Characterization of the ectodomain shedding of the beta-site amyloid precursor protein-cleaving enzyme 1 (BACE1).

Generation of the amyloid peptide through proteolytic processing of the amyloid precursor protein by beta- and gamma-secretases is central to the etiology of Alzheimer's disease. beta-secretase, known more widely as the beta-site amyloid precursor protein cleaving enzyme 1 (BACE1), has been identified as a transmembrane aspartic proteinase, and its ectodomain has been reported to be cleaved and secreted from cells in a soluble form. The extracellular domains of many diverse proteins are known to be cleaved and secreted from cells by a process known as ectodomain shedding. Here we confirm that the ectodomain of BACE1 is secreted from cells and that this processing is up-regulated by agents that activate protein kinase C. A metalloproteinase is involved in the cleavage of BACE1 as hydroxamic acid-based metalloproteinase inhibitors abolish the release of shed BACE1. Using potent and selective inhibitors, we demonstrate that ADAM10 is a strong candidate for the BACE1 sheddase. In addition, we show that the BACE1 sheddase is distinct from alpha-secretase and, importantly, that inhibition of BACE1 shedding does not influence amyloid precursor protein processing at the beta-site.

enzyme (TACE) (10,11), which cleaves the ectodomain of pro-TNF␣. TACE, also known as ADAM17, is a member of the a disintegrin and metalloproteinase (ADAM) family of metalloproteinases (12,13) and has been reported to mediate ectodomain shedding of other proteins such as L-selectin (14), interleukin-1 receptor (15), and APP by cleavage at the ␣-site (16,17). In addition to TACE, two other members of the ADAM family of metalloproteinases, ADAM9 and ADAM10, have also been proposed as ␣-secretases (18,19). ADAM9 and ADAM10, in common with other sheddase activities, exhibit regulation by protein kinase C and inhibition by hydroxamic acid-based metalloproteinase inhibitors.
Proteolytic processing of the APP appears to be central to the etiology of Alzheimer's disease (AD). Cleavage of APP by ␤and ␥-secretases generates the amyloidogenic peptide A␤, which aggregates to form senile plaque (20). In an alternative nonamyloidogenic pathway, cleavage of APP by ␣-secretase within the ␤-amyloid region of APP precludes the release of intact A␤. ␤-Secretase has been identified as a novel transmembrane aspartic proteinase, BACE1, and this proteinase fulfills all the key criteria for ␤-secretase (21)(22)(23)(24)(25)(26). BACE1 is highly expressed in neurons in the brain, and overexpression of BACE1 in cultured cells and transgenic mice using a neuronal specific promoter results in an increased production of ␤-secretase-derived APP cleavage products (21,22,27). In addition, BACE1 knockout mice appear phenotypically normal (28,29), making BACE1 a very attractive therapeutic target for the treatment of AD.
Recently it was reported that the extracellular domain of BACE1 is cleaved and released from cells in a soluble form (30). The extent to which BACE1 shedding is regulated or the possible impact this may have on the amyloidogenic processing of APP is not known. However, it has been suggested that inhibition of BACE1 shedding could have value as an alternative strategy in the treatment of AD (30). Here we show that BACE1 shedding is indeed increased by agents that stimulate protein kinase C and that a metalloproteinase, possibly ADAM10, is involved in this process. As BACE1 is known to be re-internalized and targeted to the endosomes (31), which are a major site of APP processing, it is possible that an alteration in the level of BACE1 at the cell surface would alter the levels of BACE1 available for re-internalization. This would clearly have implications for the processing of APP at the ␤-site and, hence, A␤ production. To address this issue, we identified BACE1 sheddase-selective hydroxamate inhibitors and investigated the effect of inhibiting BACE1 shedding on the processing of APP. Interestingly, abolishing the ectodomain shedding of BACE1 appeared to have no effect on the processing of APP at the ␤-site.

EXPERIMENTAL PROCEDURES
Expression of BACE1 Constructs-BACE1-MycHis 6 /pcDNA3.1 has been described previously (21). Mutant constructs were generated by QuikChange site-directed mutagenesis (Stratagene) as described by the manufacturer and sequenced. Human embryonic kidney (HEK) 293 cells expressing wild type or mutant BACE1 or APP 695 were generated using LipofectAMINE Plus Reagent (Invitrogen). Stable cell lines were generated by selection with Geneticin (G418; 600 g/ml).
