Capsular Types of Klebsiella pneumoniae Revisited by wzc Sequencing

Capsule is an important virulence factor in bacteria. A total of 78 capsular types have been identified in Klebsiella pneumoniae. However, there are limitations in current typing methods. We report here the development of a new genotyping method based on amplification of the variable regions of the wzc gene. Fragments corresponding to the variable region of wzc were amplified and sequenced from 76 documented capsular types of reference or clinical strains. The remaining two capsular types (reference strains K15 and K50) lacked amplifiable wzc genes and were proven to be acapsular. Strains with the same capsular type exhibited ≧94% DNA sequence identity across the variable region (CD1-VR2-CD2) of wzc. Strains with distinct K types exhibited <80% DNA sequence identity across this region, with the exception of three pairs of strains: K22/K37, K9/K45, and K52/K79. Strains K22 and K37 shared identical capsular polysaccharide synthesis (cps) genes except for one gene with a difference at a single base which resulted in frameshift mutation. The wzc sequences of K9 and K45 exhibited high DNA sequence similarity but possessed different genes in their cps clusters. K52 and K79 exhibited 89% wzc DNA sequence identity but were readily distinguished from each other at the DNA level; in contrast, strains with the same capsular type as K52 exhibited 100% wzc sequence identity. A total of 29 strains from patients with bacteremia were typed by the wzc system. wzc DNA sequences confirmed the documented capsular type for twenty-eight of these clinical isolates; the remaining strain likely represents a new capsular type. Thus, the wzc genotyping system is a simple and useful method for capsular typing of K. pneumoniae.

Capsule is a major virulence factor of K. pneumoniae, and capsular types are related to the severity of infection [17,18]. The prevalence of capsular types in each K. pneumoniae-related disease could be crucial for disease control and prevention. However, determination of capsular types often is difficult due to the limitations of traditional serotyping [19,20]. The results of serotyping also are inconsistent, except in patients with community-acquired PLA [8,13,19,[21][22][23][24].
Molecular methods based on the capsule polysaccharide synthesis (cps) region have been developed for K. pneumoniae capsular typing. For example, polymerase chain reaction-based genotyping of the capsular polysaccharide synthesis region, cps (wzy)-PCR genotyping, was first adopted for K. pneumoniae type K1 [6,7,[25][26][27][28], and subsequently applied for other capsular types related to community-acquired PLA [19,29,30]. However, only capsular types with known sequences of capsule specific genes (e.g., wzy) can be typed, and a separate pair of primers is needed for each type. PCR amplification of the cps gene cluster (,20 kb) followed by restriction enzyme digestion, i.e., cps PCR-restriction fragment length polymorphism (RFLP) analysis, is another commonly used method. Capsular types can be distinguished based on distinct RFLP profiles (C-patterns) [31]; however, amplifications of the cps region can be very difficult in some strains. In addition, different C-patterns have been observed in some strains that share same capsular type.
As described here, we have developed a new method for capsular typing of K. pneumoniae based on the sequence of the variable region of a gene, wzc, that encodes a capsule synthesisrelated tyrosine kinase.

Ethics statement
The clinical strains used in this study were provided from the strain collection of National Taiwan University Hospital, En Chu Kong Hospital, Far Eastern Memorial Hospital, Chang Gung Memorial Hospital in Taiwan. The Ethics Committee confirmed that no formal ethical approval was needed to use these clinically obtained materials, because the strains were remnants from patient samples, and the data were analyzed anonymously.

Bacterial strains
A total of 77 K-serotype Klebsiella reference strains purchased from Statens Serum Institute, Copenhagen, Denmark. An additional strain (A1517) of novel type KN1 was identified in a previous study from our laboratory [19]. Another eleven K. pneumoniae clinical isolates were obtained from Taiwanese and overseas clinical laboratories, including National Taiwan [19]. Together, strains representing the 78 known capsular types were included for wzc sequencing.
