Cell surface localization of proteolysis of human endothelial angiotensin I-converting enzyme. Effect of the amino-terminal domain in the solubilization process.

Angiotensin-converting enzyme (ACE) belongs to the type I class of ectoproteins and is solubilized by Chinese hamster ovary cells transfected with the full-length human ACE cDNA. ACE release in Chinese hamster ovary cells involves a proteolytic cleavage occurring in the carboxyl-terminal region, between Arg-1137 and Leu-1138. The subcellular localization of ACE proteolysis was established by pulse-chase experiments, cell surface immunolabeling, and biotinylation of radiolabeled mature proteins. The proteolysis of ACE takes place primarily at the plasma membrane. The solubilization of ACE is less than 2% within 1 h, is increased 2.4-fold by phorbol esters, but is not influenced by ionophores. An ACE mutant lacking the transmembrane domain and the cytosolic part (ACE), is secreted at a faster rate without a carboxyl-terminal cleavage, and phorbol esters or ionophores have no effect on its rate of production in the medium. Therefore, the proteolysis of ACE is dependent on the presence of the membrane anchor and suggests that the secretase(s) involved is also membrane-associated. An ACE mutant lacking the amino-terminal domain (ACE) is secreted 10-fold faster compared with wild-type ACE. The solubilization of ACE occurs at the plasma membrane and is stimulated 2.7-fold by phorbol esters, and the cleavage site is localized between Arg-1227 and Val-1228. The amino-terminal domain of ACE slows down the proteolysis and seems to act as a “conformational inhibitor” of the proteolytic process, possibly via interactions with the “stalk” of ACE and the secretase(s) itself.

Proteolytic release of the ectodomain of carboxyl-terminally anchored proteins (type I class ectoproteins) leads to the production of the protein in the extracellular space and, in some cases, in biological fluids. It is a widespread phenomenon (1), concerning several proteins with different biological functions such as membrane-anchored growth factors like pro-transforming growth factor ␣ (TGF-␣) 1 (2), the c-kit ligand (3), growth factor receptors like the tumor necrosis factor receptor (4), cell adhesion molecules (1), ectoenzymes such as angiotensin I-converting enzyme (ACE) (1) or lactase-phlorizin hydrolase (5) and the amyloid precursor protein (APP) (6,7). However, little is known about the molecular basis of this secretion. A common mechanism involving a processing enzyme named "secretase" has been proposed (1,2,6). The characteristics of this putative enzyme, which could be itself membrane-associated are broad sequence specificity (2,6) and cleavage at a certain distance from the membrane (8).
ACE (EC 3.4.15.1) is a zinc-dipeptidyl carboxypeptidase (9) that also displays endopeptidase activity on some peptides (10). ACE is an ectoenzyme found in most mammalian tissues bound to the external surface of the plasma membrane of endothelial, epithelial, neural, and neuroepithelial cells.
There are two ACE isoforms derived from a single gene, by the transcription from two alternative promoters (11,12), which are expressed in a tissue-specific fashion. In somatic tissues, ACE is expressed as a glycoprotein composed of a single polypeptide chain with an apparent molecular mass of 170 kDa, containing two large homologous domains, called the N and C domains, each domain bearing an active catalytic site (13,14); in male germinal cells, ACE is synthesized as a lower molecular mass form of 110 kDa containing only the active C domain of endothelial ACE (15). Both isoforms belong to the type I class of integral transmembrane ectoproteins (1,16) with a large amino-terminal extracellular domain, a 17-amino acid hydrophobic anchor located 30 amino acid residues from the cytosolic carboxyl terminus (13). The role of this ␣-helix in the membrane anchoring of ACE has been demonstrated by studying a carboxyl-terminal truncated cDNA (ACE ⌬COOH ), which is secreted faster in the medium of CHO cells (17).
ACE is solubilized and circulates in many body fluids such as serum (18). The source of circulating ACE in plasma is thought to be derived mainly from endothelial pulmonary cells. Wei et al. (17) showed that the full-length endothelial recombinant ACE cDNA, expressed in CHO cell lines, is secreted into the culture medium from the membrane form by a post-translational proteolytic cleavage occurring in the carboxyl-terminal region. Recently (19), we have established by carboxyl-terminal microsequencing the same carboxyl terminus sequences AGQR for secreted recombinant ACE and for human plasma ACE, which corresponds to a cleavage site between Arg-1137 and Leu-1138 (see Fig. 1).
Sen et al. (20) showed that the rabbit testicular ACE cDNA expressed in a mouse epithelial cell line is secreted in the culture medium by a proteolytic processing of its carboxylterminal region. The cleavage of the rabbit testicular ACE also occurs at a monobasic site between Arg-663 and Ser-664 (21). This cleavage site corresponds in endothelial human ACE to positions Arg-1203 and Ser-1204, downstream from the cleavage site we observed for the full somatic ACE (Fig. 1). The mutation of Arg-1137 to a glutamine residue did not prevent the secretion of ACE (19). Altogether, these findings suggest several hypotheses since the enzyme(s) responsible for ACE cleavage is (are) not identified. 1) A single enzyme of broad specificity could be involved, which could accommodate an Arg to Gln substitution in the P1 site of the human ACE. 2) Several related enzymes could be implicated. 3) Species differences could account for the differences in the solubilization of human and rabbit ACE observed in the two cell lines.
