Carriage of Extended Spectrum Beta Lactamase-Producing Escherichia coli: Prevalence and Factors Associated with Fecal Colonization of Dogs from a Pet Clinic in Lower Saxony, Germany

Simple Summary Among its role within the commensal bacterial flora, Escherichia coli (E. coli) is known as a cause of intestinal or extraintestinal diseases in pets and their owners. In order to reveal factors associated with the carriage of extended spectrum beta-lactamase-producing Escherichia coli in dogs, rectal swabs from 1000 dogs visiting a pet animal clinic in northern Germany within one year were tested. Additional data were sampled regarding, on the one hand, the dog’s health and husbandry conditions and, on the other hand, information about the owner´s medical history. Thus, we were able to define seven factors associated with extended spectrum beta-lactamase-producing Escherichia coli carriage. The high number of dogs tested and the exceptional data volume concerning the dog and owner itself, as well as those habits and interactions, underline the importance of our study to avoid the carriage and spread of pathogens, especially regarding the One Health aspect. Abstract Extended spectrum beta-lactamase (ESBL)-producing Escherichia coli are an emerging problem in veterinary and human medicine. Our study concentrated on the estimation of the prevalence and factors associated with the carriage of ESBL-producing E. coli in dogs who visited a veterinary clinic in northern Germany in 2017. For this reason, 1000 patients (healthy and sick dogs) were tested, resulting in 1000 samples originating from rectal swabs. Additional data were collected using a self-reported questionnaire that was completed by the dog owner. Factors associated with ESBL carriage were considered for further modeling if p < 0.05 using a two-sided Fisher test. Using a backward elimination procedure, the variables for the final multivariable logistic regression model were identified. In total, 8.9% of the dogs tested were positive for carriage of ESBL-producing E. coli. Seven factors were associated with the colonization of dogs with ESBL-E. coli within the multivariable model, namely husbandry system (p = 0.0019, OR = 3.00; 95% CI: 1.50–6.00), contact with puppies (p = 0.0044, OR = 2.43; 95% CI: 1.32–4.46), feeding of raw meat (p = 0.011, OR = 2.28; 95% CI: 1.21–4.31), food residues (p = 0.0151, OR = 2.31; 95% CI: 1.18–4.53) and food supplements (p = 0.0487, OR = 0.426; 95% CI: 0.18–0.96), and antibiotic treatments of dogs (p = 0.0005, OR = 3.030; 95% CI: 1.62–5.68) or owners (p = 0.041, OR = 2.74; 95% CI: 1.04–7.19) prior to the study. These factors refer to the animals themselves as well as to the owners and their habits or medical treatments. Although the causality and direction of transmission from owners to their dogs cannot be proven, the factor of antibiotic treatment of the owner is clearly associated with the dog’s status.


Study Design and Inclusion of Participants
This cross-sectional study investigated the prevalence of ESBL-producing E. coli in the dog population of a clinic in northern Germany. This pet clinic serves mostly as a referral center, with a smaller part serving as a general practitioner with 50 employees and treating about 6500 patients annually. Between October 2016 and December 2017, every owner who visited the clinic received a flyer giving information about the study and a personal briefing on the aims of the study and the study design. Dog owners agreed to participate by signing an informed consent form, including a privacy statement. This is in line with German ethical guidelines. No other inclusion or exclusion criteria were applied. Rectal swab samples were taken from 1000 dogs regardless of whether they were healthy or not.

