Dual Role of α-Secretase Cleavage in the Regulation of γ-Secretase Activity for Amyloid Production*

Processing of the amyloid precursor protein (APP) by β- and γ-secretases generates pathogenic β-amyloid (Aβ) peptides associated with Alzheimer disease (AD), whereas cleavage of APP by α-secretases precludes Aβ formation. Little is known about the role of α-secretase cleavage in γ-secretase regulation. Here, we show that α-secretase-cleaved APP C-terminal product (αCTF) functions as an inhibitor of γ-secretase. We demonstrate that the substrate inhibitory domain (ASID) within αCTF, which is bisected by the α-secretase cleavage site, contributes to this negative regulation because deleting or masking this domain turns αCTF into a better substrate for γ-secretase. Moreover, α-secretase cleavage can potentiate the inhibitory effect of ASID. Inhibition of γ-secretase activity by αCTF is observed in both in vitro and cellular systems. This work reveals an unforeseen role for α-secretase in generating an endogenous γ-secretase inhibitor that down-regulates the production of Aβ. Deregulation of this feedback mechanism may contribute to the pathogenesis of AD.

The amyloid precursor protein (APP) 2 is sequentially cleaved by ␤and ␥-secretases to generate A␤ peptides, which are widely considered to play a causative role in the pathogenesis of Alzheimer disease (AD) (1). This pathway also generates soluble APP␤ (sAPP␤) and APP intracellular domain. Alternatively, APP can be processed by ␣-secretase that cuts within A␤ peptides, which precludes the formation of toxic peptides. Similarly, the ␣/␥ pathway leads to formation of soluble APP␣ (sAPP␣), the P3 peptide, and the APP intracellular domain. It appears that ␥-secretase cleavage of substrates first requires another protease that removes the majority of the extracellular domain of the substrate. The sequential cleavage of APP is characteristic of regulated intramembrane proteolysis (RIP) (2). RIP generally requires two proteolytic steps, whereby the second intramembrane cleavage is dependent on the first cleavage, such as in Notch and sterol regulatory element binding proteins signaling. RIP has been found to govern sterol regulation, cell fate determination, unfolded protein responses, growth factor activation, mitochondria membrane remodeling, and apoptosis (2)(3)(4)(5).
A distinctive feature of APP processing compared with other RIPs is the existence of two proteases (␤-and ␣-secretase) that are both capable of executing the first cleavage. Recent studies have shown that ␤-secretase cleavage removes the large extracellular domain of APP, freeing the N terminus of ␤CTF, which is recognized by nicastrin (6). Recruiting ␤CTF to the docking site makes the substrate accessible to the active site of ␥-secretase for hydrolysis. However, the role of nicastrin for substrate binding has been controversial (7)(8)(9). Nevertheless, ␤-secretase cleavage converts a latent substrate of ␥-secretase into an active one. Whether ␣-secretase plays a role similar to that of ␤-secretase in the regulation of APP processing by ␥-secretase is unknown. It has been suggested that ␣-secretase competes with ␤-secretase for the APP substrate and thereby reduces the production of A␤ (10 -13). Therefore, elevation of ␣-secretase activity has been suggested as a therapeutic strategy to reduce the production of A␤ (14). Also, it has been reported that AD patients have reduced levels of ␣-secretase activity (as reflected by sAPP␣ levels in cerebrospinal fluid) relative to that of healthy controls, but there was no difference in the level of sAPP␤ (15,16), which results from ␤-secretase cleavage. These observations of reduced sAPP␣ levels together with normal sAPP␤ levels in certain populations of AD patients indicates that ␣-secretase can modulate the rate of A␤ production without competing with ␤-secretase. In addition, transgenic mouse studies showed that ␣-secretase competes for APP substrate with ␤-secretase when ADAM10, an ␣-secretase candidate, was highly overexpressed (13). However, when a moderate level of ADAM10 was expressed in mice, A␤ was reduced, and no statistical difference of sAPP␤ between transgenic and control animals was detected (13), suggesting that there is no significant competition for the APP substrate between ␣and ␤-secretase. These studies suggest that ␣-secretase can modulate A␤ production by an undefined mechanism in addition to depletion of the APP substrate. Recently, we identified a substrate inhibitory domain (ASID) within ␤CTF that negatively modulates ␥-secretase activity for A␤ production (17). Furthermore, the APP Flemish mutation within ASID sequence reduces its inhibitory effect on ␥-secretase, resulting in increased A␤ production (17). After ␣-secretase cleavage, ASID becomes exposed at the N terminus of ␣-CTF. How ASID within ␣CTF modulates ␥-secretase is unknown and whether this domain is involved in ␥-secretase activity for A␤ production remain to be explored.
