α1-Antitrypsin Portland Inhibits Processing of Precursors Mediated by Proprotein Convertases Primarily within the Constitutive Secretory Pathway*

We studied the extent of cellular inhibitory activity of α1-antitrypsin Portland (α1-PDX), a potent inhibitor of proprotein convertases of the subtilisin/kexin type. We compared the inhibitory effects of α1-PDX on the intracellular processing of two model precursors (pro-7B2 and POMC) mediated by six of the seven known mammalian convertases, namely furin, PC1, PC2, PACE4, PC5-A, PC5-B, and PC7. The substrates selected were pro7B2, a precursor cleaved within the trans-Golgi network (TGN), and pro-opiomelanocortin, which is processed in the TGN and secretory granules. Biosynthetic analyses were performed using either vaccinia virus expression in BSC40, GH4C1, and AtT20 cells, or stable transfectants of α1-PDX in AtT20 cells. Results revealed that α1-PDX inhibits processing of these precursors primarily within the constitutive secretory pathway and that α1-PDX is cleaved into a shorter form by some convertases. Evidence is presented demonstrating that in contrast to the full-length α1-PDX (64 kDa), the cleaved (56 kDa) secreted product does not significantly inhibit furin activity in vitro. Cellular expression of α1-PDX results in modified contents of mature secretory granules with increased levels of partially processed products. Biosynthetic and immunocytochemical analyses of AtT20/α1-PDX cells demonstrated that α1-PDX is primarily localized within the TGN, and that a small proportion enters secretory granules where it is mostly stored as the cleaved product.

In view of the potential clinical and pharmacological role of the convertases (2,14), it was of interest to design PC inhibitors. The proposed strategies involved the development of either peptidebased PC inhibitors (7,8,15,16), or protein-based inhibitors (17,18). In humans, the 394-amino acid ␣1-antitrypsin (␣1-AT) is the physiological inhibitor of neutrophil elastase (19,20). A naturally occurring mutation, known as ␣1-AT Pittsburgh (␣1-AT Pittsburgh), at the ␣1-AT reactive site AIPM 358 into AIPR 358 , changed the specificity of this serpin from an inhibitor of elastase into an inhibitor of thrombin (21). Consistent with the critical importance of Arg at the P1 and P4 positions of substrates recognized by furin (10,22), a second mutation in ␣1-AT Pittsburgh giving the sequence RIPR 358 was engineered by Anderson et al. (17), resulting in a potent inhibitor of this convertase. This new variant, called ␣1-AT Portland (␣1-PDX), was shown to inhibit 50% of the furin-catalyzed in vitro cleavage of a small fluorogenic substrate at 0.5 nM concentrations (17). Vollenweider et al. (23), demonstrated that in AtT20 cells, ␣1-PDX only partially inhibits the endogenous processing of gp160 or exogenous processing by furin, PACE4, and PC5-B.
Pro7B2 is a widely expressed endocrine and neuroendocrine precursor (3,24,25), recently shown to be a PC2-specific binding protein (3, 26 -28). Pro7B2 is processed to 7B2 early along the secretory pathway, likely in the trans-Golgi network (TGN) (29), by a PC-like enzyme (27,29), and such processing can be equally accomplished in constitutively secreting cells such as BSC40 cells, as well as in cells exhibiting the presence of dense core secretory granules, e.g. the regulated GH4C1 and AtT20 cells (27). POMC, the precursor of ACTH and ␤-endorphin is processed early into ␤-LPH in the TGN which is then later cleaved to ␤-endorphin within immature secretory granules (1,9).
The present work is aimed at rationalizing the observed difference in the extent of inhibition of the intracellular enzymatic activities of the PCs by ␣1-PDX, in constitutively secreting versus regulated cells. To reach this objective, we compared in both cell types the effect of ␣1-PDX on the processing of two model precursors, pro7B2 and POMC, using the vaccinia virus expression system. Biosynthetic data suggested that ␣1-PDX is primarily secreted via the constitutive secretory pathway and that it can be cleaved by some PCs. Immunocytochemical analysis of ␣1-PDX stably transfected into AtT20 cells demonstrated that this serpin is primarily localized within the TGN.

