The study protocol was reviewed and approved by the Institutional Animal Care and Use Committee of Khon Kaen University, Thailand (approval no. IACUC-KKU-111/67). All procedures involving animals were conducted in strict accordance with national guidelines for animal research and welfare, and in compliance with internationally accepted standards, including the ARRIVE 2.0 guidelines and the principles outlined in the Guide for the Care and Use of Laboratory Animals.

This study was based on non-invasive sampling procedures conducted under routine surveillance conditions. Cloacal swab samples were collected by trained veterinarians using standardized handling techniques to minimize stress, discomfort, and potential injury to the birds. No experimental infection, invasive intervention, or euthanasia was performed as part of this study. All sampling procedures adhered to the principles of replacement, reduction, and refinement (3Rs), ensuring that the number of animals involved was limited to the minimum required to achieve scientific validity while maintaining high standards of animal welfare.

The chickens included in this study originated from commercial farms operating under the Thai Agricultural Standard TAS 6901-2017 and were handled within standard production and processing environments. Permission to access animals and collect samples at slaughterhouses and retail markets was obtained from facility operators and relevant authorities prior to sampling. All sampling activities were carried out in a manner that did not interfere with routine farm or slaughterhouse operations.

Biosafety and biosecurity measures were strictly implemented throughout the study. All laboratory procedures involving Salmonella were conducted in a biosafety level 2 laboratory in accordance with institutional biosafety regulations. Personnel were trained in microbiological safety practices, and appropriate personal protective equipment was used at all times to prevent occupational exposure and environmental contamination.

Overall, the study design ensured full compliance with ethical standards for animal use in research while maintaining scientific rigor and data reliability.

This study was a cross-sectional seasonal surveillance study conducted along the poultry supply chain in Khon Kaen Province, Thailand. Over one year (March 2023 to February 2024), we simultaneously monitored and compared Salmonella contamination at two critical nodes, slaughterhouses and retail markets, across three climatic seasons (summer, rainy, and winter). The comparative design aimed to identify site-specific and season-specific risk patterns within the local poultry production and retail system.

The chickens sampled in this study originated from standardized commercial farms operating under the Thai Agricultural Standard TAS 6901-2017, which mandates veterinary oversight and restricts the use of certain critically important antimicrobials, and which had been approved by the Department of Livestock Development. All broilers were of a mixed breed, aged 35–40 days, and sourced from a single integrated production company. This common origin ensures a consistent baseline of farm management and biosecurity. Birds were transported to the slaughterhouses in dedicated vehicles over a short travel duration (1–2 h). The two participating slaughterhouses were small scale, manually operated facilities processing approximately 500–800 birds per day, utilizing municipal water and implementing basic daily cleaning routines. Following processing, carcasses were distributed to the two traditional wet markets included in the study, where they were displayed and sold at ambient temperature on open stalls, with no dedicated cold chain facilities.

A total of 689 swabs (including cloacal swab and carcass swab) were collected during different seasons from slaughterhouses (n = 2) and markets (n = 2) in Khon Kaen Province, Northeast Thailand. Three seasons were defined as summer (March–May, 32°C–38°C, humidity rate 66%–82%), rainy (June–October, 27°C–30°C, humidity rate 82%–88%), and winter (November–February, 20°C–28°C, humidity rate 50%–75%). The sampling period spanned one year (March 2023 to February 2024), covering a temperature–humidity index range of 65–80, representing 16 discrete intervals.

A systematic, repeated sampling strategy was employed: the same four sites were sampled once per quarter across all three seasons. Based on a reported Salmonella prevalence of 23.3% in chicken slaughterhouses [6], the sample size was calculated using the formula n = Z² × P × (1 − P) / e², with a 95% confidence interval and 5% margin of error. Using Z = 1.96, e = 0.05, and p = 0.233, the required sample size was 228. For markets, with a reported prevalence of 21.33% [17], the required sample size was 258. The total sample collected in this study was 689, which exceeded the minimum calculated sample size. The number of samples collected per season and site is shown in Table 1.

