Biosynthesis of α1-Proteinase Inhibitor by Human Lung-derived Epithelial Cells

Destruction of components of the extracellular matrix of the lung by neutrophil elastase is believed to be a critical event in the development of obstructive lung disease. The local synthesis of α1-proteinase inhibitor, the controlling inhibitor of this enzyme, may provide a partial mechanism for neutrophil elastase regulation, especially during inflammation, when proteolytic enzymes are released from phagocytes. In this study, we show that lung-derived epithelial cells not only have the capacity to synthesize functional α1-PI but also to increase the rate of its production when stimulated by specific inflammatory mediators, including oncostatin M, interleukin-1, and the glucocorticoid analogue, dexamethasone.

the extrahepatic production of specific serpins is likely to be more significant than had previously been believed. For example, Achy, which increases dramatically in concentration in plasma during the acute phase reaction can be produced by epithelial cells originating from jejunum (4), lung, breast, or skin (5). Indeed, model lung-derived epithelial HTB55 cells synthesize comparable amounts of Achy to those from hepaticderived HepG2 cells. Furthermore, such synthesis can be upregulated by a number of factors, including oncostatin M (OSM), interleukin-1 (IL-1), and glucocorticoid analog, dexamethasone (5).
The human airway epithelium, with a surface of 1-2 m 2 , has the potential of exposure to a variety of factors that may trigger an inflammatory response (6). However, the contribution of lung epithelial cells to the local production of ␣ 1 -PI has never been clarified. In the current study, a variety of inflammatory mediators were tested as potential stimulators of ␣ 1 -PI in a lung-derived HTB55 cell line, as well as in normal bronchial epithelial cells.
Cell Culture-HTB55 (Calu-3) human lung adenocarcinoma and HepG2 human hepatoma cell lines were obtained from the American Type Culture Collection (Rockville, MD). Normal human bronchial epithelial cells were purchased from Clonetics (San Diego, CA). HTB55 cells were cultured in Eagle's MEM supplemented with 0.1 mM nonessential amino acids and 1 mM sodium pyruvate (all from Life Technologies, Inc., whereas HepG2 cells were cultured in Dulbecco's modified Eagle's medium. Both media contained, in addition, 100 units/ml penicillin G, 100 g/ml streptomycin (Life Technologies, Inc.), and 10% fetal bovine serum (Atlanta Biologicals, Norcross, GA). Bronchial cells were cultured in serum-free bronchial epithelial cell basal medium containing 0.5 ng/ml human epidermal growth factor, 5 g/ml insulin, 0.5 g/ml each hydrocortisone and epinephrine, 10 g/ml transferrin, 0.5 ng/ml triiodothyronine, and 0.4% v/v bovine pituitary extract (all from Clonetics, San Diego, CA). Cells were plated, allowed to grow to confluency before assay, and then treated with various stimulating factors.
Northern Blot Analysis-Total RNA was isolated as described previously (7,8). Northern blot analysis was carried out by electrophoresis of RNA samples in 1% agarose gels containing 2.2 M formaldehyde, followed by capillary transfer (9) to Hybond-N membranes (Amersham Corp.). Hybridization with 32 P-labeled probes was performed overnight at 65°C in a mixture containing 1 M NaCl, 1% SDS, and 10% dextran sulfate. The following probes were used; 1.4-kilobase EcoRI-EcoRI re-* This work was supported by Grants HL 26148, HL 37090 from the National Institutes of Health and KBN (Poland) Grant 6 PO4C 016 11. 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.
Determination of Protein Secretion-Culture medium was collected either at 24 or 48 h after factor(s) addition. In the latter case, however, the culture medium was replaced at 24 h by fresh medium containing stimulating factors. Thus, in all experiments, the actual, quantitated stimulation of ␣ 1 -PI production was measured during a 24-h incubation period. The amount of secreted ␣ 1 -PI was determined by rocket immunoelectrophoresis using polyclonal rabbit antibodies against human ␣ 1 -PI (Dako, Carpinteria, CA). Purified human plasma ␣ 1 -PI (kindly provided by Miles, Inc., Berkeley, CA) was used as a standard.
