The carboxyl termini of beta-amyloid peptides 1-40 and 1-42 are generated by distinct gamma-secretase activities.

We have studied the effects of peptide aldehyde protease inhibitors on the secretion of β-amyloid peptide 1-40 (Aβ(1-40)) and Aβ(1-42) by HEK 293 and COS-1 cells expressing β-amyloid precursor protein with the Swedish double mutation. A multiphasic SDS-polyacrylamide gel electrophoresis system was used for the discrimination of Aβ(1-40) and Aβ(1-42). Calpain inhibitor I, carbobenzoxyl-Leu-Leu-leucinal, and calpeptin were found to reduce the amount of Aβ(1-40) released into the medium in a dose-dependent manner. The reduction of Aβ(1-40) after treatment with 50 μM calpain inhibitor I or 5 μM carbobenzoxyl-Leu-Leu-leucinal was accompanied by a slight increase of Aβ(1-42) released into the medium. These observations suggest that the cleavages at residues 40 and 42 are accomplished by different enzyme activities.

Characteristic pathological findings in the brains of Alzheimer's disease (AD) 1 patients include amyloid deposits containing the ␤-amyloid peptide (A␤) as a major component. A␤ is a proteolytic fragment of the ␤-amyloid precursor protein (APP), a large transmembrane glycoprotein with a single membranespanning region that exists in different isoforms (1). Proteolytic cleavage of APP just outside the transmembrane region within the A␤ sequence (2) by an unknown enzyme called ␣-secretase releases the soluble ectodomain of APP (3). This cleavage precludes A␤ production. Soluble A␤ peptides are generated by an alternative proteolytic processing pathway during normal metabolism of cells expressing APP (4 -6). They can also be detected in cerebrospinal fluid (7). A double mutation within the APP gene (NL670/671) found in a Swedish early onset AD family (8) increases A␤ production (9 -12). The enzymes cleaving APP to produce the amino and carboxyl termini of A␤ are referred to as ␤and ␥-secretases, respectively (13). None of these secretases has yet been identified.
Analyses of the A␤ peptides in cerebrospinal fluid (14), in conditioned media from HEK 293 cells transfected with an APP cDNA containing the Swedish double mutation (15) and from IMR 32 cells (16) revealed a series of A␤ peptides of various lengths. A minor portion of those has a longer carboxyl terminus and ends at amino acid residue 42 or 43 of the A␤ sequence. Point mutations located in codon 717 of the APP gene (APP770 numbering) found in some familial forms of AD increase the proportion of the longer A␤ peptides generated by transfected cells (17,18). The longer peptides aggregate more rapidly (19) and were reported to be deposited preferentially in amyloid plaques (20 -22) and at an early stage (23). Taken together, these findings strongly suggest that the peptides ending at residue 42 or 43 are critical for amyloidogenesis in AD.
Recent findings indicate that A␤ production can be reduced by inhibition of ␥-secretase cleavage of APP. Higaki et al. (24) observed inhibition of A␤ formation with carbobenzoxyl-Val-Phe-alaninal in Chinese hamster ovary cells transfected with a wild type APP cDNA. We previously reported the reduction of A␤ secretion from HEK 293 cells expressing APP with the Swedish double mutation after treatment with N-acetyl-Leu-Leu-norleucinal (calpain inhibitor I) (25). Since no discrimination between A␤(1-40) and A␤(1-42) was accomplished in these studies, the observations presumably reflect the effects of the peptide aldehydes on the major A␤ species. In this study we investigated the influence of three related peptide aldehydes, calpain inhibitor I, carbobenzoxyl-Leu-Leu-leucinal (MG 132) and carbobenzoxyl-Leu-norleucinal (calpeptin) on the secretion of A␤(1-40) and A␤(1-42) from HEK 293 and COS-1 cells expressing APP with the Swedish double mutation. For the discrimination of A␤(1-42) from A␤(1-40), we used Bicine/Tris SDS-polyacrylamide gels containing 8 M urea (26) and an antiserum specific for the carboxyl terminus of A␤(1-42).

