Cleavage of Amyloid- (cid:1) Precursor Protein and Amyloid- (cid:1) Precursor-like Protein by BACE 1*

Site-specific proteolysis of the amyloid- (cid:1) precursor protein (APP) by BACE 1 and (cid:2) -secretase, a central event in Alzheimer disease, releases a large secreted extracellular fragment (called APP S ), peptides of 40–43 residues derived from extracellular and transmembrane sequences (A (cid:1) ), and a short intracellular fragment (APP intracellular domain) that may function as a transcriptional activator in a complex with the adaptor protein Fe65 and the nuclear protein Tip60. APP is closely related to APP-like protein (APLP) 1 and APLP2, but only APP is known to be cleaved by BACE 1 and to be involved in Alzheimer disease. We now demonstrate that similar to APP, APLP1 and APLP2 are also cleaved by BACE 1 but not by ADAM 9, another APP protease, and also transactivate nuclear Tip60 in a complex with Fe65. Paradoxically, although BACE 1 cleavage appears to be specific for APP and APLPs, their cleavage sequences exhibit no homology, and a short sequence (7 amino acids) from APP that when placed close to the membrane converts a membrane protein that is normally not cleaved by BACE 1 into a BACE 1 substrate. Our data demonstrate that APLPs and APP are processed simi-larly to act via the same nuclear target, suggesting that BACE 1 cleavage regulates a common function of APP and APLPs in neurons. The amyloid- precursor protein of Alzheimer disease

The amyloid-␤ precursor protein (APP) 1 of Alzheimer disease is an ubiquitous membrane protein (1-3) that is physiologically processed by site-specific proteolysis (reviewed in Refs. 4 -7, see Fig. 1A). First, cleavage by ␣or ␤-secretases releases the large extracellular part of APP (called APP S ) leaving the CTF (Cterminal fragment), which is composed of a small remaining extracellular stub, the TMR, and the cytoplasmic tail to remain. The CTF of APP is then cut by ␥-secretase at multiple positions in the TMR (see Refs. 8 and 9). ␥-Cleavage liberates an intracellular cytoplasmic fragment referred to as AICD (10,11) that may function as a transcriptional activator (12,13) and have other signaling roles (14,15). In addition, ␥-cleavage generates small peptides that are derived from the TMR and adjacent extracellular sequences. These peptides include A␤40 and -42, the major components of amyloid fibrils in Alzheimer disease (14,15) (reviewed in Refs. 4 and 5).
The AICD of APP binds to several cytoplasmic adaptor proteins including Fe65 and its isoforms (16 -19). Transcriptional activation by the AICD, as measured using heterologous DNA binding domains that are fused to APP, requires binding of Fe65 to the AICD (12). Fe65 contains three conserved domains that are necessary for transcriptional activation, an N-terminal WW domain whose ligand is unknown, a central phosphotyrosine binding domain that binds to the nuclear chromosome remodeling protein Tip60, and a second C-terminal phosphotyrosine binding domain that mediates the interaction of Fe65 with the AICD (12). The potent transactivation of transcription by the AICD-Fe65 complex via binding to Tip60 suggests a model whereby the cytoplasmic tail of APP, liberated by the ␥-cleavage, translocates to the nucleus as a complex with Fe65 and binds to nuclear Tip60 to transactivate transcription. Although the endogenous DNA binding domains (beyond those associated with Tip60 (20)) have not been identified, this model is supported by three findings. First, the cytoplasmic tail of APP is detectable in the nucleus of cultured cells, albeit at low levels (10,11). Second, unmodified APP together with Fe65 is a potent transactivator of Gal4 fused to Tip60, which is inactive in transcription without Fe65 and APP (12). Third, a direct target gene (the KAI1 gene encoding a tetraspanin) for the AICD-Fe65-Tip60 complex was identified in a study on how interleukin-1␤ activates transcription (21), thereby demonstrating the in vivo relevance of the original experiments using heterologous DNA binding domains.
