RcsB regulation of the YfdX-mediated acid stress response in Klebsiella pneumoniae CG43S3

In Klebsiella pneumoniae CG43S3, deletion of the response regulator gene rcsB reduced the capsular polysaccharide amount and survival on exposure to acid stress. A comparison of the pH 4.4-induced proteomes between CG43S3 and CG43S3ΔrcsB revealed numerous differentially expressed proteins and one of them, YfdX, which has recently been reported as a periplasmic protein, was absent in CG43S3ΔrcsB. Acid survival analysis was then conducted to determine its role in the acid stress response. Deletion of yfdX increased the sensitivity of K. pneumoniae CG43S3 to a pH of 2.5, and transforming the mutant with a plasmid carrying yfdX restored the acid resistance (AR) levels. In addition, the effect of yfdX deletion was cross-complemented by the expression of the periplasmic chaperone HdeA. Furthermore, the purified recombinant protein YfdX reduced the acid-induced protein aggregation, suggesting that YfdX as well as HdeA functions as a chaperone. The following promoter activity measurement revealed that rcsB deletion reduced the expression of yfdX after the bacteria were subjected to pH 4.4 adaptation. Western blot analysis also revealed that YfdX production was inhibited by rcsB deletion and only the plasmid expressing RcsB or the nonphosphorylated form of RcsB, RcsBD56A, could restore the YfdX production, and the RcsB-mediated complementation was no longer observed when the sensor kinase RcsD gene was deleted. In conclusion, this is the first study demonstrating that YfdX may be involved in the acid stress response as a periplasmic chaperone and that RcsB positively regulates the acid stress response partly through activation of yfdX expression. Moreover, the phosphorylation status of RcsB may affect the YfdX expression under acidic conditions.


Introduction
The nosocomial pathogen Klebsiella pneumoniae causes suppurative lesions, septicemia, and infections of the urinary and respiratory tracts in immunocompromised patients [1,2]. In PLOS  been identified in Shigella flexneri and Brucella abortus, but not in Salmonella typhimurium and Vibrio cholera [21]. Although few studies have focused on HdeD, it has been shown to play a role in high-cell-density AR [17]. Different from E. coli, a sequence search for the genomes of the K. pneumoniae clinical isolates MGH78578 [22], NTUH-K2044 [23], and CG43 (NCBI Taxonomy ID: 1244085) revealed no AR2 and AR3 encoding genes. Nevertheless, hdeB and hdeD were clustered with the twocomponent system (TCS) coding genes kvhAS [24][25][26] and hdeB1 (Fig 1). Notably, although not identified in the genome of MGH78578 and NTUH-K2044, hdeA was found next to kvgAS [24] and a distantly located homolog of hdeDB (designated as hdeD1B2) in the CG43 genome (Fig 1). K. pneumoniae CG43 and NTUH-K2044 are heavily encapsulated liver abscess isolates and have recently been classified as hypermucoviscosity strains [23,27]. Given that HdeA may confer CG43 a higher AR activity than NTUH-K2044, we compared the acid stress survivals between CG43S3 and NTUH-K2044S3. However, S1 Fig revealed that NTUH-K2044S3 exhibited a stronger AR than CG43S3 at either the exponential or stationary phase. The TCS KvgAS and KvhAS, homologs of E. coli EvgAS, have previously been reported to be involved in the regulation of the virulence, drug resistance, stress response and capsular polysaccharide biosynthesis [24][25][26]. A possibility of KvhAS regulation on the expression of yfdX and hdeDB has also been speculated [25].
Bacteria are equipped with a complex regulatory pathway to ensure an appropriate and rapid response to environmental acid stress. In E. coli, the TCS EvgAS is a major determinant in the regulation of the acid stress response and multidrug resistance [28][29][30]. The RcsBCD TCS also plays a role in the acid stress response through the regulation of AR2 expression. In the absence of GadE, RcsB represses the expression of gadABC, whereas it activates the expression of the Gad operon in the presence of GadE by forming an RcsB-GadE heterodimer [31][32][33][34]. In addition, the deletion of rcsB in E. coli O157:H7 impairs the expression of HdeA, which leads to increased acid sensitivity [35,36].
The Rcs system is composed of three core proteins, innermembrane sensor kinase RcsC, phosphotransferase RcsD, and response regulator RcsB. In addition, the outermembrane lipoprotein RcsF is an auxiliary protein for receiving extracellular stimulations. The signal transduction of Rcs phosphorelay begins with the autophosphorylation of RcsC upon receiving specific stimuli, and then RcsD transfers the phosphoryl group from RcsC to the cytoplasmic regulator RcsB [37]. The phosphoryl group is transferred to the conserved aspartate residue (D56) of RcsB, and the phosphorylation status of RcsB determines its regulatory property [38,39].
In K. pneumoniae, apart from playing a role in regulating capsule production [40,41], RcsB has also been reported to play a role in resistance to polymyxin B [42,43]. In this study, we investigated the involvement of RcsB in regulating the acid stress response in K. pneumoniae CG43S3 by analyzing the deletion effect of the rcs system on the bacterial resistance to acid stress and used a comparative proteomic approach to identify the downstream genes regulated by RcsB. Table 1 lists the bacterial strains and plasmids and Table 2 lists the primers used in this study. Bacteria were grown in Luria-Bertani [44] broth at 37˚C. The antibiotics used included ampicillin (100 μg/ml), kanamycin (25 μg/ml), streptomycin (500 μg/ml), and chloramphenicol (35 μg/ml).

