Tobacco smoke control of mucin production in lung cells requires oxygen radicals AP-1 and JNK.

In smokers' lungs, excessive mucus clogs small airways, impairing respiration and promoting recurrent infection. A breakthrough in understanding this pathology was the realization that smoke could directly stimulate mucin synthesis in lung epithelial cells and that this phenomenon was dependent on the cell surface receptor for epidermal growth factor, EGFR. Distal steps in the smoke-triggered pathway have not yet been determined. We report here that the predominant airway mucin (MUC5AC) undergoes transcriptional up-regulation in response to tobacco smoke; this is mediated by an AP-1-containing response element, which binds JunD and Fra-2. These transcription factors require phosphorylation by upstream kinases JNK and ERK, respectively. Whereas ERK activation results from the upstream activation of EGFR, JNK activation is chiefly EGFR-independent. Our experiments demonstrated that smoke activates JNK via a Src-dependent, EGFR-independent signaling cascade initiated by smoke-induced reactive oxygen species. Taken together with our earlier results, these data indicate that the induction of mucin by smoke is the combined effect of mutually independent, reactive oxygen species activation of both EGFR and JNK.

The primary cause of morbidity in chronic bronchitis is mucin overproduction, a phenomenon for which the molecular pathogenesis is unknown. Inflammatory cells are abundant in smokers' airways (1)(2)(3) and are capable of stimulating mucin production (4 -7), suggesting that at least some of the excessive mucin in smokers' lungs is secondary to inflammation.
In addition, however, smoke itself can induce mucin synthesis in lung cells (8,9). The question of how this occurs is complex in that smoke, a composite of irritant molecules including acetaldehyde, hydroquinone, formaldehyde, benzo-[a]pyrene, cresol, nicotine, catechol, acrolein, coumarin, anthracene, nitrogen oxides, and heavy metals (10,11) may act on lung epithelial cells in diverse ways. For example, the induction of cytochrome P450 by tobacco smoke (12) is mediated by binding of the aryl hydrocarbon nuclear receptor to a dioxin response element in the 5Ј-flank of the gene, but the induction of the ␥-glutamylcysteine synthetase heavy subunit (␥-GCS-HS) gene is mediated by the binding of a c-Jun/c-Jun homodimer to an AP-1-like response element (13).
Previous reports have implicated the receptor for epidermal growth factor (EGFR) 1 in the induction of mucin gene MUC5AC by smoke (9). Consistent with a role for EGFR in mucin induction, an EGF response element has been identified 200 bp upstream of the MUC5AC gene (14). The response of this element to EGFR ligands EGF and transforming growth factor-␣ is mediated by Sp1. One might predict from these data that the induction of MUC5AC by smoke would depend on interaction between the EGF response element at Ϫ200 bp and Sp1.
In contrast, in the present study we show that MUC5AC is controlled principally by a smoke response element ϳ3 kb upstream of the EGF response element. This element is AP-1dependent and is bound by JunD/Fra-2 heterodimers. Such heterodimers typically require both JNK and ERK kinases for transcriptional activity. Our results indicate that the ability of smoke to stimulate MUC5AC through AP-1 strongly depends on both JNK and ERK. Although ERK activation was a consequence of EGFR activation, JNK activation was not. Instead, JNK activation by smoke was mediated by Src in an EGFRindependent manner triggered by reactive oxygen species (ROS). The results indicate that both EGFR-dependent and independent mechanisms cooperate to stimulate mucin production in response to tobacco smoke.

MATERIALS AND METHODS
Reagents-The chemical inhibitors AG1478 and PP2 were obtained from Calbiochem (EMD Biosciences, Inc. La Jolla, CA). SP600125 was purchased from Biomol (Plymouth Meeting, PA). Recombinant human EGF was from Invitrogen (Carlsbad, CA). All other chemicals were from Sigma. Antibodies specific to JunD, Fra-2, and Src were from Santa Cruz Biotechnology (Santa Cruz, CA). Antibodies to phospho-JNK and phospho-ERK were from New England Biolabs (Beverly, MA). The antibody specific to phospho-Src was from BIOSOURCE (Camarillo, CA). The JNK1 dominant negative mutant was a gift from Dr. Roger Davis (Howard Hughes Medical Institute, Worcester, MA).
