Role of the cAMP-Dependent Carbon Catabolite Repression in Capsular Polysaccharide Biosynthesis in Klebsiella pneumoniae

K. pneumoniae is the predominant pathogen isolated from liver abscesses of diabetic patients in Asian countries. Although elevated blood glucose levels cause various immune problems, its effects on K. pneumoniae virulence are unknown. This study investigated the regulation of capsular polysaccharide (CPS) biosynthesis, a major determinant for K. pneumoniae virulence, in response to exogenous glucose. We found that K. pneumoniae produce more CPS in glucose-rich medium via reduction in cyclic AMP (cAMP) levels. Individual deletion of cyaA or crp, which respectively encode adenylate cyclase and cAMP receptor protein in K. pneumoniae, markedly increased CPS production, while deletion of cpdA, which encodes cAMP phosphodiesterase, decreased CPS production. These results indicate that K. pneumoniae CPS biosynthesis is controlled by the cAMP-dependent carbon catabolite repression (CCR). To investigate the underlying mechanism, quantitative real-time PCR and promoter-reporter assays were used to verify that the transcription of CPS biosynthesis genes, which are organized into 3 transcription units (orf1-2, orf3-15, and orf16-17), were activated by the deletion of crp. Sequence analysis revealed putative CRP binding sites located on Porf3-15 and Porf16-17, suggesting direct CRP-cAMP regulation on the promoters. These results were then confirmed by electrophoretic mobility shift assay. In addition, we found putative CRP binding sites located in the promoter region of rcsA, which encodes a cps transcriptional activator, demonstrating a direct repression of CRP-cAMP and PrcsA. The deletion of rcsA in mutation of crp partially reduced CPS biosynthesis and the transcription of orf1-2 but not of orf3-15 or orf16-17. These results suggest that RcsA participates in the CRP-cAMP regulation of orf1-2 transcription and influences CPS biosynthesis. Finally, the effect of glucose and CCR proteins on CPS biosynthesis also reflects bacterial resistance to serum killing. We here provide evidence that K. pneumoniae increases CPS biosynthesis for successful infection in response to exogenous glucose via cAMP-dependent CCR.


Introduction
Klebsiella pneumoniae is a Gram-negative pathogen which causes suppurative lesions, bacteremia, and urinary as well as respiratory tract infections mostly in patients with underlying diseases [1]. In Asian countries, especially in Taiwan and Korea, K. pneumoniae is the predominant pathogen responsible for pyogenic liver abscess in diabetic patients [2,3,4]. In recent years, reports of Klebsiella liver abscess (KLA) in western countries have also been accumulating [5]. Among the virulence factors identified in K. pneumoniae, capsular polysaccharide (CPS) is considered as the major determinant for K. pneumoniae virulence. Pyogenic liver abscess isolates often carry heavy CPS loads that could protect the bacteria from phagocytosis and killing by serum factors [6,7]. The capsular serotypes of K. pneumoniae have been classified into more than 77 known types [8,9]. In Taiwan, a high prevalence of the K1 and K2 serotypes of K. pneumoniae was documented in liver abscess of diabetes mellitus patients [10]. However, the exact mechanism of the tight association between K. pneumoniae, liver abscess, and diabetes mellitus remains unclear.
Diabetic patients have been reported to be more susceptible to infections [11,12]. It has also been demonstrated that K. pneumoniae strains are more virulent in diabetic than in normal mice [13]. The increased risk of bacterial infection in diabetic patients has been studied with regard to host immune system defects [14,15,16]; however, the alteration of gene expression patterns of pathogenic bacteria in response to elevated blood glucose levels awaits further investigation. First studied in Escherichia coli but highly conserved across bacteria, the carbon catabolite repression (CCR) regulates uptake of glucose and repression of genes required for utilization of less preferred carbon sources [17,18,19]. The CCR is generally controlled by the second messenger cyclic AMP (cAMP), which has a fundamental role in global gene regulation [20]. Bacteria grown in glucose show inhibited cAMP production, while bacteria grown in less-preferred carbon sources produce elevated levels of cAMP [17,19,21]. To balance intracellular cAMP levels, the adenylate cyclase CyaA and the cAMP phosphodiesterase CpdA, are required for cAMP biosynthesis and degradation, respectively [17,19,22,23]. The cellular target for cAMP signalling is the cAMP receptor protein (CRP). To regulate mRNA transcription, CRP consists of a homodimer with cAMP and exhibits DNA-binding activity to the CRP binding site (TGTGA-N6-TCACA and TGCGA-N6-TCGCA) in promoter regions [24,25,26,27]. In E. coli, CRP-cAMP acts as a global regulator of gene expression by controlling the expression of almost 200 operons [28,29,30]. In addition to the regulation of carbon metabolism genes, cAMP signalling has been demonstrated to regulate the expression of various genes encoding virulence factors, such as flagella, fimbriae, protease, exotoxin, and secretion systems in bacteria [31][32][33][34][35][36][37][38][39][40].
Sequence analysis revealed a high similarity between CCR proteins (CyaA, CpdA, and CRP) in E. coli and K. pneumoniae, suggesting a conserved regulatory mechanism. In the previous study, CRP-cAMP has been demonstrated to regulate the expression of citrate fermentation genes in K. pneumoniae under fermentative conditions [41]. However, the role of CCR proteins in K. pneumoniae pathogenesis is large uncharacterized. In this study, we aimed to examine the effect of glucose levels and CCR proteins on the regulation of K. pneumoniae CPS biosynthesis. Individual strains of K. pneumoniae CG43, a highly virulent liver abscess isolate of the K2 serotype, in which cyaA, cpdA, and crp had been deleted were constructed for the assessment of CPS production, and the regulatory mechanism of cAMP-dependent CCR in cps transcription was analysed.

