Introduction

Escherichia coli is a common inhabitant of the gastro-intestinal tract of humans and animals [1]. In particular, E. coli is typically the most common facultative anaerobe in the lower intestine of mammals and its presence in the environment is usually considered to reflect fecal contamination [1, 2]. The E. coli population is multiclonal and fluctuates in its predominance depending on diet, exposure to antibiotics or interactions with the host endogenous microbiota [1].

The intestinal microbiota predominantly comprises strict anaerobic bacteria, especially in the colon. E. coli exists in a symbiotic relationship with strict anaerobes: E. coli ferments monosaccharides generated by the degradation of polysaccharides by anaerobes ( E. coli being unable to synthesize the necessary hydrolase enzymes) and in turn, E. coli is able to consume oxygen and therefore to favor the strict anaerobe multiplication by creating a more anaerobic environment [2, 3]. Similarly, the host- E. coli relationship is mutualistic: the intestinal environment promotes efficient E. coli survival and multiplication and in turn, the E. coli population produces vitamins K and B12, which are required by mammalian hosts, and competitively excludes pathogens from the host intestinal tract [2]. E. coli strains are able to colonize various locations in the mammalian gastro-intestinal tract, but they are mainly found on the mucus layer used by E. coli as an essential nutritional source [4]. Successful colonization of the gastro-intestinal tract by E. coli depends upon several factors: competition for nutrients with the autochthonous microbiota, production of adhesins to bring the bacteria closer to the epithelia, penetration of the mucus layer, rapid growth and biofilm formation ability [1, 2, 4]. If E. coli growth does not exceed the turnover rate of the mucus layer, the bacterial cells are sloughed off into the intestine lumen and then eliminated in the feces [4]. Therefore, E. coli must display metabolic flexibility and grow in biofilm in order to succeed in this very competitive biotope [4].

Although considered as commensal in the mammalian gut, E. coli also causes a broad range of intestinal or extra-intestinal diseases due to the acquisition of mobile genetic elements encoding virulence factors. Among pathogenic E. coli , STEC is the major food-borne pathogen responsible for hemorrhagic colitis and hemolytic uremic syndrome [5]. In particular, a STEC strain subgroup EHEC belonging mostly to the serotype O157:H7 is responsible for serious public health concern and financial burden [5]. STEC strains are mainly transmitted to humans through contaminated meat or unpasteurized milk consumption [6]. It is of interest to note that healthy ruminants, mainly cattle, are the principal reservoir for E. coli O157:H7 strains, but cattle lack the Shiga-toxin vascular receptor, which explains why they are Shiga-toxin tolerant [6].

The cost of whole genome sequencing has decreased drastically and it is now possible to sequence a large number of isolates and use bioinformatic approaches to extract strain relatedness and gene carriage data. E. coli strains involved in human infections have been extensively studied and many whole genome sequences of E. coli associated with human illness are now available, allowing exploration of pathogenicity processes and identification of virulence factors. Due to cattle STEC dissemination, a significant number of whole genomes of E. coli O157:H7 strains isolated from bovine have also been sequenced. While previous genome sequencing efforts with commensal intestinal E. coli have focused on human strains [7,8,9], such data are scarce concerning commensal E. coli strains isolated from the bovine gastro-intestinal tract. It would be valuable to have recent and reliable genomic data on bovine commensal strains to be used as reference genomes.

In this study, we report the draft genome sequence and preliminary functional annotation of the commensal E. coli strain BG1 isolated from the digestive tract of a cow. The strain BG1 has been previously included in studies concerning the adaptation of pathogenic and commensal E. coli strains in the bovine gastro-intestinal tract [10, 11]. This study aimed to characterize the genomic features of the BG1 strain in order to provide information for future genomic scale (whole genome) comparative analyses. The organism is not part of a larger genomic survey project.

Organism information

Classification and features

As described for the genus Escherichia , E. coli BG1 is a Gram-negative, rod-shaped bacterium belonging to the Enterobacteriaceae family (Table 1). E. coli is a facultative anaerobe that is motile by means of flagella (Fig. 1). E. coli strains are typically able to grow over a wide temperature range (15–48 °C) with optimum growth from 37 to 42 °C and within a pH range of 5.5–8.0 (the best growth occurs at pH 7) [1] (Table 1). Like typical members of the E. coli species, the commensal strain BG1 utilizes D-glucose, D-mannitol, L-rhamnose, D-saccharose, D-melibiose and L-arabinose. Unlike most pathogenic O157:H7 EHEC strains, the strain BG1 is able to use sorbitol as a carbon source. In addition, E. coli BG1 is positive for arginine dihydrolase, ornithine decarboxylase, β-galactosidase and indole production.

