Antimicrobial resistance in fecal Escherichia coli and Salmonella enterica isolates: a two-year prospective study of small poultry flocks in Ontario, Canada

Background Although keeping small poultry flocks is increasingly popular in Ontario, information on the antimicrobial susceptibility of enteric bacteria of such flocks is lacking. The current study was conducted on small poultry flocks in Ontario between October 2015 and September 2017, and samples were submitted on a voluntary basis to Ontario’s Animal Health Laboratory. From each submission, a pooled cecal sample was obtained from all the birds of the same species from the same flock and tested for the presence of two common enteric pathogens, E. coli and Salmonella. Three different isolates from each E. coli-positive sample and one isolate from each Salmonella-positive sample were selected and tested for susceptibility to 14 antimicrobials using a broth microdilution technique. Results A total of 433 fecal E. coli isolates (358 chicken, 27 turkey, 24 duck, and 24 game bird) and 5 Salmonella isolates (3 chicken, 1 turkey, and 1 duck) were recovered. One hundred and sixty-seven chicken, 5 turkey, 14 duck, and 15 game bird E. coli isolates were pan-susceptible. For E. coli, a moderate to high proportion of isolates were resistant to tetracycline (43% chicken, 81% turkey, 42% duck, and 38% game bird isolates), streptomycin (29% chicken, 37% turkey, and 33% game bird isolates), sulfonamides (17% chicken, 37% turkey, and 21% duck isolates), and ampicillin (16% chicken and 41% turkey isolates). Multidrug resistance was found in 37% of turkey, 20% of chicken, 13% of duck, and 8% of game bird E. coli isolates. Salmonella isolates were most frequently resistant to streptomycin, tetracycline, and sulfonamides. Resistance to cephalosporins, carbapenems, macrolides, and quinolones was infrequent in both E. coli and Salmonella isolates. Cluster and correlation analyses identified streptomycin-tetracycline-sulfisoxazole-trimethoprim-sulfamethoxazole as the most common resistance pattern in chicken E. coli isolates. Turkey E. coli isolates compared to all the other poultry species had higher odds of resistance to tetracycline and ampicillin, and a higher multidrug resistance rate. Conclusions Escherichia coli isolates were frequently resistant to antimicrobials commonly used to treat poultry bacterial infections, which highlights the necessity of judicious antimicrobial use to limit the emergence of multidrug resistant bacteria.


Background
Non-commercial poultry flocks (denoted as "small flocks") are increasingly popular in urban, suburban, and rural areas in North America [1][2][3]. Small flocks can pose a health risk to their owners by exposing them to zoonotic pathogens [4][5][6][7] through consumption of contaminated meat or eggs [8,9], or direct contact with infected birds [10] or their environment [11]. Antimicrobial resistance (AMR) in zoonotic pathogens adds to this risk [12][13][14][15] because infections with antimicrobial resistant bacteria are more difficult to treat, and result in higher morbidity and mortality [16,17]. Inappropriate antimicrobial use has been shown to be one of the main causes for the development of AMR in commensal and pathogenic bacteria of poultry [17,18]. Exposure to an individual antimicrobial may cause the bacteria to develop resistance to multiple antimicrobials if resistance genes are located on mobile genetic elements [19]. Furthermore, these acquired resistance determinants can persist even after the antibiotic selection pressure ends [12].
Health Canada categorizes antimicrobials based on their importance in human medicine: I -very high importance; II -high importance; III -medium importance; and IV -low importance [20]. The classification system considers the antimicrobial's indication (e.g., preferred choice for treatment of serious human infections) and availability of replacements (e.g., limited substitutes available) [20]. Under an amended Canadian regulation that came into effect on December 1, 2018, all medically important antimicrobials (Categories I, II, and III) used in food animals require a veterinary prescription in order to help limit the development and spread of AMR [21].
In Canada, AMR of Escherichia coli and Salmonella isolates obtained from commercial broiler chicken and turkey flocks is monitored by the Canadian Integrated Program for Antimicrobial Resistance Surveillance [22]. Emergence of resistance to antimicrobials commonly used to treat bacterial infections in commercial poultry flocks in Ontario is well documented [22][23][24]. In contrast, only one study [25], which was conducted in provincially inspected slaughter plants, has documented resistance to antimicrobials in small flocks in Ontario. Thus, the objectives of this study were to evaluate AMR patterns of fecal E. coli and Salmonella enterica isolates of chickens, turkeys, waterfowl, and game birds from Ontario small flocks submitted for laboratory diagnostic testing because of morbidity or mortality, and to determine differences in AMR patterns among different poultry species.