Glucose Stimulates CPS Biosynthesis
To analyse if exogenous glucose affects K. pneumoniae CPS biosynthesis, CG43S3 was grown in LB broth supplemented with increasing amount of glucose for the quantification of CPS. As shown in Fig. 1, the CPS level increased when bacteria were grown in LB broth supplemented with 0.25% and 0.5% glucose, while the addition of 0.1% glucose did not have an obvious effect. Since the presence of glucose in the growth medium has been demonstrated to inhibit the cAMP production in many bacteria [17,21], we examined whether the elevated CPS production was regulated by cAMP. Increasing amounts of exogenous cAMP were added to LB broth supplemented with 0.5% glucose, and bacterial CPS production was determined. The result showed that the addition of exogenous cAMP repressed the effect of glucose on CPS production, suggesting that environmental glucose can activate K. pneumoniae CPS biosynthesis through a reduction of cAMP level.

CCR Proteins Affect CPS Biosynthesis
To confirm that K. pneumoniae CPS biosynthesis is regulated by cAMP, individual strains with deletion of cyaA and cpdA, which respectively encodes adenylate cyclase and cAMP phosphodiesterase from CG43S3, were constructed, and the effects of the deletions on CPS production were analysed. As shown in Fig. 2A, compared to wild type (WT) strain, we found that CPS levels increased in DcyaA, and introduction of pcyaA, but not the empty vector control (pACYC184), into DcyaA could reverse the effect of cyaA mutation. In contrast, the deletion of cpdA caused a decreased in CPS levels, which could be complemented by introducing a plasmid-carried cpdA (pETQ-cpdA) into the DcpdA strain. These results confirmed that cAMP can act as a signalling molecule for regulation of CPS biosynthesis. In addition, since cAMP affects gene transcription through its effector protein CRP, we assessed the effect of deletion of crp on CPS levels. As shown in Fig. 2B, compared to WT, Dcrp produced higher levels of CPS. Introduction of the complement plasmid pcrp, but not the empty vector control (pACYC184), into Dcrp reversed the effect of the deletion. This result indicates that the CRP-cAMP signalling pathway is involved in the regulation of CPS biosynthesis, and that CRP acts as a negative regulator of CPS biosynthesis. In addition, since the functions of CyaA, CpdA, and CRP in controlling the cAMP level in K. pneumoniae have not yet been demonstrated, enzyme-linked immunosorbent assays were performed to determine the intracellular cAMP level upon the deletion of cyaA, cpdA, or crp. Compared to WT (9.7560.35 nM), the intracellular cAMP level was almost undetectable in DcyaA (,1 nM), whereas a higher cAMP level was found in DcpdA (3862.8 nM) (Fig. 2C). In addition, a slight increase in the cAMP level was found in Dcrp (14.560.7 nM). These result confirmed that CyaA and CpdA are responsible for cAMP biosynthesis and degradation, respectively, in K. pneumoniae.

