The Protease Inhibitor, MG132, Blocks Maturation of the Amyloid Precursor Protein Swedish Mutant Preventing Cleavage by b -Secretase*

Amyloid (A b ) peptides found aggregated into plaques in Alzheimer’s disease are derived from the sequential cleavage of the amyloid precursor protein (APP) first by b - and then by g -secretases. Peptide aldehydes, which inhibit cysteine proteases and proteasomes, reportedly block A b peptide secretion by interfering with g -secre-tase cleavage. Using a novel, specific, and sensitive en-zyme-linked immunosorbent assay for the b -secretase-cleaved fragment of the Swedish mutant of APP (APPSw), we determined that the peptide aldehyde, MG132, prevented b -secretase cleavage. This block in b -secretase cleavage was not observed with clasto -lacta-cystin b -lactone and thus, cannot be attributed to proteasomal inhibition. MG132 inhibition of b -secretase cleavage was compared with the serine protease inhibitor, 4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride (AEBSF). AEBSF inhibition of b -secretase cleavage was immediate and did not affect a -secretase cleavage. With MG132, inhibition was delayed and it decreased secretion of a -cleaved APPSw as well. Fur-thermore, MG132 treatment impaired maturation of full-length APPSw. Both inhibited intracellular forma-tion of the b -cleaved product. These results suggest that aldehydes as have multiple effects on the maturation and processing of APP. We conclude that the MG132-induced decrease in b -secretase cleavage of APPSw is due to a block in maturation. This is sufficient to explain the previously reported peptide al-dehyde-induced decrease in A b peptide secretion.

Evidence continues to accumulate supporting the hypothesis that amyloid plaques in the brain have a causative role in the generation of Alzheimer's disease (for review, see Ref. 1). Increased brain levels of amyloid peptide and cognitive decline are strongly correlated (2). Amyloid plaques largely consist of peptides of 40 (A␤40) 1 and 42 (A␤42) amino acids in length that are derived by the enzymatic processing of a type I transmembrane protein called amyloid precursor protein (APP). Two enzymatic cleavages of APP are necessary to produce amyloid peptides. First, ␤-secretase cleaves APP to create the aminoterminal end of the peptide. A double mutation (K651N/ M652L; 751 isoform numbering) just amino-terminal to this ␤-secretase cleavage site has been identified in a Swedish pedigree of familial Alzheimer's disease (3). This double mutation of APP, known as the "Swedish" mutation (APPSw), elevates intracellular and secreted levels of A␤ peptide from 6-to 8-fold (4). This appears to be a consequence of increased cleavage of APPSw by ␤-secretase compared with wild type APP. Following ␤-secretase cleavage, ␥-secretase subsequently cleaves the COOH-terminal membrane-bound fragment (CTF) of APP within the transmembrane sequence to release the A␤ peptide. Thus, inhibitors that specifically block the cleavage of APP by these secretases have enormous therapeutic potential.
Several laboratories have now cloned an enzyme that cleaves APP and APPSw at the ␤-secretase site (5)(6)(7). Referred to as BACE (␤-site APP cleaving enzyme), this enzyme is a member of a unique family of transmembrane aspartic proteases. A second related protein designated ASP1 or BACE2 has also been identified (6,8,9). However, the expression pattern of BACE2 in the brain appears to exclude it from playing a major role in Alzheimer's disease (9). The mature, fully glycosylated form of BACE has a half-life in the cell of greater than 9 h (10). BACE appears to localize to the Golgi apparatus (5). Despite about 40% amino acid similarity between BACE and pepsin proteases (9), the cysteine residues in BACE involved in intramolecular disulfide bonds are not conserved with other pepsin family members (10). Such fundamental structural differences may explain why ␤-secretase is insensitive to pepstatin, a specific inhibitor of pepsin proteases (7). The search for such a specific inhibitor of ␤-secretase cleavage of APP as a possible treatment for Alzheimer's disease has intensified with the discovery of BACE.
Before the cloning of the aspartic protease, BACE, cellular studies using a serine protease inhibitor, 4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride (AEBSF), suggested that it blocked A␤ peptide generation by directly inhibiting ␤-secretase activity (11). There is little information published about other potential specific inhibitors of ␤-secretase. Several studies have shown that peptide aldehyde protease inhibitors affect the secretion of A␤ peptides (12)(13)(14). Of these peptide aldehydes, MG132 was one of the most potent inhibitors of A␤ peptide secretion (15). Studies focusing specifically on the secretion of A␤40 and A␤42 revealed peptide aldehydes to have a complex effect on APP processing. Curiously, at low concentrations, these peptide aldehydes produced an increase in A␤40 and A␤42 peptide secretion, whereas, at higher concentrations, a decrease in A␤ peptide secretion was observed (14 -17). The increase in A␤ peptide secretion at low concentrations of peptide aldehydes has been postulated to result from inhibition of degradation of the CTFs generated by BACE cleavage, making more of them available for ␥-secretase cleavage (17). The inhibition of A␤ peptide secretion with higher concentrations of peptide aldehydes is attributed to an impairment of ␥-secretase cleavage of the CTFs (12,13,17,18).
By developing an ELISA specific for the soluble ␤-secretasecleaved amino-terminal fragment of APPSw (APPSw␤; see Fig.  1A), we now report that the peptide aldehyde, MG132, prevents ␤-secretase cleavage of APPSw in a concentration-dependent manner. Furthermore, MG132 is not inhibiting secretion of APPSw␤ into the medium since it blocked intracellular production of APPSw␤ as well. MG132 is compared with AEBSF and the specific proteasomal inhibitor, clasto-lactacystin ␤-lactone, on ␤-secretase cleavage. MG132 impairs maturation, blocking ␤-secretase cleavage of APPSw in the late Golgi apparatus. This offers an alternative explanation as to how higher concentrations of a peptide aldehyde can decrease A␤ peptide secretion.

