The study was approved by the Research Ethics Committee (REC) of the National Taiwan University Hospital (NTUH) (IRB approval #: 9561701018). All animal procedures were approved under application number 20140062 of the Institutional Animal Care and Use Committee (IACUC) of the National Taiwan University College of Medicine (NTUCM). Procedures were consistent with the recommendations of the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health and Taiwan's Animal Protection Act.

The bacterial strains used in this study are listed in Table S1. The K. pneumoniae NTUH-K2044 was isolated from a patient with liver abscess-complicated meningitis and endophthalmitis (Fang et al., 2004). The K-antigen Klebsiella reference strains were obtained from the Statens Serum Institute (Copenhagen, Denmark). A total of 74 clinical isolates of K. pneumoniae were collected from 1997 to 2003 in the National Taiwan University Hospital, as described previously (Hsieh et al., 2008). K. pneumoniae and E. coli strains were cultured in Luria-Bertani (LB) medium or LB medium supplemented with appropriate antibiotics, including 100 μg/mL ampicillin, 50 μg/mL kanamycin, or 100 μg/mL chloramphenicol.

The magA deletion mutant was co-incubated with water from rivers of Taiwan in LB broth overnight. After centrifugation, the supernatant was filtered using a 0.45-μm filter and spotted onto LB plates overlaid with the magA mutant to detect phage plaques. An agar overlay method was used for isolation of a pure phage preparation and to determine phage titers as described elsewhere (Pieroni et al., 1994).

A mutant library of the magA-deletion strain was constructed using random mutagenesis as previously described (Fang et al., 2004).

Lytic activity of O1-1 bacteriophage against the magA-deletion transposon mutants was measured with a spot test. A volume of 1 μL of 10³ plaque forming units (PFU) of the O1-1 bacteriophage was used; the sensitivity of the magA-deletion transposon mutants against the O1-1 was observed by formation of a clear circular zone.

Transposon-insertion sites were determined by semi-random PCR and DNA sequencing as described elsewhere (Chun et al., 1997). The primer pairs for semi-random PCR are listed in Table S1.

5′-Rapid Amplification of cDNA Ends (RACE) -PCR was performed using the SMARTer RACE cDNA Amplication Kit (Clontech Laboratories) following the manufacturer's instructions.

K. pneumoniae mutated in wbbY, wbbZ [polar and non-polar (NP)], magA wbbY (double mutant), or wza wzb (double mutant) were constructed using the previously described unmarked deletion method (Hsieh et al., 2008), which employs electroporation and selection with a temperature-sensitive vector (pKO3-Km) containing flanking regions for each target gene (Simon et al., 1983). For trans-complementation, wbbY, wbbZ with the intergenic region, wbbY with its intact promoter, or the entire wbbY-wbbZ region from K. pneumoniae NTUH-K2044 and E. coli F8188-41 were amplified by PCR and cloned into the pACYC184 vector (Fermentas). The recombinant pACYC184-derived plasmids were transformed into the wbbY, wbbZ, wbbZ-NP, and wbbY magA (double-mutant) strains by electroporation. The primer pairs for the deletion and complementation constructs are listed in Table S1. All sequence data for the wbbY-wbbZ regions from NTUH-K2044 and F8188-41 have been deposited (GenBank accession numbers KJ451390 and KJ451391).

An 18-h culture of each strain was used to inoculate each 5-ml LB broth aliquot at a ratio of 1:100. Each culture was grown at 37°C and growth was monitored hourly by serial dilution and plating to LB agar with next-day quantitation of colony-forming units (CFU).

The amount of K1 CPS from each strain was determined by assaying uronic acid content as described by Domenico (Domenico et al., 1989).

The exopolysaccharide (EPS) extracts (containing both CPS and LPS) were purified by a modified hot water–phenol extraction method, as described previously (Chuang et al., 2006; Hsieh et al., 2012). Samples were separated by 14% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) containing 4 M urea (Tsai and Frasch, 1982) and visualized by silver staining (Tsai and Frasch, 1982). For immunoblots, samples were separated by SDS-PAGE and blotted to a Hybond-C membrane (Amersham, Little Chalfont, UK). Anti-D-Gal II antiserum was obtained from a rabbit immunized with a magA-mutant K. pneumoniae as described elsewhere (LTK BioLaboratories, TW); this serum was diluted 1/50,000 for LPS D-Gal II analysis.

