Genomic characterization of ESBL-producing Escherichia coli isolates belonging to a hybrid aEPEC/ExPEC pathotype O153:H10-A-ST10 eae-beta1 occurred in human diarrheagenic isolates, meat, poultry and wildlife

Different surveillance studies (2005-2015) on the presence of ESBL-producing E. coli in the northwest Spain revealed that eae-positive isolates of serotype O153:H10 were periodically detected in meat (of beef, chicken and pork), and also implicated in human diarrhea. This study aimed: i) to characterize the degree of relatedness between human and animal isolates; ii) to know if this was a geographically restricted or disseminated genetic lineage. Thirty-two isolates were conventionally typified as O153:H10-A-ST10 fimH54, fimAvMT78, traT and eae-beta1, being 21 of those CTX-M-32 or SHV-12 producers. PFGE comparison of their macrorestriction profiles showed high similarity (>85%). The plasmidome analysis revealed a stable combination of IncF (F2:A-:B-), IncI1 (STunknown) and IncX1 plasmid types, together with non-conjugative Col-like. Besides, the core genome investigation based on the cgMLST scheme from Enterobase, proved close relatedness between isolates of human and animal origin. Our results demonstrate that a hybrid MDR aEPEC/ExPEC of clonal group O153:H10-A-ST10 (CH11-54) would be playing a successful role in spreading ESBLs (CTX-M-32) in our region within different hosts, including wildlife. It would be potentially implicated in human diarrhea via food (meat) transmission. Importantly, we proved genomic evidence of a related hybrid aEPEC/ExPEC in other countries.


Introduction
Escherichia coli is a normal inhabitant of the human and animal intestinal tract. However, E. coli can also act as a pathogen in a broad range of conditions, from diarrheagenic diseases to extraintestinal infections. Unlike extraintestinal pathogenic E. coli (ExPEC), with no specific virulence determinants for each subtype, the diarrheagenic E. coli (DEC) categories are characteristically defined by certain pathotype-specific virulence markers [1,2]. Thus, enteropathogenic E. coli (EPEC) are carriers of the eae gene, as part of the pathogenicity island LEE, codifying a protein called intimin. The intimin is responsible for the intimate adherence of the bacteria to the enterocyte membranes and, eventually, for the attaching and effacing (AE) lesion of the brush-border microvilli [3]. The variable C-terminal-encoding sequence of eae defines more than 30 distinct intimin types and subtypes associated with tissue tropism [4,5]. EPEC are further classified as typical (tEPEC), when they carry an EPEC adherence factor (EAF) plasmid that encodes adherence mediated by the bundle forming pilus (BFP), while atypical EPEC (aEPEC) produce the AE lesion but do not express BFP [4,6]. Currently, aEPEC isolates are emerging enteropathogens detected worldwide and isolated from different niches (animal species, environment, and food samples), while the main reservoir of tEPEC isolates are humans [7,8].
Antimicrobial resistance is a serious global concern which involves the health care system, food production and environmental integrity [9]. In fact, it is assumed that antimicrobial drug use in the livestock sector plays an important role in the spread of extended-spectrum beta lactamases (ESBL)-producing E. coli thought the food chain to humans [10,11]. The genomic plasticity of E. coli is the consequence of the important role played by mobile genetic elements (MGEs) such as plasmids, bacteriophages, pathogenicity islands, transposons and insertion sequence elements in the evolution of the bacteria [12]. As a result, hybrid E. coli pathotypes unpredictably emerge, given the mobility of most of the genes encoding virulence and antimicrobial resistance (AMR) [12,13]. Since 2011, when a novel Shiga-toxin-producing E. coli (STEC) belonging to serotype O104:H4, with virulence features (VF) common to the enteroaggregative E. coli (EAggEC ), and CTX-M-15 producer was identified as the one involved in the large German outbreak [14], the concept of pathotype has been questioned. Currently, classical and new approaches, such as whole genome sequencing (WGS), are being used to enhance the understanding the evolution of this highly adaptable species [13,15].
