Tumor Necrosis Factor Modulates Airway Smooth Muscle Function via ...

5 downloads 2713 Views 430KB Size Report
Apr 9, 2003 - using Olympus 1X70 (Hitech Instruments, Inc., Edgemont, PA). Transfection of .... IRF-1 acts as a transcriptional activator of IFN gene (25), we.
THE JOURNAL OF BIOLOGICAL CHEMISTRY © 2003 by The American Society for Biochemistry and Molecular Biology, Inc.

Vol. 278, No. 50, Issue of December 12, pp. 50615–50623, 2003 Printed in U.S.A.

Tumor Necrosis Factor ␣ Modulates Airway Smooth Muscle Function via the Autocrine Action of Interferon ␤* Received for publication, April 9, 2003, and in revised form, August 18, 2003 Published, JBC Papers in Press, September 30, 2003, DOI 10.1074/jbc.M303680200

Omar Tliba, Samira Tliba, Chien Da Huang, Rebecca K. Hoffman, Peter DeLong, Reynold A. Panettieri, Jr., and Yassine Amrani‡ From the Pulmonary, Allergy and Critical Care Division, Department of Medicine, University of Pennsylvania Medical Center, Philadelphia, Pennsylvania 19104

Current evidence suggests that tumor necrosis factor ␣ (TNF␣) and the family of interferons (IFNs) synergistically regulate many cellular responses that are believed to be critical in chronic inflammatory diseases, although the underlying mechanisms of such interaction are complex, cell-specific, and not completely understood. In this study, TNF␣ in a time-dependent manner activated both janus tyrosine kinase 1 and Tyk2 tyrosine kinase and increased the nuclear translocation of interferon-regulatory factor-1, STAT1, and STAT2 in human airway smooth muscle cells. In cells transfected with a luciferase reporter, TNF␣ stimulated ␥-activated site-dependent gene transcription in a time- and concentration-dependent manner. Using neutralizing antibodies to IFN␤ and TNF␣ receptor 1, we show that TNF␣-induced secretion of IFN␤ mediated ␥-activated site-dependent gene expression via activation of TNF␣ receptor 1. In addition, neutralizing antibody to IFN␤ also completely abrogated the activation of interferon stimulation response elementdependent gene transcription induced by TNF␣. Secreted IFN␤ acted as a negative regulator of TNF␣induced interleukin-6 expression, while IFN␤ augmented TNF␣-induced RANTES (regulated on activation normal T cell expressed and secreted) secretion but had little effect on TNF␣-induced intercellular adhesion molecule-1 expression. Furthermore TNF␣, a modest airway smooth muscle mitogen, markedly induced DNA synthesis when cells were treated with neutralizing anti-IFN␤. Together these data show that TNF␣, via the autocrine action of IFN␤, differentially regulates the expression of proinflammatory genes and DNA synthesis.

TNF␣1 is now considered to be one of the most pleiotropic cytokines in mediating inflammatory and immune responses in * This work was supported by National Institutes of Health Grants 2R01-HL55301 (to R. A. P.) and 1P50-HL67663 (to R. A. P.) and by American Lung Association Grant RG-062-N (to Y. A.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ‡ A Parker B. Francis fellow in pulmonary research. To whom correspondence should be addressed: Pulmonary, Allergy and Critical Care Division, University of Pennsylvania Medical Center, 421 Curie Blvd., 848 BRB II/III, Philadelphia, PA 19104-6160. Tel.: 215-573-9851; Fax: 215-573-4469; E-mail: [email protected]. 1 The abbreviations used are: TNF␣, tumor necrosis factor ␣; TNFR, TNF␣ receptor; ASM, airway smooth muscle; JAK, janus tyrosine kinase; IFN, interferon; STAT, signal transducers and activators of transcription; IRF, interferon-regulatory factor; GAS, ␥-activated site; ISRE, interferon stimulation response element; PDGF, platelet-derived growth factor; IFNAR, IFN␣ receptor; RANTES, regulated on activaThis paper is available on line at http://www.jbc.org

chronic lung diseases. In vivo studies using selective inhibitors of TNF␣ activity demonstrate that TNF␣ plays a major role in antigen-induced airway inflammation (leukocyte infiltration) and airway hyper-responsiveness in animal models of asthma (1, 2). Others who used receptor knock-out mice confirmed the importance of both TNF␣ receptors (TNFRs), TNFR1 and TNFR2, in the abnormal airway changes induced by allergen challenge in sensitized animals (3–5). A potential site for TNF␣ deleterious action in the lungs is airway smooth muscle (ASM), a primary effector tissue thought to only regulate bronchomotor tone (6). In human cultured ASM cells that retain physiological responsiveness and express both TNF␣ receptors (7), TNF␣ alters proinflammatory gene expression that in turn may play an important role in the pathogenesis of allergic asthma. In previous reports, we showed that TNF␣ increased expression of ICAM-1, IL-6, and RANTES by selectively activating TNFR1, although TNFR2 was also involved in TNF␣induced RANTES secretion (8 –10). TNFR1-associated gene expression has been involved in the development of bronchial hyper-responsiveness (6, 11, 12). TNF␣ also cooperates with other cytokines such as interferon ␥ (IFN␥) to regulate the expression of cytokines (IL-1 and IL-5) and chemokines (IL-8, eotaxin, and RANTES) as well as adhesion molecules such as ICAM-1, vascular cell adhesion molecule-1, and CD40 (for a review, see Ref. 10). Because ASM expresses receptors for a variety of cytokines and chemokines, investigators suggest that many secreted cytokines such as IL-1 and IL-5, in an autocrine manner, may modulate ASM function to elicit a “proasthmatic phenotype” (13, 14). Whether Janus tyrosine kinase (JAK)/signal transducers and activators of transcription (STAT)-dependent signaling pathways mediate the cooperative interaction between TNF␣ and IFN␥ in human ASM cells remains unknown. JAKs and STATs are the central components of IFN receptor signaling. Ligands stimulate IFN-receptor complexes (types I and II) and activate the receptor-associated tyrosine kinase JAK, specifically JAK1 and Tyk2 by IFN␣/␤ (type I) or JAK1 and JAK2 by IFN␥ (type II) (15). JAKs then phosphorylate STAT proteins that assemble in dimeric or oligomeric forms, translocate to the nucleus, and regulate gene expression. STAT proteins play an important function in regulating immunological and inflammatory responses (16). The role of the JAK/ STAT-dependent pathways in promoting airway diseases is supported by evidence that STAT proteins regulate a number of inflammatory responses associated with allergic diseases such as Th1 and Th2 differentiation, IgE regulation, and cytokine expression (for a review, see Ref. 17). Increased levels of tion normal T cell expressed and secreted; ICAM-1, intercellular adhesion molecule-1; IL, interleukin; r (prefix), recombinant; TRAF2, tumor necrosis factor receptor-associated factor 2; cds, coding sequence.

