Replacement of the Transmembrane Anchor in Angiotensin I-converting Enzyme (ACE) with a Glycosylphosphatidylinositol Tail Affects Activation of the B2 Bradykinin Receptor by ACE Inhibitors*

To investigate further the relationship of angiotensin I-converting enzyme (ACE) inhibitors to activation of the B2 bradykinin (BK) receptor, we transfected Chinese hamster ovary cells to stably express the human receptor and either wild-type ACE (WT-ACE), an ACE construct with most of the cytosolic portion deleted (Cyt-del-ACE), or ACE with a glycosylphosphatidylinositol (GPI) anchor replacing the transmembrane and cytosolic domains (GPI-ACE). BK or its ACE-resistant analogue were the agonists. All activities (arachidonic acid release and calcium mobilization) were blocked by the B2 antagonist HOE 140. B2 was desensitized by repeated administration of BK but resensitized to agonist by ACE inhibitors in the cells expressing both B2 and either WT-ACE or Cyt-del-ACE. In GPI-ACE expressing cells, the B2 receptor was still activated by the agonists, but ACE inhibitors did not resensitize. Pretreatment with filipin returned the sensitivity to inhibitors. In immunocytochemistry, GPI-ACE showed patchy, uneven distribution on the plasma membrane that was restored by filipin. Thus, ACE inhibitors were inactive as long as GPI-ACE was sequestered in cholesterol-rich membrane domains. WT-ACE and B2 receptor in Chinese hamster ovary cells co-immunoprecipitated with antibody to receptor, suggesting an interaction on the cell membrane. ACE inhibitors augment BK effects on receptors indirectly only when enzyme and receptor molecules are sterically close, possibly forming a heterodimer.

Subsequently, it became obvious that inhibitors of ACE affect the metabolism of both peptides (6). The successful clinical applications of ACE inhibitors have gone far beyond controlling elevated blood pressure (7,8), but questions remain regarding which of the beneficial effects are due to inhibiting angiotensin II activation and which are caused by blocking the enzymatic breakdown of bradykinin (BK) or kallidin. The very extensive clinical applications of ACE inhibitors, not only in treating hypertension but also in treating cardiac conditions, (e.g. congestive heart failure or after myocardial infarction), and in diabetic nephropathies (9 -11), have kept attention focused on this issue. In laboratory experiments and in some clinical studies (12,13), many effects of ACE inhibitors were abolished by the BK B 2 receptor blocker HOE 140. Although it was assumed that these effects were due to inhibiting the inactivation of BK, early bioassays already indicated that substances that did not prolong the half-life of BK still potentiated its actions on the isolated guinea pig ileum (14). Experiments on isolated guinea pig atria demonstrated that ACE inhibitors can resensitize the heart tissue desensitized by a B 2 receptor agonist (15). More precisely formulated, an unresolved issue was whether the potentiation of BK effects by ACE inhibitors (12,16,17) was caused only by blocking its enzymatic hydrolysis. Investigations using cultured cells transfected with cDNA of the human B 2 receptor provided evidence that ACE inhibitors do not act on the B 2 receptor directly but enhance BK effects and resensitize the receptor to BK only if ACE is also present on the plasma membranes of cells (18,19). The enhancement of the activity of the peptide and the reactivation of the receptor (thus reversing tachyphylaxis) was the same when, instead of BK, its ACEresistant peptide analogues were used. Results similar to those in transfected cells were obtained with endothelial cells that constitutively express the enzyme and the receptor (18,19). This resensitization of the desensitized receptor by ACE inhibitors occurs with the agonist still present in the medium and without adding more peptide, making it difficult to attribute these results to only blocking the breakdown of BK. This effect is also observed with some endogenous peptides that act as inhibitors or slow substrates of ACE (19). Thus, ACE inhibitors and some endogenous peptides augment BK effects on the receptor by inducing cross-talk between ACE and the B 2 receptor (19).
We reported elsewhere the effects on B 2 receptor activation of a recombinant, mutant ACE that combined part of the Nterminal end with the C-domain of ACE but deleted most of the N-domain. 2 We continued these studies as described below to explore the importance of the cytosolic and transmembrane domains of ACE in the activation of the B 2 receptor by ACE inhibitors and, as a consequence, its steric relation to the receptor. Here, we report the results with two ACE constructs: one containing the transmembrane domain with most of its cytosolic portion deleted, and a chimeric ACE construct that results in its expression on plasma membranes with a glycosylphosphatidylinositol (GPI) anchor. These experiments indicate that the cytosolic and transmembrane domains are not required for activation of the B 2 receptor by ACE inhibitors, but ACE has to be in the immediate vicinity of the receptor for the activation to occur, possibly due to heterodimer formation.
