Raw meat diets are a major risk factor for carriage of third-generation cephalosporin-resistant and multidrug-resistant E. coli by dogs in the UK

Introduction Raw-meat diets (RMD) for dogs, comprising unprocessed or non-heat-treated animal material, are increasingly popular. However, RMDs have been demonstrated to be contaminated with antimicrobial resistant (AMR) bacteria, and there is concern that such diets may pose a zoonotic disease risk. Additionally, dogs fed RMD may shed more AMR- fecal bacteria compared to those fed conventional cooked diets. Data from the UK remain limited; the present study investigated the presence of AMR-Escherichia coli in the feces of RMD and non-RMD (NRMD)-fed dogs in the UK, the E. coli AMR gene complement, and the lifestyle risk factors associated with AMR- E. coli carriage. Methods Fecal samples from UK-owned dogs (N = 193 RMD, N = 239 NRMD) and questionnaires discussing lifestyle factors, were obtained between October 2020-August 2021. Samples underwent culture and antimicrobial susceptibility testing to determine the presence of AMR-E. coli. Whole genome sequencing determined AMR gene carriage. Risk factors for the presence of AMR-E. coli were determined by multivariable modeling. Results RMD dogs carried significantly more fecal AMR E. coli (p < 0.001), including third-generation cephalosporin resistant, extended-spectrum beta-lactamase (ESBL) producing, and multidrug resistant isolates and multivariable modeling confirmed raw-meat diets to be a significant risk factor. The blaCTX–M–15 gene was the most frequently identified blaESBL gene. The blaCTX–M–55 and blaSHV–66 genes were also prevalent and were only found in RMD dogs. The mobile colistin resistance gene, mcr-4 was identified in one ESBL-producing E. coli isolate from a NRMD-fed dog. Conclusion This study has shown that dogs fed RMD in the UK are significantly more likely to shed E. coli which is resistant to highest priority critically important antibiotics, and multidrug resistant E. coli, than dogs fed NRMD. Additionally, AMR-E. coli isolates from RMD-fed dogs harbor multiple, diverse, and novel AMR genes. Therefore, provision of RMD to dogs could pose an important potential threat to human and animal health, especially given the close nature of the relationship many owners share with their pets. Awareness of these findings should be shared with pet owners, veterinary and medical professionals, pet food manufacturers and public health to mitigate potential risks.


Introduction
Raw meat diets (RMD) for pets remain a popular alternative diet choice, and while conventional cooked kibble-based diets continue to be a staple for the majority of dogs, RMD is increasingly fed as at least a constituent part of the diet for many (Dodd et al., 2020;Morgan et al., 2022;PDSA, 2022).The 2022 PDSA PAW report estimated that 7% of UK dogs were fed RMD, equating to 790,000 dogs (PDSA, 2022).RMDs are comprised of muscle, bone, skin, cartilage, tendon and organs from livestock and wild animals, which have not undergone heat treatment or cooking during the food production process (Freeman et al., 2013;Davies et al., 2019), and may be provided in a commercial pre-prepared food, or home-prepared.RMDs for dogs and cats have been demonstrated to harbor pathogenic and zoonotic organisms, including E. coli O157:H7, Salmonella spp., Listeria monocytogenes, Campylobacter spp., amongst others (Davies et al., 2019;Kaindama et al., 2020) and such bacteria have been found to be shed by dogs and cats fed RMD globally (Morley et al., 2006;Finley et al., 2007;Lefebvre et al., 2008;Leonard et al., 2015;Baede et al., 2017;Runesvärd et al., 2020;Viegas et al., 2020;Groat et al., 2022).
Furthermore, antimicrobial-resistant (AMR) bacteria have been isolated from samples of pre-prepared RMD for pets in mainland Europe (Nilsson, 2015;Baede et al., 2017;Nüesch-Inderbinen et al., 2019) and the UK (Morgan et al., 2024) and shedding of AMR bacteria by companion animals fed RMD is of increasing concern.Provision of RMD has been identified as a risk factor for canine fecal carriage of AMR E. coli in mainland Europe (van den Bunt et al., 2020), with greater proportions of dogs fed RMD shedding extended-spectrum beta-lactamase producing (ESBL)-E.coli demonstrating resistance to highest priority critically important antibiotics (HPCIAs), including thirdgeneration cephalosporins and quinolones than those fed non-raw diets (NRMD) (Runesvärd et al., 2020).
In the UK, provision of RMD has previously been identified as a risk factor for fecal carriage of AMR E. coli in both healthy, non-veterinary visiting dogs (Schmidt et al., 2015) and those visiting veterinary practices (Wedley et al., 2017).Additionally, feeding RMD was identified as a risk factor for carriage of thirdgeneration cephalosporin resistant (3GCR) E. coli in rural-living dogs (Sealey et al., 2022).A small study indicated that AMR, 3GCR and multidrug resistant (MDR) E. coli were significantly more likely to be shed by dogs fed RMD, compared to those fed NRMD, with 31% of dogs fed RMD shedding 3GCR-E.coli, compared to 4% of dogs fed NRMD (Groat et al., 2022).