MATERIALS AND METHODS
Vaccinia Virus Expression-The isolation of recombinant vaccinia viruses (VV) expressing each PC or mouse pro7B2 as well as radiolabeling experiments were previously reported (9,11,23,27,29). For PC7 we made two constructions, one with the full-length rat PC7 (5) starting with the native initiator sequence CTGATGC (VV:PC7), or with a modified GTGATGG (VV:PC7 k ) in which the DNA sequence surrounding its initiator methionine at positions Ϫ3 and ϩ4 better fits Kozak's rules (30). For VV:m7B2  and VV:m7B2  , stop codons were introduced in place of amino acids 151 or 172 (25) and the mutants transferred to the PMJ601 transfer vector and the recombinant viruses obtained as described (11,23,27,29). VV:mPOMC and VV:␣1-PDX, VV:␣1-AT Pittsburgh, and VV:␣1-AT were generous gifts of Dr. Gary Thomas (Vollum Institute, Oregon Health Science University, Portland, OR). BSC40, GH4C1, and AtT20 cells were infected with the specified plaque forming units (pfu)/cell of VV:PC, VV:7B2, and VV:␣1-PDX. The media and cell extracts were immunoprecipitated with a 7B2 (24,27), a ␤-endorphin (9), or an ␣1-AT (Sigma) antibody and the immunoprecipitates resolved by SDS-PAGE (9,11). All biosynthesis experiments were performed at least three times.
Enzymatic Activity Determinations-Enzymatic activity was assessed using the fluorogenic substrate pERTKR-MCA (Peptides International) as described previously (31). Essentially, varying amounts of the concentrated cell media, obtained following VV:␣1-PDX infections, were added to the reaction mixture (containing 50 mM Tris acetate, pH 6.5, 1 mM CaCl 2 , and soluble before cystein-rich domain (BRCD)-hfurin (31)) and allowed to incubate for 30 min at room temperature. Following addition of the fluorogenic substrate, fluorescence was monitored at 0, 30, 60, and 90 min using a Perkin-Elmer spectrofluorometer (model LS 50B). The rate of substrate hydrolysis was constant for at least 60 min. In those assays requiring inhibition of secreted shed furin, concentrated media were first treated with 100 M RVKR-chloromethyl ketone (RVKR-CK) (Bachem) for 30 min at room temperature and the reaction stopped with 0.2% bovine serum albumin followed by four washes on a Centricon-30 (Amicon).
Stable Transfectant of ␣1-PDX in AtT20 Cells-Full-length ␣1-PDX cDNA cloned into the pcDNA3 vector (InVitrogen) or an empty pcDNA3 vector were transfected into AtT20 cells. Three G418-resistant clones expressing ␣1-PDX were isolated. One of the clones expressing the highest amount of ␣1-PDX was selected for further analysis. Stimulation of secretion with 5 mM 8-Br-cAMP was done as described (32). Following separation of media and cell extracts by SDS-PAGE, radioimmunoassays using a ␤-endorphin antibody were performed on material extracted from 1-mm thick gel slices, as described (9).
Immunocytochemical Analysis of ␣1-PDX-expressing Cells-Localization of ␣1-PDX in transfected AtT20 cells was determined by indirect immunofluorescence (33). The ␣1-AT antibody (Sigma) was used at a dilution of 1:100. Guinea pig anti-TGN38 antibody (kindly donated by Dr. W. Garten, Marburg, Germany) was used at a 1:50 dilution and the guinea pig anti-ACTH at 1:100. When primary antibodies were from guinea pig, the immunoreaction was revealed using a fluoresceinelabeled secondary antibody raised in goat (Jackson ImmunoResearch Labs) diluted 1:10 in 10% normal goat serum. Samples were examined using a Zeiss microscope with an epifluorescence attachment, equipped with a Plan-Neofluor 40 ϫ 0.75 objective. Treatment of AtT20 cells with 20 M of the lysosomal inhibitor methionine methyl ester (Sigma) for 4 h was performed as described (33).

␣1-PDX Inhibits Pro7B2 Processing in Several Cell
Lines-To define the extent by which ␣1-PDX inhibits the intracellular processing of precursors, we have co-expressed this serpin with the model substrate pro7B2 (24,25), which is known to be cleaved by one or more PCs within the TGN (29). The processing of pro7B2 (30 kDa) into 7B2 (23 kDa) in BSC40 cells was studied following a pulse of 30 min with [ 35 S]Met and a chase of 60 min. The data show that this cleavage is effectively blocked by ␣1-PDX (Fig. 1). In contrast, ␣1-PDX only partially inhibits pro7B2 processing in GH4C1 and AtT20 cells which contain secretory granules, and the extent of inhibition is greater in the latter cells. As measured by biosynthetic analysis, the expression levels of either 7B2 (Fig. 1) or ␣1-PDX (not shown) in the three cell lines were comparable.
