Proteolytic processing of amyloid-beta precursor protein by secretases does not require cell surface transport.

Cleavage of amyloid-beta precursor protein (APP) by alpha-,beta-, and gamma-secretases releases an extracellular fragment called APPS, small Abeta peptides, and a short APP intracellular domain that may provide a transcriptional signal analogous to the Notch intracellular domain. Notch cleavage is activated by extracellular ligands on the cell surface, but the cellular localization of APP cleavage remains unclear. We now show that in transfected cultured cells, the plasma membrane SNARE protein syntaxin 1A, when expressed as a full-length protein, disrupts the Golgi apparatus and blocks trans-Golgi traffic and exocytosis. In contrast, truncated syntaxin 1A1-243 selectively abolishes exocytosis but has no apparent effect on trans-Golgi traffic or Golgi structure, whereas further truncated syntaxins 1A1-236 and 1A1-230 have no effect on either exocytosis or Golgi traffic. Using these syntaxin 1A fragments, we demonstrated that blocking trans-Golgi traffic greatly impairs APP cleavage and AICD-dependent nuclear signaling, whereas blocking exocytosis alone does not affect either process, even though secretion of APPS and Abeta40 peptide is abolished. Our data suggest that, different from Notch, cleavage of APP is independent of cell surface regulation by extracellular ligands but may be controlled by intracellular signaling.

Amyloid-␤ precursor protein (APP) 1 of Alzheimer's disease is a ubiquitous membrane protein that is physiologically processed by site-specific proteolysis (1)(2)(3)(4). First, cleavage of APP by ␣or ␤-secretases releases a large fragment called APP S that contains most of the extracellular sequences of APP. A small extracellular stub, the transmembrane region, and the cytoplasmic tail of APP (referred to as "AICD" for APP intracellular domain) remain in the membrane after ␣/␤-cleavage. These APP sequences are subsequently cleaved by ␥-secretase at multiple sites in the transmembrane region (5,6). ␥-secretase cleavage results in the intracellular release of the AICD and the extracellular release of small peptides, including A␤40 and A␤42, the major components of amyloid fibrils in Alzheimer's disease (see reviews cited above).
The processing pathway of APP resembles that of Notch, a cell surface protein that functions as a ligand-dependent regulator of cell fate (7,8). Notch exerts its regulatory effects by transcriptional activation of target genes, which is directly mediated by the released Notch intracellular domain (NICD). The cleavage of Notch that produces the NICD is probably performed by the same ␥-secretase complex that also generates the AICD of APP (7,8). Furthermore, the AICD may function as a transcriptional activator similar to the NICD (9), although it likely also has other signaling roles (10 -15). These similarities suggest that APP may be cleaved on the cell surface by a ligand-regulated mechanism, but the subcellular localization of APP cleavage has not been established. Extensive evidence suggests that ␣and ␤-secretases cleave APP after it has traversed the Golgi complex. For example, in cultured cells, APP is only cleaved after it has been fully glycosylated (16), and in neurons, APP is cleaved after it has entered into the fast axonal transport pathway (17)(18)(19)(20)(21)(22). However, several alternative subcellular localizations of cleavage have been suggested (23), and even the studies that report cleavage in a post-Golgi compartment do not address the question whether cleavage occurs prior to exocytosis or on the cell surface. The cellular localization of ␥-secretase cleavage is also unclear. Presenilins, a major component of ␥-secretase, are localized to the endoplasmic reticulum, the cell surface, and/or endocytic vesicles (24 -26). ␥-Cleavage of APP has been linked to cholesterol-rich lipid rafts on the cell surface similar to Notch (27) or to endocytosis (28,29). Another hypothesis indicates that a large amount of APP ␥-cleavage occurs in a pre-Golgi compartment (23), possibly by a mechanism that does not depend on presenilins (30 -32).