Genes encoding 3GCR (such as bla ESBL genes including those of bla CTX−M group 1 and bla CMY ) and resistance to quinolones (such as qnr genes), may also be co-harbored and plasmid encoded, thus mobile, increasing the potential for transmission of MDR.Plasmidmediated AMR genes, have been identified in E. coli isolated from companion animals, and have been associated with those fed RMD (Groat et al., 2022;Mounsey et al., 2022;Sealey et al., 2022).
Dogs and their owners share close and frequent contact, therefore the risk posed by RMD with regards to zoonotic disease and AMR is a potential public health concern.Despite the popularity and interest surrounding RMD, data from larger scale studies surrounding the potential AMR risks associated with their provision as a diet for dogs, particularly in the UK, remain limited, and associated AMR-E.coli genome sequencing data is sparse.

Aims
The aims of this study were to determine the presence of AMR E. coli in the feces of dogs fed either RMD or NRMD in the UK, with focus on 3GCR-E.coli, ESBL-producing E. coli and MDR-E.coli, alongside investigation of the AMR genes harbored by E. coli isolates via whole genome sequencing (WGS).Additionally, this study aimed to determine the dog and owner lifestyle risk factors for the carriage of such AMR E. coli in canine feces.

Materials and methods
This study was cross-sectional in design.Data were collected between October 2020-August 2021.Participant recruitment was via email contact of dog owners who had previously participated in related studies (Morgan et al., 2022) and had agreed to be contacted further, and additionally, through social media.Following recruitment, participating households were sent a questionnaire and a fecal sample collection kit via post.Dog owners were requested to collect one sample from a freshly evacuated stool at one time point from their dog.For multi-dog households, owners were requested to select one dog at random to participate in the study.Completed questionnaires and collected canine fecal samples were received by the laboratory by prepaid first class return post.Samples were stored at 4 • C and tested within 48 h of sample delivery.Participant details were anonymised, and each sample and corresponding questionnaire was assigned a unique identification number.
The questionnaire asked about dog lifestyle and clinical factors including diet, recent antibiotic treatment, and veterinary visits, recent diarrhea, and treatment, contact with other animals and access to communal areas such as dog kennels, dog shows and public parks.It also collected data on owner factors including age, location in the country, receipt of antibiotics and place of work.
Questions were multiple choice, with additional free text boxes included for owners to expand on their answers where appropriate.
Based on prior research from the UK (Groat et al., 2022), we hypothesized that the percentage of RMD-fed dogs carrying ESBLproducing E. coli was 30%, compared to 5% (or less) in dogs fed NRMD.A sample size of at least 32 in each group would provide 80% power to detect this difference in ESBL-producing E. coli with 95% confidence.

Microbiological methods
A 1 g sample of each fecal sample was homogenized in 4 ml buffered peptone water (BPW) at room temperature and incubated aerobically at 35 • C ± 1 for 18-20 h.Following incubation, a 5 µl loopful of the homogenate was inoculated onto plain chromogenic Harlequin E. coli/Coliform Agar (HECA) (Neogen, UK) and HECA infused with 1 µg/ml cefotaxime, a third-generation cephalosporin (HECA+Cx); all plates were incubated at 35 • C ± 1 for 18-20 h.Following incubation, plates were analyzed for the presence of typical E. coli colonies (dark blue-violet colonies, 0.1-2 mm diameter).To explore wider AMR diversity without selection, four typical colonies, where present, were randomly picked from each HECA plate, and two colonies were picked from each HECA+Cx plate to select for cephalosporin resistant E. coli specifically, then individually plated onto nutrient agar (NA) (Neogen, UK) plates and incubated at 35 • C ± 1 for 18-20 h.
Following incubation, zones of inhibition (ZOI) for each antibiotic disk were measured to the nearest millimeter.Human clinical breakpoints used for interpretation were as recommended by EUCAST (EUCAST, 2022) for all antibiotics other than amoxycillin-clavulanate, where the breakpoint used for interpretation was as recommended by the Clinical and Laboratory Standards Institute (CLSI, 2020).Multidrug resistance (MDR) was defined as demonstrated phenotypic resistance to three or more classes of antibiotics (Magiorakos et al., 2012).
