Knowledge of the genetic organization of acquired AmpC should allow us to understand the molecular mechanisms of ampC gene transfer. Some ampC genes are located within or near mobilizing elements, such as transposons or integrons (23, 46). Integrons have been implicated in the horizontal transfer of class C β-lactamases, but not as gene cassettes, as has been seen for the Ambler class A, B, and D β-lactamases (17). Indeed, several ampC genes, such as bla_DHA-1, bla_FOX-4, bla_CMY-1, bla_CMY-8, and bla_MOX-1, have been described in a particular class 1 integron, originally identified in In6 and In7, which contains two partial copies of the 3′ conserved segment (3′-CS1 and 3′-CS2) surrounding a common region (CR) and an antibiotic resistance gene (32, 35, 42, 44). Other antibiotic resistance genes have now been reported adjacent to the CR of similar integrons, such as qnr in In37 (47), bla_CTX-M-9 in In60 (38), dfrA10 in In34 (32), dfrA10 in a variant Salmonella genomic island (SGI1) from Salmonella enterica serovar Agona (8), bla_CMY-9 in pCMXR1 from Escherichia coli (13), and catA2 in pAr-32 from Aeromonas salmonicida (41). Partridge and Hall showed that CR1 containing open reading frame (ORF) orf513, CR2 (containing orfA), and CR3 (containing orf2) form a family of genetic elements involved in the mobilization of resistance genes in gram-negative bacteria (32). The incorporation of a resistance gene at the 3′ end of the CR is probably catalyzed by orf513, but this has yet to be established experimentally.

This study reports on the characterization of 12 Klebsiella sp. strains with an inducible DHA-type enzyme. These were isolated between 1998 and 2003 in the Paris, France, region. The aim of this work was to describe the epidemiological features of these strains and to explore the genetic organization of bla_DHA-1.

Bacterial strains. Twelve clinical isolates from 12 patients in eight hospitals in the Parisian region between 1998 and 2003 were studied. Eleven K. pneumoniae isolates and one Klebsiella oxytoca isolate were identified by using the API system (bioMerieux, Marcy l'Etoile, France). The characteristics of the strains are summarized in Table 1.

TABLE 1. Origins and epidemiological features of the clinical strains investigateda

Previously described DHA-1-type β-lactamase-producing strain Salmonella serovar Enteritidis KF92 (5) and M. morganii SLM01 (4) were used as controls.

Escherichia coli K-12 strain J53-2 (Rif^r), E. coli K-12 strain C600 (Nal^r), and E. coli strain DH10B (Invitrogen SARL, Cergy-Pontoise, France) were used as recipient strains for conjugation and electroporation experiments.

Plasmid vector pBKCMV (Stratagene, La Jolla, CA) was used for cloning.

Antibiotic susceptibility. The MICs for several β-lactams were determined by the standard agar dilution method on Mueller-Hinton agar (Bio-Rad, Marnes la Coquette, France). The resistance phenotype was determined by the disk diffusion method on Mueller-Hinton agar. Disk antagonism tests with imipenem or cefoxitin as inducing agents were used to detect the inducibilities of the β-lactamases (22). The antagonism was detected by the blunting of the oxyimino β-lactam zone of inhibition surrounding the imipenem or cefoxitin disk.

Class A extended-spectrum β-lactamases (ESBLs) were detected by the standard double-disk synergy test with amoxicillin-clavulanic acid and cefotaxime, ceftazidime, or cefepime (9).

Transfer of resistance. Mating experiments were done by mixing equal volumes of each test strain and either E. coli K-12 C600 (Nal^r) or E. coli K-12 J53 (Rif^r). β-Lactam-resistant E. coli transconjugants were then selected on Drigalski agar (Bio-Rad) containing either nalidixic acid (50 μg/ml) or rifampin (250 μg/ml) and either cefoxitin (10 μg/ml) or ceftazidime (2 μg/ml).

Plasmid DNA was isolated by alkaline lysis (QIAGEN Plasmid Purification, Courtaboeuf, France) and was used to transform E. coli DH10B cells by electroporation, according to the manufacturer's instructions (Bio-Rad). Transformants were incubated for 2 h at 37°C and plated on Drigalski agar supplemented with cefoxitin (10 μg/ml).