EXPERIMENTAL PROCEDURES
Antibodies and Protease Inhibitors-The mouse monoclonal antibodies referred to in Fig. 1A as 22C11 and BIOSOURCE were obtained from Roche Molecular Biochemicals and BIOSOURCE International (monoclonal antibody P2-1), respectively. The mouse monoclonal antibody 6E10, which recognizes an epitope in the first 17 amino acids of the A␤ peptide, was obtained from Senetek, Inc. The anti-APP COOHterminal rabbit polyclonal antibody was obtained from Chemicon International, Inc., and the mouse monoclonal antibody to the amino terminus of APP, LN27, was purchased from Zymed Laboratories Inc. The mouse monoclonal antibody, 8E5, was a generous gift from Dr. Dale Schenk (Elan Pharmaceuticals). The rabbit polyclonal antibody, 945, was raised against a synthetic peptide corresponding to the last 19 amino acid residues of the carboxyl terminus of APP (CMQQNGY-ENPTYKFFEQMQN) that was cross-linked to keyhole limpet hemocyanin via an amino-terminal cysteine. The polyclonal antibodies, 931 and 932, were raised against a synthetic peptide corresponding to 19 amino acid residues (CRPGSGLTNIKTEEISEVNL) just amino-terminal to the ␤-secretase cleavage site of APPSw that was similarly crosslinked to keyhole limpet hemocyanin via an amino-terminal cysteine.
The peptide aldehyde, MG132, was dissolved in dimethyl sulfoxide (Me 2 SO) at a concentration of 10 mM (Peptides International). The proteasomal inhibitor, clasto-lactacystin ␤-lactone (Calbiochem), was also dissolved in Me 2 SO at 2 mM. Me 2 SO was used in all experiments as a vehicle control. The serine protease inhibitor AEBSF (Sigma) was dissolved in sterile water at 0.2 M.
Cell Culture and Transfection-The Chinese hamster ovary cell line, CHOK1, and the HEK293 cell line used for transient transfections were obtained from the American Type Culture Collection. The cell lines stably expressing the 695 isoforms of either APPSw (CHOAPPSw) or APP wild type were a generous gift from Taraneh Haske (Pfizer Pharmaceuticals, Ann Arbor, MI). All CHO cell lines were grown in DMEM supplemented with 10% heat-inactivated fetal calf serum, glutamine, nonessential amino acids, and penicillin/streptomycin/fungizone as described (19). Experiments involving transient expression of APPSw or APP were conducted using cDNAs encoding for the 751 isoforms that were cloned into the pCDNA3 mammalian expression vector (Invitrogen, Inc.). Transfections of HEK293 cells grown on 60-mm plates were conducted using LipofectAMINE as described by the Life Technologies, Inc. Forty hours following transfection, cells were metabolically labeled or lysed for immunoblot analysis as described below.
Metabolic and Pulse-Chase Labeling-Cells stably or transiently expressing APP or APPSw were preincubated in methionine/cysteine-free medium for 15 min prior to labeling. Cells were metabolically labeled by incubating in 2 ml of medium containing [ 35 S]methionine and [ 35 S]cysteine (Tran 35 S-label; ICN Pharmaceuticals) at 50 Ci/ml for 1 h. In pulse-chase studies, cells were preincubated in methionine/cysteinefree medium for 15 min and then pulsed for 12 min with Tran 35 S-label (100 Ci/ml). Cells were then washed and incubated in complete medium containing excess methionine and cysteine for the chase times shown. After labeling and chase were complete, the conditioned medium was collected and cells washed in PBS. Cells were lysed in 1 ml of lysis buffer (0.5% Nonidet P-40, 0.5% deoxycholate in 50 mM Tris, 150 mM NaCl, and 5 mM EDTA, pH 8.0) and insoluble cell debris removed by centrifugation as described (19). The resulting cleared supernatant of the cell lysate was then subjected to further analysis.
Immunoprecipitation, Immunoblotting, and Gel Electrophoresis-Full-length APP and APPSw were isolated from the cell lysate supernatants by immunoprecipitation using the 945 rabbit antisera to the carboxyl terminus of APP. Except where noted, lysates were incubated with 4 l of 945 antisera for 90 min at 4°C and protein-antibody complexes were isolated by incubation with protein A-Sepharose for 30 min at 4°C. Intracellular APPSw␤ was similarly isolated from cell lysates using 4 l of 931 rabbit antisera. The 931 antisera was also used to isolate secreted APPSw␤ from 0.5 ml of conditioned medium that had Nonidet P-40 and deoxycholate added to a final concentration of 0.5%. Immunoprecipitations using the mouse monoclonal antibodies, 8E5 or 6E10, were conducted as described above except that protein-antibody complexes were isolated using protein G-agarose (Roche Molecular Biochemicals).
Isolated proteins were resolved by SDS-polyacrylamide gel electrophoresis (PAGE) using an 8% separating gel. Radiolabeled proteins in SDS gels were detected by fluorography using Amplify (Amersham Pharmacia Biotech). Immunoblot analysis of isolated proteins was conducted essentially as described (19). Briefly, immunoprecipitated proteins were resolved by SDS-PAGE and transferred to PROTRAN (Schleicher & Schuell). Membranes were subsequently blocked in gelatin wash buffer (0.1% gelatin, 15 mM Tris, pH 7.5, 130 mM NaCl, 1 mM EDTA, and 0.1% Triton X-100). Membranes were subsequently incubated with mouse monoclonal antibody 22C11 to detect the aminoterminal end of APP molecules from cell lysates or conditioned medium. Membranes were alternatively incubated with 6E10 to detect APPSw␣ in conditioned medium. Membranes were subsequently washed and incubated with a sheep anti-mouse IgG antibody conjugated to horseradish peroxidase as described by the manufacturer (Amersham Pharmacia Biotech). The membranes were again washed and signals detected by chemiluminescence using the ECL system (Amersham Pharmacia Biotech).