E. coli isolates from HECA+Cx plates initially underwent the extended-spectrum beta-lactamase (ESBL) double-disk test using cefotaxime 5 µg, cefotaxime 5 µg +clavulanic acid 10 µg, ceftazidime 10 µg and ceftazidime 10 µg +clavulanic acid 10 µg disks (EUCAST ESBL detection set, MAST Group Ltd, Liverpool UK).Plates were incubated at 35 • C ± 1 for 18-20 h.For isolates where the ZOI surrounding the cephalosporin +clavulanic acid disk was a minimum of 5 mm larger than the ZOI for the corresponding cephalosporin disk alone for ≥ 1 antibiotic pair, an ESBL-producing phenotype was confirmed, and isolates underwent full AST as described above.Non-ESBL producing third-generation cephalosporin-resistant (3GCR) E. coli isolates which did not demonstrate a typical positive result for ESBL production on the double disk test, but which demonstrated a pattern suggestive of AmpC production whereby there was no, or minimal, ZOI present surrounding the clavulanic acid disk(s), were also subject to full AST.
All isolates were confirmed as E. coli by PCR of the uspA gene on cell lysates (Anastasi et al., 2010).Primers used were CCGATACGCTGCCAATCAGT (forward) and ACGCAGACCGTAGGCCAGAT (reverse), with an amplicon size of 884 base pairs, and 5× HOT FIREPol R Ready To Load Master Mix (Solis Biodyne, Estonia).

Whole genome sequencing
DNA extraction was performed on ESBL-producing E. coli isolates using the QIAamp R DNA mini kit (Qiagen, Crawley, UK).
Genomic DNA samples were submitted to the Centre for Genomic Research, University of Liverpool for Illumina NEBNext Ultra II FS DNA Library Prep, completed following the manufacturer's protocol.Each library was quantified using Qubit and the size distribution assessed using the fragment analyser.These final libraries were pooled in equimolar amounts using the Qubit and fragment analyser data.The quantity and quality of the pool was assessed by Bioanalyzer and subsequently by qPCR using the Illumina Library Quantification Kit from Kapa (KK4854) on a Roche Light Cycler LC480II according to manufacturer's instructions.
Following calculation of the molarity using qPCR data, template DNA was diluted to 300pM and denatured for 8 min at room temperature using freshly diluted 0.2 N sodium hydroxide (NaOH) and the reaction was subsequently terminated by the addition of 400 mM TrisCl pH = 8.To improve sequencing quality control 1% PhiX was spiked in.The libraries were sequenced on the Illumina R NovaSeq 6000 platform (Illumina R , San Diego, USA) following the standard workflow over 1 lane of an S4 flow cell, generating 2 × 150 bp paired end reads.
Quality-and adapter-trimmed reads were assembled using SPAdes v3.13.1 (Bankevich et al., 2012).Contigs shorter than 200 bp were removed and assemblies were included in the analyses if they had (1) assembly size +/− 50% of 4.5Mb, (2) genome completeness > 90% and duplication < 10% using BUSCO v4.0.4 (Simão et al., 2015) with the gammaproteobacteria database and (3) < 10% sample read assignment to non-Enterobacteriaceae taxa using MetaPhlAn v2.8.1 (Segata et al., 2012).MLST profiles and allele sequences were obtained from pubmlst.org.Allele sequences were aligned to assemblies using Bowtie2 version 2.3.5.1 (Langmead and Salzberg, 2012) in sensitive mode, and best aligned alleles for each locus were selected and used to determine sequence type.Sequence types were used to infer eBURST groups, using goeBURST, 1 where group members shared at least 2 ST locus alleles.Genes were predicted using PROKKA version 1.14.0 (Seemann, 2014) and used to reconstruct the core genome across samples, using Panaroo v1.2.4 (Tonkin-Hill et al., 2020) in "moderate" mode.Samples were filtered to only include those with fewer than 30% of core bases missing with subsequent recalculation.Phylogenetic estimation was carried out using the Panaroo core gene alignment, with IQ-TREE v2.0 (Nguyen et al., 2015), with 1,000 bootstrap replicates using the GTR model.Tree visualization was carried out at microreact.org.

Data analysis
Data analysis was undertaken in SPSS 27 [IBM Corp. (released 2020).IBM SPSS Statistics for Windows, Version 27.0.Armonk, NY: IBM Corp.].Descriptive analysis was undertaken to determine the frequency and percentage (with 95% confidence intervals) of antimicrobial-resistant E. coli present at sample and isolate level for dogs fed RMD or NRMD.Comparisons between dogs fed RMD and NRMD were undertaken using the chi-square test (Fisher's exact test for groups of N < 5), and significance was set at p < 0.05.
Descriptive analysis of categorical questionnaire response data (frequency, percentage) was undertaken.Based on the accompanying laboratory results, three outcomes were analyzed; "presence of ESBL-producing E. coli, " "presence of 3GCR-E.coli" and "presence of MDR-E.coli."Odds ratios and 95% confidence intervals were generated by univariable logistic regression to identify explanatory variables associated with the three outcomes.Variables with a liberal p-value of < 0.3 were selected for inclusion into each multivariable model.Correlations between each variable were assessed, and where a high correlation coefficient (> 0.7) was identified, only the variable deemed most suitable was selected for inclusion into the model.Multivariable logistic regression models were built using a backward elimination method to sequentially remove variables with a p-value of > 0.05 until all remaining variables were significant at p < 0.05.Variables which had been eliminated were individually reinserted back into the model and checked to ensure that any confounding or significant variables had not been omitted.Plausible interactions between variables were also tested in the model to ensure no significant interactions had been missed and then "goodness of fit" of the final model was tested using the Hosmer-Lemeshow test.