Construction of ACE with Cytosolic Domain Deleted (Cyt-del-ACE)-ACE cDNA fragment 3743-3886 was obtained by polymerase chain reaction with two primers, ACED1 (5Ј-AGACAGCGGCCGCGTCAG) and ACED2 (5Ј-GCTCTAGAAGAGGCTGCGGTGGCGGA). After digestion with NotI and XbaI, the fragment encoding base pairs 3742-3886 of the ACE sequence, followed by a codon for Phe and a stop codon in the reading frame, was ligated with NotI-and XbaI-digested wild-type ACE-pcDNA3. The polymerase chain reaction-amplified part of the construct was sequenced to ensure that no spurious mutations were introduced.
Construction of Chimeric ACE with a GPI Anchor Signal (GPI-ACE)-A DNA fragment encoding the C-terminal 24 residues of carboxypeptidase M (20, 21) with a NotI cleavage site on the 5Ј side and an XbaI site at the 3Ј end was obtained by using carboxypeptidase M cDNA as template and two primers, MGPI-1 (5ЈACAGCGGCCGCCCAGAC-CACTCAGCTGCA) and MGPI-2 (5Ј-GCTCTAGAGGTGGTGATGT-GGGTTGA), in polymerase chain reaction amplification. The amplified product encodes residues 403-426 of carboxypeptidase M, containing the complete GPI anchor signal, the putative anchor attachment site (Ser 406 ), and the 3 residues N-terminal to this site. It fits into the correct reading frame of human ACE through the connecting NotI site. An ACE/GPI-pcDNA3 construct was obtained by ligating the ϳ100-base pair polymerase chain reaction fragment digested with NotI and XbaI with a gel-purified 9-kilobase pair fragment of ACE-pcDNA3 digested with NotI and XbaI. The ACE/GPI-pcDNA3 construct was sequenced to prove that it contains the coding region for the extracellular domain of ACE (base pairs 1-3748, encoding residues 1-1242, counting the signal peptide, yielding a mature sequence of 1213 residues from ACE) fol-lowed by the coding sequence of the GPI anchor signal of carboxypeptidase M in the same reading frame. The C-terminal sequence of ACE/ GPI is as follows: . . . . LPDSGRPDHS*AATKPSLFLFLVSLLHIFFK (underlining indicates the carboxypeptidase M sequence; S* indicates the putative GPI anchor attachment site). The plasmid DNA of ACE/GPI-pcDNA3 was purified through a Qiagen minicolumn and used to transiently transfect HEK293 cells with Superfect reagent (Qiagen) according to the manufacturer's protocol.
To establish cell lines with stable expression, CHO cells were transfected with human mutant or wild-type ACE cDNA in a pcDNA3 expression vector (carrying the neomycin resistance gene) using the Superfect method with serum-free Ham's F-12 medium without antibiotic, as described (19). Clones were isolated by cloning rings and grown to confluence (19).
Screening of Clones for ACE Activity-Individual clones were evaluated both for cell-associated and released ACE activity. Cells were incubated in fresh serum-free medium; after 24 h, medium was collected, and the cells were washed and lysed in 3 ml of 8 mM Chaps in phosphate-buffered saline, pH 7.4, or 3 ml of 13 mM Chaps in 50 mM Hepes, pH 7.4, with 100 mM NaCl. Both the supernatants and the lysates were centrifuged (for 15 min at 900 ϫ g) and their aliquots were assayed for ACE activity in 50 mM Tris-maleate, pH 7.4, with 150 mM NaCl, with Hip-His-Leu or Z-Phe-His-Leu (1 mM) as described (22,23). The reaction was terminated by addition of 0.28 M NaOH, and the assay mixture was then incubated with 100 l of 20 mg/ml o-phthaldialdehyde in methanol for 10 min at room temperature to form a fluorescent derivative of the His-Leu product. This reaction was stopped by the addition of 200 l of 3 M HCl. Fluorescence of the samples was determined at wavelengths of 365 nm excitation and 500 nm emission.
Transfection with Human B 2 Receptor cDNA-Clones with highest levels of ACE expression were transfected with a human B 2 BK receptor cDNA cloned in pcDNA3 together with pCEP4 DNA at a ratio of 10:1, using the Superfect transfection method (19). Following transfection, cells were selected in Ham's F-12 medium containing 0.5 mg/ml hygromycin B (pCEP4 contains the hygromycin B resistance gene). After selection, different clones were harvested and propagated using cloning rings (19). Cells expressing both human B 2 receptor and wild-type ACE were designated CHO/AB cells.
Radioligand Binding on Selected Clones-Clones with the highest expression of B 2 receptors were selected by [ 3 H]BK saturation binding on whole cell monolayers. Equilibrium binding of 0.05-20 nM [ 3 H]BK with or without 10 M unlabeled BK was done in Ham's F-12 cell culture medium for 1 h at 37°C (18). Bound radioactivity was separated from excess [ 3 H]BK by washing. The cells were then solubilized in 0.5 ml of a solution containing 0.1 ml of NaOH, 0.1 ml of NaHCO 3 , and 1% SDS and transferred to 20-ml liquid scintillation vials to be counted. Clones with high expression of B 2 receptors on the cell surface were chosen and used further.
