Functions of Some Capsular Polysaccharide Biosynthetic Genes in Klebsiella pneumoniae NTUH K-2044

The growing number of Klebsiella pneumoniae infections, commonly acquired in hospitals, has drawn great concern. It has been shown that the K1 and K2 capsular serotypes are the most detrimental strains, particularly to those with diabetes. The K1 cps (capsular polysaccharide) locus in the NTUH-2044 strain of the pyogenic liver abscess (PLA) K. pneumoniae has been identified recently, but little is known about the functions of the genes therein. Here we report characterization of a group of cps genes and their roles in the pathogenesis of K1 K. pneumoniae. By sequential gene deletion, the cps gene cluster was first re-delimited between genes galF and ugd, which serve as up- and down-stream ends, respectively. Eight gene products were characterized in vitro and in vivo to be involved in the syntheses of UDP-glucose, UDP-glucuronic acid and GDP-fucose building units. Twelve genes were identified as virulence factors based on the observation that their deletion mutants became avirulent or lost K1 antigenicity. Furthermore, deletion of kp3706, kp3709 or kp3712 (ΔwcaI, ΔwcaG or Δatf, respectively), which are all involved in fucose biosynthesis, led to a broad range of transcriptional suppression for 52 upstream genes. The genes suppressed include those coding for unknown regulatory membrane proteins and six multidrug efflux system proteins, as well as proteins required for the K1 CPS biosynthesis. In support of the suppression of multidrug efflux genes, we showed that these three mutants became more sensitive to antibiotics. Taken together, the results suggest that kp3706, kp3709 or kp3712 genes are strongly related to the pathogenesis of K. pneumoniae K1.


Introduction
Bacterial pathogenicity has been shown to be due to different causes, including the structures of capsular polysaccharides (CPS; the K antigen), lipopolysaccharide (LPS; the O antigen), secreted toxins, drug resistance, and genetics [1,2,3,4,5]. Klebsiella pneumoniae is an opportunistic pathogen of the Enterobacteriaceae and usually causes pneumonia or urinary tract infections [6]. In addition, the hospital outbreak of multidrug resistant Klebsiella spp., especially the so-called extended-spectrum beta lactamase (ESBL) and Klebsiella pneumoniae carbapenemase (KPC) subtypes, has draw much attention in recent years [6,7,8].
The CPS of K. pneumoniae is complex acidic polysaccharide consisting of repeating units of 3-6 sugars. The type of sugars seems to correlate with the virulence, and 78 capsule types have been identified [9]. In the past two decades, a number of K. pneumoniae strains have been found to cause primary pyogenic liver abscess (PLA) [10,11,12,13], with the capsular serotype K1 being the most virulent [14,15]. The K1 structure has been reported previously (lacking the acetyl-decoration on fucose) to possess two unique features -a fucose subunit (also found only in K54 and K63), and a unique cyclic 2,3-(S)-pyruvate appendix differing from a commonly seen 4,6-(R)-pyruvate in CPS repeat units [16,17].
It has been reported previously that magA (mucoviscosity associated gene A) in the cps locus of NTUH-K2044, a PLAcausing serotype K1 strain from National Taiwan University Hospital [18], is associated with the hypermucoviscosity of the strain and considered a virulence factor [18]. A 31-Kb fragment that covers regions from genes galF (kp3726) to ugd (kp3701) was further identified genetically as the K1 cps cluster ( Figure 1A) [15]. Recently, the whole genome for this PLA strain has been sequenced and annotated, where a previously unidentified acetyl transferase gene was shown in the cps cluster [19]. Our collaborators have re-determined the CPS structure ( Figure 1B), and demonstrated that the acetylation occurs on either C2 or C3 of the fucose (unpublished results). In this work, we took a systematic approach to study the biosynthetic genes of the CPS of NTUH-K2044. Our goal is to identify the functions of the genes and their relevance to pathogenicity.

