Pseudomonas aeruginosa Inhibition of Flagellin-Activated NF-κB and Interleukin-8 by Human Airway Epithelial Cells

ABSTRACT Pseudomonas aeruginosa-induced activation of NF-κB and secretion of proinflammatory cytokines by airway epithelial cells require that the bacteria express flagellin. We tested whether P. aeruginosa and human airway epithelial cells secrete factors that modulated this response. Experiments were performed with both the Calu-3 cell line and primary cultures of tracheal epithelial cells. P. aeruginosa strain PAK ΔfliC (flagellin knockout) did not activate NF-κB or interleukin-8 (IL-8) but inhibited flagellin-activated NF-κB by 40 to 50% and IL-8 secretion by 20 to 25%. PAK ΔfliC also inhibited NF-κB induced by IL-1β and Toll-like receptor 2 agonist Pam3CSK4. Similar inhibitions were observed with strains PAK, PAO1, and PA14. The inhibitory factor was present in conditioned medium isolated from PAK ΔfliC or Calu-3 plus PAK ΔfliC, but it was not present in conditioned medium isolated from Calu-3 cells alone or from PAK ΔfliC that had been heat treated. Inhibition by PAK ΔfliC-conditioned medium was exerted from either the apical or the basolateral side of the epithelium, was enhanced in simple Ringer's solution over that in tissue culture medium, and did not result from altered pH or depletion of glucose. The inhibitory effect of conditioned medium was abolished by boiling and appeared from filtration studies to result from effects of a factor with a molecular mass of <3 kDa. These and further studies with isogenic mutants led to the conclusion that the NF-κB and IL-8 response of airway epithelial cells to P. aeruginosa results from a balance of proinflammatory effects of flagellin and antiinflammatory effects of a small (<3-kDa), heat-sensitive factor(s) that is not lipopolysaccharide, C12 homoserine lactone, alginate, CIF, or exotoxin A, S, T, U, or Y.

The gram-negative bacterium Pseudomonas aeruginosa commonly infects the lungs of cystic fibrosis patients and contributes to triggering the excessive inflammation in the lungs of these patients. This inflammatory response is initiated by the airway epithelial cells responding to the bacteria by activating cell signaling (including activation of the key transcription factor NF-B) and secretion of proinflammatory cytokines and chemokines (e.g., interleukin 8 ) that recruit neutrophils to the infected region (18,33). Although airway epithelial cells express many Toll-like receptors (TLR) (1,22) and could therefore be activated by many bacterial products, previous experiments have focused on the important roles of flagellin and TLR5. Activation of the airway epithelial innate host defense response by P. aeruginosa is dependent on expression of flagellin in the bacteria (1,10,30,35,36) and TLR5 in the epithelial cells (5,8,19,26).
The magnitude of this response could potentially be regulated by other factors released by the bacteria and epithelial cells. For example, flagellin triggers release of ATP from human airway epithelial cells (NCIH292) (20), and ATP synergizes with ATP in activating proinflammatory responses (6). Epithelial or neutrophil-secreted lipoxins may inhibit proinflammatory responses of these cells (13). There may also be antiinflammatory regulatory roles for bacterium-produced products. Various beneficial effects of probiotic bacteria in the intestine (9) and stomach (3) may be mediated through effects to reduce inflammation. The gram-positive bacteria Bifidobacterium breve and Streptococcus thermophilus, as well as intestinal commensal bacteria, release low-molecular-mass (Ͻ3-kDa) metabolites that reduce lipopolysaccharide (LPS)-stimulated release of tumor necrosis factor alpha by human intestinal epithelial cells (HT29-19A) (21). Recent experiments showed that P. aeruginosa-produced homoserine lactone (C12) inhibits LPS-stimulated NF-B signaling and activation of macrophages (14).
