A 20-Year Study of Capsular Polysaccharide Seroepidemiology, Susceptibility Profiles, and Virulence Determinants of Klebsiella pneumoniae from Bacteremia Patients in Taiwan

ABSTRACT In this study, we selected bacteremic Klebsiella pneumoniae isolates from the Taiwan Surveillance of Antimicrobial Resistance program. A total of 521 isolates were collected over a period of 2 decades, including 121 from 1998, 197 from 2008, and 203 from 2018. Seroepidemiology showed that the top five capsular polysaccharide types were serotypes K1, K2, K20, K54, and K62, constituting 48.5% of the total isolates, and the respective ratios at each time point have remained similar over the past 2 decades. The antibacterial susceptibility tests showed that K1, K2, K20, and K54 were susceptible to most antibiotics, while K62 was relatively resistant compared to other typeable and nontypeable strains. In addition, six virulence-associated genes, clbA, entB, iroN, rmpA, iutA, and iucA, were predominant in K1 and K2 isolates of K. pneumoniae. In conclusion, serotypes K1, K2, K20, K54, and K62 of K. pneumoniae are the most prevalent serotypes and carry more virulence determinants in bacteremia patients, which may indicate their invasiveness. If further serotype-specific vaccine development is performed, these five serotypes should be included. Since the antibiotic susceptibility profiles were stable over a long duration, empirical treatment may be predicted according to serotype if rapid diagnosis from direct clinical specimens is available, such as PCR or antigen serotyping for serotype K1 and K2. IMPORTANCE This is the first nationwide study to examine the seroepidemiology of Klebsiella pneumoniae using blood culture isolates collected over a period of 20 years. The study found that the prevalence of serotypes remained consistent over the 20-year period, with high-prevalence serotypes associated with invasive types. Nontypeable isolates had fewer virulence determinants than other serotypes. With the exception of serotype K62, the other high-prevalence serotypes were highly susceptible to antibiotics. If rapid diagnosis using direct clinical specimens, such as PCR or antigen serotyping, is available, empirical treatment can be predicted based on serotype, particularly for K1 and K2. The results of this seroepidemiology study could also help the development of future capsule polysaccharide vaccines.

determine the seroepidemiology of K. pneumoniae. We hope that these findings will be beneficial for the further development of vaccines in Taiwan, the choice of effective empirical antibiotics, and virulence determinations for further diagnosis of this bacterium.
Antimicrobial susceptibility of the five most prevalent serotypes in comparison to other serotypes and nontypeable isolates. According to serotyping, all isolates were intrinsically resistant to ampicillin (data not shown). The 4 most prevalent serotypes, K1, K2, K20, and K54, were highly susceptible to different antibiotics (Table 2). For serotypes K1 and K2, over 91% susceptibility was observed for all tested antibiotics except cefazolin (CFZ), which had 83.7% and 72.8% susceptibility, respectively. Serotypes K1 and K2 were 100% susceptible to imipenem (IMP) and meropenem (MER) over 2 decades. Similar to serotypes K1 and K2, serotypes K20 and K54 were highly susceptible to the tested antibiotics, and .86% of all isolates were susceptible to all tested antibiotics except CFZ, with 77.7% and 68.9% susceptibility observed for K20 and K54, respectively (Table 2). K62 was the fifth most frequently identified serotype among all isolates. Unlike the top 4 serotypes, 42.8% of serotype K62 isolates were susceptible to the b-lactam/b-lactamase inhibitor (AUG). For cephalosporins, 33.3%, 52.3%, 47.6%, 52.3%, and 57.1% of the isolates were susceptible to CFZ, cefoxitin (FOX), cefotaxime (FTX), ceftazidime (CAZ), and cefepime, respectively, indicating that K62 was more resistant to cephalosporins than other serotypes. Over 95% of K62  serotype isolates were still susceptible to IMP and MER. Serotype K62 isolates were also less susceptible to aminoglycosides (GEN and AMK; 52.2% and 71.4%) and quinolone (CIP; 52.2%). Susceptibility tests showed that in addition to the 5 most prevalent serotypes, other serotypes and nontypeable isolates had similar susceptibility profiles. They were more resistant to CFZ (56.5% for other serotypes and 58.8% for nontypeable), in comparison to the most common serotypes of K1, K2, K20, and K54. In contrast, serotypes other than K1, K2, K20, and K54 and nontypeable isolates were more susceptible to CFZ, GEN, and AMK than K62 serotype isolates. Overall, the isolates in this study were highly susceptible to carbapenems and tigecycline, with susceptibility in over 95% (Table 2). There was no significant change in the antibiotic resistance rate from 1998 to 2018 among all isolates. Distribution of virulence-associated determinants among different serotypes of isolates. According to the serotype distribution, the prevalence of virulence-associated determinants is listed in Table 3. The virulence-associated determinants varied in prevalence among different serotypes of isolates. With reference to the serotypes, there was no extreme difference in the gene prevalence rate among isolates from  (Table 3). Since iucA and iutA are functional genes for aerobactin expression, they appeared together. Only very few isolates contained either iucA or iutA. A similar phenomenon to iroN was also observed in iucA and iutA; the prevalence was highest in serotypes K1 and K2, followed by K20, K54, and K62. A low prevalence of iucA, 17%, and iutA, 18%, was observed for K types other than the 5 most prevalent serotypes. Almost zero identification was found for iucA (1%), and iutA was undetected in the nontypeable isolates (Table 3). Similar to iroN, iucA, and iutA, the prevalence of rmpA also decreased, from 90% and 91% of serotypes K1 and K2, to 80% for K20, 65% for K54, 38% for K62, and 15% for other K types. No rmpA was found in nontypeable isolates. The prevalence of rmpA2 was similar to that of rmpA but slightly less than that for rmpA (Table 3). Also, the prevalence of rmpA2 decreased from serotype K1 (87%) and K2 (85%) to 83% for K20, 65% for K54, and 52% for K62. Only 1 nontypeable isolate (1%) was carrying rmpA2. These four genes, iroN, rmpA, iucA, and iutA, had a direct correlation with the serotype prevalence rate. clbA was distributed more frequently in serotypes K1, K2, K20, and K62 and was almost undetected in other isolates, including the highly prevalent serotype K54.

