By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
Corresponding author: X. Nassif, INSERM U1002, Pathogénie des Infections Systémiques, Faculté de Médecine, Hôpital Necker-Enfants Malades, Université Paris Descartes, AP-HP, Paris, France E-mail: email@example.com
Extraintestinal pathogenic Escherichia coli (ExPEC) strains, a major cause of bacteraemia, typically belong to phylogenetic group B2 and express diverse accessory traits that contribute to virulence in mouse infection models. However, their high genomic diversity obscures the relationship between virulence factors and severity of infection in patients. In this study, we analysed concomitantly the strain’s expression of virulence in a mouse model, genomic determinants and the clinical severity of infection in 60 bacteraemic patients. We show that bacterial virulence based on an animal model study and virulence factor determination is not correlated with pejorative outcome of E. coli human blood infections.
The Escherichia coli species comprise commensal bacteria, which belong to the normal gut microbiota of humans and many animals, as well as highly virulent intestinal (InPEC) and extra-intestinal (ExPEC) pathogenic variants. ExPEC form a heterogeneous group of bacteria responsible for a large spectrum of infectious diseases such as cystitis or pyelonephritis, meningitis, pneumonia, and bone or wound infections . ExPEC are also the leading cause of bacteraemic infections .
Phylogenetic analyses have established that E. coli can be divided into four main groups (A, B1, B2 and D) and a more recently described group E . Classification of pathogenic E. coli is classically based on the presence or absence of specific DNA regions, referred to ‘pathogenicity islands’ (PAIs), acquired by horizontal gene transfer . These DNA regions encode a large number of virulence factor (VFs) such as adhesins, iron-acquisition systems, capsules and toxins . It is clearly established that these accessory genetic traits contributed to intrinsic virulence of E. coli strains in a mouse model of infection . The B2 strains possess the highest level of virulence factors .
Epidemiological studies have demonstrated that most clinical strains of ExPEC belong to the phylogenetic group B2, and to a lesser extent to the group D . Also, several VFs are known to be associated with the ExPEC phenotype . However, ExPEC exhibit a considerable genomic diversity  and none of the known VFs is constantly expressed in all ExPEC strains .
In the present study, we conducted a concomitant analysis of (i) the intrinsic virulence of bacteraemic ExPEC strains based on the presence or absence of specific DNA sequences and virulence in mice and (ii) the severity of the clinical infection caused by these strains to test whether a correlation between these elements exists.
Materials and Methods
Patients, strains, macroarray assay
Sixty strains used for this study were selected among an unbiased collection of 161 ExPEC strains corresponding to all consecutive episodes of E. coli bacteraemia in two major university hospitals in Paris [7,8]. This subset of 60 bacteraemic ExPEC was selected as being representative both of the clinical characteristics of the sepsis and the bacterial characteristics of the entire collection . The considered clinical characteristics were the portal of entry of the sepsis, the severity of the disease , the immune status (immunodeficiencies included infection with human immunodeficiency virus (HIV) with CD4+ count <200/mm3, administration of immunosuppressive therapy (corticosteroid therapy >1 mg/kg/day for >30 days, immunomodulating agents and anti-neoplastic chemotherapy) and neutropenia defined as an absolute granulocyte count of <500/μL) and the associated co-morbidities (underlying solid tumour or haematological malignancy, HIV-positive serology, diabetes mellitus (defined as requiring insulin or hypoglycaemiant therapy) and renal failure (defined as a creatinine clearance of <30 mL/min)). The bacterial characteristics encompassed the phylogenetic diversity of the strains based on PCR phylogrouping  and MLST analysis . These 60 bacteraemic ExPEC were also characterized for the presence of nine well-documented VFs (papC, papGIII, papGII, sfa, hlyC, cnf1, iut, iroN, fuyA) using a multiplex PCR assay , and for the 40 most frequently recovered O-antigens by an allele-specific polymerase chain reaction as previously described [12,13]. To evaluate further the accessory genomic content of each strain, a DNA macroarray containing a large E. coli flexible gene pool was used . Briefly, DNA probes were selected using the genome sequence of the O6:K15:H31 uropathogenic E. coli strain 536, the O6:K2:H1 uropathogenic E. coli strain CFT073, and E. coli K-12 (MG1655). DNA sequences with more than 90% of homology in the three genomes were excluded. In addition, specific sequences of the meningitis-associated strain RS218, with less than 90% homology in the above three strains, were included. This flexible gene pool DNA macroarray contains a total of 2324 probes that could be classified in functional clusters according to their annotated function (https://www.genoscope.cns.fr/agc/mage); that is, known virulence factors (VF_sequences, n =208), cell structure membrane proteins (MB_sequences, n =133), putative functional enzymes (Met_sequences, n =500), genomic regulator (Transcriptional regulator_sequences, n =104) and unknown hypothetic protein (ORF_sequences, n =1033). Genomic DNA preparation and hybridizations were performed as previously described . The macroarray data were analysed using the ArrayVision software (Imaging Reseach, St Catharines, Canada) for signal quantification.
