Pneumonia-Specific Escherichia coli with Distinct Phylogenetic and Virulence Profiles, France, 2012–2014

In a prospective, nationwide study in France of Escherichia coli responsible for pneumonia in patients receiving mechanical ventilation, we determined E. coli antimicrobial susceptibility, phylotype, O-type, and virulence factor gene content. We compared 260 isolates with those of 2 published collections containing commensal and bacteremia isolates. The preponderant phylogenetic group was B2 (59.6%), and the predominant sequence type complex (STc) was STc73. STc127 and STc141 were overrepresented and STc95 underrepresented in pneumonia isolates compared with bacteremia isolates. Pneumonia isolates carried higher proportions of virulence genes sfa/foc, papGIII, hlyC, cnf1, and iroN compared with bacteremia isolates. Virulence factor gene content and antimicrobial drug resistance were higher in pneumonia than in commensal isolates. Genomic and phylogenetic characteristics of E. coli pneumonia isolates from critically ill patients indicate that they belong to the extraintestinal pathogenic E. coli pathovar but have distinguishable lung-specific traits.

In a prospective, nationwide study in France of Escherichia coli responsible for pneumonia in patients receiving mechanical ventilation, we determined E. coli antimicrobial susceptibility, phylotype, O-type, and virulence factor gene content. We compared 260 isolates with those of 2 published collections containing commensal and bacteremia isolates. The preponderant phylogenetic group was B2 (59.6%), and the predominant sequence type complex (STc) was STc73. STc127 and STc141 were overrepresented and STc95 underrepresented in pneumonia isolates compared with bacteremia isolates. Pneumonia isolates carried higher proportions of virulence genes sfa/foc, papGIII, hlyC, cnf1, and iroN compared with bacteremia isolates. Virulence factor gene content and antimicrobial drug resistance were higher in pneumonia than in commensal isolates. Genomic and phylogenetic characteristics of E. coli pneumonia isolates from critically ill patients indicate that they belong to the extraintestinal pathogenic E. coli pathovar but have distinguishable lung-specific traits.
N osocomial infections remain a major threat for patients and a burden on healthcare institutions, hampering the public health economy. In intensive care units (ICUs), the most common life-threatening nosocomial infection is ventilator-associated pneumonia; the attributable mortality rate is ≈13%, partly because of increased durations of mechanical ventilation and ICU stays (1,2), all of which generate considerable additional costs (3).
Until the early 2000s, Enterobacteriaceae were not considered as major pathogens responsible for ventilatorassisted pneumonia (2); as such, pathophysiological studies focused mainly on Pseudomonas aeruginosa, Staphylococcus aureus, and Acinetobacter baumannii. However, recent data have consistently shown that Enterobacteriaceae are now frequent etiologic agents of ventilator-assisted pneumonia, more frequent than P. aeruginosa and S. aureus (4)(5)(6). According to the World Health Organization, Enterobacteriaceae, including Escherichia coli, are among the critical priority antibiotic-resistant bacteria (7). Therefore, to optimize patient management, in-depth epidemiologic knowledge of the phenotypic and genotypic characteristics of these bacteria is warranted. Although most E. coli responsible for symptomatic extraintestinal infections (8), including the urinary tract (9), bloodstream (10,11), cerebral spinal fluid (12,13), and peritoneum (14), have been extensively studied, less is known about E. coli strains responsible for pneumonia (15), especially in the context of highly virulent and resistant clones such as sequence type (ST) 69 and ST131 (16).
Our previous monocentric prospective study in France found a predominance of B2 phylogenetic group (66%) among E. coli pneumonia isolates, one third of which belonging to ST complex (STc) 127 (15). To obtain further insights in the physiopathology of ventilator-assisted pneumonia, we conducted a multicentric prospective epidemiologic study of genomic and phylogenetic characteristics of E. coli strains responsible for pneumonia across France.

Patients and E. coli Isolates
We conducted this prospective study in 14 ICUs throughout France, in collaboration with their hospital laboratories. We selected these ICUs to guarantee appropriate geographic coverage of metropolitan France. The same geographic selection was also applied to the ICUs in the Paris area. The ethics committee of the French Intensive Care Society approved the study (SRLF-CE 12-388), which was registered at ClinicalTrials.gov (NCT03303937). Formal consent was not required because of the observational, noninterventional study design (no change in practices, and all procedures already routinely performed). Patients or family/relatives were informed of the nature of the study and its purpose and objectives and had the option of declining participation.
During a 38-month period (2012-2014), any E. coli isolate responsible for pneumonia in a mechanically ventilated patient was collected, regardless of the method of sampling (quantitative cultures of tracheal suctioning, bronchoalveolar lavage, or protected telescoping catheter). Demographic and clinical data for the patient were recorded. We defined ventilator-assisted pneumonia as pneumonia occurring >48 hours after initiation of invasive ventilation, hospital-acquired pneumonia as pneumonia occurring >48 hours after hospital admission but within the first 48 hours of invasive ventilation, and community-acquired pneumonia as pneumonia that occurred either before or within the first 48 hours of hospitalization (17). Each E. coli isolate was stored at -80°C in brainheart infusion broth containing glycerol 20% (the COLO-COLI collection).