We next compared the inhibitory profile in the above 3 cell lines in which 7B2 and ␣1-PDX were co-expressed with each PC or with the wild type (WT) vaccinia virus control ( Fig. 2A). The convertase PC2 was not studied, as the C-terminal of pro7B2 is an effective inhibitor of this enzyme (26 -28). First, biosynthetic analyses demonstrated that the protein expression levels of furin, PC1, PC5-A, PC5-B, and PC7 are comparable to within a factor of 1.5 in each cell type (not shown). Only the PACE4 immunoreactivity could not be tested, since antibodies recognizing the mature human enzyme were not available. Second, the extent of ␣1-PDX inhibition of the PC-generated processing of pro7B2 is maximal in BSC40 cells as compared with GH4C1 and AtT20 cells ( Fig. 2A). Third, in all three cell types and in the presence of ␣1-PDX, pro7B2 processing by furin and especially by PACE4 and PC5, results in the appearance of a secreted 25-kDa product ( Fig. 2A). These data agree with those of Fig. 1 showing the presence of a 25-kDa intermediate in the media of GH4C1 cells expressing ␣1-PDX. GH4C1 cells are known to express significant amounts of these three PCs (3). Fourth, in BSC40 cells PC5-A produces an intracellular 28-kDa intermediate. Fifth, pro7B2 processing into the 23-kDa 7B2 by PC7 is scarcely inhibited by ␣1-PDX ( Fig. 2A). Sixth, when similar experiments were performed with either ␣1-AT Pittsburgh or ␣1-AT, no inhibitory effect on the processing of pro7B2 by any PC was observed in the above cell lines (not shown), emphasizing the PC specificity of ␣1-PDX.
The N-terminal immunoreactive 25-and 28-kDa intermediates observed in the presence of PC5 and ␣1-PDX are likely to represent C-terminally truncated forms of pro7B2, possibly resulting from cleavage at either one of the only two potential pairs of basic residues QGKR 165 and VAKK 173 found in the segment 156 -186 of mouse pro7B2 (25). Since C-terminal sequencing of radiolabeled proteins is not trivial, we attempted to identify these intermediates by generating C-terminally truncated mutants of pro7B2. Thus, in Fig. 2B we compare the results of co-expression of PC5-A with pro7B2 (1-186) (30 kDa), pro7B2 (1-171) (28 kDa), or 7B2 (1-150) (23 kDa) in the absence or presence of ␣1-PDX. The data show that PC5 does not process 7B2  , but cleaves pro7B2  and pro7B2  to produce 23-and 25-kDa forms of 7B2, suggesting cleavage at both RRKRR 155 and QGKR 165 (25), respectively. The migration position of pro7B2 (1-171) (28 kDa) resembles that of the product generated by PC5 in the presence of ␣1-PDX ( Fig. 2A), indicating that the 28-kDa form is obtained following cleavage at VAKK 173 .
Thus, the data in Figs. 1 and 2 lead us to conclude that in the constitutively secreting BSC40 cells, ␣1-PDX is effective in inhibiting the cleavage of pro7B2 into the 23-kDa 7B2 by all convertases tested except for PC7. However, alternative processing into a 25-kDa form of 7B2 by furin, PACE4, and PC5 is not effectively inhibited by ␣1-PDX. In contrast, in regulated cells and especially in AtT20 cells, ␣1-PDX is only partially able to inhibit the PC-mediated processing of pro7B2, possibly because of the participation of endogenous convertases sorted to secretory granules, such as PC1.
Biosynthetic Analysis of ␣1-PDX and Its Processing by PCs-Since ␣1-PDX was derived from ␣1-AT by mutagenesis of its enzyme-binding site into an Arg 355 -X-X-Arg 358 sequence (17), which is the minimal recognition motif of PCs (12,22,34), it was plausible that the selectivity of ␣1-PDX for PCs (17,23) reflects their ability to recognize this motif and possibly to cleave it. Accordingly, we undertook the analysis of the biosynthetic products of ␣1-PDX in BSC40 versus AtT20 cells in the absence or presence of co-expressed furin or PC7, which are differentially inhibited by ␣1-PDX in BSC40 cells ( Fig. 2A).
Cells were pulse-labeled with [ 35 S]Met for 10 min followed by chase periods of 10, 20, and 45 min (AtT20 cells) or 45, 90, and 180 min (BSC40 cells) (Fig. 3). In both cell types, ␣1-PDX is first synthesized as a 59-kDa protein which is progressively transformed into a 64-kDa (AtT20 cells) or 62-kDa (BSC40 cells) secretable protein. Interestingly, the data show that the rate of cellular trafficking in AtT20 cells is much more rapid than in BSC40 cells, since the secreted ␣1-PDX is first detected in these cells at 45 versus 90 min of chase, respectively. Notice the presence of very low levels of a 52-54-kDa form of ␣1-PDX at the 10-min pulse. We believe that this represents either an unglycosylated form of ␣1-PDX (see Fig. 6), as observed for PC1 and PC2 (11), or a nonspecific protein recognized by our antibody.
