Association between rmpa and maga Genes and Clinical Syndromes Caused by Klebsiella pneumoniae in Taiwan

Transcription

1 MAJOR ARTICLE Association between rmpa and maga Genes and Clinical Syndromes Caused by Klebsiella pneumoniae in Taiwan Wen-Liang Yu, 1,6 Wen-Chien Ko, 4 Kuo-Chen Cheng, 1 Hsin-Chun Lee, 4 Der-Shin Ke, 2 Ching-Chien Lee, 3 Chang-Phone Fung, 5 and Yin-Ching Chuang 2,3 Departments of 1 Intensive Care Medicine, 2 Medicine, and 3 Medical Research, Chi-Mei Medical Center, and 4 Department of Medicine, National Cheng-Kung University Medical College, Tainan, and 5 Section of Infectious Diseases, Department of Medicine, Taipei Veterans General Hospital and National Yang-Ming University, and 6 Department of Medicine, Taipei Medical University, Taipei, Taiwan (See the editorial commentary by Kawai on pages ) Background. The association of the maga gene with the hypermucoviscosity phenotype relevant to the pathogenesis of Klebsiella pneumoniae liver abscess has been reported in Taiwan. Similarly, the rmpa gene, known as a positive regulator of extracapsular polysaccharide synthesis that confers a mucoid phenotype, may be another candidate gene causing hypermucoviscosity. However, the association of rmpa with K. pneumoniae clinical syndromes is unreported. We aimed to investigate the clinical correlation between rmpa and primary Klebsiella abscess, focusing on sites other than the liver. Methods. From July 2003 through December 2004, a total of 151 K. pneumoniae isolates recovered from 151 patients with bacteremia were collected from 2 large medical centers in southern Taiwan. Clinical data were collected from medical records. The genes rmpa and maga were amplified by polymerase chain reaction using specific primers. Results. The prevalences of hypermucoviscosity, rmpa, and maga were 38%, 48%, and 17%, respectively. As determined by statistical multivariate analysis, strains carrying rmpa were significantly associated with the hypermucoviscosity phenotype, and there was a significant correlation with purulent tissue infections, such as liver abscess and lung, neck, psoas muscle, or other focal abscess. Conclusion. Our data support a statistical correlation between the rmpa gene and virulence in terms of abscess formation for these hypermucoviscous K. pneumoniae strains. Hypermucoviscosity associated with rmpa, together with a thorough physical examination, may be helpful as a guide to carry out appropriate diagnostic tests on patients with an initially unknown source of K. pneumoniae bacteremia, particularly when looking for the occurrence of an underlying abscess. Klebsiella pneumoniae is a common pathogen of community-acquired and nosocomial infections [1 6]. It causes a wide spectrum of infections, including septicemia, pneumonia, urinary tract infection, meningitis, and purulent abscess at various sites, especially liver abscess. In particular, a distinctive clinical syndrome, which is characterized by community-acquired K. pneu- Received 8 September 2005; accepted 19 December 2005; electronically published 11 April Presented in part: 5th International Symposium on Antimicrobial Agents and Resistance, Seoul, Korea, April 2005 (abstract CM 05). Reprints or correspondence: Dr. Yin-Ching Chuang, Dept. of Medical Research, Chi-Mei Medical Center, 901 Chung Hwa Rd., 710 Yungkang City, Tainan County, Taiwan (chuangkenneth@hotmail.com). Clinical Infectious Diseases 2006; 42: by the Infectious Diseases Society of America. All rights reserved /2006/ $15.00 moniae bacteremia with primary liver abscess, metastatic meningitis, and endophthalmitis, has been recognized in Taiwan [2]. Furthermore, a novel chromosomal gene, maga (mucoviscosity-associated gene A), causes hypermucoviscosity, as defined by