Functional Genomic Screen Identifies Klebsiella pneumoniae Factors Implicated in Blocking Nuclear Factor κB (NF-κB) Signaling

Background: There is limited knowledge of Klebsiella pneumoniae determinants implicated in the blocking of the NF-κB signaling pathway. Results: A high-throughput genetic screen led to the identification of 114 putative K. pneumoniae genes that are associated with suppression of NF-κB activation. Conclusion: CPS, LPS, and the T2SS-secreted PulA are needed for immune evasion. Significance: A new therapeutic approach to treat Klebsiella infections will be the prevention of immune evasion.

are avirulent and unable to cause pneumonia and urinary tract infections (3)(4)(5). Lipid A of lipopolysaccharide (LPS), the outer membrane proteins OmpA and OmpK36, iron-scavenging systems, and several adhesins are other virulence determinants that have been characterized (6 -10). Nonetheless, there is still limited knowledge of the exact role of individual virulence factors in Klebsiella infections.
A wealth of evidence indicates that the activation of early inflammatory responses is essential to clear Klebsiella infections (11)(12)(13)(14). Any interference with this response leads to a more severe infection (15), thus in turn, augmenting the immune response with exogenous inflammatory mediators decreases the mortality associated with K. pneumoniae infection (16 -19). Collectively, this evidence suggests that Klebsiella tries to counteract the induction of these host defense responses. Indeed, we (8, 20 -22) and others (23) have provided compelling evidence for this notion.
At the molecular level, we have demonstrated that K. pneumoniae abrogates the activation of inflammatory responses by targeting NF-B and MAPK signaling pathways (22,24). Klebsiella antagonizes the activation of NF-B via the deubiquitinase CYLD and blocks the phosphorylation of MAPKs via the MAPK phosphatase MKP-1 (22). CYLD and MKP-1 are normally regulated to return to homeostasis after inflammation to protect the host from an overwhelming inflammatory response (25,26). To exert this anti-inflammatory effect, K. pneumoniae affects the membrane association of the receptor NOD1 (22). This is dependent on Klebsiella-triggered Rho GTPase Rac1 inhibition (22). To identify additional host factors involved in the anti-inflammatory effect, we applied an unbiased highthroughput siRNA gain-of-function screen to interrogate the human kinome (24). Follow-up validation revealed that K. pneumoniae exploits an EGF receptor (EGFR)-PI3K-AKT-PAK4-ERK-GSK3␤ signaling pathway to induce the expression of the deubiquitinase CYLD to attenuate the cytokine-dependent nuclear translocation of NF-B (24). Our group uncovered a role for CPS in the activation of EGFR and EGFR-dependent signaling (24). However, because CPS does not activate NOD1dependent responses (22), K. pneumoniae may employ additional factors to attenuate NF-B activation.
This study was designed to identify K. pneumoniae determinants implicated in blocking the activation of the NF-B signaling pathway. To take a systematic approach toward the identification of these bacterial factors, we performed a high-throughput genetic screen to mine a transposon mutant library of K. pneumonia strain 52145. This is a reference strain of serotype K2 highly virulent Klebsiella strains from which important virulence factors, including the large virulence plasmid harboring the regulator of mucoid phenotype (rmpA) and the aerobactin cluster, have been identified (27). We found 114 mutants that, in sharp contrast to the wild-type strain, activated the NF-B signaling pathway. Further characterization confirmed the critical role of K. pneumoniae CPS in blocking NF-B activation and uncovered the role of the LPS polysaccharide section and the pullulanase type II secretion system (T2SS) in immune evasion.

Bacterial Strains, Growth Conditions, and Reagents
Strains and plasmids used in this study are listed in Table 1. K. pneumoniae 52145 (hereafter Kp52145) is a clinical isolate (serotype O1:K2) described previously (3,28). Bacteria were grown in Luria-Bertani (LB) medium at 37°C unless indicated otherwise. When appropriate, antibiotics were added to the growth medium at the following concentrations: ampicillin, 100 g/ml; rifampicin, 50 g/ml; kanamycin, 50 g/m; cloramphenicol (Cm), 25 g/ml; and trimethoprim (Tp), 100 g/ml. FeSO 4 , and FeCl 3 were used at a concentration of 10 M, and 2-2Ј-dipyridyl (Sigma) was used at a concentration of 200 M. The sequences of the primers used in this study are available from the authors upon request.

Cell Culture and Infection
Monolayers of A549 (ATCC CCL185) were grown as described previously (22). For infections, A549 cells were seeded to 90% confluence (3 ϫ 10 5 cells/well) in 24-well tissue culture plates. Cells were serum-starved for 16 h before infection. Bacteria were prepared as described (22), and infections were performed using a multiplicity of infection of 100 bacteria/cell unless indicated otherwise. For incubation times longer than 120 min, bacteria were killed by the addition of gentamicin (100 g/ml), which was not removed until the end of the experiment. Cell viability, assessed by trypan blue dye exclusion, was Ͼ95% even after 4 h of infection.

Mutant Library Construction
Kp52145 was made electrocompetent following the method described by Sharma and Schimke (29). Transposon mutagenesis was performed by electroporating ϳ1 g of the pRL27 mini Tn5 transposon (30) to 50 l of Kp52145 electrocompetent cells using a Gene Pulser Xcell (Bio-Rad) followed by a 60-min recovery in SOC (super optimal broth) medium at 37°C. The transposon carries a kanamycin resistance cassette and an R6K origin of replication (30). Transposants were plated on LB-rifampicin-kanamycin plates, and the transposon mutant library was generated from six independent electroporation rounds. A total of 5320 mutants were arrayed in a 96-well plate format and stored at Ϫ80°C in 20% (v/v) glycerol (master plate). One well without bacteria was kept as a negative control in all plates (typically well H12). Random transposon insertion was checked in ϳ200 clones by direct genomic sequencing. Genomic DNA was extracted using the Realpure spin kit (Real). Sequencing reaction was performed in an iCycler (Bio-Rad) containing 2 l of BigDye Master Mix, 2 l of 5ϫ buffer, 1 l of betaine (5 M), 1 M primer (tpnRL17), and 2-5 g of genomic DNA. The thermal profile consisted of 5 min of a denaturalization step at 95°C followed by 100 cycles of 95°C for 30 s, 50°C for 20 s, and 60°C for 4 min. Sequencing reactions were purified following the manufacturer's instructions, resuspended in 10 l of Hi-Di formamide, and run in an ABI 3730 genome sequencer (Applied Biosystems).

