If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
* This work was supported by Canadian Institutes of Health Research (CIHR) Group Grant MGC-11474, by CIHR Operating Grants MOP-14466 (to N. G. S. and M. C.) and MT 14766 (to C. L.), and by the Protein Engineering Network of Centres of Excellence Program supported by the Government of Canada.The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Processing of the β-amyloid precursor protein (βAPP) by β- and γ-secretases generates the amyloidogenic peptide Aβ, a major factor in the etiology of Alzheimer's disease. Following the recent identification of the β-secretase β-amyloid-converting enzyme (BACE), we herein investigate its zymogen processing, molecular properties, and cellular trafficking. Our data show that among the proprotein convertase family members, furin is the major converting enzyme of pro-BACE into BACE within the trans-Golgi network of HK293 cells. While we demonstrate that the 24-amino acid prosegment is required for the efficient exit of pro-BACE from the endoplasmic reticulum, it may not play a strong inhibitory role since we observe that pro-BACE can produce significant quantities of the Swedish mutant βAPPsw β-secretase product C99. BACE is palmitoylated at three Cys residues within its transmembrane/cytosolic tail and is sulfated at mature N-glycosylated moieties. Data with three different antibodies show that a small fraction of membrane-bound BACE is shed into the medium and that the extent of ectodomain shedding is palmitoylation-dependent. Overexpression of full-length BACE causes a significant increase in the production of C99 and a decrease in the α-secretase product APPsα. Although there is little increase in the generation of Aβ by full-length BACE, overexpression of either a soluble form of BACE (equivalent to the shed form) or one lacking the prosegment leads to enhanced Aβ levels. These findings suggest that the shedding of BACE may play a role in the amyloidogenic processing of βAPP.
brain-derived neurotrophic factor
full-length BACE with a FLAG-epitope at the C terminus
[BACEF]V5full-length BACE with a V5-epitope at the C terminus
[BACEF]FG/V5, full-length BACE with a FLAG-epitope at the N terminus and a V5-epitope at the C terminus
full-length BACE mutated at the active site Asp93 into Ala and containing a FLAG-epitope at the C terminus
full-length BACE mutated at the active site Asp93 into Ala and containing a FLAG epitope at the N terminus and a V5-epitope at the C terminus
full-length BACE lacking the prodomain and containing a FLAG epitope at the C terminus
full-length BACE mutated at Arg45 into Ala and containing a FLAG epitope at the C terminus
full-length BACE mutated at Arg42 into Ala and containing V5 and FLAG epitopes at the N and C termini, respectively
[BACEF-R45A]FG/V5 full-length BACE mutated at Arg45 into Ala and containing V5 and FLAG epitopes at the N and C termini
cytosolic tail Cys-mutants included single [BACEF-C478A]FG
[BACEF-C482A]FG, [BACEF-C485A]FG, double [BACEF-C482A,C485A]FG, and triple [BACEF-C478A,C482A,C485A]FG Cys to Ala mutants flagged at the C terminus
polyacrylamide gel electrophoresis
β-amyloid precursor protein
a disintegrin and metalloprotease
Alzheimer's disease is a progressive degenerative disorder of the brain characterized by mental deterioration, memory loss, confusion, and disorientation. Among the cellular mechanisms contributing to this pathology are two types of fibrous protein deposition in the brain, intracellular neurofibrillary tangles consisting of polymerized tau protein, and abundant extracellular fibrils largely composed of β-amyloid1 (for reviews see Refs.
). β-Amyloid, also known as Aβ, arises from proteolytic processing of the β-amyloid precursor protein (βAPP) at the β- and γ-secretase cleavage sites. The cellular toxicity and amyloid-forming capacity of the two major forms of Aβ (Aβ40 and especially Aβ42) have been well documented (
An alternative, anti-amyloidogenic cleavage carried out by α-secretase(s) is located within the Aβ peptide sequence of βAPP, thus precluding the formation of intact insoluble Aβ. This cleavage by α-secretase within the (His-His-Gln-Lys↓Leu-Val) sequence of βAPP is the major physiological route of APP maturation. The products of this reaction are a soluble 100–120-kDa N-terminal fragment (βAPPsα) and a C-terminal membrane-bound ∼9-kDa segment (C83). In several recent reports, metalloproteinases such as ADAM9, -10, and –17 were shown to be involved in the α-secretase cleavage of βAPP (
). Enzymes within this family are typically synthesized as inactive zymogens that subsequently undergo prodomain cleavage and activation in the trans-Golgi network (TGN). Evidence has been presented showing that several ADAMs are activated in a nonautocatalytic manner by other enzymes such as the proprotein convertases (PCs) (
). Thus, it is conceivable that such enzymes may participate in a cascade leading to the activation of α-secretase. In support of this proposal, we recently demonstrated that inhibition of PC-like enzymes in HK293 cells by the α1-antitrypsin serpin variant α1-PDX (