Identification of specific genes as virulence factors
We examined the functions of individual genes in CPS biosynthesis, aiming to verify the cps locus annotation ( Figure 1A) and to establish the connection between the chemical structure and the biosynthetic genes in their contributions to pathogenicity. First, the genes for NDP-sugar biosyntheses (manB, manC, wcaH, gnd, wcaG, gmd, galF, ugd and uge) (Figure 2A-B) were functionally verified by over expressing and purifying recombinant enzymes and identifying their products by high-pressure liquid chromatography (HPLC) and mass spectrometry. The HPLC analyses of the reaction catalyzed by GalF are shown in Figure 2C as an example. The rest of the HPLC analyses are shown in Figures S1 and S2. The kinetic parameters of Ugd and Gnd are shown in Table S1, and the specificity of ManC and GalF are shown in Tables S2, S3. Detailed procedures are described in Results S1.
Then in-frame deletion mutants were constructed using a modified pKO3 system [20] for ugd, gnd, wbaP, wcaI, wcaH, wcaG, gmd, atf, wzy (magA), ptf, wzc, wzb, and wza in the proposed cps locus Figure 1. Structure of NTUH-K2044 CPS trisaccharide and summary of our work. A, The upper scheme shows the annotated cps cluster and its upstream and downstream genes reported previously. In the middle scheme, the blue bar indicates the CPS gene cluster re-delimited in this work, from gnd (kp3701) to galF (kp3726). In the lower scheme, the grey bar indicates the suppressed gene region [from wbaP (kp3705) to fbaB (kp3767)] in DwcaI, DwcaH, and Datf deletion strains identified in this work. (The color codes used in the cps locus represent the following genes: orange -biosynthesis of glucose in the trisaccharide formation; dark green -biosynthesis of fucose; purple -decoration of acetyl moiety; pinkbiosynthesis of glucuronic acid; light green -decoration of pyruvyl moiety; dark blue -regulation of the cps locus and assembly of the long train CPS.) B, Re-determined CPS trisaccharide structure. doi:10.1371/journal.pone.0021664.g001 (Tables 1). All mutants were obtained except Dugd and Dgmd. Based on both anti K1 serum test by double immunodiffusion assay ( Figure 3A) and string test [15], all mutants obtained (except Dgnd and DwcaH) lost the K1 serotype and mucoviscosity while remaining O1 serotype positive ( Table 2), suggesting that these mutants produce little or no CPS. Moreover, deletion of glf and uge in the proposed lps locus led to loss of O1 serotype, confirming the proposed boundary between cps and lps loci shown in Figure 1A.
Since the deletion mutants make little or no CPS, they are expected to lose pathogenicity also. Animal inoculation experiments were performed for three mutants as examples. As shown in Table 2, the results indicate that deletion of the acetyltransferase gene (Datf or D3712), the pyruvyltransferase gene (Dptf or D3715), or the mucoviscosity associated gene (DmagA, Dwzy, or D3714) was sufficient to cause a significant decrease in the virulence of the NTUH-K2044 strain (LD 50 .10 7 CFU in intra-peritoneal infection, relative to ,10 2 for WT).
These results taken together have verified the functions of individual genes and proteins. Even though the CPS itself can be considered as a virulence factor, our results suggested that the individual genes or proteins responsible for CPS synthesis can also be considered as virulence factors, and thus are potential targets for designing inhibitors against the pathogen. We further examined the properties of the deletion mutants as described below.