When either furin or PC7 are co-expressed with ␣1-PDX, a novel 56-kDa (AtT20 cells) or 54-kDa (BSC40 cells) form of the serpin became apparent. We also note that overexpression of PC7 k (see "Materials and Methods") results in a much lower level of recovered radioactivity, suggesting a destruction of ␣1-PDX epitopes either directly or indirectly, e.g. via the activation of one or more latent proteinases. In view of the engineered Arg 355 -X-X-Arg 358 site of ␣1-PDX   (17), and as no other putative PC-cleavage site exists within the sequence of this human serpin (19), it is likely that these lower molecular weight forms of ␣1-PDX represent products preferentially cleaved at this engineered site. In agreement with this hypothesis, furin did not cleave ␣1-AT in BSC40 cells (not shown), whereas it did so for ␣1-PDX (Fig. 3).
The comparative ability of each PC to process ␣1-PDX in AtT20 versus BSC40 cells was analyzed following a pulse of 20 min with [ 35 S]Met and a chase of 60 min (Fig. 4). In both cell types, PC7, furin, PACE4, and PC5-B cleave ␣1-PDX to produce the 54/56-kDa form. PC1 partially (Ͻ10%) cleaves this serpin only in AtT20 cells and PC2 does not cleave ␣1-PDX in either cell type. Interestingly, even though some reduction in ␣1-PDX levels is seen when either furin or PC2 are present in AtT20 cells, we consistently observed that co-expression of ␣1-PDX with either PC5-A in both cell types or of PC7 k in BSC40 cells, caused the virtual disappearance of the ␣1-PDX immunoreactivity. This may mean that, like PC7 in BSC40 cells (Fig. 3), PC5 may activate in both cell types a latent intracellular proteinase(s) which causes the destruction of the ␣1-PDX epitopes.
To define the intracellular site where cleavage of ␣1-PDX occurs, pulse-chase experiments were performed in the presence of brefeldin A (BFA) or at 20°C. The fungal metabolite BFA is known to cause the disassembly of the Golgi complex and fusion of the cis, medial, and trans-Golgi (but not the TGN) with the endoplasmic reticulum (35). The 20°C incubation traps secretory proteins at the level of the TGN (36). As shown in Fig. 5, when AtT20 cells were pulse-labeled for 20 min followed by chase periods of 45 and 90 min, the majority of ␣1-PDX was transformed into its mature 64-kDa form, which in the presence of furin is partially cleaved intracellularly into a secretable, broad 56-kDa form. Neither the 64-nor 56-kDa products appear in the presence of BFA, where an endoglycosidase H-sensitive 57-kDa form is observed (see Fig. 6A). This suggests that the furin-mediated cleavage of ␣1-PDX takes place in a secretory compartment following the trans-Golgi. At 20°C and in the presence of overexpressed furin, the level of the 64-kDa protein is reduced with a concomitant slight increase in the amount of the 56/59-kDa forms. This suggests that the processing compartment could be, in part, the TGN.
To define more precisely the intracellular localization of the various glycosylated (19) ␣1-PDX forms observed, we analyzed their sensitivities to digestion with endoglycosidase H (endo H) and N-glycosidase F (endo F). In both AtT20 and BSC40 cells, following a pulse of 20 min, the major 59-kDa immunoreactive ␣1-PDX is sensitive to digestion by both endo H and endo F (Fig. 6A), suggesting its presence in the endoplasmic reticulum where it has not acquired its mature complex carbohydrate structure. Following a chase of 90 min in AtT20 cells, the majority of ␣1-PDX is sensitive to endo F and is endo H resistant. These characteristics suggest that after a 90-min chase most of the ␣1-PDX left the endoplasmic reticulum and reached late Golgi compartments. As previously observed (Fig. 3), the traffic rate of ␣1-PDX in BSC40 cells is slower than in AtT20 cells, explaining the relatively larger proportion in BSC40 cells at the 90-min chase which is still endo H sensitive. However, in the media all secreted forms of ␣1-PDX are endo H resistant and endo F sensitive, emphasizing their mature complex Nglycosylation pattern. In AtT20 cells at 20°C, the mature 64-kDa ␣1-PDX was no longer sensitive to endo H. Only the 64and 56-kDa forms (AtT20 media) and the 62-kDa forms (BSC40 media and faintly in cells) are sulfated (Fig. 6B). Since the sulfotransferases are localized in the TGN (37), this strongly argues that these are the final mature products of ␣1-PDX. In the presence of BFA about 40% of ␣1-PDX was resistant to endo H in AtT20 cells, whereas in BSC40 cells all of ␣1-PDX was sensitive to this glycosidase digestion, again emphasizing the difference in trafficking rates between these two cell types. The sum of the data presented in Figs. 3-6, demonstrate that in AtT20 cells the mature, secretable glycosylated 64-kDa ␣1-PDX can be processed in the TGN into a 56-kDa form.