Effect of cAMP-dependent CCR on cps Transcription
The K2 cps gene cluster of K. pneumoniae Chedid contains 19 open reading frames (ORFs) organised into 3 transcription units, namely, orf1-2, orf3-15, and orf16-17 [42]. To investigate the effect of glucose and cAMP-related proteins on the expression of the 3 cps gene clusters, the mRNA level of orf1 (named galF), orf3 (named wzi), and orf16 (named manC) were measured by qRT-PCR in WT grown in LB medium containing 0.5% glucose with or without 1 mM cAMP. In addition, the effect of cyaA, cpdA, andcrp mutation strains on the mRNA levels of galF, wzi, and manC were also determined. As shown in Fig. 3A, we found that the mRNA levels of galF, wzi and manC was increased in glucose-rich medium (LB+0.5% glucose), whereas addition of 1 mM cAMP to the glucose-rich medium could restore the galF, wzi and manC expression, similar to the trends observed in the WT strain. Furthermore, the mRNA levels of galF, wzi, and manC revealed an apparent increase in the DcyaA and Dcrp strains as compared to that observed in the WT strain. In contrast, a slight reduction in the mRNA level of cps genes was found in the DcpdA strain. This result indicates that cps mRNA expression is regulated by cAMPdependent CCR, and CRP may acts a repressor of cps expression. To further confirm whether CRP acts as a transcriptional repressor of the promoter activity of galF, wzi, and manC, the reporter plasmids pOrf12 (P orf1-2 ::lacZ), pOrf315 (P orf3-15 ::lacZ), and pOrf1617 (P orf16-17 ::lacZ), each carrying a lacZ reporter gene transcriptionally fused to the putative promoter region of the K2 cps gene cluster [43], were used to transform the K. pneumoniae strains CG43S3DlacZ and DlacZDcrp. The promoter activity measurements shown in Fig. 3B revealed that the deletion of crp in the DlacZ strain apparently increased the promoter activities of galF, wzi, and manC. These results verify that CRP represses cps expression at the transcriptional level.

Determination of Transcriptional Start Sites on 3 Transcriptional Units in the K2 cps Gene Cluster
Until now, the transcriptional start sites of the cps gene cluster had not been characterized. 59 rapid amplification of cDNA ends (59 RACE) was first performed to determine the transcriptional start sites of the 3 transcription units in the K2 cps gene cluster. A single DNA band was obtained for galF, wzi, and manC from the 59 RACE analysis using either primer pair (data not shown). As shown in Fig 4A, B, and C, sequence analysis of a total of 10 clones each from galF, wzi, and manC revealed a transcriptional start site at the A nucleotide at position 261 relative to the translational start site of galF, at the G nucleotide at position 2470 relative to the translational start site of wzi, and at the G nucleotide at position 256 relative to the translational start site of manC. The conserved 210 and 235 promoter sequence of s70 could be readily identified and is shown in Fig 4. To further investigate the mechanism of CRP-cAMP regulation of cps gene transcription, the sequences of the E. coli CRP binding sites (TGTGA-N6-TCACA and TGCGA-N6-TCGCA) were used in searching the promoter sequence of the K2 cps gene cluster of K. pneumoniae Chedid. The maximum number of possible mismatched nucleotides was set at 3, and only the intergenic regions of the 3 transcriptional units were analysed in this search. Using these criteria, no typical CRP binding sites were located the upstream of galF (Fig. 4A). However, 3 CRP binding sites were found in the sequence upstream of wzi (Fig. 4B), and one CRP binding site was found to be located at position 2140 to 2125 relative to the transcriptional start site of manC ( Fig. 4C and D). Among the 3 CRP binding sites were located in the intergenic regions of orf2 and wzi, one was located in the 59 mRNA region that is 137 bp upstream of the translation start site of Wzi (at position +319 to +344 relative to the transcriptional start site of wzi), the other 2 CRP binding sites were found to be located at position 229 to 214 and 287 to 272 relative to the transcriptional start site of wzi (Fig. 4D). The result implies that CRP could bind directly to the CRP binding sites that are located in P wzi and P manC for controlling cps expression.

CRP Directly Binds to the CRP Binding Sites in P wzi and P manC
To demonstrate whether CRP binds directly to the upstream sequence of wzi and manC via the CRP binding sites, electric mobility shift assay (EMSA) was performed using the recombinant His 6 -CRP protein and different DNA fragments containing truncated forms of Pwzi (Pwzi-1, Pwzi-2, Pwzi-3, and Pwzi-4) and PmanC (PmanC-1 and PmanC-2), as described in the Materials and Methods. Using Pwzi fragments of different lengths, binding of His 6 -CRP could be observed for Pwzi-1, Pwzi-2, and Pwzi-3 but not for Pwzi-4 ( Fig. 5A). As shown in Fig. 5A, DNAprotein-binding complexes were observed after the incubation of 100 nM purified His 6 -CRP with 10 ng Pwzi-1 or Pwzi-2. However, formation of the Pwzi-3/CRP complex required the incubation with 200 nM purified His 6 -CRP. This suggests that the lower binding ability of His 6 -CRP in Pwzi-3 is due to the location of only one CRP binding site in Pwzi-3 as compared to 2 and 3 CRP binding sites in Pwzi-1 and Pwzi-2, respectively. In addition, we found that His 6 -CRP was able to bind to the PmanC-1 DNA fragment, but not to the PmanC-2 fragment in which the CRP binding site had been removed (Fig. 5B). This result indicates that the recombinant His 6 -CRP protein can bind directly to the predicted CRP binding sites located in P wzi and P manC .