Drug/Inhibitor Treatments-Cells were grown in a serum-free medium in the absence or presence of 2 M 12-phorbol-3-myristate (PMA; Calbiochem) for 1 h or the indicated concentration of hydroxamic acidbased metalloproteinase inhibitors for 20 h. The medium was harvested and centrifuged at 100,000 ϫ g for 10 min prior to analysis. Media from the 1-h treatments were concentrated 30-fold using Vivaspin 500 concentrators (Vivascience Ltd). Cells were lysed in 10 mM Tris/HCl, pH 7.4, and 1% Triton X-100 plus a mixture of proteinase inhibitors (Roche Applied Science) at 4°C for 30 min.
SDS-PAGE and Western Blotting-Cell lysates (10 g of protein) and equivalent volumes of medium samples were resolved on Tris-glycine SDS-polyacrylamide gels (Invitrogen) for Western blot analysis with monoclonal anti-Myc antibody (9E10, Sigma-Aldrich), anti-BACE1 monoclonal antibody 9B21, anti-APP antibody Ab54, sAPP␣ antibody 6E10, or sAPP␤ antibody G26. Monoclonal antibody 9B21 was raised to the catalytic domain of BACE1 using the BACE1-Fc fusion protein as an immunogen. Endoglycosidase H and N-glycosidase F treatment of media was carried out at 37°C for 16 h as described by the manufacturer (Roche Applied Science).

RESULTS
Ectodomain Shedding of BACE1 from HEK 293 Cells-To determine whether the extracellular domain of BACE1 is cleaved and released from cells, we generated HEK 293 cells stably expressing BACE1 with a C-terminal MycHis 6 tag (BACE1-MycHis 6 ). Western blot analysis of whole cell lysates with an anti-Myc antibody confirmed BACE1 expression in these cells (Fig. 1a). Subsequent analysis of the medium with antibody 9B21, which recognizes the catalytic domain of BACE1, allowed detection of soluble shed BACE1. Shed BACE1 in the medium was not immunoreactive with anti-Myc antibody (data not shown) and migrated as a band of lower M r than full-length BACE1 in cells, confirming that the extracellular domain of BACE1 is cleaved and released from cells. The appearance of shed BACE1 in the medium should be concomitant with the generation of an intracellular C-terminal fragment (CTF). Western blot analysis of whole cell lysates with anti-Myc antibody failed to detect a BACE1 CTF even upon prolonged exposure, suggesting that it may be rapidly turned over in the cell (data not shown). To eliminate any effect of the MycHis 6 tag at the C terminus of BACE1 on its ability to be cleaved and released from cells, medium from HEK 293 cells expressing untagged BACE1 (Fig. 1b) was also subjected to Western blot analysis with antibody 9B21. Shed BACE1 was detected in the medium from these cells, indicating that the C-terminal tags do not influence BACE1 cleavage and release from cells. In addition, both the tagged and non-tagged forms of the protein expressed in cells gave the same overall pattern of mature and immature forms ( Fig. 1), indicating that the tag had no effect on the trafficking and processing of the proteins in the endomembrane system.
To determine which form of BACE1 is shed from cells, deglycosylation experiments were carried out. Shed BACE1 in the medium is resistant to endoglycosidase H treatment ( Fig.  1c), whereas treatment with N-glycosidase F causes a shift in M r , indicating that it is the mature form of BACE1 in cells that is subject to cleavage. As BACE1 is a proteinase, we went on to investigate whether BACE1 could mediate its own ectodomain shedding. HEK 293 cells were transfected with either wild type BACE1-MycHis 6 or mutants in which the aspartic acid residues in the active site had been mutated to asparagine (D93N and D289N), which, as we have shown previously, abolish the catalytic activity of the proteinase (21). Both wild type and mutant BACE1 were expressed in cells to similar levels (Fig.  1d, top panel). Analysis of the medium from these cells revealed a similar level of soluble shed BACE1 from both wild type and mutant protein (Fig. 1d, bottom panel), indicating that BACE1 does not mediate it own cleavage and release from cells.
