Role of hyaluronan and reactive oxygen species in tissue kallikrein-mediated epidermal growth factor receptor activation in human airways.

In human airways, oxidative stress-induced submucosal gland cell hypertrophy and hyperplasia, histological features of chronic bronchitis, have been linked to epidermal growth factor receptor (EGFR) activation. To explore mechanisms of oxidative stress-induced EGFR activation and signaling, primary cultures of human tracheal submucosal gland (SMG) cells were used to assess EGFR ligand release, EGFR phosphorylation, p44/42 MAPK phosphorylation, and mucin 5AC synthesis in response to reactive oxygen species generated by xanthine/xanthine oxidase (X/XO). Exposure to X/XO increased release of epidermal growth factor (EGF) from these cells, thereby activating EGFR, phosphorylating MAPK, and increasing mucin 5AC production. The importance of EGF was confirmed by transfection of small interfering RNA inhibiting pro-EGF production, which resulted in inhibition of EGFR and MAPK phosphorylation despite X/XO exposure. Blocking signaling by using specific protease inhibitors showed that tissue kallikrein (TK) processed pro-EGF in response to X/XO. Airway TK is bound and inactivated by luminal hyaluronan (HA), and treatment of submucosal gland cells with X/XO induced HA depolymerization and TK activation. These events were blocked by reactive oxygen species scavengers and addition of exogenous excess HA and TK inhibitors. Thus, HA plays a crucial role in regulating airway TK activity and thereby TK-mediated release of active EGF from human SMG cells. Sustained HA depolymerization is expected to cause TK activation, EGF release, and EGFR signaling and to lead to SMG cell hypertrophy and hyperplasia as well as mucus hypersecretion with subsequent airflow obstruction.

Mucus hypersecretion and characteristic changes in epithelial cell morphology, including goblet cell hyperplasia in major airways and metaplasia in small airways as well as submucosal gland hypertrophy, are pathophysiological and histological hallmarks of chronic bronchitis (1,2). Recent key observations have suggested that activation of the epidermal growth factor (EGF) 1 receptor (EGFR) signaling cascade is responsible for at least some of the morphological changes in the airway epithelium (3)(4)(5)47). Cigarette smoke, the major cause of chronic bronchitis in humans, has also been shown to induce mucin secretion and mucous cell hyperplasia, at least in part, via EGFR-mediated signaling (6). Because oxidative stress causes EGFR activation (7) and cigarette smoke is a major source of direct and indirect oxidative stress (8,9), it is not surprising that many of the effects of cigarette smoke on mucin secretion and mucous cell hyperplasia can be blocked by anti-oxidants (6). However, the mechanisms by which oxidative stress stimulates EGFR signaling are not fully understood.
The classic EGFR, c-ErbB1, is activated by different ligands including EGF, transforming growth factor-␣ (TGF-␣), heparin binding EGF-like growth factor (HB-EGF), amphiregulin, epiregulin, and betacellulin (10 -14). EGFR ligand precursors are synthesized as transmembrane proteins that are cleaved to their active form by extracellular proteases (15). Whereas TGF-␣, HB-EGF, amphiregulin, epiregulin, and betacellulin are processed by metalloproteases of a disintegrin and metalloprotease (ADAM) family (e.g. Ref. 16), pro-EGF is processed by a serine protease. This serine protease has been identified to be tissue kallikrein (TK) in salivary and mammary glands, and its activity has been found to be consistent with TK in the kidney (17)(18)(19).
Although TK is present in the airway (20,21), its potentially crucial role in EGF precursor activation and EGFR-mediated signaling has not been examined. Bronchial TK is in fact uniquely regulated in the airway. Studies from this laboratory have shown that TK is secreted with hyaluronan (HA) from submucosal glands, which inhibits its catalytic activity (21). We have also shown that HA immobilizes inactive bronchial TK at the epithelial surface, creating a pool of readily available yet inactive TK on the airway surface (22) and that HA degradation results in a dramatic increase in TK activity (21). Such a TK/HA pool also exists within the lumen of submucosal gland ducts as they contain secretions not yet expelled into the airway lumen. Because HA is depolymerized by oxidative stress (e.g. Ref. 23), increased airway TK activity is expected in diseases associated with increased airway ROS production. In fact, airway TK activity is increased in chronic bronchitis and other diseases associated with mucus hypersecretion compared with normal subjects (24,25), possibly due, at least in part, to continued degradation of HA by ROS (26 -28).
Based on these findings, we hypothesized that ROS-induced HA degradation results in TK activation and pro-EGF processing in the airways, providing a direct link between oxidative stress and activation of EGFR signaling. To examine this hypothesis, we used primary cultures of human airway submucosal gland (SMG) cells to study EGFR ligand release and EGFR signaling during oxidative stress.

Materials
All materials were purchased from Sigma unless otherwise specified.