APP belongs to a gene family that includes two additional members, APLP1 and APLP2 (22)(23)(24). APP, APLP1, and APLP2 are closely related and exhibit the same domain structure (see Fig. 1A). Knock-out mice revealed that APP, APLP1, and APLP2 are functionally redundant (25). Similar to APP, APLPs appear to be cleaved by extracellular proteases followed by proteolytic processing by ␥-secretase (26 -28). Furthermore, Gal4 fusion proteins of APLPs also transactivate Gal4-dependent transcription in an Fe65-dependent manner (27,28). These observations suggest that APLPs may be proteolytically processed by mechanisms similar to APP and may have similar functions. However, not all evidence is consistent with this hypothesis. It is noticeable that although the extracellular domains and the AICDs of APP and APLPs are highly homologous, the linker sequences connecting these domains exhibit no homology (Fig. 1B). Because this linker sequence in APP is the substrate for ␣and ␤-secretases, it is unclear whether the same extracellular proteases digest APP and APLPs. This is a particularly important question for BACE 1, the primary ␤-secretase (29 -33), because it is the major extracellular processing enzyme for APP in neurons, and because it initiates production of A␤ from APP (4 -6). This question is not only important for understanding BACE 1 function but also for insight into the biological role of APP and APLPs, because a common cleavage of these proteins by BACE 1 would support the notion that proteolytic processing of these proteins by a common set of enzymes serves to regulate a physiological function such as transcriptional activation.
In the present study, we have investigated whether APLPs can be substrates for BACE 1 even though APLPs exhibit no sequence homology to the BACE 1 cleavage site in APP and whether BACE 1 cleavage has an effect on the putative transcriptional activation function of APP and APLPs. Our data demonstrate that BACE 1 alters proteolytic processing of APLPs in transfected cells, suggesting that APLPs are substrates for BACE 1 in vivo. We demonstrate that APLP1 is cut by BACE 1 at a position that is within 13 residues of the TMR resulting in the secretion of essentially the entire extracellular sequences of APLP1. In the same assay, ADAM 9 (which is one of several candidate ␣-secretases for APP (34,35)) specifically cleaved APP but not APLP1. Furthermore, we show that BACE 1 substrate recognition is specific because other cell surface proteins are not cleaved in the same assay and that a short sequence from APP (seven residues) is sufficient to confer BACE 1 cleavage onto a normally non-cleaved protein. These results indicate that BACE 1 may have a coordinate function in regulating APP and APLP processing in neurons.
Transfection and Transactivation Assays-These assays were performed essentially as described (12,40). Briefly, HEK293, HeLa, and COS7 cells were transfected at 50 -80% confluency in 6-well plates , an acidic sequence that chelates metal ions, a Kunitz domain (Ku) that is only present in APP and APLP2 and is alternatively spliced (2, 3), a cysteine-poor but conserved sequence (central APP domain, CAPP-domain), and a C-terminal linker that connects the extracellular domains to the transmembrane region (TMR). Positions of the ␣-, ␤-, ␥-, and ⑀-cleavage sites in the linker and the TMR of APP are indicated. ␥-Cleavage and ⑀-cleavage (which are presumably carried out by the same presenilin-dependent enzyme (8,9)) release the cytoplasmic tail to generate the intracellular fragment (AICD). The percent identity (fully conserved residues) between human APP, APLP1, and APLP2 is shown only below those domains (the CRD, CAD, and AICD) that are significantly conserved among all three proteins. The location of the secreted extracellular APP/APLP fragment (APP S /APLP S ) and the CTF generated by ␣-/␤-cleavage is indicated. B, alignment of the C-terminal sequences of APP, APLP1, and APLP2 starting with the C-terminal half of the CAD. Identical residues are highlighted with a domain-based color code (blue, CAD; yellow, linker; green, TMR; black, AICD). The alternatively spliced sequences in the linker are boxed on a pink background, and excision of this sequence creates a recognition site for chondroitin sulfate glycosaminoglycans, which are attached to the serine residue marked with an asterisk (56 -58). The A␤ sequence in APP is shown in red and underlines; the ␣-, ␤-, ␥-, and ⑀-cleavage sites are marked by arrows. The C-terminal sequences in APP and APLPs and the internal APLP1 sequence that were used to raise antipeptide antibodies are boxed. The position of the Fe65 binding site in the AICD is indicated. Note that although the CAD and the AICD are highly conserved, the linker between the extracellular domain and the TMR exhibits no sequence similarity between APP, APLP1, and APLP2.
SDS-PAGE and Immunoblotting-These experiments were performed using standard procedures with either regular SDS or Tricine gels (41)(42)(43). Lanes were loaded with ϳ20 g of protein from whole cell extracts harvested in sample buffer and analyzed by immunoblotting after SDS-PAGE using antibody dilutions of 1:1000 -2000 for primary antibodies and 1:10,000 horseradish peroxidase-conjugated goat antirabbit secondary antibodies (ICN). For immunoblotting analysis of secreted proteins, supernatants from transfected cells (30 l with ϳ5 g of protein) were loaded/lane.

