Metabolism of the "Swedish" amyloid precursor protein variant in neuro2a (N2a) cells. Evidence that cleavage at the "beta-secretase" site occurs in the golgi apparatus.

The 4-kDa beta-amyloid peptide (Abeta), a principal component of parenchymal amyloid deposits in Alzheimer's disease, is derived from amyloid precursor proteins (APP). To identify potential intracellular compartments involved in Abeta production, we expressed human APP-695 (APPwt) and APP-695 harboring the Swedish double mutation (APPswe) associated with familial early-onset Alzheimer's disease, in mouse N2a cells. We demonstrate that cells expressing APPswe secrete high levels of Abeta peptides and beta-secretase-generated soluble APP derivatives (APP s beta) relative to cells expressing APPwt. In addition, we observed a concomitant diminution in the levels of alpha-secretase-generated soluble APP derivatives (APP s alpha). Our interpretation of these findings is that beta-secretase cleavage occurs in an intracellular compartment and disables those substrates which would normally be cleaved by alpha-secretase. As anticipated, the levels of APPswe are diminished relative to the steady-state levels of surface-bound APPwt; moreover, surface-bound APPswe and APPwt molecules are released from the plasma membrane after cleavage by alpha-secretase, but not by beta-secretase. Finally, by examining the rate of appearance of specific APP metabolites generated by beta-secretase, we now unequivocally demonstrate that beta-secretase cleavage of APPswe occurs within the Golgi apparatus, as early as the medial compartment.

Alzheimer's disease, a progressive neurodegenerative disorder of the elderly, is characterized by the presence of parenchymal deposits of A␤, 1 a 39-to 43-amino acid peptide derived from APP (1)(2)(3)(4). APP are integral membrane glycoproteins that mature through the secretory pathway (5). A fraction of newly synthesized APP appears on the cell surface (6 -9) and some of these molecules are cleaved by ␣-secretase (7,8) within the A␤ sequence (10 -13), resulting in the release of the APP ectodomain (APP s␣ ). In a cell-type specific manner, APP s␣ is generated in the trans-Golgi network or other late compartments of the constitutive secretory pathway (14 -17). A fraction of APP are directly sorted, or reinternalized from the cell surface, to endosomal/lysosomal compartments (7), where a complex set of A␤-containing membrane-bound fragments accumulate (7,18). Finally, it is fully established that A␤ (ϳ4 kDa) and a truncated form of A␤ (ϳ3 kDa) are released constitutively in vitro and in vivo (19 -23).
The biochemical mechanism(s) and cellular compartments involved in A␤ production have not been fully elucidated. Despite earlier excitement created by the discovery of potential amyloidogenic fragments generated in endosomal-lysosomal pathways, several lines of evidence now suggest that lysosomal degradation of APP is unlikely to contribute to the production of A␤ (reviewed in Ref. 24). However, agents that interfere with pH gradients (i.e. ammonium chloride and chloroquine) inhibit the production of A␤ (19,21), suggesting that A␤ may be generated in acidic compartments (i.e. endosomes and/or late Golgi). Indeed, biochemical studies by Koo and Squazzo (23) confirmed that A␤ production and release involves the endocytosis of full-length APP from the cell surface and subsequent recycling. In this model, ␤-secretase cleavage occurs within endocytic compartments while ␥-secretase cleavage of the residual ϳ100-amino acid membrane fragment occurs virtually simultaneously with A␤ formation and release. However, the stoichiometry of A␤ contributed by reinternalized APP to A␤ generated in the biosynthetic pathway is not known.