Enzymatic Release of the Membrane Anchor of ACE-HEK293 cells transfected with ACE/GPI-pcDNA3 or CHO cells stably transfected with full-length somatic ACE were harvested by scraping into ice-cold phosphate-buffered saline and were washed by centrifugation at 4°C. For studies of the release of ACE, cells were resuspended in 10 mM Hepes, pH 7.4, containing 150 mM NaCl, 0.25 M sucrose, and protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 10 M trans-epoxysuccinyl-L-leucylamido-(4-guanidino)butane, 100 M leupeptin, 1 M pepstatin A, and 1 g/ml aprotinin) with or without 30 milliunits/ml phosphatidylinositol-specific phospholipase C (PI-PLC) or with 0.01% trypsin treated with L-1-tosylamido-2-phenylethyl chloromethyl ketone (Sigma) in the same buffer without protease inhibitors. Cells were incubated for 2 h at 37°C and then centrifuged at 14,000 ϫ g for 30 min. The supernatant was saved, and the pellet was resuspended in the same buffer containing 0.5% Chaps detergent by mixing for 15 min at room temperature. ACE activity was measured in both the pellet and supernatant.
To determine whether enzymatic release of ACE removed the hydrophobic anchor, Triton X-114 partitioning was used (24). In these experiments, cells were lysed by sonication, and cell homogenates were centrifuged at 10,000 ϫ g for 10 min and then at 100,000 ϫ g for 60 min to obtain a final membrane pellet. To solubilize membrane-bound ACE, the membrane fraction was mixed with 0.1% Chaps detergent overnight at 4°C and then centrifuged at 50,000 ϫ g for 1 h to pellet insoluble material. To release the hydrophobic membrane anchor, solubilized membrane fractions were treated at 37°C for 2 h with phosphatebuffered saline alone (control) or 30 milliunits/ml PI-PLC in the presence of the same protease inhibitors as above or with 0.01% trypsin without protease inhibitors. Triton X-114 partitioning was carried out by diluting samples 10-fold in phosphate-buffered saline followed by mixing with a 1% final concentration of Triton X-114 at 4°C for 1 h. Samples were warmed to 37°C and centrifuged for 1 min in a microcentrifuge to separate the phases. The detergent pellet and supernatant were separated for determination of enzyme activity.

Measurement of Changes in [Ca 2ϩ ] i and [ 3 H]AA-Free cytosolic calcium [Ca 2ϩ
] i was measured using a microspectrofluorometer (PTI Deltascan, Princeton, NJ) or Attofluor Ratiovision with fura-2/AM reagent. Cells were grown to confluence on glass coverslips and then incubated with 2-5 M fura-2/AM for 1 h at 37°C (19). Cells were washed with buffer, incubated for an additional 15 min, and then mounted in a Sykes-Moore chamber (Bellco, Vineland, NJ) at room temperature. Cellular fluorescence at 510 nm was measured following excitation at wavelengths of 340 or 380 nm. Changes in [Ca 2ϩ ] i were detected by the ratio of intensities at 340 and 380 nm, and the amount of free calcium in the cytosol was calculated. In some experiments, cytosolic calcium concentrations were recorded in a single cell or in 100 cells simultaneously using an Attofluor Ratiovision microspectrofluorometer (25). [ 3 H]AA release upon receptor stimulation was measured as before (18), following incorporation of [ 3 H]AA into membrane phospholipids.
Desensitization and Resensitization of the B 2 Receptor-After desensitization of the receptor by initial exposure of cells to a kinin (15,18), the restoration of sensitivity to the agonist (resensitization) was measured either by [ 3 H]AA release or by mobilization of [Ca 2ϩ ] i . For example, monolayers of transfected CHO cells loaded with [ 3 H]AA were stimulated with 1 M BK or its ACE-resistant analogue for 30 min. Then, without removal of the agonist, cells were exposed to either 5 nM enalaprilat or 5 nM ramiprilat without adding more kinin to show resensitization of the receptor, or as a control, to demonstrate desensitization, a second dose of kinin was added instead of the ACE inhibitor.
The amount of [ 3 H]AA released was determined by taking AA released during the first 30 min as baseline and normalizing to the amount released during the reactivation by adding buffer alone for 5 min.
[Ca 2ϩ ] i mobilization was measured in cells first exposed to 10 -100 nM BK. After the initial [Ca 2ϩ ] i response, without removal of BK from the medium, CHO cells were exposed either to BK again to confirm desensitization or to enalaprilat or another agent to resensitize the receptor.
Filipin Treatment-CHO cells expressing GPI-ACE and B 2 receptor or B 2 receptor alone were treated in monolayers with Ham's F-12 medium (ϩ10% fetal bovine serum) containing 10 nM filipin for 30 min at 37°C.