Expression of rcsA is Controlled by cAMP-dependent CCR and Directly Repressed by CRP
Because no CRP binding site was found in the sequence upstream of galF, the expression of galF also appeared to be controlled by cAMP-dependent CCR, implying that CRP repression of galF transcription is indirect and that other transcription factors are involved in the CRP regulon controlling cps transcription. According to previous studies, multiple regulatory proteins, which include Fur, RcsA, RcsB, RmpA, RmpA2, KvgA, and KvhR have been shown to mediate K2 cps expression [43,44,45,46]. To further investigate whether these cps regulatory proteins are involved in CRP regulon, the CRP binding site was qRT-PCR analyses of the expression of the K2 cps genes (orf1, orf3, and orf16) for WT, DcyaA, DcpdA, and Dcrp strains in LB or indicated LB medium. (B) b-galactosidase activities of K. pneumoniae CG43S3DlacZ and the isogenic strain (DlacZDcrp) carrying the reporter plasmid pOrf12 (P orf1-2 ::lacZ), pOrf315 (P orf3-15 ::lacZ), or pOrf1617 (P orf16-17 ::lacZ) were determined using log-phase cultures grown in LB medium. The results shown are an average from triplicate measurements in one single experiment representative of three independent experiments. Error bars indicate standard deviations. *P,0.05 and **P,0.01 compared to the indicated group. doi:10.1371/journal.pone.0054430.g003 searched in the upstream sequence of fur, rcsA/B, rmpA/A2, kvgA, and kvhR. However, we found the 2 CRP binding sites (rcsA-1 and rcsA-2) are located at 2192 to 2177 and 240 to 225 relative to the translation start site of RcsA (Fig. 6A), but no typical CRP binding site was found in other upstream sequence of fur, rcsB, rmpA/A2, kvgA, and kvhR. Therefore, we suggest that RcsA is a CRP-regulated transcription factor and is involved in cAMPdependent CCR control of cps expression.
To verify this possibility, the effect of glucose and cAMPdependent CCR on the mRNA level of rcsA was first determined by qRT-PCR. As shown in Fig. 6B, addition of 0.5% glucose to LB medium apparently increased the mRNA level of rcsA, while addition of exogenous 1 mM cAMP to glucose-rich medium could restore the level of rcsA expression level to the same as that observed in the WT strain. In addition, the mRNA level of rcsA increased in the DcyaA strain, while the deletion of cpdA reduced on rcsA expression. This result indicates that rcsA expression is controlled by the intracellular cAMP level in response to exogenous glucose. In addition, the deletion of crp also increased the expression of rcsA, suggesting that CRP acts a transcriptional repressor of rcsA expression. Furthermore, measurement of promoter activity confirmed the suggestion that the deletion of crp caused a higher level of expression of P rcsA (Fig. 6C).
To further investigate whether CRP binds directly to the rcsA promoter region, EMSA was performed. As shown in Fig. 6D, DNA-protein binding complexes were observed after the incubation of 100 nM purified His 6 -CRP with 10 ng PrcsA-1, but not with PrcsA-2 in which one of the CRP binding sites was removed. Therefore, we suggested that CRP-cAMP binds directly to the CRP binding site (rcsA-1) in P rcsA to repress rcsA transcription.

Role of RcsA in Regulation of CRP on CPS Biosynthesis
To understand whether RcsA participates in CRP regulation of CPS biosynthesis, the level of CPS was determined in WT, DrcsA, Dcrp, and DcrpDrcsA strains. As shown in Fig. 7A, the deletion of rcsA resulted in a slight reduction in CPS level as compared to WT strain. However, the deletion of rcsA partially restored CPS production in the Dcrp strain. In addition, introducing the complementary plasmid prcsA into the DcrpDrcsA strain increased CPS levels as compared to the strain carrying the empty vector control (pRK415). These results indicate that RcsA participates in CRP regulation of CPS biosynthesis.
To further understand the role of RcsA in CRP regulation of cps transcriptions, the promoter activity of galF (pOrf12), wzi (pOrf315), and manC (pOrf1617) was measured in DlacZ, DlacZDcrp, DlacZDrcsA, and DlacZDcrpDrcsA strains. As shown in Fig. 7B, the deletion of rcsA in DlacZ caused an apparent reduction on the promoter activity of galF, but no effect on the promoter activity of wzi and manC was observed, indicating that RcsA plays a positive role in galF transcription. In addition, the deletion of rcsA in the DlacZDcrp strain caused a slight reduction in the promoter activity of galF, unlike that in the Dcrp strain, but still showed higher promoter activity than the DlacZ strain. The result implies that RcsA participates in CRP regulation of galF transcription. However, we suggest that other unknown transcriptional regulator(s) are also involved in this regulation.