50615

50616

TNF␣ Activates STAT1 and STAT2

FIG. 1. TNF␣ activates phosphorylation and nuclear translocation of STAT1 and STAT1-dependent gene expression. A, cells were stimulated with TNF␣ (10 ng/ml) or IFN␥ (100 units/ml) for the indicated times and lysed, and total cell lysates were prepared and assayed for either phosphorylated STAT1 (Tyr-701) (P-STAT1) or STAT1 by immunoblot analysis (15% SDS-PAGE). B, cells stimulated with TNF␣ (10 ng/ml) for 3 h were fixed, permeabilized, and incubated with mouse anti-phosphorylated STAT1 (Tyr-701) followed by a fluorescein isothiocyanateconjugated anti-mouse antibody. This result is representative of three separate experiments. C and D, cells transfected with a luciferase reporter construct containing GAS motifs were then stimulated with TNF␣ (10 ng/ml, C) or IFN␥ (100 units/ml, D) for the indicated times. Cells were lysed, and the luciferase activity in cell extracts was normalized for ␤-galactosidase activity as described under “Experimental Procedures.” Data are representative of three separate experiments. *, p ⬍ 0.05 compared with untreated cells.

STAT1 and STAT1-dependent genes such as IFN-regulatory factor IRF-1 or ICAM-1 in bronchial epithelium from subjects with asthma correlate with an accumulation of T cells in the airways, a defining feature of asthma (18). Although the factors responsible for STAT activation in asthma still remain unknown, these data suggest that altered STAT-dependent gene expression may be a key factor driving airway inflammation. In the present study, we investigated whether the modulatory effect of TNF␣ on gene expression in human ASM cells involves activation of the JAK/STAT signaling pathways. EXPERIMENTAL PROCEDURES

Smooth Muscle Cell Culture and Characterization—Human ASM cell culture was performed as described previously (7). Human trachea was obtained from lung transplant donors in accordance with procedures approved by the University of Pennsylvania Committee on Studies Involving Human Beings. A segment of trachea just proximal to the carina was removed under sterile conditions, and the trachealis muscle was isolated. The muscle was then centrifuged and resuspended in 10 ml of buffer containing 0.2 mM CaCl2, 640 units/ml collagenase, 1 mg/ml soybean trypsin inhibitor, and 10 units/ml elastase. Enzymatic dissociation of the tissue was performed for 90 min in a shaking water bath at 37 °C. The cell suspension was filtered through 105-␮m Nytex mesh, and the filtrate was washed with equal volumes of cold Ham’s F-12 medium supplemented with 10% fetal bovine serum (HyClone, Logan, UT). Aliquots of the cell suspension were plated at a density of 1.0 ⫻ 104 cells/cm2. The cells were cultured in Ham’s F-12 medium supplemented with 10% FBS, 100 units/ml penicillin, 0.1 mg/ml streptomycin, and 2.5 ␮g/ml amphotericin B, and this was replaced every 72 h. Human ASM cells in subculture during the second through fifth cell passages were used since the cells retain native contractile protein expression as demonstrated by immunocytochemical staining for smooth muscle actin and myosin.

Thymidine Incorporation Assays—DNA synthesis was evaluated by measuring thymidine incorporation as described previously (19). Cells were growth-arrested by incubating the cultures in serum-free medium consisting of Ham’s F-12 medium with 0.1% bovine serum albumin. After 48 h in serum-free medium, the cells were stimulated with 1 unit/ml thrombin in the presence or absence of 10 ng/ml TNF␣ with or without neutralizing anti-IFN␤. After 16 –18 h of stimulation, human ASM cells were labeled with 3 ␮Ci/ml [methyl-3H]thymidine (40 – 60 Ci/mmol, Amersham Biosciences) for 24 h. The cells were then scraped and lysed, and the protein and DNA were precipitated with 10% trichloroacetic acid. The precipitant was aspirated onto glass filters, extensively washed, dried, and counted. Flow Cytometry Analysis—Flow cytometry was performed as described previously (20). Antibodies used for IFN␤ receptor expression (anti-IFNAR1 and anti-IFNAR2) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Fluorescein isothiocyanate-conjugated secondary antibodies were bought from Jackson ImmunoResearch Laboratories (West Grove, PA). Enzyme-linked Immunosorbent Assay—Cytokines in supernatants were measured using antibodies obtained from R&D Systems (Minneapolis, MN) (IL-6 and RANTES) and BD Pharmingen (IFN␥) as indicated by the manufacturers’ instructions (9). As a positive control for IFN␥ enzyme-linked immunosorbent assays, the renal carcinoma cell line (REN) was infected with either adenovirus encoding for IFN␥ and ␤-galactosidase (Ad-IFN␥) or a control adenovirus encoding for ␤-galactosidase alone (Ad) (21). Levels of IFN␤ were measured using a commercially available human IFN␤ enzyme-linked immunosorbent assay kit (catalogue number 41400-1, R&D Systems). The range of detection using IFN␤ standard was found to be between 125 and 10 000 pg/ml (R2 ⫽ 0.9911). Immunoblot Analysis—Immunoblot analysis for cyclin D1 and p27Kip1 was performed as described previously (20, 22). In brief, cells were lysed in buffer containing 10 mM Tris, pH 7.5, 100 mM NaCl, 1%

