Evolution of an Incompatibility Group IncA/C Plasmid Harboring bla_CMY-16 and qnrA6 Genes and Its Transfer through Three Clones of Providencia stuartii during a Two-Year Outbreak in a Tunisian Burn Unit

Abstract

During a 2-year period in 2005 and 2006, 64 multidrug-resistant Providencia stuartii isolates, including 58 strains from 58 patients and 6 strains obtained from the same tracheal aspirator, were collected in a burn unit of a Tunisian hospital. They divided into four antibiotypes (ATB1 to ATB4) and three SmaI pulsotypes (PsA to PsC), including 49 strains belonging to clone PsA (48 of ATB1 and 1 of ATB4), 11 strains to clone PsB (7 of ATB2 and 4 of ATB3), and 4 strains to clone PsC (ATB3). All strains, except for the PsA/ATB4 isolate, were highly resistant to broad-spectrum cephalosporins due to the production of the plasmid-mediated CMY-16 β-lactamase. In addition, the 15 strains of ATB2 and ATB3 exhibited decreased quinolone susceptibility associated with QnrA6. Most strains (ATB1 and ATB3) were gentamicin resistant, related to an AAC(6′)-Ib′ enzyme. All these genes were located on a conjugative plasmid belonging to the incompatibility group IncA/C₂ of 195, 175, or 100 kb. Despite differences in size and in number of resistance determinants, they derived from the same plasmid, as demonstrated by similar profiles in plasmid restriction analysis and strictly homologous sequences of repAIncA/C₂, unusual antibiotic resistance genes (e.g., aphA-6), and their genetic environments. Further investigation suggested that deletions, acquisition of the ISCR1 insertion sequence, and integron cassette mobility accounted for these variations. Thus, this outbreak was due to both the spread of three clonal strains and the dissemination of a single IncA/C₂ plasmid which underwent a remarkable evolution during the epidemic period.

INTRODUCTION

Sepsis is the most common cause of death in burn patients. Staphylococcus aureus is the main bacterial species involved, followed by Pseudomonas aeruginosa. Providencia stuartii, a member of the Enterobacteriaceae occasionally found in the human digestive flora and in the environment, is a rare etiological agent of burn wound infections (10, 11). Indeed, this opportunistic pathogen is essentially responsible for urinary tract infections, particularly in long-term care populations (33). P. stuartii is intrinsically resistant to aminopenicillins and narrow-spectrum cephalosporins due to a chromosomally encoded Ambler class C cephalosporinase (12, 22, 53). Overexpression of AmpC usually confers to this species a low-level resistance to broad-spectrum cephalosporins such as ceftazidime (12). Higher levels of ceftazidime resistance are primarily associated with the acquisition of extended-spectrum β-lactamases (ESBLs) (15, 49).

In recent years, a variety of plasmid-mediated antibiotic resistance mechanisms have emerged in Enterobacteriaceae, particularly toward β-lactams and fluoroquinolones. Thus, cephalosporinase-encoding genes, which are naturally present on the chromosome of some Enterobacteriaceae species and environmental organisms, have been mobilized to conjugative plasmids that have further disseminated in pathogenic strains (22). The plasmid-mediated AmpC enzymes degrade all β-lactams except for carbapenems. In contrast to ESBLs, they poorly affect zwitterionic oxyiminocephalosporins such as cefepime and cefpirome and are inhibited by cloxacillin but are not inhibited by clavulanic acid (22). Currently, about 120 plasmid-mediated AmpC β-lactamases (http://www.lahey.org/Studies) that fall in six groups on the basis of their percentage of similarity have been described (35). Of them, CMY-2 is the most prevalent and widely distributed enzyme (22). Similarly, the first plasmid-encoded Qnr protein, QnrA1, is thought to have been transferred from the chromosome of the environmental Gram-negative bacillus Shewanella algae (24, 36). Qnr proteins protect the type II topoisomerases from quinolone inhibition, conferring a low-level quinolone resistance (30, 44, 48). At present, more than 50 qnr genes have been assigned and have been divided into five families (A to D and S) sharing limited homology (http://www.lahey.org/qnrStudies) (21). Among them, qnrA and qnrB are the qnr determinants most frequently and broadly encountered around the world (39, 44). These genes have often been described on the same conjugative plasmid as β-lactamase-encoding genes, mainly ESBLs (6). However, a genetic link between Qnr and plasmid-mediated AmpC has also been reported. Thus, the first plasmid found to carry qnrA1 also harbored the bla_FOX-5 gene (51). To date, the most frequent association is between qnrB4 and bla_DHA-1 (6). In contrast, very few strains coexpressing QnrA-like and CMY-2-like proteins have been described (18, 37), and the molecular basis of this coexpression has not been well documented.