positive results of string test, and in one study was deemed responsible for 52 of 53 primary K. pneumoniae liver abscesses [1]. In contrast, maga is rarely found in other miscellaneous K. pneumoniae related infections without liver abscess. Despite this prevalence, however, our clinical experience has included many cases involving fulminant abscess formation at nonhepatic sites, such as the lung, pleural space, kidney, and prostate, that has resulted in severe morbidity (figure 1) and that were primarily caused by K. pneumoniae with the hypermucoviscosity phenotype. In addition, the mortality rmpa and Purulent Klebsiella Infection CID 2006:42 (15 May) 1351

2 Figure 1. Chest radiograph and CT of the abdomen of a 55-year-old man with alcoholism indicating pleural empyema (A and B), left-sided renal abscess (C; white arrow), and prostate abscess (D; white arrow), without evidence of liver abscess. Klebsiella pneumoniae characterized by hypermucoviscosity, negative for maga, and positive for rmpa was isolated from the cultures of specimens obtained from blood and right-sided pleural empyema. rate for Klebsiella bacteremic pneumonia ( 60%) [5] is far higher than that for Klebsiella bacteremic liver abscess (11.3%) [6]. However, these nonliver pyogenic organisms seem to be maga negative, suggesting that genes other than maga may be involved in the hypermucoviscosity phenotype and cause invasive virulence. Therefore, we believe that the association between maga and these devastating K. pneumoniae infections remains poorly characterized. While searching other hypermucoviscosity-associated gene(s) in the literature, we found that rmpa (regulator of the mucoid phenotype), a gene known as an extracapsular polysaccharide synthesis regulator, can positively control the mucoid phenotype of K. pneumoniae [7]. This mucoid phenotype is distinct from capsule production and results from overproduction of extracellular polysaccharide, which is encoded by the chromosome but is positively controlled by rmpa located on a plasmid [7]. Although not directly shown by a positive string test result, we believed that the glistening mucoidy and viscid consistency of the colonies conferred by rmpa [7] was similar to the so-called hypermucoviscosity phenotype described by Fang et al. [1]. Therefore we suggest rmpa as a potential hypermucoviscosity-associated gene. Although plasmid-mediated rmpa was identified more than a decade ago, the relationship between rmpa and K. pneumoniae related clinical syndromes has not been clarified. Detailed experiments on gene knockout and restoration of rmpa resulting in the loss and return of the mucoviscous phenotype have been well documented elsewhere [7]. Therefore, in this study, we aimed to assess the prevalence of maga versus rmpa in a collection of K. pneumoniae isolates recovered from patients with bacteremia and to compare their association with the hypermucoviscosity phenotype and K. pneumoniae clinical syndromes, particularly focusing on the abscesses at sites other than the liver, by application of molecular methods and statistical analysis. MATERIALS AND METHODS Primary organisms studied. A total of 151 nonrepetitive K. pneumoniae isolates from cases of bacteremia were consecutively collected from 2 medical centers (capacity, beds) in southern Taiwan from July 2003 through December Each isolate was recovered from a different patient. These isolates were identified by routine microbiological methods at each laboratory, and the species was confirmed using the API 20E system (biomérieux) in hospital A and the BD Phoenix System (Becton Dickinson) in hospital B. All isolates were subcultured and frozen at 70 C until used in the study. Of these, 46 were 1352 CID 2006:42 (15 May) Yu et al.

