Pathogenic Escherichia coli in Dogs Reveals the Predominance of ST372 and the Human-Associated ST73 Extra-Intestinal Lineages

Escherichia coli is a ubiquitous commensal and pathogen that has also been recognized as a multi-sectoral indicator of antimicrobial resistance (AMR). Given that latter focus, such as on resistances to extended-spectrum cephalosporins (ESC) and carbapenems, the reported population structure of E. coli is generally biased toward resistant isolates, with sequence type (ST)131 being widely reported in humans, and ST410 and ST648 being reported in animals. In this study, we characterized 618 non-duplicate E. coli isolates collected throughout France independently of their resistance phenotype. The B2 phylogroup was over-represented (79.6%) and positively associated with the presence of numerous virulence factors (VFs), including those defining the extra-intestinal pathogenic E. coli isolates (presence of ≥2 VFs: papA, sfaS, focG, afaD, iutA, and kpsMTII) and those more specifically related to uropathogenic E. coli (cnf1, hlyD). The major STs associated with clinical isolates from dogs were by far the dog-associated ST372 (20.7%) and ST73 (20.1%), a lineage that had commonly been considered until now as human-associated. Resistance to ESC was found in 33 isolates (5.3%), along with one carbapenemase-producing isolate, and was mostly restricted to non-B2 isolates. In conclusion, the presence of virulent E. coli lineages may be the issue, rather than the presence of ESC-resistant isolates, and the risk of transmission of such virulent isolates to humans needs to be further studied.


INTRODUCTION
Escherichia coli is both a commensal microorganism and a major pathogen in humans and animals. E. coli has also been recognized as a multi-sectoral indicator of the antimicrobial resistance (AMR) burden due to its large distribution in a wide range of ecosystems. Concerning AMR, most data on E. coli have focused on resistance to extended-spectrum cephalosporins (ESCs), conferred either by extended-spectrum beta-lactamases (ESBLs) or plasmid-mediated cephalosporinases (pAmpCs), and numerous cross-sectional studies have investigated ESCresistant (ESC-R) E. coli to decipher possible transfers across sectors. In particular, several studies have demonstrated the transmission of ESC-R E. coli between humans and dogs, mostly within households (Ljungquist et al., 2016;Gronthal et al., 2018). Overall, the distribution of ESC-R E. coli in dogs shows a wide diversity of lineages together with a few dominant clones, namely of sequence type (ST)648, ST410, and to a lesser extent, the human-associated ST131 (Ewers et al., 2014;Falgenhauer et al., 2016;Melo et al., 2019).
More globally, humans and dogs share common diseases caused by extra-intestinal pathogenic E. coli (ExPEC) (LeCuyer et al., 2018;Manges et al., 2019), of which urinary tract infections (UTIs) are by far the most common. Transmission of ExPEC between dogs and their owners has also been described (Damborg et al., 2009;Reeves et al., 2011;Damborg et al., 2018), and contact with dogs or dog feces is considered a risk factor for the acquisition of resistant ExPEC (Ukah et al., 2018). Nonetheless, due to frequent focus on AMR, the population structure of pathogenic E. coli that have not been studied for their resistance phenotype remains largely under-reported. A recent study showed the dominance of specific ExPEC lineages in dogs in the United States, such as the primarily dog-associated ST372 or the human-associated ST73, ST12, or ST127 (LeCuyer et al., 2018). Uropathogenic E. coli (UPEC), which are a sub-pathotype of ExPEC mostly found in human infections, have often been associated with four major ST complexes: ST14, ST69, ST73, and ST95 (Lau et al., 2008;Liu et al., 2015;Fibke et al., 2019). ST73 in particular may be an underestimated human UPEC because it is not often associated with ESC-R determinants (Manges et al., 2019). Furthermore, ESC-R E. coli show lower virulence scores, except for the iutA gene (LeCuyer et al., 2018;Zogg et al., 2018b). Interestingly, a reduced virulence load in multidrug-resistant E. coli, concomitant to an increased prevalence of virulence genes in susceptible E. coli belonging to the phylogroup B2, has also been reported in isolates from Germany and Switzerland (Wagner et al., 2014;Zogg et al., 2018b).
To characterize the E. coli lineages causing diseases in French dogs and to assess to what extent this population structure may include ExPEC and/or UPEC lineages causing diseases in humans, we investigated 618 non-duplicate clinical E. coli isolates collected from dogs throughout France in 2017. This was done independently of their resistance phenotype to correctly estimate the resistance and virulence load which can potentially be transmitted to (or from) humans.