Proton spectra were carried out at 500 MHz at 25°C on solutions formulated in D₂O by using a spectral width of 8.5 KHz, a 32 K data block, and a 90° pulse in a Bruker AVII-500 LC-NMR (Billerica, Massachusetts, USA) spectrometer with standard pulsed sequences and Bruker software.

In order to detect wbbY and wbbZ homologs in various species, PCR was performed using primer pairs specific for wbbY and wbbZ. Primers used in this study are listed in Table S1. PCR was performed as described elsewhere (Chuang et al., 2006).

The survival of exponential-phase bacteria in non-immune human serum was measured as previously described (Hsieh et al., 2012). Briefly, a log-phase inoculum of 2.5 × 10⁴ CFU was mixed at a 1:3 or 3:1 vol/vol ratio with mixed non-immune human serum donated by 5 healthy volunteers. The final mixture, comprising 75 or 25% non-immune serum by volume, was incubated at 37°C for 3 h. For time-course studies, the bacteria/serum mixture, comprising 75 or 25% non-immune serum by volume, was incubated at 37°C for 1, 2, and 3 h. The colony count was determined by plating of serial dilutions on LB agar, and the mean survival ratio was plotted. A mean survival ratio ≥1 corresponds to serum resistance.

Virulence was evaluated by mortality in a murine model of septicemia generated by intraperitoneal (IP) injection. Groups of five-week-old female BALB/c mice were infected IP with the K. pneumoniae NTUH-K2044 (or the isogenic mutants) in 0.1 mL of 0.95% saline (1 × 10² CFU/mouse; 8 mice for each group). The exact inoculation dose was confirmed by serial dilution and plating to LB agar and survival rate of mice was monitored for 4 weeks. For in vivo competition, the wbbY mutant strain and the fully virulent placZ deletion mutant were grown (separately) to log-phase in LB broth, mixed at a 1:1 ratio (the initial dose was 1 × 10³ CFU each) in 100 μL sterile saline, and inoculated IP into 5-week-old BALB/c mice (8 for each group). The mice were sacrificed at 24 h post-inoculation, and the spleen and liver were removed and homogenized in 1x PBS. The concentrations of bacteria in each sample were determined as described previously (Hsieh et al., 2010). The output/input ratio of the test strain to the virulent strain, the competitive index (CI), was interpreted as the in vivo survival capacity. For active immunization, five-week-old BALB/cByl mice were immunized three times by once-weekly IP injection with the indicated dose of the live magA single mutant or magA wbbY double mutant. Age-matched, unimmunized control mice (10 mice per group) were primed with an equivalent volume of saline on the same schedule. After 4 weeks, immunized and unimmunized control mice were challenged with the lethal dose 1 × 10³ CFU of NTUH-A4528 (O1:K2). The challenged mice were observed for 28 days for mortality and clinical signs. Survival was analyzed by Kaplan-Meier analysis with a log-rank test; a P value <0.05 was considered to be statistically significant. For passive protection, 100 μL of serum derived from immunized and unimmunized control mice were administered to five-week-old BALB/cByl mice by IP injection. One hour after serum injection, mice were infected IP with 1 × 10³ CFU of NTUH-A4528 (O1:K2) or 1 × 10⁶ CFU of E. coli F8188-41 per mouse. Three hour after infection, the mice were euthanized and bacterial counts from liver and spleen were determined.

Data are presented as means ± standard deviations (SDs). Statistical significance was assessed by a two-tailed Student's t test using Prism 5 (Graphpad) software. Prevalence was analyzed by chi-square test using SPSS version 12.0 software. Survival was analyzed by Kaplan-Meier analysis with a log-rank test. P values of <0.05 were considered significant.

