Population Structure and Antimicrobial Resistance of Canine Uropathogenic Escherichia coli

Escherichia coli is the most common cause of human and canine urinary tract infection (UTI). Clonal groups, often with high levels of antimicrobial resistance, are a major component of the E. coli population that causes human UTI.

ABSTRACT

Escherichia coli is the most common cause of human and canine urinary tract infection (UTI). Clonal groups, often with high levels of antimicrobial resistance, are a major component of the E. coli population that causes human UTI. While little is known about the population structure of E. coli that causes UTI in dogs, there is evidence that dogs and humans can share fecal strains of E. coli and that human-associated strains can cause disease in dogs. In order to better characterize the E. coli strains that cause canine UTI, we analyzed 295 E. coli isolates obtained from canine urine samples from five veterinary diagnostic laboratories and analyzed their multilocus sequence types, phenotypic and genotypic antimicrobial resistance profiles, and virulence-associated gene repertoires. Sequence type 372 (ST372), an infrequent human pathogen, was the predominant sequence type in dogs at all locations. Extended-spectrum β-lactamase-producing isolates with bla_CTX-M genes were uncommon in canine isolates but when present were often associated with sequence types that have been described in human infections. This provides support for occasional cross-host-species sharing of strains that cause extraintestinal disease and highlights the importance of understanding the role of companion animals in the overall transmission patterns of extraintestinal pathogenic E. coli.

INTRODUCTION

Escherichia coli is the bacterial species most commonly isolated from human urinary tract infections (UTIs). UTIs are the third most common infection experienced by people (after respiratory and gastrointestinal infections), and as a result, a UTI is a very frequent indication for the prescription of antimicrobial drugs (1, 2). The antimicrobial resistance of extraintestinal pathogenic E. coli (ExPEC) isolates associated with human UTI has increased dramatically since the early 2000s, in large part due to the emergence of the dominant clonal strain sequence type 131 (ST131), which alone accounts for up to one-third of all UTIs (3, 4). Even when ST131 is not the predominant strain of E. coli within a given patient population, other clonal E. coli strains are similarly dominant across a national, continental, or even global scale (5, 6).

E. coli is also the most frequently isolated organism from canine UTIs, and some of the clonal types associated with human UTIs have been isolated from dogs, including ST131 (7–13). Case reports also suggest the potential for zoonotic transmission of UTI-associated E. coli; e.g., a longitudinal study within a household of five humans and a dog showed that there were at least six between-host transfers of the ExPEC strain that caused an episode of cystitis in the dog during the study period (14). Additionally, exposure to dogs and/or dog feces has been identified as a risk factor for the development of drug-resistant E. coli UTI in women (15).

Relatively little is known about the epidemiology of ExPEC in companion animals. A recent study in Scotland comparing multilocus sequence types of antimicrobial drug-resistant and -susceptible canine UTI E. coli isolates failed to identify a predominant sequence type (ST) among the 33 isolates included in the study (16). A study of UTI caused by E. coli in cats identified ST73 as the predominant feline strain (17). A better understanding of the epidemiology of ExPEC in dogs is essential to understanding the role of disseminated clones in UTI, especially those with high levels of antimicrobial resistance. Additionally, an understanding of the clonal distribution in dogs may inform us about the epidemiology of ExPEC in general, including strains that cause significant human morbidity.

We tested the hypothesis that uropathogenic E. coli in dogs has a clonal population structure with a predominant canine-specific clade. To test this hypothesis, we designed a multisite survey of antimicrobial resistance, virulence factors, and sequence types of E. coli isolates associated with UTI in dogs in the United States. We prospectively obtained E. coli isolates obtained from canine urine samples from five veterinary diagnostic laboratories, assessed the population structure of these isolates by multilocus sequence typing (MLST), determined the prevalence of a group of virulence-associated genes linked with ExPEC, and explored the resistance genotypes of isolates resistant to third-generation cephalosporins.

MATERIALS AND METHODS

Sampling time frame.