Co-Immunoprecipitation of B 2 Receptors and ACE-Confluent monolayers of CHO/AB cells were treated with buffer alone, 1 M enalaprilat, 1 M BK or 1 M BK ϩ 1 M enalaprilat for 30 min at 37°C. Following this, cellular monolayers were solubilized with 10 mM Chaps for 1 h at 4°C. Samples were then centrifuged at 100,000 ϫ g for 15 min, and soluble supernatants were saved. Rabbit antisera to the human B 2 receptor were added at 1:1000 (v/v) dilution and samples were incubated overnight with shaking at 4°C. Insoluble protein A was then added to each sample to 10%, and the incubation was continued for 2 h at 4°C. Beads with immune complexes were precipitated by centrifugation at 1,000 ϫ g for 15 min. The presence of B 2 receptors was verified by [ 3 H]BK binding. Precipitates were eluted with Laemmli buffer (26) and subjected to 10% SDS-polyacrylamide gel electrophoresis. Proteins were electrotransferred to a nitrocellulose membrane, which was then subjected to immunoblotting with rabbit polyclonal anti-ACE antibod-ies for 2 h at 1:1000 (v/v) dilution. Goat anti-rabbit antibodies coupled to alkaline phosphatase were used as secondary antibodies in the same way, and the bands were visualized with an alkaline phosphatase detection kit (Sigma). The densities of ACE bands were quantitated by scanning densitometry.
Statistics-The data in the figures and text are expressed as means Ϯ S.E. of n observations (n ϭ Ն3). When some previously reported experiments were repeated with the same results, they were done once or twice in duplicate and not pursued further. Intracellular free calcium levels ([Ca 2ϩ ] i ) are reported in nM. Statistical evaluation was performed by one-way analysis of variance for matched values. Values of p Ͻ 0.05 were considered statistically significant.

Desensitization and Resensitization of the BK B 2 Receptor-To
show that the agonist BK uniformly desensitizes the B 2 receptor in transfected cells, we stimulated B 2 receptors in CHO/AB cells with 100 nM BK to elevate [Ca 2ϩ ] i and monitored it simultaneously in 100 cells. Fig. 1A illustrates the individual cellular responses to the first application of BK and that the second dose of the peptide was inactive. Fig. 1B shows the computer calculated mean values from 100 cells. The sensitivity to BK present in the medium was restored by adding the ACE inhibitor ramiprilat (1 M) (Fig. 1). Thus, the agonist BK indeed desensitized the receptor, and an ACE inhibitor resensitized it. In these and in subsequent experiments, 1 M HOE 140 (12) was always used as a control (not shown), and it blocked all the primary effects of BK and the resensitization of the receptor to BK as reported (18,19).
Effect of Changes in ACE Structure-To investigate the im- portance of the cytoplasmic and transmembrane domains of the ACE molecule (27)(28)(29) in augmenting BK effects on B 2 receptors, we created two different ACE constructs (Fig. 2). The first is Cyt-del-ACE; the second one (GPI-ACE) is a chimeric ACE in which the sequence encoding the transmembrane and cytoplasmic domains was deleted from the cDNA and replaced with a sequence from carboxypeptidase M encoding the C-terminal 24 residues, comprising the signal for attachment of a GPI-membrane anchor (20). These mutants were stably transfected and expressed in CHO cells together with human BK B 2 receptor.
Cyt-del-ACE-The mutant was obtained to test whether or not the potentiation of B 2 receptor by ACE inhibitors, and thus the cross-talk between the enzyme and the receptor, would be mediated by an intracellular mechanism involving the intracellular portion of ACE. For this purpose, a human ACE mutant, Cyt-del-ACE, was used. This construct lacks most of the intracellular domain of WT-ACE (27, 28); 18 carboxyl-terminal amino acids were deleted, and an extra Phe was added to the C terminus as a consequence of the construction method (Fig. 2). The truncation removes three out of five serine residues, which may serve as potential phosphorylation sites within the cytoplasmic domain (30). The extracellular and transmembrane domains of Cyt-del-ACE were unchanged.
Seven clones were tested for stable expression of B 2 receptor and ACE in CHO cells. The selected clone expressed 82,000 B 2 receptors per cell and 13,000 ACE molecules per cell.
The effect of enalaprilat on B 2 receptor number was tested on Cyt-del-ACE-expressing cells at two concentrations of the inhibitor. Enalaprilat (5 nM) increased B 2 receptor number insignificantly 1.4 Ϯ 0.2-fold (n ϭ 3), whereas 1 M enalaprilat raised the number 2.9 Ϯ 0.3-fold (n ϭ 3; p Ͻ 0.05). The 5 nM enalaprilat inhibited 72 Ϯ 4% (n ϭ 3) of ACE activity measured on the plasma membrane of intact cells expressing WT-ACE using Z-Phe-His-Leu substrate. In a dose-response curve, previously reported, (19), the ACE inhibitor started to increase B 2 receptor number at 0.1 M concentration of the inhibitor and reached a maximum at about 1 M.
