A Whole-Genome Sequencing Approach To Study Cefoxitin-Resistant Salmonella enterica Serovar Heidelberg Isolates from Various Sources

ABSTRACT This study characterized cefoxitin-resistant and -susceptible Salmonella enterica serovar Heidelberg strains from humans, abattoir poultry, and retail poultry to assess the molecular relationships of isolates from these sources in Québec in 2012. Isolates were collected as part of the Canadian Integrated Program for Antimicrobial Resistance Surveillance (CIPARS). All isolates were subjected to antimicrobial susceptibility testing, PCR for CMY-2, pulsed-field gel electrophoresis (PFGE), and whole-genome sequencing (WGS). A total of 113 S. Heidelberg isolates from humans (n = 51), abattoir poultry (n = 18), and retail poultry (n = 44) were studied. All cefoxitin-resistant isolates (n = 65) were also resistant to amoxicillin-clavulanic acid, ampicillin, ceftiofur, and ceftriaxone, and all contained the CMY-2 gene. PFGE analysis showed that 111/113 (98.2%) isolates clustered together with ≥90% similarity. Core genome analysis using WGS identified 13 small clusters of isolates with 0 to 4 single nucleotide variations (SNVs), consisting of cefoxitin-resistant and -susceptible human, abattoir poultry, and retail poultry isolates. CMY-2 plasmids from cefoxitin-resistant isolates all belonged to incompatibility group I1. Analysis of IncI1 plasmid sequences revealed high identity (95 to 99%) to a previously described plasmid (pCVM29188_101) found in Salmonella Kentucky. When compared to pCVM29188_101, all sequenced cefoxitin-resistant isolates were found to carry 1 of 10 possible variant plasmids. Transmission of S. Heidelberg may be occurring between human, abattoir poultry, and retail poultry sources, and transmission of a common CMY-2 plasmid may be occurring among S. Heidelberg strains with variable genetic backgrounds.

Infection with S. Heidelberg is linked primarily to the consumption of poultry and is rarely transmitted from person to person (3,5). Within the Canadian agriculture sector, this serovar is repeatedly isolated from farm, abattoir, and retail poultry samples but less frequently from bovine and porcine samples (6).
Extended-spectrum cephalosporins are used in veterinary medicine for the treatment and prevention of disease in livestock (7). In Canada, ceftiofur, a third-generation cephalosporin, has been employed in an extralabel manner in broiler chickens to prevent omphalitis caused by Escherichia coli. The Canadian Integrated Program for Antimicrobial Resistance Surveillance (CIPARS) (8) observed a strong correlation (r ϭ 0.91, P Ͻ 0.0001) between the prevalence of ceftiofur-resistant S. Heidelberg isolates found in retail chicken samples and those found in humans (9). After a voluntary withdrawal of the antimicrobial by provincial hatcheries in 2005, a decrease in ceftiofurresistant S. Heidelberg isolates from chicken meat and humans was observed, which was subsequently followed by an increase in resistant isolates in 2007 that paralleled a partial reinstitution of ceftiofur use (9). Classified by Health Canada as antimicrobials of "high importance" (category 2) (cefoxitin) and "very high importance" (category 1) (ceftriaxone) in human medicine, cefoxitin and ceftriaxone are broad-spectrum secondgeneration and third-generation cephalosporins, respectively, that are used to treat a wide variety of infections (9,10).
Resistance of microorganisms to beta-lactam antimicrobials, including cephalosporins, is commonly due to the production of beta-lactamase enzymes that hydrolyze the antimicrobial (11). AmpC beta-lactamases, which belong to Ambler class C enzymes, encode resistance to penicillins, cephalosporins (including cephamycins), and monobactams (12). Although AmpC genes are found intrinsically in some microorganisms, in Salmonella they are due to acquired mechanisms (13). The AmpC beta-lactamase bla CMY-2 , which is commonly found on plasmids, encodes resistance to cefoxitin and is a significant source of beta-lactam resistance in Salmonella found worldwide (14,15). The bla CMY-2 -containing plasmids can also harbor resistance genes to other classes of antimicrobials, thus potentially conferring multidrug resistance (13).