E. coli isolates obtained from canine urine samples were prospectively collected from the Washington Animal Disease Diagnostic Laboratory, Washington State University, Pullman, WA (WADDL; 9 April to 15 May 2015 and 1 October 2015 to 17 May 2016), North Dakota State University Veterinary Diagnostic Laboratory, Fargo, ND (NDSU VDL; 13 to 23 April 2015 and 1 October 2015 to 17 May 2016), Indiana Animal Disease Diagnostic Laboratory, Purdue University, West Lafayette, IN (IADDL; 10 April to 19 May 2015 and 23 September 2015 to 17 May 2016), The Ohio State University Veterinary Medical Center Microbiology Laboratory, Columbus, OH (OSU VMC; 1 October 2015 to 17 May 2016), and the University of California Davis Veterinary Medical Teaching Hospital Laboratory, Davis, CA (UCD VMTH; 1 October 2015 to 17 May 2016). All viable E. coli isolates saved during the respective study periods were included in the analysis. If a given patient had more than one laboratory submission during the study periods, only the first isolate was included in the analysis. Urine samples were submitted to the diagnostic laboratories for clinical diagnostic purposes. Generally, a single isolate was saved by the laboratories and was considered to be representative of the bacterial population causing the infection. However, if multiple E. coli colony morphologies (for example, both hemolytic and nonhemolytic colonies) were observed, a representative colony from each morphology was collected at the source laboratory and underwent molecular analysis; the isolate was included in the final analysis only if it was of a genotype that was distinct from that of the other isolate in the same patient.

Isolate preservation and DNA preparation.

Isolates were stabbed into LB agar, grown overnight at 37°C, and then stored at 4°C until shipment to Washington State University (WSU). Upon arrival at WSU, isolates were recovered on Columbia blood agar and then stored for further analysis by suspending a colony in brain heart infusion (BHI) broth with 20% buffered glycerol at −80°C. Template DNA for molecular procedures was obtained by suspending overnight growth from 1 ml Miller Luria-Bertani broth (Fisher Scientific, Fair Lawn, NJ) in sterile water, followed by heating (100°C, 10 min).

Laboratory identification and susceptibility testing.

E. coli isolates were obtained and identified using standard procedures at each laboratory. Identification at four laboratories (WADDL, IADDL, OSU VMC, and UCD VMTH) was performed by matrix-assisted laser desorption ionization–time of flight mass spectrometry (Biotyper; Bruker Daltonics, Billerica, MA, USA). At one laboratory (NDSU VDL), biochemical identification was based on colony morphology and lactose fermentation on MacConkey agar (Hardy Diagnostics, Santa Maria, CA, USA), positive spot indole test (Indole spot reagent; Hardy Diagnostics), and lactose and sucrose fermentation with no hydrogen sulfide production (Difco triple sugar iron agar; Becton Dickinson, Sparks, MD, USA). Susceptibility testing was performed at the primary source laboratories for all isolates by determining the MICs of drugs by broth microdilution using commercial susceptibility plates (Sensititre; Thermo Fisher Scientific, Waltham, MA, USA) and following manufacturer-recommended protocols. The plate formats used for susceptibility testing varied by time and laboratory location and included Thermo Scientific Sensititre products COMPAN1F, COMPAN2F, CMV1BURF, COMPGN1F, and CMV3AGNF. The MICs for the following drugs were included in the analysis: ampicillin, amoxicillin-clavulanic acid, trimethoprim-sulfamethoxazole, and enrofloxacin, based on their importance for the treatment of canine and/or human UTIs; these drugs were consistently tested at comparable MIC ranges across all plate types used for this study (18, 19). Cefpodoxime or ceftiofur was used to assess phenotypic susceptibility to third-generation cephalosporins. Susceptibility interpretation was based on CLSI VET01S 3rd edition breakpoints for Enterobacteriaceae, with canine urine breakpoints used when available (20). These canine breakpoints are very similar to human Enterobacteriaceae breakpoints for ampicillin, amoxicillin-clavulanic acid, and cefpodoxime; human breakpoints are used for trimethoprim-sulfamethoxazole (SXT) susceptibility interpretation in dogs (20, 21). Bovine interpretations were used for ceftiofur since there are no canine-specific breakpoints (20). Isolates with missing susceptibility data were tested by broth microdilution after shipment to Washington State University. If the only information missing was susceptibility information to third-generation cephalosporins, disk diffusion testing with ceftiofur (30-μg disk) using standard methods was performed (22).

MLST.