Table 1 Classification and general features of E. coli BG1 [58]
Full size table
Fig. 1
figure 1

Transmission electron micrograph of E. coli BG1. The strain BG1 is a rod-shaped bacteria with a length of 1.5–2 μm and a diameter of 1 μm. It moves via peritrichous flagella. The magnification rate is 20,000×. The scale bar indicates 500 nm.

Full size image

In silico serotyping using SerotypeFinder (version 1.1) [12] revealed that E. coli BG1 belongs to the serotype O159:H21. The whole genome of E. coli BG1 lacked all the genes encoding antimicrobial resistance screened using ResFinder (version 2.1) [13]. E. coli strains can be divided into different phylogroups (A, B1, B2, D and E) commonly used to investigate the evolution and diversity of E. coli strains [14]. Phylogrouping was performed in silico using the quadruplex method described by Clermont et al. [14] and the primersearch program from the EMBOSS open software suite [15]. E. coli BG1 belongs to the phylogroup B1, which is commonly distributed among both bovine commensal and human pathogenic E. coli strains [16, 17].

A whole genome phylogenetic analysis based on single nucleotide polymorphism (SNP) differences in E. coli BG1, bovine and human commensal E. coli strains, bovine pathogenic E. coli strains and bovine O157:H7 STEC strains (Additional file 1: Table S1) was conducted using CSI Phylogeny (version 1.4) [18]. Published E. coli genomes representing different E. coli pathotypes were selected for genomic comparison (Additional file 1: Table S1). In addition, two reference E. coli strains, one of which is the E. coli type strain (NCTC9001 T), were also included in this study. As shown in Fig. 2, the bacterial strains were clustered according to the phylogroup classification: BG1 was clustered with commensal and pathogenic E. coli strains belonging to phylogroup B1 (EHEC, STEC, ETEC, EAEC, APEC and E. coli responsible for postpartum metritis in dairy cows). The closest relative strains to BG1 were E. coli K71 isolated from the environment of a cow shed and E. coli W26 isolated from bovine feces, both of which belong to the phylogroup B1 (Fig. 2). In contrast, BG1 was more distantly clustered to pathogenic bovine and human E. coli strains (Fig. 2). However, E. coli KCJ852 (phylogroup B1), which is responsible for metritis, was more closely clustered to BG1 than the P4 and VL2732 strains associated with bovine mastitis (phylogroup A) (Fig. 2). It is of interest to note that i) the bovine E. coli strains of commensal origin (BG1, K71 and W26) were distantly related to bovine STEC O157:H7 strains (phylogroup E) and ii) the SNP-based phylogeny analysis failed to cluster the commensal E. coli strains according to their human or animal origin.

Fig. 2
figure 2

Phylogenetic tree highlighting the position of E. coli BG1 relative to other E. coli strains. The whole genome SNP based phylogeny was established with CSI phylogeny 1.4 [28] using the genome of K71 as a reference and standard input parameters. The tree was midpoint rooted and plotted using Seaview (version 4.6.1) [56]. Each strain is identified as H (Human), B (Bovine), A (Avian), F (Food) or K12 (Laboratory strain), and its clinical or non-pathogenic (NP) characteristic is specified.

Full size image

Genome sequencing information

Genome project history

Bovine commensal E. coli strains are poorly documented. Therefore, the E. coli BG1 strain was selected for genome sequencing to provide valuable genetic information for future genomic scale (whole genome) comparative analysis. E. coli BG1 has been used as a reference strain in studies related to carbon and nitrogen nutrition of E. coli strains in the bovine gastro-intestinal tract [10, 11]. The strain BG1 was isolated from the small intestine content of a cow at the slaughterhouse in January 2009. The animal was raised and slaughtered in accordance with the guidelines of the local ethics committee and current INRA (National Institute for Agricultural Research) ethical guidelines for animal welfare (Slaughterhouse Permit number: 63,345,001). The bovine intestinal samples were collected after the slaughter of animals required for experiments specifically approved by the “Comité d’éthique en matière d’expérimentation animale en Auvergne” (Permit number: CE22-08) in the experimental slaughterhouse of the “Herbipole”, INRA Saint-Genès-Champanelle, France. The Whole Genome Shotgun project was deposited at DDBJ/ENA/GenBank under the accession MOAH00000000 (Oct 31, 2016). A summary of the sequencing project information is provided in Table 2.