TNF␣ Activates STAT1 and STAT2 Triton X-100, 0.1% deoxycholate, 10 ␮g/ml leupeptin, 100 ␮M phenylmethylsulfonyl fluoride, 10 ␮g/ml aprotinin, 5 mM EDTA, 10 mM NaF, and 2 mM Na3VO4 for 20 min at 4 °C. Postnuclear extracts were obtained by centrifugation of lysates at 14,000 ⫻ g for 10 min. Immunoprecipitations using the Tyk2, JAK1, and STAT2 antibodies (Santa Cruz Biotechnology) were performed as indicated by the manufacturer’s instructions. Equal amounts of proteins were analyzed by 4 –12% SDSpolyacrylamide gel electrophoresis and blotted onto a nitrocellulose membrane. The membranes were blocked in 5% milk or 5% bovine serum albumin (anti-phosphoprotein antibodies) in Tris-buffered saline and then incubated with either of the following antibodies: anti-phospho-STAT1 (Tyr-701), anti-STAT1, anti-Tyk2, and anti-JAK1 (Santa Cruz Biotechnology); anti-phospho-Tyk2 (Tyr-1054/1055) and antiphosphotyrosine Tyr(P) (Cell Signaling, Beverly, MA); and anti-phospho-JAK1 (Tyr-1022/1023) (BioSource International, Camarillo, CA). Detection of IRF-1 proteins (Santa Cruz Biotechnology) was performed on nuclear extracts. After incubation with the appropriate peroxidaseconjugated secondary antibody (Roche Applied Science), the bands were visualized by the enhanced chemiluminescence system (Amersham Biosciences) and autoradiographed. Immunocytochemistry of STAT1, STAT2, and IRF-1—Immunostaining for nuclear translocation experiments was performed as described previously (8) with the exception of the following antibodies: anti-phospho-STAT1 (Tyr-701), anti-STAT2, and anti-IRF-1 (Santa Cruz Biotechnology). Isotype-matched antibodies (rabbit and mouse IgG from R&D Systems) were used as negative controls. After staining, the glass coverslips were mounted onto glass slides, examined under epifluorescence microscopy (Nikon, Tokyo, Japan), and photographed using Olympus 1X70 (Hitech Instruments, Inc., Edgemont, PA). Transfection of Human ASM Cells—Transfection of human ASM cells was performed as reported previously (9). In brief, human ASM cells were transfected with 10 ␮g of pGAS-Luc or 10 ␮g of pISRE-Luc to monitor the transcriptional activities of STAT1 and STAT2, respectively (Stratagene, La Jolla, CA), and 5 ␮g of a pSV-␤-galactosidase control vector was used to normalize transfection efficiencies (Promega, Madison, WI). Forty-eight hours after transfection, the cells were rendered quiescent in medium containing 0.1% fetal bovine serum for 24 h and exposed to 10 ng/ml TNF␣ or 500 units/ml IFN␤. In neutralizing experiments, human ASM cells were preincubated with anti-TNFR1 or anti-TNFR2 (20 ␮g/ml, 60 min) anti-IFN␤ or isotype-matched IgG (same concentration). Cells were then harvested, and luciferase and ␤-galactosidase activities were assessed with a Promega kit according to the manufacturer’s instructions. RNA Isolation and Reverse Transcriptase PCR Analysis—Human ASM cells were serum-deprived in medium containing 0.1% fetal bovine serum for 24 h and exposed to 10 ng/ml TNF␣ at different times (0, 0.5, 1, 2, 3, and 4 h). Total RNA was isolated using TRIzol reagent (Invitrogen) according to the manufacturer’s instructions. Reverse transcriptase PCR analysis of IFN␣, IFN␤, IFN␥, platelet-derived growth factor A (PDGF)-A, PDGF-B, and glyceraldehyde-3-phosphate dehydrogenase expression was then performed as reported previously (6). Consensus primers (upstream primer, 5⬘-TGATGGCAACCAGTTCCAGAAGGCTCAAG-3⬘; downstream primer, 5⬘-ACAACCTCCCAGGCACAAGGGCTGTATTT-3⬘ (23)) were used to detect multiple human IFN␣ subtypes (GenBankTM accession numbers, based on the BLAST analyses): NM_024013, Homo sapiens interferon ␣ 1 (IFNA1) mRNA; AF439447, H. sapiens interferon ␣ 1b gene, partial cds; Y11834, H. sapiens IFNA2 gene; AY255838, H. sapiens interferon ␣ 2b mRNA, complete cds; NM_021068, H. sapiens interferon ␣ 4 (IFNA4) mRNA; X02955, human interferon ␣ gene IFN␣ 4b; X02956, human interferon ␣ gene IFN␣ 5; X02958, human interferon ␣ gene IFN␣ 6; X02960, human interferon ␣ gene IFN␣ 7; X03125, human interferon ␣ gene IFN␣ 8; NM_002171, H. sapiens interferon ␣ 10 (IFNA10) mRNA; X00803, human interferon ␣ gene IFN␣ 13; X02959, human interferon ␣ gene IFN␣ 14; X02957, human interferon ␣ gene IFN␣ 16; NM_021268, H. sapiens interferon ␣ 17 (IFNA17) mRNA; NM_002175, H. sapiens interferon ␣ 21 (IFNA21) mRNA; K01900, human lymphocyte interferon ␣ type B mRNA, complete cds; V00532, human gene for leukocyte (␣) interferon C; J00210, human leukocyte interferon (IFN␣) ␣-d gene, complete cds; X00145, human mRNA for interferon ␣-F; V00533, human gene for leukocyte (␣) interferon H; V00531, human interferon genes LeIF-L and LeIF-J; M34913, human interferon ␣-J1 (IFN␣-J1) mRNA, complete cds; M27318, human interferon (IFN␣-M1) mRNA, complete cds; X00140, human mRNA for interferon ␣-N; K02055, human interferon ␣-WA gene, complete cds, clone ␭-85; M28585, human leukocyte interferon ␣ mRNA, complete cds. Other primers for IFN␤, IFN␥, PDGF-A, PDGF-B, and glyceraldehyde-3-phosphate dehydrogenase detection were identical to those reported previ-