Construction of the Expression Vector pCMVAPP751(SWE)-The
Swedish double mutation NL670/671 (APP770 numbering) was introduced into wild type APP751 cDNA, and an optimal Kozak consensus sequence (GCC GCC ATG G) (27) was generated at the initiator codon as described (28). The APP751(SWE) cDNA was subsequently cloned into an expression vector with the cytomegalovirus promoter.
Cell Culture, Transfections, and Pulse-Chase Experiments-HEK 293 cells were routinely grown in 10-cm plates in minimum essential medium (Life Technologies, Inc.) supplemented with 10% fetal calf serum (Life Technologies), COS-1 cells were cultivated in Dulbecco's modified Eagle's medium and 10% fetal calf serum. Prior to experiments, penicillin and streptomycin were added to final concentrations of 100 IU/ml and 100 g/ml, respectively. HEK 293 cells were transiently transfected with calcium phosphate DNA precipitates formed in BES (30). In most experiments, the cells were replated 16 -20 h after addition of the DNA into 60-mm dishes. COS-1 cells were transfected with Transfectam (Promega) according to the manufacturer's instructions. 40 -44 h after transfection the cells were pulse labeled for 60 min at 37°C with 0.1 mCi/ml [ 35 S]methionine in methionine-free Dulbecco's modified Eagle's medium (Sigma), 100 IU/ml penicillin, 100 g/ml streptomycin, 20 mM HEPES, pH 7.4, and 2% dialyzed fetal calf serum and chased for 3 h in the presence of the indicated compounds with medium containing 10% fetal calf serum, penicillin, streptomycin, and an excess of cold methionine. Immunoprecipitations were performed as * 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 U.S.C. Section 1734 solely to indicate this fact.
Immunoprecipitates were separated on Bicine/Tris SDS gels containing 8 M urea or 17.4% glycerol (26). Gels were fixed, stained with Coomassie Blue, impregnated with Amplify fluorographic reagent (Amersham Corp.), dried, and analyzed by fluorography using Kodak Biomax MR films. Quantification of the signals was performed with a computerized image analysis system (Microcomputer Imaging Device, Imaging Research Inc., St. Catherines, Ontario, Canada). After digitalization, areas of interest were manually outlined and calculated into relative absorbance values. The experiment shown in Fig. 2 was also quantified on a PhosphorImager (Molecular Dynamics). This was not possible for the other experiments, since the sensitivity of the Phosphor-Imager was not sufficient for the quantification of A␤(1-42).
Calpain inhibitor I was obtained from Boehringer Mannheim, MG 132 (31) and calpeptin (32) were synthesized by the following sequence. First the corresponding N-benzyloxycarbonyl-protected di-and tripeptide methylesters were synthesized by standard coupling techniques, and then these were reduced at the carboxyl terminus to the corresponding alcohols using 1.2 eq lithium chloride/1.2 eq sodium borohydride in a mixture of ethanol/tetrahydrofuran (4:3) at room temperature. In the last step the corresponding alcohols were oxidized to the corresponding aldehydes with 1.5 eq of 1-hydroxy-1,3-dioxo-3Hbenzo[d] [1,2]iodoxol-3-one in dimethylsulfoxide at room temperature. Stock solutions (25 mM) were prepared in N,N-dimethylformamide (DMF) and stored at Ϫ20°C.