Construction of the gene deletion mutants and the gene complement strains
Specific gene deletion was introduced to the chromosome of K. pneumoniae CG43S3 by using an allelic-exchange strategy essentially as described [40]. In brief, the DNA fragments of 1 kb  flanking both ends of rcsB, rcsC, rcsD, rcsF, yfdX, hdeDB, hdeB1, hdeA and hdeD1B2 gene were amplified using PCR with the primer sets in Table 2. The two amplified DNA fragments were cloned into the suicide vector pKAS46 [45]. The resulting plasmid was transformed into E. coli S17-1λpir and then mobilized by conjugation to the streptomycin-resistant strain, K. pneumoniae CG43S3. Several kanamycin-resistant transconjugants, with the plasmid integrated into the chromosome through homologous recombination, were selected from M9 agar plates supplemented with kanamycin and propagated in 2 ml of LB broth overnight. A small aliquot of the culture was then plated on LB agar containing 500 μg/ml of streptomycin. Lastly, the streptomycin-resistant and kanamycin-sensitive colonies were isolated, and the specific gene deletion of rcsB, rcsC, rcsD, rcsF, yfdX, hdeDB, hdeB1, hdeA and hdeD1B2 were verified with PCR analysis. For complementation analysis, the DNA region containing rcsB, yfdX, hdeB, hdeD, hdeDB, and hdeA were amplified using PCR with respective primer pairs in Table 2, and the DNA fragments cloned into pRK415, and transferred to the specific gene deletion mutant by conjugation.

Site-directed mutagenesis
The site-directed mutation plasmids pRK415-hdeA-F44A, pRK415-rcsB-D56A and pRK415-rcsB-D56E were generated using PCR-based mutagenesis with the plasmid pyT&A-hdeA and pyT&A-rcsB as template to substitute the phenylalanine at residue 44 of HdeA with alanine and aspartic acid at residue 56 of RcsB with glutamic acid or alanine. pyT&A-rcsB and pyT&A-hdeA were amplified with the point mutation primer sets rcsB- encompassing the mutation site by using PfuUltra II Fusion HS DNA polymerase (Agilent Technologies) to generate mutant alleles of rcsB and hdeA, respectively. The PCR product was resolved on an agarose gel, recovered, treated with DpnI to remove the template plasmid and transformed into E. coli JM109. The point mutation allele of the recombinant plasmid was later confirmed by DNA sequencing. The mutated fragment rcsB-D56E, rcsB-D56A and hdeA-F44A were subcloned into plasmid pRK415 to yield pRK415-rcsB-D56E, pRK415-rcsB-D56A and pRK415-hdeA-F44A, respectively. The site-directed mutation plasmids were then individually mobilized from E. coli S17-1 λpir to the K. pneumoniae CG43S3 strain by conjugation.