A previous study reported on an outbreak due to 20 multidrug-resistant strains of P. stuartii in a burn unit of Hospital Aziza Othmana of Tunis (Tunisia) (41). Despite the implementation of control measures, the outbreak persisted. The aim of the present work was to analyze the 64 strains of P. stuartii collected over a 2-year period (2005 and 2006) in this unit in order to (i) assess the epidemiological relatedness of these isolates, (ii) elucidate the mechanisms of their β-lactam and low-level quinolone resistance, and (iii) analyze the genetic support of the acquired antibiotic resistance genes.

MATERIALS AND METHODS

Bacterial strains and clinical data.

All multidrug-resistant nosocomially acquired P. stuartii strains isolated in the burn unit of the Hospital Aziza Othmana of Tunis (Tunisia) between January 2005 and November 2006 were collected. Only the first strain isolated from each patient was conserved and included in the study. Clinical cases were retrospectively analyzed to elicit patients' characteristics, diseases, treatment, and evolution. Various potential sources of this bacterium in the unit, particularly water supplies and equipment, were investigated for the presence of P. stuartii. Strains were identified by the API 20E system (bioMérieux, France) and stored in long conservation medium before they were sent to the Laboratory of Microbiologie Fondamentale et Pathogénicité of Bordeaux University (Bordeaux, France) for analysis.

Antimicrobial susceptibility testing.

The antibiotic susceptibility patterns of the strains and their transconjugants were determined by the disk diffusion method in Mueller-Hinton (MH) agar using 27 antibiotic disks (Bio-Rad) in the absence or presence of 250 mg/liter of cloxacillin (http://www.sfm.microbiologie.org). A synergy test was performed by the combined disks containing broad-spectrum cephalosporins and clavulanic acid (Oxoid). MICs of β-lactam agents (ticarcillin, cefoxitin, ceftazime, cefotaxime, and cefepime) alone or supplemented with 250 mg/liter of cloxacillin and quinolones (nalidixic acid, norfloxacin, ofloxacin, and ciprofloxacin) were determined by the agar dilution method (http://www.sfm.microbiologie.org).

Molecular typing of the strains.

The clonal relationship of all P. stuartii isolates was investigated by pulsed-field gel electrophoresis (PFGE) as described previously (41). In brief, after overnight culture in MH broth, the cell suspensions were immobilized into 2% low-melting-point agarose before incubation in a lysis buffer including lysozyme (2 mg/ml) and proteinase K (4 mg/ml). After the washing steps, the P. stuartii total DNA was digested with the endonuclease SmaI. Restriction fragments were separated by PFGE using a contour-clamped homogeneous electric field (CHEF) apparatus (CHEF DRIII; Bio-Rad) containing 0.5× TBE (Tris-borate-EDTA) buffer at 6 V/cm for 21 h at 14°C, with pulse times ranging from 2 to 40 s. Gels were stained with a 2-mg/liter ethidium bromide solution. Strains representing each of the five pulsotype (Ps)/antibiotype (ABT) groups (strains PsA/1, PsA/4, PsB/2, PsB/3, and PsC/3 representing the groups PsA/ATB1, PsA/ATB4, PsB/ATB2, PsB/ATB3, and PsC/ATB3, respectively) were selected at random for further molecular analysis purposes.

General DNA procedures.

Total DNA was extracted as previously described (42). The Southern blot hybridization experiments were achieved with the commercialized reagents for the probe labeling, the hybridization, and the chemiluminescence detection systems recommended by the manufacturer (Roche Applied Sciences, France). PCR amplifications were carried out using GoTaq DNA polymerase (Promega SA, France) or Phusion DNA polymerase (Finnzymes, Finland), according to the suppliers' instructions. Oligonucleotides used for PCR amplifications were provided by Eurofins MWG Operon (France) and are listed in Table 1. Sequencing was performed using an ABI Prism BigDye Terminator cycle sequencing ready reaction kit and an ABI 3130xl sequencer (Applied Biosystems, Perkin Elmer, France). Sequences were analyzed with BioEdit software, using the BLAST program of the National Center for Biotechnology Information website (http://www.ncib.nlm.nih.gov/BLAST).

Table 1.