RESULTS
Characterization of the Antibodies-HEK 293 cells were transiently transfected with APP751(SWE) and metabolically labeled overnight with 0.1 mCi/ml [ 35 S]methionine. The conditioned media were collected and subjected to immunoprecipitation with 4 l of the polyclonal rabbit antiserum NT11 (raised against A␤(1-40)), 5 g of the mouse monoclonal antibody ␤1 (raised against A␤(1-40)), or 8 l of the rabbit antiserum 42-14 (raised against A␤(36 -42) and preabsorbed with synthetic A␤ ). The samples were separated by 15% T, 5% C Bicine/Tris/urea SDS-polyacrylamide gel electrophoresis (Fig. 1A). A␤(1-42) migrates faster on these gels than A␤(1-40) ( Fig. 1B; Ref. 26). The rabbit antiserum NT11 precipitated radiolabeled peptides comigrating with synthetic peptides corresponding to A␤(1-40), A␤(1-42), the p3 peptide, and peptides moving slightly slower than A␤(1-40). The monoclonal antibody ␤1, which recognizes an epitope located within resi-dues 1-16 of A␤ (29), precipitated peptides comigrating with A␤(1-40) and A␤ . In addition, the slower moving peptides were recognized. Immunoprecipitation with the preabsorbed antiserum 42-14 revealed a single peptide band, comigrating with the synthetic A␤(1-42) peptide. HEK 293 cells expressing the Swedish mutant APP secrete primarily A␤ peptides starting at Asp-1 (11,15). Therefore, we conclude that the p3 secretion in a dose-dependent manner ( Fig. 2A). In agreement with our previous observations (25), quantification of the experiment shown on a PhosphorImager indicated that 50 M calpain inhibitor I reduced the amount of A␤ released into the medium by approximately 45% (Fig. 2A). MG 132 and calpeptin were more potent (Fig. 2B). At concentrations of 5, 25, and 50 M, calpeptin reduced A␤ secretion by approximately 45, 65, and 84%, respectively. MG 132 decreased the amount of radiolabeled A␤ by approximately 39% at a concentration of 5 M and by approximately 87% at a concentration of 25 M and almost completely blocked A␤ secretion at 50 M. Similar values were obtained by quantification of fluorographs corresponding to Fig. 2, A and B, with a computerized image analysis system (Table I). When A␤ secretion was studied with stock solutions of calpeptin or MG 132 that had been stored for a prolonged time at Ϫ20°C, comparable but smaller effects were observed, indicating that the peptide aldehydes are not stable under these conditions (data not shown).

The Secretion of A␤(1-40) and A␤(1-42) Is Affected Differently by the Peptide Aldehydes Calpain Inhibitor I and MG
132-To study the effects of the peptide aldehydes in more detail, secreted A␤ peptides were analyzed on 15% T, 5% C Bicine/Tris gels containing 8 M urea (Fig. 3). A␤(1-40) and A␤(1-42) can be separated under these conditions. We observed a reduction of A␤(1-40) after treatment with 5-50 M calpain inhibitor I but a simultaneous increase of A␤(1-42) ( Fig. 3A and Table I). In a similar fashion, incubation with 5 M MG 132 caused a marked decrease of A␤(1-40), accompanied by an increase of A␤(1-42). (Fig. 3B and Table I To confirm that the peptide species we assigned as A␤(1-42) contains the longer carboxyl terminus, transfected cells were pulse labeled and chased for 3 h in the presence of calpain inhibitor I or MG 132 in 100-mm dishes. The conditioned media were split into two aliquots, and these were subjected to immunoprecipitation with 5 g of the monoclonal antibody ␤1 and 10 l of the preabsorbed antiserum 42-14, respectively (Fig. 4). Again, immunoprecipitation with ␤1 revealed that 50 M calpain inhibitor I and 5 M MG132 decreased the amount of A␤(1-40) and simultaneously increased a peptide comigrating with A␤(1-42) ( Fig. 4A and Table 1). Following immunoprecipitation with the 42-14 antiserum only this particular peptide comigrating with synthetic A␤(1-42) was detected (Fig.  4B), indicating that it contains the longer carboxyl terminus. We also tested the effect of MG 132 on cells transfected with a Swedish mutant APP695 cDNA, which encodes APP lacking the Kunitz-type protease inhibitor domain (1). Similar results were obtained with this APP isoform (data not shown).