RESULTS
Coordinate Transcriptional Activity of APP, APLP1, and APLP2-Gal4-Tip60, although part of a nuclear protein complex that binds to DNA (20), is unable to mediate Gal4-dependent transcription but requires both Fe65 and APP for transactivation (12). The Gal4-Tip60 assay is a more stringent test for the transcriptional function of APP than the use of APP-Gal4 fusion proteins, because in the Gal4-Tip60 assay APP is not modified. To determine whether APLP1 and -2, which were shown previously as Gal4-fusion proteins to transactivate transcription in an Fe65-dependent manner (27), can substitute for APP in the Gal4-Tip60 assay and to test whether Tip60 is also a potential target for APLPs, we co-transfected into HEK293 cells increasing amounts of expression plasmids encoding APP, APLP1, or APLP2 and a constant amount of plasmids expressing Gal4-Tip60 and Fe65. We then measured Gal4-dependent transactivation using a co-transfected reporter plasmid and corrected all transactivation data for ␤-galactosidase activity produced by a co-transfected control expression vector to account for differences in transfection efficiency (Fig. 2).
For all three proteins (APP, APLP1, and APLP2), we observed a bell-shaped dose-response curve of transactivation that depended on APP or an APLP ( Fig. 2A). We next performed a similar analysis with a constant amount of APP or APLP1 and APLP2 but with increasing amounts of Fe65 (Fig.  2B). Again, we detected dose-dependent, specific transactivation of Gal4-dependent transcription. The absolute amount of transactivation varied depending on the precise combination of plasmids that we co-transfected. The highest levels of transactivation (400-fold increase) were reached when maximal amounts of Fe65 were co-transfected with APLP1 (Fig. 2B). To test whether transactivation mediated by APP, APLP1, or APLP2 in these assays involves ␥-secretase cleavage, we examined the effect of the ␥-secretase inhibitory DAPT on transactivation (Fig. 2C). Two relatively low concentrations of DAPT (2 and 10 M) were used. We observed significant inhibition of transactivation by DAPT for all three proteins (APP, APLP1, and APLP2). Inhibition was incomplete, possibly because the doses of DAPT do not fully inhibit ␥-secretase, or because a DAPT-independent proteolytic pathway contributes to transactivation. Together these results document that APLP1 and -2 can efficiently substitute for APP in Tip60-dependent transactivation, suggesting that consistent with a common transcriptional function for APP and APLPs, Tip60 is a common nuclear target. A, transactivation of transcription mediated by Gal4-Tip60 in HEK293 cells that were co-transfected with a constant amount of Fe65 plasmid (0.5 g) and increasing amounts of APP, APLP1, or APLP2 plasmids as indicated. B, same as in A, except that constant amounts of APP (0.5 g), APLP1 (0.25 g), and APLP2 plasmids (0.5 g) were co-transfected with increasing amounts of Fe65 plasmid. C, same as in A, except that constant amounts of APP (0.5 g), APLP1 (0.25 g), and APLP2 plasmids (0.5 g) were cotransfected with constant amounts of Fe65 plasmid, but the transfected cells were then treated with increasing amounts of the ␥-secretase inhibitor DAPT. In A and B, the numbers below the bars indicate the amount of the variable plasmid transfected and in C indicate the concentration of DAPT in the medium. All cells were additionally transfected with Gal4-Tip60 expression vector (0.5 g), the Gal4 reporter plasmid pG5E1B-luc (0.15 g), and a control plasmid to correct for transfection efficiency (0.10 g of pCMV-LacZ). All data were corrected for transfection efficiency and normalized to the amount of transactivation observed in the absence of APP or APLPs. Data shown are means Ϯ S.E. of triplicates from a representative experiment. known to be cut by extracellular proteases and a presenilin-dependent ␥-secretase similar to APP (26 -28). In non-neuronal cells, several proteases that belong to the ADAM family of metalloproteases are thought to cleave APP as ␣-secretases, whereas in neurons, APP is probably primarily cleaved by the ␤-secretase enzyme BACE 1 in which activity is responsible for production of the pathogenic A␤ peptides (4 -6). Although APLPs are cleaved extracellularly in non-neuronal and neuronal cells, it is not known whether the same enzymes cleave APP and APLPs. This question is particularly important for BACE 1 for which only two other substrates, the sialyltransferase ST6Gal and P-selectin glycoprotein ligand 1, besides APP are currently known (44,45). In preparation to examining APLP cleavage, we generated antibodies to the cytoplasmic C-terminal sequences of APP, APLP1, and APLP2. In addition, we produced an antibody to a peptide from the linker region of APLP1 that connects the TMR to the central domain (Fig. 1B, boxed sequence). We then tested the specificity of these antibodies by the immunoblotting of HEK293 cells that had been transfected with APP, APLP1, or APLP2 expression vectors (Fig. 3). We found that the antibodies reacted specifically with their cognate proteins except for a small degree of cross-reactivity of the APLP2 antibody with APP and APLP1.