Ethics statement
All participation was anonymous and ethical approval was granted by the University of Liverpool Veterinary Ethics Committee (approval number VREC935).
Of the dogs fed RMD, approximately one third shed 3GCR-E.coli in their feces and a quarter shed ESBL-producing E. coli.Additionally, 17% of RMD-fed dogs shed MDR ESBL-producing E. coli, compared to 1% of those fed NRMD.
Eighty-seven (N = 75 RMD, N = 12 NRMD) 3GCR E. coli isolates demonstrated unique resistance profiles on AST within a sample.Twenty separate resistance profiles were identified from RMD isolates, and just seven from NRMD isolates.The most frequently observed resistance profiles are presented in Table 2.The most frequently observed profile in both RMD (14.7%, N = 11) and NRMD (N = 25.0%,N = 3) isolates was resistance to ampicillin, amoxycillin-clavulanate, cefotaxime and ceftazidime.
These 87 isolates underwent whole genome sequencing (WGS) and included ESBL-producing isolates, and those which were potential pAmpC based on their cephalosporin resistance profile but were not confirmed as ESBL-producing.
Multiple and varied AMR genes were identified on WGS and full results are presented in Supplementary Appendix Table 1 with a summary presented in Table 3.The E. coli isolates detected from dogs fed RMD had a wider variety of AMR genes than dogs fed NRMD (Table 3).The predominant ESBL-genes in both RMD and NRMD isolates were bla CTX−M , with bla CTX−M−15 being the most frequently found (19%, 14/75 RMD; 25% 3/12 NRMD).A wide range of bla CTX−M genes was present in RMD isolates, with 12 different genes being identified, compared to two different bla CTX−M genes identified in NRMD isolates (bla CTX−M−15 and bla CTX−M−1 ).Within the RMD isolates, bla CTX−M−55 was the second-most frequently isolated ESBL-gene (12%, 9/75), however, this gene was not present in NRMD-originating isolates.Multiple bla TEM genes were identified in the isolates, with bla TEM−1 being the most frequently isolated (Supplementary Appendix Table 1), however, in terms of ESBL-producing bla TEM genes, bla TEM−52 and bla TEM−60 were isolated in RMD E. coli isolates only.Additionally, two inhibitor-resistant bla TEM genes were identified, bla TEM−78 (N = 3 RMD, N = 1 NRMD) and bla TEM−185 (N = 3 RMD only).The ESBL-producing bla SHV−66 gene was only identified in RMD isolates.The ESBL bla OXA−45 gene was infrequently observed and was identified in one isolate each from RMD and NRMD-fed dogs.
Additionally, pAmpC genes were mainly observed in RMD isolates.By far the most frequently observed pAmpC gene was bla CMY−2 , present in 21% (16/75) of RMD E. coli isolates, whereas this gene was only identified in one NRMD isolate.With regards to genes associated with plasmid-mediated quinolone resistance, five separate qnr genes were observed in RMD isolates, and only two in NRMD isolates.The qnrS1 gene was most frequently isolated, Both RMD and NRMD isolates demonstrated the presence of variants of the parC and gyrA genes which mediate quinolone resistance.Concerningly, one RMD isolate carried the aac(6 )-Ib-cr gene, which can simultaneously result in resistance to fluoroquinolones with a piperazinyl group (e.g., ciprofloxacin) and aminoglycoside resistance.The mobile colistin-resistance encoding mcr-4 gene was identified in one NRMD-originating isolate.
Multiple plasmid-mediated AMR genes were often observed concurrently, particularly within RMD-isolates, however, presence of the AMR-genes was not always associated with phenotypic resistance.The presence of the bla CTX−M−15 gene was frequently associated with the presence of qnrS1 across a range of STs.This was the case for 9 isolates (N = 7 RMD, N = 2 NRMD), and of these, 8 isolates demonstrated MDR on AST.One isolate (ST533), from a single RMD-fed dog harbored both qnrS1 and qnrS15, was MDR, and demonstrated FQR and 3GCR.It was, however, not associated with the presence of bla CTX−M genes, but blaSHV −66 and bla CMY−2 were both present.A further isolate of interest from an RMD-fed dog (ST351) carried bla CTX−M−27 , bla CTX−M−123 , bla TEM−185 and qnrS1.and was phenotypically MDR, with FQR and 3GCR.Both isolates which carried the bla DHA−1 gene (N = 1 RMD, N = 1 NRMD) also concurrently carried qnrB4, and were the only isolates associated with the carriage of this qnr gene.Both isolates demonstrated phenotypic resistance to amoxycillinclavulanate, but only one demonstrated phenotypic FQR (RMDfed).All ST101 and ST752 isolates harbored the bla CTX−M−55 gene.All but one of the isolates which carried bla CTX−M−55 demonstrated phenotypic MDR to combinations of ampicillin, ciprofloxacin, TMS, cefotaxime and ceftazidime.All isolates which harbored the bla TEM−78 gene demonstrated phenotypic amoxycillin-clavulanate resistance, alongside being MDR.
