Metalloendopeptidase EC 3.4.24.15 Is Necessary for Alzheimer’s Amyloid-β Peptide Degradation*

We have investigated the functional relationship between metalloendopeptidase EC 3.4.24.15 (MP24.15) and the amyloid precursor protein involved in Alzheimer’s disease (AD) and discovered that the enzyme promotes Aβ degradation. We show here that conditioned medium (CM) of MP24.15 antisense-transfected SKNMC neuroblastoma has significantly higher levels of Aβ. Furthermore, synthetic-Aβ degradation was increased or decreased following incubation with CM of sense or antisense-transfected cells, respectively. Soluble Aβ1–42 was degraded more slowly than soluble Aβ1–40, while aggregated Aβ1–42 showed almost no degradation. Pretreatment of CM with serine proteinase inhibitors 4-(2-aminoethyl)benzenesulfonyl fluoride and diisopropyl fluorophosphate completely inhibited Aβ degradation. Additionally, α1-antichymotrypsin (ACT), a serpin family inhibitor tightly associated with plaques and elevated in AD brain, blocked up to 60% of Aβ degradation. Interestingly, incubation of CM of MP24.15-overexpressing cells with ACT formed an SDS-resistant ACT complex, suggesting an ACT-serine proteinase interaction. Recombinant MP24.15 alone did not degrade Aβ. 14C-Diisopropyl fluorophosphate-radiolabeled CM from MP24.15-overexpressing cells contained increased levels of several active serine proteinases, suggesting that MP24.15 activates one or more Aβ-degrading serine proteases. Thus, ACT may cause Aβ accumulation by inhibiting an Aβ-degrading enzyme or by direct binding to Aβ, rendering it degradation-resistant. Identification of the Aβ-degrading enzyme and MP24.15’s role in its activation is underway. Pharmacological modulation of either enzyme may provide a means of regulating Aβ in the brain.

Alzheimer's disease (AD) 1 is a progressive neurodegenerative disorder and the most common form of dementia in the elderly (1). Evidence indicates that accumulation of amyloid-␤ (A␤) deposits in senile plaques and in cerebrovasculature is associated with the pathophysiology of AD (2). The A␤ peptide is composed of 40 -42 amino acids (3). The events leading to its formation from the transmembrane amyloid precursor protein (APP) involve proteolytic cleavage by enzymes that have been termed: 1) ␤-secretase, which cleaves at the amino terminus of A␤, and 2) ␥-secretase, which cleaves at the carboxyl terminus. In the senile plaques A␤ is associated with a number of proteins, including ␣ 1 -antichymotrypsin (ACT) (4), which is a serine proteinase inhibitor of the serpin family as well as an acute-phase protein (5). Thus, we hypothesized 10 years ago that ACT may play a role in APP processing (4,6).
Soluble A␤ peptide can be detected in the conditioned media of a variety of cultured mammalian cells in vitro (7)(8)(9), as well as in serum and cerebrospinal fluid in vivo (10). The majority of secreted A␤ is 40 amino acids in length (A␤1-40), but approximately 10% of all A␤ is 42 amino acids in length (A␤1-42) (11). Little is known about how secreted A␤ is degraded and cleared from the extracellular milieu. The excessive cerebral accumulation of A␤ that occurs in AD could be explained in part by a decreased ability of the brain to degrade and clear A␤. If neural cells can be shown to release specific A␤-degrading proteinases, changes in the activity of such proteinases and/or their upregulation could represent a therapeutic approach to AD.
We have explored the role of the metalloendopeptidase MP24.15 in the degradation of A␤. MP24.15 is a thiol-dependent enzyme that was purified to homogeneity from AD brain as a candidate ␤-secretase (12). McDermot et al. (13) also identified a partially purified MP24.15 as a potential ␤-secretase using synthetic peptide substrates. In vitro, MP24.15 has been shown to be involved in the inactivation of a number of neuropeptides, including somatostatin, bradykinin, substance P, and neurotensin (14 -16). The cDNA coding for MP24.15 was subsequently cloned from a human brain library (17,18). In an attempt to further examine the ␤-secretase properties of MP24.15 under physiologic conditions, we transfected human neuroblastoma cells with MP24.15 cDNA in the sense and antisense directions to test its activity on the endogenous, membrane-bound APP. A␤ amounts produced by cells containing the mock-, sense-, or antisense-transfected cDNA were compared. Contrary to our expectations, we observed higher A␤ levels in the conditioned medium (CM) of antisense-transfected cells than in the CM of sense-or mock-transfected cells, while APP levels remained unchanged. These unexpected results prompted us to further investigate MP24. 15. Here we report that MP24.15 is involved in the degradation of A␤ rather than its production.