In the next set of experiments we employed the newly raised antibodies to probe for APP, APLP1, and APLP2 cleavage in cells that either expressed these proteins alone or in combination with BACE 1 (Fig. 4). As negative controls, we examined two other neuronal cell surface proteins, neurexin 1␤ (Nrx1␤) and SynCAM (36,37,39). For each of the five proteins studied, six conditions were investigated: control cells that were transfected without or with BACE 1 (Fig. 4, lanes 1 and 2) and test cells that expressed APP, APLP1, APLP2, neurexin 1␤, or SynCAM either alone or together with BACE 1, which were additionally either left untreated or incubated with the ␥-secretase inhibitor DAPT (lanes 3-6). The DAPT treatments were performed to stabilize the CTFs, proteolytic intermediates that are produced by extracellular ␣-/␤-cleavage of a cell surface protein and are subsequently digested by ␥-secretase (Fig. 1). We then analyzed the transfected proteins by immunoblotting to monitor the full-length proteins and their CTFs as a function of BACE 1 and DAPT treatment.
Immunoblotting of control cells showed that transfected BACE 1 by itself did not produce immunoreactivity with any of the antibodies (Fig. 4, lanes 1 and 2). In transfected test cells incubated without DAPT, CTFs were observed only for APP and the APLPs but not for neurexins 1␤ or SynCAM (Fig. 4.,  lanes 3 and 4). Under these conditions, co-transfection of BACE 1 had little effect on full-length APP or APLPs but altered their  4 and 6). Transfected cells were incubated in the absence (lanes 1-4) or presence of the ␥-secretase inhibitor DAPT (2 mM; lanes 5 and 6). Transfected proteins (indicated on the right) were examined by immunoblotting (20 g of cell protein/lane) using regular SDS-PAGE (top panel for each protein) to monitor the full-length proteins (FL), or by Tricine gel electrophoresis to detect their CTFs (bottom panel for each protein). All transfected proteins are type I transmembrane proteins, and all immunoblotting was carried out with antibodies to the cytoplasmic C termini of the proteins. Nontransfected cells and cells transfected only with BACE 1 (lanes 1 and 2) were used as controls to ensure that the various antibodies did not recognize an endogenous or BACE 1-dependent signal. Asterisks in lane 4 for APP, APLP1, and APLP2 indicate BACE 1-dependent CTFs. Numbers on the left display positions of molecular weight markers. Data shown are from a representative experiment independently repeated multiple times.
CTFs. For APP and APLP2, the size of the CTF was shifted up, whereas for APLP1 a distinctive CTF was first induced by BACE 1 expression (Fig. 4, asterisks in lane 4). When DAPT was added, the CTFs for APP, APLP1, and APLP2 became very abundant, independent of BACE 1 expression, making it difficult to detect an effect of BACE 1 (Fig. 4, lanes 5 and 6). Surprisingly, CTFs were also observed for neurexin 1␤ upon DAPT treatment (Fig. 4). The neurexin 1␤ CTFs could reflect a physiological extracellular cleavage of neurexin 1␤ by an enzyme different from BACE 1, because BACE 1 had no effect on the CTFs under any conditions. However, studies of primary cultures of neurons incubated with DAPT failed to yield evidence for a physiological extracellular cleavage of ␤-neurexins (data not shown), and at present it is unclear whether or not neurexin 1␤ is normally cleaved. Together these data suggest that in the absence of BACE 1, APP and APLPs are efficiently cleaved in transfected cells by secretases, likely endogenous ␣-secretases, but that BACE 1 expression alters cleavage of all three proteins consistent with the notion that all three proteins are BACE 1 substrates.
To confirm that BACE 1 cleaves APLPs, we examined the effect of BACE 1 co-transfection on Gal4-Tip60 dependent transactivation by APLP1 and Fe65 similar to the assay described in Fig. 2. Co-transfecting a small amount of APLP1 with Fe65 and Gal4-Tip60 resulted in a small degree of transactivation (3-fold). However, adding increasing amounts of BACE 1 expression plasmid to the co-transfection mix caused a proportional increase in transactivation (Fig. 5), indicating that BACE 1 cleavage of APLP1 contributes to Gal4-Tip60-dependent transcription. Control experiments without APLP1 showed no such increase demonstrating that BACE 1 alone does not transactivate Gal4-Tip60 by inducing cleavage of endogenous APP.