The cellular localization of APP cleavage is of central importance for APP function. Does APP act as a ligand-activated cell surface receptor analogous to Notch, or does APP itself constitute a ligand, either in a cell surface bound form as uncleaved APP or in a secreted form as the proteolytic products APP S and A␤ peptides? Addressing this question is not only essential for understanding the function of APP but also has implications for the role of the AICD, which may be activated either by an extracellular ligand or by intracellular signals. To our knowledge, no manipulation is currently available that selectively inhibits exocytosis and would allow addressing this question directly. Tetanus toxin light chain effectively impairs regulated exocytosis of synaptic vesicles by inactivating synaptobrevin/ vesicle-associated membrane protein (33,34) and also cleaves a ubiquitous synaptobrevin homolog called cellubrevin (35). However, overexpression of tetanus toxin light chain does not impair constitutive exocytosis (36) and thus cannot be used to interfere with all exocytosis. In this paper, we exploit the unexpected observation that a fragment of the plasma membrane SNARE protein syntaxin 1A is a selective inhibitor of all exocytosis but does not alter trans-Golgi traffic, whereas fulllength syntaxin 1A blocks traffic through the Golgi complex as described previously (37). This allows for the first time testing of whether surface expression of APP is required for cleavage. We find that in cells deficient in constitutive exocytosis, ␣/␤cleavage and ␥-cleavage of APP proceed normally in contrast to cells where Golgi traffic is disrupted, demonstrating that the biology of APP is fundamentally different from that of Notch.
Transfection and Transactivation Assays-For APP trafficking and glycosylation analyses, HEK293 cells were transfected in 6-well plates with 200 ng of APP expression vector and 1 g of empty vector or various syntaxin 1A vectors/well using FuGENE 6 transfection reagent (Roche Applied Science). For BACE experiments, HEK293 cells were transfected with 0.5 g of APP expression vector, 1.5 g of empty vector or various syntaxin 1A vectors, and 50 ng of BACE expression vector or empty vector. Transfected cells and medium were collected 36 -48 h post-transfection. The cells were washed in PBS and lysed in SDS-PAGE loading buffer for immunoblotting or processed for treatment with PNGFase (New England Biolabs), O-glycosidase, and neuraminidase (Sigma) according to the manufacturer's recommendations. To detect secreted APP, 1-ml aliquots of the cell culture medium were collected, and proteins were precipitated with trichloroacetic acid. Transactivation assays were performed in HEK293 and COS cells essentially as described in Refs. 9 and 38. The cells were co-transfected in 6-well plates with the following 4 plasmids: well, and the luciferase and ␤-galactosidase activities were determined with the luciferase assay kit (Promega) using a microplate luminometer (Orion, Berthold Detection Systems), and the standard O-nitrophenyl-D-galactopyranoside (Sigma) method, respectively. The luciferase activity was standardized by the ␤-galactosidase activity as a control for transfection efficiency and general effects on transcription. Values shown are averages of transactivation assays carried out in duplicate, and repeated at least three times.
Measurement of A␤40 Peptide Levels-HEK293 cells and medium were collected 48 h after transfections. The cells were washed with PBS, lysed in extraction buffer (5 M GnCl, PBS, pH 8.0) at 50 l/well, and sonicated. The cell extracts were diluted 20-fold with sample buffer (PBS, pH 8.0, containing 5% bovine serum albumin, 5 mM EDTA, and complete protease inhibitor mixture (Roche Applied Science)), and cleared by centrifugation. Total proteins in 1-ml aliquots of cleared cell culture medium were trichloroacetic acid-precipitated, dissolved in 50 l of extraction buffer, and diluted 20-fold with sample buffer. A␤40 standards were prepared by dispersing the synthetic peptide in sample and extraction buffer mixture (20:1). A␤40 levels in samples and standards were measured in duplicates by colorimetric enzyme-linked immu-nosorbent assay (BIOSOURCE) according to the manufacturer's recommendations.
