Molecular Characteristics of Extraintestinal Pathogenic E. coli (ExPEC), Uropathogenic E. coli (UPEC), and Multidrug Resistant E. coli Isolated from Healthy Dogs in Spain. Whole Genome Sequencing of Canine ST372 Isolates and Comparison with Human Isolates Causing Extraintestinal Infections

Under a one health perspective and the worldwide antimicrobial resistance concern, we investigated extraintestinal pathogenic Escherichia coli (ExPEC), uropathogenic E. coli (UPEC), and multidrug resistant (MDR) E. coli from 197 isolates recovered from healthy dogs in Spain between 2013 and 2017. A total of 91 (46.2%) isolates were molecularly classified as ExPEC and/or UPEC, including 50 clones, among which (i) four clones were dominant (B2-CH14-180-ST127, B2-CH52-14-ST141, B2-CH103-9-ST372 and F-CH4-58-ST648) and (ii) 15 had been identified among isolates causing extraintestinal infections in Spanish and French humans in 2015 and 2016. A total of 28 (14.2%) isolates were classified as MDR, associated with B1, D, and E phylogroups, and included 24 clones, of which eight had also been identified among the human clinical isolates. We selected 23 ST372 strains, 21 from healthy dogs, and two from human clinical isolates for whole genome sequencing and built an SNP-tree with these 23 genomes and 174 genomes (128 from canine strains and 46 from human strains) obtained from public databases. These 197 genomes were segregated into six clusters. Cluster 1 comprised 74.6% of the strain genomes, mostly composed of canine strain genomes (p < 0.00001). Clusters 4 and 6 also included canine strain genomes, while clusters 2, 3, and 5 were significantly associated with human strain genomes. Finding several common clones and clone-related serotypes in dogs and humans suggests a potentially bidirectional clone transfer that argues for the one health perspective.


Phylogenetic Grouping
Assignment to the main phylogroups (A, B1, B2, C, D, E, and F) was based on the protocol of Clermont et al. [44].

Serotyping
The determination of O and H antigens was carried out using the method previously described by Guinée et al. [45] with all available O (O1 to O181) and H (H1 to H56) antisera. Isolates that did not react with any antisera were classified as O non-typeable (ONT) or H non typeable (HNT) and those non motile were denoted as HNM.

Multilocus Sequence Typing (MLST)
The sequence types (STs) were established following the MLST scheme of Achtman by gene amplification and sequencing of the seven housekeeping genes (adk, fumC, gyrB, icd, mdh, purA, and recA) according to the protocol and primers specified at the E. coli MLST web site (http://mlst. warwick.ac.uk/mlst/dbs/Ecoli) [46].

Whole Genome Sequencing (WGS)
The WGS of 23 ST372 isolates from our LREC collection was performed under the protocol of the Genomics and Bioinformatics Core Facility (Centre for Biomedical Research of La Rioja) as it was described previously [55]. The assembly information of draft genomes, database sources and input parameters can be found in Table S1 (NCBI Bioproject accession PRJNA627579).
PLACNET webserver [56,57] was used for the genome reconstruction after which Prokka [58] was used to annotate the assembled genetic elements. Primary in silico analyses were carried out using the Center for Genomic Epidemiology (CGE) (http://www.genomicepidemiology.org/) services with home-made databases, the CGE databases, and other complementary databases to explore the resistance and virulence factors. Plasmid typing was complemented by subtyping relaxases with the method defined by Alvarado et al. [59] and integrative conjugative elements (ICEs) typing was complemented by in silico analyzing the ICE-harbouring contigs with ICEberg (ICEfinder and VRprofile) (information provided in Table S1). Additionally, the ICE-harbouring contigs were analyzed with Easyfig, a comparative genomic tool that allows for visualizing homologies and similarities between contigs using BLAST [60].
Besides, we performed a single nucleotide position (SNP) tree analysis of the 23 ST372 genomes sequenced in this study plus 174 ST372 full-genome references retrieved from NCBI bioproject and EnteroBase. The SNP-tree was done using the CSI Phylogeny 1.4 server from the CGE with J22 strain as reference (ID: GCA_009497315). After analyzing the SNP matrix, we took all the ST372 genomes from human strains plus some representative genomes from canine strains to make a tree visualization using EnteroBase [61]. The accession number of all the genomes included in this study can be found in Table S2.

