Superoxide regulation of endothelin-converting enzyme.

Reactive oxygen species (ROS) act as signaling molecules in the cardiovascular system, regulating cellular proliferation and migration. However, an excess of ROS can damage cells and alter endothelial cell function. We hypothesized that endogenous mechanisms protect the vasculature from excess levels of ROS. We now show that superoxide can inhibit endothelin-converting enzyme activity (ECE) and decrease endothelin-1 synthesis. Superoxide inhibits ECE but hydrogen peroxide and nitric oxide do not. Superoxide inhibits ECE by ejecting zinc from the enzyme, and the addition of exogenous zinc restores enzymatic activity. Superoxide may inhibit other zinc metalloproteinases by a similar mechanism and may thus play an important role in regulating the biology of blood vessels.

Endothelins are polypeptides that play important signaling roles in the cardiovascular system (1,2). Endothelins have many physiological activities, stimulating smooth muscle cell contraction and proliferation, inducing the release of neuropeptides, and regulating bone resorption (reviewed in Ref. 3). Endothelins may play a role in a variety of diseases, including hypertension, vasospasm, congestive heart failure, and renal failure (3).
The three isoforms of endothelin are expressed in endothelial cells and a variety of other cells as well (1,2). Endothelin-1 (ET-1) 1 is synthesized from a large precursor, prepro-ET-1 (4). A signal peptidase cleaves prepro-ET-1 into pro-ET-1, which in turn is cleaved by a furin-like enzyme into big ET-1. Endothelin-converting enzyme (ECE) then cleaves big ET-1 into ET-1, a 21-amino acid residue polypeptide. Cleavage of big ET-1 into ET-1 is essential for its activity, because big ET-1 is biologically inactive.
The endothelin-converting enzyme (ECE) was first purified from rat lung and was subsequently discovered to be expressed in endothelial cells of many organs (5,6). ECE is also found in adrenal cells, pancreatic cells, the testis, and brain. ECE-1 and ECE-2 are encoded by separate genes and share 59% homology. Human ECE-1 has three isoforms, ECE-1a, ECE-1b, and ECE-1c (7)(8)(9)(10). All three ECE-1 isoforms are encoded by a single gene; they share a common carboxyl-terminal portion but three alternate promoters regulate expression of unique amino-terminal portions.
The ECEs are members of a larger zinc metalloproteinase family that includes neutral endopeptidase. ECE is a type II membrane glycoprotein. ECE shares with other zinc metalloproteinases a zinc-binding HEXXH motif. An arginine residue, Arg 129 , plays a role in binding of big ET. ECE-1 is a homodimer, with a single disulfide bond between the Cys 412 residue of each monomer, linking two 130-kDa ECE-1 monomers together.
The precise mechanisms that regulate ECE-1 expression, localization, and activity are partially understood. Differences in the response elements of the three alternate promoters of ECE-1 may account for differences in the expression of the isoforms. For example, ECE-1c is expressed predominantly in human smooth muscle, whereas human ECE-1a is expressed in endothelial cells. Differences in the subcellular localization of ECE-1 isoforms may be due to differences in amino-terminal targeting motifs. Human ECE-1a and ECE-1c are found mostly in the plasma membrane, and human ECE-1b is localized in an intracellular compartment. Phosphoramidon, an inhibitor of ECE, induces internalization of ECE. However, little is known about the post-translational regulation of ECE activity.
Reactive oxygen species (ROS) such as superoxide or hydrogen peroxide are produced in the cardiovascular system during a variety of physiological and pathophysiological conditions (11)(12)(13)(14). Every cell type in blood vessels can produce ROS, including endothelial cells, fibroblasts, and smooth muscle cells; and neutrophils and macrophages that infiltrate into vessels can also produce ROS (15)(16)(17)(18)(19)(20). Low levels of ROS can serve as second messengers in various signal transduction pathways. For example, hydrogen peroxide mediates signal transduction of platelet-derived growth factor (21). Superoxide mediates p21 ras regulation of the cell cycle (22). ROS activate a variety of other redox-sensitive intracellular cellular targets as well (reviewed in Refs. 11,13,14,[23][24][25][26]. Thus ROS have been shown to play important roles in the regulation of cell growth and migration.