Questionnaire
Every dog owner was asked to complete a questionnaire containing three parts. The first and second parts included information about the dog, such as breed, age, gender, country of origin in case of imported dogs, the reason for the visit, husbandry conditions and contact with other animals or people working in healthcare settings, prior diseases and treatment, diet, and general customs. Husbandry conditions have been divided into keeping within the household ("indoor dogs") and keeping only outdoors ("outdoor dogs"). Regarding diet, the questionnaire differentiated, inter alia, dry and wet food, raw food, and food residues. Supplements were defined as, e.g., vitamins and minerals. The method of keeping only outdoors includes husbandry in a shelter or garden as well as in any auxiliary building, such as a stable. The third part contained information about the owner regarding general chronic diseases and prior antibiotic treatments, contact with diarrhea patients, nutritional habits, occupation, gender, and living area. Following the interview, the data were immediately checked by the veterinarian responsible for the study for missing information and general queries from the owner. The complete questionnaire (version in the German language) is available at https://www.tiho-hannover.de/kliniken-institute/institute/ bioepi/publikationen/zusatzmaterial-publikationen/ (accessed on 3 October 2022).
The questionnaire was audited beforehand, and comprehensibility for the dog owner was assessed. The duration of the interview was less than 10 min.
All data were entered into a Microsoft Access 2003 study database.

Sample Collection and Laboratory Processing
Rectal swabs (Amies-Medium with charcoal, Mastaswab, Mast Diagnostica Reinfeld, Germany) were taken exclusively for this study by the responsible veterinarian prior to the examination. All swabs were kept at 8 • C until further processing. Bacterial isolation and identification were performed by the Institute of Microbiology and Epizootics, Centre for Infection Medicine, Freie Universität Berlin. All swabs were streaked onto self-made CHROMagar Orientation plates prepared from commercial chrome agar powder (Mast Diagnostica, Reinfeld, Germany) with added cefotaxime (2 µg/mL, Merck, Darmstadt, Germany). Cefotaxime was added after autoclaving and cooling the medium close to solidification, followed by mixing to ensure a correct concentration. Freshly prepared plates were incubated at 37 • C for 24 h. According to the manufacturer's protocols, colonies showing an E. coli-like phenotype were subcultured. Purification of the colonies was performed using the identically prepared ChromAgar Orientation plates.
Confirmatory tests were performed applying the standard CLSI guideline method [CLSI guideline VET01 2018]. For the testing of beta-lactam antibiotics, the common standard reference and quality control strain E. coli ATCC 35218 [28] was used.
Molecular confirmation of ESBL production was performed for all isolates, using specific ESBL and plasmid ampC primers by polymerase chain reaction [29].

Statistical Analysis
Based on the microbiology results, specimens were classified as "positive" for verified ESBL carriers and "negative" for cases of no verification. These outcomes were analyzed statistically for their associations with the questionnaire responses calculating odds ratios applying the two-sided Fisher test. If variables with fewer than five observations per category occurred, the categories were aggregated, or the variable was excluded from further modeling due to sparse data. Afterward, a multivariable logistic regression model without interaction terms was considered with all variables from the univariable analyses with a p-value < 0.05, and a backward elimination process was applied for the final model. To avoid multi-collinearity within the model, Cramer's V was calculated for all pairwise associations of the factors included. However, no V > 0.4 was observed. Therefore, all factors were used for modeling.
A resulting p-value of <0.05 was considered statistically significant with no multiple adjustments due to the exploratory nature of the study. All data were analyzed using the software SAS, version 9.4 TS Level 1M5 (Statistical Analysis System ® , SAS Institute Inc., Cary, NC, USA).