Discovery of a broad gene silencing effect related to fucose biosynthesis
We first examined the deletion mutants at the transcriptional level by Q-PCR. Surprisingly, expressions for genes between wcaI and galF, most of the genes in the cps gene cluster, were found completely silenced in d3706, d3709 and d3712 ( Table 3), while they expressed normally (except the deleted gene) in the other deletion mutants (Table S4). To estimate the scopes of the gene silencing effect, genes upstream of the cps locus were also examined and found to be silenced up to kp3767 in these three mutants ( Table 4, left 4 lanes). The total region influenced by the effect was about 70 Kb, including 15 of the 20 cps genes and 37 upstream genes ( Figure 1A and Table S5). Interestingly, the functions of these three genes are all related to the fucose residue of the trisaccharide repeat unit -WcaI (KP3706) is likely the fucosyl transferase, WcaG (KP3709) is responsible for GDP-fucose synthesis, and Atf (KP3712) is for fucose acetylation.
Further support for the broad gene silencing effect That the broad gene silencing effect is novel and real is further supported by three experimental approaches: (a) Complementation experiments with plasmids carrying the deleted gene restored only the expression of the deleted gene, not the other silenced genes ( Table 4, right 3 lanes). This result suggests that the silencing effect is caused by changes at the level of genomic DNA, not simply due to protein expression. (b) The effect of gene silencing was observed from fucose-related metabolites. Analyses of cell fluids extract showed that GDP-Fuc were hardly detectable in these three mutants, while clearly present in two control mutants d3705 and d3715 ( Figure 3B). The results were also verified by MS analyses ( Figure S3). This result supports the silencing effect since only KP3709 is involved in GDP-fucose biosynthesis; the functions of KP3706 and KP3712 occur after We then examined whether the broad silencing effect described above could be due to any known regulatory mechanism. The first to consider was whether the observed silencing effect is a form of well known ''polar effect'' [21] (suppression of a small number of downstream genes) often observed for insertion mutants. Even though the in-frame deletion mutants in our studies were designed to avoid polar effects [20], some insertion mutants from our previous work [15] were examined to see if they display the broad silencing effect as described above for the deletion mutants. As shown in Table 5, polar effects were observed for i3706, i3709, which are distinctly different from the broad silencing effect in the deletion mutants. Furthermore, the K1 serotype of insertion mutants can be restored by complementation as shown previously [15], but not that of the three deletion mutants ( Figure 3A-u,w and Table 4).
Several other known regulatory mechanisms are related to the cps gene cluster, but not involved in the expression of upstream genes: (i) regulators of capsule synthesis (rcs) [22], which sense the extracellular signals then regulate the expression of cps gene cluster; (ii) the transcriptional antiterminator rfaH [1,3], which recognizes Just Upstream of Many Polysaccharide Starts (JUMPstar) to ensure the expression of distal genes; (iii) wzb/wzc [1], which are believed to be involved in the CPS polymerization; and (iv) rmpA2, which is a transcriptional activator and its absence would only lower the capsule production [15]. The expression level of these regulatory genes shows no significant difference between wild-type and mutants.
Small RNA (sRNA) is specifically used to represent bacterial non-coding RNA. Since it has been suggested that sRNA [23]  effects could be dependent on the temperature of bacterial growth [24] or the state of growth [25], we examined the gene silencing effect at different temperatures (25, 42 and 45uC) ( Table S6) and at the stationary phase of the growth ( Table S7). The results indicate that these factors did not affect the broad gene silencing effect observed for the three deletion mutants.

