Degradation of the Alzheimer's Amyloid β Peptide by Endothelin-converting Enzyme*

Deposition of β-amyloid (Aβ) peptides in the brain is an early and invariant feature of all forms of Alzheimer's disease. As with any secreted protein, the extracellular concentration of Aβ is determined not only by its production but also by its catabolism. A major focus of Alzheimer's research has been the elucidation of the mechanisms responsible for the generation of Aβ. Much less, however, is known about the mechanisms responsible for Aβ removal in the brain. In this report, we describe the identification of endothelin-converting enzyme-1 (ECE-1) as a novel Aβ-degrading enzyme. We show that treatment of endogenous ECE-expressing cell lines with the metalloprotease inhibitor phosphoramidon causes a 2–3-fold elevation in extracellular Aβ concentration that appears to be due to inhibition of intracellular Aβ degradation. Furthermore, we show that overexpression of ECE-1 in Chinese hamster ovary cells, which lack endogenous ECE activity, reduces extracellular Aβ concentration by up to 90% and that this effect is completely reversed by treatment of the cells with phosphoramidon. Finally, we show that recombinant soluble ECE-1 is capable of hydrolyzing synthetic Aβ40 and Aβ42 in vitro at multiple sites.

Alzheimer's disease (AD) 1 is the most common cause of dementia in the elderly and is characterized pathologically by the accumulation of ␤-amyloid peptides (A␤) in the brain in the form of senile plaques. A␤ is normally produced from the ␤-amyloid precursor protein (␤APP) through the combined proteolytic actions of ␤and ␥-secretase and is then secreted into the extracellular milieu (1,2). The degree of A␤ accumulation is dependent not only on its production but also on the mechanisms responsible for its removal. While considerable effort has been directed at elucidating the enzymes and pathways con-tributing to the production of A␤, much less is known regarding A␤ catabolism.
A␤ catabolism is likely to involve proteases at multiple sites, both intracellular and extracellular. Proteases acting at the site of A␤ generation and/or within the secretory pathway may degrade the peptide intracellularly, thus limiting the amount of the peptide available for secretion. The concentration of secreted A␤ may be further regulated by direct degradation by extracellular proteases and by receptor-mediated endocytosis or phagocytosis followed by lysosomal degradation. Catabolism of A␤ peptides at each of these steps would limit the accumulation of extracellular A␤, and disruption of this catabolism may be a risk factor for AD. Additionally, the identification of enzymes that degrade A␤ intracellularly and extracellularly may lead to development of novel therapeutics aimed at reducing A␤ concentration by enhancing its removal.
Recent reports suggest a role for both insulin-degrading enzyme and neprilysin (NEP) in the degradation of extracellular A␤ (3)(4)(5)(6)(7)(8)(9)(10). Matrix metalloproteinase-9, EC 3.4.24.15, and ␣ 2macroglobulin complexes have also been reported to play a role in A␤ degradation (11)(12)(13). In this report, we describe the identification of endothelin-converting enzyme-1 (ECE-1) as a novel A␤-degrading enzyme. The endothelin converting enzymes are a class of type II integral membrane zinc metalloproteases (active site luminal) named for their ability to hydrolyze a family of biologically inactive intermediates, big endothelins (big ETs), exclusively at a Trp 21 -Val/Ile 22 bond to form the potent vasoconstrictors endothelins (14). In addition to this specific cleavage event, ECE-1 has been reported to hydrolyze several biologically active peptides in vitro, including bradykinin, neurotensin, substance P, and oxidized insulin B chain by cleaving on the amino side of hydrophobic residues (15,16).
Two different endothelin-converting enzymes have been cloned. The first identified, ECE-1, is abundantly expressed in the vascular endothelial cells of all organs and is also widely expressed in nonvascular cells of tissues including lung, pancreas, testis, ovary, and adrenal gland (17)(18)(19). A comprehensive analysis examining both ECE activity and expression in human brain has not been reported. Studies have, however, detected human ECE-1 immunoreactivity in fibers within the glial limitans and neuronal processes and cell bodies of the cerebral cortex (18). In rats, ECE-1 immunoreactivity has been detected in pyramidal cells of the hippocampus and in cultured primary astrocytes (20).
Four isoforms of human ECE-1 differing only in the cytoplasmic tail are produced by a single gene located on chromosome 1 (1p36) through the use of alternate promoters (17,(21)(22)(23)(24)(25). The four isoforms cleave big ETs with equal efficiency but differ primarily in their subcellular localization and tissue distribution (24,25). Human ECE-1a is localized predominantly to the plasma membrane (24,25). Human ECE-1c and ECE-1d have also been reported to be localized predominantly to the plasma membrane with additional intracellular expression detected (24,25). In contrast, human ECE-1b appears to be localized exclusively intracellularly. Co-immunolocalization studies performed by Schweizer et al. (24) on human ECE-1b-transfected CHO cells indicate the presence of this isoform in the trans-Golgi network (TGN). Azarani et al. similarly demonstrated that human ECE-1b was located in an intracellular compartment when expressed in Madin-Darby canine kidney cells (26), and Cailler et al. (27) demonstrated that a dileucine motif in the cytosolic tail of ECE-1b was probably responsible for its intracellular localization. In an endogenous ECE-1b-and ECE-1c-expressing cell line, ECV304, ECE-1 immunoreactivity was detected in intracellular Golgi-like structures as well as at the cell surface (24).
