Diversity of Antimicrobial Resistance Phenotypes in Salmonella Isolated from Commercial Poultry Farms

Salmonella remains the leading cause of foodborne illness in the United States, and the dissemination of drug-resistant Salmonellae through the food chain has important implications for treatment failure of salmonellosis. We investigated the ecology of Salmonella in integrated broiler production in order to understand the flow of antibiotic susceptible and resistant strains within this system. Data were analyzed from a retrospective study focused on antimicrobial resistant Salmonella recovered from commercial broiler chicken farms conducted during the initial years of the US FDA’s foray into retail meat surveillance by the National Antimicrobial Resistance Monitoring System (NARMS). Sixty-three percentage of Salmonella were pan-susceptible to a panel of 19 antimicrobials used by the NARMS program. Twenty-five antimicrobial resistance phenotypes were observed in Salmonella isolated from two broiler chicken farms. However, Salmonella displaying resistance to streptomycin, alone, and in combination with other antibiotics was the most prevalent (36.3%) antimicrobial resistance phenotype observed. Resistance to streptomycin and sulfadimethoxine appeared to be linked to the transposon, Tn21. Combinations of resistance against streptomycin, gentamicin, sulfadimethoxine, trimethoprim, and tetracycline were observed for a variety of Salmonella enterica serovars and genetic types as defined by pulsed-field gel electrophoresis. There were within and between farm differences in the antibiotic susceptibilities of Salmonella and some of these differences were linked to specific serovars. However, farm differences were not linked to antibiotic usage. Analysis of the temporal and spatial distribution of the endemic Salmonella serovars on these farms suggests that preventing vertical transmission of antibiotic-resistant Salmonella would reduce carcass contamination with antibiotic-resistant Salmonella and subsequently human risk exposure.

food chain has important public health implications considering the potential for treatment failure when cases of gastroenteritis require medical intervention, especially in children, the elderly, and the immunocompromised (6). In addition, infections with antimicrobial resistant bacteria including Salmonella have been associated with higher rates of morbidity and mortality (7)(8)(9).
The use of antibiotics in food animal production has been implicated as a contributing factor to the emergence of drug resistance in human foodborne pathogens (6,10). The emergence and rapid worldwide spread of the multiple drug-resistant Salmonella enterica Typhimurium phage-type DT104 clone and ceftriaxone-resistant S. enterica serovars Heidelberg, Newport, and Typhimurium have underscored the threat to both animal agriculture and human health posed by multiple drug-resistant pathogens (11)(12)(13)(14)(15). Antimicrobial resistance genes are widely disseminated in pathogenic, commensal, and environmental bacteria (16,17). Furthermore, it has been shown that once antimicrobial resistance has been introduced into an ecosystem, resistance can spread and persist without continuing selection pressure from antibiotics (18,19). In addition, the reservoir of antimicrobial resistance genes is larger than previously thought (20). It is in this environment that the potential exists for Salmonella to acquire antimicrobial resistance genes from resident poultry microbiota due to selection pressure from therapeutic and non-therapeutic antibiotic usage. It follows then that the longer Salmonellae persists in the environment of an animal production facility, the chance of acquiring resistance genes increases.
We took advantage of the integrated nature of poultry production to observe the antimicrobial resistance phenotypes acquired by salmonellae during broiler chicken production in order to identify potential critical control points for Salmonella contamination and antimicrobial resistance development; ultimately in order to provide information relevant to reducing the level of carcass contamination with antibiotic-resistant Salmonella. Data were analyzed from a retrospective study focused on antimicrobial-resistant Salmonella recovered from commercial broiler chicken farms conducted during the initial years of the US FDA's foray into National Antimicrobial Resistance Monitoring System (NARMS) retail meat surveillance (4). Despite the diversity of antimicrobial resistance profiles, poultry Salmonella recovered from these farms in 2003 were generally susceptible to the tested antimicrobials of animal and human health significance. Vertical transmission appeared to be the most important factor in chicken carcass contamination with antibiotic-resistant Salmonella.
