Identification of CTX-M-Type Extended-Spectrum-{beta}-Lactamase Genes Using Real-Time PCR and Pyrosequencing

CTX-M extended-spectrum ß-lactamases (ESBLs) are increasinglyprevalent worldwide among Escherichia coli bacteria, mostlyin community-acquired urinary tract infections. Finding a fastand reliable technique for identification of CTX-M enzymes isbecoming a challenge for the microbiology laboratory. A fastreal-time PCR amplification technique, using degenerated primersspecific for all the bla_CTX-M alleles, coupled to real-timepyrosequencing was developed. The five CTX-M groups were unambiguouslyidentified by pyrosequencing a 13-bp DNA region. Further sequencingof an additional 16-bp region allowed further division intosubgroups. Phylogenetic trees constructed with the entire bla_CTX-Mgenes and with both pyrosequenced regions (29 bp) gave similarresults, suggesting that this technique, termed the real-timedetection and sequencing method, has a powerful discriminatoryability. This high-throughput technique has been evaluated byscreening 48 ESBL-producing E. coli isolates recovered fromthe Bicêtre hospital (France) in 2004. Forty-four of thesestrains were CTX-M positive by real-time PCR detection and directpyrosequencing of the PCR products, which identified CTX-M-15as the main CTX-M-type ß-lactamase. Pulsed-field gelelectrophoresis analysis of these strains revealed that severalclones, of which one CTX-M-15-positive clone was predominant(60%), were identified both in nosocomial and in community-acquiredisolates. The combination of real-time PCR with pyrosequencingrepresents a powerful tool for epidemiological studies of CTX-Mproducers. This assay has the potential to be used in a diagnosticlaboratory since up to 96 bacterial isolates may be screenedin less than 3 h.

INTRODUCTION

Top

INTRODUCTION

Extended-spectrum ß-lactamases (ESBLs) of the CTX-Mtype have been reported increasingly in gram-negative rods,mostly in Escherichia coli (4, 5, 9, 13, 17, 24, 31, 42). Thefirst CTX-M ß-lactamase (CTX-M-1/MEN-1) was characterizedin E. coli strains isolated from German and Italian patients(2, 3). More than 50 variants, sharing 71 to 98% amino acidsequence identities, have been described and are divided nowinto five phylogenetic groups based on amino acid sequence identity.They are the CTX-M-1 group (including 22 members), the CTX-M-2group (including 10 members), the CTX-M-8 group (including 3members), the CTX-M-9 group (including 18 members), and theCTX-M-25 group (with 4 members). The prevalences of these differentgroups are variable depending on the geographical localization(31).

ß-Lactamases of the CTX-M groups are structurallyrelated to the naturally produced ß-lactamases ofvarious Kluyvera species (8, 14, 29, 31, 32). The CTX-M enzymesusually have higher activities against cefotaxime than ceftazidime,but several enzymes, such as CTX-M-15 and CTX-M-19, also hydrolyzeceftazidime efficiently (15, 31, 33), which may complicate thephenotypic identification of CTX-Ms. The bla_CTX-M ß-lactamasegenes are mostly located on plasmids, but Toho-1-like ß-lactamasegenes have also been identified on chromosomes (4).

Apart from standard PCR and gene sequencing, molecular detectiontechniques for ESBLs in general have been used for a few applicationsmost suited for the TEM- and SHV-type ESBLs that were mostlyprevalent among nosocomial isolates and not community-acquiredpathogens (1, 4, 6, 20, 23, 27, 28, 35, 41). Recent innovationsin real-time PCR have simplified PCR considerably. Pyrosequencingis a genetic analysis method based on the detection of pyrophosphatethat is released during the synthesis of DNA, thus allowingreal-time sequence determination (36). Pyrosequencing that allowsfast identification of short DNA sequences (up to 100 bp [36])has been used already for detection of resistance-conferringgenes in various bacteria (10, 12, 39, 44).

This study describes the first application of the LightCyclerreal-time PCR assay for detection of CTX-M variants, combinedwith real-time pyrosequencing for discrimination between thedifferent bla_CTX-M genes in clinically relevant gram-negativespecies. By use of this combined technology, the clusters ofCTX-M enzymes were accurately identified among a collectionof 48 ESBL-positive E. coli clinical isolates. Further divisionwithin a given CTX-M cluster into pyrosequencing subgroups wasalso possible.