Previous reports have shown that rat microglia, astrocytes, and human THP-1 monocyte cells are important in the formation (19) or clearance of A␤ and amyloid fibers (20) when A␤ is added to the culture medium, but that rat neurons do not possess this degradative potential (21). Proteinase inhibition studies have also been utilized to determine which class of proteinase is responsible for the degradation of A␤ in the culture media of microglial cells. Mentlein et al. (22) determined * This work was supported by a Zenith Award from the Alzheimer's Association and National Institutes of Health Grants AG09905 and AG00001 and Training Grant NS07152. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  that lipopolysaccharide-activated microglial cells release a 200-kDa metalloproteinase that preferentially degrades soluble, but not polymerized, A␤. Naidu et al. (23) found that the levels of soluble A␤ peptide released to the media of Chinese hamster ovary cells are reduced over time due to the activity of multiple types of proteinases, including those from the metallo, aspartyl, and thiol classes. Banks et al. (24) identified an erythrocyte-derived metalloproteinase that shows a loss of A␤ degradation activity in the presence of aluminum. Other proteinases including the lysosomal enzyme, cathepsin D (25), and extracellular proteinases, including matrix metalloproteinase 2/gelatinase-A (26) have been reported to be involved in the processing of A␤. Insulin degrading enzyme (IDE), a 110-kDa metalloproteinase, has additionally recently been shown to bind to and degrade radioiodinated A␤ peptide (27)(28)(29). This degradation by IDE occurs in rat brain and liver, human brain synaptic membrane fractions, and microglial cells in culture.
Additionally, there have been several reports of serine proteinases that degrade the A␤ peptide. One such 68-kDa proteinase isolated from human brain partially degrades A␤ purified from human brain (30). Another 28-kDa serine proteinase was found in conditioned medium of Chinese hamster ovary cells transfected with APP695. Although this serine proteinase originated from the trypsin cell passing solution, it was shown to complex with ␣ 2 -macroglobulin (␣ 2 M) serum component and to degrade both A␤1-40 and 1-42 (31).
In parallel to the direct proteolysis of A␤, alternative methods of A␤ trafficking and processing have been studied. The serpinenzyme complex receptor was identified in cells for its ability to bind to ACT and other serine proteinase inhibitors, and investigations have also shown that serpin-enzyme complex receptor mediates A␤ internalization and degradation (32). The proteinase inhibitor ␣ 2 M was shown to bind A␤ directly, possibly mediating A␤ uptake by the ␣ 2 M uptake receptor, or low-density lipoprotein receptor-related protein (33,34). The receptor for advanced glycation end products has also been shown to interact with A␤ and it may contribute to neurotoxicity that results in dementia (35). It is likely that different cells utilize different mechanisms for the clearance of A␤ whereas microglial cells may use IDE as their major A␤-degrading enzyme while neuronal cells may use a serine proteinase.
Identification of the proteinases involved in A␤ catabolism therefore is critically important for the development of therapeutics to prevent or treat AD. Our approach has been to follow up on the metalloproteinase MP24.15. Its role in the proteolytic cascade leading to A␤ degradation holds great promise for direct intervention in the pathophysiology of AD.

EXPERIMENTAL PROCEDURES
Cell Cultures and Transfections -SKNMC human neuroblastoma cell line (ATCC, MD) were cultured in minimal Eagle's medium-␣ supplemented with various batches of 10% heat-inactivated fetal bovine serum (FBS). Cells were regularly monitored for growth and viability by trypan blue exclusion assay. The MP24.15 cDNA was obtained from Dr. G. Huber (Hoffman la Roche, Basel, Switzerland (18). A construct containing the full-length MP24.15 cDNA (2551 base pairs) in the pCDNA3.1/Zeo plasmid (Invitrogen, CA) was used for transfection of the SKNMC cells. The cells were plated at 10 6 cells/100-mm dish and allowed to attach for 18 -24 h. The medium was then changed, and the cells were transfected 2 to 4 h later with 20 g of plasmid DNA by the calcium-phosphate method using a Calcium Phosphate Mammalian Cell Transfection Kit (5 Prime 3 3 Prime, Inc., Boulder, CO). The initial incubation of the plates (4 h, 37°C, 5% CO 2 ) was followed by a 15% glycerol shock (3 min at room temperature). After washing with serum-free Dulbecco's modified Eagle's medium, the cells were allowed to recover overnight in minimal Eagle's medium-␣ containing 10% FBS. The following day the cells were split into 5 plates and were incubated with minimal Eagle's medium-␣ containing 10% FBS and 400 g/ml Zeocin (Invitrogen) for selection. Medium was changed every 3-4 days and stably transfected clones were collected. MP24.15 antisense con-structs were cloned into the pCDNA3.1/Zeo plasmid as well. These constructs contain the KpnI/SmaI fragment of the MP24.15 cDNA (1191 base pairs) and were subcloned into the plasmid in the antisense direction. The SKNMC cells were transfected with the antisense construct as above. All transfected cells were grown in Dulbecco's modified Eagle's medium containing 10% FBS and Zeocin. After several passages, the MP24.15 antisense transfected cells were cloned by limiting dilution under Zeocin selective conditions, and 12 clones were collected. Mock-transfected cells were transfected with a pCDNA3.1/Zeo plasmid that lacked a DNA insert.