However, high levels of ROS can be cytotoxic. For example, vessel injury can lead to the adherence of neutrophils to the endothelium, the activation of the neutrophil NADPH oxidase, and the generation of extracellular superoxide, which can damage endothelial cells (reviewed in Refs. 11,13,[26][27][28]. Furthermore, ischemia and reoxygenation can lead to the production of high levels of ROS inside endothelial and smooth muscle cells, which can also damage cells. ROS may also play a pathophysiological role in hypertension. Superoxide levels are increased in vessels of spontaneously hypertensive rats compared with wild-type rat vessels (29). Griendling and co-workers (18,30) have shown that ROS play an important role in angiotensin signaling. Angiotensin binding to its receptors leads to intracellular production of superoxide; superoxide can combine with nitric oxide (NO), decreasing NO levels, and in theory leads to vasospasm. Thus excessive levels of ROS can be potentially harmful to the vasculature.
Although superoxide may be detrimental to the vasculature, endogenous counter-regulatory mechanisms may exist that inhibit the harmful effects of superoxide upon the vessel wall. We hypothesized that ROS inhibit the activity of ECE. A reduction in ET levels might then lead to an increase in local blood flow, which would increase perfusion of damaged tissues. Thus ROS inhibition of ECE activity might counteract some of the potentially harmful effects of ROS. To test this hypothesis, we prepared endothelial cell membranes that contain ECE, and we examined the effect of various ROS upon ECE activity.
Cell Culture-Bovine aortic endothelial cells (BAEC) were isolated using methods previously described (31). BAEC were grown on gelatincoated plates at 37°C in a humidified atmosphere of 95% air and 5% CO 2 . Individual clones were established and subcloned to obtain pure cell populations (32). Cells were fed every 2 days with RPMI 1640 medium supplemented with 15% calf serum, 100 units/ml penicillin, and 100 g/ml streptomycin. Studies were performed on confluent monolayers at passages 2-7, made quiescent by serum deprivation. Toxicity was evaluated by the trypan blue dye exclusion method and measurement of lactate dehydrogenase activity in the incubation media.
Immunoblotting-Confluent monolayers of BAEC were grown in 10-cm dishes. Cells were homogenized in 1 ml of homogenization buffer (20 mmol/liter Tris-HCl, pH 7.5, 5 mmol/liter MgCl 2 , 0.1 mmol/liter phenylmethylsulfonyl fluoride, 20 mol/liter leupeptin, 20 mol/liter aprotinin) (33). The homogenates were sonicated and then centrifuged at 100,000 ϫ g for 45 min. The membranes were washed three times in homogenization buffer. The protein concentration in each supernatant was determined according to standard methods (BCA protein assay, Pierce). 100 g of protein extracts were incubated with or without 0.1 mM xanthine plus 1 milliunit/ml xanthine oxidase (XXO) for 2 h at 37°C. When catalase or superoxide dismutase (SOD) were used, they were added 15 min before XXO at 80 and 100 units/ml, respectively. Following incubation, sample buffer was added in the presence or no ␤-mercaptoethanol to study ECE-1 protein in reducing or non-reducing conditions, respectively. Proteins were separated on 6% SDS-PAGE (100 g of protein/lane) and transferred onto a nitrocellulose membrane. The membrane was blocked in 5% nonfat dry milk in phosphatebuffered saline for 1 h at 22°C, incubated for 1.5 h at 22°C with 10 g/ml of the monoclonal antibody against bovine ECE-1 (33), a generous gift from Dr. Kohei Shimada (Biological Research Laboratories, Sankyo Co., Ltd., Tokyo, Japan), washed in TTBS (20 mmol/liter Tris-HCl, 0.9% NaCl, 0.05% Tween 20), and incubated with 200-fold diluted peroxidase-conjugated rabbit anti-mouse IgG at 22°C for 1 h. The immunoreactive bands were visualized with the SuperSignal detection system (Pierce) after 30 s of exposure to film.
ECE Activity Assay-Membrane proteins were extracted from confluent layers of BAEC as described above, and 30 g of this homogenate were incubated in the presence or absence of different radical donors for 1 h. Bovine Big ET-1 (100 ng) was then added, and the mixture was incubated for 4 h at 37°C in 250 l of a reaction mixture containing 50 mM Tris-HCl buffer, pH 7 (34,35). The reaction was stopped by adding 600 l of ice-cold ethanol. After centrifugation at 10,000 ϫ g for 10 min, the resulting supernatant was lyophilized. The dried residues were reconstituted with assay buffer, and the production of ET-1 in each sample was measured by an ELISA for ET-1 using a 96-well microtiter plate reader (Amersham Pharmacia Biotech) (35). To generate a standard curve for ET-1, serial dilutions of ET-1 (Amersham Pharmacia Biotech) ranging from 1 to 16 fmol (2.49 -39.9 pg) was added to assay buffer and placed in the ELISA wells. A cubic spline curve was fitted to the standards and unknown values interpolated from the standard curves automatically.
