Ceftiofur Resistance in Salmonella enterica Serovar Heidelberg from Chicken Meat and Humans, Canada

Use of this drug in chickens may limit effectiveness of cephalosporins in treating human infections.

losporins, such as ceftriaxone and cephamycins (16)(17)(18)(19)(20), a group of antimicrobial agents used to treat a wide variety of human infections. Among other indications, ceftriaxone is the drug of choice for treating severe or invasive salmonellosis in children and pregnant women (16,17) where fl uoroquinolones are not approved and treatment options are limited. Accordingly, third-generation cephalosporins have been classifi ed as Critically Important Antimicrobials in Human Medicine by the World Health Organization (21) and as Class 1 Very High Importance in Human Medicine by the Canadian Veterinary Drugs Directorate, Health Canada (22).
In Canada, ceftiofur resistance in bacteria from healthy animals or food is mainly reported in chicken Salmonella Heidelberg isolates originating from farm, abattoir, and retail samples and in chicken abattoir and retail generic E. coli isolates (11,12). It also is occasionally reported in Salmonella isolates from sick animals or in bovine and porcine abattoir or retail E. coli isolates but at much lower frequency (12).
The objective of this study is to highlight the correlation between ceftiofur-resistant Salmonella Heidelberg isolated from retail chicken and the incidence of ceftiofur-resistant Salmonella Heidelberg infections in humans across Canada. Public health concerns raised by publication of the CIPARS 2003 annual report, specifi cally the higher rates of ceftiofur resistance in Salmonella Heidelberg isolates from chicken meat than from humans, prompted Québec broiler chicken hatcheries to voluntarily interrupt the extralabel in ovo use of ceftiofur during 2005-2006 (23). This study therefore also describes variations in ceftiofur resistance among chicken and human Salmonella Heidelberg and chicken E. coli strains in that province before, during, and after the voluntary withdrawal.

Materials and Methods
CIPARS is a national program led by the Public Health Agency of Canada (PHAC) dedicated to the preservation of effective antimicrobial drugs for humans and animals through the collection, integration, analysis, and communication of trends in antimicrobial resistance in selected bacterial organisms. Data presented here were collected during 2003-2008 from CIPARS' surveillance of human clinical Salmonella isolates and E. coli and Salmonella isolates from retail chicken. Detailed methods for sample collection, bacterial isolation, antimicrobial resistance testing, and data analysis are described in CIPARS's reports (12).

Sample Collection Human Salmonella Isolates
Hospital-based and private clinical laboratories isolated and forwarded human Salmonella isolates to their Provin-cial Public Health Laboratory (PPHL). PPHLs forwarded Salmonella isolates to the Enteric Diseases Program, National Microbiology Laboratory (NML), PHAC, for phage type characterization and antimicrobial resistance testing. All isolates (outbreak and nonoutbreak) received passively by the Saskatchewan PPHL were forwarded; the more populated provinces (British Columbia, Ontario, and Québec) forwarded isolates received from days 1-15 of each month. Only 1 isolate per patient was kept for the analysis.

Retail Meat Samples
To use a similar geographic scale as CIPARS surveillance of human clinical Salmonella isolates and because we expected a certain level of provincial clustering in food distribution, we designed the study of CIPARS retail surveillance to provide a representative measurement of what consumers from each province were exposed to through ingestion of improperly cooked raw meat or cross-contamination. Randomization and weighted allocation of samples according to demography of the human population ensured that the data generated through retail sampling were representative and reliable at the provincial level. Retail raw chicken samples (most often chicken thigh with skin on) were collected as part of CIPARS retail program that purchases samples weekly (Ontario and Québec) or biweekly (Saskatchewan, British Columbia) from chain, independent, and butcher stores in 15-18 randomly selected census divisions in each participating province. Retail surveillance was initiated in Ontario and Québec in mid-2003 and at the beginning of 2005 in Saskatchewan. Surveillance also was conducted during part of 2007 and all of 2008 in British Columbia.