Western Blot Analysis-MP24.15 overexpressing cells, MP24.15 antisense-transfected cells, or mock-transfected cells were grown to confluence in minimal Eagle's medium-␣ containing 10% FBS in 100-mm dishes. Once confluent, the plates were washed twice with serum-free and phenol-red free Dulbecco's modified Eagle's medium and then incubated for 18 -24 h with the same medium at 37°C (in some experiments, N 2 supplement (Life Technologies, Inc.) was added to the serumfree medium). The following day, the CM was collected and centrifuged at 1500 rpm to pellet the cell debris. The plates were washed twice with ice-cold phosphate-buffered saline, and the cells were harvested using cell scrapers (Nalge Nunc Intl., Naperville, IL) in the presence of 300 l of homogenizing/lysis buffer containing 1 mM EGTA, 1 mM EDTA, 1 mM AEBSF, 1 g/ml leupeptin, and 1% Triton X-100 in phosphate-buffered saline. Cell lysates were sonicated until all visible particular matter disappeared. Protein levels in each of the CM and cell lysate samples were measured using the BCA protein assay reagent kit (Pierce, Rockford, IL). 15-20 g of protein from each sample was electrophoresed on a 10% Tris glycine SDS-polyacrylamide gel (Bio-Rad) or on 10 -20% Tris-Tricine gels (Novex, San-Diego, CA). Gels were then soaked for 20 min in blotting buffer and blotted onto Immobilon-P (Millipore, Bedford, MA) for 2 h at 200 mA for the glycine gels or 2 h at 100 V for the Tris-Tricine gradient gels. After transfer the blots were incubated with 5% milk in TBST (10 mM Tris-HCl, pH 8.0, 150 mM NaCl, and 0.05% Tween 20). Blots were incubated with the specific primary antibody as indicated in each figure. After 3 washes with TBST (15 min each), secondary antibody (Sigma) conjugated to alkaline-phosphatase was applied, and the blots were washed again three times as above. Reactions were developed in the presence of alkaline-phosphatase premixed substrate reagents kit (Bio-Rad). For the detection of MP24.15, we used monoclonal antibody IVD6 generated in our laboratory (36). For the IDE detection, we used the monoclonal antibody that was kindly provided by Dr. Richard A. Roth, Stanford University. For detection of ACT, we used sheep anti-human ACT antibodies (Atlantic Antibodies, Stillwater, MN).
Immunoprecipitation-Conditioned medium supernatants (1-3 ml) were precleared by incubation with 50 l of goat anti-mouse conjugated affinity gel-Sepharose (Cappel, Organon Teknika Corp. West Chester, PA) for 1 h at 4°C on a rocker platform. Samples were centrifuged at 6000 rpm for 5 min, at 4°C, and the precleared supernatants were incubated with the 6E10 (1:300) monoclonal antibody against A␤1-16 (37) for 5 or 18 h on a rocker at 4°C. The goat anti-mouse antibody affinity gel (50 l) was added, and incubation on the rocker continued for an additional 4 h. The immunoprecipitated complex was collected by centrifugation (same as above) and washed three times 20 min each with STEN buffer containing 100 mM Tris, pH 7.6, 300 mM NaCl, 4 mM EDTA, 0.4% Nonidet P-40 as follows: the first wash contained STEN buffer ϩ 0.5 M NaCl; the second wash STEN buffer ϩ 0.1% SDS; and the third wash, STEN buffer alone. All buffers also contained 20 mM phenylmethylsulfonyl fluoride and a proteinase inhibitor mixture (Sigma). The pellet obtained following the last wash and centrifugation was resuspended in 12 l of 2 ϫ loading buffer, boiled for 5 min, and loaded on a 10 -20% gradient Tris-Tricine gel. Gels were then blotted and probed with the 6E10 antibody or other antibodies as indicated in the figures.
Activation of ␣ 2 M with Methylamine-Human ␣ 2 M (Athens Research and Technology, Athens, GA) was incubated at room temperature with 0.23 M methylamine for 110 min to yield methylamine-activated ␣ 2 M. Incubation was performed in a buffer consisting of 0.2 M HEPES, pH 8.0, 50 mM NaCl, and 2 mM EDTA. Excess methylamine was removed by desalting on Sephadex G-25M cartridge column (Amersham Pharmacia Biotech).
