Processing of beta-secretase by furin and other members of the proprotein convertase family.

The amyloid peptide is the main constituent of the amyloid plaques in brain of Alzheimer's disease patients. This peptide is generated from the amyloid precursor protein by two consecutive cleavages. Cleavage at the N terminus is performed by the recently discovered beta-secretase (Bace). This aspartyl protease contains a propeptide that has to be removed to obtain mature Bace. Furin and other members of the furin family of prohormone convertases are involved in this process. Surprisingly, beta-secretase activity, neither at the classical Asp(1) position nor at the Glu(11) position of amyloid precursor protein, seems to be controlled by this maturation step. Furthermore, we show that Glu(11) cleavage is a function of the expression level of Bace, that it depends on the membrane anchorage of Bace, and that Asp(1) cleavage can be followed by Glu(11) cleavage. Our data suggest that pro-Bace could be active as a beta-secretase in the early biosynthetic compartments of the cell and could be involved in the generation of the intracellular pool of the amyloid peptide. We conclude that modulation of the conversion of pro-Bace to mature Bace is not a relevant drug target to treat Alzheimer's disease.

The brain of patients suffering from Alzheimer's disease (AD) 1 is characterized by the presence of amyloid plaques composed mainly of the 39 -42 amino acid amyloid ␤ (A␤) peptide (1,2). A␤ derives from a type I single membrane-spanning protein termed amyloid precursor protein (APP) by post-translational proteolytic cleavage (3). Two cleavages by ␤-and ␥-secretases, respectively, are required to release A␤ from APP.
Only recently the molecular identity of these enzymes has been elucidated. ␥-Secretase is apparently a large complex, with presenilin being an essential component of it (4 -7). ␤-Secretase has been identified independently by 5 groups and was named Bace (beta-site APP cleaving enzyme), Asp-2, or memapsin 2 (membrane-anchored aspartic protease of the pepsin family) (8 -12). Bace is a type I integral membrane protein, with a typical aspartyl protease motif in its luminal domain. Bace fulfills most of the requirements expected for a candidate ␤-secretase. It has broad tissue distribution with higher expression in the brain (8 -10). It localizes mainly in Golgi and endosomes (8,9). Bace overexpression increases, and treatment of cells with antisense oligonucleotides complementary to Bace mRNA decreases ␤-secretase cleavage of APP (8 -12). Bace is a transmembrane protein whose predicted topology is correct with respect to the ␤-secretase cleavage site in APP. It cleaves more efficiently APP carrying the Swedish mutation than wild-type APP (9 -12). The purified enzyme cleaves synthetic APP substrates encompassing the ␤-secretase site (9 -12). Finally, Bace has an acidic pH optimum and is resistant to the aspartic protease inhibitor pepstatin A (9,10).
Although this evidence is impressive, only limited information is available on the cell biology of Bace. Bace is an Nglycosylated transmembrane protein encoded in a 501-amino acid open reading frame, from which the first 21 amino acids correspond to the signal peptide. N-terminal sequencing of Bace purified from human brain revealed that the mature protein starts at glutamic acid 46 (10), indicating that Bace is further processed after its translocation into the endoplasmic reticulum. Other proteases, e.g. proprotein convertases (PCs) and members of the ADAM family, are also synthesized as inactive proenzymes that require the removal of the propeptide to become active (13). It has recently been shown that pro-Bace is predominantly located in the endoplasmic reticulum and that constitutive propeptide cleavage takes place in the Golgi apparatus C-terminal to the Arg-Leu-Pro-Arg motif (14,15), suggestive for the involvement of members of the PC family in this process. PCs are subtilisin-like serine proteases involved in the activation of many neuropeptides, peptide hormones, growth and differentiation factors, membrane-associated receptors, adhesion molecules, blood coagulation factors, plasma proteins, and some pathogenic proteins like viral coat proteins and bacterial toxins (13,16,17). Precursors are usually cleaved C-terminal to basic motifs like Lys/Arg-(X) n -Lys/Arg, where n ϭ 2, 4, or 6 and X is essentially any amino acid but Cys and rarely Pro (13). Seven members have been thus far isolated as follows: furin, PC1 (also called PC3), PC2, PC4, PC6 (also called PC5), PACE4, and LPC (also called PC7 or PC8). All enzymes have a specific, albeit partially overlapping, expression pattern and similar but not identical substrate specificities.