Description of submissions
Over the 2-year period, the Animal Health Laboratory received 160 small flock submissions, with a median of 1 bird per submission (range = 1-5), from flocks ranging in size from 1 to 299 birds (median 25) and birds ranging in age from 6 days to 7 years (median 7 months). The majority of submissions were chickens (84%, 134 submissions), although a few turkey (10), duck (8), and game bird (8) submissions were also received [26].
In the chicken E. coli isolates, there was a high frequency of resistance (≥40% of isolates) to tetracycline, a moderate frequency of resistance (15-39% of isolates) to streptomycin, sulfisoxazole, and ampicillin, and a low frequency of resistance (5-14% of isolates) to trimethoprimsulfamethoxazole and gentamicin ( Table 1). All of the other antimicrobials tested had a very low frequency of resistance (< 5%). In the turkey E. coli isolates, there was a high frequency of resistance to tetracycline and ampicillin, and a moderate frequency of resistance to streptomycin and sulfisoxazole. In the duck E. coli isolates, there was a high frequency of resistance to tetracycline, and a moderate frequency of resistance to sulfisoxazole and trimethoprimsulfamethoxazole. In the game bird E. coli isolates, there was a moderate frequency of resistance to tetracycline and streptomycin.
A high (≥40%) proportion of E. coli-positive samples were resistant to tetracycline (62.50% of the chicken, 100% of the turkey, 50% of the duck, and 57.14% of the game bird samples), streptomycin (42.50% of the chicken, 55.56% of the turkey, and 42.86% of the game bird samples), ampicillin (55.56% of the turkey samples), and sulfisoxazole (55.56% of the turkey samples) (Fig. 1).
Single-linkage clustering dendrograms with Jaccard distances for E. coli resistance are presented in Fig. 2. A relatively high proportion (i.e., a cluster) of the chicken E. coli isolates were resistant to streptomycin, tetracycline, sulfisoxazole, and trimethoprim-sulfamethoxazole; a second cluster of chicken E. coli isolates was resistant to cefoxitin and ceftriaxone. Other notable clusters included resistance to streptomycin, sulfisoxazole, ampicillin, and tetracycline (turkey E. coli isolates), amoxicillin-clavulanic acid, cefoxitin, and ceftriaxone (turkey E. coli isolates), sulfisoxazole and trimethoprim-sulfamethoxazole (duck E. coli isolates), and streptomycin and tetracycline (game bird E. coli isolates). The turkey, duck, and game bird E. coli isolates were pan-susceptible to several antimicrobials.

Differences in AMR of fecal E. coli isolates between poultry species
The odds of resistance to tetracycline (odds ratio = 5.89, 95% CI = 1.71-20.29, P = 0.005) and ampicillin (odds ratio = 4.06, 95% CI = 1.24-13.25, P = 0.020) were significantly higher in turkey E. coli isolates compared to isolates from all the other poultry species. The rate of MDR was significantly higher (incidence rate ratio = 1.99, 95% CI = 1.16-3.40, P = 0.012) in turkey E. coli isolates compared to isolates from all the other poultry species.