A␤ Degradation Assay-Synthetic A␤1-40 and A␤1-42 (Bachem, Torrance, CA) were dissolved in dimethyl sulfoxide at 2 mg/ml (0.5 mM). For radioiodination, 8 g of A␤ were mixed with 90 l of 100 mM Tris-HCl, pH 7.4, 2 l of Na 125 I (0.2 mCi), and one Iodo-Bead (Pierce). The mixture was incubated for 20 min and the iodinated peptide separated from the unbound free iodine on a Sepharose G-25 spin column (5 Prime 3 3 Prime, Inc.). CM (10 l) of MP24.15 sense-, antisense-, or mock-transfected cells were incubated with 10 l (5 M) iodinated A␤1-40 or 1-42 (50 cpm/l) at 37°C for different lengths of time (as specified for each experiment). Following incubation, Laemmli sample buffer was added to the samples, which were then boiled and electrophoresed on a 10 -20% Tris-Tricine gel. The gels were then fixed in 50% methanol, 10% acetic acid solution, dried, and exposed to x-ray film for 1-3 h as needed. When proteinase inhibitors were used in the degradation assay, the CM were incubated for 30 min at 37°C with the different inhibitors prior to the addition of the iodinated A␤ peptide. The following inhibitors were used: 0.15 mM Cpp-Ala-Ala-Phe-pAB-HCl (a specific inhibitor for MP24.15), 2 mM AEBSF (a serine proteinase inhibitor), 2 mM Zincov (a zinc metalloproteinase inhibitor), 10 mM 1,10-phenanthroline (a metalloproteinase inhibitor), 1.5 M purified ACT (a serine proteinase inhibitor) (Athens Research and Technology, Athens, GA). In some experiments radiolabeled A␤ was left to aggregate at room temperature for 1 week prior to the addition of CM. Afterward, 10 l of CM was added, the samples were incubated in a 37°C in a water bath for 5-18 h. Sample buffer was added, and the samples were separated on 10 -20% Tris-Tricine gels as above. The gels were dried and exposed to x-ray film. Autoradiographs were analyzed by computerassisted densitometry from which percent inhibition was calculated for each inhibitor. For each CM, the percent of the undegraded A␤ was calculated with and without inhibitor.
14 C-DFP Labeling-CM (9 l) from the various transfected cells was incubated with 1.5 l of 1 M Tris-HCl buffer, pH 7.4, for 5 min in a 37°C water bath. Then, 1 l of 14 C-diisopropyl fluorophosphate (DFP, NEN Life Science Products Inc.) was added, and the samples incubated at 37°C for 1 h. Laemmli sample buffer was then added and the samples were boiled and separated on 10% SDS-PAGE. Gels were fixed for 40 min in 50% methanol, 10% acetic acid, treated for 30 min at room temperature with Enlightning solution (NEN Life Science Products Inc.), and dried. The dried gels were then exposed to x-ray film (Kodak) to view the DFP-labeled bands.
ACT  (Fig. 3A). A␤ incubated with Tris buffer or with medium containing 10% heat-inactivated fetal bovine serum, as a control, is not detectably degraded during the 32-h incubation period (Fig. 3A). The difference in A␤ degradation rates of CM of MP24.15-overexpressing cells and CM of mock-transfected cells is more clearly demonstrated in the shorter time course experiment described in Fig. 3B.
Serum-free CM did not exhibit any A␤-degrading ability (data not shown), but in the presence of serum the amount of A␤ degradation directly correlates with MP24.15 levels. The effect of the serum was independent of the lot used. Comparison of the degradation rates of A␤1-40 and 1-42 clearly demonstrates that A␤1-42 catabolism is slower than that of A␤1-40 under the same conditions (Fig. 3A). Additionally, A␤1-42 partially aggregates during the incubation time in 10 mM Tris buffer and appears as a high molecular weight band in the lanes that contain A␤1-42 alone in Tris buffer, but not in those lanes containing medium with 10% serum as control (data not shown). Interestingly, in all lanes where A␤ degradation occurs an approximately 30-kDa band is seen, suggesting that iodinated A␤ binds to a CM protein that may represent an A␤-degrading serine proteinase.
The Inhibitory Profile of the A␤ Degrading Activities-To further characterize the A␤ degrading activity found in the SKNMC CM, we examined the effects of different proteinase inhibitors using the A␤ degradation assay described above.
Using concentrations known to inhibit other proteinases, we incubated the various inhibitors with the CM of the sense-, antisense-, or mock-transfected cells for 30 min at 37°C prior to the addition of the iodinated A␤ peptide for an additional incubation of 3 h. As seen in Table I, the inhibitory profile clearly demonstrates that both a metalloproteinase and a serine proteinase are involved in A␤ degradation. DFP and AEBSF, two serine proteinase inhibitors, completely inhibited the A␤ degradation. ACT, a serine proteinase inhibitor of the serpin family, inhibited 46 -60% of the A␤ degrading activity. The inhibitor Cpp-Ala-Ala-Phe-pAB, which is a non-permeable specific inhibitor of MP24.15, had a partial inhibitory effect on the A␤ peptide degradation, while Zincov had only a slight inhibitory effect. 1,10-Phenanthroline (4 mM) and insulin (1-10 M) which are both known to completely inhibit the IDE (29,38,39), had low inhibitory effects on A␤ degradation (Table I).