Statistical Analysis
All the p values were calculated using Fisher's exact test, except for the comparison of the means that was performed using the one-way ANOVA test. p values < 0.05 were considered statistically significant.

Antimicrobial Resistance in the 197 Canine Isolates
In total, 28 (14.2%) of the 197 analyzed canine faecal E. coli isolates were classified as MDR. Multidrug resistance was significantly associated with isolates belonging to B1, D, and E phylogenetic groups (Table S5). Furthermore, only eight (28.6%) of MDR isolates showed the ExPEC and/or the UPEC status (Table S6).
In total, 10 of the 28 MDR isolates produced an ESBL enzyme: CTX-M-1 (four isolates), CTX-M-14 (four isolates), CTX-M-55 (one isolate) and SHV12 (one isolate). Besides, 10 other isolates produced a plasmid-mediated AmpC β-lactamase of CMY-2 type.    A total of 50 clones were identified among the 91 canine isolates classified as ExPEC and/or UPEC, with 11 of them including at least two isolates and only four, at least four isolates i.e., B2-CH14-180-ST127 (four isolates), B2-CH52-14-ST141 (four isolates), B2-CH103-9-ST372 (25 isolates), and F-CH4-58-ST648 (five isolates) ( Table 3). In recent studies conducted by our research group [3,21,62], we had identified, as indicated in Table 3 (Table 3). Among these 31 human isolates, 28 belonged to B2 phylogroup clones and three to F phylogroup clones identified among canine isolates. These B2 clones were distributed into five ST lineages, including four lineages currently dominant in humans (ST73, ST127, ST141, and ST1193,) and the lineage currently established as the dominant lineage in dogs, namely lineage ST372. In dogs, we found three clones in the lineage ST73 with the same serotype (O6:H1). The eight human isolates sharing this lineage with dogs were distributed into the same three clones and displayed serotype O6:H1. In dogs, we found two clones in the lineage ST127 displaying two serotypes with one (O6:HNM) of them present in the two clones. The four human isolates sharing this lineage with dogs were distributed into the same two clones but displayed the common serotype (O6:HNM). In dogs, we found two clones in the lineage ST141 with the same serotype (O2:H6). The 11 human isolates that shared this lineage with dogs were distributed into the same two clones and displayed the same serotype as human isolates. In dogs, we found one clone in lineage ST1193 with one serotype (O75:HNM). Three human isolates shared this clone and serotype with dogs. Concerning the lineage ST372, we found five clones in dogs with one of them including isolates displaying six serotypes. The two human isolates sharing the lineage ST372 with dogs belonged to this multiple-serotype clone and both displayed one of the six serotypes (O83:H31). Concerning the three F group human isolates, they belonged to one of the three F group clones (F-CH32-41-ST59) identified in dogs and showed the same serotype (O1:H7).
Among the 28 canine MDR isolates, we observed 24 different clones, of which nine had also been identified among the above cited 394 isolates causing infections in humans (Table 4) Bold highlights those canine clones and serotypes also detected among E. coli isolates causing extraintestinal infections in humans.
The main objectives were to get more insights into the E. coli ST372 lineage that appears as one of the most prevalent E coli lineages among the canine faeces E. coli populations and to elucidate if there is any relation between canine and human ST372 strains.
To infer the phylogeny, we performed an SNP-tree with 197 genomes of ST372 strains (23 from this study (labelled LREC strains) and 174 obtained from public databases) corresponding to 151 genomes from canine strains and 46 genomes from human strains. A total of 70% of these genomes corresponded to strains collected between 2017 and 2019 while the remaining 30% corresponded to strains isolated between 1995 and 2016. Regarding geographical distribution, 46 genomes (23.4%) were from strains collected in Europe and 143 (72.6%) from strains collected in North America.
The SNP analysis of the E. coli ST372 lineage revealed a wide and heterogeneous population, allowing us to describe six clusters. Figure 1 only includes 97 representative genomes (including the 23 LREC genomes sequenced in this study and the 46 genomes from human strains) of the 197 analyzed so that it is possible to visualize all the information.