We also investigated whether enalaprilat can resensitize the desensitized B 2 receptor through interaction with Cyt-del-ACE (Fig. 3). The B 2 receptor was desensitized by the initial appli-cation of 1 M BK, as shown by lack of a response to an additional dose of 1 M BK (Fig. 3). Enalaprilat (5 nM) resensitized B 2 receptor-mediated [ 3 H]AA release, resulting in a 6.5 Ϯ 0.8-fold (n ϭ 3; p Ͻ 0.005) increase over buffer alone, and 1 M enalaprilat potentiated the response 8.4 Ϯ 1.2-fold (n ϭ 3; p Ͻ 0.005).
Enalaprilat (1 M) also resensitized the B 2 receptor to BK present in the medium, when a rise in [Ca 2ϩ ] i of 2.7 Ϯ 0.6-fold was recorded (n ϭ 6; p Ͻ 0.005; Fig. 4). In the absence of BK, enalaprilat was inactive (not shown).
These results, which agree with the data obtained with fullsized WT-ACE (18,19), suggest that ACE inhibitors can potentiate BK effects on B 2 receptors acting through Cyt-del-ACE, which lacks most of the intracellular domain. This is taken as an indication that for an interaction between ACE and the B 2 receptor, either the transmembrane anchor of ACE, the extracellular portion, or both, but not most of the intracellular portion, are important.
GPI Anchored ACE-To investigate the role of the transmembrane anchor, we employed a chimeric ACE cDNA encoding the extracellular domain of ACE (residues 1-1213 of the mature protein) followed by the C-terminal 24 residues of carboxypeptidase M (20, 21) (Fig. 2). Thus, the final 64 residues of ACE containing the transmembrane and cytoplasmic domains were replaced with the 24-residue GPI anchor signal sequence of the carboxypeptidase. When this construct was expressed in transiently transfected HEK293 cells, high ACE activity was present in membrane-bound form, as determined after subcellular fractionation and immunofluorescent staining (not shown).
To prove that GPI-ACE was indeed anchored by the GPI tail, HEK293 cells transfected with WT-ACE or GPI-ACE were treated either with buffer alone as control or with 30 milliunits/ml PI-PLC or 0.01% trypsin. As shown in Table I, both PI-PLC and trypsin released the GPI-ACE from the cells into the supernatant, whereas buffer alone did not. In contrast, WT-ACE was not released by PI-PLC or buffer but was cleaved from the membrane with 0.01% trypsin (Table I). To prove that cleavage by PI-PLC or trypsin removed the hydrophobic membrane anchor from GPI-ACE, membrane fractions from transfected HEK293 cells that were first solubilized with Chaps detergent were treated with either buffer alone, 30 milliunits/ml PI-PLC, or 0.01% trypsin and then subjected to Triton X-114 partitioning (24). Membrane fractions from CHO cells stably transfected with WT-ACE were used as controls. In these experiments, only 15% of the GPI-ACE partitioned into

FIG. 2. Construction of GPI-ACE and Cyt-del-ACE.
The domain structure of human somatic ACE (WT-ACE) (28) and the two mutant constructs used in this study are shown schematically. The C-terminal sequence of each construct is given below the diagram. Cyt-del-ACE lacks the C-terminal 18 residues of the cytosolic tail and contains an extra Phe, at the C terminus, that is not found in the native sequence. In GPI-ACE, the C-terminal 64 residues, including the transmembrane and cytosolic domains, were replaced with the 24 residue GPI anchor signal sequence (shown in italics) of human carboxypeptidase M (CPM). For further details, see text.
the aqueous phase in the buffer control, indicating that the majority of the enzyme contained the hydrophobic GPI anchor, whereas after PI-PLC treatment, 76% and after trypsin treatment, 60% of the enzyme partitioned into the aqueous phase (mean values from two separate experiments). In contrast, 68% of WT-ACE treated with trypsin partitioned into the aqueous phase, but only 26 or 15% was found in the aqueous phase after PI-PLC or buffer treatment. Taken together, the above data show that the C-terminal hydrophobic region of carboxypeptidase M functions as a signal for addition of the GPI tail to anchor the extracellular domain of ACE to the plasma membrane, which is normally anchored by a transmembrane peptide.