Serum-free Medium Does Not Degrade A␤-To test whether recombinant human MP24.15 (active against the specific fluorescent substrate Mcc-Pro-Leu-Gly-Pro-D-Lys(Dnp)-OH) can degrade radiolabeled A␤, we mixed recombinant MP24.15 with 125 I-A␤. No A␤ degradation was observed (data not shown). We observed that CM of sense-transfected cells grown in serumfree medium, with or without the N 2 supplement, exhibit no A␤ degrading activity. Our trial to reconstitute the A␤ degradation activity in serum-free CM with recombinant MP24.15 was un- successful, indicating that serum is needed in addition to MP24.15 for A␤ degradation to occur (data not shown).
Dependence on one or more serum factors was also observed and reported earlier by Qiu et al. (31), who described A␤ degradation from Chinese hamster ovary cells by a serine proteinase complexed with ␣ 2 M. We tested whether the factor missing in the serum-free medium was ␣ 2 M by adding methylamineactivated ␣ 2 M to our serum-free CM. In our system no A␤ degrading activity was present in the serum-free CM regardless of the presence of ␣ 2 M (data not shown). Altogether, these results indicate that a serine proteinase and MP24.15 must be present for A␤ degradation to occur in SKNMC cells, but that MP24.15 does not directly degrade A␤.

IDE Is Not a Major A␤-degrading Enzyme in SKNMC Cells-
Recently it has been reported that IDE is the enzyme responsible for A␤ degradation by microglial cell cultures. In order to determine whether IDE is present in our SKNMC cells and their CM, and to determine whether it is involved in A␤ degradation, cell lysates and CM of the sense-, antisense-, and mock-transfected cells were analyzed by Western blot using monoclonal antibodies against IDE. As shown in Fig. 4, although there is a significant difference in the amount of MP24.15 between MP24.15 sense-, antisense-, and mock-transfected cells, all samples contained comparable amounts of IDE in both their cell extracts and CM. Thus, MP24.15, but not IDE, levels correlated with A␤ degrading ability. Furthermore, insulin inhibited A␤ degradation completely only at 100 -200 M (Table I). This concentration is 20 times higher than that of 125 I-A␤ (5 M) in the assay, and thus this effect is likely nonspecific. Insulin at 1 or 10 M, which is known to completely inhibit IDE activity, was only a weak inhibitor of A␤ degradation by the neuroblastoma CM (Table I).
Levels of MP24. 15 and IDE in HEK293 and IMR-32 Cells-We analyzed by Western blot analysis the levels of MP24.15 and IDE expression in CM and cell lysates of two other non-transfected cell lines, human embryonal kidney (HEK293) and human neuroblastoma (IMR-32), in order to compare their endogenous expression of MP24.15 and their ability to degrade 125 I-A␤ peptide with the corresponding properties of SKNMC. We found that HEK293 and IMR-32 cells produce equivalent levels of MP24.15 and IDE when compared with SKNMC cells (data not shown). The cell extracts possess the majority of the MP24.15 and IDE, but each of the CM shows detectable levels of these enzymes as well. Furthermore, the A␤ degrading activity of these cells was similar to that of SKNMC cells (data not shown).
Degradation of Aggregated A␤ by CM of MP24. 15 Transfected Cells-We wished to test the A␤ degrading activity of the various CM on pre-aggregated A␤. 125 I-A␤1-40 and 1-42 were aggregated for 7 days at room temperature before incubation with CM of the MP24.15 sense-or mock-transfected cells for 4 or 18 h. Mixtures were then separated on 10 -20% Tris-Tricine gels. Samples of the pre-aggregated 125 I-A␤1-40 and 1-42 were spun at maximum speed in an Eppendorf centrifuge, the supernatant separated from the pellet, and both the pellets (Fig. 5, lanes 7) and the supernatants (lanes 8) were run on the gels. A␤1-40 had smaller soluble aggregates and most of the radioactivity stayed in the supernatant, while most of the A␤1-42 radioactivity was in the pellet and was seen on the gel as high molecular weight aggregates. Consistent with previous results by others, and as demonstrated in the autoradiograph in Fig. 5, A␤1-42 had higher aggregation ability than A␤1-40 during this period of time, and was much more resistant to degradation. A␤1-40 exhibited almost no aggregation and was consequently degraded by the CM.