Oligonucleotides used in this study

Primer use and primera	Sequence (5′ → 3′)	Source or reference
blaCMY-16 environment analysis
CMY2-F1	CGGAACTGATTTCATGATG	This study
CMY2-R5	TTATTGCAGCTTTTCAAGAA	This study
Down_SugE	ACCTGCAAACGCTTCTTTC	This study
TraA-downF	GTTGTGCAACTCAGCAATGGC	This study
ISEcp1	CCTAGATTCTACGTCAGTACTT	This study
InvISEcp1	TTCAATAAAATCAAAAATCCCA	This study
CMY-16downF	TGTTGTCATCTACACTTAAC	This study
TraCupR	ATCAACGTTTCTTCGAGTTTC	This study
qnrA6 and aac(6′)-Ib′environment analysis:
5′CS	GGCATCCAAGCAGCAAG	27
qnrA6F	CAGCAAGAGGATTTCTCACG	This study
qnrA6R	CTAATCCGGCAGCACTATGAC	This study
TniBR	ATGGACCTCGATGCCCGAACGG	This study
C6T3R	GCTATCAGTTGAACGAGC	This study
DownqnrF1	AAGCACTTGGCGCTGTCTT	This study
GmF1	GCATTCGCCAGTGACTGG	This study
GmR1	TACAGCATCGTGACCAACAG	This study
APH(3′)-IF	ATGAGCCATATTCAACGGG	This study
APH(3′)-IR	AGAAAAACTCATCGAGCATC	This study
Other antibiotic resistance gene analysis
APH(3′)-VI-F	CTGCTGCACGTATTTCTCG	This study
APH(3′)-VI-R	CGGAAACAGCGTTTTAGAGC	This study
FloF	TGATCCAACTCACGTTGAGC	This study
FloR	GAACGCAGAAGTAGAACGCG	This study
TetDF-FW	GGAATATCTCCCGGAAGCGG	1
TetDF-RV	CACATTGGACAGTGCCAGCAG	1
StrABF	GACAAGAGTACGCCGCAGCTC	This study
StrABR	ATGATGCAGATCGCCATGTAG	This study
sul2-F	GCGCTCAAGGCAGATGGCATT	4
sul2-B	GCGTTTGATACCGGCACCCGT	4

^aOnly the primers which gave a positive PCR amplification are given. Oligonucleotides used for sequencing are not indicated.

Antibiotic resistance gene characterization and analysis of the gene environment.

Plasmid-mediated AmpC β-lactamase and quinolone resistance qnr (qnrABCDS) genes were searched for in all strains using either a multiplex (35) or a simplex (9, 39) PCR protocol, respectively. Amplicons were sequenced for representative strains, as indicated above. Other resistance genes [aphA-6, floR, strAB, tetA(D), sul2, sul1] were screened by PCR amplifications and sequenced.

The genetic environment of bla_CMY-16 was explored by PCR screening for sequences of widely disseminated genetic elements generally associated with the bla_CMY-2-like genes, including the ISEcp1-like element, and the adjacent sequences were further characterized in selected isolates by a PCR mapping assay based on known sequences (17, 45). Sequences surrounding the qnrA6 and aac(6′)-Ib′ genes were determined after cloning experiments using DNA from PsB/3 partially digested with Sau3AI and using the BamHI-linearized and dephosphorylated pBK-CMV vector (Stratagen, France). The competent cells of Escherichia coli TOP10 were electrotransformed with the ligation mixtures and then plated on MH agar containing kanamycin (50 mg/liter) with either ciprofloxacin (0.05 mg/liter) for qnrA6 or gentamicin (10 mg/liter) for aac(6′)Ib′ cloning experiments, respectively.

Conjugation experiments and plasmid content analysis.

Conjugative transfers were performed by a filter method using a nalidixic acid-resistant mutant of Escherichia coli strain K-12 as the recipient at a 1:2 donor-to-recipient cell ratio. Transconjugants were selected on MH agar containing nalidixic acid (100 mg/liter) plus ceftazidime (2 mg/liter) (2). For P. stuartii strain PsA/4, three independent conjugative transfer assays were carried out at 30°C and 37°C, with the transconjugant selection performed on plates supplemented with nalidixic acid (100 mg/liter) and florfenicol (8 mg/liter) or tetracycline (10 mg/liter). For the five representative strains and their transconjugants, plasmid typing was achieved by a PCR-based assay which recognizes FIA, FIB, FIC, HI1, HI2, I1-Ig, L/M, N, P, W, T, A/C, K, B/O, X, Y, and FII replicons (7). Amplicons were confirmed by DNA sequencing and used as probes in subsequent hybridization experiments. Then, plasmid DNA from transconjugants was extracted using an alkaline lysis method (42) and was digested with the PstI enzyme (Promega SA). Furthermore, the unrestricted plasmid profiles were visualized after DNA linearization with S1 nuclease, followed by PFGE (S1-PFGE), as reported elsewhere (3). After S1-PFGE, plasmid DNA was transferred onto a nylon N membrane and labeled as described above. Purified labeled DNA products obtained from PCR of the replicon IncA/C and the qnrA6 genes were used as probes. Transconjugants from representative strains and the parental strain PsA/4 were also tested for the presence of 13 specific backbone markers, including the repA gene, as previously described for IncA/C plasmid characterization (52).

