A tendency toward excessive inflammation in cystic fibrosis (CF) patients often accompanies lung infections with Pseudomonas aeruginosa. We tested the cytokine response to P. aeruginosa in two pairs of human airway epithelial cell lines matched except for CF transmembrane conductance regulator activity. The 9/HTEo−CF-phenotypic cell line produced significantly more interleukin (IL)-8, IL-6, and granulocyte-macrophage colony-stimulating factor but not regulated on activation normal T cell expressed and secreted (RANTES) in response to Pseudomonas than the 9/HTEo−control line, and the differences widened over time. Similarly, a 16HBE cell line lacking transmembrane conductance regulator activity showed enhanced IL-8 and IL-6 responses compared with the control cell line. The pharmacology of the cytokine response also differed because dexamethasone reduced cytokine production to similar levels in the matched cell lines. The protracted proinflammatory cytokine response of the CF-phenotypic cell lines suggests that the limiting mechanisms of normal cells are absent or attenuated. These results are consistent with in vivo observations in patients with CF and suggest that our novel cell lines may be useful for further investigation of the proinflammatory responses in CF airways.
- cystic fibrosis
- granulocyte-macrophage colony-stimulating factor
in cystic fibrosis(CF), the chloride transport defect in airway epithelium and submucosal glands is somehow translated into chronic bacterial infection and excessive inflammation that are the proximate causes of lung destruction and, ultimately, the death of the patient. The mechanisms underlying this translation are probably multiple. Several hypotheses have been proposed to explain the propensity to bacterial infection in the lungs, including dysfunction of bactericidal molecules in the epithelial lining fluid of abnormal electrolyte composition (25), failure of epithelial cells to ingest and thereby inactivate bacteria (20, 21), and increased binding of specific bacteria to the surface of CF airway epithelial cells (31). Recently, however, it has become clear that in addition to the propensity toward infection in CF patients, there is also tendency toward excessive inflammation. Several pieces of data support this concept. Inhibition of inflammation by high-dose ibuprofen or steroids, far from allowing the infection to progress, is beneficial to the lung disease of patients with CF (8, 13). Increased interleukin (IL)-8, a chemoattractant cytokine, has been found in the airways of young children with CF compared with that in other young children without CF but with a comparable airway burden of bacteria (17). Inflammation has been detected very early in the life in patients with CF, even those in whom no infection can be documented. Moreover, CF mice die earlier and to a greater extent from chronic Pseudomonas infection than their non-CF littermates, and this increased death rate is accompanied by increased proinflammatory cytokines in bronchoalveolar lavage fluid but not by an increased bacterial burden (30).
One possible link between the tendency toward infection and the excessive inflammation is the response of the airway epithelial cell to contact with Pseudomonas aeruginosa, the most common infecting organism in patients with CF. This organism has increased binding to CF cells in culture and also induces greater release of IL-8 from CF cells than from normal cells (7, 10). It has been suggested that this excessive response may arise from the interaction of pilin on the surface of Pseudomonas with asialo-GM1 on the surface of CF airway epithelial cells (23). In addition, the increased binding ofPseudomonas to CF airway epithelial cells has been associated with the presence of the ΔF508 CF transmembrane conductance regulator (CFTR) allele (31). To investigate this issue further, we used an airway epithelial cell line with the CF phenotype developed by overexpression of the regulatory (R) domain of CFTR. At the appropriate phosphorylation conditions and relative concentrations of R domain and native CFTR, this results in inhibition of CFTR chloride transport activity but persistent expression of CFTR, at least at the mRNA level (19). Thus these cells express the CF chloride transport phenotype but lack mutant CFTR. Our laboratory has shown that these cells have abnormal surface properties as assessed by lectin binding (14) and also by binding of Pseudomonas (2). We used these cells to investigate the cytokine response of normal and CF-phenotypic cells to intact Pseudomonas and isogenic mutants lacking pilin or pilin and flagellin. We found that the responses of CF-phenotypic cells differ from those of normal cells and that these abnormalities may contribute to the excessive inflammatory response observed in the lungs of CF patients.
To test the hypothesis that CF-phenotypic epithelial cells respond toPseudomonas infection with an increase in proinflammatory cytokine response compared with that in normal cells, we tested another pair of airway epithelial cell lines clonally derived from parental 16HBE14o− cells (4). The cell line that demonstrates a CF phenotype, confirmed by a lack of cAMP-stimulated chloride secretion in 36Cl efflux assays and no increase in short-circuit current when grown as monolayers and studied in an Ussing chamber, disrupts CFTR function through expression of a 131-bp antisense (AS) sequence to CFTR (16HBE-AS). The control cell line was transfected with the same expression vector containing the corresponding sense (S) strand to CFTR (16HBE-S) and maintained the parental chloride efflux responses to cAMP agonists. We found that IL-8 and IL-6 secretion after Pseudomonas infection in these CF-phenotypic cells is significantly greater than that in the control cells. Our results in two matched airway epithelial cell lines in which the CF phenotypes are engineered in very different ways suggest that it is the absence of CFTR function that predisposes the CF epithelium to respond to infection with exaggerated proinflammatory responses.