CHO cells were stably transfected with the GPI-ACE construct and selected by Geneticin (G-418) resistance. Ten clones were tested for ACE activity using 1 mM Z-Phe-His-Leu as the substrate, and a clone with the highest expression was chosen (21-29 nmol/min/mg protein; 50,000 ACE molecules/cell). These cells were subsequently co-transfected with a cDNA encoding the B 2 receptor and a pCEP4 vector and selected using hygromycin B resistance. Six clones were tested by  (n ϭ 8). These results are similar to those obtained with CHO/AB cells expressing WT-ACE (18,19). However, in GPI-ACE-containing cells, 1 M ramiprilat failed to resensitize the B 2 receptors desensitized by the agonist (see below). This is in contrast to results routinely obtained with WT-ACE (18,19) or as reported above with Cyt-del-ACE. Thus, BK analogue in the medium in the presence of an ACE inhibitor did not release [ 3 H]AA or elevate [Ca 2ϩ ] i after the receptor was desensitized by the first dose of the agonist (Figs. 5 and 6).
These experiments indicate that GPI-ACE is unable to interact with the BK B 2 receptor, which could be explained by two possible mechanisms. The first one is that the transmembrane domain of ACE directly interacts with the BK B 2 receptor. Alternatively, the steric relationship of GPI-ACE to B 2 might be altered. GPI-anchored proteins are known to partition into sphingolipid and cholesterol-rich microdomains on the plasma membrane (32)(33)(34)(35)(36)(37). GPI-ACE may be sequestered in these rafts or similar lipid structures and therefore unable to interact with the BK B 2 receptor. In addition, GPI anchors confer higher lateral mobility compared with transmembrane domains (34).
To distinguish between these possibilities, we attempted to restore the steric relationship between the B 2 receptor and GPI-ACE by employing a cholesterol-depleting agent, filipin (38), which is known to disperse GPI-anchored proteins from the cholesterol rich domains in the plasma membrane.
The CHO cells expressing GPI-ACE and B 2 receptor were pretreated in a monolayer with 10 nM filipin for 30 min at 37°C. Viability of the cells was preserved after this treatment, and they responded to stimulation by 10 nM BK analogue with a typical [Ca 2ϩ ] i mobilization response (Fig. 6). In control experiments without added ramiprilat, cells did not respond to a second dose of BK (n ϭ 4; not shown). Fig. 6B, a recording from a single CHO cell stimulated with an ACE-resistant BK analogue, shows that filipin restored the ability of ramiprilat to resensitize the receptor. In eight additional experiments done with approximately 100 individual cells each, filipin restored the activity of ramiprilat (1 M) to resensitize the receptor in 84 Ϯ 5% of the cells to the BK analogue (10 nM) present in the medium. The mean [Ca 2ϩ ] i response to the BK analogue after filipin and resensitization by ramiprilat was 120 Ϯ 8% of the original response. As in all other experiments, the ACE inhibitor did not enhance the [Ca 2ϩ ] i level in the absence of an agonist (not shown).
We also tested whether ramiprilat resensitized the desensitized B 2 receptor to release [ 3 H]AA. In four experiments, in CHO cells expressing GPI-ACE and B 2 receptors, 1 M ramiprilat failed to resensitize the receptor to stimulate the release of [ 3 H]AA after desensitization by 1 M BK analogue (Fig. 5). However, if the cells were pretreated with 10 nM filipin for 30 min at 37°C, 1 M ramiprilat resensitized [ 3 H]AA release by 1 M BK analogue present in the medium, increasing it over basal release by 5.7 Ϯ 0.8-fold (n ϭ 4; Fig. 5).
Immunocytochemistry-The localization of ACE on the plasma membrane of CHO cells stably expressing either WT-ACE or GPI-ACE was studied by immunostaining and confocal microscopy. WT-ACE was distributed uniformly on the plasma membrane of CHO cells (Fig. 7A). In contrast, in the cells expressing GPI-ACE, immunostaining with antiserum to human ACE showed a patchy distribution and aggregated clusters on the plasma membrane (Fig. 7C). When cells expressing WT-ACE were treated with 10 nM filipin for 30 min prior to fixation and immunostaining, filipin did not alter the membrane-associated staining pattern of WT-ACE (Fig. 7B). However, pretreatment with filipin dispersed the patchy immunostaining pattern of GPI-ACE on the cells (Fig. 7D) to reveal a more uniform distribution of GPI-ACE. These data are consistent with the hypothesis that filipin, by binding to cholesterol, disrupts GPI-anchored protein clusters accumulated on the membrane in areas rich in cholesterol (32)(33)(34)(35)(36)(37).