Miscellaneous Procedures-For measurements of constitutive exocytosis, HEK293 cells were co-transfected in 6-well plates using FuGENE 6 reagent with pHGHCMV5 and a control vector or various syntaxin 1A constructs at 1:20 ratio (typically 0.05 g of pHGHCMV5 and 1 g of a test plasmid/well). Constitutive hGH secretion was measured as described previously (36,45) and is expressed as fold-over cellular hGH. Immunocytochemistry of transfected HeLa cells was performed essentially as described by Cao and Sü dhof (9) using polyclonal syntaxin antibodies and monoclonal antibodies to GM130 (Transduction Laboratories) and goat anti-rabbit or goat anti-mouse secondary antibodies coupled with Alexa Fluor 488 and Alexa Fluor 546 (Molecular Probes). The labeled cells were imaged on a Leica TCS SP2 confocal microscope.

Stage-specific Inhibitors of Secretory Membrane Traffic
Based on Syntaxin 1A-syntaxin 1A is a SNARE protein of the plasma membrane that mediates Ca 2ϩ -induced synaptic vesicle exocytosis and possibly other forms of exocytosis (46,47). syntaxin 1A is composed of an N-terminal three-helical H abc domain, a SNARE motif that participates in core complex formation with other SNARE proteins, and a C-terminal transmembrane region (Fig. 1A). To test whether expression of different fragments of syntaxin 1A interferes with secretory membrane traffic, we employed four syntaxin 1A proteins, fulllength syntaxin 1A 1-288 , and truncated syntaxin 1A 1-243 , 1A 1-236 , and 1A 1-230 . These proteins were co-expressed with hGH as a reporter protein that is constitutively secreted in non-neuronal cells (48).
We found that full-length syntaxin 1A 1-288 or C-terminally truncated syntaxin 1A 1-243 severely inhibited constitutive hGH secretion in HEK293 and HeLa cells, whereas truncated syntaxin 1A 1-236 and syntaxin 1A 1-230 , which contain only 7 or 13 fewer residues, respectively, than syntaxin 1A 1-243 , had no significant effect ( Fig. 1B and data not shown). The seven residues that are present in the inhibitory syntaxin 1A 1-243 protein, but absent from the innocuous syntaxin 1A 1-236 protein, are located in the middle in the SNARE motif. This suggests that the inhibitory effect of syntaxin 1A 1-243 may be due to the formation of inappropriate SNARE complexes, which requires a minimal length of the SNARE motif for SNARE complex formation. However, it is surprising that full-length syntaxin 1A is also inhibitory because it should behave "normally." To explore whether the two inhibitory syntaxin 1A constructs act by distinct mechanisms, we examined the effect of transfected syntaxin 1A constructs on intracellular organelles using immunofluorescence staining for syntaxin 1A and organelle-specific markers.
In agreement with previous studies (37), full-length syntaxin 1A 1-288 caused a disorganization of the Golgi apparatus in HeLa cells as visualized with an antibody to the Golgi protein GM130 (49) (Fig. 1C). Truncated syntaxin 1A 1-243 , however, had no effect on the morphology of the Golgi complex or any other organelle examined, despite the nearly equal inhibition of hGH secretion caused by this syntaxin construct (Fig. 1, B and C, and data not shown). Furthermore, other truncated syntaxins also induced no change (data not shown). Transfection of different syntaxins into HEK293 and Vero cells had identical effects on Golgi morphology, suggesting that these results are not dependent on the cell type used (data not shown). This observation indicates that syntaxin 1A 1-243 may inhibit hGH secretion by selectively interfering with exocytosis, whereas full-length syntaxin 1A 1-288 disrupts trans-Golgi traffic, and shorter truncated syntaxins, such as syntaxin 1A 1-236 , have no effect. The selective effect of syntaxin 1A 1-243 appears to operate in all cells tested by a mechanism that involves Munc18 proteins. 2 Differential Effects of Syntaxin 1A Constructs on APP Cleavage and Glycosylation-We co-transfected APP with control vectors or the various syntaxin 1A constructs and examined the effect of the syntaxin fragments on the glycosylation and cleavage of APP in HEK293 cells. We used immunoblotting to estimate the levels of full-length APP and of the C-terminal cleavage products of APP in the cells and measured some of the cleavage products using quantitative immunoblotting in which protein amounts are determined by 125 I-labeled secondary antibodies (44). Furthermore, to assess intramembranous ␥-cleavage of APP, we compared untreated transfected cells with cells that had been treated with DAPT, an inhibitor of presenilin-dependent ␥-secretase (50). Finally, we employed two gel systems, a standard gel system (41) to resolve larger proteins and a tricine gel system (42) to separate small peptides such as the C-terminal fragments.