In Situ Hybridization-Human smoker or non-smoker bronchial tissue was obtained at autopsy from pathologist Dr. Walt Finkbeiner (University of California San Francisco). The tissues were rinsed in sterile PBS to remove secretions and fixed in 4% paraformaldehyde in PBS overnight at 4°C. Fixed tissue was embedded in paraffin, sectioned, and placed on prepared slides. Labeled RNA probes were prepared by incorporating [ 35 S]UTP into MUC5AC-specific sense and antisense transcripts using T3 and T7 polymerases. Hybridizations were performed as described previously (15). Briefly, sections were dewaxed, rehydrated, and rinsed in PBS for 5 min and then fixed in 4% paraformaldehyde for 20 min. After being rinsed twice in PBS, they were treated with 20 g/ml Pronase in TE (1 mM Tris HCl, pH 8.0, 0.1 mM EDTA) for 7.5 min. The fixation step was repeated for 5 min, and the slides were rinsed twice in PBS. The sections were acetylated in acetic anhydride in 0.1 M triethanolamine-HCl (pH 7.5), rinsed, and dehydrated through ethanol. Sections were hybridized in 50% deionized formamide, 0.3 M NaCl, 20 mM Tris HCl, pH 8.0, 5 mM EDTA, 10 mM NaPO 4 , pH 8.0, 10% dextran sulfate, 1ϫ Denhardt's solution, 0.5 mg/ml yeast tRNA, and a final probe concentration of 1 ϫ 10 4 cpm/l overnight at 50°C. After hybridization, sections were rinsed in 50% deionized formamide, 2ϫ SSC (NaCl, Na citrate), 0.1 M dithiothreitol for 30 min at 65°C and then three times for 10 min each in 0.5 M NaCl, 10 mM Tris-HCl, pH 8.0, 5 mM EDTA at 37°C. After treatment with RNase A (20 g/ml) for 30 min at 37°C, the sections were rinsed in 2ϫ SSC, 15 min at 37°C, and 0.1ϫ SSC, 15 min at 37°C, and dehydrated. Autoradiography was performed using Ilford K.5 nuclear research emulsion (Polysciences, Inc., Warrington, PA).
Cell Culture-NCI-H292 cells (lung mucoepidermoid carcinoma) and BEAS-2B cells (normal bronchial epithelial cells, virus transformed) were maintained in RPMI 1640, 10% fetal calf serum with 100 units/ liter penicillin and 100 mg/ml streptomycin. HM3 cells (colon carcinoma) were maintained in DME-H21 with 10% fetal calf serum and antibiotics as above. Most experiments were done in NCI-H292 cells. The exceptions are indicated in the figure legends.
Preparation of Smoke-containing Medium-Smoke particulates were generated in specially designed animal exposure chambers operated by Dr. Kent Pinkerton at the University of California, Davis (16). Pallflex® borosilicate filters (Pall Gelman, East Hills, NY) were inserted in line during the operation of the chambers. Total suspended smoke particles deposited on each filter were calculated by weighing filters before and after smoke exposure. Filters were mailed to our laboratory where they were stored at 4°C until use. Prior to each experiment, we incubated filters for 1 h at 37°C, 5% CO 2 in a volume of serum-free medium (SFM) to provide a final concentration of 1.0 mg of smoke particulate per ml of SFM, diluted as needed.
Luciferase Assay-Epithelial cells were seeded at 35,000 cells/cm 2 in 48-well plates and allowed to recover overnight. Transfections were performed using 0.2 g of DNA/well and FuGENE 6 (Roche Applied Science) at a ratio of 1:6 for 18 h in RPMI 1640 containing 10% fetal calf serum without antibiotics. Cells were recovered in low serum medium RPMI 1640 containing 1% fetal calf serum and antibiotics for 6 h. For smoke exposure, cells were incubated in smoke-containing medium at the indicated concentrations for 1 h and then chased in low serum medium for 18 h. The plates were rinsed once in PBS, lysed in 100 -150 l of luciferase lysis buffer, and frozen at Ϫ80°C until read in a Monolight 2010 luminometer.
Cloning and Generation of Deletion/Mutation Constructs-The 5Јflanking region of the human MUC5AC gene was cloned as described previously (17). Briefly, the 5Ј sequence of the human gastric mucin cDNA HGM-1 (18) was used to obtain the transcription start site of the hMUC5AC gene using repetitive rounds of RACE (rapid amplification of cDNA ends)-PCR. Using primers from the 5Ј-untranslated region, genomic DNA was amplified from DNA provided in the human Pro-moterFinder TM DNA walking kit (Clontech, Palo Alto, CA) according to instructions provided by the manufacturer. Approximately 3.7 kb of the 5Ј-flanking region of the human MUC5AC gene was obtained, sequenced, and cloned into pGL3-Basic luciferase vector (Promega Corp., Madison, WI). The nucleotide sequence of the MUC5AC flanking region is available in the GenBank TM data base under accession number AF016834. Deletion constructs were obtained by combining restriction digestion with PCR using gene-specific primers. Internal mutations were made by PCR amplification using primers containing the mutant sequence. The resulting fragments were cloned into pGL3-Basic vector. All constructs were verified by DNA sequencing.