DISCUSSION
In this study, a long-duration epidemiological study of K. pneumoniae was analyzed. Associations among serotypes, antimicrobial susceptibility, and virulence-associated determinants were evaluated. These data can help to delineate the changes in bacterial species after long-term environmental and clinical selection, including parallel changes in the practice of antibiotic use and new antibiotic development. Our data have shown that the predominant serotypes of blood culture K. pneumoniae in the 2 decades from 1998 to 2018 were almost the same, indicating the possibility of serotype-associated invasiveness. Serotypes K1 (16.5%), K2 (15.5%), K20 (6.9%), K54 (5.6%), and K62 (4.0%) were the five most prevalent serotypes in our study. Although there is a slight difference in the serotype prevalence compared to a previous study conducted in 1993 to 1997 (18), the isolates were from a single hospital and from different clinical specimens. Serotypes K20 and K62 were not listed in the five most predominant serotypes in the single-hospital study. However, serotypes K1, K2, and K54 were compatible in both studies; these serotypes were listed in the five most common serotypes, and serotypes K1 and K2 were the most prevalent serotypes among blood culture isolates (18). In one study with isolates from North America and Europe, serotype prevalence differed from those in our study (24). The predominant serotypes of 703 isolates from blood were K2 (8.9%), K21 (7.8%), K55 (4.8%), K53 (2.8%), K25 (2.8%), and K68 (2.5%). Another study from Australia showed that the predominant serotypes of 293 isolates from different sources were K54 (17.1%), K28 (4.1%), K17 (3.1%), and K1 (2.0%) (25). In China, the predominant serotypes of 348 isolates from different sources were K14 (16.4%), K64 (16.4%), K1 (14.6%), K2 (8.0%), and K57 (5.5%) (26). Those studies and our present data have shown that serotype prevalence varies from region to region. However, unlike our study, the isolates were collected from different sites. Whether the collection site differences contributed to the difference in serotype prevalence needs to be further studied. The reason for the serotype difference could perhaps be confirmed by further study from our Taiwan Surveillance of Antimicrobial Resistance (TSAR) program by using different sites of isolation other than blood culture isolates.
For antimicrobial susceptibility, the 4 most prevalent serotypes of isolates and nontypeable isolates showed a relatively stable susceptibility to all tested antibiotics. K62, which belongs to multilocus sequence type ST11, was the most resistant serotype in this study. Other K types or nontypeable isolates showed relatively more resistance to most antibiotics except to carbapenems, IMP and MER.
Carbapenems are still the most effective antibiotics against all isolates. The 4 most prevalent serotypes were 100% sensitive. The rest of the other isolates had .95% susceptibility to carbapenems. Overall, serotypes K1, K2, K20, and K54 and nontypeable isolates were relatively susceptible to most antibiotics, except for CFZ and FOX. Thus, using third-generation cephalosporins in combination with an aminoglycoside could be a safe empirical treatment for K. pneumoniae when the susceptibility profile is not available in a timely manner. Recently, rapid serotype K1 and K2 diagnostic kits have been developed and used for clinical applications in direct clinical specimens (27,28). They have shown a high predictive value for K. pneumoniae infection. Combining rapid diagnosis with local epidemiology data may give physicians a crude idea of the use of empirical treatment until the susceptibility data from the clinical laboratory confirm whether changing the treatment is necessary. Currently, serotypes K1 and K2 are the two most prevalent and invasive serotypes in community-acquired infections, and susceptibility to these two serotypes has been stable in the past 20 years. Rapid diagnosis may not only help in disease identification but also provide hints for empirical treatment.
For the correlation between serotypes and virulence determinants, we observed that 100% of isolates carried entB. This gene seems to be intrinsic in K. pneumoniae and may not contribute to the virulence differences in K. pneumoniae. In future studies, this gene could be excluded, since it cannot be used as a marker to differentiate virulence or disease severity. One intriguing observation is that the nontypeable isolates contained almost no tested virulence determinants except intrinsic entB. In addition, serotypes other than the 5 most prevalent serotypes also had a relatively low prevalence of the tested virulence determinants, which could indicate that carrying more virulence factors provides more opportunities to cause diseases due to their invasiveness. Unfortunately, there are no clinical details of the patients' immune status for analysis. An invasive syndrome of K. pneumoniae has been previously suggested, where diabetic patients infected with serotype K1 or K2 are at high risk for complications (29). In addition to antibiotic treatment, information on virulence markers and patients' immune status can assist physicians in anticipating the potential complications of endophthalmitis, meningitis, or necrotizing fasciitis. If clinical data are available, the relationship between virulence markers and the occurrence of complications could be analyzed (30). Previously, a 24-variant K. pneumoniae CPS vaccine was developed in Switzerland, but it did not include the predominant capsules of K1, K20, and K54 found in this study. In the future, development of K. pneumoniae vaccines, as with Streptococcus pneumoniae, the serotype distributions from various regions should be obtained to increase the vaccine's suitability for different regions. Recently, a conjugated K. pneumoniae vaccine was also developed for serotypes K1 and K2 (31). This type of vaccine has been shown to provide effective protection and induce a sufficient immune response even in children with nonmature immunity. If there is further development of a K. pneumoniae-conjugated vaccine in Taiwan, including K20, K54, and K62 and the globally spreading carbapenem-resistant serotypes may be suitable for a high possibility of coverage.
In conclusion, serotypes K1, K2, K20, K54, and K62 were the top 5 prevalent serotypes in blood culture K. pneumoniae in Taiwan, and the prevalence was stable for 20 years. These serotypes should be included if a CPS vaccine is to be developed. The virulence-associated gene entB was present in all K. pneumoniae isolates and is possibly intrinsic in K. pneumoniae. The detection of entB may be unnecessary in the future. Virulence genes, iroN, rmpA, iucA, and iutA are almost undetected from nontypeable isolates. There was an association between antimicrobial susceptibility and serotypes. Serotypes K1, K2, K20, and K54 were highly susceptible to most of the antibiotics, and serotype K62 was more resistant than other K. pneumoniae serotypes when a susceptibility test result was not available. The choice of effective empirical antibiotic treatment could be considered if rapid diagnosis of serotype tests is available.