A mouse septicaemia model previously described  was used to assess the intrinsic extra-intestinal virulence of all 60 human ExPEC E. coli. Between the isolation and the mouse inoculation, these strains were not subcultured more than three times. Briefly, 10 outbred female Swiss mice (6–8 weeks old, 25–30 g) were challenged subcutaneously in the abdomen with 109 cfu/mL of log-phase bacteria in 0.2 mL Ringer solution . Mortality was assessed over 7 days post-challenge. It was previously established that when surviving mice were euthanized at 7 days, the organs (blood, kidney, liver and spleen) were sterile. Thus, a longer observation was not contributive . Each experimental series included a positive control (urosepsis strain CFT073) and a negative control (commensal derived strain K-12 MG1655). In this model, lethality is a rather clear-cut parameter and strains were usually classified either as non-killer (MNK, strains killing none or one mouse out of 10) or killer (MK, strains killing nine or ten mice out of 10) . Strains that did not fall into these two categories were considered as being intermediate killer (MIK). No significant difference between the survival curves obtained with MK and MIK strains was observed.
Animal experimentations were carried out according to the authorization no. 6665 given by the Ministère de l’Agriculture, France.
Comparisons were based on the chi-square test or Mann–Whitney test for categorical variables and the Fisher exact test when numbers were below five. All tests were two-tailed and p <0.05 was considered significant. Factorial analysis of correspondence (FAC), a method suitable for studying a very large amount of data, was used to describe associations among these data . FAC is an eigenvector method of ordination that uses a covariance matrix based on chi-square distances. It describes the dispersion and shape of a cloud of n objects (here, the E. coli strains) or p variables (here, the studied variables) in a multidimensional space, by replacing the original dataset with a new set of orthogonal linear coordinates in a space of significantly lower dimension. The explained variances of the elements of the dataset (the strains and the variables) are in decreasing order of magnitude with respect to these new coordinates. The variables used for FAC are categories. The computation determines a plane defined by the first two principal axes of the analysis; the first axis, F1, accounts for most of the variance, and the second axis, F2, orthogonal to F1, accounts for the largest part of the variance that is not accounted for by F1.
Several FACs were conducted with SPAD.N software (Cisia, Saint Mandé, France) from five two-way tables. These tables had 60 rows, one for each E. coli strain and 230, 140, 270, 53 and 268 columns corresponding to the same numbers of variables included in the five tables considering the VF_, MB_, Met_, ORF_ and Transcriptional regulator_ sequences, respectively, studied in the macroarray plus the mouse lethality data, the phylogenetic groups of the strains and the characteristics of the patients. Within each table, for each column, each strain was coded as a binary code: present = 1, absent = 0.
A strong association between the phylogenetic group, the virulence factor content of each strain and virulence in mice is observed
Of the 60 selected strains, 27 (45%) belonged to the B2 group whereas 15 (25%), 11 (18%) and 7 (12%) strains belonged to the D, A and B1 groups, respectively. In agreement with previous results, strains of the B2 group appeared as being the most virulent strains in mice (44% of B2 group strains are killer strains), followed by the D group strains (33% killer strains). At the opposite, A and B1 group strains were non-virulent in mice (27% and 0% killer strains, respectively). It is well documented that pathogenic E. coli strains exhibit considerable genome diversity [3,16]. Here, we investigated the genome diversity of the 60 bacteraemic ExPEC strains by macroarray analysis, as described in the materials and methods section. We searched for a correlation between the presence of VF-related sequences and the high virulence expressed in mice. To this end, we compared the profiles of genes between killer (MK) and non-killer (MNK) strains. Thirty-eight per cent of VF-related genes (80 of 208 sequences, described in reference ) were significantly more prevalent in MK than in MNK (p comprised between <0.01 and 0.05). A large number of the genes that were over-represented in MK strains encoded for iron capture systems (n =38; 76% in MK vs. 55% in MNK, p 0.029) and capsule determinants (n =27; 44% vs. 26%, p 0.003). Finally, we determined the O-types of the 60 studied strains by an allele-specific PCR assay [12,13]. Using this approach, 48 strains (80%) were typable, with a clear association between specific O-types and phylogroups (Table S1), as previously reported [13,17–19]. Altogether, these results show a strong association between the phylogenetic group, the O-type, the virulence factor content of each strain and virulence in mice.
In order to confirm this association, we conducted an FAC on the VF genes, the phylogenetic groups and the mouse lethality data. The projections of the variables on the plane F1/F2, which accounted for 39.9% of the total variance, distinguished the B2 group, most virulence genes (including the genes identified by PCR: cnf, sfa, kps, iroN, papC and hly) on the negative values of the first factor F1 from the A and B1 groups and a very few VF on the positive values of this factor. The phylogenetic group D was projected on to the negative values of the second factor F2 (Fig. 1). The variables MK and MIK were projected on to lower negative values of F1 than the B2 group, whereas the variable MNK was projected on to lower positive values of F1 than the A and B1 groups. The four FAC conducted on the MB_, Met_, Transcriptional regulator_ and the ORF_sequences retrieved similar results: a clear distinction of the four phylogenetic groups, indicating a strong clonal structure of the species, and a correlation between the B2 group and the mouse lethality (data not shown). Importantly, these results show that genomic diversity in bacteraemic ExPEC was determined by VF_sequences but also by other classes of genes such as metabolic-related sequences (Met_sequences).