Primary Cultures of Human Submucosal Gland Cells
Human tracheas and main bronchi from donor lungs rejected for transplantation were obtained through the University of Miami Life Alliance Organ Recovery Agency with approval from the local Institutional Review Board. The trachea and main bronchi were opened at the membranous portion, and the mucosa was dissected off the cartilage. Mucosal strips were digested with 0.05% protease (type 14) in Dulbecco's modified Eagle's medium (DMEM; Invitrogen) and incubated overnight at 4°C for 24 h to release ciliated epithelial cells (which were used for different experiments after shaking them off the strips). To release gland cells, the remaining tissue was incubated for another 24 h in 0.01% dispase in DMEM supplemented with penicillin (100 units/ml), streptomycin (0.25 g/ml), amphotericin B (2.5 g/ml), and gentamicin (50 g/ml) (referred to as DMEM-AB). Each strip was scraped after digestion to release the remaining gland cells. Cells were washed once in DMEM-AB, resuspended in DMEM-AB containing 0.25% trypsin-EDTA, and triturated to dissociate the remaining "clumps" of gland acini. Digestion was stopped with soybean trypsin inhibitor (SBTI, 500 g/ml), and cells were resuspended in DMEM-AB. After cell counting and viability determination by trypan blue exclusion (viability was always Ͼ95%), 5 ϫ 10 5 cells were plated on placental collagen-coated, 24-mm T-clear membrane filters inserts (Costar, Cambridge, MA). Cultures were maintained in submerged conditions by adding media to both the basolateral and apical sides consisting of 50% DMEM-AB, 50% Ham's F-12 supplemented with hydrocortisone (0.5 g/ml), insulin (5 g/ml), transferrin (10 g/ml), epinephrine (0.5 g/ml), triiodothyronine (6.5 ng/ml), and EGF (25 ng/ml). Cultures were maintained at 37°C in an atmosphere of 5% CO 2 in air. After 10 days of growth, 81 Ϯ 14% of the cells (100 cells counted per well; mean Ϯ S.E.) stained positive with Alcian blue-periodic acid-Schiff, a feature consistent with a glandular origin of the cells. Measured over an 18-h period, these SMG cells secreted TK (a specific serous cell product) and HA (3.51 Ϯ 0.20 g/mg cell lysate protein and 4.05 g/mg cell lysate protein, respectively).

Protocols
Confluent cell monolayers with resistivities of 926 Ϯ 13.5 ohms⅐cm were used for all studies. Exogenous EGF was removed from the media 48 h before experimentation. Immediately prior to all experiments, the basal medium was changed to RPMI, and the apical medium was replaced with PBS. After the studies, apical supernatants were collected, and cells were lysed with 20 mM sodium phosphate, 150 mM NaCl, 5 mM EDTA, 50 mM HEPES, 1% Triton X-100, 50 mM NaF, 1 mM sodium orthovanadate, 5 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, and 10 g/ml aprotinin, pH 7.8 (lysis buffer), for 30 min at 4°C. To remove insoluble material, cell lysates were centrifuged at 14,000 rpm for 5 min at 4°C. Supernatants and cell lysates were frozen at Ϫ20°C for later analysis.
In experiments designed to test the effects of reactive oxygen species (ROS) on EGFR ligand processing and signaling, SMG cells were exposed to X/XO (0.6 mM xanthine plus 0.05 units of xanthine oxidase) for 30 min in the absence or presence of the following: the ROS scavengers catalase (150 units/ml) or superoxide dismutase (SOD, 150 units/ml); the serine protease inhibitors aprotinin (100 g/ml) or soybean trypsin inhibitor (SBTI; 100 g/ml); HA (40 g/ml, vitreous humor, average molecular mass ϳ200 kDa, Seikagaku America, Falmouth MA); and the metalloprotease inhibitor GM6001 (25 M; Chemicon International, Temecula, CA). The use of these serine protease inhibitors allows for the distinction of TK from other serine proteases because TK is sensitive to aprotinin (K i 4.6 nM) but insensitive to SBTI up to 100 M (29). In additional experiments, the specific TK peptide inhibitor RPGLPVR-FESPLRINIIKE-NH 2 (30) was used to confirm that the ROS effect on pro-EGF processing was mediated by TK. This TK-specific peptide inhibitor was custom-synthesized by Biopeptide Co., LLC (San Diego, CA), and used at a concentration of 10 M.
To determine whether ROS effects were mediated by EGFR, functionally blocking anti-EGFR antibodies (1 g/ml, Calbiochem) were used. In experiments designed to define pro-EGF processing under base-line conditions, SMG cells were incubated for 18 h in the absence (control) or presence of aprotinin (100 g/ml), SBTI (100 g/ml ϭ 4.5 M), or GM6001 (25 M).
Because the inhibitory profile and the use of specific TK inhibitors in the above experiments suggested that TK is the main protease that processes pro-forms of EGFR ligands in these cells, experiments were designed to test the ability of TK to process pro-EGF and induce EGFR activation and signaling. SMG cells were exposed for 18 h to PBS (control) or recombinant TK (rTK, 1 M in PBS see below) in the absence or presence of aprotinin, SBTI, GM6001, or HA.

EGF Immunofluorescence
Paraffin-embedded sections of human trachea or cell cultures were labeled with a monoclonal anti-EGF antibody (5 g/ml; Calbiochem) by using overnight incubation at 4°C. After washing with PBS, horseradish peroxidase-conjugated secondary anti-mouse IgG antibody was added for 45 min. Enhanced fluorescence was achieved with the Tyramide Signal Amplification kit (Molecular Probes, Eugene, OR) according to the manufacturer's instructions.

Recombinant TK
rTK was produced in Pichia pastoris as described (31). Briefly, P. pastoris (strain GS115) was transfected with a plasmid encoding rTK (pPIC9pro-TK) generously donated by Dr. Hedy Chan (Axys Pharmaceuticals) using a Pichia Expression kit (Invitrogen) according to the manufacturer's instructions. The plasmid contains the full coding sequence of human salivary pro-TK behind the alcohol dehydrogenase promoter and a signal sequence to allow induction of synthesis and secretion into the media. This pro-form is unstable and is converted to active kallikrein by autocatalysis in the culture media. Cells were removed by centrifugation, and culture medium was dialyzed against 50 mM imidazole, pH 7.0. TK was purified using a benzamidine-Sepharose column (5 ml, Amersham Biosciences), equilibrated in 50 mM imidazole, pH 7.0. Recombinant TK was eluted with a guanidine gradient (50 -500 mM in 50 mM imidazole, pH 7.0) by using a Duo Flow high pressure liquid chromatography system (Bio-Rad). Purity of the protein was confirmed by PAGE and by specific activity. Purified fractions were dialyzed against sodium acetate, pH 7.0, and stored at Ϫ20°C. The system yielded about 30 mg of rTK per liter of culture media.