Bacterial Isolates
Escherichia coli isolates (n = 618) were collected from four veterinary diagnostic laboratories in France between January and November 2017. The four laboratories were asked to provide all non-duplicate E. coli isolates causing diseases in dogs over a continuous 3-month period of time in 2017. Two laboratories were located in the Parisian suburbs, one in the Rhône département and one in the Alpes-Maritimes département. Three laboratories were referee laboratories (the two Parisian laboratories and the one from the Rhône département), receiving samples from all across France, such that the sampling of this study was considered representative of clinical E. coli isolates circulating in dogs in France. Most E. coli isolates were recovered from UTIs (n = 403, 65.2%) (Supplementary Table S1), but also unexpectedly in a few cases of diarrhea (n = 29, 4.7%). Considering the ubiquitous presence of E. coli, we cannot completely exclude that a few isolates included in this study were not pathogenic. However, maximum precautions were taken to avoid this bias, since samples were all taken by veterinarians suspecting infectious diseases and since polycontaminated samples (≥3 bacteria growing on one plate) were not processed. Bacterial isolation was performed on Eosin Methylene Blue agar (EMB; ThermoFisher Diagnostics, Dardilly, France), Brillance TM UTI agar (Oxoid), chromID R CPS R Elite agar (BioMérieux, Marcy l'Etoile, France) or blood agar, and identification was performed using MALDI-TOF mass spectrometry or API R gallery-based identification according to each veterinary laboratory's methodology. All isolates were sent to the ANSES laboratory in Lyon for further characterization.

Resistance Genes
PCRs were performed using specific primers for the detection of the ESBL genes bla CTX−M group 1 and group 9, and bla CTX−M−2 (Shibata et al., 2006;Dierikx et al., 2010). For all bla CTX−M group 1 and group 9 isolates, additional PCRs were performed to identify the exact bla CTX−M gene variant using the ISEcp1L1/P2D and MA1/MA2 primers, respectively. The bla CMY genes were detected using CF1/CF2 primers (Eckert et al., 2004). All positive amplicons were sequenced (Genewiz, London, United Kingdom).
Isolates carrying at least two of the following genes papA, sfaS-focG, afaD, kpsMTII and iutA were classified as ExPEC as described . E. coli isolates that possessed at least two ExPEC determinants and the additional haemolysin secretion protein D gene hlyD and the cytotoxic necrotizing factor cnf1 were classified as UPEC, as previously defined (Johnson, 2003).

Genetic and Molecular Typing of the Strains
Phylogenetic grouping of the E. coli isolates was performed using a published improved method (Clermont et al., 2013). The B2-O25b-ST131 clone of E. coli was detected using a published PCR-based assay (Clermont et al., 2009). Multilocus sequence typing (MLST) of all ESBL/pAmpC/carbapenemproducing isolates (n = 34) was performed according to the Achtman protocol described on the E. coli MLST website 1 . The complete MLST scheme was also performed on a randomly selected subset of 89 non-ESBL isolates belonging to the B2 phylogroup. For the remaining E. coli isolates belonging to the B2 phylogroup (n = 392), the fumC allele was sequenced as a reliable indicator of the MLST types ST372 (fumC103), ST73 (fumC24), ST141 (fumC52), and ST131 (fumC40). The purA and icd alleles were further sequenced for a subset of isolates presumably belonging to ST12 (fumC13 and purA10), ST127 (fumC14 and icd36), or ST961 (fumC13 and purA126). ST131 clones were screened using PCR (Clermont et al., 2009). For ST131-positive isolates, allelic variants of fimH were identified by PCR and sequencing. The presence of a C1-M27 cluster was determined using previously published PCRs targeting the cluster-specific prophage-like region (M27PP1) (Birgy et al., 2017).