Effect of cAMP-dependent CCR on Susceptibility to Normal Human Serum
Since CPS has been demonstrated to protect K. pneumoniae from killing by serum factors, we suggest that glucose, cAMP, and cAMP-related proteins may also affect the ability of K. pneumoniae, through modulation of CPS levels, to resist the bactericidal effects of serum. To test the hypothesis, the effects of exogenous glucose and cAMP on survival rate were first determined in treatment with 75% normal human serum. We found that the survival rate of WT grown in LB supplemented with 0.5% glucose (43.263.9%) increased about 2-fold as compared to the survival rate of WT in LB alone (21.262.3%). Addition of 1 mM cAMP diminished the higher survival rate of WT when grown in LB with 0.5% glucose (34.360.1%). This implies that the intracellular cAMP level decreased in K. pneumoniae grown in LB with 0.5% glucose, which then increased the serum resistance in K. pneumoniae. In addition, the deletion of cyaA in WT apparently increased the survival rate (40.760.7%), while the deletion of cpdA reduced the survival rate (12.760.6%). It also confirmed that the intracellular cAMP level could affect the ability of K. pneumoniae to resist the bactericidal effects of serum. Finally, the deletion of crp in WT had a higher survival rate (28.561.9%) in treatments with 75% normal human serum as compared to WT strain. Thus, the results imply that K. pneumoniae could resist serum killing in response to higher glucose levels via the trigger of cAMP-dependent CCR.