TK Enzyme Activity
TK enzyme activity was determined using DL-Val-Leu-Arg-p-nitroanilide as a substrate as described previously (21). In the presence of SBTI, this assay is highly specific for TK (29). Briefly, samples (100 l) were incubated in an ultra low binding 96-well plate with 10 l of trypsin (5 g/ml) for 15 min at 37°C to activate any pro-kallikrein. After adding 40 l of SBTI (1 mg/ml) and 100 l of DL-Val-Leu-Arg-pnitroanilide (2 mM; ICN, Irvine, CA) in 200 mM Tris, pH 8.2, absorbance was monitored at 412 nm using a microplate reader (Molecular Devices, Sunnyvale, CA). Enzyme concentrations were calculated by interpolating activities from the rTK standard curve.

Quantitative Analysis of EGFR Ligands
The apical media of cultured SMG cells were analyzed for the following soluble EGFR ligands: EGF, TGF-␣, HB-EGF, amphiregulin, and betacellulin. EGF and TGF-␣ were measured using commercially available ELISA kits according to the manufacturer's guidelines (Quantikine kit, R & D Systems, Minneapolis, MN; and Calbiochem, respectively).
ELISAs were also used to measure HB-EGF, amphiregulin, and betacellulin. Briefly, 50 l of sample (supernatant or standard) was incubated overnight with 50 l of bicarbonate/carbonate buffer, pH 9.2, at 4°C in a 96-well plate. Then plates were washed three times with 20 mM Tris, 500 mM NaCl, 0.05% Tween 20, pH 7.5 (TTBS), and blocked with 1% bovine serum albumin (BSA) in TTBS for 6 h at room temper-ature. After washing, plates were incubated with 100 l of mouse monoclonal anti-HB-EGF antibody (R & D Systems, 10 g/ml diluted into TTBS), goat polyclonal anti-amphiregulin antibody (R & D Systems, 5 g/ml), or mouse monoclonal anti-betacellulin antibody (4 g/ ml). After 2 h, the wells were washed, and 100 l of goat anti-mouse IgG (0.2 g/ml) or mouse anti-goat IgG (0.5 g/ml), both conjugated to alkaline phosphatase (Kirkegaard & Perry Laboratories, Gaithersburg, MA), were dispensed into each well and incubated for 2 h. Color was developed with p-nitrophenyl phosphate and stopped with 3 N NaOH. Absorbance was read at 410 nm. EGFR ligand concentrations were determined by interpolation from the standard curves.

Quantitative Analysis of HA
HA content of the supernatants was estimated using a biotinylated HA-binding protein (b-HABP) assay as described by Bray et al. (32). Briefly, the wells of a microtiter plate were coated overnight with 100 g/ml of HA isolated from bovine vitreous humor (Worthington). Supernatants or HA standards (100 l) were pre-incubated for 1 h at 37°C with 16 l of b-HABP (Seikagaku, 15 ng/ml) and 84 l of PBS containing 2% BSA and 0.2% Tween 20. The pre-incubated supernatants/ standards were added to the wells and incubated for 90 min at 37°C. After washing with PBS, the wells were incubated with streptavidin coupled to alkaline phosphatase (1 g/ml, Bio-Rad) for 2 h at room temperature. Color was developed with p-nitrophenyl phosphate and stopped with 3 N NaOH. Absorbance was read at 410 nm. Concentrations were determined by interpolation from a standard curve.

HA⅐TK Complex Disruption by X/XO
HA⅐TK complexes were generated in vitro by incubating HA (average molecular size 200 kDa, 40 g/ml) with rTK (10 g/ml) in 1 ml of PBS for 3 h at 37°C. Complex formation was confirmed by determining inhibition of TK activity (resulting in 50% of base-line activity under these conditions) and by electrophoresis using an overlay membrane method to visualize enzyme activity and molecular distribution as described (21). Exposure to X/XO (0.6 mM xanthine plus 0.05 units of xanthine oxidase) for 30 min was used to disrupt TK⅐HA complexes. Complex disruption was confirmed by TK activity measurements.

Visualization of HA Depolymerization
To confirm that X/XO induced HA depolymerization under the used experimental conditions, supernatants of SMG cultures treated with X/XO, X/XO plus catalase, or PBS as described above were collected, pooled, and digested with proteinase K (125 g/ml, 2 h at 60°C). Digested samples were precipitated with 85% ethanol, and pellets used for agarose electrophoresis and fluorophore-assisted carbohydrate electrophoresis (FACE).
Agarose Gel Electrophoresis-One set of pellets was resuspended in water and run in 0.7% agarose-Tris borate-EDTA gels as described (33). After electrophoresis, samples were transferred to a Biodyne B nylon membrane (Pall Gelman Laboratory, Ann Arbor MI) and probed with b-HABP (1 g/ml, Seikagaku), followed by streptavidin-alkaline phosphatase (Bio-Rad). Color was developed with bromochloroindolyl phosphate/nitro blue tetrazolium.
FACE-To determine whether X/XO treatment resulted in the release of HA oligosaccharides from cultured cells, another set of pellets was resuspended in 0.1 M ammonium acetate, pH 7.5, and incubated overnight at 37°C with 20 milliunits of chondroitinase ABC at pH 7.5 (were hyaluronidase activity is negligible), 5 milliunits of keratanase, and 5 milliunits of endo-␤-galactosidase. Hyaluronidase was omitted in order to detect ROS-induced appearance of small hyaluronan fragments. Digested samples were ethanol-precipitated, dried, derivatized with 2-aminoacridone (AMAC, Molecular Probes), and run in MONO TM (Prozyme, San Leandro, CA) composition gels with MONO TM (Prozyme) gel running buffer as described by Calabro et al. (34). After precipitation, glycosaminoglycan disaccharides (e.g. chondroitin sulfate or keratan sulfate) were expected to remain in the supernatant (34). Thus, only ethanol-precipitated HA fragments should be available for derivatization with AMAC. Gels were photographed under 302 nm UV light using a ChemiDoc XRS imaging system and Quantity One data analysis software (both from Bio-Rad).

