Protein Kinase C-dependent α-Secretase Competes with β-Secretase for Cleavage of Amyloid-β Precursor Protein in the Trans-Golgi Network*

The release of amyloidogenic amyloid-β peptide (Aβ) from amyloid-β precursor protein (APP) requires cleavage by β- and γ-secretases. In contrast, α-secretase cleaves APP within the Aβ sequence and precludes amyloidogenesis. Regulated and unregulated α-secretase activities have been reported, and the fraction of cellular α-secretase activity regulated by protein kinase C (PKC) has been attributed to the ADAM (a disintegrin and metalloprotease) family members TACE and ADAM-10. Although unregulated α-secretase cleavage of APP has been shown to occur at the cell surface, we sought to identify the intracellular site of PKC-regulated α-secretase APP cleavage. To accomplish this, we measured levels of secreted ectodomains and C-terminal fragments of APP generated by α-secretase (sAPPα) (C83) versus β-secretase (sAPPβ) (C99) and secreted Aβ in cultured cells treated with PKC and inhibitors of TACE/ADAM-10. We found that PKC stimulation increased sAPPα but decreased sAPPβ levels by altering the competition between α- versus β-secretase for APP within the same organelle rather than by perturbing APP trafficking. Moreover, data implicating the trans-Golgi network (TGN) as a major site for β-secretase activity prompted us to hypothesize that PKC-regulated α-secretase(s) also reside in this organelle. To test this hypothesis, we performed studies demonstrating proteolytically mature TACE intracellularly, and we also showed that regulated α-secretase APP cleavage occurs in the TGN using an APP mutant construct targeted specifically to the TGN. By detecting regulated α-secretase APP cleavage in the TGN by TACE/ADAM-10, we reveal ADAM activity in a novel location. Finally, the competition between TACE/ADAM-10 and β-secretase for intracellular APP cleavage may represent a novel target for the discovery of new therapeutic agents to treat Alzheimer's disease.

Posttranslational processing of the amyloid-␤ precursor pro-tein (APP) 1 is implicated in the pathology of Alzheimer's disease (AD). The amyloidogenic amyloid-␤ peptide (A␤) fragment is generated through an initial APP cleavage by ␤-secretase(s) to define the NH 2 terminus of A␤. This cleavage generates sAPP␤, a secreted ectodomain of APP, and C99, the 99-amino acid C-terminal fragment that contains the transmembrane and cytoplasmic domains of APP. C99 is further cleaved by ␥-secretase(s), releasing the A␤ peptide, which is either 40 or 42 amino acids long (i.e. A␤  and A␤  ; reviewed in Ref. 1). Because the majority of secreted A␤ appears to be generated in the trans-Golgi network (TGN), this is likely to be a major locus for ␤and ␥-secretase activities (2). In contrast to secreted A␤, intracellular A␤  is generated in the endoplasmic reticulum (ER)/intermediate compartment (IC) (3)(4)(5)(6). This suggests that both ␤-secretase and an A␤ 1-42 generating ␥-secretase are also active in the ER/IC.
In addition to the TGN and ER/IC amyloidogenic processing pathways, APP can be processed by ␣-secretase, which cleaves APP at position 16 within the A␤ domain to generate sAPP␣ (the ectodomain of APP ending at the ␣-secretase cleavage site) and C83 (the 83-amino acid COOH-tail of APP). ␥-Secretase cleavage of C83 does not produce A␤ but generates a 3-kDa fragment composed of residues 17-40 or 17-42 of the A␤ peptide (p3). This pathway is thought to be nonamyloidogenic and may not contribute substantially to AD pathology (1). Cell surface radioiodination experiments have shown that ␣-secretase cleavage of APP can occur at the plasma membrane (7).
Because the proportion of APP processed by ␤-secretase versus ␣-secretase may affect the amount of A␤ produced, the regulation of these two pathways may be critically important to the pathogenesis of AD. Indeed, mutations in APP found in a Swedish familial AD pedigree map to the ␤-secretase cleavage site in APP and favor ␤-secretase cleavage of APP (8,9). Thus, cells expressing these mutations secrete increased amounts of A␤ and decreased amounts of p3 as compared with cells ex-pressing wild-type APP. In contrast to the Swedish mutation, which increases ␤-secretase cleavage, activation of protein kinase C (PKC) has been shown to favor ␣-secretase cleavage at the expense of ␤-secretase cleavage. For example, treatment of cells with the PKC activator phorbol 12-myristate 13-acetate (PMA) increases secretion of sAPP␣ and decreases secretion of A␤ (10,11). Transgenic mice engineered to produce high levels of human A␤ also have decreased levels of brain A␤ following PMA treatment, suggesting that stimulation of ␣-secretase cleavage may be a useful intervention in AD (12).