Full-length APP and APPSw␤ Sandwich ELISAs-The 945 and 931 antisera were used as capture antibodies for the full-length APP and APPSw␤ sandwich ELISAs, respectively. Each was first affinity-purified against the appropriate peptide that had been cross-linked to CNBr-activated Sepharose 4B (Amersham Pharmacia Biotech) following the manufacturer's instructions. The anti-APP COOH-terminal antibody, 945, was affinity-purified with the same 20-amino acid synthetic peptide used above to inoculate rabbits. The 931 antibody was affinitypurified using a synthetic peptide identical to the last seven amino acids of the neoepitope of the ␤-secretase-cleaved APPSw soluble fragment (CEIESVNL). The 945 and 931 antibodies that bound to the immobilized peptide were eluted using ActiSep Elution Medium (Sterogene). The eluted antibodies were desalted using a PD-10 column (Amersham Pharmacia Biotech) as directed by the manufacturers. The affinity-purified capture antibodies (945 and 931) were diluted to 1.2 g/ml in PBS and 100 l added per well to a 96-well Nunc-Immuno Maxisorp plate (Nalge Nunc International). After incubation to allow binding of antibody, SuperBlock (Pierce) was added to each well to block nonspecific binding sites. The wells were repeatedly rinsed and then stored at 4°C in PBS, 0.05% Tween 20 (PBS/T) until ready for use.
For the full-length APP ELISA, the protein concentration of the cell lysate supernatant was determined using the BCA protein assay (Pierce). Ten micrograms of protein from the cell lysates were aliquoted into each well. The total volume in the well was adjusted to 100 l with PBS, and samples were incubated overnight at 4°C. On the following day, the samples were incubated an additional 1 h with constant shaking. The wells were then washed four times with PBS/T and then incubated with 100 l of diluted detector antibody (the mouse monoclonal antibody, 8E5, at 0.25 g/ml except where noted). All antibodies used in the ELISA were diluted using a solution of 10% SuperBlock and 90% PBS/T. The detector antibody was incubated with each sample for 4 h with constant shaking. The wells were again washed four times with PBS/T and subsequently incubated for 1 h with rabbit anti-mouse IgG conjugated to horseradish peroxidase (diluted 1:4000; Southern Biotechnology Associates, Inc.). Following three washes with PBS/T and two washes with PBS alone, 100 l of 3,3Ј,5,5Ј-tetramethylbenzidine (Pierce) solution was added to each well. The reaction was stopped by the addition of an equal volume of 2 M sulfuric acid. The relative amount of full-length APP in the sample was then quantified colorimetrically at 450 nm. The levels of secreted APPSw␤ were assayed in the exact same way except that plates coated with affinity-purified 931 antibody were used to capture the ␤-secretase-cleaved protein and just 25 l of conditioned medium was loaded per well. Each experiment was repeated at least three times, and the indicated values are averages of triplicate measurements Ϯ S.D.

Characterization of Polyclonal Antibodies to APP and
APPSw␤-As shown in Fig. 1, ␤-secretase cleavage of APPSw generates a large soluble amino-terminal fragment (APPSw␤) that is secreted into the medium and a CTF that is subsequently cleaved by ␥-secretase to derive the A␤ peptide. Two novel rabbit polyclonal antisera were generated to perform the experiments described below. The binding sites of these and other antibodies are also shown in Fig. 1. The first polyclonal antisera (945) was raised to the last 20 amino acids of the APP carboxyl terminus. The specificity of the anti-COOH-terminal antibody, 945, is shown in Fig. 2A. CHO cells stably expressing APPSw (CHOAPPSw) were metabolically labeled for 1 h with Tran 35 S-label. Following cell lysis, equal amounts of supernatant were incubated with 1, 2, or 4 l of 945 antiserum or with 4 l of preimmune serum followed by protein A-Sepharose. The immunoprecipitated APPSw was compared with that obtained using the Chemicon anti-COOH-terminal antibody by SDS-PAGE. The resulting autoradiograph shows that 945 specifically immunoprecipitates the N-glycosylated immature (I) and completely glycosylated mature (M) forms of APPSw. The second antisera specifically recognizes only the carboxyl terminus of the soluble amino-terminal fragment (APPSw␤) created when ␤-secretase cleaves APPSw just COOH-terminal to Leu652 (751 numbering). Results shown in Fig. 2B demonstrate that antisera raised in two rabbits (931 and 932) against the 20-amino acid sequence just amino-terminal to the ␤-secretase cleavage site are capable of immunoprecipitating a soluble APPSw fragment from the conditioned medium of CHOAPPSw cells. Culture medium conditioned for 24 h by CHOAPPSw cells was divided into equal aliquots and incubated with preimmune sera from rabbit 931, 931 antisera, or 932 antisera. The mouse monoclonal antibody, 8E5, which recognizes an epitope in the lumenal region of APP ( Fig. 1), served as a positive control. The immunoprecipitates were resolved by SDS-PAGE and the APPSw amino-terminal fragments were identified by Western blot analysis using the mouse monoclonal antibody 22C11. Both 931 and 932 antisera, but not preimmune sera, immunoprecipitated a specific APP soluble fragment from the conditioned media. The 931 antisera appeared to have a higher titer than 932. Thus, all subsequent experiments utilized only 931.
