Pseudomonas aeruginosa Flagellin and Alginate Elicit Very Distinct Gene Expression Patterns in Airway Epithelial Cells: Implications for Cystic Fibrosis Disease¹

Pseudomonas aeruginosa, a Gram-negative bacillus commonly present in the environment, acts as an opportunistic pathogen in a variety of settings (1). A number of P. aeruginosa virulence factors, including flagella, pili, LPS, quorum-sensing molecules, proteases, toxins, and others, are critical in the establishment of acute infections, as well as in chronic lung infections associated with cystic fibrosis (CF)⁴ (1, 2). This repertoire of virulence factors promotes adherence to host cells, damages host tissues, elicits inflammation, and possibly disrupts host defenses by altering gene expression in host cells (3, 4, 5). P. aeruginosa environmental strains are usually flagellated and therefore motile, in contrast to many CF isolates (6). Thus, most P. aeruginosa acute infections are by strains producing flagellin, a virulence factor that directs a proinflammatory program in epithelial and other cell types (7, 8). However, a prominent feature of P. aeruginosa strains infecting CF patients is the conversion to a mucoid,exopolysaccharide alginate-overproducing phenotype (9). This phenomenon has been associated with the establishment of the chronic P. aeruginosa respiratory infections that plague the CF patient. The overproduction of alginate by P. aeruginosa may be advantageous for the bacteria by impeding phagocytosis, and providing protection against reactive oxygen species and antibiotics (10, 11, 12). The subsequent intense neutrophil-dominated airway inflammation and progressive lung disease are major causes of morbidity and mortality in this disease (13, 14). In vivo studies suggest that clearance of mucoid strains from murine lungs is diminished compared with nonmucoid strains, indicating improved survival of alginate-producing strains in the respiratory tract (15, 16, 17, 18). Alginate enhances mucin secretion by tracheal epithelial cells (19), and may inhibit neutrophil migration to the sites of infection (20). Interestingly, the production of flagellin and alginate by P. aeruginosa are inversely regulated by the alternative sigma factor AlgT, which is a positive regulator of mucoidy and a negative regulator of flagella-mediated motility (21).

During normal growth and infection, many bacteria secrete flagellin, the structural component of the bacterial flagellum (22). In epithelial cells, flagellin from different bacterial species elicits a strong inflammatory program including IL-8 secretion, inducible NO synthase activity (23, 24, 25, 26), and induced expression of innate host defense genes, such as matrilysin and human β-defensin-2 (h-BD-2) (27, 28, 29). Furthermore, secretion of flagellin is involved in the activation of proinflammatory signaling pathways and neutrophil trans-epithelial migration (30). Other cell types, including monocytes, respond to flagellin inducing the production of proinflammatory cytokines (31). Furthermore, flagellin plays a role in triggering adaptive immune responses by stimulating chemokine secretion and migration and maturation of dendritic cells (32, 33), and by modulating T cell activation in vivo (34). Flagellin is the ligand of TLR-5 (35, 36), although additional receptors may modulate signaling (37, 38). By contrast, alginate signals through TLR-2 and TLR-4 to induce cytokine expression in monocytes and macrophages (39), but the molecular mechanisms mediating the effects of alginate on epithelial cells, which constitute a first line of defense against pathogens in the airways, are unknown.

In this work, we have examined genomic responses in airway epithelial cells exposed to isogenic motile and mucoid P. aeruginosa strains, the two phenotypes relevant in acute and chronic respiratory tract infections, respectively. The responses of airway epithelial cells to these bacterial phenotypes are qualitatively and quantitatively different. We show that infection with flagellated motile strains specifically results in the increased expression of inflammation and host defense genes. By contrast, infection with mucoid alginate-overproducing strains results in an overall attenuation of host responses to P. aeruginosa and a decrease in apoptosis. Our findings show that flagellin is a critical proinflammatory determinant of this bacterium and suggest that other factors, independently of P. aeruginosa mucoidy, may contribute to the persistent inflammation characteristic of CF.