Isolation and identification of Salmonella were performed according to ISO 6579-1:2017 [18]. After collection, swabs stored in transfer medium were placed in Buffered Peptone Water (BPW) for pre-enrichment and incubated at 37°C for 24 h. Subsequently, three loopfuls of the culture were transferred to Modified Semisolid Rappaport Vassiliadis Medium (MSRV) plates for selective enrichment and incubated at 42°C for 24 h. Positive colonies were streaked onto Xylose Lysine Deoxycholate agar (XLD) plates and incubated for 24 h [19]. Presumptive Salmonella colonies were confirmed using Triple Sugar Iron (TSI, Difco) and Motility Indole Lysine (MIL, Difco) tests. All the above media and reagents were obtained from Difco Laboratories, Detroit, MI, USA. All isolates were stored in Luria–Bertani (LB) medium containing 20% glycerol at −80°C.

Serological identification was performed using diagnostic polyvalent and monovalent Salmonella “O” and “H” antisera (Denka Seiken, Tokyo, Japan). At least one isolate per positive sample was selected for serotyping. Isolates were subsequently stored in tryptic soy broth (Difco Laboratories) containing 15% glycerol at −80°C.

The panel included agents with reported usage in Thai poultry (e.g., ampicillin, tetracycline, streptomycin) alongside representatives of the WHO highest-priority critically important antimicrobials (HPCIA) (e.g., fluoroquinolones, third-generation cephalosporins, carbapenems). Colistin was excluded based on its restricted use in Thai food animals following the 2017 national ban on antimicrobial growth promoters.

Antimicrobial susceptibility testing was performed against 14 antibiotics using the disk diffusion method on Mueller–Hinton agar (Difco), in accordance with Clinical and Laboratory Standards Institute (CLSI) guidelines [20]. The antimicrobial agents (Oxoid, Liofilchem, UK) tested included streptomycin (S, 10 μg), norfloxacin (NOR, 10 μg), tetracycline (TE, 30 μg), gentamicin (GM, 10 μg), sulfamethoxazole/trimethoprim (SXT, 23.75 μg), imipenem (IPM, 10 μg), cefotaxime (CRO, 30 μg), chloramphenicol (C, 30 μg), ciprofloxacin (CIP, 5 μg), nalidixic acid (NA, 30 μg), ampicillin (AM, 10 μg), ceftazidime (CAZ, 30 μg), ceftriaxone (CTX, 30 μg), and amoxicillin/clavulanic acid (AMC, 20/10 μg). Escherichia coli ATCC 25922 (ATCC, Manassas, VA, USA) was used for quality control. Results were accepted only when the zone diameters for the quality control strain fell within CLSI-specified reference ranges. Multidrug resistance was defined as resistance to agents from three or more distinct antibiotic classes [21].

To investigate the genetic relatedness of the dominant Salmonella population, PFGE was performed on a subset of isolates comprising all available strains of the most prevalent serotype, S. Agona (n = 34). PFGE was performed according to the PulseNet standard protocol [22], using CHEF DRIII (Bio-Rad, Hercules, CA, USA). Briefly, isolates were grown on Nutrient Agar at 37°C for 16 h. Bacterial cells were suspended in TE buffer (100 mM Tris-HCl, 100 mM EDTA, pH 8.0) and adjusted to an OD610 of 1.4. The cells were embedded in agarose plugs and lysed in lysis buffer (50 mM Tris-HCl, 50 mM EDTA, 1% Sarkosyl, 1 mg/mL proteinase K) at 55°C for 16-20 h. Plugs were washed 4 times with TE buffer and digested with 50 U of XbaI (New England Biolabs) at 37°C overnight. PFGE was performed in 1% agarose gel using 0.5× TBE buffer at 14°C for 18 h [23]. Plug slices of 3–5 mm width were placed in a 1.5 mL microcentrifuge tube containing 200 μL of 1× restriction buffer (NE Buffer supplemented with 50 U of XbaI-digested restriction enzyme) and incubated overnight at 37°C. Plug slices were washed for 30 min with 0.5× TBE buffer and subsequently loaded into the wells of a 1% agarose gel, and the wells were overlaid with 1% agarose dissolved in 0.5× TBE. After solidification, the gel was electrophoresed for 18 h.