The present experiments were designed to test whether, in addition to producing and releasing proinflammatory flagellin, P. aeruginosa may also produce other pro-or antiinflammatory products. Our initial approach was to treat the human airway epithelial cell line Calu-3 with flagellin and other factors that activate proinflammatory NF-B signaling and release of IL-8 and then also in the presence of P. aeruginosa strain PAK ⌬fliC, which is defective in producing the flagellin structural gene and is therefore ineffective at activating NF-B and IL-8 (10,31). Flagellin-stimulated NF-B and IL-8 were inhibited by PAK ⌬fliC. We tested for the specificity of this inhibitory effect by determining whether PAK ⌬fliC inhibited NF-B stimulated by IL-1␤ and Pam3CSK4 (TLR2 agonist). Different strains and isogenic mutants of P. aeruginosa were used to test whether the inhibitory effect is mediated by the type III secretion system (TTSS) and TTSS toxins (32), C12 homoserine lactone (14), exotoxin A (4,34), CIF (an epoxide hydrolase; see reference 17), and alginate (23). We also tested whether the TLR4 agonist LPS modulates flagellin-stimulated activation. Finally, we tested whether P. aeruginosa or the epithelial cells were producing and secreting the inhibitory factor by testing heat-killed bacteria and supernatants or filtrates of su-pernatants isolated from bacteria or from epithelial cells exposed to the bacteria.

MATERIALS AND METHODS
Tissue culture. Calu-3 cells, a human gland epithelial cell line homozygous for the wild-type cystic fibrosis transmembrane conductance regulator (16,29), was cultured in ATCC Eagle's minimum essential medium supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin. For most experiments, cells were passaged at a 1:2 dilution and the remaining cell suspension was seeded directly onto a 24-or 12-well tissue culture plate (BD Falcon, Bedford, MA). In some experiments, cells were passaged onto either 1.0-or 4.2-cm 2 Transwell membranes (0.4-or 1-m pore size; BD Falcon) and then grown until the cells formed confluent monolayers. Cells grow equivalently on either of these filters (15). In other experiments, Calu-3 cells were grown in typical tissue culture medium and then incubated during experiments in Ringer's solution containing 145 mM NaCl, 2 mM KCl, 1.5 mM K 2 HPO 4 , 1 mM MgSO 4 , 10 mM HEPES, 2 mM CaCl 2 , and 10 mM glucose.
Human bronchial primary cultures were kindly provided by Walter E. Finkbeiner, Department of Pathology, University of California, San Francisco, San Francisco General Hospital, and cultured as previously described (31). In brief, strips of epithelium were removed from the underlying tissue and treated with protease overnight. Cells were plated on permeable filter supports (Snapwell 3407 0.4-m-pore-size polycarbonate membrane; Corning Costar, Cambridge, MA) precoated with human placental collagen (15 g/cm 2 ) at a density of ϳ10 6 /cm 2 . Cells were grown in Dulbecco modified Eagle medium/F12 culture medium supplemented with 2% Ultroser G (Biotechnics, Paris, France).
Bacterial culture and preparation of bacterial-cell-conditioned and bacterialand epithelial-cell-conditioned solutions. The strains of P. aeruginosa used were PAK, PAK ⌬fliC, PAK ⌬exoA, PAK ⌬algC, PAO1, PAO1 ⌬exoSTY, PAO1 ⌬popB, PAO1 ⌬lasI, PA14, and PA14 ⌬cif ( Table 1). All P. aeruginosa strains were grown overnight in Luria-Bertani culture medium at 37°C with shaking. We prepared these bacterial suspensions for addition to epithelial cells by washing them twice with phosphate-buffered saline (by pelleting them at ϳ16,000 ϫ g and then resuspending them in the same volume of phosphate-buffered saline) and then resuspending them in tissue culture medium minus antibiotics or in Ringer's solution (145 mM NaCl, 1.2 mM MgSO 4 , 2.4 mM K 2 HPO 4 , 0.6 mM KH 2 PO 4 , 10 mM HEPES, 10 mM glucose, pH 7.4) for addition to Calu-3 cells. The bacterial concentration was adjusted to an optical density at 600 nm of 1 (10 9 CFU/ml), and then the bacteria were diluted to 10 6 to 10 8 CFU/ml as needed. In some experiments, 10% FBS was added to Ringer's solution, while in others, glucosefree Ringer's solution was used. For experiments with heat-killed bacteria, a suspension was made of bacteria at 10 8 CFU/ml in antibiotic-free medium as described above and then heated at 85°C for 1 h, cooled to 37°C, and used with or without flagellin.