E. coli Phylotyping, O-Typing, and Virulence Factor Gene Content Determination
We used quadruplex PCR to determine the E. coli phylogenetic group (A, B1, B2, C, D, E, F) or belonging to clade I (18). Among the strains, we determined the B1 clonal complex 87 (CC87) (Institut Pasteur MLST schema nomenclature corresponding to the ST58 and 155 in the Achtman schema [19]), the 10 main B2 subgroups (20), and the clonal group A (clonal group A) from the D phylogroup (21). The exhaustive correspondence between this typing approach and STc membership according to the currently used Achtman MLST schema is available in (22).
We used PCR to determine the most anticipated serotypes in isolates from patients with extraintestinal infections (Table 1) (23). Multiplex PCR was used to detect genes encoding for 11 frequently encountered extraintestinal virulence factors (sfa/foc, papC, papGII, papGIII, fyuA, iroN, aer, traT, neuC, hlyC, and cnf1), which belong to the main classes of virulence factors (adhesins, toxins, iron acquisition systems, and protectins) (11). For each isolate, we calculated the virulence score, which was defined by the number of virulence factors present among the 11 tested.

Antimicrobial Resistance Phenotypes
We determined the antimicrobial susceptibility of each isolate by using the disk-diffusion method according to the French Society of Microbiology (https://www.sfm-microbiologie.org) ( Table 2). Resistance score was defined as the sum of inactive in vitro antimicrobial agents for each isolate. A score of 1 indicates resistant; 0.5, intermediary; and 0, sensitive. A higher score indicates a more resistant isolate. Detection of gene sequences coding for the TEM, SHV, and CTX-M enzymes was performed by PCR with genomic DNA. The oligonucleotide primer sets specific for the β-lactamase genes used in the PCR assays have been published (24,25).

Characteristics of Other E. coli Strain Collections
To learn more about E. coli pneumonia strains, we compared the isolates from the COLOCOLI collection with those of 2 published collections, originating from the Paris area in France. In 2010, a total of 280 E. coli strains were isolated from fecal samples from community adults and can be considered as commensal strains (COLIVILLE collection) (26). In 2005, a total of 373 E. coli strains were isolated from the blood of 373 in-patients in 14 hospitals during the course of bacteremia (COLIBAFI collection) (27). Of note, 20.6% of isolates from patients with bacteremia were nosocomial and 57% were of urinary origin. Among patients with bacteremia, the portal of entry was pulmonary for <2% (most patients were not in an ICU). For these strains, we determined the phylogroup/subgroup membership, the presence of the 11 virulence factors cited above, and the susceptibility to 6 antimicrobial drugs (amoxicillin, amoxicillin/clavulanic acid, cefotaxime, amikacin, ofloxacin, and cefoxitin). For all strains, we calculated a virulence score.

Statistical Analyses
For our analyses we used GraphPad Prism7 software (https://www.graphpad.com). For quantitative variables, results are presented as the median and range, and for categorical variables, as frequency and proportion. Variables were compared according to whether they were nosocomial or community isolates and whether they were of phylogenetic group B2 or not B2. As we compared the virulence factor gene content of the 3 collections, we also compared the proportion of phylogenetic groups and subgroups and of resistant strains. We used the Student t test to compare continuous variables and the Fisher exact test to compare categorical variables. Because of the observational design of the study and its exploratory aim, we did not adjust for multiple testing (28). We considered p<0.05 to be significant.