Approximately 20,000 chicks were placed per house on Farm Two. No litter amendments were used on Farm Two (22). At the hatchery, gentamicin (0.2 mg/chick) was injected subcutaneously into day-of-hatch chicks. Chicks were reared on starter feed containing bacitracin (25 g/ton), and salinomycin (50 g/ton) for the first 2 weeks, then grower feed containing bacitracin (25 g/ton) and salinomycin (50 g/ton) for 2 weeks, then finisher feed without growth promotant or coccidiostat for 1-2 weeks. Withdrawal feed without antibiotic or coccidiostat was fed for the last week of grow-out. Feed was withdrawn for 16 h prior to shipment. Escherichia coli airsacculitis was diagnosed in house B during week six of Flock 3 on Farm Two, and oxytetracycline was administered in drinking water at 10.4 mg/kg weight for 1 day and at 5.1 mg/kg weight for 4 days. In this work, we sampled chick box liners, the poultry environment, and chicken carcasses. The latter was provided to us by the participating poultry companies. We did not physically interact with chickens raised on these farms and, therefore, we were exempt from university guidelines and USDA/NIH regulations regarding animal use.

genotypic and Phenotypic characterization of Poultry Salmonella isolates
The 289 Salmonella strains, examined in this study, were isolated, serotyped, phage-typed, and strain-typed as previously described (21). Presence of aadA1 and merA was determined as described by Bass et al. (23).

statistical analysis
The Fisher's exact test with α = 0.05 and Mantel-Haenszel chisquared test were used to test for non-random associations between specific data values. Salmonella Typhimurium PFGE types 1.1, 1.2, and 1.3 were ranked with regard to multiple drug resistance as determined by fitting linear model: log (μi) = βo = β1*PFGE typei, μi = mean number antimicrobial resistances or resistance type, with assumption that data conformed to Poisson distribution.

resUlTs antibiotic susceptibility and Diversity of antimicrobial resistance Phenotypes in Poultry Salmonella
There is ample opportunity for antibiotic-resistant Salmonella to emerge on poultry farms due to the combination of on farm antibiotic usage and the significant reservoir of antimicrobial resistance genes present in poultry litter. We examined the antibiotic susceptibility of Salmonella collected from two commercial broiler farms in northeast Georgia in relation to on-farm antibiotic usage. The majority of Salmonella isolates (62.6%; n = 172) were susceptible to all 19 antimicrobials tested, with the remainder displaying resistance to streptomycin (30.9%), gentamicin (12.6%), sulfadimethoxine (20.9%), tetracycline (13.9%), and trimethoprim/sulfamethoxazole (8.6%) ( Table 1). Salmonella resistance to streptomycin alone was the most prevalent antimicrobial resistance phenotype (30.9%) (Tables 1 and 2).
A diversity of antimicrobial resistance phenotypes (n = 25) was observed among the Salmonella isolated from commercial broiler chicken farms ( Table 2). Twenty percentage of our poultry Salmonella isolates were resistant to three or more antibiotics ( Table 1). The most common antimicrobial resistance phenotypes identified were to streptomycin (36.28%); streptomycin and sulfadimethoxine, alone or in combination with other antibiotics (41.59%); and streptomycin, sulfadimethoxine, and gentamicin, alone or in combination with other antibiotics (28.32%) ( Table 2). There was a statistically significant association between Salmonella isolates displaying resistance to streptomycin and sulfadimethoxine; and streptomycin, sulfadimethoxine, and gentamicin (chisquared test: p < 0.05). While antimicrobial resistance phenotype diversity was high (Reciprocal Simpson's Index: 1.20), evenness in distribution of these phenotypes among Salmonella was low (0.26). The low evenness score may be a reflection of the broad distribution of certain antimicrobial resistance phenotypes compared to others [streptomycin resistance, alone (41 strain types); streptomycin, sulfadimethoxine, and gentamicin resistance (17 strain types); sulfadimethoxine, trimethoprim/sulfamethoxazole, and tetracycline resistance (11 strain types); streptomycin, sulfadimethoxine, gentamicin, and tetracycline resistance (8 strain types); streptomycin, sulfadimethoxine trimethoprim/ sulfamethoxazole, and tetracycline resistance (7 strain types)]. Two of the common antimicrobial resistances identified, streptomycin and sulfadimethoxine resistance, are commonly associated with the transposon, Tn21. The resistance genes merA and aadA1 are resident on this mobile genetic element and the distribution of these loci was 17.86 and 10.56%, respectively, in the recovered poultry isolates. There was a significant association between these resistance genes and resistance to streptomycin or sulfadimethoxine (chi-squared test; p < 0.05).

