Phylogenetic Distribution of CTX-M- and Non-Extended-Spectrum-β-Lactamase-Producing Escherichia coli Isolates: Group B2 Isolates, Except Clone ST131, Rarely Produce CTX-M Enzymes

Abstract

Escherichia coli is the species most frequently associated with clinical infections by extended-spectrum-β-lactamase (ESBL)-producing isolates, with the CTX-M ESBL enzymes being predominant and found in genetically diverse E. coli isolates. The main objective of this study was to compare, on the basis of a case-control design, the phylogenetic diversity of 152 CTX-M-producing and 152 non-ESBL-producing clinical E. coli isolates. Multilocus sequence typing revealed that even though CTX-M enzymes were largely disseminated across the diversity of E. coli isolates, phylogenetic group B2 showed a particularly heterogeneous situation. First, clone ST131 of group B2 was strongly associated with CTX-M production (55 [79%] of 70 isolates), with CTX-M-15 being predominant. Second, the remaining members of group B2 were significantly less frequently associated with CTX-M production (9 [12%] of 75) than E. coli phylogenetic groups A, B1, and D (88 [55%] of 159). CTX-M-producing ST131 E. coli isolates were significantly more frequent in patients hospitalized in geriatric wards or long-term care facilities. Besides, the non-ESBL ST131 isolates significantly more frequently showed resistance to penicillins than the non-ESBL, non-ST131 isolates did. In conclusion, the present study emphasizes the particular antimicrobial resistance and epidemiologic characteristics of clone ST131 within group B2, which could result from the higher antibiotic exposure of this clone, as it is the predominant clone of group B2 carried in the human gut.

INTRODUCTION

Escherichia coli is the Enterobacteriaceae species that causes the largest number of infections, but this species is also a widespread gut commensal of humans and animals. Strains of E. coli are therefore frequently exposed to antimicrobial agents, which has resulted in the emergence and rapid diffusion of antibiotic-resistant E. coli clinical isolates. Extended-spectrum β-lactamases (ESBLs), which confer resistance to extended-spectrum cephalosporins, were, until the year 2000, produced essentially by Klebsiella pneumoniae and Enterobacter sp. isolates responsible for nosocomial infections. However, E. coli has become dominant among the ESBL-producing enterobacterial species. This is especially worrisome as the human gut carriage of E. coli favors the dissemination of ESBLs in the community (31). Interestingly, the switch of dominant species among the ESBL-producing enterobacterial isolates occurred concomitantly with the emergence of novel ESBL enzymes that belong to the CTX-M family. The CTX-M enzymes have superseded the TEM and SHV ESBLs (2) and are found in genetically diverse E. coli isolates. Among CTX-M enzymes, CTX-M-15 predominates and has been associated with the dissemination of a particular E. coli clone of sequence type 131 (ST131) (23, 27).

The species E. coli is genetically diverse, and strains have been classified into a number of phylogenetic groups, the most frequent ones being A, B1, B2, and D (12, 14, 29, 32). The ST131 clone belongs to phylogenetic group B2 and is associated with extraintestinal infections, including urinary tract infection, bacteremia, and meningitis (27). However, ESBL-producing E. coli is genetically diverse.

In a previous study, entitled Coli β, we performed a case–double-control study to determine the risk factors associated with the acquisition of CTX-M-producing E. coli clinical isolates (24). We have shown that, in addition to health care-related factors, patient origin and living in collective housing were independently associated with the isolation of a CTX-M-producing E. coli clinical isolate.

In the present study, the main objective was to compare the phylogenetic and clonal diversities of the CTX-M-producing and non-ESBL-producing E. coli isolates collected during the Coli β study. The second objective was to assess the association of particular CTX-M enzymes with the clonal background of strains. The third objective was to compare the distribution of the ST131 isolates with that of the non-ST131 isolates with regard to medical wards, clinical samples, and antibiotic susceptibility in order to better understand the epidemiology of this prevalent clone that often produces CTX-M-15 enzymes (1, 6, 7, 11, 15, 25, 26, 30).