Bacterial-cell-conditioned solutions were used to test effects of bacterial-cellsecreted factors on Calu-3 cells. For experiments testing P. aeruginosa-conditioned solutions on Calu-3 cells, PAK ⌬fliC cells (10 8 CFU/ml) were suspended in medium or Ringer's solution (as described above), incubated at 37°C for 4 or 18 h, and then centrifuged at 16,000 ϫ g for 2 min. The supernatant was collected and run through a 0.2-m-pore-size syringe filter (for further clearance of bacteria) and then added to Calu-3 cells. For experiments testing P. aeruginosaepithelial-cell-conditioned solutions, Calu-3 cells were grown to confluence on 24-well plates in typical tissue culture medium. Antibiotic-free tissue culture medium or a suspension of PAK ⌬fliC (10 8 CFU/ml) in the same medium was placed on Calu-3 cells and incubated at 37°C for 4 h. After this 4-h incubation, the medium or bacterial suspension was collected, combined to minimize variation, and centrifuged and filtered as described above. For the preparation of boiled supernatant, the bacterial-cell-epithelial-cell-conditioned solution was kept in a sealed tube, placed in a boiling water bath for 15 min, and cooled to 37°C at use.
For experiments in which molecular-weight-based separation of the bacterial supernatant was performed, bacterial-cell-conditioned solution was generated by incubating PAK ⌬fliC (10 8 CFU/ml) in Ringer's solution for 18 h at 37°C and then centrifuging and filtering it as described above to remove whole bacteria. This solution then was run through molecular mass cutoff (Microcon YM-3 for 3 kDa and YM-30 for 30 kDa) filters (Millipore, Bedford, MA). Due to the volume limitation of the filters, the bacterial supernatant was separated into aliquots and filtered separately.
The filtrate (containing molecules smaller than the cutoff size) and retentate (left over on filters containing molecules larger than the cutoff size) were collected and combined. The volume of the filtrate (flowthrough) was close to that of the original solution and was then used for treatment directly with or without flagellin. The retentate (left on the filter) was diluted in Ringer's solution to a volume equal to half the original volume (to compensate for the volume lost during filtration) and used for treatment of Calu-3 cells.
Flagellin and LPS. P. aeruginosa or Salmonella enterica serovar Typhimurium flagellin (10 Ϫ4 g/ml in water; Invivogen, San Diego, CA) was stored at Ϫ80°C and diluted from the stock in the incubation medium at the stated concentrations. This solution was vortexed vigorously and heated to 37°C before being added to the solutions to ensure dispersal as monomers. Previous experiments showed that LPS contamination of this preparation is small and cannot account for the activating effects of flagellin on NF-B and IL-8 secretion. P. aeruginosa LPS (Sigma Chemical Co., St. Louis, MO) was used for some experiments by being dissolved in the medium at 10 Ϫ6 g/ml as previously done (31).
NF-B activation: luciferase adenovirus and immunoblot analysis. A recombinant adenoviral vector expressing a luciferase reporter gene driven by transcriptional activation (adv-NF-B-luc) (27) was used for studies to determine effects on NF-B activation as described previously. This vector contained the luciferase gene driven by four tandem copies of the NF-B consensus sequence and was stored in 10 mM Tris with 20% glycerol at Ϫ80°C. The virus was added to Calu-3 cells at a multiplicity of infection (MOI) of 100, and they were returned to the incubator for 24 h. Cells were then washed to remove viruses and left to grow for another day. Previous experiments showed that the adenovirus elicits expression in Ͼ75% of the cells (28).
For experiments to measure NF-B activation and IL-8 secretion, luciferaseexpressing cells were incubated in either antibiotic-free tissue culture medium (Eagle's minimum essential medium; Mediatech) or Ringer's solution as described. Treatments were performed by diluting stock solutions of reagents in corresponding medium. Normal experiments involved exposing epithelial cells to the various agonists for 4 h. Cells were then washed and processed with the luciferase assay system by using Reporter Lysis Buffer (Promega, Madison, WI) to measure NF-B-mediated transcriptional induction according to the manufacturer's protocol. Measurements of luciferase activity (relative light units) were performed in triplicate for each sample and normalized to the protein concentration (Bradford assay). Averages were then expressed relative to the average control value in the epithelial cells, which was set equal to 1.0. In experiments in which responses were somewhat variable from day to day, responses were normalized to the maximal response to the stimulant elicited in the experiment, usually 10 Ϫ6 g/ml flagellin.