In this study, we have demonstrated that ␣-secretase cleavage of APP generates ␣CTF that functions as a ␥-secretase inhibitor for A␤ production. Our studies suggest that ASID within ␣CTF is responsible for this negative regulation. There-fore, ␣-secretase plays a dual role in modulation of ␥-secretase for A␤ production.

EXPERIMENTAL PROCEDURES
Expression of Biotinylated Recombinant Proteins-AviTag, a specific peptide sequence that can be biotinylated with biotin ligase, was cloned into a pIAD16 vector (18) to generate a pIAD16Avi plasmid. APP fragments with FLAG tag were inserted into the pIAD16Avi vector. For expression of biotinylated protein, pIAD16Avi-APP and pACYC164, which encodes biotin ligase, were co-transformed in BL21 (DE3) cells. When bacteria growth reached 0.4 -0.8 at A 600 , isopropyl 1-thio-␤-D-galactopyranoside (100 M) was added to induce target protein expression in the presence of 50 M biotin. Cells were pelleted and lysed by French press. The soluble fraction was subjected to standard amylose affinity chromatography through a maltose-binding protein (MBP) tag. The purified protein was treated with thrombin at 16°C overnight to cleave between the MBP tag and AviTag. The biotinylation of target proteins was verified by LC-MS/MS.
In Vitro ␥-Secretase Assay-The recombinant proteins were incubated with ␥-secretase (40 g/ml) in the presence or absence of 1 M ␥-secretase inhibitor L-685,458. The assays were performed as previously described (17,19,20) using electrochemiluminescence (ECL) technology. The amount of product was determined using synthetic peptide or recombinant standards.
Cell-based A␤ Production Assay-HEK293 cells were cultured in Dulbecco's modified Eagle's medium (DMEM), supplemented with 10% fetal bovine serum and penicillin. HEK293 cells stably transfected with APP were maintained in DMEM with 10% fetal bovine serum plus 1 g/l puromycin. When the cells reached 70% confluence in 6-well plates, they were transiently transfected with 20 ng of pmaxGFP (Amaxa) and 2-g target plasmids as indicated in individual experiments. After transfection, fresh media with or without ␥-secretase inhibitor were added to the cells. After a 48-h total incubation, conditioned media were collected and diluted in radioimmuneprecipitation assay buffer. Secreted A␤ Overexpression of TACE, an ␣-secretase like protease, would enhance the production of sAPP␣ and ␣CTF, reduce the production of sAPP␤ and ␤CTF, and have little effect on the total ␥-secretase products (A␤ ϩ P3). G, model of dual function of ␣CTF. In addition to substrate competition, ␣-secretase cleavage negatively modulates ␥-secretase activity that concurrently leads to an accumulation of ␤CTF and a reduction of A␤ and the total ␥-secretase products (A␤ ϩ P3).
peptides were detected by ECL assay using biotinylated 6E10 or biotinylated 4G8 and ruthenylated G2-10 antibodies (20). Conditioned media and cell lysates were analyzed by Western blotting using sAPP␤, 6E10, W0-2, and CT15 antibodies (see corresponding figures for antibody recognition site of APP).
Immunoprecipitation and ␤CTF ECL Assay-The mock (empty vector) or tumor necrosis factor-␣-converting enzyme (TACE)-transfected HEK293-APP cells (1 well of a 6-well plate each) were collected and lysed with 200 l of radioimmuneprecipitation assay buffer with a mixture of protease inhibitors. After rotating at 4°C for 1 h, cell lysates were centrifuged at 13,000 rpm for 5 min, and the supernatants were subjected to immunoprecipitation with W0-2 or 4G8 antibody and blotted with varying antibodies.
The mock-or TACE-transfected HEK293-APP cell lysate were also incubated with biotinylated 6E10, APPc, and ruthenylated anti-rabbit secondary antibodies for 3 h; magnet streptavidin beads were then added and incubated for 30 min. The assay mixtures were analyzed by ECL technology.