To measure the effect of superoxide upon the substrate big ET-1, 100 ng of big ET-1 in 50 mM Tris-HCl, pH 7.0, was incubated with buffer or with 0.04 mM xanthine and 0.4 milliunits/ml xanthine oxidase for 2 h at 37°C. SOD (100 units/ml) was then added to scavenge further superoxide anions, and the mixture was incubated for 10 min at 37°C. BAEC membrane proteins (30 g) were then added to pretreated big ET-1 and incubated for 4 h at 37°C, and the production of ET-1 was measured as above.
To measure the effect of superoxide upon the product ET-1, ET-1 was incubated with buffer, 0.04 mM xanthine, and 0.4 milliunits/ml xanthine oxidase for 1 h at 37°C. SOD (100 units/ml) was added to some mixtures and incubated for 4 h at 37°C. An ELISA was used to measure ET-1.
Quantitation of Superoxide Production-The amount of superoxide generated by xanthine and xanthine oxidase was measured by the SOD inhibitable rate of cytochrome c reduction (36,37). In brief, 75 M cytochrome c (Sigma) was mixed with buffer or 20 units/ml SOD and then mixed with xanthine (0.04 mM) plus xanthine oxidase (0.4 milliunits/ml). The amount of reduced cytochrome c was measured as A 550 nm in aliquots of reaction mixture analyzed at various time points and compared with a standard curve; the amount of superoxide generated by xanthine oxidase with xanthine was calculated from the difference in amounts of cytochrome c reduced in the absence and the presence of SOD.
Zinc Assay-Zinc measurements were performed by Instrumental Neutron Activation Analysis at the Massachusetts Institute of Technology's Research Reactor, MITR-II. The protein samples were placed into small acid-washed polyethylene vials using an additional 0.2 ml of buffer solution to ensure complete transfer. These vials and two zinc reference samples (NIST Standard Reference Material 2710 Montana Soil, certified zinc concentration 6952 Ϯ 91 mg/kg) were placed in a polyethylene "rabbit" to enable them to travel in the reactor's pneumatic irradiation facility. The samples and standards were irradiated together for 12 h at a neutron flux of 8 ϫ 10 12 Newtons/cm 2 s, were transferred to irradiated vials using two additions of 0.5 ml of unirradiated buffer, and were then allowed to decay for approximately 4 weeks. The presence of the activation product 65 Zn (half-life ϭ 243.8 days, gamma energy ϭ 1115.5 keV) was measured in the samples and standards using four High Purity Germanium detectors coupled to a dedicated personal computer (all hardware and software from Canberra Industries Inc., Meriden, CT). The zinc levels in the proteins were calculated based on the 65 Zn activities in the protein samples and the reference materials, the reference sample masses, and the certified zinc reference concentrations. Measurement uncertainties were calculated using the nuclear analytical counting uncertainties and the uncertainty in the reference zinc concentrations.
Statistical Analysis-In every case, the data shown are the mean Ϯ S.E. In the endothelin-1 ELISA, each individual data was the mean of three wells. As the number of data was always under 10, and normality test are inadequate for this number of data, non-parametric statistics for more than two groups (Friedman test) were performed. A p Ͻ 0.05 was considered statistically significant.

RESULTS
Xanthine Oxidase with Xanthine Inhibits ECE Activity-To explore the role of radicals in regulating ECE, we incubated membranes isolated from bovine aortic endothelial cells (BAEC) with various substances that release radical molecules. Big ET-1 was then added to the mixture, and the amount of ET-1 released from the mixture was measured by an ELISA. Incubation of the membranes with xanthine and xanthine oxidase inhibits the generation of ET-1 (Table I). In contrast, incubation of the membrane with hydrogen peroxide, or with glucose oxidase and glucose that generates hydrogen peroxide, or with several nitric oxide donors does not affect ECE activity. (To confirm that ECE activity is being assayed, the ECE inhibitor phosphoramidon was added, which reduces endothelin production as expected.) Xanthine oxidase in combination with xanthine inhibits ECE activity in a dose-dependent manner (Fig. 1).
To exclude the possibility that superoxide inhibits ECE processing of big ET-1 by modifying the substrate big ET-1, we incubated big ET-1 with buffer or with xanthine and xanthine oxidase, then added SOD and continued the incubation, and finally added BAEC membrane proteins. We then measured the amount of ET-1 produced. BAEC membrane proteins process big ET-1 whether or not big ET-1 is previously exposed to superoxide donors (Table II). These data suggest that the target of superoxide is ECE and not big ET-1.