3 recovered from patients with nosocomial infections, and 105 were recovered from patients with community-acquired infections. Primary bacteremia, defined as no original infectious focus identifiable ( n p 47) and catheter-related infections ( n p 4), constituted 51 (34%) of 151 cases in the sample. The remaining 100 infections with secondary bacteremia included a nonabscess group ( n p 61), including pneumonia without abscess ( n p 29), urinary tract infection ( n p 13), biliary tract infection ( n p 11), primary peritonitis ( n p 4), and cellulitis ( n p 4); and an abscess group ( n p 39), including liver abscess ( n p 16), secondary peritonitis ( n p 5), deep neck abscess ( n p 3), mycotic aneurysm ( n p 3), meningitis ( n p 3), lung abscess/empyema ( n p 4), psoas muscle abscess ( n p 2), gall- bladder empyema ( n p 2), and infectious spondylitis ( n p 1). No cases of secondary (or metastatic) K. pneumoniae meningitis originating from a liver abscess were identified. Supplementary organisms for comparison in the study. A total of 24 nonrepetitive K. pneumoniae isolates had been previously collected from patients with bacteremia and liver abscess at hospital B from January 2001 to December All isolates were stored at 70 C before retrieval for testing. Hypotheses. We put forward 2 clinical hypotheses: organisms causing primary bacteremia are less virulent than those causing secondary bacteremia in terms of expression of the hypermucoviscosity phenotype; and organisms causing purulent abscess infection are more virulent than those causing nonabscess infection in terms of expression of the hypermucoviscosity phenotype. Definition of virulence and hypermucoviscosity of the bacteria and clinical syndromes. Strains with the hypermucoviscosity phenotype were defined as being virulent. To determine hypermucoviscosity phenotype, a standard bacteriologic loop was used to stretch a mucoviscous string from the colony. Strains were defined as hypermucoviscosity positive when the viscous strings were 15 mm (a positive string test result), as described elsewhere [1]. We have previously noted that the degree or strength of viscosity seems to be slightly diminished for colonies at their first subculture from frozen isolates, but their phenotype is well recovered at the second subculture and is in accordance with that from primary culture plates. Therefore, all isolates required 2 subcultures to get well-grown colonies before hypermucoviscosity testing, which was performed by an experienced laboratory scientist (C. C. Lee) at hospital B to avoid interobserver discrepancy. Primary bacteremia, nosocomial infection, and each infectious clinical syndrome were defined as described elsewhere [2, 4]. Primary Klebsiella abscess was defined by the presence of an abscess in an organ or tissue that was primarily caused by a single pathogen of K. pneumoniae without other copathogen. Detection of hypermucoviscosity-associated loci and other virulence genes. With genomic DNA used as the template, PCR was performed to amplify the chromosome-mediated maga, wabg (involved in the biosynthesis of the outer core lipopolysaccharide), and uge (encoding uridine diphosphate galacturonate 4-epimerase, responsible for biosynthesis of the capsule and smooth lipopolysaccharide) genes, as well as other virulence-associated genes, such as kfu (iron-uptake system) and fimh (fimbrial gene encoding type 1 fimbrial adhesin). The plasmid-mediated rmpa was amplified by PCR using plasmid DNA as the template. Specific primers used to detect the alleles of the target gene sequences are shown in table 1. The PCR product from the first strain tested was DNA-sequenced and showed a high level of identity (198% homology) with the published target sequence [1, 7, 9 13]; this was selected as the positive control for the subsequent PCR experiments. One clinical isolate of Escherichia coli was selected as the negative control for both the rmpa and maga PCR systems. Statistical analysis. The x 2 and Fisher s exact tests were used whenever appropriate for statistical evaluation. A P value!.05 was considered to be statistically significant. Univariate and multivariate analyses were used to determine the independent risk factors associated with the hypermucoviscosity phenotype and abscess formation and were performed using stepwise logistic regression models. The statistical software used was SPSS for MS Windows, release version (SPSS). RESULTS General prevalence of the hypermucoviscosity phenotype and