Study Population
Dog owners who presented their dogs to the veterinary clinic during the recruitment period (October 2016-December 2017) were asked for permission to have their dogs participate in this study. Approximately 40% of all dogs treated were enrolled to take part in the study. Those dogs that were presented for prevention were more likely to participate. A total of 1000 dogs were enrolled. Nearly half of the tested dogs were male (n = 461). The median dog age was 59.9 months (range 1-198 months). Most dogs were purebred (63.5%), and the Labrador Retriever was the most common breed (n = 65). A total of 89 different breeds were represented. Almost 75% (n = 774) of the dogs included in the study were raised in Germany. Approximately half (48.7%) of the tested dogs belonged to a household with more than one dog.
Approximately half of the patients (n = 566) presented due to any disease, whereas 434 dogs visited the clinic for preventive reasons, such as deworming or vaccination. Ophthalmic (n = 111) and orthopedic diseases (n = 96) were most frequent, which were followed by internal diseases (n = 87), follow-up consultation (n = 71), and surgery (e.g., soft tissue, joints, and fractures, n = 45). Other reasons for the visits were dental (n = 19), dermal (n = 19), and gastroenterological diseases (n = 33), as well as patients who presented to the departments of gynecology (n = 33), oncology (n = 36) and the least, were emergency cases (n = 26, e.g., trauma or intoxication).
The majority of dogs lived in rural areas (n = 693), and slightly less than 10% lived in large cities (n = 58). In total, 12.8% of the owners worked in the agricultural sector.
The percentage of dogs that were referred by other veterinarians was 15.7%, and 29.9% suffered from an acute disease. Antibiotic treatment during the previous three months prior to sampling was administered to 176 of the 1000 studied dogs.
Dog owners were mainly female (72.2%). Regarding medical issues, 320 owners had previous contact with (human or animal) patients with diarrhea in the last 12 months, while 109 of these were with human patients and 238 were with animals. Less than 5% of the owners had suffered from diarrhea four weeks prior to the study, and 43 of all owners had received antibiotic treatment two months prior to taking their dogs part in the study (for further details, see Table 1). Considering the age distribution, the prevalence was higher in younger dogs. Dogs aged under 12 months (n = 185) showed a prevalence of 11.9% (n = 22), and dogs aged between 12 and 24 months (n = 134) had a prevalence of 7.4% (n = 10). During the middle age between 24 and 48 months (n = 174), the prevalence drops to 5.2% (n = 9) and increases again in advanced age over 48 months (n = 508) to 9.6%.

Status of ESBL-Carriage and Basic Questionnaire Variables
All 1000 owners completed the questionnaires and gave full access to dogs' samples. From these, ESBL-producing E. coli (ESBL-E. coli) was isolated from 89 rectal swabs. The results of the univariable analysis of the questionnaire variables with the outcome of the sample results are summarized in Table 1 (see supplements for detailed results for all variables).
More than half of the variables related to the husbandry conditions of the tested dogs appeared to be statistically significant (p < 0.05) in the univariable model. Among those, there are some outstanding p-values of factors such as "keeping the dog outdoors' (p > 0.0001), 'contact with puppies' (p < 0.0001), 'staying in a shelter' (p = 0.0004), and 'multidog household' (p = 0.0007).
In particular, those variables that concerned feeding practices resulted in small p-values. According to that, 'feeding raw meat (p = 0.0029), food residues (p = 0.0044), feeding treats (p = 0.0002), and supplements (p = 0.02) were statistically significant factors.
Considering medical histories, a p-value of <0.0001 was observed for contact between a tested dog and any diarrhea patient (e.g., human, dog, or other animals).
Regardless of whether the dog or owner had received antibiotic treatment within the last three months prior to sampling, both were associated with ESBL-E. coli carriage of the dog (dog treated: p = 0.0003; owner treated: p = 0.0185).
The final multivariable model revealed seven factors that were associated with colonization with ESBL-producing E. coli (see Table 2). These factors can be assigned to the areas of how dogs were kept, feeding habits (especially raw meat, leftover food, and food supplements), and antimicrobial treatment of dogs and owners. Feeding habits such as the feeding of raw meat, food residues, and food supplements play an important role as risk factors for the carriage of ESBL E. coli. In addition, treatment with antibiotics of the dog or the owner also presents relevant risk factors in the multivariable model. Similarly, husbandry conditions, like keeping dogs outdoors and contact with puppies, were two relevant factors for ESBL-E. coli isolation from dog samples (see Table 2)

