ESBL-Producing Escherichia coli Carrying CTX-M Genes Circulating among Livestock, Dogs, and Wild Mammals in Small-Scale Farms of Central Chile

Antibiotic-resistant bacteria of critical importance for global health such as extended-spectrum beta-lactamases-producing (ESBL)-Escherichia coli have been detected in livestock, dogs, and wildlife worldwide. However, the dynamics of ESBL-E. coli between these animals remains poorly understood, particularly in small-scale farms of low and middle-income countries where contact between species can be frequent. We compared the prevalence of fecal carriage of ESBL-E. coli among 332 livestock (207 cows, 15 pigs, 60 horses, 40 sheep, 6 goats, 4 chickens), 82 dogs, and wildlife including 131 European rabbits, 30 rodents, and 12 Andean foxes sharing territory in peri-urban localities of central Chile. The prevalence was lower in livestock (3.0%) and wildlife (0.5%) compared to dogs (24%). Among 47 ESBL-E. coli isolates recovered, CTX-M-group 1 was the main ESBL genotype identified, followed by CTX-M-groups 2, 9, 8, and 25. ERIC-PCR showed no cluster of E. coli clones by either host species nor locality. To our knowledge, this is the first report of ESBL-E. coli among sheep, cattle, dogs, and rodents of Chile, confirming their fecal carriage among domestic and wild animals in small-scale farms. The high prevalence of ESBL-E. coli in dogs encourages further investigation on their role as potential reservoirs of this bacteria in agricultural settings.


Introduction
The current increase of antimicrobial resistance (AMR) is considered a main global threat to human and animal health [1,2]. AMR is responsible for thousands of human fatalities annually [3] and large economic losses that could reduce global GDP in 1-4% by 2050 [2,4]. The intense use of antibiotics in livestock production and humans is the main genes including bla CTX-M , bla TEM , and bla SHV , and (iii) use high resolution molecular typing to assess potential ESBL-E. coli transmission within farms or between different species.  [53]. Farms were randomly selected from a list provided by the Municipality's agrarian unit, accounting for areas overlapping with the known territory of wildlife as previously described [52]. Our sampling focused mainly on cattle because they had the highest potential of overlapping with wild mammals since they often free-ranged within wildlife habitat during our study period.

Sample Collection
aims of this study were (i) to estimate and compare the prevalence of ESBL-E. c carriage between livestock, dogs, and wild mammals living in the same agricultu ting of central Chile, (ii) to detect the presence of the most common ESBL genes in blaCTX-M, blaTEM, and blaSHV, and (iii) use high resolution molecular typing to assess p ESBL-E. coli transmission within farms or between different species.

Sample Collection
Fresh fecal samples were collected between March 2019 and September 201 livestock, dogs, and wildlife in and around 13 farming localities located in the mu ities of Colina (33.1045° S, 70.6159° W) and Lampa (33.2827° S, 70.8793° W) of th cabuco province in the Metropolitan Region of central Chile, in the peri-urban are Santiago Capital City (Figure 1). A farming locality was either a single private farm area where livestock from different owners grazed together and received the same treatments. The province of Chacabuco includes mainly small-to medium-scale f with an estimated livestock population of 10,662 cattle (mean: 38 animals/farm) pigs (587/farm), 5490 goats (59/farm), 4441 sheep (42/farm), and 2897 horses (4/farm Farms were randomly selected from a list provided by the Municipality's agraria accounting for areas overlapping with the known territory of wildlife as previou scribed [52]. Our sampling focused mainly on cattle because they had the highest p of overlapping with wild mammals since they often free-ranged within wildlife during our study period.  We focused on sampling the most common wild mammals encountered in those farms including several species of endemic and invasive rodents, the invasive European wild rabbit and the Andean fox, who predates these herbivore species [54,55]. These species were previously determined by discussions with farmers and the municipality's agrarian unit during preliminary visits to the farms. Peri-urban and wild rodents were live captured, sampled, and released using Sherman traps. Fifty traps were placed in and around each sampled farm for at least 4 consecutive days and checked for captured rodents daily. Rectal swabs were collected from alive individuals immobilized, using gloves and protective equipment. Rodents were identified at the genus or species level based on morphological characteristics. Fresh fecal samples from European rabbits were collected early in the morning by identifying rabbit dens in areas where farmers commonly observed rabbits. To avoid sampling the same individual twice, we only collected fresh sample feces from the same den if they were more than 4 m apart, and only sampled each den once. Fresh fecal samples from foxes were collected by walking known paths where foxes were previously captured in the area [56]. Fresh samples from foxes were identified and differentiated from dog feces by their distinct 'fruit' seeds and morphology contained on the sample. To avoid sampling the same individual twice, we only collected a fresh sample in localities that were more than 5 km apart, considering 5 km 2 as the average home range size of foxes in this area [52]. Dogs were sampled by directly taking rectal swabs or waiting until the dog defecated, depending on whether the owner considered that the dog could be aggressive or not during sampling. For all samples taken from the ground, we only collected the portion that was not in contact with the ground to avoid bacterial contamination from the soil. This study was approved by the Ethical Committee of the Universidad Andrés Bello (permit number: 018/2018). The capture and sampling of rodents were also approved by the Servicio Agricola Ganadero (permit number: 2118/2019).