Discussion
Clinically isolated K. pneumoniae strains usually produce a large amount of CPS, which confers not only a mucoid phenotype to the bacteria but also resistance to engulfment by phagocytes or to serum bactericidal factors [7,47]. The degree of mucoidy has also  been positively correlated with successful establishment of infection [48,49]. Although CPS has been repeatedly proven to play an important role in K. pneumoniae infections and multiple CPS regulators have been found, the environmental stimuli that modulate CPS biosynthesis has remained largely unknown. Our previous studies have reported that extracellular ferric ion could repress K. pneumoniae CPS production through Fur regulation [44,46]. Moreover, in the present study, we found that environmental glucose stimulated CPS production, which was regulated by cAMP (Fig. 1) and CCR proteins (Fig. 2), resulting in an increased resistance to serum killing. These findings imply that, in response to elevated blood glucose levels in diabetic patients, K. pneumoniae could produce more CPS to facilitate its persistence in the blood.
In E. coli, the expression of cps::lacZ was activated when cells were grown at a low temperature (20uC) in the presence of glucose (0.4%) as a carbon source [50]. As shown in Fig. 1, in K. pneumoniae CG43, a K2 serotype strain, we found that higher CPS production in glucose-rich medium is dependent on reducing the intracellular cAMP concentration. In addition, the activation of CPS in response to exogenous glucose was also found in K. pneumoniae NTUH-K2044, a highly virulent liver abscess isolate of K1 serotype, and addition of increasing amounts of exogenous cAMP in glucose-rich medium could completely reverse the effect of glucose on CPS production (Fig. S1). Therefore, we suggest that the regulatory role of glucose and cAMP-dependent CCR in CPS biosynthesis is conserved in K. pneumoniae stains of K1 and K2 serotype. Besides, the addition of cAMP did not completely reverse the glucose-activated CPS biosynthesis in K. pneumoniae CG43 (Fig. 1), suggesting that glucose could activate CPS biosynthesis through mechanism(s) other than repression of the cAMP-CRP regulation in K. pneumoniae CG43, which awaits further investigations.
In bacteria, CyaA and CpdA are responsible for cAMP production and degradation [22,23]. In this study, we also found that the deletion of cyaA abolished the ability of cAMP to increase CPS biosynthesis, while the deletion of cpdA elevated the cAMP level and decreased CPS biosynthesis (Fig. 2). Although the Dcrp strain has a slightly higher cAMP level than WT, CPS production is obviously high in the Dcrp strain. This result indicated that cAMP-mediated repression of CPS biosynthesis is required for CRP regulatory activity. In E. coli, CRP can repress cyaA transcription to down-regulate cAMP production [51]. In K. pneumoniae, a typical CRP binding box (59-TGTTA-AATTGA-TCACG-39) was located at 2144 to 2129 relative to the translation start site of cyaA. In addition, the mRNA level of cyaA increased more than 16-fold in the Dcrp stain, unlike that in the WT strain (data not shown), indicating that the deletion of crp could increase cyaA expression to elevate the cAMP level in K. pneumoniae, consistent with the findings for E. coli.
To further identify the regulatory mechanism of cps transcription, the transcriptional start sites of 3 transcriptional units in the cps gene cluster were first determined (Fig. 4). CRP binds directly to the predictive CRP binding sites and represses the transcription of wzi and manC ( Fig. 3 and 5), while indirectly repressing the transcription of galF via RcsA ( Fig. 3 and 7B). According to the position of the CRP binding site relative to RNA polymerase, the CRP-activated promoter is divided into the 3 classes in E. coli [27,52,53]. However, the CRP-repressed promoter displays a greater range of binding site positions [54]. In K. pneumoniae, we found CRP could directly bind to the CRP binding site of wzi-2 and rcsA-1 centred at and near the -10 and 235 boxes, respectively, suggesting that CRP may interfere with RNA polymerase regulation of gene transcription (Fig. 4D and 6A). However, K. pneumoniae CRP could also repress transcription by binding directly to the predicated CRP binding site of manC, which was located at a relatively long distance upstream of the 210 and 235 boxes (2140 to 2125) ( Fig. 3 and 4D). Whether other additional transcriptional factors are required for CRP repression of manC transcription needs further investigation.
Multiple regulators including RcsA/B, RmpA/A2, KvhA, KvgA, and Fur have been demonstrated to control the transcription of K. pneumoniae CPS biosynthesis genes [43,44,45,46]. In this study, we also found that cps transcription is regulated by cAMPdependent CCR. In E. coli, CRP has been reported to regulate the expression of fur at the transcriptional as well as at the posttranscriptional level [55,56,57]. In addition, functional interaction of CRP and Fur has been demonstrated to coordinate the transcriptional regulation of iron and carbon metabolism [58]. In K. pneumoniae, CPS biosynthesis was modulated by iron availability and Fur has been shown to repress the transcription of rmpA/A2 and rcsA to control CPS biosynthesis [46]. As shown in this study, the expression of rcsA is also regulated by glucose and cAMPdependent CCR (Fig. 6). In addition, RcsA is involved in the CRP regulon in regulating galF expression and CPS biosynthesis (Fig. 7). By analysing the promoter sequence of galF, a typical RcsAB binding box [59] (59-TAAGATTATTCTCA-39) was found to be located at position 2119 to 2107 relative to the transcriptional start site of galF, indicating that RcsA could directly activate galF transcription. In addition, we noted that the deletion of rcsA in the DlacZDcrp strain retains a higher galF promoter activity compared to the DlacZ strain, implying that CRP could directly or indirectly regulate galF expression. Although no obvious CRP binding site was found in the promoter sequence of galF, an EMSA was performed to investigate whether CRP could directly bind to P galF . As shown in Fig. S2, DNA-protein binding complexes were observed after the incubation of 150 nM purified His 6 -CRP with 10 ng P galF , implying that CRP-cAMP could directly repress the galF transcription. However, the exact binding site of CRP in P galF needs to be further investigated. In addition, we also found that introduction of a plasmid carrying His6-CRP encoding gene into Dcrp strain could complement the effect of crp mutation on CPS biosynthesis (Fig. S3). The result confirmed that His6-CRP is functional in vivo. Taken together, these findings revealed a complex regulatory circuit in these CPS regulators, which then modulate the transcription of cps genes in coordination, in response to various environmental stimuli.
In K. pneumoniae, CPS is considered to be an important virulence factor that protects the bacteria from serum killing and phagocytosis [6,7]. In this study, cAMP-dependent CCR was demonstrated to protect K. pneumoniae against serum killing, and the results suggest it plays a role in the regulation of CPS production. In addition, E. coli strains lacking cAMP-CRP are highly resistant to reactive oxygen species (ROS) containing hydrogen peroxide (H 2 O 2 ) and hypochlorous acid (HOCl) [60]. Large amounts of ROS are generated by phagolysosomes to inhibit bacterial colonization and survival [61]. Therefore, we suggest that cAMP-dependent CCR in K. pneumoniae not only regulates CPS production to protect the bacteria from serum killing and phagocytosis, but also alters bacterial resistance to oxidative stress to enhance the survival rate in the phagosome, and we are currently working to demonstrate this possibility. In addition, CPS and adherence factors, such as type 1 and type 3 fimbriae, have been demonstrated to play important roles in biofilm formation and pathogenesis, and their expression could be co-regulated [62]. Biofilms are surface-attached bacteria embedded in a selfproduced matrix, composed mainly of polysaccharide, but also containing proteins and nucleic acids [63]. Biofilm formation promotes encrustation and protects the bacteria from the hydrodynamic forces of urine flow, host defences and antibiotics [64]. In E. coli and Serratia marcescens, glucose/CRP-cAMP has been described to regulate the expression of type 1 fimbriae and bacterial biofilm formation [36,37]; however, this regulation has not been proven in K. pneumoniae. Since glucose has also been described to repress K. pneumoniae biofilm formation [65], in addition to CPS, it is possible that glucose/CRP-cAMP is able to regulate the expression of adherence factors, which we are currently investigating.
In this study, we provide important evidence that glucose stimulates CPS biosynthesis in K. pneumoniae to protect the bacteria from serum killing, and cAMP-dependent CCR plays a profound regulatory role in CPS expression in response to glucose levels in the environment. In diabetes mellitus patients, the higher glucose level in the bloodstream is thought to have a major impact on bacterial virulence. We suggest that K. pneumoniae could evade the immune response via the regulation of cAMP-dependent CCR on CPS biosynthesis, especially during infection of diabetes mellitus patients. Future studies will include determining the role of cAMPdependent CCR in modulating CPS biosynthesis, fimbria production, and biofilm formation in response to environmental stimuli.