Nucleotide sequence accession numbers.

The 3,002-bp sequence of bla_CMY-16 and its genetic environment from PsC/3 (other name Ps59) and those of the 17,719-bp sequence of qnrA6 from PsB/3 have been deposited in the GenBank database (http://www.ncbi.nlm.nih.gov/GenBank) under accession numbers FJ855437 and JN193567, respectively. The partial sequence of the aphA-6 gene and the sequence of the class 1 integron containing the aac(6′)-Ib′ and aphA1-IAB genes from PsB/3 are available under GenBank accession numbers JN193566 and JN193568, respectively. The class 1 integron with an array of five gene cassettes (dhfr14, arr-2, cmlA5, bla_OXA-10, and aadA1) was deposited in the Integrall database under accession number In633.

RESULTS

Clinical data.

A total of 64 P. stuartii strains collected over a 2-year period in the burn unit of Hospital Aziza Othmana of Tunis were examined, including 58 strains from 58 patients and 6 strains from the same tracheal aspirator (Fig. 1). The 58 patients exhibited a sex ratio of 1.0 and a mean age of 31.6 years, with ages ranging from 9 to 71 years. Clinical strains were collected from blood (50.0%), skin (34.5%), blood catheter (6.0%), urine (6.0%), urinary catheter (1.75%), and respiratory tract (1.75%). The mean duration of hospitalization was 64 days (standard deviation [SD], 59.0 days), and the mean total body surface area (TBSA) of burns was 44.4% (SD, 17.3%). Many patients (76.5%) harbored medical devices, i.e., blood and urinary catheters (48.9%), tracheal intubation devices (22.4%), and additional wound drains (5.2%). Thirty-four patients received antibiotics, including 26 receiving two or more antibiotics. Imipenem was the most frequently given antibiotic (94.1%), followed by amikacin (52.9%) and gentamicin (47.1%). Most of the patients developed sepsis (87.9%), with the mortality rate reaching 37.9%. Six P. stuartii strains were isolated from the same aspirator, the use of which was concomitant with the beginning of the outbreak (41).

Fig 1.

Temporal relationship among the 64 P. stuartii isolates of the burn unit. At the top, the blue full arrows indicate the 48 isolates belonging to the clone PsA/ATB1, and the orange dotted line arrow indicates strain PsA/4. At the bottom, the red full arrows indicate the seven isolates of clone PsB/ATB2, the black double-headed arrows represent the four strains of clone PsB/ATB3, and the green dotted line arrows represent the four isolates of clone PsC/ATB2. The open star-like symbol indicates the six environmental strains. In June 2005, three strains were collected from reservoirs 1 and 2 and from the tube of the tracheal aspirator. In February 2006, two strains were isolated from reservoirs 1 and 2. The closed diamonds indicate the representative strains.

Antibiotype and pulsotype data.

Antimicrobial susceptibility testing showed that the 64 strains divided into four antibiotypes (ATB1 to ATB4) (Table 2). As confirmed by MIC determination, all of them except for the single strain of ATB4 demonstrated low-level resistance to ticarcillin (MICs, 128 to 256 mg/liter) and high-level resistance to ceftazidime (MICs, 32 to 256 mg/liter versus 8 mg/liter for a P. stuartii strain of our collection overexpressing its chromosomally encoded cephalosporinase). Susceptibility to β-lactams, in particular, to ceftazidime, was restored by the addition of cloxacillin (MICs, 1 to 8 mg/liter), but not by the addition of clavulanic acid. All strains were also resistant to chloramphenicol/florfenicol, tetracycline, sulfonamides, and trimethoprim and to aminoglycosides, including kanamycin (K), tobramycin (T), netilmicin (Nt), amikacin (A), neomycin (Nm), isepamicin (I), streptomycin (S), and spectinomycin (Sp), while only strains of ATB1 and ATB3 were gentamicin (G) resistant. In addition, strains of ATB2 and ATB3 exhibited decreased fluoroquinolone susceptibility (Table 2).

Table 2.