Between 2004 and 2006, Twenty-nine strains were collected from the blood of patients admitted to NTUH with bacteremia. To evaluate the wzc typing system in typing strains with unknown capsular types, all of the 29 K. pneumoniae clinical isolates of unknown capsular type were screened by wzc sequencing.

Sequencing of cps region
Since we failed to amplify wzc genes from reference strains K15 and K50, we instead amplified the cps region from these strains using conserved primers CPS-1 (located in the wzi gene) and rCPS (located in gnd ), as previously described [19]. To permit comparison among the cps regions of selected strains, the corresponding regions were amplified from strains K22, K37, K45, K79, and novel type strain 1461 using primers CPS-1 and rCPS as well. PCR amplifications were performed with the Long and Accurate PCR system. The cycling program consisted of one denaturation step of 2 min at 94uC and 10 initial cycles of 10 s at 98uC, 30 s at 63uC, and 12 min at 68uC, followed by 20 iterative cycles of 10 s at 98uC, 30 s at 63uC, and 12 min plus 20 s for each new cycle at 72uC. A final elongation step was performed for 10 min at 72uC. To extend upstream and downstream from the conserved regions (from galF to gnd), primers pre-galF-F and yegH (located in the sequences at the upstream end of cps) and post-gnd R and ugd (located in the sequences at the downstream end of cps) were used to amplify the flanking sequences [19]. The PCR cycling program for these reactions consisted of 96uC for 3 min, followed by 30 cycles of 96uC for 30 s, 52uC for 15 s, and 72uC for 2-5 min. The products were sequenced by primer walking, providing complete sequences for the cps regions (from galF to gnd, extending approximately 20 kb). The resulting sequences were deposited to Genbank as Accession Numbers AB819892-AB819894, AB819896, AB819897, AB819895, and AB822494). Genes were annotated by NCBI-blast.
Primers specific for the wzy gene of strain 1461 were designed (Table 1) with the intent of confirming the presence of cps genes distinct from the 78 documented capsular types. In parallel to PCR with strain 1461, cps-PCR genotyping using the same primers was performed in 77 K-serotype reference strains (Statens Serum Institute) and KN1 (A1517). Primers pair 1461-wzyF and 1461-wzyR were used in 1461 wzy-PCRgenotyping.

Alcian blue staining
Extracellular polysaccharides, including both capsule and lipopolysaccharide, were isolated as previously reported [33]. Briefly, bacteria were cultured overnight in 1 mL Luria-Bertani (LB) medium and then harvested and resuspended in 150 mL of water. An equal volume of phenol (pH 6.6; Amresco) was added, and the mixture was vortexed. After incubation at 65uC for 20 min, samples were extracted with chloroform and centrifuged. The extracted samples were separated by 10%-sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and capsule was detected with Alcian blue as previously described [34,35]. In brief, after electrophoresis, the gel was washed three times (5 min, 10 min, and 15 min; at 50uC for each step) with fix/wash solution (25% ethanol, 10% acetic acid in water). The gel then was soaked (15 min in the dark at 50uC) in 0.125% Alcian blue dissolved in fix/wash solution, and finally destained (overnight at room temperature) with fix/wash solution. CPS was visualized as blue-stained material.

Capsular type Strain
primer pair 2 yielded products from 76 of the 78 documented capsular types (97%). The exceptions were capsular type K15 and K50 strains, which did not yield product by any of the four primer pairs. PCR products amplified by primer pair 1 were sequenced by the reverse primer KP-wzc-CR1 in 75 of the 81 strains excluding reference strains of K15, K32, K50, K59, K67 and K79; PCR products amplified by primer pair 2 were sequenced by the reverse primer KP-wzc-CR1 in reference strains K32, K59, K67 and K79. The PCR amplicons were sequenced to the start codon of wzc gene by primer walking. All of the sequences deposited in Genbank were from start codon of wzc to CD2 domain (GANNTNNCNNTNNA) which is located in the downstream of VR2 ( Figure 3) and exhibit conservation among 76 capsular types. Thus, the sequences obtained from published cps sequences and from our sequencing results together constitute a 96-strain database of wzc sequences (Table 2). Comparison among the amino acid and DNA sequences of wzc revealed high levels of similarity ( §99% identity both by amino acid sequences and DNA sequences) derived from strains with same capsular type.