The human lung carcinoma cell lines Calu-3 and A549 were obtained from the American Type Culture Collection (ATCC, Manassas, VA), and routinely maintained in RPMI 1640 medium supplemented with 10% FBS without antibiotics. P. aeruginosa FRD1 (mucA22) is a mucoid strain isolated from a CF patient, and produces the exopolysaccharide alginate, but not flagellin. The isogenic strain FRD440 (mucA22 algT::Tn501) produces flagellin and not alginate. The strains FRD875 (mucA22 algD::xylEaacC1) and FRD1234 (mucA22 algT::Tn501 fliC:: xylEaacC1) are nonmotile and do not produce alginate (40, 41) (Table I). All the mutations in FRD1-derived strains were generated using nonpolar cassettes to minimize effects on other genes. The mutations in fliC and algD are in gene clusters that affect the flagellar and alginate pathways only. Motile P. aeruginosa 56173 and 10145 were obtained from ATCC. Alginate-producing strains CF91, CF103, CF1025, and CF1028, are part of a collection of mucoid CF isolates used in previous studies (42). The anti-flagellin polyclonal Ab (43) was provided by Dr. A. Prince (Columbia University, New York, NY). Bacteria were routinely grown overnight at 37°C in 3% tryptic soy broth with the appropriate antibiotics. Gentamicin, FBS, and chemicals were obtained from Sigma-Aldrich (St. Louis, MO). Cathepsin G was obtained from Elastin Products (Owensville, MO).

Calu-3 human lung epithelial cells were seeded onto 6-well plates and grown to ∼90% density. Epithelial monolayers were infected with 10⁸ CFUs per milliliter of each bacterial strain for 60 min and then washed extensively with PBS. The cultures were subsequently incubated in fresh RPMI 1640 medium supplemented with 10% FBS, 100 μg/ml gentamicin, and 10 μg/ml chloramphenicol. After 6 h, total RNA from the cells was prepared with RNAzol B (Tel-Test, Friendswood, TX), and further purified with RNeasy (Qiagen, Valencia, CA) for microarray hybridization.

Adherence and invasion assays were performed essentially as described in (44). Briefly, airway epithelial cells were seeded into 6-well plates and grown to confluency. Following infection at a multiplicity of infection (MOI) of 50 for 1 h, epithelial cells were washed five times in PBS, and lysed in 0.1% Triton X-100 in distilled H₂O. Bacteria were plated on tryptic soy agar plates, incubated overnight at 37°C, and counted to determine the number of adhered bacteria. To calculate the total number of bacteria per well, sets of duplicate wells were lysed by the addition of 20 μl of Triton X-100. Bacteria in these lysates, representing the total number of bacteria present, both intra- and extracellularly, were titrated. Adherence frequencies were calculated as the number of bacteria recovered after PBS washes divided by the total number of bacteria present in each well. To determine invasion frequencies, after the 1-h initial incubation and PBS washes, cells were incubated for 4 h in the presence of 100 μg/ml gentamicin or amikacin (Sigma-Aldrich) to eliminate extracellular bacteria. Cells were washed five times with PBS, lysed in 1 ml of 0.1% Triton X-100 in distilled H₂O, and bacteria were plated on tryptic soy agar plates. Invasion frequencies were calculated as the number of bacteria surviving incubation with antibiotics divided by the total number of bacteria present just before the addition of antibiotics. Cytotoxicity was determined using a lactate dehydrogenase-based in vitro toxicology kit (Sigma-Aldrich), according to the manufacturer’s instructions. Data were analyzed by ANOVA and Bonferroni-type multiple t test. A p-value <0.01 was considered significant.

Quantification of cytochrome c release from the mitochondria was performed by enzyme-linked immunoabsorbent assay (Oncogene Research Products, San Diego, CA), according to the manufacturer’s instructions. For these experiments, Calu-3 cells were infected for 4 h at a MOI of 50 with the different P. aeruginosa strains, and total cell extracts and cytosolic fractions were prepared by differential centrifugation as described in Ref. 45 . Data were analyzed by ANOVA and Bonferroni-type multiple t test. A p-value <0.01 was considered significant. In similar experiments, cytochrome c distribution was analyzed by Western blotting analysis of cytosolic and total cell extract samples (46), with a monoclonal anti-cytochrome c Ab (BD Biosciences-Pharmingen).

Apoptosis was analyzed by annexin V binding with commercial reagents, according to the manufacturer’s instructions (R&D Systems, Minneapolis, MN). Briefly, Calu-3 cells were infected for 4 h at a MOI of 50, extensively washed, and stained with annexin V-FITC. Simultaneous staining with propidium iodide was used to detect necrosis. Staining was analyzed by using an inverted microscope (Nikon Eclipse TE2000-E; Nikon, Melville, NY) at a ×100 magnification and QCapture software (Quantitative Imaging Corporation, Burnaby, British Columbia, Canada).

Total RNA was extracted from Calu-3 cells after exposure to four different strains of P. aeruginosa: FRD1, FRD440, FRD875, and FRD1234. Briefly, biotin-labeled RNA was hybridized to Affymetrix Human Genome U133A (HG-U133A; Affymetrix, Santa Clara, CA) probe arrays, stained with streptavidin-PE conjugate (Molecular Probes, Eugene, OR), and the fluorescence intensities were measured with a laser confocal scanner (Affymetrix GeneScanner), according to the manufacturer’s instructions. Four independent infection-hybridization experiments were performed for each strain with the exception of FRD1234, which was subjected to three replicate experiments. Four independent control hybridization experiments were also performed on noninfected Calu-3 cell samples.