hypermucoviscosity-associated genes. The prevalences for the hypermucoviscosity phenotype and rmpa and maga genes were Table 1. Specific primers used for amplification of the target alleles of various virulence genes of Klebsiella pneumoniae. Target gene Primer Reference(s) maga [1] Forward 5 -GGTGCTCTTTACATCATTGC-3 Reverse 5 -GCAATGGCCATTTGCGTTAG-3 rmpa [7, 8] Forward 5 -ACTGGGCTACCTCTGCTTCA-3 Reverse 5 -CTTGCATGAGCCATCTTTCA-3 wabg [9] Forward 5 -ACCATCGGCCATTTGATAGA-3 Reverse 5 CGGACTGGCAGATCCATATC-3 uge [10] Forward 5 -TCTTCACGCCTTCCTTCACT-3 Reverse 5 -GATCATCCGGTCTCCCTGTA -3 kfu [11] kfub-f GAAGTGACGCTGTTTCTGGC-3 kfuc-r TTTCGTGTGGCCAGTGACTC-3 fimh [12, 13] Forward 5 -TGCTGCTGGGCTGGTCGATG-3 Reverse 5 -GGGAGGGTGACGGTGACATC-3 rmpa and Purulent Klebsiella Infection CID 2006:42 (15 May) 1353

4 Table 2. Association between the presence of rmpa and/or maga genes and the hypermucoviscosity phenotype. Variable Hypermucoviscosity (n p 151) rmpa positive/ maga positive (n p 23) rmpa positive/ maga negative (n p 49) rmpa negative/ maga positive (n p 3) rmpa negative/ maga negative (n p 76) Positive (n p 58) Negative (n p 93) OR (95% CI) 33 ( ) 29 (10 82) 1 NOTE. Data are no. of isolates, unless otherwise indicated. 38% (58 of 151 isolates), 48% (72 of 151 isolates), and 17% (26 of 151 isolates), respectively. Of the 58 hypermucoviscositypositive isolates, 52 (90%) possessed rmpa, but only 17 (29%) contained maga. Hypermucoviscosity was positively predicted by 52 (72%) of 72 rmpa-positive and 17 (65%) of 26 magapositive isolates. Coexistence of rmpa and maga significantly increased the frequency of expression of the hypermucoviscosity phenotype in K. pneumoniae isolates (OR, 33; 95% CI, ); however, 6 isolates positive for both rmpa and maga did not express the hypermucoviscosity phenotype (table 2). Additionally, 6 isolates with the characteristic hypermucoviscosity phenotype were negative for rmpa and maga. Overall, the presence of rmpa, disregarding the presence of maga, was significantly associated with expression of the hypermucoviscosity phenotype compared with that in strains without both genes. Although univariate analysis revealed a statistically significant association of rmpa and maga with the hypermucoviscosity phenotype, multivariate analysis determined rmpa to be a truly independent factor (table 3). Other capsule- or lipopolysaccharide-associated genes (uge and wabg) did not show a statistically significant association with either hypermucoviscosity-positive or hypermucoviscosity-negative isolates. Prevalence of the hypermucoviscosity phenotype and hypermucoviscosity-associated genes in organisms from bacteremic liver abscess. Detailed distributions of rmpa, maga, and the hypermucoviscosity phenotype in K. pneumoniae isolated from various infection sites are presented in table 4. Among 16 isolates causing liver abscess, 14 (88%) carried rmpa, and only 7 (44%) possessed maga. In addition, among 8 kfu-positive isolates causing liver abscess, 6 strains were maga positive. For further comparison, we identified rmpa (24 [100%] of 24 isolates) and maga (13 [54%] of 24 isolates) among an additional 24 K. pneumoniae bacteremic isolates (20 of 24 with hypermucoviscosity phenotype) that had been previously recovered from patients with liver abscess between January 2001 and December 2002 at hospital B. Overall, 20 maga-positive isolates (50%) were identified among 40 K. pneumoniae isolates causing liver abscess. All 3 isolates causing primary K. pneumoniae meningitis without liver abscess were maga negative, but 2 of the isolates were rmpa positive. Prevalence of the hypermucoviscosity phenotype and hypermucoviscosity-associated genes in organisms from patients with primary versus secondary bacteremia. The K. pneumoniae isolates that caused overall primary bacteremia exhibited less hypermucoviscosity than did those that caused secondary bacteremia (25% vs. 45%; P p.02) (table 3). For example, central venous catheter related K. pneumoniae infections were usually caused by strains not expressing the hyper- Table 3. Factors associated with expression of hypermucoviscosity phenotype in Klebsiella pneumoniae. Variable Gene No. of isolates (n p 151) Hypermucoviscosity Univariate analysis Multivariate analysis Positive (n p 58) Negative (n p 93) OR (95% CI) P OR (95% CI) P rmpa ( )! ( )!.001 maga ( ).003 wabg ( ) uge ( ).555 fimh ( ).696 kfu ( ).822 Secondary bacteremia ( ).021 Abscess ( )! ( )!.001 Hospital-acquired infection ( )! CID 2006:42 (15 May) Yu et al.