The PMA-stimulated APP cleaving activity has recently been attributed to the ADAM (a disintegrin and metalloprotease) family member tumor necrosis factor-␣-converting enzyme (TACE) (13)(14)(15). TACE has been shown to cut APP at the ␣-secretase cleavage site in vitro, and inhibitors of TACE, such as tumor necrosis factor-␣ protease inhibitor (TAPI ((N-R-(2hydroxyaminocarbonyl)methyl)-4-methylpentanoyl-L-naphthylalanyl-L-alanine 2-aminoethyl amide)) and IC-3, block PMAstimulated secretion of sAPP␣ from cells. Furthermore, fibroblasts derived from TACE knockout mice generate lower levels of sAPP␣, and fail to show any regulated sAPP␣ secretion (13). The fact that TACE knockout cells and cells treated with TACE inhibitors still secrete some sAPP␣ (in the presence or absence of PKC activators) suggests that there may be multiple ␣-secretases involved in the cleavage of APP. Indeed, another closely related disintegrin metalloprotease, ADAM-10, can also cleave APP at the ␣-secretase site and may contribute to the cleavage of APP following phorbol ester stimulation (16).
Whereas these ADAM family members may account for the majority of PKC-regulated APP ␣-secretase cleavage (␣ reg ), there is an additional constitutively active ␣-secretase activity that is PKC-independent and is not inhibited by disintegrin metalloprotease inhibitors (␣ unreg ). Here, we examined the subcellular localizations of these two ␣-secretase cleavages and determined how activation of TACE/ADAM-10 further affects processing of APP by ␤and ␥-secretases.

EXPERIMENTAL PROCEDURES
Cell Culture-CHO cells stably transfected with the wild-type 695amino acid isoform of APP (APP 695 ), a familial AD-associated APP mutant in which the two amino acids immediately flanking the N terminus of the A␤ domain were mutated to asparagine and leucine (APP ⌬NL ) (17), and CHO Pro5 cells were grown and passaged three times per week in ␣-MEM (Life Technologies, Inc.) containing 10% fetal bovine serum and penicillin/streptomycin according to standard protocols.
Preparation of Semliki Forest Virus and Infection of CHO Cells-Semliki Forest virus (SFV) vectors expressing APP 695 (SFV-APP WT ), a TGN-targeting APP mutant in which a furin tail replaced the cytoplasmic domain of APP (SFV-APP TGN ; a gift from A. Chyung), or an APP mutant in which the third and fourth amino acids from the C terminus of APP have been changed to lysines (SFV-APP ⌬KK ) were prepared and titered as described previously (3,17,18). CHO-Pro5 cells were infected in serum-free medium at a multiplicity of infection of 10. After 1.5 h, complete growth medium was replaced and infection was allowed to proceed for another 1.5 h.
Metabolic Labeling, Immunoprecipitation, and Gel Electrophoresis-Cultured CHO cells were methionine-deprived by incubation in methionine-free Dulbecco's modified Eagle's medium (Life Technologies, Inc.) for 30 min before adding [ 35 S]methionine (250 or 500 Ci/ml in methionine-free Dulbecco's modified Eagle's medium ϩ 5% dialyzed fetal bovine serum; NEN Life Science Products). Cells were pulse-labeled either for 10 min (for C-terminal fragment pulse-chase analysis) or for 1 h and chased with growth medium for different lengths of time as indicated in the figure legends. The metalloprotease inhibitor TAPI (10 M in Me 2 SO; Peptides International) was added for 30 min prior to chase as well as during the chase. The peptide aldehyde protease inhibitor MG-132 (200 M in Me 2 SO; Peptides International) was added in a similar manner. PMA (Sigma) was included at a concentration of 10 M (in Me 2 SO).
Sandwich ELISA-Sandwich ELISA was performed as described previously using mAbs specific for different species of A␤ (19,20). BAN-50 (a mAb specific for the first 10 amino acids of A␤) was used as a capturing antibody, and horseradish peroxidase-conjugated BA-27 (a mAb specific for A␤  ) and horseradish peroxidase-conjugated BC-05 (a mAb specific for A␤  ) were used as reporter antibodies. To calibrate the sensitivity of the ELISA for detecting A␤, synthetic A␤ 1-40 and A␤ 1-42 peptides (Bachem Bioscience Inc., King of Prussia, PA) were used to generate standard curves. The BAN50, BA-27, and BC-05 mAbs were prepared and characterized as described previously (20,21).
