Horizontal Gene Transfer in a Polyclonal Outbreak of Carbapenem-Resistant Acinetobacter baumannii

Clinical samples.The Westmead (WM) sample collection consists of 177 pure cultures of Acinetobacter spp. collected from patients, mostly in the ICU, at Westmead Hospital (Sydney, Australia) between 1995 and 2000. A small number of strains were selected to represent each ApaI PFGE type and were subjected to detailed analysis, including repeat PFGE; these were renamed (e.g., WM96 for a representative isolate from 1996). Nine carbapenem-resistant Acinetobacter isolates from ICU patients at Prince of Wales (PW) Hospital (Sydney, New South Wales, Australia) and Royal Brisbane (RB) Hospital (Brisbane, Queensland, Australia) were also included (PW01a to -c and RB01 from 2001; RB02a to -e from 2002).

Strain and species identification and antibiotic susceptibility testing.After initial identification in source laboratories, Acinetobacter typing and species identification were performed at the Leiden University Medical Center (LUMC) by random amplification of polymorphic DNA (RAPD) analysis and by high-resolution genomic fingerprinting by AFLP and amplified rRNA gene restriction analysis (ARDRA). RAPD analysis for strain identification (typing) used primers M13 and DAF4 as described previously (16). ARDRA was performed by restricting the 16S rRNA gene with CfoI, AluI, MboI, RsaI, and MspI (9). AFLP fingerprints and the combined ARDRA restriction profiles were compared against AFLP and ARDRA libraries of reference strains in the LUMC collection, which had been identified by DNA-DNA hybridization. MICs were determined by Phoenix NMIC-ID (Becton Dickinson, NJ) and by Etest (AB Biodisk, Solna, Sweden).

Pulsed-field gel electrophoresis.Genomic DNAs of all Acinetobacter isolates were prepared in agarose blocks and digested with the restriction enzyme ApaI (New England Biolabs, MA). DNA fragments were separated by electrophoresis in a CHEF-DR III (Bio-Rad Laboratories, CA) system, and results were interpreted according to the criteria of Tenover et al. (45).

Cell fractionation.The cell fractionation method used was modified from that of Osborne and Silhavy (31). Logarithmic-phase cells from 100 ml of nutrient broth containing 20 μg/ml ampicillin were pelleted (at 10,000 × g for 20 min at 4°C) and resuspended in 1 ml of 30 mM Tris-HCl-20% sucrose (pH 8.1); then 100 μl lysozyme (100 mg/ml in 0.1 M EDTA [pH 7.3]) was added, and the mixture held on ice for 30 min. After centrifugation to pellet spheroplasts (at 10,000 × g for 15 min at 4°C), the supernatant, enriched for periplasmic content, was stored at −20°C. The spheroplast pellet was held at −70°C for 30 min; then it was reconstituted in 3 ml of 3 mM EDTA (pH 7.3) and sonicated at a 50% cycle for 60 s. After centrifugation (at 35,000 × g for 60 min at 4°C), the supernatant (cytoplasmic extract) was stored at −20°C, and the pellet (membrane fraction) was resuspended in 1 ml of 0.25 M Tris (pH 7.5). Protein concentrations were determined using the Coomassie Plus protein assay according to the manufacturer's instructions (Pierce Biotechnology Inc., IL). Subfractions were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis to ensure uniformity and to confirm that there was no significant contamination of periplasmic extracts with cytoplasmic proteins.

Hydrolysis assay (Masuda testing).A disc assay adapted from the method of Masuda et al. (25), as modified by Bou et al. (4), was used for the detection of a carbapenem-hydrolyzing activity in resistant isolates. An antibiotic disc (10 μg imipenem [IPM], 10 μg ampicillin or 30 μg aztreonam; BBL, MD) was placed at the center of a Mueller-Hinton agar plate (Oxoid, United Kingdom) spread with a sensitive Escherichia coli indicator strain. Filter discs loaded with increasing amounts (100 μg, 150 μg, 200 μg) of the periplasmic extract or with 10 μl of phosphate-buffered saline (PBS) were placed 10 mm from the antibiotic disc, within the predicted zone of inhibition, and the plate was incubated overnight at 37°C. Growth of the indicator strain within the zone of inhibition indicated hydrolysis of the antibiotic.

EDTA synergy tests.EDTA disc synergy tests were performed as previously described (23). Briefly, test or control isolates (Klebsiella pneumoniae expressing the IMP-4 metalloprotease or carbapenem-sensitive Pseudomonas aeruginosa) suspended in 0.9% (wt/vol) saline (a McFarland standard of approximately 0.5; 1.5 × 10⁸ CFU/ml) were swabbed onto Mueller-Hinton plates. A disc soaked in EDTA (10 mM) was placed 10 mm from a 10-μg IPM disc in the center of each plate. The IPM inhibition zone, evident after overnight incubation at 37°C, was augmented by EDTA in metalloenzyme-expressing strains.