Bovine ECE-1b, which corresponds to human ECE-1c, is also localized predominantly on the plasma membrane (28). However, in contrast to human ECE-1a, bovine ECE-1a has convincingly been shown to be constitutively targeted to the lysosome (28). This difference between the localization of human and bovine ECE-1a may be due to the fact that the isoformspecific N-terminal region of ECE-1a is poorly conserved between the species. In fact, Emoto et al. (28) identified lysosometargeting signals in the N-terminal tail of bovine ECE-1a that are not conserved in human ECE-1a.
ECE-2 is a homologous enzyme with catalytic activity similar to that of ECE-1. Bovine ECE-2 has been cloned and is encoded by a separate gene from ECE-1 (29). The sequence and chromosomal location of the human ECE-2 gene have not been reported. However, Nagase et al. (30) recently reported the cloning of an unidentified human brain cDNA, KIAA0604, that shares 93% identity with the bovine ECE-2 gene and is located on human chromosome 3. Given the similarity to bovine ECE-2, this cDNA probably represents human ECE-2.
ECE-2 is localized intracellularly and has an acidic pH optimum (29). Immunocytochemical analysis of endogenous ECE-2 in HUVECs revealed a punctate pattern of staining consistent with expression of ECE-2 in acidic intracellular vesicles of the constitutive secretory pathway (31). Northern blot analysis of bovine tissues revealed that ECE-2 is most abundantly expressed in neural tissues including cerebral cortex, cerebellum, and adrenal medulla, with low level expression detected in many other tissues (29). In mouse brain, ECE-2 is expressed in heterogeneous populations of neurons in the thalamus, hypothalamus, amygdala, dentate gyrus, and CA3 (32). Like ECE-1, ECE-2 cleaves big ET-1 most efficiently among the three big ETs (24,29). Another member of the ECE family, ECE-3, has recently been purified from bovine iris microsomes and is highly specific for the conversion of big ET-3 (33). The enzyme responsible for this activity has not yet been cloned.
The role that ECE may play in Alzheimer's disease has not been previously explored. Here we present both pharmacological and biochemical evidence that ECE-1 can hydrolyze A␤ both in vitro and in vivo. These data indicate a potential role for this enzyme family in A␤ catabolism.

EXPERIMENTAL PROCEDURES
Analysis of A␤ Concentration by Sandwich ELISA-Human A␤ was measured by sandwich ELISA as previously described (34), using the BAN50/BA27 and BAN50/BC05 antibody systems (Takeda) to detect A␤40 and A␤42, respectively. Hamster A␤ derived from CHO cells was measured using the BNT77/BA27 and BNT77/BC05 antibody systems. The BNT77 antibody (Takeda) was raised against amino acids 11-28 and thus may recognize amino-terminally modified or truncated peptides (35). A␤ concentration was determined by comparing values obtained for samples with those obtained for synthetic A␤40 and A␤42 standards (Bachem).
RNA was prepared from HUVEC cells using the Qiagen RNeasy miniprep kit and was reverse transcribed using Superscript II reverse transcriptase and an oligo(dT) primer (Roche Molecular Biochemicals). The 5Ј fragment of ECE-1a was amplified using primers 1 and 3. The 5Ј fragment of ECE-1b was amplified using primers 2 and 3. The 3Ј fragment of ECE-1, which is common to both isoforms, was amplified using primers 4 and 5. Pfu polymerase (Stratagene) was used for all amplifications. The 5Ј and 3Ј fragments were ligated together at the PvuII site and subcloned into pcDNA3 (Invitrogen) using primer-encoded EcoRI and XbaI sites. The sequences of the constructs were confirmed by dideoxy sequencing by the Mayo Molecular Biology Core Facility.
Cell Culture and Transfections-Unless otherwise noted, cell culture reagents were purchased from Life Technologies, Inc., and cell lines were purchased from ATCC. HUVECs were cultured in Kaighn's F12K medium (ATCC) supplemented with 10% fetal bovine serum, 0.1 mg/ml heparin, 0.03 mg/ml endothelial cell growth supplement (Sigma), 100 units/ml penicillin, and 100 g/ml streptomycin. CHO cells were cultured in Ham's F-12 medium (BioWhittaker) supplemented with 10% newborn calf serum, 100 units/ml penicillin, and 100 g/ml streptomycin. H4 cells (human neuroglioma origin) were cultured in Opti-MEM supplemented with 10% fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomycin. For passaging of cells prior to experiments in which ECE activity was to be measured, a highly purified trypsin (Sigma T-7418) solution was used (17). CHO cells were transfected with FuGENE 6 (Roche Molecular Biochemicals) according to the directions of the manufacturer. Stable lines were generated by selecting pcDNA3-transfected cells (ECE-1a and ECE-1b) with 1 mg/ml Geneticin and pSecTag-transfected cells (solECE-1) with 0.8 mg/ml Zeocin (Invitrogen).
