From August 2004 to May 2005, a total of 66 ESBL-producing K. pneumoniae were isolated from patients in an intensive care unit (ICU) of the Kangbuk Samsung Hospital (Seoul, Korea), which is a tertiary care hospital with 645 beds. Of these, 43 isolates were preserved and included in this study (Table 1). Twenty-nine isolates were from sputum, and 11 isolates were from urine; the others were isolated from blood or bile. Thirty-eight K. pneumoniae isolates, which were isolated at another three Korean tertiary hospitals as a part of a surveillance study that was conducted in 1999 and 2004, were also investigated in this study for comparisons (Table 2). Thirty-one isolates were collected from the Samsung Medical Center (Seoul, Korea), and others were from the Dong-A University Hospital (Busan, Korea) and the Chungbuk National University Hospital (Cheongju, Korea). Of these 38 surveillance isolates, 15 produced ESBL and the others did not. For screening for ESBL, the GNS-650 card (Vitek system, bio-Merieux, Hazelwood, MO, U.S.A.), which is based on the use of cefotaxime and ceftazidime with or without clavulanic acid, was used. For a reference, K. pneumoniae ATCC 13883 was also analyzed. Thus, a total of 82 K. pneumoniae isolates were analyzed.

We selected five housekeeping genes to establish the MLST scheme; rpoB (RNA polymerase β-subunit), gyrA (DNA gyrase subunit A), gapA (glyceraldehydes 3-phosphate dehydrogenase A), groEL (GroEL protein), and gyrB (DNA gyrase subunit B). The primers we used for amplification of the gene fragments are shown in Table 3. PCR amplifications were carried out under the following conditions: 35 cycles of denaturation at 94℃ for 30 sec, annealing at 50-55℃ for 30 sec, and extension at 72℃ for 1 min; this was all preceded by a 5 min denaturation at 94℃ and then it was all followed by a 5 min final extension at 72℃. The PCR products were purified using a PCR purification kit (Bioneer, Daejeon, Korea), according to the manufacturer's recommendations, and then the products were sequenced on an ABI3700 DNA sequencer (Applied Biosystems, Foster City, CA, U.S.A.).

The raw sequences were concatenated and edited by using the EditSeq and MegAlign programs (DNASTAR, Madison, WI, U.S.A.). For each locus, distinct allele sequences were assigned as an arbitrary allele number. Each isolate was characterized by its allelic profile, which was represented as a series of 5 integers corresponding to the alleles at each of the loci, in the order of rpoB, gyrA, gapA, groEL, and gyrB. The sequence type (ST) was designated for each unique allelic profile. Clusters of related STs sharing three or more identical alleles out of the five were defined as clonal complexes (CCs), which were determined by the eBURST program (http://eburst.mlst.net/loci.asp) (15). The d_N/d_S ratios, with d_N being the number of nonsynonymous substitutions per nonsynonymous site and d_S being the number of synonymous substitutions per synonymous site (16), were determined by using the START program (http://outbreak.ceid.ox.ac.uk/software.htm). The index of association (I_A) was also calculated with the START program to test to what extent the K. pneumoniae population was clonal (17). I_A is a measure of the variance in the number of allelic mismatches relative to that expected under the hypothesis of panmixia (random association), where the mean number of pairwise allelic mismatches is a measure of genetic distance. When the I_A value does not deviate significantly from zero, the alleles are in linkage disequilibrium, i.e., there is frequent recombination or non-clonal population structure.

The 43 ESBL-producing K. pneumoniae isolates recovered from an outbreak at the Kangbuk Samsung Hospital were characterized by PFGE as described elsewhere (18, 19). Agarose plugs containing genomic DNA were digested with SmaI (Gibco, BRL, Gaithersburg, MD, U.S.A.), according to the manufacturer's recommendations. The DNA restriction fragments were separated with using a CHEF-Mapper apparatus (Bio-Rad Lab.), at 6 V/cm for 22 hr. The PFGE patterns were interpreted with using the published criteria (20). A divergence in more than three bands was interpreted as indicative of different PFGE types.

The bla genes related to ESBL enzymes were assayed by PCR with the corresponding primers for the TEM, SHV, CTX-M, and OXA ESBL types, as described previously (21). To specify the subtype of the bla genes, the amplified PCR products were sequenced on both strands as was done in the MLST analysis. The amino acid sequences were deduced from the nucleotide sequences using the MegAlign program (DNASTAR), and they were compared with the database of the website (http://www.lahey.org/Studies/).

The allele sequences of K. pneumoniae ATCC13883 have deposited in the GenBank database under accession numbers DQ673324 to DQ673328.