50617

FIG. 2. Concentration-dependent effects of TNF␣ on GAS-dependent gene expression: role of TNFR1. Cells transfected with a luciferase reporter construct containing GAS enhancer elements were stimulated with the indicated concentration of TNF␣ for 4 h (A) or were first preincubated with neutralizing anti-TNFR1 or anti-TNFR2 (20 ␮g/ml, 60 min) antibodies before adding 10 ng/ml TNF␣ for 4 h (B). Cells were lysed, and the luciferase activity in cell extracts was normalized for ␤-galactosidase activity as described under “Experimental Procedures.” Data are representative of three separate experiments. *, p ⬍ 0.05 compared with untreated cells. #, p ⬍ 0.05 when compared with cells treated with TNF␣ alone. NS, non-significant. ously (23, 24) and were designed to amplify at least one intron in the genes to exclude contamination of cDNA with genomic DNA. PCR products were separated on 1% agarose gels and stained with ethidium bromide. Materials and Reagents—Tissue culture reagents and primers used for PCR were obtained from Invitrogen. Human rTNF␣ was provided by Roche Applied Science. rIFN␥, rIFN␤, anti-TNFR1 neutralizing antibody, anti-IFN␤ neutralizing antibody (sheep polyclonal antibody, catalogue number 31400-1, 5 ␮g/ml, 15 min), isotype-matched goat or mouse IgG were purchased from R&D Systems. Cycloheximide and thrombin were purchased from Sigma and Calbiochem, respectively. The anti-TNFR2 neutralizing antibody was obtained from Cell Sciences Inc. (Norwood, MA). Statistical Analysis—To compare differences between treatment means (expressed as mean ⫾ S.E.), all data were subjected to one- or two-way analysis of variance when experiments were of a factorial design. After analysis of variance, Fisher’s method of protected least significant differences was used as a multiple comparison test. Comparison of two populations was made with Student’s t test. Values of p ⬍ 0.05 were sufficient to reject the null hypothesis for all analyses. RESULTS

TNF␣ Stimulates STAT1 Activation and ␥-Activated Site (GAS)-dependent Gene Expression in Human ASM Cells—In human ASM cells, TNF␣ and IFNs synergize to regulate the expression of a number of proinflammatory genes (10). Here we tested the hypothesis that TNF␣ modulates IFN-associated signaling pathways. We found that TNF␣ stimulates STAT1 activation at 3 and 4 h as shown by the increased STAT1 phosphorylation at tyrosine residue 701 in cytokine-treated

50618

TNF␣ Activates STAT1 and STAT2

FIG. 3. TNF␣, via the autocrine action of secreted IFN␤, activates STAT1-dependent gene expression and IFN␤ receptor-associated proteins. A, ASM cells were stimulated with TNF␣ for the indicated times. Total mRNA (2 ␮g) was subjected to reverse transcriptase PCR with the primers for glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and IFN␤. PCR products were separated on a 1% agarose gel and stained with ethidium bromide. Data are representative of mRNA obtained from three different experiments. B, cells were stimulated with 10 ng/ml TNF␣ for the indicated time, lysed, immunoprecipitated with anti-JAK1 or anti-Tyk2, and assayed for JAK1, phospho-JAK1 (P-JAK1), Tyk2, or phospho-Tyk2 (P-Tyk2) by immunoblot analysis (gradient 4 –12% SDS-PAGE). Results are representative of three separate experiments. C, ASM cells were transfected with luciferase reporter construct containing GAS enhancer elements and stimulated for 4 h with TNF␣ (10 ng/ml) or IFN␤ (500 units/ml) in the presence or absence of neutralizing anti-IFN␤ (5 ␮g/ml, 15 min). Cells were lysed, and luciferase activity in cell extracts was normalized for ␤-galactosidase activity as described under “Experimental Procedures.” Data are representative of four separate experiments. *, p ⬍ 0.05 when compared with control cells. #, p ⬍ 0.05; ##, p ⬍ 0.05 when compared with cells treated with IFN␤ or TNF␣, respectively.

nuclear extracts. In comparison, STAT1 phosphorylation by IFN␥ was swifter and more robust occurring at 30 min (Fig. 1A). Fig. 1B shows that TNF␣ treatment for 3 h also stimulated the nuclear translocation of activated STAT1 in human ASM cells, an effect that was completely abrogated by cycloheximide, a protein synthesis inhibitor (data not shown). To address the functional consequences of TNF␣-induced STAT1 activation, ASM cells were transfected with a reporter construct containing GAS motifs that bind activated STAT1 (15). As shown in Fig. 1C, TNF␣ stimulated luciferase activity in a time-dependent manner, although a slight decrease in GAS reporter activity was observed at 2 h (45%) when compared with untreated cells. Interestingly the effect of TNF␣ on GAS-dependent gene expression temporally correlated with the increase in the phosphorylation of STAT1 induced by TNF␣ (Fig. 1A). The effect of TNF␣ on GAS-mediated gene expression was concentration-dependent (Fig. 2A) with significant increases at 1, 10, and 30 ng/ml with a 3.49 ⫾ 0.6-, 5.2 ⫾ 0.54-, and 13 ⫾ 0.37-fold increase over basal, respectively. In contrast to TNF␣, IFN␥ differentially regulates GAS-dependent gene expression, characterized by a swifter time course as early as 2 h with a 5 ⫾ 0.7-fold increase over basal and more robust activation of reporter activity reaching an 18 ⫾ 1.3-fold increase over basal at 4 h (Fig. 1D), suggesting that both cytokines activate GAS-dependent gene expression via potentially disparate pathways. Using antagonistic antibodies against TNFR receptors (9), we showed that neutralizing anti-TNFR1, but not anti-TNFR2,

abrogated TNF␣-induced GAS-dependent transcription (Fig. 2B), while an isotype-matched IgG had no effect on TNF␣induced STAT1 activation (data not shown). Together these results show that TNF␣, via TNFR1, induces a delayed activation of GAS-mediated gene expression in human ASM cells. STAT1 Activation by TNF␣ Involves the Autocrine Action of Secreted IFN␤—The molecular mechanism(s) by which TNF␣ regulate(s) STAT1 activation in ASM cells remain(s) unknown. Cycloheximide, a protein synthesis inhibitor, blocked TNF␣induced STAT1 phosphorylation (data not shown), suggesting the involvement of a de novo synthesized protein in the delayed time course of TNF␣-induced STAT1 activation. Since IFN␥ is the prototypical ligand that stimulates GAS-dependent gene transcription (15), we next examined whether TNF␣ stimulated the expression of IFN␥. TNF␣ did not induce secretion of IFN␥ in ASM cells, while a human malignant mesothelioma cell line (REN cells) infected with adenovirus encoding for IFN␥ (21) constitutively secreted IFN␥ as determined by enzymelinked immunosorbent assay (data not shown). We also investigated whether TNF␣ stimulated expression of IFN␣, IFN␤, or PDGF (A and B) since these ligands reportedly activate STAT1 in many cell types (15). Using a reverse transcriptase PCR technique, we found that TNF␣ significantly increased, in a time-dependent manner, levels of IFN␤ mRNA starting at 1 h, reaching a plateau at 2–3 h, and declining after 4 h (Fig. 3A). TNF␣, however, had no effect on mRNA expression of PDGF-A, PDGF-B, or IFN␣ (data not shown). TNF␣ also induced secre-