Strains belonging to distinct capsular types exhibited lower levels of similarity (40-80% identity by amino acid sequences and 60-80% identity by DNA sequences), with three exceptions. Specifically, the K22 and K37 type strains had wzc sequences that were identical to each other; the K9 and K45 type strains shared 99% amino acid or DNA sequence identity; and the K52 and K79 type strains shared 93% amino acid sequence identity (89% DNA sequence identity). In order to make the method more easily to be used for capsular type identification, conserved regions, CD1 (TNANNGTNTANNC) and CD2 (GANNT-NNCNNTNNA), nearby VR2 were identified in 76 capsular types. The CD1-VR2-CD2 region (115-151 bp in length from different capsular types) was selected for comparison ( Figure 3 and File S1). Therefore, only one-run sequencing using KP-wzc-CR1 (,350 bp from CD2) or KP-wzc-CR2 (,60 bp from CD2) can cover this region for further comparison. The CD1-VR2-CD2 region from distinct capsular types in our wzc database showed ,80% DNA identity and the region derived from strains with same capsular type shared §97% DNA identity with the  K15 and K50 were found to have transposase insertions that precludes capsule expression As noted above, PCR amplification of the ,2.5-3 kb wza-wzbwzc region using the wza and wzc primers failed in reference strains K15 and K50. We therefore amplified and sequenced the full cps region by PCR. The resulting sequences (Accession Numbers AB819895 and AB822494) revealed that both the wzb and wzc genes were replaced by genes encoding transposases both in K15 and K50 (Figure 4). We further designed additional specific primer pairs based on the sequences of the wzy gene of K15 (primers K15-wzyF and K15-wzyR) and the sequences of a gene encoding a glycosyltransferase homolog in K50 (primers K50-gly1F and K50-gly1R) ( Table 1 and Figure 4). PCR performed on each of the 78 capsular type strains confirmed that these primers were specific for the K15 and K50 capsular types (data not shown). Therefore, although wzc genotyping was not successful for capsular type K15 and K50, type-specific primers can be used to genotype the K15 and K50 strains.
Moreover, the wzb and wzc genes are thought to be essential for capsule synthesis in Klebsiella, suggesting the loss of capsule in the K15 and K50 strains. Therefore, we used Alcian blue staining to determine the capsular status of these strains. Our results revealed the absence of CPS in reference strains K15 and K50, as also seen with NTUH-K2044 DmagA, a known capsule-deficient mutant; in contrast, CPS (visualized as high-molecular weight Alcian blue stained material at the top of an SDS-PAGE gel) was observed in positive controls, including a K1 strain (NTUH-K2044) and an isogenic DwbbO (O-antigen-deficient) mutant ( Figure 5). Thus, the reference strains K15 and K50 are acapsular.
cps regions of K22/K37, K9/K45, and K52/K79 As noted above, sequencing of wzc revealed higher than expected DNA sequence similarities between the type strains for K22 and K37 (100% identity at wzc), K9 and K45 (99% identity), and K52 and K79 (89% identity). We therefore further explored the genetic structure of the cps regions in these strains. The sequences (Accession Numbers AB819893 and AB819894) showed that K22 and K37 not only have the same wzy gene which is thought to be distinct among different capsular types, but also have indistinguishable cps regions with the exception of a sequence difference in the ORF downstream of gnd. In K22, this ORF encodes a putative acetyltransferase; in K37, the ORF is truncated as a result of a frameshift mutation (single nucleotide deletion) relative to K22 ( Figure 6). Interestingly, this result is consistent with the previous finding that the capsule structures of K22 and K37 differ only by the presence of acetyl group in K22 CPS [36]. We designed two primers (K22-acylF and K22-acylR) appropriate for amplification of the acetyltransferase gene (Table 1). Sequencing of the resulting amplicon is expected to reveal the status of the putative acetyltransferase-encoding gene, permitting the distinction between K22 and K37 despite identity in both wzc and wzy.