Synergy between the fucose-related virulence factors and drug resistance
To test whether the broad gene silencing effect is related to the pathogenicity of the NTUH-K2044 strain, we examined possible functions of the silenced genes. Importantly, the 37 genes upstream of the cps locus affected by the gene silencing effect include many regulators and multidrug efflux genes (from kp3742 to kp3747) ( Figure 1A and Table S5). Since multidrug efflux pumps are known to contribute to drug resistance in Gram-negative bacteria [26], this finding led us to predict that the three mutants d3706, d3709, and d3712 would be less drug-resistant to some of the antibiotics. As shown in Figure 4A, in the absence of antibiotics there were no significant differences in growth rates between K2044 and its mutants, confirming that CPS is important for pathogenicity but not for growth (Duge or d3699, a slow-growing mutant, was used as a positive control as Uge is involved in the biosynthesis of LPS). Then we examined the effects of various antibiotics on the growth rates of these strains. Two of the antibiotics tested, zeocin (a member of the bleomycin/phleomycin family of antibiotics known to bind and cleave DNA) ( Figure 4B) and erythromycin (a macrolide family of antibiotics known to interfere with protein synthesis) ( Figure 4C) showed significant inhibition against DwcaI, DwcaG, and Datf (d3706, d3709, and d3712, respectively) but had limited effect against wild-type and the other mutants. In contrast, tetracycline, sulfamethoxazole, ciprofloxacin, and geneticin, which are classified as members of tetracyclines, sulfonamides, quinolones, and aminoglycosides antibiotics, respectively, did not show specific effects on the three mutants ( Figure S4A-S4D).
If the effects of zeocin or erythromycin on the three deletion mutants are indeed caused by silencing of multidrug efflux genes, we should be able to identify a single efflux gene, construct the deletion mutant of the single gene, and show that the mutant is sensitive to the antibiotics. To test this possibility, a kp3742 (yegM) deletion mutant was constructed. The inhibition assays showed that the growth of the DyegM strain was inhibited by zeocin ( Figure 4D) but not erythromycin ( Figure 4E) or the other antibiotics (not shown).
There has been growing evidence that fucose [27] and acetylation [28,29,30] are important factors in bacterial pathogenicity and in other diseases. For example, Helicobacter pylori is a well known primary cause of gastritis, duodenal ulcer, and gastric cancer, which infects about one-half of human population. Liu et al. demonstrated the important connection between human Lfucosidase (FUCA2) and H. pylori on the adhesion pathogenesis and escape of host surveillance [31]. In addition, Coyne et al. also demonstrated that bacteroides in mammalian intestine use a mammalian-like pathway to decorate numerous capsular polysaccharides and glycoproteins with fucose [32]. No mechanism has been suggested about the special roles of fucose. Our results add an example in the potential biological significance of fucose, and we have discovered some link between fucose biosynthetic genes and gene silencing. However the specific functions of fucose or fucose biosynthetic genes remain to be further investigated. Figure 3. Characterization of the in-frame deletion mutants (dxxxx), complementation strains of the deletion mutants (cxxxx), and insertion mutants (ixxxx). A, Double immunodiffusion assay for the K1 serotype. a: wild-type, b: d3695, c: d3699, d: d3704, e: d3713, f: d3714, g: i3705, h: d3705, i: i3706, j: d3706, k: i3709, l: d3709, m: d3712, n: i3715, o: d3715, o: d3718, q: d3719, r: d3720, s: wild-type, t: d3708, u: c3706, v: c3709, w: c3712. Only wild-type, d3695, d3699, d3704, and d3708 remained K1 positive. B, HPLC analysis of NDP-sugars. All peaks were determined by the standard retention time and confirmed by MS ( Figure S3). The GDP-fucose peak is indicated by the red arrow. The data for the wild-type was not obtained due to the high mucoviscosity, which prohibited extraction of metabolites. doi:10.1371/journal.pone.0021664.g003 In summary, our results on the CPS of the NTUH-K2044 strain of the PLA K. pneumoniae add to the emerging evidence that fucose [32] and O-acetylation [28] are important factors in bacterial pathogenicity, and our genetic analyses of deletion mutants uncover a possible link between three fucose-related genes and the multidrug efflux genes via an unprecedented broad gene networking. This phenomenon potentially represents a newly uncovered pathogenesis mechanism. These results led us to hypothesize that the broad silencing effects observed for the three in-frame deletion mutants are caused by DNA structure alteration in the deletion mutants of the three fucose-related genes, and that the silencing effects may be relevant to the specific roles of the fucose and its acetylation in the pathogenicity of the strain. The validity of the hypothesis, and the mechanism of the broad silencing effect, are important subjects for future studies.

Bacterial strains and plasmid vectors
Bacterial strains and plasmids used in this study are listed in Table 1. Clinically isolated K. pneumoniae strains were collected at National Taiwan University Hospital (NTUH) [18].

Serum resistance assay and animal inoculation
The serum resistances of K. pneumoniae strains were determined as previously described [33]. Female BALB/cByl 5-week-old mice were used for inoculation. BALB/cByl mice were bred and housed in specific pathogen-free rooms within animal care facilities of the Laboratory Animal Center at the National Taiwan University College of Medicine (NTUCM) with free access to food and water. All procedures were approved by the National Taiwan University College of Medicine and College of Public Health Institutional Animal Care and Use Committee (IACUC), following with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health and Taiwanese Animal protection act. IACUC approval number: 20060139. K. pneumoniae inoculation consisting of 10 2 -10 6 mid-logarithmic growth phase CFUs were diluted in 100 ml normal saline and injected intraperitoneally [34,35]. Four mice were used to test the effects of each inoculum. After inoculation, the mice were observed for 30 days. The LD 50 was calculated using the method established by Reed and Muench [35].