Construction of a NF-B Reporter Cell Line
A549 cells were seeded into 24-well plates at a density of 4 ϫ 10 4 cells/well and transfected 24 h later with 5 g of pNifty2-SEAP (InvivoGen) using Lipofectamine 2000 (Invitrogen) lipofection reagent according to the manufacturer's instructions. In the pNifty2-SEAP vector, the proximal promoter of the endothelial cell-leukocyte adhesion molecule (ELAM-1; E-selectin), containing three NF-B sites and lacking an AP1/CREB (cAMP-response elementbinding protein) site found in the full-length promoter, controls the reporter gene, SEAP (secreted form of the human embryonic alkaline phosphatase). Twenty-four hours after transfection, cells were selected with 500 g/ml Zeocin for 2 weeks (InvivoGen). Cells were singularized by serial dilutions in 96-well plates, and monoclonal cell lines were propagated for 2 weeks. Stably transfected A549-SEAP cells were grown in complete RPMI medium supplemented with 100 g/ml Zeocin. A2 was the clone selected for screening purposes.

High-throughput Screening
Cell Seeding and Bacterial Culture-A549-SEAP A2 cells from the same passage number were seeded into 96-well plates at 1-2 ϫ 10 4 cells/well in 100 l of RPMI 1640. Transposon mutants and controls were grown in 96-well plates. 5 l from the bacteria master plate were used to inoculate 150 l of LB supplemented with the appropriate antibiotics. Plates were incubated overnight at 37°C without shaking. Bacterial growth was determined by measuring absorbance at 600 nm (A 600 ) using a microplate reader (Biotek PowerWave HT).
Infection-A549-SEAP A2 cells were washed twice with 200 l of PBS, and then 150 l of RPMI 1640 supplemented with 10% fetal calf serum (FCS) was added. Five l of the bacterial overnight culture was used to infect the cells (multiplicity of infection, ϳ100:1). Only the inner 60 wells of the plate were used. Two replicates of each control (negative/positive (IL-1␤, 10 ng/ml) cells infected with Kp52145, 52OmpA, 52145⌬wca K2 , and 52145⌬wca K2 -OmpA) and 48 mutants were run in each plate. After 3 h, wells were washed twice with 200 l of PBS, and then 100 l of RPMI 1640 containing 100 g/ml of gentamicin was  (37) added to each well. Plates were incubated overnight in a humidified CO 2 incubator. Colorimetric Assay-30 l of the supernatant was mixed with 200 l of QUANTI-Blue reagent (InvivoGen), and the mixture was incubated at 37°C for 24 h. The levels of SEAP were measured at an absorbance at 625 nm (A 625 ).
Statistical Analysis-The Z score was calculated for each of the 48 wells of a plate as described previously (31). For a given well, the Z score was calculated by subtracting the mean value of the wells on that plate from the value of the well and dividing by the standard deviation value for all the plate wells. A Z score Ն 2 was considered significant. As means and standard deviations are greatly influenced by statistical outliers (31), which in the context of a screening are putative hits, we also selected mutants that induced SEAP levels higher than A 625 Ͼ 0.5, which represents a 2-fold A 625 value induced by the wildtype strain.
Hit Validation-Four independent validation rounds were run for each selected hit. A hit was considered validated if the SEAP levels were A 625 Ͼ 0.5 in at least two of four validation runs.
Transposon Insertion Site Identification-Genomic DNA was purified using the Realpure spin kit (Real). Direct genomic sequencing was carried out as described above (see "Mutant Library Construction").
Growth Deficiency Determination-To identify those hits with growth deficiencies, transposon mutants were plated on LB plates or M9 plates supplemented with either 0.2% glucose or 10 mM citrate and incubated at 37°C for 48 h.

CPS Purification and Quantification
Cell-associated CPS from K. pneumoniae strains were purified using the hot phenol-water method exactly as described previously (32). CPS was quantified by determining the concentration of uronic acid in the samples using a modified carbazole assay (33).

LPS Extraction and Analysis by SDS-PAGE
The LPS from K. pneumoniae strains was extracted by a modified phenol-water purification method (34). The LPS was run on a 12% SDS-PAGE and visualized using a Pro-Q Emerald 300 lipopolysaccharide gel stain Kit (Invitrogen).

K. pneumoniae Mutant Construction
The entF, irp1, iutA, and pulA mutants were constructed by insertion mutagenesis using the pKNOCK-Cm suicide vector (35). An internal fragment of each gene was amplified using genomic DNA from wild-type K. pneumoniae 52145, Vent polymerase (New England Biolabs), and the corresponding primer pairs. PCR products were phosphorylated and cloned into the SmaI-digested and dephosphorylated pKNOCK-Cm vector to obtain pKNOCK-Cm_entF, pKNOCK-Cm_irp1, pKNOCK-Cm_iutA, and pKNOCK-Cm_pulA, respectively. The glf mutant was also constructed by insertion mutagenesis using the pir replication-dependent plasmid pSF100 (36). Plasmids were introduced into Klebsiella by conjugation and PCR, or Southern blot was used to identify strains in which plasmids were inserted into the genome by homologous recombination.