TNF␣ Activates STAT1 and STAT2

50619

FIG. 4. TNF␣ stimulates the nuclear translocation of STAT2 and induces ISRE-dependent gene expression. A, cells stimulated with TNF␣ for the indicated time were lysed, immunoprecipitated with anti-STAT2 antibody, and assayed for phosphotyrosine (pTyr) or STAT2 by immunoblot analysis (gradient 4 –12% SDS-PAGE). B, cells stimulated with TNF␣ (10 ng/ml) for 3 h in the presence or absence of neutralizing anti-IFN␤ (5 ␮g/ml, 15 min) were fixed, permeabilized, and incubated with mouse anti-STAT2 followed by a secondary fluorescein isothiocyanateconjugated anti-IgG antibody. This result is representative of three separate experiments. C and D, ASM cells were transfected with luciferase reporter construct containing ISRE enhancer elements and stimulated for the indicated time with TNF␣ (10 ng/ml) or IFN␤ (500 units/ml) in the presence or absence of neutralizing anti-IFN␤ (5 ␮g/ml, 15 min). Cells were lysed, and luciferase activity in cell extracts was normalized for ␤-galactosidase activity as described under “Experimental Procedures.” Data are representative of three separate experiments. *, p ⬍ 0.05 when compared with control cells. #, p ⬍ 0.05; ##, p ⬍ 0.05 when compared with cells treated with TNF␣ or IFN␤, respectively.

tion of IFN␤ protein with net levels of 168 ⫾ 31 IU/ml over basal at 3 h (p ⬍ 0.05). In support of IFN␤-mediated regulation of TNF␣-induced STAT1 activation, we also found that TNF␣ stimulated IFN␤ receptor-associated signaling molecules JAK1 and Tyk2. TNF␣ increased tyrosine phosphorylation of both JAK1 and Tyk2, an effect that was apparent at 1 h and increased in a time-dependent manner as shown in Fig. 3B. In addition, blocking anti-IFN␤ antibody completely abrogated both TNF␣and IFN␤-induced GAS-dependent gene expression as shown in Fig. 3C. These data suggest that TNF␣ regulates GAS-dependent gene expression via the autocrine action of IFN␤. TNF␣ Activates STAT2 and IRF-1 in Human ASM Cells— Because STAT2 is a major component of IFN␤ signaling pathways, we next determined whether TNF␣ activates STAT2 in human ASM cells. Immunoprecipitation studies showed that TNF␣ increased STAT2 phosphorylation at 2 h (Fig. 4A) and nuclear translocation (Fig. 4B) at 3 h, an effect that was completely blocked by neutralizing anti-IFN␤ antibodies (Fig. 4, A and B) and by cycloheximide, a protein synthesis inhibitor (data not shown). To address the question as to whether STAT2 activation was transcriptionally active, we studied the effect of TNF␣ on a luciferase reporter that is controlled by a promoter containing ISRE binding elements. As shown in Fig. 4C, TNF␣ activates ISRE-dependent luciferase activity in a time-dependent manner (2– 4 h). Neutralizing antibody to IFN␤ completely prevented the induction of ISRE-dependent gene expression in response to both TNF␣ and exogenous IFN␤ (Fig. 4D). Because

IRF-1 acts as a transcriptional activator of IFN␤ gene (25), we determined whether TNF␣ induces IRF-1 expression in human ASM cells. Both immunoblot analysis and immunostaining showed IRF-1 expression in the nucleus of cells treated with TNF␣, observed as early as 1 h (Fig. 5, A and B) and sustained for up to 4 h, an effect that was not inhibited by neutralizing anti-IFN␤ (Fig. 5B). These data suggest that in addition to IRF-1, TNF␣ activates major components of IFN␤ receptor signaling pathways (STAT1 and STAT2) in human ASM cells. TNF␣ Has Little Effect on IFN␤ Receptor Expression—Most IFN␤ effects are mediated by activating the type I receptor, which is composed of two subunits called ␣ (IFNAR1) and ␤ (IFNAR2) (26 –28). The role of such proteins in modulating ASM cell function remains unknown. Because cytokines, such as TNF␣ or IL-1␤, increase the expression of IFNAR2 in hepatocellular carcinoma cell lines (29), we examined whether similar effects occur in ASM cells. As shown in Fig. 6, both IFN␤ receptor subunits are expressed in ASM cells. However, pretreating cells with TNF␣ had no effect on the expression of IFNAR1 (Fig. 6A) or IFNAR2 (Fig. 6B). These data suggest that the stimulatory effect of TNF␣ on the IFN␤-associated signaling pathway does not involve a modulation of IFN␤ receptor expression. Secretion of IFN␤ Differentially Regulates TNF␣-induced Expression of Proinflammatory Genes—Although TNF␣ regulates a variety of proinflammatory genes in human cells that may play an important role in the pathogenesis of asthma (9, 20,

TNF␣ Activates STAT1 and STAT2

50620

FIG. 5. TNF␣ stimulates the nuclear translocation of IRF-1. A, cells stimulated with TNF␣ for the indicated time were lysed, and nuclear extracts were prepared and assayed for IRF-1 by immunoblot analysis (15% SDS-PAGE). B, cells stimulated with TNF␣ (10 ng/ml) for 3 h in the presence or absence of neutralizing anti-IFN␤ (5 ␮g/ml, 15 min) were fixed, permeabilized, and incubated with mouse or anti-IRF-1 followed by a secondary fluorescein isothiocyanate-conjugated antiIgG antibody. This result is representative of three separate experiments.