Measurement of ECE Activity in Cell Membrane Fractions-Cell membrane fractions were prepared as described by Xu et al. (17). Treatment of Cells with Metalloprotease Inhibitors-Cells were passaged into six-well plates 1 day prior to treatment and were utilized at confluence. Triplicate wells were washed twice with Hanks' balanced salt solution and then incubated for 17-24 h with 1 ml of growth medium containing phosphoramidon (Roche Molecular Biochemicals), thiorphan (Sigma), or captopril (Sigma) at the indicated concentrations. Control cells were incubated in growth medium containing an equal concentration of vehicle (phosphate-buffered saline). After treatment, the culture medium was harvested and spun at 14,000 ϫ g, and the supernatant was analyzed for A␤40 and A␤42 by sandwich ELISA as described and for secreted APP by Western blot. To assess cellular toxicity of the compounds, MTS assays (CellTiter 96 ® , Promega), which measure the conversion of MTS to formazan by metabolically active cells, were performed on the cells after the indicated time points. Culture medium was subjected to electrophoresis on 10 -20% Tricine gels (Novex) and was subsequently transferred to Immobilon P (Millipore Corp.). Western blots on CHO cells were performed using 22C11 antibody (Roche Molecular Biochemicals) to detect secreted APP. Bound antibody was detected by incubation with the appropriate horseradish peroxidase-linked secondary antibody (Amersham Pharmacia Biotech) followed by ECL Western blotting reagents (Amersham Pharmacia Biotech) and exposure to x-ray film.
The solECE-1 construct was subcloned using primer-encoded EcoRI and NotI sites into pSecTag2B (Invitrogen), which incorporates a leader sequence onto the N terminus of the protein for secretion by mammalian cells and sequential c-Myc and His 6 tags at the C terminus to facilitate detection and purification. The sequence of the construct was confirmed by dideoxy sequencing by the Mayo Molecular Biology Core Facility. A stable solECE-1 secreting cell line was generated by transfecting the construct into CHO cells and selecting with Zeocin (0.8 mg/ml). At confluence, the cells were washed twice with Hanks' balanced salt solution and then cultured for 48 h in serum-free medium (CHO-S-SFM II) containing 1 mM sodium butyrate. Conditioned medium was filtered through a 0.2-m filter and dialyzed against binding buffer, 0.05 M sodium phosphate, pH 8.0, 0.3 M NaCl, prior to purification via the C-terminal His 6 tag using Ni 2ϩ -nitrilotriacetic acid-agarose (Qiagen). Bound protein was eluted from the Ni 2ϩ -nitrilotriacetic acidagarose with binding buffer containing 100 mM imidazole.
The solECE-1 was initially estimated to be Յ25% pure, as assessed by SDS-polyacrylamide gel electrophoresis and Coomassie staining (data not shown). The concentration of active solECE-1 was further estimated by determining the second order rate of big ET-1 hydrolysis and dividing this rate by the published k cat /K m . Specifically, big ET-1 (0.1 M) was reacted at 37°C with solECE-1 in 50 mM MES, pH 6.5, containing 100 mM NaCl, 0.05% bovine serum albumin, 1 mM PMSF, 100 M leupeptin, and 20 M pepstatin. The reactions were stopped by the addition of EDTA (5 mM), and mature ET-1 peptide was measured by sandwich ELISA (Amersham Pharmacia Biotech). The second order rate was determined as the apparent rate constant k from Equation 1, where y is the fraction of substrate hydrolyzed and t is the time. Since k ϭ k cat /K m ϫ [E], the calculated k was then divided by the reported k cat /K m for big ET-1 hydrolysis by solECE-1 (2.52 ϫ 10 4 M Ϫ1 s Ϫ1 at pH 6.5) (16) to determine the active enzyme concentration. The active enzyme concentration was estimated by this method throughout this report and was ϳ30-fold lower than that originally estimated from the protein concentration and Coomassie-stained gel, suggesting either that our estimation of purity by Coomassie staining was incorrect or that a majority of the solECE-1 was in an inactive form. To control for the presence of co-purifying native CHO proteins, conditioned serumfree medium from nontransfected CHO cells was purified as above, and the eluted proteins were used in control experiments.
ELISA Analysis of solECE-1-mediated A␤40 and A␤42 Degradation-SolECE-1 was preincubated for 15 min at room temperature with the ECE inhibitor PD069185 (synthesized according to published methods (38) by the Organic Chemistry Core Facility at the Mayo Clinic Jacksonville) or phosphoramidon (150 M) or an equal volume of vehicle (Me 2 SO and phosphate-buffered saline, respectively), prior to incubation with synthetic A␤40 or A␤42 (0.01 M) in 50 mM MES, pH 6.5, containing 0.01% C 12 E 10 , 1 mM PMSF, 100 M leupeptin, and 20 M pepstatin. The amount of solECE used in this reaction was estimated to be ϳ6 nM. As a control, A␤ was incubated with Ni 2ϩ -nitrilotriacetic acid-purified proteins from nontransfected CHO cells or in reaction buffer alone. Following incubation at 37°C for the indicated time points, the reactions were stopped by the addition of 5 mM EDTA. A␤ concentration was then analyzed using a highly specific sandwich ELISA, which captures A␤ via binding of antibody BAN50 to the N terminus and detects full-length peptides ending at position 40 (BA27) or 42 (BC05). Thus, A␤ that has been cleaved will not be detected in this assay.
HPLC Analysis of solECE-1-mediated Degradation of 3 H-Labeled A␤40 -To further analyze the degradation of A␤ by solECE-1, the enzyme was incubated as above with 3 H-radiomethylated A␤40 (0.56 M, 2 ϫ 10 4 dpm), a gift from Dr. T. L. Rosenberry. A␤ peptide and fragments were resolved by reversed-phase HPLC on a Vydac C4 column using a linear gradient of 0 -100% B in 60 min (A buffer ϭ 0.1% trifluoroacetic acid in water; B buffer ϭ 0.1% trifluoroacetic acid in acetonitrile). Fractions were collected and counted in a Beckman LS 6500 scintillation counter.