Western Blot Analyses-CHO Pro 5 cells were lysed as described above. Proteins were resolved on 7.5% Tris-glycine gels and probed with a mAb raised against the catalytic domain of TACE or C-15 (goat polyclonal antiserum raised against the C-terminal domain of TACE; Santa Cruz). Anti-mouse or anti-goat HRP-conjugated secondary antibodies were used followed by visualization with ECL.
Trypsin and Triton X-100 Treatment of Cells-Cells were treated with and without PMA for 15 min. Cells were then treated with trypsin (10 g/ml in phosphate-buffered saline without calcium or magnesium) or with trypsin and 0.1% Triton X-100 on ice. Cells were washed in phosphate-buffered saline containing 100 g/ml soybean trypsin inhibitor and lysed for Western blotting or immunoprecipitation as before.

PKC Stimulation Increases ␣-Secretase
Cleavage of APP While Reducing ␤-Secretase Cleavage and A␤ Production in an ADAM-dependent Manner-The effects of PMA and the ADAM protease inhibitor TAPI on sAPP␣ secretion were examined in CHO cells stably transfected with APP 695 (CHO 695 ). CHO 695 cells express high levels of APP and constitutively secrete easily measurable amounts of sAPP␣. As early as 15 min following PMA treatment, secretion of sAPP␣ was increased by approximately 3-5-fold (data not shown). The effects of PMA on sAPP␣ secretion were cumulative for approximately 1 h, but over longer time periods (3-5 h), this effect waned, and sAPP␣ secretion returned to baseline levels (data not shown). Therefore, we examined APP secretion during the first hour following PMA treatment by metabolically labeling CHO 695 cells with [ 35 S]methionine for 1 h followed by a chase in the presence or absence of PMA (Fig. 1, A and C). sAPP␣ was quantitatively immunoprecipitated and visualized on Tris-glycine gels. Addition of PMA induced sAPP␣ secretion by approximately 3-fold. However, the protease inhibitor TAPI abolished the PMA effect and even reduced sAPP␣ levels to below control levels, suggesting that TACE/ADAM-10-mediated cleavage of APP accounts not only for the PKC-regulated ␣-secretase activity (␣ reg ) induced by PMA, but for much of the constitutive ␣-secretase activity as well. Because TAPI did not fully inhibit sAPP␣ release, another TAPI insensitive ␣-secretase may also cleave APP in both stimulated and unstimulated cells. To eliminate the possibility that the sAPP␣ recovered during the 1-h chase was produced prior to the addition of TAPI, TAPI was added as early as 1 h prior to PMA treatment. The effects of TAPI on APP secretion were maximal when TAPI was added 30 min prior to chase. No decrease in TAPI insensitive secretion was noted with longer pretreatments or higher concentrations of TAPI.
If ␤-secretase activity were limited by the availability of APP, then increased cleavage of APP by other secretases (such as ␣ reg ) could decrease ␤-secretase cleavage of APP and hence A␤ production. To determine whether or not increased ␣ reg led to reduced ␤-secretase activity, we examined the effects of PMA induced ADAM activity on ␤-secretion of APP, using an antibody that specifically recognizes sAPP␤ for immunoprecipitation (18). Fig. 1, B and D, shows that PMA reduced secretion of sAPP␤ by 23%. To determine whether activation of TACE/ ADAM-10 was involved in the decreased secretion of sAPP␤, we tested the effects of TAPI on sAPP␤. As shown in Fig. 1B, TAPI pretreatment restored sAPP␤ secretion to baseline levels. Increased ␣ reg activity also reduced production of A␤. CHO 695 cells were treated with PMA and/or TAPI and levels of secreted A␤ 1-40 and A␤ 1-42 were quantitated by sandwich ELISA. Fig.  2 shows that PMA decreased secretion of A␤ 1-40 and A␤ 1-42 by 51 and 31%, respectively. In contrast, pretreatment with TAPI restored secretion of both A␤ isoforms to control levels. Taken together, these data suggest APP is a limiting substrate for which both ␣ reg and the ␤-secretase activities compete. Furthermore, because enhanced ␣ reg cleavage of APP reduces both ␤and ␥-cleavage of APP, ␣ reg -cleavage of APP must occur prior to or concomitant with ␤-secretase cleavage.