Immunoblotting for Phosphorylated and Total EGFR
Duplicate aliquots of cells lysates containing equal amounts of protein were used for estimating total and phosphorylated EGFR (pEGFR) by Western blotting. For pEGFR, samples were immunoprecipitated using a monoclonal anti-EGFR antibody (Ab-3, 1 g/ml; Calbiochem) and protein A-agarose beads (Santa Cruz Biotechnology, Santa Cruz, CA). Incubation occurred overnight at 4°C. Immunoprecipitates were washed four times with cold PBS containing 0.05% Tween 20. Beads were resuspended in 20 l of Laemmli sample buffer (Bio-Rad), heated to 95°C for 5 min, and centrifuged. Supernatants from immunoprecipitation as well as cell lysates were electrophoresed on 4 -15% Tris-HCl Ready Gels (Bio-Rad) and then transferred electrophoretically to polyvinylidene difluoride membranes (Millipore, Billerica, MA). The membranes were incubated with 1% gelatin in TBS containing 0.05% Tween 20 for 1 h followed by mouse monoclonal anti-phosphotyrosine antibody (PY99, 2 g/ml; Santa Cruz Biotechnology) or anti-EGFR antibody. Secondary antibody was a mouse anti-goat IgG conjugated to alkaline phosphatase (0.5 g/ml, Kirkegaard & Perry Laboratories) that was visualized using bromochloroindolyl phosphate/nitro blue tetrazolium as a substrate. Developed blots were scanned, and the pEGFR to EGFR ratio was quantified using Quantity One analysis software (Bio-Rad).

Immunoblotting for p44/42 MAPK
Aliquots of cells lysates containing equal amounts of protein were suspended in Laemmli sample buffer and boiled for 5 min. Proteins were separated on 4 -15% Tris-HCl Ready gels and transferred to membranes as described above. Samples were examined for p44/42 MAPK (Erk1 and Erk2) using the PhosphoPlus p44/p42 MAP kinase (Thr-202/Thr-204) antibody kit according to the manufacturer's instructions (Cell Signaling Technology, Beverly, MA). Developed blots were scanned, and the p44/42 MAPK to p44/42 MAPK ratio was quantified using Quantity One analysis software (Bio-Rad).

Immunoassay for MUC 5AC Protein
Cell lysates (50 l) were incubated with bicarbonate-carbonate buffer, pH 9.2 (50 l), for 1 h at 37°C in a 96-well plate (Dynex, Chantilly, VA) or until dry. Plates were washed three times with TBS and blocked with 2% BSA for 1 h at room temperature. Plates were again washed three times with TBS and then incubated with 50 l of a monoclonal anti-MUC 5AC antibody (10 g/ml, Chemicon International) diluted into PBS containing 0.05% Tween 20. After 1 h, wells were washed with TBS, and 100 l of goat anti-mouse IgG conjugated to alkaline phosphatase (0.2 g/ml, Kirkegaard & Perry Laboratories) was dispensed into each well. After 1 additional hour, plates were washed with TBS. Color was developed with p-nitrophenyl phosphate and stopped with 3 N NaOH. Absorbance was read at 410 nm. Results were expressed as % changes in A 410 above the PBS control. This was necessary as pure MUC 5AC was not available to calibrate the assays.

Inhibition of Pro-EGF mRNA Expression
Double-stranded siRNA for the DNA target sequence of human pro-EGF (5Ј-AATCCTTATGAGGAGTCGAGC-3Ј) and control, non-silencing RNA (5Ј-AATTCTCCGAACGTCTCACGT-3Ј) were bought from Qiagen-Xeragon (Germantown, MD) and transfected into human SMG cells with TransMessenger TM Transfection Reagent (Qiagen) according to the manufacturer's instructions. Two days after transfection, the cells were exposed to X/XO in the absence or presence of catalase or PBS for 30 min. Cell lysates were analyzed for EGFR and p44/42 MAPK activation as described above.
To confirm that the siRNA silenced pro-EGF expression, experiments using siRNA and control RNA were repeated on SMG cells grown on coverslips. EGF protein expression was assessed by immunofluorescence as described above. Sample volume was insufficient for determination of EGF by ELISA.

Statistical Analysis
Data were expressed as mean Ϯ S.E. Differences between multiple groups were compared using a one-way analysis of variance followed by the Tuckey Kramer honestly significant difference test. Levene test was used to analyze the homogeneity of variances. Significance was accepted at p Ͻ 0.05.