Characterization of A␤40 Proteolytic Fragments by HPLC, Mass Spectrometry, and Edman Sequencing-solECE-1 (ϳ20 nM) was incubated as above with synthetic A␤40 (100 M) for 17 h at 37°C. The digest samples were run on an Applied Biosystems 130A Separation System (Applied Biosystems, Foster City, CA) using a C4 reversedphase column (Brownlee Aquapore BU-300; 2.1 ϫ 220 mm). The peptides were eluted with a gradient of 5-80% B over a period of 69 min (A buffer ϭ 0.1% trifluoroacetic acid; B buffer ϭ 80% acetonitrile, 20% water, 0.09% trifluoroacetic acid). The flow rate was 200 l/min, and absorbance was monitored at 215 nm. Half-minute fractions were collected, and peaks were analyzed by mass spectrometry and Edman sequencing. The masses of the collected peptides were determined with a PerSeptive Biosystems MALDI-TOF Voyager DE-RP Mass Spectrometer (Framingham, MA) operated in the delayed extraction and reflector mode using ␣-cyano-4-hydroxycinnamic acid. The first two amino acids of each peptide were determined using an Applied Biosystems Procise 492 sequencer.
Kinetics of A␤ Hydrolysis by solECE-1-In an attempt to determine the K m of A␤40 hydrolysis by solECE-1, we incubated the enzyme (ϳ17 nM) with synthetic A␤40 (3, 10, and 20 M) for 0 or 6 h at 37°C in 50 mM MES buffer, pH 6.5, containing 0.05% bovine serum albumin, 1 mM PMSF, 100 M leupeptin, and 20 M pepstatin. Control reactions were carried out in the absence of enzyme or in the presence of solECE-1 inhibited by phosphoramidon (100 M). The reactions were stopped by the addition of EDTA (5 mM), and A␤40 concentration was determined by sandwich ELISA using the BAN50/BA27 system. The rate of A␤ hydrolysis in these assays was linear with respect to substrate concentration, precluding a determination of K m and V max .
We next determined the second order rate constant, k cat /K m , for A␤40 hydrolysis relative to that for big ET-1 hydrolysis by solECE-1. The second order rate for hydrolysis of each substrate was determined at substrate concentrations well below K m . Specifically, A␤40 (2.5 M) and big ET-1 (0.1 M) were incubated in triplicate alone or with solECE-1 (0.3-8.3 nM) at 37°C at either pH 6.5 or pH 5.6 in 50 mM MES buffer containing 100 mM NaCl, 0.05% bovine serum albumin, 1 mM PMSF, 100 M leupeptin, and 20 M pepstatin. The reactions were stopped by the addition of EDTA (5 mM). Remaining full-length A␤40 was determined by BAN50/BA27 sandwich ELISA, and the amount of ET-1 peptide generated was measured by ET-1 sandwich ELISA (Amersham Pharmacia Biotech). The apparent rate constant k was again determined from Equation 1. Under these conditions, no detectable ET-1 peptide was generated in the absence of solECE-1. However, we consistently observed some loss of A␤40 in the absence of enzyme, presumably due to adsorption to the reaction tube or other nonspecific mechanism. Therefore, the rate of loss in the absence of enzyme was determined and subtracted from the rate determined in the presence of solECE-1. For subsequent calculations the average k Ϯ S.E. for A␤40 hydrolysis at each pH was determined from three experiments, with triplicate reactions in each experiment. For big ET-1 hydrolysis, the k is an average of two separate experiments, with triplicate reactions in each experiment. From the enzyme concentration [E] determined from the k for big ET-1 at pH 6.5 as outlined under Equation 1, we determined the k cat /K m for A␤40 hydrolysis at pH 6.5 and 5.6 and for big ET-1 at pH 5.6 with the equation k cat /K m ϭ k/[E].

RESULTS
Phosphoramidon, but Not Thiorphan or Captopril, Increases A␤ Accumulation by H4 Neuroglioma Cells-Our group and others have previously shown that treatment with metalloprotease inhibitors, in particular phosphoramidon, results in a rapid 2-3-fold increase in the concentration of A␤40 and A␤42 in the conditioned medium of neuronal cell lines without affecting the concentration of secreted APP (sAPP) (39,40). Importantly, this increase in A␤ concentration is as large or larger than that seen in cells expressing most AD-causing mutations (2). The enzyme(s) responsible for the phosphoramidon-induced elevations in A␤ has not previously been reported. Phosphoramidon is known to inhibit several metalloproteases including NEP (IC 50 ϭ 0.034 M), angiotensin-converting enzyme (ACE; IC 50 ϭ 78 M), ECE-1 (IC 50 ϭ 1-3.5 M), and ECE-2 (IC 50 ϭ 0.004 M) (29, 41) but does not inhibit insulin-degrading enzyme (42). A role for NEP in extracellular A␤ catabolism has been highlighted in a recent report from Iwata and colleagues (9). Infusion of the metalloprotease inhibitor thiorphan into the hippocampus of rats resulted in a significant increase in the amount of A␤ and in the deposition of the longer more amyloidogenic form, A␤42, reportedly through the inhibition of A␤ degradation by NEP. NEP and ACE have been reported to reside predominantly on the cell surface, although a soluble form of NEP is also present in serum and cerebral spinal fluid (43)(44)(45)(46). A recently identified thiorphan-sensitive NEP homo-logue SEP/NL1/NEPII is expressed both as a membrane-bound and secreted protease (47)(48)(49). We evaluated a role for NEP and ACE in H4 neuroglioma cells using more selective inhibitors of these enzymes, thiorphan and captopril, respectively. We have not yet been able to similarly analyze ECE, since a more selective inhibitor of ECE is not commercially available.