MATERIALS AND METHODS

Bacterial isolates.

The 304 E. coli isolates studied, which comprised 152 CTX-M producers and 152 non-ESBL producers, were prospectively collected from 17 November 2008 to 30 June 2009 at 10 hospitals (7,554 beds) in the area of Paris (Assistance Publique—Hôpitaux de Paris, Paris, France). They were collected within the framework of the Coli β study, a case–double-control study aimed at determining the factors associated with a CTX-M-producing E. coli clinical isolate (24). This study was approved by the ethics committee of the groupe hospitalier Hôpitaux Universitaires Paris Nord Val de Seine (institutional review board no. IRB 00006477). Each of the 152 CTX-M-producing clinical isolates studied came from a case to which was assigned a control patient with a non-ESBL-producing isolate detected on the day of case detection or within the next 3 days. Control isolates caused infections but not necessarily the same as those caused by the CTX-M-producing isolates. The clinical origins of the E. coli isolates studied and the wards where the patients were hospitalized were recorded (Tables 1 and 2). CTX-M-producing E. coli clinical isolates detected within the first 48 h of hospitalization were deemed imported.

Table 1.

Distribution of ST131and non-ST131 CTX-M producers among the participating hospitals and the number of wards involved in each hospital by the patients infected with these isolates

Hospital no. (type)a	No. of CTX producers (no. of STs)/no. of wards involved
	ST131	Non-ST131
I (STCF)	3/3	15 (11)/9
II (LTCF)b	9/4	9 (8)/5
III (LTCF)b	8/5	4 (4)/3
IV (STCF)	2/2	5 (5)/4
V (STCF)	11/10	20 (14)/15
VI (Pediatric)	2/2	4 (4)/4
VII (STCF)	12/8	16 (13)/9
VIII (STCF)	1/1	12 (9)/8
IX (Pediatric)	1/1	1/1
X (STCF)b	6/5	11 (7)/5

^aSTCF, short-term care facility; LTCF, long-term care facility.

^bHospital with a rehabilitation department.

Table 2.

Distribution of ST131 and non-ST131 E. coli strains among CTX-M producers and non-ESBL producers according to medical ward

Ward	No. (%) of CTX-M producer isolates		Adjusted P valuea	No. (%) of Non-ESBL producer isolates		Adjusted P valuea
	ST131	Non-ST131		ST131	Non-ST131
Medicine	15 (27)	47 (49)	0.1	6 (40)	12 (9)	0.2
Geriatric/long-term care	15 (27)	6 (6)	0.01	4 (27)	21 (15)	0.9
Rehabilitation	10 (18)	9 (9)	0.4	2 (13)	59 (43)	0.5
Surgery	7 (13)	16 (17)	0.7	2 (13)	29 (21)	1.0
Obstetric/gynecologic	5 (9)	5 (5)	0.7	1 (7)	7 (5)	1.0
Pediatrics	2 (4)	1 (1)	0.6	0 (0)	6 (5)	1.0
Intensive care	1 (2)	13 (13)	0.1	0 (0)	3 (2)	1.0

^aFisher exact test (adjusted for multiple comparisons).

Antibiotic susceptibility.

The antibiotic susceptibility of the 304 E. coli isolates was tested by using the disk agar diffusion method and interpreted according to the 2008-2009 French Antibiogram Committee recommendations (www.sfm-microbiologie.org) for the following antibiotics: ampicillin or amoxicillin, amoxicillin-clavulanic acid, ticarcillin, ticarcillin-clavulanic acid, piperacillin, piperacillin-tazobactam, cephalothin, cefoxitin, cefotaxime or ceftriaxone, ceftazidime, cefepime, imipenem, gentamicin, amikacin, nalidixic acid, ciprofloxacin, cotrimoxazole, and fosfomycin.

β-Lactamase detection.