More severe issue of sepsis in human E. coli infection is not related to virulence in the mouse model
To examine whether an association might exist between the severity of the sepsis in humans and the virulence of these strains in mice, an FAC was conducted on the phylogenetic groups, the mouse lethality data and the clinical data. The projections of the variables on the plane F1/F3, which accounted for 30.27% of the total variance, distinguished a first group of variables (B2 group, MK variable, urinary tract origin of the sepsis, immunocompetent character of the patients, and presence of co-morbidity factor) by negative values of the factor F3 from a second group (phylogenetic groups A and B1, MNK variable and digestive origin of the sepsis) by the positive values of the two factors (Fig. 2). The severity of the bacteraemia and the death of the patients were orthogonally projected (i.e. independently) from the other variables, according to the negative values of the first factor F1. Similarly, the severity of the sepsis in humans was not correlated with the O-types of the strains (data not shown). Altogether, these data strongly suggest that there is no correlation between the virulence of ExPEC in mice and the severity of the sepsis. More severe issue of sepsis in human E. coli infection is not related to virulence in the mouse model.
In this study, we analysed concomitantly the intrinsic virulence of ExPEC strains assessed by the bacterial genomic content and the lethality in a mouse model [5,14], and the severity of bacteraemia including diverse portal of entry in patients. We showed that B2 strains are more likely to be virulent in mice and that VF sequences were most highly represented in this B2 phylogenetic group of E.coli. Similarly, several rodent models of bacteraemia using various routes of inoculation [5,14,20,21] of urinary tract infection  and of meningitis , as well as non-rodent models, zebra fish , locust , Caenorhabditis elegans  and Dyctostelium discoidum , have found a correlation between the B2 phylogenetic group, the presence of VF genes and the experimental pathogenicity. However, these works were not focused on the relationships between the intrinsic virulence and the clinical severity of the infection. Epidemiological studies showed that B2 strains were over-represented in clinical pathogenic strains. Also, specific accessory traits admitted as virulence factors were described as highly prevalent and responsible for a specific pathogenic profile of ExPEC strains [1,27]. Nevertheless, it is well documented that pathogenic strains of E. coli exhibit considerable genomic diversity [3,16].
No correlation has been evidenced between the presence of a set of VFs and the outcome of an infection in the host , or between the severity of sepsis and phylogenetic groups of strains [8,28]. In a large prospective study including 1051 consecutive E. coli bacteraemia, severity of clinical outcome was more strongly associated with specific characteristics of the patient, especially immune status, and with the portal of entry, than with bacterial determinants . In another study, Johnson et al.  have shown that associations between bacterial factors of E. coli bacteraemic isolates and host characteristics were highly variable, depending on the primary source of bacteraemia.
In this study, we show that the ability to be septicaemic in humans does not correlate with virulence in mice. Numerous strains isolated from blood cultures are not virulent in the mouse model. No link was observed between the severity of sepsis and the killer status of strains in mice. Additionally, the definition of bacterial virulence based on animal model studies and VF determination was not correlated with more severe evolution of E. coli human blood infections. The evolution of human blood infection is dependent on many, often heterogeneous, host factors. These include immune status, which may explain the lack of correlation with observations in animal models. Finally, classical genes so-called ‘virulence factors’ that are described from experimental data obtained in mice may not be the determinants of the ExPEC’s intrinsic virulence explaining the severity of infection in humans.
With subcutaneous bacterial injection, the mouse model tests essentially the ability of a strain to be able to survive in the extracellular fluid. Indeed, the usual VFs associated with the ExPEC phenotypes are those that allow the bacteria to resist phagocytosis by polymorphonuclear neutrophils, killing by complement, and promote bacterial ability to chelate the ferric iron from iron-containing molecules. One point that the mouse model does not address is the ability of the bacteria to disseminate from the portal of entry. The above data showing no correlation between the severity of the sepsis on the one hand and the mortality in mice, or the expression of usual virulence factors, on the other hand, may suggest that survival in the extracellular fluids is not as important as initially believed from the data obtained in mouse models of septicaemia .
We thank Dr Eric Fontas and Dr Maddugoda Madhavi for critical comments on the manuscript and Jérémy Glodt for the O-typing.
LL, FJ, BP, SB and XN designed the study. LL, FJ, EF, GG, SG, OC and EC conceived the experiments and gathered data and isolates. LL, FJ, EF, BP, ED and XN analysed the data and provided statistical expertise. FJ, LL, ED, BP, EL, SB and XN wrote the manuscript. All the authors read and approved the final manuscript.
The authors have no conflict of interests. No specific financial support has been used.