Plasmid DNA fingerprinting. Plasmid DNA was purified from transconjugants or transformant cells by using a Plasmid Midi kit (QIAGEN), according to the manufacturer's recommendations. For fingerprinting analysis, plasmid DNA was digested with the EcoRI restriction enzyme (New England Biolabs Inc., Saint Quentin en Yvelines, France) and subjected to electrophoresis in a 1% agarose gel at 80 V for 4 h.

PFGE analysis. Genomic DNA, prepared as described by Decré et al. (11) and digested with XbaI (New England Biolabs), was subjected to pulsed-field gel electrophoresis (PFGE) in a CHEF DRIII device (Bio-Rad). DNA fragments were separated in a 1% (wt/vol) agarose gel in 0.5× Tris-borate-EDTA buffer at 200 V for 20 h with pulse times ranging from 5 to 30 s.

PCR and DNA sequencing: characterization of bla genes. Total DNA was extracted by using a QIAamp DNA Mini Kit (QIAGEN). Primers OT3 (5′-ATGAGTATTCAACATTTCCG-3′) and OT4 (5′-CCAATGCTTAATCAGTGAGG-3′) were used to amplify putative bla_TEM genes, and primers OS5 (5′-TTATCTCCCTGTTAGCCACC-3′) and OS6 (5′-GATTTGCTGATTTCGCTCGG-3′) were used to amplify bla_SHV genes. The bla_TEM and bla_SHV genes were characterized by sequencing and were then analyzed by using tables available on the website www.lahey.org/Studies. The entire region containing the bla_DHA gene and its regulator gene, ampR, was amplified by using three pairs of primers generating overlapping PCR fragments on a 2,243-bp region: DHA upper1 and DHA lower1, DHA upper2 and DHA lower2, and DHA upper3 and DHA lower3 (Table 2, PCRs M, N, and O, respectively). The PCR conditions were as follows: 1 min at 94°C, 1 min at 55°C, and 1 min at 72°C. All enzymes were purchased from Sigma (St. Quentin Fallavier, France).

TABLE 2. Sequences of oligonucleotides used for PCR mapping

Cloning of ampC and ampR genes and DNA sequencing. Plasmids pKp760 and pTN60013 were isolated from the E. coli C600 transconjugant by alkaline lysis (QIAGEN Plasmid Purification).

The plasmids were totally digested with EcoRI, and the fragments were ligated into the EcoRI site of pBKCMV. Each recombinant plasmid was introduced into E. coli DH10B. Transformants were selected on Mueller-Hinton agar supplemented with cefoxitin (50 μg/ml) and kanamycin (25 μg/ml).

The insert was sequenced as described by Sanger et al. (39), using an ABI 373 DNA sequencer (PE Applied Biosystems, Foster City, CA).

The BLAST and the FASTA programs were used to search databases for similar nucleotide and amino acid sequences. The ClustalX program was used to align multiple protein sequences (1, 2, 34).

PCR mapping and DNA sequencing: genetic organization of bla_DHA-1. Previously described PCR primers for integron pSAL-1 were used to map the unknown region surrounding the ampC and the ampR genes in DHA-1-producing strains. PCRs A, B, F, D, and G were performed (46). The PCR conditions were as described above for all fragments except those that could exceed 3 kb, most notably, for PCR B, which included the gene cassettes. In this case the extension time was increased to 5 min and the DNA polymerase was adapted for long-time PCR: DyNAzyme EXT (Finnzymes, Espoo, Finland).

In strains RBDHA and MPDHA, the unknown sequences downstream from ampC or downstream from CR1 were amplified by using the Universal Genome Walker kit (Clontech Laboratories, Palo Alto, CA) (40). Genomic DNA was digested with EcoRV, DraI, PvuII, StuI, NruI, and AviII; and the DNA fragments were ligated with a Genome Walker adapter. Then, the following steps were performed according to the manufacturer's recommendations by using ampC-specific or CR1-specific primers (Table 2).

New PCR primers were designed in this study to map the bla_DHA-1 region by using as templates the DNA sequences of the clones obtained from pKp760 or pTN60013, the results obtained for RBDHA and MPDHA with the Universal Genome Walker kit, or, finally, the DNA sequences of various In6- and In7 type-integrons. The primers are listed in Table 2.