). Correspondingly, overexpression of a PC (i.e. PC7) increased α-secretase activity. Of the previously mentioned candidate α-secretases, our ontogeny and tissue expression analyses suggest that, in adult human and/or mouse brain neurons, ADAM10 is a more plausible α-secretase than ADAM17 (
The amyloidogenic pathway of βAPP processing begins with β-secretase(s). This enzyme generates the N terminus of Aβ by cleaving βAPP within the Glu-Val-Lys-Met-↓-Asp-Ala sequence, or by cleaving the Swedish mutant βAPPsw within the Glu-Val-Asn-Leu-↓-Asp-Ala sequence. In addition, cleavage has been reported to occur within the Aβ sequence Asp-Ser-Gly-Tyr10-↓-Glu11-Val, generating Aβ11–40/42 (
). Comparative modeling of the three-dimensional structure of BACE complexed with a substrate suggested that BACE would preferentially cleave substrates having a negatively charged residue at P1′ and a hydrophobic residue at P1 (
). In fact, this is the case for the β-secretase sites in βAPP and βAPPsw as well as for the site leading to the formation of the Aβ11–40 peptide. Both BACE and BACE2 are type I membrane-bound proteins with a prodomain that, at least for BACE (
), is rapidly cleaved intracellularly. However, little else is known about the mechanism of zymogen processing of these enzymes, including whether their activation is autocatalytic or carried out by other enzymes. Recent data derived from BACE overexpressed in bacteria (
), is not autocatalytic; rather it is effected by another proteinase(s). Moreover, our developmental analysis of the comparative tissue expression of mouse BACE and BACE2 suggested that BACE, but not BACE2, is a good candidate β-secretase in the brain (
The second step in the amyloidogenic pathway of βAPP maturation involves cleavages at the γ-secretase sites (Val-Val-↓-Ile-Ala-↓-Thr-Val) to generate either Aβ40or Aβ42. Recently, in neuronal N2a cells, Aβ40 was shown to be produced within the TGN and subsequently packaged into post-TGN secretory vesicles, suggesting that the TGN is the major intracellular compartment within which the Aβ40-specific γ-secretase is active (
). Although some insoluble, N-terminally truncated Aβx-42originates in the endoplasmic reticulum (ER), Aβ40 and Aβ42 are formed primarily in the TGN. This compartment is composed of the major source of the constitutively secreted pool of Aβ that is deposited as extracellular amyloid plaques. Furthermore, the generation of either peptide requires that βAPP or its membrane-bound, β-secretase cleavage product C99, passes at least once through endosomal compartments (
). Thus, βAPP trafficking to or retention in particular cellular compartments may critically influence its processing. Although the identification of the γ-secretase(s) has not yet been conclusively established (
In the current study, we investigate whether PCs are responsible for the cleavage of the prosegment of BACE, as well as the consequences of blocking this maturation. In addition, we examine several post-translational modifications of BACE and their possible influence on the processing of βAPP and the generation of amyloidogenic Aβ peptides.
Mouse BACE and Its Mutants
Full-length mouse BACE (mBACEF) was cloned from AtT20 cells by reverse transcriptase-polymerase chain reaction (Titan One-Tube, Roche Molecular Biochemicals) using the following nested sense (S) and antisense (AS) oligonucleotides: S1 = AAGCCACCACCACCCAGACTTAGG; S2 = CTCGAGCTATGGCCCCGGCGCTGCGCTG (XhoI site at 5′) and AS1 = GAGGGTCCTGAGGTGCTCTGG; AS2 = CCTCCTCACTTCAGCAGGGAGATG. The final product (1519 base pairs) was completely sequenced, matched with the published structure (
), and then subcloned into the expression vector pcDNA3.1/Zeo (Invitrogen). To detect recombinant BACEF, we added, in phase (by polymerase chain reaction), either a V5 (GKPIPNPLLGLDST; [BACEF]V5) or FLAG (DYKDDDDK; [BACEF]FG) epitope to the C-terminal amino acid of the cytosolic tail of mouse BACE. We also prepared a BACEF construct in pIRES2-EGFP (Invitrogen) in which the FLAG epitope was introduced just after the signal peptide cleavage site (giving the sequence … GMLPA↓DYKDDDDK–QGTHL … ) and the V5 epitope at the C terminus of the molecule [BACEF]FG/V5. Other BACE constructs were also prepared as follows: 1) an active site D93A mutant single- [BACEF-D93A]FG or double-tagged [BACEF-D93A]FG/V5; 2) a prosegment deletion mutant [BACEF-Δp]FG in which the signal peptide ending at Ala19 was fused directly to the sequence … Met-Leu-Pro-Ala19-↓-Glu46-Thr-Asp-Glu-Glu-; 3) a prosegment deletion mutant of the active site mutant [BACEF-Δp-D93A]FG; 4) PC cleavage site (42R LP R45, where boldface with underlines indicate sequence position of arginine) mutants [BACEF-R45A]FG as well as the double-tagged [BACEF-R42A]FG/V5 and [BACEF-R45A]FG/V5; and 5) cytosolic tail Cys mutants, including single [BACEF-C478A]FG, [BACEF-C482A]FG, [BACEF-C485A]FG, double [BACEF-C482A,C485A]FG, and triple [BACEF-C478A,C482A,C485A]FG Cys substitutions. Soluble forms of BACE (BACES) were also prepared by deleting the transmembrane domain and cytosolic tail (CT), leaving the sequence … TDEST454 followed by a V5 epitope. These constructs included [BACES]V5, [BACES]FG/V5, [BACES-R42A]FG/V5, and [BACES-R45A]FG/V5.