MATERIALS AND METHODS
pCEP and pCEP-R cell lines.
Human tracheal epithelial cells (9/HTEo−) derived from SV40-transformed human tracheal epithelial cells (kindly provided by Dieter Gruenert, University of California, San Francisco) were transfected with the LIPOFECTIN reagent (GIBCO BRL, Life Technologies, Gaithersburg, MD) with either empty vector (pCEP4, Invitrogen, San Diego, CA) or vector cloned with the R domain of CFTR (pCEP-R) as a CF-phenotypic cell line. Cells were grown in Dulbecco's minimal Eagle's medium (DMEM) supplemented with 10% fetal bovine serum and 2.5 mM l-glutamine and maintained under selection with 40 μg/ml of hygromycin at 37°C in an atmosphere of 95% air-5% CO2.
Cells transfected with pCEP-R were previously characterized and shown to overexpress the R domain while retaining expression of endogenous CFTR mRNA (19). These cells completely inhibited cAMP-stimulated chloride transport while retaining Ca2+-stimulated chloride transport. To further characterize the effect of R domain expression on CFTR processing, the R domain was coexpressed with CFTR in a heterologous expression system (HEK293 cells). Coexpression of equal molar amounts of CFTR and R domain produced no changes in the ratio of band C (fully processed) to band B (core-glycosylated) CFTR, indicating no interference by the R domain in CFTR processing (14). Overproduction of the CFTR R domain in 9/HTEo− cells also led to increased levels of asialo-GM1 and increased P. aeruginosa binding (2).
16HBE14o− antisense and sense cell lines.
To establish the matched cell lines, plasmids containing the first 131 nucleotides of human CFTR were transfected into 16HBE14o−cells (provided by D. Gruenert) in the sense (16HBE-S) and antisense (16HBE-AS) orientations and maintained under selection. Both cell lines polarized and formed tight junctions on a filter support. The CF-phenotypic 16HBE-AS cell line did not respond to cAMP agonists with increased chloride secretion, whereas the 16HBE-S cell line demonstrated a significant response (22). 16HBE-S and 16HBE-AS cells were maintained in DMEM supplemented with 10% fetal bovine serum, 2 mM l-glutamine, and 200 μg/ml of G-418.
The nonmucoid laboratory isolate PAO1 and the isogenic derivative strains used for cytokine assays were kindly provided by Alice Prince (Columbia University, New York, NY). The phenotypic properties of an isogenic strain of PAO1 lacking pili (PA/NP) and an isogenic strain lacking both pili and flagella (PA/NP/fliA−), unrelated to the expression of pilin and flagellin, were the same as those of the parental strain (28). The Pseudomonas strains were grown in trypticase soy broth (TSB) overnight at 37°C with shaking.
Tumor necrosis factor-α (TNF-α) and Il-1β (Sigma, St. Louis, MO) were used at concentrations of 100 ng/ml in Hanks' balanced salt solution (HBSS). Water-soluble cyclodextrin-encapsulated dexamethasone (Sigma) was used at a concentration of 3 μg/ml.Escherichia coli and Pseudomonaslipopolysaccharides (LPSs) were also from Sigma and were used at 10 μg/ml to 1 mg/ml. Neutrophil elastase (Elastin Products, Owensville, MO) was used at 10 μg/ml.
Stimulation of Cytokine Production by Cells
9/HTEo− cells transfected with either empty vector (pCEP4) or vector with the R domain of CFTR (pCEP-R) were plated at a density of 1 × 106 cells/well on vitrogen-coated 24-well plates. Twenty-four hours after being plated, the cells were switched to serum-free medium for 18 h.
16HBE14o− cells transfected with the pBKCMV vector expressing CFTR sense (16HBE-S) or antisense (16HBE-AS) were plated at an original density of 1 × 106 cells on Millicell-HA 0.45-μm filters (12-mm diameter; Millipore, Bedford, MA) as inserts in 24-well plates and grown for 12–14 days until tight monolayers were established. Both apical and basolateral surfaces were bathed in 0.5 ml of medium. The cells were incubated in serum-free medium for 18 h before the experiments. The permeability of the monolayer was determined at the beginning and end of the experiment by applying 20 μg/ml of FITC-inulin (Sigma) to the apical side of the filter and measuring the percentage that passed through to the basolateral solution in 2 h. Less than 1.5% of the amount applied to the apical surface was recovered from the basolateral medium. In addition, under experimental conditions, known concentrations of human IL-8 (Sigma) were added to unstimulated triplicate filters, and the change in concentration of IL-8 in the basolateral compartment compared with that in nonspiked control cells was also determined. Not more than 6–7% appeared in the basolateral compartment.