To identify potential intracellular compartments involved in A␤ production, we examined the metabolism of APP with a double mutation at codons 670 and 671 (of APP-770) described in two large Swedish pedigrees with familial Alzheimer's disease. Cells expressing this mutant APP secrete ϳ6 -8-fold higher levels of A␤ relative to cells expressing wild-type APP (25,26). In preceding efforts, we examined the trafficking and metabolism of APPwt and APPswe in polarized epithelial cells (MDCK) (27). We demonstrated that APPswe was cleaved at the ␤-secretase site and that the resulting soluble derivative, termed APP s␤ , was detectable in cell lysates. These results lead to the suggestion that "␤-secretase" cleavage occurs in an intracellular compartment. In the present report, we extend the observations in MDCK cells by characterizing the maturation of APPswe in transiently and stably transfected N2a cells, a mouse peripheral neuroblastoma line. We demonstrate that N2a cells expressing APP harboring the Swedish mutations secrete high levels of A␤ peptides and APP s␤ derivatives relative to cells expressing APPwt. Interestingly, we observed a concomitant diminution in the levels of secreted APP s␣ . One scenario to explain this intriguing result is that normal ␣-secretase substrates are disabled by the prior action of ␤-secretase. This model predicts that ␤-secretase must cleave APPswe in an intracellular compartment proximal to the cellular site(s) of ␣-secretase activity. We now confirm that ␤-secretase cleavage occurs within the Golgi apparatus, as early as the medial compartment.

MATERIALS AND METHODS
Plasmid Construction and Expression-The construction of expression plasmids pAPPwt and pAPPswe were described previously (27). Briefly, to generate pAPPwt, a DNA fragment encoding wild-type human APP-695 with an carboxyl-terminal epitope tag of 12 amino acids from the c-Myc oncoprotein (MEQKLISEEDLN), at the COOH terminus of APP, was subcloned downstream of a cytomegalovirus promoter in plasmid pCB6. Plasmid pAPPswe (27) encodes Myc epitope-tagged human APP-695 that harbors the Swedish FAD-specific amino acid substitutions (K595N and M596L). To generate plasmid pEF-VSVG, a 1.7-kilobase XhoI fragment encoding the VSVG protein from plasmid pJC/G (28) was subcloned downstream of the elongation factor 2␣ promoter in plasmid pEF-Bos (29).
Mouse N2a neuroblastoma cells were transfected using a high-efficiency CaPO 4 co-precipitation procedure (30). To generate cell lines that express human APPwt and APPswe, N2a cells were transfected with pAPPwt or pAPPswe and stable transfectants were selected in medium containing 0.4 mg/ml G418 (Life Technologies, Inc.) (the pCB6 vector contains a gene encoding resistance to the neomycin analog, G418). Expression of c-Myc-modified human APP was assayed by immunoblotting and immunoprecipitation with a polyclonal antiserum Myc-I raised against the synthetic peptide MEQKLISEEDLN (31).
Metabolic Labeling and Immunoprecipitation Analysis-For metabolic labeling, N2a cells were starved for 20 min in methionine-free Dulbecco's modified Eagle's medium (Life Technologies, Inc.) and then labeled with 250 Ci/ml [ 35 S]methionine (DuPont NEN) in methioninefree Dulbecco's modified Eagle's medium supplemented with 1% dialyzed fetal bovine serum (Life Technologies, Inc.) for either 3 h or 10 min. For pulse-chase analysis, parallel dishes of cells were labeled for 5 min; at the end of the labeling period, one dish of cells was washed and then lysed in immunoprecipitation buffer containing detergents and protease inhibitors (11), while the remaining dishes were washed once and incubated for various periods of time in Dulbecco's modified Eagle's medium containing 1% dialyzed fetal bovine serum and 1 mM L-methionine.
Cell-associated and soluble APP derivatives were immunoprecipitated as described previously (11). To immunoprecipitate full-length APP and APP COOH-terminal fragments, we used 369, a polyclonal antiserum raised against the entire cytoplasmic tail of APP (32), or Myc-I, a polyclonal serum raised against a 12-amino acid epitope in the c-Myc oncoprotein (see above). To immunoprecipitate full-length and soluble APP derivatives, we used a monoclonal antibody (mAb), P2-1, which recognizes an epitope in the NH 2 -terminal cysteine-rich region of human APP (33). 6E10, a mAb that recognizes residues 1-17 of human A␤ (34) was used to immunoprecipitate soluble APP generated following cleavage by ␣-secretase, while 4G8, a mAb that recognizes residues 17-28 of A␤ (34) was used to immunoprecipitate A␤ and related p3 peptides. Antibody 54 was used to immunoprecipitate the soluble APP derivative generated following cleavage of APPswe (35,36). Polyclonal anti-VSV serum (28) was used to immunoprecipitate VSVG protein.