MATERIALS AND METHODS

Top

MATERIALS AND METHODS

Bacterial strains. The following well-characterized CTX-M producer strains wereused as controls to develop and optimize the novel detectionmethod: Salmonella enterica serovar Typhimurium NC (bla_CTX-M-1)(17), E. coli Jab (bla_CTX-M-2) (18), Enterobacter aerogenesKer (bla_CTXM-3) (18), Kluyvera georgiana CUETM 4246-74 (bla_CTX-M-8)(32), Enterobacter cloacae Rms1 (bla_CTX-M-9) (gift from C. Deschamps),E. coli Mai (bla_CTX-M-14) (18), E. coli Urh (bla_CTXM-15) (18),E. coli Rms2 (bla_CTX-M-16) (Gift from C. Deschamps), Klebsiellapneumoniae ILT-3 (bla_CTX-M-19) (34), and E. coli ESBL530 (bla_CTXM-25)(22). Four Klebsiella oxytoca clinical isolates from the Bicêtrehospital with chromosomal ß-lactamases related toCTX-M (e.g., KOXY/K1) were used as controls. E. coli DH10B wasused as a CTX-M-negative control strain. Forty-eight ESBL-producingE. coli isolates were identified in clinical samples (one isolateper patient) during 2004 at the Bicêtre hospital (Le Kremlin-Bicêtre,France) (Table 1).

View this table:
[in this window]
[in a new window]

TABLE 1. Amplification and pyrosequencing primers for the bla_CTX_-M-like genes used in this study

Antimicrobial agents and susceptibility testing. Routine antibiograms were determined by the disk diffusion methodon Mueller-Hinton agar (Bio-Rad, Marnes-La-Coquette, France),and the susceptibility breakpoints were as recommended by theClinical and Laboratory Standards Institute (6) and by the FrenchSociety for Microbiology guidelines (SFM) (http://www.sfm.asso.fr/nouv/general.php?pa=2).The presence of ESBLs was inferred by a synergy image, usingthe double-disk synergy test that was performed with cefotaximeor ceftazidime and ticarcillin-clavulanic acid disks (26).

Nucleic acid extractions. Whole-cell DNAs were extracted either from standard strainsusing a QIAamp DNA mini kit (QIAGEN, Les Ulis, France) or fromclinical isolates and standard strains by a boiling extractprocedure, using a few colonies of each bacterial strain resuspendedin 100 µl of distilled water. After heating at 100°Cfor 10 min, freezing at –80°C for 10 min, and boilingfor five additional min, the suspensions were centrifuged (5min, 10,000 x g) and the recovered supernatant was frozen at–20°C until use.

PCR and Sanger DNA sequencing. PCR was performed on an ABI 2700 thermocycler, using laboratory-designedprimers for detection and sequencing of the entire bla_CTX-M-likegenes, as described previously (16, 18, 38) (Table 1). Two microlitersof boiling extract supernatant was used as a template. PCR experimentswere performed with 35 cycles consisting of 45 s of denaturationat 94°C, 45 s of annealing at 57°C, and 60 s of extensionat 72°C. Both strands of the PCR products were sequencedwith an Applied Biosystems sequencer (ABI 3100). The nucleotideand deduced amino acid sequences were analyzed and comparedto sequences available over the Internet at the National Centerfor Biotechnology Information website (http://www.ncbi.nlm.nih.gov).

Sequence alignment and phylogenetic tree construction. Multiple sequence alignments were performed online at the EBIbioinformatics website (http://www.ebi.ac.uk/clustalw/) usingthe program ClustalW, which calculates the best match for theselected sequences. The evolutionary relationships between sequences,based on the phylogenetic calculations of the neighbor-joiningmethod of Saitou and Nei (37), have been displayed as phylograms,where the branch lengths are proportional to the amounts ofinferred evolutionary change.