In this study, we examined the molecular characteristics of S. Heidelberg to assess the relationships of isolates from human, abattoir poultry, and retail poultry sources from Québec by using high-quality core genome single-nucleotide variant (hqSNV) analysis and bla CMY-2 plasmid analysis.
Macrorestriction analysis of S. Heidelberg. Macrorestriction analysis using pulsedfield gel electrophoresis (PFGE) of the 113 isolates revealed that all were closely related with Ն80% similarity, with 111 (98.2%) clustering at Ն90% similarity (see Fig. S1 in the supplemental material). Overall there were a total of 16 groups, labeled A1 to A16,    where A1 to A4 contained the majority of isolates (n ϭ 99, 87.6%). The maximum band difference between A1 and the remaining groups was 6, where most groups (9/16, 56.3%) differed by only 1 to 3 bands. WGS analysis of S. Heidelberg isolates. Of the 113 isolates that were analyzed by whole-genome sequencing (WGS), all had an average coverage of 134 times with a range of 56 to 346 times. hqSNV analysis was performed on all isolates, which included S. Heidelberg strain 12-4374 as the reference, as this genome was fully closed in a previous study (16). A maximum of 151 hqSNVs were identified between all isolates, with 95.9% of the reference included in the core genome.
All 113 isolates belonged to multilocus sequence type (MLST) ST15 (Table 1). Predicted antimicrobial resistance genes were identified using WGS data, and the majority of isolates (n ϭ 112, 99.1%) had genes that correlated to the antimicrobial phenotypes determined by broth microdilution (Table 1). For the remaining isolate (12-5643), although initial antimicrobial susceptibility testing revealed an increased MIC for nalidixic acid, Resfinder (17) did not detect a corresponding resistance gene for this phenotype. Subsequent retesting using broth microdilution indicated that this isolate was susceptible to nalidixic acid.
CMY-2 plasmid analysis. Using WGS data, 63/65 (96.9%) of the isolates containing bla CMY-2 plasmids belonged to replicon type IncI1. For the remaining 2 isolates (12-6245 and N13-01320), the bla CMY-2 gene resided on the largest contig (Ͼ747 kb) corresponding to the chromosome, suggesting that these plasmids might have integrated into the chromosome. An alignment of the previously characterized IncI1 bla CMY-2 plasmid, S. Kentucky pCVM29188_101 (GenBank accession number CP001121.1) (18), against the chromosomal contigs identified part of the bla CMY-2 plasmid (approximately 80.5 kb), which had inserted into the chromosome (Fig. 2). The plasmid inserted 386 bp downstream of the chromosomal rlmM gene and into the transcriptional repressor glcR, causing the terminal 91 bp to be deleted. No direct or inverted repeats flanking the insertion sites were found. Nucleotide sequence analysis of both 12-6245 and N13-01320 identified ISEcp1 approximately 200 bp downstream of the plasmid insertion site, followed by bla CMY-2 and a portion of the remaining plasmid. Several genes did not integrate with the plasmid into the chromosome, including those belonging to pil and tra operons, which are associated with transfer pili (Fig. 2). The plasmid replication initiation gene repA also did not insert into the chromosome, which may explain the lack of an IncI1 replicon type found in these isolates. The chromosomal insertion site for both isolates was confirmed using PCR. Using the hqSNV analysis, there were 12 SNVs found between isolates 12-6245 and N13-01320 containing the chromosomally located bla CMY-2 (Fig. 1).