Characteristics of BACE1 Shedding-The ectodomain shedding of membrane proteins can be up-regulated by agents that activate protein kinase C (32). To investigate whether ectodomain shedding of BACE1 is controlled by a similar mechanism, HEK 293 cells expressing BACE1-MycHis 6 were grown in the absence or presence of the phorbol ester PMA, which is known to activate protein kinase C. Treatment of cells with PMA led to an increase of ϳ2.5-fold in the level of shed BACE1 in the medium (Fig. 2a, bottom panel) indicating that activation of protein kinase C increases BACE1 shedding. Hydroxamic acid-based metalloproteinase inhibitors have been reported to inhibit the shedding of other cell surface proteins (1,5,6). We went on to determine whether the ectodomain shedding of BACE1 is sensitive to inhibitors of this class by exposing HEK 293 cells expressing BACE1-MycHis 6 to two hydroxamate inhibitors, SB-202968 or SB-242113 (Fig. 2b). The presence of either SB-202968 or SB-242113 reduced the release of soluble BACE1 in the medium to Ͻ20% of the level seen in untreated cells, clearly indicating the involvement of a metalloproteinase in the ectodomain shedding of BACE1. Taken together, this set of results indicates that BACE1 shedding shows all the characteristics of ectodomain shedding reported previously for other proteins, namely up-regulation by phorbol esters and cleavage by a metalloproteinase.
Inhibition of BACE1 Shedding Does Not Regulate APP Processing at the ␤-Site, and BACE1 Sheddase Is Distinct from ␣-Secretase-The phorbol ester and hydroxamate inhibitors described so far are non-selective and are known to affect the ectodomain shedding of other cell surface proteins, including ␣-secretase cleavage of APP (5). To determine whether inhibition of BACE1 shedding could influence APP processing at the ␤-site (for example, by reducing the amount of BACE1 reinternalized to the endosomes, which is a major site of APP processing), we needed to identify fully selective BACE1 sheddase inhibitors, which had no effect on ␣-secretase processing of APP. Therefore, a range of hydroxamate-based inhibitors were screened for their effects on BACE1 shedding and APP processing by ␣-secretase.
A number of hydroxamate inhibitors were found to inhibit ectodomain shedding of BACE1 from HEK 293 cells (Fig. 3a).
To determine whether these compounds affected APP processing, HEK 293 cells expressing BACE1-MycHis 6 cells were transiently transfected with APP 695 followed by treatment with the inhibitors. None of the hydroxamate inhibitors had any effect on the processing of APP by ␣-secretase, as similar levels of sAPP␣ and intracellular CTF␣ were detected in the control and treated cells (Fig. 3b). This different inhibition profile for the BACE1 sheddase and ␣-secretase suggests that the BACE1 sheddase is distinct from ␣-secretase. This was further confirmed by the identification of a hydroxamate inhibitor, SB-451916. which had no effect on BACE1 shedding (Fig. 3c) but significantly reduced the level of sAPP␣ (Fig. 3d).
The selective BACE1 sheddase inhibitors were subsequently used to determine whether inhibition of BACE1 shedding affects ␤-secretase cleavage of APP in these cells. None of the hydroxamate inhibitors had any effect on secreted sAPP␤ levels or intracellular CTF␤ levels, indicating that the inhibition of BACE1 shedding has no effect on APP processing at the ␤-site (Fig. 3b). In addition, the inhibitors had no effect on the level of A␤40 and 〈␤42 peptides secreted into the medium from these cells (data not shown).
ADAM10-selective Hydroxamate Inhibitor Blocks Ectodomain Shedding of BACE1-Members of the ADAM family of metalloproteases are known to mediate the ectodomain shedding of a variety of cell surface proteins (34). To establish the identity of the BACE1 sheddase, the potency and selectivity of two hydroxamate inhibitors to ADAM10 and TACE in vitro were exploited (33). GW4023 is a highly potent and selective ADAM10 inhibitor (IC 50 , 5 nM) being 100-fold less potent against TACE (IC 50 , 541 nM). In contrast, GW0264 shows similar potency for ADAM10 (IC 50 , 11 nM) and TACE (IC 50 , 8 nM) (Fig. 4a) Treatment of HEK-BACE1 cells with the indicated concentrations of either inhibitor caused a dose-dependent inhibition of the release of shed BACE1 (Fig. 4b). Densitometric analysis of these blots allowed concentration response curves and EC 50 values for inhibition of BACE1 shedding to be generated (Fig. 4c). Both compounds inhibited the generation of shed BACE1 with potencies similar to those described for these inhibitors against ADAM10 in vitro. As GW4023 abolished BACE1 shedding at concentrations far below the reported IC 50 for this compound against TACE, it is unlikely that TACE mediates the ectodomain shedding of BACE1. Rather, our data show ADAM10 to be a potential candidate for the BACE1 sheddase.

