(cid:1) -Secretase Cleavage at Amino Acid Residue 34 in the Amyloid (cid:1) Peptide Is Dependent upon (cid:2) -Secretase Activity*

The amyloid (cid:1) peptides (A (cid:1) ) are the major components of the senile plaques characteristic of Alzheimer’s disease. A (cid:1) peptides are generated from the cleavage of amyloid precursor protein (APP) by (cid:1) - and (cid:2) -secretases. (cid:1) -Secretase (BACE), a type-I transmembrane aspartyl protease, cleaves APP first to generate a 99-amino acid membrane-associated fragment (CT99) containing the N terminus of A (cid:1) peptides. (cid:2) -Secretase, a multi-protein complex, then cleaves within the transmembrane region of CT99 to generate the C termini of A (cid:1) peptides. The production of A (cid:1) peptides is, therefore, dependent on the activities of both BACE and (cid:2) -secretase. The cleavage of APP by BACE is believed to be a prerequisite for (cid:2) -secretase-mediated processing. In the present study, we provide evidence both in vitro and in cells that BACE-mediated cleavage between amino acid residues 34 and 35 (A (cid:1) -34 site) in the A (cid:1) region is dependent on (cid:2) -secretase activity. In vitro, the A (cid:1) -34 times with 100 m M NH 4 HCO 3 , pH 7.8. After the final wash, the cap- tured products were eluted from the beads with a minimal volume of a saturated solution of (cid:3) -cyano-4-hydroxycinnamic acid dissolved in 50% acetonitrile/0.1% trifluoroacetic acid. One microliter of the bead eluate was then spotted on to the an NP-2 SELDI sample target (Ciphergen Biosystems) or a stainless steel target for an ABI Voyager and allowed to air dry. The dried targets were placed into the SELDI-TOF and Voyager mass spectrometers and matrix/analyte coprecipitates were desorbed with N 2 laser at 337 nm. The laser power was attenuated and the detector sensitivity manipulated such that spectra, which were the average of 100 laser shots, provided a qualitative assessment of the immunoprecipitation-captured cleavage products.

In cells, the activity of BACE and ␥-secretase to process APP has been observed in the membrane fractions, particularly in the trans-Golgi network (1, 8 -11). Production of A␤ has also been detected in both the endoplasmic reticulum and the trans-Golgi network (12)(13)(14)(15). Several lines of experimental evidence suggest that cellular APP processing is sequential and that cleavage of APP by ␣or ␤-secretase is a prerequisite for ␥-secretase-mediated processing. First, only two species of APP N-terminal products cleaved by endogenous secretases were reportedly detected in cells: the BACE-cleaved product sAPP␤ and the ␣-secretase-cleaved product sAPP␣ (1)(2)(3)(4)(5)(6)(7). No ␥-secretase cleaved N-terminal APP, or so-called sAPP␥, without prior ␣or ␤-secretase cleavage, was ever detected or reported. Second, in an in vitro cell-free assay system, the N-terminally truncated form of APP, CT99 (or CT100), has been demonstrated as an efficient substrate for ␥-secretase (16,17). In contrast, the full-length APP is a very poor substrate for ␥-secretase. 2 Therefore, removal of the N-terminal region of APP seems to be essential for ␥-secretase-mediated cleavage.
In the present study, we found that BACE-mediated cleavage at the A␤-34 site is dependent on the ␥-secretase activity. BACE cleaves the A␤-34 site in vitro only within A␤ 1-40 peptide, a product from ␥-secretase, not the site in the noncleaved CT99 fragment. Moreover, the cellular production of A␤ 1-34 species induced by overexpression of BACE1 or BACE2 was blocked by known ␥-secretase inhibitors in a concentration-dependent manner, whereas the ␥-secretase inhibitors had no effect on enzymatic activity of BACE1 or BACE2 in vitro. Our results suggest that the cleavage activity of BACE at the A␤-34 site, both in vitro and in cells, is dependent on ␥-secretase activity. Therefore, BACE and ␥-secretase activity can be mutually dependent.