In Vitro Inhibitory Activity of the 64-and 56-kDa Forms of ␣1-PDX-To test the inhibitory activity of the 56-kDa form of ␣1-PDX, we overexpressed this serpin in AtT20 cells with or without furin. Western blots revealed that in the absence of furin, the 64-kDa product represents more than 85% of secreted ␣1-PDX, whereas in its presence the 56-kDa product is the major immunoreactive protein secreted (Fig. 7A). Molecular sieving fractionation of the secreted product in the ␣1-PDX/furin coexpression experiment, followed by an ␣1-AT-specific radioimmunoassay (not shown), also revealed that the major immunoreactive secreted form of ␣1-PDX migrates as a 56-kDa protein (Fig. 7A). Analysis of the in vitro furin-mediated cleavage of a pentapeptide fluorogenic substrate (7,8) was used to evaluate the inhibitory potency of the 64-kDa ␣1-PDX and its 56-kDa product. In the case of the ␣1-PDX/furin coexpression, where enzymatically active furin is shed into the medium (not shown), it was necessary to first inhibit the secreted furin activity with 100 M RVKR-chloromethyl ketone, followed by removal of the unreacted inhibitor (see "Materials and Methods"). This process did not affect the inhibitory potency of the 64-kDa ␣1-PDX (Fig.   FIG. 6. N-Glycosylation state of the molecular forms of ␣1-PDX. Cells were infected with 2 pfu/cell of VV:PDX. A, the cells were then pulse-labeled with [ 35 S]Met for 20 min (P20Ј) and then chased for 90 min in the presence or absence of 2.5 g/ml BFA. The cell extracts and media were then immunoprecipitated with an ␣1-AT antibody and the immunoprecipitates either incubated with buffer (control), endo H, or Nglycosidase F (endo F), as described (11,23). The products were separated on an SDS-PAGE gel as in Fig. 4. B, to define the molecular forms of PDX reaching the TGN, the cells were pulse labeled with Na 2 [ 35 SO 4 ] for either 90 min (P90) or 20 min followed by a chase of 90 min (P20ЈC90Ј). Notice that in both cell types most of the ␣1-PDX is secreted in the medium as a mixture of 64 kDa (major) and 56 kDa (minor) forms (AtT20 cells) or as a 62-kDa protein (BSC40 cells). 7B). The results indicate that only when the 64-kDa form of ␣1-PDX is predominant can we detect significant inhibition of furin in vitro (Fig. 7B), suggesting that the secreted 56-kDa form is not an effective inhibitor of this enzyme.
␣1-PDX Does Not Affect the Processing of ProPC1 or Profurin and Partially Inhibits That of ProPC2-Since ␣1-PDX did inhibit the activity of some of the pro7B2 convertases in AtT20 cells (Figs. 1 and 2), it was possible that it could either bind to these enzymes in the TGN itself or earlier along the secretory pathway. Accordingly, we investigated whether ␣1-PDX could affect the zymogen cleavage of the convertases. It is now accepted that profurin (38) and proPC1 (2,11,13) are converted into furin and PC1, respectively, within the endoplasmic reticulum. In contrast, proPC2 to PC2 conversion occurs in the TGN (2,11,13) and continues in immature secretory granules (39). Co-expression of ␣1-PDX with furin or PC1 in AtT20 cells did not affect their zymogen conversion, the secretion of PC1 or that of shed furin (not shown). Similarly, in AtT20 cells, ␣1-PDX did not affect the conversion of [ 35 SO 4 ]proPC2 (77 kDa) into either the 71-or 68-kDa PC2 (11, 28) (Fig. 8). In contrast, in BSC40 cells which do not express endogenously the PC2binding protein 7B2 (3) and where proPC2 is mostly converted into the 70-kDa PC2 intermediate (27,34), ␣1-PDX partially inhibited this cleavage performed by either endogenous en-zyme(s) or by co-expressed furin (33), resulting in a higher level of secreted 77-kDa proPC2 (Fig. 7). Thus, processing of proPC2 to the 71/70-kDa form which occurs in the TGN of constitutive cells could be partially blocked by ␣1-PDX. However, in regulated cells which express 7B2 endogenously and where proPC2 processing into the final 68-kDa form occurs in immature secretory granules (39), ␣1-PDX has no significant effect on this zymogen cleavage.