The criterion established to define a cluster was that it should include genomes with less than 200 SNPs distance between them. An exception to this rule was the inclusion of the genome ECOL-19-VL-SD-MI-0018, with a maximum of 391 SNP distance, in cluster 4. Five genomes did not reach this criterion, having more than 400 SNP distance between them and could form five other clusters. However, we have included those genomes in only one category (undefined) to simplify the following analysis.
According to the phylogenetic tree built from the genome of the 197 strains, cluster 1 comprised 147 (74.6%) of the 197 analyzed genomes. This cluster was mostly composed of genomes of canine strains (138 genomes; 93.9%). Genomes of canine strains were also included in clusters 4 (nine genomes) and 6 (two genomes) while only human strain genomes were included in clusters 2 (28 genomes), 3 (three genomes) and 5 (three genomes) (Table 5). Thus, cluster 1 comprised significantly more canine strain genomes (p < 0.00001) while clusters 2 (p < 0.00001), 3 (p = 0.01209), and 5 (p = 0.01209) comprised significantly more human strain genomes. A total of 20 of the 21 genomes of the Spanish canine strains belonged to cluster 1, whereas, the genome of the remaining Spanish canine strain (LREC_356) belonged to cluster 4. The genomes of the Spanish and French human strains (LREC_341 and LREC_342) belonged to cluster 2. The criterion established to define a cluster was that it should include genomes with less than 200 SNPs distance between them. An exception to this rule was the inclusion of the genome ECOL-19-VL-SD-MI-0018, with a maximum of 391 SNP distance, in cluster 4. Five genomes did not reach this criterion, having more than 400 SNP distance between them and could form five other clusters.  Both clusters 1 and 2 were the most frequent clusters observed among the studied E. coli ST372 strains (canine and human) isolated in Europe and North America. However, cluster 1 was significantly associated with North America strains (p = 0.02476), while cluster 2 was especially associated with Europe strains (p = 0.01233) ( Table 6).  To compare the virulence profile of the 197 canine and human ST372 strains, we in silico investigated the presence of 32 VF-encoding genes in the 197 strains and defined their ExPEC and UPEC status. We also investigated the distribution of those VF-encoding genes according to the classification of the strains into the six defined clusters. Table 7 summarizes the results obtained from the mentioned analysis. Microbiological, geographical, and genomic data of each of the 197 studied strains are available in Table S2.
We also described 11 plasmids (four conjugative plasmids, six mobilizable plasmids and one plasmid with no relaxase suggesting that it is not mobilizable) which belonged to the following relaxase families (MOB) and incompatibility groups (Inc.): MOB P3 /IncX1 (n = 3); MOB P1 /nd (n = 2); MOB F12 /IncFII-pCD1 (n = 2); MOB F12 /IncFII-IncFIB (n = 1); MOB H11 /IncHI2 (n = 1); MOB Qu /ColRNAI (n = 1); nd/p0111 (n = 1). To predict plasmid transferability, we investigated the presence of mating pair formation (Mpf) system proteins. These proteins were present in all the previously described MOB F12 and MOB H11 conjugative plasmids. Furthermore, in silico analysis showed that these plasmids did not carry resistance or virulence encoding genes except for the cba and cma genes that were found in plasmid pLREC354_1 and a bla TEM gene found in pLREC346_1. Table 10 summarizes the MGE content of the 23 ST372 genomes.
We in silico investigated the presence of 189 VF-encoding genes, 87 antibiotic-resistance encoding genes (ARGs), and 18 types of point mutations (Table S8). Through this analysis, the 23 ST372 strains were shown with an UPEC status and harbouring a wide variety of VF-encoding genes, reaching an average number of 80. In contrast, these 23 ST372 strains were shown as carrying very few ARGs. However, genes encoding drug efflux were detected but only in the two human strain genomes (LREC_341 and LREC_342) that also harboured antibiotic-resistance encoding genes: bla TEM-1A , sul1, aadA1, dfrA1, and mdf(A). These results were in agreement with those previously obtained by conventional methods.