Treatment of H4 cells with phosphoramidon (34 M) resulted in a greater than 2-fold elevation in A␤40 accumulation (Fig.  1), with a half-maximal effect occurring at a dose of ϳ7.5 M (data not shown). However, treatment with thiorphan or captopril at concentrations greater than 1000 times the reported IC 50 for the target enzymes in in vitro studies (50,51), but less than that required to inhibit ECE, failed to result in increases in extracellular A␤ (Fig. 1), indicating that the phosphoramidon-induced effect in H4 cells is not likely to be due to inhibition of NEP or ACE. Similar results were obtained for A␤42 (data not shown). Since NEP and ACE are localized mainly on the cell surface, the membrane permeability of these compounds is not relevant to the inhibition of the known protease, although we cannot rule out the possible presence of an intracellular form of NEP or ACE based on these results. Similarly, we cannot rule out the possibility that the phosphoramidoninduced effect in these cells is due to an as yet unidentified enzyme that is insensitive to treatment with thiorphan and captopril. We did, however, find endogenous ECE activity in solubilized membranes of H4 cells using a big ET conversion assay (17, 36) (data not shown). Collectively, these data led us to investigate ECE more closely.
Overexpression of Endothelin-converting Enzyme-1 Results in a Significant Decrease in Extracellular A␤ Concentration That Is Completely Reversed by Treatment with Phosphoramidon-Evidence implicating a potential role for ECE in modulating A␤ concentration came further from the casual observation that CHO cells, which have no endogenous ECE activity (17), produce very high levels of A␤ when compared with most other cell types 2 and fail to respond to phosphoramidon (see Fig. 2). Conversely, HUVECs, which have high levels of endogenous ECE (52), accumulate very little A␤ unless treated with high concentrations of phosphoramidon (data not shown). To further investigate the role of ECE in A␤ accumulation, we cloned and stably transfected CHO cells with human ECE-1b and ECE-1a. ECE activity, determined using a big ET-1 conversion assay (17,36), was confirmed to be present in solubilized membranes from the ECE-1-transfected cells and absent in vector-transfected cells (data not shown). The amount of ECE-1 activity in each of our stable ECE-1 lines was similar. Overexpression of either ECE-1a or ECE-1b in CHO cells, which lack endogenous ECE activity, resulted in a striking 75-90% reduction in A␤40 and a 45-60% reduction in A␤42 (Fig. 2). No significant changes were observed in the amount of sAPP accumulation in ECE-1-transfected cells compared with the vector controls, indicating that the cells were similarly viable and that general secretion is not affected by ECE-1 overexpression. The reduction in A␤ concentration in ECE-1aand ECE-1b-transfected cells was completely reversed by treatment with phosphoramidon, indicating that the observed phenotype was probably due to the enzymatic activity of the overexpressed ECE-1.
Increased Removal of Exogenous A␤ Is Apparent Only in ECE-1a-transfected Cells-While the exact mechanism of the phosphoramidon-induced increase in A␤ concentration in neuronal cells was unknown, it has been suggested that it may be the result of inhibition of intracellular degradation of the peptide (40). Consistent with this hypothesis, treatment with phosphoramidon is reported to result in a 2-fold increase in extracellular A␤ concentration and an increase in cell-associated A␤ without affecting sAPP levels in SY5Y cells treated with the 2 C. B. Eckman, unpublished observation.  2. Overexpression of ECE-1a or ECE-1b in CHO cells reduces extracellular A␤ concentration without affecting secretion of sAPP. A␤40 (A) and A␤42 (B) concentration in the conditioned media of stable ECE-1a-and ECE-1b-transfected CHO cell lines was determined by sandwich ELISA (BNT77/BA27 and BNT77/BC05, respectively) following a 24-h incubation with or without 100 M phosphoramidon. Data are plotted as mean Ϯ S.E. of triplicate wells. Western blot analysis (C) was also performed on the conditioned media of cells incubated with or without phosphoramidon (phos), using 22C11 antibody to detect sAPP. compound (40). As we have shown, treatment of H4 neuroglioma cells with phosphoramidon also results in a significant increase in extracellular A␤ concentration (see Fig. 1). To examine whether the phosphoramidon-induced increases in A␤ were likely to be due to inhibition of a cell-surface or secreted protease, we performed a spike experiment in which exogenous A␤ was added to the medium bathing H4 cells in the absence or presence of phosphoramidon. As shown in Fig. 3, exogenous A␤42 was removed equally well in the presence or absence of phosphoramidon. The inset shows the results from a sister set of culture wells where synthetic A␤ was not added, indicating that phosphoramidon was indeed promoting endogenous A␤ accumulation in this experiment. Similar data were obtained using synthetic A␤40 (data not shown).
To determine whether extracellular A␤ removal could account, at least in part, for the dramatic decrease in extracellular A␤ concentration in the ECE-1-transfected cell lines, we next spiked synthetic A␤40 into the culture medium in the presence or absence of phosphoramidon and determined the percentage of removal by sandwich ELISA at 6 and 24 h. After a 6-h incubation, removal of A␤ was similar in the culture medium of vector-and ECE-1-transfected CHO cells (Fig. 4A) and was not affected by phosphoramidon treatment (data not shown), although endogenous A␤ accumulation by phosphoramidon-treated ECE-1-transfected cells was increased 1.5-2fold during the same time period (Fig. 4B).