The subcellular localization of the proteolytic cleavage of endothelial ACE in CHO cells is unknown and could occur intracellularly or at the plasma membrane. The molecular mechanisms of its regulation are also unknown.
The aim of the present study was to identify the subcellular localization of the proteolytic cleavage of ACE and its regulation by metabolic labeling in CHO cell lines permanently transfected with the cDNA of endothelial ACE. This study shows that the solubilization takes place at the plasma membrane and that ACE solubilization under basal conditions is low and enhanced by phorbol esters but not by ionophores, implying a calcium-independent protein kinase C. The solubilization rate also depends on the presence of the N domain; an aminoterminal truncated mutant (ACE CF ) was secreted 10-fold faster compared with the wild-type enzyme comprising both domains, and its cleavage site was different.

Construction of Expression Plasmids
The construction of the expression plasmids has been described previously (14,17). Expression plasmids, peCHO-ACE containing the fulllength ACE cDNA, peCHO-ACE ⌬COOH a carboxyl-terminal truncated cDNA in which the last 47 amino acids at the carboxyl terminus and the putative transmembrane domain are not translated (17), and peCHO-ACE CF lacking the entire amino-terminal domain (14), were cotransfected with the neomycin resistance plasmid pSV2neo in CHO cells by calcium phosphate precipitation. Single colonies of primary G418-resistant transformants were assayed for the expression of ACE activity using p-benzoyl-L-glycyl-L-histidyl-L-leucine (Bachem, Switzerland) as substrate (22). The detection and quantification of the liberated hippuric acid was performed by high performance liquid chromatography (23). Cell lines expressing ACEs were selected and purified by subcloning using the limiting dilution technique. Cells were maintained in Ham's F-12 medium supplemented with 10% fetal calf serum, penicillin (100 units/ml), and streptomycin (100 g/ml).

Metabolic Labeling and Immunoprecipitation
Wild-type and mutant ACE CHO cell lines with similar cellular contents of ACE activity were seeded in 60-mm dishes at a density of 2 ϫ 10 6 cells/dish and grown to confluency. Cells were then incubated at 37°C for 30 min, in methionine-, cysteine-, and serum-free medium (Ham's F-12). The medium was removed, and the cells were incubated in 2 ml of the same medium containing 50 Ci/ml [ 35 S]methioninecysteine for 15 or 30 min. The radioactive medium was then replaced with serum-free UltraCHO medium (Bio-Whitaker), and the cells were incubated at 37°C for the indicated chase periods.
After metabolic labeling and chase periods, the culture medium was collected and centrifuged at 1,000 ϫ g for 5 min at 4°C, and EDTA was added to the resultant supernatant to a final concentration of 10 mM. Cell extracts were prepared by washing labeled cells with ice-cold phosphate-buffered saline (PBS) and were solubilized in 0.5 ml of lysis buffer/dish, 1% Triton X-100 in 50 mM Tris-HCl buffer, pH 7.4, containing 150 mM NaCl, 10 mM EDTA, and 0.15% SDS for 30 min at 4°C. Cell lysates were collected, and insoluble material was removed by centrifugation for 10 min at 10,000 ϫ g at 4°C. Cell lysate supernatants and media were used immediately or stored at Ϫ70°C until further use.
Three different rabbit antisera were used for immunoprecipitation (see Fig. 1). Antiserum Y1 was obtained from rabbits immunized against pure human kidney ACE (13). Antisera 28A and Clo were raised in rabbits against synthetic peptides corresponding to amino acids 1258 -1277 at the carboxyl terminus and 1-18 at the amino terminus of the human endothelial ACE sequence, respectively (14,19). Cell lysates or culture media were incubated overnight at 4°C with undiluted antisera and protein A-Sepharose (Pharmacia Biotech Inc.) (50 l of a 50% suspension in lysis buffer). The immune complex protein A-Sepharose was collected by centrifugation and washed 4 times with 0.1% FIG. 1. Wild-type ACE, carboxyl-, and amino-terminally truncated mutants. A, diagram of human ACEs encoded by the cDNA constructions. Human wild-type ACE, encoded by the full-length cDNA, is composed of 1277 residues, comprising the N and C homologous domains (egg-shaped boxes) in the extracellular region and the hydrophobic segment of 17 amino acids near the carboxyl terminus (square boxes). The carboxyl-terminally truncated ACE (ACE ⌬COOH ) comprises 1230 residues. The 47 amino acids deleted from the carboxyl terminus correspond to the transmembrane and the cytosolic regions. The amino-terminally truncated ACE (ACE CF ) comprises residues 1-4 of the amino terminus followed by residues 572-1277 of mature ACE. Antiserum Y1 directed against pure human kidney ACE is shown. Sequences of synthetic peptides used for the production of antisera 28A, Clo, 5, and 3 are also shown in the diagram. B, carboxyl-terminal sequence of ACE from Cys-1114 to Ile-1254. Cleavage sites of human endothelial ACE (R1137), rabbit testicular ACE (R1203), and ACE CF (R1227) are shown by an arrow. The hydrophobic domain is boxed.