NF-B signaling activity was also assayed by immunoblot analysis of NF-B p65, I〉␣, and IKK␣ and phosphorylated NFB p65, I〉␣, and IKK␣/␤. Calu-3 cells were grown in wells; incubated in Ringer's solution; treated with flagellin (10 Ϫ6 g/ml), PAK ⌬fliC (10 8 CFU/ml, MOI of 100), or flagellin plus PAK ⌬fliC for up to 4 h; and then lysed in M-PER mammalian protein extraction reagent (Pierce, Rockford, IL) containing 5 g/ml leupeptin, 5 g/ml pepstatin, 1 mM phenylmethylsulfonyl fluoride, and 50 nM calyculin A. Protein concentrations were determined with Bradford reagent (Bio-Rad, Hercules, CA). Immunoblot analysis was performed by first separating protein (10 to 50 g/lane) electrophoretically by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and subsequently transferring it to polyvinylidene difluoride membranes. Membranes were blocked (5% nonfat dried milk) in 20 mM Tris-HCl (pH 7.5)-150 mM NaCl-0.05% Tween 20 and then incubated with specific antibodies. Primary antibodies (diluted 1:1,000) for NF-B p65, phospho-p65 (serine 536), I〉␣, phospho-I〉␣ (serine 32), IKK␣, and phospho-IKK␣/␤ (serine176/180) were purchased from Cell Signaling (Danvers, MA). Blots were first probed for phosphorylated proteins. The membranes were then stripped and probed with antibodies for nonphosphorylated proteins. Immunostaining of ␤-actin was performed to control for the amount of protein in each sample. Binding of primary antibodies was visualized by enhanced chemiluminescence with horseradish peroxidase-conjugated secondary antibodies (1:2,000) and Renaissance Chemiluminescence Reagent Plus (Perkin-Elmer Life Sciences). Quantitation was performed with NIH Image on blots with the background subtracted and normalization to maximal intensities measured during the experiment. Enzyme-linked immunosorbent assay of IL-8 secretion. Samples of incubation medium were collected from plastic wells in which epithelial cells were grown. Samples from control or treated cells were collected, cleared of any P. aeruginosa or cellular debris by centrifugation (5 min, 1,000 ϫ g), stored at Ϫ20°C until use, thawed, diluted 1:100 or 1:200 in 100 l of Assay Diluent (BD Pharmingen, San Diego, CA), run in triplicate in accordance with the manufacturer's protocol (OptEIA Human IL-8 Set; BD PharMingen), and read at 450 nm with an EL X 808 Ultra Microplate Reader (Bio-Tek Instruments, Winooski, VT). Averages of the triplicates are reported.
Statistics. Quantitative comparisons were made with a t test for paired or unpaired samples, as appropriate. Differences were considered statistically significant at P Ͻ 0.05.

PAK and PAK ⌬fliC inhibit flagellin-, IL-1␤-, and Pam3CSK4activated NF-B and IL-8.