Recovery of Isolates from Retail Chicken Meat
Primary isolations of E. coli and Salmonella spp. were conducted at the Laboratory for Foodborne Zoonoses, PHAC. Every retail chicken meat sample received was cultivated for Salmonella, but only 1 of every 2 samples was systematically selected to be tested for generic E. coli isolation. Incubated peptone rinses of chicken meat samples were streaked on eosin-methylene blue agar (Becton Dickinson, Sparks, MD, USA). Presumptive E. coli colonies were identifi ed by using the Simmons' citrate and indole tests. Colonies showing negative indole results were identifi ed by using the API 20E (bioMérieux Clinical Diagnostics, Marcy l'Étoile, France). All chicken samples were tested for Salmonella with a modifi ed MFLP-75 method of the Compendium of Analytical Methods (24). Incubated peptone rinses were injected into modifi ed semisolid Rappaport-Vassiliadis media. Presumptive E. coli colonies were injected into triple sugar iron and urea agar slants and subjected to the indole test. Presumptive Salmonella isolates were verifi ed by slide agglutination using PolyA-I and Vi Salmonella antiserum (Difco, Becton Dickinson). Salmonella isolates were shipped between laboratories on a tryptic soy agar slant by priority courier. No selective media were used to isolate ceftiofur-resistant bacteria.

Serotyping, Phage Typing, and Susceptibility Testing
Human and chicken Salmonella isolates were serotyped and phage typed by using published methods (25)(26)(27)(28). MICs were determined by the NML (human isolates) and the Laboratory for Foodborne Zoonoses, PHAC (chicken isolates) by the broth microdilution method (Sensititre Automated Microbiology System; Trek Diagnostic Systems Ltd., Westlake, OH, USA). Salmonella and E. coli isolates were tested by using the National Antimicrobial Resistance Monitoring System custom susceptibility plate for gramnegative bacteria. The breakpoint used to determine ceftiofur resistance was >4 μg/mL (29).

Data Analysis
We analyzed data using SAS version 9.1 (SAS Institute Inc., Cary, NC, USA). The yearly proportion of retail chicken samples contaminated with ceftiofur-resistant Salmonella Heidelberg (or E. coli) and the incidence rate of human infection with ceftiofur-resistant Salmonella Heidelberg was calculated as described in CIPARS 2006 annual report (12). The Pearson product-moment correlation was used to verify the correlation between ceftiofur-resistant Salmonella Heidelberg isolated from retail chicken and human incidence estimates by using the Pearson option in the PROC CORR procedure in SAS. We computed the overall correlation coeffi cient using data across all provinces under study and computed a specifi c coeffi cient for provinces with >5 observations (30) To describe ceftiofur resistance changes by quarter and reduce the noise around the estimate caused by the small number of observations per quarter, we computed a nonweighted rolling average of the prevalence of ceftiofur resistance using data from the current quarter and the previous 2 quarters for chicken E. coli, chicken Salmonella Heidelberg, and human Salmonella Heidelberg isolates from the province of Québec. We tested differences in ceftiofur resistance between years with SAS using χ 2 or Fisher exact tests when appropriate.

Ceftiofur-Resistant Salmonella Heidelberg Isolated from Retail Chickens and from Humans
Across Canada, the annual percentage of chicken samples contaminated with ceftiofur-resistant Salmonella Heidelberg correlated strongly with the annual incidence of human cases related to this type of isolate (r = 0.91, p<0.0001) (Figure 1). This strongly signifi cant correlation held across time and within different Canadian provinces (Ontario, r = 0.93, p<0.01; Québec, r = 0.89, p = 0.01).
Changes in ceftiofur resistance alone did not explain a number of the temporal changes in exposure (12). For example, in Ontario, the decrease in the prevalence of retail chicken contaminated with ceftiofur-resistant Salmonella Heidelberg isolates during 2004-2008 ( Figure 1) was linked to a decrease in ceftiofur resistance from 58% to 14% (Table) and a decrease in the prevalence of Salmonella Heidelberg in chicken from 61% to 15% of all Salmonella isolates. In Québec, the decrease in contamination of chicken with ceftiofur-resistant Salmonella Heidelberg strains from 2003 to 2004 ( Figure 1) was related mainly to a decrease in the prevalence of Salmonella Heidelberg (from 71% to 48%) in chicken, whereas the decrease from 2004 to 2006 was attributable mainly to a drop in ceftiofur resistance from 62% to 7% (Table). In British Columbia, the low level of chicken contamination with ceftiofur-resistant Salmonella Heidelberg strains resulted mainly from the rarity of Salmonella Heidelberg (only 11% of all Salmonella in 2007-2008), and low exposure levels in Saskatchewan were related mainly to low ceftiofur resistance among Salmonella Heidelberg (Table).

