Rapid Genotyping of CTX-M Extended-Spectrum β-Lactamases by Denaturing High-Performance Liquid Chromatography

Abstract

Denaturing high-performance liquid chromatography (dHPLC) is a powerful technique which has been used extensively to detect genetic variation. This is the first report of the application of dHPLC for rapid genotyping of bacterial β-lactamase genes. The technique was specifically developed to genotype members of all bla_CTX-M DNA homology groups. Thirteen well-defined bla_CTX-M extended-spectrum β-lactamase (ESBL)-producing strains were used to develop and optimize the dHPLC genotyping assay. Further evaluation was carried out with a blinded panel of 62 clinical isolates. The results of bla_CTX-M genotyping achieved by dHPLC were comparable to the typing results obtained by DNA sequencing. Applying the newly developed dHPLC-based genotyping method, we successfully genotyped all 73 bla_CTX-M ESBL-producing strains from the 4-month survey study. Furthermore, we found the first reported cases in the United Kingdom of clinically significant disease caused by CTX-M-14- and CTX-M-1-producing Escherichia coli strains. We conclude that the novel dHPLC assay is highly accurate, rapid, and cost-effective for the genotyping of bla_CTX-M-producing ESBLs and has great potential for determining the clinical relevance of different and new bla_CTX-M genotypes, as well as for epidemiological studies and surveillance programs.

The emergence of extended-spectrum β-lactamases (ESBLs) among both hospital- and community-acquired gram-negative bacteria has become a serious problem worldwide (4). The CTX-M-type ESBLs, which are non-TEM and non-SHV derivatives, represent a new and rapidly growing family of molecular class A ESBLs (4). On the basis of amino acid sequence similarities, they have been classified into five distinct phylogenetic groups: groups 1, 2, 8, 9, and 25/26. To date, more than 50 CTX-M β-lactamases have been characterized (http://www.lahey.org). In the United Kingdom, although the first report was of a CTX-M-9-producing isolate (group 9) of Klebsiella oxytoca in Leeds in 2000 (1), the first outbreak involved CTX-M-26 (group 25/26)-producing Klebsiella pneumoniae at City Hospital, Birmingham (7). In late 2002 a sharp increase in ESBL-producing isolates of Escherichia coli from distant parts of the United Kingdom was noted, together with a major outbreak in Shropshire (3). Interestingly, a survey of fecal carriage of CTX-M-type ESBL-producing bacteria from York showed a diversity of molecular subtypes, including CTX-M-9, -14, and -15 (34). The first isolation of CTX-M-3 (group 1), CTX-M-17 and -18 (group 9), and CTX-M-8 and -40 (group 8) β-lactamases has also recently been reported in the United Kingdom (3, 16, 35).

The diversity and increasing prevalence of CTX-M-type ESBLs presents an increasingly complex molecular epidemiological scenario in the United Kingdom (23). While CTX-M-15 has been dominant, we have noted an increasing number of the genotypes, particularly CTX-M-14 (Li Xu, unpublished observation). There is, therefore, a pressing need for high-throughput, low-cost methods for the identification of specific genotypes of CTX-M β-lactamase genes. Various molecular methods, including multiplex PCR, have successfully been applied to the detection of all members of all five groups (36, 39). However, without full DNA sequencing, precise genotype characterization of CTX-M-type β-lactamase genes remains a challenge. Currently, sequencing is the only method for definitive identification, but it is time-consuming and expensive.

Denaturing high-performance liquid chromatography (dHPLC) is a relatively new, powerful technique for the detection of genetic variation. The principle behind this technology has been described in detail elsewhere (38). It has successfully been used extensively for the analysis of mutations in disease-related genes (13, 20). In recent years, the technique was first developed for bacterial identification and molecular typing and has subsequently been applied to the detection of a variety of mutations in antibiotic resistance-conferring genes, such as mutations in gyrA and grlA in Staphylococcus aureus, Salmonella, Yersinia pestis, and Neisseria gonorrhoeae (10, 14, 17). The identification of mutations associated with antituberculosis drug resistance in rpoB, katG pncA, rpsL, and embB has also been reported (9, 32). The aim of our study was to establish a sensitive, rapid, and high-throughput genotyping assay for the screening of CTX-M-type ESBLs. The dHPLC genotyping method consists of three steps: separate amplification of both reference and clinical isolates of DNA by multiplex PCR, heteroduplex formation between reference and sample PCR products, and analysis of heteroduplex DNA by dHPLC. We first developed the technique with bla_CTX-M-producing control strains from the four groups and then evaluated the assay with a blinded panel of 62 previously characterized clinical isolates (39). Finally, the dHPLC assay was applied to explore the prevalence of diverse CTX-M-type ESBLs in our hospital in a 4-month survey carried out from September to December 2005 in Heartlands Hospital, Birmingham, United Kingdom.