Plasmid Characterization
Plasmids were typed using PCR-based replicon typing (PBRT) according to the published PBRT kit scheme using a commercial kit (Diatheva, Cartoceto, Italy) (Carattoli et al., 2005). Plasmids carrying the ESBL/pAmpC/carbapenme-resistance genes were detected using PFGE-S1 gels (6 V/cm for 20 h with an angle of 120 • at 14 • C with pulse times ranging from 1 to 30 s) followed by Southern blot using adequate probes (Dierikx et al., 2010). After electrical transfer of DNA to nylon membranes, Southern blots were performed according to the manufacturer's protocol (Roche Diagnostics, Meylan, Germany). After overnight hybridization at 37 • C, all membranes were treated with maleic acid and blocking solution for 45 min, then with purified anti-digoxigenin-AP Fab fragments, washing buffer and finally stained with NBT/BCIP. Plasmid co-localization was assessed by comparison between the bands corresponding to the resistance gene and those corresponding to the Inc type of the plasmid. The chromosomal location was identified by PFGE on I-Ceu1-digested DNA, followed by Southern blot hybridization using a 16S rDNA probe and probes specific to resistance genes.

Statistical Analysis
Fisher's exact test was used to determine whether proportions for one variable (antimicrobial susceptibility ( Table 1) or multidrugresistance (Table 2) or the distribution of virulence gene (Table 3) are different between phylogroups. The tests were considered significant when p-values were ≤0.05. When the Fisher's exact test was significant, an analysis and graphic visualization of the proportions and their 95% confidence intervals (CI) were performed. A proportion was considered significant when its CI did not overlap at least one other CI. Ceftiofur 10 (28.6) 4 (7.5)  Univariate logistic regression model was implemented to assess the associations between ESBL or AmpC or pathology (here UTI) status and the presence of virulence genes: the ESBL (or AmpC or pathology) status was the dependent variable and each model included a bimodal explanatory variable related to the studied virulence gene regardless of the presence (or not) of other genes. The p-values, odds ratio (OR) and 95% confidence intervals were calculated by a logistic regression model univariate ( Table 4). A multivariate model (taking into account the 12 virulence genes detected in this study as well as their interactions) could not be implemented since there were too many parameters to estimate in relation to the available data and the model would not converge. Associations were considered significant at p ≤ 0.05. R software version 3.6.0 (2019-04-26) was used for statistical analyses (R Core Team, 2019). When the Fisher's exact test was significant, an analysis and graphic visualization of the proportions and their 95% confidence intervals (CI) was performed. A proportion was considered significant when its CI did not overlap at least with one other CI. 2 VS is the Median virulence-associated gene Score calculated as the sum of virulence genes found positive using PCR. IQ: interquartile range of the virulence gene scores. ExPEC status was defined as the presence of ≥2 of the following genes: papA, focG, kpsMTII, sfaS, afaD, and iutA. UPEC status was defined as the presence of the hlyD and cnf1 genes among ExPEC isolates. The virulence genes cnf2, stx1, stx2, aggR, and cdt3 were absent from this population of isolates. The p-values, odds ratio and 95% confidence intervals (CI) were determined by univariate logistic regression.