Distribution of the P. stuartii strains according to their epidemiological relationship, antimicrobial resistance phenotype, and IncA/C plasmid type

No. of strains	Pulsotype	ATBa no.	Antimicrobial resistance phenotype of P. stuartii strains and their transconjugantsb												IncA/C PCR profilec	Plasmid size (kb)d
			AM	Cz	S	Sp	G	T	A	Su	Tp	C	Te	Cp
48e	A	1	AM	Cz	S	Sp	G	T	A	Su	Tp	C	Te	—g	1 2 3 4 5 6 - - 9 10 11 12	175
1	A	4	—	—	S	Sp	—	T	A	Su	Tp	C	Te	—	1 2 3 4 5 - - - - 10 11 12	100
7f	B	2	AM	Cz	S	Sp	—	T	A	Su	Tp	C	Te	Cp	1 2 3 4 5 6 7 8 9 10 11 12	195
4	B	3	AM	Cz	S	Sp	G	T	A	Su	Tp	C	Te	Cp	1 2 3 4 5 6 7 8 9 10 11 12	195
4	C	3	AM	Cz	S	Sp	G	T	A	Su	Tp	C	Te	Cp	1 2 3 4 5 6 7 8 9 10 11 12	195

^aATB, antibiotype.

^bAM, ampicillin; Cz, ceftazidime; S, streptomycin; Sp, spectinomycin; G, gentamicin; T, tobramycin; A, amikacin; Su, sulfonamides; Tp, trimethoprim; C, chloramphenicol; Te, tetracycline; Cp, ciprofloxacin.

^cAccording to Welch et al. (52).

^dOnly the plasmids giving a positive hybridization with the repAIncA/C₂ probe are indicated.

^eIncluding five environmental strains.

^fIncluding one environmental strain.

^g—, sensitive phenotype except for ampicillin and gentamicin because the strains exhibited low-level resistance due to the basal expression of their chromosomal species-specific enzymes, AmpC (12) and AAC(2′)-I (16).

PFGE analysis showed that the 64 strains divided into three Sma pulsotypes, i.e., pulsotype PsA (49 strains, including 5 environmental strains), pulsotype PsB (11 strains, including the remaining environmental strain), and pulsotype PsC (4 strains) (data not shown; Table 2). Strains belonging to clone PsA consisted of the 48 strains of ATB1 and the single isolate of ATB4. The 11 strains with the profile PsB comprised isolates of ATB2 (7 gentamicin-susceptible strains) and ATB3 (4 gentamicin-resistant strains). The four strains with the pattern PsC exhibited the antibiotype ATB3. The distribution of the strains in the burn unit according to the time period showed that strains belonging to the five pulsotype/antibiotype groups overlapped (Fig. 1).

Characterization of antibiotic resistance determinants and transfer analysis.

After multiplex PCR (35), a bla_CMY-2-like gene was identified in all strains, except for PsA/4. In four representative strains (PsA/1, PsB/2, PsB/3, and PsC/3), sequencing allowed identification of bla_CMY-16. Analysis of the genetic environment of the bla_CMY-16 gene by a Southern blot hybridization experiment demonstrated that it was located on a similar large EcoRI fragment of ca. 60 kb (data not shown). In the 15 strains of pulsotypes PsB and PsC, screening by PCR amplifications showed the presence of a qnrA gene. Sequencing in three representative strains (PsB/2, PsB/3, and PsC/3) allowed characterization of qnrA6.

All representative strains (PsA/1, PsB/2, PsB/3, and PsC/3) harboring bla_CMY-16 transferred this gene to E. coli K-12 by conjugation at frequencies ranging from 10⁻⁴ to 10⁻⁵ transconjugants (Tc) per donor cell. In contrast, the transfer of florfenicol and tetracycline resistance from PsA/4 (the strain lacking CMY-16) remained negative, even when the mating assays were performed at 30°C (38). The various antibiotic resistances found in the representative clinical strains of ATB1, ATB2, and ATB3 cotransferred with the bla_CMY-16 gene, in particular, the quinolone resistance (qnrA6 gene) for the ATB2 and ATB3 isolates. The aminoglycoside resistance genes, such as aphA-6 (phenotype KNmAI), aphA1-IAB (KNm), strAB (S), and aadA1 (S/Sp), were also detected in all strains and their transconjugants. It is noteworthy that the APH(3′)-VIa enzyme, essentially found in Acinetobacter spp. (25), presented only 99% identity with the closest sequence deposited in the GenBank database, i.e., that found in an Alcaligenes faecalis strain (29), differing by 3 amino acid substitutions (S94P, F154L, and E155D). All the representative strains, including PsA/4, possessed this new variant. In contrast, the aac(6′)-Ib′ gene, conferring the phenotype GTNt (26), was detected only in the gentamicin-resistant strains of ATB1 and ATB3 and their transconjugants. Other genes contributing to the remaining resistances were also detected by PCR amplifications in P. stuartii clinical strains and their transconjugants, i.e., floR (chloramphenicol/florfenicol resistance), tetA(D) (tetracycline resistance), and sul1 and sul2 (sulfonamide resistance).