Acid stress survival assessment
Overnight-grown bacteria diluted 1:20 in LB broth were incubated at 37˚C to OD 600 of 0.6~0.7 (exponential phase) or 1.0~1.1 (stationary phase). An aliquot of the bacteria was collected by centrifugation, resuspended in the acidic LB broth (adjusted to pH 4.4 by HCl) and subjected to 37˚C incubation under shaking cultured condition (130 rpm) for 1 h adaptation before the acid challenge. After adaptation, the bacteria (approximately 5x10 8~1 x10 9 CFU/ml) were transferred to the acidic M9 medium (adjusted to pH 2.5 by HCl) and incubated at 37˚C for 30 min under shaking cultured condition (130 rpm). After the acid stress treatment, the bacteria were diluted serially to 10 −6 and 10 μl of each sample was spotted onto LB agar plate and incubated at 37˚C overnight. The presented results are representative of at least three independent experiments. The survival was calculated by dividing the number of colonies after acid treatment by that before the treatment (after pH4.4 acid adaptation). Each sample was assayed in triplicate, and at least three independent experiments were conducted. The data were calculated from three independent experiments and are shown as the mean and standard deviation from that samples. Student's t-test was used to determine differences between groups and values of P<0.05 and P<0.01 were considered statistically significant difference.

2D-PAGE analysis
Overnight-grown bacteria diluted 1:20 in LB broth were incubated at 37˚C to OD600 of 0.6~0.7. Since cells died profoundly upon pH 2.5 treatment, the bacteria were transferred to the acidic LB broth (pH 4.4) and the incubation continued for 1 h before protein collection. Bacteria were finally collected by centrifugation at 3000 rpm for 30 min, washed three times with wash buffer (10 mM Tris-HCl pH 7.5, 250 mM sucrose), resuspended in 3 ml lysis buffer (10 mM Tris-HCl pH 7.5). The bacteria were lysed by sonication followed by centrifugation at 3000 rpm for 40 min. The supernatants were treated by DNase (20 units) and RNase (20 units) at 37˚C for 45 min, followed by centrifugation at 15,000 rpm for 30 min to remove the insoluble portions. Finally, the supernatants were passed through the 10 kDa microcon (Millipore), and the filtrates were freeze-dried and stored at -80˚C before use. Aliquot of sample (250 μg) was dissolved in 250 μl rehydration buffer (2 M thiourea, 7 M urea, 2% 3-[(3-Cholamidopropyl) dimethylammonio]-1-propanesulfonate (CHAPS), 1% immobilized pH gradient (IPG) buffer, 0.002% bromophenol blue, 0.28% Dithiothreitol (DTT)) followed by centrifugation at 15,000 rpm for 20 min. Each sample was added into holder, and then the strip (pH 4-7, 13 cm) was placed and an appropriate amount of dedicated mineral oil was added. The holder was inserted in IPGphor (GE Healthcare) to execute isoelectric focusing. After finishing isoelectric focusing, the strip was soaked into the equilibration buffer (50 mM Tris-HCl pH 8.8, 6 M urea, 30% glycerol, 2% SDS, 0.002% bromophenol blue) containing 1% DTT for 15 min and then soaked into the equilibration buffer containing 2.5% idoacetamide (IAA) for 15 min. After the pretreatment of strip, 12.5% polyacrylamide gel was used to run 2D-PAGE. Finally, Sypro Ruby (Invitrogen) and Typhoon 9200 (GE Healthcare) were respectively used to stain and scan the gel, and Image Master 2D platinum 6.0 (GE Healthcare) was used to detect and quantify the protein spots. Student's t-test was used to determine relative volume of the protein spots between CG43S3 and CG43S3ΔrcsB. The protein spots with more than 1.5 fold differences of the relative volume between the two proteomes were isolated and subjected to mass spectrometry analysis.