Gel (Electrophoretic Mobility Shift) Assay-Nuclear extracts from NCI-H292 cells were prepared using the NE-PER® nuclear extraction kit (Pierce, Rockford, IL) according to the manufacturer's instructions. Prior to extraction, serum-deprived cells were exposed to smoke-containing or control medium. The protein concentration of the cell extract was determined using a Bradford protein assay kit (Bio-Rad, Hercules, CA) with bovine serum albumin as standard. Various double-stranded oligonucleotide probes as indicated in each experiment were synthesized on the basis of the results of luciferase assay experiments described under "Results." Oligonucleotides were labeled by [␥-32 P]ATP (PerkinElmer Life Sciences) and T4 polynucleotide kinase and purified on a MicroSpin TM Sephadex G-25 spin column (Amersham Biosciences). Probes were incubated at 30°C for 15 min with nuclear extract (5-10 g of protein) in a solution containing 10 mM Tris-HCl (pH 7.5), 50 mM NaCl, 1 mM MgCl 2 , 0.5 mM EDTA, 0.5 mM dithiothreitol, 0.05 mM poly(dI-dC)⅐poly(dI-dC) and 4% (v/v) glycerol. For cold competition assays, a 25-fold molar excess of unlabeled oligonucleotide was added to the reaction and incubated for 30 min on ice before adding radiolabeled probe. For supershift analysis, 1 l of the antibody of interest (Santa Cruz Biotechnology) was added to the reaction mixture preincubated as described above. Samples were applied to polyacrylamide gels under native conditions in high ionic strength buffer. The dried gel was exposed to x-ray film at Ϫ70°C for 2-12 h with double intensifying screens.
Phospho-JNK Translocation by Immunofluorescence-Human immortalized bronchial epithelial cells (BEAS-2B) were grown on glass coverslips and synchronized in low (1%) serum. The cells were incubated for 20 min in the presence or absence of smoke-containing medium. The cells were then washed with cold PBS, fixed in methanol: acetone (50:50), permeabilized, and immunoreacted with antibody specific to phospho-JNK (Thr-183/Tyr-185). The phospho-JNK was visualized with Cy3 conjugated to goat anti-rabbit IgG (Amersham Biosciences) prior to viewing with a Nikon Eclipse E600 microscope using an NCF Fluor 40 objective lens.
Semi-quantitative RT-PCR in Cultured Cells-The expression of MUC5AC by human lung epithelial cells was assessed after stimulation with serum-free medium or smoke-containing medium for 1 h followed by low serum medium for 18 h. QuantumRNA TM 18 S Classic II Internal Standards (Ambion Inc., Austin, TX) were used as the endogenous standards and were amplified in the same linear range, at the same level, and under the same conditions as the MUC5AC mRNA with forward primer (5Ј-CACTTCTCAACGTTTGACGGGAAG-3Ј) (ϩ1350) and reverse primer (5Ј-GTGCTGAGCATACTTCTCATTCTC-3Ј) (ϩ1893). Starting with 2.0 g of total RNA for each sample, RT-PCR was carried out using the SuperScript first-strand synthesis system for RT-PCR (Invitrogen), as described by the manufacturer, with random hexamer primers. A total of 100 ng of RNA served as a template for PCR with 400 nM forward and reverse MUC5AC primers added to pre-mixed Ready-To-Go TM PCR beads (Amersham Biosciences) on ice for a final volume of 25 l. QuantumRNA 18S Classic II Internal Standards were used as controls. The endogenous standards were amplified in the same linear range, at the same level, and under the same conditions as the MUC5AC mRNA. After reaching a denaturing temperature of 94°C, tubes were added and cycled 39 times using 94°C for 30 s to denature, 59°C for 30 s to anneal, and 72°C for 30 s to extend. PCR products were analyzed by agarose gel electrophoresis and visualized with ethidium bromide staining.