The magA deletion mutant, which was derived from the NTUH-K2044 (K1:O1), lacks the outermost capsule, thereby exposing the O1 antigen (Hsieh et al., 2012). In the present study, we isolated a K. pneumoniae O1-1-specific bacteriophage using the magA mutant. A magA mutant lawn was successfully lysed by bacteriophage O1-1, yielding visible plaques from phage titers as low as 10 PFU (Figure 1A). In contrast, the wild-type, the wbbO mutant (LPS mutant), and the magA wbbO double-mutant (CPS- and LPS-lacking) strains were all resistant to lysis by bacteriophage O1-1, even when the viral titer was elevated to 10⁵ PFU. This result demonstrates that bacteriophage O1-1 did not infect the encapsulated K. pneumoniae bacteria, presumably due to masking of the receptor by CPS.

We constructed a magA deletion strain transposon library containing 2500 mutants to screen for genes associated with LPS biosynthesis. Phage O1-1-resistant mutants derived from the magA mutant were classified into two types, distinguished by their sensitivity to the O1-1 phage on spot tests. Among 17 phage O1-1-resistant mutants, 16 isolates (94%) were fully resistant to lysis by phage O1-1, even when the viral titer was elevated to 10⁵ PFU. In Figure 1A, this class is represented by the 16-1C mutant. In contrast, 1 of the 17 phage O1-1-resistant mutants, designated 17-9F, showed visible plaques when the viral titer was raised to 10² PFU (Figure 1A). The transposon-insertion sites of these 17 phage O1-1-resistant mutants were determined by semi-random PCR and DNA sequencing. The resulting sequences were analyzed by comparison to the genome sequence of strain NTUH-K2044 (accession number AP006725). The analyses showed that 15 mutants harbored insertions within the wb cluster: five in wbbM, four in wbbN, two in wbbO, two in wzt, one in wzm, and one in glf (Table 1, and Figure 1B). Mutant 16-1C had an insertion in wecA, which initiates LPS biosynthesis by catalyzing the transfer of GlcNAc-1-phosphate onto undecaprenyl phosphate, generating undecaprenyl-pyrophosphoryl-GlcNAc (Figure 1C) (Lehrer et al., 2007). Mutant 17-9F was disrupted in a hypothetical protein-encoding gene; consistent with The Bacterial Polysaccharide Gene Nomenclature scheme (BPGN) (Reeves et al., 1996), this ORF is designated as wbbZ (Figure 1D).

Analysis using the BlastX program showed that the predicted WbbZ protein shares a conserved domain with CsaB of Bacillus subtilis, which is a pyruvyl transferase required for polysaccharide biosynthesis (Mesnage et al., 2000). The gene adjacent to wbbZ, wbbY, encodes a putative glycosyltransferase that might be involved in the synthesis of oligosaccharides, polysaccharides, and glycoconjugates by transferring the sugar moiety from a sugar donor onto an acceptor (Figure 1D). The nearby flanking genes, kp0660, kp0661, and kp0664, are all putative transposase-encoding genes that are likely not associated with the biosynthesis of LPS. Moreover, WbbY and WbbZ share 99% amino acid sequence identity with EcSMS35_2255 and EcSMS35_2256 of E. coli SMS-3-5; these E. coli gene products have been conjectured to contribute to LPS biosynthesis, though the proteins were not further characterized (Fricke et al., 2008). This genetic evidence encouraged us to investigate the association between the wbbY-wbbZ region and LPS biosynthesis of K. pneumoniae.

The sequence features of the wbbY-wbbZ region suggested that wbbY and wbbZ are transcribed as a divergent gene pair; however, these two genes are only separated by a very short intergenic region consisting of 97 base-pairs between the postulated ATGs of these two ORFs. To avoid the polar effects resulting from the single mutations, that is, that deletion of wbbY affected the function of wbbZ and vice versa. Therefore, we determined the transcription initiation sites of both the wbbY and wbbZ genes by 5′-RACE (5′-Rapid Amplification of cDNA Ends) method. The transcription start site and promoter region of wbbZ were located within the short intergenic region between the wbbY and wbbZ genes. In contrast, the transcription start site and promoter region of the wbbY gene were located in the coding sequences of wbbZ (Figure 2A).