Stable Transfectant of ␣1-PDX in AtT20 Cells-To better define the intracellular localization of ␣1-PDX and assess its ability to inhibit processing of precursors within the regulated secretory pathway, we obtained three stable transfectants of ␣1-PDX in AtT20 cells. For detailed analyses, we selected the one with the highest expression of ␣1-PDX (Figs. 9-11). Following a 4-h pulse, the majority of the ␣1-PDX is found in the medium (Fig. 9). Furthermore, cAMP, which is known to stimulate secretion from mature secretory granule, has little effect on the release of the 64-kDa form, while exerting a 2.5-fold stimulation of the secretion of the 56-kDa form (Fig. 9). This suggests that the secretion route of the major 64-kDa ␣1-PDX is mostly from the constitutive and/or constitutive-like secretory pathways (40,41), and that the minor 56-kDa form (Ͻ15% of the total ␣1-PDX content) is stored in granules.
We next investigated whether ␣1-PDX can inhibit processing of POMC, a precursor normally cleaved in immature secretory granules, resulting in products stored in dense core granules. The tissue-specific processing of POMC has now been explained by the relative expression levels of PC1 and PC2 (9,42,43). Thus, PC1 is mostly responsible for the production of ␤-LPH and ACTH in anterior pituitary corticotrophs, while PC2 transforms these products into ␤-endorphin and ␣-melanotropin-stimulating hormone, respectively (9). Our control AtT20 cells express PC1 which is responsible for the production of ␤-LPH (9) and to a lesser extent PC2 (42,43), explaining the production of ␤-endorphin (9). Upon cAMP stimulation, we observed a 3-fold increase in the level of secreted ␤-endorphin from wild type AtT20 cells with no significant change in that of ␤-LPH. This result suggests that ␤-endorphin is stored in secretory granules and that ␤-LPH and unprocessed POMC are mostly secreted from the constitutive or constitutive-like pathways. However, for AtT20/␣1-PDX cells, cAMP stimulated the release of ␤-LPH and ␤-endorphin, demonstrating that in these cells both polypeptides are stored in granules. Furthermore, in contrast to control cells, POMC can be detected in the medium of stimulated AtT20/␣1-PDX cells, suggesting that in these cells a proportion of POMC is stored in granules.
To better quantitate the levels of POMC products, media and cells extracts were run on SDS-PAGE, followed by radioimmunoassay analyses using a ␤-endorphin antibody (Table I). Similar to radiolabeling results (Fig. 9), the presence of ␣1-PDX causes a 7-fold reduction in the basal secretion level of ␤-LPH. Stimulation of secretion by cAMP (ϩ) results in a 1.3-and 5-fold increase in the percentage of secreted ␤-LPH in control and AtT20/␣1-PDX cells, respectively. The data also show that ␣1-PDX does not significantly affect the degree of cAMP-stimulated secretion of ␤-endorphin from either cells. In contrast, whereas cAMP does not affect the percentage of secreted POMC in AtT20 control cells, it causes a 2-fold increase in the percentage of secreted POMC from AtT20/␣1-PDX cells, suggesting an accumulation of POMC in secretory granules. This phenomenon could be explained by the delayed processing of POMC in ␣1-PDX-expressing cells until it reaches immature granules where it can be processed consecutively by PC1 into ␤-LPH and then by PC2 into ␤-endorphin (9,43).
The latter conclusion is further reinforced by the analysis of the processing products of proPC1 and pro7B2 within AtT20/ ␣1-PDX cells (Fig. 9). The convertase proPC1 is known to be first processed into PC1 (84 kDa) within the endoplasmic re-ticulum (11) and then transported to the TGN and immature granules where its transformation into the C-terminally truncated 66-kDa form occurs (11,44). In control cells, the 66-kDa PC1 is the major form stored in mature granules and secretable by cAMP, and the 84-kDa form is primarily secreted constitutively. In contrast, in the presence of ␣1-PDX, cAMP stimulates the secretion of both the 84-and 66-kDa forms of PC1. In a similar vein, in the presence of ␣1-PDX the secretion of both pro7B2 and 7B2 are stimulated by cAMP, whereas in control cells only the 23-kDa form of 7B2 is secreted (Fig. 8). The latter results are compatible with a delayed processing of the 84-kDa PC1 and the 30-kDa pro7B2 caused by ␣1-PDX. Similar analyses with a clone expressing lower amounts of ␣1-PDX revealed reduced inhibitory levels (not shown).