As others have demonstrated, the ␣and ␤-cleaved soluble fragments are difficult to electrophoretically resolve from one another (20,21). Therefore, it was unclear from this Western blot analysis of conditioned medium whether the soluble APPSw fragment immunoprecipitated by 931 was limited to secreted APPSw␤ or also included ␣-secretase-cleaved APPSw (APPSw␣). Furthermore, it was not known if 931 could immunoprecipitate wild type APP␤ and APP␣. Consequently, HEK293 cells were transiently transfected with either pC-DAPP or pCDAPPSw. Medium that had been conditioned for 36 h by these transiently transfected cells was collected and soluble APP fragments were immunoprecipitated with 931 or the mouse monoclonal antibody 8E5. Since the epitope recognized by 8E5 is amino-terminal to the ␤-secretase cleavage site, it is capable of immunoprecipitating APP␣, APP␤, APPSw␣, and APPSw␤. Immunoprecipitates were resolved by SDS-PAGE and immunoblotted using the mouse monoclonal antibody, 22C11 to detect all forms of soluble APP and APPSw (see Fig. 1). As expected, 8E5 was able to immunoprecipitate both secreted APP and APPSw from the conditioned media (Fig. 2C,  panel 1). However, 931 only immunoprecipitated soluble APPSw amino-terminal fragments from the conditioned medium of cells transiently transfected with pCDAPPSw and did not recognize wild type secreted APP fragments. A set of immunoprecipitations from conditioned medium identical to that conducted in Fig. 2C (panel 1) was immunoblotted with the mouse monoclonal antibody 6E10, which detects only secreted APP and APPSw soluble fragments that have been cleaved by ␣-secretase. Although both APP␣ and APPSw␣ were detected in the 8E5 immunoprecipitates, neither form was immunoprecipitated using the 931 antibody (Fig. 2C, panel 2). Taken together, these results demonstrate that the 931 antibody specifically recognizes only the neoepitope derived with ␤-secretase cleavage of APPSw. It does not cross-react with full-length APPSw (compare Fig. 4, A and C, chase at 0 min), soluble APP␣, APP␤, or APPSw␣.
Characterization of ELISAs for Detection of Full-length APP and the Soluble ␤-Secretase Cleaved APPSw Amino-terminal Fragment-Since 931 specifically recognized the neoepitope of ␤-secretase-cleaved APPSw in conditioned medium, it was used to create an enzyme-linked immunosorbent assay (ELISA) to quantitatively measure the amount of secreted APPSw␤. A similar ELISA was also developed to measure the amount of full-length APP or APPSw present in cell lysates using 945. To identify the best detector antibody to use in the ELISA, the affinity-purified 945 was coated on 96-well plates to capture full-length APP from cell lysates of either CHO cells or CHO cells stably expressing APPSw. After rinsing, triplicate wells were incubated with no detector, 8E5 (0.25 g/ml), BIO-SOURCE (0.25 g/ml), or LN27 (0.5 g/ml) mouse monoclonal antibodies and developed as described under "Experimental Procedures." The signal level observed using CHO cell lysates with these detector antibodies did not significantly differ from background, demonstrating the specificity of the ELISA for the stably expressed, human APPSw (Fig. 2D). Full-length APPSw was specifically detected in CHOAPPSw lysates with 8E5 and the BIOSOURCE mouse monoclonal antibodies showing the greatest sensitivity. A total protein concentration of 10 g of cell lysate was found to be optimal to detect full-length APPSw (data not shown).
The 8E5 and BIOSOURCE mouse monoclonal antibodies were also evaluated as detector antibodies in the 931 ELISA to measure secreted APPSw␤. Both 8E5 and BIOSOURCE antibodies were sensitive detectors of APPSw␤ captured in wells coated with affinity-purified 931 (Fig. 2E). The addition of the eight-amino acid synthetic peptide corresponding to the carboxyl terminus of APPSw␤ to the conditioned medium from CHOAPPSw cells blocked detection of the ␤-cleaved product. When using 6E10 as the detector antibody, no signal above background was observed (data not shown). This indicated that no APPSw␣ was captured in the ELISA by the affinity-purified 931. Together, these results demonstrate the high specificity of this novel ELISA for APPSw␤. Intracellular APPSw␤ Is Detected by Pulse-Chase Analysis-Experiments were next conducted to determine whether 931 recognized intracellular APPSw␤. Intracellular APPSw␤ was detectable in the cell lysate as a slightly diffuse band migrating just above a nonspecific band (Fig. 3A, panel 2). This nonspecific 35 S-labeled band was detected in both CHOAPP and CHOAPPSw cell lysates (data not shown). Therefore, an additional experiment was conducted to verify that this protein was not recognized by antibodies specific for the APPSw␤ neoepitope. CHOAPPSw cells were pulse-labeled with Tran 35 Slabel for 12 min and chased for 45 min as described under "Experimental Procedures." Conditioned media and cell lysates were isolated and divided in half. An excess of a synthetic peptide with a sequence corresponding to the last eight amino acids of the neoepitope (EISEVKNL) in APPSw␤ was added to one set of samples. Secreted and intracellular APPSw␤ were then immunoisolated using 931 from conditioned media and cell lysates, respectively. Full-length APPSw was subsequently immunoprecipitated from the same cell lysates using 945. Fulllength APPSw was still immunoprecipitated in the presence of the neoepitope peptide, indicating that the peptide was not causing a general block in the immunoisolation of proteins (Fig.  