Real-time PCR. The oligonucleotide primers for PCR targeted two regions ofCTX-M genes. The primers and probes were obtained from SigmaProligo (Saint-Quentin Fallavier, France). The sequences forthe primers are shown in Table 1 and Fig. 1. Primer pairs CTX-M-A1bio/CTX-M-A5and CTX-M-A8bio/CTX-M-A6 have been designed in order to amplifytwo fragments of 224 bp and 175 bp, respectively, from eachof the known bla_CTX-M genes (Table 1 and Fig. 1). The CTX-M-A1bioand CTX-M-A8bio primers were biotin labeled at their 5' ends.All the primers and probes for real-time PCR and pyrosequencingwere purified by high-performance liquid chromatography (SigmaProligo).

View larger version (72K):
[in this window]
[in a new window]

FIG. 1. Alignment of the nucleotide sequences of genes encoding ß-lactamases CTXM-1, CTXM-2, CTXM-8, CTXM-9, and CTXM-25 based on the Lahey website (http://www.lahey.org/Studies/other.asp#table%201). PCR amplification primers (CTXM-A1bio, CTXM-A5, CTXM-A6, and CTXM-A8bio) and sequencing primers (S1 and S2) are indicated with open boxes, and the sense and antisense directions are indicated with arrows. Double slashes have been introduced into the sequences in order to show sequence interruptions. The sequenced regions are indicated in gray letters, and the polymorphic nucleotides in that region are highlighted. Nucleotide numbering of the bla_CTX-M genes is indicated on the right-hand side of the nucleotide sequences.

PCR was performed using the LightCycler 2.0 system (Roche Diagnostics,Meylan, France) by rapid cycling in a reaction volume of 20µl with amplification primers, each at a final concentrationof 0.5 µM. LightCycler-FastStart DNA Master plus SYBRgreen I (Roche) was used. To complete the PCR mixtures, 15 µlof the master mixture and 5 µl of a DNA preparation wereloaded into glass capillary cuvettes (Roche). After a shortcentrifugation (3,000 x g for 10 s), the sealed capillarieswere placed into the LightCycler rotor.

The reaction protocol was as follows: an initial FastStart TaqDNA polymerase activation phase at 95°C for 10 min; a 45-cycleamplification phase consisting of a 95°C denaturation segmentfor 10 s, a 57°C annealing segment for 10 s, and a 72°Cextension segment for 10 s; a melting phase from 55 to 95°Cwith a temperature transition rate of 0.1°C/s; and a rapidcooling phase. The quantity of the amplified product was monitoredby detection of fluorescence energy emitted at 530 nm by SYBRgreen.

Melting curve analysis. The LightCycler system combines real-time PCR with a postamplificationmelting point analysis. The ability of the LightCycler systemto perform melting curve analyses also allowed confirmationof the specificity of the amplified product and allowed fordifferentiation of CTX-M groups. The melting curve for eachspecimen was analyzed manually to determine the melting temperature(T_m). The T_m is the peak of the curve of the derivative of fluorescence(dF) with respect to temperature (dF/dT). By using the manualT_m function of the LightCycler software, the T_m was definedas the temperature at which the cursor just covered the highestpoint on the melting curve (the dF/dT curve). If the highestpoint was a plateau, the midpoint of the plateau was taken tobe the T_m.

Sample recovery and control analysis of the amplification product. In order to verify the sizes of selected amplification productsand the numbers of products for both amplification primer pairs,the capillaries were opened after the run in the LightCyclerinstrument and placed upside down in Eppendorf tubes. Aftera brief centrifugation, 5 µl of each sample was analyzedby 1.5% agarose gel electrophoresis (38).

Real-time pyrophosphate sequencing. Pyrosequencing was performed by using a PSQ 96 sample preparationkit and a PSQ 96MA analyzer (AB Biotage, Uppsala, Sweden), followingthe manufacturer's instructions. Unpurified amplified productsrecovered after real-time PCR were captured and separated byusing streptavidin-Sepharose beads, and the resulting single-strandedDNA was used as a template for pyrosequencing with appropriatepyrosequencing probes (Fig. 1). The overall experiment is basedon a protocol detailed on the Pyrosequencing Inc. website. Briefly,in pyrosequencing, a sequencing primer is annealed to a single-strandedPCR product, and nucleotides are added to the reaction mixture.Incorporation of the nucleotide by DNA polymerase leads to therelease of pyrophosphate, which is further processed by sulfurylaseand luciferase, producing light in proportion to the amountof pyrophosphate. The amount of light released at each extensionstep is presented as a pyrogram, and the relative numbers ofa given nucleotide are consequently reflected by the relativepeak heights on data traces. Excess nucleotides are enzymaticallydegraded before the following nucleotide is added.

