In April 2008 a carbapenem-nonsusceptible E. coli strain, marked as Eco2, was recovered from the gall bladder drainage of a 91-year-old man with dialysis-dependent end-stage renal disease, congestive heart failure, anemia, and peptic ulcer disease. A month earlier, the patient had been hospitalized with sepsis that developed after an infected heel wound had required amputation of the left leg below the knee. The patient was treated with ertapenem, metronidazole, colistin, and vancomycin. Acute cholecystitis developed, and the patient underwent cholecystostomy. During his hospital stay, the patient underwent screening for carriage of carbapenem-resistant Enterobacteriacae (CRE) as part of a routine infection control program aimed at limiting the spread of CRE. Two rectal swabs were collected 1 week apart. The first swab specimen was negative for CRE by culture, and the second swab specimen showed a carbapenem-resistant K. pneumoniae strain (marked Kpn1), which was PCR positive for bla_KPC. One month after the patient’s admission, a carbapenem-nonsusceptible E. coli (Eco2) was isolated from drainage at the cholecystectomy site, which prompted this study.

Microbiologic and molecular investigations (pulsed-field gel electrophoresis [PFGE], DNA isolation, isoelectric focusing analysis [IEF], PCR detection of resistance genes, plasmid isolation, transformation, and Southern analysis) were performed as described (2,4,7). The carbapenem-nonsusceptible E. coli isolated from the clinical specimen (Eco2) was initially identified by Vitek-2 (bioMérieux, Marcy-l’Etoile, France) as resistant to imipenem (MIC>16 mg/L). Further antimicrobial-drug susceptibility testing of the strain by using agar dilution and Etest (AB Biodisk, Solna, Sweden) showed MICs in the resistant range for ceftriaxone and aztreonam; in the intermediate range for ceftazidime and piperacillin/tazobactam; and in the susceptible range for cefepime, ertapenem, meropenem, imipenem, aminoglycosides, quinolones, tigecycline, and colistin (Appendix Table). IEF identified 2 β-lactamases with isoelectric pH values of 5.4 and 6.7, corresponding to those of TEM-type and KPC. β-lactamase gene PCR screening and sequencing indicated the presence of bla_TEM-1 and bla_KPC-3. Results of screening for other β-lactamase genes were negative.

Transformation of plasmids purified from Eco2 into an E. coli DH10B recipient strain (Eco2-T) indicated transfer of a single plasmid that encoded these bla_TEM-1 and bla_KPC-3 (Figure 1, panel A), and increased the MICs of the recipient strains to broad-spectrum cephalosporins and carbapenems (Appendix Table). PFGE identified the genetic similarity between the colonizing Klebsiella (Kpn1) and a representative KpnST258 (isolate Kpn557 described previously [4]) (Figure 2). Susceptibility testing of Kpn1 reflected the extremely drug-resistant phenotype characteristic of isolates belonging to this clone (12) (Appendix Table).

We compared plasmids of Kpn1 and Eco2. Kpn1 carried 4 different plasmids, whereas Eco2 carried 1 plasmid that correlated with the 105-kb plasmid of Kpn1. Experiments to transform Eco2 and Kpn1 plasmids into an E. coli DH10B recipient, followed by selection on plates containing 100 µg/mL ampicillin and screening for bla_KPC-positive colonies, showed that DH10B was transformed with the105-kb KPC-3–encoding plasmid (Figure 1, panel A). This plasmid correlated in size with that of pKpQIL, the KPC-3-encoding plasmid of Klebsiella ST258 in Israel (13). Plasmid DNA restriction fragment length polymorphism showed that band patterns of the 2 KPC-3–encoding plasmids of Kpn1 and Eco2 were highly similar (98% similarity) (Figure 1, panel B), and Southern analysis with a bla_KPC-3 probe showed the same hybridization pattern (Figure 1, panel C).

We aimed to determine whether the patient’s gut was colonized with a carbapenem-susceptible E. coli strain, which would ultimately serve as the in vivo recipient of the bla_KPC-3-encoding plasmid. Thus, the first broth culture prepared (obtained before the patient was colonized with KPC-3–producing E. coli) was processed. Aliquots (0.1 mL) were directly plated onto a MacConkey agar plate (Hy-Labs, Rehovot, Israel). E. coli colonies isolated from the plate were restreaked onto a MacConkey agar plate, yielding an E. coli strain 7364 (Eco1) that was susceptible to all antimicrobial drugs tested (Appendix Table). PFGE DNA fingerprinting showed that Eco1 was 100% identical to the KPC-3-producing clinical strain Eco2, isolated from the clinical specimen (Figure 2). Plasmid analysis of this strain, however, proved that it lacked plasmid pKpQIL (results not shown).

Elements belonging to KPC transposon Tn4401, including tnpA, tnpR, ISKpn6, and ISKpn7 (14), were identified by PCR and sequencing on both KPC-3–encoding plasmids originating from Kpn1 and Eco2. These genetic determinants were absent in the susceptible Eco1. These data suggest that Eco1 has acquired pKpQIL from Kpn1 in the patient’s gut, leading to the formation of Eco2. Although acquisition of the plasmid increased MICs for imipenem, meropenem, and ertapenem considerably, it did not confer full resistance (Appendix Table) presumably due to copy number of the plasmid or the expression level of bla_KPC-3 in E. coli. Curing of pKpQIL from Eco2 was performed by sequential transfers at an elevated temperature (42°C). The cured strain, which lacked the KPC-encoding plasmid, showed full susceptibility to all antimicrobial drugs tested, similar to the Eco1 strain isolated from the patient’s gut flora.

The patient received a combination of 4 antimicrobial agents concomitantly (ertapenem, metronidazole, colistin, and vancomycin) during the period in which Eco1 acquired in vivo the plasmid pKpQIL, thereby becoming Eco2. We believe that the selection pressure imposed by these antimicrobial agents contributed to the sequence of events that led to plasmid transfer. We hypothesize that interspecies conjugation and antimicrobial pressure led to the preferential selection of Eco2, rather than Eco1, as a determinant of infection in this patient.

Interspecies KPC transfer can presumably occur through the dissemination of mobile genetic elements as has been described for transfer of the mecA gene between strains of Staphylococcus aureus (15). bla_KPC may spread through transfer by virtue of its location on the Tn4401 transposon (14), or by dissemination of the intact KPC-encoding plasmid, likely through natural conjugation. Multiple attempts to mimic the natural transfer of pKpQIL, the KPC-3-encoding plasmid from Kpn1 into Eco1 by using conjugation experiments were not successful. The isolation of an isogenic, antimicrobial drug–susceptible E. coli clone enabled us to decipher the natural order of the interspecies genetic transfer event

With increasing global spread of KPC-producing K. pneumoniae ST258, the likelihood increases of interspecies transfer of drug-resistance determinants into a highly fit E. coli clone. Such an event may have severe public health consequences, leading to elimination of any effective antimicrobial drug treatment against the most common human bacterial pathogens.