Quantification of immunoprecipitated polypeptides was performed by phosphorimaging using Molecular Dynamics software.
A␤ Immunoprecipitation and Radiosequencing Analysis-A␤ and p3 were immunoprecipitated with 4G8, separated on 16% Tris-Tricine gels and transferred to Immobilon filters. Radiosequencing of immunoprecipitated A␤ was performed as described (19).
Cell Surface Labeling Studies-To examine the levels of cell surfacebound APP, duplicate dishes of cells were metabolically labeled with [ 35 S]methionine for 3 h. At the end of the labeling period, conditioned medium was collected and cell monolayers were rinsed three times in ice-cold phosphate-buffered saline containing 1 mM each of CaCl 2 and MgCl 2 . Cell surface proteins were biotinylated with 0.5 mg/ml NHS-SSbiotin (Pierce) in 10 mM borate buffer, pH 9.0 (37), for 45 min at 4°C. Cells were washed four times with phosphate-buffered saline contain-ing 1 mM CaCl 2 , 1 mM MgCl 2 , and 25 mM NH 4 Cl to quench unreacted biotin. Ice-cold Dulbecco's modified Eagle's medium supplemented with 1% fetal bovine serum was added to one set of cells and kept on ice; prewarmed medium was added to the other set of cells and placed at 37°C for 10 min. After 10 min, the medium from cells held at 4°C or at 37°C were collected and detergent-soluble extracts were prepared from cell monolayers after lysis in immunoprecipitation buffer.
To assess the level of cell surface-bound APP, radiolabeled cell lysates were subject to immunoprecipitation with 369 antiserum. The immunoprecipitates were boiled in immunoprecipitation buffer containing 1% SDS to dissociate APP-related polypeptides from the antibody. The resulting soluble fraction was subsequently incubated with streptavidin-agarose beads (Pierce) to recover biotinylated APP, as described previously (27). To examine the fraction of soluble derivatives which were biotinylated, i.e. that were released from the cell surface, equivalent aliquots of medium collected from cells incubated at 37°C following biotinylation was incubated with either mAb 22C11, an antibody that recognizes a region between residues 66 and 81 of APP (6,38) or mAb 6E10 (see above). Immunoprecipitated proteins were dissociated by boiling in immunoprecipitation buffer containing 1% SDS. The resultant soluble fraction was adjusted to 1 ϫ immunoprecipitation buffer and subjected to a second round of binding to streptavidin-agarose beads.
Endoglycosidase H Digestion-[ 35 S]Methionine-labeled APP was immunoprecipitated from detergent soluble lysates using Myc-I antiserum. The immune complexes were boiled for 3 min in a buffer containing 50 mM Tris-HCl, pH 7.6, 1% SDS, and 0.2% ␤-mercaptoethanol to dissociate APP-related proteins from the antibody and sodium citrate was added to a final concentration of 0.2 M. The samples were split into two 50-l aliquots and incubated in the presence or absence of 2 milliunits of endoglycosidase H (Boehringer Mannheim) at 37°C for 16 h.