Table 2 Genome sequencing project information for E. coli BG1
Full size table

Growth conditions and genomic DNA preparation

E. coli BG1 was inoculated in Luria-Bertani broth from a single colony and incubated at 37 °C with shaking (200 rpm) to early stationary phase. The bacterial suspension was then centrifuged (10,000 g for 15 min) and the total DNA was extracted from the bacterial pellet using the DNeasy Blood and Tissue Kit following the manufacturer’s recommendations (Qiagen). DNA was quantified using a Nanodrop spectrophotometer and DNA integrity was electrophoretically verified by ethidium bromide staining.

Genome sequencing and assembly

Whole genome sequencing was performed at the GeT-PlaGe core facility (INRA Toulouse, France). DNA-seq libraries were prepared according to Illumina’s protocols using the Illumina TruSeq Nano DNA LT Library Prep Kit. Briefly, DNA was fragmented by sonication using a Covaris M220 and adapters were ligated to be sequenced. Eight cycles of PCR were applied to amplify libraries. Library quality was assessed using the Agilent Bioanalyzer and libraries were quantified by qPCR using the Kapa Library Quantification Kit. DNA-seq experiments were performed on an Illumina MiSeq using a paired-end read length of 2 × 250 bp with the Illumina MiSeq Reagent Kits v2. The raw reads were stored in ng6 [19] and quality was checked using fastqc [20]. They were assembled with SPAdes (version 3.1.1) [21] using standard parameters.

Genome annotation

The assembled contigs were annotated with Prokka (version 1.10) [22] using standard parameters. Predicted genes were also assigned to functional categories of Clusters of Orthologous Groups (COGs) of proteins using blastp against the NCBI COG 2014 database [23]. Additional gene features were predicted using TMHMM Server 2.0 [24], SignalP Server (version 4.1) [25], CRISPRfinder (last update 2016–09-01) [26] and ISsaga (version 2.0) [27]. PHASTER [28] was then used to identify prophage regions in the BG1 genome. A prophage region was considered to be intact if the associated completeness score was above 90, questionable if the score was between 70 and 90 and incomplete if the score was less than 70 [28].

Genome properties

The genome of E. coli BG1 consists of 4,782,107 bp with no discernible plasmid (no match retrieved with PlasmidFinder version 1.3 [29]), and a G + C content of 50.7%. The genome has been assembled into 84 contigs. Of the 4562 predicted genes, 4465 coded for protein and 97 were RNA-related (including eight 5S rRNA genes, suggesting the presence of 8 rRNA operons, and 86 tRNA genes). In addition, 22 pseudo genes were identified. Among the 4465 protein coding genes, 3831 (85.8%) had an assigned function while the 634 remaining genes (14.2%) encoded proteins annotated as hypothetical or unknown. In addition, the BG1 genome contained 38 predicted insertion sequences (ISs), 4 intact and 1 questionable prophage regions, and 2 CRISPR elements suggesting possible genetic crosstalk, such as horizontal gene transfer among the E. coli population. The genome properties are presented in Table 3. The distribution of genes into COGs functional categories is summarized in Table 4.

Table 3 Genome statistics
Full size table
Table 4 Number of genes associated with general COGs functional categories
Full size table