MATERIALS AND METHODS

Bacterial strains and antimicrobial susceptibility testing.

Thirteen well-defined bla_CTX-M-producing strains were used as controls to develop and optimize the dHPLC genotyping assay (Table 1). To evaluate the accuracy of the optimized assay, a collection of 62 bla_CTX-M-positive isolates from our previous study (39), including 43 from group 1 (40 CTX-M-15 isolates and 3 CTX-M-3 isolates), 16 belonging to group 9 (10 CTX-M-14 isolates and 6 CTX-M-9 isolates), and 3 CTX-M-26 producers were tested in the preliminary stage. To test the field applicability of the dHPLC method, 78 ESBL-producing consecutive nonduplicate clinical isolates from both community- and hospital-based patients were collected during the period from September to December 2005. Community-acquired isolates were from patients who had not been hospitalized in the preceding 3 months, residents of nursing homes, and outpatients; hospitalized patients with a first positive culture obtained after 48 h of hospital admission were considered to have nosocomial infections. These isolates included 58 E. coli isolates, 15 Klebsiella spp., and 5 Enterobacter cloacae isolates.

TABLE 1.

Control strains used to evaluate the dHPLC genotyping assay

Isolate	Species	bla gene (group)	Source
GZ3	E. coli	CTX-M-3 (1)	Guangzhuo, China
QE15	E. coli	CTX-M-15 (1)	Birmingham, United Kingdom
F1	E. coli	CTX-M-1 (1)	Paris, France
SP32	E. coli	CTX-M-32 (1)	La Coruna, Spain
KE12	K. pneumoniae	CTX-M-12 (1)	Kenya
JA1	E. coli	Toho-1 (2)	Tokyo, Japan
F2	E. coli	CTX-M-2 (2)	Paris, France
GR5a	Salmonella enterica serotype Typhimurium	CTX-M-5 (2)	Athens, Greece
Y19	E. coli	CTX-M-9 (9)	York, United Kingdom
Y15	E. coli	CTX-M-14 (9)	York, United Kingdom
GZ13	K. pneumoniae	CTX-M-13 (9)	Guangzhuo, China
ESBL530	E. coli	CTX-M-25 (25/26)	Manitoba, Canada
H610	K. pneumoniae	CTX-M-26 (25/26)	Birmingham, United Kingdom

^aGR5 was kindly sent to us by L. Tzouvelekis as a bla_CTX-M-4 producer; however, when the PCR amplicon was sequenced, its sequence was identical to that of bla_CTX-M-5 submitted by the group of K. Bush (GenBank accession no. U95364) (6). Therefore, for the purpose of this study, this strain is renamed as a bla_CTX-M-5 producer.

All isolates were first tested for their susceptibility to cefpodoxime by the British Society for Antimicrobial Chemotherapy disk diffusion method (http://www.bsac.org.uk). ESBL production was confirmed phenotypically by using cefpodoxime, ceftazidime, and cefipirome with and without clavulanic acid.

Multiplex PCR detection of bla_CTX-M genes.

All isolates were initially screened for the presence of bla_CTX-M by our previously described mutiplex PCR protocol (39), except that Optimase proofreading DNA polymerase (Transgenomic Inc., Omaha, NE) was used. Four sets of primers were used to amplify fragments of bla_CTX-M open reading frames of CTX-M-type ESBLs, which are designed to give product sizes of 341 bp (group 2), 293 bp (group 9), 255 bp (group 1), and 207 bp (groups 25/26 and 8).

Heteroduplex formation and dHPLC analysis.