Recently furin was implied in the production of amyloidogenic peptides in familial British dementia (18,19). This ob-servation stimulated us to ask whether proteases of the PC family could be involved in the regulation of the activity of Bace. Although the answer to this question is important from a cell biological point of view, we obviously also speculated that new insights in the regulation of Bace activity could foster new ideas for therapeutic intervention in AD.
We investigate here the posttranslational maturation of Bace in cells in culture, and we demonstrate that mainly furin but, in addition, although to a lesser extent, other PCs like PACE4, LPC, PC6A, and PC6B could cleave the Bace propeptide in vivo, indicating some redundancy in this controlling step. We find also that Bace activity on APP is not significantly affected by the absence of furin or by PC inhibitors, strongly suggesting that pro-Bace can process APP. We conclude that it is unlikely that the proteolytic maturation of pro-Bace is a valid therapeutic target.

EXPERIMENTAL PROCEDURES
Cloning and Mutagenesis of Bace-Two primers were designed based on the sequence of mouse Bace/Asp2 cDNA (GenBank TM accession number AF200346 (11)), corresponding to positions 1-23 (5Јatggc-cccagcgctgcactggct3Ј, sense primer) and 1483-1507 (5Јtcacttgagcagg-gagatgtcatc3Ј, antisense primer) and used for amplification. The PCR product was cloned into pGEM-T (Promega). This was subsequently used as template to introduce a C-terminal Myc tag in Bace and sBace. The latter encompasses the entire ectodomain of Bace but lacks the transmembrane domain and cytoplasmic tail. The sense primer contained the start codon, preceded by a BamHI restriction site (5Јctcg-gatccatggccccagcgctgcactgg3Ј), the antisense primer contained the Myc tag, followed by a stop codon, and an EcoRI restriction site (Bace, 5Јctcgaattctacaagtcctcttcagaaatgagcttttgctccttgagcagggagatgtcatc3Ј; sBace, 5Јctcgaattcctccaagtcctcttcagaaatgagcttttgctcataggctatggtcataag-tg3Ј). The PCR products were digested with BamHI and EcoRI and cloned in pcDNA3 (Invitrogen). Bace-ALPA and sBace-ALPA, in which the propeptide cleavage site RLPR2 was mutated into ALPA, were made using QuikChange Site-directed Mutagenesis Kit (Stratagene), according to the suppliers guidelines and using Bace-Myc as template. Note that these constructs also contain the Myc tag. All constructs were verified by sequencing.
Antibodies-A polyclonal antibody was raised in New Zealand White rabbits against a synthetic polypeptide (ETDEEPEEPGRRGSFV) corresponding to the region immediately C-terminal to the propeptide, coupled to keyhole limpet hemocyanin. Generation of antibody GM 190, directed against the propeptide of Bace, has been described before (15). Mouse anti-Myc monoclonal antibody clone 9E10 was used to detect the Myc-tagged Bace proteins. All antibodies used for immunodetection of the different members of the proprotein convertase family were obtained from Alexis Biochemicals. The polyclonal antibodies against the APP C terminus and against the ectodomain of APP have been described elsewhere (4). Monoclonal 4G8 and 6E10 antibodies were raised against A␤- (17)(18)(19)(20)(21)(22)(23)(24) and A␤-(1-16), respectively, and were obtained from Senetek. Polyclonal rabbit 53/4 specifically recognizes the ␤-secretasegenerated neoepitope and was kindly provided by Dr. Savage, Cephalon (20).
Immunofluorescence-Hippocampal neurons were cultured from embryonic day 17 C57 black embryos and co-cultured with a glial feeder layer. At day 15 post-plating, neurons were fixed using 4% paraformaldehyde in 0.1 M phosphate buffer for 30 min at room temperature followed by ice-cold methanol/acetone incubation to permeabilize the cells (36). After blocking (4°C, overnight), neurons were incubated with primary antibodies anti-Bace polyclonal and anti-␤-COP monoclonal antibody (clone MAD, Sigma) for 2 h at room temperature. For detection, Alexa488 and Alexa546 dyes coupled to secondary antibodies (Molecular Probes) were used (1 h at room temperature). Analysis was done on a NIKON inverted microscope DIAPHOT 300 (PlanApo 60/1.40 oil) connected to a Bio-Rad MRC1024 confocal microscope, and images were captured by Lasersharp (version 3.2) and processed using Adobe Photoshop 5.0 (Adobe, CA).