BACE1 and BACE2 Expression and
Purification-Cloning of human BACE1 and BACE2 cDNA and purification of human BACE1 and BACE2 full-length protein from transiently transfected human embry-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
onic kidney 293T cells have been described previously (18). Expression and purification of truncated version of BACE1 from a recombinant baculovirus expression system was performed as described previously (19).
BACE Activity in Vitro Assay-A␤ 1-40 was purchased from Enzyme Systems Products (Livermore, CA). Human cathepsin D was purchased from Calbiochem. The production of recombinant APP-CT100-Flag was described previously (16). Treatment of CT100-Flag with aminopeptidase (Sigma) resulted in a CT99-Flag with the amino acid residue Asp at its N terminus. CT99-flag was confirmed by mass spectrometric analysis and by its reactivity with a neo-epitope-specific antibody, FCA-18, purchased from Dr. F. Checler (20). Assays were performed for the indicated time in the presence of 50 mM ammonium acetate, 0.15 M NaCl, pH 4.5 or 5.0, and 0.1 mg/ml BSA at 37°C. The reaction was terminated by heating samples at 75°C for 5 min. Samples were analyzed by mass spectrometry or/and by reversed-phase high pressure liquid chromatography. BACE1 or BACE2 activity in vitro was also assessed by cleavage of a FRET substrate encompassing P8-P4Ј of the APPsw ␤ site, TAMRA-5-CO-EEISEVNLDAEF-NH-QSY7, similar to the one described by Ermolieff et al. (21). The reaction condition is the same as described above and the cleavage product was measured by a LJL Analyst AD instrument (LJL BioSystems, Sunnyvale, CA) with excitation at 530 nm and emission at 580 nm.
Cell Culture, Transfection, and Analysis of APP Processing Products-HEK293T cells stably expressing human APP695 were seeded in 100-mm dishes and transfected with either BACE1 or BACE2 DNA as described above. Transfection was performed using LipofectAMINE reagent (Invitrogen) according to the manufacturer's instructions. Media were changed on the next day, and either Me 2 SO control or compound-1 was added to the new media. After an additional 24 h of incubation at 37°C, the media and cells were harvested. A␤ level in media was assayed by Origen system, as described previously, using biotinylated 6E10 antibody (Senetek, Maryland Heights, MO) and ruthenylated G2-10 antibody (16) licensed from the University of Heidelberg.
Mass Spectrometric Analysis of BACE-cleaved Products-MALDI mass analysis was performed with a surface-enhanced laser desorption/ ionization-time-of-flight (SELDI-TOF) mass spectrometer from Ciphergen Biosystems (Fremont, CA) and a Voyager DE/RP MALDI-TOF from Applied Biosystems. Typically, 1 l of reaction mixture was mixed with 1 l of a saturated MALDI matrix solution (␣-cyanohydroxycinnamic acid in 1:1 water/acetonitrile with 0.1% trifluoroacetic acid) and was then spotted onto Ciphergen NP2 MALDI chips. After drying and crystal formation, sample spots were typically analyzed with 100 laser shots.
In all figures containing MALDI mass spectrometric data, the generic notation A␤ X-X refers to the length of the A␤ fragment, such as A␤ 1-40 or A␤ 1-34. This notation is used to denote the A␤ fragment corresponding to the mass of the singly charged peptide ion (MϩH) ϩ . In addition, in cases in which the doubly charged ion (Mϩ2H) 2ϩ is visible, as in Fig. 2, the generic notation (A␤ X-X ϩ 2H) 2ϩ , denotes the A␤ species corresponding to the mass of the doubly charged peptide ion.