3A, panel 1). However, the synthetic neoepitope peptide blocked immunoprecipitation of secreted APPSw␤ (Fig. 3A,  panel 2) and intracellular APPSw␤ (Fig. 3A, panel 3) using 931 antibody. In contrast, the nonspecific band was still detectable in the presence of synthetic peptide. Brefeldin A (BFA) treatment of HEK293 cells to block transport of proteins beyond the Two newly derived rabbit polyclonal antisera (931 and 932) were used to immunoprecipitate APP soluble fragments from conditioned medium collected from CHOAPPSw cells. This was compared with APP fragments immunoprecipitated by the 8E5 antibody by immunoblotting with 22C11, which recognizes both APPSw␣ and APPSw␤ fragments. C, immunoblot analysis indicates that the rabbit polyclonal antisera 931 specifically immunoprecipitates only APPSw␤ from conditioned media. Conditioned media collected from HEK293 cells transiently transfected with either pCDAPP or pCDAPPSw were incubated with either 8E5 or 931 and protein-antibody complexes immunoabsorbed using protein G-or protein A-agarose, respectively. Duplicate samples of immunoprecipitated proteins were resolved by SDS-PAGE and transferred to nitrocellulose. Panel 1, one set of samples was immunoblotted with the 22C11 mouse monoclonal antibody to detect secreted APP␣, APPSw␣, APP␤, and APPSw␤. Panel 2, a duplicate set of immunoprecipitated samples was immunoblotted for APP␣ and APPSw␣ using the mouse monoclonal antibody, 6E10. D, two ELISAs described under "Experimental Procedures" were derived to measure APPSw␤ in conditioned media and full-length APP or APPSw in cell lysates. Affinity-purified 945 antibody was used to capture full-length APPSw molecules from the lysate of CHOAPPSw cells. Three different mouse monoclonal antibodies that recognize epitopes in the lumenal region of human APP were screened as detector antibodies. Lysates of CHO cells expressing only endogenous hamster APP served as controls for specificity. E, affinity-purified 931 antibody was used to capture secreted APPSw␤ molecules from the conditioned media of CHOAPPSw cells. CHO cell conditioned medium served as a negative control. Both of the anti-APP lumenal domain antibodies, 8E5 and BIOSOURCE, were effective detector antibodies. Addition of the CEIESVNL peptide to the CHOAPPSw conditioned medium prevented detection of secreted APPSw␤.
Golgi apparatus inhibited ␤-secretase cleavage of APPSw (20,22). Thus, a block in maturation of APPSw by BFA does not permit access to BACE in the late Golgi. We sought to confirm these findings using the 931 antisera. CHOAPPSw cells were metabolically labeled with Tran 35 S-label for 1 h in the presence or absence of 20 g/ml brefeldin A. Labeled conditioned medium was collected and cells lysed as above. Intracellular and secreted APPSw␤ were isolated by immunoprecipitation using the 931 antisera from cell lysates and conditioned media, respectively. Full-length APPSw was subsequently immunoisolated from cell lysates using the 945 antibody. As expected, BFA treatment blocked the appearance of APPSw␤ in the conditioned medium (Fig. 3B, panel 3). Completely glycosylated, mature full-length APPSw was not detected in BFA-treated cell lysates indicating that maturation was blocked (panel 1). Similar to earlier reports (20,22), our studies found that intracellular APPSw␤ was not detected in CHOAPPSw cells treated with BFA, yet the nonspecific band was still detectable (Fig.  3B, panel 2). Thus, blocking maturation of APPSw prevents it from being cleaved by ␤-secretase.
Taken together, the results described above indicated that iAPPSw␤ is detectable in the lysates of radiolabeled CHOAPPSw cells. Consequently, a pulse-chase study was conducted to evaluate the time course of APPSw␤ production and secretion by CHOAPPSw cells. Cells were pulsed for 12 min with Tran 35 S-label and subsequently chased in complete medium containing excess methionine and cysteine for 0 -90 min. At each indicated chase time, conditioned medium was col-lected and the PBS-washed cells were lysed. Intracellular APPSw␤ was immunoprecipitated from the cleared supernatants using 931. Full-length APPSw was subsequently immunoisolated from the supernatants using 945. Secreted APPSw␤ was immunoprecipitated from the conditioned media with 931. These radiolabeled proteins were resolved by SDS-PAGE, and the resulting autoradiographs are shown in Fig. 4. A 12-min pulse was sufficient to label N-glycosylated APPSw (I in Fig.  4C) but not fully glycosylated, mature APPSw (M). Mature APPSw was detectable after 7.5 min of chase and peaked after 25-35 min. After a 90-min chase, no mature APPSw was detectable and only a trace of immature protein was still present. Note that a large portion of APPSw fails to chase into the fully mature form over the chase period, but instead appears to remain as incompletely glycosylated APPSw. This may represent misfolded protein that is retained in the ER for eventual degradation by the proteasome in these APPSw-overexpressing cells. Fig. 4A shows that iAPPSw␤ was readily detectable after chasing for about 25 min and corresponded with the appearance of the completely glycosylated, mature APPSw. The level of iAPPSw␤ reached a maximum at 35-45 min and was barely detectable after 90 min of chase. Thus, as expected, the appearance of iAPPSw␤ preceded the secretion of APPSw␤ into the medium (Fig. 4B). Only a trace of intracellular APPSw␤ was detected after 35 min, while secreted APPSw␤ continued to accumulate after chasing for 90 min.