Transfections and Biosynthetic Analyses
All transfections were done with 2–4 × 105 HK293 cells using Effectene (Qiagen) and a total of 1–1.5 μg of BACE construct cDNAs subcloned into the vector pIRES2-EGFP. Two days post-transfection, the cells were washed and then pulse-incubated for various times with either 200 μCi/ml [35S]Met; 400 μCi/ml Na2[35SO4], [3H]Leu, [3H]Arg, [3H]Ser; or 1 mCi/ml [3H]palmitate (PerkinElmer Life Sciences) (
). The cells were lysed in immunoprecipitation buffer (150 mm NaCl, 50 mm Tris-HCl, pH 6.8, 0.5% Nonidet P-40, 0.5% sodium deoxycholate, and a protease inhibitor mixture (Roche Molecular Biochemicals)), after which the lysates and media were prepared for immunoprecipitations (
). The monoclonal antibodies used were directed against either the FL (FLAG-M2; 1:500 dilution; Stratagene) or V5 (1:1000 dilution; Invitrogen) epitopes. Rabbit polyclonal antisera included those directed against aa 122–131 (PA-(1–756)) and 485–501 (PA-(1–757)) of human BACE (Affinity Bioreagents Inc., Golden, CO), both used at a 1 μg/ml concentration; aa 1–16 of human Aβ (produced in our laboratory, used at a 1:200 dilution); anti-β-amyloid, recognizing mostly the C-terminal part of Aβ40 (Sigma A8326, used at a 1:200 dilution); FCA18, recognizing all Aβ peptides beginning with the N-terminal Asp; FCA3340, specifically recognizing the C terminus of Aβ40; and FCA3542, recognizing the C terminus of Aβ42 (
Pro-BACES/BACES preparations were obtained by concentrating the media (5–10-fold) of HK293 cells transiently transfected with the cDNAs of [BACES]FG/V5 or [BACES-R45A]FG/V5. Proprotein convertases were obtained from the media of BSC40 cells infected with vaccinia virus recombinants of either human furin, human PACE4, mouse PC5-A (
In lieu of a precise analytical means of quantitating our pro-BACES/BACES preparations, we used identical aliquots of these HK293 preparations to ensure that the tubes within each incubation series received exactly the same starting quantity of pro-BACES.
of pro-BACES/BACES-containing HK293 media with the appropriate volume of PC-containing BSC40 medium for 1–4 h at 37 °C in a final volume of 200 μl (adjusted to 50 mm Tris acetate, pH 7.0, 2 mm CaCl2 and 0.1% Triton X-100 (v/v)). PC activity-inhibited controls comprised identical incubations for 4 h along with 1 μm of the appropriate purified PC cognate prosegment (
). Western blot analyses of the reaction products were carried out following 10% SDS-PAGE using either the FG (1:1000 dilution) or V5-HRP (1:5000 dilution) monoclonal antibodies (Stratagene). The secondary antibody for FG consisted of anti-mouse HRP-coupled IgGs (Roche Molecular Biochemicals). The disappearance of the N-terminal FLAG epitope on the Western blot was taken to represent the degree of prosegment removal by the PCs.
In Vitro Enzymatic Activity Assays
β-Secretase activity was evaluated using a 20-aa synthetic peptide (SW20) that spans the cleavage site (KTEEISEVNL↓DAEFRHDSGY) of βAPPsw. Reactions were carried out for 4–18 h at 37 °C in 100 μl of 50 mm NaOAc (pH 4.5), 10–30 μm of SW20, and 10 μg/ml of leupeptin (to inhibit low levels of a non-β-secretase proteolytic activity). The digestion products were separated via reverse transcriptase-high pressure liquid chromatography using a 1%/min trifluoroacetic acid/acetonitrile gradient (15–45%) on a C-18 column (Vydac). Peaks were quantitated according to their absorption values at 210 nm and were definitively identified using matrix-assisted laser desorption ionization/time of flight mass spectroscopy (Voyager/PerkinElmer Life Sciences). Digestions by PCs of the BACE maturation site-spanning peptide (LGLRLPR↓ETDEESEEPGRRG; pro-BACE-(39–58)) were carried out as described above for the pro-BACES/BACES Western blot preincubations (except in 100 μl), whereas digestions of this peptide by BACE were as for β-secretase activity determinations (either at pH 4.5 or 6.5). Incubations with the peptide comprising the entire prosegment of mBACE (THLGIRLPLRSGLAGPPLGLRLPR; pro-BACE-(22–45)) were carried out as described for the β-secretase activity measurements, first preincubating at either pH 4.5 or 6.5 with 10–30 μm of this peptide. Again, the digestion products were quantitated by reverse transcriptase-high pressure liquid chromatography and matrix-assisted laser desorption ionization/time of flight-mass spectroscopic analyses.