Complementation of pulA Mutant
To complement the pulA mutant, a DNA fragment of 4,008 bp containing the putative promoter region and coding region of pulA was PCR-amplified (primer pair pulACom.F-pul ACom.R) using Vent polymerase, gel-purified, and cloned into SmaI-digested pUC18R6KT-mini-Tn7TKm (37) to obtain pUC18R6KT-mini-Tn7TKm_pulACom. The pTSNSK-Tp plasmid, which contains the transposase tnsABCD, necessary for Tn7 transposon mobilization (38), was electroporated into the pulA mutant. Then, the pUC18R6KT-mini-Tn7TKm_pulACom plasmid was mobilized into this strain by triparental conjugation using the helper strain Escherichia coli HB101/pRK2013. Colonies were screened for resistance to kanamycin and sensitivity to ampicillin. Because the ampicillin cassette is located outside of the Tn7 region on the vector, sensitivity to ampicillin denotes the integration of the Tn7 derivative at the attTn7 site instead of incorporation of the vector into the chromosome. Confirmation of integration of the Tn7 transposon at the established attTn7 site located downstream of the glmS gene was verified by a multiplex PCR using primers glmS.UP, glmS.DW, Tn7.L, and Tn7.R. PCR reactions were performed in a final volume of 25 l containing 50 ng of DNA, 1.5 mM MgCl 2 , 0.2 mM dNTPs, each primer at 0.2 M, and 1 unit of GoTaq polymerase (Promega). The thermal profile was as follows: an initial denaturation step at 94°C for 5 min followed by 5 cycles of 94°C for 45 s, a touchdown from 62 to 58°C for 45 s and 72°C for 45 s, and then 25 cycles of 94°C for 45 s, 58°C for 45 s, and 72°C for 45 s, and a final extension step at 72°C for 5 min. Correct insertion of the Tn7 transposon yielded two amplicons of 462 and 216 bp amplified with primer pairs glmS.UP-Tn7.L and glmS.Dw-Tn7.R, respectively. Only one amplicon of 455 bp, corresponding to an internal fragment of the glmS gene, was amplified when used as the template DNA from wild-type Klebsiella. pTSNSK-Tp from the recipient strain was cured by growing bacteria at 37°C due to the plasmid-thermosensitive origin of replication pSC101. Plasmid removal was confirmed by susceptibility to trimethoprim.

Construction of gfp Reporter Fusions
A trimethoprim resistance cassette was obtained as a SmaI fragment from plasmid p34S-Tp (40). The cassette was cloned into SphI-digested and blunt-ended plasmid pPROBEЈgfp[LVA] to obtain pPROBEЈ-gfp[LVA]Tp. This vector contains a promoterless gfp gene (41). DNA fragments containing the promoter regions of entC, ybtA, psn, and iucA were amplified by PCR using Vent polymerase, gel-purified, and cloned into SmaI-digested pPROBEЈ-gfp[LVA]Tp plasmid. Plasmids were introduced in E. coli DH5␣-pir and then mobilized into Klebsiella by triparental conjugation using the helper strain E. coli HB101/pRK2013.

Luciferase Activity and GFP Fluorescence Measurements
Overnight cultures of the reporter strains were diluted 1/10 in fresh LB medium and incubated for 3 h at 37°C on an orbital incubator shaker (180 rpm). The cultures were harvested (4000 ϫ g for 20 min) and resuspended to an A 600 of 1.0 in PBS. Luciferase activity and GFP fluorescence were determined as described previously and expressed as relative light units (RLU) or relative fluorescence units (RFU) (42). All measurements were performed in triplicate on at least three independent occasions.

Adhesion and Internalization Assays
Bacteria adhesion and internalization to A549 cells were determined as our group described previously (8). The results are expressed as cfu/well. Experiments were carried out in duplicate on at least three independent occasions.

IL-8 Stimulation Assay
A549 cells, seeded in 24-well plates, were infected for 2 h and then washed twice with PBS, and fresh medium plus gentamicin (100 g/ml) was added. Supernatants were recovered after 12 h, cell debris was removed by centrifugation, and samples were frozen at Ϫ80°C. IL-8 levels in the supernatants were determined using a commercial ELISA (eBioscience) with a sensitivity of Ͻ2 pg/ml. Experiments were run in duplicate and repeated at least three independent times.

Small Interfering RNA (siRNA)
Transfection of siRNAs was performed at the time of cell seeding in a 96-well plate format (2 ϫ 10 4 cells/well). Lipofectamine TM 2000 transfection reagent (Invitrogen) was used following the manufacturer's instructions. Transfection experiments were carried out in Opti-MEM reduced serum medium (Invitrogen). siRNAs were used at a concentration of 20 nM, and experiments were carried out 48 h after transfection. The knockdown efficiency of the siRNAs targeting NOD1, TLR2, TLR4, and EGFR was verified previously (22,24).

Murine Infection Model
Mice were treated in accordance with the Directive of the European Parliament and the Council on the Protection of Ani-mals Used for Scientific Purposes (Directive 2010/63/EU) and in agreement with the Bioethical Committee of the University of the Balearic Islands. This study was approved by the Bioethical Committee of the University of the Balearic Islands (authorization number 1748).
Five-to-seven-week-old female CD-1 mice (Harlan) were infected as described previously by our group (8). Non-infected mice were inoculated with 20 l of PBS. At 24 h post-infection, mice were euthanized by cervical dislocation. Tissues were rapidly dissected for bacterial load determination, and half of the lungs were immediately frozen in liquid nitrogen and stored at Ϫ80°C until purification of RNA was carried out.

RNA Purification and Real Time Quantitative PCR (RT-qPCR)
RNA from lungs was purified as described by us (8). cDNA was obtained by retrotranscription of 1 g of total RNA using the M-MLV reverse transcriptase (Sigma). The reaction included one step to eliminate traces of genomic DNA. RT-qPCR analyses were performed using an iCycler real-time PCR instrument (Bio-Rad). Reactions were performed in 25 l containing 12.5 l of KAPA SYBR FAST qPCR kit master mix (KapaBiosystems), each primer at 0.2 M, and 50 ng of cDNA. Relative quantities of mRNAs were obtained using the comparative threshold cycle (⌬⌬C T ) method by normalizing to hprt1.
Bacteria were grown at 37°C in 5 ml of LB medium on an orbital incubator shaker (180 rpm) until an A 600 of 0.3 was reached. RNA was purified following an established protocol (43). cDNA was obtained by retrotranscription of 2 g of total RNA using commercial M-MLV reverse transcriptase (Sigma) and random primers mixture (SABiosciences, a Qiagen company). 50 ng of cDNA was used as a template in a 25-l reaction. RT-PCR analyses were performed with a Smart Cycler realtime PCR instrument (Cepheid, Sunnyvale, CA) using a KapaSYBR Fast qPCR kit as recommended by the manufacturer (Cultek). SYBR Green dye fluorescence was measured at 521 nm. cDNAs were obtained from two independent extractions of mRNA, and each one was amplified by RT-qPCR on two independent occasions. Relative quantities of wabM mRNAs were obtained using the comparative threshold cycle (⌬⌬C T ) method by normalizing to the rpoD gene.