Comparable effects were observed when radiolabeled A␤ peptides were precipitated with the ␤1 antibody from conditioned media of COS-1 cells expressing APP751(SWE) (Fig. 5 and Table 1). A␤(1-40) secretion was reduced by 50 M calpain inhibitor I, 5-25 M MG 132, and 25 M calpeptin. Again, 50 M calpain inhibitor I and 5 M MG 132 decreased the main A␤ species, whereas they simultaneously increased a peptide species moving slightly faster on Bicine/Tris/urea gels. DISCUSSION One proposed potential therapeutic strategy for preventing or slowing the pathogenetic mechanism in AD involves the reduction of A␤ generation by partial inhibition of ␤and/or ␥-secretase (1). Previous results from our group (25) and from others (24) showed that A␤ secretion from transfected cells can be reduced with certain cell-penetrating peptide aldehydes that were found to inhibit ␥-secretase activity. In the present study, we confirmed the reduction of A␤ and p3 secretion from transfected HEK 293 cells overexpressing APP carrying the Swedish double mutation by calpain inhibitor I. In addition, we present two related peptide aldehydes, MG 132 and calpeptin, that are  more potent. These observations about the secretion of total A␤, however, primarily reflect the influences of the inhibitors on the generation of the main A␤ species, which is A␤  in the case of HEK 293 cells expressing the Swedish mutant APP (15). Several lines of evidence suggest that A␤ peptides ending at residue 42 or 43 are of particular importance in AD: (i) missense mutations in codon 717 of the APP gene responsible for rare familial forms of the disease increase the proportion of the longer peptides (17,18); (ii) in vitro experiments indicate a faster rate of polymerization for peptides including the critical additional residues Ile-41 and Ala-42 (19); and (iii) the longer A␤ peptides were reported to present a major component of senile plaques (20 -22) and were shown to be initially deposited (23).
In view of these observations we investigated whether selected peptide aldehyde inhibitors affect the secretion of A␤ peptides (1-40) and (1-42) in the same way. A␤ peptides generated by HEK 293 cells expressing the Swedish mutant APP primarily start with Asp-1 (11,15). To discriminate between the peptides ending at Val-40 or Ala-42 we used Bicine/Tris gels containing 8 M urea (26). A␤(1-42) was identified by comigration with synthetic A␤(1-42) on these gels and by immunoprecipitation with a rabbit antiserum raised against A␤(36 -42) and preabsorbed with synthetic A␤ . Our data show that the secretion of A␤(1-40) and A␤(1-42) is differentially affected by calpain inhibitor I. Although 50 M calpain inhibitor I clearly reduced the amount of immunoprecipitable A␤(1-40) released into the medium, A␤(1-42) was increased. A different peptide aldehyde, MG 132, had similar effects but was more potent. At 5 M, MG 132 clearly decreased A␤(1-40) and simultaneously increased A␤(1-42) (or at least did not alter the amount of A␤(1-42)). At higher concentrations (25-50 M), MG 132 reduced both A␤(1-40) and A␤ . Comparable findings were obtained when radiolabeled A␤ peptides secreted from COS-1 cells and immunoprecipitated with the ␤1 antibody were analyzed.
Taken together, our results indicate that the cleavages of APP at A␤ residues 40 and 42 exhibit different sensitivities toward calpain inhibitor I and MG 132. We conclude that the carboxyl termini of A␤(1-40) and A␤(1-42) are generated by at least two different proteolytic activities, a ␥(40)-secretase and a ␥(42)-secretase. Future experiments could be directed toward the development of inhibitors that are specific for the ␥(42)secretase activity cleaving at Ala. In view of the accumulating data indicating a critical role for the longer A␤ peptides in AD, we consider this secretase activity the most attractive target among the enzymes involved in A␤ formation.