The resulting PCR products were sequenced and then analyzed with the BLAST, FASTA, and ClustalX programs. AMIGene software was used to characterize the most likely coding sequences in a large bacterial genome sequence (7).

Several PCR products obtained from strains TN58647, JDDHA, STA1DHA, TN26033, STA2DHA, IGR1DHA, and IGR2DHA were not sequenced but only compared by their restriction fingerprints. Two aliquots of the products of PCRs J, K, L, P, and S were totally digested, one with TacI (New England Biolabs) and the other with HaeIII (New England Biolabs). The digestion products were separated on 2.5% agarose gels and compared to that obtained from strain TN60013 under the same conditions.

Nucleotide sequence accession numbers. The nucleotide sequences in strains Kp760, RDDHA, RBDHA, MPDHA, and TN60013 have been submitted to GenBank and have been assigned accession numbers AJ971341, AJ971342, AJ971343, AJ971344, and AJ971345, respectively.

Antibiotic susceptibility. According to the criteria of the French Society for Microbiology (www.sfm.asso.fr/), the 12 isolates were highly resistant to amoxicillin, ticarcillin, and cefoxitin. Clavulanic acid did not significantly restore susceptibility to penicillins. The isolates were also resistant or showed decreased susceptibility to ceftazidime and cefotaxime, whereas they remained susceptible to cefepime and imipenem (Table 3). Antagonism between imipenem or cefoxitin and cefotaxime or ceftazidime was apparent in disk diffusion assay. This phenotype suggested the presence of an inducible AmpC-type β-lactamase. In four K. pneumoniae isolates (isolates TN58647, RDDHA, JDDHA, and TN60013), synergy between cefepime and amoxicillin-clavulanic acid was also observed, suggesting the presence of a class A extended-spectrum β-lactamase.

TABLE 3. MICs of β-lactams and inhibitors

Transfer of resistance. E. coli J53-2 or E. coli C600 transconjugants were obtained by mating out with rifampin or nalidixic acid and cefoxitin selection (Table 1). Electroporation of plasmids extracted from K. pneumoniae STA1DHA and STA2DHA into E. coli DH10B successfully transferred cefoxitin resistance. The transconjugants and transformants exhibited an antagonism between imipenem and either cefotaxime or aztreonam. No transfer of resistance was obtained from K. oxytoca MPDHA or K. pneumoniae TN26033 either by mating out or by electroporation. Moreover, the synergy between cefepime and amoxicillin-clavulanic acid observed in four donor strains was conserved in the transconjugant for one strain only, K. pneumoniae RDDHA, suggesting the transfer of both class A and class C β-lactamase-encoding genes.

Transconjugants showing typical synergy between clavulanic acid and broad-spectrum β-lactams were obtained by mating out strains TN58647, JDDHA, and TN60013 in E. coli C600 by using ceftazidime for selection.

Epidemiological analysis. According to the interpretation criteria of Tenover et al. (43), of the 11 K. pneumoniae isolates analyzed in this study, 9 different PFGE profiles were obtained (data not shown). Three isolates, isolates TN58647, JDDHA, and TN60013, had identical patterns, indicating their clonal origin (Table 1).

Plasmid DNA fingerprinting carried out with transconjugants or transformants identified similar plasmids in seven isolates (Fig. 1; Table 1).

FIG. 1. Plasmid DNA fingerprinting after restriction with EcoRI. Lane 1, molecular size marker (Bio-Rad); lane 2, transconjugant E. coli J53(pJDDHA); lane 3, transformant E. coli DH10B(pSTA1DHA); lane 4, transconjugant E. coli J53(pTN58647); lane 5, transconjugant (more ...)

Characterization of β-lactamases: PCR and DNA sequencing of bla genes. PCR experiments were carried out with the clinical isolates to detect bla_DHA and bla_TEM genes.

PCRs were positive for the three overlapping DHA fragments with all isolates. The analysis of the 2.2-kb DNA sequence yielded two ORFs in opposite orientations: ampC and ampR. This region was 98% identical to the M. morganii nucleotide sequence. For each isolate, the deduced amino acid sequence of the AmpC-like enzyme was 100% identical to that of DHA-1.

By PCR and sequence analysis, 4 of the 12 isolates were also shown to be TEM-1 positive (Table 1).