Discussion
To get more insights into the population structure of canine E coli, we investigated those harboured in the intestinal tract of 104 healthy Spanish dogs by using different approaches, knowing that the gut is the reservoir of the great majority of E. coli causing extraintestinal infections. The phylogenetic group, VF-encoding gene and antibiotic susceptibility analyses, showed that among the 197 canine faecal isolates obtained from the 104 dogs, 84 (42.6%) belonged to B2 phylogroup, 91 (46.2%), mostly B2 group isolates, were classified as ExPEC and/or UPEC, and 28 (14.2%), mostly non-B2 group isolates, as MDR. This strongly suggests that the intestinal tract of healthy dogs might be an important reservoir of ExPEC and/or UPEC isolates, and in a lesser extent, of MDR E. coli isolates. However, some studies that focused on antibiotic-resistant canine isolates suggested that dogs might also be an important reservoir for antibiotic-resistant strains [63][64][65][66][67][68][69][70][71][72][73][74][75][76][77][78][79][80][81], notably for those producing ESBLs or CMY-2 [64,66,67,70,72]. Although there was a low prevalence of MDR isolates among the 197 studied isolates, we found, as previously described that they produced ESBLs or CMY-2.
MLST assigned the 91 Spanish canine faecal isolates with an ExPEC and/or UPEC status to 34 STs. among which six were displayed by 67% of the 91 isolates: ST372 (31.9%), ST12 (9.9%), ST127 (8.8%), ST648 (6.6%), ST141 (5.5%), and ST73 (4.4%). Few studies have been carried out so far to characterize the ST structure of canine ExPEC and/or UPEC isolates. In the USA, LeCuyer et al. [11] analyzed 295 E. coli isolates from canine UTI. They found that ST372, which is uncommon among the human E. coli pathogens [3,12,62,82,83], was the predominant ST in canine UTI isolates (21.7%), and this was well ahead of the five other most frequent STs: ST12 (6.4%), ST73 (6.4%), ST127 (4.1%), ST131 (4.1%), and ST297 (3.7%). A total of 170 (57.4%) of these isolates met the criterion to be classified as ExPEC, and, except for ST297, the most prevalent STs were associated with ExPEC status. In France, Valat et al. [14] analyzed 618 canine E. coli isolates collected from diagnostic laboratories, including 403 (65.2%) from UTIs. B2 phylogroup was over-represented (79.6%) and positively associated with the presence of numerous VFs, including those defining the ExPEC status. MLST of a randomly chosen subset of 89 isolates belonging to B2 phylogroup revealed five dominant STs: ST372 (17.9%), ST73 (17.9%), ST12 (10.1%), ST141 (7.9%), and ST961 (5.6%). In Australia, Kidsley et al. [10] focused their study on the canine fluroquinolone-susceptible E. coli clinical isolates (n = 449) that were identified during a nation-wide survey of antibiotic resistance in Australian animals between January 2013 and January 2014. They found that these isolates mostly (n = 317; 71%) belonged to B2 phylogroup. By using the RAPD typing system, they found a distribution of the 317 B2 group isolates into 35 main clusters. To pursue their molecular investigation, they sequenced and analyzed the whole genome of 77 representatives of the B2 group fluoroquinolone-susceptible isolates. Thus, they found that the 77 sequenced isolates were assigned to 24 STs, among which four were dominant: ST372 (31%), ST73 (17%), ST12 (7%), and ST80 (7%). In sum, the present study and those previously published show that three STs (ST372, ST12, and ST73) are the dominant ST in healthy and infected dogs irrespective of the countries (the USA, France, Australia and Spain) and strain sources (clinical samples and faeces). Such a finding might argue for the "prevalence" theory with regard to UTI pathogenesis (most UTIs are opportunistic infections caused by bacteria that predominate in the faecal microbiota) in dogs. Nevertheless, the fact that the canine isolates belonging to the most dominant ST, either present in all studied countries (ST372, ST12, and ST73) or present in some studied countries (ST127 in the USA and Spain, and ST141 in France and Spain) were shown to harbour numerous VF-encoding genes might also argue for the "special pathogenicity" theory in dogs.