To confirm that superoxide inhibits ECE processing without modifying the substrate big ET-1, we incubated membranes isolated from BAEC with buffer or with xanthine and xanthine oxidase for 1 h, then added SOD and incubated the mixture for 10 min, and then added the substrate big ET-1 and incubated the mixture for 4 h. Pretreatment with xanthine oxidase and xanthine inhibits the ability of BAEC membranes to process big ET-1, even when radicals are scavenged during the processing of big ET-1 (Table III).
To exclude the possibility that superoxide somehow modifies ET-1, decreasing the sensitivity of the ET-1 ELISA, we next incubated ET-1 with buffer, xanthine oxidase and xanthine, or xanthine oxidase and xanthine followed by SOD. Treated and untreated ET-1 was then measured by ELISA. Superoxide treatment of ET-1 does not affect the sensitivity of the ELISA (Table IV). These data suggest that superoxide does not interfere with the ELISA by modifying ET-1.
To measure the amount of superoxide generated by xanthine oxidase, we incubated cytochrome c and xanthine oxidase and xanthine with or without SOD, and we measured the A 550 nm at various times. This assay shows that 0.04 M xanthine and 0.4 milliunits/ml xanthine oxidase produce 920 nmol/liter of superoxide over 1 h, 1070 nmol/liter over 2 h, and 2040 nmol/liter over 4 h.
Superoxide Inhibits ECE Activity-Xanthine oxidase synthesizes superoxide (O 2 . ) using xanthine as a substrate. To determine whether or not O 2 . mediates xanthine oxidase inhibition of ECE, SOD was added to mixtures of BAEC membranes, xanthine oxidase, and xanthine. ECE activity was then assayed as above. As before, xanthine oxidase with xanthine inhibits ECE activity. SOD reverses this inhibition; catalase does not (Fig.  2). These data suggest that superoxide inhibits ECE. However, hydrogen peroxide does not, since catalase has no effect. Furthermore, SOD or catalase alone do not have an effect on ECE activity (data not shown). Superoxide Does Not Affect ECE Homodimerization-We next explored the mechanism by which superoxide inhibits ECE. Superoxide might inhibit ECE by interfering with ECE homodimerization. ECE homodimerization is maintained by a single disulfide bond between each ECE monomer. Oxidants could indirectly disrupt this disulfide bond, converting active ECE homodimers into inactive monomers; or oxidants could inhibit the formation of this disulfide bond, decreasing the amount of ECE homodimers and increasing the amount of monomers. To test this hypothesis, we treated BAEC membranes with or without xanthine oxidase and xanthine and then fractionated the membranes on denaturing and non-denaturing PAGE. Treatment of membranes with xanthine oxidase and xanthine did not affect the mobility of ECE. The relative mobility of ECE on denaturing PAGE is approximately  1. Dose-dependent inhibition of ECE by xanthine oxidase with xanthine. BAEC membranes were incubated with increasing doses of xanthine with xanthine oxidase, and then big ET-1 was added to the mixture, and ET-1 production was assayed by an ELISA. (n ϭ 3; * for p Ͻ 0.05 compared with xanthine 0 mM and xanthine oxidase 0 milliunits/ml.)

TABLE II
Superoxide does not affect big ET-1 Big ET-1 was pretreated or not with xanthine oxidase and xanthine. Membrane proteins from BAEC were then added to big ET-1, and an ELISA was used to measure the amount of ET-1 generated (n ϭ 6 Ϯ S.E.). mU indicates milliunits.

TABLE IV Superoxide does not affect detection of ET-1
ET-1 was treated with buffer, xanthine oxidase and xanthine, or xanthine oxidase and xanthine followed by SOD. An ELISA was used to measure the amount of ET-1 (n ϭ 3 Ϯ S.E.).  (Fig. 3). Thus superoxide does not inhibit ECE by converting active ECE homodimers into inactive monomers. Superoxide Ejects Zinc from ECE-We next hypothesized that superoxide inhibits ECE by removing Zn 2ϩ from ECE. ECE is a metalloproteinase that contains Zn 2ϩ as a cofactor. Zn 2ϩ is bound to ECE through an HEXXH motif. We first confirmed that Zn 2ϩ is necessary for ECE activity by determining the effect of EDTA upon ECE activity. BAEC membranes were incubated with and without EDTA, and then ECE activity was measured. EDTA decreases ECE activity, and adding Zn 2ϩ to membranes incubated with Zn 2ϩ restores ECE activity (Fig.  4A). These data confirm the importance of Zn 2ϩ to ECE activity, as others (38 -40) have shown. Next we treated BAEC membranes with xanthine and xanthine oxidase. Xanthine with xanthine oxidase also decreases ECE activity, and Zn 2ϩ restores ECE activity (Fig. 4B). These data suggest that O 2 .