To confirm our phage screening results, we constructed in-frame deletion mutant of wbbY, wbbZ, a non-polar deletion mutant of wbbZ that retained the upstream end of the wbbZ ORF (encoding 28 N-terminal amino acids of WbbZ, and corresponding to the DNA sequences of the presumed minimal wbbY promoter) and examined the production of CPS and LPS in these three mutants (Figure 2B). Growth rate and CPS production by in-frame deletion of wbbY, wbbZ or the non-polar wbbZ mutant strain was similar to that of the wild-type strain (Figures 2C,D). In the case of LPS expression, the wild-type and the magA mutant strains displayed similar patterns in both silver staining and immunoblotting, producing both HMW-LPS and LMW-LPS (Figure 3A). In contrast, the wbbO mutant strain completely lost its ability to produce O-antigen, displaying only the core and lipid-A in silver staining. The wbbY mutant was completely depleted in HMW-LPS. The in-frame wbbZ deletion mutant was impaired in the production of HMW-LPS, such that HMW-LPS was detected only at low levels. Unexpectedly, the non-polar wbbZ mutant expressed HMW-LPS and the molecule was detected by anti-D-Gal II antiserum (Figure 3B).

To compensate for the deficiency in D-Gal II synthesis, we transformed the wbbY and wbbZ single-mutant strains with plasmids containing either (i) wbbY, (ii) wbbZ with the intergenic region between these two genes, or (iii) the wbbY-wbbZ region, and then examined HMW-LPS production in the transformants. Compared to the vector control, the LPS of wbbY mutant harboring the wbbY-expressing plasmid shifted to a region of higher molecular weight, and this molecule was detected by anti-D-Gal II antiserum (Figure 3A). An even higher molecular weight LPS was obtained by introducing the wbbY-wbbZ-expressing plasmid into the wbbY mutant; the size of the LPS produced by this strain resembled that of the wild-type strain. Additionally, introduction of the wbbY-expressing plasmid resulted in a shift to a higher molecular weight of LPS in the wbbZ mutant. The wbbZ-expressing plasmid did not compensate for the lack of production of HMW-LPS in the wbbZ mutant. As expected, the wbbY-wbbZ-expressing plasmid fully restored the production of LPS in the wbbZ mutant. These results indicated that the deletion of the wbbZ gene have additional polar effects on wbbY, presumably by removing the promoter of wbbY. Thus, these results implied that only wbbY was essential for the production of D-Gal II in the K. pneumoniae O1 strain.

To investigate whether the wbbY-wbbZ region was sufficient for the expression of D-Gal II, we individually introduced plasmids containing (i) wbbY, (ii) wbbZ with the intergenic region between these two genes, (iii) wbbY with its intact promoter sequence, or (iv) the entire wbbY-wbbZ region into the Klebsiella K72 reference and the wbbY single-mutant strains. The Klebsiella K72 reference strain is an isolate that produces O2 antigen, which is composed only of D-Gal I and runs as LMW-LPS and is not recognized by anti-D-Gal II antiserum (Figure 3C). Introduction of a plasmid carrying wbbY with its intact promoter sequence enabled the K72 and the wbbY single-mutant strains to produce partial D-Gal II antigen of increased molecular weight that was weakly detected by anti-D-Gal II antiserum (Figure 3D). Interestingly, the introduction of the entire wbbY-wbbZ region into the K72 strain enabled the bacterium to produce D-Gal II. This results parallels the full restoration of D-Gal II antigen production to the wbbY mutant strain upon complementation with the plasmid containing the entire wbbY-wbbZ region.

Additionally, we compared the ¹H-NMR spectra of LPS obtained from the wbbY-deletion mutant, from the wbbY mutant harboring the wbbY-expressing plasmid, and from the K72 strain. To rule out the possibility of CPS contamination during LPS analysis, we constructed these mutants in the K-antigen-deficient background. Consistent with the previous NMR assignment (Whitfield et al., 1991), we found that the LPS of the magA wbbY double-deletion mutant strain contained primarily D-Gal I [H-1a: -3)-β-D-Galf-(1- at 5.18 ppm) and H-1c: -3)-α-D-Galp-(1- at 5.06 ppm], resembling the LPS derived from the K72 wza wzb double-deletion mutant strain (D-Gal I). In contrast, the LPS of the magA wbbY double-deletion mutant harboring the wbbY-expressing plasmid and that of the magA mutant strain contained both D-Gal II [H-1b: -3]-α-D-Galp-(1- at 5.15 ppm and H-1d: -3)-β-D-Galp-(1- at 4.66 ppm] and D-Gal I (Figure 3E). In addition, we tried to express the K. pneumoniae O1 antigen in E. coli. Co-transformation of E. coli DH5α with two plasmids containing the wb cluster and the entire wbbY-wbbZ region enabled the bacterium to express a typical smooth lipopolysaccharide that could be detected by anti-D-Gal II antiserum, though the molecular weight was slightly lower than that produced in K. pneumoniae (Figure 3C). No HMW-LPS or smooth LPS was detected in those bacteria containing plasmids carrying the wb cluster only or with vector alone. Taken together, these genetic and chemical analyses indicate that both wbbY and wbbZ were sufficient for the expression of D-Gal II.