Farm Variability in antimicrobial susceptibilities of Poultry Salmonella
Differences were observed within and between poultry farms in antibiotic susceptibilities of Salmonella isolates. Antimicrobial susceptibility patterns differed between farms as Salmonella isolates from Farm One were more likely to be resistant to streptomycin, sulfadimethoxine, and trimethoprim/sulfamethoxazole compared to those recovered from Farm Two (chi-squared test: p < 0.05) ( Table 1). There were also differences in antibiotic susceptibilities among certain Salmonella serovars within farms as well as between farms. S. Typhimurium isolated from Farm One were less susceptible to antibiotics, tested in this study, than S. Enteritidis isolated from the same farm. Salmonella Kentucky isolated from Farm One exhibited significantly more antimicrobial resistance than other Salmonella isolated from the same farm as well as S. Kentucky isolated from Farm Two ( Table 1). Following tetracycline treatment on Farm Two, Salmonella isolates were less likely to be resistant to tetracycline, as determined using one-sided, Fisher's exact test at α = 0.05 (p = 0.0046), or to other antibiotics (Cochran-Mantel-Haenszel method, p = 0.0046). The therapeutic treatment of E. coli airsacculitis with tetracycline did not seem to selectively enrich for antimicrobial resistance in Salmonella isolated from subsequent flocks. In addition, there was no statistically significant difference in Salmonella isolates displaying resistance to tetracycline between the two poultry farms (chi-squared test; p = 0.34).
is horizontal or Vertical Transmission responsible for spread of antibioticresistant Salmonella to Poultry Meat?
S. Typhimurium (n = 159) was the most prevalent serovar isolated in this study, and this serovar was frequently isolated from Farm One. Serovar Typhimurium isolates were largely pan-susceptible (66.6%); however, the most prevalent antimicrobial resistance phenotypes were to streptomycin (6.6%), sulfadimethoxine (12.4%), gentamicin (9.4%), and tetracycline (6.4%) ( Table 3). Resistance to the other 14 antimicrobials tested was not observed that often (≤5%). Eleven percentage of S. Typhimurium isolates were resistant to three or more antibiotics. The most prevalent S. Typhimurium resistance phenotypes observed were as follows: streptomycin alone (23.7%) and the multi-drug resistant phenotype to streptomycin, gentamicin, sulfadimethoxine, and tetracycline (5.3%). A diversity of antimicrobial resistance phenotypes (n = 9) was observed for the three related S. Typhimurium strain types identified by PFGE ( Table 3). Combinations of resistance against streptomycin, gentamicin, sulfadimethoxine, and tetracycline, accounted for 85.3% of the resistance phenotypes ( Table 3). There was no significant difference in resistance phenotypes between the three S. Typhimurium genetic types isolated from Farm One with the exception that PFGE type T1.3 was significantly more likely to be ampicillin resistant (α = 0.05).