Extended insights

Genome repertoire comparison

It is admitted that bacterial genome sequences show significant diversity due to horizontal gene transfers, gene loss and other genomic rearrangements [1]. In this report, characteristics of whole genome datasets of a selection of E. coli strains were compared with those of E. coli BG1 (Table 5). Our main objective was to compare the genome of BG1 with that of bovine (K71 and W26) and human (SE15 and Nissle) commensal E. coli strains, but we also included a bovine pathogenic strain (VL2732) and a human EHEC pathogen (Sakai), as the bovine intestine is the main reservoir of EHEC [10]. A human uropathogenic strain (NCTC9001 T), which is also the E. coli type strain, was also included as reference. These strains were assigned to different phylogroups (Additional file 1: Table S1; Fig. 2). As expected, the greatest difference in genome size was observed between BG1 and the EHEC strain Sakai (the genome size of BG1 is 812,370 bp smaller than the Sakai genome [17.0% of the BG1 genome]). This difference could be explained by the number of mobile genetic elements: the Sakai genome contains 18 prophage regions (at most 5 in the BG1 genome) and 80 insertion sequences (38 in the BG1 genome) [30]. About half of the Sakai-specific sequences are of bacteriophage origin and carry the genes involved in EHEC pathogenesis (bloody diarrhea, hemolytic uremic syndrome) [30]. More surprisingly, the chromosome length of the commensal E. coli Nissle 1917 is 659,093 bp larger than the BG1 genome (13.8% of the BG1 genome). E. coli Nissle 1917 is a human commensal strain known to be a successful colonizer of the human gut and used as a probiotic for the treatment of various intestinal disorders [31]. It is well documented that the Nissle genome carries at least three genomic islands (GEIs) inserted at different tRNA sites (serX, argW and pheV) probably acquired by horizontal gene transfer [32, 33]. These GEIs contained genes encoding proteins considered as fitness factors (microcins, iron uptake systems, proteases …) contributing to survival of E. coli Nissle and successful colonization of the human body [32, 33]. These GEIs were found in non-pathogenic E. coli strains but were also frequently distributed among ExPEC strains [32]. Sequence comparison showed that the genes carried by Nissle 1917 GEIs (mch, mcm, iro, iuc, sat, iha, ybt) are absent in the BG1 genome, suggesting the absence of these GEIs in BG1.

Table 5 Characteristics of whole genome datasets of different E. coli strains
Full size table

In accordance with the differences in genome size, the highest number of tRNA genes, described as common sites for integration of foreign DNA elements (bacteriophages, genomic islands), were detected in the genome of E. coli strains Nissle and Sakai (121 and 103 tRNA genes, respectively while only 86 were identified in the BG1 draft genome (Table 5). The genome of the remaining strains carried 62 (in the type strain NCTC 9001 T) to 85 tRNA-encoding genes (Table 5). These numbers may be slightly different depending on the annotation pipeline used for the draft genome sequences.

Virulence factors

The genes encoding virulence factors in the E. coli BG1 genome were analyzed using blastn against the Virulence Factors Database genomic dataset [34]. A total of 164 genes encoding virulence factors were identified in BG1 (Additional file 2: Table S2), while 181 and 202 genes encoding virulence factors were found in the reference strains NCTC86 and NCTC9001 T, respectively. In-depth analysis of the BG1 genome showed that most of these genes are involved in bacterial adherence to the host epithelium, iron acquisition systems (siderophores) and flagella synthesis. As expected, genes coding for toxins produced by pathogenic E. coli strains responsible for diarrhea or intestinal damage in mammals (Shiga-toxin, heat stable [ST] toxin, heat-labile [LT] toxin, heat-stable enterotoxin 1 [EAST1], cytotoxic necrotizing factor 1 [CNF1]) are absent in the BG1 genome. The E. coli BG1 genome also lacks the genes encoding α-hemolysin and enterohemolysin which are involved in the virulence of pathogenic E. coli strains.

Adherence systems

A total of 49 genes coded for the synthesis of organelles involved in adherence of E. coli to host intestinal epithelium (Additional file 3: Table S3). Accordingly, the transmission electron micrograph of E. coli BG1 showed numerous fimbriae surrounding the bacteria (Fig. 1). Removal of partial genes and incomplete gene clusters revealed that BG1 possessed the genetic information required to encode 12 potentially functional full adherence systems (Table 6). All these systems are known to be produced by pathogenic E. coli and to adhere in vitro to different cells lines (Table 6) (for reviews see [35,36,37]). These adherence systems reflect the ability of commensal E. coli to colonize distinct niches during its transit through the different compartments of the bovine gastro-intestinal tract. It is also of interest to note that some of these adherence systems possess characteristics corresponding to physiological conditions encountered in the bovine gastro-intestinal tract: i) eaeH expression is induced at 39 °C, the internal bovine temperature, but not at 37 °C [38] ii) the pili HCP is involved in adherence of E. coli to bovine gut explants [39] and iii) the F9 fimbriae are essential for in vivo colonization of calves [40]. Furthermore, the stg and F9 gene clusters are strongly associated with E. coli belonging to phylogenetic group B1 [41, 42]. To broaden these results, in silico analysis of adherence systems carried by additional E. coli strains (human and bovine commensal and pathogenic isolates) (Additional file 1: Table S1; Additional file 4: Figure S1) was also performed. A hierarchical clustering based on the presence/absence of 78 distinct adherence systems encoding genes was built using R (version 3.3.1) [43]. As shown in Additional file 4: Figure S1, bovine and human E. coli strains were not separately distributed (the closest relative strains to BG1 were the human E. coli strains S11 and IAI1 [Additional file 4: Figure S1]) suggesting that the adherence systems are associated with the adaptation of E. coli to a specific habitat (i.e. the digestive tract) rather than host specificity. As expected, the uropathogenic strain NCTC9001 T possesses the pap ACDEGHIK genes which are specific to UPEC strains [44].