How Specific Is BACE 1 Cleavage Measured by Immunoblotting in Transfected Cells?-The possible cleavage of APLP1 and -2 by BACE 1 is surprising considering the lack of sequence homology in the APP cleavage site for BACE 1 (Fig. 1B). Although the co-transfection/cleavage assay for BACE 1 cleavage has been used previously to confirm BACE 1 enzyme activity (46), the specificity of BACE 1-mediated cleavage under these conditions has not been investigated, and the substrate specificity of BACE 1 and of other secretases has been studied primarily with short synthetic peptides (47). The data in Fig. 4 demonstrate that BACE 1 does not randomly cleave overexpressed cell surface proteins, because neurexin 1␤ and Syn-CAM were not digested. However, the data raise the question about what determines the cleavage specificity of BACE 1. Is cleavage a property of a specific sequence in APP, possibly in combination with the proximity of that sequence to the membrane, or are the extra-or intracellular domains of APP involved in directed BACE 1 cleavage? To address these questions, we constructed a hybrid neurexin 1␤/APP protein in which the extracellular sequence of neurexin 1␤ is fused to the C-terminal sequences of APP at a position just N-terminal to the normal BACE 1 cleavage site (Fig. 6A). Comparison of the cleavage of APP and of the neurexin/APP hybrid protein in transfected cells revealed that they were almost identically processed, with a similar production of CTFs that were significantly shifted to larger sizes by co-expression of BACE 1 (Fig.  6B, lanes 3, 5, 7, and 9). For both APP and the APP/neurexin hybrid, DAPT induced a massive accumulation of CTFs consistent with normal digestion of the CTFs by ␥-secretase (Fig.  6B, lanes 4, 6, 8, and 10). Thus the extracellular domains of APP N-terminal to the ␤-secretase cleavage site are dispensable for BACE 1 cleavage.
Does the Intracellular AICD or the TMR of APP Direct BACE 1-dependent Cleavage?-We tested this question by inserting into full-length neurexin 1␤ a seven-residue sequence from APP that encompasses its normal BACE 1 cleavage site (sequence EVKMDAE) or the corresponding sequence from the Swedish APP mutant that is preferentially cleaved by BACE 1 (sequence EVNLDAE (48)). The insertion was placed just outside of the TMR, corresponding to the normal position of this sequence in APP, but no other APP-related sequences were present in the neurexin 1␤ derivatives (referred to as Nrx1␤*Wt or Nrx1␤*Sw). We then tested the effect of the insertions on the cleavage of Nrx1␤ by BACE 1 (Fig. 6C).
An analysis of co-transfected cells revealed that BACE 1 did not induce a decrease in the levels of full-length Nrx1␤ that did not contain an insertion (Fig. 6C, lanes 3-6). BACE 1 caused the appearance of low levels of a CTF from Nrx1␤ that was almost undetectable and could only be seen as a faint band in the presence of DAPT (Fig. 6C, asterisk in lane 6). Thus overexpressed BACE 1 cleaves co-expressed Nrx1␤ at extremely low rates. However, insertion of the seven-residue sequence was sufficient to convert neurexin 1␤ into a full-fledged BACE 1 substrate (Fig. 6C, lanes 7-14). With both Nrx1␤* constructs, BACE 1 co-expression resulted in a dramatic loss of full-length Nrx1␤*. This loss was because of the complete digestion of Nrx1␤* by BACE 1, because the addition of DAPT induced the accumulation of high levels of CTFs (Fig. 6C, lanes 10 and 14). Interestingly, Nrx1␤*Sw was cleaved at the ␤-secretase cleav- age site in the transfected cells even in the absence of cotransfected BACE 1, suggesting that low levels of endogenous BACE 1 are sufficient to partially digest Nrx1␤*Sw (Fig. 6C,  lane 8). These data demonstrate that even in the context of a resident membrane protein, a very short substrate sequence is sufficient to confer BACE 1 cleavage onto a protein.