Co-Immunoprecipitation of B 2 Receptors and ACE-The above data are in agreement with a physical interaction between ACE and the B 2 receptor requiring the extracellular domain of ACE. To further test the steric relationship between B 2 receptor and ACE and their possible interaction, we immunoprecipitated the two proteins with anti-B 2 receptor antibodies. CHO/AB cells in monolayers were exposed to buffer alone, enalaprilat (1 M), BK (1 M), or BK and enalaprilat for 30 min. (Fig. 8). Cells were then solubilized with detergent, the lysates immunoprecipitated with anti-B 2 receptor antibodies (see under "Experimental Procedures"), and then the precipitate was analyzed by Western blotting with rabbit polyclonal anti-ACE antibodies. ACE was present in all precipitates as detected with the antibodies, showing that the enzyme and receptor formed a complex (Fig. 8). When immunoprecipitation was performed with normal rabbit serum or with antiserum to the

TABLE I Release of ACE from HEK293 cells transfected with the wild type (WT-ACE) or chimeric enzyme containing the C-terminal tail of carboxypeptidase M (GPI-ACE)
Transfected cells were scraped, washed, and then incubated either with buffer, with PI-PLC and protease inhibitors, or with trypsin, for 2 h at 37°C. Cells were centrifuged at 14,000 ϫ g for 30 min, and ACE activity in the supernatant and pellets was measured with Z-Phe-His-Leu. Results shown are mean values from two or three separate transfections. Total ACE activity among the different transfections did not differ by more than 2-fold and ranged from 40 -80-fold above the background activity measured in nontransfected cells.  (18,19), no immunoreactivity to anti-ACE antibodies was detected (not shown). In control experiments, Western blotting of cell lysates without immunoprecipitation revealed a dark band of the expected molecular weight in cells expressing WT-ACE, whereas cells expressing only the B 2 receptor were negative (Fig. 8). Thus, WT-ACE and B 2 receptor, when transfected into CHO cells, form a complex, very likely a heterodimer, owing to the very close steric relationship. This gross complex formation was sustained even in the presence of agonist and ACE inhibitor.

DISCUSSION
Human ACE is a single chain protein that can be roughly divided into three parts: 1) a large extracellular portion containing two homologous active sites on the N and C domains; 2) a hydrophobic domain near the C terminus that functions as the transmembrane anchor; and 3) a short cytoplasmic domain (27,28) (Fig. 2). Our previous studies have shown that ACE inhibitors potentiate BK and reactivate the B 2 receptor to respond to BK by a mechanism other than protecting BK from degradation (15,18,19,23,25,39). This implies that ACE and the BK receptor interact and that the cross-talk can be altered by binding of inhibitors, slowly cleaved substrates, or active site-directed antibodies of ACE (19).
We report here that the cytoplasmic and transmembrane domains of ACE are not required for the resensitization and reactivation of the BK B 2 receptor by ACE inhibitors but that the receptor and ACE have to be in close proximity, possibly forming a heterodimer mediated through the extracellular domains of ACE and the receptor. Two different, active ACE inhibitors, enalaprilat and ramiprilat, gave the same results, showing that their effect was group-specific (i.e. their ability to bind ACE) and not related to any unique aspect of their structures (12).
Receptors can form dimers via interactions between their cytosolic domains (40). To determine whether the cytoplasmic domain of ACE was important in our studies, we created Cytdel-ACE, where the major part of the cytosolic domain was deleted, including three out of five of the possible serine phosphorylation sites (30). This deletion had no effect on the potentiation of BK and resensitization of its receptor by an ACE inhibitor.
To establish whether the transmembrane domain was critical for this interaction, we expressed a chimeric ACE cDNA containing the two extracellular active site domains and the C-terminal GPI anchor signal sequence from carboxypeptidase M (20,21,24) to replace the cytoplasmic and transmembrane domains of ACE. This resulted in a chimeric ACE that was still membrane-associated via a GPI-anchor but completely lacked transmembrane and cytoplasmic domains. In these cells, BK and an ACE-resistant BK analogue (31) activated the B 2 receptor as before, but in contrast to cells expressing WT-ACE and the B 2 receptor, ACE inhibitor did not resensitize the receptor to kinin. This may mean that the transmembrane domain of ACE is crucial for its interaction with the B 2 receptor, implying that the cross-talk takes place within the membrane. A more likely explanation is based on the following. It has been established that GPI-anchored proteins are concentrated in cholesterol-and sphingolipid-rich microdomains, also termed lipid rafts, on the cell membrane that can exclude many transmembrane proteins (32)(33)(34)(35)(36)(37). GPI-ACE would be sequestered in such lipid rafts. Furthermore, GPI-anchors in general enhance the lateral mobility of proteins on cell membrane (34). This should also apply to GPI-ACE, if it is compared with WT-ACE anchored by transmembrane and cytosolic portions to the cell membrane (27)(28)(29). Consequently, the distance of GPI-ACE from the B 2 receptor would increase on the cell surface and deny an enzyme receptor interaction. When the cells expressing both GPI-ACE and B 2 receptor were depleted of cholesterol by filipin, which resulted in the dispersion of GPI- anchored proteins, it led to the full restoration of the response to ACE inhibitors. This indicates that the lack of reactivation with GPI-ACE is due to sequestration in lipid rafts and not necessarily to the increased lateral movement. The absence of resensitization of the B 2 receptor in cells expressing GPI-ACE and restoration of the response by filipin also indicate that close physical proximity is required for cross-talk between B 2 receptors and ACE.