Mass Spectrometric Analysis of APP Processing Products in Media-For analysis of APP processing species secreted into media, the col-lected media samples were immunoprecipitated with 6E10 antibody and protein G beads (Protein G UltraLink; Pierce Chemical). After incubation at room temperature for 1 h, the supernatants were removed, and the immunoprecipitation beads were first washed three times with a buffer containing 150 mM NaCl, 10 mM Tris, 2 mM EDTA, 0.1% Triton X-100, and 0.1% Igepal CA630, pH 7.8, then washed twice with a buffer consisting of 500 mM NaCl, 10 mM Tris, 2 mM EDTA, 0.1% Triton X-100, and 0.1% Igepal CA630, pH 7.8, and finally washed three times with 100 mM NH 4 HCO 3 , pH 7.8. After the final wash, the captured products were eluted from the beads with a minimal volume of a saturated solution of ␣-cyano-4-hydroxycinnamic acid dissolved in 50% acetonitrile/0.1% trifluoroacetic acid. One microliter of the bead eluate was then spotted on to the an NP-2 SELDI sample target (Ciphergen Biosystems) or a stainless steel target for an ABI Voyager and allowed to air dry. The dried targets were placed into the SELDI-TOF and Voyager mass spectrometers and matrix/analyte coprecipitates were desorbed with N 2 laser at 337 nm. The laser power was attenuated and the detector sensitivity manipulated such that spectra, which were the average of 100 laser shots, provided a qualitative assessment of the immunoprecipitation-captured cleavage products.

BACE1 or BACE2, but Not Cathepsin D, Cleaves the A␤-34
Site in Vitro-The activity of purified human full-length BACE1, BACE2, or cathepsin D, a closely related aspartyl protease, was tested for cleavage of a synthetic A␤ peptide 1-40 in vitro.
When BACE1 or BACE2 was incubated with a synthetic A␤ peptide 1Ϫ40 in vitro, a cleavage site located between A␤ residues 34 and 35 was clearly identified through mass spectrometric analysis as shown in Fig. 1, B1, B3, and B4. The theoretical mass value for 〈␤ 1-34 is 3786.9 and the observed mass values for the cleaved 1-34 products were 3785.6 and 3786.6 for BACE1 and BACE2, respectively. However, when A␤ 1-40 peptide was incubated with cathepsin D, A␤ species cleaved at residues between 19 and 20 were observed by mass spectrometric analysis, but not the A␤ 1-34 product (Fig. 1B, 1  and 2). The absence of A␤ 1-34 species in cathepsin D-treated sample is not the result of a secondary cleavage of A␤ 1-34, because a shorter incubation time and with less enzyme showed similar results (data not shown). In contrast, both purified full-length human BACE1 (Fig. 1B, 3) and BACE2 (Fig. 1B, 4) processed the A␤-34 site. Thus, the cleavage of A␤-34 site is specific for BACE1 and BACE2 not for cathepsin D.
BACE1 Lacking the Transmembrane and Cytoplasmic Regions Cleaves the A␤-34 Site in Vitro-One of the major differences between BACE and cathepsin D is that cathepsin D does not contain a transmembrane domain and cytoplasmic tail. To investigate whether the transmembrane and cytoplasmic regions of BACE contribute to its specific activity of processing the A␤-34 site, we tested the activities of truncated forms of BACE1 lacking both the transmembrane and cytoplasmic domains using the A␤ 1-40 substrate. Results from mass spectrometric analysis showed that truncated BACE1, both the pro-(amino acids 22-460) and the mature (amino acids 46 -460) forms (19), cleaves the A␤-34 site in vitro (Fig. 2). Therefore, the differential cleavage of A␤-34 site between BACE and cathepsin D is probably because of the difference in their enzyme active sites. It is interesting to note that the truncated forms of BACE1 did not cleave the A␤-20 site as seen with the fulllength enzyme.