P. aeruginosa strains PAO1, PA/NP, and PA/NP/fliA− were grown in TSB for ∼18 h until there was an optical density at 600 nm of ∼1.0 or ∼1 × 109colony-forming units (cfu)/ml (determined by dilution plating). Aliquots of the overnight cultures were sedimented and washed twice by resuspension in HBSS (GIBCO BRL) to a final dilution in HBSS of the appropriate colony-forming units per milliliter. Washed bacterial aliquots (0.5 ml/well) were then immediately incubated for 60 min, unless otherwise indicated, with the confluent monolayers of epithelial cells at 37°C. The doses of bacteria indicated are per well of cells. There was no growth of Pseudomonas in the serum-free culture medium at 24 or 48 h after experimental treatment as tested by plating aliquots of the medium on TSB plates. Nontreated control wells were processed similarly. For 16HBE14o− cells on filters,Pseudomonas and other treatments were applied in 0.5 ml of HBSS to the apical surface only. As a positive control, the cells were stimulated for 1 h with IL-1β (100 ng/ml) and TNF-α (100 ng/ml; both from Sigma). The cell monolayers were then washed three times in HBSS followed by incubation for the indicated time points in 0.5 ml of serum-free cell culture medium containing 100 μg/ml of gentamicin.
To test for the ability of killed Pseudomonas to stimulate cytokine production, 109 cfu/ml of PAO1, PA/NP, or PA/NP/fliA− were either killed by heating to 65°C for 30 min or lightly fixed in 0.5% glutaraldehyde for 30 min and washed. Aliquots of the bacterial suspensions were prepared on grids for negative staining and electron microscopy. Triplicate wells from each experiment were also fixed in 1% paraformaldehyde after being washed for visualization by electron microscopy.
Collected medium was assayed for human cytokine IL-8, IL-6, granulocyte-macrophage colony-stimulating factor (GM-CSF), and regulated on activation normal T cell expressed and secreted (RANTES) concentrations with enzyme-linked immunosorbent assay (ELISA) kits (R&D Systems, Minneapolis, MN). Epithelial cell monolayers were lysed with 300 μl of 1× lysis buffer (Promega, Madison, WI), and protein concentration was determined with the bicinchoninic acid method (Pierce). The protein concentration displayed a linear relationship to cell number over the range of 105 to 2 × 106 cells. Each treatment was performed in triplicate wells, and duplicates of each sample were assayed. To test for cytokine release, IL-8 concentration was measured 24 h after PAO1 stimulation in the medium and epithelial cell lysates. To combine multiple experiments, the secreted cytokine concentration (in pg/mg protein) of 109 cfu of PAO1-stimulated pCEP-R cells at 24 h was set to 100% for each experiment, and other concentrations are expressed relative to this value.
Cell Viability Assays
The viability of pCEP, pCEP-R, 16HBE-S and 16HBE-AS cells was assessed by trypan blue exclusion and by measuring the release of the cytoplasmic enzyme lactate dehydrogenase (LDH) from cells with compromised membrane permeability before and after PAO1 or TNF-α-IL-1β treatments. The supernatants were recovered, and LDH activity was determined with a LDH kit (Sigma) following the manufacturer's instructions. Control cell lysates were measured for LDH activity. LDH released by stimulated and nonstimulated cells is expressed as (activity in supernatants from experimental wells/LDH activity in control cell lysates) × 100. Cytotoxicity as determined by trypan blue exclusion assays was performed by incubating the cells with 0.4% trypan blue (Sigma) in PBS and assessing the percentage of stained cells within each well by microscopy and cell counting.
Where applicable, statistical analysis of results and differences was determined by SigmaStat (SPSS, San Rafael, CA) with Student's t-test; if equal variance testing failed, the Mann-Whitney rank sum test was applied. Results shown are means ± SE. Results were considered significant when P ≤ 0.05.