Following a 24-h incubation, we did observe a significant increase (p ϭ 0.0495) in the removal of the spiked-in A␤ in the medium of ECE-1a-transfected cells compared with the vector controls (Fig. 4A). No significant change in exogenous A␤ removal was observed in cells expressing ECE-1b. The ECE-1ainduced increase in A␤ removal could be completely attenuated by phosphoramidon treatment, indicating that the effect was probably due to the enzymatic activity of ECE-1a (Fig. 4A,  inset). In the same ECE-1a cells, phosphoramidon treatment resulted in an ϳ600% increase in endogenous A␤ accumulation at the 24-h time point (Fig. 4B).
Partially Purified solECE-1 Degrades A␤ in Vitro-Recom-binant, soluble forms of ECE-1 (solECE-1) lacking the intracellular and transmembrane domains have been reported to hydrolyze big ET-1 with activity comparable with that of membrane-bound ECE-1a (16,36,37). The soluble ECE-1 preparation described by Korth et al. (37) was active in a broad pH range from 5 to 7, with an optimum of pH 6.6 -6.8 for big ET-1. Under the conditions assayed, Ahn et al. (36,53) found that their soluble ECE-1 preparation had a somewhat narrower pH

FIG. 4. Removal of exogenous synthetic A␤ by ECE-overexpressing CHO cells.
A, synthetic human A␤40 (150 pM) was added to confluent CHO cells stably transfected with ECE-1a, ECE-1b, or the control vector and incubated for 6 and 24 h. Human A␤40 was measured at the indicated time points using the BAN50/BA27 sandwich ELISA, which does not detect endogenous CHO A␤. Data shown represent the mean Ϯ S.E. of triplicate wells that were incubated with synthetic A␤40. The concentration of A␤ remaining after 24 h is significantly lower in ECE-1a cells than in vector controls (p ϭ 0.0495, Mann-Whitney). The inset graph shows the percentage of exogenous A␤ removed by ECE-1a and vector-transfected cells after 24 h in the presence of phosphoramidon (34 M). B, a second set of cells was incubated with or without phosphoramidon (34 M) for the same time period to determine the accumulation of endogenous A␤. Endogenous A␤40 was measured at the indicated time points using the BNT77/BA27 sandwich ELISA. Data are plotted as mean Ϯ S.E. of triplicate wells. Given that BNT77 was raised against amino acids 11-28 of A␤, this assay can also detect amino truncated peptides and may lead to an overestimation of full-length A␤40. optimum, with an optimal pH for big ET-1 of 6.5. To examine whether ECE-1 is capable of direct catabolism of A␤, we generated a soluble ECE-1 similar to those previously described. Incubation of synthetic A␤40 and A␤42 with this enzyme resulted in a nearly complete loss of the full-length peptides as detected by sandwich ELISA (Fig. 5). This reduction was completely blocked by incubation with phosphoramidon and also with a more selective ECE-1 inhibitor, PD069185. (This inhibitor, while very useful for in vitro studies, is not informative in cell-based studies due to its toxicity (38)). To confirm that the loss of A␤ was indeed due to A␤ catabolism and not to A␤ binding or some other phenomenon, we analyzed the effect of solECE-1 on A␤ by HPLC with a radiolabeled A␤ reporter molecule. Incubation of 3 H-labeled A␤40 with solECE-1 resulted in loss of the full-length peptide and formation of at least three novel peaks detected by reversed-phase chromatography (Fig. 6). The formation of these peaks was completely blocked by treatment with PD069185.
Kinetic Analysis of A␤40 Cleavage by solECE-1-SolECE-1 has been reported to hydrolyze big ET-1 with a K m of ϳ2-4 M (16, 36, 37) and bradykinin with a K m of 340 M (16). Despite a much higher K m , the catalytic efficiency of bradykinin hydrolysis by solECE-1 actually exceeds that of big ET-1, with a second order rate constant of 6.6 ϫ 10 4 M Ϫ1 s Ϫ1 for bradykinin compared with 2.5 ϫ 10 4 M Ϫ1 s Ϫ1 for big ET-1 at pH 6.5 (16). To attempt to determine the K m for A␤40 hydrolysis by solECE-1, we initially examined the rate of hydrolysis of A␤ at various substrate concentrations up to 20 M at the reported pH opti- Identical reactions were carried out with co-purifying proteins isolated from nontransfected CHO cells (Control). After the incubation, the remaining A␤ was detected using the BAN50/BA27 and BAN50/BC05 sandwich ELISA systems, which detect full-length A␤40 or A␤42 peptides, respectively. Data are plotted as mean Ϯ S.E. of triplicate reactions.