Expression and Metabolism of Human APP and the "Swedish" Variant in N2a
Cells-Expression constructs encoding c-Myc tagged APPwt or APPswe (27) were transiently transfected into N2a cells and recipient cells were labeled continuously with [ 35 S]methionine for 3 h. APP-related species were immunoprecipitated from detergent lysates using an antibody, 369, raised against the entire cytoplasmic domain of APP. Essentially identical levels of accumulated "immature" ϳ105-kDa and "mature" ϳ120-kDa species were detected in lysates of cells expressing APPwt of APPswe (Fig. 1A, compare lanes 2 and 3), which were clearly overexpressed over endogenous APP species (Fig. 1A, lane 1). As expected, ϳ100-kDa soluble APP-related species (APP s ) were immunoprecipitable from conditioned medium of cells expressing APPwt using antibody P2-1, specific for an epitope in the human APP amino terminus (Fig. 1A, lane 5). On the other hand, we consistently immunoprecipitated elevated levels of soluble APP derivatives from conditioned medium of cells overexpressing APPswe (Fig.  1B, lane 6) relative to derivatives in medium of cells expressing APPwt. Moreover, the vast majority of the soluble derivatives generated from APPswe migrated with somewhat accelerated electrophoretic migration relative to the ϳ100-kDa soluble form of APPwt, suggesting that APPswe was cleaved upstream of the ␣-secretase site. The latter result is consistent with our earlier findings in MDCK cells in which we demonstrated that APP s␤ species derived from APPswe failed to be recognized by antibodies specific for amino-terminal epitopes of the A␤ sequence (27). To assess the levels of APP s␣ and APP s␤ , we immunoprecipitated APP s with an antibody, 6E10, which recognizes an epitope between residues 1 and 17 of A␤. As anticipated, 6E10 immunoprecipitated ϳ100-kDa APP s␣ from conditioned medium of cells expressing either APPwt or APPswe (Fig. 1A, lanes 8 and 9, respectively). However, and in sharp contrast to the results obtained with antibody P2-1, we consistently observed that the accumulated level of APP s␣ was considerably lower in medium from cells expressing APPswe relative to cells expressing APPwt (Fig. 1A, compare lanes 9 and 8, respectively). Thus, we argue that the vast majority of solu-ble derivatives derived from APPswe which are detected by antibody P2-1 (Fig. 1A, lane 6) are likely generated following cleavage upstream of the ␣-secretase site, and most likely at the ␤-secretase site (i.e. at the amino terminus of A␤). These latter observations are not the result of overloading of trafficking pathways in transient transfection assays as similar results have been obtained following transfection of as little as 250 ng of expression plasmids (not shown) and by analysis of stably transfected N2a cells that constitutively express APPwt or APPswe (see below). Finally, we demonstrate that cells Finally, and to confirm that endoproteolytic cleavage by ␤-secretase occurs between Leu 596 and Asp 587 of APPswe, we characterized the carboxyl terminus of the secreted soluble derivative and the amino terminus of A␤. To confirm the authenticity of the ␤-secretase-generated soluble derivative, we subjected parallel aliquots of conditioned medium shown on  Fig. 1A, lanes 4 -6, to immunoprecipitation analysis with antibody 54, specific for APP s␤ (35,36). This antibody fails to detect ␣-secretase-cleaved APPswe (35,36), and hence is specific for an epitope of the APPswe soluble derivative which is "exposed" upon truncation of the full-length precursor by ␤-secretase. As anticipated, antibody 54 recovered high levels of ϳ95-kDa species from medium of cells expressing APPswe (Fig. 1B, lane 3); this result is in sharp contrast to the isolation of nonspecific polypeptides of ϳ85 and ϳ95 kDa from medium of cells transfected with empty vector or wtAPP cDNA (Fig. 1B, lanes 1 and  2, respectively). To analyze the amino terminus of A␤ derived from cells expressing APPwt and APPswe, we radiolabeled cells with [ 3 H]phenylalanine and immunoprecipitated A␤ and p3 from the conditioned medium; the isolated 4-kDa band was subjected to radiosequencing. Predominant peaks of radioactivity in cycles 4, 19, and 20 clearly show that Ͼ95% of the 4-kDa peptides secreted by cells expressing either APPwt and APPswe initiate at Asp 1 (Fig. 1B, panels APPwt and APPswe). Taken together with studies demonstrating that the membrane-bound fragment generated by ␤-secretase initiates at Asp 1 (39), our demonstration that the carboxyl terminus of the soluble derivative extends to Leu 596 fully confirms our view that ␤-secretase cleaves APPswe between Leu 596 and Asp 597 .