Determination of Sugar Heterogeneity-The microheterogeneity of both ␣ 1 -PI and ␣ 1 -acid glycoprotein (AGP) was analyzed by crossed affinoimmunoelectrophoresis (CAIE) with ConA, using 1% agarose in Tris-barbital buffer (11). At first, proteins were resolved by electrophoresis in agarose gel (3 h at 10 V/cm) incorporated with ConA (1. 5 mg/ml). A second dimension electrophoresis (overnight, 2 V/cm), vertical to the first separation, was carried out in a gel containing either specific anti-␣ 1 -PI or anti-AGP antibodies and 60 mg/ml methyl-␣-Dglucopyranoside. The concentration of antibodies and the amounts of sample were adjusted in different experiments to produce precipitation arcs of comparable intensity and height. Gels were stained with Coomassie Blue to demonstrate precipitin lines. The area covered by ␣ 1 -PI glycoforms was determined by planimetry and the amount of ConA-nonreactive fraction expressed as a percentage of total ␣ 1 -PI.
Separation of Isoforms of ␣ 1 -PI-The separation of glycoforms of ␣ 1 -PI was achieved by affinity chromatography using ConA-4B Sepharose column. After application of aliquots of serum, LLF, or HTB55 cell medium, the ConA-nonbinding fraction was eluted with Tris-HCl buffer, pH 7.6, containing 0.15 M NaCl and 1 mM each of MnCl 2 , MgCl 2 , and CaCl 2 . ConA-bound material was desorbed with 0.2 M ␣-methyl-Dmannoside in the same buffer. Fractions were dialyzed against 25 mM ammonium bicarbonate, lyophilized, and dissolved in 1:10 of the original volume. The total amount of recovered ␣ 1 -PI in each fraction was measured by rocket immunoelectrophoresis using a specific antiserum against ␣ 1 -PI, whereas the composition of glycoforms was determined by affinoimmunoelectrophoresis with ConA.
Enzymatic Treatment and Western Blot Analysis-Aliquots of serum, LLF, the ConA-nonbinding fraction of LLF, and HTB55 cell medium, all containing defined amounts of ␣ 1 -PI, were incubated with equimolar amounts of HNE (a generous gift of Athens Research and Technology, Athens, GA) for 30 min at room temperature in 20 mM Tris-HCl, pH 7.5, 0.15 M NaCl. Untreated or HNE-treated samples were then subjected to 9% SDS-PAGE (12). ␣ 1 -PI was visualized by enhanced chemiluminescence (ECL; Amersham) after electrotransfer to a nitrocellulose membrane and incubation with antibodies. 1:2000 dilution of rabbit anti-␣ 1 -PI antibodies (Dako) and 1:2000 dilution of donkey anti-rabbit IgG antibodies conjugated to horseradish peroxidase (Amersham) were used for Western blot analysis.
Biosynthetic Labeling, Immunoprecipitation, and Fluorography-Confluent monolayers of HTB55 or normal bronchial cells were stimulated for 18 h with 50 ng/ml OSM. The cells were then rinsed and incubated for 4 h in the presence of methionine-free medium containing OSM and 200 Ci/ml [ 35 S]methionine/cysteine (Tran 35 S-label) (ICN Biomedicals, Inc., Costa Mesa, CA). Aliquots of medium were pretreated with preimmune serum and Pansorbin (Calbiochem-Novabiochem, La Jolla, CA), as described previously (9). The supernatants were then incubated overnight at 4°C in 20 mM Tris-HCl, pH 7.5, 140 mM NaCl, 1% Triton X-100, with excess anti-␣ 1 -PI antibody. Immune complexes were precipitated with protein A-agarose (Santa Cruz Biotechnology, Inc., Santa Cruz, CA), washed, released by boiling in Laemmli sample buffer, and examined on SDS-PAGE. Bands were detected by fluorography as described elsewhere (13).