Sequence data were obtained for the five selected loci from all 82 K. pneumoniae isolates included in this study. The nucleotide sequences at each locus ranged from 271 bp (gapA) to 752 bp (gyrA). The number of different alleles ranged from 9 (gapA) to 20 (gyrA). The proportion of sites at which nucleotide variation was observed within each locus that was examined ranged from 3.93% (rpoB) to 8.49% (groEL) (Table 4). By calculating the d_N/d_S ratio, the degree of selection operating on each locus was evaluated. No nonsynonymous substitutions were found in the gapA and gyrB loci, so the d_N/d_S ratios could not obtained for them. The d_N/d_S ratios for the other loci were significantly less than 1 (Table 4), indicating there was no strong positive selective pressure on the selected genes. Thus, these five loci could be appropriate for the K. pneumoniae MLST scheme. The index of association (I_A) was used to test for linkage disequilibrium between the alleles at the five loci. The I_A value was 1.951 for all the K. pneumoniae isolates. This value was reduced to 0.524 when including only one representative isolate for each sequence type (STs), and this reflects the moderate linkage equilibrium or clonal population structure of the K. pneumoniae isolates included in this study.

Based on the nucleotide variations of the five genetic loci, a total of 37 different STs could be identified among 82 K. pneumoniae isolates (Table 1, 2). The majority of these (29 out 37 STs, 78.4%) were represented by single isolate. Among the STs shared by multiple isolates, the most frequently encountered were ST2 (25 isolates, 30.5%), ST20 (8 isolates, 9.8%), and ST15 (5 isolates, 6.1%). Other STs (ST9, -11, -13, -14, and -21) represented two to four isolates. These STs were grouped, using the eBURST program, into CCs that consisted of all the isolates of the STs that shared three or more identical alleles (Fig. 1). As a result, six CCs were identified among the 82 K. pneumoniae isolates. CC20 included six STs (ST17, -18, -20, -21, -26, and -32), of which ST20 was assumed as a founder. Fourteen isolates (17.1%) were included in the CC20. CC2 included four STs (ST2, -13, -15, and -37), of which ST2 was assumed as a founder. Thirty-four isolates (41.5%) belonged to the CC2, but most of them were isolated at the Kangbuk Samsung Hospital during an outbreak. CC36 also included 7 isolates (8.5%) of four STs (ST4, -9, -31, and -36), of which ST36 was assumed as a founder. The other three CCs each consisted of two STs (Table 1, 2, Fig. 1).

PFGE was performed for 43 outbreak isolates from Kangbuk Samung Hospital. Among the 43 outbreak isolates, seven PFGE types (A to G) were identified (Fig. 2 and Table 1). The PFGE patterns were well concordant with the STs in the MLST. Two subtypes, AA and DD, could be differentiated from A and D, respectively. All 31 isolates of the PFGE type A belonged to CC2, which were differentiated into ST2 or ST15. ST2 and ST15 were single-locus variants of each other, differing at rpoB. One isolate (KS412-6) of PFGE type AA belonged to ST13, a single-locus variant of ST2 with a different allele in gyrA. PFGE type B was correlated with ST36 of CC36. Isolates of the PFGE type D and DD showed ST11 and ST14, respectively, both of which belonged to CC11-14. Both isolates of the PFGE type G belonged to CC20, but they could be differentiated into ST17 and ST18, which are different at gyrA. The other three PFGE types, C, E, and F, represented single isolates.

Of the 43 ESBL-producing K. pneumoniae isolates recovered from the Kangbuk Samsung Hospital, all but one produced SHV enzymes (29 isolates of SHV-1, 5 of SHV-5, 8 of SHV-11, and one each of SHV-12 and SHV-14). All but four isolates produced CTX-M-14 and 10 isolates were positive to bla_TEM-1 (Table 1). One isolate (KS412-4) was negative on PCR assays for the four bla genes tested in this study.

All 33 outbreak isolates of CC2 and PFGE type A produced SHV and CTX-M enzymes. All the isolates produced CTXM-14, and all but three isolates produced SHV-1. The other three isolates produced SHV-5. Nine K. pneumoniae isolates produced TEM-1. Except for one isolate (KS502-5) of CC3-16 and PFGE type F, all the TEM-1-producing isolates belonged to CC2 and they showed PFGE type A. Four isolates of CC36 and PFGE type B produced SHV-11 and CTX-M-14. Of the three isolates of CC11-14, two of ST11 and the PFGE type D produced SHV-11 and CTX-M-14, while one of ST13 and PFGE type DD produced SHV-1 and CTX-M-14. Two isolates of CC20 and PFGE type G produced both SHV-11 and CTX-M-14, or only SHV-11.

Of the 38 K. pneumoniae isolates collected from three Korean hospitals as a part of a surveillance program, 15 isolates produced ESBL. All 15 ESBL-producing isolates produced SHV enzymes. Among these surveillance isolates, SHV-12 was the most frequent (10 isolates), and this was in contrast to the outbreak isolates, of which SHV-1 was the most common. Five isolates produced CTX-M-14, and 8 isolates were positive to bla_TEM-1 (Table 2). Of the 15 ESBL-producing isolates, nine belonged to CC20, six isolates belonged to ST20, two belonged to ST21, and one belonged to ST26. There might be no obvious correlation between CC and the bla genes. Twenty-three K. pneumoniae isolates did not produce ESBL. These ESBL-negative isolates were more diverse than were the ESBL-producers. Only three isolates belonged to CC20 (ST20 and ST32), and three isolates belonged to CC2 (ST13 and ST37). Most of the isolates formed singletons in MLST analysis.