Cell Lines and DNA Transfer-Medium, serum, and supplements used for the maintenance of cells were obtained from Life Technologies, Inc. Chinese hamster ovary (CHO) and the furin-deficient derivative RPE.40 cells (21), N2A, and COS cells were maintained in Dulbecco's modified Eagle's medium/F12 (1:1) supplemented with 10% fetal calf serum. 8 -10 ϫ 10 5 cells/10-cm 2 culture plate were transfected with 2 g of DNA and 6 l of Fugene (Roche Molecular Biochemicals) and were used for experiments the next day (CHO and RPE.40) or after 2 days (N2A and COS cells).
Radiolabeling and Immunoprecipitation-Cells (8 -10 ϫ 10 5 cells/10 cm 2 ) were starved for 1 h in methionine-free RPMI 1640 medium and then labeled in the same medium containing 100 Ci/ml [ 35 S]methionine and chased with Dulbecco's modified Eagle's medium/F12 (1:1) for the times indicated in the figure legends. In case of overnight labeling, 5% dialyzed fetal calf serum was added to the labeling medium, and starvation was omitted. For immunoprecipitation of Bace and PCs, cells were lysed in 1 ml of DIPA (50 mM Tris/HCl, pH 7.8, 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS). Immunoprecipitations and endoglycosidase H and F digestions were performed as described (22,23). Immunoprecipitation of Bace and Bace mutants with either anti-Myc 9E10 or anti-Bace antibody gave identical results (data not shown). The figure legends indicated which antibody was used.
To study APP processing, cells were cotransfected with plasmids encoding APP, Bace constructs, furin, and PDX as indicated in the figures. Twenty four (for CHO and RPE.40 cells) or 48 h (N2A and COS cells) after transfection, cells were pulse-labeled for 4 h and immediately lysed. APP full-length and C-terminal fragments (CTFs) were immunoprecipitated from the cell extracts, whereas A␤ and total secreted APP (APPs) were immunoprecipitated from the conditioned medium as described (4,24).
To discriminate between APPs originating from ␣versus ␤-cleavage, samples from the conditional medium were resolved by 10% PAGE. Western blotting was subsequently performed with either 6E10 or 53/4 antibodies.

RESULTS AND DISCUSSION
Bace Subcellular Localization in Hippocampal Neurons-We used confocal microscopy to study the intracellular localization of Bace in primary cultures of hippocampal neurons (Fig. 1). Mouse neurons indeed express Bace, and some overlap in the distribution of Bace and the Golgi marker ␤-COP is observed. Bace is, however, also present in ␤-COP negative vesicles, most likely endosomes. Previous studies have addressed the issue of Bace subcellular distribution in non-neuronal cells using Bace overexpressed from transfected cDNA (8,9,12,14,15). Our data confirm these previous findings at the endogenous levels of expression and indicate that transfected Bace localizes to the relevant subcellular compartments. Since the levels of endogenous expression of Bace are very low (results not shown), further biochemical analysis to characterize the maturation and the activity of Bace was performed in transfected cells.
Biosynthesis of Bace-We next cloned and sequenced the cDNA encoding Bace from a brain-specific mouse cDNA library (Stratagene) and confirmed its identity to the sequence published by Yan et al. (11). To characterize the biosynthesis and maturation of Bace, pulse-chase experiments were performed. Transiently transfected CHO cells were radiolabeled and chased for various times ( Fig. 2A). Immediately after the pulse labeling, three specific protein bands were observed as follows: a major one migrating with an approximate mass of 65 kDa and two minor ones migrating at 50 and 75 kDa, consistent with previous reports (14,15). The 50-kDa protein disappeared during the chase, whereas the amount of 65-kDa protein decreased but remained prominent and the 75-kDa increased during the first 2 h of chase and remained constant afterward. Deglycosylation experiments with endoglycosidase H (Fig. 2B) demonstrated that the 75-kDa protein but not the 65-kDa protein carries complex-type oligosaccharides. Removing all N-linked sugars using endoglycosidase F resulted in a single band of 50 kDa. This indicates that the three protein species contain the same polypeptide backbone. The 50-kDa protein is therefore the unglycosylated Bace precursor; the 65-kDa protein is Bace containing simple N-linked oligosaccharides; and the 75-kDa species finally is the fully complex glycosylated protein. The large shifts in molecular weight upon glycosylation indicate that the four potential glycosylation sites of Bace are probably all utilized, although some heterogeneity is possible (14).