Biosynthesis and Processing of BACE
To characterize the biosynthetic pathway of BACE and its post-translational modifications, we first cloned the enzyme from the mouse corticotrophic cell line AtT20. The resultant, fully sequenced 1519-base pair product corresponded to the published mouse sequence (
). To detect membrane-bound pro-BACE or BACE, we used the V5 epitope at the C terminus of the cytosolic tail. Alternatively, we employed the N-terminal FLAG epitope (FG) immediately following the signal peptidase cleavage site to detect specifically pro-BACE. This double-tagged, full-length protein ([BACEF]FG/V5) was coexpressed in human kidney epithelial cells (HK293) either with a control (CTL) (brain derived neurotrophic factor (BDNF)) or α1-PDX cDNA. Two days after transfection, the cells were pulse-labeled with [35S]Met for 15 min (P15). They were then chased for 1 or 2 h in the presence or absence of the fungal metabolite brefeldin A (BFA), which promotes fusion of thecis-, medial, and trans-Golgi (but not the TGN) with the ER (
). Cell extracts were immunoprecipitated with either FG or V5 monoclonal antibodies and analyzed by SDS-PAGE (Fig.1). In the absence of BFA and α1-PDX at P15 (Fig. 1A), the FG epitope reveals a 66-kDa pro-BACE form that is gradually transformed first into a 64-kDa (C1 h) and then into a minor 72-kDa (C2 h) pro-BACE form. The 72-kDa form is not visible in the presence of BFA, the major band appears at 63 kDa. In contrast, the 72-kDa form is greatly enriched in the presence of α1-PDX (Fig. 1B). Treatment with endoglycosidases revealed that the 63- and 64-kDa pro-BACE forms are sensitive to both endoH and endoF, whereas the 72-kDa form is sensitive only to endoF (not shown). These data suggest that the 63- and 64-kDa bands represent immature (likely ER-resident),N-glycosylated pro-BACE, whereas the 72-kDa form represents maturely glycosylated pro-BACE. Only in the presence of α1-PDX does pro-BACE immunoreactivity accumulate in the Golgi apparatus. In immunoprecipitation experiments employing the V5 epitope, the 2-h chase period revealed mainly a 68-kDa band (Fig.1C). In the presence of α1-PDX (Fig.1D), we observed an accumulation of a 72-kDa protein reminiscent of pro-BACE (Fig. 1C). An additional control, using wild-type α1-antitrypsin to replace of α1-PDX, gave the same results as the BDNF controls presented here (not shown).
) was carried out on SDS-PAGE-purified immunoprecipitates. The C-terminally flagged 72-kDa [pro-BACEF]FG, labeled with [3H]Leu and produced in the presence of α1-PDX, had a Leu at positions 3, 7, 9, and 13 (not shown). This is consistent with the protein starting at Thr22(Ala-Gln-Gly21-↓-Thr22-His-L-Gly-Ile-Arg-Leu-Pro-Leu-Arg-Ser-Gly-Leu) just after the signal peptidase cleavage site (
). The corresponding 68-kDa protein, labeled with [3H]Ser, revealed a Ser at position 6 (not shown), compatible with the protein being mature BACE obtained following removal of the prosegment (aa 22–45) at the Arg-Leu-Pro-Arg45-↓-Glu46-Thr-Asp-Glu-Glu-Ser-Glu-Glu sequence (
To determine whether a proprotein convertase(s) could carry out the processing of pro-BACE to BACE, we transiently coexpressed in HK293 cells the double-tagged [BACEF]FG/V5 with an array of PC inhibitors including α1-PDX (
). In addition, we prepared mutant forms of BACE in which the PC consensus cleavage site Arg residues in the prosegment were replaced by Ala at positions 42 or 45 (R42A or R45A, respectively). The transfected cells were pulse-labeled for 20 min with [35S]Met and then chased for 90 min without label. Following immunoprecipitation of the cell lysates with a FG antibody, the material was analyzed by SDS-PAGE. We first note that these treatments do not affect the lower ER-associated form of pro-BACE (66 kDa) but only modulate the relative level of the pro-BACE (72 kDa) associated with the Golgi, where processing to BACE takes place. When BACE was coexpressed with either α1-PDX, pro-Fur, pro-PC5 or α2M-F, the quantity of the 72-kDa pro-BACE (pBACEG, Golgi form) was elevated (Fig. 2A). Similar results were seen for the both the R42A or R45A prosegment cleavage site mutants. In contrast, the 72-kDa pro-BACE was barely detectable in the control, pro-PC7, pro-SKI-1, or α2M coexpressions. Parallel control experiments (not shown) verified that the prosegments of PC7 (
) were able to inhibit processing of appropriate substrates by their cognate enzymes. These data strongly support the hypothesis that a PC-like enzyme may be involved in the processing of pro-BACE into BACE. The prosegment results implicate furin and PC5 as likely PC candidates, whereas PC7 and SKI-1 appear unlikely to mediate this process. The finding that the Arg residues at the predicted42RXXR45↓ site are essential for pro-BACE processing is also consistent with the reported cleavage specificities of furin and PC5 (
To define better the region of the Golgi where pro-BACE processing occurs, we coexpressed in HK293 cells [BACEF]FG/V5 with either furin or α1-PDX and then labeled the cells for 2 h with Na2[35SO4]. SDS-PAGE analyses of the FG or V5 immunoprecipitates are shown in Fig. 2B. By using the FG antibody, we observed that pro-BACE is weakly sulfated (CTL, better seen on overexposed gels). In the presence of α1-PDX, the intensity of the 72-kDa35SO4-labeledpro-BACE (pBACEG) was greatly enhanced. The V5 immunoprecipitates clearly demonstrated that BACE is sulfated and further revealed that furin digestion appears to lower the average apparent mass of sulfated BACE from 72 (pBACEG) to 68 kDa (BACEG). Finally, the data suggest that processing of pro-BACE by a PC-like enzyme into BACE occurs at the TGN or in a subsequent compartment. Not only are sulfotransferases located in this region of the secretory pathway (