PFGE. Whole-cell DNA from bla_CTX-M-producing E. coli isolates wasanalyzed by pulsed-field gel electrophoresis (PFGE) using XbaI(Amersham Biosciences, Les Ulis, France) with a CHEF DRII apparatus(Bio-Rad) as previously described (25). XbaI macrorestrictionpatterns were digitized and analyzed using Taxotron software(Institut Pasteur, Paris, France) and interpreted accordingto Tenover et al. (40).

RESULTS

Top

RESULTS

Real-time PCR detection of bla_CTX-M genes in control strains. Fluorescence monitoring of product accumulation and detectionby SYBR green of the bla_CTX-M genes were achieved during DNAamplification. The fluorescence signals rose above the backgroundlevels after about 20 cycles, when ca. 50 ng of QIAGEN kit-extractedwhole-cell DNA was used as a template. No increase in the fluorescencesignal was observed with the four K. oxytoca clinical isolatesor with the negative controls, i.e., the E. coli DH10B isolateand the Tris-EDTA solution, during the whole run of 45 cycles(data not shown).

The melting curves for the amplification products with the primerpair CTX-M-A1bio and CTX-M-A-5 were systematically analyzed.The peaks obtained after mathematical transformation were 86.5°Cfor bla_CTX-M-3/15, 86.9°C for bla_CTX-M-1, 87.5°C forbla_CTX-M-8, 87.8°C for bla_CTX-M-25, 88.2°C for bla_CTX-M-2,and 88.5°C for bla_{CTX-M-9/14/16/19}. Even though very slight,these differences have been obtained reproducibly. Similar resultswere obtained with primer pair CTX-M-A6 and CTX-M-A8bio (datanot shown). Additionally, monitoring by gel electrophoresis(data not shown) revealed one major amplification product ofhigh intensity and expected length for each amplification primerpair (224 bp for CTX-M-A1bio/CTX-M-A5 and 175 bp for CTX-M-A6/CTX-M-A8bio).These products were absent from the K. oxytoca isolates, theE. coli DH10B isolate, and the Tris-EDTA negative control (datanot shown).

Real-time sequencing. The high intensities of the PCR products obtained from eachamplification primer pair were consistent with a use for pyrosequencing.The entire 20 µl of each PCR product was used for pyrosequencingtemplate preparation, without initial purification. The amountof light released at each extension step is directly proportionalto the amount of nucleotide added, the relative numbers of agiven nucleotide being consequently reflected by the relativepeak heights. Since the amplification product was of high intensity,the emitted light at each extension step was also of high intensityand clearly above the background level. In a few minutes, pyrosequencingdistinguished among the genes encoding the five groups of CTX-Mß-lactamases used as controls. The pyrograms obtainedwith three control CTX-M producer strains (CTX-M-1, CTX-M-3,and CTX-M-15) are illustrated in Fig. 2. Using the second sequencingprimer, the S2 primer, differences in nucleotide sequences withina given group were obtained, especially within the CTX-M-1 group,allowing differentiation of the genes encoding CTX-M-1-, CTX-M-3-,and CTX-M-15-type enzymes (Fig. 2).

View larger version (43K):
[in this window]
[in a new window]

FIG. 2. (Left panels) Pyrosequencing of PCR products obtained with primers CTX-M-A1bio and CTX-M-A5 and DNAs extracted from standard strains bla_CTX-M-1 (A), bla_CTX-M-3 (C), and bla_CTX-M-15 (E), using the S1 sequencing primer. (Right panels) PCR products obtained with CTX-M-A6 and CTX-M-A8bio and DNAs extracted from standard strains bla_CTX-M-1 (B), bla_CTX-M-3 (D), and bla_CTX-M-15 (F), using the S2 sequencing primer. The scale on the left-hand side indicates the relative levels of fluorescence emitted during the pyrosequencing reaction. The corresponding nucleotide sequence is indicated above the obtained pyrogram for the bla_CTX-M-1 sequence (A and B). For the remaining programs, only nucleotides different from those in the bla_CTX-M-1 sequence are indicated.