␤-Secretase Cleavage of Swedish APP Variant
In order to further characterize intracellular trafficking and processing of APPwt and APPswe APP in N2a cells, we generated 20 and 18 independent stably transfected cell lines which expressed variable levels of APPwt or APPswe, respectively. We did not observe any toxicity-related phenotypes in the N2a lines expressing either APPwt or APPswe. Biosynthetic rates of c-Myc epitope-tagged transgene-derived polypeptides in individual cell lines were assessed by short (3 min) pulse-labeling with [ 35 S]methionine and immunoprecipitation analysis with Myc-1 antibody. Two cell lines, 695.12 and Swe.1, that express essentially indistinguishable levels of newly synthesized AP-Pwt and APPswe, respectively (Fig. 1C, compare lanes 1 and 2, respectively), were selected for further analysis. Immunoprecipitation of APP-related species using APP NH 2 -terminal antibody, P2-1, from lysates of 695.12 and Swe.1 cells labeled for 3 h, revealed the presence of accumulated ϳ105and ϳ120-kDa forms (Fig. 1C, lanes 3 and 4). Interestingly, we also detected a species with an apparent molecular mass of ϳ95 kDa in lysates of cell line Swe.1 (Fig. 1C, lane 4), similar to a species observed in lysates of MDCK that constitutively express APPswe (27). Analysis of soluble APP derivatives in radiolabeled conditioned medium of lines 695.12 and Swe.1 (Fig. 1C, lanes 5-8) fully supported our findings in transient assays. For example, relative to the level of soluble derivatives in 695.12 cell medium (Fig. 1C, lane 5) the total level of soluble derivatives in Swe.1 cell medium (Fig. 1C, lane 6) was clearly elevated; by phosphorimaging analysis, the total level of secreted derivatives generated by line Swe.1 was increased ϳ3-fold relative to line 695.12, and the vast majority of these species exhibited accelerated electrophoretic mobility. Moreover, the levels of APP s␣ in Swe.1 cell medium was clearly diminished relative to the level of APP s␣ in 695.12 cell medium (Fig. 1C, compare lanes 8  and 7, respectively); phosphorimaging analysis showed ϳ3-fold higher levels of APP s␣ in 695.12 cell medium compared to Swe.1 cell medium. Hence, we argue that the principal contributor to the elevated level of accumulated derivatives secreted by Swe.1 cells are species which are generated following cleavage at the ␤-secretase site. Finally, a 4-fold elevation of accumulated ϳ4-kDa A␤ was evident in Swe.1 cell medium relative to 695.12 cell medium. However, and in contrast to our findings in transiently transfected N2a cells, a corresponding diminution in the levels of the p3 fragment was not observed. At present, there is not a satisfactory explanation for this phe-nomenon. Earlier studies have suggested that the penultimate precursor for p3 is the membrane-retained fragment generated by ␣-secretase (24). It is also conceivable that a fraction of ␤-secretase-generated, carboxyl-terminal derivatives are substrates for cleavage by ␣-secretase and that this truncated fragment is ultimately cleaved by ␥-secretase to generate p3. Thus, we would propose that the trafficking and proteolysis of the ␤-secretase-generated fragment is intrinsically different in stable and transiently transfected N2a cells, and may reflect differences in expression levels of transgene-encoded APPswe or the corresponding ␤-secretase-generated carboxyl-terminal fragment. Despite this inconsistency in p3 levels between stable and transiently transfected N2a cells, these data do not detract from our principal findings that the total levels of secreted derivatives generated by the stable APPswe cell line are increased relative to cells expressing APPwt and that over 80% of these species exhibited accelerated electrophoretic mobility. Moreover, concomitant dimunition in APP s␣ levels in medium of the stable cell line expressing APPswe strongly supports our view that potential ␣-secretase substrates are disabled by the prior action of ␤-secretase.