To get more insight into the potential link between the canine ExPEC and/or UPEC and the E. coli isolates causing extra-intestinal infections in humans, we determined which clone (defined by the association of phylogroup, clonotype and ST) and which serotypes characterized the 91 canine ExPEC and/or UPEC in order to compare them with human E. coli clinical isolates collected in 2015 and 2016 in Spain and France and characterized for these two traits [3,21,62]. This approach allowed us to found that among the 50 clones identified in the 91 Spanish canine ExPEC and/or UPEC isolates, 15 were present in the human collection accounting for 49 (18,8%) of the human isolates. However, only 31 of the 49 human ExPEC and/or UPEC isolates presented the same O:H serotype as the canine ones. By coupling clonal type and serotype for each E. coli ST lineage shared by dogs and humans, we observed various features about the distribution of the human isolates when the lineages included several clones and several clone-related serotype in dogs. This feature shows that it is difficult to make hypotheses about the relationship between canine and human isolates sharing a given clone-serotype couple in a given lineage without knowing the structure of the clone-serotype couples in the given lineage in humans. For example, we had found [3] that the human ST73 isolates were distributed into four clones, of which two here were identified in dogs. In human, three of the four clones comprised, each, isolates with different serotypes but one serotype (O6:H1) was exhibited by isolates distributed into the four clones. In dogs, the ST73 isolates exhibited only serotype O6:H1. This suggests that serotype might be an ecological niche marker, meaning, in this case, that isolates of the lineage ST73 exhibiting serotype O6:H1 are adapted to both dogs and humans. However, the Kidsley et al.'s study [10] in which a phylogenetic tree was built with the genome of ST73 strains from dogs, cats, and humans, seems to contradict this hypothesis. Indeed, the 13 studied Australian canine isolates of the lineage ST73 exhibited four serotypes among which serotype O6:H1 was exhibited by only one isolate that formed an animal-specific cluster (containing cat O6:H1 ST73 isolates) distinct from the four main clusters of human O6:H1 ST73 isolates. Nevertheless, the hypothesis that we made for serotype O6:H1 with regard to Spanish canine and human ST73 isolates could be made for serotype O2:H1 with regard to Australian canine and human ST73 isolates as this serotype was shared by clustered canine and human isolates. By extending the comparison of the structure of clone-serotype couples to the other human-specific-human ST lineages (ST127, ST141 and ST1193) shared by the Spanish studied dogs and humans, we observed that the clone-serotype couples shared by dogs and humans comprised mostly the serotype the most frequent in the human clones. This feature seems to indicate that serotype frequency might be a variable involved in the E. coli exchanges between dogs and humans. Concerning the dog-specific lineage ST372, we had found only one clone comprising to isolates in Spanish humans, while we found here five clones in the 29 Spanish dogs. Among the six serotypes exhibited by these 29 canine isolates, the serotype exhibited by the two human ST372 isolates corresponded to one (O18:H31) of the two dominant serotypes in dogs that was, on the other hand, exhibited by canine isolates belonging to three different clones. Thus, the suggestions that we made about the fact that serotype could be an ecological niche marker and that serotype frequency could shape the E coli exchange between dogs and humans seems to be able to be applied to the lineage ST 372.
Interestingly, concerning the 24 clones identified in the 28 canine MDR isolates, which were mostly non-B2-group isolates, we observed that if there were some clones (n = 9) shared by the Spanish human (35 of 394) and canine (10 of 197) isolates there was only one isolate that shared the same clone and the same serotype as one canine isolate.
To better understand the potential relationship between canine and human E. coli isolates with regard to the lineage ST372, we turned to the whole genome sequencing and analysis of 197 ST372 strains (151 from dogs and 46 from humans). The SNP analysis of the core genome of these 197 strains revealed an extensive phylogenetic diversity of the ST372 isolates that was segregated into six clusters. Cluster 1 comprised 91.4% of canine strains while cluster 2 comprised 60.9% of human strains. Cluster 2 was specific of human strains associated with serotypes O18:H31 and O45:H31, the latter serotype being exclusively found in human ST372 strains. Three other serotypes were the most prevalent serotype among strains belonging to cluster 1, including O4:H31 and O15:H31 associated with canine strains, and O83:H31 identified in similar proportion among canine and human strains. Overall, the WGS analysis suggests that canine strains of clone B2-CH103-9-ST372, belonging to cluster 1 and having serotype O83:H31 might cause extraintestinal infections in humans and dogs, as already suggested by the clone-serotype couple analysis, whereas strains of this clone belonging to cluster 2 and having serotypes O18:H31 and O45:H31 might cause only human extraintestinal infections. Molecular epidemiological studies on E. coli ST372 in human extra-intestinal infections are required to confirm these suggestions.
