Genetic characterization of Shiga toxin-producing Escherichia coli O26:H11 strains isolated from animal, food, and clinical samples

The Shiga-toxin producing Escherichia coli (STEC) may cause serious illness in human. Here we analyze O26:H11 strains known to be among the most reported STEC strains causing human infections. Genetic characterization of strains isolated from animal, food, and clinical specimens in Argentina showed that most carried either stx1a or stx2a subtypes. Interestingly, stx2a-positive O26:H11 rarely isolated from cattle in other countries showed to be an important proportion of O26:H11 strains circulating in cattle and food in our region. Seventeen percent of the isolates harbored more than one gene associated with antimicrobial resistance. In addition to stx, all strains contained the virulence genes eae-β, tir, efa, iha, espB, cif, espA, espF, espJ, nleA, nleB, nleC, and iss; and all except one contained ehxA, espP, and cba genes. On the other hand, toxB and espI genes were exclusively observed in stx2-positive isolates, whereas katP was only found in stx1a-positive isolates. Our results show that O26:H11 STEC strains circulating in Argentina, including those isolated from humans, cattle, and meat products, present a high pathogenic potential, and evidence that cattle can be a reservoir of O26:H11 strains harboring stx2a.


INTRODUCTION
Shiga toxin-producing Escherichia coli (STEC) are E. coli strains that can cause human diseases, like gastrointestinal illnesses and hemolytic uremic syndrome (HUS). E. coli O157 has been the most commonly reported STEC serogroup since it was identified in the 1980s. However, the number of non-O157 STEC infections has increased substantially in the last years. The O26 is one of the few serogroups frequently reported in non-O157 cases (Rivas et al., 2010;EFSA, 2013;Gould et al., 2013;ISPCH, 2014). In particular, strains of serotype O26:H11/-have been associated with severe human diseases (Gerber et al., 2002;Zimmerhackl et al., 2010;Käppeli et al., 2011).
All STEC strains are characterized by the ability to produce Shiga toxins (Stx). The Stx family consists of two major types: Stx1 and Stx2, which can be further divided into several subtypes. A single STEC strain may carry one or more Shiga toxinencoding genes (stx) which are generally carried by prophages. Epidemiological studies indicate that different subtypes of stx are related to different clinical manifestations after STEC infection (Krüger and Lucchesi, 2015). Particularly, the stx 2a subtype is associated with highly virulent strains and HUS. Characterization of stx genotypes showed that STEC O26 strains isolated from patients can harbor stx 1a , stx 2a or both, however, strains harboring only stx 2a were significantly associated with HUS . Furthermore, Bielaszewska et al. (2013) identified a new highly virulent stx 2a -positive O26 clone as an emerging cause of HUS in Europe.
The production of Stx seems to be essential but not solely responsible for STEC pathogenicity. Other known and putative virulence factors are usually present in pathogenic STEC strains. Some of them, like the adhesin intimin encoded on a bacterial chromosomal pathogenicity island and the enterohemolysin EhxA encoded on a plasmid have been found in association with severe clinical disease in humans (Boerlin et al., 1999;Aldick et al., 2007).
It is a well-known fact that ruminants are the main reservoir of STEC strains (Naylor et al., 2005). Some studies suggest that domestic animals serve as reservoir for human pathogenic O26 STEC strains (Leomil et al., 2005). The O26 STEC strains also seem to be widely distributed in cattle, since those strains have been isolated from bovines belonging to different animal categories and production systems (Monaghan et al., 2011;Fernández et al., 2012;Paddock et al., 2014;Bonardi et al., 2015;Ison et al., 2015). Moreover, O26:H11 strains have been isolated from beef and dairy products (Bosilevac and Koohmaraie, 2011;Madic et al., 2011;Mohammed et al., 2014). Several studies from Europe and the United States report that O26 strains isolated from food and cattle generally carry stx 1 or both stx 1 and stx 2 genes; moreover, O26 strains harboring only stx 2a have been rarely isolated from cattle and food (Pearce et al., 2006;Geue et al., 2009;Bonanno et al., 2015;Ison et al., 2015).
Our aim was to characterize O26:H11 STEC strains isolated from cattle, food and humans to contribute to the global knowledge of virulence profiles and epidemiology of O26 strains circulating in Argentina.

Bacterial Strains and Growth Conditions
E. coli O26:H11 strains were selected from STEC collections in Argentina. The strains had been isolated between 1995 and 2013, from cattle, meat, human, and farm environment ( Table 1). Most of the strains had been previously characterized by PCR regarding the presence of stx 1 , stx 2 , eae, ehxA, and saa genes. Strains were stored at −70 • C with 20% (v/v) glycerol and when necessary grown in Luria Bertani broth at 37 • C overnight.  Geue et al., 2010). The array contained 87 probes targeting virulence genes and 102 probes targeting antimicrobial resistance associated genes. Visualization of hybridization was achieved using the ArrayMate instrument (CLONDIAG GmbH) and signals were analyzed automatically. The results were converted into a binary numerical format (1-present, 0-absent) and further analyzed using BioNumerics (Version 6.6; Applied Maths).
stx Subtyping Specific PCR reactions were performed to identify stx 1a , stx 1b , and stx 1c subtypes (Scheutz et al., 2012). The presence of stx 2b , stx 2e , stx 2f , and stx 2g subtypes was evaluated with the oligonucleotide microarray. This assay also detects stx 2a , stx 2c , and stx 2d subtypes but does not discriminate among them. Therefore, strains positive with the probe that detects stx 2a,c,d subtypes where further subtyped with specific PCR reactions (Scheutz et al., 2012).

