Antimicrobial use and resistance in Escherichia coli from healthy food-producing animals in Guadeloupe

Background Selection pressure exerted by use of antibiotics in both human and veterinary medicine is responsible for increasing antimicrobial resistance (AMR). The objectives of this study were to better understand antimicrobial use in pigs, beef cattle, and poultry on farms on Guadeloupe, French West Indies, and to acquire data on AMR in Escherichia coli in these food-producing animals. A cross-sectional survey was conducted at 45 farms on Guadeloupe, and practical use of antimicrobials was documented in declarative interviews between March and July 2018. A total of 216 fecal samples were collected between January 2018 and May 2019, comprising 124 from pigs, 75 from beef cattle, and 17 from poultry litter. E. coli isolates were obtained for further testing by isolation and identification from field samples. Antimicrobial susceptibility testing and screening for blaCTX-M, blaTEM, tetA, and tetB resistance genes by polymerase chain reaction on extracted genomic DNA were performed. Results The study showed rational use of antimicrobials, consisting of occasional use for curative treatment by veterinary prescription. Tetracycline was the most commonly used antimicrobial, but its use was not correlated to E. coli resistance. Extended-spectrum β-lactamase (ESBL) E. coli isolates were detected in 7.3% of pigs, 14.7% of beef cattle, and 35.3% of poultry. blaCTX-M-1 was the predominant gene found in ESBL-E. coli isolates (68.8%), followed by blaCTX-M-15 (31.3%). Conclusion Despite rational use of antimicrobials, the rate of ESBL-E. coli in food-producing animals in Guadeloupe, although moderate, is a concern. Further studies are in progress to better define the genetic background of the ESBL-E. coli isolates. Graphical abstract

spectrum β-lactamase (ESBL)-producing Enterobacteriaceae, which hydrolyze key antimicrobials, such as the expanded-spectrum cephalosporins cefotaxime, ceftriaxone, ceftazidime, and cefepime, is due mainly to the selective pressure of antibiotics used in both human and veterinary medicine [7]. In 2014, the antimicrobial consumption was higher in animals (152 mg of active substance per kg of estimated biomass) than in humans (124 mg/kg) in Europe. Consumption of 3rd-and 4thgeneration cephalosporins was associated with resistance in E. coli in humans. Tetracyclines and polymyxins resistance in E. coli from animals was associated with corresponding antimicrobial consumption in animals [8]. The resistance is mediated mainly by acquired ESBL genes located on mobile genetic elements and is frequently associated with resistance genes against several families of antimicrobial agents [9].
Guadeloupe, a French overseas department in the Caribbean, has been classified as a very high-resource country [10]. Less than one third of the surface of this small island is devoted to agriculture, and the livestock in 2018 comprised 14,500 pigs, 44,900 cattle, and 507,000 laying hens and broilers [11]. The latest of the few studies on AMR on Guadeloupe showed a low prevalence of ESBL-producing Enterobacteriaceae in human community-acquired urinary tract infections [12] and in wastewater treatment plants [13]. As there is close contact between humans and domestic animals, the animals may be reservoirs of resistance genes; however, the only data on antibiotic resistance in local domestic animals is a study on horses, which showed the emergence of various ESBL-producing E. coli clones, some of which persisted for more than a month after antibiotic treatment [14].
The main objective of this work was to obtain information on antimicrobial use on farms on Guadeloupe and on AMR levels in the zoonotic indicator bacteria E. coli in pigs, beef cattle, and poultry.