Multilocus sequence types were assigned to all isolates as previously described by Achtman et al. (23). Briefly, PCR products of seven housekeeping genes were directly sequenced either bidirectionally (adk) or unidirectionally (fumC, gyrB, icd, mdh, purA, and recA) with the Sanger sequencing services of Functional Biosciences, Inc. (Madison, WI). Previously undescribed alleles were sequenced in both directions to confirm the nucleotide sequence. MLST types were assigned using BioNumerics 6.6 (Applied Maths, Austin, TX) and EnteroBase (http://enterobase.warwick.ac.uk). During the study period, EnteroBase stopped accepting new alleles and sequence types based on Sanger sequencing, so these new alleles and STs are listed in the supplemental material (Table S1). Those STs that could not be assigned were assigned temporary placeholder ST identifiers for the purpose of this analysis.

Virulence genotyping.

A multiplex PCR and capillary electrophoresis protocol was developed with the following eight targets: papA, papC, sfa-foc, afa-dra, iutA, kpsMTII, ireA, and fyuA. Primer sequences were as previously described, with the addition of a new kpsMTII primer (5′-AGGCCGATGAACAGGGTGACCA-3′) (Table S2), with fluorescent-labeled forward primers (Applied Biosystems, Foster City, CA) (24–26). The method was validated using a control set of previously characterized isolates provided by James Johnson (Veterans Affairs Medical Center, Minneapolis, MN) (24).

Each PCR (25 μl) contained 3.0 μl PCR H₂O, 2.5 μl 10× PCR buffer, 2.0 μl 50 mM MgCl, 2.0 μl dinucleoside triphosphates (dNTPs), 0.25 μl Platinum Taq polymerase (Invitrogen/Thermo Fisher Scientific, Waltham, MA, USA), 14.25 μl of the primer pool, and 1 μl of DNA lysate as the template. One microliter of a 1:10 diluted PCR product was added to 18.5 μl formamide and 0.5 μl GeneScan LIZ 1200 ladder (Applied Biosystems, Foster City, CA, USA) and was run on an ABI 3730 analyzer at the WSU Molecular Biology and Genomics Core (Pullman, WA). GeneMarker version 2.4.0 (SoftGenetics, State College, PA, USA) was used to analyze capillary electrophoresis data using expected peak sizes based on previous reports (24–26). Isolates were classified as ExPEC if they had at least two of the following virulence gene combinations: papA or papC, sfa-foc, afa-dra, iutA, and kpsMTII, as described by Johnson et al. (27). Virulence gene scores were calculated by assigning one point for the detection of each of the eight virulence-associated genes assessed.

Broad-spectrum β-lactamase genotyping.

Isolates showing decreased susceptibility to a third-generation cephalosporin (ceftiofur MIC ≥4 or zone of inhibition <24 mm, cefpodoxime MIC ≥8) were screened for the presence of bla_CMY-2 and bla_CTX-M genes (28). Primers for bla_CMY-2 detection were designed in-house based on a published bla_CMY-2 sequence, that with GenBank accession no. JN714983.1 (CMY2F, 5′-CCTCTTTCTCCACATTTGCTGC-3′, and CMY2R, 5′-AAGTGCAGCAGGCGGATACC-3′), and bla_CTX-M was detected using previously described primers (29). Isolates that were positive on pan-bla_CTX-M PCR were grouped using previously described bla_CTX-M grouping primers (30). Isolates positive for bla_CTX-M group 9 were sequenced in both directions using group 9 sequencing primers (31). Group 1 isolates were sequenced bidirectionally using upstream and downstream bla_CTX-M group 1 primers (32). Sequencing was performed by Functional Biosciences, Inc. (Madison, WI), and sequences were analyzed with the Sequencher 5.0 software (Ann Arbor, MI, USA) and were compared to published sequences of bla_CTX-M using the NCBI Basic Local Alignment Search Tool (reference GenBank accession numbers LC259308, GQ274933, LC259307, KF155155, KX023260, and KU510263). Phenotypic extended-spectrum β-lactamase (ESBL) production, using cefotaxime (30 μg), cefotaxime-clavulanic acid (30/10 μg), ceftazidime (30 μg), and ceftazidime-clavulanic acid (30/10 μg) disks (BD BBL Sensi-Disc; Thermo Fisher Scientific, Waltham, MA, USA), was confirmed in all bla_CTX-M-positive isolates, and none of the screened isolates that were negative for bla_CTX-M and bla_CMY-2 showed an ESBL phenotype (21).