PFGE profiles were analyzed using BioNumerics software (version 4.5, Bio-Rad). Dendrograms were constructed using the unweighted pair group method with arithmetic mean, with the Dice similarity coefficient, optimization set to 1%, and tolerance set to 1%. Isolates with 100% similarity were considered to belong to the same PFGE type.

Data were processed using Excel, SPSS 20.0 (IBM Corp., Armonk, NY, USA), and GraphPad Prism 10.1.3. Categorical data were analyzed using the chi-square test, with statistical significance set at p < 0.05 and high statistical significance at p < 0.01.

The prevalence of Salmonella was calculated for each season, and differences between seasons were evaluated by the chi-square test (χ²). Relative risk (RR) was used to assess the likelihood of Salmonella occurrence associated with seasonal factors. In addition, the independent effects of season and location type were assessed using multivariable logistic regression. Results are reported as adjusted odds ratios (aOR) with 95% confidence intervals (CI).

This study found an overall Salmonella prevalence of 32.37% (223/689), but its distribution showed significant spatial heterogeneity and seasonal clustering. Spatially, contamination levels varied significantly across sampling sites (χ² = 25.48, p < 0.0001). The overall prevalence at retail markets (particularly Market-1, 60.83%) was significantly higher than that at slaughterhouses (18.23%–22.97%). Seasonally, the distribution of positive samples was generally non-uniform within most individual sites (most p₁ < 0.05). Overall, positive samples were predominantly concentrated in the rainy (45.29%) and winter (37.22%) seasons, which together accounted for more than 82% of total detections. Further testing indicated that this seasonal clustering pattern was particularly pronounced in winter and differed significantly among sampling sites (p₂ < 0.0001). These results collectively demonstrate that Salmonella contamination status is significantly influenced by both the “sampling site” and “season” (Table 2).

To quantify the contamination risk across seasons, this study calculated the RR for the rainy and winter seasons, using summer as the reference. The analysis was stratified by sample type, revealing distinct seasonal risk patterns depending on the sample source. For anal swabs (reflecting animal carriage), the risk of Salmonella detection in the rainy season was 2.29 times higher than in summer, a statistically significant increase. In contrast, the risk in winter was not significantly different from that in summer. Conversely, for carcass swabs (reflecting processing contamination), although the detection rates in the rainy and winter seasons were numerically higher than those in summer, neither RR reached statistical significance. This indicates that the influence of seasonal factors on Salmonella prevalence is primarily evident at the farm/transport stage (animal carriage), while its effect may be masked or mitigated at the processing stage, likely by other factors such as slaughterhouse hygiene practices (Table 3).

To assess the independent effects of each risk factor, this study further constructed a multivariate logistic regression model. After adjusting for the confounding effects of season and sampling location, the analysis revealed that retail markets (compared to slaughterhouses) and the rainy season and winter (compared to summer) each represented independent risk factors for Salmonella contamination (Table 4). Specifically, the contamination risk for market samples was 4.82 times higher than that for slaughterhouse samples (aOR = 4.82, 95% CI: 3.42–6.79); while the risks during the rainy season and winter were 3.81 times (aOR = 3.81, 95% CI: 2.40–6.04) and 2.80 times (aOR = 2.80, 95% CI: 1.73–4.53) higher than in summer, respectively. The model demonstrated good fit (Hosmer-Lemeshow test, p = 0.62). Collectively, the univariate and multivariable analyses demonstrate that Salmonella contamination is independently and statistically significantly associated with both spatial (retail market location) and temporal (rainy and winter seasons) factors, with market samples and the rainy season posing the highest respective contamination risks.