Mass spectrometry analysis
The gel digestion method was modified from the previous research [47]. Protein spots were excised from the gel and washed using wash buffer (50% acetonitrile, 25 mM NH 4 HCO 3 ) for 15 min. After removing the wash buffer, the gel pieces were shrunk by dehydration in 100% acetonitrile, swelled by rehydration in NH 4 HCO 3 (100 mM) for 5 min, and shrunk again by addition of 100% acetonitrile. The liquid phase was removed, and the gel pieces were completely dried at room temperature. Then, 2 μl of the digestion buffer (25 mM NH 4 HCO 3 , 20 ng/μl trypsin (Promega) was added and the gel pieces were placed at 4˚C for 1 h. Then, a sufficient volume of NH 4 HCO 3 (25 mM) was added to keep the gel pieces wet, and the gel pieces were placed at 37˚C overnight for enzymatic cleavage. Peptides were extracted by addition of 2 μl of the buffer containing 100% acetonitrile and 1% trifluoroacetic acid and sonication of the gel pieces for 10 min. The extraction step was repeated three times. Then, the collected samples were subjected to MALDI-TOF analysis by Academia Sinica Proteomic mass spectrometry common facility (http://www.ibc.sinica.edu.tw/Facility_Mass_E.asp).

Chaperone assay
As described in ref [20,48], the chaperone assay used alcohol dehydrogenase (ADH) as the target substrate for chaperone protein. Initially, ADH (10 mM) was mixed with different concentrations of the purified recombinant YfdX protein. Then, each of the mixtures was treated with acidic distilled water (adjusted to pH 1.0 with HCl) for 15 min, followed by centrifugation at 15,000 rpm for 10 min to remove the insoluble portions. Finally, the supernatants were analyzed by SDS-PAGE.

Measurement of promoter activity
The putative promoter region of kvhAS, yfdX, hdeDB, hdeB1, hdeA, and hdeD1B2 were PCRamplified using primers sets in Table 2. The amplicons were then cloned into placZ15 [25] to generate P yfdX -lacZ, P hdeDB -lacZ, P hdeB1 -lacZ, P kvhAS -lacZ, P hdeA -lacZ, and P hdeD1B2 -lacZ. The promoter-reporter plasmids were individually mobilized into K. pneumoniae CG43S3ΔlacZ strains through conjugation from E. coli S17-1 λpir. The β-galactosidase activity was measured for the bacteria grown to exponential phase with OD 600 of 0.6~0.7 [26]. The promoter activity was expressed as Miller units. Each sample was assayed in triplicate, and at least three independent experiments were conducted. The data were calculated from three independent experiments and are shown as the mean and standard deviation from that samples. Student's t-test was used to determine differences between groups and values of P<0.05 and P<0.01 were considered statistically significant difference.

RcsB expression plasmid construction
The coding regions of rcsB was PCR-amplified with the primer pairs listed in Table 2. The amplified DNA was cloned into cloning vector yT&A (Yeastern Biotech Co., Ltd.), and the resulting recombinant plasmids named pyT&A-rcsB. For protein expression and purification, the coding regions from pyT&A-rcsB was subcloned into pET30a (Novagen) and resulting the plasmids pET30a-rcsB.

Expression and purification of the recombinant proteins
The plasmids pET30a-rcsB was individually transformed into E. coli NovablueDE3, and protein production was induced with 0.5 mM isopropyl-β-D-thiogalactopyranoside (IPTG) for 5 h at 37˚C. The overexpressed protein was then purified from the soluble fraction of the cell lysate by affinity chromatography using His-Bind resin essentially according to the QIAexpress expression system protocol (Qiagen). The purified RcsB protein was dialyzed against Trisbuffered saline (pH 7.4) containing 10% glycerol at 4˚C overnight, followed by condensation with PEG 20000. The protein purity was determined using SDS-PAGE.

YfdX antisera preparation
The yfdX coding sequence was amplified using PCR from K. pneumoniae CG43S3 genome, and ligated into expression vector pET30a (Novagen). The plasmid pET30a-yfdX was transformed into E. coli JM109, and the gene expression of the recombinant protein His 6 -YfdX was induced with 0.5 mM IPTG for 5 h at 37˚C. The soluble His 6 -YfdX protein was purified using a nickel column (Novagen, Madison, WI, USA). Then Then the polyclonal YfdX antisera was prepared by LTK BioLaboratories (Taoyuan, Taiwan, ROC). The procedure was as follows: 1 mg purified protein emulsified with 500 μl complete Freund's adjuvant was used to immunize New Zealand white rabbits weighing 2.0~2.5 kg by intramuscular injection. The rabbits were boosted three times at 2 week intervals with 500 μg purified YfdX recombinant protein. The YfdX antisera was obtained by intracardiac puncture 8 weeks later.