Triton X-100 in 50 mM Tris-HCl containing 1 mM EDTA, 0.15% SDS and once with 20 mM Tris-HCl, pH 6.8, and then dried. Samples were heated for 5 min at 100°C in 25 l of SDS-PAGE electrophoresis sample buffer. The Sepharose was removed by centrifugation, and proteins were resolved by 6 or 10% SDS-polyacrylamide gel electrophoresis (24) and revealed by autoradiography. Proteins were quantified from autoradiograms using a video densitometer (MC View Color, Agfa).
In some experiments a two-step immunoprecipitation method was used. Cells were pulse-labeled as described above and then chased in some cases in a serum-free medium for 4 or 16 h. Cell lysates (20 l) were immunoprecipitated overnight with the antibody 28A, at the concentrations indicated in the figure legends. The immune complex protein A-Sepharose was removed, and the supernatant was heated at 100°C for 5 min and immunoprecipitated for 4 h with antibody Y1 (3 l).

Cell Surface Localization
Cell Surface Immunolabeling-CHO cell lines were labeled as described above and chased for different periods in UltraCHO medium. Cells were then washed 4 times with ice-cold PBS. After washing, 10 l of antibody Y1 or 28A in 2 ml of ice-cold PBS were added to the cell cultures. After 1 h of incubation with gentle agitation at 4°C, the cells were washed 4 times with PBS. Cells were lysed in lysis buffer as described previously, and the lysates containing antigen-antibody complexes were immediately adsorbed on protein A-Sepharose (50 l) at 4°C for 4 h and analyzed by SDS-PAGE. To test the presence of the carboxyl-terminal part of the cell surface ACE immunolabeled with antibody Y1, antigen-antibody complexes from direct immunoprecipitation were heat-denatured at 100°C for 5 min and then immunoprecipitated by protein A-Sepharose coupled to antiserum 28A (3 l).
Biotinylation of Cell Surface ACEs-For cell surface biotinylation experiments, the three CHO cell lines were metabolically labeled as described above and chased for 24 h (wild-type ACE) or 4 h (ACE ⌬COOH and ACE CF ). At the end of the chase period, subsequent steps were performed at 4°C. Cell were washed 4 times with PBS containing 1 mM CaCl 2 and 1 mM MgCl 2 (PBS-CM) for 10 min. Cell surface proteins were biotinylated (1 mg/ml NHS-LC-Biotin: Pierce) for 30 min and washed 3 times with PBS-CM containing 10 mM glycine to quench biotinylation. One group of cells was solubilized, and another group was incubated at 37°C for 24 h in UltraCHO medium. Finally, solubilized cells and media were immunoprecipitated as described above. ACEs were eluted from washed antibody-protein-A Sepharose with 2% SDS in 50 mM Tris-HCl, pH 6.8. Biotinylated proteins were isolated by incubation with streptavidin-Sepharose beads (Sigma) for 30 min at 4°C. The streptavidin-Sepharose beads were washed 3 times with 1% Triton X-100, 5 mM EDTA, 250 mM NaCl in Tris-HCl 25 mM, pH 7.8. Pellets were boiled in electrophoresis sample buffer for 10 min, and proteins were resolved by SDS-PAGE.

Deglycosylation
For the deglycosylation of ACEs, cell lysates and media from the three cell lines were immunoprecipitated as described above. The immune complex with protein A-Sepharose was suspended in 20 l of 1% SDS and heated at 100°C for 3 min. Water was added (150 l), and the sample was heated at 100°C for 1 min. The supernatant was adjusted to 50 mM sodium phosphate, pH 7.5, 1% Triton X-100, 100 mM EDTA, 0.1% SDS and acidified to pH 6 with sodium acetate (1 M). The samples were divided into two aliquots; one was treated with 1 milliunit of endoglycosidase H (endo-H) (Boehringer Mannheim) overnight at 37°C, and the other was mock treated. Both samples were dried and then dissolved in sample buffer and analyzed by SDS-PAGE and autoradiography.