Although the wzc genes of K9 and K45 showed high DNA sequences similarity (99% identity), genes located in the cps regions differed between these two capsular types (Figure 7; Accession Numbers AB371293 and AB819892). We designed primers K9-wzyF, K9-wzyR, K45-wzyF, and K45-wzyR based on the sequences of the wzy genes of K9 and K45 (Table 1 and Figure 7), and demonstrated that PCR amplification with K9-wzyF and K9-wzyR was detected in the K9 capsular type strain but not in the other 77 capsular type strains. Likewise, K45-wzyF and K45-wzyR also showed specificity for capsular type K45 (data not shown). Therefore, these two type-specific primer pairs can be used to distinguish K9 and K45, despite highly similar wzc sequences.
The full sequenced cps regions of K52 and K79 type strains revealed that these type strains possessed different genes in the clusters, although the strains shared 89% identity in wzc sequences (Figure 8; Accession Numbers CP000647 and AB819896). Furthermore, the wzc sequences of the two K52 strains in our panel exhibited 100% identity at the DNA level, suggesting that K52 and K79 can still be distinguished despite similarities in wzc sequences.

wzc genotyping of clinical isolates with unknown capsular types
To evaluate the wzc genotyping system, capsular types of 29 K. pneumoniae blood isolates (obtained from patients admitted to NTUH) were determined by our method. The four primer pairs described above were used for PCR amplifications of the wza-wzbwzc regions of these strains. The four primer pairs provided typing by PCR amplification in 90% (26/29) of these strains using primer pair 1 (KP-wza-CF1 and KP-wzc-CR1), 97% (28/29) using primer pair 2 (KP-wza-CF2 and KP-wzc-CR1), 59% (17/29) using primer pair 3 (KP-wza-CF1 and KP-wzc-CR2), and 62% (18/29) using primer pair 4 (KP-wza-CF2 and KP-wzc-CR2). The combination of primer pairs 1 and 2 permitted typing of all 29 of the tested strains. The amplified PCR products by use of primer pairs 1 were sequenced by the reverse primer KP-wzc-CR1 in 26 of the 29 strains, whereas the primer pair 2-amplicons were subjected to sequencing with KP-wzc-CR1 in the remaining three strains, my1684, 5872, and 5982-2. Sequences from CD1 to CD2 covered VR2 region (115-151 bp) were used for comparing with our wzc database. The results revealed that among the 29 strains, 28 strains showed high DNA sequence similarity ( §94% identity) with documented capsular types in CD1-VR2-CD2 region. Based on the DNA sequences, these 28 strains were classified as capsular type K1 (n = 6), K2 (5) (1). We further confirmed the results by wzy-PCR genotyping using type-specific wzy primers for K1, K2, K14, K16, K20, K23, K39, K54, K62, and KN1 [19,25]. The results demonstrated that wzc genotyping provided results consistent with wzy genotyping ( Table 3). The one (out of 29) remaining strain showed relatively low DNA sequence similarity in CD1-VR2-CD2 region (,70% identity) with the documented capsular types in our wzc panel, suggesting that this strain represented a novel wzc sequence. Therefore, we further evaluated this strain to determine whether this strain represented a new capsular type distinct from the previously described 78 types. Specifically, the cps region of strain 1461 was amplified and the variable regions of cps gene cluster was analyzed (Figure 9) (Accession Number AB819897). Specific primers 1461-wzyF and 1461-wzyR were designed based on the novel wzy sequence (Table 1 and Figure 9); PCR genotyping with this primer pair provided detection only for strain 1461, and not for any of the 78 documented capsular types (data not shown). Based on these results, we infer that this strain likely represents a novel capsular type.