Immunocytochemical Localization of ␣1-PDX-Immunoreactivity of ␣1-PDX in stably transfected AtT20 cells was analyzed by conventional immunofluorescence methods. Results (Fig.  10A), revealed the presence of heavily stained paranuclear structures resembling the Golgi apparatus. Intracellular punctate-like granulation was also observed within the cytoplasm and in the tips of some cellular extensions. In an effort to identify these organelles, we first compared the localization of ␣1-PDX to the POMC-derived endocrine marker adrenocorticotrophic hormone (ACTH) using an ACTH antiserum which recognizes both POMC and ACTH (Fig. 10B). This was achieved within the same cells with a combination of rhodamine labeling of anti-rabbit antibodies for ␣1-PDX and fluoresceine labeling (arrows) of anti-guinea pig antiserum for ACTH. The black and white representation of the labeling is presented in Fig. 10, A and B, respectively. The data demonstrate that immunoreactive POMC/ACTH and ␣1-PDX exhibit a partial co-localization. This is especially evident within paranuclear Golgi-like structures (heavy arrows). However, while numerous intra-cytoplasmic granular structures were ACTH positive, no equivalent labeling for ␣1-PDX was detected (small light arrows). In addition, differences were evident at the tips of some cellular extensions which labeled for ACTH and not for ␣1-PDX (open arrows).
To verify the hypothesis that the common organelles containing both ACTH and ␣1-PDX are primarily limited to Golgi-like structures, we tested the co-localization of ␣1-PDX with the TGN marker TGN38 (45) (Fig. 11). Within compact TGN structures in paranuclear positions (heavy arrows) and in cytoplasmic punctate structures (light arrows), ␣1-PDX and TGN38 are remarkably co-localized (Fig. 11, A and B). In control cells, immunoreactivity of TGN38 is also evident (Fig. 11D), and no ␣1-PDX is detected (Fig. 11C). Thus, these data strongly sug-  Fig. 4. The AtT20 cells' endogenous POMC, PC1, and 7B2 products were immunoprecipitated with their respective antibodies.

TABLE I
Comparison by radioimmunoassay analysis of the production of POMC and its metabolites (␤-LPH and ␤-endorphin) in AtT20 cells Samples from transfected cells (see legend to Fig. 9) were processed and separated via SDS-PAGE as described (9). A ␤-endorphin antibody (9) was used to detect the fractionated products. AtT20

␣1-PDX/AtT20
Ϫ a Values (as a percentage) represent the fraction of the immunoreactivity of each species divided by the total amount of immunoreactivity measured in the sample (cells plus medium).
gest that the cytoplasmic punctate structures are TGN-derived and that ␣1-PDX immunoreactivity is primarily limited to structures containing TGN38. However, cAMP stimulation results ( Fig. 9), and the presence of ␣1-PDX immunoreactivity at the tips of cellular extensions (Fig. 10A), suggest that a small proportion of ␣1-PDX can enter secretory granules. Furthermore, our results with the lysosomal enzyme inhibitor methionine methyl ester, known to block endocytotic catabolism (33), also demonstrate that a fraction of ␣1-PDX can enter endocytotic and/or lysosomal compartments, where it is probably degraded (not shown).

DISCUSSION
Tissue-specific processing of inactive precursors by convertases into active polypeptides is a general mechanism to generate and regulate the level of biological diversity achieved with a given proprotein (1)(2)(3)(4)(5). As one or more of convertases is responsible for the activation of host cell pathogens such as viruses and toxic bacteria (1,2,12), as well as growth factors involved in cellular proliferation and adhesion, these serine proteinases are considered targets for the development of inhibitor-based therapeutic approaches (14). Various DNA-and protein-based strategies have been suggested to silence the expression of the PCs (17,20,34). The protein-inhibitor based approach has so far led to the development of ␣1-PDX which was shown to inhibit convertasecatalyzed processing (17,23,46).