Pseudomonas Strains and Cytokine Secretion From pCEP and pCEP-R Cells
We performed multiple experiments to measure the proinflammatory response of pCEP (control) and pCEP-R (CF-phenotypic) epithelial cells to the most common bacterial colonizer in CF lungs, P. aeruginosa, by incubating serial dilutions of three isogenic strains of the laboratory isolate PAO1 (PAO1, PA/NP, and PA/NP/fliA−) for 1 h with 1 × 106epithelial cells/well. After the initial exposure, the cells were washed and incubated in serum-free medium containing gentamicin for 24 h. Concentrations of cytokines secreted into the medium were assayed by ELISA. As a positive control, cells were stimulated with TNF-α-IL-1β, also for 1 h. Baseline cytokine secretion was measured from wells treated with similar procedures without bacteria or cytokines. Cell viability was checked at the conclusion of some experiments by measuring release of LDH and by trypan blue exclusion. Untreated cells or cells stimulated with TNF-α-IL-1β had a viability of ∼95% by either criterion. Cells incubated with 109 cfu of PAO1 had a viability of ∼93 (pCEP-R) and 88% (pCEP).
Figure 1 illustrates one representative experiment (of two similar experiments) that measured the potent proinflammatory chemokine for neutrophils, IL-8, 24 h after an initial exposure of 1 h to various strains ofPseudomonas. Both cell lines showed a dose-dependent release of IL-8 in response to PAO1 (Fig. 1 A) from 0.1 to ∼100 bacteria/epithelial cell, a saturating concentration for the available asialo-GM1 receptors on the cell surface (2,10). The IL-8 response of the CF-phenotypic or pCEP-R cells was significantly greater than that of the control cell line at all bacterial dilutions that elicit responses above background (109 and 108 cfu, P ≤ 0.001; 107 cfu, P = 0.004; 106 cfu, not significant). The Pseudomonas strain that lacks pili (PA/NP), a major adherence component to asialo-GM1 cell surface receptors, also provoked IL-8 secretion in a dose-dependent response to levels ∼30–45% of that induced by intact PAO1 (Fig.1B). The CF-phenotypic cells still showed a significant increase over control cells (109 cfu, P = 0.015; 108 cfu, not significant; 107 cfu,P = 0.016; 106 cfu, not significant). When both pili and flagella were eliminated in the isogenicPseudomonas strain PA/NP/fliA− (Fig.1 C), very little IL-8 was secreted (only 3–8% of the PAO1-stimulated values), but still pCEP-R cells produced significantly more (109 cfu, P = 0.037; 108cfu, P = 0.016; 107 cfu, not significant).
Figure 2 shows the results from the same experiment for the pleiotropic inflammatory mediator IL-6. Similar to the PAO1-stimulated IL-8 response, both cell lines exhibited a dose-dependent increase in IL-6 secretion measured 24 h after the initial exposure, with significant increases in the pCEP-R cell line compared with the pCEP control cell line (109 and 108 cfu, P ≤ 0.001; 107 cfu, not significant; Fig. 2 A). The PA/NP strain also evoked a dose-dependent response, which reached 25–45% of PAO1-stimulated values and was significantly higher in the CF-phenotypic cells at 109 cfu (Fig. 2 B). There was no IL-6 secretion above that in nonstimulated control wells for either cell line with the PA/NP/fliA− (Fig. 2 C). Although, for the particular experiment shown in Figs. 1 and 2, it appears that the constitutive IL-8 and IL-6 secretion (the “none” condition) was greater from pCEP-R cells, combined results from 13 experiments showed no significant difference between the cell lines in unstimulated conditions (IL-8: pCEP, 3.6 ± 0.9 pg/mg protein; pCEP-R, 3.1 ± 1.0 pg/mg protein; IL-6: pCEP, 5.2 ± 1.7 pg/mg protein; pCEP-R, 3.9 ± 0.9 pg/mg protein).
To determine whether the differences in stimulated IL-8 secretion between the cell lines were due to differences in cytokine release rather than production, we compared intracellular and secreted concentrations of IL-8 before and after PAO1 stimulation. Intracellular IL-8 was 8–12% of the secreted IL-8 in both cell lines, indicating that there was no difference in the ability to secrete the cytokine after stimulation.
These results taken together demonstrate that the CF-phenotypic cells responded differently to the same Pseudomonas inoculum, and although pili were a major contributor to the IL-8 and IL-6 secretion in response to Pseudomonas stimulation, factors other than the asialo-GM1-pili interaction must also be important in eliciting the inflammatory response as well as the differential response between the CF-phenotypic and non-CF-phenotypic cells.