In untreated transfected cells, APP was present in multiple bands that likely correspond to partially glycosylated immature and fully glycosylated mature species (6,51). In addition, a small amount of the C-terminal APP fragments was detected ( Fig. 2A). As expected, treatment with the ␥-secretase inhibitor DAPT increased the abundance of the C-terminal fragments but had no effect on the banding pattern of full-length APP. When we co-expressed full-length syntaxin 1A 1-288 with APP, no C-terminal fragment of APP was detectable, even after the addition of DAPT ( Fig. 2A). Furthermore, glycosylation of APP was altered, as evidenced by the loss of the mature forms of APP on the immunoblots. In contrast, when we co-transfected truncated syntaxin 1A 1-243 (which inhibits exocytosis) or syntaxin 1A 1-236 (which has no effect on exocytosis), we observed no change in the production of the C-terminal fragment or the apparent glycosylation of APP ( Fig. 2A).
To ensure that the truncated syntaxin 1A 1-243 did not cause a kinetic impairment in APP cleavage, we quantified the amount of accumulated C-terminal fragment of APP in cells after 20 min, 60 min, and 180 min of DAPT treatment (Fig. 2B). Quantitations were performed with 125 I-labeled secondary antibodies and phosphorimaging detection. We observed no difference in the rate of APP cleavage between cells in which APP had been transfected alone or co-transfected with syntaxin 1A 1-243 or syntaxin 1A 1-236 , whereas no C-terminal fragments were detectable after transfection with full-length syntaxin 1A ( Fig. 2B and data not shown).
APP is glycosylated by both N-and O-linked sugars (51). To obtain a more detailed characterization of the glycosylation of APP after either trans-Golgi traffic or exocytosis were inhibited by co-expression of full-length syntaxin 1A 1-288 or truncated syntaxin 1A 1-243 , respectively, we examined the effect of deglycosylating enzymes on the mobility of APP. Proteins from cells co-transfected with control vector or the various syntaxin expression vectors were treated with PNGFase (which removes N-linked sugars), with O-glycosidase (which digests O-linked sugars), and with neuraminidase (which cleaves sialic acid) (Fig. 2C). We found that the changes in electrophoretic mobility of APP induced by such treatments were identical between 2 M. Khvotchev  control-transfected and syntaxin 1A 1-243 -expressing cells, suggesting similar glycosylation patterns. In contrast, APP from cells expressing full-length syntaxin 1A 1-288 exhibited only a small PNGase F-dependent change in electrophoretic mobility (Fig. 2C). This result indicates that similar to other treatments affecting the integrity of the Golgi apparatus or the exit of APP from the endoplasmic reticulum (52,53), full-length syntaxin 1A 1-288 interferes with O-glycosylation of APP.
Effect of Syntaxin 1A Fragments on BACE1 Cleavage of APP-In brain, the major ␤-secretase for APP is BACE1 (54 -57). To test whether transfection of full-length syntaxin 1A 1-288 also inhibits proteolytic cleavage of APP by BACE1, we cotransfected APP, BACE1, and control or syntaxin 1A vectors in various combinations into HEK293 cells. Transfection of BACE1 in control experiments resulted in two distinct C-terminal fragments (CTFs) that are of higher molecular weight than the CTF produced by endogenous secretases present in HEK293 cells (Fig. 3A) (58) and increased the total amount of APP cleavage ϳ3-fold, as measured by quantitative immunoblotting (Fig. 3B). Co-expression of full-length syntaxin 1A 1-288 completely blocked APP cleavage not only by endogenous secretases, but also by BACE1, whereas truncated syntaxin 1A  had no significant effect on BACE1-dependent cleavage (Fig. 3,  A and B).