Statistical Analysis
Results were analyzed by analysis of variance (ANOVA) or with a one-tailed t test followed by a Bonferroni contrast correction for multiple testing using GraphPad Prism software (GraphPad Software Inc.). Results are given as the means Ϯ S.D. A p value of Ͻ0.05 was considered statistically significant.

Results
High-throughput Screening to Identify K. pneumoniae Loci Involved in Attenuating NF-B Activation-We developed an assay suitable for identifying the K. pneumoniae loci required to block activation of the NF-B pathway. A new indicator cell line was engineered by transfection of the human epithelial cell line A549 with the reporter vector pNIFTy2-SEAP (InvivoGen). Three clones were selected, A549-SEAP A2, D2, and D4, and control experiments were carried out to determine whether different stimuli known to activate NF-B induce the secretion of SEAP (Fig. 1A). Clone A2 secreted higher amounts of SEAP than the other two clones after cells were challenged with different stimuli (Fig. 1A). Time course experiments showed that clone A2 secreted higher levels of SEAP over time than the other two clones upon challenge with IL-1␤ (10 ng/ml) (Fig.  1B). Western blot analysis of IB␣ levels showed no differences between A549 cells and A549-SEAP A2 after stimulation with different agonists (Fig. 1C). Clone A2 was selected for screening purposes.
SEAP levels induced by Kp52145 were not significantly different than those observed in non-infected cells ( Fig. 2A). In contrast, the cps, ompA mutants, and the cps-ompA double mutant did trigger the secretion of SEAP. We have already reported that these strains elicit the activation of NF-B (8). The differences observed between strains are consistent with data from our laboratory showing the relative importance of CPS and OmpA to attenuate NF-B activation (8).
The assay was then transferred from a 24-well format to a 96-well format by optimizing culture conditions, cell number/ well, multiplicity of infection, time of contact, and time of detection of secreted SEAP. To determine the quality of the assay, the ZЈ factor (31) was calculated taking as a negative control SEAP levels secreted by cells infected with the wild-type strain and as a positive control SEAP levels secreted by cells infected with the ompA mutant or cells stimulated with IL-1␤.  In our case, the resulting ZЈ factors were 0.73 and 0.86 for the respective comparisons. The ZЈ factor is a screening window coefficient indicating the capability of a hit identification for a given assay at the defined screening conditions. ZЈ factors higher than 0.5 indicate that an assay is robust enough for a high-throughput screening (31).
To identify Klebsiella loci involved in attenuating NF-B activation, we examined a bank of 5320 transposon mutants. Each mutant was tested individually. For candidate selection the standard Z score normalization procedure was applied and mutants with a Z score Ն 2 were selected. We also selected those mutants that induce SEAP levels higher than A 625 Ͼ 0.5, which represents a 2-fold increase in the SEAP levels induced by the wild type. The distribution of A 625 values and the Z scores of the mutant library are shown in Fig. 2, B and C. Following the first round of screening, 522 mutants (10% of the transposon library) were considered candidates for validation.
To validate the selected candidates, each one was tested individually in four independent screening rounds. Candidates were considered validated if they induced SEAP levels higher than A 625 Ͼ 0.5 in at least two of four independent experiments upon infection of the A549-SEAP A2 cells. Using this validation criterion, 114 Kp52145 mutants were considered validated ( Table 2). Control experiments showed that all mutants grew in LB agar plates, LB broth, and RPMI 1640 complete medium, whereas five and three mutants displayed growth deficiencies when cultivated in M9 agar plates supplemented with glucose and citrate, respectively ( Table 2).
Transposon insertion sites were determined by genomic sequencing using primer tpnRL17. The sequences obtained were compared with the Kp52145 genome and plasmid sequences (GenBank TM accession numbers FO834904, FO834905, and FO834906) and to the reference sequence of K. pneumoniae strain MGH78578 (GenBank TM accession number CP000647.1). Ninety-nine insertions were found in the chromosome, whereas nine and six were located in plasmid 1 and plasmid 2, respectively. The MicrobesOnline and STRING databases were interrogated to annotate the loci. Thirty of them were related to metabolism, twenty were included in the category outer membrane and envelope-related genes, nine in transport, eight in regulation of transcription, seven in the stress response category, and four in the category of DNA-related processes (Table 2). Interestingly, six of the insertions were located in regions of the Kp52145 genome not present in other Klebsiella genomes (27). Nineteen mutants had the transposon inserted in non-coding regions of the genome, and 17 were located within the coding region of hypothetical proteins.
In summary, our high-throughput genetic screen mining a K. pneumoniae transposon library led to the identification of 114 mutants that triggered the activation of NF-B. In the following sections we describe the characterization of the mutants on the enterobactin siderophore, WaaL encoding the O-antigen ligase and pullulanase T2SS. We selected the enterobactin and ligase mutants because iron-scavenging systems and LPS are two well characterized Klebsiella virulence determinants, although their contributions to attenuating inflammation have not been evaluated previously. On the other hand, we chose to investigate the Klebsiella pullunase T2SS because its role in Klebsiella virulence had not been established.
Analysis of the Enterobactin Mutant-One transposon insertion was identified in the enzyme EntF, a component of the biosynthetic machinery of the enterobactin siderophore (44). Enterobactin is one of the three siderophores expressed by Kp52145 (27). Enterobactin is expressed in the lungs of infected mice, and an enterobactin mutant is attenuated (9). As anticipated (9), the doubling times of the entF mutant slowed when grown in LB supplemented with the iron chelator 2-2Ј-dipyridyl (Fig. 3A). Similar rates of growth were observed when the wild type and the mutant were grown in LB or RPMI 1640 (data not shown).
We sought to determine whether the other two siderophores encoded by Kp52145, aerobactin and yersiniabactin, are implicated in attenuation of inflammation. However, the SEAP levels induced by the aerobactin mutant, strain 52iutA, and the yersiniabactin mutant, strain 52irp1, were not significantly different than those induced by the wild-type strain (Fig. 3B). In contrast, SEAP levels induced by the entF mutant were higher than those induced by Kp52145, which in turn, were not significantly different than those found in the supernatant of non-infected cells (Fig. 3B). In the canonical NF-B activation pathway, nuclear translocation of NF-B is preceded by phosphorylation and subsequent degradation of IB␣ (45). Immunoblot analysis demonstrated that the entF mutant induced the degradation of IB␣ in A549 cells (Fig. 3C).
