exhibited limited variation in MIC distributions, suggesting relatively uniform selective pressure within the studied aquaculture environment. The MIC₅₀ of AgNO₃ (16 µg/mL) observed in the present study was higher than that previously reported for E. coli isolated from tilapia in Zimbabwe, possibly reflecting adaptation to anthropogenic silver-containing biocides in Thai aquaculture systems [44]. Likewise, the MIC₅₀ values for CuSO₄ (2,048 µg/mL) and ZnCl₂ (1,024 µg/mL) exceeded those reported for E. coli isolated from fishponds in Nigeria [45]. These findings raise concerns regarding inappropriate concentrations of heavy metals in commercial fish feeds, which may contribute to environmental accumulation and promote AMR [46]. The limited regulatory oversight of heavy metal use may further increase the ecological and public health risks associated with these environmental contaminants.

Strong biofilm production was generally uncommon, indicating limited biofilm-forming potential among E. coli isolates recovered from tilapia aquaculture. Nevertheless, a small subpopulation of strong biofilm-producing isolates was identified, consistent with previous findings in fish from Vietnam [47]. Biofilms protect bacteria against disinfectants and antimicrobials, thereby facilitating the persistence of AMR and multidrug-resistant (MDR) strains [48]. Previous studies have also demonstrated a positive association between biofilm formation and resistance to QACs [49]. Collectively, these findings highlight the importance of implementing targeted strategies to mitigate biofilm-mediated resistance and reduce its environmental and public health impacts.

The E. coli isolates in the present study lacked all five targeted heavy metal resistance genes encoding metal efflux pumps, consistent with the very low prevalence of these genes reported in bacteria isolated from fish-cage sediments in Turkey [50]. This finding suggests that heavy metal tolerance in these aquaculture environments may be mediated through alternative resistance mechanisms. Resistance to QACs was primarily associated with chromosomally encoded efflux pump genes. Among the disinfectant resistance genes investigated, ydgE and mdfA were the most prevalent, consistent with findings reported for E. coli isolated from retail fish in India [44]. The mdfA gene confers cross-resistance by exporting both tetracyclines and BKC, thereby reducing the effectiveness of both antimicrobials and disinfectants [20]. Other chromosomally encoded genes, including sugE(c) and emrE, which are associated with high-level resistance to QACs, were also detected. In contrast, the prevalence of plasmid-associated QAC resistance genes, including qacE, qacEΔ1, qacF, qacG, and sugE(p), was very low. These findings suggest that chromosomally encoded mechanisms may play a greater role than plasmid-mediated mechanisms in disinfectant resistance among E. coli isolates from tilapia aquaculture.

Despite the observed heavy metal tolerance, none of the targeted heavy metal resistance genes were detected. This discrepancy suggests that the observed phenotype may result from intrinsic metal homeostasis, regulatory co-expression, cross-resistance involving other efflux systems, or additional resistance determinants that were not investigated in the present study [51]. These findings emphasize the importance of monitoring both intrinsic and acquired resistance mechanisms in aquatic environments.

Cluster analysis demonstrated that the chromosomally encoded genes ydgE, sugE(c), and emrE predominated within specific clusters, although each cluster remained heterogeneous with respect to sample type, farm origin, and overall resistance gene profile. The predominance of these genes within individual clusters suggests possible co-selection under environmental pressures, such as repeated exposure to disinfectants. In contrast, the plasmid-associated genes qacE, qacEΔ1, and qacF were detected across all clusters, suggesting widespread dissemination that may be facilitated through horizontal gene transfer (HGT), as reported in a previous study [52]. The co-occurrence of chromosomally encoded efflux pump genes and plasmid-mediated resistance determinants further highlights the interaction between intrinsic and acquired mechanisms in shaping disinfectant resistance profiles. Furthermore, the heterogeneous clustering patterns may reflect multiple contamination sources and diverse evolutionary pathways of resistance acquisition within aquaculture systems. Collectively, these findings illustrate the complexity of resistance gene distribution and emphasize the influence of environmental selective pressures on the maintenance and dissemination of antimicrobial and biocide resistance in aquatic ecosystems.