PKC-regulated ␣-Secretase Competes Directly with ␤-Secretase for Cleavage of APP-The ability of TAPI and PMA to modulate the production of sAPP␣, sAPP␤, and A␤ suggested that the ␣ reg and ␤-secretases compete for limiting amounts of APP. If true, then PMA and TAPI should also affect the production of intracellular C-terminal fragments of APP produced by ␣and ␤-secretases. CHO 695 cells were labeled for 90 min followed by 45 min of chase in the presence or absence of PMA or TAPI. APP C-terminal fragments were immunoprecipitated with an antibody specific for the last 40 amino acids of APP (2493). As shown in Fig. 3, lane 1, the products of ␣and ␤-secretase cleavage (C83 and C99, respectively) were recovered at similar levels. In addition, a band corresponding to the last 89 amino acids of APP (C89) was recovered (as determined by radiosequence analysis), suggesting efficient cleavage of APP at residue 11 of A␤ in CHO cells. Following PMA treatment, levels of C99 and C89 decreased by 40%, whereas C83 levels increased 1.6-fold. Thus, changes in levels of intracellular C-terminal fragments of APP mirrored changes seen in sAPP␣ and sAPP␤ levels, confirming that activation of PKC resulted in increased ␣-secretase cleavage of APP at the expense of ␤-secretase cleavage. Similar to secreted sAPP␣, C83 levels were decreased by 46 Ϯ 6% by TAPI. As before, TAPI pretreatment completely blocked the effects of PMA, decreasing C83 recovery by 53 Ϯ 7% (Fig. 3, lane 4), eliminating the possibility that a non-metalloprotease-mediated event (such as a general effect of PMA on APP trafficking) is responsible for this effect.
To test whether the C83 generated by ␣ reg and constitutively generated C83 are both substrates for ␥-secretase, we treated cells with MG132, which has been shown to inhibit the intracellular ␥-secretase activity that produces A␤  . As expected, MG132 treatment resulted in significantly increased recovery of C99, C89, and C83 (Fig. 3, lanes 5-8), indicating that all of these species can be turned over by a MG132-sensitive pathway, such as the A␤ 1-40 producing ␥-secretase. However, because MG132 is known to have other effects, such as proteasome inhibition, we cannot eliminate the possibility that these fragments are also turned over by other degradative processes. Interestingly, MG132 had similar effects on C83 levels in both unstimulated and PKC-stimulated cells, suggesting that C83 produced by ␣ reg and ␣ unreg are both turned over by a MG132sensitive ␥-secretase. Finally, TAPI pretreatment rescued C99 recovery to baseline levels in the presence of PMA, further supporting the hypothesis that that ␣ reg competes with ␤-secretase for cleavage of APP.
Because ␣ reg -cleavage decreased ␤-secretase cleavage of APP, we hypothesized that ␣ reg and ␤-secretase cleavage should either occur contemporaneously or sequentially, with ␣ reg cleavage preceding ␤-secretase cleavage. Because steady-  1. PMA stimulation of APP ␣-secretase cleavage causes decreased ␤-secretase cleavage due to metalloprotease activation. CHO 695 cells were pulse-labeled for 1 h and chased with and without PMA or TAPI for 45 min. sAPP␣ (A) and sAPP␤ (B) were immunoprecipitated from medium with 6E10 and 53, respectively, and were resolved on 7.5% Tris-glycine gels. Representative gels of three separate experiments are shown. sAPP␣ (C) and sAPP␤ (D) levels were quantitated using a Phos-phorImager. Means and S.E. are shown for three experiments. Single factor analysis of variance revealed p Ͻ .001; post hoc analysis showed p Ͻ 0.05 (*) and p Ͻ 0.01 (**).
state labeling experiments cannot address the temporal sequence of ␣and ␤-secretase cleavage of APP, we used a pulsechase paradigm to test this hypothesis. To do this, CHO ⌬NL cells were pulse-labeled for 10 min in the presence or absence of PMA, and C-terminal APP intracellular fragments were immunoprecipitated at various time points during the chase period. Due to the low levels of C99 produced under these conditions, we used CHO cells that express the APP ⌬NL mutant, which is cleaved more efficiently than APP WT by ␤-secretase. As shown in Fig. 4, C99 and C83 were made contemporaneously in the absence of PMA. PMA treatment shifted APP processing away from ␤-secretase cleavage and toward ␣-secretase cleavage (as evidenced by C83 accumulation at the expense of C99), although without altering the total level of C-terminal fragment generation. This argues against a PMA induced global alteration in APP processing (such as would be expected from alterations in APP trafficking) and supports direct competition between these secretases. Importantly, even with favored ␤-secretase cleavage of APP in CHO ⌬NL cells, C99 was not produced prior to C83, arguing against the possibility that PMA stimulation results in C99 turnover to C83.