Characterization of the Serine Proteinase That Degrades A␤ Peptide in SKNMC Neuroblastoma Cell Medium-We performed 14 C-DFP labeling experiments in order to identify active serine proteinases in the CM. We have repeatedly found that 14 C-DFP strongly labels a band of 26 -28 kDa and several high molecular mass bands at about 200 to 230 kDa and higher in serum-containing CM of MP24.15 overexpressing cells (Fig.  6). The same bands are labeled in the mock-and antisensetransfected cell CM, but to a much lesser extent. Thus, there is a direct correlation between the amount of MP24.15 that is expressed in the cells and found in CM and the intensity of the 14 C-DFP-labeled serine proteinases (Fig. 6). Serum-free CM of mock-, sense-, and antisense MP24.15-transfected cells did not contain any 14 C-DFP-labeled serine proteinases (data not shown). In addition, we found the same pattern of 14 C-DFP labeling in CM of non-transfected SKNMC, HEK293, and IMR-32 cells (data not shown).

ACT Inhibits A␤ Degradation and Forms an SDS-resistant Complex with a Proteinase in the CM of MP24.15-transfected
Cells-As mentioned above (Table I) we found that the serine proteinase inhibitors AEBSF and DFP completely inhibit A␤ degradation by neuroblastoma CM. We therefore tested ACT, a serine proteinase inhibitor whose levels we previously found to be elevated in AD brain, for a possible inhibitory role in A␤ degradation. Fig. 7 demonstrates that A␤ degrading activity by CM of both mock-and MP24.15-transfected cells was partially inhibited by 5 M ACT (Fig. 7, lanes 3 and 4). Since ACT is a serine proteinase inhibitor from the serpin family and members of the serpin family form SDS-insoluble complexes with their serine proteinases, we examined the ability of CM to form complexes with ACT. These complexes can be detected by Western blot analysis using antibodies to ACT. We incubated CM of non-transfected, mock-transfected, and MP24.15-transfected cells with 0.75 M ACT for 30 min at 37°C. The samples were then boiled in Laemmli sample buffer and separated on a 7.5% gel. ACT and ACT-proteinase complexes are demonstrated on the blot with anti-human ACT antibodies. As a positive control, a sample of ACT was incubated in parallel with cathepsin G, a a Note that 1 mM 1,10-phenanthroline is known to completely inhibit IDE, but in our system 4 mM 1,10-phenanthroline inhibits only 25-30% of the A␤ degrading activity.
b Also, 1 M insulin has been shown to completely inhibit IDE, while in our system 1 M insulin had no effect at all and 10 M insulin only partially inhibited A␤ degradation. Insulin at 100 M is 20 times more concentrated than the 125 I-A␤ used in the assay (5 M) and the inhibition is believed to be nonspecific. known substrate for ACT inhibition, at a 1:1 molar ratio. Fig. 8 demonstrates an ACT complex of approximately 100 kDa formed with a factor from CM of MP24.15 sense-transfected cells (panel A, lane 4 and panel B, lanes 3 and 4) and a complex of 90 kDa formed with cathepsin G (panel B, lane 2). In order to compare the ACT-cathepsin G complex with the ACT⅐CM proteinase complex, we added ACT and cathepsin G to the CM of mock-or MP24.15 sense-transfected cells. Two different ACT complexes are observed. An ACT-cathepsin G complex of 90 kDa is seen in panel B, lane 2, which contains medium, serum, ACT, and cathepsin G . In panel B, lanes 3 and 4, which contain ACT, cathepsin G, and CM of mock-and sense-transfected cells, respectively, the amount of the 90-kDa ACT-cathepsin G complex is lower, and a new complex of approximately 100 kDa is seen . In panel B, lane 4, where the level of MP24.15 is higher than that in lane 3, the level of ACT-CM proteinase is also higher. Note that as the ACT-cathepsin G complex diminishes in panel B from lanes 2 to 4, the ACT⅐CM proteinase complex increases, suggesting that the CM proteinase competes with cathepsin G for binding to ACT.  lanes 3 and 4) is correlated with more ACT-CM proteinase at 100 kDa and less ACT-cathepsin G complex at 90 kDa, suggesting that cathepsin G competes with the CM serine proteinase for binding to ACT.

DISCUSSION
Excessive deposition of A␤, the main constituent of the extracellular amyloid plaques, is an early pathological hallmark of AD. Understanding the normal clearance mechanism for secreted A␤ and finding methods to accelerate this process could be extremely beneficial for the prevention, delay, or treatment of AD. The data presented in this article demonstrate that the metalloendopeptidase EC 3.4.24.15 (MP24.15) is a necessary element in the pathway that leads to the degradation of A␤ by human neuron-like cells. Our results show a direct correlation between the expression levels of MP24.15 of healthy SKNMC cells and the degradation of naturally produced or radiolabeled A␤ by their CM. Media from cells overexpressing MP24.15 degrade A␤ more rapidly than CM of mock-transfected cells, which contain only the endogenous MP24.15. Antisense-transfected cells express only very low levels of MP24.15 and, likewise exhibit the lowest levels of A␤ degradation. This correlation is striking in light of the fact that the concentration of this enzyme declines with age. In addition, MP24.15 levels are decreased further in brain regions with the greatest accumulation of A␤. In AD this decline is even more pronounced (40).