The APPSw␤ ELISA Is a Specific and Sensitive Detector of Inhibitors of ␤-Secretase Cleavage-The serine protease inhibitor AEBSF, inhibits A␤-peptide secretion in a concentrationdependent manner in both neuronal and non-neuronal cell lines stably expressing APP or APPSw by presumably blocking ␤-secretase cleavage (11). The APPSw␤ ELISA enabled us to compare the concentration dependence of the AEBSF-induced inhibition of ␤-secretase activity to that reported for inhibition of A␤ peptide secretion. CHOAPPSw cells plated the previous day were rinsed in PBS and incubated for 5 h with fresh medium containing increasing concentrations of AEBSF (0.1-1.0 mM). The conditioned medium was then collected, and APPSw␤ levels were quantified using the 931 ELISA. AEBSF produced a concentration-dependent inhibition of APPSw␤ secreted into the medium (Fig. 5A). CHOAPPSw cells treated with 1 mM AEBSF had secreted only 43% of the APPSw␤ detected in conditioned medium from untreated cells. This supports the observation that AEBSF inhibits ␤-secretase cleavage of APPSw by a direct or indirect mechanism.
The highly specific proteasomal inhibitor, lactacystin, in-

FIG. 3. Immunoprecipitation of radiolabeled intracellular and secreted APPSw␤ is blocked by a synthetic peptide of the Swedish mutant and brefeldin A treatment.
A, CHOAPPSw cells were pulse labeled as described and chased for 45 min in complete DMEM. Conditioned media and cell lysates were isolated and divided in half. 931 was used to immunoprecipitate secreted and intracellular APPSw␤ in the presence or absence of excess synthetic peptide corresponding in sequence to the carboxyl terminus of APPSw␤ (CEIESVNL). 1, the excess peptide did not interfere with immunoprecipitation of full-length APPSw with 945. 2, the CEIESVNL peptide blocked immunoprecipitation of iAPPSw␤ (indicated by an arrow), but not the nonspecific protein just below it. 3, secreted APPSw␤ was not immunoprecipitated by 931 when the CEIESVNL peptide was added to the conditioned medium. B, CHOAPPSw cells were metabolically labeled for 1 h with Tran 35 S-label in the presence (ϩ) or absence (Ϫ) of 20 g/ml brefeldin A (BFA). Conditioned medium was collected and cell lysates isolated as described. Full-length APPSw was immunoprecipitated with 945 antisera (panel 1). Intracellular (indicated by arrow in panel 2) and secreted (panel 3) APPSw␤ were immunoprecipitated from cell lysate and conditioned media, respectively, with 931 antisera. creases APP␣ and intracellular and secreted A␤42 (18,23,24). Other investigators reported that lactacystin increased secretion of both A␤40 and A␤42 from SH-SY5Y cells (14). The proposed explanation for these increases is that inhibition of the proteasome prevents degradation of both full-length and COOH-terminal fragments of APP making more of each available for cleavage by the ␣and ␥-secretases, respectively (18). We reasoned that, if lactacystin disrupted ER-associated degradation of APPSw in a manner similar to other transmembrane proteins (25)(26)(27)(28), then more full-length protein would also be available for cleavage by ␤-secretase. To test this hypothesis, CHOAPPSw cells were incubated for 6 h with increasing concentrations of the active form of lactacystin, clastolactacystin ␤-lactone (␤-lactone). This cell-permeable, irreversible inhibitor of the proteasome has a 5-10 fold greater potency than lactacystin with an IC 50 in intact cells of 1 M (29). ELISA measurements of secreted APPSw␤ revealed that a concentration of 10 times the IC50 (i.e. 10 M) still had no effect on the levels of secreted APPSw␤ (Fig. 5B). Only at a concentration of 20 M ␤-lactone was a decrease in ␤-secretase-cleaved APPSw detectable (about 85% of control levels). Thus, proteasomal inhibition has little influence on ␤-secretase cleavage of APPSw.
Intriguingly, peptide aldehyde protease inhibitors, which are capable of inhibiting both the proteasome and cysteine proteases, increase A␤ peptide secretion at low concentrations and inhibit A␤ secretion at high concentrations (13)(14)(15)(16)(17). MG132 was one of the most potent peptide aldehyde inhibitors of A␤ peptide secretion (15,18). We hypothesized that MG132 was inhibiting A␤ peptide secretion at high concentrations by interfering with ␤-secretase cleavage of APP. To test this hypothesis, CHOAPPSw cells were incubated for 6 h with MG132 at concentrations demonstrated to inhibit A␤ peptide secretion (20 -100 M). The 931 ELISA was used to quantify the amount of APPSw␤ secreted by MG132-treated cells and compared with untreated controls. MG132 treatment produced a concentration-dependent decrease in the levels of APPSw␤ detected in the conditioned media (Fig. 6A). Cells incubated with 100 M MG132 secreted APPSw␤ at 35% the level secreted by un-treated cells. As a control, MG132 was added to a final concentration of 100 M following collection of media that had been conditioned for 6 h by untreated cells (100ϩ). The level of APPSw␤ was identical to untreated samples, indicating that MG132 was not interfering with the ELISA itself. To determine whether MG132 was inhibiting APPSw␤ levels by decreasing the levels of full-length APPSw, the 945 ELISA was performed on equal amounts of lysates of the treated cells. No significant difference in full-length APPSw levels was detected with MG132 treatment (Fig. 6B). Thus, MG132 treatment caused a concentration-dependent decrease in the amount of secreted APPSw␤. To examine the time course of this block in APPSw␤ production, CHOAPPSw cells were incubated for increasing time with or without 80 M MG132. Conditioned medium was collected at the times indicated and the amount of secreted APPSw␤ determined by ELISA (Fig. 6C). Our results show that inhibition of APPSw␤ secretion could not be detected until 2 h following addition of the peptide aldehyde inhibitor. Following this incubation period, APPSw␤ continued to be secreted into the media by MG132-treated cells but at a much slower rate compared with untreated controls.