30), the underlying mechanisms remain unknown. To investigate the functional relevance of IFN␤ secretion on gene expression in ASM cells, we examined the effect of neutralizing antiIFN␤ on TNF␣-induced expression of ICAM-1 and secretion of IL-6 and RANTES. As shown in Fig. 7, A and B, anti-IFN␤ antibodies markedly reduced TNF␣-induced RANTES secretion by 64% (from 8,088 ⫾ 376 to 3,118 ⫾ 198 pg/ml), while IL-6 secretion was increased by 185% (from 7,626 ⫾ 836 to 19,438 ⫾ 601 pg/ml) as compared with that obtained in cells treated with TNF␣ alone. Interestingly TNF␣-induced ICAM-1 expression was unaffected (Fig. 7C). An isotype-matched IgG had no effect on TNF␣-induced RANTES and IL-6 secretion (data not shown). These data suggest that autocrine secretion of IFN␤ differentially mediates TNF␣-induced expression of proinflammatory genes in human ASM cells. Secretion of IFN␤ Suppresses Both Basal and Mitogen-stimulated Cell Mitogenesis—Evidence suggests that TNF␣ suppresses agonist-induced cell proliferation (31), although the exact mechanism remains unclear. Because IFNs inhibit proliferation of many cell types (16), including ASM cells (32), we investigated the possibility that TNF␣-induced secretion of IFN␤ regulated both basal and mitogen-induced cell mitogenesis. As shown in Fig. 8, TNF␣ alone had little effect on ASM cell proliferation. Interestingly, in the presence of neutralizing anti-IFN␤ but not isotype-matched antibody (data not shown), TNF␣ induced a significant increase in [3H]thymidine incorporation (Fig. 8A), an effect that correlates with an increase in cyclin D1 expression and degradation of the cell cycle inhibitor p27Kip1 (Fig. 8B). In addition, neutralizing anti-IFN␤ completely prevented the inhibitory effect of TNF␣ on thrombininduced ASM cell mitogenesis (Fig. 8A). Together these data show that TNF␣ exerts a growth-suppressive effect on ASM cell mitogenesis via the secretion of IFN␤. DISCUSSION

By its ability to secrete a variety of proinflammatory cytokines and chemokines, airway smooth muscle, an important effector tissue that regulates bronchomotor tone, may also regulate airway inflammation in chronic lung diseases (10). Our report shows that TNF␣, a proinflammatory cytokine involved

in the pathogenesis of asthma (6, 7), regulates mitogenic responses and gene expression in human ASM cells via the autocrine activation of IFN␤-associated signaling pathways. Previous reports from our laboratory revealed that TNFR1associated gene expression in ASM cells may play a central role in two features of asthma, i.e. airway inflammation and airway hyper-responsiveness (6 –9, 11, 33). Although TNFR1 initiates signal transduction cascades by recruiting the adapter protein TRADD (TNF receptor-associated death domain) that in turn activates the signaling proteins TRAF2 and FADD (Fas-associated death domain) (for a review, see Ref. 33), our findings now suggest that IFN␤ signal transducers represent active components in some but not all cellular responses induced by TNF␣. We now show that TNF␣, via activation of TNFR1, not only induces gene and protein expression of IFN␤ but also activates IFN␤ receptor-associated signaling molecules JAK1 and Tyk2 tyrosine kinases in human ASM cells. The molecular mechanism by which TNF␣ stimulates IFN␤ expression in ASM cells remains unknown, but investigators showed that transcription factors IRF-1 and NF-␬B, both activated in ASM cells (present study and Refs. 9 and 20), play a major role in the transcriptional activation of the IFN␤ promoter (34). This raises the possibility that both NF-␬B and IRF-1 may induce IFN␤ expression by TNF␣, although evidence suggests that IRF-1 could also act as a competitive inhibitor of NF-␬B-dependent gene expression (35). The effect of TNF␣ on IFN␤-dependent signaling molecules, however, was not associated with changes in the expression of IFN␤ receptor subunits, IFNAR1 or IFNAR2. Others reported that TNF␣ increased expression of IFNAR2 in hepatocellular cell lines (29). In a variety of cell lines, activation of TNFR1 leads to a rapid activation of JAK/ STAT pathway unlike that observed in ASM cells. In adipocytes and B cells, TNFR1 physically recruits both JAK1 and JAK2 within 5 and 15 min after stimulation (36, 37), while in HeLa cells both STAT1 and JAK2 are constitutively associated with both TNFR1 and TNFR2 in basal conditions, and this association was further increased by TNF␣ at 15–30 min (36, 38). In ASM cells, however, we found that activation of JAK1 and Tyk2 by TNF␣ was delayed and was associated with acti-

TNF␣ Activates STAT1 and STAT2

50621

FIG. 6. TNF␣ does not affect expression of IFN␤ receptor subunits IFNAR1 and IFNAR2. Cells were stimulated with TNF␣ (10 ng/ml) for 3 h, and levels of IFNAR1 (A) and IFNAR2 (B) were assessed by flow cytrometry as described under “Experimental Procedures.” The results shown are representative of three independent experiments.

vation of both STAT1 and STAT2 at 2 h, while STAT3 was not activated by TNF␣ (data not shown) in contrast to that reported in human B cells (37). Another major difference in our study was that nuclear translocated STAT1 and STAT2 induced by TNF␣ were transcriptionally active as evidenced by the ability of TNF␣ to activate GAS- and ISRE-dependent gene expression. In HeLa cells as well as in 3T3-L1 adipocytes, there was no DNA binding activity attributable to activated STAT1 in response to TNF␣ (36, 38). The authors suggested that STAT1 may be acting as a negative regulator of TNF␣-induced NF-␬B activation via a blockade of TNFR1-associated signaling molecule TRAF2 (38). Another interesting finding was the inhibitory effect of TNF␣ on GAS-dependent gene expression noticed at 2 h (but not observed at 3 and 4 h), while the ISRE-dependent transcription was induced by TNF␣. The reasons for the differential effect of TNF␣ on gene expression controlled by GAS but not by ISRE binding elements at early time points remain unclear, and further studies are required to determine whether physical and/or functional interaction between members of TNF␣ signal transduction and JAK/STAT pathways as shown in HeLa cells is involved (38). A recent

study using STAT1-deficient cells U3A and STAT1 stably transfected cells U3APSG91 showed that STAT1 played a minor role in TNF␣-induced cell cytotoxicity (39), suggesting that the role and the type of STAT proteins activated by TNF␣ are complex and highly cell-specific. Although TNF␣ increased IFN␤ mRNA (40) or protein (41) in vascular and bronchial smooth muscle cells alone or in cells infected with Chlamydia pneumonia, the physiological consequence of IFN␤ secretion by TNF␣ was not examined. In our study, IFN␤, in an autocrine manner, promoted TNF␣-induced expression of RANTES while acting as a negative regulator of IL-6 expression. Previous reports showed that IFN␤ induced RANTES expression in T cell lymphoma cell lines and in human macrophages (42) but not in endothelial cells (43), while in human fibroblasts IFN␤ also requires the presence of TNF␣ to induce RANTES expression (44). Interestingly secreted IFN␤ does not affect TNF␣-induced ICAM-1 expression, although investigators using dominant negative proteins or reporter plasmids and their 5⬘ deletion derivatives demonstrated the importance of STAT1 binding elements in the transcriptional activation of the ICAM-1 promoter (45). Others also showed