Multiple combinations of plasmid groups were associated with bla ESBL gene carriage in the present study (Table 4), including IncF, IncB/O/K/Z, IncI, IncH and IncX groups.There were some associations of plasmids with specific STs; plasmid IncFII was identified in all ST752 isolates.

Survey data analysis
A total of 432 surveys were received.Participant demographics and univariable logistic regression results are presented in Supplementary Appendix Tables 2-4.Multivariable analysis demonstrated several dog and owner lifestyle risk factors for carriage of 3GCR-E.coli, ESBL-producing E. coli and MDR E. coli by dogs (Table 5).
There were also some common risk factors across all three outcomes, dogs fed a raw diet and dogs which had received antibiotics in the last 3 months were significantly more likely  to shed 3GCR, ESBL-producing and MDR E. coli.Dog owners were asked to report the type of antibiotic prescribed (if known), these descriptive results are shown in Supplementary Appendix Table 5.The most frequently prescribed antibiotic was amoxycillinclavulanate (N = 19 dogs), followed by metronidazole (N = 7 dogs).
Veterinary visits in the last 3 months were also common to all outcomes, dogs which had visited for an emergency appointment were more likely to shed 3GCR and ESBL-producing E. coli, whereas dogs which had attended a veterinary clinic in general were more likely to shed MDR E. coli.Dogs which were fed shop bought cooked treats/biscuits were less likely to shed 3GCR, ESBL-producing or MDR E. coli.There were some risk factors which were unique for 3GCR and ESBL-producing E. coli carriage.Dogs which attended dog shows or whose owner worked in a nursery were more likely to carry 3GCR-E.coli, however, dogs were less likely to carry 3GCR-E.coli with increasing age.Dogs that visited care homes, for example "Pets as Therapy" (PAT) dogs were more likely to carry ESBL-producing E. coli in their feces.

Discussion
This study specifically aimed to investigate the effect of diet on fecal AMR E. coli carriage by dogs and to determine the dog and owner lifestyle risk factors associated with canine carriage of AMR E. coli.It has provided strong evidence that provision of RMD to dogs in the UK is a significant risk factor, and that dogs fed RMD are significantly more likely to shed MDR E. coli and E. coli demonstrating resistance to HPCIAs in their feces than those fed a cooked diet.
In the present study, a quarter of dogs fed RMD carrying ESBL-producing E. coli, and approximately one third of dogs fed RMD carrying 3GCR-E.coli.The provision of raw meat to dogs has been identified as a risk factor for AMR E. coli carriage globally (Lefebvre et al., 2008;Baede et al., 2015;Leonard et al., 2015;Runesvärd et al., 2020;van den Bunt et al., 2020) as well as in previous smaller studies in the UK (Wedley et al., 2017;Sealey et al., 2022).Additionally, small studies in Sweden and the UK have also observed significantly greater 3GCR-and ESBLproducing E. coli carriage in dogs fed RMD than those fed NRMD (Runesvärd et al., 2020;Groat et al., 2022).The findings of the present study support those of previous studies, albeit on a larger scale, highlighting the concerning prevalence of E. coli demonstrating resistance to HPCIAs in the feces of dogs fed RMD in the UK, and providing further detail surrounding the phenotypic AMR profiles of E. coli carried by RMD-fed dogs, and the associated genotypes.
There was a significantly greater prevalence of phenotypic resistance to ampicillin, amoxycillin-clavulanate, TMS and ciprofloxacin in the AMR-E.coli isolated from dogs fed RMD than NRMD.High levels of phenotypic resistance to ampicillin, amoxycillin-clavulanate and/or TMS have been reported in dogs fed RMD previously in the UK and Sweden (Schmidt et al., 2015;Runesvärd et al., 2020;Groat et al., 2022), and a UK study of 16-week-old puppies identified that provision of a raw diet was the most substantial risk factor for FQR E. coli carriage (Mounsey et al., 2022).The findings of the present study are interesting as two previous studies demonstrated no (Groat et al., 2022) or uncommon (Schmidt et al., 2015) phenotypic fluoroquinolone resistance in E. coli isolated from healthy adult dogs in the UK.However, in the present study approximately 11% of RMD-fed dogs carried ciprofloxacin-resistant ESBL-producing E. coli (compared to < 1% of NRMD-fed dogs), and this was frequently associated with MDR.No carbapenem resistance was demonstrated in dogs fed either diet in the present study, a finding which echoes that of Runesvärd et al. (2020).