ESBL production was detected by the double-disk synergy test as recommended by the French Antibiogram Committee (13). The E. coli strains screened as ESBL producers were tested for ESBL-encoding bla genes by PCR as previously described (21). The amplified fragments were sequenced by using primers specific for the bla_CTX-M, bla_TEM, and bla_SHV genes as previously described (21).

MLST.

Multilocus sequence typing (MLST) was carried out on the basis of an international MLST scheme (32) by using previously described primers and guidelines as specified at http://mlst.ucc.ie/. The allele and profile assignments were determined on the basis of the central E. coli MLST database at the above website.

Data analysis.

Sequence chromatograms were edited and stored by using BioNumerics v6.6 (Applied Maths, Sint-Martens-Latem, Belgium). To achieve high levels of confidence in each nucleotide substitution, all of the nucleotides within the internal gene portion chosen for MLST analysis were supported by at least two sequence chromatograms. Microevolutionary relationships among STs were investigated by comparing allelic profiles by using the eBURST-like minimum spanning tree (MStree) method in BioNumerics.

To determine the deep phylogenetic relationships among strains and assign them to the main E. coli phylogenetic groups, we first constructed a neighbor-joining tree based on the concatenated sequences of the seven MLST genes for the 304 isolates, as well as 230 isolates of known phylogenetic groups, including the ECOR reference collection, previously analyzed with the same genes (32) (http://mlst.ucc.ie). On the basis of their clustering positions relative to reference strains, the isolates were assigned to phylogenetic groups A, B1, B2, and D. The latter group was broadly defined as including strains previously labeled as F (14) or ABD (32). To confirm the phylogenetic group assignments using a recombination-aware method, ClonalFrame (8) analysis was performed. Two runs, each with a total of 150,000 iterations, were performed, and the first 50,000 were discarded as burn-in; the two runs gave consistent results. Finally, the Structure program was used to determine for each isolate the proportion of ancestry of nucleotides from various numbers of ancestral populations. These results were used to confirm the group assignments determined as described above. Finally, the triplex PCR pattern (4) was determined for some isolates to control the correspondence between our sequence-based groupings and the triplex PCR method. Genotypic diversity was estimated by the Simpson index computed from http://www.comparingpartitions.info.

To compare the distributions of categorical variables between the paired CTX-M-producing E. coli isolates and non-ESBL-producing E. coli isolates, the nonparametric Cochran-Mantel-Haenszel paired test was used. To compare the distributions of the categorical variables between nonpaired ST131 isolates and non-ST131 isolates among the CTX-M producers or the non-ESBL producers with regard to medical wards, clinical samples, and antibiotic susceptibility, the nonparametric Fisher exact test was used. If the overall distribution of categorical variables differed significantly, an adjustment for these two tests was used to adjust P values for multiple comparisons class by class (by using SAS proc MULTTEST). All statistical analyses were performed with SAS software, version 9.1 (SAS Institute, Cary, NC). All tests were two sided at the 0.05 significance level.

RESULTS

Phylogenetic diversity.

To test whether CTX-M-producing isolates were distributed differently from non-ESBL-producing isolates in the main E. coli groups, we determined the phylogenetic positioning of the 304 isolates on the basis of the seven gene sequences. Internal portions of the seven MLST genes (a total of 3,423 nucleotides) were obtained for the 304 isolates. A total of 286 (8.4%) nucleotide positions were variable, among which 225 were parsimony informative, providing a large quantity of information for phylogenetic analysis. For selected isolates, we also determined the triplex PCR pattern. Overall, the numbers of isolates classified into groups A, B1, B2, and D on the basis of gene sequences were 70 (23%), 42 (14%), 145 (48%), and 47 (15%), respectively. The two phylogenetic group assignment methods disagreed for six strains (2%) that were clearly assigned to group A or B1 on the basis of gene sequences but to B2 or D on the basis of triplex PCR. As sequence-based assignment is more reliable, we used the assignments based on this method in subsequent analyses.