Likewise, the extended-spectrum β-lactamase characterized in the transconjugants obtained from strains TN58647, JDDHA, and TN60013 by selection with ceftazidime was SHV-2a and that of the transconjugant E. coli C600(pRDDHA) was SHV-2 (Table 1).

Characterization of the genetic contexts of bla_DHA-1. As bla_DHA-1 from KF92 is inserted in a sul1-type integron that includes orf513, it was hypothesized that bla_DHA-1 in the other isolates would exist in the same genetic configuration. Therefore, we used PCR to look for the presence of orf513 in all isolates (PCR D) (46). PCR D was negative for eight isolates and positive for the remaining four isolates (Table 4). Therefore, we explored the genetic organizations of the bla_DHA-1 genes in several ways: (i) isolates that were orf513 positive were considered to include a backbone of an In6- and In7-like sul1-type integron, and they were explored by PCR mapping and cloning experiments. (ii) We considered that isolates that were orf513 negative belonged to the same DHA-1 plasmid fingerprint and that, in these cases, bla_DHA-1 would have the same genetic organization. TN60013, an orf513-negative strain, was then chosen for the cloning and sequencing of the ampC and ampR regions. Subsequently, the other isolates were tested by PCR mapping.

TABLE 4. Results for PCR mapping: length of PCR products

Isolates carrying orf513: characterization of the In6- and In7-like integrons. We used PCRs A, B, and F to successfully amplify fragments from the four orf513-positive isolates (isolates Kp760, RDDHA, RBDHA, and MPDHA; Table 4). These fragments overlapped a region typical of a complex sul1-type integron, from the int gene to CR1, with variable gene cassette arrays (Fig. 2) (46).

FIG. 2. Comparison of regions surrounding orf513 and blaDHA-1. Arrows with slashes, M. morganii region; vertical bars, IR sequences.

(i) In the integron from Kp760 we detected two gene cassettes, aadB and aadA2.

(ii) In the integron from RDDHA we detected three gene cassettes, dhfr, arr2, and cmlA7, and a fused gene cassette oxa10-aadA1. An open reading frame, orf437, of unknown function was found downstream of dhfr in an opposite orientation; this ORF encodes a 437-amino-acid protein that is 79% identical to that of orfII (GenBank accession number AAK54203). The cmlA7 and oxa10 genes are interrupted by two putative insertion sequences, IS_X and IS_Y, respectively (1,202 bp and 1,039 bp, respectively), that are homologous to ISXCD1 (68% protein identity) found in Xanthomonas campestris pv. dieffenbachiae (GenBank accession number AF263433) and to ISXo2 (45% protein identity) found in Xanthomonas oryzae pv. oryzae (GenBank accession number AAF61102). The transposase in both putative insertion sequences (ISs) was surrounded by two inverted repeats (IRs), with a direct repeat site found on the target flanking the IS: GGGC for cmlA and TGACCTTA for oxa10.

(iii) In the integron from strain RBDHA we detected four gene cassettes: aac(6′)-Ib, bla_OXA-30, catB3, and arr-3.

(iv) In the integron from strain MPDHA we detected two gene cassettes, aac(6′)-IIc and ereA2. ereA2 is interrupted at position 188 by a 3,924-bp region surrounded by a 4-bp direct repeat site, TTAA. This 3,924-bp region included an insertion sequence, IS1247, previously described in Xanthobacter autotrophicus (45), which was characterized with imperfect IRs on either side. IS1247 was followed by two ORFs whose protein products were 79% and 72% identical to aminoglycoside acetyltransferase aac3-Vb from S. marcescens (GenBank accession number Q01515) and a putative rifampin ADP-ribosyltransferase from Rhodopseudomonas palustris (accession number NP_948241), respectively.

We attempted PCR G with the four orf513-positive isolates to cover the region between CR1 and bla_DHA-1. However, PCR G was positive only for strain RDDHA. Thus, bla_DHA-1 from RDDHA was mobilized from a 3,085-bp region of the M. morganii chromosome and inserted downstream from CR1 in an In6- and In7-like sul1-type integron with a particularly long (9-kb) gene cassette region (Fig. 2).