FIG. 6. HPLC analysis of 3 H-labeled A␤40 degradation by soluble ECE-1. Soluble ECE-1 (ϳ6 nM) was incubated for 24 h at 37°C with 3 H-labeled A␤40 (0.56 M, 2 ϫ 10 4 dpm). As a control, the enzyme was preincubated for 15 min with the ECE inhibitor PD069185 (150 M) prior to the addition of A␤. The resulting peptides were separated by reversed-phase chromatography (C4 column). 1-min fractions were collected, and 3 H dpm was determined by liquid scintillation counting. For comparative purposes, synthetic A␤ alone elutes as a single peak at fraction 21, identical to that observed for A␤ incubated with solECE-1 in the presence of PD069185 (data not shown) .   FIG. 7. Determination of sites of A␤40 cleavage by solECE-1. A, unlabeled synthetic A␤40 (100 M) was digested with solECE-1 (ϳ20 nM) overnight at 37°C. The cleavage products were separated by reversed-phase HPLC using a C4 column. Half-minute fractions were collected, and absorbance was monitored at 215 nm. B, amino acid sequence of A␤40 showing principal solECE-1 cleavage sites. The identities of HPLC peaks 1-5 were determined by mass spectrometry and NH 2 -terminal sequencing (see Table I). The open arrows under the sequence indicate previously determined NEP cleavage sites (8). mum for big ET-1 cleavage. The rate of A␤ hydrolysis was linear with respect to substrate concentration up to 20 M (data not shown). We were concerned that the use of significantly higher concentrations of A␤ in these experiments would complicate the kinetic measurements, since A␤40 peptide is prone to aggregate and precipitate in the high micromolar range. Thus, we were unable to determine the K m and V max . We were able, however, to calculate the k cat /K m by measuring the rate of A␤ hydrolysis by solECE under second-order conditions. Under these conditions, when the substrate concentration is well below K m , the rate of substrate hydrolysis is equal to the k cat /K m multiplied by the enzyme concentration (see "Experimental Procedures"). Using this method, the k cat /K m for A␤40 hydrolysis by solECE-1 was determined to be (1.7 Ϯ 0.6) ϫ 10 3 M Ϫ1 s Ϫ1 at pH 6.5. This value is 15-fold lower than that for big ET-1 hydrolysis under the same conditions.
Intrigued by a recent report indicating that solECE-1 cleaves bradykinin and substance P with an acidic pH optimum of ϳ5.6 compared with the optimum of pH 6.5 for big ET-1 (53), we next compared the efficiency of solECE-1 hydrolysis of A␤40 and big ET-1 at pH 5.6. Similar to bradykinin and substance P, solECE-1 cleaves A␤40 more efficiently at pH 5.6, with a k cat /K m determined to be (2.0 Ϯ 0.5) ϫ 10 4 M Ϫ1 s Ϫ1 , ϳ12 times greater than the value determined at pH 6.5. The k cat /K m for big ET-1 hydrolysis by our preparation of solECE-1 at pH 5.6 under these conditions was determined to be ϳ6.1 ϫ 10 4 M Ϫ1 s Ϫ1 at pH 5.6, slightly greater than that determined at pH 6.5. This value of k cat /K m for big ET-1 hydrolysis at pH 5.6 is only 3-fold greater than that for A␤40 at the same pH. It is important to note that the k cat /K m values reported here rely on a determination of enzyme concentration based on the reported k cat /K m of big ET-1 hydrolysis by solECE-1 at pH 6.5 (see "Experimental Procedures"), and thus the absolute values may differ based on any differences in actual enzyme concentration from our estimates. However, the relative k cat /K m for A␤40 hydrolysis versus big ET-1 hydrolysis was determined in parallel experiments with the same enzyme and would not change even if the estimated concentration of enzyme were slightly different. DISCUSSION Recently, there has been considerable debate over the enzyme or enzymes that contribute most to A␤ catabolism in the human brain. Both insulin-degrading enzyme and neprilysin have been argued to be major proteases involved in the degradation of secreted A␤ (5,9). However, with the data available at this time, it is impossible to determine which, if any, of the identified proteases contributes most to A␤ degradation in the intact brain. It is likely that multiple proteases, both intracellular and extracellular, may play a role in determining A␤ concentration. The relative contribution of A␤-degrading enzymes and other mechanisms of A␤ removal may vary in different regions of the brain and may also differ for A␤40 and A␤42. A decrease in the activity of any of these mechanisms, whether they are major or minor, may potentially result in increased A␤ accumulation and the development of AD pathology. Conversely, an increase in the activity of any enzyme capable of degrading A␤ may result in decreased accumulation of the peptide, potentially reducing the risk for AD.
Taken together, the data presented in this report indicate that ECE-1 activity can dramatically affect A␤ concentration, probably by direct degradation of the peptide. Recombinant soluble ECE-1 has a substrate specificity in non-ET peptides similar to that of NEP, with preferential cleavage on the amino side of hydrophobic residues (16). In contrast to NEP, however, ECE-1 appears not to cleave peptides smaller than 6 amino acids in length. Given this specificity, there are ϳ13 potential ECE/NEP cleavage sites in the A␤40 peptide. At least five of these sites are reported to be cleaved by recombinant NEP in vitro (8). Using HPLC, mass spectrometry, and NH 2 -sequence analysis, we have determined that soluble ECE-1 cleaves synthetic A␤40 at least three sites, resulting in the formation of A␤ fragments 1-16, 1-17, 1-19, and 20 -40. Consistent with the known substrate specificity of ECE-1, each of these observed cleavages by solECE-1 occurred on the amino side of hydrophobic residues (Leu 17 , Val 18 , and Phe 20 ). Given that ECE-2 is highly homologous to ECE-1 and shares similar catalytic activity (29), we hypothesize that ECE-2 may also be capable of degrading A␤. Since ECE-2 is abundantly expressed in the central nervous system, this activity may also potentially be relevant to the accumulation of A␤ in the brain.