The present work examined several aspects of the cellular biology of ␣1-PDX and its ability to inhibit precursor processing by PCs. In general, we find that ␣1-PDX inhibits precursor processing within the constitutive secretory pathway in more than one cell type. Thus, in the constitutively secreting BSC40 cells, this variant serpin inhibits the processing of pro7B2 into its 23-kDa product by most convertases (Fig. 2A). In contrast, in cells endowed with secretory granules such as GH4C1 and AtT20 cells, the ability of ␣1-PDX to inhibit the pro7B2 processing is much reduced (Figs. 1 and 2). Notably, PC7 activity is not well inhibited by ␣1-PDX even in BSC40 cells ( Fig. 2A), a result which was confirmed in vitro (not shown). Moreover, small fluorogenic substrates containing an ␣1-PDX-like RXXR sequence are well cleaved by furin but only poorly by PC7 (31). In view of the reduced trafficking rate of ␣1-PDX in BSC40 cells as compared with AtT20 cells (Fig. 3), it is possible that the enzyme and inhibitor spend more contact time along the secretion route of BSC40 cells as compared with that in AtT20 cells, which could provide a rationale for the observed increased inhibitory potency in the former cells. We also examined the biosynthetic forms of this inhibitor in both cell types. Our results (Figs. 3, 4, and 9) demonstrated that ␣1-PDX is produced as a mature 62/64-kDa precursor which can be cleaved by furin, PC7, or PACE4 into a 54/56-kDa form in all cells examined, and by PC1 in AtT20 cells. Since only ␣1-PDX is cleaved by PCs, whereas ␣1-AT is not, the engineered Arg 355 -X-X-Arg 358 site in ␣1-PDX is the likely cleavage site. We also examined the possibility that the truncated 56-kDa form of ␣1-PDX can still inhibit the PCs, as is the case of the inhibition of elastase by cleaved ␣1-AT (20). The available data (Fig. 7B) suggest that the secreted, full-length ␣1-PDX is active, whereas its truncated form is essentially inactive. In contrast to the complex formed between ␣1-AT and serine proteases of the chymotrypsin type (20), it appears that the complex of ␣1-PDX or its truncated form and a PC is not a covalently stabilized structure, since it can be dissociated by SDS. Thus, our data showing cleavage of ␣1-PDX, and the reported ability of 0.5 nM ␣1-PDX to inhibit 50% of furin activity in vitro (17), suggest that ␣1-PDX could effectively compete with endogenous substrates for the occupancy of the catalytic pocket of the PCs, thus acting as a competitive inhibitor. Transfection of ␣1-PDX in AtT20 cells subtly changed the composition of the granules, enriching them with unprocessed and/or partially processed products, suggesting a delayed processing (Fig. 9). We do not know whether the ␣1-PDX which enters granules can still retain inhibitory activity, but based on our in vitro data the cleaved 56-kDa form is not expected to be active (Fig. 7). Thus, the ineffective inhibition of POMC processing may be due to the low level of serpin which enters granules and to its processing within. In this context, recent kinetic data with ␣1-AT revealed that more than four molecules of serpin are needed to productively inhibit subtilisin (47). That ␣1-PDX actually binds to some PCs is supported by results which show the co-immunoprecipitation of PC1, PC5, or PC7 with ␣1-PDX under non-denaturing conditions (not shown). Our data revealed that ␣1-PDX is primarily targeted to the constitutive or constitutive-like secretory pathways (40,41), but that some of the protein does enter granules where it can be processed. This conclusion is supported by the insensitivity of the majority of ␣1-PDX secretion to cAMP stimulation, except for the processed 56-kDa form (Fig. 9). In addition, immunocytochemical data demonstrated that the intracellular localization of ␣1-PDX is remarkably similar to that of TGN38 (Fig. 11, A and B) and shows important differences from that of the granule-associated ACTH (Fig. 10).
Unexpectedly, ␣1-PDX allowed us to define a novel processing site of pro7B2, proposed to be at the sequence QGKR 165 and which is especially preferred by PC5 and PACE4 ( Fig. 2A). Such cleavage could lead to the formation of a C-terminally amidated 7B2 (1-162-amide) . Interestingly, this dibasic site is not conserved across species, e.g. the QGKR 165 sequence is replaced by QGQR 165 in human and Xenopus 7B2 (48). It has been reported that the pro7B2 CT-peptide representing the fragment 7B2 (156 -186) inhibits the activity of PC2 (26), and that internal cleavage and trimming of the CT-peptide at the conserved KK 173 (25,34,48) by PC2 inactivates its effect (49). Therefore, it is possible that cells which express PC5 together with PC2 and 7B2 (e.g. pancreatic glucagon cells (32)) may exhibit an early release of inactive CT-peptide fragments cleaved at either KK 173 and/or KR 165 and hence liberate active PC2 to act earlier than in other cells. It should be noted that our results differ from those recently reported by Mains et al. (50), who used a spliced form of rat PACE4 which has 2 and 13 amino acids missing in the catalytic and P-domains, as compared with the full-length human PACE4 (called PACE4-A) used in our study. Their data show that ␣1-PDX does not inhibit rPACE4 activity in vitro, a result which differs from ours where ␣1-PDX inhibits the activity of hPACE4-A both in vitro (46) and ex vivo (Ref. 23, and this work).
In conclusion, we demonstrated that ␣1-PDX may not inhibit the processing of all precursors to a similar extent and that processing inhibition occurs primarily within the constitutive secretory pathway. Therefore, ␣1-PDX is a very useful lead protein to inhibit processing of precursors including endogenous growth factors and imported viral surface glycoproteins in constitutive cells, and may exhibit a limited toxicity to cells in vivo. Finally, further variations in the structure of this serpin may lead to a more specific inhibitor which may discriminate between the convertases.