Plasmid analysis.

Multiplex PCR detected the repA gene characteristic of the plasmid incompatibility group IncA/C in all P. stuartii strains and their transconjugants, including PsA/4. repA sequencing in the five representative strains revealed the presence of an identical gene, repAIncA/C₂ (17). The S1-PFGE analysis showed the presence of a single plasmid of 195 kb in TcPsB/2, TcPsB/3, and TcPsC/3, while a smaller plasmid of 175 kb was detected in TcPsA/1. Two plasmids, of 100 kb and 50 kb, respectively, were visualized in PsA/4 (Fig. 2a). All the 195-, 175-, and 100-kb plasmids provided positive hybridization with the repAIncA/C₂ probe (Fig. 2b), while only the 195-kb plasmid gave a signal for qnrA6 (Fig. 2c). The IncA/C plasmids were further investigated by using a backbone-specific PCR assay (52). Plasmids from TcPsB and TcPsC were positive for all 12 specific amplifications, while plasmids from TcPsA/1 and PsA/4 were negative for 2 (R7 and R8) and 4 (R6 to R9) amplifications, respectively (Table 2). The loss of the R6 region was associated with the absence of the traA and traC genes in PsA/4, as confirmed by PCR amplifications (data not shown). The analysis of the IncA/C₂ plasmids from TcPsA/1, TcPsB/2, TcPsB/3, and TcPsC/3 after PstI digestion showed that they had similar profiles (data not shown).

Fig 2.

S1-PFGE with hybridization with the repAIncA/C₂ and qnrA6 probes. Lanes 1 to 5, TcPsB/2, TcPsB/3, TcPsC/3, TcPsA/1, and PsA/4, respectively; lanes 6, E. coli K-12; lanes M, PFGE marker. (a) S1-PFGE; (b) Southern hybridization with the repAIncA/C₂ probe; (c) Southern hybridization with the qnrA6 probe.

Genetic environment of bla_CMY-16, qnrA6, and aac(6′)Ib′ genes.

For the four representative producing strains, analysis of the bla_CMY-16 environment revealed the presence of the insertion sequence ISEc9 (an ISEcp1-like element) upstream and of the blc (bacterial lipocalin) and sugE1 (quaternary ammonium compound resistance) genes downstream. Further analysis showed that ISEc9 was adjacent to the traA gene of the IncA/C plasmid (primer pair TraA-downF/InvISEcp1, 422 bp) and that the sugE gene was near the traC gene (primer pair CMY-16downF/TraCupR, ca. 7,800 bp) (Fig. 3A).

Fig 3.

Characterization of multidrug resistance regions identified in the IncA/C2 plasmids carrying bla_CMY-16 (A), the class 1 integron with the same gene cassette array and differing by presence or not of qnrA6 (B), and the class 1 integron harboring aac(6′)-Ib′ (C). Open reading frames (ORFs) correspond to arrow-shaped boxes; the arrows express the direction of transcription of the corresponding ORFs. Boxes bordered by a dotted line indicate ORFs detected by PCR amplification but not sequenced. The names of the ORFs are indicated below the arrows. Black boxes show the genes of interest, i.e., bla_CMY16 (A), qnrA6 (B), and aac(6′)-Ib′ (C). (A and B) Dark gray boxes, genes implicated in transposition and/or gene mobilization mechanisms; crosshatched boxes, resistance genes; horizontal dashed lines, sequence of the cloned fragment from quinolone-resistant clone C6 (B); horizontal dotted lines framed by triangles, PCR mapping strategy used to characterize the multidrug resistance region; arrowheads, position of primers used for the PCR amplifications; black and white circles, attI and attC sites implicated in gene cassette mobilization by the integrase IntI1, respectively; horizontal bold lines, identity with similar regions available in the GenBank database; arrow, transcription promoter upstream from qnrA6 (pOUT). (C) Arrows and squared rectangles, direct repeat sequence framed by the aac(6′)-Ib′ gene (DR1 and DR2).