DISCUSSION
The extracellular domains of many diverse cell surface proteins are cleaved and released from cells by a process known as ectodomain shedding. Here we have shown that the ectodomain of BACE1 undergoes a similar cleavage, resulting in the re-lease of soluble shed protein from cells. As reported for other shed proteins, ectodomain shedding of BACE1 is stimulated by agents that activate protein kinase C and is inhibited by hydroxamic acid-based metalloproteinase inhibitors, suggesting a common processing mechanism. Hence, BACE1 belongs to that group of transmembrane proteins that is subjected to regulated ectodomain shedding and includes APP, TNF␣, and L-selectin.
Using potent and selective ADAM10 inhibitors, we demonstrate that ADAM10, but not TACE, is a strong candidate for the BACE1 sheddase in HEK cells. However, we can not rule out the possibility that other metalloproteinases that may be sensitive to the hydroxamate inhibitors described in this study could also contribute to the BACE1 sheddase activity in these cells. In addition, by testing a range of hydroxamate inhibitors we show that the BACE1 sheddase is distinct from ␣-secretase. Several members of the ADAM family of metalloproteinases (ADAM10, ADAM9, and TACE) have been proposed as candidates for ␣-secretase (16 -19). Overexpression of ADAM10 in cells increases basal and PKC-inducible ␣-secretase cleavage of APP, whereas a catalytically inactive form of ADAM10 acts as a dominant negative, precluding the generation of sAPP␣ (19). However, ␣-secretase activity is preserved in fibroblasts from ADAM10-deficient mice (35). Overexpression of TACE and ADAM9 also results in increased PKC-inducible ␣-secretase cleavage of APP (17,18). In addition, fibroblasts derived from TACE-deficient mice exhibit reduced PKC inducible ␣-secretase activity (16). All of these findings suggest that multiple ADAMs contribute to ␣-secretase activity in different tissues. Our findings suggest that ADAM10 is not the major ␣-secretase in HEK cells but may constitute the BACE1 sheddase activity in these cells. Future studies will allow the role of ADAM10 in the ectodomain shedding of BACE1 to be further elucidated.
A key question to address was whether regulation of BACE1 shedding could influence APP processing at the ␤-site, as this may have therapeutic implications. APP is known to transit along the secretory pathway to the cell surface where it can be processed by ␣-secretase (36). APP at the cell surface is reinternalized and targeted to the endosomes where ␤-secretase cleavage of APP occurs (37). The trafficking of BACE1 in cells follows a similar route, with the protein residing at the cell surface and undergoing re-internalization to endocytic compartments (31). Inhibition of BACE1 shedding could lead to the accumulation of increased levels of full-length BACE1 at the cell surface that could then be re-internalized, perhaps in conjunction with APP, and targeted to endosomes. In contrast, stimulation of BACE1 shedding would have the opposite effect and reduce the level of full-length BACE1 that is in cells and available for re-internalization and, therefore, decrease APP processing at the ␤-site. The possibility that the ectodomain shedding of BACE1 may enhance the amyloidogenic potential of this proteinase was proposed in a previous study that employed a secreted form of the protein (30). Expression of a soluble form of BACE1 in cells that resembled the shed protein led to a marked increase in A␤40 levels but, interestingly, a reduction in CTF␤ levels. However, it is possible that the increase in the A␤40 level reported in this study was merely a reflection of the faster rate of transit in cells of soluble BACE1 as compared with that of the full-length protein rather than an indication that shed BACE1 is more amyloidogenic. To address this point in a more physiological manner, we exploited the availability of selective and potent inhibitors of the BACE1 sheddase and clearly established that none of these compounds have any significant effect on ␤-secretase cleavage of APP. However, subtle changes in ␤-secretase activity due to inhibition of BACE1 shedding cannot be ruled out.
In conclusion, we have shown that BACE1, a key therapeutic target for treatment of AD, is subject to ectodomain shedding. Although the physiological significance of shed BACE1 is unclear at present, it appears to have no major functional implications for APP processing, and, therefore, inhibition of BACE1 shedding is unlikely to have therapeutic potential in Alzheimer's Disease.