Regulation of ␣ 1 -PI Synthesis in HTB55
Cells-Both inflammatory cytokines and the glucocorticoid analog dexamethasone were tested for their ability to regulate the synthesis of ␣ 1 -PI in lung adenocarcinoma HTB55 cells. As demonstrated in Fig. 1, A and B, expression of ␣ 1 -PI was stimulated with OSM and to a lesser extent by dexamethasone. Although the up-regulation of ␣ 1 -PI synthesis by OSM was also noted at the protein level, the effect of dexamethasone on ␣ 1 -PI protein synthesis was variable as demonstrated by a high S.D. value (Table I). As might be expected, these two factors acted additively for the FIG. 1. Stimulation of ␣ 1 -PI synthesis in HTB55 cells. Cells were incubated in serum-free MEM supplemented with 50 ng/ml IL-6, 100 units/ml IL-1, 10 units/ml LIF, 50 ng/ml OSM, and 10 Ϫ6 M dexamethasone. At 24 h, total cellular RNA was isolated and subjected to Northern blot analysis. The blots were hybridized to an ␣ 1 -PI probe followed by rehybridization to the GAPDH probe (A). Audoradiographs were then scanned by densitometry, and ␣ 1 -PI mRNA was normalized to amounts of GAPDH (B). Samples of cell medium collected at 24 h were subjected to rocket immunoelectrophoresis using antiserum to ␣ 1 -PI. Gel was stained with Coomassie Blue to demonstrate precipitin lines (C). up-regulation of ␣ 1 -PI expression, which was also observed at the protein level ( Fig. 1; Table I). However, no stimulation of ␣ 1 -PI synthesis was obtained when IL-1 was given alone, although this cytokine appeared to enhance the effect of either OSM or dexamethasone. In fact, the most significant up-regulation of ␣ 1 -PI synthesis was noted when OSM, dexametha-sone, and IL-1 were given in combination. No significant stimulation of ␣ 1 -PI expression was observed following either IL-6 or LIF treatment (Fig. 1), although both cytokines were capable of stimulating ␣ 1 -PI production in HepG2 cells (Table I).

Stimulation of ␣ 1 -PI Expression in Normal Bronchial
Epithelial Cells-Northern blot analysis of RNA extracted from normal bronchial epithelial cells indicated that ␣ 1 -PI mRNA levels were substantially increased by OSM or a combination of OSM and IL-1, and somewhat by IL-1 alone, whereas IL-6 had no effect (Fig. 2A). The stimulatory effect of OSM and OSM in combination with IL-1 on ␣ 1 -PI message expression correlated with corresponding changes in secretion (Fig. 2C). In the case of OSM, the up-regulation of ␣ 1 -PI synthesis was observed as early as 2 h after addition of this cytokine, with maximum stimulation occurring after 8 h, and no further significant increase in ␣ 1 -PI production after prolonged time of cell exposure to OSM (Fig. 3). The increase in synthesis of ␣ 1 -PI by OSM-exposed cells was also concentration-dependent and occurred only at cytokine concentrations above 10 ng/ml, reaching a maximum at about 50 ng/ml and remaining approximately at the same level at higher OSM concentrations (Fig. 3A).
Analysis of ␣ 1 -PI Secreted by Hepatic and Epithelial Cells-CAIE with ConA revealed that three microheterogeneous forms of ␣ 1 -PI were secreted by liver-derived HepG2 cells, whereas only one was secreted by lung-derived HTB55 cells (Fig. 4). ConA binds with bi-but not with tri-or tetra-antennary carbohydrate structures, and in the case of multiheteroglycan proteins, the degree of reactivity with this lectin depends on the number of bi-antennary structures present on the molecule (14). The three variants of ␣ 1 -PI present in HepG2 medium can thus be described as either nonreactive (the electrophoretic migration of this form in a gel containing ConA is not affected), weakly reactive (retarded electrophoretic mobility in ConA-agarose gel), and strongly reactive with ConA (immobilized in ConA-agarose gel electrophoresis), respectively. The ␣ 1 -PI secreted by HTB55 cells was only of the ConA-nonbinding type, as confirmed by affinity chromatography using ConA-4B Sepharose, where all of the applied ␣ 1 -PI passed directly through the column (0.6 ϫ 7 cm) without being bound or even retarded.