␣ reg Cleavage of APP Occurs in the Golgi Compartment-Because we established that ␣ reg competes with ␤-secretase to cleave APP and because ␤-secretase activities have been detected in the ER and the TGN, we sought to identify the site of ␣ reg cleavage. To do this, we utilized two organelle-specific APP targeting constructs: APP ⌬KK and APP TGN . APP ⌬KK is an APP construct that has been modified to contain the dilysine ER retention motif; expression of APP ⌬KK in cells restricts APP to the ER/IC. APP TGN is an APP/furin hybrid that contains the extracellular and transmembrane domains of APP, fused to the cytoplasmic domain of rat furin. 2 The cytoplasmic domain of furin is sufficient for targeting furin to the TGN (22) and serves to target APP TGN to the TGN as well. CHO cells infected with SFV-APP WT , SFV-APP TGN , or SFV-APP ⌬KK were metabolically labeled with [ 35 S]methionine for 1 h, pretreated with TAPI for 30 min, and chased with or without PMA for an additional 45 min. sAPP␣ was immunoprecipitated with 6E10, resolved on Tris-glycine gels, and quantitated using a PhosphorImager (Fig. 5A). In order to standardize results for varying levels of SFV infection, sAPP␣ levels were normalized to levels of fulllength APP (Fig. 5, B and C).
As shown in Fig. 5A, infection of cells with SFV-APP ⌬KK completely abrogated both constitutive and PMA-stimulated sAPP␣ secretion, suggesting a post-ER/IC locus for both ␣-secretases. In contrast, infection of cells with SFV-APP TGN resulted in a 49% decrease in sAPP␣ secretion as compared with cells infected with SFV-APP WT . This decrease could be due to increased ␤-secretase cleavage in the TGN because this is the major site that produced secreted A␤. Alternatively, the decrease in sAPP␣ secretion could be explained if the constitutively active ␣-secretase were located in a post-TGN compartment. Indeed, following TAPI treatment, sAPP␣ secretion was reduced by 85% in APP TGN -expressing cells, indicating that ␣ reg accounted for the bulk of ␣-secretase activity in the TGN and that most of the non-TACE/ADAM-10-mediated ␣-secretase activity is confined to a post-TGN locus. Following PMA stimulation, sAPP␣ secretion was increased to similar extents in APP TGN -and APP WT -expressing cells. Because APP TGN is predominantly retained in the TGN, this suggests that ␣ reg cleavage of APP occurs in the TGN. In contrast, in the absence of PMA treatment, ␣-secretase cleavage of APP occurs both in the TGN, where TACE/ADAM-10 is required for cleavage, and post-TGN, where cleavage of APP is TACE/ADAM-10-independent.
In order to confirm that the TGN is the site in which ␣ reg competes with ␤-secretase for cleavage of APP, we quantitated production of intracellular and secreted A␤ species in CHO cells expressing APP WT , APP TGN , and APP ⌬KK (Fig. 6). PMA attenuated A␤ 1-40 secretion by a TAPI-sensitive mechanism in both APP WT -and APP TGN -expressing cells. This further confirms that ␣ reg and ␤-secretase both can cleave APP in the TGN. Interestingly, PMA treatment also decreased levels of intracellular A␤ 1-40 but not intracellular A␤ 1-42 . This may reflect the different sites of production for intracellular A␤ 1-40 and A␤ 1-42 , because A␤ 1-40 is largely produced in the TGN, whereas intracellular A␤ 1-42 is produced in the ER/IC (3,4). Indeed production of intracellular A␤ 1-42 by ER retained APP ⌬KK was unaffected by PMA or TAPI.
Mature TACE Is Found Intracellularly-Because experi- FIG. 3. Normal APP processing is restored by ADAM inhibition following PMA treatment. CHO 695 cells were pulse-labeled for 90 min and chased for 45 min in the presence or absence of PMA, TAPI, and MG132. Intracellular APP C-terminal fragments were recovered from cell lysates by immunoprecipitation with 2493 and were resolved on a 10%/16% discontinuous gradient Tris-Tricine gel. A representative of three experiments is shown, with bands corresponding to full-length APP, C99, C89, and C83 labeled. Contrast has been enhanced on the lower portion of the gel in order to emphasize the APP C-terminal fragments.