Analysis of the N terminus of Bace reveals a potential propeptide of 24 amino acids, ending in the sequence Arg-Leu-Pro-Arg 45 . Basic motifs are often recognized and cleaved by PCs (13, 16, 17). Since Bace purified from human brain or transfected cells starts at Glu 46 (9 -11), it is therefore likely that Bace is processed by a member of this PC family. Since the basic residues are essential for recognition by PCs (13), we substituted the arginine residues at the P4 and P1 positions in the basic motif (Arg-Leu-Pro-Arg 3 Ala-Leu-Pro-Ala). We called this mutant protein accordingly "Bace-ALPA." Upon expression in CHO cells, Bace-ALPA is apparently synthesized and N-glycan maturated in a way indistinguishable from Bace-WT (Fig. 2C).
We also generated the soluble forms of Bace-WT and Bace-ALPA (sBace-WT and sBace-ALPA, respectively) lacking the transmembrane and cytoplasmic domains. Upon transfection with both constructs, soluble Bace is synthesized and secreted and can be recovered from the conditioned medium (Fig. 2D). The soluble secreted sBace-WT and sBace-ALPA are complex glycosylated and an endoglycosidase H glycosidase-sensitive precursor can be detected in the cells (data not shown). This indicates that sBace matures during its traffic through the secretory pathway in a way similar to the wild-type full-length protein, validating it as a tool for our further investigations.
Cleavage of the propeptide should decrease the molecular mass of pro-Bace by 2,560 Da. From the experiments described above it is clear that this small difference between pro-Bace and processed Bace is too small to allow discrimination in the gel electrophoresis system used. We noticed, however, that the propeptide of Bace contains 4 arginines and no acidic residues, whereas the mature Bace protein is acidic in nature. We therefore decided to use isoelectric focusing (IEF) to separate processed from unprocessed Bace.
We used the furin-deficient CHO cell strain RPE.40 (21) to determine the possible role of furin in the processing of both Bace and sBace. We studied first whether sBace became processed by the PCs, because the pattern of protein bands to be analyzed is much less complex than that for wild type Bace (see Fig. 2). Cotransfection of sBace with furin cDNA resulted in further migration of the immunoprecipitated protein toward the anode, at the bottom of the gel (Fig. 3A), suggesting that the propeptide had indeed been removed by furin. This was confirmed using the propeptide-specific antibody ␣-pro-Bace which should not react with mature sBace and which, in fact, failed to recognize the bands generated in the presence of furin (Fig.  3B). The heterogeneity of the bands corresponding to mature sBace is commonly observed with other proteins in IEF (25). This can be explained by the heterogeneity in the glycosylation pattern of sBace together with the asparagine to aspartic residue conversion that occurs during the deglycosylation prior to IEF. Similar results were obtained with full-length Bace. Upon furin cotransfection part of pro-Bace was converted to the mature Bace form (Fig. 3, C and D; see also Fig. 5, A and C). The fastest migrating band in Fig. 3C could be immunoprecipitated by antibodies directed against mature Bace but not by antibodies directed against the propeptide (D). This indicates that this band represents processed Bace. We observed, however, that a fraction of pro-Bace was resistant to furin treatment (indicated as pro-Bace PRE ). Based on the pulse-chase experiments shown in Fig. 2, from which it is clear that a large part of wild-type Bace remains endoglycosidase-sensitive after 4 h of chase, we conclude that this fraction represents newly synthesized, immature glycosylated Bace. Since this pool is localized in the early compartments of the secretory pathway, it is not accessible for furin, which is only active in the late Golgi apparatus. The fraction of protein that is labeled Pro-Bace in Fig. 3, C and D, on the other hand, represents the fully glycosylated protein that has reached the Golgi compartment and is therefore sensitive to furin cleavage. To our surprise, both sBace-ALPA and Bace-ALPA are efficiently processed in RPE.40, furin-deficient cells (Fig. 3), and their processing is not inhibited by the PC inhibitor ␣ 1 -PDX (not shown) in contrast to their wild-type Bace counterparts. Furthermore, preliminary results using protease inhibitors indicate that this cleavage is performed by a trypsin-like protease (data not shown). It is therefore likely that the double ALPA mutation creates a novel site that becomes artificially cleaved by a non-PC, trypsin-like protease. Regardless of the identity of this protease and the precise site of cleavage, it is clear that this processing event is physiologically irrelevant. 