50622

TNF␣ Activates STAT1 and STAT2

FIG. 7. TNF␣, via the secretion of IFN␤, differentially regulates ICAM-1, IL-6, and RANTES expression. Cells were stimulated for 24 h with TNF␣ (10 ng/ml) in the presence or absence of neutralizing anti-IFN␤ (5 ␮g/ml, 15 min). Secretion of IL-6 (A) and RANTES (B) or expression of ICAM-1 (C) levels were assessed as described under “Experimental Procedures.” Values shown are mean ⫾ S.E. and are significantly different from controls. *, p ⬍ 0.05 when compared with basal. #, p ⬍ 0.05 when compared with cells treated with TNF␣ alone.

FIG. 8. TNF␣, via the secretion of IFN␤, suppresses ASM cell proliferation and cell cycle regulators cyclin D1 and p27Kip1. Cells were stimulated for 40 h with thrombin (Thr) (1 unit/ml) and/or TNF␣ (10 ng/ml) in the presence or absence of neutralizing anti-IFN␤ (5 ␮g/ml, 15 min). A, [3H]thymidine incorporation was performed as described under “Experimental Procedures.” The experiment represents the mean ⫾ S.E. from three replicates per condition performed in three separate experiments. #, p ⬍ 0.05; *, p ⬍ 0.05 when compared with cells treated with diluent or thrombin alone, respectively (determined by analysis of variance with Bonferroni-Dunn correction). B, cells were lysed, and immunoblot analysis was performed with monoclonal antibodies specific to cyclin D1, p27Kip1, and ␤-actin (lower panel) for equal amounts of protein loading. These data are representative of two experiments.

TNF␣ Activates STAT1 and STAT2 that IFN␤ inhibits TNF␣-induced ICAM-1 expression in brain endothelial cells (46). Our study shows that IFN␤ negatively regulates cytokine-induced IL-6 gene expression, suggesting that the regulation of gene expression by IFN␤ appears to be cell-specific and modulated by other cytokines. Further studies will provide essential information regarding the transcription factors involved in the differential effect of IFN␤ on cytokineinduced gene expression. Because TNF␣ regulates a variety of cytokines and/or chemokines in human ASM cells such as IL-8, eotaxin (47), or MCP-1 (48), IFN␤ may also serve as an important regulator of other TNF␣-induced proinflammatory genes. Finally we showed that secreted IFN␤ inhibited ASM cell proliferation supporting our recent findings that exogenous IFNs are potent suppressors of ASM cell mitogenesis (32). Previous reports also showed that TNF␣ inhibits thrombininduced ASM cell proliferation, although the underlying mechanisms remained unclear (31). Blockade of IFN␤ revealed that TNF␣ exerts a growth-suppressive effect on both basal and agonist-induced proliferative responses via the autocrine action of IFN␤. These data suggest that IFN␤ is a novel suppressor of ASM cell proliferation, and the alterations in the TNF␣IFN␤ pathways may play an important role in regulating changes in ASM mass seen in asthma (49). Our findings also raise the possibility that, in addition to IFNs, TNF␣ represents a potential cytokine involved in the activation of STAT1 and IRF-1 transcription factors previously described in the airways of patients with asthma (18). In summary, we showed that human ASM cells treated with TNF␣ are a novel source of active IFN␤ in the airways. Secreted IFN␤, in an autocrine manner, modulates cell mitogenesis as well as TNF␣-induced expression of proinflammatory genes RANTES and IL-6 that have been associated with chronic inflammatory diseases. Further studies are needed to determine the molecular mechanisms by which TNF␣ promotes secretion of IFN␤ as well as the pathophysiologic consequences of IFN␤ secretion by ASM in asthma where TNF␣ is thought to play a major role. Acknowledgments—We acknowledge Laura Kester for excellent technical assistance in immunoblot analyses and Mary McNichol for assistance in preparing the manuscript. REFERENCES 1. Lukacs, N. W., Strieter, R. M., Chensue, S. W., Widmer, M., and Kunkel, S. L. (1995) J. Immunol. 154, 5411–5417 2. Renzetti, L. M., Paciorek, P. M., Tannu, S. A., Rinaldi, N. C., Tocker, J. E., Wasserman, M. A., and Gater, P. R. (1996) J. Pharmacol. Exp. Ther. 278, 847– 853 3. Peschon, J. J., Torrance, D. S., Stocking, K. L., Glaccum, M. B., Otten, C., Willis, C. R., Charrier, K., Morrissey, P. J., Ware, C. B., and Mohler, K. M. (1998) J. Immunol. 160, 943–952 4. Kanehiro, A., Lahn, M., Makela, M. J., Dakhama, A., Joetham, A., Rha, Y. H., Born, W., and Gelfand, E. W. (2002) J. Immunol. 169, 4190 – 4197 5. Broide, D. H., Stachnick, G., Castaneda, D., Nayar, J., and Sriramarao, P. (2001) Am. J. Respir. Cell Mol. Biol. 24, 304 –311 6. Amrani, Y., Chen, H., and Panettieri, R. A., Jr. (2000) Respir. Res. 1, 49 –53 7. Amrani, Y., and Panettieri, R. A., Jr. (2002) Curr. Opin. Allergy Clin. Immunol. 2, 39 – 45 8. Amrani, Y., Lazaar, A. L., Hoffman, R., Amin, K., Ousmer, S., and Panettieri, R. A., Jr. (2000) Mol. Pharmacol. 58, 237–245