To examine the components of PAO1 necessary for this response, we tested LPS, which stimulates nuclear factor-κB activation in some airway epithelial cell systems (6). NeitherPseudomonas nor E. coli LPS in concentrations from 10 μg/ml to 1 mg/ml stimulated IL-8 and IL-6 release in these cell lines. The requirement for motile Pseudomonas with pili for proper adherence and signaling was examined by killing the bacteria by heating to 65°C for 30 min before exposure. Heat-killed PAO1 did not evoke a proinflammatory cytokine response even at the highest concentration applied, 109 cfu (data not shown). Heat-killed PAO1 prepared for electron microscopy with negative staining showed that the bacteria had lost most flagella and appeared shriveled compared with non-heat-treated but lightly glutaraldehyde-fixed bacteria (data not shown). Mild fixation of PAO1 preserved morphology, including the presence of flagella. Application of 109 cfu of these fixed bacteria to the epithelial cells for 1 h also did not elicit an IL-8 response (data not shown). In contrast, even 10 min of exposure of epithelial cells to live PAO1 (109 cfu) was sufficient to elicit an IL-8 but not an IL-6 response in both pCEP and pCEP-R cells assayed 24 h after the initial exposure. The IL-8 response was much less after 10 min of exposure compared with a 1-h PAO1 exposure (∼1,400 pg/mg protein compared with up to 50,000 pg/mg protein for 1 h of exposure) and was not significantly different between the cell lines (data not shown). These data taken together indicate that the killing of thePseudomonas was effective and that any killed bacteria retained on the monolayer had a much reduced effectiveness in stimulating cytokine production.
Time Course of Cytokine Secretion in pCEP and pCEP-R Cells
We measured the kinetics of the secreted cytokines 6, 12, 24, 36, and 48 h after an initial incubation with 109 cfu of PAO1 or with TNF-α-IL-1β in four separate experiments (Figs. 3 and4, respectively). To combine data from different experiments, concentrations were normalized by setting the pCEP-R-secreted concentration at 24 h to “100%” within each experiment. After a 1-h incubation of epithelial cells with 109 cfu of wild-type PAO1, IL-8 accumulation continued for 48 h in the CF-phenotypic cells but appeared to slow in the pCEP control cells at 24 h (Fig. 3 A). The least difference in IL-8 secretion between the cell lines was at the earliest time points, but the difference widened over time, reaching significance at 24 h (P = 0.009). The production of IL-6 after PAO1 exposure also continued to increase over 48 h for pCEP-R cells, whereas the normal cell line almost ceased further secretion after the first 6 h (Fig. 3 B). For IL-6, the differential secretion between the cell lines was significant even at the earliest time point (P = 0.003). When known concentrations of exogenous IL-8 and IL-6 were added to the medium, 90% or more was recovered at 48 h (data not shown), so the protein was stable in the medium in the presence of the cells. Therefore, the values in Fig. 3 represent cumulative secretion.
We also examined the time course of production of GM-CSF, an epithelial cell-derived neutrophil chemoattractant that enhances neutrophil survival by delaying apoptosis. (3, 29). There was no increase in GM-CSF secretion after 6 h from normal cells, but there was a continuous secretion of GM-CSF from the CF-phenotypic cells over 24 h, accentuating the differences in the epithelial cell responses over time.
The mixture of TNF-α and IL-1β, potent proinflammatory cytokines, stimulated production of IL-8, IL-6, and GM-CSF in both cell lines. For IL-8, the only significant difference between the cell lines occurred at the 24-h time point (results of 13 combined experiments normalized to pCEP-R PAO1-stimulated values: pCEP, 106 ± 11.037; pCEP-R, 189.87 ± 23.544; P = 0.013; Fig.4 A), and there was no continued secretion from the CF cells between 24 and 48 h. IL-6 secretion from the CF-like cells was significantly increased at all time points after TNF-α-IL-1β stimulation (6, 24, 36, and 48 h, P ≤ 0.001; 12 h, P = 0.010), and there was no further accumulation of IL-6 from the normal cells after 6 h (Fig. 4 B). GM-CSF secretion continued to increase between 6 and 24 h from pCEP-R cells, yet the pCEP cells did not continue to secrete this cytokine 6 h after treatment, resulting in a very significant difference at 24 h (P ≤ 0.001; Fig. 4 C).
RANTES is a potent chemoattractant produced by lung epithelial cells for various important inflammatory cells, including eosinophils, and is involved in the pathophysiology of inflammation-associated lung injury (26). We found that neither the pCEP nor pCEP-R cell line responded with RANTES secretion when stimulated with PAO1 (Fig.5), yet TNF-α and IL-1β elicited a response that did not differ between the CF-phenotypic and non-CF cell lines (n = 3 experiments).