These findings suggested that enterobactin, but not the other siderophores, was expressed under our infection conditions. To monitor the transcription of the siderophores quantitatively, transcriptional fusions were constructed in which a promoterless gfp gene was under the control of the iutA, psn, ybtA, and entC promoter regions (see "Experimental Procedures"). iutA has been used previously to study aerobactin expression and psn and ybtA to analyze yersiniabactin expression, whereas entC has been used to determine enterobactin expression (9). gfp fusions were introduced into Kp52145, and fluorescence was determined after growing the reporter strains in LB, LB supplemented with iron chelator 2-2Ј-dipyridyl, or RPMI 1640 (Fig. 3D). The entC::gfp fusion was the only one expressed under iron-limited conditions, LB plus 2-2Ј-dipyridyl and RPMI 1640 (Fig. 3D). It has been reported previously that bacteria up-regulate iron-scavenging systems when grown in tissue culture medium (46). Further confirming this finding, the addition of FeCl 3 to RPMI 1640 decreased the expression of the entC::gfp fusion by 75% (9106 Ϯ 850 RFU and 2190 Ϯ 323 RFU, respectively; p Ͻ 0.05, one-tailed t test).
Collectively, our data are consistent with a model in which K. pneumoniae experiences a poor iron environment when infecting A549 cells, hence leading to the activation of the siderophore enterobactin. There are reports suggesting that siderophores may inhibit immune responses (47), hence making it possible that enterobactin may directly attenuate NF-B activation. However, it has been also shown that K. pneumoniae CPS expression is down-regulated in iron-poor media (48). Therefore, the involvement of enterobactin in attenuating NF-B activation would be indirect considering that CPS is known to prevent NF-B activation (20,21). Indirectly support-  ing the latter evidence, the entF mutant expressed 50% less cellbound CPS than the wild-type strain (6.11 Ϯ 2.99 mg/10 7 cells and 11.49 Ϯ 1.91 mg/10 7 cells, respectively; p Ͻ 0.05, one-tailed t test), and the activity of the transcriptional cps::lucFF fusion was 80% lower in the entF mutant background than in the wildtype one (3.1 ϫ 10 6 Ϯ 16,000 RLU and 1.7 ϫ 10 7 Ϯ 34,000 RLU, respectively; p Ͻ 0.01, one-tailed t test). Further confirming the inverse correlation between the CPS levels and the adhesion and internalization to epithelial cells (49), the entF mutant adhered to and was internalized by A549 cells in higher numbers than the wild type (Fig. 3E). Altogether, this evidence suggests that entF-induced NF-B activation could be explained by the reduced CPS levels expressed by the mutant. To validate this hypothesis, we asked whether the SEAP levels induced by the entF mutant were significantly different than those triggered by the cps mutant or the double cps-entF mutant. The data shown in Fig. 3F demonstrate that indeed there were no differences in the SEAP levels triggered by any of the three strains.
In aggregate, our findings revealed that enterobactin was activated under our tissue culture infection conditions and that the lack of the siderophore was linked to a reduced CPS expression. In turn, the low CPS levels underlined the NF-B activation induced by the entF mutant.  (50). LPS analysis confirmed that neither of the mutants express OPS, and the LPS pattern was similar to that displayed by a nonpolar waaL mutant (50) (Fig. 4A). waaL is encoded within the operon responsible for the synthesis of the LPS core (50), hence making it possible that the transposon insertion would affect the expression of downstream genes. However, RT-qPCR experiments showed that the expression of the downstream gene wabM was not significantly different between the wild type and the transposon mutants (data not shown). WabM encodes the glycosyltransferase, adding the first glucose residue of the core where the OPS is attached (50). The SEAP levels induced by the ligase mutants were significantly higher than those induced by the wild type (Fig. 4B). No significant differences were found in the SEAP levels induced by any of the mutants (Fig. 4B).
To rule out the possibility that the non-ligated OPS exerted an unanticipated effect on Klebsiella physiology, we tested the glf glycosyltransferase mutant, strain 52glf. Glf is a UDP-galactopyranose mutase essential for OPS biosynthesis (51). Immunoblot analysis revealed that the ligase and OPS mutants triggered the degradation of IB␣ in A549 cells in contrast to the  wild-type strain (Fig. 4C). Many cellular stimuli also activate MAPK pathways. The activation of the three MAPKs (p38, JNK, and p44/42) occurred through phosphorylation of serine and threonine residues. All strains induced the phosphorylation of the three MAPKs; however the phosphorylation of p38 and p44/42 was more apparent in cells infected with the ligase and OPS mutants than in those infected with the wild type (Fig.  4D). No differences were observed between both mutants (Fig.  4, C and D). Altogether, these results support the view that Kp52145 OPS is required to attenuate inflammation. waaL mutants express similar levels of CPS as the wild type (Ref. 50 and data not shown). To assess the relative contribution of CPS and OPS to attenuate NF-B activation, we investigated whether the absence of OPS increases SEAP levels induced by the cps mutant. Results shown in Fig. 4E confirmed that this was the case.
To further analyze the contribution of the LPS polysaccharide section to the attenuation of NF-B activation, we sought to determine whether the LPS core section plays any role in Klebsiella attenuation of NF-B activation. wabM, wabK, and wabH mutants lack, in addition to the OPS, the first, second, and third sugars of the LPS core, respectively (50), but they express similar levels of CPS as the wild type (50). The three mutants triggered the secretion of SEAP by infected cells (Fig.  4E). SEAP levels were higher than those induced by the wild type; the levels induced by the wabM and wabK mutants were also significantly higher than those induced by the waaL mutant (Fig. 4E). No significant differences were found between the wabM and wabK mutants (Fig. 4E). In the genetic background of the cps mutant, the wabM and wabK mutants induced similar levels of SEAP, which were higher than those triggered by the waaL mutant (Fig. 4E). SEAP levels induced by the wabH mutant were not significantly different than those induced by the waaL mutant (Fig. 4E). Comparatively similar results were found when the secretion of IL-8 was evaluated (Fig. 4F). Collectively, these data show that the CPS, the OPS, and the first glucose of the core are bacterial factors required to attenuate the activation of NF-B by epithelial cells upon K. pneumoniae infection.
Recently, we have shown that Kp52145 attenuates proinflammatory mediator-induced NF-B activation (22). This process requires bacteria-cell contact, and removal of bacteria by washing followed by a 1-h gentamicin treatment renders the cells responsive to agonist-induced IL-8 secretion (22). To exert this anti-inflammatory effect, Kp52145 engages NOD1 and EGFR receptors (22,24). To explore whether the OPS could account for the Kp52145 anti-inflammatory effect, we determined the effect of the waaL mutant on IL-1␤-induced IB␣ degradation using the assay that we described previously (22,24). Similar to Kp52145, waaL attenuated IL-1␤-induced IB␣ degradation (Fig. 4G). As with the wild type, this effect was abrogated in cells in which NOD1 and EGFR were knocked down using siRNA (Fig. 4H).