Multivariable logistic regression analysis provided novel insights into the co-occurrence of resistance determinants in E. coli isolated from Thai tilapia aquaculture. The presence of int1 was significantly associated with strA, tetA, aadA2, qacF, emrE, and mdfA. These associations indicate that int1-positive isolates may harbor broader and more diverse resistance profiles, reflecting the combined effects of mobile genetic elements, genetic linkage, and co-selection by antimicrobials and biocides. Classical integrons typically contain qacEΔ1 and sul1, whereas non-classical integrons may harbor alternative genes, including qacH and sul3 [53]. However, the multivariable analysis did not support either of these classical integron configurations. This finding suggests that alternative mechanisms, including plasmid-mediated transfer, transposons, and co-selection driven by disinfectants and heavy metals, may contribute to the dissemination of disinfectant resistance genes in aquatic environments. From a One Health perspective, simultaneous monitoring of int1 alongside antimicrobial and biocide resistance genes in fish, farm water, sediments, and surrounding aquatic environments could improve early detection and surveillance of resistance dissemination across environmental, animal, and human interfaces.

This study provides the first evidence of reduced TCS susceptibility among E. coli isolates recovered from Thai tilapia aquaculture and proposes the first ECOFFs for multiple disinfectants and heavy metals in this production system. Unlike previous investigations focused primarily on terrestrial livestock, this study integrates MIC distributions, biofilm-forming capacity, and resistance gene profiling to provide a comprehensive assessment of biocide susceptibility in an aquatic production environment. The predominance of chromosomally encoded resistance determinants over plasmid-mediated mechanisms suggests that intrinsic resistance mechanisms play a major role in disinfectant tolerance in these isolates. The identification of non-wild-type populations with reduced susceptibility to TCS, reflected by a bimodal MIC distribution, is an important finding indicating a shift in susceptibility under natural aquaculture conditions. Such reduced susceptibility raises concerns regarding the long-term effectiveness of commonly used disinfectants and the potential for co-selection of AMR. Although HGT likely contributes to resistance dissemination, the present findings suggest that its role is secondary to intrinsic chromosomal mechanisms. Furthermore, the observed resistance gene clustering indicates potential selection under prolonged disinfectant exposure, whereas the presence of strong biofilm-forming isolates suggests that biofilms may serve as protective niches that facilitate the persistence and dissemination of resistant bacteria in aquaculture environments.

Several limitations should be considered when interpreting the findings of this study. First, the isolates were obtained from a limited geographic area and a relatively small number of tilapia farms, which may limit the generalizability of the findings to other aquaculture production systems. Second, the targeted PCR approach focused on selected disinfectant and heavy metal resistance genes and therefore may not have identified additional resistance mechanisms involved in reduced susceptibility. Third, environmental concentrations of disinfectants, biocides, and heavy metals were not quantified, precluding direct assessment of the relationship between environmental exposure and the observed MIC distributions. Furthermore, the cross-sectional sampling design did not permit evaluation of temporal changes in susceptibility or the emergence of non-wild-type populations. Finally, clonal relatedness among isolates was not investigated; therefore, it was not possible to determine whether similar resistance profiles resulted from clonal dissemination or the independent acquisition of resistance determinants.

Despite these limitations, this study provides valuable baseline data on disinfectant and heavy metal susceptibility in E. coli from Thai tilapia aquaculture and establishes proposed ECOFFs for future surveillance. Future investigations should incorporate longitudinal and multicenter sampling, quantify environmental concentrations of disinfectants and heavy metals, and employ whole-genome sequencing and comparative genomic analyses to characterize resistance mechanisms, mobile genetic elements, and transmission pathways. Integrating environmental monitoring with phenotypic and genomic surveillance will improve understanding of the drivers of disinfectant tolerance and AMR dissemination and will support evidence-based risk assessment and sustainable aquaculture management within a One Health framework.