Typing of fimH.

ST131 isolates underwent typing of the fimH gene to assess similarity to human ST131 strains. A forward primer described by Weissman et al. (33) and a reverse primer from Johnson and Stell (24) were used to amplify the fimH region. The sequences of the 498-bp fimH typing region were used to assign fimH allele numbers using FimTyper 1.0 (available at https://cge.cbs.dtu.dk/services/FimTyper/) and sequence alignment using Sequencher 5.0 (Ann Arbor, MI, USA) (34). Sequencing was performed by Functional Biosciences, Inc. (Madison, WI, USA).

Patient information.

Patient information that accompanied the urine submission to the source laboratory or that was available to the source laboratory via electronic medical record was collected. These parameters included age, sex, breed, spay/neuter status, method of urine collection, history of urinary tract infection, previous urine submissions to the source laboratory, if the animal was being treated with antimicrobials when the sample was collected, if the animal received antimicrobial therapy in the month prior to sample collection, comorbidities (neurologic, neoplastic, and/or urogenital), treatment with potentially immunosuppressive drugs (corticosteroids, chemotherapy, other immunomodulating drugs), if the sample was submitted by a specialist or general practice veterinarian, urinalysis findings, and if the patient had clinical signs associated with urinary tract infection. Because of the large number of different breeds represented in the patient population, breeds were divided into the following three groups based on average adult weight: small breed (≤9.1 kg), medium breed (9.2 to 27.3 kg), and large breed (≥27.4 kg). Urinalysis parameters included information on the presence of pyuria (>5 leukocytes/high-powered field on urine sediment exam), hematuria (observed grossly or >5 erythrocytes/high-powered field on urine sediment exam), isosthenuria (urine specific gravity, 1.008 to 1.012), and proteinuria (more than a trace protein on urine dipstick or >25 mg/dl protein by photometer). Laboratory parameters, such as the quantity of E. coli isolated, number of bacterial species isolated, and number of colony morphologies of E. coli isolated, were also recorded.

For the purpose of this study, urinary tract infection was defined as the isolation of E. coli from a urine sample. Most (86%) of the dogs that had clinical data available had at least one of the following: clinical signs consistent with UTI (dysuria, pollakiuria, and increased urgency), pyuria, or hematuria. However, we cannot rule out subclinical bacteriuria in a portion of these patients, nor can we completely rule out the possibility of clinically insignificant contamination in free-catch or catheter-collected urine samples, especially in some of the patients that had missing clinical data. However, based on the isolates that had clinical data, it appears that most of the isolates truly represent urinary tract infections and not bacteriuria.

Statistical analysis.

Statistical analyses were performed with R version 3.4.3 (35) and R Studio version 1.1.423 (Boston, MA, USA). Diversity was calculated using the R packages simboot version 0.2-6 and vegan version 2.4-4 (36, 37). Phylogenetic inference using a global optimal eBURST (goeBURST) was performed using PHYLOViZ 2.0 (38). Nonparametric tests were used to compare virulence gene scores, as the scores were not normally distributed (Shapiro-Wilk normality test, P < 0.001). Mann-Whitney U tests were used for pairwise comparison of virulence gene distributions, and a Kruskal-Wallis test was used to compare multiple groups. All hypothesis tests were considered significant when P values were ≤0.05.

Logistic regression.

Multivariate models were created to assess the association between bla_CTX-M and bla_CMY-2 status with the presence of each of the eight individual virulence-associated genes papA, papC, sfa-foc, afa-dra, iutA, kpsMTII, ireA, and fyuA. Predictor variables were selected in a stepwise manner, using backwards selection.

Poisson regression with robust standard errors.