In the next set of experiments, we attempted to demonstrate directly if PCs could process pro-BACE in vitro. In preliminary work, we first tested which of the PCs expected to be active in the constitutive secretory pathway could correctly cleave a peptide (pro-BACE-(38–54)) spanning the N-terminal furin consensus site. The best processing rates were observed with furin and PC5 (not shown), followed distantly by PACE4, whereas PC7 could barely cleave this sequence even when a 10-fold excess (as assessed by pERTKR-MCA hydrolysis) of activity was employed. At the same time, we observed no detectable cleavage of this peptide by either crude or partially purified soluble BACE [BACES]V5 (not shown), lending further support to the view that the BACE does not autocatalytically remove its own propeptide. We next examined the PC-mediated processing of a double- tagged soluble (S) form of pro-BACE [pro-BACES]FG/V5 expressed in HK293 cells. Western blots of the secreted enzyme probed by the FG antibody revealed that some of the enzyme was still in the form of pro-BACES. We thus incubated identical aliquots of pro-BACES from concentrated HK293 cell media with equivalent hydrolytic activities (estimated using the fluorogenic substrate pERTKR-MCA) of partially purified furin, PC5, PACE4, and PC7 for 1–4 h. The digestion products were then run on SDS-PAGE and revealed by Western blotting using either the FG or V5 antibodies. The results demonstrated that furin could completely process pro-BACE into BACE within 2 h, whereas PC5 and PACE4 had failed to complete this cleavage even after 4 h (Fig.3). PC7 is barely, if at all, able to perform this reaction. As confirmation of the identity of the enzyme(s) carrying out this conversion, we treated the 4-h pro-BACE digestion reaction with 1 μm of purified PC prosegments (pPCs) produced in bacteria as reported previously (
). Correspondingly, the pPCs of furin, PC5, and PACE4 inhibited pro-BACE processing. Analysis of the R45A mutant (Fig. 3, right-hand side) of pro-BACES with both the V5 and FG epitopes indicated that none of the PCs tested could cleave this form, consistent with processing occurring at Arg45 of the42RXXR45↓ PC consensus site. Similar results were obtained using the R42A mutant (not shown). Finally, coexpression of [BACEF]FG in furin-deficient LoVo cells (
) with each of the above PCs or with the yeast PC homologue kexin revealed that furin, kexin, and, to a lesser extent, PC5 could best mediate efficient intracellular processing of pro-BACE into BACE (not shown).
Post-translational Modifications of BACE and Their Effects on β-Secretase Activity
To investigate the functions of the prosegment and the transmembrane/cytosolic tail of BACE, we prepared a series of mutants singly tagged at the C terminus with a FG or V5 epitope. The first construct was a truncated form of full-length BACE in which the prosegment was removed (BACE-Δp). We also created Ala mutants of three Cys residues located within the cytosolic tail of BACEF that are potential Cys-linked palmitoylation sites (
). Accordingly, we made three single (Cys478, Cys482, and Cys485), as well as double (C482A,C485A) and triple (C478A,C482A,C485A) mutants. As described previously, transiently transfected HK293 cells were pulse-labeled for 20 min with [35S]Met followed by a chase of either 1 or 2 h. SDS-PAGE analysis of the FG-immunoprecipitated products (Fig.4A) revealed that, in contrast to the wild-type [BACEF]FG, the truncated [BACE-Δp]FG remains mostly in the ER, with only small amounts reaching the TGN (seen on overexposed autoradiograms). This mutant also demonstrated a high level of endoH sensitivity and a very low level of sulfation (not shown). Furthermore, microsequencing of the [3H]Arg [BACE-Δp]FG revealed an Arg11,12 sequence, clearly showing that the signal peptide was removed (not shown). These data suggest that the majority of BACE-Δp remains in the ER, with only a small fraction reaching the TGN and being sulfated. This was further corroborated by immunocytochemical evidence showing that the majority of BACE-Δp immunoreactivity was concentrated in the ER (not shown). Hence, the prodomain appears to be critical for the efficient exit of pro-BACE from the ER. On the other hand, BACESpasses rapidly through the secretory pathway, as evidenced by its accumulation in the medium after 1 h of chase (Fig. 4A) and the relatively low amounts of pro-BACES in the ER (endoH-sensitive, lower band in cells; not shown) after either 1 or 2 h of chase. By transfecting [BACES]FGinto HK293 cells and then labeling for 2 h with Na2[35SO4], we were able to examine the intramolecular site(s) at which sulfation of BACE occurs. Equal aliquots of the FG-immunoprecipitated media were digested with endoH, endoF, or arylsulfatase. Only endoF removed the35SO4 label (Fig. 4B), demonstrating that sulfation occurred on one or more matureN-glycosylation sites (