From different in-house surveillance studies on the presence of ESBL-producing E. coli in the northwest of Spain (2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015), we noticed that eae-positive isolates of the serotype O153:H10 were periodically recovered from meat, wildlife, and avian farm environment. We also found them involved in human diarrhea. This study aimed: i) to characterize the degree of relatedness between human and animal isolates; ii) to know if this was a geographically restricted or disseminated genetic lineage.

Results
Thirty-two eae-positive E. coli (21 ESBL and 11 non-ESBL) belonging to the serotype O153:H10 constituted the collection of study. They were detected within different surveys in the period 2005 to 2015: 14 from human stools, eight from beef meat, seven from chicken meat, and one each of pork meat, wildlife (fox feces) and poultry farm environment (Table S1). Table 1 summarizes the main traits determined by conventional typing for the 32 isolates. All were positive for the intimin eae-beta1, but negative for bfpA gene, conforming the aEPEC pathotype. Other virulence genes defining verotoxigenic (VTEC), enteroinvasive (EIEC), enteroaggregative (EAggEC) or enterotoxigenic (ETEC) pathotypes were not detected; however, the fimAvMT78 gene, which is a virulence locus that codify a fimA variant MT78 of type 1 fimbriae [16] was present in all isolates. Besides, the traT gene that codifies an outer membrane protein implicated in serum survival [17] was also present in 17 of the isolates (Table 1). By means of the serotype, phylogroup, ST and clonotyping, the isolates were assigned to the clonal group O153:H10-A-ST10 (CH11-54).  The PFGE comparison of the XbaI-macrorrestriction profiles of the ESBL-producing aEPEC isolates revealed high similarity. Thus, all but one clustered with an identity >85% in the dendrogram shown in Figure 1. It is of note that three human clinical isolates, recovered in different years, clustered each with a fox (95.2% of similarity) and with two beef meat isolates (100% and 97.6% of similarity, respectively).

Whole genome sequencing (WGS)
To further investigate the virulence profile, resistome, plasmid content and relatedness, 17 representative aEPEC/ExPEC isolates of different origins were WG sequenced. The de novo assembled contigs were then typed in silico using the Enterobase tools (Table S2), as well as the Center for Genomic Epidemiology (CGE) databases (Table 2).
SerotypeFinder and Enterobase predictions corroborated O and H antigens, with the exception of LREC-120 and LREC-121, for which O153 was solved by serotyping. MLST (CGE and Enterobase), CHtyper and ClermonTyping also confirmed conventional data for ST (10), CH  and phylogroup (A) ( Table 2, Table S2). Additionally, the wgST, cgST, and rST of the genomes were determined using the schemes of Enterobase based on 25,002; 2,513 and 53 loci, respectively (Table  S2). WgMLST and cgMLST are powerful schemes with extreme and high resolution, respectively, which determined different STs for each of the 17 genomes analyzed, while rST (medium resolution) established the same ST (2021) for all genome but for LREC-127 (58738) ( Table S2).
VirulenceFinder corroborated the hybrid pathotype nature of the isolates, predicting in all genomes the eae gene (intimin) together with other components encoded in the LEE pathogenity island, as well as the increased serum survival gene iss recognized for its role in ExPEC virulence [18]. Besides, the astA gene, which encodes the heat-stable enterotoxin 1, was also present in all 17 isolates (Table 2).
ResFinder identified the genes associated to resistances observed in vitro (acquired resistances for beta-lactams, aminoglycosides, and point mutations for quinolones). Only, the blaCTX-M-32 was not predicted in silico for LREC-112 and LREC-119, but by conventional sequencing. Furthermore, ResFinder determined other acquired resistances which had not been tested in vitro, such as to phenicols and macrolides in all genomes, and to tetracyclines in 16 out of the 17 genomes (Table 2).
It is of note that we have detected this clonal group in subsequent and current studies on meat sampled in supermarkets of our city. In fact, we recovered aEPEC/ExPEC from 15 out of 100 poultry meat samples (2016-2017); from those, five were carriers of isolates belonging to the clonal group O153:H10-A-ST10, being one CTX-M-32 carrier (unpublished data). Recently, Zhang et al. [29] reported a 2.75% prevalence of aEPEC in retail foods at markets in the People's Republic of China, being the beta-1 intimin and the ST10 the second intimin and ST most prevalent within their isolates. According to the authors, the presence of virulent and MDR aEPEC in retail foods poses a potential threat to consumers.