In order to evaluate the discriminatory power of pyrosequencing,the 29 pyrosequenced nucleotides from each of the known bla_CTX-Mgenes have been aligned (Fig. 3). It is obvious from this alignmentthat pyrosequencing may rapidly distinguish between the fiveCTX-M groups but may also extend CTX-M differentiation withinindividual CTX-M groups. As an example, CTX-M group 1 may befurther divided into eight pyrosequencing subgroups (named Ato H). Each of these subgroups may contain one or several bla_CTX-Mgenes. Pyrosequencing 29 bp has of course less discriminatorypower than sequencing the entire gene. However, in order tocompare both techniques, phylogenetic trees were constructedusing either the two pyrosequenced regions, representing 29nucleotides, or the entire gene for all the bla_CTX-M gene sequences(Fig. 4) available at the Lahey Clinic (http://www.lahey.org/Studies)and/or the GenBank database (http://www.ncbi.nlm.nih.gov). Theobtained phylogenetic trees were very similar and grouped theß-lactamases in the same general manner (Fig. 4),suggesting that by pyrosequencing as few as 29 nucleotides,high discrimination not only among the different groups butalso within a given group can be achieved.

View larger version (65K):
[in this window]
[in a new window]

FIG. 3. Nucleotide sequence alignment of 57 bla_CTX-M genes covering only the two pyrosequenced regions, S1 and S2. The orientations of the sequencing primers are indicated above the nucleotide sequences with arrows. Sequences whose GenBank accession numbers are different from the numbers assigned to them at the Lahey website (http://www.lahey.org/Studies/other.asp#table%201) are represented with a superscript "a". Boldfaced names represent sequences present in the GenBank database and absent from the Lahey website. Accession numbers are indicated. Polymorphic nucleotides are highlighted in gray. The corresponding pyrosequencing subgroups are indicated in the right-hand column.

View larger version (12K):
[in this window]
[in a new window]

FIG. 4. Dendrogram obtained for 57 CTX-M enzymes by ClustalW analysis (37) of the two pyrosequenced regions, corresponding to 29 bp (A) and the entire bla_CTX-M nucleotide sequence (B). Branch lengths are drawn to scale and are proportional to the numbers of changes.

Screening of ESBL-producing E. coli clinical isolates. In 2004, out of 4,200 clinical samples containing Enterobacteriaceaeat the Bicêtre hospital in France, 86 (2.1%) had an ESBLphenotype according to the results of the double-disk antibioticsusceptibility tests. Since E. coli was by far the most prevalentESBL-producing species (56%) and since CTX-M-type enzymes aremostly found in that species (4, 31), our work focused exclusivelyon E. coli. These 48 ESBL-producing E. coli isolates, representing2% of the E. coli isolates (one E. coli isolate per patient)identified from clinical samples taken in 2004, were recoveredmostly from urine (86%) and were isolated in locations scatteredthroughout the entire hospital, most predominantly in the gerontology(40%) and the emergency (21%) units (Table 2).

View this table:
[in this window]
[in a new window]

TABLE 2. Screening of bla_CTX_-M-like genes in ESBL-producing E. coli clinical isolates from the Bicêtre hospital isolated in 2004

LightCycler PCR revealed that 44 out the 48 ESBL-producing E.coli isolates harbored a bla_CTX-M gene (Table 2). Using pyrosequencing,the CTX-M enzymes detected belonged to three of the five CTX-Mclusters, i.e., the CTX-M-1, CTX-M-2, and CTX-M-9 clusters,and to five pyrosequencing subgroups, i.e., CTX-M-1(C) (n =37), CTX-M-1(E) (n = 1), CTX-M-2(A) (n = 1), CTX-M-9(A) (n =1), and CTX-M-9(B) (n = 4) (Table 2). Using the LightCyclertechnology, up to 30 isolates were screened in one run of 50min; thus, two runs were necessary to screen 48 isolates withone primer pair, and two additional runs were necessary forthe second primer pair. The 44 CTX-M-positive E. coli isolateswere genotyped with both pyrosequencing primers in less than1 hour (30 min for sample preparation and 20 min for the pyrosequencingrun). Overall, it took about 3 h to analyze 44 ESBL-producingE. coli isolates with one primer pair and less than 4 h withboth primer pairs; since pyrosequencing allows up to 96 sequencingreactions in a single run, both PCR products from the 44 isolatescould be pyrosequenced simultaneously.

