APPLIED AND ENVIRONMENTAL MICROBIOLOGY, Mar. 2009, p. 1373–1380 0099-2240/09/$08.00⫹0 doi:10.1128/AEM.01253-08 Copyright © 2009, American Society for Microbiology. All Rights Reserved.
Vol. 75, No. 5
Associations between Antimicrobial Resistance Phenotypes, Antimicrobial Resistance Genes, and Virulence Genes of Fecal Escherichia coli Isolates from Healthy Grow-Finish Pigs䌤 Leigh B. Rosengren,1* Cheryl L. Waldner,1 and Richard J. Reid-Smith2,3 Department of Large Animal Clinical Sciences, Western College of Veterinary Medicine, University of Saskatchewan, 52 Campus Drive, Saskatoon, Saskatchewan S7N 5B4, Canada1; Antimicrobial Resistance Surveillance Unit, Laboratory for Foodborne Zoonoses, Public Health Agency of Canada, 160 Research Lane, Unit 103, Guelph, Ontario N1G 5B2, Canada2; and Department of Population Medicine, Ontario Veterinary College, University of Guelph, Guelph, Ontario N1G 2W1, Canada3 Received 5 June 2008/Accepted 2 January 2009
Escherichia coli often carries linked antimicrobial resistance genes on transmissible genetic elements. Through coselection, antimicrobial use may select for unrelated but linked resistance or virulence genes. This study used unconditional statistical associations to investigate the relationships between antimicrobial resistance phenotypes and antimicrobial resistance genes in 151 E. coli isolates from healthy pigs. Phenotypic resistance to each drug was significantly associated with phenotypic resistance to at least one other drug, and every association found that the probability of observing the outcome resistance was increased by the presence of the predictor resistance. With one exception, each statistical association that was identified between a pair of resistance genes had a corresponding significant association identified between the phenotypes mediated by those genes. This suggests that associations between resistance phenotypes might predict coselection. If this hypothesis is confirmed, evaluation of the associations between resistance phenotypes could improve our knowledge of coselection dynamics and provide a cost-effective way to evaluate existing data until large-scale genotypic data collection becomes feasible. This could enable policy makers and users of antimicrobials to consider coselection in antimicrobial use decisions. This study also considered the unconditional relationships between resistance and virulence genes in E. coli from healthy pigs (aidA-1, eae, elt, estA, estB, fedA1, stx1, and stx2). Positive statistical associations would suggest that antimicrobial use may select for virulence in bacteria that may contaminate food or cause diarrhea in pigs. Fortunately, the odds of detecting a virulence gene were rarely increased by the presence of an antimicrobial resistance gene. This suggests that on-farm antimicrobial use did not select for the examined virulence factors in E. coli carried by this population of healthy pigs. the physical relationships between genes that dictate coselection. Secondly, associations among resistance phenotypes were analyzed and compared to the associations among resistance genes. Identifying similar relationships in these analyses would suggest that phenotypic resistance data could also generate hypotheses about coselection. Finally, associations between resistance and virulence genes were investigated. Other investigators have reported more frequent resistance, physical linkages, and statistical associations between resistance and virulence genes in pathogenic E. coli (10, 41). This has created concern that antimicrobial use may contribute to the persistence and spread of virulence in E. coli (24, 27, 46). Describing similar associations in commensal E. coli would indicate a potential for antimicrobial use to increase virulence in E. coli carried by healthy pigs.
Antimicrobial resistance (AMR) is a serious concern in human and veterinary medicine. Resistant bacterial infections are associated with increased morbidity, mortality, and treatment expense compared to their susceptible counterparts (3, 4, 40). Resistant commensal E. coli, while rarely causing direct disease, is a reservoir of AMR genes. These genes can be transferred to zoonotic pathogens, such as Salmonella, or to other gram-negative bacteria in the gut (42, 43). Pathogenic and commensal strains of E. coli have different rates of resistance and can carry different genes (10, 41). Therefore, describing the AMR genes in commensal E. coli isolates documents the diversity of genes available for dissemination to other bacteria. In a previous paper, we described the frequency and patterns of phenotypic AMR in E. coli from healthy grow-finish pigs in western Canada. Of 1,439 isolates, 21% were susceptible to all of the 16 drugs considered, while 57% were resistant to two or more antimicrobials (33). This paper describes the presence of resistance and virulence genes in a subset of those isolates. The data were explored through three sets of analyses. First, unconditional associations between resistance genes were assessed; these associations generate hypotheses about
MATERIALS AND METHODS Bacterial isolates. A total of 151 E. coli isolates were selected from 1,439 isolates that had been collected from healthy grow-finish pigs in 20 herds in western Canada and previously characterized for phenotypic antimicrobial susceptibility to 16 drugs (33). Isolate selection was purposive to ensure that each of the eight following outcomes occurred in a minimum of 15 isolates: resistant to ampicillin, chloramphenicol, kanamycin, streptomycin, sulfamethoxazole, tetracycline and trimethoprim-sulfamethoxazole and susceptible to all 16 of the tested antimicrobials. These outcomes were selected because more than 5% of the 1,439 available E. coli isolates were resistant to these antimicrobials (33). Only these seven resistance phenotypes and genes encoding resistance to these seven
* Corresponding author. Mailing address: Box 451, Midale, Saskatchewan S0C 1S0, Canada. Phone: (306) 458-2967. Fax: (306) 4582267. E-mail:
[email protected]. 䌤 Published ahead of print on 9 January 2009. 1373
1374
ROSENGREN ET AL.
APPL. ENVIRON. MICROBIOL.
TABLE 1. AMR gene primers, amplicon sizes, and sources for 28 genes investigated Antimicrobial family and gene
Beta-lactam genes blaTEM blaSHV Aminoglycoside genes ant(2⬙)-Ia (aadB) aac(3)-IIa (aacC2) aac(3)-IV aph(3⬘)-Ia (aphA1) aph(3⬘)-IIa (aphA2) Tetracycline genes tet(A) tet(C) tet(C) Phenicol genes catI floR cmlA Trimethoprim genes dhfrI dhfrV dhfrVII dhfrIX dhfrXIII Sulfonamide genes sul1 sul2 sul3 Streptomycin genes strA strB a
Primer a
Direction
Sequence (5⬘ 3 3⬘)
F R F R
GAGTATTCAACATTTTCGT ACCAATGCTTAATCAGTGA TCGCCTGTGTATTATCTCCC CGCAGATAAATCACCACAATG
F R F R F R F R F R
Amplicon size (bp)
GenBank accession no.