Representative control strains (confirmed by DNA sequencing) GZ3 (CTX-M-3), Y19 (CTX-M-9), J1 (Toho-1), and ESBL530 (CTX-M-25) were used as reference standards for groups 1, 9, 2, and 25/26, respectively. PCR products (5 μl) of either the control or the clinical isolates were mixed with an equal amount of DNA amplified from a reference standard to ensure that consistent peak heights (absorbance, minimum of 2 mV) were observed between different runs. The mixtures were heated to 95°C for 5 min and then gradually cooled to 25°C at a rate of 1°C/min to form homo- and heteroduplex DNA molecules.

Aliquots of 5 to 8 μl of each duplexed PCR product were loaded onto a DNAsep column in a WAVE 4500 DNA fragment analysis system (Transgenomic Inc.). The linear dHPLC acetonitrile gradient consisted of buffer A (0.1 M triethylamine acetate) and buffer B (0.1 M triethylamine acetate, 25% acetonitrile). The injected DNA was eluted at a flow rate of 1.5 ml/min for 3.3 min with a specific concentration of buffer B and was detected by measurement of the absorbance at 260 nm.

The gradient conditions and the heteroduplex analysis temperature required to obtain a successful resolution of the heteroduplex were predicted with Navigator software (Transgenomic, Inc.) and the dHPLC Melt Program developed at Stanford University (http://insertion.stanford.edu/melt.html) by using the reference amplicon sequences. The results were shown in the form of chromatographic peaks. The chromatographic peaks for controls, the blind test evaluation panel, and the unknown bla_CTX-M type were compared with each other by visual inspection and with the reference homoduplex peak, which was obtained by reannealing of an equal amount of reference PCR products.

DNA sequence analysis.

The dHPLC genotyping results were further confirmed by DNA sequencing of a limited number of isolates (10 of 78) from the 4-month survey study. PCR fragments covering the whole open reading frame of the relevant bla_CTX-M gene were cleaned by using a QIAquick PCR purification kit (QIAGEN). One hundred nanograms of the PCR product was used as the template for TaqCycle sequencing with an ABI Prism BigDye Terminator (version 3.0) cycle sequencing kit. The cycle sequencing products were analyzed on an ABI PRISM 3700 DNA analyzer (Functional Genomics Laboratory, University of Birmingham).

RESULTS

Development and optimization of the dHPLC genotyping assay.

The sequences of the PCR amplicons of 13 control strains from four bla_CTX-M groups (groups 1, 2, 9, and 25/26) were aligned and analyzed. The sequence diversity within the PCR fragments of the control strains for each group is summarized in Table 2. When the 15 sequence variations within the PCR fragments from four bla_CTX-M groups were analyzed, it was found that 8 led to a change in the amino acid sequence and that 7 were silent mutations.

TABLE 2.

DNA sequence variations among PCR amplicons of bla_CTX-M groups generated by multiplex PCR

Strain and blaCTX-M group	Sequence (amino acid)
Group 1
GZ3 (CTX-M-3)a	CGG (Arg 194)	GGC (Gly 203)	GCT (Ala 226)	GAC (Asp 242)
F1 (CTX-M-1)	CGT (−)b	GGT (−)
V12 (CTX-M-12)			GCA (−)
QE15 (CTX-M-15)				GGC (Gly)
SP32 (CTX-M-32)	CGT (−)	GGT (−)		GGC (Gly)
Group 9
Y19 (CTX-M-9)a	GAG (Glu 124)	GCG (Ala 157)
Y15 (CTX-M-14)	GAA (−)
GZ13 (CTX-M-13)	GAA (−)	GAG (Glu)
Group 2
J1 (Toho-1)a	GTA (Val 233)	GAA (Glu 255)	AGG (Agr 275)
F2 (CTX-M-2)			AGC (Ser)
GR5 (CTX-M-5)	GGA (Gly)	GCA (Ala)	AGC (Ser)
Group 25/26
ESBL530 (CTX-M-25)a	CAG (Gln 225)
H610 (CTX-M-26)	CGG (Arg)

^aThe reference strain for the group.

^b−, silent mutation (no change in amino acid).