Calu-3 cells were grown on plastic plates and left untreated or treated overnight with adenovirus to express NF-B-regulated luciferase. Cells expressing NF-B-regulated luciferase were treated with different combinations of PAK, PAK ⌬fliC, and flagellin and then processed for luciferase activity measurement. Medium from some of these cells was sampled at the start of the experiment and then after 4 h for measurement of IL-8 by enzyme-linked immunosorbent assay. Results from experiments testing the dose-dependent effects of PAK, PAK ⌬fliC, and heat-killed PAK ⌬fliC on Calu-3 cell NF-B-luc activity under control conditions and during simultaneous exposure to flagellin are shown in Fig. 1A and B. PAK activated NF-B of Calu-3 cells at 10 6 CFU/ml, and there was further, apparently maximal, activation of NF-B at 10 7 and 10 8 CFU/ml. PAK ⌬fliC at 10 6 to 10 8 CFU/ml or heat-killed PAK ⌬fliC at 10 8 CFU/ml had little effect on NF-B of Calu-3 cells (Fig. 1A). Calu-3 cells that had been treated with 10 Ϫ7 g/ml flagellin (intermediate concentration for activating TLR5 [references 6 and 30]) exhibited sixfold activation of NF-B (Fig. 1B). In the presence of flagellin, adding PAK caused a slight activation of NF-B at 10 6 CFU/ml but reductions in NF-B at 10 7 and 10 8 CFU/ml (Fig. 1B). PAK ⌬fliC caused only reductions in the NF-B activity of flagellin-treated Calu-3 cells (Fig. 1B). At 10 8 CFU/ml, PAK ⌬fliC reduced flagellin-stimulated NF-B activation by about 60%. This dose of PAK ⌬fliC reduced flagellin-stimulated IL-8 secretion by both Calu-3 cells and primary cultures of human tracheal epithelial cells by about 30% (Fig. 1C). In contrast, heat-killed PAK ⌬fliC (10 8 CFU/ml) did not alter flagellinstimulated NF-B activation of Calu-3 cells (Fig. 1B). These results indicated that PAK and PAK ⌬fliC both inhibited flagellin-stimulated NF-B activity and IL-8 secretion by airway epithelial cells, and this inhibitory effect was abolished by heating the bacteria.
Further analysis of NF-B signaling was performed by Western blotting of NF-B (p65), IB␣, and IKK and the phosphorylated versions of these proteins during the treatment of Calu-3 cells with flagellin, PAK ⌬fliC, or flagellin plus PAK ⌬fliC. Typical blots are shown in Fig. 2, and quantitation of the changes is shown in Fig. 2A to F. Flagellin, but not PAK ⌬fliC, stimulated the phosphorylation of p65 ( Fig. 2A), IB␣ (Fig.  2C), and IKK (Fig. 2E), beginning within 10 min. There were also reductions in the levels of p65 (Fig. 2B), likely resulting from migration into the nucleus that occurred within the first 30 min. Levels of IB␣ decreased over 10 to 30 min but then recovered, likely resulting from its degradation and resynthesis (Fig. 2D). These results were consistent with rapid activation of NF-B signaling by flagellin but not by PAK ⌬fliC. Adding flagellin plus PAK ⌬fliC together elicited effects that were intermediate between those of flagellin and PAK ⌬fliC. The most dramatic effects of PAK ⌬fliC were to inhibit the secondphase increase in IB␣ phosphorylation and resynthesis of IB␣ that occurred at 120 and 240 min ( Fig. 2C and D). PAK ⌬fliC also inhibited p65 phosphorylation at 120 and 240 min ( Fig. 2A). These effects were consistent with PAK ⌬fliC-induced reduction of NF-B-luciferase activity (Fig. 1).
Are type III secretion and bacterial-cell-epithelium contact required for the inhibitory effect? P. aeruginosa could have mediated the inhibitory effect through the TTSS injection of exotoxin S, T, or Y into the epithelial cells. PAK does not produce ExoU, so this exotoxin was not involved. The TTSS requires bacterial contact with epithelial basolateral membranes (15). The roles of ExoS, -T, and -Y and of the TTSS apparatus were tested with strain PAO1, which, like PAK, does not produce ExoU. PAO1 and mutants that do not express exoS, exoT, or exoY (PAO1 ⌬exoSTY) or the TTSS apparatus itself (PAO1 ⌬popB; 10 8 CFU/ml) (32) caused five-to sixfold increases in Calu-3 cell NF-B activity, while PAO1, PAO1 ⌬exoSTY, or PAO1 ⌬popB inhibited flagellin-stimulated NF-B responses by about 50% (Fig. 5). Thus, PAO1 on its own activated NF-B in Calu-3 cells but inhibited flagellinstimulated NF-B. This inhibition did not require ExoS, -T, -U, or -Y or the TTSS apparatus.