Source of DNA
857
AF309824
R. C. Levesque
768
AF148850
R. C. Levesque
TCCAGAACCTTGACCGAAC GCAAGACCTCAACCTTTTCC CGGAAGGCAATAACGGAG TCGAACAGGTAGCACTGAG GTGTGCTGCTGGTCCACAGC AGTTGACCCAGGGCTGTCGC ATGGGCTCGCGATAATGTC CTCACCGAGGCAGTTCCAT GAACAAGATGGATTGCACGC GCTCTTCAGCAATATCACGG
198
AJ238349
Pasteur Institute
591
X75562
R. C. Levesque
705
J05162
R. C. Levesque
701
X92506
A. Huletsky
700
X04555
R. C. Levesque
F R F R F R
GTGAAACCCAACATACCCC GAAGGCAAGCAGGATGTAG CCTTATCATGCCAGTCTTGC ACTGCCGTTTTTTCGCC ACTTGGAGCCACTATCGAC CTACAATCCATGCCAACCC
740
X54723
D. Sandvang
627
X01385
J. Harel
600
M18329
J. Harel
F R F R F R
AGTTGCTCAATGTACCTATAACC TTGTAATTCATTAAGCATTCTGCC CGCCGTCATTCCTCACCTTC GATCACGGGCCACGCTGTGTC TTGCAACAGTACGTGACAT ACACAACGTGTACAACCAG
680
V00618
J. Harel
888
X00006
J. Harel
293
AB212941.1
P. Boerlin
F R F R F R F R F R
AAGAATGGAGTTATCGGGAATG GGGTAAAAACTGGCCTAAAATTG CTGCAAAAGCGAAAAACGG AGCAATAGTTAATGTTTGAGCTAAAG GGTAATGGCCCTGATATCCC TGTAGATTTGACCGCCACC TCTAAACATGATTGTCGCTGTC TTGTTTTCAGTAATGGTCGGG CAGGTGAGCAGAAGATTTTT CCTCAAAGGTTTGATGTACC
774
J01830
J. Harel
881
J01749
J. Harel
827
X65876
S. B. Levy
853
L06940
M. C. Roberts
823
AF070999
M. C. Roberts
F R F R F R
TTCGGCATTCTGAATCTCAC ATGATCTAACCCTCGGTCTC CGGCATCGTCAACATAACC GTGTGCGGATGAAGTCAG GAGCAAGATTTTTGGAATCG CATCTGCAGCTAACCTAGGGCTTTGGA
547
M62822
J. Harel
543
X53796
Pasteur Institute
880
AY494779.1
P. Boerlin
F R F R
CCTGGTGATAACGGCAATTC CCAATCGCAGATAGAAGGC ATCGTCAAGGGATTGAAACC GGATCGTAGAACATATTGGC
546
EF090911.1
P. Boerlin
509
EF090911.1
P. Boerlin
F, forward; R, reverse.
drugs were considered in this study. Clustering was minimized by selecting a maximum of one isolate per pen while maintaining the original herd distribution. AMR gene detection. DNA hybridization was used to test for 28 AMR genes (De´partement de Pathologie et Microbiologie, Faculte´ de Me´decine Ve´te´rinaire, Universite´ de Montre´al, Saint-Hyacinthe, Que´bec, Canada). The strains used as
positive controls and templates for DNA amplification were obtained from different laboratories (Table 1) (10, 25, 26, 41). These strains were maintained as frozen stocks at ⫺80°C in tryptic soy broth medium containing 10% glycerol (vol/vol). They were propagated on Luria-Bertani broth or agar containing one of the following antimicrobial agents at the appropriate concentration(s): ampi-
VOL. 75, 2009
ASSOCIATION OF AMR PHENOTYPES AND RESISTANCE GENES
1375
TABLE 2. Gene primer sequence, amplicon size, and source for eight virulence factors tested in E. coli Gene
eae estA estB stx1 stx2 fedA1 aidA-1 elt a
Primer namea
Primer sequence (5⬘33⬘)
Length (bp)
eae-F eae-R STa-F STa-R STa-F STb-R VT1-F VT1-R VT2-F VT2-R F18-F F18-R AIDA1-F AIDA1-R LT-F LT-R
ATC TTC TGC GTA CTG CGT TCA CAT TAT GGA ACG GCA GAG GT TCC CCT CTT TTA GTC AGT CAA CTG GCA CAG GCA GGA TTA CAA CAA AGT GCA ATA AGG TTG AGG TGA T GCC TGC AGT GAG AAA TGG AC TTA GAC TTC TCG ACT GCA AAG TGT TGT ACG AAA TCC CCT CTG CTA TAT CTG CGC CGG GTC TG AGA CGA AGA TGG TCA AAA CG GTG AAA AGA CTA GTG TTT ATT TC CTT GTA AGT AAC CGC GTA AGC ACA GTA TCA TAT GGA GCC A TGT GCG CCA GAA CTA TTA TTA CGG CGT TAC TAT CCT CTC TA GGT CTC GGT CAG ATA TGT GAT TC
790 171 377 530 327 510 585 275
E. coli positive control
STJ348 O157:H7 P97–2554B O149:K91 P97–2554B O149:K91 STJ348 O157:H7 STJ348 O157:H7 P88–1199 O139:K82 2787 2787 P97–2554B O149:K91
Reference(s)
5 28, 37 23 44 44 21 6, 28 18
F, forward; R, reverse.