ECE-1 activity is localized both intracellularly and on the cell surface in many cell types, with ECE-2 appearing to reside exclusively within the cell. In CHO cells overexpressing human ECE-1a, we observed increased phosphoramidon-sensitive removal of exogenous A␤ after 24 h of incubation, indicating that ECE-1a may also contribute slightly to the extracellular degradation of the peptide. In CHO cells overexpressing ECE-1b and in H4 neuroglioma cells expressing endogenous ECE, degradation of exogenous A␤ was not sensitive to phosphoramidon. These results raise the possibility that the dramatic increase in A␤ accumulation by these cells in the presence of phosphoramidon may be due to inhibition of intracellular degradation of the peptide. Even in ECE-1a-expressing CHO cells, the dramatic increase in A␤ concentration upon treatment with phosphoramidon does not appear to be accounted for by the modest increase in exogenous A␤ degradation. We cannot, however, rule out the possibility of a local event at the cell surface upon secretion of endogenous A␤ that might not be evident in our spike experiments, where the peptide is diluted directly into the culture medium. While ECE-1a has been reported to be localized predominantly to the cell surface, this isoform has been shown to process big ET-1 intracellularly in CHO cells, most likely in secretory vesicles, as well as at the cell surface (54). Therefore, ECE-1a may similarly degrade A␤ intracellularly in CHO cells as it is being trafficked to the cell surface.
While detailed co-localization studies have not been performed, separate studies indicate that ECE and A␤ are present in the same cellular compartments. Human ECE-1b has been reported to be present in the TGN, a proposed site of A␤40 generation in neuronal cells (24,55). Interestingly, we found that solECE-1 hydrolyzed A␤40 more efficiently at pH 5.6 than at pH 6.5, with a k cat /K m at pH 5.6 only 3-fold lower than that for a known physiological substrate, big ET-1. This result may be particularly relevant, since the TGN, where ECE-1b appears to be expressed, has an acidic pH. ECE-2 is also likely to be present in the TGN and vesicles of the constitutive secretory pathway (29,31). Consistent with the hypothesis that phosphoramidon may inhibit intracellular degradation of A␤, Fuller et al. (40) have reported that phosphoramidon treatment of SY5Y neuronal cells results in an increase in cell-associated A␤. Unfortunately, we have not been able to convincingly detect intracellular A␤ in H4 cells either in the presence or absence of phosphoramidon, presumably due to rapid secretion of the peptide. However, the fact that removal of exogenous A␤ by H4 cells is insensitive to phosphoramidon suggests that the compound may exert its effect on A␤ through an intracellular event.
Phosphoramidon itself has a polar structure that would appear to make it poorly cell-penetrant (56). However, this compound has been shown to be able to inhibit the intracellular conversion of big ET-1 to mature endothelin (17,29,57). It is possible that phosphoramidon may elevate A␤ accumulation through inhibition of intracellular ECE, cell surface ECE, or both, depending on the expression of cell surface and intracellular ECE in various cell types.
The 2-3-fold increase in endogenous A␤ accumulation by cell lines treated with phosphoramidon is as large as or larger than that seen in cells expressing most AD-causing mutations (2), suggesting that disruption of phosphoramidon-sensitive A␤ degradation may be a significant risk factor for AD. While many factors in addition to ECE probably contribute to endothelin levels, and the results are controversial, it is noteworthy that decreases in endothelin levels have been reported in the cerebral spinal fluid of AD patients when compared with nondemented control individuals (58). A reasonable test for the degree of involvement of ECE or any protease in the brain is to examine the effect of animals null for these enzymes. ECE-1 knockout mice have craniofacial and cardiac abnormalities resulting in embryonic lethality, while ECE-2 knockout mice develop normally and are viable (32,59). Surprisingly, large amounts of mature ET-1 peptide are detected in both lines. In ECE-1/ECE-2 double knockout mice, the cardiac defects are even more severe; however, mature ET-1 peptides are still detected, indicating compensation by as yet unidentified proteases (32,59). The apparent redundancy of endothelin-converting enzyme activity may limit the utility of the knockouts to examine the contribution of ECE in A␤ accumulation in vivo. Nonetheless, these experiments are under way and may provide additional insight into the role of ECE in endogenous A␤ accumulation in the brain.
Regardless of the extent of the in vivo role of ECE activity in the amount of A␤ accumulation in the brain, these results are highly significant for several reasons. First, ECE inhibitors have received a large amount of pharmaceutical interest for their potential as anti-hypertension drugs (60). If degradation of A␤ by ECE does, as we hypothesize, contribute to the extracellular concentration of the plaque-forming peptide in the brain, inhibiting this activity may lead to the development of AD and/or accelerate the disease in susceptible individuals. Because most ECE and NEP inhibitors will inhibit both enzymes at certain concentrations, the use of these drugs may be particularly risky if ECE and NEP both play physiological roles in the degradation of A␤ in the brain. ET receptors may be a safer target for pharmacological interference with the endothelin system to reduce hypertension without the side effect of decreased A␤ catabolism.
Second, up-regulation of ECE activity may be useful therapeutically for the treatment of AD. One obvious concern with this treatment method is that patients may become hypertensive. Up-regulation of ECE activity in the periphery of mice injected with a construct to increase ECE expression, however, does not appear to result in increased circulating endothelin levels, indicating that ECE is not likely to be rate-limiting in the conversion of big ET to ET (61). Further, even if increased ECE activity does augment endothelin levels, endothelin recep-tor antagonists could be given in parallel to reduce or block any effect of increased endothelin levels.
Third, mutations in ECE may be identified that are causative of Alzheimer's disease in certain individuals. In this regard, it is worthwhile to note that the sib-pair analyses of genetic factors contributing to late onset AD have not excluded the region on chromosome 1 where the ECE-1 gene is located (62). Equally important, however, is the possibility that there may be individuals with normally high levels of ECE activity who are at a reduced risk for the disease. A careful analysis of ECE activity in AD and control individuals is necessary to determine the extent of the involvement of this enzyme family in the development of AD.