Distribution of Phylogroups and Lineages of Pathogenic E. coli From Dogs
A total of 618 E. coli isolates were collected, of which 35 belonged to phylogroup A (5.7%), 53 to B1 (8.6%), 492 to B2 (79.6%) and 38 to phylogroup D (6.1%). Most isolates (65.2%) were associated with UTIs, but several other diseases were diagnosed (Supplementary Table S1). In all diseases, B2 was by far the predominant phylogroup, ranging from 58.6% (n = 17, digestive tract infections) to 90% (n = 54, otitis) of the isolates (Supplementary Table S1). According to this over-representation of the B2 phylogroup, complete MLST of a randomly chosen subset of 89 B2 E. coli revealed dominant STs, with ST372 (n = 16, 17.9%) and ST73 (n = 16, 17.9%) being the most frequent (Figure 1). ST12 (n = 9), ST141 (n = 7), and ST961 (n = 5) were also repeatedly identified, and ST131 was detected in two animals. Sixteen STs were identified once, three twice and two STs were detected three times. Finally, six STs were not found in the Achtman database. To further decipher the major lineages composing the 492 B2 E. coli group, partial MLST was performed: the fumC allele was sequenced on all 403 remaining isolates, and the purA and icd alleles were subsequently sequenced on a subset of isolates to refine the analysis. Over the total number of B2 E. coli, 102 belonged to ST372 (102/492, 20.7%) and 99 to ST73 (20.1%), slightly modifying the previously estimated proportions of both STs (both initially estimated at 17.9%). The other STs were found in smaller numbers, with 37 isolates belonging to ST141 (7.5%), 13 to ST12 (2.6%), 12 to ST127 (2.4%), 9 to ST131 (1.8%), and 6 to ST961 (1.2%). Among the nine ST131, eight showed FIGURE 1 | Phylogenetic proximity of 89 randomly chosen B2 isolates, as inferred using goeBURST (phyloviz). Each circle is a pie chart showing the pathologies from which the isolates were obtained. Digestive disease is colored in red, UTIs in dark blue, unknown pathology in light blue, skin infection in orange, otitis in light orange, reproduction disorder in green, respiratory disease in light green. The node size of the goeBURST is proportional to the number of isolates. Nodes are in log scale. Scaling factor = 2.
an unusual fimH variant, and the last one showed the fimH30 allelic variant.
ESC-resistance was identified in 33 isolates (5.3%), including 14 ESBL-and 19 AmpC-producing E. coli. All these latter isolates produced the CMY-2 enzyme and all ESBL-producing E. coli produced a CTX-M-type enzyme. CTX-M-15 was the most frequent (n = 10), followed by CTX-M-27 (n = 2), CTX-M-1 (n = 1) and CTX-M-3 (n = 1). The 19 CMY-2-producing E. coli presented very diverse genetic backgrounds, since 13 different STs were identified. Only E. coli ST12 and ST4608 were each detected in three different dogs. On the contrary, ST410 was identified in 7/14 of the ESBL-producing E. coli. Of note, one ST131 isolate producing CTX-M-27 belonged to the C1-M27 sub-clade. A single isolate was resistant to ertapenem (0.2%) and identified as an OXA-48 producer belonging to ST1485. All bla CMY−2 genes were carried by IncI1 plasmids. The bla CTX−M−27 and bla CTX−M−1 genes were located on IncF plasmids and the bla OXA−48 was carried by an IncL plasmid. In contrast, all bla CTX−M−15 genes were found on the bacterial chromosome.
The univariate logistic regression model showed that iutA was most frequently present in ESBL-producing isolates, but the iroN and kpsMTII genes were less frequently present in ESBL-or AmpC-positive isolates. The cnf1, focG, papA, sfaS, and hlyD genes mostly found in B2 were also associated with UTIs (Table 4).