Where Is APLP1 Cleaved by BACE 1?-It is puzzling that BACE 1 cleavage is relatively specific in that, for example, neurexin 1␤ is not normally a substrate but is cut by BACE 1 when the seven-residue substrate sequence is inserted. At the same time, BACE 1 cleavage appears to be relatively nonspecific because APLP1 and -2 are also cleaved by BACE 1 at a site that exhibits no sequence similarity with the APP cleavage site for BACE 1. Do all extracellular proteases that cleave APP also digest APLPs, and do these proteases cleave APLPs at a position comparable with the APP cleavage position? We addressed these questions by a more detailed study of the cleavage of APLP1 (chosen because it is less similar to APP than APLP2) and by comparing the activity of BACE 1 with that of ADAM 9, a disintegrin metalloprotease that may be one of several ␣-secretase enzymes (43,46). To validate the activity of BACE 1 and ADAM 9, we first examined APP (Fig. 7).
Transfection of BACE 1 or ADAM 9 alone into HEK293 cells had no significant effect on the processing of endogenous APP probably because the majority of the endogenous APP is from non-transfected cells (Fig. 7, lanes 1-3). Without BACE 1 or ADAM 9, a single major CTF was observed from transfected APP at steady-state (Fig. 7, lanes 4 and 7). Co-transfection of APP with BACE 1 at two concentrations (Fig. 7, lanes 5 and 6) or ADAM 9 at two concentrations (lanes 8 and 9) increased the abundance of the CTF from APP, consistent with cleavage of APP by both transfected BACE 1 and ADAM 9. In parallel, APP S , the secreted extracellular fragment of APP that results from ␣-/␤-cleavage, was monitored with antibodies to the extracellular APP sequences in the medium from the transfected cells. A dramatic increase in APP S production was observed both after BACE 1 and after ADAM 9 co-transfection (Fig. 7). The CTF that was produced by ADAM 9 in the transfected cells appeared to be larger than expected for regular ␣-secretase cleavage of APP, suggesting that ADAM 9 may cleave APP at a site that is not the normal ␣-secretase site. However, ADAM 9 unequivocally enhanced production of CTFs in these experiments, suggesting that even if ADAM 9 is not the physiological ␣-secretase, it can still serve as a control for BACE 1 because it is another enzyme that cleaves APP in the extracellular domain. Thus together these results demonstrate that in co-FIG. 6. Determinants of BACE 1 cleavage of APP. A, diagrams of the structures of APP, the neurexin 1␤/APP hybrid protein (Nrx␤/APP), and the modified neurexin 1␤ (Nrx1␤*). Nrx1␤*Sw contains a sevenresidue insertion from the Swedish mutant of APP (sequence EVNL-DAE), whereas Nrx1␤*Wt contains the corresponding insertion from wild-type APP (sequence EVKMDAE). Insertions were placed into neurexin 1␤ just N-terminal to the TMR. In the neurexin 1␤/APP hybrid protein, the nearly complete extracellular sequence of neurexin 1␤ (residues 1-375) is fused to APP such that only a short sequence of the extracellular linker from APP followed by the TMR and AICD from APP are included (residues 494 -695 of APP 695 ). B, comparison of the effect of BACE 1 and DAPT on CTFs produced in HEK293 control cells (lanes 1 and 2) or transfected cells expressing APP (lanes 3-6) or the neurexin 1␤/APP hybrid (lanes 7-10). Cells transfected and treated as indicated (see also legend to Fig. 4) were examined by immunoblotting with antibodies to the C terminus of APP after standard gel electrophoresis (top panel) or Tricine gel electrophoresis (bottom panel). C, same as in B, except that the effect of BACE 1 on the cleavage of neurexin 1␤ and neurexin 1␤* was examined. Note that under the conditions of the experiment, BACE 1 induces a large decrease in the steady-state levels of full-length neurexin 1␤* but not of normal neurexin 1␤, suggesting that neurexin 1␤* is cleaved quantitatively. In B and C, numbers on the left indicate positions of molecular weight markers, and the BACE 1-dependent CTFs are identified by asterisks.  2 and 3, respectively), cells transfected either with APP alone (lanes 4 and 7) or together with two concentrations of BACE 1 plasmid (lanes 5 and 6), or ADAM 9 plasmid (lanes 8 and 9). Blots show the analysis of cellular proteins to detect full-length APP (FL) separated on regular SDS gels (top) and CTFs separated on Tricine gels (middle panel). In addition, the media of the cells were collected and analyzed for the secreted proteolytic APP fragment (APP S , bottom panel). The top two blots were obtained with antibodies to the cytoplasmic C terminus of APP, and the bottom blot was obtained with antibodies to extracellular APP sequences. transfected cells, both BACE 1 and ADAM 9 cleave APP and increase secretion of APP S . We next performed similar experiments for APLP1 (Fig. 8), probing the intracellular sequence with the antibody to the C terminus used above and the secreted extracellular domain sequences (APLP1 S ) with a new polyclonal antibody that we raised to a short synthetic peptide derived from the extracellular sequence of APLP1 (Fig. 3) just outside of the TMR (Fig.  1B, boxed). We tested three different cell lines (HEK293, HeLa, and COS cells) to exclude cell-specific artifacts. Indeed, we observed slight