Glucocorticoid Inhibition of IL-8 and IL-6 Secretion of pCEP and pCEP-R Cells
Glucocorticoids exert multiple anti-inflammatory activities, including the inhibition of lymphocyte migration and inhibition of transcription of cytokines dependent on nuclear factor-κB activation (5). To investigate the pharmacology of glucocorticoid activity in our cell lines, we applied dexamethasone in the serum-free medium after the 1-h Pseudomonas exposure. Dexamethasone significantly abrogated the excess IL-8 secretion of the CF-phenotypic cell lines in response to 109 cfu of PAO1 beginning at 12 h, effectively reducing the response to the levels of normal cells (Fig. 3 A). Dexamethasone inhibited IL-6 secretion of pCEP-R cells after PAO1 exposure but did not reduce it to normal levels (Fig. 3 B).
TNF-α-IL-1β-stimulated IL-8 responses were inhibited by dexamethasone at all time points observed (Fig. 4 A), inhibiting the secretion by both normal and CF-phenotypic cells to approximately the same level, eliminating the excess secretion from pCEP-R cells. Dexamethasone did not inhibit the IL-6 response of normal cells to TNF-α-IL-1β stimulation but inhibited the IL-6 secretion from the pCEP-R cells so that over time there was no increase in secretion (Fig. 4 B).
16HBE14o− Antisense and Sense Cell Lines and Cytokine Production
16HBE14o− antisense cells are a human bronchial epithelial cell line made to assume a CF phenotype by stable transfection of antisense oligonucleotides to CFTR. The matched control cell line was transfected with sense oligonucleotides to CFTR (16HBE14o− sense) and retained a cAMP-stimulated chloride efflux as previously reported (22). We tested these two cell lines, which polarize and form tight junctions when grown on filters, for differences in proinflammatory cytokine production in response to PAO1 and TNF-α-IL-1β. When 109 cfu of PAO1 were incubated on the apical surface for 1 h and IL-8 was measured from both the basolateral and apical media 24 h after the exposure, the combined results of four separate experiments showed a significant increase in secretion of IL-8 (P ≤ 0.001) and IL-6 (P = 0.038) from the CF-phenotypic cells (Fig.6 A). We found no significant difference between the sense and antisense lines in GM-CSF or RANTES production when stimulated with Pseudomonas (Fig.6 A). At the conclusion of the experiments, viability of the cells was checked with LDH release and trypan blue exclusion. PAO1 or TNF-α-IL-1β treatment did not increase the amount of released LDH in either polarized cell line (1–5%).
Apical application of TNF-α-IL-1β stimulated similar amounts of IL-8, IL-6, and GM-CSF secretion from the sense and antisense lines measured 24 h after the original induction, although this stimulated release was less than that from PAO1 stimulation. However, the TNF-α-IL-1β stimulation of RANTES was significantly greater in normal cells (P ≤ 0.001; Fig. 6 B).
Glucocorticoid Inhibition of Cytokine Secretion
Dexamethasone significantly inhibited IL-8, IL-6, and GM-CSF release from 16HBE14o− antisense and 16HBE14o− sense cell lines when applied to the apical cell surface after a 1-h incubation with PAO1, preventing the cytokine responses almost completely (Fig. 6 A).
Polarity of Cytokine Secretion
We did not observe consistent excess cytokine secretion to either surface before or after stimulation. Ratios of apical to basolateral secretion did not differ significantly from 1. However, one might speculate that in the native condition in vivo, where apical surface fluid volume is very small, higher local concentrations of cytokines might result on the apical surface if the same amount of cytokine was secreted in both directions. Because, in our culture model, tight monolayers do not permit transepithelial passage of cytokines, one might speculate that such high concentrations in airway surface liquid might be sustained.
Airway epithelial cells of the CF phenotype produce significantly more IL-8, IL-6, and GM-CSF in response to several different isogenic mutants of the laboratory strain of Pseudomonas PAO1, including those lacking pilin and both pilin and flagellin, than comparable normal cells. The time course of response differs between the normal and CF-phenotypic cells. For these three cytokines, the CF-phenotypic cells continued to increase production long after the normal cells had ceased or slowed production. With a low-level or brief stimulus or at early time points, the CF phenotype cells had a similar cytokine production compared with that in the normal cells. However, with more sustained stimuli or longer response times, the CF-phenotypic cell lines continued to produce large quantities of cytokines well after the normal cells had ceased or slowed production.