Discussion
Serotyping has been used for determination of K. pneumoniae Ktypes since 1926 [37]. However, several studies have suggested that a substantial proportion (ranging from 23% to 75% in different laboratories) of strains are non-typable by serotyping. [20,22,23]. These observations could reflect limited assay sensitivity, or could reflect limited assay specificity (e.g., serological cross-reactivity between different capsular types). In addition, the high cost and limited sources of anti-sera and tedious experimental procedures of serotyping make the practice of serotyping difficult. Therefore, capsular genotyping methods that bypass the use of anti-sera have become more widely used in discriminating the capsular types of K. pneumoniae [6,7,19,[25][26][27][28][29][30]. PCR-based cps genotyping is a rapid and accurate method for detecting cps genotype [25]. Since the gene layout and DNA sequences of variable regions in the cps synthesis loci are distinct in different capsular types, type-specific primers (located in wzy-like genes or other genes of the cps gene cluster) can be used for distinguishing capsular types. However, this method does not permit detection of all capsular types, because classification cannot be performed unless the DNA sequences of the entire cps gene cluster are available. One study reported a novel capsular genotyping method, cps PCR-RFLP analysis, that permitted typing with high discriminatory power [31]. In this method, capsular types are determined according to the distinct RFLP profiles (C-patterns). In addition, this technique permits distinction among strains with the same K serotype, because subtle differences in DNA sequences can be detected based on variations in cps PCR-RFLP pattern. However, this increased complexity may complicate interpretation of capsular genotyping. Moreover, these two capsular genotyping methods (cps-PCR genotyping and cps PCR-RFLP) require the amplification of the entire ,20 kb capsule synthesis region; such long PCR products can be difficult to obtain. By comparison, the wzc genotyping method (developed in the present study) requires amplification of a ,2.5-3 kb PCR fragment and ,350 bp of DNA sequencing can cover the variable region for comparison. As demonstrated by our PCR analysis of 78 capsular type strains, along with multiple clinical isolates, PCR amplicons were obtained in more than 90% of strains screened with primer pair 1 alone, and in up to 100% of strains screened with the combination of primer pairs 1 and 2. Therefore, our method is expected to be convenient and useful in clinical settings; most isolates will be identifiable using only one or two primer pairs, with few strains requiring testing with additional primers.  Our results indicated that wzc CD1-VR2-CD2 sequences were highly similar ( §94% DNA identity) among strains with the same capsular type. Relatively low levels of similarity (,80% identity) were observed among strains of different capsular types, with the exceptions of K22/K37 (100% wzc identity), K9/K45 (98% identity), and K52/K79 (90% identity). Since K52 and K79 can still be discriminated based on differences in wzc sequences, our proposed typing method is expected to discriminate 74 types (including type K22/K37 and K9/K45). Therefore, only the differentiation between types K22 and K37 (requiring sequencing of the putative acetyltransferase-encoding gene), between K9/K45 (requiring cps-PCR genotyping) and in wzc-deficient K15 and K50 (requiring cps-PCR genotyping) would require the use of additional assays.
After the cps regions of K22 and K37 were resolved, we found that K22 and K37 shared same cps genes for their capsule synthesis. This result was consistent with previous observation that the cps PCR-RFLP patterns of K22 and K37 were  indistinguishable [38] and that K22 and K37 were usually crossreactive by serotyping [39]. Interestingly, we also observed the truncation of a putative acetyltransferase-encoding ORF in K37 compared to K22, providing an explanation for the virtually identical (except for an acetylation modification) capsule structures of K22 and K37 [36]. This phenomenon is similar to that of pneumococcus 9V and 9A, which differ from each other only in the acetylation of capsule. Serotype 9V was found to possess an intact acetyltransferase-encoding gene, while the equivalent gene of serotype 9A was disrupted by a frameshift mutation (deletion of guanine at nucleotide 726) [40].