X/XO Induces EGFR Activation and Signaling in Human
Airway SMG Cells-To test if primary cultures of human SMG respond to oxidative stress as described for cell lines of epithelial origin (35)(36)(37), cultures were treated with X/XO and assessed for activation of the EGFR signaling cascade. X/XO treatment of SMG cell cultures (n ϭ 12 cultures obtained from four different donors) induced EGFR activation as measured by EGFR phosphorylation and induced EGFR signaling as deter-mined by MAPK phosphorylation (Fig. 1, a-d). The ROS scavengers, catalase and SOD, reduced X/XO-induced EGFR and p44/42 MAPK phosphorylation (Fig. 1, a-d). Functionally blocking anti-EGFR antibodies reduced EGFR phosphorylation and to a lesser degree p44/42 MAPK phosphorylation, possibly suggesting an alternative pathway of activating p44/42 MAPK. Although MUC5 AC amounts inside SMG cells did not change over this 30-min incubation time, exposure to X/XO for 18 h induced MUC 5AC significantly compared with PBS control (16.4 Ϯ 3.1% above PBS; p Ͻ 0.05; not shown).
EGF Is the Ligand Responsible for X/XO-induced EGFR Activation-Because all EGFR ligands are synthesized as transmembrane precursors that need to be proteolytically cleaved to release active ligands, protease inhibitors were used to assess which proteases were involved in the activation of EGFR. The metalloprotease inhibitor GM6001 was used to inhibit the members of the ADAM family responsible for the activation of HB-EGF, TGF-␣, amphiregulin, and betacellulin, whereas the serine protease inhibitors aprotinin and SBTI were used to inhibit the activation of EGF. In addition, HA was used, as it is the endogenous inhibitor of TK, an important serine protease in the airway. As seen in Fig. 1, a-d, aprotinin and HA but not SBTI or GM6001 inhibited EGFR and MAPK phosphorylation. Because aprotinin but not SBTI reduced activation, this inhibitory profile was consistent with TK being the enzyme responsible for the processing of the ligand that activates EGFR. Also, it suggests that EGF is the produced ligand because other ligands are processed by metalloproteases of the ADAM family that are sensitive to GM6001.
To confirm these findings, TK and EGF were measured in SMG cell culture media after X/XO treatment as described under "Experimental Procedures" (n ϭ 12, from four different donors). Compared with PBS control, X/XO increased the concentration of active TK (4.3 Ϯ 0.2 g/mg protein versus 1.2 Ϯ 0.4 g/mg; p Ͻ 0.05) and of soluble EGF (4.4 Ϯ 0.4 pg/mg protein versus 2.2 Ϯ 0.4 g/mg; p Ͻ 0.05). Catalase and SOD as well as aprotinin and an excess of HA brought TK activity and soluble EGF concentration back to base-line levels (Fig. 2). In contrast, X/XO did not modify the levels of TGF-␣ (10.4 Ϯ 3.2 pg/mg cell lysate protein versus 11.4 Ϯ 3.9 pg/mg; p Ͼ 0.05) or HB-EGF (1.6 Ϯ 0.2 pg/mg cell lysate protein versus 1.8 Ϯ 0.1 pg/mg; p Ͼ 0.05) in SMG cell culture supernatants. Amphiregulin and betacellulin were undetectable by our assays (Ͻ50 pg/ml and Ͻ1 ng/ml, respectively) in the cell supernatants under all experimental conditions. To confirm that the X/XO effect on pro-EGF processing was due to increased endogenous TK activity, a specific peptide inhibitor of TK was used. This peptide inhibitor (PI) blocked the X/XO-induced EGF release significantly (2.7 Ϯ 0.1 pg/ml for PBS; 4.1 Ϯ 0.2 pg/ml for X/XO, and 2.9 Ϯ 0.3 pg/mg for PI, n ϭ 4 obtained from two different lung donors, p Ͻ 0.05) and inhibited X/XO-induced EGFR and MAPK activation (Fig. 3).

FIG. 1. ROS increases EGFR and MAPK activation in SMG cells cultures.
Human SMG cells were exposed to PBS (control), X/XO, or X/XO plus catalase (Cat), SOD, aprotinin (Apr), SBTI, GM6001, HA ,or functionally blocking anti-EGFR antibody (RAb). After 30 min, cell lysates were analyzed for EGFR phosphorylation (a and b) and MAPK activation (c and d).

X/XO Induces HA Depolymerization and TK Activation in
Human SMG Cells in Vitro-Because bronchial TK is secreted by serous cells of airway submucosal glands together with HA, and because HA binds to and inhibits the bronchial enzyme activity of TK (21), TK⅐HA complexes generated in vitro, as described under "Experimental Procedures," were exposed to X/XO to disrupt HA-TK association (by causing HA depolymerization), thereby releasing active TK. After 3 h of incubation with HA, TK showed a 50% inhibition of enzyme activity. X/XO exposure for 30 min restored TK activity to 100% (Fig. 4), whereas PBS did not (not shown). Additional experiments showed that X/XO alone did not affect purified rTK activity (16.8 Ϯ 1.2 to 17.1 Ϯ 2.7 milliunits/mg protein) or TK activity in cell supernatants previously treated with hyaluronidase (0.4 Ϯ 0.03 (supernatants) to 1.9 Ϯ 0.82 (ϩhyaluronidase) and 2.1 Ϯ 0.3 (ϩhyaluronidase followed by X/XO) milliunits/mg protein.
ROS generated by X/XO during a 30-min exposure of the apical side of SMG cultures increased HA concentrations in apical cell supernatants when compared with PBS control (1.87 Ϯ 0.31 g/mg cell lysate protein versus 0.27 Ϯ 0.05 g/mg; p Ͻ 0.05), consistent with HA breakdown and release from the cell surface (Fig. 5a). Concurrent treatment with catalase prevented HA release (0.78 Ϯ 0.05 g/mg; p Ͼ 0.05 compared with control and p Ͻ 0.05 compared with X/XO). Analysis of HA molecular weight distribution in these samples showed that X/XO decreased the average molecular size from ϳ1,800 to ϳ300 kDa (increased mobility and decrease in HA size heterogeneity, Fig. 5b) and caused HA fragments to appear as seen in FACE (Fig. 5c). Together, these results suggested that X/XO induced HA degradation and caused active TK release from HA⅐TK complexes generated in vitro and in vivo on human tracheal SMG cells in culture.
EGF Is Present in Airway Submucosal Glands of Human Tracheal Tissue Sections-To confirm EGF protein expression by SMG (38), immunofluorescence of human tracheal sections was performed as described under "Experimental Procedures." Immunostaining of tracheal sections showed that SMG cells synthesize proteins with an EGF domain both at their apical and basolateral membranes (Fig. 6).
An Enzyme with the Biochemical Characteristics of TK Processes Pro-EGF in Human SMG Cells under Resting Conditions-To define the protease responsible for pro-EGF activa-

FIG. 2. ROS increases TK activity and EGF release in SMG cell cultures.
Human SMG cells (n ϭ 12 cultures from four different lungs) were exposed to PBS (control), X/XO, or X/XO plus catalase (Cat), SOD, aprotinin (Apr), SBTI, GM6001, HA, or functionally blocking anti-EGFR antibody (RAb). After 30 min, supernatants were analyzed for TK activity (g/mg cell lysate protein) and EGF levels (pg/mg cell lysate protein). * indicates p Ͻ 0.01 compared with PBS (control); ** indicates p Ͻ 0.01 compared with X/XO; and *** indicates p Ͻ 0.01 compared with both PBS and X/XO.