Host Characteristics
During the study period, we collected 260 E. coli isolates from 243 patients with a median age of 64 years (interquartile range 52-73 years) ( Table 3). Of these isolates, 117 were responsible for ventilator-assisted pneumonia, 61 for hospital-acquired pneumonia, and 82 for communityacquired pneumonia. The main reasons for ICU admission ‡Resistance score was defined as the sum of inactive in vitro antimicrobial agents for each isolate. Tested antimicrobials were gentamicin, amikacin, minocycline, nalidixic acid, ofloxacin, ciprofloxacin, fosfomycin, furans, trimethoprim, amoxicillin, amoxicillin-clavulanic acid, ticarcillin, piperacillin, piperacillin-tazobactam, imipenem, cefotaxime, and ceftazidime. A score of 1 was attributed for a resistant, 0.5 for an intermediary, and 0 for a sensitive isolate; a higher score thus indicated a more resistant isolate. For each antimicrobial, resistance is defined by the sum of resistant or intermediary isolates. We identified the O-type of 163 strains. We identified B2 phylogroup strains at the clonal level as having a combination of subgroup and O-type, as previously described (27) (Table 1). Among B2 strains, clones II-O6a were predominant, followed by IV-O2b, I-O25b (which belongs to ST131), III-O6a (which belongs to the highly virulent archetypal strain 536 [30]), and VI-O4.
Pneumonia E. coli isolates were more resistant than commensal isolates to all tested antimicrobial drugs except amikacin. Pneumonia isolates and bacteremia isolates did not differ in terms of antimicrobial drug susceptibility, except for resistance to cefotaxime (pneumonia isolate resistance 10.8%) and cefoxitin (pneumonia isolate resistance 11.5%) compared with bacteremia isolates (5.1%; p = 0.009 for cefotaxime, p = 0.004 for cefoxitin). Of note, when the 260 pneumonia isolates were compared with the 220 bacteremia strains of urinary tract origin, they were still distinguishable in terms of phylogroups/subgroups and virulence factor content (Appendix).