Univariate log binomial analyses were used to identify potential predictor variables for use in multivariate models. Those predictors with a Wald P value of ≤0.3, along with biologically plausible predictors and risk factors described in the literature, were used to build multivariate models (39, 40). Predictor variables were selected in stepwise manner, using backwards selection with a 10% change-in-estimate criterion for inclusion in the final model. Multivariate models were also compared using Akaike's information criterion (AIC), with a preference for models with a lower AIC. If predictors were highly correlated, the predictor variable with the lower P value in the univariate analysis was selected for inclusion in the multivariable model, and/or each predictor was assessed individually in the multivariate model and AIC values were compared. Multivariate models were created using modified Poisson regression with robust standard errors, and missing data were estimated using multiple imputations (n = 200) for the dogs that had at least baseline clinical data (age, sex, and breed) available (41). The missing data estimated by multiple imputation included spay/neuter status, quantification of E. coli, if two colony types were present in the urine, history of urinary tract infection, history of antimicrobials within the month before submission, if the animal was on antimicrobials at the time of urine collection, if there were previous urine submissions to the laboratory, if the animal received immunosuppressive therapies, if the patient had other comorbidities (cancer, including transitional cell carcinoma, or neurologic disease), if the animal had clinical signs associated with UTI, and urinalysis findings. Summaries of the raw data are in the supplemental material (Table S3). Data were assumed to be missing at random for this analysis. Multiple imputations were performed using R package Amelia version 1.7.4 (42, 43). Robust standard errors were calculated with R packages lmtest version 0.9.35 and sandwich version 2.4.0 (44, 45). A set of imputations and a multivariable model were created for each of two outcomes: (i) isolation of ExPEC status E. coli and (ii) isolation of bla_CTX-M-positive E. coli.

RESULTS

A total of 295 isolates were analyzed, with 59 isolates from the Washington Animal Disease Diagnostic Laboratory, 44 isolates from the North Dakota State University Veterinary Diagnostic Laboratory, 64 isolates from the Indiana Animal Disease Diagnostic Laboratory, Purdue University, 31 isolates from The Ohio State University Veterinary Medical Center Microbiology Laboratory, and 97 isolates from the University of California Davis Veterinary Medical Teaching Hospital Laboratory. Specialist veterinarians submitted 170 of the samples, and general practice veterinarians submitted 120 (41.3%) of the cases; the practice type of the veterinarian was unknown in 5 cases.

Population structure.

ST372 was the predominant ST at all five locations, comprising 21.7% of the 295 isolates (Fig. 1). Almost half of the isolates fell within the six most frequent STs, ST372, ST12, ST73, ST127, ST131, and ST297 (Table S3). Half of the 12 ST131 isolates were of the H30 subclone (Table 1). Thirty isolates (10.1%) were classified as new sequence types, and each new ST was detected only once. Thirty-three STs included at least two isolates, and 72 (24.3%) isolates were singletons that represented the only isolate of their respective ST; the Simpson's diversity index (1 − D) of the STs of all isolates was 0.95. Thirty-three STs included at least two isolates, and 72 (24.3%) isolates were singletons that represented the only member of their respective ST. The phylogenetic relatedness of the isolates is pictured in Fig. 2A. While ST372 was predominant at all locations, there was variability in ST distribution across locations (Fig. 1).

FIG 1.

Dominant sequence types by laboratory source of isolates. Every ST that included three or more isolates at that given location is listed. The hatched boxes represent the dominant sequence type at all five locations, ST372.

TABLE 1.

Typing of fimH in the 12 ST131 canine UTI isolates, along with their laboratory source and antimicrobial susceptibility information

fimH type	Source	Third-generation cephalosporin resistance gene	Susceptibility by antimicrobiala
			AMP	AMC	ENRO	SXT
H30	IADDL		S	S	S	S
	WADDL		S	S	S	S
	WADDL		R	R	R	R
	UCD VMTH	blaCMY-2	R	R	R	S
	UCD VMTH		R	S	R	S
	UCD VMTH		R	S	S	S
H22	WADDL		R	S	S	S
	IADDL		S	S	R	S
	UCD VMTH		S	S	S	S
H41	OSU VMC		R	S	S	S
H89	UCD VMTH		R	R	S	S
H298	NDSU VDL		R	S	R	S

^aAMP, ampicillin; AMC, amoxicillin-clavulanic acid; ENRO, enrofloxacin; SXT, trimethoprim-sulfamethoxazole; S, susceptible; R, resistance.

FIG 2.