The phylogenetic relationships of the 304 isolates (Fig. 1) clearly demarcated group B2 and isolates assigned to group D. However, the branching order among groups B1 and A and the branch that included ST648, which was previously described as group F (14), was not resolved. This lack of resolution is consistent with previous work that demonstrated a high frequency of recombination between E. coli groups, in particular between groups A and B1 (32).

Fig 1.

Phylogenetic distribution of CTX-M enzymes in E. coli. The phylogeny was constructed on the basis of the seven MLST gene sequences using ClonalFrame software. CTX-M-producing isolates are indicated by circles colored according to the CTX-M enzyme type as indicated. Phylogenetic groups A, B1, B2, and D are identified by a gray background. STs of particular interest (see text) are indicated. ST117 belongs to group ABD (32), whereas ST648 belongs to group F (14).

The CTX-M-producing E. coli strains were grouped as follows: 44 (29%) in group A, 19 (13%) in group B1, 64 (42%) in group B2, and 25 (16%) in group D (Fig. 1). The non-ESBL-producing isolates were grouped as follows: 26 (17%) in group A, 23 (15%) in group B1, 81 (53%) in group B2, and 22 (15%) in group D. The distribution into phylogenetic groups was not significantly different between the two populations of E. coli.

Genotypic diversity.

To compare the genotypic diversity of CTX-M-producing E. coli with that of non-ESBL-producing E. coli, the MLST data were used to group isolates into STs and clonal complexes (CCs). To group closely related STs into meaningful clonal groups, we used the standard definition of CCs as groups of STs differing by only one gene from at least one other member of the group (10). The 304 E. coli isolates were grouped into 116 distinct STs. Of these, 36 were novel STs, which were assigned ST numbers ST1648 to ST1682 and ST1806. The most frequent STs were ST131 (group B2) with 70 isolates (23%), ST10 (group A) with 22 isolates (7%), and ST73 (group B2) with 16 isolates (5%). In contrast, 87 STs (including all of the novel STs) comprised a single isolate. There were 13 CCs, including 4 that were represented at high frequencies: CC131 with 71 isolates (23%), CC10 with 46 (15%), CC73 with 18 (6%), and CC23 with 15 (5%) (Fig. 2). CC131 included only two STs, ST131 (70 isolates) and the newly defined ST1680 (1 isolate), which differed from ST131 at the gyrB gene. In contrast, CC10 (11 STs) and CC23 (5 STs) were more heterogeneous (Fig. 2). Of note, CC10, CC23, CC73, and CC131 correspond to the frequent CCs CC2, CC4, CC66, and CC43, respectively, found previously among bacteremia isolates by using an alternative MLST scheme (14) (www.pasteur.fr/mlst).

Fig 2.

Clonal compositions of CTX-M-producing and non-ESBL-producing populations. The minimum spanning tree shown was constructed on the basis of MLST allelic profiles by using BioNumerics v6.6.4. Each circle corresponds to an ST, the number of which is indicated beside it. Circle size corresponds to the number of isolates of the respective ST. Pie section colors within circles correspond to CTX-M enzyme types as indicated in the upper right corner and correspond to the colors in Fig. 1 (white, non-ESBL-producing isolates). STs that belong to a single CC are surrounded by gray shading. The CC numbers of the major CCs are indicated with a gray background. The links between circles depict the number of allelic mismatches between STs as follows: solid black, one mismatch; solid gray, two mismatches; plain gray, three mismatches; dashed gray, four mismatches; dotted gray, five or more mismatches.