differences in the properties of the CTFs produced in the different line lines, consistent with differences between these cells in the presence of various ␣-secretases (see below). Nevertheless, in all cell lines co-expression of BACE 1 caused a large change in CTFs, whereas co-expression of ADAM 9 had no detectable effect (Fig. 8). Furthermore, no APLP1 S was observed in the medium from transfected cells except after BACE 1 expression. Because the antibody to the extracellular APLP1 sequences used in this experiment is to a sequence that is very close to the TMR (Fig. 1B), we cannot rule out that an APLP1 S fragment is produced in the absence of BACE 1 in APLP1-transfected cells by a secretase that cleaves N-terminal to the antibody used. However, the clear-cut appearance of the APLP1 S fragment after BACE 1 expression demonstrates that BACE 1 cleaves APLP1 in the 13 residues that are located between the epitope of the antibody and the TMR (Fig. 1B). DISCUSSION Our data support the notion that APP is functionally redundant with APLP1 and -2 as described in the survival of knockout mice (25) and the similar transactivation obtained with APP-and APLP-Gal4 fusion proteins (27) and extended here with the demonstration that APLPs similar to APP act via nuclear Tip60 (Fig. 2). Furthermore, we show that all three proteins are cleaved by BACE 1, the enzyme that is responsible for the ␤-secretase activity of APP cleavage and is the most abundant neuronal APP cleavage protein. The evidence that BACE 1 cleaves APLP1 and -2 can be summarized as follows. Co-transfection of BACE 1 with APLP1 or -2 results in the production of a new CTF (Fig. 4). Increasing amounts of BACE 1 enhance transactivation of Gal4-Tip60 mediated by APLP1 (Fig. 5). Co-expression of BACE 1 with APLP1 results in a large and specific increase in the secretion of APLP1 S (Fig. 8). A newly generated antibody against a short extracellular sequence of APLP1 that is adjacent to the TMR only recognizes a secreted APLP1 S proteolytic fragment when BACE 1 is cotransfected with APLP1, localizing the BACE 1 cleavage site to a position next to the TMR (Fig. 8). The majority of this evidence rests on the use of a transfection assay whereby BACE 1 is co-transfected with APP or an APLP, and the effect of BACE 1 on the production of proteolytic fragments from the co-transfected protein is examined. This assay was validated by demonstrating that two other neuronal cell surface proteins, neurexin 1␤ and SynCAM (34 -36), are not cleaved by BACE 1 (Fig.  4), and by the finding that a seven-residue sequence from APP, when inserted into neurexin 1␤ next to the TMR, is sufficient to convert neurexin 1␤ into a BACE 1 substrate (Fig. 6). The fact that APLPs are BACE 1 substrates similar to APP, that BACE 1 is the major extracellular neuronal secretase for APP, and that APLPs and APP are functionally redundant suggests that BACE 1 cleavage is an integral component of the functions of APP and APLPs in neurons. BACE 1 does not seem to have a stringent substrate recognition sequence. In APP, BACE 1 has two cleavage sites in which the sequences suggest that BACE 1 prefers a hydrophobic residue preceding the cleavage site and an acidic residue following the cleavage site (47). However, the ␤-cleavage sequences are not conserved in APLP1 or -2, and no similar sequence motif can be readily identified in APLP1 or -2. The 13-residue sequence in APLP1 that must be cleaved by BACE 1, based on the antibody data shown in Fig. 8, best resembles the second C-terminal BACE cleavage site of APP. Because of overall sequence similarity between the cleaved sequences in APP and APLPs, the cleavage of both APP and APLPs by BACE 1 was unexpected, leading us to hypothesize initially that additional specificity for BACE 1 cleavage may be provided by other parts of APP and APLPs, especially the extracellular cysteine-rich domain (CRD) and the central APP domain, which are highly conserved among APP and APLPs (Fig.  1). However, replacement of a large part of APP with neurexin 1␤ had little effect on the cleavage by BACE 1, whereas insertion of only seven amino acids from the ␤-secretase cleavage site of APP conferred BACE 1 cleavage onto neurexin 1␤. Therefore, as long as a BACE 1 site is accessible to the enzyme, BACE 1 cleavage does not require any additional sequence information in vivo. If so, how is the BACE 1 cleavage regulated? One possibility is that the accessibility of the BACE 1 site could be regulated by a ligand or a protein that interacts with APP. Binding may physically block the access of BACE 1 directly or induce a conformational change so that the ␤-site is  2 and 3, respectively), or cells transfected either with APLP1 alone (lanes 4 and 7) or together with two concentrations of BACE 1 plasmid (lanes 5 and 6) or ADAM 9 plasmid (lanes 8 and 9). Three different cell lines were analyzed to control for possible cell-type-specific effects on cleavage. Blots show analysis of cellular proteins to detect full-length APLP1 (FL) separated on regular SDS gels (top) and CTFs separated on Tricine gels (middle panel). In addition, the cell media were collected and analyzed for the secreted proteolytic APLP1 fragment (APLP1 S ) in the bottom panel.