Since the occurrence of the major outbreak of HUS in Europe caused in 2011 by an EAggEC/STEC O104:H4, other hybrid pathotypes have been recognized, and new are expected, either by novel assemblies of E. coli virulence determinants or through acquisition of new virulence genes from other bacterial species [13]. In Norway, Lindstedt et al. [30], expressed their concern regarding the detection of E. coli from human fecal content with a combination of intestinal and ExPEC virulence genes (IPEC/ExPEC) in a high frequency (64.3%). Several other studies have also identified STEC-and ETEC-associated virulence genes coexisting in E. coli isolates from humans, animals or environmental origin [31,32]. But probably one of the most outstanding is the EPEC/STEC O80:H2-ST301, emerged in France over the last few years and diffused within Europe, associated with invasive infections, which combines intestinal VFs (stx2d, eae-xi and ehxA genes) and extraintestinal genes characteristic of the plasmid pS88 [33,34]. To highlight in this O80 clone, the location of MDR and pS88 genes in the same plasmid; and in addition to this plasmid, another two (a carrier of ehxA gene and a cryptic one) were described within the isolates [33,34]. The clonal group described here poses also the threat of being MDR and characteristically associated with ESBL type CTX-M-32. CTX-M-32 enzyme is derived from CTX-M-1 by a single amino acid replacement, being probably an ancestor among CTX-M-1 and CTX-M-15 [35]. The blaCTX-M-32 gene was first described in 2004 in an Escherichia coli isolate in our Health Area (A Coruña, northwest Spain) [35]. Furthermore, it was described in three human isolates O25b:H4-ST131 ibeA-positive of our region, as early as in 2008 [11]. Of the 2,427 E. coli bloodstream isolates recovered in the hospital of our city (HULA) in the period 2000-2011, 96 were positive for ESBL production, from which 4.2% were CTX-M-32 and 4.2% SHV-12 [36]. The same prevalence was observed in this hospital in 2015 (unpublished data). The in silico analysis of 17 representative genomes O153:H10-A-ST10 corroborated the main traits determined by conventional typing. In a recent study, we had proved the good correlation and usefulness of SerotypeFinder or Enterobase predictions [22,37]. Here, only the serotype of two genomes could not be predicted in silico, probably due to the limitation of the assembly based on Illumina short reads [38]. MLST, CHTyper from CGE and Enterobase also confirmed conventional results. Like in the previous study, we found that VirulenceFinder properly identifies E. coli pathotypes (hybrid in this case), although based on different traits for the ExPEC pathotype. Thus, this clonal group O153:H10-A-ST10 typically carries the locus that codify a fimA variant MT78 of type 1 fimbriae [16] and the traT gene for an outer membrane protein implicated in serum survival [17]. Both VFs are not included in the VirulenceFinder scheme, and so they were not predicted. On the contrary, CGE tool identified in all genomes the increased serum survival gene iss, recognized for its role in ExPEC virulence [18], which was not determined by PCR. This is because CGE database predicts 14 variants of the iss gene [39], including the one described in E. coli IAI1 (CU928160), and harbored by the O153:H10-A-ST10 genomes. Our specific PCR detects the plasmid-borne iss allele (designated type 1), which is highly prevalent among avian pathogenic E. coli and neonatal meningitis-associated E. coli isolates but not among uropathogenic E. coli isolates [18]. The phenotypic AMR determined in vitro correlated with the results based on ResFinder databases, with the exception of blaCTX-M-32 not predicted in two genomes, but solved by conventional sequencing. Based on this and previous studies [37,40], we consider both conventional and genomic-based analysis complementary for a better understanding and characterization of emerging isolates.