Effect of Syntaxin 1A Fragments on the Secretion of APP S and A␤40 Peptide That Are Produced by APP Cleavage-APP cleavage by ␣or ␤-secretases produces a large secreted fragment called APP S . Subsequent ␥-cleavage releases small peptides, which include A␤40, A␤42, and p3 (reviewed in Refs. [1][2][3][4]. If syntaxin 1A 1-243 generally blocks constitutive exocytosis, secretion of these APP cleavage products should also be inhibited, even when cleavage proceeds normally. To test this, we examined the secretion of APP S from transfected HEK293 FIG. 2. Effect of syntaxin 1A fragments on APP cleavage and glycosylation. A, HEK293 cells were co-transfected with APP expression vector encoding full-length APP 695 and an empty control vector or syntaxin 1A vectors. Transfected cells were cultured in the absence (Ϫ) or presence (ϩ) of the ␥-secretase inhibitor DAPT as indicated, and APP was then analyzed by immunoblotting with polyclonal anti-APP antibody (U955) to the APP C terminus using two different gel systems: an 8% Tris-glycine gel to resolve different glycosylation variants of APP (top) and a 15% Tricine gel to visualize the C-terminal ␣/␤-cleavage products (CTF ␣/␤ ; bottom). Numbers on the left indicate positions of molecular mass markers. B, quantitation of C-terminal ␣/␤-cleavage products in HEK293 cells that were co-transfected with APP and a control plasmid or plasmids encoding the C-terminally truncated syntaxins 1A 1-243 and 1A 1-236 . The amounts of C-terminal fragment and of full-length APP were measured by immunoblotting with 125 I-labeled secondary antibodies and phosphorimaging device detection performed in quadruplicates. The levels of the C-terminal fragment were normalized to the total APP signal (full-length APP ϩ C-terminal fragment) to control for differences in expression and are expressed as a percentage of the amount at t ϭ 180 min. Data shown are means Ϯ S.E. from a single representative experiment repeated three times. C, glycosylation of APP in HEK293 cells co-transfected with APP expression vector and an empty control vector or various syntaxin 1A vectors. Cell extracts were prepared 48 h after transfection and treated with the de-glycosylating enzymes as indicated. Samples were analyzed by immunoblotting with polyclonal anti-APP antibody (U955). The dashed lines in A and C indicate the fully glycosylated APP forms; the dotted lines in C mark the deglycosylated form.

FIG. 3. BACE1-dependent cleavage of APP is blocked by fulllength (but not by truncated) syntaxin 1A.
A, immunoblot of HEK293 cells that were co-transfected with APP expression vector, an empty control or syntaxin 1A vectors, and BACE1 as indicated. Note that transfected BACE1 generates two distinct C-terminal fragments (CTF␣/␤) of APP. B, quantitation of the results shown in A. Relative levels of full-length APP and CTFs were measured using 125 I-labeled secondary antibodies and phosphorimaging detection. Combined levels of both CTFs were used for the samples transfected with BACE1. Each sample was loaded in triplicates, and results shown are means Ϯ S.E. from a single experiment independently performed multiple times. n.d., not detectable. cells that either express APP alone or co-express APP with various syntaxin 1A proteins. As expected, we found that fulllength syntaxin 1A 1-288 (which disrupts Golgi apparatus and blocks APP cleavage) abolished APP S secretion (Fig. 4A). Cotransfection of syntaxin 1A 1-243 also blocked APP S secretion, whereas syntaxin 1A 1-236 or the syntaxin-binding protein Munc18-1 had no significant effect.