Tests of the role of bacterial contact with the epithelium were performed by growing NF-B-luc-expressing Calu-3 cells on filters with 0.4-m (prevents bacterial passage) or 1.0-m (allows bacterial passage) pores (15) and exposing cells to flagellin (10 Ϫ6 g/ml) on the apical surface and also to PAK (10 8 VOL. 77, 2009 ANTIINFLAMMATORY EFFECTS OF P. AERUGINOSA ON EPITHELIA CFU/ml) on the apical or basolateral surface. As shown in Fig.  6, Calu-3 cells on both 0.4-and 1.0-m-pore-size filters responded to apical flagellin with increases in NF-B activity. The larger luciferase responses from cells grown on 1.0-versus 0.4-m-pore-size filters may have resulted from greater access of the adenovirus to the cells through the larger pores and consequently higher luciferase expression. Flagellin-stimulated NF-B activity was reduced by about 33% when PAK was added to the apical surface of these monolayers, and there was an even larger inhibition when PAK was added to the basolateral surface. This inhibitory effect was similar for Calu-3 cells grown on 0.4-and 1.0-m-pore-size filters. These results showed that PAK inhibited flagellin activation of NF-B during exposure to either the apical or the basolateral surface of the epithelium, and contact of the bacteria with the epithelium was not required for this inhibitory effect. Do bacteria secrete an antiinflammatory factor(s) into the medium? Recent experiments showed for macrophages that the quorum-sensing molecule C12 homoserine lactone was responsible for inhibiting LPS/TLR4-induced activation of NF-B and tumor necrosis factor alpha secretion (14). We therefore compared responses to flagellin, PAO1, and/or PAO1 ⌬lasI (lasI encodes the enzyme responsible for the synthesis of C12). As summarized in Table 2, flagellin (10 Ϫ6 g/ml) elicited 9.5-fold activation of NF-B by Calu-3 cells, while PAO1 and PAO1 ⌬lasI (each at 10 8 CFU/ml) activated NF-B only 2.0-to 2.5-fold. PAO1 and PAO1 ⌬lasI both reduced flagellin-stimulated NF-B activity by approximately 50%, showing that C12 was unlikely to explain the P. aeruginosainduced inhibition of flagellin-stimulated NF-B signaling in Calu-3 cells.
Similar experiments were performed to test other factors that could have been secreted or released to reduce flagellintriggered proinflammatory signaling, including CIF, which has epoxide hydrolase activity (17) and could have altered regulatory factors secreted by the epithelial cells. As summarized in Table 2, PAK and exotoxin A mutant strain PAK ⌬exoA (10 8 CFU/ml) both activated NF-B sixfold and reduced flagellinstimulated NF-B activity by about 25%. Alginate mutant PAK ⌬algC activated NF-B by only 50% and reduced the flagellin-stimulated response by about 85%. PA14 and the CIF mutant PA14 ⌬cif activated NF-B by 2.5-fold and inhibited flagellin-activated NF-B by about 60%. LPS (10 Ϫ6 g/ml, activates TLR4) had no effect on NF-B activation when added alone to Calu-3 cells (data not shown), consistent with previous data showing no effects of LPS on inflammatory responses of primary human airway epithelial cells (31), and there was similar activation of NF-B by flagellin (10 Ϫ6 g/ml) and flagellin plus LPS (data not shown). These experiments showed that ExoA, alginate, CIF, and LPS were not responsible for P. aeruginosa inhibition of flagellin-stimulated inflammatory signaling. We tested whether the inhibitory effects of P. aeruginosa on Calu-3 cells are elicited by bacterium-or epithelium-derived products and whether these products are sensitive to boiling and filtration. Flagellin elicited similar activation of NF-B in Calu-3 cells incubated in fresh medium or in medium that had been exposed to the cells for 4 h, indicating that the epithelial cells alone were not producing an inhibitory factor (data not shown). As summarized in Fig. 7, supernatant isolated from PAK ⌬fliC or from epithelial cells exposed to PAK ⌬fliC inhibited flagellin-stimulated NF-B responses of Calu-3 cells by 30 to 40%, similar to the 60% inhibition elicited by intact PAK ⌬fliC. Inhibitory effects of epithelium-plus-PAK ⌬fliC-conditioned medium were abolished by boiling. These results were consistent with the idea that PAK ⌬fliC secreted a heat-sensitive factor into the medium that was inhibitory to flagellin activation of NF-B in Calu-3 cells.