Analysis of ␣ 1 -PI in Serum and LLF-␣ 1 -PI appears to enter FIG. 2. Effect of inflammatory cytokines on ␣ 1 -PI expression in normal bronchial epithelial cells. Cells were incubated in bronchial epithelial cell basal medium supplemented with 50 ng/ml IL-6, 100 units/ml IL-1, and/or 50 ng/ml OSM. In A, total cellular RNA isolated at 24 h was subjected to Northern blot analysis. The blot was then hybridized to an ␣ 1 -PI probe and to demonstrate equal loading rehybridized to the GAPDH probe. In B, after 18-h treatment with the indicated factors, cells were subjected to radiolabeling for 4 h with [ 35 S]methionine/ cysteine. Aliquots of cell medium were then immunoprecipitated with human anti-␣ 1 -PI, and immunoprecipitates were subjected to SDS-PAGE followed by fluorography. C, densitometer scans of data presented in A (␣ 1 -PI mRNA normalized using GAPDH as an internal control) and B (secreted ␣ 1 -PI).

FIG. 3. Dose-and time-dependent stimulation of ␣ 1 -PI expression in normal bronchial epithelial cells treated with OSM.
Cells were incubated in bronchial epithelial cell basal medium containing indicated amounts of OSM for 24 h (A) or were stimulated with 50 ng/ml OSM for the indicated times (B). Total cellular RNA was then isolated and subjected to Northern blot analysis. Blots were hybridized to an ␣ 1 -PI probe followed by rehybridization to the GAPDH probe. Right panels (A and B) demonstrate ␣ 1 -PI mRNA densitometer scans normalized using GAPDH mRNA as an internal control. the lung by diffusion from blood plasma (2). However, CAIE analysis of ␣ 1 -PI from serum and LLF of the same healthy donor demonstrated a higher proportion of the ConA-nonbinding ␣ 1 -PI glycoform in LLF, which may reflect its local synthesis (Fig. 5). Using planimetry, it was found that the amount of the ConA-nonbinding ␣ 1 -PI fraction, expressed as percentage of total ␣ 1 -PI, was about 12% higher in LLF (17.2 Ϯ 1.5 in LLF versus 4.8 Ϯ 0.9 in serum, mean Ϯ S.D. of three separate determinations), and the difference was independent of the concentration of ConA and methyl-␣-D-glucopyranoside, parameters known to affect the separation on CAIE (11). ConA CAIE analysis of serum mixed with HTB55 cell medium resembled that of LLF with respect to ␣ 1 -PI (Fig. 5). A decrease in ConA reactivity appeared to be specific for ␣ 1 -PI, because there was no difference in the ConA reactivity of AGP derived from the same serum and LLF (Fig. 5). It is important to note that, in contrast to ␣ 1 -PI, AGP was not synthesized by lung-derived HTB55 cells (data not shown). To exclude a possibility that differences in ConA reactivity observed between ␣ 1 -PI from serum and LLF were due to inactivation of this inhibitor, ␣ 1 -PI from LLF was separated on a ConA-Sepharose 4B column, and the nonbinding fraction of the inhibitor was tested for the ability to form a complex with HNE. As demonstrated in Fig. 6, this fraction of ␣ 1 -PI did form such a complex.
Inhibition of HNE by ␣ 1 -PI Derived from Lung Epithelial Cells-To examine if lung epithelial cell-derived ␣ 1 -PI exhibits inhibitory activity comparable to serum ␣ 1 -PI, the formation of an SDS-stable complex with HNE was studied in detail. For this experiment, the condition medium from OSM-stimulated HTB55 cells was subjected to rocket immunoelectrophoresis to determine the concentration of ␣ 1 -PI, and the medium was then incubated with various amounts of HNE. At 0.2 HNE: ␣ 1 -PI molar ratio, a shifted (␣ 1 -PI-HNE complex) band appeared (Fig. 7A). At higher concentrations of HNE, the intensity of the shifted band increased, and native ␣ 1 -PI decreased, with all detectable ␣ 1 -PI being in complex with HNE when the concentration of this enzyme was equimolar to ␣ 1 -PI (Fig. 7A). This result was similar to that observed when human serum ␣ 1 -PI was reacted with HNE ( Figs. 6 and 7).