An interesting trait of our isolates was the concomitant presence of IncF (F2:A-:B-), IncI1 (STunknown) and IncX1, together with non-conjugative Col156-like plasmids. Although carriage of plasmids means a fitness cost on the hosts [41], different studies support the hypothesis that interference between conjugative plasmids may reduce fitness costs by decreasing the efficiency of transfer. However, the mechanisms of such inhibitory systems need further investigation [42]. On the other hand, small plasmids was shown to increase its stability in cells containing big plasmids [41].
Another objective in this study was to know if this was a restricted genetic lineage. For this purpose, we searched related genomes uploaded in Enterobase based on the HierCC Cluster ID. As a result, we found a hybrid aEPEC/ExPEC pathotype A-ST10 eae-beta1 within its database associated to five human, one avian, and one unknown isolates (Table S3). Of note, the two human isolates (Code Name: 853984 and 866428) from United Kingdom, which clustered with the 17 Spanish genomes in the HC100 HierCC group (37600) (Table S3, Figure S1). The in silico analysis of these two genomes showed they belonged to the clonal group O153:H10-A-ST10 CH11-54 eae-beta1, were MDR carried similar virulence traits (conforming hybrid aEPEC/ExPEC pathotype), and plasmid combination: IncF (F2:A-:B-), IncX1, Col156-like (Table S6). To highlight that six of the seven genomes were carriers of IncF (F2:A-:B-) and Col156-like plasmids (Table S6). As above suggested, it would be necessary further investigation on the interaction between these plasmids and other mobile genetic elements affecting their transmission and persistence, as well as their role in the maintenance/acquisition of resistance genes.
In summary, our results demonstrate that a hybrid MDR aEPEC/ExPEC of clonal group O153:H10-A-ST10 (CH11-54) would be playing a successful role in spreading ESBLs (CTX-M-32) in our region within different hosts, including wildlife. It would be potentially implicated in human diarrhea via food (meat) transmission due to the genomic relatedness of isolates. Importantly, we proved the presence a related hybrid aEPEC/ExPEC in other countries.

E. coli collection
During the period of 2005 to 2015, different surveillance studies performed at the Reference Laboratory of Escherichia coli (LREC), in Lugo, Spain, aimed the detection of ESBL-producing E. coli within different sources of our region. These studies included samples from chicken, beef and pork meat, as well as poultry farm environment and wildlife. Briefly, the confluent growth of the MacConkey Lactose plates from each sample was screened by PCR for the presence of specific bla genes using the TEM, CIT, SHV, CTX-M-1 and CTX-M-9 group-specific primers [43]. Then, up to 10 individual colonies from positive plates were re-analyzed. Those confirmed for the bla genes were further characterized by PCR for the presence of VF eae, stx1, stx2, ipaH, pcDV432, eltA, estA or estB associated with the main intestinal pathotypes (enteropathogenic, verotoxigenic, enteroinvasive, enteroaggregative and enterotoxigenic) of E. coli. Likewise, specific extraintestinal VF were tested: fimH, fimAvMT78, papC, sfa/focDE, afa/draBC, cnf1, cdtB, sat, hlyA, iucD, iroN, kpsM II (establishing neuC-K1, K2 and K5 variants), kpsM III, cvaC, iss, traT, ibeA, malX, usp and tsh (Table S7, Table S8, Table S9).
On the other hand, human diarrheanic E. coli isolates, mainly from the Hospital Universitario Lucus Augusti (HULA) of our city (Lugo, northwest Spain), were routinely analyzed in our laboratory for intestinal VF, and those positive, complementary analysed for extraintestinal traces and ESBL genes, as described in the preceding paragraph.
All isolates were serotyped using the method previously described by Guinée et al. [44] employing O1 to O185 and H1 to H56 antisera. As a result, 32 eae-positive E. coli (21 ESBL and 11 non-ESBL) belonging to the serotype O153:H10 constituted the collection of study (Table S1).

Phylogenetic assignment and PFGE comparison
Phylogroup and ST assignment was performed following the Clermont et al. [46] and Achtman MLST [47] schemes´, respectively. The clonotyping was based on the internal 469-nucleotide (nt) and 489-nt sequence of the fumC and fimH genes, respectively, to define the CH type [48]. The molecular similarity within the collection was established comparing the XbaI-PFGE profiles of the isolates obtained following the PulseNet protocol, and imported into BioNumerics (Applied Maths, St-Martens-Latern Belgium) to perform a dendrogram with the UPGMA algorithm based on the Dice similarity coefficient and applying 1% of tolerance in the band position.