DISCUSSION
In this study, we showed that the population structure of E. coli isolated from diseased dogs was clearly dominated by the B2 phylogroup, which accounted for 79.6% of all E. coli isolates. The predominance of the B2 phylogroup was most likely associated with the study design. Other studies on diseased animals with no AMR criteria have also reported high proportions of the B2 phylogroup (Liu et al., 2016;LeCuyer et al., 2018), and studies from healthy animals and/or focusing on ESC-resistant E. coli tended to report non-B2 phylogroups (Wagner et al., 2014). A recent exception to this pattern is a study on clinical ESC-R E. coli in dogs in France, which reported a high prevalence of multidrug-resistant B2 E. coli, but mainly linked to a high prevalence of ST131/H30 isolates (Melo et al., 2019). The low prevalence of ST131/H30 in our study confirms this clone as a human-associated pathogenic E. coli lineage, which occasionally transfers to companion animals.
In the present study, ST372 and ST73 were by far the dominant STs representing 20.7 and 20.1% of the B2 isolates, respectively. The seven ST372, ST73, ST141, ST12, ST127, ST131, and ST961 comprised 56.3% of all B2 isolates, which is very similar to what has been described in the United States in a study with comparable settings (LeCuyer et al., 2018). Similar findings in these two geographic areas strongly suggest a dog-specific distribution of pathogenic E. coli clones rather than the effect of regional factors. Of note, among the nine ST131 identified in this study, eight did not belong to the human-associated ST131 sublineages and carried unknown fimH variants. The last isolate produced a CTX-M-27 enzyme and belonged to the newly described C1-M27 clade. This clade first emerged in Japan before being identified in humans in other countries and continents. It has only recently been described in dogsconcomitantly in France and the United Kingdom -and this new occurrence of a C1-M27/ST131 isolate in a dog proves that this clade is not restricted to humans (Bortolami et al., 2019;Melo et al., 2019).
ST372, which has already been reported in dogs and much more sporadically in humans (Adler et al., 2012;Wagner et al., 2014), is thus emerging as a major dog-associated ST that occasionally carries resistance genes, as described here for ESBLs. Of note, ST372 was also recently reported as an OXA-48 producer (Melo et al., 2017). Its frequency in dogs most likely relies upon specific host adaptation abilities, allowing its spread in France and the United States despite the fact that the two countries differ in terms of antibiotic treatments, especially since France has set up a National Action Plan in 2012 to reduce antibiotic use (ANSES, 2018). Nonetheless, an Australian study on dogs reported the absence of ST372, which may be due to the strong border policies to avoid importation of foreign animals, but also most likely to the fact that only healthy and not clinical animals were sampled (Bourne et al., 2019). Apart from ST372, other frequent clones were human-associated STs, such as ST73, ST12, or ST127. Though usually considered a human-specific lineage, ST73 is most probably an ubiquitous urinary pathogen, because it has recurrently been described in animals, both in diseased and healthy contexts (Wagner et al., 2014;Liu et al., 2015;Bourne et al., 2019). Also, ST73 may well be under-reported in the literature because most papers have focused on ESC-R genes, which are less frequently found in this ST than in other ones such as ST131 or ST38 (Hertz et al., 2016;Manges, 2016;Van Hout et al., 2020). The prevalence of ESC-R E. coli in this study was 5.3% (33/618), which is fairly low compared with what has recently been described in the United Kingdom (18.4%), Switzerland (32.1%) or even in France in 2011 (18.5%) (Zogg et al., 2018a;Bortolami et al., 2019). This low prevalence may result from the strong policy measures on AMR in the vet sector since 2012 through the National Action Plan Ecoantibio, which resulted in a 65.5% decrease in the use of last-generation cephalosporins in cats and dogs between 2013 and 2017 (ANSES, 2018). Yet, despite these efforts, carbapenem-resistant E. coli isolates are emerging in dogs with no apparent risk factors. ST1485 OXA-48 E. coli found here is the second report of an OXA-48-producing E. coli after the occurrence of an OXA-48 ST372 E. coli in a healthy dog in 2015 (Melo et al., 2017). It questions the origin of these isolates since, unlike in other countries, the use of carbapenems is totally forbidden in France, even in emergency veterinary medicine. The emergence of OXA-48 in dogs has also been demonstrated in neighboring countries such as Germany (Pulss et al., 2018), so that monitoring of carbapenem-resistance in companion animals is now strongly recommended.