FIG. 3. Specific TK peptide inhibitors block EGF release, EGFR signaling, and MAPK activation induced by X/XO.
Human SMG cells (n ϭ 8 from four different lung donors) were treated with PBS control, X/XO, or X/XO plus TK peptide inhibitors (PI). After 30 min, supernatants were analyzed for EGF content (pg/mg cell lysate protein, a), and cell lysates were analyzed for EGFR phosphorylation (b and c) and MAPK activation (d and e). * indicates p Ͻ 0.01 compared with PBS (control); # indicates p Ͻ 0.01 compared with X/XO. tion under resting conditions, i.e. without X/XO stimulation, EGF release into the culture media was measured in the absence or presence of serine and metalloprotease inhibitors. Cells (n ϭ 12 cultures from four different donors), incubated apically in PBS for 18 h and bathed basolaterally in medium containing no EGF, released EGF into the apical supernatant (10.5 Ϯ 1.2 pg/mg cell lysate protein; Fig. 7). Aprotinin reduced EGF release (4.0 Ϯ 0.9 pg/mg; p Ͻ 0.05 versus control), and SBTI or the metalloprotease inhibitor GM6001 had no effect (9.3 Ϯ 0.7 and 11.0 Ϯ 1.1, respectively; both p Ͼ 0.05 versus control; Fig. 7). This inhibitor profile (aprotinin sensitivity and SBTI insensitivity) is specific for a serine protease of the tissue kallikrein family. These findings were supported by the fact that active TK in the apical media correlated with EGF release (r 2 ϭ 0.91) as follows: 3.5 Ϯ 0.2 g/mg in control cells; 1.0 Ϯ 0.3 g/mg in aprotinin-treated cells (p Ͻ 0.05 versus control); 3.1 Ϯ 0.1 g/mg in SBTI and 4.3 Ϯ 0.6 g/mg in GM60001-treated cells (for both p Ͼ 0.05 versus control; Fig. 7).
None of the inhibitors modified low base-line release of HB-EGF by SMG cells (1.5 Ϯ 0.1 pg/mg cell lysate protein). SMG cells also released TGF-␣ (70.4 Ϯ 7.9 pg/mg cell lysate protein). The TGF-␣ release was significantly decreased by GM6001 (to 45.7 Ϯ 9.4 pg/mg, p Ͻ 0.05 versus control) but not by SBTI (61.4 Ϯ 10.2 pg/mg, p Ͼ 0.05 versus control). Surprisingly, aprotinin also decreased TGF-␣ release (to 25.4 Ϯ 6.5 pg/mg, p Ͻ 0.05 versus control) suggesting that, under unstimulated conditions, TK plays a role in pro-TGF-␣ processing. Amphiregulin and betacellulin remained undetectable in our assays under base-line conditions. The sensitivity of our assay was borderline in detecting betacellulin at the EC 50 levels (1.4 nM, see Ref. 39); however, a significant role of betacellulin in EGFR activation after stimulation with X/XO was excluded by the inefficiency of GM6001 to block EGFR activation and signaling. Because the affinity of EGFR for amphiregulin is lower (Ͼ1 M) than for EGF or TGF-␣ (1.2 nM) (11), it is unlikely that amphiregulin is involved in EGFR activation under these conditions. In addition to its lower affinity, amphiregulin cannot dimerize EGFR unless other accessory molecules (such as heparin) are present to stabilize its functional association with EGFR (11).
In summary, these data support the hypothesis that a TKlike enzyme is the endogenous activator of pro-EGF in human SMG cells under resting conditions, similar to the findings in the kidney as well as salivary and mammary glands (17)(18)(19).
Recombinant TK Cleaves Pro-EGF, Releases EGF, and Activates EGFR in Human SMG Cells-To confirm that TK cleaves pro-EGF from SMG and to test that the release of mature EGF activates EGFR signaling, SMG cells were incubated for 18 h with PBS (control) or rTK in the absence or presence of protease inhibitors (aprotinin, SBTI, or GM6001), HA, or functionally blocking anti-EGFR antibodies (R-Ab) (each n ϭ 12, cells from four different donors). Whereas EGF was quantified in apical culture supernatants, cell lysates were probed for tyrosine phosphorylation of EGFR and p44/42 MAPK. Cell lysates were also used for MUC 5AC quantification. rTK increased EGF release from SMG cells compared with base line (18.5 Ϯ 1.0 pg/mg cell lysate protein versus 10.5 Ϯ 1.2 pg/mg; p Ͻ 0.05; Fig. 8). EGF release induced by rTK was blocked by aprotinin (3.1 Ϯ 0.9 pg/mg; p Ͻ 0.05 compared with rTK and PBS) and HA (7.0 Ϯ 1.3 pg/mg; p Ͻ 0.05 compared with rTK and PBS), whereas SBTI and GM6001 had no effect (16.6 Ϯ 2.6 and 19.7 Ϯ 2.7 pg/mg, respectively; both p Ͼ 0.05 compared with rTK). In addition, incubation with functionally blocking anti-EGFR antibodies increased EGF levels in the incubation media (31.7 Ϯ 4.5 pg/mg), probably due to inhibition of EGF binding to its receptor by the blocking antibody.
In parallel to the EGF increase in culture media, EGFR as well as p44/42 MAPKs were phosphorylated upon exposure of SMG cells to rTK. Aprotinin as well HA reduced EGFR phosphorylation, whereas SBTI and GM6001 did not (Fig. 9). A similar profile of stimulation and inhibition was evident in downstream signaling as evidenced by p44/42 MAPK phosphorylation (Fig. 9) and MUC 5AC protein expression (Fig. 10). Apical compartments of human SMG cells cultures (n ϭ 6 from three different lung donors) were exposed for 30 min to PBS or X/XO alone or with catalase. Soluble HA was measured and normalized to the simultaneously recovered cell lysate protein (a). * indicates p Ͻ 0.01 compared with PBS (control); ** indicates p Ͻ 0.01 compared with X/XO. Agarose electrophoresis followed by transblotting using a b-HABP and FACE analysis (n ϭ 4 from two different lung donors) shows that X/XO exposure induced a decrease in HA average molecular weight (b) and appearance of HA fragments (c) in culture media. y axis in b shows standards in kDa, as estimated using DNA standards, and in c are AMAC-derivatized disaccharides. MUC 5AC protein expression in cell lysates increased by 20.3 Ϯ 3.0% above base line in response to rTK (p Ͻ 0.05; Fig.  10). The rTK-induced increase in intracellular MUC 5AC was inhibited by aprotinin, HA, and functionally blocking anti-EGFR antibodies (decreased by 4.0 Ϯ 4.0%, 1.7 Ϯ 6.5%, and 6.2 Ϯ 3.9% below base line, respectively; all p Ͻ 0.05 compared with rTK stimulation and all p Ͼ 0.05 compared with base line) but not by SBTI or GM6001 (22.0 Ϯ 4.9% and 11.2 Ϯ 3.2 above base line, respectively; all p Ͻ 0.05 compared with base line). These results confirm that TK induces EGF release from SMG cells and that this release results in EGFR activation and downstream signaling.