Western Blot Analysis
Western blot analysis was performed using a milliblot SDE (semidry transfer cell, Millipore) to transfer ACE to a polyvinylidene difluoride membrane (Immobilon P, Millipore). Immunoelectrophoretic blot analysis was carried out as described by Wei et al. (17). Three different rabbit antisera were used to characterize the ACE isoenzymes. Antisera Y1 and 28A described above. Rabbit antisera 5 and 3 were directed against synthetic peptides containing the sequence 1126 -1137 and 1214 -1227, respectively, corresponding to the two putative cleavage sites of human endothelial ACE, as described previously (19), see Fig. 1.

Subcellular Localization of Wild-type ACE Proteolysis
CHO cells stably transfected with the cDNA encoding fulllength human endothelial ACE (17) contained in the expression vector pECE were used to investigate the subcellular localization of the proteolytic cleavage.

Intracellular Transport of the Wild-type ACE
Proteins localized in the endoplasmic reticulum (ER) are characterized by their endo-H sensitivity, and mature forms localized in the post-ER compartment are identified by their endo-H resistance. The intracellular transport of newly synthesized ACE from the endoplasmic reticulum to the plasma membrane was followed by metabolic labeling. Incubation of CHO cells in [ 35 S]methionine-cysteine for 30 min at 37°C was followed by a chase period of 0.5-16 h in serum-free medium.
Endoplasmic Reticulum-The endo-H sensitive (EHS) form of ACE without or after 0.5 h of chase was identified as a 160-kDa protein and migrated as a 140-kDa protein after endo-H treatment, corresponding to the expected molecular size of nonglycosylated ACE. This 160-kDa form of ACE crossreacted with antibodies 28A and Y1, suggesting that the cleavage is not localized in the ER (Fig. 2A).
The two-step immunoprecipitation (tsIP) of EHS form of ACE showed that the protein was entirely immunoprecipitated with increasing concentrations of the 28A antibody (Fig. 3); all expressed EHS forms of ACE were immunoprecipitated with 5 l of the 28A antibody since the subsequent immunoprecipitation step using the Y1 antibody (3 l) did not detect ACE, indicating that all of the protein is in a membrane-anchored form.
Brefeldin A (BFA) induces a resorption of the Golgi into the ER, blocks intracellular transport, and increases the protein pool from the ER into an intermediate compartment (25,26). BFA was used to confirm the absence of ER proteolytic cleavage. A 4-h pulse-chase with BFA (5 g/ml) showed that ACE  (Fig. 4A). In cells, BFA-treated ACE cross-reacted with Y1 and 28A antibodies. In the medium, BFA treatment inhibited the secretion of ACE (data not shown). The 28A antibody (5 l) immunoprecipitated all BFA-treated ACE cell lysates, as shown by the two-step immunoprecipitation (Fig. 4A).
These results show that the ACE precursor forms are not carboxyl-terminally truncated, which excludes ACE proteolysis in the ER compartment.
Golgi Compartment-After a 16-h chase, 90% of initially 35 S-labeled ACE has a molecular mass of 170 kDa and was resistant to the endo-H treatment. This endo-H resistant (EHR) form of ACE cross-reacted with the two antibodies Y1 and 28A ( Fig. 2A). The results of the two-step immunoprecipitation of these EHR proteins (data not shown) were superimposable to those shown in Fig. 3, indicating that the mature form of ACE was not cleaved intracellularly.
When the cells were cultured at 20°C, the newly synthesized proteins were retained in the trans-Golgi network, reducing the access to the plasma membrane (27,28). When CHO cells were pulse-chased at 20°C for 4 h, the rate of biosynthesis and glycosylation was reduced by approximately 20 -30% compared with a pulse-chase at 37°C but was not qualitatively modified. Cellular ACE was present as two forms of 160 and 170 kDa (Fig. 4B), the tsIP show that cellular ACE at 20°C was entirely immunoprecipitated with the antibody 28A (5 l).
These results show that the proteolytic cleavage of ACE does not occur in the Golgi compartment or between the post-Golgi compartment and the plasma membrane since neither intracellular EHS or EHR forms of ACE are carboxyl-terminally truncated. The soluble ACE found in the culture medium is the only species that lacks the cytoplasmic domain as shown by its absence of reactivity with the 28A antibody.

Plasma Membrane Localization of Wild-type ACE Proteolysis
This was studied by cell surface immunolabeling and biotinylation of cell surface proteins.
Cell Surface Immunolabeling-Cell surface immunolabeling (4°C during 1 h) was performed after a pulse-chase of 16 h. It was verified that the cell surface immunolabeling conditions used did not detect intracellular proteins. The 160-kDa EHS form of ACE (Fig. 5A, lane 2) is not immunolabeled by the Y1 antibody (Fig. 5A, lane 1). Less than 50% of the mature forms of ACE (170 kDa) were immunolabeled by the Y1 antibody (Fig.  5A, compare lane 1 with lane 2). Y1 immunolabeled ACE crossreacted with the 28A antibody (Fig. 5A, lane 3). Moreover, the 28A antibody added in the medium in place of the Y1 antibody under the same conditions failed to detect any form of ACE (data not shown).