To determine whether dHPLC can be used as a rapid method of screening or identifying CTX-M-type ESBLs within a group, we first selected as the reference strain one strain from each group and formed a homoduplex and a heteroduplex with the PCR amplicons representing the different members of bla_CTX-M genes from the relevant group. Then a qualitative comparison of the dHPLC elution profiles between the control and the reference strains was made. A total of 13 control strains (including 1 reference strain for each group), 5 from group 1, 3 from group 9, 3 from group 2, and 2 from group 25/26, were used to develop and optimize the dHPLC genotyping assay. The detection of sequence variation by dHPLC relies on a comparison of the chromatogram elution patterns between the sample (containing heteroduplex molecules) and the reference (containing homoduplex molecules). The initial temperatures for successful resolution of homo- and heteroduplex molecules were obtained by the simulation of the melting behavior of the reference sequence by using both Navigator software (Transgenomic Inc.) and the dHPLC Melt Program developed at Stanford University. The predicted temperatures of the entire strands of the bla_CTX-M PCR fragments for distinguishing the molecules by use of the Navigator software/the dHPLC Melt Program were 63.5°C/62°C, 63.8°C/63.3°C, 63.1°C/62°C, and 64.6°C/62.8°C for groups 1, 9, 2, and 25/26, respectively.

To obtain the optimal temperature for the detection of single nucleotide or multiple sequence variations between reference bla_CTX-M strains and other representative control strains within the group, all samples were initially analyzed at a wider range of temperatures up to ±3°C of the predicted highest and lowest temperatures in order to achieve the best resolution of heteroduplexes. The reference strains analyzed in the four groups produced a single peak at all temperatures tested (i.e., a true homoduplex). The dHPLC patterns of PCR amplicons containing sequence variations showed significant deviations from the corresponding elution profile of the reference amplicon. Alterations in the dHPLC chromatogram profile were observed at different temperatures. As the analysis temperature increases, the peak reduces in height and broadens in width. An example of a single nucleotide substitution (AGG to AGC), which resulted in an Arg-to-Ser 275 change between bla_CTX-M-Toho-1 (the reference strain) and bla_CTX-M-2 within group 2, is shown in Fig. 1. In contrast to the samples containing the bla_CTX-M-Toho-1 reference sequence, which shows a single peak of a homoduplex DNA profile at all temperatures tested (Fig. 1A), Fig. 1B depicts the bla_CTX-M-2 dHPLC signatures (three or two peaks) when the melting temperature was raised from 62°C to 63 and 65°C. Although the predicted temperatures were 63.1°C/62°C, the highest detection efficiency for bla_CTX-M-2 was observed to be 64°C.

FIG. 1.

Effect of column temperature on sequence variation detection by dHPLC. dHPLC profiles. (A) bla_CTX-M-Toho-1 (the homoduplexed reference strain) with a single peak observed at all temperatures between 62°C and 65°C; (B) heteroduplexed bla_CTX-M-2. The single nucleotide substitution (AGG to AGC) was detected at both 64°C and 65°C as three and two peak profiles, respectively.

The temperature window at which sequence variations of the bla_CTX-M amplicons within a group can be identified was ascertained (Table 3). The optimal temperatures for obtaining dHPLC typing signatures ranged from a single temperature of 65°C for bla_CTX-M-32 to six temperatures ranging from 60 to 65°C for bla_CTX-M-5. Figure 2 presents the dHPLC chromatograms at these optimal temperatures for all control strains of the four bla_CTX-M groups investigated in this study. For the bla_CTX-M group 1 amplicons analyzed, bla_CTX-M-12 and bla_CTX-M-15 contained single base substitutions; bla_CTX-M-1 and bla_CTX-M-32 harbored two and three mismatches (Table 2) from the sequence of the bla_CTX-M-3 reference strain. Although the optimal temperature of 64°C alone is sufficient for obtaining a distinct dHPLC profile for each bla_CTX-M genotype, the dHPLC signature for bla_CTX-M-32 was resolved into clearer double peaks at 65°C (Fig. 2A). To achieve 100% sensitivity and 100% specificity, it would be necessary to analyze amplicons at two temperatures (64°C and 65°C) for group 1 isolates. The single optimal temperature of 64°C for distinguishing bla_CTX-M member strains for groups 2 and 25/26 was achieved (Fig. 2B and D, respectively). However, for group 9, although a single temperature was attained for differentiation, the higher temperature of 67°C, which is 3°C higher than the predicted temperature, was required (Fig. 2C).