Raw data from the hybridization experiments was processed using the Affymetrix Microarray Suite Version 5.0 (MAS 5.0; Affymetrix) to extract transcript detection calls and intensities. All arrays were globally scaled to the same target intensity and scaling factors checked for consistency according to standard Affymetrix protocols. To smooth within-strain biological and empirical variation, the independent control and strain array data sets were analyzed as separate groups of replicates. Additional information on statistical methods is available at www.wfubmc.edu/genomics/publicationdata.htm.

We combined the individual array MAS 5.0 detection statistics (gene transcript present/absent call and p-value) into an overall statistic for each group to classify gene transcripts (i.e., probe sets on the array) as present or absent at the group level. Changes in expression levels of gene transcripts were detected through two separate tests. Gene transcripts that were detected as present in one control/strain group of arrays, but absent in another, were classified as showing absolute change. For example, a transcript was identified as significantly up-regulated in the control × FRD strain group comparison if the transcript (array probe set) was detected as absent in the control group and present in the FRD strain group (i.e., an absolute change up). Absolute down-regulation (change down) is the converse situation. Gene transcripts that were detected as present in both compared groups were classified as showing relative change. Using the presence/absence tests, we filtered the gene transcripts within each strain group to remove genes that were absent. Because absent genes have low signal intensity, this effectively also removes weakly expressed genes. We applied a two-sample t test using empirical (permutation-derived) p-values to identify gene transcripts showing a significant relative change in expression between compared strain groups. The maximum p-values for significant t tests were set to 0.0011 (control × FRD1), 0.012 (control × FRD440), 0.0055 (control × FRD875), and 0.00087 (control × FRD1234). These p-value thresholds were chosen to ensure a uniform false discovery rate (47) of 20% across all four group comparisons. Genes were mapped to probes sets and classified by molecular function (ontology classification) using EASE (http://david.niaid.nih.gov/david) and Netaffyx (http://www.affymetrix.com).

Total RNA samples were separated by electrophoresis in 1.2% agarose-formaldehyde gels, and blotted onto Hybond nylon filters (Amersham Pharmacia Biotech, Buckinghamshire, U.K.). The integrity of the RNA in the different samples was ascertained by direct visualization of the gels under UV light. Northern hybridization for matrilysin and GAPDH mRNAs was done as described before (27). For the analysis of the expression of h-BD, total RNA samples were reverse transcribed using random hexamer primers (PerkinElmer, Branchburg, NJ), and cDNAs were then amplified by PCR as described before (29). The sizes of the amplified products for h-BD-2 and h-BD-1 are 241 and 258 bp, respectively. For the analysis of Nckap1 (Nap1), cDNAs were amplified for 21 cycles using the primers and conditions described in (48). The size of the amplified product was 184 bp. Reactions were analyzed on 3% agarose gels or 6% acrylamide gels.

Calu-3 and A549 human lung epithelial cells were seeded onto 6-well plates and grown to ∼90% density. Monolayers were infected with 10⁸ CFUs of each bacterial strain (corresponding to a MOI of 50) in 1 ml of RPMI 1640 medium without serum or antibiotics, and incubated at 37°C for a period of 60 min. Conditioned medium samples from the 60-min infection period were collected, centrifuged at 10,000 × g for 10 min to remove debris, concentrated 10-fold by lyophilization, and analyzed by Western blotting with flagellin specific Abs (43), as described below. In other experiments, cells were washed extensively after infection and further incubated in the presence of antibiotics for 24 h postinfection. Conditioned media were then collected for the analysis of cytokine and chemokine secretion by enzyme-linked immunoabsorbent assay with Quantikine reagents (R&D Systems), according to the manufacturer’s instructions. Data are reported as the means ± the SD. All determinations of cytokine and chemokine secretion were done in duplicate and repeated in at least three independent experiments. Data were analyzed by ANOVA and Bonferroni-type multiple t test. A p-value <0.01 was considered significant.

P. aeruginosa flagellin was purified from overnight culture supernatants as described before (27). For some experiments, flagellin was further purified by using polymyxin B beads (Sigma-Aldrich), according to the manufacturer’s instructions. Removal of endotoxin to <0.06 endotoxin U/ml was verified by using the Limulus amebocyte lysate detection kit (BioWhittaker, Walkersville, MD).