The overall genotypic diversity of the 304 isolates was 93.5% and was homogeneous across the 10 hospital centers, with frequent STs being recovered from many of them. Notably, the genotypic diversity of CTX-M-producing E. coli and non-ESBL-producing E. coli was distinct, as Simpson's index of diversity was 85.7% for the first population of E. coli but was much higher, 96.9%, for the second one. The lower diversity among CTX-M-producing E. coli isolates was, in large part, explained by the high frequency of ST131 (55 CTX-M-producing ST131 isolates, 36%) and ST10 (13 isolates, 8.6%). In addition, non-ESBL-producing E. coli corresponded more frequently to STs with a unique isolate (36%) than CTX-M-producing E. coli did (21%, P = 0.005).

The clonal compositions of CTX-M-producing and non-ESBL-producing E. coli populations differed significantly both at the level of STs and at the level of CCs (P < 10⁻⁴). Four individual group B2 STs were distributed differently between CTX-M-producing and non-ESBL-producing E. coli strains: ST131 (36.2% versus 9.9%, P < 10⁻⁴), ST73 (0% versus 10.5%, P = 0.0003), ST95 (0% versus 5.3%, P = 0.01), and ST141 (0% versus 4.6%, P = 0.02). Similarly, when the distributions of the 13 recognized CCs were compared, 2 of them, belonging to group B2, had a significantly distinct distribution between CTX-M-producing and non-ESBL-producing E. coli: CC131 (36.8%versus 9.9%, P < 10⁻⁴) and CC73 (0% versus 11.9%, P = 0.005). CC10 (group A) was slightly more frequent among CTX-M producers than among non-ESBL producers (21.7% versus 8.6%, respectively, P = 0.08). In contrast, CC23 (group A), the third most frequent CC among CTX-M producers (after CC131 and CC10), was also frequent in non-ESBL producers (P = 0.9).

CTX-M enzyme distribution among phylogenetic groups and CCs.

Of the 152 CTX-M-producing isolates, 10 different variants of CTX-M enzymes were detected. There were 119 (78%) enzymes belonging to group CTX-M-1, including 78 (51%) CTX-M-15 and 36 (24%) CTX-M-1 enzymes. There were 30 (20%) enzymes in group CTX-M-9, including 18 (12%) CTX-M-14 enzymes. Finally, there were 3 (2%) group CTX-M-2 enzymes.

CTX-M enzymes were distributed among all of the phylogenetic groups, with frequencies per group ranging from 44% (B1 and B2) to 53% in group D and 62.5% in group A. However, the distribution of the CTX-M variants showed striking differences among and within groups (Fig. 1). First, a sharp contrast was observed within group B2 between ST131, where CTX-M-15 and other CTX-M enzyme producers predominated, and the remaining isolates of group B2, which were mostly negative for CTX-M enzymes: 55 out of 70 ST131 isolates were CTX-M producers, whereas the other genotypes within group B2, 66 out of 75, were not ESBL producers (P < 10⁻⁴). Among the CTX-M enzymes, CTX-M-15 was clearly distributed differently (P = 0.0003) than the other enzymes. CTX-M-15 enzymes were found predominantly in group B2 (56%), whereas this group accounted for only 25.7% of the other CTX-M enzymes (P = 0.0007). In particular, ST131 was significantly associated with the production of CTX-M-15 rather than other CTX-M enzymes (P < 10⁻⁴).

Second, when ST131 was put aside, the remaining genotypes of group B2 were less associated with CTX-M enzyme production than the other E. coli groups were. When ST131 was excluded, the proportion of ESBL producers within B2 (9 [12%] of 75) was strikingly lower than the proportion in groups A, B1, and D (89 [55%] of 162) (P < 10⁻⁴). Notably, the frequent B2 group CC73 contained no CTX-M-producing isolate. These observations show that, with the remarkable exception of ST131 (and in fact, CC131, as the only isolate of ST1680, was a CTX-M-15 producer), group B2 as a whole is significantly less strongly associated with CTX-M production than the other E. coli groups are. Some heterogeneity was also observed within other groups. For example, within group A, CC10 showed a high proportion of CTX-M producers (Fig. 1).

Antibiotic susceptibility.