Ceftiofur-Resistant E. coli Isolated from Retail Chicken
Retail chicken generally was more frequently contaminated with ceftiofur-resistant commensal E. coli than with ceftiofur-resistant Salmonella Heidelberg isolates ( Figure  1). The proportion of chicken contaminated with ceftiofurresistant E. coli (Figure 1) closely followed changes in ceftiofur resistance (Table) because commensal E. coli was recovered from almost all (89%-100%) chicken samples collected. Exposure to ceftiofur-resistant E. coli strains appeared to have increased in recent years in Canada (Figure 1). In 2008, exposure to ceftiofur-resistant E. coli strains was highest in British Columbia and lowest in Québec.

Temporal Changes in Ceftiofur Resistance in the Province of Québec, 2003-2008
In 2003-2004, >60% of the chicken Salmonella Heidelberg isolates were ceftiofur resistant, and ceftiofur resistance among chicken E. coli and human Salmonella Heidelberg isolates varied from 30% to 40% (Figure 2). Ceftiofur resistance declined sharply immediately after the fi rst quarter of 2005 among chicken E. coli and Salmonella Heidelberg isolates, and a similar decline began in the next quarter among human Salmonella Heidelberg isolates ( Figure 2). This decline steadily continued until the end of 2006. As a result, the prevalence of ceftiofur resistance signifi cantly decreased from 2004 to 2006 among chicken (62% to 7%; p<0.001) and human (36% to 8%; p<0.0001) Salmonella Heidelberg isolates and chicken E. coli isolates (34% to 6%; p<0.0001 [Table]). Then, from 2006 to 2008, the prevalence of ceftiofur resistance signifi cantly increased among chicken E. coli isolates (6% to 18%; p = 0.002), and prevalence of ceftiofur resistance increased, but not signifi cantly, among Salmonella Heidelberg from chicken (7% to 18%; p = 0.32) and human (8% to 12%; p = 0.41) isolates (Table).