These results indicate that only mature cellular ACE proteins reached the plasma membrane and that the intracellular carboxyl terminus of the protein is not exposed to the extracellular space.
Cell Surface Biotinylation-Biotinylation of cell surface proteins (4°C during 1 h) was performed after a pulse-chase of 16 h. Biotinylated ACE was localized at the plasma membrane and cross-reacted with the two antibodies Y1 and 28A (Fig. 5B  lanes 1 and 2). It was verified that the method was selective for membrane proteins and not for cellular proteins since intracellular EHS forms or BFA-treated ACE were not biotinylated (data not shown). After biotinylation, 15% of the initially biotinylated proteins appeared in the medium after 16 h (Fig. 5B,  lane 3). The soluble biotinylated proteins did not cross-react with the 28A antibody (Fig. 5B, lane 4), indicating that a proteolytic cleavage occurred at the plasma membrane.
Using this method, the biotinylated ACE solubilization at 1 and 16 h was estimated at 1.9 Ϯ 0.3% and 13.8 Ϯ 3.1%, respectively. Furhermore, ACE solubilization quantified by the ratio of total soluble ACE in the medium after 1 and 16 h of chase over the total initially labeled cellular ACE at the chase time 0 was superimposable, estimated at 1.5 Ϯ 0.2 and 18 Ϯ 3%, respectively. Thus, the secretion of ACE in the medium closely corresponds to that of proteolytic cleavage occurring at the plasma membrane.

Regulation of the Solubilization by Phorbol Esters
The time-course of ACE secretion in the medium was studied by following ACE enzymatic activity under the influence of different pharmacological reagents after verification that these products did not modify the enzymatic activity. In absence of phorbol 12-myristate 13-acetate (PMA), ACE solubilization expressed as the ratio of enzymatic activity in the medium after 60 min over the cellular activity is measured at 2.25 Ϯ 0.05% (n ϭ 6) (Fig. 6). PMA at 10 Ϫ6 M increased ACE solubilization 2.4 Ϯ 0.3-fold (n ϭ 6). This stimulation is specific since it is not observed with the inactive analogue of PMA at 10 Ϫ6 M (data not shown). Furthermore, this stimulation is inhibited (95%) by staurosporin (10 Ϫ6 M). In contrast, the ionophore A23187 at 10 Ϫ6 M has no effect on solubilization (Fig. 6).
To test whether PMA treatment altered the proteolytic process of ACE, a pulse-chase of 16 h was used. After this period, cells were washed and then incubated for 1 h in the absence or presence of PMA (10 Ϫ6 M). The secretion of ACE was increased with PMA in the same range, 2.2 Ϯ 0.3 (n ϭ 3), and soluble proteins did not cross-react with the 28A antibody (data not shown). These results suggest that the proteolysis of ACE in CHO cells is controlled by a calcium-independent protein kinase C.

Effects of Membrane Anchoring and of the Amino-terminal Domain on the Solubilization Process
To test whether the membrane anchoring and the aminoterminal domain of the protein influence solubilization, two truncated mutants stably expressed in CHO cell lines were used. ACE ⌬COOH is a carboxyl-terminal truncated mutant in which the transmembrane domain and the cytosolic part of ACE are not translated; ACE CF is a amino-terminal truncated mutant lacking the entire amino-terminal domain (Fig. 1).

Role of Membrane Anchoring: Biosynthesis and Secretion of Metabolically Labeled ACE and ACE ⌬COOH
It was established previously that enzymatically active ACE ⌬COOH was secreted in large amounts from transfected cells (17). In order to determine whether or not the plasma membrane insertion influences the intracellular traffic and the secretion, the biosynthesis of ACE ⌬COOH was compared with that of the wild-type.
The nonglycosylated form of ACE ⌬COOH migrated as a 135-kDa protein after endo-H treatment (Fig. 2B). Intracellular traffic of ACE ⌬COOH is similar to that of the wild-type ACE. The EHS form of ACE ⌬COOH (155 kDa) is entirely converted in the EHR form (165 kDa) after 16 h of chase (Fig. 2B). 8 h were required for conversion of half of the 30-min pulse-labeled ACE and ACE ⌬COOH molecules to the EHR forms.
The secreted form of ACE appeared in the medium after a 4-h chase, whereas the secreted form of ACE ⌬COOH appeared after a 1-h chase, in the same experimental conditions (Fig. 7,  A and B). Quantification of secretion using the ratio of total soluble ACEs after a 16-h chase, over initially labeled cellular ACEs at chase time 0, revealed that 97 Ϯ 17% of the labeled glycosylated ACE ⌬COOH was secreted compared with 18.5 Ϯ 1.8% of ACE after a chase of 16 h when 90% of cellular proteins are mature. ACE ⌬COOH is not found at the plasma membrane by cell surface immunolabeling and biotinylation (data not shown). Western blot analysis showed that both antisera 5 and 3 crossreact with soluble forms of ACE ⌬COOH (Fig. 8), suggesting that ACE ⌬COOH is secreted into the medium without modification of its carboxyl-terminal end.