TABLE 3.

Temperature titration used to identify optimum temperature at which clear dHPLC chromatograms are seen

Group (representative strain) and blaCTX-M	Titration required at following temp (oC)a:								OpTb (oC)
	60	61	62	63	64	65	66	67
Group 1 (CTX-M-3)c
CTX-M-1	No	No	Yes	Yes	Yes	Yes	No	No	63-65
CTX-M-12	No	Yes	Yes	Yes	Yes	Yes	No	No	63-65
CTX-M-15	No	No	No	No	Yes	Yes	No	No	64-65
CTX-M-32	No	No	No	No	Yes	Yes	No	No	65
Group 2 (Toho-1)
CTX-M-2	No	No	No	No	Yes	Yes	No	No	64-65
CTX-M-5	Yes	Yes	Yes	Yes	Yes	Yes	No	No	60-65
Group 9 (CTX-M-9)
CTX-M-13	No	No	No	No	No	Yes	Yes	Yes	65-67
CTX-M-14	No	No	No	No	Yes	Yes	Yes	Yes	65-67
Group 25/26 (CTX-M-25)
CTX-M-26	No	Yes	Yes	Yes	Yes	Yes	Yes	Yes	62-66

^aYes, heteroduplex detected (two or more peaks detected); No, no heteroduplex detected (single peaks only).

^bOpT, optimal temperature at which chromatograms are significantly different from that of the homoduplex reference.

^cThe reference strain for the group.

FIG. 2.

dHPLC genotyping of bla_CTX-M genes at different column temperatures. The single optimal temperature for differentiation of bla_CTX-M variants within each group is boxed. The single peak profile of the reference strain (marked with an asterisk) for each group is shown only at the optimal temperature. (A) Genotyping of group 1 bla_CTX-M genes; (B) genotyping of group 2 bla_CTX-M genes; (C) genotyping of group 9 bla_CTX-M genes; (D) genotyping of group 25/26 bla_CTX-M genes.

bla_CTX-M PCR amplicons from group 1 were analyzed under the same temperature conditions in three separate runs, and dHPLC produced highly reproducible chromatograms (data not shown).

dHPLC evaluation with clinical isolates.

A total of 62 clinical isolates carrying sequences confirmed to be known bla_CTX-M genes from a previous study (39) were subjected to blinded dHPLC analysis in order to assess the reproducibility and the sensitivity of the dHPLC genotyping method. The collection included clinical isolates from groups 1, 9, and 25/26. No group 2 clinical isolates were available to us. Our dHPLC assay correctly genotyped all 62 isolates, with 53 showing characteristic chromatogram profiles that were different from the profile for the reference strain of each group and from each other. All isolates carrying the same bla_CTX-M gene had indistinguishable profiles (results not shown). For the remaining nine isolates, dHPLC typing showed single peak profiles corresponding to those for the reference strains: six isolates with profiles corresponding to bla_CTX-M-9 and three isolates with profiles corresponding to bla_CTX-M-3. The overall dHPLC typing results were 100% compatible with the data obtained by sequencing in our previous study (39).

Molecular typing of bla_CTX-M genes from the 4-month survey by dHPLC analysis.