MATERIALS AND METHODS
Collection of isolates. Taiwan Surveillance of Antimicrobial Resistance (TSAR) is a surveillance program that is run by the National Health Research Institutes. Every other year, bacterial isolates are collected from hospitals that participate in the TSAR program (see the Appendix). The program was initiated in 1998, and isolates have been continually collected up to the present. In this study, we selected 2 months of nonrepeated isolates of blood cultures K. pneumoniae in 1998, 2008, and 2018 from TSAR. There were 121 isolates, 197 isolates, and 203 isolates collected from 1998, 2008, and 2018, respectively (Fig. 1).
Serotyping by rapid antigen tests, PCR typing, and sequencing. The isolates were serotyped by rapid cassette to classify the K1, K2, and non-K1/K2 groups (32). We used serotype-specific primer sets to perform multiplex PCR for serotyping. If nontypeable isolates were observed from the first two serotyping tests, then wzi, wza, and wzc gene sequencing was performed to confirm the serotype. PCR and sequencing of these genes were performed according to methods reported previously (33)(34)(35)(36) and summarized in Table S1 in the supplemental material.
Virulence-associated gene detection and antibiotic susceptibility testing. The genes iroN, clbA, entB, rmpA, rmpA2, iutA, and iucA were detected to study the virulence of K. pneumoniae. The primer set is listed in Table S2 (7). The Taq polymerase we used for entB was amaR (One PCR HotStar, Taiwan), and for the other virulence factors, iroN, clbA, rmpA, rmpA2, iutA, and iucA, we used a Q-Amp 2Â ScreeningFire Taq master mix (Taiwan).
Antimicrobial susceptibility testing. We tested the susceptibility of isolates to amikacin, gentamicin, amoxicillin-clavulanate, piperacillin-tazobactam, ampicillin, cefazolin, ceftazidime, cefepime, cefoxitin, cefuroxime, cefotaxime, imipenem, meropenem, ciprofloxacin, trimethoprim-sulfamethoxazole, and aztreonam by means of the broth microdilution method. Susceptibility of isolates to tigecycline and colistin was determined by Etest. The MICs were interpreted following the standards of the Clinical and Laboratory Standards Institute (37), except for tigecycline and colistin. The Food and Drug Administration breakpoint was used for tigecycline, and for colistin, we used the European Committee on Antimicrobial Susceptibility Testing breakpoint (38).

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, DOCX file, 0.1 MB.