INTRODUCTION
Klebsiella pneumoniae is responsible for a variety of diseases in humans and animals but the majority of infections are associated with hospitalization. However, serious community infections due to K. pneumoniae can affect previously healthy persons and recently, K. pneumoniae serotype K1 has been identified as a major aetiological agent in liver abscesses, most prominently in Taiwan and Korea [Reference Brook and Frazier1–Reference Fung4]. Typing of the capsular (K) antigen has been the classical method for characterizing K. pneumoniae strains but it is time-consuming and requires a large number of specific antisera. However, relatively few serotypes, particularly K1 and K2, appear to be associated with invasive strains. The rapid development of molecular methods available for epidemiological purposes, such as multilocus sequence typing (MLST), has given new insights into genotype-virulence relationships in several bacterial species and demonstrated genetic heterogeneity within strains of uniform serotypes [Reference Jaureguy5, Reference Sukhnanand6]. However, there are only a few reports on the clones of K. pneumoniae causing bacteraemia [Reference Brisse7].
It has long been recognized that bacterial isolates expressing a mucoid phenotype on agar culture are generally considered to have more virulence, greater lethality, and increased resistance to phagocytes in animal experiments [Reference Moses8]. K. pneumoniae strains often form large mucoid hypermucoviscous colonies (HV phenotype) on agar culture which tend to coalesce with prolonged incubation and form a ‘string’ of growth when picked up on a loop from the agar surface.
The magA gene was reported to be a novel virulence gene related to the HV phenotype in 2004 [Reference Fang9] but was later found to be restricted to serotype K1 and part of the capsular polysaccharide gene cluster for this serotype [Reference Struve10, Reference Yeh11]. RmpA and rmpB genes positively control the mucoid phenotype in of K. pneumoniae [Reference Nassif12] and both rmpA and magA are strongly associated with the HV phenotype [Reference Yu13]. This phenotype is now an accepted marker of virulence in clinical strains causing liver abscess and is characterized by serotype K1 and sequence type (ST) 23 [Reference Pomakova14, Reference Turton15]. However, the clinical consequences of the HV phenotype in bacteraemia have remained elusive. We therefore investigated the frequency of the HV phenotype in bacteraemic K. pneumoniae isolates and compared their characteristics (virulence genes, serotypes, MLST, antimicrobial resistance, clinical manifestations) with non-HV phenotypes.
METHODS
Bacterial isolates
From June to August 2010, a total of 33 K. pneumoniae isolates were each recovered from individual patients with bacteraemia hospitalized in 13 tertiary-care hospitals located in 10 provinces of Korea: Seoul St Mary's Hospital (n = 6), Yoido St Mary's Hospital (n = 1), Bucheon St Mary's Hospital (n = 1), St Vincent Hospital (n = 5), Uijeongbu St Mary's Hospital (n = 1), Incheon St Mary's Hospital (n = 4), Chung-ang University Yongsan Hospital (n = 2), Chung-ang University Hospital (n = 2), Chungju Medical Center (n = 1), Soonchunhyang University Kumi Hospital (n = 4), Ulsan University Hospital (n = 1), Chosun University Hospital (n = 2), and Yeungnam University Medical Center (n = 3). Medical records of all the patients enrolled in the study were retrospectively reviewed to identify clinical characteristics including the underlying illness, acquisition mode of infection (community vs. hospital), source of bacteraemia, and clinical signs and symptoms.
Case definitions
Community-acquired bacteraemia was defined as a positive blood culture taken on admission or within 48 h of admission [Reference Yu16] and fatal underlying illness was according to the McCabe classification [Reference McCabe and Jackson17]. Patients without a clinically evident source of bacteraemia were considered to have primary bacteraemia. Liver abscess was diagnosed by evidence of an intrahepatic abscess cavity by ultrasonography or computed tomography (CT).
Microbiological tests
K. pneumoniae isolates were sent by the participating hospitals on nutrient agar slants to a central laboratory in Seoul for confirmation of species identity and susceptibility to 15 antimicrobial agents (cephalothin, cefuroxime, cefoxitin, cefotaxime, ceftazidime, cefepime, piperacillin, piperacillin-tazobactam, imipenem, meropenem, amikacin, gentamicin, ciprofloxacin, trimethoprim-sulfamethoxazole, aztreonam) by the Vitek GNI system (bioMérieux Vitek, USA). Extended spectrum β-lactamase (ESBL) production was detected by an inhibitor-potentiated disk diffusion method, as described previously [Reference Song18]. The isolates were cultured on 5% sheep blood agar plate overnight at 37 °C and classified as HV or non-HV phenotype by the string test performed by two experienced technologists, using a bacterial loop to stretch a string of growth from the bacterial colony. Formation of a mucoviscous string of >5 mm was defined as a positive string test [Reference Fang9].