Discussion
Overall, an occurrence rate of 8.9% of ESBL-E. coli-positive samples was observed within this study population. The occurrences of ESBL-E. coli has been associated with factors related to husbandry, feeding habits, and the medical histories of dogs as well as their owners.
Since 1000 dogs were systematically recruited for this study, to the best of the author s knowledge, this is one of the largest studies that identifies the factors associated with ESBL-producing E. coli carriage in an ordinary dog population visiting a veterinary clinic (including preventive examinations).
However, due to the voluntary participation of the dog owners, the results of our study are possibly prone to selection bias. Therefore, the study population was compared with the overall clinic population. Dogs that were presented for prevention were more likely to participate (43.3%). Normally, only 25% of patients visit the clinic for preventive care and show no clinical signs of disease (calculation based on data from the clinic patient management software). This may lead to a bias in the resistance prevalence. The average age in the study population was slightly younger (59 months vs. 69 months in usual patients). Regarding the age distribution, the prevalence drops from younger dogs to middle age before it increases in dogs older than 48 months. Thus, our study might have underestimated the prevalence.
However, nearly 45% of the presented dogs were male, which was a proportion similar to the usual clinical population. Furthermore, the percentage of cross-breed dogs was very similar to the entire clinical population (34.7% of the clinical population versus 36.5% of the tested dog population). Overall, it may be stated that the sample is a representative sub-group of the clinic population, which itself is identified as a typical set of dog patients in the northern parts of Germany.
Due to the diagnostic procedures used, the risk of bias based on the microbiological methods used in contrast to other study results was low. The framework of this study includes only the testing of dogs without considering the possible ESBL E. coli carriage of the owner. Thus, the transmission from owner to dog or vice versa cannot be proven.
The questionnaire s basic concept has been successfully used in different studies [30,31] and was therefore elaborated based on these experiences. Therefore, the occurrence of any information bias due to the questionnaire information was avoided.
To deal with multi-collinearity, Cramer's V was calculated for all pairs of factors to identify pairwise associations. Within this procedure, no extended V > 0.4 was observed, which led to the inclusion of all factors in the multivariable model.
Various factors that were associated with ESBL-producing E. coli colonization in the univariable model could not be confirmed by the multivariable model, which generally emphasizes the need for confounding adjustment by means of multivariable statistical models. For example, there was a noticeable p = 0.0012 when dogs originated from abroad in the univariable model. However, this factor was not statistically significant in the multivariable model. This is in line with Rzewuska et al., as well as with other studies from neighboring countries, which confirm an increase in the pan-European presence of ESBL-producing E. coli, even if there were deficiencies in the comparisons among uni-and multivariable statistical analyses in these investigations [32,33].
It is a significant finding that the odds ratios that remained in the final multivariable model were stable when compared to the univariable calculations. Along with the statistical procedures, these results indicate that the included factors suggest only a little or no confounding, i.e., the factors do not influence each other. In addition, the OR of all factors remained over 2, which is a general twofold effect or greater for all factors identified.
Though the prevalence of 8.9% of ESBL-producing E. coli in dogs is generally in agreement with other studies (see Table 3), these studies used different inclusion or exclusion criteria, and the kind of study population varied considerably. In our study, the healthy subpopulation showed a higher prevalence than the diseased ones. This is not in line with other studies which reported higher prevalence within a diseased dog population [9,53].
Concerning the geographical aspect, Albrechtova et al. [54] and Wedley et al. [34] examined a dog population in a rural and semirural area with different outstanding prevalence of 75% and 0.5%. Focusing on these two extremes, the region of living could be quite relevant regarding the carriage of ESBl-producing E. coli.
Comparing dog populations visiting veterinary practices, Wedley et al. [34] found a quite low prevalence of 1.9%, which is lower than our result. Hospitalized dogs were excluded, and samples were taken from many practices, which does not allow a direct comparison with our study. Higher than in our study is the prevalence in a study completed in a veterinary clinic in Turkey [30], Tunisia [51], and China [53]. Here the prevalence were considerably higher than in our study (16.8%, 19.5%, and 41.3%). One reason could be different treatment regimens and resulting resistance prevalence.
Even German study results vary greatly. Schmiedel et al. (2014) reported a high prevalence of 52% when testing infected dogs (e.g., isolates taken from different sources, for example, routine screening for ESBL-producing E. coli, blood cultures, or urine samples) [47], while Ewers et al. (2010) reported that ESBL-producing E. coli had been isolated from 10.7% of clinical samples that were collected from dogs with urinary tract infections (UTIs), wound infections and diarrhea [41]. In contrast to these results, Schmidt et al. reported a prevalence of 1.4% when testing clinically healthy dogs [55]. It needs to be considered that these studies handled clinical samples while we studied the colonization of dogs.
In addition, isolation methods and the detection of the ESBL-producing phenotypes/genotypes have been described insufficiently in many studies, which limits direct comparisons of the study results [56].
Considering the heterogeneity of the study in Table 3, it gets apparent that the majority of publications of infected and possibly treated dogs show higher percentages of ESBL-E. coli carriers than studies that include healthy, non-treated dogs. This emphasizes the urgent need for a standardized protocol that enables a comparison of risk factors, possibly within a systematic review.
Seven factors were statistically significantly associated with the carriage of ESBLproducing E. coli in the multivariable model: keeping methods, contact of dogs with puppies, feeding of raw meat and food residues, antibiotic treatment of dog and owner, and, surprisingly, the protective role of giving feeding supplements.