Alginate was purified from the strain FRD1 as described in Ref. 49 , with some modifications. Briefly, P. aeruginosa FRD1 was grown overnight at 37°C in 3% tryptic soy broth medium. Following the addition of 1 vol of saline, alginate was precipitated from the supernatant of this culture (which had an obvious mucoid appearance) by the addition of an equal volume of 2% cetylpyridinium chloride (Sigma-Aldrich). After a centrifugation step at 25,000 × g for 30 min, the pellet was resuspended in the initial volume of 1 M NaCl. Finally, alginate was precipitated by the addition of 1 vol of chilled isopropanol, resuspended in PBS, and quantified in a colorimetric assay using alginic acid (Sigma-Aldrich) to plot a standard curve (50).

A549 cells were seeded in 96-well plates at ∼90% confluency in RPMI 1640 medium containing 10% FBS 24 h before the transfection. The reporter plasmid pNF-κB-Luciferase (Stratagene, La Jolla, CA) was used to analyze the effect of external stimuli on NF-κB activity in A549 cells. Transfections were conducted using 200 ng of DNA and 1 μl of Lipofectamine 2000 (Invitrogen Life Technologies, Carlsbad, CA) per well. After an incubation of 48 h to allow maximal expression of the transgene, cells were stimulated with flagellin (10⁻⁷-10⁻⁹ M) or alginate (20–40 μg/ml) for different periods of time. Cells were lysed in Glo-Lysis buffer (Promega, Madison, WI) for the analysis of luciferase expression with Bright-Glo reagents (Promega), according to the manufacturer’s instructions. Luciferase activity in each sample was measured with a Reporter Microplate Luminometer (Turner Designs, Sunnyvale, CA). Data are reported as the means ± the SD. All transfection experiments were done in triplicate and repeated at least three times. Data were analyzed by ANOVA and Bonferroni-type multiple t test. A p-value <0.01 was considered significant. Transfection efficiency was assessed by cotransfection with a plasmid containing Renilla luciferase under the control of the SV40 viral promoter (phRL; Stratagene), and did not vary >25% among the different samples. However, because the expression of this internal standard was modified by the ligands used in our experiments, Renilla expression data were not used to correct the firefly luciferase expression data (51).

Conditioned medium samples were separated on 12% SDS-polyacrylamide gels and transferred by semidry electrophoretic transfer to nitrocellulose membranes (Hybond ECL; Amersham Pharmacia Biotech), and probed with anti-flagellin Abs as described before (27, 29).

To gain insight into the molecular processes underlying the interaction of airway epithelial cells with P. aeruginosa phenotypes relevant in lung disease, we conducted a genechip analysis of airway epithelial cell responses to P. aeruginosa mucoid and motile strains. For this, we used an in vitro model of Calu-3 human lung epithelial cells infected with the alginate-producing CF isolate FRD1 and a series of isogenic mutant strains, including a motile strain that does not express alginate (FRD440), and mutants that lack the expression of alginate and flagellin (FRD875), and alginate, flagellin, and the alternative sigma factor AlgT (FRD1234) (Table I). Calu-3 cells were exposed to the different strains at a MOI of 50, and gene expression examined at 6 h postinfection. Fig. 1 summarizes the number of unique annotated genes statistically detected in each treatment group of arrays. There were a total of 612 statistically significant changes (313 up-regulated, 299 down-regulated) in gene expression in response to the motile strain, FRD440, while only 48 statistically significant changes were observed in response to the isogenic fliC mutant, FRD1234. Therefore, the presence of flagellin in the bacteria correlates with ∼500 changes in gene expression in infected airway epithelial cells. Furthermore, the conversion of P. aeruginosa to a mucoid phenotype, such as what takes place in CF (strain FRD1), resulted in only 67 statistically significant changes (39 up-regulated, 28 down-regulated) in gene expression. Remarkably, our transcriptional profile analysis shows that a nonmucoid and nonmotile strain (FRD875) regulates the expression of 231 genes in airway epithelial cells. Interestingly, strain FRD875 is a nonalginate-producing mutant derived from the mucoid isolate FRD1. The alternative sigma factor AlgT, which regulates the expression of many virulence factors by P. aeruginosa (52), is active in both strains FRD1 and FRD875 (Table I). Therefore, our results show that the production of alginate by P. aeruginosa actually attenuates the magnitude of the host response to this bacterium. Taken together, these results demonstrate that a P. aeruginosa motile phenotype has a much more extensive effect on host gene transcription than a mucoid phenotype, and suggest that flagellin and alginate direct substantially different patterns of gene expression in airway epithelial cells. Consistent with this hypothesis, the lowest number of host gene expression changes (only 48 statistically significant changes) was observed in response to the strain FRD1234, a nonmucoid, nonmotile, algT mutant strain (Fig. 1). Full data sets for all microarrays analyzed in this study are available in the National Center for Biotechnology Information Gene Expression Omnibus (NCBI GEO) database (www.ncbi.nlm.nih.gov/geo). The genomic data have the following GEO accession numbers: GSM14498-GSM14516 (GSE923, NCBI tracking system no. 15031016).