5 Table 4. Distribution of rmpa, maga, and the hypermucoviscosity phenotype in Klebsiella pneumoniae isolates from various infection sites. Infection Secondary bacteremia No. of cases (n p 151) rmpa-positive isolates (n p 72) maga-positive isolates (n p 26) HMV-positive isolates (n p 58) All Urinary tract infection Intra-abdominal infection Spontaneous bacterial peritonitis Secondary peritonitis Cholangitis Gall bladder empyema Primary liver abscess Cellulitis Lower respiratory infection Nonabscess pneumonia Lung abscess/empyema Other primary focal abscess a Primary bacteremia All Unknown focus Catheter-related infection Community-acquired infection Hospital-acquired infection NOTE. HMV, hypermucoviscosity. a Mycotic aneurysm ( n p 3), deep neck abscess ( n p 3), meningitis ( n p 3), psoas muscle abscess ( n p 2), and infectious spondylitis ( n p 1). mucoviscosity phenotype (table 4). Similarly, there was a trend toward a lower incidence of rmpa in primary bacteremia versus secondary bacteremia isolates, although this was not statistically significant (37% vs. 53%; P p.067; table 4). For further clarification, when isolates from urinary tract infection ( n p 13) were excluded, the incidence of rmpa in the organisms causing other secondary bacteremia (49 of 87 isolates) was significantly higher than for strains from primary bacteremia (56% vs. 37%; P p.047). No statistical difference in the presence of maga was found between primary and secondary isolates (12% vs. 20%; P p.205; table 4). Diverse prevalence of the hypermucoviscosity phenotype and hypermucoviscosity-associated genes in organisms from the secondary bacteremia subgroups. It seems that isolates causing nonabscess infections, such as cellulitis, cholangitis, or urinary tract infections, were less prone to express the hypermucoviscosity phenotype (table 4). Statistical multivariate analysis determined that rmpa and abscess formation were truly independent factors with respect to the expression of the hypermucoviscosity phenotype (table 3). To precisely clarify the correlation of hypermucoviscosity, hypermucoviscosity-associated genes, and other known virulence genes with purulence, we conducted additional stratified analysis focusing on the secondary bacteremic cases. Primary bacteremia was excluded because of the possibility of an undiagnosed focal abscess. In addition, we excluded 1 possible confounding factor: secondary peritonitis ( n p 5), which oc- curs after intestinal rupture and infection with mixed organisms; this might not truly reflect the suppurative nature of the K. pneumoniae strain isolated in that case. Thus, we subclassified the remaining isolates ( n p 95) from cases of secondary bacteremia into organisms causing primary abscess diseases ( n p 34) and nonabscess infections ( n p 61). The incidences of kfu, rmpa, maga, and the hypermucoviscosity phenotype were significantly higher in the abscess group than in the nonabscess group, as determined by univariate analysis (P p.033, P!.001, P p.002, and P!.001, respectively). Multivariate analysis determined rmpa and the hypermucoviscosity phenotype to be independent risk factors for abscess formation (table 5). The results of the subanalysis of other infection types are summarized in figure 2. The differences in rmpa, maga, and the hypermucoviscosity phenotype between the subgroups of liver abscess versus abscesses at other sites were not statistically significant. Prevalence of the hypermucoviscosity phenotype and hypermucoviscosity-associated genes in organisms from patients rmpa and Purulent Klebsiella Infection CID 2006:42 (15 May) 1355