Serotyping of 223 Salmonella isolates revealed that Salmonella Agona (15.25%) and Salmonella Saintpaul (14.79%) were the most predominant serotypes, while Salmonella Singapore (10.31%) was also widely distributed across all sampling sites. The serotype distribution exhibited significant seasonal variation: although the rainy season accounted for the highest total number of isolates (68.16%), several important serotypes, including S. Agona, S. Bareilly, and S. Rissen, were predominantly detected during the winter season (Table 5).

Among positive isolates, S. Agona, as the predominant circulating serotype, exhibited significant seasonal variation in prevalence (χ² = 15.64, p < 0.001, Table 6). Its prevalence during winter (37.70%) was markedly higher than that during the rainy season (9.78%) and summer (5.56%). These findings indicate that seasonal variation not only affects the overall prevalence of Salmonella but also significantly shapes its serotype composition. The distinct seasonal dominance of specific serotypes indicates clear seasonal patterns within the local poultry production chain. Therefore, it is imperative to implement monitoring and control strategies tailored to these seasonal serotype variations.

The prevalence of AMR (to at least one drug) among Salmonella isolates varied significantly across seasons (χ² = 7.21, p = 0.027, Table 6). The resistance rate was highest in winter (85.5%), followed by summer (74.4%) and the rainy season (58.4%). Further comparative analysis revealed that, relative to summer, isolates recovered in winter had a significantly higher risk of being drug-resistant (RR = 1.15, 95% CI: 1.00–1.32, p = 0.044). In contrast, the risk in the rainy season was lower but did not reach statistical significance (RR = 0.78, 95% CI: 0.61–1.01, p = 0.058).

The heat map (Figure 1A) indicates that isolates from all locations exhibited the highest resistance rates to streptomycin (S) (39.7%–50.0%), with particularly pronounced resistance to ampicillin (AM) at Market-2 (57.1%) and Market-1 (35.6%); Slaughterhouse-1 exhibited significantly higher resistance rates to gentamicin (GM, 37.0%) and NA (48.1%) compared to other sites, whereas no resistance was detected to critically important antimicrobials such as CIP, NOR, or IMP. This illustrates severe resistance to first-line drugs, with distinct resistance patterns across locations, suggesting that the emergence and persistence of AMR are closely linked to site-specific environmental factors.

Comparisons revealed that isolates from market sources exhibited higher average resistance rates, particularly to β-lactams and tetracyclines than those from slaughterhouse sources, while isolates from slaughterhouse sources demonstrated higher average resistance rates to aminoglycosides and quinolones (Figure 1B). This indicates distinct antimicrobial selection pressures along the supply chain from slaughterhouses to markets, potentially reflecting variations in antimicrobial usage practices or contamination sources at different stages.

Radar charts revealed unique AMR patterns across locations (Figure 1C), with Slaughterhouse-1 exhibiting the broadest resistance spectrum (the highest levels of multidrug resistance) and Market-2 showing particularly pronounced resistance to certain antimicrobial classes (e.g., β-lactams). These findings confirm significant location-specific resistance profiles, underscoring the need for targeted prevention and control strategies tailored to different settings rather than generic universal approaches.

Analysis of high-risk drugs with overall resistance rates >10% confirmed streptomycin (S), ampicillin (AM), NA, and tetracycline (TE) as primary threats. It clearly showed that Slaughterhouse-1 is the main location contributing to high resistance rates for gentamicin (GM) and NA, while Market-2 is a hotspot for ampicillin (AM) and tetracycline (TE) resistance (Figure 1D). This analysis identifies high-risk antibiotics requiring priority intervention and their core sources, providing direct evidence for implementing targeted surveillance and control measures.

To further investigate the distribution of MDR Salmonella, we found that they accounted for 32.43% of total positive cases (n = 79). The prevalence of MDR exhibited significant seasonal patterns, with the highest detection rate occurring in winter (Figure 2). Slaughterhouses (particularly Slaughterhouse-1, with a detection rate of 65.38%) were identified as critical risk points, exhibiting significantly higher levels of resistance and variability compared to market environments (Figure 2). These findings indicate that the slaughter process acts as a key node for the generation and transmission of MDRbacteria, necessitating enhanced targeted surveillance and biosecurity measures during the winter season.