Thus, the proteolysis of wild-type ACE depends on the presence of its membrane-anchor and possibly on a plasma membrane-associated secretase.

Role of the N-Terminal Domain: Biosynthesis and Secretion of Metabolically Labeled ACE CF
The possible effect of the amino-terminal domain on ACE solubilization was first investigated in different clones expressing ACE CF , where a 10-fold increase in the solubilization process was observed.
Since, the rate of biosynthesis of ACE CF was faster than that of ACE, the conditions of the metabolic labeling were adapted. Cells were labeled for 15 min with [ 35 S]methionine-cysteine and chased in serum-free medium for 0.5, 1, 2, and 4 h (Fig.  7C). After 4 h of chase, 90% of the initially labeled ACE CF was converted to the EHR form of the protein. Under these conditions, the biosynthesis and secretion of ACE CF could be compared with that of a 16-h chase of ACE, where 90% of the initially labeled ACE was converted to the EHR form of ACE.
The nonglycosylated form of the ACE CF migrated as a 75-kDa protein after endo-H treatment (Fig. 2C). ACE CF was initially synthesized as an EHS protein (95 kDa), which was  1 and 2). Cells were maintained at 37°C for 16 h, and then the medium (m 16 ) was removed and immunoprecipitated with Y1 or 28A (lanes 3 and 4). Biotinylated proteins were revealed with streptavidin after immunoprecipitations. ¢, cell lysates; m, medium. entirely converted to an EHR form (105 kDa) after 4 h of chase (Fig. 2C). 2 h were required for conversion of half of the 15-min pulse-labeled initial molecules to the EHR form.
The secreted ACE CF appeared in the medium (Fig. 7C), after a 30-min chase. After 4 h of chase, when 90% of initially labeled proteins were mature, 35% of ACE CF was secreted as a carboxyl-terminal truncated protein with a decrease of 5 kDa in molecular mass (100 versus 105 kDa). Western blot analysis (Fig. 8) showed that both antisera 5 and 3 cross-react with the soluble form of ACE CF , suggesting that the cleavage site of ACE CF is localized between Arg-1227 and Val-1228 (see Fig. 1). This cleavage did not occur intracellularly, as shown by the two-step immunoprecipitation (Fig. 9A).
Like ACE, ACE CF is expressed at the cell plasma membrane (Fig. 9B, lane 1) and was studied by biotinylation (Fig. 9B,  lanes 2-6). After 4 h of chase and 1 h of incubation at 37°C in a serum-free medium, 17.7 Ϯ 9.2% of initially ACE CF biotinylated proteins appeared in the medium (Fig. 9B, lanes 3 and 4). This protein did not cross-react with the 28A antibody.
Solubilization of biotinylated ACE CF in the medium 1 and 4 h after the biotinylation was 17.7 Ϯ 9.2 and 35.1 Ϯ 2%, respectively. The solubilization, quantified by the ratio of soluble ACE CF in the medium after 1 and 4 h of chase over the initially labeled cellular ACE CF at the chase time 0, was similar, estimated at 19.4 Ϯ 7% and 33.03 Ϯ 2.7%, respectively. Thus, as for wild-type ACE, the proteolytic cleavage of ACE CF takes place at the plasma membrane and occurs at the carboxylterminal end.
ACE CF solubilization is also stimulated by phorbol esters since PMA (10 Ϫ6 M) increases the solubilization 2.7 Ϯ 0.8-fold, which is inhibited by staurosporin (10 Ϫ6 M). The ionophore A23187 has no effect on ACE CF solubilization (data not shown).
In order to verify that PMA stimulates the proteolysis of membrane-bound ACE CF , cells were pulse-labeled, chased 4 h and biotinylated as described above. After 1 h in the presence of PMA (10 Ϫ6 M), the solubilization of biotinylated ACE CF was increased (Fig. 9B, lane 5) compared with control (Fig. 9, lane  3). Moreover, soluble biotinylated ACE CF , after PMA treatment, did not cross-react with the 28A antibody (Fig. 9, lane 6).