Fig. 4C shows the results of SDS-PAGE analysis of FG-immunoreactive BACE following a 2-h labeling with [3H]palmitate of HK293 cells transiently overexpressing either BACEF, its cytosolic tail Cys mutants, BACE-Δp or BACES. Both BACEF (68 kDa) and the ER-concentrated BACE-Δp (64 kDa) were palmitoylated. When each of the three Cys residues was individually mutated, we observed a significant decrease in the degree of palmitoylation (not shown). The double (C482A,C485A) mutant had ≤30% as much palmitoylation as the wild-type BACEF, whereas the triple mutant C478A,C482A,C485A was barely palmitoylated. We verified that each of the mutants was expressed to similar degrees based on their FG-immunoprecipitated reactivities following a 2-h pulse labeling with [35S]Met (not shown). These data demonstrate that palmitoylation can occur at all three of the Cys (478, 482, and 485) residues within the cytosolic tail of BACEF. Predictably, soluble BACES was not palmitoylated. The fact that the 64-kDa BACE-Δp was palmitoylated, as opposed to the mature 68-kDa BACEF, suggests that this type of post-translational modification can begin at the level of the ER (
The β−secretase activity of [BACEF]FG was first tested in HK293 cells transfected with βAPPswcDNA. Following a 3-h pulse labeling with [35S]Met (Fig. 5), the cells were exposed to either BFA, bafilomycin (an inhibitor of vesicular acidification) (
). Fig. 5A shows that BFA and the 20 °C incubation prevented FG-immunoprecipitated 66-kDa pro-BACE from escaping the ER and becoming either the 72-kDa pro-BACE or mature, endoH-resistant BACE (not shown), whereas bafilomycin exerted a retarding effect in the ER (compared with untreated cells). As shown in Fig. 5B, coexpression of wild-type BACEF and βAPPsw leads to the production of a membrane-bound ∼10-kDa intracellular product (C99) that was detected by a polyclonal antiserum raised against the N-terminal 16 aa of Aβ. This band was also observed using the Aβ N-terminal-specific antibody FCA18 (
), confirming that this cleavage product began with the correct N terminus of Aβ (starting at the β-secretase cleavage site sequence653DAEFRHDS … ) and likely ended at the C terminus of βAPP, as reported previously (
). Unexpectedly, regardless of the relative levels of BACE and pro-BACE, βAPPsw was well processed in the ER. However, C99 accumulation in the presence of bafilomycin may also result from a decreased turnover due to lysosomal alkalinization (
). In other pulse and pulse-chase experiments we observed that the maximal amount of C99 product was generated by BACEF after a 20-min pulse, consistent with production of C99 in an early secretory compartment, most likely the ER.
We next tested whether BACEF could be transformed into a soluble shed form. As shown in Fig.6A, we could indeed detect a small amount of an ∼6-kDa cellular form of [35S]Met FG-immunoreactive BACEF but not FG-labeled BACES. This suggested that at least a small amount of shedding of membrane-bound BACEF could occur. Furthermore, a similar [3H]palmitate Cys-labeled product was also observed (not shown), supporting the notion that it represents a C-terminal stump of BACE. To confirm this and to test the importance of palmitoylation of the cytosolic tail Cys in the shedding of BACE, we used three different antibodies. As shown in Fig.6B, the ∼6-kDa stump is detected by either the C-terminal (CT) FLAG epitope antibody or a commercially purchased polyclonal antiserum directed against the 485–501 CT sequence of BACE. The latter antiserum also detected the ∼6-kDa stump using wild-type BACEF that does not carry a FLAG epitope (not shown). In addition, we consistently observed that the cytosolic tail Cys-to-Ala triple mutant C478A,C482A,C485A BACEF, which was minimally palmitoylated, produced a significant increase (about 3-fold) in the level of the ∼6-kDa stump (Fig. 6B). This result suggests that palmitoylation diminishes the shedding of BACE or enhances the clearance of the ∼6-kDa stump. Finally, to verify the presence of the extracellular shedding product, we used a third antibody directed against the N-terminal (NT) 122–131 sequence of BACEF. This allowed us to detect the secreted shed form of BACE in the media of HK293 transfectants (Fig. 6C), the level of which is also enhanced in absence of Cys palmitoylation. Consistent with these data, the apparent mass of the shed form of BACE (∼58 kDa) is smaller than that of its cellular counterpart.
Wild-type and Mutant BACE Processing of βAPPsw
In the next set of experiments (Fig.7), wild-type BACE and selected BACE mutants were coexpressed with βAPPsw. As shown in Fig.7A, C99 production was evident in cells coexpressing wild-type BACEF and βAPPsw following pulse labeling for 4 h with [35S]Met. Unexpectedly, the same band, although less intense, was also obtained with the mutants [BACEF-R45A] and BACEF-Δp (Fig.7A), as well as with the [BACEF-R42A], [BACEF-C482A,C485A], and [BACEF-C478A,C482A,C485A] mutants (not shown), indicating that all of these isoforms have at least some β-secretase activity. The absence of C99 production by the active site mutant [BACEF-D93A] confirms that these activities actually correspond to BACE and its mutant forms (Fig. 7A). Notably, the soluble form of BACES produced much less C99 compared with any of the other active forms analyzed, even though similar amounts of immunoreactive BACE were expressed (not shown).
We next analyzed the secreted βAPP cleavage products using a polyclonal antibody developed against Aβ40 as well as the antibodies FCA3340 (not shown) recognizing the C terminus of Aβ40 (
). Amazingly, BACES and, to a lesser extent, BACE-Δp were by far the forms of β-secretase that ultimately lead to the highest formation of Aβ40 (Fig. 7B). In addition, BACESdid not lead to the production of Aβ11–40, suggesting that the latter reaction requires a membrane-bound form of BACE. Indeed, overexpression of either BACEF or BACER45A (as well as the Cys mutants [BACEF-C482A,C485A] and [BACEF-C478A,C482A,C485A], not shown) resulted in an elevation of the level of Aβ11–40 product with no significant change in that of Aβ40. Again, as expected, [BACEF-D93A] was inactive.