6 Table 5. Factors associated with risk for abscess among 95 Klebsiella pneumoniae isolates from secondary bacteremia. Variable Liver (n p 16) Abscess Univariate analysis Multivariate analysis Any abscess (n p 34) No abscess (n p 61) OR (95% CI) P OR (95% CI) P Hypermucoviscosity ( )! ( ).002 Gene rmpa ( )! ( ).043 maga ( ).002 kfu ( ).033 fimh ( ) wabg ( ) uge ( ).676 Hospital-acquired infection ( ) ( ).011 NOTE. Statistical comparisons are between the any abscess group and the nonabscess group. with community-acquired infection versus nosocomial bacteremia. The prevalences for rmpa, maga, and the hypermucoviscosity phenotype were significantly higher in isolates recovered from patients with community-acquired infections than in those from patients with nosocomial infections ( P p.002, P p.021, and P!.001, respectively; table 4). In general, nosocomial strains were associated with a significantly higher rate of primary bacteremia than were community-acquired strains (47% vs. 27%; P p.015). This was associated with the nosocomial strains expressing a statistically significant lower frequency of hypermucoviscosity (table 3). Additional statistical subanalysis of secondary bacteremia demonstrated that nosocomial strains were significantly less prone to abscess formation (table 5). DISCUSSION In this pilot study of K. pneumoniae bacteremic isolates, we found that rmpa, in addition to maga, may play a major role in expression of the hypermucoviscosity phenotype and purulent infection. Furthermore, multivariate analysis determined that rmpa was an independent factor associated with the expression of the hypermucoviscosity phenotype, whereby it significantly predisposes to abscess formation, supporting our initial hypothesis. However, some discordance exists between the expression of the 2 genes. The 6 strains that were positive for rmpa and maga but negative for the hypermucoviscosity phenotype may be explained either by a lack of another positive regulator for rmpa, such as the rmpb gene, which is responsible for full expression of rmpa, or by the presence of a negative regulator of Lon protease at the posttranscriptional level [7]. Also, 6 isolates with the hypermucoviscosity phenotype did not seem to contain either rmpa or maga, implying that other regulator genes may be able to switch on the expression of the hypermucoviscosity phenotype. Overall, our data suggest that a complex regulatory pathway controls the expression of the hypermucoviscosity phenotype and that this needs to be clarified by further experimental studies. The results that other capsule or lipopolysaccharide-associated genes (such as uge and wabg) are not statistically correlated with the hypermucoviscosity phenotype, in accordance with the findings of Nassif et al. [7] that mucoidy itself is distinct from capsule production. A variant of the mucoid K2 capsular strain, after losing the rmpa-encoded plasmid, no longer expressed the mucoid phenotype, but it was still able to encapsulate [7]. This may explain why there was no statistically significant correlation of the uge and wabg genes with either the hypermucoviscosity-positive or hypermucoviscosity-negative isolates. Thus, it is unnecessary to conduct further investigation of other capsule genes, such as the lipopolysaccharide (wb) cluster and capsular polysaccharide (cps) cluster, which were Figure 2. Statistical analysis between various groups of 95 Klebsiella pneumoniae isolates recovered from patients with secondary bacteremia. P values from univariate analysis are presented. HV, hypermucoviscosity. *Significant ( P!.05), as determined by forward stepwise procedure CID 2006:42 (15 May) Yu et al.