The changes in biochemical and cellular pathways in response to P. aeruginosa exposure are summarized in Table II. The most significant changes in gene expression, both qualitative and quantitative, were elicited by the motile strain FRD440. The isogenic fliC mutant, FRD1234, had no significant effect on many of these bioprocesses. Thus, flagellin specifically directs gene expression changes in pathways related to the inflammatory and innate immune responses, chemoattractant activity, cell cycle control, and response to external stimuli. By contrast, the conversion of P. aeruginosa to mucoidy results in a very limited effect on pathways coordinating the immune and inflammatory responses, as seen in the response to the strain FRD1. Furthermore, the number of genes that showed differential expression in response to the strains FRD1 and FRD875 (nonalginate producing) in the broad functional categories of protein metabolism, intracellular localization, and physiological processes illustrates the fact that the production of alginate by the mucoid strain FRD1 modulates, and in fact strongly attenuates, the extent of the host response to P. aeruginosa. Tables III and IV present a summary of changes in expression for selected genes in response to each bacterial strain. Confirming our microarray analysis, up-regulation of the chemokine CCL20, ICAM-1, and the acute phase reaction protein pentraxin 3 by flagellin has been recently described (32, 53, 54). Interestingly, the most significant up-regulating changes in response to FRD1 exposure were in genes related to apoptosis (Table III). Genes down-regulated in response to bacterial exposure were included in various cell signaling categories (Table IV).

To further characterize the interaction between P. aeruginosa and Calu-3 airway cells, we determined the adherence and invasion frequencies of the bacterial strains listed in Table I. As shown in Table V, FRD440 had a higher rate of attachment and invasion than the other strains, suggesting that the presence of flagellin and/or motility may favor these interactions. However, and more importantly, there were no significant differences in adherence and invasion between the mucoid strain FRD1 and the isogenic nonalginate-producing strain FRD875. Therefore, the data suggest that alginate does not significantly modulate these interactions of P. aeruginosa with Calu-3 cells, while very significantly affecting gene expression (Fig. 1, Table II). Furthermore, in our experimental design of 1-h infection followed by a 5-h or 23-h postinfection incubation in the presence of antibiotics, no significant cytotoxicity was caused by the different P. aeruginosa strains, as determined by lactate dehydrogenase (LDH) release (Table VI).

To further examine the regulation of gene expression by P. aeruginosa phenotypes relevant in airway infection, we exposed Calu-3 human lung epithelial cells to the strains listed in Table I. We have previously shown that bacterial exposure, and specifically Gram-negative flagellin, up-regulates the expression of matrilysin, a matrix metalloprotease involved in host defense (27, 29, 55). Exposure to the flagellated strain FRD440 resulted in a 5-fold induction in the expression of matrilysin (Fig. 2,A). By contrast, neither the alginate-producing P. aeruginosa strain FRD1 or the double mutants lacking alginate and flagellin production (FRD875 and FRD1234) had any effect on matrilysin expression in human lung carcinoma cells (Fig. 2,A, data not shown). To further explore the regulation of host defense gene expression, we examined the expression of the h-BDs, h-BD-1 and h-BD-2, in infected Calu-3 cells. As shown in Fig. 2,B, the expression of h-BD-2 is specifically up-regulated by infection with flagellated P. aeruginosa strains (ATCC 51673 and FRD440), but not by the CF isolate, alginate-producing strain FRD1, or the nonmucoid/nonmotile strains, FRD875 and FRD1234 (data not shown). The expression of h-BD-1, which is constitutive (56), did not change in response to exposure to any of these strains and thus served as an internal control. The expression of Nckap1 was induced by FRD1, but not the other strains (Fig. 2,C), in agreement with the microarray data. We examined the response of additional markers of inflammation to P. aeruginosa virulence factors in airway epithelial cells. For this, we determined the levels of secretion of IL-8 and GM-CSF, which are involved in neutrophil recruitment and survival in the airways (57) (Fig. 2,D). Infection with the flagellated strains ATCC 51673 and FRD440 resulted in a 5- to 10-fold induction in IL-8 and GM-CSF secretion at 24 h (Fig. 2,D). By contrast, infection with the alginate-producing strain FRD1, as well as the nonmotile, nonmucoid strains FRD875 and FRD1234, had no significant effect on chemokine secretion (Fig. 2,D, data not shown). Furthermore, the increased expression of host defense genes correlated with the presence of soluble flagellin released by P. aeruginosa motile strains to the supernatant of infected epithelial cells (Fig. 2,E) (27). In fact, the degree of gene expression increase in these experiments correlates with the amount of soluble flagellin detected in the supernatant of infected cells, which varies among P. aeruginosa strains (Fig. 2 F). Alginate, at a concentration between 0.1–0.5 μg/ml, was detected in the 1-h-conditioned medium of cells exposed to FRD1 (mucoid strain).