cillin (50 g/ml), gentamicin (30 g/ml), kanamycin (50 g/ml), tetracycline (10 g/ml), chloramphenicol (10 g/ml), trimethoprim (10 g/ml), and sulfamethazine (200 g/ml). The PCR primers used in this study have all been previously described, as have their amplified product sizes and the references for the corresponding strains used as amplification templates (Table 1) (10, 25, 26, 41). Amplifications were performed with 5 l of supernatant from bacterial preparations that had been boiled for 10 min (15). The PCR mixture (total volume, 50 l) included 29.6 l of H2O, 5.0 l of 10⫻ PCR buffer (Amersham Pharmacia Biotech Inc., Piscataway, NJ), a 10 mM concentration of each of the four deoxynucleoside triphosphates, 1 U of Taq DNA polymerase (Amersham Pharmacia Biotech Inc.), 25 pmol of each primer, and 5 l of template. DNA amplification was carried out in a GeneAmp PCR system 9700 (Perkin-Elmer, Foster City, CA) by using the following conditions: 5 min at 94°C, followed by 30 cycles of 94°C for 30 s, 50°C for 30 s, and 72°C for 1.5 min. A 3-l aliquot of the PCR product was verified for size and purity by gel electrophoresis (1.2% [wt/vol] agarose in 1⫻ Tris-acetateEDTA buffer). Colony hybridizations were performed as previously described (20). Briefly, each bacterial colony was grown on LB-agarose medium and applied to filter paper (541; Whatman Inc., Florham Park, NJ). Colonies were lysed, dried, exposed to amplified probes that had been labeled with [␣-32P]dCTP by using a DNA labeling beads kit (Amersham Pharmacia Biotech Inc.), and washed. Film was exposed to the marked colonies and developed. Black spots corresponding to a colony were interpreted as a positive result (20). Virulence gene detection. Eight virulence factors that are carried by E. coli causing gastrointestinal disease in humans (AIDA1, intimim, VT1, and VT2) or pigs (intimin, F18, LT, STa, and STb) were selected for testing (2, 6, 9, 45). Virulence gene testing was conducted by Prairie Diagnostic Services, Saskatoon, Saskatchewan, Canada. Each strain was grown up on blood agar plates, and two to four E. coli colonies were randomly selected and mixed in 400 l of D-Solution (4 M guanidinium thiocyanate, 25 mM Na citrate at pH 8.0, 0.5% sarcosyl, 0.1 M -mercaptoethanol [700 l in 100 ml]) (Sigma, Oakville, Ontario, Canada). To each tube, 100 l of Tris-EDTA buffer-saturated phenol (Sigma) and 100 l of chloroform (Sigma) were added, and the contents of the tube were mixed. The DNA was lysed by incubation at ⫺20°C for 10 min and then centrifuged for 5 m at 4°C and 15,000 ⫻ g, and the aqueous layer was removed. The process of washing, mixing, and centrifuging was repeated at least once or until the interface was clear. A total of 500 l of 95% salted ethanol (VWR, Luterworth, Leicestershire, England) was added, inverted to mix, and incubated for 1 to 12 h at ⫺20°C. The tube was then centrifuged for 15 m at 4°C at 15,000 ⫻ g, and the ethanol was decanted off. The DNA pellet was dried for 5 to 10 min at 30 to 35°C and dissolved in 80 to 100 l of sterile water. Oligonucleotide primers were used for the detection of virulence-associated genes (Table 2). Bacterial DNA amplification was performed in 30 l of sterile, ultrapure water, 5 l of 10⫻ PCR buffer, 4 l (25 mM/l) of MgCl2, 0.5 l of the deoxynucleoside triphosphates (25 mM/l), 0.5 l of Taq polymerase (5 U/l), and four primers (2 l per primer, or 20 pmol/l) (Fermentas, Burlington, ON, Canada). Two microliters of DNA was dispensed into each tube, and the tubes were centrifuged for 30 s and then immediately
placed into the preheated cycler. The cycler denatured the DNA for 2 min at 94°C and then amplified it by 35 cycles as follows: denaturing for 30 s at 94°C, annealing of primers for 30 s at 60°C, and extension for 30 s at 72°C. Final extension occurred for 10 min at 72°C, and the reaction was concluded at 10°C. The PCR amplicons were visualized following electrophoresis on 1.25% agarose gel and staining with ethidium bromide. Amplicons were compared to a sterile negative control and a positive reference strain (Table 2). Results were recorded with an Alpha Imager documentation camera (Fisher Scientific, Ottawa, Ontario, Canada). Statistical analyses. MIC, AMR gene, and virulence factor data were maintained in a relational database (Microsoft Access; Microsoft Corporation, Redmond, WA). For the purpose of this analysis, isolates with intermediate MICs were considered susceptible. Descriptive and nonparametric statistical analyses were performed with commercially available software (SPSS, version 14.0, for Windows; SPSS, Chicago, IL) while tests of association were conducted using generalized estimating equations (PROC GENMOD; SAS, version 9.1, for Windows; SAS Institute, Cary, NC). Model outcomes were at the isolate level and were adjusted for clustering of resistance within herds. The association between the number of drugs in each isolate’s resistance phenotype (outcome) and the number of resistance genes detected (predictor) was investigated using a Poisson distribution, a log link function, and an exchangeable correlation structure. Associations were considered statistically significant at a P value of ⬍0.05 and were reported as a risk ratio (equal to exp) with 95% confidence intervals (CI). Resistance to each drug was described in two ways: (i) an isolate had a “resistant” phenotype if the MIC was greater than the breakpoint, and (ii) an isolate had a “resistant” genotype if it carried a gene encoding for resistance to that drug. The agreement between these descriptions was evaluated using the kappa statistic (), and unconditional associations (odds ratios [OR]) (16). Each resistance phenotype was evaluated to determine whether it was associated with a summary description of the resistance genes (presence or absence of any resistance genes for that drug) and then whether it was associated with each individual resistance gene for that drug. Three additional sets of analyses were conducted. The first set evaluated the unconditional associations between individual AMR genes. The second set described the unconditional associations between the various resistance phenotypes. The final set of analyses described the unconditional association between each AMR gene and the presence of each virulence factor. All models had a logit link function, binomial distribution, and exchangeable correlation structure. The association between each variable of interest and outcome was considered significant at a P value of ⬍0.05 and was reported as an OR (exp) with a 95% CI (16). Genes observed in less than 2% of the isolates were not considered in any analyses to avoid problems with model power, stability, and convergence. Associations of interest with a zero in the contingency table were noted and evaluated for significance with a two-sided Fisher exact test (significant at a P value of ⬍0.05).
1376
ROSENGREN ET AL.
APPL. ENVIRON. MICROBIOL.
TABLE 3. Number of drug resistances, AMR genes, and virulence genes in commensal E. coli isolates No. of isolates positive fora:
No. of resistance genes or drug resistances
Drug phenotype
Resistance gene(s)
Virulence gene(s)
0 1 2 3 4 5 6 7 8
16 3 24 37 35 25 11 0 0
13 24 29 17 34 21 10 1 2
49 28 45 22 3 2 1 1 0
a
The total number of isolates was 151.