7 also universally present in all tested strains in the report of Fang et al. [1]. Unlike the report of Fang et al. [1], in which maga was responsible for hypermucoviscosity of 98% of K. pneumoniae liver abscesses, we found that rmpa (88%) in addition to maga (44%) may have been responsible for the hypermucoviscosity phenotype (81%) of our sample. The discrepancy in the frequency of maga in the 2 studies was further confirmed in our supplementary isolates that caused liver abscess (54% of magapositive isolates). Because maga is specifically restricted to the gene cluster of K. pneumoniae capsule serotype K1 but is not found in isolates of non-k1 serotypes [14], the prevalence of maga among our isolates causing liver abscess may reflect the prevalence of the K1 serotype, which is in accordance with a previous report of the K1 serotype (63.4%) of K. pneumoniae causing liver abscess at 2 large medical centers in Taiwan [15]. The relatively low frequency of maga among our isolates may be responsible for the low frequency of kfu, an ironacquisition system on the chromosome of K. pneumoniae that has been frequently associated with maga-positive strains causing liver abscess [11]. Although significantly correlated with abscess formation by univariate analysis, kfu was not found to be significant by multivariate analysis. Nevertheless, the pathogenic role of kfu should not be neglected, because coexistence of kfu and maga may potentially enhance the propensity for liver abscess. Moreover, in contrast to a previous report of the strong association between maga and metastatic K. pneumoniae meningitis secondary to liver abscess [1], our isolates causing primary K. pneumoniae meningitis without liver abscess were all maga-negative (but 2 of 3 isolates were rmpa positive). Together with the finding of Ma et al. [11] that 2 of 4 K. pneumoniae isolates causing primary meningitis were maga negative, this implies that different virulent genes were probably responsible for the 2 distinct entities, metastatic and primary K. pneumoniae meningitis (maga and rmpa, respectively). However, because of the limited number of cases, further investigation is required. Expression of the hypermucoviscosity phenotype and the presence of the rmpa and maga genes were generally predominant in cases of secondary bacteremia and purulent disease. The proportion of nonpurulent infections, such as cellulitis or urinary tract infections, in secondary bacteremia may potentially confound the statistical variation of the hypermucoviscosity phenotype or rmpa overall between strains from secondary and primary bacteremia. In addition, the occurrence of unrecognized focal abscesses (such as rapid death without adequate diagnosis) in patients with primary bacteremia might also confound the statistical significance. Statistical multivariate analysis of secondary bacteremia determined that the hypermucoviscosity phenotype and rmpa were independent factors correlated with primary abscess formation, again in accordance with our initial hypothesis. Nosocomial bacteremic strains of K. pneumoniae were statistically significantly less virulent in terms of hypermucoviscosity expression and abscess formation than were communityacquired strains, possibly accounting for the lack of association between liver abscess and nosocomial K. pneumoniae bacteremia in a previous study [4]. This study also reported a higher proportion of infections from an unknown source among patients with nosocomial cases, compared with those with community-acquired cases (58% vs. 16%; P!.001). Fang et al. [1] reported a higher prevalence of the hypermucoviscosity phenotype in invasive strains of liver abscess than in noninvasive strains without liver abscess (98% vs. 17%). Although rmpa may be associated with the hypermucoviscosity phenotype, only maga displayed a significantly higher prevalence in invasive strains (98% vs. 29%), whereas rmpa was universally present in all tested strains [1]. This prevalence for rmpa in hypermucoviscosity-negative strains (100%) was higher than in our study (22%); the reason for this discrepancy is difficult to explain but might depend on methodology, because