In each participating hospital, the susceptibility of isolates to amoxicillin or ampicillin, amoxicillin-clavulanic acid, cephalothin, cefotaxime or ceftriaxone, imipenem, and gentamicin was systematically tested (Table 3). Susceptibility to the other antibiotics was not systematically tested. However, in each participating hospital, the susceptibility to the same antibiotics was tested for each CTX-M producer and non-ESBL producer control pair. The percentage of resistant isolates was significantly higher among CTX-M producers than among non-ESBL producers for all of the antibiotics tested (P = 7.10⁻⁴ to <10⁻⁴), except for imipenem (no resistant isolate in either population) and fosfomycin (0% for CTX-M producers and 1% for non-ESBL producers; P = 0.2). The percentages of resistant isolates ranged from 14% (cefoxitin) to 100% (amoxicillin or ampicillin, ticarcillin, piperacillin, cephalothin, cefotaxime or ceftriaxone, ceftazidime, and cefepime) for CTX-M producers and from 0% (cefepime) to 51% (penicillins) for non-ESBL producers (Table 3).

Table 3.

Comparison of antibiotic susceptibilities of E. coli CTX-M producers versus non-ESBL producers and of ST131 versus non-ST131 strains among CTX-M producers and non-ESBL producers

Antibiotic	E. coli isolates intermediately susceptible or resistant
	No. of pairs tested	% of CTX producers	% of non-ESBL producers	P valuea	% of CTX producers		P valuea	% of non-ESBL producers		P valueb
ST131					ST131	Non-ST131		ST131	Non-ST131
Amoxicillin or ampicillin	152	100	51	<10−4	100	100	1	80	47	0.03
Amoxicillin-clavulanic acid	152	90	23	<10−4	93	89	0.6	33	22	0.3
Ticarcillin	148	100	50	<10−4	100	100	1	80	47	0.03
Ticarcillin-clavulanic acid	137	88	19	<10−4	91	86	0.4	27	18	0.4
Piperacillin	134	100	51	<10−4	100	100	1	91	48	0.009
Piperacillin-tazobactam	127	52	10	<10−4	67	44	0.015	10	10	1
Cephalothin	152	100	20	<10−4	100	100	1	27	19	0.5
Cefoxitin	129	14	2	0.0007	9	16	0.3	0	3	1
Cefotaxime or ceftriaxone	152	100	2	<10−4	100	100	1	0	2	1
Ceftazidime	148	100	2	<10−4	100	100	1	0	2	1
Cefepime	87	100	0	<10−4	100	100	1	0	0
Imipenem	152	0	0		0	0		0	0
Gentamicin	152	34	5	<10−4	31	38	0.6	20	4	0.03
Amikacin	148	22	1	<10−4	34	15	0.01	0	1	1
Cotrimoxazole	133	65	33	<10−4	43	77	<10−4	29	34	1
Nalidixic acid	147	78	29	<10−4	96	68	<10−4	73	24	<10−4
Ciprofloxacin	149	70	19	<10−4	94	57	<10−4	67	13	<10−4
Fosfomycin	118	0	1	0.15	0	0		14	1	0.1

^aCochran-Mantel-Haenszel paired test.

^bFisher exact test (adjusted for multiple comparisons).

Epidemiological and antimicrobial resistance characteristics of ST131.