Furthermore, we localized ICEs in the chromosome of 22 of the ST372 sequenced genomes and confirmed that all ICEs belong to a yersiniabactin synthesis-associated ICE type (ICEKp1 family) with relaxase type MOB Q. In contrast, we found very few plasmids. Moreover, we found that the number of plasmids retrieved from the human ST372 strains was higher than that of plasmids found in the canine strains (four plasmids in the two human strains versus seven plasmids in the 21 canine strains). Interestingly, the genome of canine LREC_356 strain from cluster 4 carried two plasmids and was the canine strain genome the most similar to human strain genomes. Those plasmids were not rich with genes encoding of antibiotic resistance and virulence-factors. Nonetheless, a high number of virulence factor encoding-genes were found in the chromosome of the ST372 genomes and we hypothesized that the origin of the UPEC status of ST372 strains is due to the acquisition of ICEs harbouring the genes associated with this status. Although there is still limited knowledge about the origin of genomic islands, like ICEs or pathogenicity islands (PAIs), it has been speculated that they derive from the integration of plasmids or phages into the chromosome. Further, genomic research has shown that genomic islands have played a major role in the transformation of avirulent into virulent bacteria. Besides, most VFs of ExPEC are encoded by ICEs and PAIs [92][93][94][95].

Conclusions
The intestinal tract of healthy dogs appears as an important reservoir of ExPEC and/or UPEC, and, in a lesser extent, of MDR E. coli isolates. However, the canine MDR isolates could be a good reservoir of ESBLs and CMY-2 because most of them produce these enzymes. Among the canine isolates displaying an ExPEC and/or UPEC status, clone B2-CH103-9-ST372 was dominant. This canine clone and 14 others, also displaying an ExPEC and/or UPEC status, had been identified in isolates previously published as causing extraintestinal infections in human suggesting a zoonotic potential of these clones. WGS analysis suggests that canine strains of clone B2-CH103-9-ST372, belonging to cluster 1 and having serotype O83:H31 might cause extraintestinal infections in both humans and dogs, whereas those strains of this clone belonging to cluster 2 and serotypes O18:H31 and O45:H31 might cause only human infections. Taking into consideration that Kidsley et al. have recently characterized the phylogenetic relationship between canine, cat, and human isolates of the lineage ST73 [10], such studies are still required for the other ST lineages and clones that we showed in this study to be shared by canine and human isolates in order to clarify their potential role in infection occurrence in both dogs and humans.
Supplementary Materials: The following are available online at http://www.mdpi.com/2076-2607/8/11/1712/s1, Figure S1: Comparison of contigs harbouring integrative conjugative elements (ICEs) from 22 ST372 E. coli genomes, Table S1: Bioproject accession (PRJNA627579) and assembly genome information, Table S2: SNP matrix and VF-encoding genes of 197 ST372 genomes, Table S3: Prevalence of the phylogenetic groups in the 197 canine E. coli isolates, Table S4: Comparison of the distribution of the phylogenetic groups among the 197 canine isolates according to the strain ExPEC and UPEC status, Table S5: Comparison of the distribution of the phylogenetic groups among the 197 canine isolates according to the strain multidrug resistant (MDR) status, Table S6: Comparison of the strain ExPEC and UPEC status among canine multidrug resistant (MDR) and non-MDR isolates, Table S7: New sequence types observed in 18 canine E. coli isolates, Table S8: In silico determination of VF-encoding genes, antibiotic-resistance encoding genes (ARGs) and point mutations in 23 ST372 E. coli genomes. Funding: This study was supported by projects: PI16/01477 from Plan Estatal de I+D+I 2013-2016, Instituto de Salud Carlos III (ISCIII), Subdirección General de Evaluación y Fomento de la Investigación, Ministerio de Economía y Competitividad (Gobierno de España) and Fondo Europeo de Desarrollo Regional (FEDER); ED431C2017/57 from the Consellería de Cultura, Educación e Ordenación Universitaria, (Xunta de Galicia) and FEDER.