FIG. 4. Stimulation of PKC with PMA decreases ␤-secretase cleavage of APP due to competition between ␣-secretase and ␤-secretase for APP. CHO ⌬NL cells were pulse-labeled with and without PMA for 10 min and chased with and without PMA for the indicated times. Intracellular APP C-terminal fragments were immunoprecipitated with 2493 and resolved on discontinuous gradient Tris-Tricine gels. Representative of three experiments is shown, with bands corresponding to full-length APP, C99, C89, and C83 labeled. Contrast has been enhanced on the lower portion of the gel in order to emphasize the APP C-terminal fragments. ments with CHO 695 cells suggested that ␣ reg , like ␤-secretase, cleaves APP inside the cell, we asked whether the subcellular localization of TACE and ADAM-10 are consistent with intracellular sites of activity. Indeed, mature ADAM-10 is found both on the cell surface and intracellularly. Although mature TACE has also been labeled at the cell surface, we asked whether or not mature TACE is also found inside the cell. To this end, CHO cells were incubated with or without trypsin and Triton X-100 for 20 min, and levels of endogenous nonprocessed ϳ110-kDa TACE and proteolytically activated ϳ85-kDa TACE were measured by immunoblot using a mAb that recognizes the catalytic domain of TACE. As shown in Fig. 7, both the immature and mature forms of TACE were recovered, and neither was susceptible to trypsin cleavage in the absence of Triton X-100, suggesting that both isoforms were present predominantly intracellularly.
Because TACE contains six potential N-linked glycosylation sites, it can be variably glycosylated, thus altering gel migration (14). In order to confirm the identity of the bands shown in Fig. 7, we documented TACE immunoreactive bands of similar molecular weights in THP-1 monocytes, a cell line previously shown to express abundant TACE (14). Further, antibodies raised against the catalytic domain of TACE and against the C-terminal domain of TACE both detected the ϳ85and ϳ110-kDa bands (data not shown).
In order to test whether PMA stimulation induced translocation of TACE to the cell surface, cells were pretreated with PMA for 15 min prior to trypsinization and quantitated by Western blot as before. PMA treatment did not alter the distribution or maturation of TACE (data not shown), thus indicating that even following PKC activation TACE remains intracellular.

DISCUSSION
In a subset of familial AD cases, increased A␤ production induced by mutations in the APP gene appears to be sufficient to cause AD. Thus, insights into the regulation of APP cleavage are crucially important to understanding the pathogenesis of AD. The APP ectodomain is cleaved by ␣-secretase between residues 16 and 17 of A␤, thereby precluding production of amyloidogenic A␤. Whereas cells constitutively cleave APP at this site, ␣-secretase cleavage is dramatically up-regulated following treatment of cells with PKC activators such as PMA. The data presented here as well as in two other recent reports (13,16) have implicated TACE and ADAM-10 as the ␣ reg because the activity of both of these enzymes are increased following PKC activation.
Currently, it is unclear whether TACE or ADAM-10 is the major ␣ reg because both TACE and ADAM-10 have been demonstrated as the ␣ reg that cleaves APP. For example, although experiments using cells generated from knockout animals showed that TACE could account for all of the regulated secretion of APP, overexpression of ADAM-10 is sufficient to increase regulated cleavage of APP under some circumstances (13,16). Thus, TACE may be the physiological catalyst for sAPP␣ secretion, but when ADAM-10 is abundant, it can also cleave APP. However, the high degree of homology between these two enzymes makes it difficult to distinguish their activities, especially because they are inhibited by similar compounds (16,24), and mutant forms of either one may have a dominant negative effect on the other wild-type enzyme (14).
In addition to increasing ␣-secretase cleavage, PKC activation also leads to decreased ␤-secretase cleavage of APP. How might PKC activation increase ␣-secretase cleavage of APP and decrease ␤-secretase cleavage of APP? Previous studies have suggested that PKC may alter APP trafficking by increasing movement of APP out of the TGN and to the cell surface via secretory vesicles (23). However, our data support a role of PKC in the activation of ␣ reg , and the evidence is as follows. First, if ␤-secretase activity were present in the TGN, and ␣-secretase were present at the cell surface, redistribution of APP from the TGN to the plasma membrane would result in increased ␣-secretase cleavage and decreased ␤-secretase cleavage. However, we did not observe any redistribution of APP to the cell surface following PMA treatment. Second, we did not detect any changes in APP trafficking by pulse-chase analysis upon PKC activation (Fig. 4). Instead, we found that PKC activation led to increased enzymatic activity of ␣ reg and that this ␣ reg competes with ␤ secretase (which has previously been localized to the TGN) for cleavage of APP. Finally, our observation that the PMA-induced reduction in secretion of sAPP␤ and A␤ was completely blocked by TAPI indicates that PMA affects APP processing by activating TAPI-sensitive metalloprotease(s) rather than by altering APP trafficking.