MP24.15 was previously examined as a candidate ␤-secretase by several groups after they discovered that this neuropeptidase cleaves test peptides containing the ␤ secretase cleavage site (12,13,41). The most rigorous of these studies showed that neither the overexpression of MP24.15 nor the use of specific inhibitors dramatically altered A␤ levels or APP processing in these cells, as expected of a ␤-secretase (41)(42)(43). In contrast, our own studies showed that MP24.15 does generate amyloidogenic fragments from recombinant APP (12); however, as we have shown here, there is no obvious or dramatic increase in A␤ production when cells genetically altered to produce high levels of the MP24.15 were analyzed. On the contrary, we were surprised to find that reducing the levels of MP24.15 using antisense techniques produced an acute accumulation of soluble A␤ in the conditioned medium. Given this new discovery, it seemed likely that MP24.15 was involved in the degradation of A␤ rather than in its generation.
The distribution of neuronal MP24.15 in the brain is widespread and enriched in areas displaying high levels of its known substrates (i.e. substance P, etc.) (44). Neuronal loss in AD may be responsible for the observed decreased activity of MP24.15. A continuum of work on the physiologic role for MP24.15 has generated a wealth of data with regard to its function in the central nervous system and to its level of expression in specific brain regions. Several of these studies have noted that the level of MP24.15 expression in the temporal cortex decreases with age and that neuropeptides, such as substance P or somatostatin, two substrates of MP24.15, decrease as well in aging brain tissues (44 -46). Additionally, when AD patients were compared with age-matched controls that lacked any overt neuropathology or psychiatric illness, their MP24.15 levels were found to be drastically reduced. The diminished levels were especially striking in the temporal and parietal cortices where A␤ deposits are heaviest. We hypothesize that these reductions in MP24.15 activity, especially in AD brains, lead to a loss of capacity to remove soluble A␤, to higher levels of soluble A␤, A␤ aggregation, and eventually more plaque formation.
There have been a number of proteinases that have been implicated in A␤ degradation, but the IDE has emerged as a strong candidate. Degradation of A␤ by IDE occurs in rat or human membrane preparations from brain homogenates and in a variety of cell types, but not neurons (27)(28)(29). We have found in our system that IDE is not the major contributor to A␤ degradation. We observe no correlation between the level of IDE present in cell lysates or CM and the rate of A␤ degradation. We also report that neither insulin (at 1 M or 10 M) nor 1,10-phenanthroline (at 1 mM), both effective inhibitors of IDE (38,47), do not substantially block A␤ degradation. Only insulin at 100 M completely inhibited A␤ degradation by the SKNMC cells CM (Table I). This concentration of insulin is 20 times higher than 125 I-A␤ in the assay (5 M) and may represent nonspecific inhibition. Similarly, 4 mM 1,10-phenanthroline inhibited only 25-30% of the A␤ degradation by CM of SKNMC neuroblastoma cells. While our serum-free CM contains both MP24.15 and IDE there is neither any A␤ degrading activity nor 14 C-DFP-labeled (i.e. active) serine proteinases. This contrasts again with Qiu et al. (29), who report that microglial IDE degrades A␤ in serum-free medium. The dependence on serum in SKNMC cells further distinguishes these two activities.
In exploring the inhibitor profile of SKNMC CM we suggest that MP24.15 is required for the activation of an A␤-degrading serine proteinase(s). The A␤ degradation by CM of MP24.15 sense-transfected or mock-transfected cells is completely inhibited by AEBSF and DFP (Table I), while Zincov, 1,10-phenanthroline, insulin, and the MP24.15 inhibitor CPP-Ala-Ala-Phe-pAB only partially inhibit A␤ degradation. The inhibitory effect of these metalloproteinase inhibitors was low (between 7 and 30%) in comparison to AEBSF or DFP, which demonstrate 90 -100% inhibition. Since serine proteinase and A␤ degrading activity correlate with the MP24.15 expression level, these results indicate that the main A␤ degrading activity in these cells involves one or more serine proteinases that are activated or induced by MP24.15. Autoradiographs of the A␤ degradation assay show a distinct band at 28 kDa which may be the serine proteinase responsible for the degradation of A␤ (data not shown). This band suggests an acyl-intermediate complex between the 125 I-labeled A␤ and the proteinase that degrades it. This band is clearly distinct from aggregates of A␤ and is only seen in lanes that show A␤ degradation.