Husbandry Conditions
The OR of 2.99 from our study and publications that have analyzed this factor [45,52,53] shows the importance of this factor in the carriage of ESBL-producing E. coli. In summary, 24 (19.8%) of the 121 dogs that were kept only outdoors were ESBL-E. coli carriers. In contrast, only 7.44% of dogs primarily kept indoors (n = 65) tested positive. Studies that have compared keeping conditions are quite rare. Usually, studies of multidog households involve breeders, who generally keep their dogs in kennels ("outdoor dogs"). De Graef et al. found more antimicrobial-resistant E. coli in kennel dogs than in privately owned dogs [57]. Similar results were reported by Belas et al. [44], who compared dogs from shelters/breeders and privately owned dogs. The findings of Harada, which indicate that the spread of multiresistant E. coli extends beyond different litters and affects the entire kennel [58], emphasize the transmission risk in multidog households. In particular, dogs originating from the same kennel have close contact in their outdoor pens and, thus, a common environment, which can contribute to infections by multiresistant bacteria [59]. Nevertheless, it has also been shown that some dogs remained ESBL-E. coli negative even though several of their littermates in the same household were confirmed to be ESBL-E. coli positive [30].

Contact with Puppies
Since the factors 'keeping predominantly outdoors' and 'contact with puppies' are associated with a higher risk of carrying ESBL-E. coli, it may be assumed as a surrogate for breeding dogs and a higher prevalence in these. In addition to the findings of higher incidences of antimicrobial-resistant E. coli in breeding dogs [57], the frequent misuse of antibiotics in breeding kennels is associated with the occurrence of multidrug-resistant bacteria [59]. Dogs represent a potential reservoir for pathogenic E. coli [60][61][62]. Thus, the transfer of bacteria or resistance determinants from mother to puppy can occur through the milk and vaginal flora as well as through contact with feces. In their study, Münnich and Lübke-Becker determined that vertical transfer was the most common route for E. coli infections in puppies. Similarly, ESBL-producing E. coli can be transferred among dogs and their owners through close contact [41,42,63].
The easy vertical transfer between bitches and puppies and, consequently, even subclinical infections of the puppies could explain why 14.6% of the 321 dogs that had contact with puppies in the last 12 months prior to the study were positive for ESBL-producing E. coli [63].
Dogs having contact with puppies are believed to have a higher risk for infection with ESBL E. coli, which is reflected in our results. It is probable that dogs kept by breeders have, on the one hand, regular contact with puppies and, on the other hand, are kept in groups. In general, keeping animals in groups or herds could increase the risk of spreading antimicrobial-resistant bacteria within such populations [58,64,65].