Further filtration tests were performed to determine the rough molecular size of the PAK ⌬fliC-secreted inhibitory factor. PAK ⌬fliC was incubated in Ringer's solution for 18 h, and the supernatant was isolated and added to Calu-3 cells directly  or filtered through a 3-or 30-kDa-cutoff filter before addition to Calu-3 cells. As summarized in Fig. 7, the Ͻ3-kDa filtrate potently inhibited flagellin-stimulated NF-B-activity, while the Ͼ3-kDa retentate caused no significant inhibition of flagellin-activated NF-B. Similar results were obtained with the 30-kDa-cutoff filter (n ϭ 3; data not shown). These data showed that the molecular mass of the PAK ⌬fliC-secreted inhibitory factor was lower than 3 kDa. We finally tested whether the inhibitory effect of P. aeruginosa was dependent on the composition of the epithelial incubation medium. NF-B-luc-expressing Calu-3 cells were incubated for 4 h in tissue culture medium, in Ringer's solution, or in Ringer's solution without glucose or supplemented with serum and treated with flagellin, PAK ⌬fliC, and/or supernatants from PAK ⌬fliC-exposed (10 8 CFU/ml) Calu-3 cells. Flagellin-stimulated NF-B luciferase responses (compared to those of untreated control cells) were larger for Calu-3 cells grown in tissue culture medium (19.2 Ϯ 0.8) than responses in Ringer's solution (11.0 Ϯ 0.4), but there was similar (though somewhat more pronounced) PAK ⌬fliC inhibition of Calu-3 cells in Ringer's solution (62% inhibition) versus those in tissue culture medium (31%) (data not shown; n ϭ 3). Further, supernatants isolated from Calu-3 cells incubated with PAK ⌬fliC (10 8 CFU/ml) inhibited the flagellin-activated NF-B activity of Calu-3 cells incubated in glucose-free (86% inhibition), glucose-containing (77% inhibition), or glucose-plus-FBS-containing (36% inhibition) Ringer's solution (data not shown; n ϭ 3 for each comparison). These results showed that the effect of the PAK ⌬fliC-secreted inhibitory factor did not require the presence of glucose or FBS, and this inhibitory effect of PAK ⌬fliC was similar in tissue culture medium and Ringer's solution.

DISCUSSION
A major conclusion from these studies was that multiple strains of P. aeruginosa elicited both stimulatory (proinflammatory) and inhibitory (antiinflammatory) effects on both Calu-3 and primary cultures of human airway epithelial cells: PAK, PAO1, and PA14 each activated NF-B and/or IL-8 secretion by Calu-3 cells but inhibited flagellin-stimulated NF-B/IL-8 activation. Based on the lack of stimulation by PAK ⌬fliC (Fig. 1) (10,31,36), it seems likely that the stimulatory effect of these different P. aeruginosa strains was mediated through their production and release of flagellin onto the Calu-3 cells. The smaller activation of NF-B and IL-8 by intact bacteria (up to 10 8 CFU/ml) than by 10 Ϫ7 or 10 Ϫ6 g/ml flagellin could have resulted from the bacteria releasing flagellin at concentrations below 10 Ϫ7 or 10 Ϫ6 g/ml. However, the responses to 10 Ϫ6 g/ml flagellin were larger than the responses to 10 Ϫ6 g/ml flagellin plus 10 8 CFU/ml P. aeruginosa, indicating that P. aeruginosa elicited effects that decreased flagellin/ TLR5-mediated activation of NF-B. This antiinflammatory effect was eliminated by heating the bacteria, indicating that the bacteria needed to be living to elicit their effects or that the factor was heat sensitive. Previous experiments monitoring changes in epithelial cell gene expression in response to heatkilled P. aeruginosa will have missed these antiinflammatory effects (25).