Control experiments revealed that the ligase and core mutants adhered to A549 cells in higher numbers than the wild type (for each comparison between Kp52145 and the mutants, p Ͻ 0.05 (one-tailed Student's t test)) (Fig. 4I). The adhesion levels of the wabM and wabK mutants were higher than that of the waaL mutant (Fig. 4I). As reported previously (49), strains lacking CPS adhered in higher numbers than CPS-expressing strains (for each comparison between CPS-expressing and CPS-negative strains, p Ͻ 0.01 (one-tailed Student's t test)) (Fig.  4I). In the cps mutant background, no significant differences in adhesion between strains were found (Fig. 4I). waaL and core mutants were internalized in higher numbers than the wild type (for each comparison between Kp52145 and the LPS mutants, p Ͻ 0.01 (one-tailed Student's t test)) ( Fig. 4I). In the cps mutant background, however, waaL, wabM, and wabK mutants were internalized in lower numbers than the cps mutant (Fig. 4I), and the levels were not significantly different than those of the wild type (for each comparison between Kp52145 and the LPS mutants, p Ͼ 0.05 (one-tailed Student's t test)). Interestingly, wabH mutant displayed internalization levels similar to those of the cps mutant (Fig. 4I).
Finally, we sought to determine the innate receptor(s) involved in the recognition of the OPS mutant by using as cellular readouts of the NF-B-dependent secretion of SEAP and the secretion of IL-8. Almost all TLRs activate cellular signaling pathways through TIR domain-mediated interactions with the adaptor molecule MyD88 (52). To explore the involvement of TLRs in waaL mutant-induced cell activation, the function of the MyD88 adaptor molecule was interrupted by siRNA. In MyD88 knocked-down cells, waaL mutant induced neither the activation of the NF-B reported construct nor the secretion of IL-8 (Fig. 5, A and B, respectively). To further dissect the contribution of TLR-dependent signaling to waaL mutant-induced cell activation, TLR2 and TLR4 were knocked down by siRNA. The results shown in Fig. 5 indicate that both receptors contributed to waaL-induced cell activation. On the whole, these data suggest that waaL mutant induced NF-B activation and IL-8 secretion is mediated by TLR4-TLR2-MyD88 pathway.
Analysis of the Pullulanase Mutants-Two transposon insertions were found in the pulC and loci. pulC is the first locus of the Klebsiella T2SS that secrets the enzyme pullulanase encoded by pulA (53). yacC encodes for a lipoprotein containing a domain related to the pulS/outS family but for which the  -tailed t test). In E and F, CPS-expressing strains, gray bars; CPS negative strains, black bars. exact function has not yet been described. To help understand the contribution of Klebsiella T2SS to NF-B attenuation, we constructed a pulA mutant, because PulA is the only known protein secreted by K. pneumoniae T2SS (54). Immunoblot analysis confirmed the absence of the enzyme pullulanse in the outer membranes of the pulC and yacC mutants and, as expected, also in the outer membrane of the pulA mutant (Fig.  6A). The amount of cell-bound CPS expressed by the mutants was quantified, and no differences were found between the CPS expressed by the wild type (11.49 Ϯ 1.91 mg/10 7 cells) and the CPS expressed by any of the mutants (52⌬pulA, 15.21 Ϯ 3.65 mg/10 7 cells; 52pulC::tn5, 9.44 Ϯ 2.91 mg/10 7 cells; and 52yacC::tn5, 13.26 Ϯ 4.97 mg/10 7 cells; for each comparison between wild-type CPS levels and mutant levels, p Ͼ 0.05, onetailed Student's t test). Control experiments revealed that the adhesion levels of the pulC and yacC mutants were higher than those of the pulA and Kp52145 strains, which in turn, were not significantly different (Fig. 6B). In contrast, no significant differences were found in the internalization to cells between any of the strains (Fig. 6B).
SEAP levels induced by the T2SS mutants were higher than those induced by the wild type (Fig. 6C), and IB␣ levels were lower in those cells infected with the T2SS mutants (Fig. 6D). Complementation of the pulA mutant restored the SEAP levels to those triggered by the wild type, and we found an increase in the levels of IB␣ in the lysates of infected cells (Fig. 6, C and D). Cells infected with the pulA mutants also secreted higher levels of IL-8 than cells infected with the wild type (Fig. 6E). Complementation of the mutant restored IL-8 secretion to wild-type levels. We also evaluated the activation of MAPKs in pulA mutant-infected cells. Western blot analysis shown in Fig. 6E revealed that infection with the pulA mutant triggered the phosphorylation of the three MAPKs. Phosphorylation levels of p38 and p44/42 were higher at 60 and 90 min post-infection in cells infected with the pulA mutant than in those infected with Kp52145 (Fig. 6E). Kp52145 and pulA mutant triggered similar levels of phosphorylated JNK (Fig. 6E). In aggregate, these findings indicate that K. pneumoniae that PulA contributes to limiting the activation of the NF-B canonical pathway and MAPKs p38 and p44/42.
siRNA-based experiments were carried out to investigate whether a TLR-MyD88-dependent pathway is implicated in the recognition of the pulA mutant. SEAP levels induced by the pulA mutant were reduced to those induced by the wild type only in MyD88 knocked-down cells and in TLR2-TLR4 knocked-down cells (Fig. 7A). Likewise, pulA mutant-triggered IL-8 secretion was also dependent on TLR4-TLR2-MyD88 (Fig. 7B). Altogether, our results demonstrated that K. pneumoniae PulA is required to attenuate the activation of NF-B by limiting the activation of a TLR4-TLR2-MyD88 pathway.
Relative Contribution of CPS, O-polysaccharide, and pulA to K. pneumoniae Attenuation of Inflammation-The fact that CPS, LPS OPS, and PulA are all involved in attenuating inflammatory responses led us to study the relative contribution of each of these factors. Therefore, we asked whether the absence of PulA further increases the SEAP levels and the secretion of IL-8 induced by the cps and waaL mutants. Data displayed in Fig. 7A show that SEAP levels induced by the double mutant waaL-pulA were higher than those induced by either the waaL or pulA single mutant. In turn, waaL-infected cells secreted higher SEAP levels than cells infected with the pulA mutant (Fig. 8A). In the cps mutant background, the strain inducing the highest SEAP levels was the double mutant waaL-pulA. No significant differences were observed between the waaL and pulA mutants. Similar picture was obtained when the IL-8 secretion was used as read out for cellular activation (Fig. 8B). Collectively, these data show that the CPS, the LPS OPS and PulA are Klebsiella factors require to limit NF-B activation and the secretion of IL-8 by infected epithelial cells.
Virulence of K. pneumoniae Mutants-To determine the ability of pulA and waaL mutants to cause pneumonia, CD-1 mice were infected intranasally, and 24 h post-infection the bacterial loads in trachea, lung, spleen, and liver homogenates were determined. No differences were found in the bacterial loads in the trachea (Fig. 9A). In contrast, the bacterial loads of the pulA and waaL mutants were significantly lower in lung, spleen, and liver than those of the wild type (Fig. 9A).
We also assessed by RT-qPCR the expression levels of kc and tnf␣ cytokines in the lungs of infected mice. The levels of kc were higher in lungs of infected mice than in lungs of non-infected animals (p Ͻ 0.05 for all comparisons versus non-infected mice; one-  tailed t test) (Fig. 9B). However, kc levels were significantly higher in the lungs of mice infected with pulA and waaL mutants than in those infected with the wild type (Fig. 9B). tnf␣ levels were higher in the lungs of mice infected with pulA mutant than in those infected with the wild type or the waaL mutant.