Discussion
Our study evaluated AMR in fecal E. coli and Salmonella enterica isolates from birds from small flocks experiencing morbidity, mortality, or production issues, and found a relatively high frequency of resistance to antimicrobials commonly used to treat bacterial infections in poultry. Differences in flock characteristics, including  [4,28], Bangladesh [29], and Chile [30]. The Salmonella serotypes identified in our study (Anatum, Indiana, Ouakam, Uganda, and Montevideo) are not among the most prevalent commercial poultry-associated serotypes (Kentucky, Enteritidis, Heidelberg, and Typhimurium) in Canada [31] or the United States [32], and they are less frequently associated with human salmonellosis cases in Canada [33]. Nonetheless, the presence of AMR in Salmonella in small flocks is concerning because of the risk that resistant salmonellae pose in human cases with bacteremia or a compromised immune system. Although the frequency of Salmonella enterica was very low in our study [26], and many of the Salmonella isolates were pan-susceptible, some were resistant to streptomycin alone, or to streptomycin, sulfisoxazole, trimethoprim-sulfamethoxazole, and tetracycline. Our findings agree with a recent study that evaluated AMR in Salmonella isolated from small flock cases at the California Animal Health and Food Safety Laboratory System, and found resistance to streptomycin, sulfonamides, and tetracycline [34].
In the fecal E. coli isolates in our study, there was a very low frequency of resistance to cephalosporins, carbapenems, macrolides, and quinolones, which are antimicrobials classified in Canada as being of very high or high importance in human medicine [20]. This is an encouraging finding from a human health viewpoint because flock owners can be exposed to antimicrobial resistant zoonotic pathogens [6,13,15] through direct contact with their birds [10] or their environment [11], or consumption of contaminated meat or eggs [8,9]. However, there was a moderate to high frequency of resistance to tetracycline, streptomycin, sulfonamides, and ampicillin; antimicrobials frequently used to treat bacterial infections in poultry [35]. These findings were in accordance with other small flock studies in Canada (tetracycline resistance 37%, streptomycin 21%, sulfisoxazole 16%, ampicillin 15%) [25] and Ecuador (tetracycline 69%, streptomycin 42%, sulfisoxazole 65%, ampicillin 45%) [36], and one study of commercial broiler chicken flocks in Canada (tetracycline 53%, streptomycin 33%, sulfisoxazole 18%, ampicillin 38%) [23].
Our cluster and correlation analyses of the chicken E. coli isolates showed that there was concurrent resistance to  streptomycin, tetracycline, sulfisoxazole, and trimethoprimsulfamethoxazole; clusters of turkey, duck, and game bird isolates included many of the same antimicrobials. Our cluster analyses also showed that there was concurrent resistance to cefoxitin and ceftriaxone in the chicken E. coli isolates and to amoxicillin-clavulanic acid, cefoxitin, and ceftriaxone in the turkey E. coli isolates. The moderate frequency of MDR in the turkey and chicken E. coli isolates (and to a lesser extent in the duck and game bird E. coli isolates), and the strong correlations in resistance of the chicken E. coli isolates to antimicrobials commonly used to treat bacterial infections of poultry, highlight the importance of judicious antimicrobial use to limit the development and dissemination of multidrug resistant bacteria in small flocks [12,13].
Our regression models showed that there were higher probabilities of resistance to tetracycline and ampicillin in the E. coli isolates of turkeys when compared to isolates obtained from all the other poultry species. Moreover, the rate of MDR was significantly higher in the turkey isolates compared to all the other species.
Differences in AMR between poultry species might be explained by variation in antimicrobial use or husbandry practices. However, these findings should be interpreted cautiously because chicken isolates were over-represented in our study. Therefore, further studies are needed to assess factors that might have a role in the development of AMR in commensal and pathogenic enteric bacteria of small flocks.
Limitations of this study include a sampling bias, as most submissions came from southwestern and eastern Ontario, which might have been the consequence of the geographic proximity to the diagnostic laboratories in Guelph and Kemptville, respectively [26]. Also, small flocks were not randomly selected, and our study included only owners who had a flock veterinarian, as ciprofloxacin; NAL = nalidixic acid; TET = tetracycline. B A cluster analysis, using the Jaccard binary similarity coefficient, was used to compare individual antimicrobials with respect to their similarity in the resistance status of E. coli. The Jaccard distance measures dissimilarity between antimicrobials and is obtained by subtracting the Jaccard binary similarity coefficient from one [27]. A high dissimilarity measure indicates that relatively few isolates were resistant to both antimicrobials, a low dissimilarity measure indicates that a relatively high proportion of isolates were resistant to both antimicrobials, and a dissimilarity measure of zero indicates that all isolates were susceptible to both antimicrobials this is a laboratory submission requirement. We also used fecal samples from diagnostic submissions and not from healthy birds. Our study might therefore overestimate the frequency of AMR because samples came from birds that might have already been treated with antimicrobials.

Conclusions
Our study enhances the knowledge on AMR of small flocks by evaluating the AMR patterns of E. coli and Salmonella isolates from chickens, turkeys, ducks, and game birds. These results can be used as a benchmark for ongoing monitoring of AMR in enteric bacteria of small flocks in Ontario, in light of the recently amended antimicrobial use regulation in Canada. Ultimately, the findings derived from this study can be used to educate veterinarians and small flock owners about issues surrounding AMR, with a goal of reducing the presence of multidrug resistant bacteria in small flocks and mitigating the risk they might pose to public health.

Study design
Samples were obtained through a prospective surveillance study of small flocks conducted in Ontario between October 2015 and September 2017, which is described in detail elsewhere [26]. In brief, a small flock was defined as a non-commercial poultry flock composed of not more than 299 broiler chickens, 99 layer chickens, 49 turkeys, 300 waterfowl, or 300 game birds. Small flock owners who had issues with production, clinical illness, or mortality in their flock were provided the opportunity to submit birds for diagnostic testing for a discounted fee. Submissions (n = 160) were made to the Animal Health Laboratory, University of Guelph through the owner's veterinarian. A maximum of 5 sick and/or dead birds of one species from the same flock constituted a submission. Live birds submitted to the lab were euthanized using carbon dioxide.