BACE1 or BACE2 Does Not Cleave A␤-34 Site Using CT99 as Substrate-To further investigate BACE in vitro cleavage specificity, we tested BACE activity using a recombinant CT99 protein substrate. CT99 or CT100 has been described previously as an efficient in vitro substrate for ␥-secretase (16,17). Control mass spectrometric analysis of CT99 alone (Fig. 3A) at mass range 1000 -5000 Da did not show any notable species (Fig. 3B). To our surprise, when CT99 was incubated with BACE1, cleavage products of CT99 at either the A␤-20 site or the A␤-34 site were not observed (Fig. 3C). This was true for both full-length and truncated forms of BACE1 (data not shown). The absence of BACE1-mediated cleavage at A␤-20 or A␤-34 sites in CT99 was not caused by substrate depletion, because the substrate peak was still observed by mass spectrometry at the end of the reaction (data not shown), nor was it caused by enzyme inactivation, because that addition of a small peptide substrate encompassing the ␤-scissile bond at the end of the CT99 reaction yielded an appropriate cleavage product (data not shown). Likewise, BACE2 also did not cleave CT99 at the A␤-34 site, although it cleaved at the A␤-19 and -20 sites (Fig. 3D). Because the CT99 used here is in solution and not embedded in the membrane, the hydrophobic domain of the CT99 might be "abnormally" folded in the buffer. To test whether addition of a limited amount of detergent could expose the A␤-34 site in CT99 for processing, the in vitro reaction with CT99 was also performed in the presence of 0.05% Triton X-100. Both BACE1 and BACE2 were fully functional in the presence of 0.05% Triton X-100 to cleave the ␤ site of APP (data not shown) (3). However, neither of them showed any cleavage of A␤-34 site in CT99 (data not shown). Therefore, unlike A␤ 1-40 peptide, CT99 is not a substrate for A␤-34 site cleavage by either BACE1 or BACE2 in vitro.
␥-Secretase Inhibitors Block the Production of A␤ 1-34 in Cells-To evaluate the effect of ␥-secretase inhibitors on the cellular production of A␤ 1-34 species, HEK293T/APP695 cells, either mock-transfected or transfected with BACE1 or BACE2, were treated with either Me 2 SO or with 10 M concentrations of each of the two known ␥-secretase inhibitors, compound 1 and compound 2 (24,25). As expected, both compounds inhibited A␤ 1-40 production in HEK293TAPP695 cells (Fig. 5A,  1-3) and the cells transfected with BACE1 (Fig. 5A, 4 -6). The effect of compounds on the cellular production of A␤ 1-40 was further confirmed by mass spectrometric analysis of the A␤ species in the conditioned media of these cells (Fig. 5B). Notably, in BACE1 overexpressing cells, these ␥-secretase inhibitors blocked the production of not only A␤ 1-40 species but also A␤ 1-34 species (Fig. 5B, 4 -6). Likewise, in BACE2 overexpressing cells, the increased generation of A␤ 1-34 species was also inhibited by these compounds (Fig. 5B, 7-9). In contrast, the generation of other A␤ species such as 1-16, 1-19, or 1-20 were largely unblocked (Fig. 5B).
Inhibition of A␤ 1-34 Production by ␥-Secretase Inhibitor Is Concentration-dependent and Coincides with the Blockade of A␤ 1-40 Production-To further evaluate the effect of ␥-secretase inhibitors on the cellular production of A␤ 1-34 species, another known ␥-secretase inhibitor with a different structural class, compound 3 (17,26), was tested in cells at various concentrations (1 and 10 M). Results from the modified A␤ ELISA assay revealed a concentration-dependent inhibition of A␤ 1-40 production in both HEK293T/APP695 cells (Fig. 6A, 1-3) and cells transfected with BACE1 (Fig. 6A, 4 -6). Furthermore, the mass spectrometric analysis of A␤ species in conditioned media of these cells showed that both the production of A␤ 1-40 and A␤ 1-34 species were blocked by compound-3 in a dose-dependent manner (Fig. 6B). Notably, in BACE1 overexpressing cells, the extent of A␤ 1-34 inhibition coincided with the blockade of A␤ 1-40 production (Fig. 6B, 4 -6). In BACE2 overexpressing cells, the generation of A␤ 1-34 was similarly inhibited by compound 3 in a dose-dependent manner (Fig. 6B, 7-9).