Construction of K. pneumoniae deletion mutant strains
The deletion mutants were generated using a modified pKO3-Km vector that contained a temperature-sensitive origin of replication and markers for positive and negative selection for chromosomal integration and excision [36,37]. The genes and its flanking regions were amplified by PCR using primers listed in Table S8 and cloned into a pGEM-T easy vector. The deletion fragment was generated by inverse PCR using primers listed in Table S8. The deletion fragments described above were cloned into a NotI site of a pKO3-Km plasmid separately [36]. The resulting constructs were then electroporated into wild-type strain. The transformants were cultured at 43uC. Five colonies were picked in 1 ml LB broth followed by serial dilution and   plating onto LB plates containing 5% sucrose cultured at 30uC. Colonies sensitive to kanamycin were screened and confirmed by PCR using appropriate combinations of primers, and DNA sequencing.

Trans-complementation
The wcaG, wcaI, and atf genes were amplified by PCR using primers (Table S1) and cloned into a pGEM-T Easy-Km plasmid [9]. These plasmids were transformed into their corresponding isogenic mutant strains by electroporation [38]; for selection of complementation strains, LB agar plates were supplemented with 50 mg/ml kanamycin or 100 mg/ml chloramphenicol.

String test
The mucoviscosity of K. pneumoniae was determined by a string test as previously described [33]. The bacteria were cultured on 5% sheep blood agar plates overnight. Colonies were tested by inoculation loop.
If the colony could be stretched to a string longer than 0.5 cm then it was defined as positive string test; otherwise as negative.

K1 and O serotyping
Capsule and lipopolysaccharide were purified as previously described [15]. The K1 serotyping was performed by the double immunodiffusion assay using a serotype K1-specific antiserum (Statens Serum Institute) as previously described [15]. After separation in 12% SDS-PAGE gel, the O antigen was detected by sera from mice infected with a capsule-deficient magA deletion mutant.

Quantitative PCR
1 mg of extracted RNA was reverse transcribed into cDNA in 20 ml reaction as the manufacturer protocol (Roche, US). 20 ml of cDNAs were then diluted into 1 ml. 5 ml of diluted cDNA samples were added into the Q-PCR reaction plate, and also the 10 ml of reaction solution. The analysis of Q-PCR result is that each gene was first normalized with 23S RNA. Each gene in each strain was then normalized with wild-type to obtain the relative gene expression pattern. The primers used for Q-PCR are listed in Table S9.

Extraction and separation of nucleotides and sugar nucleotides
For isolation of soluble fractions containing sugar nucleotide/ nucleotides from K. pneumoniae strains, cells were collected when  OD 600 was 1.0. The cells were harvested by centrifugation at 5000 rpm for 30 min at 4uC. Cell pellets were re-suspended in 1 ml phosphate-buffered saline (PBS, pH 7.4), followed by addition of a mixture chloroform/methanol (1:2) and vortexing for 10-15 min. The mixture was then centrifuged at 4000 rpm for 15 min, the pellet was removed and mixed with chloroform and ddH 2 O (1:1), and centrifuged again. The upper phase containing soluble components were collected and dried under air. It was redissolved in distilled water and further purified by Amicon filter (YM-10 cutoff), and the filtrate was collected and monitored by anion exchange HPLC using ammonium formate. The peaks were identified by comparing the retention times and MS with known standards.

Growth curves of WT and mutants and effects of antibiotics
NTUH-K2044 and knock-out strains were grown in LB broth at 37uC. For growth curves, log phase cultures were diluted to OD 600 = 0.005 in LB broth with desired antibiotics. The growth curve was plotted by measuring OD 600 periodically.