Genome sequencing, assembly and analysis
DNA from 17 isolates was extracted with the QIAamp 96 DNA Qiacube HT kit (Qiagen, Hilden, Germany) and libraries were prepared using the Nextera XT kit (Illumina). Pooled libraries were denatured following the Illumina protocol and 600 μl (approx. 20 pM) were loaded onto a MisSeq V2 -500 cycle cartridge (Illumina) and sequenced on a MiSeq to produces fastq files. Raw reads were uploaded and automatically assembled in Enterobase using SPAdes Genome Assembler v 3.5. with a contig threshold of minimum 200 nucleotides. Subsequently, the de novo assembled contigs were MLST (7 gene Achtman ST scheme, whole genome MLST, core genome MLST and ribosomal MLST) and serotyped in silico using Enterobase typing tools [49]. The raw reads were also analyzed using the CGE databases: SerotypeFinder , MLSTtyper , CHtyper , PlasmidFinder, ResFinder, and VirulenceFinder [50][51][52][53][54]. For genomic relatedness comparison, we used different approaches based on the cgMLST of Enterobase. Thus, a MSTree was inferred using the MSTree V2 algorithm and the asymmetric distance matrix based on the cgMLST scheme from Enterobase. This cgMLST scheme consists of 2,513 genes present in over 98% of 3,457 genomes, which represented most of the diversity in Enterobase https://enterobase.readthedocs.io/en/latest/pipelines/escherichia-statistics.html.
We also investigated the HierCC designations for our collection and other related genomes of Enterobase within each cluster group [49,55]. The SNP tree was also built in Enterobase, where all assemblies were aligned against LREC-113 using Last [56], and SNPs from these alignments were filtered to remove regions with low base qualities or ambiguous alignment. Specifically, any sites with low base qualities (Q < 10) or sites which could not be aligned unambiguously (ambiguity of alignment ≥ 0.1, as reported by Last) were excluded. Additionally sites were removed if disperse repetitive regions were aligned with ≥ 95% identities and longer than ≥ 100 bps according to nucleotide BLAST; or they were part of tandem repeats that were identified by TRF [57]; or within CRISPR regions, which were identified by PILER-CR [58]. After removing repetitive regions, all core SNPs were then called in the core genomic regions that were conserved in ≥ 90% of the genomes.

Supplementary Materials:
The following are available online at www.mdpi.com/xxx/s1, Table S1: Thirty-two isolates included in the study (in red) from our own collections, Table S2: Assembly data from Enterobase of the 17 O153:H10-A-ST10 genomes sequenced using Illumina NextSeq technology, Table S3: HierCC designations from Enterobase for the 17 Spanish collection and other 7 related genomes within each cluster group. SNPs of the core genomic regions, Table S4: Number of human stool samples analyzed and positive for aEPEC O153, Table S5: Twenty-three aEPEC O153 human isolates recovered in the period 2006-2012, Table S6: in silico characterization of seven E. coli related genomes from Enterobase using CGE databases, Table S7: Targets and primers associated with diarrheagenic pathotypes of E. coli, Table S8: Targets and primers associated with extraintestinal pathotypes of E. coli, Table S9: Detection and sequencing of blaTEM, blaSHV and blaCTX-M genes, Figure S1: GrapeTree inferred using the NINJA NJ algorithm and based on the cgMLST V1 + HierCC V1 scheme from Enterobase. Funding: This study was supported by projects: AGL2013-47852-R from the Ministerio de Economía y Competitividad (MINECO, Spain) and Fondo Europeo de Desarrollo Regional (FEDER); AGL2016-79343-R from the Agencia Estatal de Investigación (AEI, Spain) and FEDER; 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 and FEDER; and ED431C2017/57 from the Consellería de Cultura, Educación e Ordenación Universitaria of Xunta de Galicia and FEDER.