RESULTS
Expression of ␣-Secretase-like Activity Causes an Accumulation of ␤CTF and a Reduction of A␤ in Cells-To explore the undefined role of ␣-secretase in regulation of ␥-secretase activity, we examined the effect of TACE on the processing of APP.
Both ADAM10 and TACE have been found to exhibit ␣-secretase activity (21,22). Therefore, the expression of TACE or ADAM10 is able to promote ␣CTF production. We expressed TACE fused with a HA tag in HEK293-APP cells and examined its effect on APP processing. We first confirmed TACE protein expression by Western blot analysis using an anti-HA antibody (Fig. 1A, top panel). The expression of TACE had only a moderate effect (15% reduction) on the level of APP (Fig. 1A, middle panel). Equal protein loading was confirmed by anti-␤-tubulin Western blotting (Fig. 1A, bottom panel).
Although TACE can cleave multiple substrates, we focused on its action in APP processing. We analyzed the secreted APP species that include sAPP␣, sAPP␤, and A␤ in conditioned media. sAPP␣ and sAPP␤, which refer to the liberated N-terminal fragments of APP resulting from the cleavages of ␣secretase and ␤-secretase, were specifically recognized by the 6E10 and anti-sAPP␤ antibodies, respectively. We found elevated sAPP␣ levels, indicating that the transiently expressed TACE increases ␣-secretase-like processing of APP. Furthermore, reduced sAPP␤ suggested that there was less substrate available for ␤-secretase cleavage (Fig. 1B), which is a consequence of TACE-mediated depletion of the APP substrate. This finding is consistent with previous reports of competition between ␣-secretase and ␤-secretase for the APP substrate (10,11,13).
We next determined the level of secreted A␤40 with 6E10 and G2-10 antibodies. The amount of A␤40 was significantly reduced by 82% following TACE overexpression (Fig. 1C), which is consistent with previous results seen in ADAM10 mouse studies (13). However, when the total amount of ␥-secretase-cleaved products including A␤ and P3, collectively known as X40, were examined, we found that up-regulation of TACE activity also resulted in a 53% reduction of X40 (Fig. 1C). Clearly, higher ␣-secretase activity leads to a reduction of total ␥-secretase-cleaved products (P3 and A␤40), suggesting that an increased production of ␣CTF suppressed overall ␥-secretase activity for APP processing. In addition, we have determined the amount P3 in media using synthetic A␤ as a standard. We found that the expression of TACE led to a reduction of P3 ϳ50% relative to vector control, supporting that ␥-secretase activity for production of both A␤ and P3 is reduced.
To examine further this notion that ␣CTF functions as an endogenous inhibitor of ␥-secretase in cells, we determined the effect of TACE expression on cellular ␣CTF and ␤CTF frag- There is a thrombin cleavage site between MBP and the target proteins. The cleavage sites of thrombin and ␣and ␥-secretase are indicated by arrows. The recognition epitope of antibodies that have been used in the assay of D are indicated. B and C, analyses of ␣CTF and ␤CTF. Purified proteins were separated by SDS-PAGE and stained by Coomassie Blue (B). The protein masses were determined by LC-MS (C). The molecular masses of ␣CTF and ␤CTF were calculated through deconvolution of the mass-to-charge ratio (m/z). D, in vitro ␥-secretase activity for X40 production from ␣CTF and ␤CTF substrates. Each substrate was incubated with 4 g of HeLa membrane at 0.1, 0.5, and 1 M in the presence of 0.25% CHAPSO. After a 2.5-h incubation, the reaction was stopped by adding radioimmuneprecipitation assay buffer, and the product X40 was assayed with biotinylated 4G8 and ruthenylated G2-10 antibodies by ECL (mean Ϯ S.E. (error bars), n Ͼ 3). The amount of X40 was determined using synthetic A␤40 and P3 peptides as standards. Note: in all figures, the assay background was defined in the presence of 1 M L-685,458 for in vitro assays. ␥-Secretase activity was calculated by subtracting the assay background from the signal that was detected in the absence of inhibitor (dimethyl sulfoxide only).