Western blot analysis
Aliquots of the total cellular lysates were resolved through sodium dodecyl sulfate polyacrylamide gel electrophoresis, and the proteins were electrophoretically transferred onto a polyvinylidene difluoride (PVDF) membrane (Millipore, Billerica, MA, USA). After incubation with 5% skim milk at room temperature for 1 h, the membranes were washed 3 times using phosphate buffered saline with 0.1% Tween 20 (PBST), and were then incubated with an anti-GAPDH (1:5000 dilution) (GeneTex, GTX100118) or anti-YfdX (1:10000 dilution) antiserum at room temperature for 2 h. Again, the membranes were washed 3 times with 1X PBST, and subjected to incubation with a 1:5000 dilution of the secondary antibody, alkaline phosphatase-conjugated anti-rabbit immunoglobulin G (Millipore, AP132A), at room temperature for 1 h. Finally, the blots were rewashed, and the secondary antibodies bound on the PVDF membrane were detected using chromogenic reagents 5-bromo-4-chloro-3-indolyl phosphate (BCIP) and nitro blue tetrazolium (NBT). Data quantification was done using ImageJ 1.46r (http://imagej.nih.gov/ij/).

Deletion of rcsB or rcsD reduced acid survival
To analyze the deletion effect of rcsB on bacterial AR and whether other Rcs components were also involved in the RcsB-dependent acid stress response, we constructed rcsB, rcsC, rcsD, and rcsF deletion mutant strains and subjected them to acid treatment modified from previous studies [49,50]. The bacterial strains were cultured at a pH of 4.4 for 1 h for acid adaptation and then at a pH of 2.5 for 30 min for acid stress treatment. The survival of bacterial strains that were subjected to the pH 4.4 treatment were first analyzed to reveal that the growth of each bacterial strains was at the comparable levels. As shown on the left panel of S2 Fig, the acid adaptation (pH 4.4 for 1h) had no apparent influence on the survival of each bacterial strain. The acid survival was quantitatively and qualitatively determined to assess the deletion effects. As illustrated in Fig 2A and S2A Fig, after treatment at pH 2.5, the acid stress resistance was reduced when rcsB or rcsD was removed. By contrast, the survival of the mutants ΔrcsC and ΔrcsF was similar to that of the parental strain.

AR may be affected by the phosphorylated form of RcsB
During bacterial signal transduction, RcsD kinase receives the phosphate group from RcsC and subsequently relays it to RcsB. Because the removal of rcsD as well as rcsB deletion reduced the AR, we assumed that the relay of the phosphoryl group was blocked, thereby affecting the phosphorylated form of RcsB. To investigate whether the phosphorylation status of RcsB influences the acid stress resistance, site-directed mutants, RcsB D56A and RcsB D56E , which mimic the nonphosphorylated and phosphorylated forms of RcsB, respectively, were created. As illustrated in Fig 2B and S2B Fig, the deletion effect was complemented by introducing an RcsBexpression plasmid pRK415-rcsB into the mutant, indicating an involvement of RcsB in the regulation of the acid stress response. In addition, the acid stress sensitivity of ΔrcsB was also significantly reduced by increasing the expression of RcsB D56A . By contrast, the deficiency of ΔrcsB could not be rescued through the expression of RcsB D56E . These findings suggest that RcsB in the nonphosphorylated form activates the AR response. Notably, we also could not rule out the possibility that the point mutation influenced the protein stability or conformation of RcsB D56E and resulted in the loss of the regulation function. More experiments are warranted to support this finding.