One possible explanation for the differences we observed is that the CF cell lines bind Pseudomonas to a greater extent than normal cells via pilin-asialo-GM1 ligation, and it is the activation of signal transduction through this pathway that produces the excessive response to Pseudomonas in the CF-phenotypic cell lines. Indeed, stimulation of either normal or CF-phenotypic cell lines with mutants of PAO1 lacking pilin produces less than half the cytokine response of intact PAO1, and the mutants lacking pilin and flagellin produce still less (3–8%) IL-8 and no IL-6 at all. However, if the pilin-asialo-GM1 interaction were the sole cause of the excess cytokine production in CF-phenotypic cells, then stimulation with PAO1 mutants lacking pilin, which bind to pCEP and pCEP-R cells to the same extent (2), should stimulate cytokine production to a similar extent in normal and CF-phenotypic cells. Actually, pilin-negative and pilin- and flagellin-negative mutants stimulate production of greater amounts of IL-8 and IL-6 in the CF-phenotypic cells compared with normal cells. Thus there must be factors in addition to excessive asialo-GM1 stimulation that account for the excess cytokine production in response toPseudomonas in the CF cells. Because the CF-phenotypic and control cell lines do not differ with respect to cytokine production without stimulation, the differences must reside in the response to stimulation. Although the differences are most evident (for IL-8 at least) in the response to Pseudomonas, significant differences are also observed in the cytokine responses to TNF-α-IL-1β, indicating that pilin-asialo-GM1 ligation is not the only reason for the excessive cytokine response of the CF cells.
The temporal pattern of cytokine production is different in the normal and CF-phenotypic cell lines. Although at early time points normal and CF-phenotypic cell lines produce similar amounts of IL-8, by 24 h, the CF-phenotypic cell lines are clearly producing more. At later time points, the IL-8 accumulation in CF-phenotypic cells continues, but it levels off in the normal cells. This phenomenon is not due to increased intracellular retention of IL-8 in the normal cells because intracellular IL-8 represents ∼10% of the total in both cell lines. Similarly, IL-6 and GM-CSF production in normal cells is maximal by only 6 h but continues to accumulate in CF cells over 24–48 h.
It is possible to limit the accumulation of inflammatory cytokines pharmacologically. Dexamethasone inhibits TNF-α-stimulated IL-6, GM-CSF, and IL-8 accumulation in both normal and CF cell lines, but when the cytokines are stimulated with PAO1, dexamethasone has a greater effect on CF-phenotypic cells. These data suggest that the pharmacology of the TNF-α-IL-1β response is similar in normal and CF-phenotypic cells but that the pharmacology of the PAO1 response is different. Different signaling pathways may be activated or the mix of pathways may be different in normal and CF cell lines.
Our observations in vitro are consistent with several clinical observations. Our finding that the increased cytokine production in response to Pseudomonas is sustained much longer in CF-phenotypic cells than in non-CF cells is consistent with a report (12) that infants with CF, even those with negative bacterial cultures, have elevated IL-8 in bronchoalveolar lavage fluid. It is possible that these patients' cytokine response to a previous infection persists for a very long time, well after the stimulus is removed (12, 18), just as we observed in vitro. Our observation that bacterial stimulus induces a much greater cytokine response from CF epithelial cells than from normal epithelial cells is consistent with the observations of Noah et al. (18) that CF children have higher IL-8 in bronchoalveolar lavage fluid for any given lung burden of organisms. The excess response of both the chemoattractant IL-8 and the neutrophil survival factor GM-CSF in CF airway epithelial cells might contribute to the exuberant neutrophil accumulation in the CF airway, and the relative lack of RANTES contributes to the relative paucity of eosinophils at this site.
In vitro studies of inflammatory cytokine production in normal and CF airway epithelial cells have had mixed results. Although many studies, especially those with cells from submucosal glands (11,27), reported increased IL-8 production in response to various stimuli (or even at the basal state) in CF compared with normal samples, another study (24) found similar cytokine profiles in normal and CF cell lines, and in a recent report (16), CF cells are reportedly deficient in cytokine production. Our results may shed some light on these differing results because they show that distinguishing between normal and CF-phenotypic cell lines depends on the specific stimulus, the intensity of the stimulus, and the time after the stimulus is applied as well as the specific cytokine under study. For IL-8, there is no significant difference between normal and CF-phenotypic cells in the response to PAO1 at early time points, but by 24 h and later after exposure, the CF-phenotypic cells clearly have a marked increase in IL-8 production compared with normal cells. A briefer duration of PAO1 exposure and lower concentrations of bacteria applied to the cells also make it more difficult to reliably detect differences between normal and CF-phenotypic cells. For IL-8, differences between CF and non-CF cell lines in response to TNF-α-IL-1β are significant only at the 24-h time point under the conditions of the experiment. In contrast, for IL-6 and GM-CSF, differences between normal and CF-phenotypic cells are apparent at the 6-h time point for both PAO1 and TNF-α-IL-1β stimulation. In our studies, RANTES was not stimulated by exposure to PAO1 but was stimulated by TNF-α-IL-1β in both normal and CF-phenotypic cells. Although the CF cell lines produced less, the difference was significant only for 16HBEo− cells. These data are in general agreement with the report of Schweibert et al. (24).