The presence of genes encoding virulence factors (rmpA, aerobactin, wabG, allS, kfu, mrkD) was determined by polymerase chain reaction (PCR) using primers documented previously [Reference Brisse7]. The rmpA2 gene, a homologue of rmpA sharing 80% nucleotide sequence [Reference Wacharotayankun19], was detected by PCR as described previously [Reference Lee20]. The virulence scores were expressed as the total number of these genes positive in each strain. Capsular serotypes (K1, K2, K20, K54, K57) were determined by multiplex PCR [Reference Lin21]. MLST was performed for all isolates as described by Diancourt et al. [Reference Diancourt22] and sequence types were assigned by reference to the MLST database (http://www.pasteur.fr/recherche/genopole/PF8/mlst/Kpnemoniae.html). Sequences of alleles or sequence types not registered on the database were submitted to the curator and coded as novel. Pulsed-field gel electrophoresis (PFGE) of XbaI digests of chromosomal DNA was performed on isolates of predominant sequence types [Reference Böhm23] and DNA patterns were interpreted according to Tenover et al. [Reference Tenover24].
Statistical analysis
Fisher's exact test was used to statistically assess the frequency of microbiological and clinical variables and the Mann–Whitney test (non-parametric) was used to compare strain virulence scores. P value <0·05 was considered to be statistically significant. The statistical software used was SPSS, release version 15.0.1 (SPSS Inc., USA).
RESULTS AND DISCUSSION
The microbiological characteristics of the 33 isolates are summarized in Table 1 and show that 14 (42·4%) were positive for the HV phenotype by the string test and all formed mucoviscous strings of >10 mm. The mean virulence score for HV-positive strains was 5·5, and for non-HV strains it was 1·6 (P < 0·001) (Fig. 1) and the distribution of virulence genes was markedly different in HV and non-HV strains. Five of the seven genes investigated, (rmpA, rmpA2, aerobactin, wabG, allS) were significantly more frequent in HV strains (100%, 78·6%, 100%, 64·3%, 50%, respectively) than in non-HV strains (5·3%, 5·3%, 10·5%, 10·5%, 5·3%, respectively). It is of note that all HV strains harboured rmpA and aerobactin. rmpA is a positive regulator of extracellular polysaccharide synthesis and its isoform, rmpA2, has also been reported to enhance the colonial mucoidity of various serotypes of K. pneumonia [Reference Wacharotayankun19]. All isolates harbouring rmpA2 also had the rmpA gene. The aerobactin gene encodes high-affinity iron chelators and both this gene and rmpA have been associated with higher lethality in animal models [Reference Yu16, Reference Nassif25]. This coexistence of rmpA and aerobactin genes in HV strains is in agreement with previous reports that demonstrated the two genes were located simultaneously on a 180-kb plasmid [Reference Nassif26]. The kfu gene was also more frequent in the HV phenotype, but this was not statistically significant. The mrkD gene encoding the fimbrial tip adhesin MrkD was detected in all but two isolates (Table 1). In a previous report, the mrkD gene was not detected specifically in clonal complex (CC) 82 with serotype K1 (CC82K1) [Reference Brisse7]. In our data, not only the ST82 (CC82) isolate but also the novel clone, ST713, was negative for the mrkD gene.
Mean number of each virulence gene carried in hypermucoviscous (HV) and non-HV strains. HV strains carried significantly more virulence genes than non-HV strains (5·5 vs. 1·6, P < 0·001).
Microbiological variables of Klebsiella pneumoniae isolates from the 33 patients with bacteraemia
ESBL, Extended spectrum β-lactamase; ST, sequence type; n.a., not available, unidentified serotype.
All but one of the 14 HV strains were grouped into specific serotypes by the multiplex PCR. K1 was the most prevalent (n = 8), followed by K2 (n = 3), K54 (n = 1) and K57 (n = 1). By contrast, only two out of 19 non-HV strains were serogrouped by PCR (K20, K54). We cannot exclude the possibility that this might be due to the fact that the PCR lacks specific primers for types K4 and K5, which are also highly virulent in experimental infection in mice and are often associated with severe infections in humans and animals [Reference Simoons-Smit27, Reference Mizuta28].
Molecular genotyping by MLST identified 21 different sequence types, including six novel alleles and seven unique sequence types (ST655, ST662, ST710, ST711, ST712, ST713, ST714; Table 1). There were two, three, and two different sequence types in K1, K2 and K54 serotypes, respectively and eight sequence types in HV strains, of which ST23 was the most frequent (7/14). Of the 13 different sequence types in non-HV strains, ST11 was the most frequent (5/19). PFGE revealed that the seven ST23 isolates from six different hospitals were closely related and of the five ST11 isolates from four different hospitals, four were closely related, and one was possibly related (Fig. 2).
Pulsed-field gel electrophoresis patterns of (a) seven ST23 and (b) five ST11 isolates. Both show clonal relationships.