ments. First, cell lysates were immunoprecipitated with the 4G8 antibody and blotted with the APPc antibody (Fig. 1D, top  panel). Clearly, the expression of TACE augmented the production of ␣CTF, which is consistent with the increase in secreted sAPP␣. In contrast to reduced sAPP␤, there was an increase in ␤CTF, another product of ␤-secretase (Fig. 1, D and  E). This cannot be simply explained by the substrate competition model because a reduction in ␤CTF would be predicted (Fig. 1F). The increased accumulation of ␤CTF in the TACEexpressing cells was confirmed by multiple analyses. First, we immunoprecipitated ␤CTF with W0-2 antibody and probed with W0-2 and FCA-18 antibodies. FCA-18 specifically recognizes the ␤-secretase-cleaved N terminus of ␤CTF, and W0-2 binds to the A␤1-16 fragment. Both blots (Fig. 1D, middle and  bottom panels) showed that the expression of TACE led to increased ␤CTF species, rather than a reduction. Second, we demonstrated that the level of ␤CTF in the TACE expression cells is ϳ2.5-fold higher than that in the control cells (Fig. 1E) with 6E10 and APPc antibodies using ECL technology. All of these studies showed that there is more ␤CTF in the TACEexpressing cells than in control cells, suggesting a reduction of ␥-secretase activity for ␤CTF processing (Fig. 1G).
Although the level of both ␣CTF and ␤CTF was higher in the TACE-expressing cells than the vector-control cells, the total ␥-secretase-cleaved products, including both P3 and A␤, were diminished, indicating that expression of ␣-secretase like activity could have dual effects on A␤ production. First, competing with ␤-secretase for APP substrate results in a reduction of A␤. Second, ␣CTF functions as an endogenous inhibitor that reduces ␥-secretase activity for A␤ production. This novel role of ␣-secretase is supported by the finding that increased ␣-secretase like activity leads to a reduction of ␥-secretase activity for A␤ production and an accumulation of ␤CTF in cells despite a reduction in sAPP␤. However, whether functional inhibition of total ␥-secretase activity is attributed to the ASID (17) within ␣CTF needs to be further investigated. Nevertheless, these studies suggest that ␣CTF is a poor substrate and/or an inhibitor that negatively modulates ␥-secretase activity.
␣CTF Is a Poor Substrate of ␥-Secretase in Vitro-To compare the reactivity ␣CTF and ␤CTF with ␥-secretase directly, we overproduced recombinant proteins as MBP fusion proteins with a thrombin cleavage site ( Fig. 2A). Purified fusion proteins were treated with thrombin and analyzed by SDS-PAGE (Fig. 2B). The molecular mass of ␣CTF-FLAG and ␤CTF-FLAG was confirmed by LC-MS (Fig. 2C). The measured molecular mass of ␣CTF-FLAG and ␤CTF-FLAG was 10,621 and 12,557 kDa, respectively, and matched the calculated masses 10,617 and 12,554. Each substrate was incubated with HeLa membrane at three concentrations in the presence and absence of 1 M L-685,458, a potent ␥-secretase inhibitor (23). The signal difference between L-685,458-treated and untreated samples is attributed to ␥-secretase activity. The A␤40 site cleavage was detected by a pair of antibodies: biotinylated 4G8 and ruthenylated G2-10. The two substrates exhibit striking differences in their ability to be processed by ␥-secretase. The rate of ␥-secretase hydrolysis of ␤CTF-FLAG is 12-fold greater than for hydrolysis of ␣CTF-FLAG (Fig. 2D). This result indicates that ␣CTF is poor substrate, which supports our assertion that ␣CTF can function as an endogenous ␥-secretase inhibitor in cells.