50623

9. Amrani, Y., Ammit, A. J., and Panettieri, R. A., Jr. (2001) Mol. Pharmacol. 60, 646 – 655 10. Lazaar, A. L., and Panettieri, R. A., Jr. (2001) Curr. Opin. Pharmacol. 1, 259 –264 11. Hunter, I., Cobban, H. J., Vandenabeele, P., MacEwan, D. J., and Nixon, G. F. (2003) Mol. Pharmacol. 63, 714 –721 12. Thabut, G., El-Benna, J., Samb, A., Corda, S., Megret, J., Leseche, G., Vicaut, E., Aubier, M., and Boczkowski, J. (2002) J. Biol. Chem. 277, 22814 –22821 13. Hakonarson, H., Maskeri, N., Carter, C., Chuang, S., and Grunstein, M. M. (1999) J. Clin. Investig. 104, 657– 667 14. Lahiri, T., Laporte, J. D., Moore, P. E., Panettieri, R. A., Jr., and Shore, S. A. (2001) Am. J. Physiol. 280, L1225–L1232 15. Darnell, J. E., Jr. (1997) Science 277, 1630 –1635 16. Horvath, C. M. (2000) Trends Biochem. Sci 25, 496 –502 17. Pernis, A. B., and Rothman, P. B. (2002) J. Clin. Investig. 109, 1279 –1283 18. Sampath, D., Castro, M., Look, D., and Holtzman, M. (1999) J. Clin. Investig. 103, 1353–1356 19. Panettieri, R. A., Murray, R. K., DePalo, L. R., Yadvish, P. A., and Kotlikoff, M. I. (1989) Am. J. Physiol. 256, C329 –C335 20. Amrani, Y., Lazaar, A. L., and Panettieri, R. A., Jr. (1999) J. Immunol. 163, 2128 –2134 21. Gurubhagavatula, I., Amrani, Y., Pratico, D., Ruberg, F. L., Albelda, S. M., and Panettieri, R. A., Jr. (1998) J. Clin. Investig. 101, 212–222 22. Lazaar, A. L., Amrani, Y., Hsu, J., Panettieri, R. A., Jr., Fanslow, W. C., Albelda, S. M., and Pure, E. (1998) J. Immunol. 161, 3120 –3127 23. Chehadeh, W., Weill, J., Vantyghem, M. C., Alm, G., Lefebvre, J., Wattre, P., and Hober, D. (2000) J. Infect. Dis. 181, 1929 –1939 24. Talon, J., Horvath, C. M., Polley, R., Basler, C. F., Muster, T., Palese, P., and Garcia-Sastre, A. (2000) J. Virol. 74, 7989 –7996 25. Taniguchi, T. (1997) J. Cell. Physiol. 173, 128 –130 26. Uze, G., Lutfalla, G., and Gresser, I. (1990) Cell 60, 225–234 27. Domanski, P., Witte, M., Kellum, M., Rubinstein, M., Hackett, R., Pitha, P., and Colamonici, O. R. (1995) J. Biol. Chem. 270, 21606 –21611 28. Lutfalla, G., Holland, S. J., Cinato, E., Monneron, D., Reboul, J., Rogers, N. C., Smith, J. M., Stark, G. R., Gardiner, K., and Mogensen, K. E. (1995) EMBO J. 14, 5100 –5108 29. Takayama, A., Yano, H., Ogasawara, S., Higaki, K., and Kojiro, M. (2000) Int. J. Mol. Med. 6, 621– 627 30. Ammit, A. J., Lazaar, A. L., Irani, C., O’Neill, G. M., Gordon, N. D., Amrani, Y., Penn, R. B., and Panettieri, R. A., Jr. (2002) Am. J. Respir. Cell Mol. Biol. 26, 465– 474 31. Stewart, A. G., Tomlinson, P. R., Fernandes, D. J., Wilson, J. W., and Harris, T. (1995) Am. J. Respir. Cell Mol. Biol. 12, 110 –119 32. Amrani, Y., Tliba, O., Choubey, D., Huang, C. D., Krymskaya, V. P., Eszterhas, A., Lazaar, A. L., and Panettieri, R. A., Jr. (2003) Am. J. Physiol. 284, L1063–L1071 33. MacEwan, D. J. (2002) Cell. Signal. 14, 477– 492 34. Taniguchi, T. (1989) J. Interferon Res. 9, 633– 640 35. Sanceau, J., Boyd, D. D., Seiki, M., and Bauvois, B. (2002) J. Biol. Chem. 277, 35766 –35775 36. Guo, D., Dunbar, J. D., Yang, C. H., Pfeffer, L. M., and Donner, D. B. (1998) J. Immunol. 160, 2742–2750 37. Miscia, S., Marchisio, M., Grilli, A., Di Valerio, V., Centurione, L., Sabatino, G., Garaci, F., Zauli, G., Bonvini, E., and Di Baldassarre, A. (2002) Cell Growth Differ. 13, 13–18 38. Wang, Y., Wu, T. R., Cai, S., Welte, T., and Chin, Y. E. (2000) Mol. Cell. Biol. 20, 4505– 4512 39. Mukhopadhyay, A., Shishodia, S., Fu, X. Y., and Aggarwal, B. B. (2002) J. Cell. Biochem. 84, 803– 815 40. Palmer, H., and Libby, P. (1992) Lab. Investig. 66, 715–721 41. Rodel, J., Assefa, S., Prochnau, D., Woytas, M., Hartmann, M., Groh, A., and Straube, E. (2001) FEMS Immunol. Med. Microbiol. 32, 9 –15 42. Cremer, I., Ghysdael, J., and Vieillard, V. (2002) FEBS Lett. 511, 41– 45 43. Dhib-Jalbut, S., Jiang, H., and Williams, G. J. (1996) J. Neuroimmunol. 71, 215–222 44. Horikoshi, T., Ezoe, K., Nakagawa, H., Eguchi, H., Hanada, N., and Hamaoka, S. (1995) FEBS Lett. 363, 141–144 45. Walter, M. J., Look, D. C., Tidwell, R. M., Roswit, W. T., and Holtzman, M. J. (1997) J. Biol. Chem. 272, 28582–28589 46. Defazio, G., Livrea, P., Giorelli, M., Martino, D., Roselli, F., Ricchiuti, F., and Trojano, M. (2000) Brain Res. 881, 227–230 47. Pang, L., and Knox, A. (2001) FASEB J. 15, 261–269 48. Watson, M. L., Grix, S. P., Jordan, N. J., Place, G. A., Dodd, S., Leithead, J., Poll, C. T., Yoshimura, T., and Westwick, J. (1998) Cytokine 10, 346 –352 49. Panettieri, R. A., Jr. (1998) Am. J. Respir. Crit. Care Med. 158, S133–S140