To explore potential clonal relationships and provide genetic context for the observed spatiotemporal and phenotypic patterns, all available isolates of the most prevalent serotype, S. Agona (n = 34), were analyzed by PFGE. These isolates represented the distribution across seasons (summer, rainy, and winter) and sampling locations (Slaughterhouses-1, 2 and Markets-1, 2). Figure 3 presents the resulting dendrogram, corresponding antimicrobial susceptibility patterns, and metadata for the analyzed strains. The isolates were categorized into 13 distinct PFGE types (A-M) using a 78% similarity threshold. In line with the epidemiological data, isolates from the same location and season frequently clustered together, providing molecular evidence for localized clonal persistence or dissemination. Furthermore, the finding that identical PFGE types were shared between some isolates from different locations (e.g., a slaughterhouse and a market) offers genetic evidence consistent with cross-contamination events along the poultry supply chain.

The present study provides a comprehensive overview of the prevalence, serotype distribution, AMR, and genetic relatedness of Salmonella isolates recovered from chicken samples originating from slaughterhouses and retail markets in Khon Kaen Province, Thailand, across three distinct climatic seasons. The overall prevalence of Salmonella was 32.37%, which is consistent with findings from other developing countries where poultry serves as a major reservoir for this pathogen [24, 25]. The significantly higher contamination rate observed in markets (60.83% in Market 1) compared to slaughterhouses (22.97% in Slaughterhouse 1) underscores the role of post-processing contamination, cross-contamination, and inadequate hygiene practices in retail market environments. These findings align with previous studies conducted in China [26] and Korea [27], where market samples exhibited higher Salmonella prevalence than slaughterhouse samples, highlighting the need for enhanced sanitary measures along the poultry supply chain.

The significant seasonal variation observed, with the highest prevalence during the rainy season (45.29%), parallels findings from tropical regions worldwide [28]. Studies conducted in Bangladesh demonstrated similar seasonal patterns where monsoon conditions facilitated Salmonella proliferation in poultry environments [29]. The high temperature (27°C–30°C) and humidity (82%–88%) during Thailand’s rainy season likely create favorable conditions for bacterial survival and spread through water sources and contaminated equipment.

The most novel and operationally relevant finding of this study is the distinct seasonal dissociation between peak Salmonella prevalence and peak MDR. While the rainy season (45.29%) posed the highest risk for overall contamination, likely driven by high temperature and humidity that facilitate bacterial survival and cross-contamination [30], the prevalence of MDR peaked significantly in winter, particularly within slaughterhouses (65.38% in Slaughterhouse-1). This dissociation suggests that distinct seasonal mechanisms are responsible: contamination is primarily environmentally mediated, whereas MDR amplification is likely anthropogenic, reflecting seasonal management practices such as prophylactic antibiotic use during colder months or the consolidation of multi-source flocks prior to slaughter [31, 32]. This finding has direct and practical policy implications: rainy season interventions should prioritize hygiene and biosecurity to reduce bacterial load, while winter efforts must focus on antimicrobial stewardship and MDR surveillance within slaughterhouses to prevent the amplification and downstream dissemination of resistant clones.

Comparison with historical data from the same province reveals a notable shift in serotype predominance over the past two decades. While Angkititrakul et al. [33] reported S. Rissen and S. Stanley as the dominant serovars in retail chicken meat in Khon Kaen, our study identified S. Agona and S. Saintpaul as the predominant serotypes in the local contemporary poultry supply chain. This temporal shift may reflect changes in poultry production systems, biosecurity practices, or antimicrobial usage patterns, and thus warrants further in-depth investigation. The primary contribution of the present study, therefore, lies not in the identification of novel serotypes but in establishing an updated baseline for Northeast Thailand, a region with limited recent epidemiological surveillance data, and documenting, for the first time, the seasonal and spatial distribution patterns of these now-predominant serotypes. Notably, the predominant serotype identified in this study, S. Agona, has been consistently reported in earlier human salmonellosis surveillance in Thailand [34], and remains a clinically significant serotype of public health concern. This highlights the need for updated integrated One Health surveillance to assess current transmission dynamics between poultry and human populations.