Feeding Customs
Currently, it has become popular in many countries to feed dogs with 'BARF' ('biologically appropriate raw food' or 'bones and raw food') [66,67] that originates from game and livestock. Higher prevalence values of E. coli-positive samples have been found among commercial raw pet foods when compared with conventionally processed foods [68,69]. In addition, individual diets have been supplemented with freeze-dried products, such as pig ears or tracheas. These products are usually raw and may pose a considerable risk of contamination by pathogens [69][70][71][72]. Studies have shown a connection between feeding raw food and the carriage of ESBL-producing E. coli. Thus, Leonard et al. reported that feeding raw meat or adding anything raw to the diet is associated with a higher risk of colonization with ESBL-producing E. coli [73], which is supported by the results of this study.
Van Bree et al. were able to detect ESBL-producing E. coli in 80% of analyzed commercial raw meat-based diets [74], and Hellgren et al. found Enterobacteriaceae in 100% of their samples [75]. Storing and preparation of raw foodstuffs may also pose health risks to persons who come into contact with the feed, which may pose a risk of transmission [67], especially for patients who belong to high-risk groups [66,75].
Another possible source of bacterial transmission in addition to pet food is sharing of food that is intended for human consumption. Naziri et al. (2016) reported that feeding raw meat as well as sharing human food with their own dogs poses a high risk of infection [76]. Our results confirm the findings of previous studies related to the potential transmission of ESBL-producing E. coli within the food chain in pet food and through feeding or handling of raw meat.

Antibiotic Treatment
The selection of ESBL-producing strains is promoted by antibiotic use, which has significant implications for current therapeutic options and the epidemiologic resistance situation [77]. Our findings are supported by other studies, which showed a considerably higher risk of carrying ESBL-producing E. coli after antimicrobial treatments [6,44]. Schmidt et al. described the prevalence of ESBL-producing E. coli immediately after treatment in dog feces as well as at one and three months after treatment. Although this study found an association between antimicrobial treatments and 3GCR (third-generation cephalosporin-resistant)/AmpC, no evidence showed an association between antimicrobial treatments and ESBL-producing E. coli. Nevertheless, the study indicated a higher likelihood of carrying AMR E. coli up to one month after the end of antimicrobial therapy [78].
Similar results have also been described in several studies regarding ESBL-producing E. coli in humans after antimicrobial treatments [79,80], especially in the case of inappropriate use of antibiotics [81]. Even up to three months after antimicrobial treatments, a higher risk for carriage of ESBL-producing Enterobacteriaceae was observed [37].
Thus, the association between antibiotic treatments and the carriage of ESBL-producing E. coli has been described and emphasized in several publications, which is supported by our study.

Feeding Supplements
Our results suggest that the provision of dietary supplements (e.g., vitamins and minerals) has a protective effect on dogs. Unexpectedly, dogs receiving feed supplements had a lower likelihood of being colonized with ESBL-producing E. coli. However, the p-value of 0.0487 is quite low.
Studies regarding the protective effects of food supplements, especially in small animal medicine, are quite rare. Furthermore, the association between the habits of owners who feed supplements and their influence on the dissemination of ESBL-producing E. coli is unclear.
The reasons for this result could only be estimated. Possibly, owners who are used to feeding supplements care more for their dog's health. The interaction between dogs and supplement-feeding owners needs further investigation to better understand our findings.

Conclusions
Our analyses showed in our study population an association between different factors, such as feeding raw meat and antibiotic treatments, with ESBL-producing E. coli colonization in dogs. Moreover, we identified risk factors such as "contact with humans with diarrhea" and "antibiotic intake of the owner", which suggest that transmission from human to dog and vice versa seems likely. With regard to a One Health approach, doctors of medicine as well as veterinarians should be aware of this possible transmission and its effects.

Informed Consent Statement:
Written informed consent has been obtained from the patient(s) to publish this paper. Data Availability Statement: Full data can be accessed at https://www.tiho-hannover.de/klinikeninstitute/institute/bioepi/publikationen/zusatzmaterial-publikationen/.