During the 4-month period from September to December 2005, a total of 2,562 nonduplicate isolates of the family Enterobacteriaceae were processed in the clinical diagnostic service in the Birmingham Public Health Laboratory, Heartlands Hospital. A total of 78 ESBL producers were identified (3%), including 52 of 1,236 (4.1%) strains from hospitals and 26 of 1,326 (2%) strains from the community. The majority of the 78 clinical isolates were associated with urinary tract infections (66 [84%] strains); but they were also isolated from 6 sputum samples (7.8%), 3 cases of bacteremia, 2 wound infections, and 1 ascitic fluid sample. Multiplex PCR screening for the presence of bla_CTX-M genes yielded amplicons of the expected size from 73 of 78 isolates. The five bla_CTX-M PCR-negative isolates were further screened by a bla_SHV PCR, and three isolates were positive. The remaining two isolates may have possessed ESBLs other than CTX-M and SHV. Of the 73 bla_CTX-M-producing isolates, 71 belonged to bla_CTX-M group 1 and 2 were found to harbor group 9 bla_CTX-M genes. By using bla_CTX-M-3 as the reference strain, the dHPLC genotyping of group 1 isolates at the two column temperatures of 64°C and 65°C revealed that 70 (95.8%) had bla_CTX-M-15-specific elution profiles at 65°C and that 1 had the bla_CTX-M-1-specific three-peak pattern at both 64°C and 65°C. The two group 9 isolates depicted a bla_CTX-M-14-specific chromatogram signature when their patterns were compared with that of the bla_CTX-M-9 control (results not shown). The clinical sites, the distribution of species, and the CTX-M genotypes are summarized in Table 4. All 70 bla_CTX-M-15-producing isolates yielded identical profiles, and Fig. 3 demonstrates the signature matching between the CTX-M-15 control and the clinical isolates from the 4-month survey producing the bla_CTX-M-15 gene. Seven of 70 (10%) that typed as CTX-M-15 β-lactamase producers by dHPLC, as well as isolates which possessed bla_CTX-M-1 and bla_CTX-M-14, were sequenced. The genotypes predicted by dHPLC were verified by the sequencing results (sequences with GenBank accession no. DQ915953, DQ915954, and DQ915955).

TABLE 4.

Bacterial species, specimen type, and distribution of bla_CTX-M and bla_SHV genotypes among 78 ESBL clinical isolates

Species	Total no. of isolates	Specimen type (no.)	Specimen no. by characteristic	Patient locationa	blaCTX-M PCR	Genotype by dHPLC	blaSHV PCR result
E. coli	58	Blood (3)	3	H	Positive	CTX-M-15	Negative
		Fluid (2)	2	H	Positive	CTX-M-15	Negative
		Sputum (3)	1	C	Negative	Negative	Negative
			1	C	Positive	CTX-M-14	Negative
			1	H	Positive	CTX-M-15	Negative
		Urine (50)	21	C	Positive	CTX-M-15	Negative
			1	C	Positive	CTX-M-14	Negative
			26	H	Positive	CTX-M-15	Negative
			1	H	Positive	CTX-M-15	Positive
			1	H	Positive	CTX-M-1	Negative
Klebsiella spp.	15	Sputum (2)	2	H	Positive	CTX-M-15	Positive
		Urine (13)	2	C	Positive	CTX-M-15	Negative
			7	H	Positive	CTX-M-15	Positive
			2	H	Negative	Negative	Positive
			1	H	Negative	Negative	Negative
			1	H	Positive	CTX-M-15	Negative
E. cloacae	5	Sputum (1)	1	H	Positive	CTX-M-15	Negative
		Tissue (1)	1	H	Positive	CTX-M-15	Positive
		Urine (3)	1	H	Negative	Negative	Positive
			2	H	Positive	CTX-M-15	Negative

^aH, hospital acquired; C, community acquired; see Materials and Methods for definitions.

FIG. 3.

Overlay of dHPLC signatures to assess reproducibility of characteristic features of CTX-M-15. The top chromatogram is the CTX-M-15 control signature and is compared to signatures from two bla_CTX-M-15-producing clinical isolates from the survey.

DISCUSSION

The dHPLC technology for mutation detection has facilitated the more efficient analysis of sequence variations and has led to a better understanding of the role that such variations play in the development of diseases such as cancer (21). It is also regarded as one of the most specific and sensitive techniques for mutation detection, with a specificity and a sensitivity approaching 100% each (38). To date, the dHPLC technique has seen limited applications to both bacterial species identification and the detection of mutational changes within antibiotic resistance-associated genes. However, its potential usefulness for rapid screening of the genotypes of antibiotic resistance-associated variants has been little explored (10, 14). In this study, the use of dHPLC for the rapid screening of bla_CTX-M gene variants has been investigated.

The dHPLC technology is typically used to detect point mutations in DNA sequences by comparing the chromatograms of a wild-type (reference) strain and an experimental sample. Phylogenetic analysis of the target gene (bla_CTX-M) has allowed the definition of five distinct groups which show very limited cross-group sequence homology. In addition, within each of these groups as many as 20 mutational changes have been seen between two members. The lack of sequence homology between different phylogenetic groups also precludes identification of a single PCR target that will facilitate cross-group discrimination. To accommodate these features of bla_CTX-M variation, we have developed a multiplex PCR-based approach (39).