Phylogenetic relatedness of the isolates in the study inferred by goeBURST. The size of each node is scaled to the number of isolates of that ST, and the distance between the nodes is scaled to the number of allele differences between STs. Nodes are labeled with ST number. (A) Phylogenetic relatedness of all isolates in the study. The color represents the laboratory source of the isolate. (B) bla_CMY-2 isolates in the study and the coresistances are shown for each bla_CMY-2 isolate. The color represents the laboratory source of the isolate. (C) bla_CTX-M isolates in the study and the coresistances are shown for each isolate. The color represents the bla_CTX-M gene in each isolate.

Antimicrobial resistance.

Antimicrobial susceptibility profiles varied, sometimes quite dramatically, among E. coli isolates obtained from the five laboratories (Fig. 3). The overall prevalences of resistance to the antimicrobials used most frequently to treat canine urinary tract infections were ampicillin, 26.8%; amoxicillin-clavulanic acid, 14.2%; enrofloxacin, 10.2%; and trimethoprim-sulfamethoxazole, 5.4%. Isolates carrying bla_CTX-M often had coresistance to fluoroquinolones and folate pathway inhibitors (Fig. 2C). All isolates were susceptible to carbapenems. Most of the isolates of the predominant ST372 were susceptible to the major drug classes used to treat canine UTI included in this analysis (54/64 [84.3%]).

FIG 3.

Percentage of E. coli isolates obtained from canine urine from each source laboratory with phenotypic resistance (R) to given antimicrobials and the proportion of third-generation cephalosporin-resistant isolates positive for bla_CMY-2 and bla_CTX-M by the laboratory source of the isolate.

While phenotypic resistance to third-generation cephalosporins was observed in isolates obtained from all five laboratories, bla_CTX-M-positive isolates were only identified at three of the laboratories, and the prevalences at those laboratories ranged from 5.2 to 12.9%. Most of the bla_CTX-M isolates carried bla_CTX-M-15 (n = 10), but bla_CTX-M-14 (n = 2) and bla_CTX-M-55 (n = 2) were also identified. The most frequent STs for the bla_CTX-M-positive isolates were ST410 (n = 3), ST648 (n = 2), and ST10 (n = 2); the other bla_CTX-M-positive isolates were each from a unique ST (Fig. 2C). None of the ST131 isolates were bla_CTX-M positive.

All five laboratories obtained E. coli isolates positive for bla_CMY-2, and prevalences were similar across locations (4.5 to 8.2%). A single isolate carried both bla_CTX-M and bla_CMY-2; the isolate was a new ST and positive for both bla_CMY-2 and bla_CTX-M-14. No bla_CTX-M-positive ST372 isolates were detected. Six of 15 (40%) bla_CMY-2 isolates were ST372; all of the additional bla_CMY-2 isolates represented unique STs, except for two ST973 isolates.

Virulence-associated genes.

One or more ExPEC virulence-associated genes were detected in 257 (86.8%) isolates, and 170 (57.4%) of isolates met the criterion (≥2 of papA or papC, sfa-foc, afa-dra, iutA, and kpsMTII) to be classified as ExPEC. The most frequently identified virulence-associated gene was fyuA (78%), and afa-dra was the least common (1.7%). The distribution of virulence-associated genes was varied among and within STs (Table 2 and Fig. S1). Virulence-associated gene scores varied from 0 to 7, with a median of 3 (interquartile range, 2 to 5). The virulence-associated gene scores were similar among isolates obtained from the five laboratories (Kruskal-Wallis rank sum test, P = 0.25).

TABLE 2.

Distribution of eight virulence-associated genes among the six most frequently isolated STs across all five laboratory locations. Gene prevalence varied between and within STs

^aThe distribution of the virulence-associated genes (VG) within these STs is illustrated in Fig. S1. ExPEC status is defined as the presence of ≥2 of the following: papA or papC, sfa-foc, afa-dra, iutA, and kpsMTII. Median virulence-associated gene scores and the interquartile range (IQR) of the virulence gene scores are also presented.