Discussion
We recently showed that K. pneumoniae dampens the activation of inflammatory responses in host cells by targeting the activation of the NF-B canonical pathway (22,24). Our results revealed that K. pneumoniae CPS is necessary but not sufficient to attenuate inflammation (22,24). To identify additional Klebsiella factors required to dampen inflammation, we standardized and applied a high-throughput gain-of-function screen, mining a Klebsiella transposon mutant library. Using this approach, we identified 114 mutants that triggered the activation of NF-B. Follow-up validation uncovered that, in addition to the CPS, K. pneumoniae LPS OPS and the pullulanase T2SS are required for evasion of innate immune responses. Table 3 summarizes the findings obtained in this study. This study represents the first functional genomics-driven identification of previously unknown bacterial factors required to down-regulate the NF-B signaling pathway.
The high number of Klebsiella loci involved in the attenuation of NF-B activation was somewhat unanticipated. To the best of our knowledge the majority of the loci identified in our screening have never been implicated in NF-B control, and even the function of most of them is largely speculative, as it is based on in silico predictions. Nevertheless, it should be noted that the transposon mutant library examined is not saturated, hence making it possible that additional loci are needed for fully effective immune evasion. For example, in our screening we did not retrieve any cps mutant, despite the fact that CPS plays a crucial role in Klebsiella anti-immune strategies (8,20,24). However, this is not unprecedented in this type of studies; for example, cps mutants were not isolated in other Klebsiella screenings searching for virulence factors (4,55,56).
The two gene ontology categories including half of the loci identified in the screening are: metabolism and transport, 32% of the mutants, and envelope-related genes, 17% of the mutants. None of the metabolic and transport mutants displayed any growth defects under the screening conditions or in LB, hence making it unlikely that any gross growth defect underlies the activation of NF-B in infected cells. As this will be the subject of future studies, at present we speculated that their contribution to Klebsiella immune evasion is indirect and related to their impact on other Klebsiella factors. Supporting this hypothesis, the follow-up analysis of the enterobactin mutant revealed that the lack of the siderophore is linked to a reduced CPS expression, which in turn explains the NF-B activation induced by the mutant. The regulatory connection between iron levels and CPS expression has been already established (48). Because our wild-type strain encodes for two additional iron-scavenging systems, yersiniabactin and aerobactin, we sought to determine whether any Klebsiella siderophore mutant would induce NF-B activation. However, and in good agreement with Lawlor et al. (9), enterobactin was the only siderophore expressed under our infection conditions. Therefore, it was not unexpected to find that the mutants in the other two scavenging systems did not activate NF-B. Nonetheless, the evidence indicates that yersiniabactin is the Klebsiella siderophore that plays a dominant role in vivo (9), hence making possible that yersiniabactin would be the siderophore implicated in Klebsiella immune evasion during pneumonia.
The other major set of mutants were related to envelope structures, thereby adding further evidence to the critical role played by bacterial surface structures on host-pathogen interactions. Although in this work we have characterized only the role of the OPS and T2SS systems, it is worth discussing the possible contribution of the peptidoglycan and adhesion mutants to Klebsiella immune evasion. The seminal discoveries showing that NOD receptors recognize peptidoglycan motifs from Gram-negative bacteria provide a mechanistic explanation for the observation that peptidoglycan mutants induce NF-B activation (57,58). This is because these mutants secrete peptidoglycan fragments into the medium, which then activate NOD1 (59). Future efforts in our laboratory will confirm whether this is the case with our set of mutants. In addition, we will be eager to explore whether K. pneumoniae remodels its peptidoglycan to attenuate detection by NOD receptors.
In our previous studies we did demonstrate that Klebsiellacell contact is essential to attenuate NF-B activation (22). In this scenario, it can be predicted that any mutant deficient in adhesion will be affected on its ability to block NF-B activation. Of note, in our recent siRNA-based screening we uncovered the fact that Klebsiella may manipulate the ␤1-integrin-ILK signaling cascade to dampen inflammatory responses (24). ␤1-integrins are used by many pathogens as cellular receptors for attachment (60). It is then tempting to speculate that the adhesion-related mutants found in the present study might not be able to target the ␤1-integrin-ILK signaling cascade.
The LPS OPS is perhaps the most widely found factor in any screen designed to identify virulence factors of Gram-negative bacteria. However, to the best of our knowledge, ours is the first screening to demonstrate its involvement in attenuating NF-B activation. The fact that the double mutant lacking CPS and OPS induced higher responses than each of the single mutants indicates that the Klebsiella polysaccharides do not play redundant roles in host-Klebsiella interactions. Having established the role of the OPS, we next investigated whether LPS core residues may also participate in immune evasion. To investigate the contribution of core residues in an OPS-bearing strain, we used defined mutants that lack the OPS in addition to core residues, which in turn might suggest that the core residues will never be exposed in a wild-type strain. However, it should be noted that epidemiological data indicate that nearly 10% of Klebsiella clinical isolates do not express the LPS OPS (61), and therefore, core residues will not be masked by the OPS. Our results revealed that the first glucose residue of the LPS core is necessary to attenuate NF-B activation, because the SEAP levels induced by the wabM mutant, lacking also OPS, were higher than those triggered by the OPS mutant. Elimi-  NT, not tested; -, no capsule; ϳ, similar to the wild-type strain; 1, higher/stronger than the wild type, where the number of arrows gives an indication of relative levels compared with the wild type; 2, less than the wild type. nation of additional core residues did not further increase the SEAP levels.