To determine if ␣ 1 -PI secreted by normal bronchial epithelial cells was functionally active, we compared its ability to form a complex with HNE to that exhibited by HTB55 cell-derived ␣ 1 -PI. As illustrated in Fig. 7B, at a molar excess of HNE over ␣ 1 -PI, both HTB55-and normal bronchial cell-derived ␣ 1 -PI was detected in complex with HNE, concomitant with the disappearance of the band corresponding to native inhibitor. The additional band visible in case of bronchial cells most probably represents nonspecific binding because it can be removed by an additional preclearing step using preimmune serum (data not shown).

DISCUSSION
At sites of inflammation or tissue injury, proteolytic enzymes, released from phagocytes, may directly contribute to the host inflammatory response as well as to tissue destruction. The implication of these proteinases, including HNE, in the pathogenesis of lung emphysema as well as the finding that extrahepatic cells produce ␣ 1 -PI suggest that locally synthesized ␣ 1 -PI is likely to have an effect on the antielastase screen of lung tissue.
In this study, we demonstrated that cells originating from the respiratory tract epithelium are capable of ␣ 1 -PI synthesis and that several mediators of inflammation increase the expression of this inhibitor in these cells. However, regulation of ␣ 1 -PI in HTB55 and normal bronchial epithelial cells differs significantly from that in hepatic cells. IL-6-type cytokines (IL-6, OSM, LIF, IL-11, and ciliary neurotrophic factor) upregulate the production of ␣ 1 -PI in hepatocytes (Ref. 15; Table  I), whereas OSM and to a lesser extent IL-1, but not IL-6 or LIF, exert a significant increase in ␣ 1 -PI levels in lung-derived epithelial cells (Figs. 1 and 2). This strong effect of OSM and the lack of effect of IL-6 and LIF on production of another inhibitor of serine proteinases, Achy, was also observed in epithelial cells originating from the lung (16). However, IL-1 appears to be a much more potent stimulator of Achy than ␣ 1 -PI synthesis in these cells. It is important to note that this difference in the effect of OSM and IL-1, on Achy expression, compared with that of IL-6, was also observed in breast-derived epithelial cells (5).
Many of the overlapping biological responses of IL-6-related cytokines, including stimulation of ␣ 1 -PI production in hepatoma cells, have been explained by the presence of a common subunit, gp130, in cytokine receptors, or in the case of OSM and LIF, by the finding that both cytokines bind with high affinity to the same receptor, the latter consisting of two subunits, the Aliquots of both serum and LLF from the same healthy donor, as well as serum mixed with HTB55 culture medium collected 24 h after OSM (50 ng/ml) and dexamethasone (10 Ϫ6 M) addition (Serum HTB55), were subjected to CAIE with ConA. Specific antisera against ␣ 1 -PI or AGP were used in the second-dimension gel.
LIF receptor ␣ chain and gp130 (17). However, the regulation of ␣ 1 -PI expression in lung-derived epithelial cells, relative to hepatocytes, illustrates significant functional differences among IL-6-type cytokines. The exclusive effect of OSM on induction of several biological responses has been described previously (18). However, the mechanism of this process is not yet fully understood, and the stimulation of different patterns of tyrosine-phosphorylated proteins by OSM and LIF, as well as the finding that some cells bind OSM but not LIF, confirms the existence of an OSM-specific receptor (19). It is important to note that in lung-derived epithelial cells OSM, but not LIF or IL-6, exerted an effect on synthesis of another inhibitor of serine proteinases, Achy (16). Analysis of the mechanism(s) by which OSM wields its stimulatory action may help to explain the sensitivity of lung-derived epithelial cells to this cytokine. Apart from the differential regulation of ␣ 1 -PI synthesis in hepatic and epithelial cells, a significant difference in the glycosylation pattern of the inhibitor was also found. The latter effect was clearly demonstrated by ConA reactivity, which indicated three hepatic and a single ConA-nonbinding epithelial glycoform (glycoforms) of ␣ 1 -PI (Fig. 4). This observation may suggest a different clearance mechanism for lung-derived epithelial ␣ 1 -PI, because the half-life of the liver form of this inhibitor is, at least in part, dictated by the number and type of carbohydrate side chains (2). Furthermore, hepatocytes respond to inflammatory stimuli by changing the synthesis and type of N-glycosylation of many proteins, including ␣ 1 -PI (20,21). It has been shown previously that IL-6, LIF, and interferon-␥ decreased the ConA reactivity of HepG2-derived ␣ 1 -PI (21). A similar effect was observed when HepG2 cells were stimulated with OSM (data not shown). However, although this cytokine was the most potent in up-regulation of ␣ 1 -PI in lung-derived epithelial cells, it did not change the ConA reactivity of the inhibitor (data not shown). A significant difference in the binding to ConA of ␣ 1 -PI secreted by HepG2 and HTB55 cells suggested that it could be useful in determining the contribution of epithelial cells to the local synthesis of ␣ 1 -PI in the lung. Because ␣ 1 -PI is believed to diffuse from serum into most organs (2), the proportion of ConA-nonreactive ␣ 1 -PI glycoform should be higher in lung fluids compared with serum if lung epithelial cells are a local source of this inhibitor. Such a difference was clearly observed between LLF and serum from a single healthy donor (Fig. 5). To mimic the postulated contribution of lung-derived epithelial cells to the synthesis of ␣ 1 -PI, the serum was mixed with HTB55 cell medium, and the mixture was examined using CAIE. The results obtained clearly confirm the suggested interpretation because the separation of ␣ 1 -PI in this synthetic mixture resembled that in LLF.