In addition, isolates from RMD-fed dogs demonstrated more varied STs and a great diversity of ESBL genes.Furthermore, STs observed in E. coli isolated from RMD-fed dogs were similar those present in samples of UK RMDs (Morgan et al., 2024), and included globally disseminated uropathogenic STs 58 and 69; ST155, which has major importance in the plasmid mediated spread of ESBL-genes from animals to humans (Matamoros et al., 2017); ESBL, pAmpC and quinolone resistance associated qnr genes associated with each isolate which underwent whole genome sequencing, alongside the sequence type (ST) and clonal complex (CC) identified and phenotypic resistance demonstrated via disk diffusion.For the "raw" column, a yellow box denotes a raw-fed dog isolate, whereas a blue box denotes a non-raw fed dog isolate.For the genes, a colored box indicates presence of a gene.For the phenotypic resistance, a black box denotes a resistance, and a gray box denotes susceptible.Although amikacin, tigecycline and meropenem were all tested via disk diffusion, no resistance was observed, and they have been omitted from this figure.*Amp, ampicillin; AmxC, amoxycillin-clavulanate; Cip, ciprofloxacin; TMS, trimethoprim-sulphamethoxazole; Ctx, cefotaxime; Ctz, ceftazidime; MDR, multidrug resistance.ˆinhibitor-resistant genes.Phylogenetic tree demonstrating the eBURST groups and sequence types (ST) present, and their relationships, for E. coli isolates which underwent whole genome sequencing.For the "group" column, a pink box denotes raw-fed, blue denotes non-raw-fed isolate.
and ST602, which is commonly isolated from UK livestock (Ludden et al., 2019).The most frequently observed genes in E. coli isolates from RMD-fed dogs in the present study were bla CTX−M−15 , bla CTX−M−55 and bla SHV−66 .While bla CTX−M−15 was identified, albeit far less frequently, in dogs fed NRMD, no bla CTX−M−55 or bla SHV−66 was present in isolates from NRMD-fed dogs.The presence of bla CTX−M−15 was frequently associated with concurrent qnrS1 carriage, as well as MDR, and was present across a range of STs.There are few studies which have specifically investigated the resistance genes present in E. coli isolated from dogs fed raw diets.However, bla CTX−M−15 has been identified as the most prevalent bla ESBL gene in samples of UK raw pet food (Morgan et al., 2024), and has been isolated from poultry and pigs at slaughter in the UK (Randall et al., 2011;Veterinary Medicines Directorate, 2021, 2022) While previous studies have demonstrated a predominance of bla CTX−M−1 in the UK healthy dog population (Wedley et al., 2017;Mounsey et al., 2022), this gene was only observed in 5 isolates from RMD-fed and one isolate from NRMD-fed dogs in the present study.The bla CTX−M−15 gene is the most frequently isolated bla CTX−M gene in canine E. coli in other countries including the USA (Lv et al., 2013), Canada (Cormier et al., 2019) and Portugal (Carvalho et al., 2021).It is also the most commonly identified blaESBL gene associated with human E. coli infections in the UK (Woodford, 2008;Woodford et al., 2011).The dominance of bla CTX−M−15 across a range of STs in the present study is interesting and, along with the WGS findings from other studies (Timofte et al., 2016;Singleton et al., 2021;Sealey et al., 2022), may demonstrate an increase in this particular gene within the canine population in the UK in general, alongside a decrease in bla CTX−M−1 carriage, as well as potentially an increased risk of bla CTX−M−15 carriage in RMD-fed dogs.A recent study of canine fecal E. coli from dogs in the South West of England demonstrated a predominance of the bla CTX−M−15 gene in urban dogs, but not rural dogs, however, excretion of E. coli with bla CTX−M genes was significantly associated with RMD-feeding in both urban and rural dogs (Sealey et al., 2022).