Because ST131 accounts for a large fraction of CTX-M-producing E. coli strains, we investigated the distribution of this clone among the participating hospitals and among medical wards. CTX-M-producing ST131 isolates were identified at each of the 10 participating hospitals and distributed among various wards of each hospital, as were the non-ST131 CTX-M producers (Table 1). Considering wards (Table 2), among the CTX-M producer population, ST131 was distributed significantly differently among the seven wards (P = 2.10⁻⁴) than were the other STs. This could be attributed to the strong association of ST131 with geriatric/long-term care hospitalization (27% for ST131 versus 6% for other STs [P = 0.01]). Among the non-ESBL-producing E. coli strains, the distribution of ST131 was also globally distinct from that of the other STs (P = 0.02). The antibiotic susceptibility of ST131 isolates was compared with that of non-ST131 isolates in the two E. coli populations: CTX-M producers and non-ESBL producers (Table 3). Irrespective of the two E. coli populations, ST131 isolates were significantly more often resistant to quinolones (nalidixic acid and ciprofloxacin) than non-ST131 isolates were (P < 10⁻⁴). In contrast, the percentage of cotrimoxazole-resistant E. coli strains was significantly higher among non-ST131 E. coli isolates than among ST131 E. coli isolates when the CTX-M producers were compared (P < 10⁻⁴). Concerning the aminoglycosides, when a significant difference was observed (amikacin for CTX producers [P = 0.01] and gentamicin for the non-ESBL producers [P = 0.03]), the highest percentage of resistance always involved ST131 E. coli. With regard to the β-lactams cefoxitin and imipenem, which are classically not hydrolyzed by class A β-lactamases such as ESBL enzymes, there was no significant difference between ST131 and non-ST131 E. coli for the two E. coli populations. For the other β-lactam molecules, the analysis (which focused on the non-ESBL producers because those producing CTX-M were consequently resistant to these molecules) showed that ST131 E. coli was significantly more often resistant to penicillins (aminopenicillin, P = 0.03; ticarcillin, P = 0.03; piperacillin, P = 0.009) than non-ST131 E. coli was. The non-ESBL ST131 strains were not significantly more often hospital acquired than the non-ST131 non-ESBL producers (P = 0.16).

In contrast, with regard to clinical sources, the distribution of ST131 isolates was not significantly different from that of non-ST131 isolates within the CTX-M-producing and non-ESBL-producing E. coli strains (P = 0.3 and 0.2, respectively). Finally, we did not find that ST131 was significantly more prevalent than non-ST131 among the CTX producers isolated within the first 48 h of hospitalization (36% and 47%, respectively; P = 0.2).

DISCUSSION

In this study, we compared for the first time using a case-control format the clonal composition of CTX-M-producing and non-ESBL-producing E. coli clinical isolates. In addition, we determined the association of the different CTX-M enzymes with clonal groups and investigated the epidemiological features and antimicrobial susceptibilities of ST131, the major genotype of isolates producing CTX-M enzymes.

Both CTX-M-producing and non-ESBL-producing populations were polyclonal and were distributed across major phylogenetic groups A, B1, B2, and D in proportions that are consistent with previous studies (14, 19). The amount and nature of the clones found in this study are consistent with previous studies that showed the important contribution of ST131, CC10, and CC23 to the CTX-M-producing E. coli population (3, 11, 26).

This study demonstrates, for the first time using a case-control design, that the population of CTX-M-producing E. coli clinical isolates differs in clonal composition from non-ESBL-producing populations, with a few prevalent clones being clearly distributed nonrandomly in the two populations. In particular, within group B2, most of the CC131 isolates were CTX-M producers, whereas most CC95 isolates and all CC73 and ST141 isolates were not, as was the case for most of the remaining STs of phylogenetic group B2. As CTX-M-producing ST131 isolates were identified in each of the 10 participating hospitals and as a great variety and number of wards within each hospital were affected by ST131 isolates, it can be expected that the situation observed in these hospitals reflects, at a local level, the global ST131 clonal diversity rather than local cross-transmission. Although CTX-M-15 was predominant in ST131, there is clear evidence of several independent acquisitions of distinct CTX-M types by strains of ST131, as found previously (27).