Sample collection and bacterial isolation
All bacterial isolation and antimicrobial susceptibility testing were conducted at the Animal Health Laboratory, Guelph, Ontario. From each submission, one pooled cecal sample was collected (from all of the birds of the same species from the same flock in the submission) and tested for fecal E. coli and Salmonella spp. Cecal material was directly plated on MacConkey and Hektoen enteric agars (Oxoid Ltd., Nepean, ON) for E. coli isolation, and inoculated into buffered peptone water (Bio-Media Unlimited Ltd., Toronto, ON) for Salmonella spp. preenrichment. Aliquots of buffered peptone water were then transferred to Hajna tetrathionate (Animal Health Laboratory, Guelph, ON) and Rappaport Vasiliadis broths (Bio-Media Unlimited Ltd.) for Salmonella spp. enrichment, followed by plating on brilliant green (Bio-Media Unlimited Ltd.) and XLT-4 agars (Oxoid Ltd.). Presumptive E. coli and Salmonella spp. colonies were identified using matrix-assisted laser desorption ionization time-of-flight mass spectrometry (Bruker Ltd., Billerica, MA) [37]. Salmonella-positive isolates were submitted to the OIE (World Organisation for Animal Health) Salmonella  [38].

Data analysis
Antimicrobial susceptibility data were entered into a spreadsheet (Microsoft Excel 2016, Microsoft Corporation, Redmond, WA), reviewed for missing values, and subsequently transferred into a statistical software program (STATA Intercooled, version 14.2, Stata Corporation, College Station, TX) for analysis.
For each poultry species (chicken, turkey, duck, and game bird), estimates of the proportion of E. coli and Salmonella isolates that were resistant to each of the 14 tested antimicrobials were computed by dividing the number of isolates resistant to an antimicrobial by the total number of isolates tested for the antimicrobial. In addition, estimates of the proportion of isolates that showed multidrug resistance (MDR) were computed by dividing the number of multidrug resistant isolates by the total number of isolates tested.
Further, for each poultry species, estimates of the percentage of E. coli-positive samples that were resistant to each of the 14 tested antimicrobials were computed by dividing the number of samples resistant to an antimicrobial by the total number of E. coli-positive samples. A sample was considered to be resistant to a specific antimicrobial if at least one isolate from a pooled cecal sample was resistant. For all estimates, exact binomial 95% confidence intervals (CIs) were calculated.
To compare individual antimicrobials with respect to their similarity in the resistance status of E. coli, a cluster analysis, using the Jaccard binary similarity coefficient, was performed for each poultry species. The number of isolates that are resistant to both antimicrobials and the number that are resistant to one yet susceptible to the other are used in the calculation of the coefficient. Dendrograms were constructed using the single-linkage clustering method with the Jaccard distance. The Jaccard distance measures dissimilarity between antimicrobials and is obtained by subtracting the Jaccard binary similarity coefficient from one [27]. Thus, a high dissimilarity measure indicates that relatively few isolates were resistant to both antimicrobials, whereas a low dissimilarity measure indicates that a relatively high proportion of isolates were resistant to both antimicrobials. A dissimilarity measure of zero indicates that all isolates were susceptible to both antimicrobials.
Further, to measure the strength and direction of the relationships between individual antimicrobials with respect to resistance of E. coli at the isolate-level, Spearman's rank correlation coefficients were calculated; only chicken isolates were evaluated. A Bonferroni correction was used to adjust for multiple comparisons among antimicrobials, with P ≤ 0.0036 (α of 0.05/14) indicating a significant correlation.
To identify differences in E. coli resistance between poultry species, logistic regression was used; only antimicrobials for which ≥5% of the isolates were resistant were evaluated. Therefore, 6 of 14 antimicrobials were analyzed: ampicillin, gentamicin, streptomycin, sulfisoxazole, trimethoprim-sulfamethoxazole, and tetracycline. Four population-averaged models were built for each antimicrobial using the generalized estimating equation method, with a robust variance estimate and exchangeable correlation structure to account for sample-level clustering. In these univariable models, the binary (yes/ no) dependent variable represented the frequency of resistance to the antimicrobial, while the independent variable was poultry species (binary variable: chickens compared to all the other poultry species; turkeys compared to all the other poultry species; ducks compared to all the other poultry species; and game birds compared to all the other poultry species). This method of grouping species together was preferred over analyzing species as a categorical variable because of the limited number of isolates from turkeys, ducks, and game birds. A P-value ≤0.05 on the Wald χ 2 test indicated a statistically significant association.
In addition, four Poisson regression models were built to identify differences in E. coli MDR between poultry species using the generalized estimating equation method described above. The dependent variable was the number of antimicrobial classes to which an isolate was resistant; as seven antimicrobial classes were studied, this count potentially ranged from zero to seven. The independent variable was the poultry species (binary variable; described above).