In order to detect the production of D-Gal II, we generated and evaluated antiserum raised against LPS D-Gal II by immunization of a rabbit with the unencapsulated magA mutant (K⁻₁ O₁). Antiserum raised against the magA mutant reacted with all (20 of 20) of randomly selected Klebsiella reference strains, which are known to belong to the O1 serotype and produce D-Gal II. In contrast, this antiserum did not react with the Klebsiella K72 strain, an isolate that is known to belong to the O2 type and produce D-Gal I only (Figure 3C). Therefore, the antiserum isolated from the magA-mutant-immunized rabbit is specific to D-Gal II. Using this tool, we compared the distribution of D-Gal II by immunoblotting of samples from 74 clinical isolates of K. pneumoniae, including 42 PLA strains and 32 non-tissue-invasive strains. The D-Gal II-containing strains accounted for 63.5% of all clinical isolates, with significant differences in representation among the strains that caused PLA (38/42; 90.5%) and those that were not tissue-invasive (9/32; 28.1%) (P < 0.0001, chi-squared test).

A previous study indicated that the E. coli O19 antigen is recognized by an anti-Klebsiella O1 monoclonal antibody (McCallum et al., 1989; Fricke et al., 2008). We therefore investigated the link between the wbbY and wbbZ genes and structures based on D-Gal II synthesis in the E. coli O19-type bacterium. Specifically, we carried out PCR analysis using primer pairs located within the wbbY and wbbZ genes. Positive signals were obtained from the acapsulated E. coli O19ab K⁻:H7 strain, F8188-41; the amplicons were of the same sizes as those obtained from K. pneumoniae NTUH-K2044 and A5054, which are known O1-type strains (Figure 4A). We then carried out immunoblotting analysis using the anti-D-Gal II antiserum. All wbbY-, wbbZ-positive strains, including F8188-41, showed positive signals. In addition, D-Gal II was immunologically detected in all of the wbbY and wbbZ single-mutant and the K72 harboring the F8188-41 wbbY-wbbZ-expressing plasmid strains (Figures 4B,C). These results suggested that D-Gal II is present in the LPS of the E. coli O19 group. Moreover, we also determined the DNA sequences of the wbbY and wbbZ genes from F8188-41, which were clustered on the chromosome and showed 95% DNA sequence identity compared with those from the NTUH-K2044 and another O19-type E. coli strain, SMS-3-5. The WbbY and WbbZ homologs from K. pneumoniae NTUH-K2044 and E. coli SMS-3-5 showed 99% protein sequence identity and belong to the same branch by phylogenetic tree analysis (Figure 4D); the E. coli F8188-41 homologs showed lower protein sequence similarity and sorted as a separate phylogenetic branch. To investigate the presence of WbbY and WbbZ homologs in other bacteria, we searched the NCBI database of encoded sequences from other bacterial genomes by BLASTp. WbbY and WbbZ homologs were detected in another member of the Klebsiella genus (K. oxytoca) and in four other more distantly related Gram-negative bacteria (E. coli, Erwinia piriflorinigrans, Serratia plymuthica, and Commensalibacter intestine). In each of these species, the wbbY and wbbZ gene were adjacently located on the respective chromosome and also as divergent transcription units (Table 2). These observations suggested that the D-Gal II structure described in the present study also may be present in Gram-negative bacteria of multiple genera.