DNA purification.Crude whole-cell lysates, made from overnight cultures of bacterial isolates by resuspending cells in 1 ml water to a 3.0 McFarland standard and incubating at 96°C for 30 min, were used as DNA templates for PCR. Genomic DNA for PCR was extracted from selected isolates using a Wizard genomic DNA purification kit (Promega). Plasmid DNA for electroporation was prepared using a Wizard Plus mini-prep kit (Promega). Plasmid DNA for profiling was prepared using a modification of published methods (21, 43). Cells from overnight growth on horse blood agar plates were suspended in 50 μl of TE buffer (10 mM Tris-HCl, 1 mM EDTA [pH 8.5]), and 200 μl of lysis solution (3% [wt/vol] sodium dodecyl sulfate, 50 mM Tris [pH 12.5]) was added. After 90 min (for A. baumannii) or 60 min (for Enterobacteriaceae containing control plasmids) at 60°C, lysed mixtures were extracted with phenol-chloroform-isoamyl alcohol (25:24:1). Samples were electrophoresed through 0.6% (wt/vol) agarose in 1× Tris-acetate-EDTA buffer at 100 V for 6 h, and the buffer was replaced hourly.

PCR amplification.PCR primers are listed in Table 1. Amplification was carried out in 50-μl volumes with 1× reaction buffer, 1 U of Taq polymerase (Bioline, London, United Kingdom), 200 μM each deoxynucleoside triphosphate, 1.5 mM MgCl₂, 10 pmol of each primer, and 4 μl of lysed cells as the template. PCR conditions generally were as follows: a hot start at 94°C for 5 min; 35 cycles of 30 s at 94°C, 45 s at either 41°C (bla_GIM-1), 48°C (bla_OXA-51), 50°C (ISAba1), 52°C (bla_OXA-58), or 55°C (others), and 60 s at 72°C; and a final step of 10 min at 72°C. The intI genes of class 1 to class 3 integrons, and class 1 integron cassette arrays, were amplified as previously described (10) using the primers listed in Table 1.

Conjugation.Transfer of plasmids to E. coli UB5201Rf, a spontaneous rifampin-resistant derivative of UB5201 (8a), was attempted using a filter mating method (adapted from reference 8). Recipient cells were harvested from overnight growth on a nutrient agar plate, and donor cells were harvested from 1 ml of an overnight culture in nutrient broth. After a wash with PBS and centrifugation, cell pellets were resuspended in 0.1 ml PBS, the suspensions were mixed, and 0.1 ml was spotted onto a nitrocellulose filter placed on a blood agar plate. Following incubation at 37°C for 4 h, cells were washed off the filter with PBS, and 0.1-ml aliquots of appropriate dilutions were spread on nutrient agar plates containing 80 μg/ml rifampin (Sigma-Aldrich, MO) and 20 μg/ml ampicillin (CSL Ltd.).

Electroporation.Plasmid DNA (25 ng) was mixed with 200 μl of electrocompetent E. coli DH5α prepared by standard methods (37), and the mixture was pulsed (25 μF, 2.5 kV, 200 Ω, 4 s). After addition of 1 ml of nutrient broth and incubation for 60 min at 37°C with shaking, the mixture was plated onto nutrient agar containing 20 μg/ml ampicillin.

DNA sequencing and analysis.PCR products were sequenced using a Dye-Terminator sequencer (Applied Biosystems Inc., CA) and DNA templates prepared according to the manufacturer's specifications. Sequence comparisons were carried out using BLAST searches (www.angis.org.au ; www-is.biotoul.fr/ ).

Nucleotide sequence accession numbers.The sequences of the cassette arrays have been submitted to GenBank under the following accession numbers: PW01c (aacC1-orfP-orfQ-aadA1), EF015496; WM98a (blaP1), EF015497; WM98b (dfrA17-aadA5), EF015498; WM99c (aacC1-orfP-orfP-orfQ-aadA1), EF015498. The ISAba1-bla_OXA-23 sequence from isolate WM99c has been submitted under accession no. EF015500.

A common phenotype in unrelated cocirculating strains.Prior to about 1999, most clinically significant Acinetobacter isolates from Westmead Hospital were resistant to several antibiotics, including ticarcillin-clavulanate and gentamicin, commonly used in combination in this hospital, while occasional isolates were relatively antibiotic susceptible (see Table 2 for examples). From 1999, isolates that were resistant to multiple antibiotics, including the carbapenems imipenem and meropenem, began to appear. Nine isolates from Brisbane (RB) and elsewhere in Sydney (PW) collected after 1999 were also carbapenem resistant.