The continued increase in APPSw␤ secretion 30 min and 1 h after incubation with MG132 could be explained in two ways. Either this was the time required for MG132 to penetrate the cell and block ␤-secretase cleavage or inhibition was immediate but was not detectable by this assay because of the lag time required for secretion of APPSw␤ generated prior to the block. It was also of interest to determine whether the block in APPSw␤ production by MG132 was similar in its immediacy and specificity to that observed for AEBSF. Therefore, CHOAPPSw cells were pulsed for 12 min with Tran 35 S-label and then chased for 0, 45, or 90 min in complete medium containing Me 2 SO, 80 M MG132, or 1 mM AEBSF. The resulting cell lysates were immunoprecipitated first with 931 followed by 945. Equal aliquots of conditioned medium were incubated with 931 or 6E10 to immunoprecipitate secreted APPSw␤ and APPSw␣, respectively. When 80 M MG132 was present only during the chase period, it had little effect on full-length APPSw maturation, intracellular APPSw␤ production, or APPSw␤ secretion compared with control. MG132 treatment, however, produced a large increase in the amount of secreted APPSw␣ (Fig. 7D). In contrast to this is the effect of chasing in the presence of 1 mM AEBSF, which immediately blocked the appearance of APPSw␤ in the medium (Fig. 7C). The amount of iAPPSw␤ was markedly reduced after chasing for 45 min compared to control and was still reduced after 90 min (Fig. 7B). AEBSF treatment also stabilized the immature form of full-length APPSw, which was nearly unchanged through 90 min of chase in the presence of the inhibitor (Fig.  7A). Although AEBSF nearly completely blocked APPSw␤ secretion, secreted levels of APPSw␣ were slightly increased after 90 min of chase (Fig. 7D). The lack of an effect of MG132 during this chase period indicates that it does not produce an immediate block in cleavage as observed with AEBSF. These results suggest that, although both AEBSF and MG132 induced a block in APPSw␤ production, they differ in their molecular mechanisms.
MG132 had little effect on secretion of APPSw␤ in the first 90 min of incubation (see Figs. 6 and 7). Therefore, an additional pulse-chase experiment was conducted in which cells were preincubated for 2 h with 80 M MG132 or Me 2 SO since a block in APPSw␤ production was detected after 2 h of MG132 treatment (Fig. 6C). After the pretreatment, cells were pulsed and then chased in the presence of vehicle or MG132. Conditioned media and cell lysates were isolated and analyzed as above for full-length APPSw, iAPPSw␤, APPSw␤, and APPSw␣. A profound effect was observed on APPSw processing after 2 h of MG132 preincubation (Fig. 8). After a 45-min chase, mature APPSw was readily detectable in CHOAPPSw cells treated with Me 2 SO, while the MG132-treated cells showed only a trace of the mature form (Fig. 8A). After 90 min of chase, very little radiolabeled immature or mature APPSw remained in the control sample, whereas in the MG132-treated cells a large portion of the radiolabeled immature APPSw detected at time zero still remained. Corresponding with this impaired maturation of APPSw with MG132 treatment was a decrease in the levels of iAPPSw␤ (Fig. 8B), secreted APPSw␤ (Fig. 8C), and secreted APPSw␣ (Fig. 8D). This suggests that treatment of cells with high concentrations of peptide aldehydes may cause a more general impairment of the secretory pathway. This impairment blocks transport and processing of APPSw through the secretory pathway, thus explaining the decrease in A␤ peptide secretion observed by others (12)(13)(14). DISCUSSION In previous studies, peptide aldehydes such as MG132 were thought to inhibit A␤ and p3 peptide secretion by blocking ␥-secretase cleavage of the COOH-terminal fragment of APP (12,13,17,18). This conclusion was based on the observation that CTFs were detectable in cells incubated with peptide aldehydes. Our data suggest that peptide aldehydes, such as MG132, have a more widespread effect on APP maturation, processing and secretion. First and foremost, ␤-secretase cleavage of APPSw was blocked at concentrations of MG132 reported to impair A␤ peptide secretion (15,18). This inhibition in ␤-secretase cleavage is sufficient to account for the reported decrease in A␤ peptide secretion. Second, we demonstrated that, after 2 h of incubation with MG132, secretion of APPSw␣ was also inhibited. Finally, in pulse-chase studies on CHOAPPSw cells preincubated with MG132, fully mature APPSw was barely detectable in the cell lysates. Instead, the immature, N-glycosylated form that is located in the ER was stabilized, suggesting MG132 incubation has multiple effects on APPSw processing. Like Skovronsky et al. (18), we were unable to detect stabilization of mature APPSw and inhibition of secreted APPSw␣ and APPSw␤, when cells were only incubated with MG132 during the chase period. However, our studies, in which a 2-h preincubation with MG132 blocked production of iAPPSw␤, may better represent the effects of peptide aldehydes on A␤ peptide secretion over the long incubation period used by these and other investigators (13)(14)(15)(16)(17)(18).
Our results on MG132-induced inhibition of ␤-secretase cleavage along with results of others describing the stabilization of CTFs in cell lysates by such peptide aldehydes can be best explained if treatment with these compounds is somehow impeding progression of APPSw through the secretory pathway. Haass et al. (30) have shown that impairing maturation of APPSw by treating cells with BFA prevents production of in- tracellular APPSw␤ because transport to the site of ␤-secretase cleavage is blocked. We obtained the same results using the 931 antisera described here. Other investigators achieved similar results by blocking movement of APP from the endoplasmic reticulum (ER). In those studies, a dilysine ER retrieval motif introduced into the cytoplasmic tail of APP strongly impaired cleavage by ␤-secretase (31). We, again, have obtained equivalent results expressing APPSw bearing a dilysine ER retrieval motif in CHO cells. 2 Like BFA treatment or introduction of a dilysine retrieval motif, a 2-h MG132 treatment of CHO cells stably expressing APPSw blocked the maturation of the fulllength molecule in experiments described here. This, in turn, blocked production of intracellular APPSw␤ and its subsequent appearance in the media. We suggest that the increase in the C83 and C99 CTFs observed by others with MG132 treatment results from a similar block in access of these ␣and ␤-secretase-cleaved fragments to ␥-secretase. Regardless of the mechanism, our results showing that MG132 blocks ␤-secretase cleavage of APPSw call into question the utility of such peptide aldehydes as ␥-secretase inhibitors.