Animal Exposures to Tobacco Smoke and PCR Analysis-All procedures involving animals were consistent with National Institutes of Health guidelines and were reviewed and approved by the animal welfare committee of the Inhalation Toxicology Research Institute, Albuquerque, NM. Male F344/Crl rats were obtained at 8 -9 weeks of age (CDF (F344)/CrlBR rats, Charles River Laboratories, Raleigh, NC). Animals were placed one per cage into whole-body inhalation chambers (H2000, Laboratory Products, Maywood, NJ) and supplied with filtered air for quarantine and conditioning. The chambers were maintained at 12 air changes/hour at a temperature of 24 Ϯ 2°C and relative humidity of 40 -70%. Rats were randomly assigned by weight to experimental groups. Exposure to cigarette smoke for 10 days was begun when the animals were 11-13 weeks old. Diluted mainstream cigarette smoke was generated from 1R3 research cigarettes (Tobacco and Health Research Institute, Lexington, KY) by modified AMESA Type 1300 automated smoking machines (AMESA, Geneva, Switzerland) as described previously (19). Cigarettes were held at 24°C and 50 -70% relative humidity before use and then puffed twice per minute at a 70-ml puff volume taken over 2 s. Each cigarette was puffed 6 -7 times; the fresh smoke was diluted with filtered air and delivered to the exposure chambers. The mass concentration of cigarette smoke total particulate matter (TPM) was determined gravimetrically, CO concentrations were measured periodically using an infrared analyzer (Beckman Industries), and smoke particulate size was measured by cascade impaction. Exposures were conducted daily for 6 h/day for 10 days to a target cigarette smoke concentration of 250 mg TPM/m 3 . Filtered air controls were held in a separate chamber in the same room. At the end of the exposure the right lung (fifth generation bronchi) was dissected, and total RNA was prepared using the TRIzol method (20). RT-PCR was performed as described previously (21) using the following primers: 5Ј-GGTACCCAGGTCTACACCTACTC-3Ј, rat MUC5AC forward primer; 5Ј-CATCATTCCTGTAGCAGTAGTGAGG-3Ј, rat MUC5AC reverse primer; 5Ј-CTGGAGAAACCTGCCAAGTATGAT-3Ј, rat GAPDH for-ward primer; 5Ј-TTCTTACTCCTTGGAGGCCATGTA-3Ј, rat GAPDH reverse primer.
Western Blot-Cell lysates were prepared by adding 500 l of lysis buffer directly to cells plated in a 6-well plate. Equal amounts of lysate were sonicated for 30 s on ice, and cell debris was removed by centrifugation. For immunoblot analysis, 10 g of protein was subjected to SDS-PAGE and transferred to nitrocellulose at 100 V for 1 h. Immunostaining was done by routine methods as instructed by the manufacturer using specific antibodies as indicated.
JNK Assay-JNK activity was assayed using the SAPK/JNK assay system (Cell Signaling Technology, Beverly, MA) according to manufacturer's instructions. NCI-H292 cells grown to confluence in 6-well plates were treated as indicated. The cells were lysed in 0.5 ml of lysis buffer containing 1 mM phenylmethylsulfonyl fluoride, scraped, and sonicated four times for 5 s. The supernatant was cleared by centrifu-

FIG. 1. Smoke stimulates MUC5AC expression in vivo and in vitro.
A, in situ hybridization. Bronchial tissue obtained at autopsy was fixed in paraformaldehyde and embedded in paraffin. Sections incubated with a 35 S-labeled RNA probe for MUC5AC revealed dramatically more MUC5AC mRNA in samples from smokers with chronic obstructive lung disease than in samples from control non-smokers. L, lumen of airway. The dark areas are epithelial cells containing autoradiographic label. The epithelium is pathologically thick and convoluted in smokers' airways. B, RT-PCR showing that the endogenous MUC5AC gene is up-regulated by smoke. NCI-H292 cells were incubated in SFM or smoke condensate-containing SFM (Smoke) for 1 h followed by a 16-h chase in medium containing 1% fetal calf serum prior to RNA extraction. For RT-PCR, samples were normalized for sample loading using Ambion competimer 18 S RNA. C, induction of Muc5ac in lungs of rats exposed to smoke. Rats received whole-body smoke exposure for 10 days as described under "Materials and Methods." RNA extracted from equivalent airway generations in each animal was analyzed by RT-PCR. Sample loading was normalized using a probe for glyceraldehyde-3-phosphate dehydrogenase (GAPDH). D, luciferase reporter assay in NCI-H292 cells. Luciferase construct was driven by 3752 nucleotides of the MUC5AC 5Ј-flanking region. Activity in the presence of SFM alone is assigned a value of 1 for reference. The -fold increase in reporter activity is shown for smoke and EGF.