The top two blots were obtained with antibodies to the cytoplasmic C terminus of APLP1, and the bottom blot was obtained with antibodies to a short extracellular peptide sequence of APLP1 that is located 13 residues N-terminal to the TMR (boxed in Fig. 1). Numbers on the left indicate positions of molecular weight markers. masked. The access of ␤-site could also be controlled by posttranslational modifications of APP. Another possibility is that BACE 1 is the target for regulation. For example, the intracellular localization of BACE 1 can be manipulated so that the cleavage of APP and APLPs can be regulated. Consistent with this idea, it has been demonstrated that phosphorylation of BACE 1 at the C terminus regulates the trafficking of BACE 1 within the secretory and endocytic pathway (49).
The demonstration that APLPs are also BACE 1 substrates is consistent with the importance of BACE 1 as a processing enzyme for APP in the brain and the structural and functional similarity of APP with APLPs. Because APLPs are also BACE 1 substrates, A␤-like peptides must be produced from APLPs but are presumably shorter than A␤ given the fact that, at least in APLP1, the BACE 1 cleavage site appears to be closer to the TMR than in APP (Fig. 8). Thus cleavage of APP and APLPs produces very similar secreted large extracellular proteins (APP S , APLP1 S , and APLP2 S ) and soluble intracellular fragments (AICDs) but distinct small secreted peptides that are composed of non-homologous sequences. It remains to be established whether these diverse peptides execute distinct functions in addition to a common function of APP and APLPs mediated by their conserved domains. Another implication of our findings is that if APP and APLP cleavage is as physiologically important as suggested by transgenic experiments (50), then inhibition of BACE 1 in the brain will interfere with APP and APLP processing and may thus inhibit their common functions. As the list of substrates of ␥-secretase grows, the therapeutic strategy for Alzheimer disease has shifted from inhibiting ␥-secretase to modulating ␤and ␣-secretases. BACE 1 inhibitors have been suggested as a promising cure for Alzheimer disease because knock-out mice with BACE 1 have no detectable abnormalities, and the level of A␤ in these mice is reduced (51,52). However, inhibition of BACE 1 may abolish the functions of APP and APLPs as suggested in our study, and thus drugs that target BACE 1 may produce side effects when applied in treatment of Alzheimer patients.
The similarities between APP and APLPs make it likely that the function of APP is not directly involved in Alzheimer disease, because APLPs have not been linked to this disease. Nevertheless, their functions are probably important for Alzheimer disease. The abundance of the proteolytic fragments that are produced by APP and APLP cleavage will depend on the regulation of cleavage. If cleavage by ␤-secretase is upregulated or cleavage of ␣-secretase down-regulated, A␤ production will increase. The example of other transcription pathways suggests that the cleavage of APP and APLPs and the putative transcriptional signal caused by the cleavage may be controlled by extracellular ligands and be components of a regulatory loop (e.g. see the Notch paradigm; Refs. 53 and 54). Indeed, heparan sulfate has been proposed as a possible ligand to regulate APP cleavage (55). At present, the nature of the ligands that influence cleavage and of the intracellular regulatory proteins that may provide feedback regulation of cleavage are not well studied. However, if the regulation of cleavage and transcription is changed only slightly over long time periods such changes could lead to the accumulation of A␤ to cause Alzheimer disease. Thus identification of the signals and mechanisms involved in regulating the function of APP and APLP s is of prime importance, because one possible mechanism by which one could interfere with the pathogenesis of Alzheimer disease would be to alter these signals and mechanisms.