Inhibition of Pro-EGF mRNA Expression Is Associated with a Decrease of EGFR and MAPK Phosphorylation in Response to
Oxidative Stress-To confirm that activation of EGFR in response to ROS was due to endogenous processing of pro-EGF by TK, we inhibited pro-EGF mRNA translation using siRNA. RNA with 16/21-base overlap with a Thermotoga maritime gene and no known homology to any other mRNA in human sequence data bases was used as a non-silencing RNA control. As shown in Fig. 11, transfection of human SMG cells with pro-EGF siRNA blocked EGF protein expression in these cells as measured by immunofluorescence. In contrast to the nonsilencing RNA transfection and PBS control, X/XO treatment did not result in EGFR phosphorylation and MAPK activation (Fig. 12) in cells transfected with siRNA. These data confirm that EGF and not other EGFR ligands are responsible for EGFR activation and signaling in response to oxidative stress in human SMG, and support the hypothesis that ROS-induced HA depolymerization and TK activation are responsible for initiation of EGFR signaling in these cells. DISCUSSION The data presented here support the hypothesis that oxidative stress-induced EGFR activation of human airway submucosal gland cells is due, at least in part, to a cascade of events initiated by depolymerization of HA, activation of TK, and release of mature EGF. The data show the importance of intact HA and its inhibition of TK in the airway lumen. They also show that TK processing of pro-EGF plays a role in regulating the EGF signaling pathway in airway submucosal glands and provides a possible direct mechanistic link between sustained airway ROS production during inflammation and the development of gland hyperplasia in the airways. HA/TK interaction, previously demonstrated in sheep and human airways (21,22), is shown here to be functionally relevant in regulating EGF release from the cell membrane with subsequent EGFR activation in these cells. The integrity of HA is sensitive to cleavage by ROS (23), and the tracheobronchial tree is exposed to increased levels of ROS either by inhalation of exogenous sources (e.g. tobacco smoke) or by stimulation of endogenous production by airway epithelial and inflammatory cells (40). ROS generated by X/XO increased active TK availability by HA degradation, thereby leading to EGF release and EGFR signaling. The fact that EGF up-regulates HA synthase 2 expression (41) suggests that negative feedback might exist to control pro-EGF activation (by controlling TK activity with increased HA), thereby inhibiting uncontrolled cell growth in airway submucosal glands leading to gland cell hypertrophy and hyperplasia. Chronic HA depolymerization by oxidants might disrupt this control mechanism.
EGF receptor tyrosine kinase activation (c-ErbB 1) can be achieved by multiple EGFR ligands as follows: EGF, TGF-␣, HB-EGF, amphiregulin, epiregulin, and betacellulin, although affinities and specificities for the different Erb receptors vary between these ligands (14,38,39,42). All of these ligands are synthesized as transmembrane precursors and are cleaved to FIG. 7. Human SMG cells release TK and EGF constitutively. Human SMG cells (n ϭ 12 cultures from four different donors) were incubated in RPMI media in the basolateral compartment and with PBS alone or PBS containing either aprotinin, SBTI, or GM6001 in the apical compartment. Supernatants were analyzed for TK activity (top; g/mg cell lysate protein) and EGF levels (bottom; pg/mg cell lysate protein). TK activity levels correlated with EGF release (r 2 ϭ 0.91). TK was inhibited by aprotinin but not SBTI or GM6001. * indicates p Ͻ 0.05 compared with PBS control.

FIG. 8. Recombinant TK increases EGF release from SMG cells.
Human SMG cultures were exposed for 18 h to PBS or rTK, rTK in the presence of aprotinin (Aprot), SBTI, GM6001, HA, or functionally blocking anti-EGFR antibody (RAb) (n ϭ 12, four lungs). EGF was measured and expressed as pg/mg cell lysate protein. * indicates p Ͻ 0.01 compared with PBS (control); ** indicates p Ͻ 0.01 compared with rTK; and *** indicates p Ͻ 0.01 compared with both PBS and rTK.