The pyrosequencing-determined DNA sequence matched the DNA sequenceobtained with classical Sanger DNA sequencing (Table 2). Resultsobtained with classical PCR followed by Sanger sequencing showedCTX-M cluster distributions identical to those obtained usingthe CTX-M real-time detection and sequencing method (Table 2).Similarly, the five different types of CTX-M enzymes detected,i.e., CTX-M-15 (n = 37), CTX-M-1 (n = 1), CTX-M-2 (n = 1), CTX-M-27(n = 1), and CTX-M-14 (n = 4), matched perfectly with the fivedefined pyrosequencing subgroups, being CTX-M-1(C) (n = 37),CTX-M-1(E) (n = 1), CTX-M-2(A) (n = 1), CTX-M-9(A) (n = 1),and CTX-M-9(B) (n = 4), respectively (Table 2).

Epidemiology of ESBL-producing E. coli clinical isolates. Genotyping by PFGE revealed that 60% of CTX-M-producing E. coliisolates were clonally related. Whereas most isolates belongedto several subtypes of a given clone, differing by only oneor two bands, several strains were unrelated (Fig. 5). Whilemost of the CTX-M-15 isolates belonged to a single clone (Fig.5) and were mostly found in the gerontology unit, where theyseem to be epidemic, some clonally related isolates were foundin the entire hospital and even in the emergency ward, wherethey were isolated from patients coming directly from the community.

View larger version (46K):
[in this window]
[in a new window]

FIG. 5. Digitized PFGE patterns and phylogenetic tree of 44 CTX-M-producing E. coli isolates. XbaI macrorestriction patterns were digitized and analyzed using Taxotron software (Institut Pasteur, Paris, France) to calculate Dice coefficients of correlation and to generate a dendrogram by the unweighted pair group method using arithmetic averages. The scale is proportional to the level of pattern similarity. PFGE results were interpreted according to the criteria of Tenover et al. (40). The wards are indicated on the right.

DISCUSSION

Top

DISCUSSION

Although ESBL production may lead to failure of treatment withexpanded-spectrum cephalosporins, the MICs of expanded-spectrumcephalosporins for ESBL producers may be increased only slightlycompared to those for non-ESBL producers (30). Consequently,clinical detection of ESBLs in clinical isolates is difficult(19, 30). Indeed, in a European survey, up to 33% of ESBLs inEurope may remain undetected (19). A series of easy-to-carry-outtests, mostly based on synergy between clavulanic acid and expanded-spectrumcephalosporins, are recommended, but their main pitfall is alimited sensitivity and the requirement of an overnight incubation.Detection of ESBLs at the genetic level represents an interestingalternative, which is independent of the degree of gene expression(7). A few methods that had been designed for TEM and SHV variants,such as oligotyping of bla_TEM genes (20), single-strand conformationalpolymorphism analysis-PCR, and PCR-restriction fragment lengthpolymorphism for bla_SHV genes (23, 28), have now been overwhelmedby the diversity of types recognized since their first description.The disadvantages of the molecular-biology-based methods, suchas labor expense, costliness, and the lack of general applicability,have outweighed their advantages and have so far prevented theirbroad acceptance. Real-time PCR coupled to melting curve analysishas been developed for several SHV- and GES-type ESBLs (35,41). These techniques are interesting for investigating single-nucleotidemismatches within a sequence but are inappropriate for the studyof a highly divergent family of ß-lactamases, suchas the CTX-Ms. Very recently, a multiplex PCR for detectionof bla_CTX-M genes was able to discriminate between the fiveCTX-M-clusters based on amplicon size analysis after agarosegel electrophoresis of the amplicons (43).

The method presented here is a rapid tool for identificationof bla_CTX-M-type genes based on real sequence data, which isthe gold standard for genetic information and thus the bestassurance of a correct test. The PCR assay developed includesa first primer pair able to detect the known bla_CTX-M genesemerging in various countries. Pyrosequencing performed as acomplementary step allows the identification of the subgroupof the CTX enzyme in a few minutes. Pyrosequencing of the PCR-generatedproduct using the second primer pair allows further enzyme differentiation.DNA obtained by boiling extraction of bacterial colonies fromplates yielded high-intensity PCR products, consistent witha direct use for pyrosequencing, suggesting that accurate detectionof bla_CTX-M genes is possible regardless of the vastly variousamounts of template DNA available from clinical isolates. Furthermore,the presence of minor amplification products in some amplificationreactions did not interfere with the subsequent sequencing reaction(data not shown).