Sample Size and Prevalence Estimation
The required sample size needed to estimate the prevalence of ESBL-E. coli in livestock (defined as the number of animals harboring at least one isolate of ESBL-E. coli over the total number of sampled animals) was calculated with the program Epi Info 7.2.2.6 TM [57]. To our knowledge, no previous study has estimated the prevalence of fecal carriage of ESBL-E. coli among livestock in Chile. Thus, we assumed an expected prevalence of ESBL-E. coli of 30%, similar to a study conducted around the Lima capital in Peru with similar farm characteristics [12]. Based on this expected prevalence, a margin of acceptable error of 5% and a confidence interval of 95%, the minimum number of livestock to be sampled in the region was 323.
Based on previous studies on wildlife and dogs, we assumed an expected prevalence of 5% to estimate our sample size. In fact, 5% prevalence of ESLB-E. coli was found in wild rodents in China [34,58], no bacteria were found in a previous study conducted in European wild rabbit in Portugal [59], 4% prevalence was found in wild foxes of Portugal [60], and 8% was found in the only study conducted on dogs in Chile [41]. Based on an expected ESBL-E. coli prevalence of 5%, a margin of acceptable error of 5% and a confidence interval of 95%, the minimum number of animals to be sampled was 73. We aimed to collect 73 samples per wildlife group (e.g., foxes, rabbits, and rodents). However, giving the intrinsic lower density of foxes compared to small mammals and logistic constraints for finding foxes, we expected a much lower sample size for this species.

Microbiology Analyses
Fresh fecal samples were collected using Stuart Transport Medium (Deltalab ® ) and cultured within 3 days of sampling. Swabs were screened for cefotaxime non-susceptible E. coli by direct incubation in standard atmospheric conditions (100 kPa) at 37 • C for 24 h in a MacConkey medium containing 2 µg/mL of cefotaxime sodium salt (Sigma-Aldrich, St. Louis, MO, USA) [61]. Up to 3 isolates with different morphotypes compatible with E. coli per sample/plate were purified and then stored at −80 • C for further analyses. Bac-terial species were confirmed by matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) mass spectrometry (BioMérieux, Marcy l'Etoile, France) at the Genomics and Resistant Microbes (GeRM) Group of the Millennium Initiative for Collaborative Research on Bacterial Resistance (MICROB-R).
Extended-spectrum beta-lactamase production was confirmed in all cefotaxime nonsusceptible E. coli isolates by the double-disk synergy test [30] on Müller Hinton agar (Difco, BD, Sparks, MD, USA) with and without the AmpC inhibitor phenylboronic acid (Sigma-Aldrich). Briefly, disks of ceftriaxone (30 µg), ceftazidime (30 µg), cefepime (30 µg), and aztreonam (30 µg) were used along with a disk of amoxicillin with clavulanic acid (30 µg) placed in the center of the plate at approximately 20 mm. Inhibition zones (ghost zones) observed around any of the cephalosporin disks towards the disk containing the clavulanic acid after 18-20 h of incubation at 37 • C aerobically were considered as a positive result to produce ESBL.
The presence of the most common ESBL-encoding genes in E. coli isolates including bla CTX-M , bla TEM , and bla SHV , was tested by a previously described multiplex PCR [63]. DNA samples from reference bla CTX-M , bla TEM , and bla SHV strains stored at the Universidad de Concepción's Laboratory of Research in Antimicrobial Agents were used as positive PCR controls. The specific group of each CTX-M alleles (CTX-M groups 1, 2, 8, 9, and 25) were detected by multiplex-PCR as described previously [64]. In order to explore the phylogenetic relationships between ESBL-E. coli isolates within and between host species or localities, isolates were fingerprinted by ERIC-PCR according to Bilung et al. [65].