The predominance of S. Agona in slaughterhouses may reflect its well-documented ability to form robust biofilms on abiotic surfaces [35]. For S. Saintpaul, genetic determinants involved in surface attachment have been characterized, providing a mechanistic basis for its persistence in processing environments [36]. In contrast, the greater serotype diversity observed in markets, including S. Stanley, S. Bareilly, and S. Rissen, which are historically dominant serotypes, suggests that these serovars continue to circulate in the region, albeit at lower frequencies. This diversity indicates the presence of multiple contamination sources at the retail level, a pattern consistent with recent genomic surveillance in Thai food markets [37].

While direct farm-level antimicrobial usage (AMU) data were not available for correlation with our findings, the resistance patterns observed can be contextualized within the known landscape of AMU in Thai poultry production as reported in the literature. The resistance profiles observed, high rates to streptomycin (43.05%), ampicillin (34.53%), and NA (30.49%), are consistent with national and regional surveillance data from Thailand and neighboring countries [38–40]. The rising resistance gradient from slaughterhouses to markets parallels findings from Vietnam, suggesting additional selective pressure or cross-contamination at the retail level [41].

The absence of resistance to fluoroquinolones and carbapenems is encouraging and likely reflects Thailand’s 2017 colistin ban and subsequent restrictions on highest-priority critically important antimicrobials in food animals [42]. The notably higher ampicillin resistance in markets (57.14%) compared to slaughterhouses (11.11%) may reflect unauthorized farm-level antibiotic use or cross-contamination during market handling, a pattern observed elsewhere in the region [43].

The high MDR prevalence in slaughterhouses (65.38% in Slaughterhouse-1), peaking in winter, highlights slaughterhouses as critical nodes for resistance amplification. This seasonal pattern may reflect prophylactic antibiotic use during cold months or flock consolidation prior to winter slaughter [44, 45]. PFGE clustering of isolates by source and season supports localized clonal dissemination, while shared PFGE types across sites suggest potential cross-contamination along the supply chain [46]. These findings reinforce the need for integrated One Health surveillance that links human salmonellosis cases with poultry supply chain data to better inform public health interventions.

Several limitations of the present study should be acknowledged. First and foremost, this study did not collect AMU data from the primary source farms. Consequently, critical information regarding the specific types of antimicrobials used, the distinction between therapeutic and prophylactic applications, and potential seasonal variations in antimicrobial medication practices remains unknown. Without these farm-level AMU records, any robust causal inference between antimicrobial exposure and the observed resistance patterns remains speculative. Second, the absence of comprehensive genotypic characterization of resistance determinants limits our understanding of the underlying resistance mechanisms and their potential for horizontal gene transfer. Third, environmental samples were not collected from sampling sites, restricting our capacity to trace cross-contamination pathways. Fourth, PFGE offers lower discriminatory power than whole-genome sequencing for fine-scale bacterial transmission tracing. Fifth, the cross-sectional design only captures associations at discrete time points but cannot establish temporal causality. Finally, the single-province scope may limit the generalizability of the findings to other regions of Thailand. Despite these limitations, this study provides valuable, much-needed baseline data on the spatiotemporal dynamics of Salmonella along the poultry slaughterhouse–market continuum in this understudied region. Future research should systematically integrate longitudinal primary farm-level AMU records, including drug types, purpose, and seasonal patterns, with whole-genome sequencing-based high-resolution molecular surveillance and environmental sampling to definitively elucidate the causal pathways driving AMR emergence and transmission in the Thai poultry sector.