RESULTS Most isolates examined in this study were resistant to at least three of the seven drugs and carried at least three AMR genes (Table 3). Only 11% of the isolates were susceptible to all of the tested drugs, and 9% had no resistance genes. The number of AMR genes detected in an isolate was a significant predictor of the number of drugs in its resistance phenotype (risk ratio, 1.19; 95% CI, 1.15 to 1.24; P ⫽ 0.0001). In this sample, the most common resistance genes identified were tetA and tetB. The second-most-common resistance genes were for streptomycin; all isolates with strA also carried strB. Therefore, all analyses considered these genes as one unit, the
strA-strB gene pair. Five resistance genes were detected in less than 2% of the isolates (Table 4), while 11 were not identified in any of the E. coli isolates: blaSHV, aac(3⬘)-IIa, ant(2⬙)-1a, catI, floR, cmlA, dhfr1b, dhfrV, dhfrVII, dhfrIX, and dhfrXII. On average, 7.6 isolates were tested from each herd (range, 4 to 10). For the three drugs to which phenotypic resistance was most common, more than one type of resistance gene was identified in most herds. All three tetracycline efflux genes were detected in 2 herds, two were identified in 16 herds, and only one was noted in 2 herds. All three dihydropteroate synthetase (sul) genes were detected in 5 herds, two were detected in 10 herds, and one was detected in 4 herds. One herd had no sul genes detected even though four sulfamethoxazole-resistant isolates were detected in the herd. One herd had three streptomycin resistance determinants, 12 herds had two, and 7 herds had one. The agreement between resistance, as described by phenotype and genotype, ranged from a of 0.85 for ampicillin to 0.33 for sulfamethoxazole. No outcome had perfect agreement because every antimicrobial outcome had some isolates with a resistant phenotype but no genetic explanation (Table 4). This was most pronounced for chloramphenicol resistance; no isolates carried any of the three chloramphenicol resistance genes examined. The reverse situation also occurred. Twelve different resistance genes were identified in at least one susceptible isolate. Despite these inconsistencies, each set of resistance genes was a significant predictor of its resistance phenotype, except for resistance to chloramphenicol. When each gene was
TABLE 4. Comparison of AMR in E. coli isolates according to phenotypic and genotypic testing Agreement between resistance phenotype and genotype (n ⫽ 151)e
Characteristics of isolate Antimicrobial NP a
Gene(s) carried
NGb
P⫹/G⫺c
P⫺/G⫹d
Ampicillin
69 48 34
Streptomycin
82
60 60 0 21 16 3 2 79 35 1 51 69 22 27 40 124 52 81 2 14 7 6 1
10
Chloramphenicol Kanamycin
Any blaTEM Any Any aph(3⬘)-1a aph(3⬘)-IIa aac(3)-IV Any aadA1 aadA6 strA-strB Any sul1 sul2 sul3 Any tetA tetB tetC Any dhfrI dhfrXIII dhfrXV
1 1 0 1 0 0 1 17 10 0 7 8 1 1 6 4 3 1 1 4 3 1 0
Sulfamethoxazole
111
Tetracycline
125
Tetracycline Trimethoprim-sulfamethoxazole
a
24
48 14
20
42
5
14
95% CI
P
54–5280 54–5280
⬍0.0001 ⬍0.0001
0.85
55 32
16–182 9.4–108
⬍0.0001 ⬍0.0001
0.67
10 2.5
5.3–20 1.2–5.3
⬍0.0001 0.01
0.51
11 3.3 9.7 13 2.7 95 3.5 17 0.4 23 8.3 35
5.3–22 2.6–16 1.2–78 1.2–142 0.9–8.0 29–312 1.2–10 4.6–61 0.1–3.4 4.6–116 1.6–44 2.0–619
⬍0.0001 ⬍0.0001 0.03 0.03 0.06 ⬍0.0001 0.03 ⬍0.0001
OR
534 534
NP, number of E. coli isolates expressing phenotypic resistance to the indicated antimicrobial agent. NG, number of E. coli isolates carrying the indicated resistance gene. P⫹/G⫺, number of phenotypically resistant E. coli isolates (P⫹) with no resistance gene (G⫺) for the drug identified. d P⫺/G⫹, number of phenotypically susceptible E. coli isolates (P⫺) with a resistance gene (G⫹) for the drug identified. e Association between drug phenotypes and resistance genes and phenotype-genotype intertest agreement. n, number of isolates. b c
⬍0.0001 0.01 0.01
0.33
0.77
0.46
VOL. 75, 2009
ASSOCIATION OF AMR PHENOTYPES AND RESISTANCE GENES
TABLE 5. Significant OR between E. coli resistance genes and 95% CIs Outcome resistance gene
Outcome-predictor association
Predictor resistance gene
OR
95% CI
P
strA-strB tetB sulI tetA blaTEM sul1 sul2 tetA tetB aadA1 strA-strB tetA tetB dhfrI strA-strB aadA1 strA-strB sul1 tetB dhfrI blaTEM strA-strB tetA sul1 sul2
2.4 2.9 70.4 7.8 2.4 0.2 6.8 0.3 4.3 76.8 0.2 9.0 0.4 8.4 8.5 8.1 0.3 8.6 0.1 4.7 2.9 4.3 0.1 9.4 4.5
1.0–5.4 1.4–5.8 12.0–412 2.7–23.0 1.0–5.5 0.1–0.7 2.9–16.3 0.2–0.7 1.9–9.8 14.7–402 0.0–0.6 3.2–25.2 0.2–0.9 1.6–43.6 3.4–21.7 2.7–24.4 0.1–0.6 3.0–24.1 0.0–0.1 1.2–18.0 1.5–5.8 1.9–9.8 0.1–0.2 1.8–48.0 1.4–14.1
0.04 0.003 0.0001 0.0002 0.04 0.01 0.0001 0.003 0.0004 0.0001 0.005 0.0001 0.04 0.01 0.0001 0.0002 0.001 0.0001 0.0001 0.02 0.002 0.0005 0.0001 0.007 0.01
TABLE 6. Significant OR between E. coli drug resistance phenotypes and 95% CIs
aadA1 strA-strB
sul1
sul2 tetA
tetB dhfrI
considered individually, only sul3 was not a significant predictor of its own phenotypic resistance. Three genes, aph(3⬘)-1a, dhfrXIII, and sul3, were not associated with any other resistance gene. In contrast, sul1 was associated with five other genes. Three associations showed an increased probability of detecting sul1 in the presence of another gene, while two suggested a decreased probability (Table 5). The strongest association identified was between aadA1 and sul1. Phenotypic resistance to each drug was associated with phenotypic resistance to at least one other drug, and every identified association was positive (showed an increased probability of observing the outcome resistance in the presence of the predictor resistance) (Table 6). The OR could not be estimated for six pairs because no isolates demonstrated one of the combinations of resistance and susceptibility (i.e., 0-0, 0-1, 1-0, or 1-1, where 0 indicates the absence of the phenotype and 1 indicates its presence). However, these resistance combinations were all highly significant on nonparametric analysis (Fisher’s exact test, P ⬍ 0.0001). Importantly, with the exception of the association between tetA and dhfrI, every significant association between AMR genes (Table 5) matched a significant association between the phenotypes encoding those genes (Table 6). One virulence factor, LT (encoded by elt), was not identified, and another, VTI (stx1), occurred in one isolate. The remaining factors were identified more frequently: intimin (eae), 17 isolates; F18 (fed1), 17; STb (estB), 23; AIDA (aida-1), 49; STa (estA), 52; and VT2 (stx2 or stx2e), 58. Thirty-two percent of the E. coli isolates had no virulence factors identified while almost half carried two or more (Table 1). The number of AMR genes identified in an isolate was not
Outcome-predictor association
Resistance phenotype (drug) Outcome
blaTEM
1377
Ampicillin
Kanamycin Streptomycin
Tetracycline Sulfamethoxazole
Trimethoprimsulfamethoxazole Chloramphenicol
Predictor
ORa
95% CI
P
Streptomycin Tetracycline Trimethoprimsulfamethoxazole Streptomycin Ampicillin Kanamycin Tetracycline Sulfamethoxazole Ampicillin Streptomycin Sulfamethoxazole Streptomycin Tetracycline Trimethoprimsulfamethoxazole Chloramphenicol Ampicillin
2.8 11.0 3.5
1.6–4.7 3.4–35.3 1.1–10.4
0.0002 0.0001 0.03
2.6 3.2 3.0 * 2.0 16.6 * 7.8 1.9 8.4 *
1.3–5.0 1.8–5.8 1.1–8.4
0.01 0.0001 0.04 ⬍.0001 0.01 0.0001 ⬍.0001 0.0001 0.02 0.0001 ⬍.0001
Sulfamethoxazole Sulfamethoxazole
* 3.5 * *
1.2–3.4 3.9–70.7 3.9–15.7 1.1–3.3 4.3–16.4
1.2–10.1
⬍.0001 0.02 ⬍.0001 ⬍.0001
a The asterisk indicates that the OR could not be estimated because no isolates demonstrated one of the combinations of resistance and susceptibility (i.e., 0-0, 0-1, 1-0, or 1-1 关see text兴) (nonparametric Fisher’s exact test).