By cloning the region encompassing ampC and ampR in strain Kp760, a recombinant plasmid (pKp760-2_ind) with an 8.1-kb insert that conferred inducible resistance to ceftazidime was selected. The insert sequence was determined (Fig. 2). A 5.9-kb region at the 5′ end of the insert that contained ampC and ampR was 98% identical to the M. morganii chromosome. The 3′ end of the insert includes a 3′-CS of an integron. We designed a reverse primer that binds to the 5′ region of the insert of pKp760-2_ind (PCR H) to generate an overlap between the CR1 and recombinant plasmid pKp760-2_ind. We used the overlapping fragments from PCRs A, B, F, and H and pKp760-2_ind to map the 13.5-kb region surrounding bla_DHA-1 in pKp760 (Fig. 2). The ampC and ampR genes were found to be inserted in a 6,133-bp region identified as the M. morganii chromosome. However, the first 561 bp was unknown in nucleotide data banks. We designed primers to generate an overlap between this unknown region and the available M. morganii sequence (GenBank accession number AF055067) (PCR U). PCR U was successful for both pKp760 and M. morganii SLM01. The nucleotide sequences of these PCR products were 99% identical. Therefore, we suggest that the entire 6,133-bp region comprising ampC and ampR came from the M. morganii chromosome. This 6,133-bp region is followed by a 3′-CS with the same deletion of quacEΔ1 at the 5′ end (180 bp) observed in pSAL-1. A partial tniB gene was found downstream of orf5 at the 3′ end of the insert in a reverse orientation and contained the EcoRI site (Fig. 2).

We then explored bla_DHA-1 in strains RBDHA and MPDHA. Although we assumed that the genetic organization was similar to that in Kp760, PCR H failed for both RBDHA and MPDHA. Therefore, we explored the unknown regions of RBDHA and MPDHA downstream from CR1 and downstream from ampC with the Universal Genome Walker kit. With both strains, a 1.8-kb PCR product was obtained downstream from CR1 (Table 4, PCR i). We characterized two ORFs whose predicted protein sequences were about 90% identical to the sapB and sapA products of the sap operon (ABC transporter family) from Salmonella enterica serovar Typhimurium LT2 (GenBank accession number AE008775). We analyzed a 2-kb PCR product obtained downstream from ampC in RBDHA and MPDHA (Table 4, PCR ii). The sequence of the 3′ end was identical to that of the M. morganii chromosome. The 5′ end provided several ORFs of the psp operon: pspF, pspA, pspB, and pspC. Both the sap and the psp operons are usually found in the genomes of members of the family Enterobacteriaceae and can be adjacent to each other, as in Salmonella enterica serovar Typhimurium LT2, section 79 of 220 of the complete genome, in which sapA is directly followed by pspF. Consequently, we attempted PCR I to link the region downstream from CR1 (sapB) and the region upstream from M. morganii (pspF). The products of PCRs I, J, and K were obtained from strains RBDHA and MPDHA, and the sequences were identical in both strains. The region between the end of CR1 (TATACCC site) and the beginning of the M. morganii region was found to be 10,317 bp long (Fig. 2). It includes a part of the sap operon at the 5′ end (3,049 bp), followed by an unknown region (4,216 bp) and then part of the psp operon at the 3′ end (3,052 bp).

The structure of the 3′-CS2 region present in pSAL-1 and in the orf513-containing integrons from strains RDDHA, RBDHA, and MPDHA was investigated by a PCR mapping approach based on known sequence data (8, 32, 47). As expected, we identified the second copy of a 3′-CS (3′-CS2). However, variable structures were found downstream of orf5 (Fig. 2). In strains KF92 and MPDHA, the IS26 insertion sequence followed the 3′-CS2, as in strain TN60013, whereas a partial tniB was found downstream of 3′-CS2 in RDDHA, as in Kp760. In RBDHA, the last 314 bp of the 3′-CS2 is deleted by the insertion of an IS similar to IS4321 (33). This is followed first by two ORFs similar to orf97 (GenBank accession number AAR05756.1) and orf98 (GenBank accession number AAR05757.1), as described in S. enterica serovar Typhimurium plasmid pU302L (GenBank accession number AY333434), and then by IS6100.