Discussion
CIPARS data clearly indicate a temporal association between changing levels of contamination of retail chicken with ceftiofur-resistant Salmonella Heidelberg strains and incidence of ceftiofur-resistant Salmonella Heidelberg infection in humans. This correlation is strong and applies to different regions of Canada. Our observation is consistent with published results from outbreak investigations and case-control studies suggesting that chicken products are a source of human infection with Salmonella Heidelberg in Canada (7,8).
Although humans potentially can become infected with ceftiofur-resistant Salmonella Heidelberg from sources other than chicken, chicken appears the most likely source in Canada. Ceftiofur-resistant Salmonella Heidelberg has never been reported among CIPARS porcine Salmonella of abattoir origin, and it has not been detected among retail pork, abattoir beef, or retail beef, in which Salmonella prevalence remains <2% (12). Data generated by National Antimicrobial Resistance Monitoring System retail surveillance in the United States indicated that 17% of Salmonella Heidelberg Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 16  isolates recovered from ground turkey in 2006 were resistant to ceftiofur (13). CIPARS does not conduct ongoing surveillance of retail turkey, and we cannot ignore the possibility that retail turkey could be a source of ceftiofur-resistant Salmonella Heidelberg for humans as well. However, turkey consumption in Canada (4.7 kg per capita) was much lower than chicken consumption (33.2 kg per capita) in 2007 (31). Lastly, Salmonella Heidelberg has been reported in clinical samples from various other animal species in Canada (12), and exposure to sick animals could potentially be another source of infection. However, ceftiofur resistance in clinical Salmonella Heidelberg isolates remains anecdotal in species other than chicken and turkey (12). In 2007, the Québec broiler industry announced a potential partial reinstitution of ceftiofur use to control omphalitis in young chicks, with the intention of using the drug on a rotational basis and limiting its use to no more than 6 months per year (32). Again, CIPARS data from Québec retail chicken sampling in 2007-2008 demonstrating a reemergence of ceftiofur resistance among E. coli but at lower levels than in 2003-2004 are consistent with a partial return to ceftiofur use. The simultaneous reduction (and reemergence) in ceftiofur resistance in both retail chicken E. coli and Salmonella Heidelberg isolates and in clinical chicken E. coli and Salmonella isolates from MAPAQ surveillance support the hypothesis that the fl uctuations in ceftiofur resistance most likely were driven by a common exposure (or reduction of exposure) to ceftiofur in chicken hatcheries, rather than simply being secondary to the natural spread and disappearance of a ceftiofur-resistant clone unrelated to ceftiofur use.
Although Ontario hatcheries had never announced an offi cial withdrawal of ceftiofur use, a drop in ceftiofur resistance also was observed among chicken Salmonella Heidelberg isolates in Ontario in 2005. Although some argue that this proves the absence of an association between ceftiofur use and ceftiofur resistance in broiler chicken, movement of hatching eggs, broiler chicks (mostly from Québec to Ontario), and retail chicken meat between these 2 provinces could explain some of the similarities among Salmonella Heidelberg isolates in Ontario and Québec (33). The withdrawal in Québec might also have led Ontario broiler chicken hatcheries to temporarily decrease their use of ceftiofur in 2005.
In the absence of reliable comprehensive drug use information in the broiler chicken industry, we use resistance in commensal E. coli as a surrogate measure of the level of drug use (34). The high prevalence of ceftiofur resistance among E. coli isolates from British Columbia (almost half of the isolates in 2008 in that province), the increasing prevalence of resistance measured in Saskatchewan, and the 22% ceftiofur resistance among chicken E. coli isolates from Ontario when ceftiofur resistance prevalence was at its lowest level in Québec in 2006, indicates that ceftiofur use is unlikely to be restricted to the province of Québec. Lastly, in ovo ceftiofur use has also been reported in US chicken hatcheries (35).
Coselection of resistance to cephalosporins by exposure to other antimicrobials or to chemicals in the agricultural environment has been suggested as a confounding factor for the increase in observed resistance. Giles et al. (36) report the presence of the sugE gene on the same element as the bla CMY-2 gene in Salmonella, but the capacity of this gene to effectively confer resistance to quaternary ammonium compounds and provide coselection remains uncertain.
The levels of contamination of retail chicken with ceftiofur-resistant E. coli represent an additional concern. No selective media for ceftiofur-resistant strains was used, and the level of contamination of retail chicken with ceftiofur-resistant E. coli (and Salmonella Heidelberg) strains was most likely underestimated. Although this study describes exposure to commensal E. coli, such strains occasionally may cause infections in predisposed humans. In addition, the species E. coli includes a variety of strains commonly pathogenic for humans, and some strains from the normal fl ora of animals may carry a variety of virulence determinants that increase their potential for causing disease in humans (37). Poppe et al. (38) also demonstrated experimentally the acquisition of resistance to extended-spectrum cephalosporins by Salmonella serovar Newport from E. coli strains by conjugation in poultry intestinal tracts. In addition, molecular characterization of plasmids from fi eld isolates demonstrates that identical bla CMY-2 plasmids can be found in both Salmonella and E. coli from the same chicken (P. Boerlin et al., unpub. data). Because the bla CMY-2 gene is horizontally transferable and is frequently observed in ceftiofur-resistant isolates of chicken origin, chicken could potentially be a reservoir of this gene for human pathogens, including Salmonella and others.
Except for anecdotal information, little information is available about drugs used by broiler chicken hatcheries and growers in Canada. The absence of on-farm drug use monitoring data prevents us from fully determining the effect of subtle changes in the level of use of ceftiofur (or other drugs) on resistance among bacteria recovered from chickens in Canada. Surveillance data from turkey or other nonsurveyed commodities would be useful to adequately quantify the contribution of each commodity to the overall number of cases related to ceftiofur-resistant Salmonella Heidelberg in humans. The impact of disinfectants used by the broiler industry at the farm or processing level on the selection of ceftiofur-resistant strains also needs to be assessed. Lastly, CIPARS is planning a burden-ofillness study to measure the impact of extended-spectrum cephalosporin resistance in Salmonella Heidelberg on human health.
Efforts undertaken by Québec chicken hatcheries to voluntarily withdraw use of ceftiofur in 2005-2006 coincided with a markedly reduced prevalence of ceftiofur-resistant Salmonella Heidelberg in retail chicken. This drop also effectively reduced the number of ceftiofur-resistant Salmonella Heidelberg infections in humans in this province during the same period. This reduction suggests that control of resistance to extended-spectrum cephalosporins is possible by managing ceftiofur use at the hatchery level. The partial reintroduction of ceftiofur use in Québec chicken hatcheries in 2007 with increasing rates of ceftiofur resistance after reintroduction, and indications that ceftiofur is used for the same purpose in other Canadian provinces, is of high concern. An increasing level of exposure to E. coli strains carrying horizontally transferable genes conferring resistance to extended-cephalosporins complicates the situation. To ensure the continued effectiveness of extended-spectrum cephalosporins to treat serious human infections, multidisciplinary efforts are needed to scrutinize, and where appropriate, limit use of ceftiofur in Canadian food animal production, particularly in chicken.