Because both BACE and ␥-secretase are aspartyl proteases (2, 3, 16, 17), we performed experiments to determine whether any of the ␥-secretase inhibitors used in our cellular study blocked the activities of BACE1 or BACE2 in vitro. When BACE1 or BACE2 were incubated in vitro with a peptide substrate containing the APPsw ␤ site sequence, a steady formation of the cleavage products was observed with time (Fig. 7). The known BACE inhibitor StatV (3) inhibited the cleavage activities of both BACE1 and BACE2. However, no inhibitory effect was observed with any of the ␥-secretase inhibitors (Fig.  7), whereas at the same concentration (10 M), all three compounds completely blocked ␥-secretase activity in cells (Figs. 5 and 6).

DISCUSSION
Aberrant A␤ production is believed to be one of the major causes of AD (1). BACE and ␥-secretase are two critical enzymes responsible for the generation of A␤ peptides and have therefore been considered potential therapeutic targets for AD (1)(2)(3)(4)(5)(6)(7). Studies of the relationship between BACE, ␥-secretase, and APP substrate could shed light on APP metabolism and facilitate our understanding of A␤ generation process. Several lines of experimental evidence suggest that BACE cleavage in APP is a prerequisite for ␥-secretase-mediated processing. In the current investigation, we present evidence that, both in vitro and in cells, BACE and ␥-secretase activities can be mutually dependent. An example described here is the generation of A␤ 1-34 species.
First, we show that the A␤-34 site is cleaved by BACE both in vitro and in cells; second, we demonstrate that ␥-secretase activity is required for BACE cleavage at the A␤-34 site both in vitro and in cells.
In vitro, human BACE, both the full-length form and the A, A␤ level in the media was determined by a modified ELISA using 6E10 and G2-10 antibodies. B, mass spectrometric analysis of A␤ species in the same media captured by 6E10 antibody. A and B, 1-3 truncated form lacking transmembrane and cytoplasmic domains, cleaved the A␤-34 site within A␤ 1-40 peptide substrate. However, cathepsin D, a closely related aspartyl protease, did not cleave the A␤-34 site, although it shares many of the cleavage sites in APP with BACE, including the ␤-scissile bond 3 and the A␤-19 or -20 sites (Fig. 1). Thus, the A␤-34 site seems to be rather specific for BACE. Notably, the sequence around the A␤ 34 site, IIGL-MVGG, shows a high homology to a reported in vitro "optimized sequence" for BACE1, EIDL-MVLD (27).
In cells, overexpression of BACE1 or BACE2 increased the generation of A␤ 1-34 species (Fig. 3), suggesting that cellular production of A␤ 1-34 is dependent on BACE levels. The A␤ 1-34 species has also been observed in cells stably overexpressing BACE1 (10,28). However, in all these cases, it is difficult to completely rule out the possibility that overexpressed BACE1 increases the cleavage of the A␤-1 site, and other enzymes, such as ␥-secretase, could be responsible for the cleavage at the A␤-34 site in cells. In this aspect, our data from BACE2 overexpression studies is particularly noteworthy. The overexpression of BACE2 similarly increased A␤ 1-34 level, but abolished A␤ 1-40 production (Fig. 4B, 3). This indicates that two different enzymes must be responsible for the cleavage of A␤-34 and A␤-40 sites. In fact, we and others have observed that BACE1 prefers the A␤-1 site, whereas BACE2 prefers the internal cleavage sites within A␤ 1-40, such as the 〈␤-19 or -20 sites and the A␤-34 site (18). These "internal" cleavage activities by BACE2 would lead to a decrease in A␤ 1-40 level and, we expect, in A␤ 1-34 level if the A␤-34 site were indeed processed by ␥-secretase. Thus, taken together, our data indicate that in cells, the A␤ 34 site is cleaved by BACE1 or BACE2 not the ␥-secretase.