All representative strains contained a class 1 integron with an array of five gene cassettes (dhfr14, arr-2, cmlA5, bla_OXA-10, and aadA1) that we deposited in the Integrall database under accession number In633 (31). In PsA/1 and PsA/4, the usual 3′ conserved sequence (CS; qacEΔ1/sul1) was followed by the orf5 gene (Fig. 3B). In PsB and PsC strains, an additional sequence of 11,116 bp encompassing orf513/ISCR1 followed by the qnrA6 gene was identified by cloning and by long-range PCR (Fig. 3B). The insertion sequence IS10 (1,321 bp) was present immediately upstream from qnrA6. Between the end of ISCR1 and the beginning of the duplicated 3′ CS, there was a 6,601-bp sequence encompassing the qnrA6 gene (excluding the IS10 sequence) that shared 69% homology with the sequenced fragment (6,358 bp) harboring the chromosomal qnrA2 gene of a strain of S. algae (24), with sequence parts reaching 91% homology (Fig. 3B). However, downstream of qnrA6, a sequence of 1,240 bp showed more homology with the Shewanella woodii chromosome (69%) than with that of S. algae (55%) (Fig. 3B). The sequence analysis of gentamicin-resistant clones from PsB/3 revealed that the aac(6′)-Ib′ gene cassette was located at the first position in a class 1 integron upstream from the aph1-IAB gene cassette. This aac(6′)-Ib′ gene was framed by a perfect direct repeat of 174 bp (Fig. 3C). The other gentamicin-resistant strains showed a similar organization, since the PCR amplification with primer pair 5′CS/APH(3′)-IR gave an amplicon of ca. 1,850 bp (data not shown). In contrast, the seven gentamicin-susceptible strains harbored an integron with only the aphIA-IB gene cassette (5′CS/APH(3′)-IR, ca. 1,000 bp).

DISCUSSION

This work extends a preliminary report on the beginning of an outbreak due to several clones of multidrug-resistant P. stuartii in a Tunisian burn unit (41). In the past decades, Gram-negative organisms have emerged as the most common etiologic agents of invasive infections, and antibiotic-resistant organisms have increasingly been isolated in burn units (10). However, very few studies have reported endemic or epidemic situations due to P. stuartii in these wards (11, 41, 53). Most patients of this study suffered from severe burn injury (mean TBSA, greater than 40%) and thus exhibited risk factors for acquiring multidrug-resistant bacteria, i.e., antibiotic receipt, extended duration of hospitalization, and invasive procedures, together with immunosuppression induced by the burn injury (10, 50). In this context, P. stuartii infection was a significant cause in the global outcome of the disease, since more than 88% of the patients developed sepsis and 38% of them died. The aspirator used for suctioning respiratory secretions of the patients was shown to play a critical part in this outbreak since P. stuartii strains belonging to two of the three implicated clones were found on this device.

Most P. stuartii strains involved in this outbreak were highly resistant to broad-spectrum cephalosporins such as ceftazidime. This resistance, first attributed to an ESBL (41), was actually due to the expression of the same transferable cephalosporinase, CMY-16. Plasmid-mediated cephalosporinases are not easily detectable in natural AmpC producers that can overexpress their own enzymes. CMY-16 is a variant of the CMY-2 lineage which derives from the chromosomal species-specific AmpC of Citrobacter freundii. CMY-16 might have evolved from either CMY-4 or CMY-12 (the closest homologues) since it differs from the latter enzymes by a single amino acid substitution (A171S and N363S, respectively) and from CMY-2 by two substitutions (A171S and W221R) (14). CMY-2 is the most common plasmid-mediated AmpC enzyme worldwide. It has been most often described in E. coli and AmpC-missing Enterobacteriaceae, such as Klebsiella pneumoniae (6) and Proteus mirabilis. Thus, recently, a large outbreak probably related to the dissemination of a single P. mirabilis clone producing the CMY-16 enzyme was reported in several hospitals and long-term care facilities in Italy (28). However, three CMY-2-producing P. stuartii strains have recently been described in Algeria, a country neighboring Tunisia (20). In our study, 15 P. stuartii strains involved in this outbreak also demonstrated low-level resistance to quinolones, due to the presence of a qnrA6 gene. To date, seven alleles of a qnrA-like gene have been identified (http://www.lahey.org/qnrStudies) (21). The sequence of QnrA6 was previously described in P. mirabilis (5). In 2006, five qnrA6-producing P. stuartii strains have been isolated at the Sahloul Hospital, Sousse, Tunisia (13). Of them, three harbored a 120-kb plasmid and two a 55-kb plasmid. However, no bla_CMY gene was mentioned. Very recently, a set of genes, including the bla_NDM-1 carbapenemase together with bla_CMY-16 and qnrA6, was found to be located on a conjugative IncA/C plasmid of 150 kb isolated from a K. pneumoniae strain in Switzerland (37). Our S1-PFGE experiments associated with hybridization assays confirmed the location of qnrA6 on the same plasmid (the IncA/C plasmid) that carried the bla_CMY-16 gene of PsB and PsC strains. The qnrA family of genes has previously been described on IncA/C plasmids, usually upstream from a common region 1 that is part of sul1-type intregrons (34). Nevertheless, the few descriptions of the qnrA gene environment have shown the close presence of an ISCR1 (44). In our study, the distance between the ISCR1 and qnrA6 reached 3,585 bp. Thus, for its expression, the qnrA6 gene certainly did not use the promoter found in the 3′ end of ISCR1 (40) but probably used the −35 and −10 sequences named pOUT located at the 5′ end of IS10 (43) (Fig. 3). Several alleles, qnrA2 to qnrA5, have been found in strains of the aquatic bacterium S. algae (24, 36). However, our study is the first description of the qnrA6 environment, and our data reinforce the hypothesis that S. algae is the natural reservoir for qnrA genes.