The differential regulation of gene expression by mucoid and motile phenotypes was not restricted to the CF isolate FRD1 and FRD1-derived strains. Thus, the specificity of the increase in host defense gene expression and chemokine secretion by flagellated P. aeruginosa was further confirmed by exposure of Calu-3 airway epithelial cells to a panel of mucoid and motile P. aeruginosa strains, i.e., only the flagellated strains were able to induce significantly the expression of the examined host defense genes (Fig. 3).

Taken together, our studies demonstrate that motility and mucoidy, two critical P. aeruginosa virulence phenotypes, have very distinct and specific effects on host gene expression. Our data also show that motility, but not mucoidy, is the bacterial phenotype specifically up-regulating host defense gene expression (Table II, Figs. 2 and 3). We next investigated whether the effects of exposure to the mucoid and motile strains could be reproduced by challenging the cells with purified flagellin and alginate. In these experiments, Calu-3 cells were challenged for 1 h with purified components, and gene expression examined at 6 h posttreatment. As shown in Fig. 4 (A and B), matrilysin and h-BD-2 were induced by flagellin and not alginate. Therefore, these data correlate with the expression pattern seen in cells directly exposed to motile and mucoid strains (Figs. 2 and 3). Furthermore, treatment of Calu-3 cells with purified flagellin, but not alginate, resulted in a dose-dependent increase in chemokine and cytokine secretion (Fig. 4, C and D). In fact, treatment with 10⁻⁸ M purified flagellin resulted in levels of IL-8, IL-6, CCL20 (MIP-3α), and GM-CSF secretion very similar to those observed with infection (Figs. 2, 3, and 4, data not shown). Based on the yield of our flagellin purification procedure and the amount of flagellin detected in infected cell supernatants (27) (Fig. 2, D and E), challenge of Calu-3 epithelial cells with 10⁻⁸ M flagellin is roughly equivalent to a direct infection at a MOI of 10–50. By contrast, treatment of Calu-3 cells with purified alginate at concentrations 20–80 μg/ml had no effect on host defense gene expression and chemokine and proinflammatory cytokine secretion, even for periods of treatment up to 24 h (Fig. 4). It is worth mentioning that the concentration of alginate in CF sputum varies between 4 and 100 μg/ml (58). Furthermore, treatment of flagellin with polymyxin B did not affect chemokine secretion by airway epithelial cells, suggesting that the effect is LPS-independent (data not shown). This finding confirmed our previous observation that matrilysin induction by flagellin was not inhibited by polymyxin B, and was in fact LPS-unrelated (27). Furthermore, the effect of flagellin is completely dependent on the integrity of the protein, and flagellin bioactivity is lost when the protein is specifically cleaved by neutrophil serine proteases, including cathepsin G (Fig. 4 E) (29).

We further investigated the effects of exposure to P. aeruginosa in epithelial cells from distal lung by infecting type II pneumocyte-like A549 human cells with FRD1 and the FRD1-derived strains (Fig. 5). A549 cells do not express matrilysin or defensins (Y. S. López-Boado, unpublished observations), but responded to infection with increased secretion of IL-8 upon exposure to all the strains (Fig. 5,A). However, the presence of flagellin resulted in a further 5-fold increase in the amount of IL-8 detected in the conditioned medium of infected cells (compare the levels obtained in response to the motile strain FRD440 and the corresponding isogenic fliC mutant, FRD1234), while the presence of alginate did not augment IL-8 secretion by these cells (compare the levels in response to the mucoid strain FDR1 and thecorresponding isogenic nonalginate-producing strain FRD875). These results were further confirmed by using purified flagellin and alginate (Fig. 5 B). A549 cells challenged with flagellin responded with a 20-fold increase in IL-8 secretion, while the exopolysaccharide alginate had no significant effect. Indeed, the regulation of IL-8 and other markers of inflammation (data not shown) in the alveolar-like A549 cells in response to P. aeruginosa exposure is similar to the response observed in Calu-3 cells.