Bacterial Strains, Plasmids, and Media
Bacterial strains and plasmids used in this study are listed in Table 1. Primers used in this study are list in Table 2. Bacterial were routinely cultured at 37uC in Luria-Bertani (LB) medium supplemented with appropriate antibiotics. The antibiotics used include ampicillin (100 mg/ml), kanamycin (25 mg/ml), streptomycin (500 mg/ml), and tetracycline (12.5 mg/ml).

Detection of cAMP
Bacteria was adjusted to 1610 7 colony forming units (c.f.u.)/ml and washed twice in phosphate-buffered saline (PBS). Then, the bacteria were resuspended in 300 ml of 1X lysis buffer and lysated by sonication. The lysate was centrifuged briefly at 14,000 rpm for 10 min, and the supernatant was tested for cAMP levels by using cAMP XP TM Assay Kit (Cell Signaling Technology, Inc.) and according to manufacturer's recommendations.

Construction of the Gene-deletion Mutants and Complementation Plasmids
Specific gene deletion containing crp, cyaA, and cpdA was introduced into K. pneumoniae CG43S3 using an allelic exchange strategy as previously described respectively [45]. In brief, two approximately 1000 bp DNA fragments flanking both sides of the deleted region were cloned into the suicide vector pKAS46 [66], a suicide vector containing rpsL, which allows positive selection with streptomycin for vector loss. The resulting plasmid was then mobilized from E. coli S17-1lpir [67] to K. pneumoniae CG43S3, K. pneumoniae CG43S3DlacZ, K. pneumoniae CG43S3DrcsA or CG43S3derived strains, by conjugation. The transconjugants, with the plasmid integrated into the chromosome via homologous recombination, were selected with ampicillin and kanamycin on M9 agar plates. Several of the colonies were grown in LB broth supplemented with 500 mg/mL of streptomycin to log phase at 37uC and then spread onto an LB agar plate containing 500 mg/ mL of streptomycin. The streptomycin-resistant and kanamycinsensitive colonies were selected, and the deletion was verified by PCR and Southern hybridization (data not shown). The resulting K. pneumoniae mutants are listed in Table 1.
To obtain the complementation plasmids, DNA fragments containing the promoter and coding sequence of crp, cyaA, and rcsA were individually PCR-amplified with primer pairs GT131/ GT132, GT196/GT197, and HY001/GT145 (Table 2) and cloned into the shuttle vector pACYC184 or pRK415 to generate pcrp, pcyaA, and prcsA, respectively. To generate the cpdA complement plasmid, pETQ-cpdA, DNA fragment containing the coding sequence of cpdA was individually PCR-amplified with primer pair GT210/211 (Table 2) and cloned into the expression vector pETQ. To generate the His-crp complement plasmid, pETQ-His-crp, DNA fragment containing N-terminal His-tag fused with the coding sequence of crp was individually PCRamplified with primer pair GT132/227 (Table 2) from pET30b-CRP and cloned into the expression vector pETQ.

Extraction and Quantification of CPS
CPS was extracted and quantified as previously described [68]. The glucuronic acid content, represents the amount of K. pneumoniae K2 CPS, was determined from a standard curve of glucuronic acid (Sigma-Aldrich) and expressed as micrograms per 10 9 c.f.u. [69].

qRT-PCR
Total RNAs were isolated from early-exponential-phase grown bacteria cells by use of the RNeasy midi-column (QIAGEN) according to the manufacturer's instructions. RNA was DNasetreated with RNase-free DNase I (MoBioPlus) to eliminate DNA contamination. RNA of 100 ng was reverse-transcribed with the Transcriptor First Strand cDNA Synthesis Kit (Roche) using random primers. qRT-PCR was performed in a Roche Light-CyclerH 1.5 Instrument using LightCycler TaqMan Master (Roche). Primers and probes were designed for selected target sequences using Universal ProbeLibrary Assay Design Center (Roche-applied science) and listed in Table 2. Data were analyzed

Measurement of Promoter Activity
The promoter-reporter plasmids, pRcsAZ15, pOrf12, pOrf315, and pOrf1617, were individually mobilized into K. pneumoniae strains by conjugation from E. coli S17-1 lpir. The bacteria were grown to logarithmic phase in LB broth, and the b-galactosidase activity was measured as previously described [43].

Identification of the Transcriptional Start Sites of Three Transcriptional Units in the K2 cps Gene Cluster
For the determination of 59 mRNA ends in the three transcriptional units in the K2 cps gene cluster, a rapid amplification of PCR was performed using 59 RACE kit (Clontech) according to the manufacturer's instruction as previously described [70]. A total of ten clones each of galF, wzi, and manC, respectively, were subjected to sequence analysis, and the transcriptional start sites of galF, wzi, and manC were determined. All the sequencing results indicated the same nucleotide as the transcriptional start site of galF, wzi, and manC.