These data show that basal ACE solubilization is clearly impaired by the presence of the amino-terminal domain since deletion of this domain increases the rate of solubilization of ACE CF without altering the PMA-induced solubilization process. In each case, a tsIP was performed as described in the legend to Fig. 3 with 28A (5 l) and Y1 (3 l). B, CHO-ACE CF cells were labeled and chased for 4 h and then immunolabeled with Y1 (10 l) for 1 h at 4°C. Cell lysates were directly immunoadsorbed on protein A-Sepharose (lane 1). CHO-ACE CF cells were pulse-chased for 4 h and biotinylated as indicated in the legend to Fig. 5 (¢ 0 , lane 2). Biotinylated cells were maintained at 37°C for 1 h, in the absence (Ϫ) or presence (ϩ) of PMA (10 Ϫ6 M). Then media (m 1 ) were removed, and biotinylated proteins were revealed, after immunoprecipitation with Y1 (3 l) or 28A (3 l) (lanes 3-6) by streptavidin. DISCUSSION The general phenomenon of the solubilization of carboxylterminal membrane-anchored proteins involves a carboxyl-terminal proteolytic cleavage that takes place near or at the plasma membrane. This is the case for the secretion of the carboxyl-terminally truncated secreted APP (8,29), or TGF-␣ (2). The mechanism of ACE secretion in CHO cells involves a proteolytic cleavage of the carboxyl-terminal part of the protein between Arg-1137 and Leu-1138 (19). The present study shows that the processing of membrane-anchored ACE operates primarily at the cell surface. Indeed, all intracellular ACE retained in the ER or in Golgi compartment before the membrane, cross-reacts with the carboxyl-terminal antibody, and the two-step immunoprecipitation method failed to detect any soluble intracellular forms of ACE even when intracellular blockers (brefeldin A or lowering temperature to 20°C treatments) were used. After 16 h of chase, 90% of initially radiolabeled ACE was present as endo-H resistant form, N-glycosylated, and present at the plasma membrane. Immunolabeling of mature radiolabeled ACE shows that the amino-terminal part of ACE is exposed in the extracellular space, whereas the carboxyl-terminal end remained intracellular, excluding the possibility that the carboxyl terminus forms a hairpin loop and is exposed in the extracellular space. Biotinylation of mature radiolabeled ACE present at the cell surface clearly shows that the proteolytic cleavage of ACE takes place at the plasma membrane since the biotinylated ACE secreted in the medium is carboxyl-terminally truncated. The solubilization rate estimated for biotinylated proteins was the same as that determined for total cellular ACE. Therefore, the cleavage at the plasma membrane appears to be the predominant pathway for the generation of soluble ACE. However, we cannot exclude the possibility that some cleavage takes place in a specialized vesicular compartment near the plasma membrane via a degradation/recycling pathway, as described for APP (30,31). If this processing occurs, it would involve another sequence than the consensus endocytic sequence SXYQRL, which is not present in the cytosolic part of ACE.
Experiments using an ACE mutant with its transmembrane ␣-helix and cytosolic parts deleted (ACE ⌬COOH ) show a faster rate of secretion than ACE, without a difference in glycosylation or pattern of cellular traffic. This mutant is not expressed at the plasma membrane and is not cleaved intracellularly between Arg-1137 and Leu-1138, showing that membrane-anchoring of the protein and its cytosolic region are necessary for the proteolytic cleavage to take place at the plasma membrane, as for other proteins of this class.
Very little is known about the molecular mechanisms that may influence the secretion rate of C domain anchored proteins. One general finding, however, seems to be the increase in the solubilization rate induced by phorbol esters. The solubilization of the membrane-bound form of APP, pro-TGF-␣, and ACE in CHO cells, transfected with the cDNA of APP, pro-TGF-␣ (32), or ACE (present study) is low, being 5, 1, and 2%, respectively, after 1 h of chase. The proteolytic cleavage of pro-TGF-␣ (33)(34)(35) and APP (36 -38) is enhanced in both cases by phorbol esters or ionophores. A series of experiments were designed to evaluate this observation with respect to the ACE mutants. Proteolytic cleavage of recombinant human ACE in CHO cells is stimulated 2.4-fold by phorbol esters but is not influenced by ionophores. These results are in agreement with those published by Sen and co-workers (21) on the phorbol ester-induced stimulation of rabbit testicular ACE secretion. In the present study, the degree of this stimulation does not depend on the amino-terminal part of ACE, since phorbol ester stimulation is quantitatively similar to that observed for ACE CF . In addition, the secretion of ACE ⌬COOH is not regulated by phorbol esters and ionophores, suggesting that the PMA effect depends on the membrane-anchoring of the protein. Experiments using biotinylated mature radiolabeled ACE CF show that phorbol esters stimulate the proteolysis of the membranebound protein, suggesting that the implicated secretase(s) is probably also membrane-associated, a hypothesis consistent with the membrane-associated secretase(s) proposed for ACE (39), APP (8,40), or pro-TGF-␣ (41).
The increased solubilization of ACE (endothelial or testicular) by phorbol esters (2-fold over basal conditions) is 2.5 times less than the solubilization observed for pro-TGF-␣ or APP in CHO cells (32). Another difference is that the secretion of pro-TGF-␣ and APP in CHO cells is stimulated by ionophores involving a calcium-dependent protein kinase C, whereas a calcium-independent protein kinase C appears to be implicated in the proteolytic processing of rabbit testicular ACE (21) and endothelial ACE (present study).