The P. stuartii strains harbored multidrug resistances which were all transferred to E. coli via a conjugative IncA/C plasmid. A single strain (PsA/4) did not yield any transconjugants, but our PCR experiments confirmed the presence of a nontransferable IncA/C plasmid lacking the tra genes required for the transfer mechanism. In the literature, the bla_CMY-2-like genes are often located on multidrug resistance plasmids belonging to the IncA/C family (6), one of the replicons that is the most commonly reported worldwide (6, 45). Nevertheless, other incompatibility group plasmids have been reported, and several studies have also described chromosomal locations of bla_CMY-2-like genes (8, 14, 32).

Using comparative sequence analysis, Fricke et al. (17) suggested that the IncA/C plasmids have evolved from a common ancestor through stepwise integration of horizontal transfers. Their results indicate that recent members of the IncA/C plasmid family have acquired resistance gene arrays in a conserved plasmid backbone. In our study, the IncA/C plasmids found in the three clones were very closely related, as indicated by (i) similar PstI-digested plasmid profiles, (ii) identical location and surrounding sequences of the bla_CMY-16 gene in the 195-kb and 175-kb plasmids, and (iii) 100% identical sequences of multiple genes, i.e., repA, aphA-6, aphA1-IAB, floR, and strAB, in the five representative strains. In particular, the aminoglycoside enzyme APH(3′)-VIa, rarely found in the Enterobacteriaceae, is a new variant compared to those previously described. In addition, all plasmids of representative strains, including PsA/4, contained the integron In633. For these reasons, it is likely that the different IncA/C plasmids found in the three clones of P. stuartii are very closely related, reflecting a very recent evolution. Recombination events may explain the loss of conserved sequences used for defining the backbone (52) and variations in plasmid size. In addition, the 11-kb insertion leading to the formation of complex class 1 integron and acquisition of quinolone resistance is probably due to the mobilization of qnrA6 onto the IncA/C plasmid via the insertion sequence ISCR1 (46). Recently, Toleman and Walsh (47) suggested that the ISCR elements that belong to the IS91 family could be key players in IncA/C plasmid evolution.

Furthermore, the strains of ATB2 and ATB4 were susceptible to gentamicin. In these isolates, the aac(6′)-Ib′ gene was not present at the first position of a class 1 integron. This absence may be due to either an acquisition or an excision of the corresponding cassette catalyzed by the IntI1 integrase, depending on the selective pressure (19). In our study, IntI1 possessed the amino acids P32 and H3, and the recent work of Jové et al. (23) showed that this variant displayed higher integration than excision activities. In addition, the direct repeat sequences framing the aac(6′)-Ib′ gene might facilitate its mobility.

In conclusion, our work describes a complex outbreak due to the concomitant dissemination of three clones and of a highly transferable IncA/C₂ plasmid, confirming the contribution of these plasmids to the global spread of antibiotic resistance. It further demonstrates that IncA/C₂ plasmids tend to evolve by deletions or acquisition of mobile genetic elements over a limited period of time. The variability of some resistance genes located on this plasmid illustrates the plasticity of these genetic platforms and thus highlights the need for full multilevel characterization of the genetic environments of antibiotic resistance genes in order to understand their spread. This paper contributes to the extension of our knowledge on the dissemination, distribution, and evolution of the IncA/C plasmid and to the origin and evolution of qnrA-like genes. Furthermore, our findings emphasize the need to screen for transferable ampC in strains of natural AmpC producers exhibiting a markedly high level of resistance to broad-spectrum cephalosporins.