Results
In our survey of use of medication containing antibiotics in pigs, beef cattle, and poultry on Guadeloupe, 64.4% (29/45) of all farmers reported their use during the last year (Table 1). Beef cattle were individually treated, whereas collective treatments to the entire flock were administrated in pig and poultry farms. Antimicrobials were usually administered as curative treatment (20/29, 69.0%) and under veterinary prescription (22/29, 75.9%). The main causes for which antimicrobials were given were respiratory diseases in pigs (5/11, 45.5%), skin diseases in cattle (5/12, 41.7%), and respiratory and digestive diseases in poultry (4/6, 66.7%) ( Table 1).
A total of 216 fecal samples were collected from foodproducing animals on 28 farms and in the slaughterhouse (34 additional farms). Samples were collected from 75 adult beef cattle for meat production (51 on farms and 24 at the slaughterhouse), 124 pigs included piglets, weaned, finishers, sows and breeding males (90 on farms and 34 at the slaughterhouse), and 17 hen houses representing 53,000 poultry. The prevalence of ESBL-E. coli was higher on poultry farms (4/9, 44.4%) than on beef cattle farms (4/32, 12.5%) or pig farms (3/ 21, 14.3%); however, the poultry samples were from litter, representing more than one bird (Table 2).
On the basis of the observed frequencies of AMR phenotypes, corresponding to ESBL-E. coli isolated on ceftriaxone selective plates and on non-selective plates for non-ESBL-E. coli, the highest levels of resistance were against ampicillin, cefotaxime, streptomycin, tetracycline, and trimethoprim-sulfamethoxazole. Significant differences in the frequencies of resistant E. coli isolates were found among the three production systems (Tables 1 and 3).
We compared the rates of resistance to tetracycline, ampicillin, streptomycin, and trimethoprim-sulfamethoxazole according to the antibiotic or class of antibiotics reported in the survey, according to animal species. The number of farms in the ATB use survey is smaller than the number of farms investigated for fecal sampling. As mentioned above and shown in Table 1, although only one third of poultry farmers who used antibiotics used tetracycline as collective treatment, a high prevalence of tetracycline-resistant E. coli were found in poultry ( Table 3). Use of β-lactams also did not correspond to the level of resistance to β-lactams in poultry (17/23, 73.9%), which did not receive these drugs. The proportion of tetracycline-, ampicillin-, cefotaxime-, streptomycin-, and trimethoprim-sulfamethoxazole-resistant ESBL-E. coli was similar to non-ESBL-E. coli from poultry (Table 3). ESBL-E. coli frequency was not associated to the rate of tetracycline and streptomycin resistance occurrence despite the use of these antimicrobials by half of the pig farmers. These resistances occurred Veterinarian is the drug supplier d Skin disease and respiratory pathology 0 Skin disease, respiratory and digestive pathologies 0 Digestive pathology and other c 0 Antimicrobials used β-lactams + phenicols + colistin + macrolides independently of the detection of an ESBL. ESBL-E. coli frequency was associated to the rate of ampicillin (100.0% of ampicillin resistance in ESBL-E. coli vs 14.4 in non-ESBL-E. coli, P < 0.001) and trimethoprim-sulfamethoxazole resistance occurrence (72.7% of trimethoprim-sulfamethoxazole resistance in ESBL-E. coli vs 13.7%, in non-ESBL-E. coli P < 0.001), while β-lactams were used by half of the pig farmers and trimethoprimsulfamethoxazole was not declared to be used. Regardless of the antimicrobials used, ESBL-E. coli isolates from beef cattle, individually treated, were significantly associated with other resistance carriage (P ≤ 0.025) ( Table 3). Congruence was observed between the absence of quinolone use and a low frequency of nalidixic acid-resistant E. coli in the three animal species.
One molecule used for 2 distinct pathologies More than one molecule used for the same pathology    The ESBL isolates harbored predominantly the bla CTX-M-1 gene (22/32, 68.8%), followed by bla CTX-M-15 (10/32, 31.3%) ( Table 4). The bla CTX-M-1 gene was combined with the bla TEM-1C in two pigs and with the bla TEM-1B gene in one poultry isolate. The remaining ESBL-E. coli carried a bla CTX-M-15 gene, usually with combined cefotaximase and ceftazidimase activity. A comparison of phenotypic and genotypic profiles based on combined patterns analysis of AMR and antimicrobial resistance genes (ARG) is shown in Fig. 1. The bla CTX-M genes were carried by ESBL-E. coli isolates found in the three food-producing animal systems on four farms in distinct geographic areas. The comparative analysis generated 18 distinct patterns of the 32 ESBL-E. coli (Fig. 1), with 20 (62.5%) isolates grouped into seven clusters with similar AMR/ARG patterns (A-G) comprising two to five isolates, whereas 12 (37.5%) distinct patterns were not clustered. Three clusters (A, D, G) included nine ESBL-E. coli at the same farm, whereas the other clustered ESBL-E. coli (B, C, E, F) were not specific to a production system. Twelve clustered isolates with similar AMR/ARG profiles were found in different food-producing farms (12/32, 37.5%); e.g. one cluster of ESBL-E. coli carriers (C) consisted of two pigs and one hen house on three different farms.