FIG. 2.
Mapping of smoke response element to AP-1 sites at approximately ؊3500 bp. A, deletion mutation of MUC5AC-luciferase reporter constructs revealed a major smoke response element between Ϫ3700 and Ϫ3337 nucleotides. B, presence of the smoke response element greatly increased luciferase response to smoke but not to EGF (25 ng/ml). The experiment was performed in NCI-H292 cells. C, site-directed mutagenesis of four potential transcription factor binding sites within the smoke response element indicated probable cooperativity among two AP-1 sites (Dist, distal; Prox, proximal), one NFB site, and an RXR site. The experiment was performed in BEAS-2B cells. Wt, wild type; SMK, smoke.
gation. Lysates (250 l) were incubated with a c-Jun fusion protein linked to agarose beads to pull down SAPK enzyme. The beads were washed and then incubated with kinase buffer and 100 M ATP for 30 min at 30°C to allow SAPK phosphorylation of the c-Jun substrate. Phospho-c-Jun (Ser-63) antibody was used to measure SAPK/JNK activity by immunoblotting.

MUC5AC Mucin Is Up-regulated by Smoke in Vivo and in
Vitro-The effects of smoke on lung mucus production are evident from clinical findings showing that mucus volume increases from several milliliters/day in the airways of healthy individuals to several hundred milliliters/day in smokers with chronic bronchitis (22)(23)(24)(25). Mucin, the principal macromolecular component of mucus, is encoded by a family of genes of which MUC5AC appears to be the most highly expressed in the human lung (26). We found MUC5AC mRNA levels to be greatly elevated in the lungs of human smokers, as measured by in situ hybridization (Fig. 1A). The direct application of smoke to human lung cells essentially reproduced this phenomenon in culture (Fig 1B). Moreover, we could mimic this phenomenon in rats by exposing them to smoke for 10 days in specially designed chambers (19). Data in Fig. 1C show that whereas MUC5AC mRNA was undetectable in lung homogenates from 5 individual rats exposed to filtered air, it was clearly visible in 4 of 5 rats exposed to tobacco smoke.
These increases in steady state MUC5AC mRNA (Fig. 1, A-C) could have resulted from increased transcription or increased stability of the mRNA or both. To examine the possibility of transcriptional control, we transfected into human lung cells a luciferase reporter gene driven by 3700 nucleotides of the human MUC5AC 5Ј-flanking sequence. Upon exposure of the transfected cells to smoke extract, we observed an ϳ7-fold increase in luciferase activity (Fig. 1D), indicating that smoke induces MUC5AC at the level of transcription. Despite the involvement of EGFR in the induction of MUC5AC by smoke (8,9), saturating concentrations of EGF (25-500 ng/ml) evoked a relatively small (2.5-fold) response (Fig. 1D) in agreement with Perrais et al. (14). The EGFR ligands transforming growth factor-␣ and amphiregulin elicited responses of similarly small magnitude (not shown).
The Induction of MUC5AC by Smoke Is Mediated by AP-1 Sites-To better understand the smoke induction mechanism, we used deletion mutagenesis to map smoke response elements upstream of MUC5AC. Although preliminary studies revealed weak responsiveness within 941 nucleotides of the transcription start site (containing the EGF response element at Ϫ202 (14)), the major response element was at Ϫ3700/Ϫ3337 (Fig.  2A).
The DNA fragment containing this element was critical for the MUC5AC response to smoke but not for its response to EGF (Fig. 2B). Reference to standard data bases revealed four major transcription factor binding sites in this region: two AP-1 sites (TGACTCA), one RXR site (TGTTCAGAGGTCA), and one NFB site (GGGGCACTCCCC) (Fig. 2C). Mutations targeting each of these sites decreased responsiveness of MUC5AC (Ϫ3752/ϩ1)-luciferase to smoke, with the greatest effects resulting from mutation of the AP-1 sites (Fig. 2C).