The 931 antibody, described here for the first time, specifically recognized APPSw␤ and did not cross-react with fulllength APPSw, APPSw␣, APP␤, or APP␣. This enabled us to detect intracellular APPSw␤ in stably transfected CHOAPPSw cells. The results observed in our pulse-chase studies agree closely with previous findings of APPSw processing in HEK293 cells (22). Intracellular APPSw␤ was detectable within 25 min in pulse-chase experiments and coincided with the appearance of mature APPSw. Furthermore, we clearly detected APPSw␤ intracellularly prior to its appearance in the medium, eliminating the possibility that the APPSw␤ detected in cell lysates was due to already secreted ␤-cleaved fragments associating with the cell membrane.
Although BACE is an aspartic protease, 1 mM AEBSF (a serine protease inhibitor) reportedly blocks secretion of A␤ peptide by inhibiting ␤-secretase activity to about 44% of untreated cells (11). Our results measuring the secretion of APPSw␤ using the 931 ELISA showed that 1 mM AEBSF reduced APPSw␤ secretion to a remarkably similar 43% of control levels. Our results also agreed with theirs in that AEBSF treatment caused a slight increase in APPSw␣ secretion and stabilized full-length APPSw. Furthermore, we extended these observations by showing that this inhibition of ␤-secretase cleavage had a concentration dependence similar to that observed for A␤ peptide secretion (11). Only a small amount of iAPPSw␤ was detected in AEBSF-treated cells. This showed that AEBSF inhibition was not due to accumulation of APPSw␤ intracellularly because of a block in secretion, but due to a block in ␤-secretase cleavage of APPSw. Knowing now that BACE is an aspartic protease, it is unclear how a serine protease inhibitor such as AEBSF may block ␤-secretase cleavage, but it would argue against a direct inhibition of BACE. However, the discovery of such an inhibitor illustrates the importance of an in vivo screening method for compounds that block ␤and ␥-secretase cleavage of APP.
The reported effects of peptide aldehyde protease inhibitors on A␤ peptide secretion have been complex and conflicting. Nevertheless, results have repeatedly shown that A␤ and p3 peptide secretion is inhibited by the following peptide aldehydes: MDL23170 (12,32), ALLN (13,15,17), calpeptin (14,15), and MG132 (15,18) with the latter proving to be one of the most potent. Because increased levels of CTFs were detected in cells incubated with peptide aldehydes, the impairment in sub-sequent A␤ peptide secretion has been attributed to an inhibition of ␥-secretase activity by peptide aldehydes (12). Our results using both an APPSw␤ ELISA and pulse-chase studies revealed that ␤-secretase cleavage of APPSw was also impeded by MG132. The concentration dependence of this ␤-secretase cleavage inhibition closely corresponds to that reported for the decrease in A␤40, A␤42, and p3 peptide secretion (15,18). Inhibition of ␤-secretase cleavage was detectable after as little as 2 h of exposure to MG132. In studies by others (13)(14)(15)(16)(17)(18), A␤ peptide levels were measured after 3-16 h of incubation with peptide aldehyde. Thus, the MG132-induced block in ␤-secretase cleavage that we observed here could account for the decrease in A␤ peptide secretion. This observation indicates that the inhibitory effect by peptide aldehydes such as MG132 is not limited to inhibition of ␥-secretase activity. In addition to inhibiting cysteine proteases, MG132 is known to inhibit proteasomal cleavage with an IC 50 of a few micromolar (29). However, the potent and highly specific proteasomal inhibitor, clasto-lactacystin ␤-lactone, only caused a partial inhibition in APPSw␤ secretion in our studies at 20 times its IC 50 for proteasomal inhibition. Therefore, we conclude that the decrease in ␤-secretase cleavage of APPSw is not due to MG132 inhibition of proteasomal activity. In fact, lactacystin and low concentrations of peptide aldehydes are reported to increase A␤ peptide secretion (14). We detected no increase in ␤-secretase cleavage of APPSw with clasto-lactacystin ␤-lactone or low concentrations of MG132 (data not shown). Thus, we suggest that low concentrations of peptide aldehydes increase A␤ peptide secretion because they inhibit proteasomal degradation of ␤ and ␣-secretase-cleaved CTFs. This, in turn, makes more CTFs available for ␥-secretase cleavage. We further hypothesize that, at higher concentrations, peptide aldehydes inhibit A␤ and p3 peptide secretion by blocking cysteine proteases that have roles in protein processing and trafficking in the secretory pathway. This would explain the biphasic effect of peptide aldehydes observed on A␤ peptide secretion (14,15,17).
Although BACE activity in CHO cells was the highest of all non-neuronal cells, it was still significantly less than that observed for neuronal cell cultures (7). Therefore, the anti-APPSw␤ antibody, 931, may prove even more useful in analyzing APPSw processing in neuronal cell lines where greater levels of iAPPSw␤ may be expected. In particular, using affinity-purified 931 in the APPSw␤ sandwich ELISA described here will be immensely useful for exploring the cellular and molecular mechanisms that regulate ␤-secretase cleavage.