The bla CTX−M−55 gene is derived from bla CTX−M−15 (He et al., 2015) and is frequently identified in humans, as well as being reported in food-producing animals and pets, in China (Sun et al., 2010;Lv et al., 2013;Zhang et al., 2014).However, bla CTX−M−55 is infrequently identified in dogs in other countries, and has been reported in low numbers previously in studies from Korea (Tamang et al., 2012), Canada (Cormier et al., 2019), Portugal (Carvalho et al., 2021), France (Lupo et al., 2018), Switzerland (Zogg et al., 2018) and the Netherlands (Baede et al., 2015).Interestingly, a study of dogs fed either RMD or a conventional cooked diet in Brazil found that ESBL-producing E. coli was only shed by RMD-fed dogs, and the most commonly identified blaESBL gene was bla CTX−M−55 (Ramos et al., 2022).The high prevalence of bla CTX−M−55 in the present study (12% of RMD isolates) is a particularly intriguing finding; it has only been reported once before in dogs in the UK, in E. coli isolates from clinical samples (Bortolami et al., 2019).All isolates which carried bla CTX−M−55 in this study except one demonstrated MDR.bla CTX−M−55 has been identified in healthy pigs at slaughter in the UK (Veterinary Medicines Directorate, 2022), and was the most frequently identified bla ESBL gene in healthy broilers (Veterinary Medicines Directorate, 2021).It has also been identified in preprepared RMD containing duck in the UK (Morgan et al., 2024).Therefore bla CTX−M−55 could be an emerging bla ESBL gene of interest within Europe, as well as within the UK dog population and may be associated with provision of raw meat, particularly poultry.
The identification of bla SHV−66 in E. coli isolated from dogs fed RMD, which was not present in E. coli isolated from NRMD-fed dogs, is also of relevance.bla SHV−66 is usually 10.3389/fmicb.2024.1460143TABLE 4 Inc group plasmids associated with sequence types (STs) and ESBL genes of interest from ESBL-producing E. coli isolated from RMD-fed (N = 75 isolates) and NRMD-fed (N = 12 isolates) dog feces in the present study.
It is unsurprising that the most prevalent pAmpC gene in this study was bla CMY−2 , present across a range of STs, as this is the most frequently isolated pAmpC gene from E. coli of animal and human origin (Denisuik et al., 2013;Hansen et al., 2016).Additionally, bla CMY−2 has been demonstrated in E. coli isolated from raw pet food in the UK and mainland Europe (Nilsson, 2015;Baede et al., 2017;Morgan et al., 2024).Dogs have been frequently shown to carry E. coli which harbors bla CMY−2 in previous studies from South Korea, the Netherlands, Denmark, Costa Rica, France and the UK (Tamang et al., 2012;Baede et al., 2015;Hansen et al., 2016;Rodríguez-González et al., 2020;Haenni et al., 2022;Sealey et al., 2022).However, although it was isolated from E. coli from one NRMD-fed dog in the present study, far more E. coli isolates from RMD-fed dogs were demonstrated to carry this gene, suggesting that provision of RMD could be a risk for bla CMY−2 carriage.This finding is also supported by the multivariable model results demonstrating provision of RMD to be a risk factor for phenotypic 3GCR-E.coli carriage by dogs.
Of concern was the identification of the mcr-4 (ST4981, isolated from a NRMD-fed dog) gene in this study, which confers plasmid-mediated resistance to colistin.This isolate was also phenotypically MDR.The mcr-4 gene has previously been reported in K. pneumoniae isolated from canine feces in China (Wang et al., 2021), however, to the author's knowledge, this is the first report of isolation of this gene from canine E. coli.
Although the greatest odds for AMR E. coli shedding by dogs was associated with RMD provision, there were additional risk factors identified across the three models tested (ESBL-producing E. coli, 3GCR-E.coli and MDR-E.coli).The provision of antibiotics in the last 3 months was a significant risk factor, and has been identified as a risk factor for carriage of AMR E. coli by dogs over this timeframe previously (Gandolfi-Decristophoris et al., 2013;Wedley et al., 2017).Treatment with specific antibiotics has been linked with AMR E. coli carriage in dogs; the provision of oral cephalexin has been associated with selection of bla CMY−2 producing E. coli (Damborg et al., 2011), and carriage of MDR E. coli has been attributed to the use of fluoroquinolones (Gibson et al., 2011;Leite-Martins et al., 2014), amoxycillin-clavulanate and cefovecin (Schmidt et al., 2018).Fluoroquinolone use was not widely reported in the present study, with amoxycillin-clavulanate being the most frequently prescribed antibiotic reported.
Visiting a veterinary practice in the last 3 months was a further risk factor for AMR E. coli carriage, with an emergency visit specifically being significant for ESBL-producing and 3GCR E. coli.Previous studies have identified veterinary hospitals as sources of ESBL-producing E. coli (Timofte et al., 2016;Schmitt et al., 2021), with carriage by staff (Royden et al., 2019) and patients being reported.A further study identified frequent carriage of AMR E. coli by vet-visiting dogs, with resistance to ampicillin, tetracycline and trimethoprim most commonly detected (Wedley et al., 2017).As opposed to previous studies where hospitalization and length of stay was a significant risk factor for MDR E. coli (Gibson et al., 2011;Tuerena et al., 2016;Haenni et al., 2022), hospitalization was not significant for any of the AMR outcomes in the present study.