When we analyzed the levels of secreted APPSα generated by α-secretase using the same 1–16 Aβ antibody, we noticed an inverse relationship between the levels of C99 and those of secreted APPS. The constructs BACEF, [BACEF-R45A], BACEF-Δp generated higher amounts of C99 and Aβx-40 along with lower levels of secreted APPS, whereas control cells or cells overexpressing the inactive [BACEF-D93A] mutant secreted much more pronounced levels of APPS (Fig.7C). These data argue that the APPS measured with the 1–16 Aβ antibody is probably APPSα resulting from cleavage of βAPP by α-secretase either at the TGN or at the cell surface (
). In comparison, some of our other data (Fig. 5) showed that overexpressed BACE or its mutants process βAPPsw in an earlier compartment such as the ER and thus precede the action of α-secretase.
To examine further the possibility that pro-BACE has β-secretase activity, digestion analyses of a synthetic peptide substrate (KTEEISEVNL[/underln]↓DAEFRHDSGY) encompassing the βAPPsw β-secretase cleavage site were carried outin vitro using concentrated media of HK293 cells that overexpressed BACES. Our preliminary findings (not shown) indicate that preincubation of wild-type BACES (but not the [BACES-R42A] or [BACES-R45A] mutants) with furin appears to increase the cleavage of the SW20 synthetic peptide. This is consistent with our Western blot (Fig. 3) that confirmed that furin had removed the FG epitope from the prosegment of the wild-type but not from either the [BACES-R42A] or [BACES-R45A] mutants. Although these data imply that removal of the prosegment from pro-BACE enhances the activity of this enzyme, a definitive conclusion regarding the extent of activation should await the findings of more detailed quantitative studies using purified pro-BACE. In keeping with the hypothesis that the prosegment of BACE could act as an autocatalytic inhibitor of its cognate enzyme, we tested whether a synthetic peptide representing the full-length prosegment (pro-BACE-(22–45)) could function as an inhibitor. When preincubated with active BACE, 20 μm of this propeptide resulted in only an ∼20% inhibition of the Swedish peptide substrate (at 10 μm) cleavage. Since this was a weak inhibition in the presence of twice the amount of substrate, we concluded that this peptide is, at best, a poor inhibitor of BACE and did not pursue this point any further.
The discovery of the unique type I membrane-bound BACE has provided a new perspective in our understanding of β-secretase (
) indicate that it colocalizes with βAPP and ADAM10 in the cortex and hippocampus of adult mice and in the cortex of human presenile patients. Furthermore, the distributions of either BACE2 or ADAM17 were not compatible with them being candidate brain β- or α-secretases, respectively.
In this work we focused on BACE, the more plausible β-secretase, and we sought to define some of its molecular and cellular trafficking properties. We first showed that in HK293 cells BACE is synthesized as pro-BACE in the ER and then moves to the TGN where it rapidly loses its prosegment due to cleavage by an α1-PDX-inhibitable convertase(s). We next went on to show that, aside from α1-PDX and the furin-site mutated α2-macroglobulin, other inhibitors such as the preprosegments of furin and PC5 can also inhibit pro-BACE processing. N-terminal sequencing confirmed that this cleavage occurs at the site42RLPR45↓ of pro-BACE sulfated at one or more of its carbohydrate moieties. The observations that sulfation of sugars occurs in the TGN (
), are active only in this compartment or beyond indicated that processing of pro-BACE to BACE occurs either in the TGN or in post-TGN vesicles. This conclusion is consistent with those of Huse et al. (
) who showed that furin is a potential processing enzyme of pro-BACEF. However, ourex vivo and in vitro data complement and extend these findings by more clearly defining the candidate PCs likely to be responsible for BACE maturation. Thus, in vitro digestions of pro-BACE (Fig. 3) and ex vivo coexpression of pro-BACE with the PCs in furin-negative LoVo cells (not shown) demonstrated that zymogen processing was best performed by furin and less so by PC5. During revision of this article, we also became aware of an article by Creemers et al. (
) who reached similar conclusions regarding the involvement of several PCs in the processing of BACE, although, in our hands, PC7 is not a candidate BACE convertase.
Mutation of either of the arginines found to be critical for the prosegment removal, i.e. R42A or R45A, did not result in significant alteration of the trafficking rate of pro-BACE to the TGN, as estimated by pulse-chase (Fig. 2A) and sulfation rate analyses. While this article was in preparation, we became aware of two reports (
) regarding the prosegment removal of human BACE, their data, like ours, revealed that this processing occurs in the TGN and that BACES traffics more rapidly than BACEF toward the TGN. However, our data also differ from theirs, which suggests that the R45A mutant of human BACE does not exit the ER. Our triplicate pulse-chase data (Fig. 2A) clearly demonstrate that the exit of both pro-BACEF and pro-BACEF-R45A (or R42A) to the TGN is slow but does in fact occur to a similar extent for both forms.