Smoke Stimulates JunD and Fra-2 Binding to AP-1 Sites-Pursuing the role of each of these sites, we performed electromobility gel shift assays using probes corresponding to each site in the presence and absence of smoke. Specificity was confirmed by the use of cold (unlabeled) probe inhibitors. Clearly, smoke induced protein binding to the proximal AP-1 site (Fig. 3A, left panel). Because of the close proximity of the RXR and distal AP-1 sites (nucleotides Ϫ3577/Ϫ3571 and Ϫ3565/Ϫ3553 respectively), it was difficult to construct individual probes. Instead, we constructed a probe containing both sites. This probe underwent smoke-induced protein binding (Fig. 3A, right panel). Appropriate mutagenesis of cold probe competitors showed that mutation of the AP-1 site interfered with competitive binding much more than mutagenesis of the RXR site (Fig. 3B). This was consistent with the functional data shown in Fig. 2C. Together, these gel shifts supported the idea that both AP-1 sites play a critical role in the response to smoke.
AP-1 motifs (also known as TPA (12-O-tetradecanoylphorbol-13-acetate) response elements) commonly bind either Jun homodimers or Jun/Fos heterodimers. Although the varying quality and affinity of antibodies for Jun and Fos family members (c-Jun, JunB, JunD, c-Fos, Fra-1, and Fra-2) make it difficult to quantitatively assess the importance of each specific protein, our supershift assays indicated that smoke stimulated both JunD and Fra-2 to bind to the distal AP-1 site (Fig 3B). Smoke not only activated existing pools of JunD and Fra-2 but also stimulated synthesis of these early response genes, as shown by increased levels of each protein in cultured lung epithelial cells (Fig. 3C). Smoke also seemed to alter protein binding to the NFB site. Although the site is occupied by proteins under control conditions, smoke seems to increase the relative abundance of certain of these proteins, as reflected by increased intensity of the lower band in Fig 3D. Role of JNK and ERK-The information regarding JunD and Fra-2 suggested potential upstream signaling mechanisms. In particular, Jun family members are stimulated in a JNK-dependent manner via phosphorylation of serine or threonine residues in their N-terminal regions (27,28), whereas Fos family members are activated by ERK-dependent phosphoryl- . Cells were exposed to smoke (SMK) for 2-60 min prior to lysis, as indicated, and prepared for Western blotting. The experiment was performed in HM3 cells. C, immunofluorescence showing nuclear translocation of JNK in response to smoke. Red, phosphorylated JNK; blue, nuclei (4 -6-diamidino-2phenylindole (DAPI)); violet in merged images indicates nuclear translocation of JNK in response to smoke (SMK). The boxed area in the lower right panel is enlarged below. This experiment was performed in BEAS-2B cells.
ation of Ser or Thr in their C-terminal region (29). Based on our supershift data implicating JunD and Fra-2 in the smoke response, it seemed likely that both JNK and ERK would be stimulated by smoke and required for MUC5AC responsiveness. Indeed, Western blots measuring ERK and JNK phosphorylation (activation) showed activation of both kinases; this peaked at 5-15 min after the onset of smoke exposure (Fig. 4,  A and B). Immunofluorescence imaging extended these results by showing that smoke not only induced the phosphorylation of JNK but also its migration into the nucleus (Fig. 4C), a prerequisite for transcriptional activation through AP-1.
Smoke-induced phosphorylation of JNK and ERK suggested that these two kinases might participate in the induction of MUC5AC in lung cells. To investigate this possibility, we manipulated JNK and ERK activity and assessed the effects of these treatments on the responsiveness of MUC5AC to smoke. Using a dominant negative JNK mutant, we observed a major reduction in both the MUC5AC-luciferase response to smoke (Fig. 5A) and the response of the endogenous mucin gene (Fig.  5B). The JNK-specific inhibitor SP600125 caused a similar effect (Fig. 5C). Interestingly, this inhibitor had no effect on the modest induction of MUC5AC by EGF (Fig. 5C), supporting the idea that the smoke and EGF induction mechanisms differ. Based on the role of ERK in both activating (30) and inducing (31) Fos family proteins, it was not surprising that ERK inhi-bition (by PD98059) also reduced MUC5AC induction by smoke (Fig. 5D). ERK dependence had earlier been reported for MUC5AC induction by EGF (14). Thus, ERK is a signaling component common to the induction of MUC5AC by both smoke and EGF, whereas the role of JNK appears to be smoke-specific.