Several lines of evidence indicate that BACE cleavage of the A␤-34 site is dependent on ␥-secretase activity. In vitro, BACE cleaves only the A␤-34 site within the A␤ 1-40 peptide, the product of ␥-secretase cleavage; BACE did not process the A␤-34 site in CT99, the substrate for the ␥-secretase. In con-trast, BACE2 cleavage of A␤-19 and -20 sites occurred with both A␤ 1-40 and CT99 substrates. We were unable to observe any BACE1 cleavage product using CT99 as substrate, although the enzyme was active toward the ␤-scissile bond cleavage. One possible explanation is that other BACE1-mediated cleavages within CT99 could also be dependent on ␥-secretase activity. More experiments are needed to test this hypothesis. In cells, cleavage of the A␤-34 site induced by BACE1 or BACE2 is blocked by several different classes of ␥-secretase inhibitors. Moreover, the inhibition of A␤-34 site processing is dose-dependent and coincides with the blockade of A␤-40 site cleavage. None of the ␥-secretase inhibitors tested here showed any inhibitory effect on enzymatic activity of BACE1 or BACE2 in vitro. Therefore, the most likely explanation would be that ␥-secretase cleavage of CT99 in cells is a prerequisite for BACE-mediated processing of A␤-34 site. Additionally, two recent studies also reported that ␥-secretase inhibitors blocked the production of the A␤ 1-34 species in cells (28,29). During the manuscript revision for this article, Fluhrer et al. published a study with similar findings using presenilin dominant-negative mutants to inhibit cellular ␥-secretase activity (30).
A question that remains is how the activity of ␥-secretase affects the BACE cleavage of the A␤-34 site. Results from the in vitro experiments using either A␤1-40 or CT99 substrates suggest that certain features in CT99 fragment interfere with BACE cleavage of A␤-34 site. One possibility is that the hydrophobic domain of the CT99, normally embedded in the membrane, might be misfolded and therefore prevented the access of BACE. However, addition of a limited amount of detergent, 0.05% Triton X-100, did not promote the processing. Alternatively, CT99 fragment could still possess an "unfit" conformation or contain an inhibitory element downstream from the ␥-secretase processing sites. In any case, CT99 needs to be truncated at the ␥-secretase cleavage sites for BACE to cleave the A␤-34 site. In vitro, such a requirement is fulfilled by directly using the ␥-secretase-cleaved product and a synthetic A␤ 1-40 peptide as substrate; in cells, this is accomplished through the cellular ␥-secretase activities. A ␥-secretase inhibitor(s) blocked the "truncation" of CT99 and therefore inhibited BACE cleavage at the A␤-34 site in cells.
In summary, we show that BACE, either BACE1 or BACE2, cleaves A␤-34 site in vitro and in cells, but only after ␥-secretase-mediated processing of APP. It is interesting to note that N-terminal truncation of APP by BACE or ␣-secretase is a prerequisite for ␥-secretase cleavage around A␤-40 sites. And it is, in turn, as shown in the present study, that ␥-secretase cleavage at the C terminus of A␤ peptide is a prerequisite for BACE-mediated processing at the A␤-34 site. Our data indicate a sequential and mutual dependence of BACE and ␥-secretase activities in APP processing. We and others have observed and reported that BACE cleaves at least four sites within a stretch of 50 amino acid residues of APP, around and within the A␤ region, approximately every 10 amino acid residues (18). Results described in this study further support such a complex picture in APP metabolism: product from one enzyme can become substrate for the other and verse visa.
The biological role of the A␤ 1-34 species is still unclear. Although A␤ 1-34 species, either in soluble or insoluble form, has not been reported in AD brain, its presence in human cerebrospinal fluid has been observed (31). The contribution of A␤ 1-34 species to AD pathogenesis remains to be determined. The discovery that BACE cleavage at the A␤-34 site is dependent on ␥-secretase activity could facilitate our understanding of APP processing by these two critical enzymes.