ASID within ␣CTF Regulates ␥-Secretase Activity in Vitro-Recently, we demonstrated that ASID within ␤CTF negatively regulates the activity of ␥-secretase (17). This prompted us to determine whether this domain, which is located at the N terminus of ␣CTF, accounts for the poor reactivity of ␣CTF for ␥-secretase. Therefore, we attempted to delete or mutate this domain to assess its role. However, alteration of this inhibitory domain destroys the 4G8 antibody-binding epitope upon which our assay depends; hence, we introduced an AviTag into the N terminus of APP CTFs. AviTag, a specific 15-residue peptide, is recognized by biotin ligase that specifically catalyzes an attachment of biotin to the lysine residue within the AviTag. Biotinylation of the recombinant substrate allows us to monitor ␥-secretase activity directly using streptavidin beads and the G2-10 antibody. To facilitate protein isolation, we inserted an MBP/thrombin site ahead of the AviTag on these target proteins. First, we tested this strategy by directly fusing the transmembrane domain of APP behind the MBP/thrombin-AviTag. A target protein was co-expressed with biotin ligase in the presence of biotin. The biotinylated APP transmembrane domain, which is hereafter referred to as Sb1, was confirmed by LC-MS analysis. The Sb1 protein was incubated with HeLa membrane, and the A␤40 site cleavage was detected using the G2-10 antibody. ␥-Secretase effectively cleaved the Sb1 substrate (Fig. 3A), and therefore we constructed two additional substrates, Sb2 and Sb3, which both contain the A␤17-23 sequence, with Sb3 also including the C-terminal tail of APP (Fig. 3A). Very low 40-site specific ␥-secretase cleavage product was detected using both the Sb2 and Sb3 substrates. However, deleting the ASID (A␤17-23) as we previously reported generates a better substrate (Sb4) with a rate similar to that of the Sb1 substrate (Fig.  3A). Again, the critical question to address was whether the high reactivity of the Sb4 substrate was caused by shortening of the N-terminal portion of this ␥-secretase substrate or by a sequence-specific effect as we observed in ␤CTF. Therefore, we designed two more constructs: Sb5 and Sb6. The A␤17-23 sequence of Sb3 was replaced by A␤10 -16 (YEVHHQK) in Sb5 or a random sequence (VAGAGGN) in Sb6. Both Sb5 and Sb6 are remarkably active (Fig. 3A). Their rates of ␥-secretase proteolysis were 0.85-and 1.1-fold of Sb1 for Sb5 and Sb6, respectively. This observation strongly indicates that it is the ASID in ␣CTF that directly regulates ␥-secretase activity, rather than an effect being mediated by shortening of the N-terminal portion of the substrate. Furthermore, it appears that removal of A␤1-16 enhanced the effect of ASID, indicating that the A␤1-16 sequence may play a specific role in modulating the inhibitory effect of ASID because ␤CTF is a substrate of ␥-secretase. To examine the relationship of A␤1-16 and ASID, we synthesized A␤1-16, A␤1-7, A␤1-12, A␤8 -16, and A␤1-21 peptides and tested their effect on ␥-secretase activity for processing of the Sb3 substrate. Combinations of three concentrations (0.1, 1, and 10 M) of A␤(1-16) and three concentrations (0.5, 1, and 2 M) of Sb3 were used to test two peptide motifs for ␥-secretase activity in trans (acting from a different molecule). There is no difference among all of these combinations, i.e. little 40-site-specific ␥-secretase product was detected. However, when we compared the inhibitory potency of A␤1-21 and A␤17-21, we obtained IC 50 values for A␤1-21 and A␤17-21 of 7.8 M and 0.3 M, respectively (Fig. 3B). These studies suggest that the A␤1-16 sequence only works in cis (acting from the same molecule) by coordinating with A␤17-21 for regulation of ␥-secretase activity. Furthermore, ␣-secretase cleavage significantly enhances the inhibitory effect of ASID, indicating that ␣-secretase-like cleavage can serve as a negative feedback for ␥-secretase modulation.
We next examined whether masking ASID reduces its negative regulation of ␣CTF processing and therefore increases substrate reactivity in vitro. To test this hypothesis, we preincubated the Sb3 substrate with the monoclonal antibody 4G8, which binds directly to the LVFFAE epitope and should therefore mask the inhibitory domain. The assay background was defined in the presence of L-685,458. After subtracting the L-685,458-defined background, there was a 26-fold increase in ␥-secretase activity with the 4G8 antibody-bound substrate (Fig. 3C). Therefore, our data show that either deleting or masking the inhibitory domain of ␣CTF alters its regulatory role and results in an increase in ␥-secretase-cleaved products. Dashed arrows represent changes of ␥-secretase-cleaved products. As shown on the right, ␣CTF inhibits ␥-secretase activity for the processing of ␤CTF. B, effect of recombinant Sb4 on in vitro ␥-secretase activity. Sb4 protein, which lacks the inhibitory domain, was added at concentrations of 0.25, 0.5, and 1 M to the in vitro ␥-secretase reaction mixture in the presence of 1 M ␤CTF as substrate. The 40-site product was detected by 6E10/G2-10 antibodies (mean Ϯ S.E., n Ͼ 3). As indicated by the scheme on the right, Sb4 does not suppress the processing of ␤CTF by ␥-secretase and is even a better substrate of ␥-secretase. C, effect of ␣CTF on ␥-secretase activity for Notch1 cleavage. The Notch1 cleavage was detected with a Notch1 cleavage-specific antibody SM320 (mean Ϯ S.E., n Ͼ 3).