Research is required to investigate the potential for transmission and co-carriage of AMR E. coli between dogs, in-contact people, and the environment.Dogs and their owners frequently share close contact, especially within the home where behaviors such as sharing of soft furnishings and beds, dogs sitting on the owners lap, and dogs licking owners hands and faces occur (Westgarth et al., 2008), as well as owners kissing their pets (do Vale et al., 2021).It is this close relationship, and the behaviors associated with it, which may pose a particularly high risk for transmission of AMR-bacteria between pets and their owners.In particular, risky behaviors around food such as sharing plates, utensils and allowing pets to eat from bare hands is reported, despite owners potentially being aware of the zoonotic disease potential (Dickson et al., 2019).Dogs and humans in close contact, either within the home or within another close-contact environment such as a shelter or veterinary hospital environment, have been demonstrated to share AMR E. coli with similar resistance genes and resistance patterns (Sidjabat et al., 2006;Toombs-Ruane et al., 2020;Cozma et al., 2022;Naziri et al., 2022), and AMR E. coli of the same sequence type (Johnson et al., 2016;Grönthal et al., 2018).ESBL and AmpC-producing E. coli of the same strain has been identified between human patients with urinary tract infections and pet dogs in the same household, suggesting within-household transmission does occur (Johnson et al., 2016;Toombs-Ruane et al., 2020).
The findings of the present study highlight that such close contact should be of particular concern with dogs fed a raw meat diet, where the potential for contact with foodborne zoonotic pathogens is greater.Few studies have investigated the risks of transmission and co-carriage of AMR E. coli within a pet-owning household in relation to provision of a raw diet specifically.A study from the UK identified a common E. coli lineage (ST744) carried by a raw fed puppy and isolated from a human urinary tract infection within a local area (Mounsey et al., 2022).A previous study from The Netherlands identified co-carriage of ESBL-producing E. coli between dogs and their owners in a small number of households, and observed that provision of RMD was a risk factor for ESBLproducing E. coli carriage in dogs (van den Bunt et al., 2020).Carriage of AMR E. coli of STs which are known to be of clinical importance in human medicine has been identified in the present study to a greater degree in dogs fed RMD, associated with mobile transmissible genetic elements.Therefore, it stands to reason that dogs fed RMD could pose an increased public health risk for transmission of AMR E. coli, however, further research is required to investigate this risk.
This study relied on direct contact using email of dog owners who had previously volunteered to take part in related studies, and via social media, thus there may have been an element of bias; certain owner demographics (such as those without access to social media), and populations of dogs where raw feeding may be regularly utilized (such as hunt kennels) may be underrepresented.Owner responses within the survey could be subject to recall bias, however, this is unlikely to affected recall around the overall food type.
The HECA media used for bacterial isolation allowed easy recognition of E. coli colonies.However, it is possible some colonies could be missed if there was a slight deviation from the expected color for any reason.This could lead to underestimation of E. coli presence at sample level.A set number of E. coli picks were taken from each agar plate.This method aims to obtain a representative sample by sampling multiple colonies at random, however, does mean that there could be an over-or underrepresentation of the level of AMR present, depending on the colonies picked.Finally, the presence of the AMR-genes identified by WGS in this study was not always associated with phenotypic resistance; interpretation of the AMR genes must be undertaken with caution as their presence does not necessarily indicate that resistance will be demonstrated.Further research is needed to determine the transmissibility of genes, however, the identification of these plasmid-mediated genes should be cause for concern surrounding the potential for spread of genes capable of mediating resistance to HPCIAs to other bacteria.

Conclusion
This study has contributed to the growing body of evidence to suggest that provision of RMD to dogs is a public health concern.Dogs fed RMD were demonstrated to shed significantly greater proportions of E. coli resistant to HPCIAs than dogs fed NRMD.STs and ESBL genes were identified which are linked to those identified in livestock and humans, and associated with clinical disease in both humans and animals, and novel AMR genes not previously identified in healthy dogs were detected.This constitutes a potential One Health concern, as well as a concern for animal health and welfare.Further research is required to investigate the risks of co-carriage and transmission of AMR E. coli with respect to dogs, their owners and their environment, nevertheless, provision of RMD as a pet food choice should be considered with caution and efforts should be made to continue to engage with pet owners, pet food retailers, veterinary and medical professional with regards to the AMR bacteria risks associated with RMD feeding.

TABLE 1
Sample level data [Number (N)and percentage (%)] describing the overall phenotypic antimicrobial resistance demonstrated by E. coli isolated from the feces of dogs fed either a raw (RMD, N = 193) or non-raw (NRMD, N = 239) diet.

TABLE 3
Summary table of ESBL and pAmpC genes identified in E. coli isolates from RMD-fed (N = 75 isolates) and NRMD-fed (N = 12 isolates) via whole genome sequencing, demonstrating percentage (%) and number (N) of genes present within the isolates submitted for sequencing.
Final multivariable regression models describing explanatory variables significantly associated with dog (N = 432) fecal carriage of 3GCR-, ESBL-producing and MDR-E.coli in the present study.