Statistical Analyses
The prevalence of ESBL-E. coli was reported and 95% confidence intervals were calculated using the binom.confint function (Agresti-Coull method) in the binom package in R 3.6.1 [66]. Significant differences in prevalence between populations were tested using the Fisher's exact test in R, since the limited number of observations prevented the use of a Chi-Squared test. We constructed a dendrogram based on the ERIC-PCR electrophoretic patterns using the BioNumerics software v8.0 (Applied Maths, Belgium) and R [65,66]. An UMPGA dendrogram was built based on scaled densitometry curves from the ERIC-PCR obtained from BioNumeric using the hclust function of the dendextended R package.  A total of 47 ESBL-E. coli isolates (confirmed by the double-disk synergy test) from 33 animals were analyzed. Fourteen ESBL-E. coli isolates were obtained from 10 livestock, 32 isolates from dogs and 1 isolate from a mouse. ESBL-E. coli isolates from livestock were resistant to a median (mean) of 1 (2.6) (range: 0-6) out of 8 antibiotics tested, while ESBL-E. coli isolates from dogs were resistant to a median (mean) of 1 antibiotic (1.9) (range: 0-6) ( Figure 3A). Overall, 21% of ESBL-E. coli isolates from livestock and 31% from dogs were susceptible to all antibiotics, 36% of ESBL-E. coli isolates from livestock and 21% from dogs were resistant to one antibiotic, and 43% of ESBL-E. coli isolates from livestock and 48% from dogs were resistant to two or more antibiotics. Additionally, 43% of ESBL-E. coli isolates from livestock, 47% from dogs and an isolate from one rodent were multidrug resistant (MDR). The ESBL-E. coli isolated from a rodent sample was resistant to chloramphenicol, sulfamethoxazole, and ciprofloxacin. More than 20% of ESBL isolates were resistant to ciprofloxacin, chloramphenicol, sulfamethoxazole, and tetracycline in both dogs and livestock. In contrast, no resistance was observed against ertapenem. Among ESBL A total of 47 ESBL-E. coli isolates (confirmed by the double-disk synergy test) from 33 animals were analyzed. Fourteen ESBL-E. coli isolates were obtained from 10 livestock, 32 isolates from dogs and 1 isolate from a mouse. ESBL-E. coli isolates from livestock were resistant to a median (mean) of 1 (2.6) (range: 0-6) out of 8 antibiotics tested, while ESBL-E. coli isolates from dogs were resistant to a median (mean) of 1 antibiotic (1.9) (range: 0-6) ( Figure 3A). Overall, 21% of ESBL-E. coli isolates from livestock and 31% from dogs were susceptible to all antibiotics, 36% of ESBL-E. coli isolates from livestock and 21% from dogs were resistant to one antibiotic, and 43% of ESBL-E. coli isolates from livestock and 48% from dogs were resistant to two or more antibiotics. Additionally, 43% of ESBL-E. coli isolates from livestock, 47% from dogs and an isolate from one rodent were multidrug resistant (MDR). The ESBL-E. coli isolated from a rodent sample was resistant to chloramphenicol, sulfamethoxazole, and ciprofloxacin. More than 20% of ESBL isolates were resistant to ciprofloxacin, chloramphenicol, sulfamethoxazole, and tetracycline in both dogs and livestock. In contrast, no resistance was observed against ertapenem. Among ESBL isolates, the prevalence of resistance to each antibiotic was highly correlated between livestock and dogs (Spearman's test, Rho = 0.90, p < 0.0001), but livestock had a slightly higher prevalence than dogs for most antibiotics ( Figure 3B).
The dendrogram analysis of the ERIC-PCR results showed a high diversity of ESBL-E. coli clones within species and farm localities. No visual clustering by species nor farm localities was observed ( Figure 4). However, ESBL-E. coli isolates from a cow and a dog from the same farm locality clustered together. ESBL-E. coli isolates from dogs were only encoded by the CTX-M genotype while all isolates from livestock carried CTX-M (100%), followed by TEM (14%), and SHV (7%) genotypes ( Figure 3C). Among the most common CTX-M groups searched, 93% of ESBL-E. coli from livestock carried bla CTX-M-group 1 and 36% carried bla CTX-M-group 2 genes ( Figure 3D). Isolates from dogs carried a more diverse pool of CTX-M genotypes with 78% carrying CTX-M from group 1, followed by group 2 (63%), group 9 (12.5%), group 8 (3%, one isolate), and group 25 (3%). The ESBL-E. coli isolate found on a wild mouse carried CTX-M from group 1.
The dendrogram analysis of the ERIC-PCR results showed a high diversity of ESBL-E. coli clones within species and farm localities. No visual clustering by species nor farm localities was observed ( Figure 4). However, ESBL-E. coli isolates from a cow and a dog from the same farm locality clustered together.