Role of ROS and Src-Given the important role of JNK, we next tried to identify upstream mediators actually linking JNK to smoke exposure. Recently, we reported that one of the earliest effects of smoke on lung epithelial cells was the generation of oxygen radicals (reactive oxygen species) (32). ROS play an important role in stress-induced signaling and have been shown to stimulate Src family members (33). Because our preliminary studies suggested a role for Src in the smoke response, and Src has been shown to play a role in the activation of JNK (34,35), we considered a potential role for ROS and Src in the activation of JNK by tobacco smoke. Supporting the relevance of this cascade to smoke signaling, we observed (Fig. 6A) that Src phosphorylation on tyrosine 418 (a reflection of Src activation) was elicited by smoke. Moreover, this phosphorylation was blocked in the presence of the ROS scavenger dimethylthiourea as well as by the Src inhibitor PP2 (Fig. 6A). That ROS and Src are required for the induction of mucin was indicated by data showing that both dimethylthiourea and PP2 strongly RT-PCR was normalized for sample loading using Ambion competimer 18 S RNA. C, the JNK inhibitor SP-600125 reduced MUC5AC-luciferase response to smoke but not to EGF. D, the ERK-1/2 inhibitor PD98059 reduced the MUC5AC-luciferase response to smoke. The same inhibitor was shown to reduce the MUC5AC response to EGF in an earlier report (14).
We previously reported that smoke-induced phosphorylation of EGFR was ROS-dependent ((32)) and have also observed that it is Src-dependent. 2 This raised the possibility that ROS, Src, EGFR, and JNK might participate in a single linear pathway to stimulate MUC5AC. Opposing such a model, however, were data showing that smoke-induced JNK activity (as reflected by phosphorylation of its substrate c-Jun) was sensitive to ROS and Src inhibitors (N-acetylcysteine and PP2) but not to inhibitors of EGFR kinase (e.g. AG1478) (Fig. 6, C and D). These results, taken together with those reported earlier (32), support the idea that the induction of ROS and Src by smoke is a critical early step leading to the dual but mutually independ-ent activation of both EGFR and JNK. We propose that ERK, a downstream effector of EGFR, cooperates with Src-activated JNK to activate Fra-2 and JunD, respectively. Their combined effect mediates MUC5AC transcription in response to smoke.

DISCUSSION
Mucin overproduction is a common finding in smokers' airways and an important source of clinical symptoms. A breakthrough in understanding this phenomenon came with the realization that smoke directly stimulates mucin synthesis in epithelial cells (8,9). The signaling cascade mediating this phenomenon has been shown to involve EGFR at the cell surface, but the distal elements of the pathway have not been described. In the present study, we determined that smokeinduced mucin transcription was directly controlled by a smoke response element located at Ϫ3700/Ϫ3337 in the MUC5AC 5Ј-flanking region. Site-specific mutagenesis revealed that the major functional components of this region were two AP-1 sites. Like other AP-1-mediated transcription, mucin induction by smoke was dependent upon the activation of both ERK and JNK. Clearly, the involvement of EGFR in smoke signaling is sufficient to account for ERK activation, but our data showed that the activation of JNK by smoke could not be attributed to EGFR. The JNK activation mechanism, although still incompletely understood, appears to require the stimulation of Src by smoke-induced ROS and the subsequent phosphorylation of JNK by Src.
As shown in Fig. 2, the smoke response element contains, in addition to two AP-1 motifs (both TGACTCA), an NFB site and an RXR (retinoid receptor) site. Mutagenesis of each site reduced smoke-induced transcription in vitro, suggesting that the four sites act cooperatively in vivo. Cooperation between NFB and AP-1 has been described previously (e.g. see Ref. 36). The close proximity of the RXR and distal AP-1 sites strongly suggests that protein binding to the AP-1 site is controlled in part by the identity of proteins binding to the RXR site. In particular, the identity of RXR-binding proteins may dictate the identity of Jun and Fos family members binding to AP-1. This, in turn, can profoundly affect the activation potential of the AP-1 site and can act to customize gene activation in response to a particular stimulus (37).