ApaI PFGE fingerprint analysis performed on all 177 WM isolates distinguished eight pulsotypes (I to VIII) by the criteria of Tenover et al. (45) (Fig. 1). Three additional pulsotypes (IX to XI) were found among the nine isolates from the other two hospitals, with type X common to both sites. The differences between types VI to X were relatively minor. Pulsotypes III, IV, and V, all of which were relatively antibiotic susceptible (Table 2), were seen rarely (Table 3). The initial carbapenem-susceptible pulsotype I appeared to be gradually outnumbered by pulsotype II, although both pulsotypes were still present when the first of the carbapenem resistance pulsotypes appeared in 1999. Each of the three carbapenem resistance pulsotypes (VI to VIII) appeared to displace its predecessor within a year. Different pulsotypes thus clearly cocirculated in our ICU, and the carbapenem-resistant phenotype, once established, appeared to persist in successive clonal waves.

• Open in new tab
• Download powerpoint

FIG. 1.

ApaI PFGE of selected isolates. Representatives of different pulsotypes are shown as indicated. Lanes 1 to 8, WM; lanes 10 to 15, RB; lanes 17 to 19, PW; lanes 9 and 16, Lambda ladder pulsed-field gel electrophoresis marker (New England Biolabs).

Detailed genotypic strain and species identification by RAPD, AFLP, and ARDRA was applied to selected WM isolates chosen as representative of pulsotypes that immediately predated and/or cocirculated with the initial carbapenem-resistant strains (Table 4). By using AFLP and ARDRA library typing for species identification, most of the tested isolates proved to be A. baumannii, while isolates with the rare pulsotypes III and IV corresponded to the unnamed genomic species 3 and 13TU, respectively. The carbapenem-resistant phenotype was confined to A. baumannii. The detailed comparative typing by RAPD and AFLP generally supported the ApaI PFGE typing data, apart from the carbapenem-resistant strains WM99a (pulsotype VI) and WM99b (pulsotype VII). These isolates were on the borderline of the Tenover criteria (45) for relatedness, and RAPD and ARDRA data suggested epidemiological relatedness. Indeed, taken together, the typing data suggest that WM99a (type VI) and WM99c (nominally, type VII) are minor variants of the same strain.

Of note, cluster analysis of the AFLP profiles of WM96 and WM98 (both pulsotype I) indicated >82% relatedness (generally used as the A. baumannii clone delineation level) to strains of the widespread epidemic-associated A. baumannii EU clone II (29) (data not shown).

Carbapenem-resistant isolates contain bla_OXA-23.Outer membrane protein preparations revealed no consistent difference between carbapenem-susceptible and carbapenem-resistant isolates (data not shown), suggesting that outer membrane variation is an unlikely primary explanation for the resistant phenotype. Periplasmic extracts from selected susceptible isolates hydrolyzed ampicillin, but neither imipenem nor aztreonam, in vitro (in a Masuda test). In the same assay, resistant isolates hydrolyzed both ampicillin and imipenem but not aztreonam, consistent with the presence of a periplasmic carbapenemase of the OXA type.

EDTA did not augment the imipenem susceptibilities of representative carbapenem-resistant strains from WM, PW, and RB, arguing strongly against any contribution from a metalloenzyme. No PCR products were obtained using primers to detect the genes encoding the metalloenzymes SPM-1 and GIM-1 or some of the IMP and VIM variants that are now well established (49) (Table 1). PCR amplification using proven primers to detect bla_OXA-24 (2) and bla_OXA-58 (17) indicated that neither gene was present in any of the isolates for which results are given in Table 5. By contrast, bla_OXA-23 was present in all carbapenem-resistant strains (n = 13). Sequencing of the bla_OXA-23 genes amplified from pulsotypes VII and XI (WM99c and PW3, respectively) indicated that they were both identical to the original bla_OXA-23 gene (ARI-1; GenBank accession no. AJ132105) (11).

The “forward” primer used here, and in most other published studies, to amplify the bla_OXA-23 gene actually lies outside the gene itself in the insertion sequence ISAba1. Primers internal to ISAba1 (ISAba1L2 and ISAba1R2) (Table 1) gave the expected product with all of the A. baumannii strains examined, including those that were carbapenem sensitive, but not with the 13TU or DNA group-3 strains. PCR primers ISAba1R2 and OXA23-R (Table 1) amplified a product that included most of ISAba1 upstream and a complete bla_OXA-23 gene in all carbapenem-resistant strains listed in Table 5 but in none of the carbapenem-susceptible strains. Interestingly, this first bla_OXA-23 sequence (ARI-1; AJ132105) differs from all other bla_OXA-23 sequences in the databases, and from all those that we sequenced (e.g., WM99c; accession no. EF015500), by the absence of 7 bp (CTCTTTT) immediately adjacent to the inverted repeat of ISAba1.