Although wzc sequences were almost identical in K9 and K45, the cps gene clusters differed between the two capsular types. This could be due to recombination in the region from wzc to gnd, resulting in gene replacement across this interval. Therefore, the cps sequence similarities between K9 and K45 and between K22 and K37 may provide insights into the evolution and divergence of capsular types.
The cps sequences of reference strains K15 and K50 revealed the presence in the clusters of several genes encoding transposase homologs. Notably, the typical wzb-wzc locus of the cps region was replaced by transposase-like genes in these two capsular types. The wzb and wzc genes may have been lost during chromosomal rearrangements associated with transposition events. Wzc, a tyrosine autokinase, is dephosphorylated by its cognate phosphatase, Wzb. Wza, located in the outer membrane, is known to interact with the periplasmic domain of Wzc and is believed to act as a channel [41]. These gene products (Wza, Wzb, and Wzc) are associated with the control of capsule polysaccharide polymerization and cross-membrane translocation, and are thought to be essential for capsule synthesis in E. coli and Klebsiella sp. [42][43][44]. We demonstrated that reference strains K15 and K50 were in fact acapsular. This observation is consistent with the absence of wzb and wzc in these two strains. Capsule structures of reference strains K15 and K50 from the same origin (Statens Serum Institute) have been reported in previous studies in 1992 and 1982 respectively [45,46]. Our reference strains K15 and K50 were purchased from Statens Serum Institute in 2004 and stored at -80uC. Experiments in this study were performed using original stock in our laboratory, therefore, these two strains seemed to have lost capsule before we obtained them.
Using our proposed wzc typing method, we were able to successfully determine the capsular types of all of the clinical isolates tested, with the exception of a single strain that appears to represent a new capsular type. According to the comparison of sequences in our wzc database, strains with same capsular type shared §97% identity, but one strain among the clinical isolates of known types did not hit 97% identity. However, even though the only exception revealed 94% (,97%) DNA identity in CD1-VR2-CD2 region with the corresponding locus of capsular type K20 from our wzc database, wzy PCR genotyping confirmed that this strain was type K20. Therefore, our data suggest that strains harboring wzc CD1-VR2-CD2 sequences of §94% DNA sequence identity can be expected to share the same capsular type. Furthermore, the consistency of the results between wzcand wzy-genotyping suggests that wzc should provide genotyping as accurate as that of wzy. Since not all of the wzy genes for the documented capsular types are currently available, wzc genotyping, a simple alternative method, may be more useful for complete capsular typing. Moreover, we infer that strains with novel wzc sequences probably represent new cps genotypes. Consistent with this hypothesis, we noted that the cps region of strain 1461 was distinct from those of previously reported capsular types. Notably,  the cps gene cluster of strain 1461 was most similar to that of E. coli MS 146-1(Accession No. ADTN00000000). Previous studies had reported that Klebsiella K20 and E. coli K30 harbor identical capsule structures and highly similar cps sequences, implying that horizontal gene transfer had occurred between these strains [47].
Our results with strain 1461 provided further evidence for this phenomenon. Wzc, an inner membrane protein with a cytosolic C-terminal tyrosine autokinase domain, is believed to interact with the outer membrane protein Wza, forming a trans-envelope capsule translocation complex. In the current study, we demonstrated capsular type-specific regions in the wzc locus. And we also found that VR2 region is rich in lysine (a basic amino acid). Therefore, the variable regions of wzc genes might encode binding domains containing positively charged amino acids. The lysine-rich domains might interact with type-specific acidic capsular polysaccharides during the process of translocation.
In conclusion, we have developed a simple and useful capsular genotyping method for K. pneumoniae based on wzc sequences. We demonstrated the use of this typing method for the detection of existing and novel capsular types of K. pneumoniae. Sequencing of cps loci suggested a molecular basis (frameshift mutation) for the difference between types K22 and K37, and revealed that reference strains K15 and K50 were acapsular.

Supporting Information
Table S1 Primers used for wzc sequencing.