The rate of ESBL-producing enterobacterial isolates (2.1%) atthe Bicêtre hospital has increased threefold since 2002,when it was only 0.7% (16). Similarly, the frequency of ESBLin E. coli has increased fivefold since 2002, when it was 0.38%(16). The CTX-M enzymes detected were of five different types,i.e., CTX-M-15 (n = 37), CTX-M-1 (n = 1), CTX-M-2 (n = 1), CTX-M-27(n = 1), and CTX-M-14 (n = 4) (Table 2), belonging to threemain CTX-M clusters, i.e., the CTX-M-1, CTX-M-2, and CTX-M-9clusters. The bla_CTX-M-15 gene was predominant (37/44 CTX-Mproducers) in ESBL-producing E. coli isolates recovered fromclinical samples at Bicêtre hospital, a situation thatmirrors the current trend observed in other countries (4, 16,31, 42). Again, the bla_CTX-M genes were always associated withinsertion sequence ISEcp1-like elements located upstream ofthe ß-lactamase gene (data not shown), which are responsiblefor their expression and mobilization (33). In the present study,CTX-M enzymes represented 92% of the ESBLs produced by E. coli(44 of 48), which have increased significantly since 2002, whenthe frequency was only 60%, the other ESBLs being TEM derivatives(data not shown).

The molecular epidemiology of the CTX-M-producing E. coli isolatesin our institution was complex. Most CTX-M-15 isolates wereepidemiologically related and were sufficiently similar to beconsidered subtypes of the same clone. One CTX-M-15 producerstrain accounted for 60% of the isolates recovered from locationsscattered throughout the hospital, most of which were isolatedin the gerontology ward. Clonal diffusion of several pulsotypeswas observed, likely the result of multiple importations ofseveral unrelated clones from the community, as suggested fromthe observed diversity among strains from the Parisian area(9). The emergence of multiple CTX-M producers from the community,mostly observed here in the emergency unit, justifies by itselfthe rapid identification of CTX-M producers for subsequent isolationof strains from patients. Further development of this techniqueusing clinical samples (such as urine) may help to further reducethe time for identification of CTX-M producers. Furthermore,multiplex PCR (both amplification primer pairs in one tube)followed by pyrosequencing may also reduce the overall timeof the experiment to 3 hours. In addition, by coupling thistechnique with a rapid TEM and SHV PCR approach, most (if notall) of the currently spreading ESBLs in France may be identified.

In conclusion, ease, speed, and reliability render the real-timedetection and sequencing method a powerful tool for epidemiologicalsurveys concerning CTX-M producers and make it an interestingcandidate for implementation into routine diagnostics. In geographicalareas, such as in the United Kingdom, Spain, and France, wheredifferent CTX-M variants are known to coexist with a quite highprevalence (9, 13, 16, 21, 42), it might be useful to be ableto distinguish these enzymes rapidly, avoiding the usual sequencingdelays associated with classical methods. In addition, the pyrosequencingtechnique may also provide a rapid tool for determination ofthe molecular mechanisms of multidrug-resistant strains.

ACKNOWLEDGMENTS

We thank F. Lartigue and N. Fortineau for helpful discussions.We are grateful to S. Marin, J. Hogg, and R. England from BiotageIndustry for technical support of this work.

This work was funded by a grant from the Ministère del'Education Nationale et de la Recherche (UPRES-EA3539), UniversitéParis XI, France, and mostly by the European Community (6thPCRD, LSHM-CT-2005-018-705).

FOOTNOTES

* Corresponding author. Mailing address: Service de Bactériologie-Virologie, Hôpital de Bicêtre, 78 rue du Général Leclerc, 94275 Le Kremlin-Bicêtre cedex, France. Phone: 33-1-45-21-29-86. Fax: 33-1-45-21-63-40. E-mail: thierry.naas{at}bct.ap-hop-paris.fr.

Published ahead of print on 6 November 2006.

REFERENCES

Top