Of the three S. Typhimurium strain types (T1.1, T1.2, and T1.3) present on Farm One, there were three instances where two of these strain types were present with chicks on the broiler chicken farm (T1.1 and T1.2) and chicken carcasses derived from these flocks ( Table 4). There were also three other situations where these same S. Typhimurium strain types were only isolated from the farm environment and then chicken carcasses at processing. The only antibiotic resistant S. Typhimurium strain types found on chicken carcasses matched with those isolated from chicks at farm placement indicating that resistant S. Typhimurium strains were likely vertically transferred from the breeder flock.   (2015), half of the poultry Salmonella isolates were pan-susceptible to a panel of 12 antibiotics. Salmonella isolated from retail meats, in this survey, were resistant to tetracycline (37.3%), streptomycin (37.3%), sulfadimethoxine (8.5%), gentamicin (5.1%), and ampicillin (8.5%) (27).
There was a diversity of antimicrobial resistance phenotypes identified among our poultry Salmonella isolates. Despite this diversity, the antimicrobial resistance phenotype: streptomycin and sulfadimethoxine resistance alone or with other antibiotics was commonly encountered in Salmonella isolated from the commercial poultry farms. The genes conferring resistance to these antimicrobials are frequently found residing on mobile genetic elements which are responsible for the wide-spread dissemination of antimicrobial resistance in nature. The transposon Tn21 contains the mercury resistance gene merA; streptomycin resistance gene aadA1; and sulfadimethoxine resistance gene sul1 (28). This transposon is often responsible for dissemination of mercury and antimicrobial resistance in nature (28) and is prevalent in poultry Salmonella and E. coli (23). While we observed linkage between the resistance genes merA and aadA and streptomycin/ sulfadimethoxine resistance, only 17.72% of streptomycinresistant Salmonella had aadA1, indicating that other antimicrobial resistance gene(s) are responsible for streptomycin resistance and further illustrates the diversity underlying antimicrobial resistance phenotypes observed in these isolates. Despite the high prevalence of Tn21 in these poultry isolates, antimicrobial resistance phenotypes were not uniformly distributed among Salmonella serovars within as well as between the two commercial broiler chicken farms. Certain Salmonella serovars differed in their antibiotic susceptibility patterns. Salmonella Enteritidis tended to be pan-susceptible while S. Typhimurium exhibited a diversity of antimicrobial resistance phenotypes. Similar trends have been observed for these Salmonella serovars reported in NARMS retail meats (2003,2015) and HACCP (2003HACCP ( , 2014 surveys (27). Even within S. Typhimurium, there were differences in antibiotic susceptibilities among strain types. The S. Typhimurium PFGE subtype T1.1 from Farm One (21) was identified as phage type (PT) 193, a PT commonly associated with illnesses in humans (29)(30)(31)(32)(33)(34)(35)(36)(37)(38)(39). This Salmonella PT has also been isolated from cattle (38,40,41), poultry (31,42), pigs (31,40), and dogs (40). Like S. Typhimurium DT104, PT 193 isolates generally exhibit resistance to three or more antibiotics, but resistance phenotypes reported have been variable (40,43). The majority (68.0%) of our S. Typhimurium PT DT193 isolates from Farm One were pan-susceptible, with 32% possessing the following resistance phenotypes to: streptomycin alone; streptomycin, sulfadimethoxine, tetracycline, and trimethoprim/sulfamethoxazole; streptomycin, sulfadimethoxine, gentamicin, tetracycline, and trimethoprim/sulfamethoxazole. The other S. Typhimurium PFGE types, T1.2 and T1.3, were identified, respectively, as PTs DT107 and U302 (21). The S. Typhimurium PTs DT107 and DT193 from this study appear to be genetically related as determined by PFGE (44). Close genetic-relatedness as determined by PFGE among different S. Typhimurium and S. Enteritidis PTs has been reported by others (45,46). The S. Typhimurium DT107 isolates were similar to the S. Typhimurium DT193 isolates, in that the majority were pansusceptible (59.6%), with the most prevalent antimicrobial resistance phenotype being resistance to streptomycin only (25.4%).