An interesting observation occurred when we analyzed the rate of exit of pro-BACE from the ER at 20 °C, a temperature that normally blocks the budding of TGN vesicles but should not prevent movement from the ER to the TGN (
). We found that under these conditions, pro-BACE cannot exit the ER, as is the case with BFA and, much less so, bafilomycin treatments (Fig. 5A). This is reminiscent of the observation that αβ integrins do not exit the ER at 20 °C because of their inability to form heterodimers (
). Our data show that pro-BACE can process βAPPsw into C99 in the ER (Fig.5B), suggesting that γ-secretase activity could be the limiting factor at 20 °C. Even though the holoenzymes BACE and pro-BACE (not shown) exhibit an in vitro pH optimum of 4.5 for cleavage of synthetic peptides mimicking the β-site (
), our data argue in favor of active BACE within the neutral pH environment of the ER (Fig. 5B). The combined observations that the active site mutant [BACEF-D93A] can lose its prosegment (not shown), that BACE did not cleave the PC cleavage site spanning peptide (aa 39–58 of BACE), and that PCs such as furin and PC5 can remove the prosegment of BACE in vitro and ex vivo support the notion that BACE does not auto-activate, but likely requires a furin-like enzyme for zymogen activation. Alternatively, we cannot rule out the possibility that there are other enzymes or proteins that can interact with pro-BACE and activate it by cleavage or dislocation of its prosegment. Our finding that the BACE zymogen is activated by a PC is similar to processing of the relatively inactive prorenin to renin by PC5 (
). Modeling of mouse pro-BACE, based on the structure of the closely related human progastricsin, suggested that the full-length prosegment acts as a flap covering the active site of BACE and that the furin-processing site42RXXR45↓ is quite accessible to cleavage (not shown).
Next, we showed that full-length BACEF is palmitoylated at Cys residues 478, 482, and 485 within the cytosolic tail and that a soluble form of BACES is not (Fig. 4C). Interestingly, BACES seems to be secreted rapidly from and does not accumulate within the cell, suggesting that the cytosolic segment of BACEF must contain determinants that control cellular trafficking rates and destination. One such element could be Cys palmitoylation, since we found by pulse-chase experiments that the triple mutation C478A,C482A,C485A slows the exit of pro-BACE from the ER (not shown). However, immunocytochemical analysis of the localization of [BACEF]FG and [BACEF-C478A,C482A,C485A]FG failed to reveal gross qualitative differences in their cellular distribution (not shown). In contrast, our data show that preventing palmitoylation of these Cys residues significantly enhances the shedding of a soluble form of BACE into the medium (Fig. 6, B and C). Thus, although the role of palmitoylation of BACE, which begins to occur in the ER, remains to be fully elucidated, this modification most likely provides a second anchor to the plasma membrane, possibly directing the protein to discrete membrane microdomains and/or resulting in a remodeling of the structure of its cytoplasmic region (
In an effort to define the importance of cellular trafficking on the production of C99 and Aβ, we compared the ability of various engineered forms of BACE to process βAPPsw and ultimately to generate amyloidogenic peptides following a 3-h pulse labeling. Surprisingly, overexpression of the soluble form of BACESresulted in a very significant increase in the levels of secreted Aβ40, but not Aβ11–40, as measured by an Aβ40-specific antiserum A8326 (Fig. 6B). These data were also confirmed with the FCA18, FCA3340, and FCA3542 antibodies (see “Experimental Procedures”), which revealed that most of the processed material was indeed Aβ40 and not Aβ42 (not shown). This experiment, which was repeated four times, revealed that, in the presence of BACEF, intracellular C99 production is enhanced, and secreted Aβ11–40 becomes clearly visible. In contrast, in the presence of BACES, we note a small increase in intracellular C99 accompanied by a large enhancement in the level of secreted Aβ40 (Fig. 6B). Pulse-chase analyses demonstrated that C99 is formed early, since we could observe its maximal production within the first 20 min of radiolabeling (not shown). The only difference is that at the 20-min pulse, the level of intracellular C99 is much higher with BACEF as compared with BACES, whereas following a 90-min chase, secreted Aβ40 was only observed with BACES. These data suggest that BACEF and BACES may process βAPPsw in different intracellular compartments, ultimately leading to the accumulation of either C99 or its γ-secretase product Aβ40. In view of the rapid exit of BACES from the ER and its fast trafficking through the TGN and past the cell surface (Fig. 4A), the production of C99 by BACES may be favored in a micro-compartment close to where γ-secretase is active. An exciting extension of this model (which will require extensive verification) is that shedding may enhance the amyloidogenic potential of BACE. Indeed, a small amount of an ∼6-kDa C-terminal membrane-bound stub and a soluble 58-kDa form resulting from BACEF shedding were observed in HK293 cells (Fig. 6). Moreover, palmitoylation of BACE cytosolic tail Cys residues appears to suppress shedding (Fig. 6, B andC), perhaps by controlling the cellular localization of BACEF as has been reported for other proteins (
) did not affect the generation of either C99 or Aβ by endogenous secretases (not shown), suggesting that this mutant does not act as a dominant negative, as was the case for the active site mutant of the candidate α-secretase ADAM10 (
In conclusion, our data revealed that BACE can process βAPPsw in the ER and that furin or PC5 process the zymogen in the TGN, possibly to optimize its activity in acidic cellular compartments. BACE undergoes a number of other post-translational modifications such as carbohydrate sulfation and cytosolic tail Cys palmitoylation that may finely regulate its rate of trafficking and cellular destination(s). The latter modification seems to reduce the level of BACE ectodomain shedding, which may provide a safety mechanism to reduce the amyloidogenic potential of BACE. Since expression of soluble BACE (as opposed to full-length BACE) ultimately leads to a much higher production of the amyloidogenic peptide Aβ, inhibiting the shedding of BACE may have value as an alternative strategy in the treatment of Alzheimer's disease. Finally, thein vivo physiological function of BACE remains to be elucidated, along with the possibility that this enzyme may be part of a larger complex with other proteins, including the other secretases involved in the processing of βAPP.
We thank A. Lemieux for her technical help throughout this study. The secretarial assistance of S. Emond is appreciated.