However, we observed incomplete inhibition of ␣ reg by TAPI, which could arise from the contribution of TAPI-insensitive metalloproteases to ␣ reg or from incomplete inhibition of TACE/ ADAM-10 by TAPI. Because fibroblasts derived from TACE knockout mice do not show any TAPI-insensitive ␣ reg (13), it is likely that the small amount of residual ␣ reg activity in the presence of TAPI reflects incomplete inhibition of TACE. Thus, these data further support that TACE is the enzyme that is responsible for the majority of ␣ reg activity in our system.
The implication of ADAM family members in the ␣ reg -cleavage of APP allowed us to examine the dynamic relationship between ␣ reg cleavage of APP to generate sAPP␣ and p3 and ␤-secretase cleavage to generate sAPP␤ and A␤. We observed an inverse relationship not only between sAPP␤ and sAPP␣ secretion, but also between recovery of C83 and C99, the products of ␣and ␤-secretase activities, respectively. This inverse relationship of ␣ reg and ␤-secretase activities suggested that TACE/ADAM-10 and ␤-secretase may compete to cleave limiting amounts of APP (as illustrated in Fig. 8). Significantly, in the absence of PMA, CHO ⌬NL cells produce similar amounts of C83 and C99, but PMA increased C83 recovery while decreas- FIG. 5. PMA-stimulated, ADAM-dependent ␣-secretase cleavage of APP occurs in the TGN. SFV-infected CHO cells engineered to express APP WT , APP TGN , or APP ⌬KK were pulse-labeled for 60 min and chased for 45 min in the presence or absence of PMA and TAPI. A, sAPP␣ was recovered from the medium by immunoprecipitation with 6E10 and resolved on 7.5% Tris-glycine gels. B, intracellular APP was recovered from cell lysates by immunoprecipitation with Karen antiserum and resolved on 7.5% Tris-glycine gels. APP TGN migrates more slowly than APP WT because the furin domain is 16 amino acids longer than the normal cytoplasmic tail of APP. C, sAPP␣ levels were quantitated and normalized for intracellular APP levels. Representative gels, means, and S.E. of three separate experiments are shown. Single factor analysis of variance revealed p Ͻ .001; post hoc analysis showed p Ͻ 0.05 (*) and p Ͻ 0.01 (**).
ing C99 recovery, and this effect was inhibited by TAPI, further confirming that the competition occurred via a TACE/ADAM-10-dependent mechanism. Alternatively, these data could be explained by sequential cleavage of APP by ␤-secretase (generating C99) and then by ␣ reg (generating C83). If so, then PMA stimulation would result in increased C83 production coupled with increased C99 turnover. Two lines of evidence suggest that this is not the case. First, although ␣ reg cleavage of C99 could explain both decreased recovery of C99 and decreased production of A␤, it could not explain the parallel decreases in sAPP␤ observed with PMA treatment. Second, ␣ reg turnover of C99 is unlikely because pulse-chase analysis of APP C-terminal fragments showed the contemporaneous production of C99 and C83 in CHO 695 cells. PMA treatment did, however, result in decreased levels of total cellular C99 (Fig. 3). Because PMA treatment decreases C99 levels even in the presence of MG132, which has been shown to inhibit ␥-secretase cleavage of this species (25)(26)(27), this suggests that activated ␣ reg competes successfully with ␤-secretase for cleavage of APP before ␥-secre-  In unstimulated cells, a fraction of full-length APP is cleaved in the TGN by both ␤-secretase and a low level of regulated ␣-secretase. Additional fulllength APP is transported to the plasma membrane (PM) and cleaved by unregulated ␣-secretase. Following PMA stimulation, regulated ␣-secretase cleavage of APP in the TGN is dramatically increased, and ␤-secretase cleavage in the TGN is almost completely abrogated. tase cleavage.