Each class I ectoprotein could possess its own specific determinants to control the plasma membrane proteolysis (2,6). One mechanism involved is the primary structure of the cytoplasmic tail, since a specific role for the carboxyl-terminal valine has been clearly demonstrated for pro-TGF-␣ (42) but is not involved in the solubilization of the tumor necrosis factor receptor (43). The role of the cytoplasmic tail of ACE has not yet been described. A second mechanism involved is the extracellular ectodomain of the protein. Deletion of 285 amino acids from the amino-terminal part of APP increases the solubilization of secreted APP (8). Construction of a chimeric protein of APP, where amino acids 1-647 of APP ectodomain were replaced by a human secreted alkaline phosphatase derivative, (44) shows that the amino-terminal part of APP could inhibit the solubilization of secreted APP.
The role of the amino-terminal part of APP (8) in its solubilization as well as the increased rate of solubilization of testis ACE (21) compared with endothelial ACE (present study) prompted us to study the specific effect of the ACE aminoterminal domain in the solubilizing process. The proteolytic release of ACE is increased 10-fold by the deletion of the aminoterminal domain of ACE, reaching 35% (Fig. 10) after 4 h of chase. Several hypotheses could be envisaged to explain this increase of solubilization. 1) The intracellular traffic of ACE CF could be faster than that of ACE, resulting in an artificial increase in solubilization rate expressed as the ratio of soluble ACE CF in the medium over cellular ACE CF . However, ACE CF is expressed at the plasma membrane, and biotinylated mature ACE CF is secreted in the medium after 1 h of chase 10-fold faster than ACE, under the same experimental conditions, i.e. when 90% of all mature proteins have reached the plasma membrane. Moreover, ACE CF , like ACE, is primarily cleaved at the cell surface and not intracellularly. Thus, the increased solubilization observed with ACE CF is not a consequence of accelerated biosynthesis or another cellular proteolytic process. 2) Through structural interactions with ACE itself or the secretase(s), the amino-terminal domain of ACE could act as a "conformational inhibitor" of the proteolytic process of membrane-anchored ACE. The amino-terminal domain could act on extracellular ACE conformation, near the anchoring domain; when the amino-terminal domain is present, the region near the membrane anchor may have a conformation distinct from that when the N domain is absent. Indeed, the amino-terminal amino acids of the native protein cannot be reached by immunolabeling with the Clo antiserum, directed against the first 18 amino acids of the human endothelial sequence (data not shown). Accordingly, the amino-terminal part of the protein is more or less masked by the rest of the protein and/or in close-contact with the membrane. It is possible that a part of the amino-terminal domain interacts with the "stalk" of ACE, the region situated between the end of the globular part of the C domain and the hydrophobic membrane-spanning domain. Another possibility is that the amino-terminal domain of ACE affects the conformation of the membrane-bound secretase(s) itself, via protein-protein interactions.
Two distinct cleavage sites are involved in the proteolysis of ACE and ACE CF . ACE is cleaved between Arg-1137 and Leu-1138, whereas ACE CF is probably cleaved between Arg-1227 and Val-1228. Therefore, a model can be proposed where ACE and ACE CF are cleaved by one or two secretase(s) (Fig. 10), Arg-1137 and Arg-1227 being alternative cleavage sites. Interaction of the amino-terminal domain with the "stalk" and secretase(s) could induce a preferential cleavage in Arg-1137 with a low turnover. Alternatively, the deletion of the amino-terminal domain would involve a conformational change of ACE stalk and/or of the secretase(s), which cleaves then after Arg-1227 with a more rapid turnover.
It is possible to envisage a general model for the processing of ACE. ACE could be cleaved at the plasma membrane by a secretase(s), localized like ACE to the plasma membrane. The amino-terminal domain would act as a conformational inhibitor of the proteolytic solubilization via structural interactions between the ACE stalk and the secretase(s) itself. Under these conditions, endothelial ACE would be slowly solubilized in plasma leading to the presence of most ACE activity at the endothelial cell level. Since protein-protein interactions vary in the plasma membrane and in the extracellular space, it is possible to conceive a "conformational" change in the cleavage site of ACE, leading to an increased ACE solubilization, under certain biological circumstances. FIG. 10. Subcellular localization of ACE and ACE CF . Schematic role of the amino-terminal domain of ACE in the solubilization. All intracellular (IC) species cross-reacts with 28A (28Aᮍ). 90% of initially labeled proteins reach the membrane (pm) after 16 h of chase for ACE and 4 h of chase for ACE CF . Under these comparable conditions, solubilization of ACE from the plasma membrane is 10-fold lower than ACE CF . In the medium (extracellular; EC) ACE and ACE CF do not cross-react with 28A (28Aᮎ). The inset shows the proposed "conformational inhibitor" role of the ACE amino-terminal domain. On the left, the amino-terminal domain interacts with ACE stalk and the secretase(s) leading to a cleavage between Arg-1137 and Leu-1138. On the right, deletion of the amino-terminal domain modifies interactions of the secretase(s) with the ACE stalk, which cleaves between Arg-1227 and Val-1228.