Isolates not carrying orf513. By cloning the region encompassing ampC and ampR in strain TN60013, we selected a recombinant plasmid (pTN60013-2_ind) with a 12-kb insert that conferred inducible resistance to oxyimino-cephalosporins, and we partially determined (10.2 kb) the sequence of the insert (Fig. 3). As IS26 was found at the 3′ end of the recombinant plasmid, we looked for and detected IS26 upstream from pTN60013-2_ind using PCR T (Table 4). The PCR T product, which overlapped the recombinant plasmid (pTN60013-2_ind), was sequenced; and this allowed us to map a 12.2-kb region surrounding bla_DHA-1 in pTN60013.

FIG. 3. Genetic organization of pTN60013 in comparison with those of pMPDHA and pT948. Arrows with slashes, M. morganii region; shaded region, regions of 100% identity; vertical bars, IR sequences.

The ampC and ampR genes are inserted in a 3,437-bp region that originated from the M. morganii chromosome. Downstream of ampR there is a 1,737-bp 3′-CS, including partial qacEΔ1, sul1, and orf5 sequences, with the same length of deletion (180 bp) detected for 3′-CS2 in pSAL-1. This partial 3′-CS is followed by an insertion sequence, IS26, that includes a tnpA gene encoding a transposase oriented in the same way as orf5. Upstream from the M. morganii region, a 2,691-bp segment with five ORFs was characterized by using AMIGene software (7). Each protein product was 100% identical to the products of the psp operon in strains RBDHA and MPDHA: PspF, PspA, PspB, PspC, and PspD. Neither 3′-CS1 nor 5′-CS was found on plasmid pTN60013.

By using the results obtained with TN60013, several PCRs (PCRs J, K, L, P, and S) (Table 4) were then used to characterize the genetic context of bla_DHA-1 in the other seven orf513-negative isolates. The PCR products and the restriction profiles obtained with these PCR products were the same for each isolate (data not shown), suggesting that the genetic environment of bla_DHA-1 was at least similar, if not identical, in the eight orf513-negative isolates.

We have studied 12 DHA-1-producing clinical strains isolated between 1998 and 2003 from 12 patients in eight hospitals of the Parisian area. To date, all DHA-type class C β-lactamases described seem to be inducible. However, the expression in the strains isolated in the United Kingdom remains unknown (20). A previous study of Salmonella serovar Enteritidis KF92, which produces DHA-1, showed that ampR was mobilized together with ampC from the M. morganii chromosome and that the plasmid-encoded AmpR was functional (5). Most plasmid-mediated ampC genes are expressed constitutively, as they lack ampR mobilization (35). Few other inducible acquired non-DHA class C β-lactamases have been described: ACT-1 in a K. pneumoniae isolate from New York City in 1996 (36), CFE-1 in an E. coli isolate from Japan in 1997 (28), and CMY-13 in an E. coli isolate from Greece in 2001 (23). In these strains, ampC and ampR had the same configuration and had a high degree of identity with the corresponding chromosomally encoded genes of Enterobacter asburiae for ACT-1 (37) and C. freundii for CFE-1 and CMY-13. However, plasmid-mediated bla_AmpC originating from C. freundii is not often found with ampR. AmpC enzymes, for example, MIR-1 and CMY-2, therefore have a high level of constitutive expression (6, 18). By contrast, bla_AmpC originating from M. morganii is nearly always mobilized together with ampR. Why ampC alone or ampC and ampR are mobilized from a chromosome onto a plasmid has yet to be determined.

Many strains with plasmid-encoded AmpC enzymes also produce TEM-1, TEM-2, or even an ESBL such as SHV-5 (3, 18, 25). To date, DHA-producing K. pneumoniae strains isolated in Taiwan or Seoul, South Korea, have been shown to have acquired bla_TEM-1 and, sometimes, an additional ESBL, either a CTX-M-type or SHV-type enzyme (30, 49, 50). Our study of 12 isolates of K. pneumoniae and K. oxytoca producing DHA-type β-lactamases showed that 4 isolates produced an ESBL of the SHV type. Epidemiological results and the β-lactam resistance profiles showed that three identical isolates produced SHV-2a, which was encoded by a plasmid different from that encoding DHA-1, while with another isolate, the DHA-1- and SHV-2-encoding determinants were cotransferred.