Using plasmid multilocus sequence types (pMLST) based on WGS data, the primary IncI1 plasmid sequence type of the 63 bla CMY-2 plasmids was ST12 (n ϭ 56, 88.9%) ( Table 1). Four (6.3%) plasmids were identified as belonging to ST2, ST25, ST26, and ST66, and three (4.8%) were untypeable, as only a portion (289/343 bp) of the ardA allele was present. The ST2 plasmid was isolated from a human and was resistant to only the beta-lactam antimicrobials. The plasmid with ST26 was isolated from a chicken and was resistant to beta-lactams, chloramphenicol, and sulfisoxazole. The remaining 2 plasmids with ST25 and ST66 were isolated from retail turkey and retail chicken, respectively, and were resistant to beta-lactams. Sequence analysis of the bla CMY-2 -containing plasmids revealed high homology (95 to 99%) to a previously described plasmid (pCVM29188_101) found in S. Kentucky (18). Nucleotide sequence alignments were performed to identify differences in genes compared to pCVM29188_101. All cefoxitin-resistant plasmids in this study belonged to 1 of 10 variant plasmids (Fig. 3). The majority of the 63 plasmids belonged to 1 of 3 variants, labeled as group A (n ϭ 33, 52.4%), group B (n ϭ 3, 4.8%), and group C (n ϭ 20, 31.7%) ( Table 1). The 7 remaining plasmids were isolated from individual strains. In comparison to S. Kentucky pCVM29188_101, all plasmids were missing the IS66 transposase. Some plasmids were missing additional genes encoding proteins that included transposases, recombinases, translational repressor protein RelE, DNA polymerase III subunit epsilon, colicin 1B immunity protein, quaternary ammonium resistance protein SugE, chromosome partitioning protein ParA, part of a membrane protein, and part of shufflon protein A.
Virulence data analysis. In this study, 169 (8.4%) virulence genes were identified among the 113 strains. These genes were found to be involved in a variety of processes, including adhesion, type III secretion system (T3SS), regulation of genes, resistance to antimicrobial peptides, magnesium uptake, and regulation of stress factors. All sequenced isolates carried genes for curli, fimbriae, Salmonella pathogenicity island 1 and 2, PhoPQ, SodCl, and Mig-14. The majority of isolates (n ϭ 97, 85.8%) did not contain stfA but contained the other genes in this operon (stfCDEFG), which is involved in fimbrial adherence of the organism. Most isolates (n ϭ 110, 97.3%) also did not contain SeAg-B4893, which encodes a putative outer membrane protein that is involved in fimbrial adherence. Some isolates (n ϭ 12, 10.6%) did not carry a greater percentage of virulence genes than the others. The genes missing in these isolates were all different, and they were not associated with any one characteristic, including source, antimicrobial resistance pattern, or type of human infection.

DISCUSSION
The use of antimicrobials in agriculture is of concern to public health, as overuse or misuse of these drugs can lead to resistance that may transfer to humans (19). A

FIG 3
Nucleotide sequence alignments of the previously characterized S. Kentucky pCVM29188_101 plasmid against the 10 variant bla CMY-2containing plasmids found in all S. Heidelberg isolates in this study, generated using GView Server. The reference plasmid is represented by the black boxes that denote open reading frames. For the remaining plasmids, solid colored boxes indicate that DNA is present. The gray line toward the bottom represents the GC content of the reference. Proteins listed above the alignment indicate those that are not present in certain plasmids, except for CMY-2, which was found in all plasmids. A representative sequence of a group A plasmid has been previously published (16).
previous study has suggested that the use of ceftiofur in poultry in Québec has contributed to the elevated levels of cephalosporin resistance in S. Heidelberg from animals and humans (9). This current study utilized hqSNV analysis of S. Heidelberg from humans, abattoir poultry, and retail poultry to further understand the genetic relatedness between isolates from these sources in Québec in 2012.
As stated in previous studies (20), our findings support the fact that macrorestriction analysis using PFGE lacked the discriminatory power that was needed to identify potential relationships of this clonal serovar retrieved from human, abattoir poultry, and retail poultry sources, in contrast to hqSNV analysis using WGS.
Currently, there exists no defined range for the number of SNVs observed between 2 S. Heidelberg isolates that are genetically related. A study by Bekal et al. examined 3 epidemiologically defined outbreaks of S. Heidelberg in Québec and found a maximum number of 4 SNVs between isolates belonging to the same outbreak (20). Another study by Leekitcharoenphon et al. used hqSNV to characterize an outbreak of S. Heidelberg from the United States in 2011 and identified a maximum of 19 SNVs between their outbreak isolates (21). Using a maximum of 4 SNVs, our guideline to consider potential genetic linkages between isolates, we observed 13 clusters of 2 to 18 isolates each containing 0 to 4 SNVs. Four clusters contained up to 4 isolates each of human and retail chicken origin with 0 SNVs in the core genome. Identification of these clusters with 0 to 4 SNVs suggests that isolates within a cluster potentially originated from a common source.