Table 6 Adherence systems encoded by the E. coli BG1 genome
Full size table

Some of these adherence systems possess redundant properties: EhaB, ELF, HCP and UpaG are known to bind to laminin and curli, EhaA, EhaB, EhaC, ECP, F9, EaeH, HCP and UpaG are involved in biofilm formation (Table 6). This suggested an important role of both laminin binding and biofilm formation in survival and/or multiplication of commensal E. coli . Laminin is an extracellular matrix commonly present in the mammalian intestine which act as an interlinking molecule in connective tissues that promote bacterial adhesion and colonization to the host tissues [45]. Moreover, commensal E. coli strains can reside in mixed biofilms in the mucus layer covering the mouse intestine [4, 46]. Because the survival of E. coli depends on anaerobes that degrade polysaccharides included in the mucus layer, it has been hypothesized that the anaerobes in the mixed biofilms provide E. coli with monosaccharide locally rather than from a mixed pool available to all species [4, 46]. Therefore, the mixed biofilm formation can results in a more efficient carbon source for commensal E. coli strains in the mammalian gut [4, 46].

As discussed above, the adhesion systems encoded by the BG1 genome were associated with E. coli strains mostly isolated from clinical cases (Table 6). However, it is important to note that the BG1 genome did not carry the genes encoding the F17, F5 and F41 fimbriae and the afimbrial adhesin CS31A mainly associated with bovine pathogenic E. coli strains involved in diarrhea [47]. For example, a recent epidemiological study showed that the F5/F41 fimbriae were prevalent among bovine diarrheagenic E. coli isolated in France [48]. The genes encoding F17, F5 and F41 are not detected in the genome of the human and bovine E. coli strains included in this study suggesting that these adherence systems are specific to bovine intestinal pathogenic E. coli .

Flagella synthesis

A total of 47 genes encoding proteins required for flagella synthesis were present in the BG1 genome. Accordingly, the transmission electron micrograph of E. coli BG1 showed peritrichous flagella attached to the bacterial cell surface and clearly distinct from fimbriae (Fig. 1). Flagella are mainly locomotive organelles allowing bacterial movements. However, it is well documented that the flagella (also known as H-antigen) of some pathogenic E. coli mediate the adhesion to or invasion of epithelial cells (NMEC, aEPEC, ETEC, EAEC, EHEC, APEC) and contribute to biofilm formation (UPEC, ETEC) (for a review see Zhou et al. [49]). In particular, flagella of aEPEC, ETEC and EHEC strains specifically recognized a receptor located at the microvillus tips of human enterocytes [50]. Interestingly, E. coli BG1 possesses the genetic information required to encode the flagella H21, a H antigen type reported to be involved in the invasion of EHEC O113:H21 into HCT-8 colonic epithelial cells [49]. Also, it should be noted that STEC strains with serotype O159:H21 have been isolated from bovine as well as porcine feces [51, 52].

Iron acquisition systems

Complete genetic information required for enterobactin synthesis (entABCDEFS) and ferric-enterobactin uptake (fepABCDEFG) was present in the genome of E. coli BG1 (Additional file 2: Table S2). Siderophores, including enterobactin, are mechanisms secreted by E. coli to scavenge iron in order to survive and multiply in hosts or external environments. Siderophores are usually described as crucial for the proliferation of pathogenic E. coli in the host and have been classified as virulence factors. However, enterobactin is frequently produced by commensal E. coli isolated from healthy mammals (human and animal isolates) [53]. ent and fep genes were also found in the genome of the reference strain NCTC86 (data not shown). Accordingly, Pi et al. have demonstrated that enterobactin plays a fundamental role in the colonization of healthy mouse gastro-intestinal tract by non-pathogenic E. coli [54].