We have previously found that ACT mRNA is highly elevated in AD brain and that ACT protein is tightly associated with A␤ in both diffuse and classical plaques (4). Fitting well with our serine proteinase inhibitory profile, we report here that ACT also inhibits A␤ degradation by CM of mock-or MP24.15transfected cells. This inhibition of A␤ degradation may occur by two possible mechanisms. First, by binding to the active site of the serine proteinase, ACT may block the proteinase from degrading A␤. We demonstrate the formation of an SDS-stable ACT-serine proteinase complex in the CM of the MP24.15overexpressing cells (Fig. 8) (which possesses reduced A␤-degrading activity) (Fig. 7). Although the identity of this ACTinhibited A␤-degrading serine proteinase has not yet been determined, preliminary results from the 14 C-DFP labeling experiment (Fig. 6) and the ACT complex formation (Fig. 8) support a molecular mass of 26 -28 kDa. The formation of an ACT-proteinase complex, that migrates at 100 kDa, suggests that this proteinase does not excise the 8-kDa reactive site loop of ACT under these conditions. In contrast, the ACT-cathepsin G complex migrates at 90 kDa, as expected. The second mechanism for ACT's inhibition of A␤ degradation would be through direct binding to A␤ peptide. We have shown that ACT binds A␤ to form an SDS-stable complex rendering ACT a weaker inhibitor of chymotrypsin (48). Recently the binding of A␤ to ACT was described in structural detail (49). Binding of ACT to A␤ could render A␤ a poorer substrate for degradation. We have shown that when both ACT and cathepsin G are added to either CM of mock-or MP24.15-transfected cells, two ACT complexes are formed, one of 90 kDa and one of 100 kDa (Fig.  8). Because the amounts of these two ACT complexes are in-versely correlated it is suggested that the activated serine protease in the CM binds to ACT with a higher affinity than cathepsin G does.
It should also be noted that preliminary measurements of A␤ deposits in doubly transgenic mice expressing human ACT and human APP demonstrate a higher amyloid load than that of human APP expressing singly transgenic mice of similar age. 2 These in vivo results support the in vitro findings reported here that ACT inhibits A␤ catabolism.
The data presented in Fig. 6 demonstrate the direct relationship among MP24.15 expression levels, the concentration of 14 C-DFP-inhibitable active serine proteinase(s), and A␤ degradation. The higher the MP24.15 expression, the more labeling by 14 C-DFP of an active serine proteinase in the CM. More importantly, the higher the level of MP24.15, the greater the degree of A␤ degradation by the CM. In addition, we show that although CM of mock-and MP24.15-overexpressing cells degrade soluble A␤1-40 and 1-42 efficiently, the degradation of aggregated A␤1-42 (but not aggregated A␤1-40) is markedly reduced (Fig.  5). However, CM of MP24.15 overexpressing cells degrades aggregated A␤1-42 faster than CM of mock-transfected cells, suggesting that pharmacological up-regulation of MP24.15 could be an exciting target for the therapeutic reduction of A␤.
Qui et al. (31) reported that A␤ degradation by HEK 293 cells is caused by a chymotrypsin-like activity from the trypsin solution used to pass the cells in culture. When complexed with the proteinase inhibitor ␣ 2 M, this serine proteinase degrades A␤. We observe strong 14 C-DFP radiolabeling of a serine proteinase at 26 -28 kDa, as well as several high molecular weight bands above 200 kDa (Fig. 6). Since ␣ 2 M appears to play a role in the stabilization of the A␤ degrading activity observed by Qui et al. (31), we added methylamine-activated and native ␣ 2 M to the conditioned serum-free medium of the SKNMC cells, but the A␤ degradation activity was not restored. We conclude that in addition to MP24.15, a serum factor other than ␣ 2 M is important for the activation or induction of the serine proteinase that is degrading A␤.
It has been shown that MP24.15 is not capable of degrading large proteins such as albumin, casein, or gelatin (14,50,51). However, such studies do not preclude the likely possibility of an as yet unidentified protein substrate of greater size and specificity. Therefore, in the A␤-degradation pathway MP24.15 may activate a serine proteinase by: 1) activating the proteinase by proteolytic degradation of its zymogen precursor or 2) degrading its endogenous peptide inhibitor. Here we propose that MP24.15 is necessary as part of a proteolytic cascade for the induction or the activation of one or more serine proteinases. We believe that this MP24.14/serine proteinase axis represents the main A␤-degradation pathway in these cells and that other enzymes (i.e. metalloproteinases) have a lesser contribution to A␤ degradation. We are now pursuing the identification of the serine proteinase and the activation mechanism by MP24.15 in order to better understand the cascade of events leading to A␤ degradation by neuronal cells. Such understanding may lead to the design of pharmaceutical agents that can enhance A␤ degradation.