For ten clinical isolates, i.e. seven ESBL-producing clinical isolates collected during 2001 and three more recently collected isolates (Table ), the API20E codes 5205773 (isolates GA1, GA2, GA3, MN2, MN3 and VGM), 5205753 (isolates DHJ1 and DHJ3) or 5204673 (isolates RA and DBH) were obtained. The first code resulted in a weak identification as either E. aerogenes, K. pneumoniae or Raoultella (Klebsiella) planticola, the second code did not yield any identification, and the third code resulted in a weak but acceptable identification as K. terrigena. The isolates with the codes 5205773 and 5205753 were identified as K. pneumoniae by additional biochemical testing due to negative reactions for motility (tested in semi-solid agar) and ornithine decarboxylase. However, these isolates all possessed an inducible cefalosporinase, as detected on the antibiogram using a disk approximation test, a finding which strongly contradicts an identification as K. pneumoniae or Klebsiella sp.
Clinical data and phenotypic and genotypic identification results of the Klebsiella pneumoniae and Enterobacter aerogenes isolates used in this study
In fact, the first hint that these strains, phenotypically identified as K. pneumoniae
, were actually E. aerogenes
, came from tDNA-PCR based identification. Using this method, all isolates were identified as E. aerogenes
, and this by comparison of the obtained fingerprint – composed of amplified intergenic tRNA spacers of 101, 106, 111, 115, 121, 189, 190, 198 or 289, and 391 bp in length – with a library containing fingerprints of more than 3000 strains belonging to hundreds of species, available at http://allserv.ugent.be/~mvaneech/All_C.txt
Confirmation of this genotypic identification was obtained by 16S rRNA gene sequencing for five isolates (Table ). Analysis yielded a similarity of between 99.8% and 100% to E. aerogenes Genbank entries.
The presence of a genuine K. pneumoniae isolate in patient DHJ (Table ) further complicated the identification.
This observation lead us to carry out additional phenotypic testing. Using the hanging drop method for testing motility, a few motile cells were observed, and upon retesting in semi-solid medium, weak migration could be observed. Like most biochemical tests in the routine laboratory, ornithine decarboxylase is read after overnight or 24 hours of incubation, but when the incubation period was prolonged to up to 2–5 days, all isolates tested positive.
Because at present automated systems are frequently used for routine identification, a selection of six strains, containing four isolates of the study and two controls, i.e. phenotypically correctly identified E. aerogenes (LBV268) and K. pneumoniae (BG) isolates, were tested in two different systems, i.e. Vitek 2 (bioMérieux, Marcy l'Etoile, France) and Phoenix (BD Biosciences, Sparks, Md.). Both automated systems yielded the same results as the API20E, i.e. that the E. aerogenes isolates, aberrant due to a slow reaction for motility and ornithine decarboxylase, were misidentified as K. pneumoniae (Table ). This misidentification by the automated systems is not unexpected, since they are based on biochemical testing only and a reading time of 24 hours or less. The control strains were correctly identified.
Disk diffusion antibiotic susceptibility testing, carried out according the NCCLS guidelines, revealed basically the same resistotype for all isolates, characterized by resistance to ceftazidime and susceptibility to ceftriaxone. Additional resistance to aztreonam was observed for some isolates, reflecting the most dominant resistance patterns for the E. aerogenes
isolates in our hospital. All isolates were also found to carry high-level resistance to cefoxitin, which is highly unusual for Klebsiella
spp. Furthermore, the disk-approximation test with an amoxycillin-clavulanic acid disk close to β-lactam disks on Mueller-Hinton II agar, showed a combined pattern of synergy (broadening of the inhibition zone in the direction of clavulanic acid) and antagonism (flattening of the inhibition zone), which is suggestive for a combination of an ESBL and an inducible β-lactamase. Again, inducible β-lactamases are very rare in Klebsiella
spp. but typical for Enterobacter
spp. It should be noticed that this phenomenon will not be detected by automated MIC-determination systems like Vitek 2 and Phoenix. Using PCR and sequencing as described previously [2
], the presence of TEM-5 could be shown in the isolates of patients GA and MN, and SHV-4 in the isolate of patient DHJ.
The genotypic relationship of the phenotypically aberrant isolates was investigated with AP-PCR. Isolates GA1, GA2, GA3, MN2, MN3 and DBH were corresponding to Belgian clone BEI (Figure , pattern A), isolates VGM, DHJ1 and DHJ3 were closely related to Belgian clone BEII (Figure , pattern B, differing from pattern C, characteristic of clone BEII, by a single extra band) while isolate RA was not related to any of the others (Figure , pattern D). This genetic diversity among the phenotypically aberrant strains makes it probable that strains with this kind of aberrant phenotype are not restricted to a single clone within E. aerogenes.
Figure 1 RAPD analysis of Enterobacter aerogenes and Klebsiella pneumoniae strains used in this study Lane M: DNA molecular weight marker (100 base pair ladder). The E. aerogenes RAPD-types are indicated as A, B, C and D and the K. pneumoniae types are indicated (more ...)