The results of our experiments also provide insights on aspects of ␤-secretase(s) activities. For example, recent reports suggest that APP ⌬NL and APP WT may be cleaved by distinct ␤-secretases (17,27,28). Because the ␤-secretase that cleaves APP ⌬NL may be localized in different organelles we asked whether ␣ reg similarly competes with this secretase. Pulsechase analysis showing contemporaneous production of C99 and C83 from APP ⌬NL and the successful competition away from ␤-secretase cleavage with stimulation of ␣ reg support the idea that the APP ⌬NL cleaving ␤-secretase most likely also reside in the same organelle as ␣ reg . Another characteristic of ␤-secretase revealed by our experiments is that although increasing ␣ reg -cleavage caused decreased ␤-secretase cleavage of both APP WT and APP ⌬NL , decreasing ␣ reg -cleavage did not increase ␤-secretase cleavage beyond baseline levels. This suggests that ␤-secretase cleavage is not limited by substrate availability under steady-state conditions and that only after PKC stimulation is ␤-secretase substrate limited.
Previous studies have shown that the TGN is a major locus for ␤-secretase cleavage, and our data are consistent with ␣ reg competing with ␤-secretase for cleavage of APP in this organelle (Fig. 8). Indeed, our observation that APP targeted to the TGN by means of a cytoplasmic furin domain was readily cleaved by ␣ reg and that A␤ production was attenuated by PMA stimulation of ␣ reg supports the idea that ␣ reg -cleavage of APP occurred in this organelle. We confirmed the successful targeting of APP TGN by verifying the exclusion of APP TGN from the cell surface, thereby ensuring the reliability of this observation (data not shown). Although we cannot formally exclude the possibility that undetectable APP TGN is present at the cell surface, it seems unlikely that such a small pool of APP could be the substrate for ␣ reg cleavage given the high levels of sAPP␣ generated by PMA treatment.
Although our data on APP TGN are consistent with the notion that ␣ reg does not cleave APP beyond the trans-Golgi, APP TGN may be cleaved prior to its arrival in the TGN. However, because ␣ reg -secretase failed to cleave the ER/IC restricted APP ⌬KK , we can conclude that ␣ reg -cleavage occurs in a post-ER/IC, pre-plasma membrane compartment. Finally, because the majority of APP are found in the TGN and because the bulk of ␤-secretase activity resides in the TGN, our data best support the TGN as the locale for ␣ reg -cleavage of APP. In contrast, APP TGN was not efficiently cleaved by the non-ADAM-dependent constitutively active ␣-secretase (Fig. 6), supporting that the bulk of this ␣ unreg -cleavage occurs on the cell surface, as suggested previously (7,29).
Our demonstration of ␣ reg activity in the TGN would require that mature TACE and/or ADAM10 be in this organelle. Indeed, we found a significant amount of TACE/ADAM10 that are trypsin-resistant, supporting an intracellular location of these enzymes although the mature form of both of these proteases have been detected on the cell surface (14,16). The activity of TACE in the TGN would match what is known about MDC9 and MDC15, because both are cleaved by furin-like proprotein convertases in the late Golgi network, and are present as mature, active enzymes in the TGN (24,30). Thus, our demonstration of intracellular APP cleavage by TACE/ ADAM-10 provide novel evidence implicating ADAM family members in mechanisms of intracellular protein maturation/proteolysis.
The competition between TACE/ADAM-10 and ␤-secretase for intracellular cleavage of APP demonstrated here has potential therapeutic implications for the treatment of AD. Although we have carefully examined the regulation of ␣ reg by PMA, phorbol esters are not the only regulators of APP cleavage. For example, glutamate and the muscarinic agonist carbachol have both been shown to increase ␣ reg cleavage of APP (31)(32)(33). Although these agents may work through the common pathway of PKC activation, it is unclear how PKC in turn activates TACE. One possibility is that PKC directly phosphorylates TACE. Indeed, the ADAM family member MDC9 is phosphorylated following PMA stimulation (24). However, the role of PKC in ADAM phosphorylation and the consequences of such phosphorylation on protease activity are still unknown. Interestingly, the mitogen-activated protein kinase kinase inhibitor PD98059 and an inactive extracellular signal-regulated protein kinase mutant have been shown to partially antagonize regulated ␣-cleavage of APP (34). Whether extracellular signalregulated protein kinase is involved in regulating ADAM activity or whether it plays a role in trafficking APP or ADAM to the TGN remains to be determined. Future studies will define further how PKC increases the activity of TACE/ADAM-10, and such information may be useful in the discovery of new therapeutic strategies for the treatment of AD.