Deletion of rcsB blocked YfdX production
Either pH 4.4 or pH 5 treatment had no apparent effect on the transcriptional levels of rcsB, indicating that the rcsB expression was not acid induced (S3 Fig). Moreover, no AR2 system or conserved AFI was identified in the K. pneumoniae genome. To elucidate the RcsB-mediated regulation of the response to acid stress, a comparative proteome analysis of CG43S3 and CG43S3ΔrcsB was performed. As depicted in Fig 3A, after incubation at pH 4.4 for 1 h for acid adaptation, the spots marked 772, 817, 832, 879, 946, 972, 973, and 1064 exhibited differences between the proteomes of the mutant strain CG43S3ΔrcsB and the parental strain CG43S3. As shown in S1 Table, rcsB deletion led to 2.18-, 1.90-, and 1.52-fold decreased expression of spots 772, 972, and 973, respectively. Spots 817 and 832 were present only in CG43S3. Spot 832, which exhibited obviously different, was isolated, analyzed through mass spectrometry, and identified as YfdX. Unlike the gene in E. coli, yfdX is located next to hdeDB in K. pneumoniae CG43 (Fig 1). Sequence comparison (using Vector NTI 10.3) revealed that the YfdX of K. pneumoniae had a similarity of 56% and 37% to E. coli MG1655 and S. Typhi CT18, respectively, and a signal peptide could be predicted for all three YfdX sequences (Fig 3B). The YfdX of S. Typhi CT18, STY3178, may play a role in multidrug resistance, and the F67, Y109, Y186, and Y187 residues for STY3178's interaction with antibiotics [51] are conserved in K. pneumoniae YfdX (Fig 3B).

Deletion of yfdX, hdeD, and hdeB1 reduced acid survival
As shown in Fig 1, the clustered location of yfdX with the chaperone encoding gene hdeDB and the evgAS orthologous gene kvhAS suggested a similar functional role. To determine the involvement of YfdX and Hde proteins in the acid stress response, the specific gene-deletion  mutants were constructed, and the effects of deletion on bacterial susceptibility to acid stress were analyzed. In E. coli, the expression of hdeA is repressed by a histone-like nucleoid structuring protein (H-NS) and is activated when the bacteria enter the stationary phase [34,52,53]. Therefore, the deletion effects of bacteria grown at the exponential and stationary phases were measured. Each gene deletion was confirmed, showing no apparent effect on the bacterial growth before pH 2.5 treatment (the left panel of S4 Fig). As shown in Fig 4 and S4 Fig, the  deletion of yfdX, hdeD, and hdeB1 reduced the survival after acid treatment, whereas the deletion of hdeA, hdeB, hdeB2, or hdeD1B2 had no apparent effect on the acid survival at the exponential phase. However, no significant difference in acid survival was observed among the mutant strains at the stationary phase.

Overexpression of HdeA enhanced the AR
In E. coli, the periplasmic chaperone activity of HdeA, HdeB, and HdeD has been demonstrated [17,18,20,54,55]. A recent report demonstrated that Salmonella YfdX also exhibited a periplasmic chaperone activity [56]. To investigate whether the yfdX deletion effect could be complemented by the chaperone activity, the plasmid pRK415 carrying hdeA, hdeB, hdeD, or hdeDB was individually used to transform the ΔyfdX mutant, and then the acid survivals were determined. As illustrated in Fig 5A and S5A Fig, introducing [57]. To further confirm that the chaperone activity of K. pneumoniae HdeA may compensate for the yfdX deletion effect, the critical residue phenylalanine 44 of HdeA was selected for site-directed mutation. The plasmid pRK415 carrying yfdX, hdeA, or hdeA F44A was individually used to transform the ΔyfdX mutant, and then acid survival rates were determined. Again, no differences in growth were confirmed in each bacterial strain before the pH 2.5 treatment. As illustrated in Fig 5A and the lower panel of Fig 5B, the deletion effects of yfdX could be entirely complemented by increasing the expression of YfdX, HdeA, and HdeA F44A . This finding suggests that YfdX as well as HdeA functions as a periplasmic chaperone. The acid survival of ΔyfdX[pRK415-HdeA F44A ] was less than that of ΔyfdX[pRK415-HdeA] indicating that the phenylalanine is crucial for determining HdeA chaperone activity. Although the crystal structure of K. pneumoniae YfdX has been resolved (PDB 3DZA), its functional role in the bacteria remains unknown. As shown in Fig 6, alcohol dehydrogenase, a commonly used substrate protein for chaperone activity analysis, was insoluble and became aggregated when the protein incubation switched from pH 7 to 1. The acid-induced aggregation significantly decreased after the addition of the purified recombinant YfdX, as assessed and visualized using optical density measurement and SDS-PAGE, respectively. This supports the possibility that YfdX proteins function as periplasmic chaperones to protect bacteria from acid stress damage.