2 In any event, the fact that wild-type Bace (Bace-WT) and "wild-type" sBace are poorly cleaved in furindeficient cells, together with the rescue of the cleavage process after furin expression, indicates strongly that furin is involved in pro-Bace maturation in vivo. That other members of the PC-family could rescue the cleavage of Bace-WT cannot, however, be excluded. Therefore, all other PCs that have broad tissue distribution, i.e. PACE 4, PC6 (isoforms A and B), and LPC, were tested for their activity in sBace (Fig. 4) and Bace (Fig. 5) maturation. Expression of the enzymes was confirmed (Fig. 5B). From Fig. 4, it is obvious that only furin was able to process pro-sBace. Furthermore, endogenous processing activity on pro-sBace was observed in CHO cells (right panel) but not in RPE.40 cells (left panel), and activity could be stimulated by expression of furin and inhibited by expression of the PC inhibitor ␣ 1 -PDX. ␣ 1 -PDX is a genetically engineered serine protease inhibitor derived from the trypsin inhibitor ␣ 1 -antitrypsin and has been shown to inhibit efficiently furin and to a lesser extent PACE4, PC6A, and PC6B (26,27). The same experiment was performed with wild-type Bace (Fig. 5). Although overexpression of furin resulted in efficient processing of pro-Bace, other PCs were capable of cleaving pro-Bace to a various extent as well. This is unlikely to be a cell type-specific effect, since similar results were obtained in the neuron-based cell line N2A (Fig. 5C). On the other hand, the data obtained with sBace suggest that furin has a preponderant role in pro-Bace processing in vivo.
APP Processing by Wild-type and Mutant Bace-To determine whether the prodomain cleavage and membrane anchorage of Bace affect ␤-secretase activity, RPE.40 cells were cotransfected with plasmids encoding APP, wild-type Bace, Bace-ALPA, sBace, sBace-ALPA, and furin or ␣ 1 -PDX as indicated in Fig. 6 (A-C). Processing of APP was analyzed by in vivo labeling and immunoprecipitation of cell-associated C-terminal fragments (CTFs, Fig. 6A), total secreted APP (APPs, Fig. 6B,  bottom), and A␤ peptides (Fig. 6B, top) or by Western blotting to discriminate between APPs␣ and APPs␤ generated by ␣Ϫ and ␤Ϫsecretase, respectively (Fig. 6C). Expression in RPE.40 cells of APP alone resulted in cleavage of APP mainly at the 2 We have recently made a single amino acid substitution into the PC consensus cleavage site RLPR of Bace to generate a GLPR site. This mutant is not processed, in agreement with our hypothesis that the wild-type Bace is processed by a member of the PC family, whereas the ALPA mutant creates an artifactual cleavage site. ␣-secretase site, as shown by the accumulation of CTFs starting at the ␣-cleavage site (Fig. 6A) and by the fact that most of the secreted APP corresponded to APPs␣ (Fig. 6C, lane 1). ␤-Stubs (Fig. 6A) and APPs␤ (Fig. 6C), on the other hand, were almost undetectable, which is entirely consistent with data obtained in many other cell lines showing that endogenous ␤-secretase activity is very low, and APP processing is mainly by the nonamyloidogenic pathway in non-neuronal cells (28). Coexpression of Bace induced cleavage of APP at the 2 ␤-secretase sites (Asp 1 and Glu 11 (9)), with the expected concomitant decrease in ␣-processing (Fig. 6, A and C). Competition between ␣and ␤-secretase for the substrate APP has been reported previously (9 -11, 29 -31). It is remarkable that CTFs (Fig. 6A) and secreted peptides (Fig. 6B) starting at position Glu 11 are far more abundant than those starting at Asp 1 under the experimental conditions used. The phenomenon is not specific for RPE.40 cells, since similar results were obtained with N2A and COS cells (Fig. 6D). We speculated that Asp 1 is still the preferred ␤-secretase site under those conditions but that the 99 amino acids CTF that is generated by Asp 1 cleavage (CTF␤1) can be further processed by overexpressed Bace to yield the Glu 11 C-terminal APP fragment (CTF␤11). That this interpretation is correct was proven by two different experi-ments. First, if Bace cleaves preferentially at Asp 1 , then APPs secreted into the medium should react with an antibody that specifically recognizes the neoepitope generated after Asp 1 cleavage. As shown in Fig. 6C, Bace transfection indeed resulted in the substantial accumulation of secreted APP that contains this neoepitope (APPs␤). Second, if the relative abundance of Glu 11 cleavage in our experiments is a consequence of the high levels of Bace expression, then decreasing Bace expression should result in a switch toward Asp 1 cleavage. This is, in fact, what we observed (Fig. 7). Transfection of N2A cells with decreasing amounts of Bace cDNA resulted in decreasing levels of Bace protein expression (Fig. 7A, 2nd panel). Immunoprecipitation of peptides secreted into the medium showed that at high levels of Bace protein (1 g of transfected cDNA) most of the secreted A␤ peptides starts at Glu 11 , whereas with decreasing Bace expression levels (until they are undetectable in our assay) the majority of the secreted peptide starts at Asp 1 ( Fig. 7A and quantification in B). As expected, there was a direct correlation between the levels of Bace expression and the amount of APPs␤ recovered in the medium, and decreasing levels of APPs␤ were accompanied by increases in the amount of secreted APPs␣ (Fig. 7A, two lower panels). Altogether, these results show that the cleavage at Glu 11 is a function of the expression level of Bace. The Glu 11 position is known to be a normal cleavage site of Bace (see for example Vassar et al. (9)). Moreover, peptides starting at this position are produced by primary cultures of neurons and are also present in plaques of AD patients (Ref. 28 and references therein). We therefore consider the Glu 11 cleavage as a reliable reflection of the Bace activity in our experimental system. In addition we analyzed also the production of APPs␤, which reflects cleavage of APP at Asp 1 . Our approach therefore allows us to evaluate the proteolytic capacity of Bace at position Glu 11 as well as at position Asp 1 and therefore to determine whether propeptide processing is needed for Bace activity or not. In this regard, processing of APP was observed even in the absence of furin, suggesting that pro-Bace is an active enzyme (Fig. 6, A-C). RPE.40 cells lack furin, and little or no mature Bace is detected after overexpression (Fig. 5A, lane 2). Since other PCs, at least when overexpressed, can cleave pro-Bace (Fig. 5), one could argue that enough mature Bace is synthesized in RPE.40 cells that would explain the observed processing of APP. Although we cannot definitively rule out this possibility, we consider this as very unlikely, since coexpression of ␣ 1 -PDX, that inhibits several other PCs in addition to furin, resulted in no detectable decrease in ␤-secretase cleavage as compared with Bace or Bace plus furin (5th lane versus 3rd and 4th in Fig. 6A). Moreover, in preliminary experiments, we found that when Bace is retained in the endoplasmic reticulum by means of a KK motif, it is still capable of cleaving APP (not shown). This Bace-KK is, as expected, not complex glycosylated. Since propeptide cleavage occurs in the Golgi and trans-Golgi network, this mutant Bace-KK protein should still contain its propeptide.
The Bace-ALPA mutant is, as expected, as active as wildtype Bace in cleaving APP (Fig. 6, A and B). Expression of sBace, finally, induces cleavage of APP mainly at position Asp 1 (Fig. 6, A and B). The fact that sBace does not cleave at Glu 11 when overexpressed in vivo suggests that sBace is less efficient in reaching the APP substrate or at least the Glu 11 site, which is closer to the cell membrane than the Asp 1 site (9) and that the transmembrane domain of Bace is needed to allow for efficient cleavage at the Glu 11 site.
In conclusion, we confirm and extend previous work that demonstrated the glycosylation and maturation of Bace in the secretory pathway (14,15). We demonstrate in particular that pro-Bace is processed by furin to its mature form. There is redundancy in the proteolytic maturation of Bace, since other members of the PC family can compensate for loss of furin activity. We present evidence suggesting that this maturation step is not essential for the ␤-secretase activity of Bace on APP. We conclude that the maturation of pro-Bace has little relevance as a therapeutic target for Alzheimer's disease. Several other functions, apart from inhibiting proteolytic activity of the proenzymes, have been found for propeptides, including roles in folding and intracellular transport. Possibly the propeptide of Bace is important for folding or intracellular transport of pro-Bace. Alternatively, it is possible that APP is not the only physiological substrate of Bace and that cleavage of other yet unidentified Bace substrates is dependent on appropriate removal of the propeptide.
Finally, our finding that pro-Bace is able to cleave APP implies that it could be active as a ␤-secretase in early biosynthetic cell compartments, i.e. in endoplasmic reticulum and early Golgi. Therefore, it is likely that pro-Bace is involved in the generation of the intracellular amyloid peptide pool as well (32)(33)(34)(35). This pool is considered by some investigators as the real culprit in Alzheimer's disease.