Purification of His 6 -CRP Protein
The coding region of crp was PCR amplified with primer sets GT132/GT137 (Table 2) and cloned into the BamHI/HindIII site in pET30b (Novagen, 205 Madison, Wis). The resulting plasmid pET30b-CRP was then transformed into E. coli BL21(DE3) (New England Biolabs), and overproduction of the recombinant protein was induced by the addition of 0.1 mM IPTG for 4 h at 37uC. The recombinant proteins were then purified from the soluble fraction of the total cell lysate by affinity chromatography using His-Bind resin (Novagen, Madison, Wis). Finally, the purified proteins were dialyzed against 1X TE buffer (20 mM Tris-HCl, pH 8.0, 1 mM EDTA) containing 10% glycerol at 4uC overnight, and the purity was determined by SDS-PAGE.

EMSA
The DNA fragments of the putative promoter region of orf1-2, orf3-15, orf16-17, and rcsA were respectively PCR amplified by using specific primer sets ( Table 2). The purified His 6 -CRP was incubated with 10 ng DNA in a 15 ml solution containing 4 mM Tris-HCl (pH 7.4), 10 mM KCl, 100 mM dithiothreitol, 200 mM cAMP, and 10 mg/ml BSA at room temperature for 30 min. The samples were then loaded onto a native gel of 5% nondenaturing polyacrylamide containing 200 mM cAMP in 0.5X TB buffer (45 mM Tris-HCl, pH 8.0, 45 mM boric acid). Gels were electrophoresed with a 20-mA current at 4uC and then stained with SYBR Green I dye (Invitrogen).

Bacterial Survival in Serum
Normal human serum, pooled from healthy volunteers, was divided into equal volumes and stored at 270uC before use. Bacterial survival in serum was determined with minor modifications [45]. In brief, bacteria were grown in LB broth, and when growth reached mid-exponential phase, the bacteria were collected, washed twice with phosphate-buffered saline (PBS), and then adjusted to approximately 1610 6 c.f.u./ml. The reaction mixture containing 250 ml of the cell suspension and 750 ml of pooled human serum was incubated at 37uC for 15 min. The number of viable bacteria was then determined by plate counting. The survival rate was expressed as the number of viable bacteria treated with human serum compared to the number of pretreatment. The assay was performed triple, each with triplicate samples. The data from one of the representative experiments are shown and expressed as the mean and standard deviation from the three samples. A CPS-deficient mutant strain, K. pneumoniae CG43S3DgalU (galU encodes for an UDP-glucose pyrophosphorylase that is responsible for supplying UDP-glucose as material for CPS biosynthesis) is served as a control.

Statistical Method
An unpaired t-test was used to determine the statistical significance and values of P,0.05 and P,0.01 were considered significant. The results of CPS quantification, qRT-PCR analysis, and promoter activity measurement were derived from a single experiment representative of three independent experiments. Each sample was assayed in triplicate and the mean activity and standard deviation are presented.

Ethics Statement
For isolation of normal human serum from healthy volunteers, the procedure and the respective consent documents were approved by the Ethics Committee of the China Medical University Hospital, Taichung, Taiwan. All healthy volunteers provided written informed consent. Figure S1 Glucose and cAMP affects the CPS levels of K. pneumoniae NTUH-K2044. CPS levels of K. pneumoniae NTUH-K2044 were activated by increasing environmental glucose. Bacterial strains were grown in LB broth supplemented with glucose and cAMP as indicated at 37uC with agitation. After 16 h of growth, the bacterial glucuronic acid content was determined. *P,0.05 and **P,0.01 compared with no addition. #P,0.05 and $ P,0.01 compared to the indicated group. (TIF) Figure S2 CRP directly binds to P galF . Diagrammatic representation of the galF loci. The large arrows represent the open reading frames. The relative positions of the primer set used in PCR-amplification of the DNA probes are indicated, and the numbers denote the positions relative to the translational start site. Name of the DNA probes are shown on the left. Different concentrations of purified His 6 -CRP were incubated with 10 ng of the upstream regions of galF. Following incubation at room temperature for 30 min, the mixtures were analyzed on a 5% nondenaturing polyacrylamide gel containing 200 mM cAMP. The gel was stained with SYBR Green I dye and photographed.

Supporting Information
(TIF) Figure S3 Induced expression of His 6 -CRP complements the effect of crp mutation on CPS biosynthesis. CPS levels of K. pneumoniae strains carrying the IPTG inducible vector pETQ or pETQ-His-crp, as shown in the left panel, were determined. Bacteria were grown in LB medium with 100 mM IPTG at 37uC with agitation. **P,0.01 compared to the indicated group. (TIF)