The presence of CTX-M enzymes in all major phylogenetic groups reflects the wide dissemination of these enzymes across the diversity of E. coli strains. However, a striking observation of our study is an apparent restriction against the acquisition of CTX-M enzymes by members of group B2, since most group B2 genotypes were devoid of CTX-M enzymes, except for ST131 (Fig. 1). To our knowledge, this is the first demonstration that ST131 is more often associated with CTX-M enzymes than the other STs of group B2. In turn, non-ST131 B2 isolates were clearly less often associated with CTX-M production than other E. coli groups were. Three hypotheses might explain these observations. First, group B2 might be less exposed to antibiotics than groups A, B1, and D, which are predominant in the fecal E. coli population of healthy subjects (9, 20). Second, among the group B2 fecal E. coli isolates, those belonging to ST131 might display features different from those of non-ST131 group B2 fecal E. coli isolates. It has to be noted that ST131 has been shown to be the most prevalent clone among group B2 isolates in the dominant fecal E. coli population of healthy subjects (20). Thus, ST131 E. coli, similar to E. coli of non-B2 groups, might be more exposed to antibiotics. Third, clone ST131 might differ in some important biological properties from other group B2 E. coli isolates. Notably, ST131 was shown to belong, within phylogenetic group B2, to subgroup I (5). Interestingly, this subgroup was suggested to have a basal position within the diversity of group B2 isolates (22). Although incompletely resolved, our phylogeny (Fig. 1) is consistent with this proposal, as ST131 and closely related STs ST429, ST555, ST1680, and ST1650 formed one of the most basal branches of group B2. This basal position could imply that characteristics common to non-ST131 B2 strains could have evolved after the evolutionary separation of ST131 and related genotypes. This independent evolution could have resulted, for example, in intrinsic differences in the ability of ST131 to acquire the mobile elements that carry CTX-M-encoding genes. Thus, further studies are required to explore the greater propensity of ST131 than non-ST131 group B2 E. coli isolates to harbor CTX-M enzymes.

Another new feature of ST131, shown in this study by comparing non-ESBL ST131 isolates with non-ESBL, non-ST131 isolates, is the significantly higher percentage of ST131 isolates resistant to penicillins. In addition, ST131 was commonly resistant to fluoroquinolones, as previously reported (16). These observations add to the phenotypic resemblance of ST131 to non-B2 isolates, rather than to other group B2 members. Interestingly, ST131 has already been shown to differ from classical extrapathogenic group B2 E. coli strains with regard to its potential virulence, as assessed by the absence from ST131 of several virulence factors that are common in other B2 strains (5, 18) and by its reduced capacity to induce death or infection in different animal models (17, 18).

With regard to the clinical epidemiology of ST131, we noted that CTX-M-producing ST131 E. coli isolates were significantly associated with patients hospitalized in geriatric wards or long-term care facilities. In contrast, this characteristic was not identified for all CTX-M-producing E. coli isolates compared with non-ESBL-producing isolates (24). In this previous study, we showed that living in collective housing was a factor associated with a CTX-M-producing clinical E. coli isolate. Consistently, nursing homes in Ireland were shown to be the reservoir of ST131 isolates that produce CTX-M-15 (28). A large proportion of the patients hospitalized in geriatric wards and long-term care facilities may come from collective housing such as nursing homes. Therefore, we hypothesize that French nursing homes, similarly to Irish nursing homes, might be a reservoir of CTX-M-producing ST131 bacteria. The absence of significant dominance of ST131 bacteria in the others wards (intensive care unit and medical wards) and the absence of a difference between the proportions of ST131 among the hospital-acquired and the imported CTX-M producers reinforce the hypothesis of reservoirs of CTX-M-producing ST131 E. coli outside the hospital.

In conclusion, the present study emphasizes the particular characteristics of B2 clone ST131 with regard to its epidemiology, carriage of genes that code for CTX-M enzymes, and antimicrobial susceptibility. We hypothesize that the latter features are due to an ecological adaptation of ST131 to the gut. As clone ST131 is the predominant clone of group B2 in the human gut (9, 20), it should, as a consequence, be more exposed to antimicrobial agents. In contrast, non-ST131 members of group B2 carry CTX-M enzymes much less frequently than other E. coli groups do, which may result from a lower exposure to antimicrobial agents due to their lower frequency of carriage in the gut.