associated with the number of virulence genes detected (P ⫽ 0.97), nor was there an association between the presence of any resistance gene and any virulence gene (P ⫽ 0.92). Three unconditional associations were identified between the presence of a specific AMR gene and a virulence factor. The odds of identifying STb were 4.7 times higher (95% CI, 1.1 to 21; P ⫽ 0.04) in isolates carrying dhfrXIII. The odds of identifying VT2 were 5.2 times higher (95% CI, 1.7 to 16; P ⫽ 0.004) in isolates carrying dhfrI. The odds of identifying AIDA were decreased 0.5 times (95% CI, 0.2 to 0.9; P ⫽ 0.03) in isolates with tetB. DISCUSSION This study described the distribution of resistance and virulence genes in E. coli isolates from healthy grow-finish pigs in Alberta and Saskatchewan swine herds. It also described the associations between resistance genes, between resistance phenotypes, and between resistance and virulence genes. The absolute frequency of genes and phenotypes in these isolates do not necessarily represent the prevalence of these characteristics in the study herds or other herds in western Canada because isolate selection was not random. Therefore, absolute prevalence data should not be extrapolated beyond this study, and interpretation of the associations and relative frequencies should be carried out with some care. Despite this limitation, comparing the genes found in these isolates to other North American reports is worthwhile because, to the authors’ knowledge, other descriptions of AMR genes in E. coli from healthy pigs in western Canada are not available, and few studies have taken this approach to investigating gene associations.
1378
ROSENGREN ET AL.
Phenotypic and genotypic AMR. Tetracycline was the most common phenotypic resistance. Although at least 36 genes encode tetracycline resistance, two efflux genes explained more than 90% of the resistance in these isolates (14). This was similar to other reports of commensal and pathogenic E. coli from pigs (10, 11, 22, 26). The next most commonly identified genes encoded streptomycin resistance. In contrast to an Ontario study where E. coli from healthy pigs carried aadA genes almost twice as frequently as strA-strB (10), these isolates more commonly carried strA-strB. This finding could reflect regional differences but could also be an artifact of isolate selection. Resistant isolates (MIC ⱖ 64) were more likely to be selected. These isolates would be more likely to carry strA-strB because these genes encode a higher level of resistance (MIC ⱖ 64) than the adenylating genes (aadA), which result in MICs of ⬍64 (10, 38). The sul3 gene was first described in 2003 in Switzerland (29). It has since been reported in E. coli from pigs in Ontario (Canada), Oklahoma, and numerous European countries (8, 10, 19, 34, 39). The sul3 gene appears to be widespread in western Canada as it was the most common sulfonamide resistance gene in this study and was identified in two-thirds of study herds. Finally, resistance to ampicillin was well described by blaTEM, which was similar to a description of E. coli O149:K91 from sick pigs in Quebec, Canada (26). In contrast to the above resistance types, which were well described by the genes examined, resistance to kanamycin, trimethoprim-sulfamethoxazole, and chloramphenicol were poorly explained. Only 59% of the kanamycin and 42% of the trimethoprim-sulfamethoxazole resistant E. coli isolates carried a putative resistance gene. Future studies may need to consider more of the known dihydrofolate reductase and kanamycin resistance genes (30, 36). Even more notable was the failure to identify any chloramphenicol resistance genes. Although many different chloramphenicol resistance genes have been described, three have been reported in E. coli from North American pigs (8, 35, 41). Resistant E. coli from healthy pigs in Ontario, Canada, and sick pigs in the United States predominately carried cmlA, while catI and floR occurred less frequently (8, 41). Considering that chloramphenicol has been banned in Canadian livestock since 1985 (20a), resistance has presumably been maintained through coselection of chloramphenicol genes with other resistance and virulence genes (7, 41). Future research should investigate what chloramphenicol resistance genes are prevalent in western Canada. Identifying sul3 in various populations gave insight into how resistance can spread (8, 10, 19, 29, 34, 39). Similarly, describing differences in the chloramphenicol resistance genes between pig populations might provide insight into barriers for the spread or persistence of resistance genes. Two methods described AMR: measurement of MIC and assessment of resistance genes. These tests had fair ( of 0.2 to 0.4) to almost perfect ( of ⬎0.8) agreement (16). Yet no resistance outcome was in complete agreement. Isolates with resistance phenotypes but with no AMR genes were identified, as were isolates with susceptible phenotypes that carried resistance genes. The apparent contradiction of susceptible isolates carrying resistance genes has two possible explanations. First, there is a biological explanation. Resistance genes may be unexpressed if they are distant from or associated with a weak promoter in an integron. Similarly, free gene cassettes (not
APPL. ENVIRON. MICROBIOL.