We next investigated the effect of various syntaxin proteins on the secretion of A␤40 peptides. Co-transfection of BACE1 with APP or transfection of the Swedish APP mutant (59) did not have a major effect on the amount of A␤40 in cells but significantly enhanced A␤40 secretion (Fig. 4B) (60). In the absence of BACE1, syntaxin 1A 1-288 significantly reduced both secreted and cellular levels of A␤40. Unexpectedly, when syn-taxin 1A 1-288 and BACE1 were co-transfected, we observed a massive accumulation of A␤40 in the cells, whereas secretion of A␤40 was still blocked. Finally, when we co-transfected syntaxin 1A 1-243 (which blocks constitutive exocytosis), A␤40 accumulated in the cells (500 -600% of control levels), but the secretion of A␤40 into the medium was still severely inhibited (Fig. 4B). BACE1 co-transfection had little effect on cellular and secreted A␤40 levels in the cells transfected with syntaxin 1A  . Together these results confirm that blocking constitutive exocytosis with syntaxin 1A 1-243 inhibits secretion of APP cleavage products. The differential requirement for BACE1 in causing the accumulation of A␤40 in cells (i.e. the fact that in cells in which trans-Golgi traffic is blocked with full-length syntaxin 1A, A␤40 only accumulates when BACE1 is co-expressed, whereas in cells in which constitutive exocytosis is blocked with syntaxin 1A 1-243 , A␤40 accumulates independently of BACE1; see Fig. 4B) is probably because of the fact that endogenous BACE1 in HEK293 cells is present in a post-Golgi compartment (54 -57), and thus A␤40 is produced in cells in which trans-Golgi traffic is blocked only when BACE1 is overproduced.
Effect of Syntaxins on Transactivation Mediated by the Cytoplasmic Tail of APP-To examine the effect of syntaxin 1A on APP cleavage by an independent approach, we next measured APP-mediated transactivation as a function of syntaxin cotransfection in HEK293 cells, using a system that we had described previously for studying the function of the AICD (9,38). For these experiments, we utilized APP containing a Gal4-VP16 module that functions as a powerful transcriptional activator of Gal4-dependent promoters (61). We employed three APP constructs in which the Gal4-VP16 module was inserted just N-terminally to the AICD: full-length APP, the 99-residue C-terminal fragment of APP that is normally produced by ␤-secretase cleavage (C99), and the isolated AICD. In these experiments, we controlled for transfection efficiency in all samples with a co-transfected LacZ reporter construct. Furthermore, to examine the role of ␥-secretase in transactivation, we measured the effect of DAPT on the amount of transactivation observed. All transactivation results were normalized for the signal obtained under control conditions. Under control conditions, DAPT inhibited transactivation mediated by full-length APP/Gal4-VP16 and C99/Gal4-VP16 but not transactivation produced by the AICD/Gal4-VP16 protein (Fig. 5A). This result is consistent with a role for ␥-secretase in releasing the AICD from full-length APP and the C99 fragment (62), an activity that is not needed for the AICD/Gal4-VP16 fusion protein. Full-length syntaxin 1A 1-288 severely depressed transactivation mediated by full-length APP/Gal4-VP16 and by C99/Gal4-VP16 but again had no significant effect on transactivation produced by AICD/Gal4-VP16 (Fig. 5B). The inhibition of transactivation caused by syntaxin 1A 1-288 was more effective than that caused by DAPT, possibly because DAPT did not block all ␥-secretase activity or because other secretases can also release the AICD. However, the small amount of transactivation observed by C99/Gal4-VP16 in the presence of syntaxin 1A 1-288 was still sensitive to DAPT (Fig.  5B). The syntaxin 1A 1-243 protein, which blocks exocytosis but has no effect on cleavage (Figs. 2-4), also had no significant effect on transactivation by any of the APP-derived proteins (Fig. 5C), confirming the biochemical results. Furthermore, syntaxin 1A 1-236 (which does not alter exocytosis) also had little effect on transactivation (Fig. 5D).