All A. baumannii isolates were positive for bla_OXA-51-like genes, while bla_OXA-23 was found in all carbapenem-resistant and no carbapenem-susceptible A. baumannii strains. Primers in two different locations within ISAba1 (JVFTnox and TNORF2-1), paired with a primer in bla_OXA-51 (3′OXA51-like-all R), generated amplicons of the correct size for a direct linkage of ISAba1 upstream of bla_OXA-51 in several carbapenem-resistant strains and no carbapenem-susceptible strains. However, bla_OXA-51 was disrupted by a large (>1-kb) insertion in WM99c, and several carbapenem-resistant strains were negative by PCR for the ISAba1-bla_OXA-51 sequence (data not shown). Thus, while the presence of an intact and complete ISAba1-bla_OXA-23 sequence correlated exactly with resistance for all A. baumannii isolates examined, bla_OXA-51 was massively disrupted in at least one carbapenem-resistant strain and lacked any demonstrable association with ISAba1 (and the promoter it provides) in others.

Class 1 integrons and cassette arrays in Acinetobacter strains.Since gene cassettes in class 1 integrons are common in Acinetobacter spp. (see, e.g., references 15 and 22), all the strains in Table 5 were tested for these features. A multiplex PCR method targeting the specific integrase genes (intI) for class 1, -2, and -3 integrons (10) failed to identify intI2 (encoding the class 2 integrase) or intI3 (class 3) in any isolate. However, intI1 was identified in all carbapenem-sensitive isolates and all but two of the carbapenem-resistant isolates (WM00 [type VIII] and RB02c [type IX]). WM00 and RB02c also failed to yield cassette array products following several attempts using published primers targeting the 5′- and 3′-conserved segments of class 1 integrons (50). Cassette arrays were found in all intI1-positive isolates except for the single representatives of types III (WM97b) and V (WM98c) and one of the pulsotype VI strains (WM99a; very closely related to WM99c) (Table 5). WM97a (A. baumannii) and the unrelated WM98b (13TU) both contained the two-cassette array dfrA17-aadA5, while the other 13TU strain, WM98a, had a single cassette, blaP1 (the product of which is also known as PSE-1 or CARB-2).

A 3.1-kb amplicon was common to two almost indistinguishable carbapenem-sensitive strains (both A. baumannii, pulsotype I) and most of the carbapenem-resistant isolates (1 of 2 type VI isolates, 4 of 5 type IX isolates, and single examples of types VII and X), while PW01c (pulsotype XI) yielded a 2.6-kb amplicon. The PCR products from WW99c (3.1 kb) and PW01c (2.6 kb) were completely sequenced and found to contain closely related arrays, aacC1-orfP-orfP-orfQ-aadA1 and aacC1-orfP-orfQ-aadA1, respectively (Table 5). AvaI restriction digestion of the 3.1-kb PCR products from representatives of each relevant pulsotype resulted in identical patterns (data not shown).

Plasmid profiling.Plasmid profiling has been suggested as an epidemiological marker for Acinetobacter spp. (14), and plasmids provide efficient vehicles for transferable resistance. Electrophoresis of plasmid preparations revealed a number of high-molecular-weight plasmids, along with smaller plasmids (illustrated for selected isolates in Fig. 2), but none were both unique to carbapenem-resistant (bla_OXA-23-positive) strains and present in all of them. Repeated attempts to transfer a plasmid from strain WM99c into E. coli UB5201Rf by conjugation, or into E. coli DH5α by electroporation, with selection for ampicillin (bla_OXA-23), were unsuccessful.

• Open in new tab
• Download powerpoint

FIG. 2.

Plasmid profiling of representative Acinetobacter strains. Lane 1, WM96; lane 2, WM97a; lane 3, WM98a; lane 4, WM99c; lane 5, PW01c; lane 6, RB02d; lane 7, RB01; lane M1, 1-kb marker (Gibco); lane M2, 2.5-kb Molecular Ruler (Bio-Rad); lane i, E. coli DH5α chromosomal DNA; lane ii, uncut pBC-SK (3.4 kb; Stratagene); lane iii, EcoRI-restricted pBC-SK; lane iv, R751 (53.0 kb); lane v, R934 (72 kb).