The principle of the dHPLC typing strategy applied in this study is illustrated for group 1 bla_CTX-M isolates in Fig. 4. When dHPLC is applied to the genotyping of highly variable genes, such as bla_CTX-M, the choice of the reference strain is crucial. For the typing of group 1 isolates, GZ3 carrying bla_CTX-M-3 was selected as the reference strain because in the United Kingdom bla_CTX-M-15-producing isolates account for greater than 95% of all CTX-M-type ESBLs from both community and hospital patients. The bla_CTX-M-15 gene contains a single nucleotide substitution (A725 to G725) compared to the sequence of bla_CTX-M-3, which leads to the Asp-240-Gly change (by the numbering system of Ambler et al. [2]). This mutation has been shown to increase hydrolytic activity against ceftazidime (5) and can be readily discriminated from bla_CTX-M-3 by dHPLC (Fig. 2A). It is worth mentioning that three other group 1 strains, CTX-M-28, -29, and -33, also differ from CTX-M-3 by a single nucleotide (A750 to G750) within the PCR amplicon. However, these three strains are rare variants. CTX-M-33 and CTX-M-29 have been reported in only one E. coli strain each in Greece and China, respectively (40). Since the first report of its isolation in France, CTX-M-28 has appeared in only one Chinese hospital. Given the fact that CTX-M-15 is the predominant group 1 genotype in the United Kingdom, the chromatogram obtained by dHPLC very likely predicts the presence of bla_CTX-M-15. Therefore, we believe that our dHPLC typing technique is robust enough to serve as a rapid screening genotyping method; and in an outbreak of CTX-M-type ESBL infection, sequencing of a proportion of the isolates should be undertaken in order to confirm the bla_CTX-M genotypes of the outbreak strains (Fig. 4). In addition to allowing the sensitive identification of the bla_CTX-M-15 genotype, Fig. 2A also shows the high sequence-specific discriminatory power of the dHPLC technique, in which other subtypes of group1 bla_CTX-M variants can be distinguished from those of the reference strain and each other on the basis of their chromatogram profiles.

FIG. 4.

Algorithm for dHPLC genotyping of bla_CTX-M-producing ESBLs in a clinical laboratory.

As described above, one of the key factors influencing the efficiency of detection of sequence variation by dHPLC is the column temperature used. The correlation of our experimentally determined optima with those predicted by the two computer programs used showed that the Navigator program gave a closer correlation than the dHPLC Melt Program did, but neither of the programs was able to predict the temperature at which the maximum discrimination was achieved for all four of the bla_CTX-M group amplicons being analyzed. This limitation in the predictive value of these algorithms is possibly due to the fact that for each amplicon, multiple sequence variations at different positions in the amplicon must be differentiated. This perhaps is more demanding for the software, because some variants can be identified at several temperatures while others are resolved at only a single temperature. Previous reports (18) have shown that a 2°C increase above the predicted discriminatory optimum was necessary to maximize mutation detection. However, in this study we found that 67°C, which is 3°C higher than the predicted optimum, was required in order to achieve the best discrimination of the group 9 bla_CTX-M-producing strains evaluated. Of the PCR amplicons tested for the four groups, the single temperature was sufficient to differentiate the member strains of three groups and only the group 1 amplicon required analysis at two temperatures in order to achieve maximum specificity. In addition to the column temperature, the G+C content of the DNA analyzed, the numbers and types of mutations within the DNA fragment, and the DNA concentration all affect the efficiency of mutation detection in terms of the shapes, numbers, and heights of heteroduplex peaks. In this study DNA concentrations are standardized before analysis, but some minor variation (one of the factors causing differences in the heights of each of the peaks) may inevitably occur. Signature matching between control isolates (with known mutations) and clinical isolates (Fig. 3) is the most important key to the identification of mutations and to the interpretation of changes in the shapes or the heights of the peaks.

The blinded validation with 62 clinical bla_CTX-M producers from our previous study showed that dHPLC genotyping was able to assign all of the strains to the correct genotype.