Virulence gene scores were lower in bla_CTX-M-positive isolates (median, 1) than in bla_CTX-M-negative isolates (median, 3; Mann-Whitney U test, P = 0.003). Fluoroquinolone resistance was also associated with lower virulence scores (median, 2) than those in fluoroquinolone-susceptible isolates (median, 4; Mann-Whitney U test, P = 0.001). Using a multivariate logistic regression model to examine the association between the eight virulence-associated genes assessed in this study and resistance to third-generation cephalosporins, we observed that papC (adjusted odds ratio [OR], 76; 95% confidence interval [95% CI], 1.4 to 2,089; P = 0.03) and iutA (adjusted OR, 21.9; 95% CI, 4.27 to 112; P ≤ 0.001) were both associated with bla_CTX-M-positive isolates. The ferric aerobactin receptor gene iutA was also associated with bla_CMY-2-positive isolates (adjusted OR, 3.44; 95% CI, 1.09 to 10.9; P = 0.04).

Risk factors.

Associations between patient characteristics, clinical signs, and comorbidities and the isolation of an isolate with ExPEC-defining virulence-associated genes or the isolation of a bla_CTX-M-positive isolate were determined (Table 3). Male dogs were more likely to have UTIs due to ExPEC. Receipt of antimicrobials at the time of urine culture and presence of transitional cell carcinoma of the urinary bladder were positively associated with isolation of a bla_CTX-M isolate. All other patient, laboratory, and clinicopathologic factors for which data were available were not associated with ExPEC or bla_CTX-M status.

TABLE 3.

Patient characteristics associated with isolation of a bla_CTX-M-positive E. coli isolate from the patient's urine sample or whether the E. coli isolate obtained from the patient's urine sample was classified as ExPEC

Predictor	Crude relative risk (95% CI)	P value	Postimputation crude relative risk (95% CI)	P value	Postimputation adjusted relative risk (95% CI) (n = 237)a	P value
blaCTX-M outcome
On antibiotics at the time of urine submission	12.6 (2.73–57.7)	0.001	14.2 (4.18–48.2)	<0.001	11.1 (2.42–51.3)	<0.001
Transitional cell carcinoma of the urinary bladder	8.05 (2.06–31.5)	0.003	9.22 (2.68–31.8)	<0.001	5.58 (1.06–29.8)	0.02
ExPEC outcome
Male	1.36 (1.11–1.67)	0.003	No imputed data		1.31 (1.04–1.65)	0.01

^aPredictors for bla_CTX-M outcome adjusted for age, sex, spay/neuter status, and laboratory of origin. Predictor for ExPEC outcome adjusted for age, spay/neuter status, breed size category, and laboratory of origin.

DISCUSSION

We investigated the population structure and antimicrobial susceptibility of E. coli isolates obtained from canine urine samples at five locations and demonstrated that dogs have a predominant ST and that dogs are susceptible to UTI caused by some of the common human ExPEC STs. We also describe E. coli virulence gene repertoires and associations between clinical parameters and problematic antimicrobial resistance genotypes.

Predominant STs among canine uropathogenic E. coli isolates.

All of the most frequently occurring UTI STs (ST372, ST12, ST73, ST127, ST131, and ST297) from the dogs in this study have previously been described in association with human extraintestinal disease, and some (ST12, ST73, ST127, and ST131) are major clonal groups in human ExPEC (46–52). The ST372 prevalence that we observed in dogs is similar to that of ST131 in human-uropathogenic E. coli studies and ST73 in feline E. coli UTIs (6, 17). This suggests that each host species may have a particular ST that comprises most of the E. coli uropathogens isolated despite the evidence of broad overlap of many STs across species. While ST131 was one of the major canine ST groups in this study, all canine ST131 isolates were bla_CTX-M negative, contrasting with the major human ST131-H30-Rx bla_CTX-M-15-carrying subclone that causes a high burden of human extraintestinal disease (5, 53, 54). However, fluoroquinolone-resistant ST131-H30 isolates were present, indicating that dogs may serve as an occasional reservoir of this important human pathogen.

Virulence-associated genes.

Fewer than 60% of canine UTI E. coli isolates met our definition of ExPEC. The virulence gene with the lowest prevalence that we observed, afa-dra, is highly prevalent (>90%) in feline-UTI-source E. coli isolates (17), suggesting a possible difference in virulence requirements for E. coli binding to urothelium in different species. Because virulence-associated genes tend to be similar across isolates of the same ST, this may account for the dominant STs that vary across animal species. Compared to female dogs, males were more likely to have ExPEC isolates than non-ExPEC isolates, indicating that virulence-associated genes may be necessary to overcome the anatomical barriers to UTIs in males. The bla_CTX-M-positive isolates tended to have lower virulence scores in our canine-source isolates. The exception was the ferric aerobactin receptor gene iutA, which was associated with bla_CTX-M positive isolates, as has been previously described in human-source isolates and was also associated with bla_CMY-2 isolates (55).

bla_CTX-M.