Feedback Modulation of ␥-Secretase
OCTOBER 15, 2010 • VOLUME 285 • NUMBER 42 ␣CTF Inhibits ␥-Secretase Activity for the Processing of ␤CTF in Vitro-To examine further the notion that ASID within ␣CTF is responsible for its poor reactivity as a substrate and ␣CTF functions as an endogenous ␥-secretase inhibitor for A␤ production, we determined in vitro ␥-secretase activity for the production of X40 (P3 ϩ A␤40) and A␤40 in the presence of both the ␤CTF and ␣CTF substrates (Fig. 4A). At 0.25, 0.5, and 1.0 M, ␣CTF significantly suppresses ␥-secretase activity for A␤ production by 66, 71, and 82%, whereas X40 production is only inhibited by 24, 32, and 60%, respectively (Fig. 4A). To verify that the inhibitory domain within ␣CTF contributes to this inhibition, we determined the ␥-secretase activity in the presence of both ␤CTF and the Sb4 substrate in which the inhibitory domain was deleted. Both ␤CTF and Sb4 are processed (Fig. 4B), and the Sb4 substrate did not display inhibitory activity. Taken together, these results demonstrate that ␣CTF acts as an inhibitor to suppress ␥-secretase for ␤CTF processing and to reduce the production of X40. Clearly, deletion of the ASID within ␣CTF abolishes its inhibitory effect on ␥-secretase, and thereby the existence of the ASID at the N-terminal end of ␣CTF is prerequisite for this negative regulation. Finally, we assessed the effect of ␣CTF on Notch1 cleavage (Fig. 4C). ␣CTF at 0.25, 0.5, and 1.0 M inhibits ␥-secretase activity for Notch1 cleavage by 7, 20, and 36%, respectively, indicating that ␣CTF is less effective in inhibiting Notch1 cleavage than A␤40 production, which is consistent with our previous report using the ASID peptide (17).

DISCUSSION
␥-Secretase cleaves multiple substrates including APP, Notch, and other type I transmembrane proteins. ␥-Secretase cleavage of substrates first requires an additional protease to remove the majority of the extracellular domain of the substrate, referred to as ectodomain shedding. However, in the processing of APP, both ␤and ␣-secretase can shed the ectodomain by cleaving different regions of APP, which appears to be unique to APP processing. There has been a long and critical question: do ␤and ␣-secretase share the same mechanism for regulation of ␥-secretase? Our work suggests that ␣and ␤-secretases play a distinctive role in the modulation of ␥-secretase in the processing of APP for A␤ production. ␤-Secretase generates a ␥-secretase substrate that results in A␤ production, whereas ␣-secretase produces a modulator containing ASID that negatively regulates ␥-secretase activity through feedback inhibition. Unbalance of this system of ␤and ␣-secretase for APP processing could lead to disease states.
Present studies have discovered an unforeseen role for ␣-secretase in the processing of APP, which is different from the substrate competition model. This newly defined feedback mechanism for ␥-secretase regulation is proposed in Fig. 5. ␣-Secretase cleaves APP on the cell surface to generate the membrane-bound ␣CTF, which is then trafficked into intracellular compartments where it meets ␥-secretase to form a low productive complex. The formation of the ␣CTF⅐␥-secretase complex ultimately leads to inhibition of ␥-secretase for ␤CTF processing. ␤CTF is generated by ␤-secretase that predominantly resides in the late Golgi/trans-Golgi network. Active ␥-secretase complex has been found in multiple compartments including endoplasmic reticulum (ER) and across plasma membrane, and increased evidence indicates that A␤ is mainly produced in the trans-Golgi network and endosomes (25). Binding modes involving interactions between the ASID of ␣CTF and the E I site of ␥-secretase could well account for the behavior of ␣CTF as an inhibitor and a poor substrate of ␥-secretase. This notion is supported by the finding that coexpression of TACE and APP leads to a significant reduction in secreted X40 (A␤ ϩ P3) with a concomitant accumulation of ␤CTF in cells. In addition, the formation of the ␣CTF⅐␥secretase stable complex has been observed from the studies of ␥-secretase complex isolation (26), which supports our findings. Thus, the present cellular and biochemical studies offer a new model wherein ␣-secretase cleaves APP to generate an endogenous inhibitor of ␥-secretase that down-regulates the production of A␤, in addition to competing with ␤-secretase for the APP substrate (10,11). Previously, we have shown that the ASID within ␤CTF interacts (cis mode of inhibition) with the E I of ␥-secretase and reduces ␥-secretase activity for A␤ production (17). Both deletion and muta- The figure shows the mechanism of action of the inhibitory domain when it is engaged with ␤CTF or ␣CTF. The interaction of ␣CTF with ␥-secretase forms a low or unproductive complex that allows the inhibitory domain to modulate ␥-secretase activity both in cis and in trans and thus leads to a reduction of ␥-secretase activity for production of A␤ and P3. Concurrently, it results in an accumulation of ␤CTF. ␤CTF binds to ␥-secretase to form a productive complex for production of A␤. However, interaction of the ASID with the inhibitory domain binding site (E I ) of ␥-secretase limits the catalytic ability of active site (E AS ) for A␤ production in cis.