Among the three identical isolates, both TN58647 and TN60013 were isolated in the same medical care unit in Tenon Hospital in Paris in 1999 and 2002, respectively. Strain JDDHA was isolated in 2002 in Joffre-Dupuytren Hospital in the Parisian region. Although many patients admitted to Joffre-Dupuytren Hospital are transferred from either Tenon Hospital or St. Antoine Hospital, no unambiguous epidemiologic link could be established between these patients. The similarity of the plasmids of four other K. pneumoniae strains led us to conclude that these plasmids spread by horizontal transfer. The emergence of DHA-1 in the Parisian region is primarily due to the dissemination of one plasmid.

Previous outbreaks with AmpC-type β-lactamases (ACC-1 or CMY-type enzymes) have been described in Paris, with strains being either imported from a region of endemicity or isolated after a secondary horizontal spread inside the hospital (12, 27, 29). However, in our study, a common geographical origin could not be established for the DHA-1-producing strains, as most of the patients were French and had not traveled abroad immediately prior to infection. The higher rate of isolation in recent years suggests that DHA-1 β-lactamases are becoming more prevalent in the Parisian region, which may be a cause of concern for public health. Considering the number of hospitals affected, we believe that our study shows one of the most significant instances of the emergence of plasmid-mediated class C β-lactamases described in France.

Although DHA-1-producing strains have been present in the Far East for several years, there have been few data about the genetic organization of bla_DHA-1. We compared epidemiological data with the genetic organization for our 12 isolates. Seven orf513-negative isolates with the same plasmid fingerprint also had the same genetic organization: the 3,437-bp M. morganii-derived segment bounded by two directly repeated IS26 elements within a 12.2-kb region (Fig. 3). The plasmid fingerprint of an eighth isolate, isolate TN26033, could not be obtained, as cefoxitin resistance was not transferable either by electroporation or by conjugation. However, we classified this isolate with the other seven strains because it had the same genetic organization.

Four strains differed from the eight orf513-negative isolates in their plasmid fingerprints and by the presence of an integron-type backbone in the environment of bla_DHA-1. These four strains all have the same complex sulI-type integron, similar to In6, In7, and pSAL-1, with a CR1 that includes orf513, which encodes a putative recombinase (42, 44, 46). We propose that these integrons be named InKp760, InRDDHA, InRBDHA, and InMPDHA (Fig. 2). As with similar integrons, the variable region (M. morganii) is inserted at the same right-hand boundary of CR1, which is presumed to be a site recognized by orf513. The deletion of the 5′ part of 3′-CS2 is the same for each bla_DHA-1-containing integron (180 bp), whereas this deletion is variable in other In6- and In7-type integrons (8, 38, 42, 44). However, each of the five DHA-1-carrying integrons differs in several ways: (i) the gene cassette region varies in number and identification (Fig. 2). The integron of plasmid pSAL-1 carries the usual gene cassette aadA2, and the integron InKp760 carries the usual gene cassettes aadB and aadA2 (Fig. 2).

The integron InRDDHA is unusual by the presence of ORFs that are not well characterized (Fig. 2). The first gene cassette, dhfr, is followed by orf347, which codes for a protein close to orfII and which has been described several times, always in a reverse orientation, in integrons comprising the bla_VIM-2 gene, in Pseudomonas aeruginosa YMC 95/1/704 (GenBank accession number AY029772), or in Acinetobacter genomospecies 3 YMC 99/11/160 (GenBank accession number AF369871). Following orf347, the last four genes cassette, arr2, cmlA7, oxa10, and aadA, are in the same array as part of In53, described in E. coli MG-1, which encodes an extended-spectrum β-lactamase, bla_VEB-1 (26). A notable difference with In53 is the presence of a putative IS-type structure inserted inside cmlA7 and oxa10. Their presence most likely leads to the inactivation of both genes. To our knowledge, this is the first time that a putative IS-type structure has been found to be inserted inside a gene cassette. Their presence is surprising, as a gene cassette can simply be excised if its expression needs to be stopped. The cmlA7 gene confers nonenzymatic chloramphenicol resistance and contains a promoter-like sequence (26). The fused gene cassette oxa10-aadA1, previously described by Naas et al. (26), likely provides its own promoter.

Integron InRBDHA carries four gene cassettes, identical to that found in In37 from an E. coli strain isolated in Shanghai, China, between March 2000 and March 2001 (Fig. 4): aac6′Ib, bla_OXA-30, catB3, and arr3 (47).