Detection of eae-β Gene
Strains were tested for the presence of the eae-β subtype by specific PCR using the primer set SK1/LP4 (Oswald et al., 2000).
Allelic variants identified for each VNTR were sequenced with the same primers used to amplify those regions (Macrogen, Inc.). The sequences obtained were analyzed using Chromas 2.32 software (Technelysium Pty. Ltd.) and allele sequences of each VNTR were aligned with the software Clustal W (Larkin et al., 2007) in order to identify the number of tandem repeat units (TR). Alleles were named according to the number of TR. The absence of an amplification product was considered a null allele (−2).
The diversity index (D N ), based on Nei's marker diversity, was calculated for each locus using the formula D N = 1-(fra) 2 , where fra is the allelic frequency (Noller et al., 2003). The discriminatory power of the method was assessed using the Simpson diversity index (D S ) (Hunter and Gaston, 1988). Figure 1 shows the genes encoding virulence factors detected in the O26:H11 strains, clustered according to the Bionumerics analysis. Genes that were not found in any isolate are summarized in a footnote in the figure.

Genetic Characterization of Virulence Factors
The stx types identified with the array were in agreement with previous PCR results. Forty five percent of the isolates were stx 1positve, 52% stx 2 -positive, and one isolate (3%) was stx 1 and stx 2 -positive. By using the PCR subtyping protocol, all stx 1 genes corresponded to the stx 1a subtype and all stx 2 corresponded to the stx 2a subtype.
All tested isolates harbored the genes encoding for the adhesin Intimin (eae subtype β), and its receptor Tir (tir). Other adhesion related genes were also identified. The presence of efa1, espB, and iha genes was demonstrated in all isolates, whereas espP was found in all except the stx 1 and stx 2 -positive isolate. The toxB gene was only observed in stx 2 -positive isolates. The fasA, fedA, fedF, fim41a, nfaE, and saa genes were not found.
Among toxin-encoding genes, ehxA (encoding for a hemolysin) was present in 97% of the isolates and astA (encoding for EAST1, the enteroaggregative Escherichia coli heat-stable enterotoxin 1) was detected in 72% of the isolates. The cba and celB genes associated with colicin activity were identified in 97 and 17% of the isolates, respectively. Also, one bovine isolate was positive for the gene encoding colicin M (cma) and one isolate obtained from human with diarrhea was positive for mchB, mchC, mchF, and mcmA microcin genes.
Several type III secretion system components (translocators and effectors) were identified. All the strains were positive for cif, espA, espfO103, espJ, nleA, nleB, and nleC genes, and 76% of the isolates harbored tccP. The espI gene was only identified in stx 2positive strains. The etpD, a gene encoding for a type II secretion pathway-related protein, was not found.
The iss and hemL genes were identified in 100 and 97% of the isolates, respectively. The katP gene, encoding a catalase peroxidase, was detected in stx 1 -positive isolates only.

Antimicrobial Resistance
Five O26:H11 isolates (17%) carried more than two antimicrobial resistance genes, being bla TEM , strA-strB, and sul2 the most frequently detected ( Table 2). Genes that were not found in any of the isolates are summarized in a footnote in the table. The integron integrase encoding gene (intI1) was detected in isolate 21. The antimicrobial susceptibilities of the five isolates carrying resistance genes were determined using disk diffusion method. Four isolates were resistant to multiple antimicrobial agents ( Table 2), and most genetic resistances were phenotypically confirmed. The exception was isolate 23 (human isolate) that carried strA, strB, and sul2 genes but was susceptible for the tested antibiotics.
Simpson's index of diversity, calculated for the combined typing set, showed a value of D S = 0.96. Three main clusters were obtained: one group included the stx 2 -positive isolates and one stx 1 -positive isolate (isolate 17); a second group, most of the stx 1 -positive strains and the stx 1 and stx 2 -positive isolate (isolate 23) and a third group included three stx 1 -positive strains.