All the K1 and K2 strains showed the HV phenotype. Interestingly, however, there was a difference in the distribution of virulence genes between K1 and K2 serotypes; whereas the prevalence of wabG and allS in K1 isolates was very high (100% and 87·5%, respectively), none of the K2 isolates harboured them. In addition, the kfu gene, which is known to be associated with the development of abscesses in the liver and in the brain of mice [Reference Yu29], was found in all of K1 isolates but in only one K2 isolate (ST380). This finding is in line with a previous report, where allS and kfu genes were present in all K1 strains tested but not in K2 strains [Reference Yu29]. Considering the fact that K1 strains are emerging as a pathogen capable of causing catastrophic septic ocular or central nervous system complications from liver abscesses [Reference Fang30], further experiments are required to validate whether the presence of additional wabG, allS and kfu genes in K1 isolates contributes to invasive complications. Interestingly, the allS gene, included in a 22-kb chromosomal region and known to be present only in the K1 serotype [Reference Yu29], was found in one non-HV isolate (ST270) of an unidentified serotype.
Of particular interest, in the three cases where the source of bacteraemia was a liver abscess, one case was an ST380 isolate of the K2 serotype harbouring the kfu gene. To the best of our knowledge, this is the first report of an ST380 isolate of serotype K2 causing a pyogenic liver abscess and subsequently, bacteraemia. Although K2 strains have not received as much attention as K1 strains, severe and fatal infections due to two unrelated clones (ST86 and ST380) of the K2 serotype were reported recently in France [Reference Decré31]. In our study, about 10% of the bacteraemic isolates had the K2 serotype, and their sequence types were ST86, ST375 and ST380. Whether these clones of K2 strains will prove to be as dominant as ST23 in the K1 strains remains to be determined.
None of the HV or non-HV strains was resistant to imipenem or meropenem. HV strains showed a lower resistance rate to all the other antimicrobial agents than non-HV strains. All HV strains were susceptible to piperacillin-tazobactam and amikacin. But in non-HV strains, the resistance rates to these antimicrobials were 31% and 21%, respectively. Of the HV strains 7% were resistant to gentamicin and ciprofloxacin and 14% to trimethoprim-sulfamethoxazole compared to non-HV strains where resistance rates to these three agents were 31%, 47% and 57%, respectively.
None of the HV strains produced ESBLs but in non-HV strains, the ESBL production rate was higher (47·4%) than that (30%) of Korean isolates of K. pneumoniae from various sources in a previous study [Reference Jeong32]. This might be due to the predominance of ST11 in the non-HV strains, which is a dominant clone in ESBL-producing K. pneumoniae in Korea [Reference Ko33]. A plausible explanation for this negative association between the HV phenotype and ESBL production might be that the virulence genes were lost during the bacterial acquisition of antibiotic resistance, as it can induce phenotypic and genetic changes as a part of reduced biological fitness [Reference Kang34]. Or as many virulent isolates harbour a large (∼200 kb) plasmid [Reference Tang35], stable coexistence of another plasmid might be difficult. In addition, ST11 is a single-locus variant of ST258, which has been reported worldwide, including Korea, as the predominant clone of Klebsiella pneumoniae carbapenemase (KPC)-producing K. pneumoniae [Reference Kitchel36–Reference Qi38] and therefore, it is worrisome that ST11 predominates in bacteraemic isolates.
The clinical characteristics of the patients infected by isolates with or without HV phenotype are compared in Table 2. The frequency of fatal underlying illness was significantly lower in patients infected with the HV phenotype compared to the non-HV phenotype (14·3% vs. 68·4%, P = 0·004). Similarly, the presence of focal infections was significantly more frequent in patients with the HV phenotype (78·6% vs. 26·3%, P = 0·005). Although not statistically significant, mortality was also higher in the population infected with the HV strains (28·6% vs. 10·5%, P = 0·363). Poor immune status, fatal underlying diseases of the patients infected by non-HV strains, and higher antimicrobial resistance of these strains might have influenced the risk of mortality in this group of patients.
Clinical chracteristics of 33 bacteraemic patients by Klebsiella pneumoniae with or without the hypermucoviscosity (HV) phenotype
In summary, about 40% of bacteraemic isolates of K. pneumoniae were demonstrated to be of the HV phenotype. These isolates had a higher virulence gene complement but none harboured ESBL. In contrast to the non-HV strains, most of which were of unidentified serotypes, the majority of HV strains fell into serotypes K1 and K2. The distribution of virulence genes differed between these serotypes. The isolates were in the main genetically diverse but ST23 and ST11 were the most frequent in HV and non-HV isolates, respectively. Clinicians should be aware of the increased virulence of the HV phenotype and efforts should be made to investigate the sources of infection. Finally, considering the high frequency (21·4%) of the K2 serotype in HV strains, surveillance should be extended to this serotype and its associated sequence types to determine whether severe infections due to these strains can be spread in the community.
DECLARATION OF INTEREST
None.
ACKNOWLEDGEMENTS
This research was supported by a research fund (no. 2011 E4400400) of the Korean Centers for Disease Control and Prevention. We thank Dr Jung of Sungkyunkwan University for kindly donating positive strains of capsular K1 serotype K. pneumoniae.