In the transfected cells, the various syntaxin 1A constructs are expressed at different levels, which may at least, in part, explain their differential effects on APP processing. To address this concern, we examined the effects of increasing amounts of FIG. 4. Effect of syntaxin 1A fragments on secretion and intracellular accumulation of APP S and A␤40 peptide. A, HEK293 cells were co-transfected with APP expression vector and an empty control vector, syntaxin 1A vectors, or Munc18 -1 vector, and the amount of APP S in the medium and of full-length APP in the cells was examined by immunoblotting after 48 h in culture. B, HEK293 cells were cotransfected with an APP expression vector and an empty control vector, syntaxin 1A vectors, and BACE1 vector, as indicated. Both wild type (WT) and Swedish mutant (SW) APP expression vectors were used. The levels of A␤40 accumulated in the cells (top) and secreted into the medium (bottom) were determined by enzyme-linked immunosorbent assay 48 h after transfection. Transfections were performed in duplicates, and data shown are means Ϯ S.E. from a single representative experiment repeated three times. n.d., not detectable. transfected syntaxins on transactivation (Fig. 6). We found that even at high concentrations, truncated syntaxin 1A 1-243 or syntaxin 1A 1-236 had no effect on transactivation mediated by either full-length APP/Gal4-VP16 or by C99/Gal4-VP16. In contrast, all concentrations of full-length syntaxin 1A 1-288 inhibited transactivation (Fig. 6A). For both APP/Gal4-VP16 and C99/Gal4-VP16, inhibition reached a plateau at relatively low amounts of co-transfected syntaxin 1A 1-288 . In the case of the C99 C-terminal APP fragment, this plateau was at ϳ30% of wild type transactivation levels, suggesting that part of the C-terminal APP fragment continues to be cleaved even when the Golgi complex is disrupted (Fig. 6A). Immunoblotting of the transfected cells with syntaxin antibodies demonstrated that, although the various syntaxin constructs were expressed at distinct levels, high DNA concentrations of the least-expressing construct (syntaxin 1A 1-236 ) achieved higher protein levels than low DNA concentrations of the higher-expressing construct (syntaxin 1A 1-288 ) (Fig. 6B). At the same time, low DNA concentrations of the latter effected potent inhibition, whereas high DNA concentrations of the former did not (Fig. 6A), effectively ruling out expression problems as a cause for the differential activities of the various syntaxin fragments.
In a final experiment, we tested whether the observed effects of syntaxins on APP cleavage-dependent transactivation are specific for HEK293 cells or are generally applicable to other cell types. We addressed this question by comparing APP-dependent transactivation in human HEK293 and monkey COS cell lines (which lack regulated exocytosis) and in the rat neuroendocrine PC12 cell line (which shares many properties with neurons and exhibits regulated exocytosis) (Fig. 7). In all cell lines tested, full-length syntaxin 1A 1-288 potently inhibited APP-dependent transactivation, whereas other syntaxin fragments exerted, at most, small effects. This result confirms that the effects described here of syntaxin fragments on APP trafficking and processing are not dependent on cell type. DISCUSSION Our data demonstrate that APP cleavage by ␣-, ␤-, and ␥-secretases largely requires that APP traversed the Golgi complex but that it can occur at normal rates without APP ever reaching the cell surface. This conclusion was made possible by the development of new molecular tools that allow stage-specific disruption of secretory membrane traffic using closely related syntaxin 1A derivatives. Specifically, we found that expression of full-length syntaxin 1A 1-288 disrupted the Golgi complex and blocked trans-Golgi traffic, whereas truncated syntaxin 1A 1-243 left trans-Golgi traffic unimpaired but abolished all exocytosis, and the slightly shorter syntaxin 1A 1-236 was without effect (Fig. 1). The selective effect of truncated syntaxin 1A 1-243 on exocytosis is presumably caused by inter- ference with plasma membrane SNARE complexes, consistent with a general role for plasma membrane syntaxins in exocytosis, whereas the basis for the disruption of the Golgi complex by full-length syntaxin 1A (37) is unclear.