Discussion
This study of antimicrobial use on small-scale pig, beef cattle, and poultry farms in Guadeloupe showed moderate ESBL-E. coli in pig and beef cattle production, probably because of rational use of antimicrobials. The island adheres to the French AMR reduction plan [15] on the use of veterinary antimicrobials, and the moderate ESBL resistance may reflect its effectiveness. The frequencies were nevertheless higher than those in French national surveillance for AMR in infected animals in 2018, in which E. coli isolates resistant to third-and fourthgeneration cephalosporins were detected in 2.3% of beef cattle and in < 2.0% of pigs, poultry, and turkeys [16]. The differences may be due to sampling of diseased rather than healthy animals for detection of ESBL-E. coli. The prevalence in our study is closer to that observed in Portugal, where 5.7% of fecal samples from 35 healthy pigs and 10% of those from healthy chickens were positive for ESBL-E.
coli [17], but lower than that in Switzerland, where 15.3% of pigs and 13.7% of bovine fecal samples were positive [18]. In European countries, the occurrence of E. coli resistance in healthy animals at slaughterhouse varied from 0 to 7.9% in fattening pigs; from 0 to 5.9% in calves under 1 year of age [19]. Country-and production-specific factors may influence the occurrence of resistance [20,21]. In our study, the patterns of resistance of most of the ESBL-E. coli isolates were either farm-specific or were shared by isolates from distant farms and distinct animal species. This observation suggests that ESBL-producing E. coli and their resistance profile spread within farms or arise independently.
None of the E. coli isolates were resistant to any of the quinolones tested (enrofloxacin, ciprofloxacin, nalidixic acid), probably reflecting the low use of these antibiotics in food-producing animals. A previous study showed a significant positive correlation between antibiotic dose and the occurrence of antibiotic-resistant bacteria in animal feces [22]. In our study, the discrepancy between the use of antimicrobials and the level of resistance is striking, as antimicrobial use was not always linked to AMR level, especially for β-lactams and tetracyclines. A potential bias might be underreported use of antimicrobials, which would explain the inverse relation, but there may be other reasons. The low level of resistance to tetracycline in beef cattle, despite the number of farms that used this drug, might be due to targeted treatment in small beef cattle production rather than the collective treatment used in larger-scale pig and poultry husbandry. Antimicrobial treatment also reflects the veterinary cost, which is lower for collective administration, e.g. to poultry. Collective treatment might therefore contribute to the emergence of resistance and should therefore be more closely controlled.
The high frequency of ESBL resistance observed in broilers (35.3%) and to a lesser extent in pig and beef cattle farms with no use of third-generation cephalosporins is difficult to explain. We tested imported 7day old chicks 1 day after arrival from mainland France but found no resistance to these antimicrobials. It has been shown that a rapid increase in ESBL-E. coli prevalence in the first week of life must be due to factors other than latent contamination of the majority of birds at arrival [23]. Therefore, as no third-generation cephalosporins were administered in the production systems in which ESBL-E. coli resistance was detected, the observed resistance to these drugs was probably due to co-selection of several resistance genes in the same genetic determinant by other antibiotics, [24], but not specifically documented here. It has also been reported that tetracycline resistance in commensal E. coli is often linked to resistance to other antimicrobials, such as ampicillin and trimethoprim-sulfonamides [21]. A recent study on a Danish pig production farm showed clearly that commonly used antimicrobials such as tetracycline, which are not listed as critically important for human treatment, can promote resistance to critically important antimicrobials, limiting treatment possibilities [25]. A recent metagenomic study on bacterial communities showed that tetracycline resistance is often found in ESBL isolates and transmitted with ESBLcontaining plasmids [26]. Moreover, an integrated approach to AMR found an increased prevalence of integrons containing resistance genes in tetracyclineresistant isolates [27]. The wide use of tetracyclines in Guadeloupe may explain some of the disproportion between the prevalence of resistance and the use of third-generation cephalosporins. These results reinforce the importance of animal food-producing systems as a reservoir of mobile genetic elements carrying multiple resistance determinants.

Conclusion
Our study provides the first baseline information on levels of antimicrobial use, on the dynamics of phenotypic and genotypic resistance to tetracyclines, and on ESBL-E. coli in small-scale pig, beef cattle, and poultry production on Guadeloupe. Despite rational use of antimicrobials, E. coli resistant to third-generation cephalosporins were found on the farms. Mechanisms other than selective pressure of these antimicrobials in the emergence of AMR remain to be elucidated.

Survey design
A prospective survey on the use of antimicrobial agents in veterinary medicine and food animal production was conducted between March and July 2018 on 14 pig, 16 beef cattle, and 15 poultry production farms. The farms were selected randomly in 16 of the 32 townships of the island to ensure representative production, covering small-to large-scale cooperative or independent production. All the pig facilities were farrow-to-finishing farms, with 30-3120 head per farm, for a total of 8549 pigs, representing 59.0% of pig production on Guadeloupe. Beef cattle breeding was investigated on 13 small-scale grassland farms (≤ 90 head) and three large-scale farms, for a total of 1691 head (mean age of 4.1 ± 2.4 years), representing 4.3% of local meat production. The poultry breeder farms had 400-64,000 birds, for a total of 184, 510, representing 36.4% of local production [11].
Antimicrobial use was documented in declarative faceto-face interviews with farmers by an agronomist using a questionnaire specific for the study. Each participant provided information on farm characteristics (size, number of head) and routines for antimicrobial use, including frequency, reasons for treatment, name of the antibiotic drug used, route of administration, and estimated annual cost of treatment, including laboratory analyses, veterinary services, and drug purchase.