FIG. 4. Comparison of pRBDHA and In37. Shaded region, regions of 100% identity; vertical bars, IR sequences.

The region of gene cassettes in integron InMPDHA is unusual, with the insertion of IS1247 inside a well-characterized gene cassette, ereA2, which leads to the duplication of a 4-bp target site. We suggest that IS1247 mediated the insertion of a large region of unknown origin comprising two new antibiotic resistance genes: a putative aminoglycoside acetyltransferase and a putative rifampin ADP-ribosyltransferase. Both genes will be explored in a future study.

(ii) In InRBDHA and InMPDHA, the integron-type backbone is unusual, with the insertion, between the CR1 and the M. morganii region, of an additional 10.3-kb large region that includes the psp and sap operons (Fig. 2).

(iii) The length of the M. morganii region is variable, with the length of the segment between CR1 and ampC varying—172 bp (strain KF92), 638 bp (strain RDDHA), 989 bp (strains RBDHA and MPDHA), and 3,687 bp (strain Kp760)—while the segment downstream from ampR is always 321 bp long (Fig. 2). The sequence at the left boundary of the M. morganii region is different for each strain. Thus, its insertion downstream from CR1 seems to be randomly determined and not target site specific.

(iv) The extension of the 3′-CS2 beyond sulI was only partially sequenced. However, we established differences between the five integrons (Fig. 2). Unfortunately, PCR mapping did not allow us to characterize the boundaries of the integrons with IRi and IRt.

The variability of such a complex sulI-type integron suggests that this structure evolves. Each different integron was found in a single isolate, which suggests an overall high rate of recombination events. On the contrary, the genetic organization of the bla_DHA-1 context in the orf513-negative isolates was conserved, with a mobile genetic element apparently capable of spreading. A trace of a sul1-type integron, a 3′-CS2, was present in the genetic organization of pTN60013 (Fig. 3). However, a complete integron was not found, with there being no 3′-CS1 or int gene present in KpTN60013. The presence of two directly repeated IS26 copies suggests a composite transposon. This could have been involved in the integration of the 12.2-kb region that includes bla_DHA-1. However, the sequences beyond the IS26 elements could not be determined; and it is unknown whether an 8-bp target site is duplicated, although no marked target selectivity has been described for IS26 (21). This genetic organization is very close to that of a Chinese isolate of K. pneumoniae carrying bla_DHA-1 on the plasmid pT948 (GenBank accession number AY705809) (48). The segment comprising part of a psp operon, followed by the M. morganii region and then part of a 3′-CS2, is 100% identical in both plasmids pT948 and pTN60013 (Fig. 3). However, the orientation of IS26 and the sequences adjacent to it differ in these two plasmids. Both of these plasmids are likely to have evolved from a common structure, with the IS26 elements being inserted independently. Plasmids pMPDHA, pTN60013, and pT948 were analyzed; and the genetic organizations of bla_DHA-1 in the three plasmids were compared (Fig. 3). We suggest that pTN60013, as well as pT948, has evolved from a complex integron, such as that carried by pMPDHA, by the insertion of IS26 upstream from the psp operon, for pTN60013, or inside the psp operon, for pT948. This then leads to the deletion of the first part of a complex integron from int to CR1.

Study of the spread of bla_DHA-1 in the Parisian region allowed us to describe the genetic organization of this acquired class C β-lactamase. The involvement of In6- and In7-type integrons, especially orf513 including CR1, is clearly demonstrated in the horizontal transfer of antibiotic resistance genes. The function and mobility of orf513, which is assumed to interact with the right-hand boundary of CR1, remain to be explored. This study allowed the description of new IS or IS-like structures in the variable gene cassette region from sulI-type integrons.

This work was supported by grants from Faculté de Médecine Saint-Antoine, Université Pierre et Marie Curie, and from the European Community (6th PCRD contract LSHM-CT 2003-503335).

We thank Dominique Decré, Frédéric Fitoussi, Michèle Saladin-Allard, Houria Nebbad-Lechani, Nevin Boutros, Béatrice Salauze, Jean-Marie Libert, Guilène Barnaud, Anne-Marie Rossier, Elizabeth Chachaty, and Alain Philippon.