incorporated into an integron) are silent because the integron’s promoter is required for expression (13, 17). Both of these situations could create a susceptible isolate that carries resistance genes. Low MIC test sensitivity provides an alternative explanation. Isolates could be falsely categorized as susceptible if the MIC breakpoint is higher than the resistance imparted by the gene. Such a situation is well known with aadA genes and streptomycin resistance (10, 38). We previously described the clustering of phenotypic resistance within these herds. Based on those findings, we suggested that investigating herd-level risk factors for resistance could be rewarding (33). Finding that most herds had more than one type of resistance gene for tetracycline, sulfamethoxazole, and streptomycin has prompted us to modify that conclusion. Ideally, on-farm studies exploring the relationship between antimicrobial use and resistance should use resistance genes rather than phenotypes as outcomes. Phenotype may be too crude of a measure to adequately explain coselection since antimicrobial exposure does not appear to coselect for all genes encoding a given phenotype. For example, we previously reported that the odds of phenotypic trimethoprim-sulfamethoxazole resistance were 4.6 times higher in E. coli from herds with sulfonamide use in nursery pigs (31). However, sulfonamide exposure may not select for all types of trimethoprim resistance equally as dhfrI was associated with sul1 and sul2 while dhfrXIII was not associated with any sul gene. Exploring the relationship between antimicrobial exposures and AMR genes may even explain some of the negative associations between antimicrobial use and phenotypic AMR previously reported in E. coli from healthy pigs (1, 31). Positive statistical associations between resistance genes may reflect gene linkages and, thus, cotransmission via plasmids, transposons, or integrons while negative OR might indicate gene incompatibilities (10, 41). For example, the strong association between aadA-1 and sul1 might be explained by the presence of aadA in gene cassettes and sul1 in type I integrons, which collect gene cassettes. Plasmid incompatibilities have been proposed as an explanation for negative associations between tetA and tetB (10). In general, our findings concurred with previous reports (10, 39). As in E. coli from pigs in Ontario, two sets of resistance genes were observed (10). One set included tetA, aadA-1, and sul1 while the other included tetB, blaTEM, strA-strB, and sul2. Boerlin et al. did not consider trimethoprim or beta-lactam resistance genes. However, in E. coli isolated from various types of meat in Norway, a positive association was noted between sul1 and dhfrI (39). To our knowledge, the association between tetB and blaTEM has not been previously reported. Future studies should investigate this gene pair association. Both extended-spectrum beta-lactams and tetracycline are used in pigs in western Canada (32). If exposure to one drug class increases resistance to the other, barns that use either class or that cycle between them could be selecting for resistance to both. Interestingly, our findings were distinctly different from the associations described in E. coli isolates from sick pigs in Quebec, Canada. For example, the association between sul1 and tetA was negative in Quebec but positive in this and other studies (10, 26, 39). Different E. coli types (i.e., enterotoxigenic E. coli [ETEC], pathogenic but non-ETEC, and commensal) carry different AMR genes or the same genes at different
VOL. 75, 2009
ASSOCIATION OF AMR PHENOTYPES AND RESISTANCE GENES
frequencies (10, 41). So, it is plausible that the contradictions between this study and that of Maynard et al. are because the E. coli strain was not accounted for in the analysis (26). Multiple-drug resistance is generally described as a phenotype pattern rather than as a pairwise association between drug resistances. While patterns depict the relationships between many drugs, they are difficult to compare between populations. Our findings suggest that statistical associations between resistance phenotypes might predict coselection because almost every associated pair of resistance genes was matched by an association between those genes’ purported phenotypes. The only exception was the association between tetA and dhfrI, which lacked a corresponding association between tetracycline and trimethoprim-sulfamethoxazole resistance. Although this hypothesis requires validation, it could advance our knowledge of coselection dynamics and provide a cost-effective way to evaluate existing AMR phenotype data until obtaining genotype data becomes feasible on a large scale. Thus, future research should investigate the similarity of gene and phenotype associations in other populations and data sets. Such research would have practical applications if it allowed antimicrobial prescribers to consider the effects of the use of a specific drug on resistance to other drugs. AMR and virulence genes. Postweaning diarrheas are a problem in the North American swine industry. Herds with scours outbreaks experience increased mortality and decreased productivity, which results in financial losses (2). Diarrheas are often attributed to ETEC carrying the F18 fimbrial adhesion gene (fedA1) and the virulence factors LT, STa, and STb (2, 45). Intimin, encoded by the eae gene, is known for its role in causing attaching and effacing lesions in non-ETEC diarrheas (2, 45). E. coli virulence factor genes are often located on plasmids or transposons. These virulence factors can be linked to AMR genes (10, 41), creating the potential for antimicrobial use targeted at diarrheas to coselect for virulence genes and worsen disease severity. The extent that E. coli from healthy pigs carries linked resistance and virulence genes has not been studied extensively. To address this question, virulence genes fedA1, elt, stA, stB, and eae were examined. If these genes were commonly associated with AMR genes in E. coli carried by healthy pigs, it would indicate that there is extensive potential for antimicrobial use to select for virulence traits in pigs. Such knowledge could influence the antimicrobial use decisions of swine producers and veterinarians, particularly the decision to use antimicrobials for disease prevention or growth promotion. This study’s findings indicate that linkages between the examined resistance and virulence genes were not prevalent in this sample. Thus, on-farm antimicrobial use does not appear to be selecting for these virulence factors in commensal E. coli in the studied pig population. The hypothesis that on-farm drug use could select for more severe food-borne disease was also explored by studying the relationship between AMR and virulence genes in E. coli from healthy pigs. Verotoxigenic E. coli can cause outbreaks of diarrhea in humans, with devastating sequelae including hemolytic-uremic syndrome and hemorrhagic colitis. The primary virulence factors carried by verotoxigenic E. coli are VT1 and VT2, and a relationship between VT2 and intimin has been identified (9). The virulence factor AIDA1 is associated
1379
with infantile and neonatal diarrheas caused by enteropathogenic E. coli (6). Healthy pigs can harbor E. coli carrying these virulence genes in their gastrointestinal tracts, and as these animals are near to market, their intestinal bacteria have the potential to contaminate food and cause food-borne disease. Fortunately, we observed no positive associations between the tested AMR genes and VT1, VT2, or intimin in E. coli from healthy grow-finish pigs. Had resistance and virulence genes been commonly associated, it would have suggested that onfarm drug use could create food safety risks beyond residues and resistance. Although the findings of this study differ from previous reports of numerous associations between resistance and virulence genes, the results are not contradictory (10, 41). Travis et al. identified no associations between cmlA, the predominant chloramphenicol resistance gene in commensal E. coli, and the virulence genes tested. This was in contrast to numerous associations between catA1, the predominant chloramphenicol resistance gene in the ETEC, and the virulence factors tested (41). The relationships between resistance and virulence genes appear strain specific; in interpreting studies describing such relationships in a specific type of bacteria, investigators must be cautious to not extrapolate their findings to different bacterial populations. Limitations and conclusions. This study considered the seven resistance phenotypes observed in more than 5% of a sample of E. coli isolates collected from healthy pigs. Selection ensured that these resistance phenotypes occurred in more than 10% of these isolates. This restriction minimized problems with statistical power. Due to practical constraints, only some of the resistance genes known for each antimicrobial agent were considered. Future investigations in this region should consider additional genes for chloramphenicol, trimethoprim, and kanamycin resistance as these were poorly explained. In reality, resistance genes do not interact as isolated pairs but as parts of an interconnected system. While we do not have sufficient data to present these relationships as causal pathways or multivariate models, associations between resistance genes and phenotypes appear to provide insight into coselection. If validated, this simple statistical approach may identify unforeseen repercussions from antimicrobial use. This is an important finding, as it would allow policymakers and users of antimicrobial agents to consider coselection in decisions about antimicrobial use. ACKNOWLEDGMENTS We thank the swine producers and veterinarians who participated in this project. This study was supported by the Public Health Agency of Canada. REFERENCES 1. Akwar, H. T., C. Poppe, J. Wilson, R. J. Reid-Smith, M. Dyck, J. Waddington, D. Shang, and S. McEwen. 2008. Associations of antimicrobial uses with antimicrobial resistance of fecal Escherichia coli from pigs on 47 farrow-tofinish farms in Ontario and British Columbia. Can. J. Vet. Res. 72:202–210. 2. Amezcua, R., R. M. Friendship, C. E. Dewey, C. Gyles, and J. M. Fairbrother. 2002. Presentation of postweaning Escherichia coli diarrhea in southern Ontario, prevalence of hemolytic E. coli serogroups involved, and their antimicrobial resistance patterns. Can. J. Vet. Res. 66:73–78. 3. Barza, M. 2002. Potential mechanisms of increased disease in humans from antimicrobial resistance in food animals. Clin. Infect. Dis. 34(Suppl.):S123– S125.