RcsB regulates yfdX, hdeDB, and hdeB1 expression at the transcriptional level
To determine whether RcsB regulates the expression of yfdX and hde, the putative promoter sequences were analyzed. As depicted in Fig 7A, the predicted RcsB binding element -KMRGAWTMWYCTGS- [58] could be identified within the regions of P hdeDB , P yfdX , P kvhAS , P hdeB1 , P hdeA , and P hdeD1B2 . The rcsB deletion effects on each promoter activity were then measured using LacZ as the reporter. As shown in Fig 7B, after the bacterial strain was grown to the exponential phase and then cultured at pH 4.4 for 1 h for acid adaptation, the promoter activity of yfdX, hdeDB, or hdeB1 was reduced by the deletion of rcsB, whereas that of kvhAS, hdeA, or hdeD1B2 exhibited no apparent changes (Fig 7B).
To determine whether RcsB directly regulates the expression of yfdX and hdeDB, electrophoretic mobility shift assay (EMSA) analysis was subsequently performed. As shown in Fig  7C, the recombinant RcsB could bind to the intergenic DNA between yfdX and hdeDB (P yfdX ), or between kvhAS and hdeB1 (P hdeB1 ), and the interaction could be inhibited by an excess amount of the nonlabeled probe P yfdX or P hdeB1 demonstrating the binding specificity ( Fig 7C). As depicted in the lower panel of Fig 7C, the interaction between His-RcsB and P yfdX � was also outcompeted by an excess amount of P hdeB1 . Moreover, the binding between His-RcsB and P yfdX � or His-RcsB and P hdeB1 � was interfered by the addition of acetyl phosphate further supporting that phosphorylation status of RcsB determines its regulatory role in influencing the expression of P yfdX , P hdeDB and P hdeB1

YfdX expression may be affected by the phosphorylated form of RcsB
To investigate whether RcsB phosphorylation affects the acid response regulation as assessed by YfdX expression, Western blot analysis was conducted. Consistent with the promoter activity analysis, as shown in Fig 8A, the production of YfdX in the acidic LB broth (pH 4.4) was blocked by rcsB deletion but was unaffected by rcsC or rcsF deletion in both culture conditions. Notably, the YfdX production was reduced by the deletion of rcsD. And as shown in Fig 8B, the effect of rcsB deletion on YfdX production could be partly restored by complementation of RcsB and RcsB D56A , but not RcsB D56E . The phosphorylated form of RcsB (RcsB D56E ) could not restore the production of YfdX.
To further confirm the involvement of RcsD in the acid response regulation, the rcsD deletion effect was also determined in the complementation analysis. As shown in Fig 8C, in the rcsD deletion mutant, overexpression of RcsB D56A exhibited higher production of YfdX than  RcsB regulation of the Klebsiella acid stress response