Poultry litter contains a large reservoir of antimicrobial resistance genes. We had shown in a previous study that many of these antimicrobial resistance genes are shared among diverse bacterial species in poultry litter (ex. aadA1 in Corynebacterium and Salmonella) (20). Therefore, the potential exists for environmental transfer of antimicrobial resistance to Salmonella and subsequent horizontal transmission of emergent resistant Salmonella strains to poultry in this environment. Of the eight antibiotic resistant phenotypes solely present in S. Typhimurium isolated from the farm environment, none were identified in S. Typhimurium isolated from processed chicken carcasses. This finding suggests that despite the diversity of antibiotic-resistant S. Typhimurium resident in the broiler house environment, none of these antibiotic resistant strains were being transmitted through the processing plant to the poultry carcass. Only those antibiotic-resistant strain types present with the chicks at placement remained on birds at processing. Therefore, our data support the importance of vertical transmission routes in the dissemination of antibiotic-resistant Salmonella through the food chain.
cOnclUsiOn Therapeutic tetracycline antibiotic usage was not a significant predictor of emergent antimicrobial resistance in Salmonella. This result is not surprising, considering that the all-in, all-out production method used in the commercial poultry industry is designed to break disease cycles and should minimize antimicrobial resistance development, as long as pathogen persistence from flock-to-flock is prevented (47). However, the reservoir for antimicrobial resistance remains within the farm environment. Additional measures involving litter management and pest control may be needed to prevent future emergence of antimicrobial resistance zoonotic bacteria on treated farms. In addition, the prevalence of streptomycin resistance in poultry Salmonella was surprisingly high considering that streptomycin is rarely used in poultry production medicine and to our knowledge had not been used at these farms. This is most likely due to linkage of streptomycin resistance gene(s) with other resistance genes, or competitively advantageous genes (bacteriocins, siderophores, etc.); or its integration into the chromosome that has maintained streptomycin resistance in Salmonella, even in the absence of antibiotic selection (19). However, gentamicin is commonly used with in ovo poultry vaccines as a prophylaxis against peritonitis in chicks and therefore may explain, in part, the level of resistance to this antibiotic in Salmonella. The physical linkage of resistance genes associated with gentamicin with streptomycin resistance may also explain the persistence of streptomycin resistance in the absence of usage (19). As gentamicin was used by both poultry companies, it is uncertain whether gentamicin resistance in Salmonella will persist with time. The other antibiotics used by the poultry farms in this study, bacitracin and virginiamycin, are used to control C. perfringens infections in poultry. While these antibiotics do not affect Salmonella or other Gram-negative enterics, they do have an impact in the Gram positive, intestinal microbiota of chickens (48). It is currently not known how changes to the chicken intestinal microbiota, in response to bacitracin and virginiamycin, affect Salmonella prevalence, abundance, or antibiotic resistance patterns.
Vertical transmission from the breeder flock, rather than horizontal transmission from the environment, appears to play a significant role in carcass contamination with antibiotic-resistant Salmonella. If antibiotic usage is involved in the emergence and spread of antibiotic-resistant Salmonella to chicken meat, it may exist at the breeder, not broiler level of poultry production. One way to block transmission of antimicrobial-resistant Salmonella would be to apply an intervention such as competitive exclusion or vaccination at the breeder level (49,50). The poultry integrator for Poultry Farm One has recently instituted a company-wide Salmonella vaccination program at the broiler-breeder level. It will be interesting to see if this mitigation strategy has significantly changed antimicrobial resistance profiles of Salmonella isolated from broiler chicken farms and poultry charges, especially on Poultry Farm One. aUThOr cOnTriBUTiOns JM, CH, DW, and ML contributed to the conception and design of this study. KL and SA were responsible for the acquisition of data analyzed in this study. JM, KL, and DW were involved in the analysis and interpretation associated with this work. KL was responsible for writing the first draft. All the authors were involved in manuscript revisions and final approval of the version to be published.

acKnOWleDgMenTs
We wish to thank the United States Department of Agriculture and the College of Veterinary Medicine at the University of Georgia for support of this work and the poultry companies that participated in this study. This work, presented here, was part of KL PhD dissertation at the University of Georgia (51).