1380
ROSENGREN ET AL.
4. Barza, M., and K. Travers. 2002. Excess infections due to antimicrobial resistance: the “attributable fraction.” Clin. Infect. Dis. 34(Suppl.):S126– S130. 5. Beaudry, M., C. Zhu, J. M. Fairbrother, and J. Harel. 1996. Genotypic characterization of Escherichia coli isolated from dogs manifesting attaching and effacing lesions. J. Clin. Microbiol. 34:144–148. 6. Benz, I., and M. A. Schmidt. 1992. AIDA-I, the adhesin involved in diffuse adherence of the diarrhoeagenic Escherichia coli strain 2787 (O126:H27), is synthesized via a precursor molecule. Mol. Microbiol. 6:1539–1546. 7. Bischoff, K. M., D. G. White, M. E. Hume, T. L. Poole, and D. J. Nisbet. 2005. The chloramphenicol resistance gene cmlA is disseminated on transferable plasmids that confer multiple-drug resistance in swine Escherichia coli. FEMS Microbiol. Lett. 243:285–291. 8. Bischoff, K. M., D. G. White, P. F. McDermott, S. Zhao, S. Gaines, J. J. Maurer, and D. J. Nisbet. 2002. Characterization of chloramphenicol resistance in beta-hemolytic Escherichia coli associated with diarrhea in neonatal swine. J. Clin. Microbiol. 40:389–394. 9. Boerlin, P. 1999. Associations between virulence factors of Shiga toxinproducing Escherichia coli and disease in humans. J. Clin. Microbiol. 37:497– 503. 10. Boerlin, P., R. Travis, C. L. Gyles, R. Reid-Smith, H. Lim, N. Janecko, V. Nicholson, S. A. McEwen, R. Friendship, and M. Archambault. 2005. Antimicrobial resistance and virulence genes of Escherichia coli isolates from swine in Ontario. Appl. Environ. Microbiol. 71:6753–6761. 11. Bryan, A., N. Shapir, and M. J. Sadowsky. 2004. Frequency and distribution of tetracycline resistance genes in genetically diverse, nonselected, and nonclinical Escherichia coli strains isolated from diverse human and animal sources. Appl. Environ. Microbiol. 70:2503–2507. 12. Reference deleted. 13. Carattoli, A. 2001. Importance of integrons in the diffusion of resistance. Vet. Res. 32:243–259. 14. Chopra, I., and M. Roberts. 2001. Tetracycline antibiotics: mode of action, applications, molecular biology, and epidemiology of bacterial resistance. Microbiol. Mol. Biol. Rev. 65:232–260. 15. Daigle, F., J. Harel, J. M. Fairbrother, and P. Lebel. 1994. Expression and detection of pap-, sfa-, and afa-encoded fimbrial adhesin systems among uropathogenic Escherichia coli. Can. J. Microbiol. 40:286–297. 16. Dohoo, I., W. Martin, and H. Stryhn. 2003. Veterinary epidemiologic research. AVC Inc., Charlottetown, Prince Edward Island, Canada. 17. Enne, V. I., A. A. Delsol, J. M. Roe, and P. M. Bennett. 2006. Evidence of antibiotic resistance gene silencing in Escherichia coli. Antimicrob. Agents Chemother. 50:3003–3010. 18. Furrer, B., U. Candrian, and J. Luthy. 1990. Detection and identification of E. coli producing heat-labile enterotoxin type I by enzymatic amplification of a specific DNA fragment. Lett. Appl. Microbiol. 10:31–34. 19. Hammerum, A. M., D. Sandvang, S. R. Andersen, A. M. Seyfarth, L. J. Porsbo, N. Frimodt-Møller, and O. E. Heuer. 2006. Detection of sul1, sul2 and sul3 in sulphonamide resistant Escherichia coli isolates obtained from healthy humans, pork and pigs in Denmark. Int. J. Food Microbiol. 106:235– 237. 20. Harel, J., H. Lapointe, A. Fallara, L. A. Lortie, M. Bigras-Poulin, S. Laviriere, and J. M. Fairbrother. 1991. Detection of genes for fimbrial antigens and enterotoxins associated with Escherichia coli serogroups isolated from pigs with diarrhea. J. Clin. Microbiol. 29:745–752. 20a.Health Canada. 2007. Consolidation of the Food and Drugs Act and the Food and Drug Regulations. Health Canada, Ontario, Canada. http://www .hc-sc.gc.ca/fn-an/legislation/acts-lois/fda-lad/index-eng.php. 21. Imberechts, H., H. De Greve, C. Schlicker, H. Bouchet, P. Pohl, G. Charlier, H. Bertschinger, P. Wild, J. Vanderkerckhove, J. Van Damme, M. Van Montagu, and P. Lintermans. 1992. Characterization of F107 fimbriae of Escherichia coli 107/86, which causes edema disease in pigs, and nucleotide sequence of the F107 major fimbrial subunit gene, fedA. Infect. Immun. 60:1963–1971. 22. Lanz, R., P. Kuhnert, and P. Boerlin. 2003. Antimicrobial resistance and resistance gene determinants in clinical Escherichia coli from different animal species in Switzerland. Vet. Microbiol. 91:73–84. 23. Lortie, L. A., J. D. Dubreuil, and J. Harel. 1991. Characterization of Escherichia coli strains producing heat-stable enterotoxin b (STb) isolated from humans with diarrhea. J. Clin. Microbiol. 29:656–659. 24. Maidhof, H., B. Guerra, S. Abbas, H. M. Elsheikha, T. S. Whittam, and L. Beutin. 2002. A multiresistant clone of Shiga toxin-producing Escherichia coli O188:[H16] is spread in cattle and humans over different European countries. Appl. Environ. Microbiol. 68:5834–5842. 25. Maynard, C., S. Bekal, F. Sanschagrin, R. C. Levesque, R. Brousseau, L. Masson, S. Larivie`re, and J. Harel. 2004. Heterogeneity among virulence and antimicrobial resistance gene profiles of extraintestinal Escherichia coli isolates of animal and human origin. J. Clin. Microbiol. 42:5444–5452.