Discussion
A comparative analysis of the proteomes and the promoter activity analysis in K. pneumoniae CG43S3 demonstrated a positive control of YfdX expression by RcsB, which suggested that YfdX expression may be used as a reporter for the RcsB-mediated regulation of the acid response. The expression of yfdX is positively controlled by RcsB, and RcsD seems to be required to ensure that RcsB is in the nonphosphorylated form to activate the expression of YfdX. We have also demonstrated that YfdX as well as HdeA may function as periplasmic chaperones to protect K. pneumoniae CG43S3 from acid stress damage. In E. coli, the nonphosphorylated form of RcsB positively affects the expression of AR2 [31,32,34]. Fig 2 illustrates that nonphosphorylated RcsB (RcsB D56A ) but not phosphorylated RcsB (RcsB D56E ) may complement the rcsB deletion effect. This suggests that nonphosphorylated RcsB (RcsB D56A ) also plays a major role in the acid stress response of K. pneumoniae which lacks AR2. Proteome analysis revealed that several differential proteins regulated by RcsB were induced under acidic conditions. We isolated only spot 832 (YfdX) for further study because it exhibited significant changes. Nevertheless, we could not exclude the involvement or importance of these proteins that were not identified.
The hdeA containing gene cluster is only present in CG43 but not in the genome of NTUH-K2044 and MGH78578. We speculate that the gene cluster may have been horizontally acquired during evolution of CG43. As shown in Fig 4, hdeA deletion had no effect on acid stress survival. By contrast, hdeA overexpression enhanced the acid survival of both wild type (S5B Fig) and yfdX deletion mutant (Fig 5). This indicates that HdeA may exhibit a chaperone activity against acid stress in CG43S3. Nevertheless, we speculated that HdeA might not be expressed at exponential phase due to the repression of that in E. coli by H-NS [34,52,53], but the result revealed that HdeA was also unnecessary at stationary phase. when and how HdeA functions in the bacteria to respond to acid stress warrants further investigation. The results of acid survival analysis also indicated that HdeD and HdeB1 may play an important role in protecting CG43S3 against acid damage. As shown in S6 Fig, compared with ΔhdeB[pRK415-hdeD], ΔhdeB[pRK415-hdeDB] exhibited a higher level of acid survival. This implies a major role of HdeD while an assistant role of HdeB in AR. However, if HdeB assists HdeD in K. pneumoniae resistance to acid stress remains to be investigated.
The effect of yfdX deletion was cross-complemented by HdeA suggesting YfdX as well as HdeA functions as a chaperone ( Fig 5). As shown in Fig 6, the result that the purified recombinant protein YfdX reduced the acid-induced protein aggregation further corroborated its chaperone activity. This is supported by a recent report indicating that STY3178 exhibits a periplasmic chaperone activity [56].
Acid survival analysis in Fig 2B suggested that RcsB in nonphosphorylated form played a positive role in the AR response. The possibility was further supported by the EMSA showing that adding acetyl phosphate interfered the binding efficiency of RcsB-P yfdX � or RcsB-P hdeB1 � . As shown in Fig 8 and S7 Fig, the YfdX production in ΔrcsB strain could be induced by expression of RcsB or RcsB D56A , while that in ΔrcsD strain only be induced by RcsB D56A , suggesting that RcsD determines the nonphosphorylated form of RcsB. In Salmonella, a phosphoryl group could be transferred differentially from RcsC or RcsD to RcsB depending on specific stimuli [59]. Whether the sensor kinase RcsC is required for the phosphorelay in regulating the acid stress response remains unknown.
Deletion of rcsB or rcsD reduced the YfdX production at either pH 4.4 (Fig 8A) or pH 7 (S7A Fig). However, the deletion effect of rcsB or rcsD on YfdX production could be fully restored by expression of RcsB or RcsB D56A in bacteria grown at pH 7 (S7B and S7C Fig) but could only be partly restored in bacteria grown in pH 4.4. Since the expression of YfdX, as assessed using promoter activity and protein production levels, was substantially higher in bacteria grown after pH 4.4 adaptation, we speculate that YfdX expression at pH 4.4 may be subjected to other regulatory system besides RcsBD.
In summary, this is the first report demonstrating that YfdX may be involved in the acid stress response as a periplasmic chaperone and the expression of yfdX is positively controlled by RcsB. Moreover, RcsD is required to ensure that RcsB is in the nonphosphorylated form to activate the expression of YfdX to enable acid stress response. RcsB phosphorylation status on YfdX production. (C) Analysis of the deleting effects of rcsD sensor kinase gene on the RcsB phosphorylation-dependent control. Bacteria was cultured in LB broth (pH 7) at 37˚C for 20h, and then total proteins were collected for western blot analysis of YfdX expression using anti-YfdX antiserum. The fold change of YfdX amount calculated using ImageJ software is shown. GAPDH was probed as protein loading control. (TIF) S1 Table. Analysis of the spots which exhibited differences between the proteomes of CG43S3 and CG43S3ΔrcsB.