removes zinc from ECE. To prove that O 2 . removes Zn 2ϩ from ECE, we next measured the amount of Zn 2ϩ in ECE by instrumental neutron activation analysis. ECE was immunoprecipitated from BAEC membranes and exposed to media alone or to xanthine oxidase with xanthine. A constant amount of ECE protein was then irradiated with neutrons and allowed to decay for 4 weeks, and the amount of 61 Zn was measured. Native ECE contains 61 Zn, but ECE exposed to superoxide contains 4-fold less 61 Zn, consistent with background levels (Table V). Thus superoxide decreases the amount of Zn 2ϩ in ECE.

DISCUSSION
The major finding of our study is that superoxide inhibits ECE by ejecting zinc from the enzyme. This inhibition is due specifically to superoxide; hydrogen peroxide and nitric oxide do not affect ECE activity. Furthermore, this inhibition is reversible, since the addition of zinc restores ECE activity.
The mechanism by which superoxide ejects zinc from ECE is unclear. However, it is unlikely that superoxide destroys the structure of ECE, since the addition of zinc restores ECE activity. One possible mechanism by which superoxide could remove zinc from ECE is by reducing the residues that coordinate zinc binding. Superoxide can convert histidine into 2-oxo-histidine, but this reaction is not reversible (41)(42)(43)(44). Another possibility is that superoxide reduces Zn 2ϩ to Zn ϩ , which might interact less strongly with the HEXXH motif of ECE and disassociate from ECE. This mechanism would explain why the addition of exogenous zinc restores ECE activity, since Zn 2ϩ can interact with the HEXXH motif. This reduction and ejection of zinc from ECE by superoxide is similar to the effect of superoxide upon iron during superoxide inactivation of ironsulfur-containing (de)hydratases such as cis-aconitase (45,46).
There are over 200 zinc metalloproteinases, and they regulate a variety of physiological and pathophysiological processes, including development, reproduction, tissue remodeling or destruction, inflammation, carcinogenesis, and fibrosis (47,48). Superoxide may also regulate other zinc metalloproteinases in addition to ECE as well. Others have shown that superoxide can increase matrix metalloproteinase activity. For example, superoxide can increase activity of collagenase, MMP-2, and MMP-9 (19, 49 -52). Furthermore, superoxide dismutase inhibits matrix degradation (53,54). In contrast, our data suggest

TABLE V
Effect of superoxide donors upon zinc content of ECE ECE was immunoprecipitated (IP) from BAEC exposed or not to xanthine oxidase with xanthine, and the amount of zinc (Zn) in 300 g of immunoprecipitated protein was analyzed by Instrumental Neutron Activation Analysis. Samples were prepared from BAEC (ECE), BAEC treated with xanthine oxidase with xanthine (ECE ϩ XXO), immunoprecipitants from BAEC using a control antibody immunoprecipitation with control antibody), or immunoprecipitation buffer and antibody and xanthine oxidase with xanthine but lacking BAEC membranes. ( that superoxide inhibits ECE. Perhaps superoxide activates one set of metalloproteinases and inhibits another because of the difference in zinc-binding motifs. ECE belongs to the zinc metalloproteinase clan MA with the zinc-binding motif HEXXH and the third zinc ligand Glu. In contrast, the matrix metalloproteinase clan MB contains the zinc-binding motif HEXXHXXGXXH, and the third zinc ligand is His. What are the potential physiological consequences of superoxide inhibition of ECE? ROS might act as part of a negative feedback loop during inflammation of the vasculature. For example, neutrophils can bind to endothelial cells as a result of hypoxia or cytokine stimulation. The neutrophil NADPH oxidase can then secrete high levels of superoxide. Superoxide can damage endothelial cells, leading to endothelial cell dysfunction, decreasing endothelial NO production, which can lead to local vasospasm. However, if superoxide also decreases local synthesis of ET-1, then vasospasm might be diminished. Thus ROS which is normally produced during inflammation might actually serve to counteract vasospasm by decreasing ET synthesis. Since ET may play a role in a variety of other pathophysiological conditions, these data also suggest a novel target for inhibitors of ECE.