The invasion experiments revealed unexpected features of the contribution of CPS and LPS to Klebsiella interaction with epithelial cells. Available evidence supports the notion that Klebsiella CPS is the critical factor preventing Klebsiella internalization by epithelial cells (49). Our data further corroborated this notion but also highlighted the contribution of OPS and core residues. However, in the absence of CPS, the D-galactan OPS mediates Klebsiella internalization by epithelial cells because the internalization of those mutants lacking the OPS, mutants waaL, wabM, and wabK, was reduced to the levels of the wild-type strain. Surprisingly, the internalization levels of the wabH mutant were similar to those of the CPS mutant. Structural studies showed that a core galactose residue gets exposed in this mutant (50), which in turn could account for the mutant internalization.
The contribution of the LPS core to virulence is poorly characterized in most Gram-negative pathogens, and it has been rigorously established only for Yersinia enterocolitica and Kp52145 (50,62). In a previous study we uncovered the contribution of the Klebsiella core in preventing macrophage phagocytosis (6), which, together with the evidence presented in this work, led us to put forward the notion that core residues are important for Klebsiella immune evasion. Interestingly, this might be a general feature of LPS core sections, as it has been reported recently that Brucella abortus core acts as a shield against immune recognition (63). Future studies testing additional pathogens should help to further validate this hypothesis.
In this work, we discovered that Klebsiella pullulanase T2SS is important for immune evasion in vitro and in vivo. The Klebsiella pullulanase secretion system is the archetype of T2SS, and most of the previous research efforts have been devoted to characterizing the molecular structure and function of the secretion system (54). However, its contribution to Klebsiella virulence, if any, was unknown. Perusal of the literature indicates that PulA is the only protein secreted by K. pneumoniae T2SS, although presently we cannot rigorously rule out that other proteins yet to be found are also secreted by the system. Nonetheless, the fact that SEAP levels induced by the pulA, pulC, and yacC mutants were not significantly different may support the idea that indeed the only protein secreted by the T2SS is PulA. Taking into account the evidence available for similar enzymes from other pathogens (64,65), we speculate that PulA interacts with glycan(s) located in the epithelial surface, hence leading to limited activation of the TLR4-TLR2-MyD88 pathway upon Klebsiella infection. Future efforts of the laboratory will be devoted to identify these host glycan(s).
In this work, we have demonstrated for the first time, using a pneumonia mouse model, that PulA is important for bacterial survival in the lung. Research over the last 20 years has demonstrated a correlation between the activation of inflammatory responses and Klebsiella clearance from the airways (11)(12)(13)(14). Hence the higher inflammatory response induced by the pulA mutant may contribute to mutant clearance. It should be noted that cytokines and chemokines released by epithelial cells do increase the bactericidal activity of professional phagocytes. Studies are ongoing to determine whether PulA plays any role in the interaction of K. pneumoniae with alveolar macrophages and neutrophils.
Recently, we demonstrated that K. pneumoniae exerts an anti-inflammatory effect by engaging NOD1 and EGF receptors (22,24). Bacterial removal by gentamicin treatment renders cells again responsive to pro-inflammatory mediators (22). Considering that CPS is necessary but not sufficient to attenuate inflammation (22,24), in this work we asked whether OPS and PulA could mediate this anti-inflammatory effect. However, our findings shown in Figs. 4 and 6 do not support this. These results do not contradict the other findings reported in this work showing that OPS and T2SS mutants activate inflammatory responses. It should be noted that in most of the experiments in this study gentamicin was used to remove bacteria after a 2-h infection; for example, IL-8 in the supernatants was measured at 12 h post-infection. In turn, the fact that waaL and pulA mutant-induced responses were dependent on TLR2-TLR4-MyD88 activation suggests that OPS and PulA perturb TLR-dependent recognition of K. pneumoniae. These results further corroborate in vitro and in vivo data showing that TLR2 and TLR4 play a dominant role in detecting K. pneumoniae infections (20,66,67).
On the whole, the evidence is consistent with the notion that an essential aspect of K. pneumoniae infection biology is to thwart the TLR-dependent activation of host defense responses. Our studies have revealed the concerted action of CPS, OmpA, LPS polysaccharide, and T2SS to prevent TLR-dependent responses (this work and others (8,20,21)). Further highlighting the importance of immune evasion for Klebsiella virulence, all of these mutants are attenuated in the pneumonia model (this work and others (3,8)). Therefore, we put forward the notion that a new therapeutic approach to the treatment of Klebsiella infections will be to prevent this virulence strategy, either by using molecules directed to block some of these factors or by increasing TLR-governed defense responses. Supporting the feasibility of the latter approach, the administration of the TLR ligand CpG augments anti-Klebsiella immunity in pneumonia (17).