Among ESC-R isolates, the fluoroquinolone-resistant ST410 clone was found in 7/14 of the ESBL-producing E. coli. ST410 has recently been claimed to be an international high-risk clone, similar to ST131 in humans, for its capacity to combine resistance and virulence determinants as well as to cause hospital outbreaks (Roer et al., 2018). ST410 has been reported in dogs mostly in Europe (Germany, Switzerland, United Kingdom), but also in the United States; therefore, its emergence as a major ESBLproducing pathogen in dogs should be monitored (Falgenhauer et al., 2016;Timofte et al., 2016;LeCuyer et al., 2018;Zogg et al., 2018b;Bortolami et al., 2019). In this study, our samples originated from dogs that had no epidemiological link among each other. However, all animals visited veterinary hospitals working closely with the veterinary laboratories involved, so that it may be hypothesized that this clone is circulating in veterinary hospitals. However, this remains to be proven since other sources (such as a high prevalence in the community) cannot be excluded. The bla CTX−M−15 gene associated with ST410 was systematically carried on the bacterial chromosome, as already reported in Germany (Falgenhauer et al., 2016). Interestingly, the three remaining bla CTX−M−15 genes found on a single-locus variant of ST410, ST372, and ST648, respectively, were also found on the chromosome. Chromosomal insertion of bla CTX−M genes has long been considered as less frequent than their location on plasmids, but recent convergent studies suggest that it would probably be worth reconsidering this point. The factors driving chromosomal insertions are unclear. They may potentially involve bacterial fitness and certainly deserve further investigation. ST648 has also been described as a pandemic clone in animals, and has been recurrently associated either with ESBL or pAmpC enzymes (Ewers et al., 2014;Marques et al., 2018). In this study, it was identified twice as an ESBL carrier.
In agreement with previous studies, most VFs related to UPEC (cnf1, focG, papA, sfaS, and hlyD) were found in isolates from the B2 phylogroup. Our data confirm the low prevalence of the UPEC-associated VFs among resistant isolates, and especially in ESC-R isolates (Harwalkar et al., 2014;Hashemizadeh et al., 2017). iutA was the only exception, because this virulence gene was most frequently present in ESBL-producing E. coli, as previously suggested (LeCuyer et al., 2018). A strong correlation between the UPEC-associated VFs and the primarily dogassociated ST372 was demonstrated here. Among the VFs, cnf1, and hlyD were associated with both ST372 and ST73. Most VFs and serotypes identified in this study using the WGS analysis have been previously characterized in UPEC strains related to UTIs (Johnson and Russo, 2005;Yamamoto, 2007;Subashchandrabose and Mobley, 2015). All these results suggest a link between UPEC-associated VFs, B2 phylogroup, ST372/ST73 and UTIs in dogs.
In conclusion, to our best knowledge based on the largest pathogenic E. coli collection in dogs, our study showed that the population of E. coli in this animal species was dominated by ExPEC lineages. Although ST372 is most probably largely adapted to dogs, the human-associated ST73 appears more ubiquitous than previously thought. We also showed a clear discrepancy between resistance phenotypes and virulenceassociated genes, which clustered in the non-B2 and B2 isolates, respectively. Considering that ESC-R E. coli only account for 5.3% of the population, but B2 isolates represent 79.6% of this same population, transmission of virulent ExPEC/UPEC lineages may well be the issue, rather than the transmission of ESC-R isolates. The risk for humans of being infected or colonized by such isolates remains to be determined. Further studies on E. coli populations from various hosts, irrespective of their AMR status, are surely needed to refine the scientific opinions on risks associated with inter-sectorial transfers of pathogenic E. coli in a "One Health" approach.

ETHICS STATEMENT
Ethical review and approval was not required for the animal study because No ethical approval was needed since this study did not involve any experimentation on animals. Only bacteria collected from clinical samples were used in the frame of this study. Written informed consent for participation was not obtained from the owners because No consent was needed since the study involved clinical samples and no metadata were recorded.

AUTHOR CONTRIBUTIONS
CV, MH, and J-YM designed the experiments. SB, RP, H-JB, and VB supervised the sampling campaign. AD, CV, and MH performed the experiments. CV, MH, and J-YM analyzed the data. GC and CV performed statistical analyses. MH and CV drafted the manuscript. J-YM actively contributed to the manuscript's writing. All authors approved the final version of the manuscript.

FUNDING
This work was funded by the ANSES, the French Agency for Food Environmental and Occupational Health & Safety. The funding organization had no involvement in the study design, collection, analysis and interpretation of the data or in writing the article.