Although there is widespread prevalence of bla_CTX-M-positive E. coli in human clinical isolates, there were no bla_CTX-M E. coli isolates obtained at two of the veterinary laboratories, and the overall antimicrobial resistance rates in canine isolates were lower than rates reported in community-acquired UTIs in people (54, 56). For example, the mean ampicillin resistance prevalence in human-uropathogenic E. coli isolates across multiple studies before 2010 was almost twice that of the ampicillin resistance prevalence that we observed in dogs (57). Most of the bla_CTX-M-positive isolates carried the _CTX-M-15 allele, as is the case in human ESBL-producing ExPEC isolates (56). We also found bla_CTX-M-55 in two isolates of different STs from different locations, a bla_CTX-M allele that has not been previously described in clinical veterinary E. coli isolates in the United States.

We also identified risk factors for bla_CTX-M isolation, such as transitional cell carcinoma of the urinary bladder and antimicrobial treatment at the time of urine collection. Animals with bladder tumors lack some of the anatomical defenses against UTI, which may expose them to the administration of multiple rounds of antimicrobials for UTI therapy. These decreased defenses may also allow colonization and infection of the urinary bladder by isolates with fewer virulence-associated genes, which includes most bla_CTX-M-carrying isolates that we isolated. These animals may also be more likely to be in a hospital environment with potential exposure to nosocomial pathogens, such as ESBL-producing E. coli isolates. This information could potentially be used by veterinarians to identify high-risk dogs in similar patient populations, such as referral practices and veterinary teaching hospitals, that may be infected with bla_CTX-M E. coli of public health importance.

bla_CMY-2.

Unlike bla_CTX-M, bla_CMY-2 prevalence did not vary dramatically across locations. Most of the bla_CMY-2 isolates were from ST372, while bla_CTX-M isolates were more phylogenetically diverse. This suggests a different geographic distribution of bla_CMY-2 and bla_CTX-M within canine populations. ST372 may be acquiring widely disseminated bla_CMY-2-carrying plasmids within dogs, while the bla_CTX-M-carrying isolates tended to be seen in minor canine STs that are known to harbor ESBLs in other animal species (17, 46, 58, 59). It is possible that there are geographic differences in canine exposure to bla_CTX-M-positive isolates in other animals, which could contribute to the variability in bla_CTX-M prevalence observed in this study.

Study limitations.

The limitations of this study include the reliance upon veterinarians to submit clinical samples for diagnostic workups as a source of isolates and the large number of isolates for which clinical data were completely or partially unavailable. The population of patients and bacterial isolates evaluated here may not represent uncomplicated UTIs in dogs, many of which are treated empirically and diagnosed without urine culture. We included both laboratories that serve general practice veterinarians, as well as those that serve large referral centers to attempt to better represent the diversity of canine urine isolates. Additionally, clinical information was limited to the information that was provided to the laboratory by the submitting veterinarian or, in some cases, the data that were available in an electronic medical record. Missing data were estimated based on parameters for the dogs for which clinical data were provided to the laboratories, and this may not truly represent the clinical status of the dogs that did not have clinical data available.

Summary and conclusions.

The clonal population structure of uropathogenic E. coli isolates that has been described in people is recapitulated in dogs, but the predominant canine strain is ST372. Additionally, human-associated ExPEC STs that produce broad-spectrum β-lactamases were isolated from clinical infections in dogs, albeit infrequently. While a direct assessment of zoonotic and reverse-zoonotic transfer of uropathogens requires further study, we present evidence of host sharing of clinical UTI STs that cause both human and veterinary disease; people are clinically affected by ST372 and human-associated STs made up many of the bla_CTX-M-carrying isolates in canine UTI. However, this sharing appears to be minimal compared to the higher prevalence of extraintestinal infections caused by a predominant strain for each host species. The cross-species sharing of STs, including ST131-H30, and resistance genes of public health concern in clinical isolates highlight the importance of evaluating companion-animal pathogens, along with risk factors for their acquisition, to more completely elucidate the epidemiology of ExPEC.