Fos family proteins (c-Fos, Fra-1, and Fra-2) are substrates of the protein kinase ERK, whereas Jun family proteins are phosphorylated by JNK. Phosphorylation of these AP-1-binding proteins enhances their transactivation potential and therefore plays an important role in AP-1-dependent gene transcription. Regarding the three Jun family proteins (c-Jun, 2 H. Lemjabbar and C. Basbaum, unpublished data.  ). B, luciferase reporter assay in HM3 cells. Luciferase construct was driven by 3752 nucleotides of the MUC5AC 5Ј-flanking region. Activity in the presence of SFM alone is assigned a value of 1 for reference. The -fold increase in reporter activity is shown for smoke in the presence and absence of the ROS scavenger dimethylthiourea (100 M) and the Src inhibitor PP2 (5 M). C, JNK assay in which JNK activity is reflected by phosphorylation of the JNK substrate c-Jun. The stimulation of JNK by smoke was inhibited by the anti-oxidant N-acetylcysteine (NAC, 20 mM) and the Src inhibitor PP2 (5 M) but not by the EGFR kinase inhibitor AG1478 (AG, 5 M). D, JNK kinase assay as described in C, indicating that JNK is activated weakly by EGF (100 ng/ml) and strongly by smoke (1 mg/ml). The effect of smoke on JNK was not altered by the EGFR kinase inhibitor AG1478 (5 M). JunB, and JunD), it was initially thought that only c-Jun was efficiently phosphorylated by JNK; JunB was thought to lack suitable phosphoacceptor sites, whereas JunD was thought to lack a suitable JNK docking site (38). Subsequent reports have indicated, however, that all three are legitimate JNK substrates. Phosphoacceptor sites have been identified on the JunB protein at Thr-102 and Thr-104 (39), whereas a JNK docking site has been identified on the JunD protein at Leu-57/Leu-59 (40). Consistent with supershift data suggesting roles for both JunD and Fra-2 in the induction of mucin by smoke, our results showed that mucin induction was dependent on both JNK and ERK. Whereas ERK activation by smoke could be easily explained via the upstream activation of EGFR, the mode of JNK activation was unclear.
Based on data showing that one of the early events occurring in smoke-exposed cells is the generation of ROS (32), we examined the possibility that ROS were involved in JNK activation. At least two mechanisms have been shown to link ROS with JNK. First, it has been shown that ROS are able to modify thioredoxin, which exists in an intracellular complex with the mitogen-activated protein kinase kinase kinase, ASK-1. ROS exposure provokes thioredoxin to release ASK-1, relieving its repression and permitting it to phosphorylate and thereby stimulate JNK through its upstream kinases (41,42).
Secondly, ROS have been reported to stimulate Src (43)(44)(45), which has been shown to stimulate JNK through Cas (34). Available evidence suggests that ROS stimulate Src indirectly, i.e. by inhibiting protein tyrosine phosphatases (PTP) such as CD45. PTPs are sensitive to inactivation by ROS due to the presence of a universal, highly reactive, negatively charged cysteine residue in the PTP catalytic site (46). Impairment of PTP activity results in elevated levels of tyrosine phosphorylation on cellular proteins. In the case of Src family proteins, two key tyrosine residues control their activity; one is inhibitory and the other is excitatory. The exposure of cells to H 2 O 2 has been found to promote phosphorylation of both tyrosines on the Src family member Lck (47). The activity of Lck was increased under these conditions, implying that the activation of Src family kinases by phosphorylation of the activating tyrosine is dominant over the inhibition induced by phosphorylation of the inhibitory tyrosine (47). Based on these earlier reports and the findings of the present study, we propose that the stimulation of JNK by smoke is a consequence of ROS-dependent PTP inhibition. This, in turn, potentiates Src phosphorylation, thereby activating it to stimulate JNK through Cas (34).
Not only does smoke exposure cause an elevation in cytoplasmic ROS, but smoke itself contains ROS (48). We believe that it is the intracellularly generated ROS and not the extracellular, smoke-associated ROS that trigger the effects of smoke on epithelial cells, based on experiments showing that diphenyleneiodonium, an inhibitor of cellular NADPH oxidase, essentially abolished the presence of ROS in smoke-exposed cells (32). This led us to conclude that the ROS crucial to the effects of smoke in our in vitro studies originate in the epithelial cell itself, rather than from exogenous smoke. Notably, however, this does not rule out a potential role for smoke radicals in vivo.
In summary, the results of the present study, taken together with our results published earlier (32), indicate that intracellularly generated ROS are an early and critical mediator of pathology in the smoker's lung (Fig. 7). The earliest smoke effect for which we have evidence thus far is the stimulation of NADPH oxidase to generate ROS (32). The ROS, in turn, activate at least two downstream signaling cascades. The first involves the activation of tumor necrosis factor-␣-converting enzyme (TACE), amphiregulin, and EGFR (32), and the second activates Src and subsequently JNK (this study). Although the first pathway appears sufficient to stimulate lung cell proliferation, mucin induction depends on the cooperative effect of both pathways to activate JunD and Fra-2 and thereby to stimulate transcription of MUC5AC through AP-1. Because MUC5AC is the predominant lung mucin and is strongly up-regulated in the lungs of smokers (Fig. 1), the mechanisms we describe should help explain how smoke exposure leads to mucus hypersecretion, airway plugging, and the associated recurrent infection seen in patients with smoke-induced chronic obstructive lung disease.