FIG. 9. Treatment with recombinant TK induced EGFR and MAPK activation in SMG cells.
Human SMG cultures were exposed to PBS or rTK, in the absence or presence of aprotinin (Apr), SBTI, GM6001, or HA. rTK increased the pEGFR-EGFR ratio (a and b) and MAPK phosphorylation (c and d).

FIG. 10. MUC5 AC content is induced by rTK and specifically blocked by TK inhibitors.
Human SMG cells were exposed to rTK for 18 h in the presence or absence of aprotinin, HA, SBTI, GM6001, or functionally blocking anti-EGFR monoclonal antibodies (RAb). MUC 5AC levels were expressed relative to PBS control. rTK increased MUC 5AC levels, and this effect could be blocked with aprotinin, HA, and functionally blocking anti-EGFR monoclonal antibodies (RAb), whereas SBTI or GM6001 had no effect. * indicates p Ͻ 0.05 compared with PBS; ** indicates p Ͻ 0.05 compared with rTK. their active form by proteases. TK is responsible for release of mature EGF in mammary and salivary glands as well as the kidney (17)(18)(19) and, as shown here, in airway submucosal gland cells. Cleavage of all other EGFR ligands involves matrix metalloproteases of the ADAM family (16). ADAM17/TACE cleaves pro-TGF-␣, amphiregulin, and HB-EGF (16), whereas ADAM9, ADAM10, and ADAM12 have been implicated in HB-EGF shedding (37,43,44). Oxidative stress is likely to increase the availability of a specific subset of these EGFR ligands depending on the anatomical location or cell type investigated. For instance, in rat gastric epithelial cells, oxidative stress has been shown to increase gene expression of the ligands HB-EGF and amphiregulin (10); on the other hand, only amphiregulin but not HB-EGF has been reported to be released in response to cigarette smoke in a human pulmonary muco-epidermoid carcinoma cell line (NCI-H292) via oxidative stress (5,47). Our study suggests that EGFR activation in the airway, at least in experiments using primary cultures of human submucosal gland cells, is mediated by EGF, which is cleaved from its precursor by TK, an enzyme produced and secreted locally. In this system, we did not find measurable amounts of amphiregulin and betacellulin either at base line or upon ROS exposure, and the low levels of HB-EGF were not modified by ROS. The high concentrations of TGF-␣ under resting conditions (after 18 h of incubation) and the decreased levels after aprotinin treatment are very intriguing. It has been shown that EGFR activation by EGF can increase pro-TGF-␣ synthesis and cleavage by an unknown mechanism (45), possibly partially explaining these findings.
EGFR signaling can be specifically regulated at a cellular level either by the expression of a subset of specific ligand precursors or by the expression of a subset of proteases, specific for the cleavage of ligand precursors. These mechanisms of specificity make it possible that different stimuli induce cleavage of different ligand precursors from the same cells. In NCI-H292 cells for instance, lipoteichoic acid from Gram-positive bacteria activates ADAM10 to process HB-EGF precursors re-sulting in EGFR activation and increased mucin synthesis (37), whereas tobacco smoke extracts activate ADAM17 to cleave amphiregulin precursors, again resulting in EGFR signaling (5,47). Thus, EGFR activation pathways cannot be predicted from immunohistochemical analysis alone nor can they be extrapolated from experiments done in other cell types. Here we show that ROS generated by X/XO stimulates EGFR signaling in SMG cells via TK-mediated cleavage of EGF precursors. Although EGF was functionally important in SMG cells, TGF-␣ and HB-EGF were also expressed in vitro, consistent with a previous report (38) using immunohistochemistry. Despite their expression, these EGFR ligands did not appear to play a role in ROS-induced EGFR activation and signaling in SMG cells because the metalloprotease inhibitor GM6001 (effective in blocking the activity of all members of the ADAM family) did not modify X/XO-induced EGFR activation and because blocking pro-EGF mRNA expression with siRNA resulted in inhibition of EGFR and MAPK phosphorylation. In our experiments, however, we also found that functionally blocking EGFR antibodies did not completely inhibit MAPK phosphorylation, suggesting that an additional EGFR-independent signaling pathway in response to ROS may also be present in our cells.
We used MUC 5AC synthesis as the final end point of assessing the X/XO/TK-mediated EGF activation of EGFR signaling. Although MUC 5AC is not the major mucin produced in SMG cells, it is produced by the mucous cells in SMG (46,48). We chose to study this mucin to compare our results using primary cells with the results reported by others using cell lines, and we showed that MUC 5AC content was also increased in SMG cells upon EGFR and MAPK activation. As expected for gene induction, the increase in MUC 5AC synthesis was observed after 18 h of rTK treatment of SMG cells and was not seen after only 30 min of stimulation. These data confirm other findings that EGFR signaling leads to MUC 5AC induction, although the stimulus used in our experiments was TK-induced EGF release as opposed to TGF-␣ (4) or ligandindependent EGFR activation induced by oxidative stress (7). FIG. 11. Transfection with of pro-EGF siRNA inhibits EGF protein expression. Micrographs of EGF immunostaining of SMG cells in culture grown and labeled as described under "Experimental Procedures" are shown. Transfection with pro-EGF siRNA resulted in inhibition of EGF protein expression (d) compared with non-silencing RNA-transfected cells (c) that showed a pattern of EGF protein expression comparable with non-transfected cells (b). a is a mouse serum control. Bar, 50 m.
In conclusion, our results suggest that TK plays an important role in the airways by processing pro-EGF to mature EGF in airway submucosal gland cells. The activity of TK is regulated by HA; TK is inhibited in the presence of native HA in the airway and in the submucosal gland duct. However, upon depolymerization of HA by ROS, active TK is released to cleave EGF precursors expressed at the cell surface. The released mature EGF interacts with EGFR to initiate signaling, including MAPK kinase phosphorylation and increases of gene and protein expression of MUC 5AC. The novel recognition of the central role of HA in this signaling pathway could potentially be used for new therapeutic approaches in airway disease associated with hypersecretory states.