DISCUSSION
Genetic characterization showed that in addition to stx genes, O26:H11 strains harbored genes encoding other toxins, adhesins, and components related to the type III secretion system that contribute to their virulence. In particular, eae-β, tir, efa, iha, espB, cif, espA, espF, espJ, nleA, nleB, nleC, and iss genes were detected in all the isolates; and all except one contained ehxA, espP, and cba genes.
The analysis based on the presence/absence of genes associated with virulence identified three main clusters, one containing the stx 1a -positive isolates, another the stx 2a -positive isolates and a third one the stx 1a and stx 2a -positive isolate (Figure 1). The toxB, espI, and katP genes were differentially distributed between stx 1a -positive and stx 2a -positive groups. The toxB and espI genes, previously associated with severe disease (Mundy et al., 2004;Michelacci et al., 2014), were exclusively present in stx 2a -positive isolates. Conversely, the katP gene encoding for catalase-peroxidase was only detected in the stx 1apositive isolates.
Heterogeneity in gene composition of STEC O26:H11 plasmids has been detected by Zhang et al. (2000). Considering the plasmidic markers ehxA, katP, espP, and etpD, two major subgroups of stx 2a -positive strains were identified in Europe by Bielaszewska et al. (2013), one of them typical for the new virulent German clone. Noticeably, all the stx 2a -positive strains of our study present a distinct profile, positive for ehxA and espP, but negative for katP and etpD genes. Thus, stx 2a -positive strains of our Argentinian collection differ from the German clone, and also from the human-pathogenic strain recently identified in France , which is negative for all these plasmidic markers. The plasmid profile of stx 1a -positive strains of our study (positive for ehxA, katP, espP, and negative for etpD) is the same that Zweifel et al. (2013) identified in the human stx 1 -positive strains from patients with HUS or bloody diarrhea in Switzerland. On the other hand, the unique stx 1 and stx 2positive isolate was negative for all plasmidic genes (toxB, espP, katP, ehxA, etpD, saa, subAB, epeA) tested, suggesting the absence of a virulence plasmid like pO26-Vir, pO157, or pO113.
The MLVA profiles found in the present study do not coincide with any of the profiles identified by Brandal et al. (2012) among ovine and human O26:H11 isolates from Norway. Although loci CVN001, 004, and 007 were monomorphic in both studies, the alleles were different. The loci CVN002 and CVN003 presented a null allele for all tested samples in our study, in agreement with results obtained by Løbersli et al. (2012) for O26 isolates. Available epidemiological information suggests that isolates forming clusters with identical MLVA profiles were derived from the same source and constitute a single clone. However, one cluster contained human stx 1 -positive isolates not epidemiologically related. In addition, two isolates (isolates 20 and 21) from the same farm had the same virulence and MLVA profiles but only isolate 21 had antimicrobial resistance genes. Although there is a possibility that these isolates are from the same clone and that isolate 21 acquired antimicrobial genes, it is also possible that MLVA typing is not discriminatory enough to distinguish both isolates from each other. Interestingly, major groups obtained by MLVA analysis were similar to those obtained by the analysis of virulence factors.
Antimicrobial resistance genes were detected in five isolates obtained from two calves, two meat samples and one patient with diarrhea. Except for the human isolate, all other isolates showed phenotypic resistance profiles predicted by the corresponding genotypic profile. Among the antibiotics tested, all four isolates were resistant to ampicillin, amoxicillin/clavulanate, and tetracycline and also showed intermediate resistance to cephalotin. The two meat isolates (isolates 26 and 27) were also resistant to streptomycin, nalidixic acid, and trimethropimsulfamethoxazole. As we commented below, these two isolates also shared the same virulence and MLVA profiles. Considering the epidemiological link, as they were found in meat samples from the same meat processing plant, our results suggest that both isolates correspond to the same circulating strain.
Previous reports showed that integrons can be frequently detected in STEC strains and that most of the integrons can contain the aadA1 gene alone, or in association with the drf A1 gene (Morabito et al., 2002;Cergole-Novella et al., 2011). In our study, the intI1 gene was detected in isolate 21 which was also positive for aadA1 and drf A1 genes. Strikingly, this isolate carrying an integron and showing multiple resistances to antimicrobials was obtained from a newborn calf. Our results highlight the presence of multi-antimicrobial resistant STEC in cattle and meat in agreement with previous studies reporting the emergence and dissemination of antimicrobial resistance among STEC strains (Zhao et al., 2001;Li et al., 2011;Sasaki et al., 2012). Although antibiotic therapy is discouraged for treatment of STEC infections, the presence of antimicrobial resistant STEC strains in animals represent a risk for animal and human health. The genes coding for antimicrobial resistance could be transferred to other pathogens. Moreover, antimicrobial resistant STEC strains may have a selective advantage over other bacteria in intestines of animals under antibiotic treatments (Zhao et al., 2001). Taking into account that the same classes of antimicrobial agents are used both in humans and animals, joint efforts should be made to reduce the inappropriate use of antimicrobial agents in animals (Aidara-Kane, 2014).
In conclusion, we identified three different populations of native O26:H11 strains whose main differences were associated with genes present in mobile genetic elements. Although O26 strains harboring only stx 2a subtype have been rarely isolated from cattle and food in Europe and the United States (Pearce et al., 2006;Geue et al., 2009;Chase-Topping et al., 2012;Ison et al., 2015), stx 2a -positive strains have been an important proportion of O26:H11 strains circulating in farms in Argentina and showed to carry genes associated with high virulence, representing a potential risk for public health.