Sampling and collection
Between January 2018 and May 2019, 11 pig farms, eight beef cattle farms, nine poultry farms, and the only slaughterhouse for beef cattle and pigs on Guadeloupe, representing respectively 24 and 10 farms, were screened for E. coli. As most small herds of beef cattle were raised free in fields, sampling was more difficult than that of pigs or poultry confined in blocks and is therefore less representative of the total production (4.3%) [11]. During the study, 216 fecal samples (30 g) were collected randomly just after excretion (124 from pigs of which 34 were slaughterhouse pigs and 75 from beef cattle of which 24 were slaughterhouse beef cattle). Fecal material from 17 hen houses was sampled by walking on litter approximately 100 m around a flock in boot socks (Sterisocks Tryptone SodiBox, Nevez, France). All samples were stored and transported in sterile cups or bags on ice to the Institut Pasteur laboratory within 4 h. Samples were stored at 4°C and processed within 8 h of sampling.

Bacterial isolation and identification
A 10-μL loop of each fecal sample was mixed in Luria Bertani broth BD Difco™ (Humeau, La Chapelle-sur-Erdre, France) supplemented or not with 4 mg/L of ceftriaxone and incubated at 37°C for 24 h. Selective enrichment with 4 mg/L of ceftriaxone were streaked on chromogenic coliform agar plates (CHROMagar™, Paris, France) supplemented with 4 mg/L of ceftriaxone. Nonselective enrichments without 4 mg/L of ceftriaxone were streaked on chromogenic coliform agar plates without 4 mg/L of ceftriaxone. All plates were incubated at 37°C for 24 h. One susceptible and three resistant metallic blue colonies were randomly picked up from nonselective and selective chromogenic coliform agar, respectively and identified by matrix-assisted laser desorption and ionization time-of-flight mass spectrometry on an Axima performance spectrometer (Shimadzu Corp, Osaka, Japan).

Resistance gene screening
For molecular characterization of ARG, genomic bacterial DNA was extracted from one colony with the Insta-Gene™ Matrix kit (Biorad, California, USA), according to the manufacturer's instructions. ESBL and tetracycline resistance coding genes were screened by PCR in all E. coli tetracycline-resistant isolates. bla CTX-M multiplex PCR including phylogenetic groups 1, 2, and 9 was performed [29]. bla TEM gene was screened by simplex PCR [30]. Amplified PCR products were sequenced (Eurofins, Ivry-sur-Seine, France) and compared with known resistance gene sequences in the GenBank database by multiple-sequence alignment with the Basic Local Alignment Search Tool program for further characterization. Tetracycline-resistant isolates were screened for the presence of tetA and tetB genes with specific tetA primers designed for this study (tetA-F 5′-TAGAAGCC GCATAGATCGCC-3′ and tetA-R 5′-GCTTCATGAG CGCCTGTTTC-3′) and published specific tetB primers [31]. The duplex PCR amplification conditions for tetA and tetB were optimized as follow: 5 min at 95°C, followed by 35 cycles at 95°C for 30 s, at 62°C for 30 s, and at 72°C for 30 s, followed by a final extension at 72°C for 7 min. The amplicons were detected by gel electrophoresis. For quality control, a subsample of 10% was genotyped twice.

Combined numerical analysis
The combined numerical analysis was performed on ESBL-producing E. coli patterns with BioNumerics® v6.6 software (Applied Maths NV). Each file with experimental data from AMR and ARG screening was merged as a composite data set in the BioNumerics® database, with the similarity coefficient option taken from each experiment. The matrices from the individual experiments were averaged according to the same defined weight, and an individual similarity matrix was calculated in such a way that all characters had an equal influence on similarity. A dendrogram was drawn by using the unweighted pair group method with arithmetic averages with a tolerance of 1% to show the similarity of combined AMR and ARG patterns of the bacterial isolates.

Statistical analysis
Results are presented as means ± standard deviation, medians with the interquartile ranges for quantitative variables, and numbers and percentages for qualitative variables. Intergroup differences among farms classified according to food-producing animal category were assessed with the Kruskal-Wallis or chi-square test, as appropriate. The level of significance was defined as P < 0.05. Analyses were conducted with STATA® 11.2.