PAK-or PAK ⌬fliC-induced inhibition of inflammatory responses also occurred during treatment with IL-1␤ and the TLR2 agonist Pam3CSK4, indicating that the antiinflammatory effect of P. aeruginosa was exerted at a level besides interference with epithelial cell receptor function. It is possible that the inhibitory effects were exerted at postreceptor steps in the signaling pathways, e.g., MyD88, TRAF, etc. Further work is required to determine such mechanisms. The inhibitory effect of PAK ⌬fliC on flagellin-stimulated IL-8 secretion was smaller than that on NF-B, indicating that although there was FIG. 7. P. aeruginosa secretes a small (Ͻ3-kDa), heat sensitive factor that inhibits flagellin-stimulated NF-B activation. NF-B-luc-expressing Calu-3 cells were incubated in tissue culture medium for 4 h without treatment (control) or with flagellin (10 Ϫ6 g/ml) or flagellin plus supernatant isolated from PAK ⌬fliC (10 8 CFU/ml) (FliC Sup ϩ Flag); supernatant isolated from Calu-3 cells exposed to PAK ⌬fliC (10 8 CFU/ml) for 4 h (Epi/fliC Sup ϩ Flag); supernatant isolated from Calu-3 cells exposed to PAK ⌬fliC (10 8 CFU/ml) for 4 h, boiled for 15 min, and then cooled to 37°C (Boiled Epi/fliC Sup ϩ Flag); the same supernatant that had been filtered through a 3-kDa-cutoff filter (filtrate) (Ͻ 3kD PAKfliC Sup ϩ Flag); or the resulting retentate (Ͼ 3kD PAKfliC Sup ϩ Flag). Luciferase activities were measured and expressed relative to the magnitude of the responses to flagellin. Averages Ϯ standard deviations are shown. n ϭ 3 to 9 for each condition. The symbol # indicates significant difference from control cells. The symbol * indicates significant difference from flagellin-treated cells. Experiments performed with epithelial cells grown on largeversus small-pore filters showed that the inhibitory effect did not require contact of the bacteria with the epithelium. This effect of P. aeruginosa was therefore different from the inhibitory effects of Escherichia coli on interferon-activated signal transducer and activator of transcription 1 signaling, which required bacterial contact with the epithelial cells (12).
Instead, the present studies showed that P. aeruginosa secreted the inhibitory factor into the medium. Although the molecular identity of the factor has not been determined, studies with isogenic mutants showed that inhibitory effects did not require TTSS or TTSS-secreted exotoxin S, T, Y, or U and also did not require exotoxin A, alginate, or CIF. Filtration studies showed that the molecular mass of the bacterial-cell-secreted inhibitory factor is lower than 3 kDa. The inhibitory effect is mediated by addition to either side of the Calu-3 monolayer, indicating that the bacterial-cell-produced inhibitor may be able to permeate epithelial-cell membranes. Supernatant isolated from either PAK ⌬fliC or epithelial cells plus PAK ⌬fliC elicited similar inhibition of flagellin-stimulated NF-B, indicating that PAK ⌬fliC alone was sufficient to generate this effect. Many of these results obtained with P. aeruginosa and Calu-3 cells were similar to those elicited by lactic acid bacteria (21) or butyrate (11) in intestinal epithelial cell lines. Butyrate is unlikely to be the low-molecular-weight inhibitor observed here because the P. aeruginosa-secreted factor was inactivated by boiling. The low-molecular-weight (i.e., 297), membranepermeant, P. aeruginosa-produced quorum-sensing molecule C12 homoserine lactone inhibits NF-B in airway epithelial cells (14) and could therefore have been the antiinflammatory factor. However, PAO1 and PAO1 ⌬lasI elicited similar inhibition of flagellin-induced activation of NF-B, indicating that C12 was also unlikely to be the still-to-be identified inhibitor.
An interesting implication of our study is that the overall epithelial inflammatory response induced by P. aeruginosa may result from the different release or secretion of flagellin and also from the different production or secretion of the inhibitory factor(s). Thus, different inflammatory responses to different strains of P. aeruginosa (e.g., Table 2) may result from different amounts of flagellin and/or from different amounts of the inhibitory factor(s). Such effects may also explain the apparently higher potency of the P. aeruginosa-produced inhibitory factor when the bacteria and epithelial cells were incubated in Ringer's solution than when they were incubated in tissue culture medium. Another implication is that in intact lung airways, the composition of the airway surface liquid may affect epithelial inflammatory responses through effects on P. aeruginosa production of both flagellin and the inhibitory factor(s). The cystic fibrosis transmembrane conductance regulator plays an important role in controlling airway surface liquid composition and could therefore control proinflammatory signaling by altering the bacterial expression of both proinflammatory factors like flagellin and antiinflammatory factors that remain to be identified.