Discussion
The spread of AMR at the interface between domestic animals and wildlife remains poorly understood, particularly in low-income rural areas without specific barriers to limit the interaction between domestic and wild animals. In this study, we simultaneously estimated the prevalence of ESBL-E. coli fecal carriage among livestock, dogs, and wild mammals among small-scale agricultural localities of central Chile. The prevalence of ESBL-E. coli fecal carriage was lower in livestock (3%) and wildlife (less than 1%) compared to dogs (24%), suggesting that dogs can be an important carrier of these bacteria in agricultural settings. Dogs carried ESBL-E. coli in the three farms where ESBL-E. coli were detected in livestock, highlighting the potential sharing of these bacteria between dogs and livestock. Among ESBL-E. coli isolates, five CTX-M groups including groups 1, 2, 8, 9, and 25 were detected, with most isolates carrying CTX-M group 1. Molecular typing of ESBL-E. coli by ERIC-PCR showed no cluster of isolates by neither species nor locality, suggesting a wide range of ESBL-E. coli strains circulating on agricultural settings and highlighting the potential for cross-species transmission of either bacteria or antibiotic resistance genes.
ESBL-E. coli have been detected across livestock in South America, with prevalence in cattle ranging from 18% in Brazil to 48% in Peru [12,67]. In this study, we detected ESBL-E. coli fecal carriage in cattle, swine, sheep, and chicken, showing the widespread