Discussion
This prospective nationwide study provides data on E. coli pneumonia isolates in critically ill patients. With regard to the characteristics of E. coli pneumonia isolates, we found the following: 1) a preponderance of phylogenetic group B2 (59.6%); 2) a predominant STc73 (subgroup II) lineage and threatening emergence of ST131 (within subgroup I), STc69 (clonal group A ), and STc127 (subgroup III), along with STc141 (subgroup IV); 3) a much lower proportion of STc95 (subgroup IX) in B2 pneumonia than in bacteremia isolates; 4) a specific virulence factor gene content in pneumonia versus bacteremia strains. Taken together, these epidemiologic, phylogenetic, genotypic, and experimental data argue for inclusion of E. coli pneumonia isolates in the extraintestinal pathogenic E. coli (ExPEC) pathovar but with distinguishable lung-specific traits.
In 2010, Croxen and Finlay reviewed the molecular mechanisms of E. coli pathogenicity (31). Among ExPEC, numerous pathovars were listed, including uropathogenic and neonatal meningitis pathogenic E. coli, but none concerned the lungs. Pneumonia was not even cited as a possible disease caused by E. coli.
The situation is now clearly different. In 2012, Hamet et al. reported that Enterobacteriaceae accounted for a quarter of the 323 episodes of ventilator-assisted pneumonia occurring in their ICU (6). Our group showed that over a 5-year analysis of ventilator-assisted pneumonia episodes, finding Enterobacteriaceae as the responsible pathogen increased significantly (5). In an international multicenter study, Kollef et al. also confirmed that Enterobacteriaceae were the leading pathogens of ventilator-assisted pneumonia in the ICU (4). Among them, E. coli is a major threat, recently highlighted by the World Health Organization, because of its ever-increasing resistance to antimicrobial drugs (7).
ExPEC are characterized by pathogenic virulence factor genes coding for various combinations of adhesins, toxins, iron-acquisition systems, capsule production, and toxins that enable them to cause disease once outside the host gut reservoir (32). ExPEC virulence factors are encoded on the bacterial chromosome, where they are usually located within pathogenicity-associated islands (PAIs) or plasmids. Most ExPEC isolates belong to the B2 phylogroup and, to a lesser extent, the D phylogroup. More in-depth analysis of ExPEC strains has enabled characterization of particular STs of ExPEC isolates including ST131, ST73, and ST127 (33). Our most striking finding was the specificity of pneumonia E. coli strains, compared with bacteremia ones, within the ExPEC family (Table 5). First, although B2 phylogroup strains are preponderant in ExPEC (15,34), we found a trend toward an even greater proportion of B2 isolates among pneumonia isolates (59.6%) than among bacteremia isolates, whatever the origin (52%; p = 0.06) Pneumonia-Specific Escherichia coli (Table 3; Figure). This finding is in the range of what is observed in urosepsis isolates (62%) (11) but a little less than in neonatal meningitis strains (68%) (13). Then, among B2 phylogroup isolates (Table 4), if the predominance of subgroup II (STc73) was expected, we found high proportions of specific phylogenetic group B2 clones among other lineages, which could be worrisome (Table 1). The ST131 O25b:H4 clone represented 9.7% of the B2 phylogroup isolates, 53.3% of them producing an extended-spectrum β-lactamase. Subgroup IV (STc141) was significantly more present among pneumonia strains than among bacteremia strains. Challenging the hypothesis of a commensal character, with a low level of human invasiveness (27), this finding indicates that subgroup IV isolates may have a high affinity for the respiratory tract. Last, we must highlight the greater proportion of subgroup III among pneumonia isolates (namely STc127) compared with bacteremia and commensal isolates, in agreement with previous findings from our group (15). Contrary to isolates from the other 2 collections, pneumonia isolates were composed of fewer IX subgroup strains (STc95). Whereas subgroup IX is usually well represented among commensal and other pathogenic E. coli strains (26,27,35), these data suggest that the respiratory tract is less suitable than other tissues for subgroup IX implantation, a finding in agreement with our previous report in which subgroup IX was not represented (15). Among D phylogroup strains, the higher prevalence of clonal group A (STc69) in pneumonia isolates than in bacteremia and commensal isolates was unexpected. Among 571 D phylogroup E. coli responsible for extraintestinal infection, Johnson et al. reported only 144 clonal group A E. coli (25.2%) (29). The multidrug-resistant nature of these pathogens is of increasing concern (29,36).
Regarding virulence factor gene content, the literature suggests that PAIs involved in ExPEC causing pneumonia differ from those involved in urinary tract and blood-stream infections (37,38). Using PAI deletion mutants in a rat model of pneumonia, Phillips-Houlbracq et al. related pneumonia pathogenicity to the presence of PAIs I and III (37). In our study (Figure 1), pneumonia isolates differed from bacteremia isolates because they significantly more often carried sfa/foc, iroN (both belonging to PAI III), pap-GIII and cnf1 (belonging to PAI II), and hlyC (belonging to PAI I and II). The role of α-hemolysin in experimentally induced pneumonia in rats has been reported (15,39). Of note, the virulence factor content of the pneumonia strains still differs when we consider only the urosepsis isolates (Appendix). Although additional studies are required to confirm, these findings do suggest a coherent molecular trait for the isolates' lung specificity.
Our study should be interpreted within the context of its limitations. First, although our collection of E. coli pneumonia isolates is large, results regarding some subgroup analyses will require confirmation because of their sample size. Our choice of PCR rather than whole-genome characterization was governed by our wish to compare the 2 other collections with the same type of data. Despite these limitations, our analysis based on 14 centers representing France on a population and geographic level, its prospective design, and the comparison of our large number of isolates  +  CC87 †  +++  +++  +  B2  ++++  ++++  +++  I  +  +  +  II  ++  ++  ++  III  +  --IV  +  -+  IX - with those from recently published collections enable us to draw valid conclusions. We did not assess the functionality and expression of the encoded virulence factors in these isolates. Our team has consistently demonstrated these features in several murine models of infection (including pneumonia) and observed a strong correlation between the presence of these genes and death (15,34,37). Our data raise the question of why certain clonal lineages were overrepresented in patients with respiratory tract infection. Patients acquire E. coli infection from their own digestive tract (41). This event implies an upward retrograde motion of the bacterial cells to reach the oropharynx and the lung parenchyma and suggests particular metabolic-adaptation and response-to-stress characteristics (e.g., to overcome the acidity of the stomach). We have previously shown that some E. coli strains are capable of high growth capacities in relation to metabolic pathways (42) while others are highly resistant to stress (43). For both studies, however, no link to specific clones could be established. The specific organ tropism therefore more probably results from a combination of genetic background and virulence factors.
In summary, we identified emerging pneumonia-causing pathogenic E. coli whose main characteristics define them as ExPEC. Their specificities include a very strong proportion of B2 phylogroup isolates; a high proportion of subgroups II (STc73), I (STc131), IV (STc141), and III (STc127); and consequent proportions of clonal group A (STc69) isolates within the D phylogroup. Virulence factor gene content of pneumonia isolates also appeared to be singular compared with that of bacteremia isolates, among them urosepsis isolates. These epidemiologic data underline the specificity of pneumonia E. coli populations and may help with the design of more targeted therapies.
Pneumonia-Specific Escherichia coli also grateful to the patients and their families, and we gratefully acknowledge Dominique Poignard and Marie-Claire Hipeaux for technical assistance and all the nurses and assistant nurses of the ICU for their help in obtaining the microbiological samples.