It is possible that the difference in the proportion of ConAnonreactive ␣ 1 -PI in LLF when compared with serum is due to selective transportation into, or concentration within, the lung. This is, however, less likely because the ConA reactivity of AGP was unchanged in LLF. We also examined whether the faster migration of LLF-derived ␣ 1 -PI on ConA CAIE might be due to modifications of serum-derived ␣ 1 -PI within the lung. This was important in view of the fact that differences in the migration of ␣ 1 -PI on agarose gel electrophoresis and crossed immunoelectrophoresis in lung-derived fluids, when compared with serum, have been described, and in those studies, ␣ 1 -PI originating from the respiratory tract appeared to be inactive (22,23). However, our results of SDS-PAGE analysis demonstrate that the ConA-nonreactive fraction of ␣ 1 -PI was active as an inhibitor of HNE (Fig. 6), which suggests that although ␣ 1 -PI is modified within the lung, such a modification does not seem to be manifested by an increase in the amount of ConAnonreactive ␣ 1 -PI.
The data we have obtained also show that ␣ 1 -PI is produced as a fully active inhibitor by lung-derived epithelial cells (Fig.  7), which strongly supports a role for these cells in the local synthesis of ␣ 1 -PI. It is, however, a matter of further analysis to establish other sources of ␣ 1 -PI in the lung, as well as to examine the significance of local production of this inhibitor in protecting the lung tissue from irreversible destruction during inflammatory episodes. FIG. 6. Western blot analysis of ␣ 1 -PI. Aliquots of serum, LLF, V 0 fraction of LLF obtained by affinity chromatography on ConA-Sepharose 4B column (LLFConA), and HTB55 culture medium collected 24 h after OSM (50 ng/ml) and dexamethasone (10 Ϫ6 M) addition (HTB55) were subjected directly to 9% SDS-PAGE or were incubated with HNE (as described under "Experimental Procedures") and then separated on SDS-PAGE. The bands were visualized by immunodetection of ␣ 1 -PI, using enhanced chemiluminescence. FIG. 7. Analysis of complex formation between epithelial cellderived ␣ 1 -PI and HNE. In A, HTB55 cells were incubated for 24 h in MEM supplemented with 50 ng/ml OSM. Aliquots of cell medium or pure serum-derived ␣ 1 -PI (left panel) were incubated with indicated amounts of HNE (HNE:␣ 1 -PI molar ratio is presented) as described under "Experimental Procedures" and then subjected to Western blot analysis. Bands were visualized by enhanced chemiluminescence. In B, HTB55 and normal bronchial epithelial cells were incubated for 18 h in MEM and bronchial epithelial cell basal medium, respectively, both supplemented with 50 ng/ml OSM. Cells were subjected to radiolabeling for 4 h with [ 35 S]methionine/cysteine. Aliquots of cell medium were then incubated with the indicated amounts of HNE as in A (HNE:␣ 1 -PI molar ratio is presented) and immunoprecipitated with human anti-␣ 1 -PI. Immunoprecipitates were subjected to SDS-PAGE followed by fluorography.