APPL. ENVIRON. MICROBIOL. 26. Maynard, C., J. M. Fairbrother, S. Bekal, F. Sanschagrin, R. C. Levesque, R. Brousseau, L. Masson, S. Larivie`re, and J. Harel. 2003. Antimicrobial resistance genes in enterotoxigenic Escherichia coli O149:K91 isolates obtained over a 23-year period from pigs. Antimicrob. Agents Chemother. 47:3214– 3221. 27. Mora, A., J. E. Blanco, M. Blanco, M. P. Alonso, G. Dhabi, A. Echeita, E. A. Gonza ´lez, M. I. Berna ´rdez, and J. Blanco. 2005. Antimicrobial resistance of Shiga toxin (verotoxin)-producing Escherichia coli O157:H7 and non-O157 strains isolated from humans, cattle, sheep and food in Spain. Res. Microbiol. 156:793–806. 28. Ngeleka, M., J. Pritchard, G. D. Appleyard, D. Middleton, and J. M. Fairbrother. 2003. Isolation and association of Escherichia coli AIDA-I/Stb, rather than EAST1 pathotype, with diarrhea in piglets and antibiotic sensitivity of isolates. J. Vet. Diagn. Investig. 15:242–252. 29. Perreten, V., and P. Boerlin. 2003. A new sulfonamide resistance gene (sul3) in Escherichia coli is widespread in the pig population of Switzerland. Antimicrob. Agents Chemother. 47:1169–1172. 30. Prescott, J. F. 2000. Aminoglycosides and aminocyclitols, p. 191–228. In J. F. Prescott, J. D. Baggot and R. D. Walker (ed.), Antimicrobial therapy in veterinary medicine, 3rd ed. Iowa State University Press, Ames. 31. Rosengren, L., C. Waldner, R. Reid-Smith, P. Dowling, and J. Harding. 2007. Associations between feed and water antimicrobial use in farrow-tofinish swine herds and antimicrobial resistance of fecal Escherichia coli from grow-finish pigs. Microb. Drug Resist. 13:261–269. 32. Rosengren, L., C. Waldner, R. Reid-Smith, J. Harding, S. Gow, and W. Wilkins. 2008. Antimicrobial use through feed, water and injection in 20 swine farms in Alberta and Saskatchewan, Canada. Can. J. Vet. Res. 72: 143–150. 33. Rosengren, L. B., C. L. Waldner, R. J. Reid-Smith, S. L. Checkley, M. E. McFall, and A. Rajic. 2008. Antimicrobial resistance of fecal Escherichia coli isolated from grow-finish pigs in 20 herds in Alberta and Saskatchewan, Canada. Can. J. Vet. Res. 72:160–167. 34. Sa ´enz, Y., L. Brin ˜ as, E. Domínguez, J. Ruiz, M. Zarazaga, J. Vila, and C. Torres. 2004. Mechanisms of resistance in multiple-antibiotic-resistant Escherichia coli strains of human, animal, and food origins. Antimicrob. Agents Chemother. 48:3996–4001. 35. Schwarz, S., C. Kehrenberg, B. Doublet, and A. Cloeckaert. 2004. Molecular basis of bacterial resistance to chloramphenicol and florfenicol. FEMS Microbiol. Rev. 28:519–542. 36. Sko ¨ld, O. 2001. Resistance to trimethoprim and sulfonamides. Vet. Res. 32:261–273. 37. So, M., and B. J. McCarthy. 1980. Nucleotide sequence of the bacterial transposon Tn1681 encoding a heat-stable (ST) toxin and its identification in enterotoxigenic Escherichia coli strains. Proc. Natl. Acad. Sci. USA 77:4011– 4015. 38. Sunde, M., and M. Norstro ¨m. 2005. The genetic background for streptomycin resistance in Escherichia coli influences the distribution of MICs. J. Antimicrob. Chemother. 56:87–90. 39. Sunde, M., and M. Norstro ¨m. 2006. The prevalence of, associations between and conjugal transfer of antibiotic resistance genes in Escherichia coli isolated from Norwegian meat and meat products. J. Antimicrob. Chemother. 58:741–747. 40. Travers, K., and M. Barza. 2002. Morbidity of infections caused by antimicrobial-resistant bacteria. Clin. Infect. Dis. 34(Suppl.):S131–S134. 41. Travis, R. M., C. L. Gyles, R. Reid-Smith, C. Poppe, S. A. McEwen, R. Friendship, N. Janecko, and P. Boerlin. 2006. Chloramphenicol and kanamycin resistance among porcine Escherichia coli in Ontario. J. Antimicrob. Chemother. 58:173–177. 42. van den Bogaard, A. E., and E. E. Stobberingh. 2000. Epidemiology of resistance to antibiotics: links between animals and humans. Int. J. Antimicrob. Agents 14:327–335-355. 43. Winokur, P. L., D. L. Vonstein, L. J. Hoffman, E. K. Uhlenhopp, and G. V. Doern. 2001. Evidence for transfer of CMY-2 AmpC beta-lactamase plasmids between Escherichia coli and Salmonella isolates from food animals and humans. Antimicrob. Agents Chemother. 45:2716–2722. 44. Woodward, M. J., P. J. Carrol, and C. Wray. 1992. Detection of entero- and verocyto-toxin genes in Escherichia coli from diarrhoeal disease in animals using the polymerase chain reaction. Vet. Microbiol. 31:251–261. 45. Zhang, W., M. Zhao, L. Ruesch, A. Omot, and d. Francis. 2007. Prevalence of virulence genes in Escherichia coli strains recently isolated from young pigs with diarrhea in the US. Vet. Microbiol. 123:145–152. 46. Zhao, S., D. G. White, B. Ge, S. Ayers, S. Friedman, L. English, D. Wagner, S. Gaines, and J. Meng. 2001. Identification and characterization of integron-mediated antibiotic resistance among Shiga toxin-producing Escherichia coli isolates. Appl. Environ. Microbiol. 67:1558–1564.