The method described in this paper detects sequence variations in 200- to 300-bp fragments of the bla_CTX-M gene, which enables the differentiation of isolates into distinct CTX-M genotypes. The current “gold standard” for precise genotyping and the identification of novel enzymes is complete DNA sequencing of the entire 0.87-kb bla_CTX-M gene. To further improve the specificity of dHPLC typing technique and to enable the detection of all novel enzymes by dHPLC, two potential approaches are currently under investigation. First, a dHPLC typing assay with a whole-gene PCR amplicon could be developed; although DNA fragments of up to 1.5 kb have been used for dHPLC-based mutation detection (27), the high level of sequence diversity within the bla_CTX-M gene may complicate chromatogram interpretation. Second, and alternatively, two PCR amplicons that cover the entire CTX-M gene could be used and the simultaneous appearance of characteristic peak profiles in two fragments may provide a specific genotype fingerprint.

Since the early 1990s, the rapid expansion of the CTX-M type of ESBLs has created great concern in the public health community worldwide. The prevalence of different CTX-M types of ESBL varies geographically. While the predominant genotypes have been CTX-M-14 and -9 in Spain (15, 31), CTX-M-14 and -3 in China (40), CTX-M-1 in Italy (8), and CTX-M-2 in both Japan and Argentina (30, 33), CTX-M-15 is the most prevalent genotype in the United Kingdom, with an epidemic CTX-M-15-producing strain (strain A) being identified (37).

Our 4-month survey is the first comprehensive prospective study of the genotyping of all isolates with a bla_CTX-M-producing ESBL in a large group of teaching hospitals (1,100 beds) in the United Kingdom. Applying the newly developed dHPLC-based genotyping method, we successfully genotyped all 73 bla_CTX-M ESBL-producing strains in our sample set. Our data have confirmed the findings of the general screening of isolates referred to the United Kingdom Health Protection Agency's Antibiotic Resistance Monitoring and Reference Laboratory from 42 centers during 2003, that is, that CTX-M-15 is the predominant type of CTX-M β-lactamase in the United Kingdom (37). The 4-month survey showed that 57/58 (98%) ESBL-producing E. coli strains harbored the bla_CTX-M gene (Table 4). This much higher rate of detection of CTX-M-type ESBLs in clinically significant E. coli isolates may suggest that bla_CTX-M has been exceptionally successful in both community and hospital settings in the south Birmingham area. Since its first description in 2001 (19), CTX-M-15 has rapidly spread worldwide, including to Asia, Europe, South America, Africa, and the Middle East (11, 12, 24, 26, 29, 34). More alarmingly, CTX-M-15- producing E. coli strains have recently caused reported outbreaks in countries such as Italy, where previously only CTX-M-1 β-lactamases have been reported (25), and in France and Spain, where the clonal dissemination of CTX-M-15 in hospitals had previously never been reported (22, 28). The availability of high-throughout genotyping by our dHPLC technique will enable the spread of bla_CTX-M-15 as well as all the other major outbreak genotypes of CTX-M (bla_CTX-M-1, bla_CTX-M-2, bla_CTX-M-3, bla_CTX-M-9, bla_CTX-M-12, bla_CTX-M-13, bla_CTX-M-14, and bla_CTX-M-26) to be monitored. The dHPLC genotyping results from our survey have demonstrated the discriminatory power of our technique, as we have found the first reported cases of clinically significant disease caused by CTX-M-14- and CTX-M-1-producing E. coli strains in the United Kingdom. Since TEM-type ESBLs have rarely been encountered in clinical isolates in the United Kingdom, bla_TEM was not screened for in this study.

This is the first report demonstrating the application of the dHPLC technology to the genotyping of β-lactamases and was particularly applied to CTX-M-type ESBLs. The technique has many advantages. It is reliable, and the PCR products are processed in a 96-well autosampler unit and continuously injected onto the column for analysis without any time-consuming premanipulations, such as purification and denaturation. Other major advantages over sequencing are the high-throughput capability (with 3.3 min/injection, 100 samples can be analyzed in 6 h) and low consumables cost (at least 10 times less than the cost of consumables for sequencing). It can be applied to clinical isolates on a large scale inexpensively, and it can also be used to determine the clinical relevance of different genotypes and can be applied in epidemiological studies and surveillance programs. By using dHPLC, the worldwide distribution of bla_CTX-M genotypes and their subsequent evolution can be delineated, and furthermore, any novel mutations/insertions/deletions can potentially be detected as a new chromatograph profile.