Discussion
The spread of AMR at the interface between domestic animals and wildlife remains poorly understood, particularly in low-income rural areas without specific barriers to limit the interaction between domestic and wild animals. In this study, we simultaneously estimated the prevalence of ESBL-E. coli fecal carriage among livestock, dogs, and wild mammals among small-scale agricultural localities of central Chile. The prevalence of ESBL-E. coli fecal carriage was lower in livestock (3%) and wildlife (less than 1%) compared to dogs (24%), suggesting that dogs can be an important carrier of these bacteria in agricultural settings. Dogs carried ESBL-E. coli in the three farms where ESBL-E. coli were detected in livestock, highlighting the potential sharing of these bacteria between dogs and livestock. Among ESBL-E. coli isolates, five CTX-M groups including groups 1, 2, 8, 9, and 25 were detected, with most isolates carrying CTX-M group 1. Molecular typing of ESBL-E. coli by ERIC-PCR showed no cluster of isolates by neither species nor locality, suggesting a wide range of ESBL-E. coli strains circulating on agricultural settings and highlighting the potential for cross-species transmission of either bacteria or antibiotic resistance genes.
ESBL-E. coli have been detected across livestock in South America, with prevalence in cattle ranging from 18% in Brazil to 48% in Peru [12,67]. In this study, we detected ESBL-E. coli fecal carriage in cattle, swine, sheep, and chicken, showing the widespread dissemination of these bacteria in agricultural settings. This is the first report of ESBL-E. coli in cattle in Chile, although their prevalence was low (3%) compared to a similar study in Peru estimating a prevalence of 48% among small-scale farmers in the Lima region [12]. The observed prevalence in Chile is similar to farms in high-income countries such as France or Denmark, where the restriction of third-generation cephalosporins has been associated with a reduction in ESBL-E. coli [68,69]. The high prevalence of resistance to ciprofloxacin (over 60%) found in ESBL-E. coli isolated from domestic animals in this study is consistent with the high level of plasmid-mediated quinolone resistant found in 74% of ESBL-E. coli isolated from Chilean hospitals [70] and a high prevalence of resistance to ciprofloxacin (84%) in ESBL-E. coli recovered from intensive care units of Southern Chile [71]. The presence of ESBL-E. coli could result from low but existing selective pressure by the use of third generation cephalosporins in these farms, which requires further investigation. In a similar agricultural setting of Peru, the low use of cephalosporins [72] was associated to a high prevalence of ESBL-E. coli in livestock (50%) [12], suggesting that factors other than antibiotic use can influence AMR. For example, farm hygiene, herd size, contact with humans or other husbandry conditions such as storage of slurry in a pit have been associated with the presence of ESBL-E. coli in livestock [13,20,21].
The low prevalence of ESBL-E. coli in wildlife (less than 1%) is similar to other studies focusing on ESBL-E. coli among wildlife in Latin America and other LMICs [12,73]. For example, a previous study estimated a 4% prevalence of ESBL-E. coli among vampire bats (Desmodus rotundus) in Peru using a similar methodology for screening [12]. Previous studies conducted in Chile and Latin America have detected the presence of ESBL-E. coli on wild birds including gulls [36], Andean condors [43], and three species of owls [42]. Likewise, bla CTX-M genes have been previously detected using qPCR methods from feces in Andean foxes [52] and the guiña (Leopardus guigna) [74], although the bacteria species carrying the genes, and whether it was expressed or not, remains unknown. To our knowledge, this is the first study to report E. coli carrying CTX-M group 1 on wild mammals in Chile. The origin of ESBL-E. coli found in a rodent remains to be clarified. Given the presence of similar bla CTX-M genes among a nearby farm and a wide variety of ESBL-E. coli strains circulating, one potential explanation is the transmission of bla CTX-M from domestic animals, although other potential contamination sources (e.g., humans, water contamination) cannot be discarded.
The high prevalence of ESBL-E. coli found in dogs (24%) highlights their role as either passive 'receivers' or reservoirs of ESBL-E. coli in agricultural settings. Although there are only a limited number of studies estimating the prevalence of ESBL-E. coli among dogs, previous studies have shown a prevalence in Latin American dogs ranging from 9-30%, and a global prevalence of 7% [30,[75][76][77][78][79]. The detection of ESBL-E. coli in dogs has been associated with previous antibiotic treatment, but also close contact with livestock, implying the potential transmission of these bacteria between livestock and dogs [29,30,80]. The latest is also suggested by our study, as the three farms where we detected ESBL-E. coli in livestock also had a dog carrying ESBL-E. coli. Molecular typing by ERIC-PCR showed no cluster of ESBL-E. coli by host species, while isolates sampled from a cow and a dog at the same farm clustered together. These results suggest that bacterial strains or ESBL genes such as bla CTX-M could be exchanged between host populations. Overall, the circulation of ESBL-E. coli among dogs highlights the potential public health risk for domestic animals but also for dog owners, given the potential spillover of bacteria from dogs to humans [28,29,81]. Moreover, the higher prevalence observed in dogs compared to livestock suggests that ESBL-E. coli could be spreading from dogs to livestock, and not necessarily in the other direction, as most previous studies have assumed.
Our study constitutes one of the first One Health approaches to simultaneously address the circulation of ESBL-E. coli among livestock, dogs, and wildlife in a rural setting. However, several future research can complement our findings and provide further insight into the selection and spread of AMR among these compartments. First, the limited sample size of foxes prevented a more accurate estimation of ESBL-E. coli prevalence in this species. Thus, we could not conclude whether predators or preys are more likely to carry ESBL-E. coli in this setting. Secondly, the low selective pressure for ESBL-E. coli should be confirmed by studies on antibiotic use among farmers in these agricultural settings [72], which are currently lacking in Chile. Although the use of antibiotics in Chilean terrestrial livestock remains unknown, the national health authority (Servicio Agricola Ganadero) advises the use of fluroquinolones and cephalosporins as a last resource antibiotic in livestock, following a susceptibility test [82]. Antibiotic residues of tetracyclines, betalactams, aminoglycosides, and macrolides have been found in eggs from backyard poultry production [83]. Thirdly, although the ERIC-PCR technique used has a high resolution and allows us to differentiate among E. coli strains from the same locality and host species [65], several other molecular techniques can improve our understanding of the transmission dynamics of resistance genes and E. coli. For example, future work could determine the pathogenic potential of these strains using whole genome sequencing, or whether bla CTX-M genes are carried by specific mobile elements such as plasmids. Finally, future research should identify associated factors to ESBL-E. coli fecal carriage in each animal population (e.g., individual characteristics of dogs and cattle).

Data Availability Statement:
The data presented in this study are available within this article.