Another possible explanation for the variability of the earlier results is the use of cell lines that differ in more than just CFTR expression or function to make the comparisons. Our cell lines were constructed from a SV40-transformed normal human tracheal epithelial cell line, 9/HTEo−, or a human bronchial epithelial cell line, 16HBEo−. For the 9/HTEo− cells, transfection with empty vector (pCEP4) did not alter the chloride transport properties of the cells, although the clonal selection process did make them more uniform. Transfection with pCEP-R, an episomal expression vector containing the R domain of CFTR, completely eliminated cAMP-stimulated chloride transport (which is also eliminated by antisense to CFTR) (19). In the planar lipid bilayer system, the R domain interacts with wild-type CFTR to prevent chloride transport, which seems the mostly likely mechanism for this effect in the 9/HTEo− cells (15). The mRNA for CFTR continues to be expressed in R domain-transfected cells, and cotransfection experiments in a heterologous system indicate that CFTR processing is unaffected by free R domain protein (14,19). Because pCEP4 is an episomal, nonintegrating vector, insertional activation or inactivation of particular genes is unlikely. Thus the cell lines we studied are likely well matched except for the function of CFTR, and the CF-phenotypic cell line continues to express and process its native CFTR normally. The 16HBEo− cell lines were transfected with sense or antisense to CFTR and selected for presence of the plasmid by PCR, Southern blot, appropriate CFTR physiology by 36Cl efflux, and Ussing chamber experiments. Thus they are likely to be quite similar except for CFTR activity. One study (24) used a CF cell line (IB-3) corrected with CFTR with an NH2-terminal truncation (C38 cells). Although this mutant restored chloride transport, protein interactions involving the NH2 terminus were lost. Another study (1) has been done with corrected CF cell lines that may overexpress CFTR and not be representative of the native state, with cell lines immortalized from different people whose genetic complements outside the CFTR genotype may differ profoundly or with primary cell lines from different people. Indeed, polymorphisms in the promoter regions of some cytokine genes, which cause profound changes in the amount of cytokine produced, have been proposed as modifiers of the CF phenotype (9). It is thus especially important to control the other genes in the cell lines to be compared if the impact of CFTR activity is to be assessed.
Although our two cell line pairs have advantages as CF models, they also illustrate that cytokine responses may be cell line specific. For example, although there is a brisk GM-CSF response in the 9/HTEo− cell lines and the CF-phenotypic cell line outstrips the normal in the 16HBEo− lines, the response is quantitatively less, and there is no difference between CF and non-CF cell lines. Model systems can be selected for their suitability to study relevant activities; concordance between different models gives more confidence in the conclusions.
Our results in these two cell lines also suggest that it is not misprocessing of mutant CFTR that entrains the excess cytokine production in CF airway epithelial cells. In neither of these cell lines is mutant CFTR present. Expression of the R domain does not increase endoplasmic reticulum retention of CFTR in a heterologous system (14). Thus our data suggest that the excess cytokine responses are a consequence of impaired cAMP-dependent chloride transport.
We conclude that two different airway epithelial cell lines in culture, remote from the inflammatory milieu of the CF airway, which have been converted to CF-phenotypic cells either by overexpression of the R domain or by expression of antisense constructs, have excessive cytokine responses to bacterial stimulation. These observations in engineered cell lines with no mutant CFTR make it likely that the excessive cytokine responses are related to impaired CFTR activity and not to a particular CF genotype. Moreover, one CF-phenotypic cell line expresses wild-type CFTR, making it unlikely that the absence of protein interactions with wild-type CFTR is critical for excess cytokine response. The pattern of response suggests that mechanisms that limit the inflammatory response in normal cells are absent or attenuated in the CF cell lines but that anti-inflammatory agents such as steroids, which are moderately effective in patients with CF in limiting the inflammatory response, are effective in this cell model as well. These well-matched cell systems may be useful for further dissection of the proinflammatory responses in CF airways and for evaluating potential therapeutic agents.
We thank Alice Prince for the Pseudomonas aeruginosastrains, Aura Perez for the 9/HTEo− cell lines, and our Cystic Fibrosis Center Inflammatory Mediator Core for expert technical services.
This work supported by National Heart, Lung, and Blood Institute Grants HL/DK-49003 and HL-60293 and National Institute of Diabetes and Digestive and Kidney Diseases Grant DK-27651.
Address for reprint requests and other correspondence: D. Kube, CWRU Dept of Pediatrics, BRB 835, 2109 Adelbert Rd., Cleveland, OH 44106-4948 (E-mail:).
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2001 the American Physiological Society