Fang and colleagues used dot-blot hybridization and our study used PCR for gene detection. PCR-positive rmpa is strongly correlated with hypermucoviscosity, and this result sigificantly reflects the function of the PCR-positive rmpa. The clinical syndrome whereby K. pneumoniae strains have an ability to invade tissue may not be limited to liver abscess only. Expanded to other purulent infections, including abscess at nonliver sites, rmpa-associated hypermucoviscosity may also play an important role. Our study is the first to reveal a statistically significant clinical correlation between rmpa (with or without maga), hypermucoviscosity phenotype, and abscess formation. In a clinical setting, physicians dealing with K. pneumoniae bacteremia from an unknown focus should make an effort to identify the hypermucoviscosity status of the strain. If isolates are positive for hypermucoviscosity or rmpa, a detailed search for the underlying infectious origin, particularly abscess formation, should then be conducted, involving a new thorough physical examination and appropriate imaging tests. Acknowledgments We thank Chin-Li Lu, Department of Medical Research, Chi-Mei Medical Center (Tainan, Taiwan), for assistance with statistical analysis. Financial support. Chi-Mei Medical Center. Potential conflicts of interest. All authors: no conflicts. References 1. Fang CT, Chuang YP, Shun CT, Chang SC, Wang JT. A novel virulence gene in Klebsiella pneumoniae causing primary liver abscess and septic metastatic complications. J Exp Med 2004; 199: Ko WC, Paterson DL, Sagnimeni AJ, et al. Community-acquired Klebsiella pneumoniae bacteremia: global differences in clinical patterns. Emerg Infect Dis 2002; 8: rmpa and Purulent Klebsiella Infection CID 2006:42 (15 May) 1357

8 3. Liu YC, Cheng DL, Lin CL. Klebsiella pneumoniae liver abscess associated with septic endophthalmitis. Arch Intern Med 1986; 146: Tsay RW, Siu LK, Fung CP, Chang FY. Characteristics of bacteremia between community-acquired and nosocomial Klebsiella pneumoniae infection: risk factors for mortality and the impact of capsular serotypes as a herald for community-acquired infection. Arch Intern Med 2002; 162: Chen CW, Jong GM, Shiau JJ, et al. Adult bacteremic pneumonia: bacteriology and prognostic factors. J Formos Med Assoc 1992; 91: Wang JH, Liu YC, Lee SS, et al. Primary liver abscess due to Klebsiella pneumoniae in Taiwan. Clin Infect Dis 1998; 26: Nassif X, Honore N, Cole ST, Sansonetti PJ. Positive control of colonic acid synthesis in Escherichia coli by rmpa and rmpb, two virulenceplasmid genes of Klebsiella pneumoniae. Mol Microbiol 1989;3: Arakawa Y, Ohta M, Wacharotayankun R, et al. Biosynthesis of Klebsiella K2 capsular polysaccharide in Escherichia coli HB101 requires the functions of rmpa and the chromosomal cps gene cluster of the virulent strain Klebsiella pneumoniae Chedid (O1:K2). Infect Immun 1991; 59: Izquierdo L, Coderch N, Piqué N, et al. The Klebsiella pneumoniae wabg gene: role in biosynthesis of the core lipopolysaccharide and virulence. J Bacteriol 2003; 185: Regué M, Hita B, Piqué N, et al. A gene, uge, is essential for Klebsiella pneumoniae virulence. Infect Immun 2004; 72: Ma LC, Fang CT, Lee CZ, Shun CT, Wang JT. Genomic heterogeneity in Klebsiella pneumoniae strains is associated with primary pyogenic liver abscess and metastatic infection. J Infect Dis 2005; 192: Gerlach GF, Clegg S, Allen BL. Identification and characterization of the genes encoding the type 3 and type 1 fimbrial adhesins of Klebsiella pneumoniae. J Bacteriol 1989; 171: Schembri MA, Blom J, Krogfelt KA, Klemm P. Capsule and fimbria interaction in Klebsiella pneumoniae. Infect Immun 2005; 73: Struve C, Bojer M, Nielsen FM, Hansen DS, Krogfelt KA. Investigation of the putative virulence gene maga in a worldwide collection of 495 Klebsiella isolates: maga is restricted to the gene cluster of Klebsiella pneumoniae capsule serotype K1. J Med Microbiol 2005; 54: Fung CP, Chang FY, Lee SC, et al. A global emerging disease of Klebsiella pneumoniae liver abscess: is serotype K1 an important factor for complicated endophthalmitis? Gut 2002; 50: CID 2006:42 (15 May) Yu et al.