Serum killing assays were performed on the K. pneumoniae wbbY and non-polar wbbZ single-deletion mutants to determine the mutants' sensitivities to serum bactericidality. Exposure to human serum (75% human serum for 3 h) resulted in higher levels of killing in the wbbY mutant than in the wild-type and non-polar wbbZ mutant strains (Figure 5A); the wild-type and non-polar wbbZ mutant strains exhibited similar levels of resistance to serum killing. Moreover, introduction of plasmid carrying either wbbY alone or the entire wbbY-wbbZ region successfully restored serum resistance levels to the wbbY mutant strain. This result confirmed that resistance to serum killing correlated with the production of D-Gal II.

To delineate the contribution of CPS, D-Gal I, and D-Gal II to K. pneumoniae's serum resistance, the survival ratio of the wild-type, magA, wbbO, wbbY and non-polar wbbZ single-deletion mutant strains after killing by 75 and 25% human serum was assessed at one-hour intervals through 3 h of exposure (Figures 5B,C). Mutation of magA dramatically reduced the serum resistance of the NTUH-K2044, presumably because the mutant lacked K1 antigen. As expected, wbbO mutation resulted in the loss of O1 antigen; the mutant exhibited elevated susceptibility to serum killing compared to the wild-type, wbbY, and non-polar wbbZ mutant strains. The wbbY mutant, which lacked D-Gal II, showed more susceptible to serum killing than the wild-type and non-polar wbbZ mutant strains. These results demonstrated that the sensitivity to serum killing of the K1:O1 K. pneumoniae strain depended on the integrity of bacterial surface polysaccharides.

To evaluate the contribution of the LPS D-Gal II of K. pneumoniae in bacteremia, we challenged mice with the same dose (1 × 10² CFU) of the wild-type and wbbY deletion-mutant strains. Upon IP infection of mice, the wbbY mutant showed significantly reduced virulence compared to the parent strain (P = 0.024, Figure 5D). Previously, we generated an isogenic lacZ mutant with its promoter deletion and used it as the wild-type strain in the competition assay (Hsieh et al., 2010). In those experiments, no competitive advantage was observed for either the wild type or placZ mutant in the spleen and liver of BALB/c mice. Using the same assay in the present study, the wbbY mutant strain showed lower competitive indices than the wild type or placZ mutant (P < 0.001, Figure 5E). These results indicated that D-Gal II was required for K. pneumoniae's serum resistance and contributed to in vivo virulence.

To evaluate the protective efficacy of immunization against D-Gal II, mice were pretreated by IP injection with either magA or magA wbbY (1 × 10⁶ CFU per dose), administered once weekly for 3 weeks. On the fourth week, the animals (in groups of 10) were challenged IP with a lethal dose of highly virulent K. pneumoniae NTUH-A4528 (O1:K2). Within 6 days of NTUH-A4528 infection, 100% of the unimmunized control mice or the mice pre-immunized with the magA wbbY double-mutant strain died. Meanwhile, 70% of the mice pre-immunized with the magA-mutant (K⁻₁ O₁) survived without any symptoms of disease through 28 days after challenge (P < 0.0001, Figure 5F). Mice immunized with the magA mutant (K⁻₁ O₁) showed anti-D-Gal II immunoglobulin G (IgG) production by immunoblot; mice immunized with the magA wbbY double mutant or unimmunized control mice did not (Figure 5G).

The results of the immunization study were consistent with a separate test in which the passive protective efficacy of anti-D-Gal II antiserum was tested. In the passive immunization study, mice were pretreated with immune mouse sera (negative and positive controls, respectively). The in vivo protective capacity of the sera was tested using a K. pneumoniae or an E. coli septicemia infection model, whereby the mice were inoculated IP with 1 × 10³ CFU of the encapsulated K. pneumoniae NTUH-A4528 (O1:K2) or 1 × 10⁶ CFU of the acapsulated E. coli F8188-41 (O19) strain. The bacterial load was determined in both the liver and spleen at 3 h post-infection. Compared to animals pre-treated with the magA wbbY double-mutant immune mouse serum (IMS-NC), mice pretreated with the magA mutant immune mouse serum (IMS-PC) had significantly reduced bacterial loads (Figure 5H). These results indicated that in both active and passive immunization studies, D-Gal II-specific antiserum was able to reduce bacterial dissemination and protect against infection by encapsulated O1:K2 PLA-associated K. pneumoniae and by D-Gal II-containing E. coli in mouse models of septicemia.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.