We examined the effects of flagellin and alginate on NF-κB activity by using cells transiently transfected with an NF-κB reporter plasmid. As shown in Fig. 6,A, flagellin, but not alginate, was able to induce NF-κB activity in A549 airway epithelial cells. The effect was dose- and time-dependent, with maximum induction of NF-κB activity observed after 3 h of challenge with 10⁻⁷ M flagellin (Fig. 6, data not shown). By contrast, alginate (at concentrations between 20 and 40 μg/ml) was unable to induce NF-κB activity in transfected cells (Fig. 6,A). Altogether, these results indicate that flagellin, but not alginate, activates NF-κB-dependent pathways in airway epithelial cells. Consistent with these data, the induction of matrilysin and other host defense gene expression was specifically inhibited by the proteasome inhibitor MG132 (which blocks NF-κB activity), but not the p38-MAPK inhibitor SB203580 or the MEK1 inhibitor PD98059 (Fig. 6 B, data not shown).

Infection with P. aeruginosa causes apoptosis of airway epithelial cells, a mechanism involved in bacterial clearance by the host (59), which is accompanied of release of cytochrome c from the mitochondria (46). Previous studies have examined the effect of motile P. aeruginosa on apoptosis (46, 60, 61). To compare the effects of exposure to the mucoid and motile P. aeruginosa phenotypes on Calu-3 cell apoptosis, we determined the degree of cytochrome c release from the mitochondria in infected cells. As shown in Fig. 7,A, cytochrome c was released in response to P. aeruginosa infection. However, there was significantly less cytochrome c released to the cytosol in cells exposed to the mucoid strain FRD1, compared with the other strains. Similar results were observed when cytochrome c was detected by Western blotting in cytosolic and total extracts from infected cells (Fig. 7,B). Finally, we examined P. aeruginosa-induced apoptosis by performing an annexin V staining analysis of infected Calu-3 cells. Annexin V staining detects phosphatidyl serine flipped to the outer leaflet of the plasma membrane, an early apoptotic event during infection (62). As shown in Fig. 8, exposure to the motile strain FRD440 resulted in the staining of virtually all cells. However, infection with the mucoid strain FRD1 resulted in markedly reduced apoptosis compared with cells exposed to the other strains. No staining with propidium iodide was observed in infected cells (data not shown). Thus, these data strongly suggest that the production of alginate by P. aeruginosa results in less apparent apoptosis in infected cells.

In this work, we have analyzed the transcriptome of human airway epithelial cells exposed to P. aeruginosa phenotypes relevant in acute and chronic infections. Our model of coculture of human lung cells with isogenic strains of this bacterium has allowed usto identify changes in expression patterns that can be ascribed to specific P. aeruginosa virulence determinants. Thus, exposure to motile strains directs a response characterized by the increased expression in pathways related to inflammation and host defense. Furthermore, this effect is specifically orchestrated by flagellin, as demonstrated by the lack of effect of isogenic fliC mutants, and many features of these responses can be reproduced by challenging airway epithelial cells with purified flagellin. By contrast, the response of Calu-3 cells exposed to mucoid P. aeruginosa strains and purified alginate is not proinflammatory and is much more restricted in the number of genes whose expression was significantly changed, compared with motile strains. Therefore, our data show that P. aeruginosa mucoid strains, which chronically infect CF patients, do not elicit the expression of proinflammatory pathways in this model of airway epithelial cells. Interestingly, our microarray analysis showed FRD1-dependent up-regulation of genes with antiapoptotic effect on epithelial and other cell types (63, 64, 65), and exposure to an alginate-producing strain results in diminished apoptosis in Calu-3 airway epithelial cells. Thus, the lack of the appropriate host defense and inflammatory milieu in the airways, and impaired bacterial clearance because of reduced epithelial cell apoptosis (59, 66), may explain the increased persistence of these strains in animal models of acute infection (15, 16, 17, 18).