tion of the ASID within ␤CTF led to an increase in A␤ production (17). In fact, the Sb4 substrate has been used for the development of sensitive ␥-secretase assays (27,28) and characterization of ␥-secretase (29 -31). Another question that needs to be addressed is how P3 is produced at the cellular level. We propose that there is a putative factor that could interact with ASID to promote latency of its inhibitory activity. In fact, we found that the 4G8 antibody, which recognizes this sequence, is able to promote the cleavage of Sb3. Therefore, this unknown factor may possess a binding site that is similar to the antigen recognition site found in the 4G8 antibody. It could be critical to identify this factor and determine its role and specificity in the regulation of ␥-secretase. In addition, it is likely that this putative factor would become a novel target in the regulation of ␥-secretase activity for A␤ production.
Our studies have significant implications in AD. First, under in vivo conditions, there is no drastic overexpression of ␣-secretase activity. Therefore, this new mechanism involving ␣CTF inhibition of ␥-secretase may well be the primary mode by which ␣-secretase down-regulates the production of A␤. Postina et al. (13) found the modest expression of ADAM10 in mice led to a reduction of A␤ but had no effect on the level of sAPP␤. In other words, even if there is little competition for the APP substrate between ␣and ␤-secretase, the reduction in A␤ could result from inhibition of ␥-secretase by increased production of ␣CTF. This mechanism is also consistent with previous studies which showed that the level of sAPP␣ in cerebrospinal fluid was much lower in AD patients possessing the apoE4 allele compared with controls, but the amount of sAPP␤ remained the same (15,16). Moreover, decreased levels of sAPP␣ have been suggested as a diagnostic marker for AD (16,32). These patient studies imply that a decrease in ␣-secretase activity is associated with AD. Reduction of ␣-secretase activity, which alleviates the negative feedback ("brake") of ␥-secretase, could be one of the major pathways in the pathogenesis of sporadic AD. Indeed, ␣-secretase activity is significantly reduced in the majority of AD patients (33). Thus, our model for the inhibition of ␥-secretase has mechanistic implications that are critical to understanding the molecular basis for the diagnosis and pathogenesis of AD. Second, although we have demonstrated that the concentration of P3 generated in our in vitro system was not high enough to suppress ␥-secretase activity, we have not excluded the possibility that in vivo the A␤17-23 inhibitory domain of P3 might contribute to the inhibitory effects we are presently ascribing to ␣CTF. The possible existence of product inhibition of ␥-secretase has been previously proposed by Shen and Kelleher (24).
In summary, the present studies uncover a novel function of ␣-secretase that is opposite to ␤-secretase cleavage in the regulation of ␥-secretase activity. ␣-Secretase cleavage negatively regulates ␥-secretase activity for A␤ production. This work highlights fascinating twists in the link between ␥-secretase and ␣-secretase cleavages for APP processing and reveals an unprecedented regulatory mechanism of ␥-secretase activity that could be key to our understanding of the diagnosis of and progression of AD. Furthermore, a better understanding of the interactions that regulate A␤ production is a fundamental step toward elucidating ␥-secretase specificity and catalysis and designing specific inhibitors for AD therapies.