The conclusions reached after using filipin in a large number of cells are supported by immunocytochemistry. As shown before by immunohistochemistry (41,42), when antibody to ACE was used, WT-ACE was uniformly distributed on the plasma membrane of cells (Fig. 7). In contrast, GPI-ACE had a patchy distribution, indicative of aggregated clusters of the enzyme. After employing filipin, GPI-ACE was dispersed on the membrane of the cells and distributed more uniformly, and the aggregate was dissolved. Thus, filipin changed the distribution of GPI-ACE on the membrane to resemble more that of WT-ACE.
G-protein linked receptors, including the B 2 BK receptor, can form homodimers and even heterodimers with other receptors or different proteins (43)(44)(45). The dimer formation has been associated with receptor desensitization and endocytosis (46). The results of the present studies, showing co-immunoprecipitation of ACE and the BK receptor, indicate an interaction between these two proteins on the plasma membrane. This suggests how ACE inhibitors may affect BK receptor activity via ACE. ACE inhibitors, by stabilizing the ACE-B 2 receptor heterodimer, could inhibit B 2 receptor homodimer formation. However, we found no apparent difference in co-immunoprecipitation in the presence or absence of an ACE inhibitor. Thus, the reactivation by ACE inhibitor may reflect a conformational change in ACE transduced to the B 2 receptor in a preformed heterodimeric complex, which results in abolishing desensitization and resensitizing the receptor to the agonist. Resensitization of the receptor by ACE inhibitors can enhance the activity of the agonist up to 8-fold (Refs. 18 and 19 and see under "Results").
It was reported that ACE inhibitors reduce the sequestration of the B 2 receptor into caveolae, a location that was also found to be enriched in ACE (47). This implied that the inhibition of receptor sequestration by an ACE inhibitor would be responsible for its ability to resensitize the cellular receptor to a BKstimulated increased [Ca 2ϩ ] i level (47). These studies might have offered an interpretation for receptor reactivation. For example, ACE inhibitors may interfere with the caveolar sequestration of the ACE and B 2 receptor heterodimer complex. However, other findings argue against this interpretation.
First, although ACE has been reported to be in caveolae, caveolin-rich fractions can be contaminated with other membrane compartments, depending on the method of isolation. In fact, in the studies reporting a procedure for the isolation of highly purified caveolar fractions, ACE was found to be associated only with plasma membrane fractions and excluded from caveolar fractions, resulting in its use as a marker for membrane fractions that do not contain caveolae (49,50). Furthermore, the resensitization of the receptor to BK by ACE inhibitors is immediate, happening in seconds (18,19). This rapid response does not correlate with the physical sequestration of the receptor, which can be on the order of 5-15 min (47). All of this is taken as an indication that the effect of ACE inhibitors on caveolar sequestration is unlikely to be responsible for the reactivation of the response to BK. That could be a secondary phenomenon that occurs after the sequence of events that mediate the initial desensitization and resensitization.
Our previous experiments (19) indicate that the pathway mediating the action of resensitized B 2 receptors differs from the pathway involved in the immediate response to BK. Consequently, the receptor resensitized to the same peptide can initiate signaling through alternate pathways. 2 Judging from experiments based on AA release and [Ca 2ϩ ] i elevation, both G i -and G q -linked B 2 receptors (48,51) are involved in signaling in the reactivation process. Similarly to WT-ACE (18,19), in cells expressing the receptor and Cyt-del-ACE, the release of AA by BK was stimulated at a much lower concentration of ACE inhibitor (5 nM) than the elevation of [Ca 2ϩ ] i (0.1-1 M). In addition, in the present experiments as well, ACE inhibitors increased the BK receptor sites only when used at the higher concentration (18).
The release of prostaglandins and NO mediate the important cardiovascular effects of BK and kallidin (16,17), and this can be a consequence of the clinical application of ACE inhibitors (7)(8)(9)(10)(11). Prostaglandin release is attributed to activation of phospholipase A 2 by G␣ i -coupled B 2 receptors, followed by AA liberation, whereas phospholipase C is activated by a G␣ q complex (51). The latter reaction leads to elevated [Ca 2ϩ ] i level and subsequently to enhanced NO synthesis and release. ACE inhibitors are more potent in augmenting BK effects and resensitizing B 2 receptors if they are coupled to G␣ i than if they are coupled to G␣ q (18,19). At the end of the transduction chain, this can lead to enhanced release of prostaglandins over NO liberation in the presence of ACE inhibitors in some cells and tissues (52).
In conclusion, ACE and the B 2 receptor expressed on cell membranes are sterically closely associated, probably forming a heterodimer, and ACE inhibitors and other agents with af- finity for the active center likely alter the heterodimer interaction to promote a conformation of the B 2 receptor that can more efficiently induce signal transduction.