Our model of coculture of epithelial cells and genetically defined P. aeruginosa has allowed us to explore the complex bacteria-cell interactions shaping the response of Calu-3 human airway epithelial cells to phenotypes of this bacterium relevant in lung infections. Bacteria-epithelial cells interactions are determined by the virulence factors expressed by bacteria as well as the effect of these virulence factors on mammalian signaling pathways. Although differences in attachment or invasion may partially contribute to the very distinct cellular response of Calu-3 cells to the strains used in this study, our work demonstrates that P. aeruginosa mucoid and motile phenotypes direct fundamentally different responses in host cells, both in the number of genes and the type of cellular bioprocesses affected. Although our microarray analysis was limited to one time point after bacterial infection and we have used an immortalized lung epithelial cell line, our data show for the first time that the conversion to mucoidy by the bacterium correlates in this model with a fundamental switch in host gene expression patterns. Our work also underscores the extent to which proinflammatory and host defense responses to P. aeruginosa in the airways are dependent on the presence of flagellin, and not alginate. Our previous work identified flagellin as a P. aeruginosa virulence factor that specifically up-regulates host defense gene expression in airway epithelial cells (27, 29). Flagellin mutants were unable to induce the expression of the matrix metalloproteinase matrilysin both in vivo and in vitro (27). Furthermore, flagellin is a substrate for host proteases, and the outcome of the interaction with the pathogen is further modulated by the host via the specific cleavage of flagellin by neutrophil serine proteases and the subsequent inactivation of flagellin signaling (29).

In Pseudomonas spp., the alternative sigma factor AlgT is a global regulator of gene expression, which specifically modulates the expression of virulence factors (52), and inversely regulates mucoidy and flagellin expression (21). A remarkable finding of our study is the difference in the magnitude of host responses to the mucoid strain FRD1 (67 gene expression changes) and the nonmucoid isogenic strain FRD875 (231 gene expression changes). Both strains have an active AlgT, suggesting that the presence of alginate itself in the mucoid strain attenuates host responses and helps the bacterium to evade host detection. Finally, FRD1234, a nonmotile, nonmucoid, and algT mutant strain derived from FRD1, had a very limited effect on host gene expression in our experimental system. It is tempting to speculate that factors not yet determined in the CF airway milieu promote and select the conversion of P. aeruginosa to a mucoid phenotype. Thus, the AlgT-mediated P. aeruginosa conversion to a mucoid phenotype in CF serves a double purpose for the bacterium, and by simultaneously repressing flagellin synthesis and derepressing alginate production, the bacterium further favors chronic colonization of the airways. Interestingly, a recent study shows that flagellin expression in P. aeruginosa is regulated by factors present in CF airway fluid (67). Our future studies will examine the possibility that mutations in CF transmembrane conductance regulator (68) modulate the inflammatory responses of airway epithelial cells to mucoid and motile P. aeruginosa strains.

A recent analysis of the transcriptional response of airway epithelial cells exposed to P. aeruginosa suggests a role for specific type III-secreted factors in the regulation of host gene expression (4). However, because the nonmotile strain PA103 (69) was the genetic background for the generation of the type III-secretion mutations, this study did not detect a significant increase in proinflammatory and innate host defense pathways (4). A previous analysis of the interaction of P. aeruginosa with type II pneumocyte-like human airway epithelial cells (3) exposed A549 cells to the motile wild-type strain PAK (70) and the isogenic type IV pili mutant PAK-NP (a pilA mutant, defective in adherence to epithelial cells) (71). In this study, which was more limited than ours in the scope of genes interrogated, the increased expression of a limited number of these proinflammatory genes was dependent to some degree on adherence (3). Furthermore, a recent study of CF airway epithelial cells exposed to P. aeruginosa suggests that flagellin, and not pilin, is the factor responsible for cytokine gene up-regulation (5). Remarkably, our and other studies also point to the general unresponsiveness of lung and intestinal epithelial cells to LPS (3, 5, 8, 27). Altogether, the data indicate that epithelial cells, which constitute a first line of host defense and the initial barrier encountered by a pathogen, are geared to readily respond to flagellin with a program of “cell activation,” mediated by TLR-5 (35, 36). Indeed, TLR-5 is constitutively expressed by Calu-3 cells and mediates the responses to flagellin (Y. S. López-Boado, unpublished observations). Alginate can signal through TLR-2 and TLR-4 to activate monocytes and macrophages (39), and it is likely that alginate can act on other cell types relevant in CF airway disease. However, alginate seems unable to activate NF-κB in airway epithelial cells, although the expression of TLR-2, -4, and -5 in these cells has been reported (72). Our future studies will address the mechanism by which alginate affects gene expression in airway epithelial cells.

Finally, a recent study suggests that interspecies communication between the host microflora and P. aeruginosa modulates the expression of virulence factors in the latter organism (73). Remarkably, this study shows that the interaction between oropharyngeal flora and P. aeruginosa in CF results specifically in the up-regulation of fliC expression. In this context, our study suggests that the underlying cause of the exacerbations frequently observed in adult CF patients, which are not the result of acquisition of new strains (74), may be the potent proinflammatory activity of flagellin.

We thank Amanda Kuber (Wake Forest University Microarray Core), Rebecca Keyser, and Haiping Lu for technical help. We also thank Dr. Alice Prince for the anti-flagellin Ab.