Ethanolamine utilization

In a previous study, we demonstrated that ethanolamine present in the bovine gut is used by EHEC as a nitrogen source [11]. Furthermore, ethanolamine promotes expression of fimbrial genes and influenced EHEC adherence to epithelial cells [55]. Interestingly, E. coli BG1 is unable to degrade ethanolamine present in the bovine intestine, while the EHEC reference strain EDL933 gains a growth competitive advantage by assimilating ethanolamine in bovine intestinal content [11]. Therefore, we performed in-depth analysis of the genes involved in ethanolamine utilization in order to understand the inability of the commensal strain BG1 to use ethanolamine as a nitrogen source.

The degradation and assimilation of ethanolamine by EHEC EDL933 requires exogenous adenosylcobalamin (Ado-Cbl) and are encoded by 17 genes included in the eut operon [11]. In this study, we used blastn and Seaview (version 4.6.1) [56] to compare the eut genes of E. coli BG1 with those of EHEC EDL933. Sequence alignment showed 317 SNPs between the two eut operons (97.82% identity) (Additional file 5: Table S4). In addition, no premature stop codon was detected and only 34 amino acid changes due to non-synonymous SNPs were identified among the 17 predicted polypeptides encoded by the eut operon of BG1 (Additional file 5: Table S4). Furthermore, the presence of a 72 bp insertion was also identified in the eutT gene coding for cobalamin adenosyltransferase in the BG1 genome compared with the EDL933 genome (Additional file 6: Figure S2). It is important to note that ethanolamine ammonia-lyase, the key enzyme in ethanolamine degradation, required the Ado-Cbl cofactor produced by EutT to be active. The 72 bp insertion sequence at position 395 resulted in a modified translated polypeptide with 24 additional amino acids at position 132. The possible EutT conformation illustrated in Fig. 3 was predicted using Phyre (version 2.0) [57] and showed that 18 of the 24 amino acids encoded by the 72 bp sequence were predicted to form an additional alpha helix in the BG1 EutT protein.

Fig. 3
figure 3

Predicted secondary structure modeling of the EutT protein of E. coli BG1 obtained with Phyre version 2.0 [57]

Full size image

In summary, in view of the 34 amino acid changes due to non-synonymous SNPs among the 17 predicted polypeptides encoded by the eut operon and the prediction of an additional alpha helix in BG1 EutT cobalamin adenosyltransferase, we suspected a reduced or abolished ethanolamine ammonialyase activity, which could explain the inability of BG1 to assimilate ethanolamine in the bovine digestive tract.

Conclusion

The comparison of whole genomes provides information on gene content and organization, and gives an overview of how organisms are related. The draft genome sequence of E. coli BG1 isolated from the bovine intestine is now available and can provide valuable information at the genomic scale to explore the genetic and functional features adapted to the bovine gut. The genome of E. coli BG1 can be used as a reference for subsequent evolution and comparative studies (some examples of genome comparative analysis have already been described in this report).

As expected, the BG1 genome does not carry the genetic information encoding toxins responsible for intestinal damage. More surprisingly, the E. coli BG1 strain possesses the genetic information required to encode systems classified as “virulence factors” and produced by pathogenic E. coli . This could suggest that genes encoding virulence factors are “in transit” from commensal species that act as genetic depositories with the ability to transmit DNA fragments to pathogenic E. coli strains. However, both pathogenic and non-pathogenic E. coli strains are able to colonize the gut and seem to use similar factors to adhere to the host epithelial cells. Therefore, it is questionable whether the ability of intestinal E. coli to colonize the host gut (resistance to the intestinal flux), excrete siderophores (iron uptake from the surrounding environment) and produce flagella (capacity to move toward nutrient-rich environments) can be considered as “virulence factors”. The terms “virulence”, “fitness” and “colonization” factors appear to overlap for E. coli species. In fact, factors contributing to E. coli survival in a given environment should be considered as fitness and adaptation factors enabling successful colonization of the host rather than strict markers of pathogenesis. In contrast, the factors responsible for disease establishment or intestinal damages during infection (e.g. aqueous or hemorrhagic diarrhea), such as toxins or the type III secretion system, appear to be true virulence factors.