Two groups of isolates with 0 SNVs (found in clusters 3 and 7) had isolates that were susceptible to all antimicrobials tested, and those that contained an IncI1 bla CMY-2 plasmid conferring resistance to beta-lactams. As extrachromosomal DNA was not included in the core genome used for the analysis, the occurrence of both antimicrobial-resistant and -susceptible isolates with identical core genomes may be indicative of plasmids being gained or lost, as was previously shown with the loss of an IncI1 bla CMY-2 plasmid found in Salmonella enterica serovar Bredeney within 49 days in an antimicrobial-free environment (22).
The majority of the turkey isolates (6/7, 85.7%) did not cluster with human or chicken isolates. The turkey isolates had different resistance patterns, including being resistant to beta-lactams, streptomycin, sulfisoxazole, and tetracycline (n ϭ 2, 28.6%), resistant to only beta-lactams and either tetracycline or streptomycin (n ϭ 2, 28.6%), or resistant to only ampicillin or tetracycline (n ϭ 3, 42.9%). The fact that no human S. Heidelberg isolates clustered with turkey isolates suggested either that the isolates are less virulent in humans or that, possibly more likely, turkey is consumed less frequently than chicken, so humans are exposed to S. Heidelberg from turkey less frequently. In Canada, 85.6% of people reported consuming chicken in the previous 7 days compared to 11.8% who consumed turkey; and in Québec specifically, 86.9% of people reported eating chicken in the previous 7 days compared to 7.5% who consumed turkey (23).
All isolates were screened for virulence genes, which were defined as those that aid an organism to colonize a host, replicate, and cause inflammation, tissue damage, and consequently disease (24). There was no correlation seen between type of infection and virulence gene content.
In this study, two bla CMY-2 plasmids were found to have inserted into the chromosome. It has been suggested that ISEcp1 upstream of bla CMY-2 is involved in mediating such plasmid insertions (25). We identified 12 SNVs between the 2 isolates containing the plasmid integrations. It is unknown whether these integrations occurred separately at 2 different time points or whether they occurred once and the 2 isolates consequently evolved over time.
Comparison of our bla CMY-2 plasmids against previously characterized plasmids revealed homology to the S. Kentucky pCVM29188_101 plasmid and identified 10 plasmid subtypes. Three of the plasmids (groups A, B, and C) were found in multiple isolates of human, abattoir poultry, and retail poultry origins, and all plasmids in these groups belonged to ST12. The finding that the majority of bla CMY-2 -containing plasmids were related suggests the dissemination of a similar resistant plasmid in variable genetic backgrounds of S. Heidelberg isolates.
In conclusion, we have compared cefoxitin-resistant and -susceptible S. Heidelberg isolates from humans, abattoir poultry, and retail poultry from a specific year and for a specific region in Canada using hqSNV analysis. Our findings suggest that although there is evidence to imply highly related isolates (0 to 4 SNVs) from human and poultry, suggesting a potential common source, it would seem that the majority of cefoxitinresistant isolates studied occurred due to the dissemination of a plasmid between different S. Heidelberg genetic backgrounds. This suggests that the transmission of bla CMY-2 is due to the horizontal transfer of an IncI1 plasmid rather than clonal dissemination of a particular S. Heidelberg strain. Further WGS studies are under way to examine the temporal and spatial distribution of cefoxitin-resistant S. Heidelberg in Canada.

MATERIALS AND METHODS
Bacterial isolates. Salmonella Heidelberg isolates were collected as part of CIPARS (8). A convenience sample of 113 isolates from humans (n ϭ 51), abattoir poultry (n ϭ 18), and retail poultry (n ϭ 44) were studied, with the majority (n ϭ 103) from Québec from 2012. Since no cefoxitin-resistant abattoir isolates from Québec in 2012 were available, 10 resistant isolates from chicken from Ontario and New Brunswick in 2011 and 2012 were selected for inclusion.
Whole-genome sequencing. DNA was extracted using the EpiCentre MasterPure Complete DNA and RNA purification kit (Illumina, Madison, WI, USA). Libraries were prepared using the Nextera XT DNA Sample Prep kit (Illumina, Madison, WI, USA). Sequencing was performed using paired-end reads to obtain an average coverage of Ͼ60 times for all isolates.