Our data documented the finding that APP is cleaved by secretases at a normal rate even when transport of APP to the cell surface is completely abolished. Comparison of APP glycosylation and cleavage (Figs. 2 and 3) with the secretion of the cleavage products APP S and A␤40 (Fig. 4) provides a striking demonstration of this finding. When exocytosis was blocked with syntaxin 1A 1-243 , APP glycosylation was apparently normal, and the production of the C-terminal cleavage products of APP was unchanged. At the same time, however, no secretion of APP S or A␤40 was detectable, and A␤40 accumulated intracellularly. Our results are consistent with several prior studies that used a variety of approaches. For example, in brain, it has been shown that fast axonal transport conveys APP to nerve terminals (17,18,20) and that A␤ is secreted from the terminals (21,22). Cleavage was demonstrated to occur en route (19), supporting the notion that it takes place post-Golgi but preplasma membrane. Similarly, our data agree with in vitro studies demonstrating cleavage of APP in vesicles derived from the trans-Golgi complex (63, 64).
Our findings on the subcellular localization of APP cleavage have several implications for the function of APP and the pathogenesis of Alzheimer's disease. 1) It has been suggested that APP may be functionally similar to Notch (7,8). Ligand binding to Notch, a cell surface receptor, stimulates Notch cleavage, which in turn releases the intracellular NICD that signals in the nucleus. The observations that both Notch and APP are cleaved by the same ␥-secretase to release an intracellular fragment and that both the NICD and the AICD can mediate transcriptional activation in the nucleus establish a strong similarity between APP and Notch. However, our conclusion that APP is cleaved prior to reaching the cell surface implies that cleavage cannot be activated by an exogenous ligand, although it could still be controlled by an endogenous ligand, as suggested by the finding that F-spondin binds to APP and inhibits APP cleavage (65). Thus at least in this respect, the biology of Notch and APP are very different. 2) A critical question in understanding the pathogenesis of Alzheimer's disease is where and at what rate A␤ is produced in neurons. A substantial proportion of APP is normally not cleaved but exposed on the cell surface as a receptor-like protein. This part of APP is endocytosed in clathrin-coated vesicles (66,67) and may subsequently be degraded in lysosomes or may be recycled to the Golgi complex and cleaved in the exocytic pathway. Our data suggest that the fundamental decision whether or not to cleave APP by secretases is made between the Golgi complex and the plasma membrane. At present, little is known about what regulators may contribute to this decision, but pinpointing its location is essential for any attempt at modulating cleavage. This is potentially important for Alzheimer's disease, because sporadic Alzheimer's disease may arise by a subtle dysregulation of cleavage over decades. 3) Full-length syntaxin 1A is relatively less able to inhibit transactivation mediated by the C99 APP fragment than transactivation mediated by fulllength APP. Conversely, the ␥-secretase blocker DAPT is a relatively better inhibitor of transactivation by the C99 fragment than of transactivation by full-length APP. This result is consistent with the idea that at least some C99 can be cleaved by ␥-secretase in the endoplasmic reticulum but that ␣/␤-cleavage requires transport through the Golgi complex. Once APP has passed through the Golgi complex, however, full-length APP may be cleaved by alternative, possibly lysosomal, pathways that do not operate on C99 (presumably because it is rapidly cleaved by ␥-secretase) and that cannot be inhibited by DAPT.
In summary, our study suggests that, although APP and Notch may both function in nuclear signaling, they are controlled by distinct mechanisms. Notch is transported uncleaved to the cell surface, where the presence or absence of exogenous ligands determines whether or not intracellular Notch signaling occurs. In contrast, the signaling function of APP appears to be determined prior to its arrival on the cell surface. The factors that regulate whether or not APP is cleaved on the way to the cell surface are unknown but could theoretically act both by engaging the extracellular sequences of APP (as an intrinsic ligand) or its cytoplasmic sequences (e.g. the multiple AICD-interacting proteins that have been characterized; see Ref. 12 for a review). Cleavage then determines whether APP acts as a cell surface receptor that is subsequently internalized and degraded or is used to produce a secreted protein (APP S ) and an intracellular signaling module (AICD) that may have a variety of extra-and intracellular downstream effects. Elucidation of the mechanisms that regulate APP cleavage and of the normal functions of APP and/or of its cleavage fragments is now a major challenge in further understanding the biology of this essential and pathologically important molecule.