In the aggregate, we have demonstrated that in transient and stably transfected N2a cells, the levels of soluble derivatives and A␤ in medium of cells expressing APPswe are elevated relative to respective species secreted by cells expressing APPwt. Moreover, the vast majority of soluble derivatives in medium of cells expressing APPswe are APP s␤ , molecules generated following cleavage at the ␤-secretase site. The production of elevated levels of APP s␤ from cells expressing APPswe is essentially identical to results obtained by expression of chimeric human placental alkaline phosphatase-APP molecules harboring the Swedish substitutions in H4 glioblastoma cells. These studies demonstrated that human placental alkaline phosphatase-APPswe were sensitive to endoproteolytic cleavage at the ␤-secretase site and that high levels of chimeric human placental alkaline phosphatase-APP s␤ species were secreted into the conditioned medium (40). More importantly, we document that the total level of APP s␣ derivatives is diminished in medium of cells expressing APPswe. This result can be accommodated by a model in which a significant population of substrates which might normally be subject to cleavage by ␣-secretase are disabled by the prior action of ␤-secretase. Implicit in this model is that ␤-secretase must cleave APPswe substrates in an intracellular compartment proximal to the cellular site(s) of ␣-secretase activity. Thus, we argue that APPswe is obligatorily cleaved by ␤-secretase in an intracellular compartment wherein ␣-secretase is either inactive or absent.
Kinetics of ␣and ␤-Secretase Processing of APPwt and APPswe APP in N2a Cells-If our interpretation that ␤-secretase cleavage of APPswe occurred during transit through the secretory pathway is correct, then we anticipated recovery of a specific membrane-bound, carboxyl-terminal fragment containing A␤ sequences in cell lysates prior to appearance of an ␣-secretase-generated carboxyl-terminal fragment. We prepared detergent lysates from N2a cells transiently transfected with cDNA encoding Myc-tagged APPwt or APPswe and metabolically labeled with [ 35 S]methionine for 3 h. Lysates were subjected to immunoprecipitation with Myc-1 or 369 antibodies, specific for epitopes in the carboxyl terminus of transgenederived polypeptides. Both antibodies recovered similar steadystate levels of a common ϳ12-kDa peptide from either APPwt ( Fig. 2A, lanes 2 and 5) or APPswe ( Fig. 2A, lanes 3 and 6) cell lysates and a specific fragment of ϳ13.5 kDa in lysates of cells expressing APPswe (Fig. 2, lanes 3 and 6). Since the half-life of the ␣-secretase generated carboxyl-terminal fragment of ϳ12 ␤-Secretase Cleavage of Swedish APP Variant kDa has not been determined, it is not possible to interpret the significance of the finding that cells expressing APPwt and APPswe accumulate similar levels of this fragment. Thus, the accumulation of this fragment is silent with respect to its rate of production. Moreover, this fragment does not necessarily derive from ␣-secretase processing of the precursor; for example, high levels of peptides that migrate with the bona fide ␣-secretase-generated carboxyl-terminal fragment are apparent in lysates of cells which express APP with a lysine to valine mutation at position 612 (8). In this case, APP s␣ secretion is diminished by over 80%. 2 Hence, we argue that proteolysis of APP in endosomal/lysosomal compartments may also contribute to accumulated ϳ12-kDa species. On the other hand, the ϳ13.5-kDa fragment is likely generated following cleavage at the ␤-secretase site, and is analogous to the carboxyl-terminal 100-amino acid fragment that is enriched in human M17 cells expressing APP harboring the Swedish mutations (39).
To evaluate the kinetics of cleavage of APPwt and APPswe at the ␣or ␤secretase sites specifically, we performed pulsechase analyses. Parallel dishes of N2a cells were transiently transfected with cDNAs expressing APPwt or APPswe. Cell monolayers were pulse-labeled with [ 35 S]methionine for 10 min, then chased at 37°C for varying periods of time up to 45 min. At the end of each chase period, medium was saved and detergent soluble lysates were prepared from cell monolayers. Lysates or conditioned medium were subjected to immunoprecipitation with Ab369 or P2-1 antibodies, respectively. Fig. 2, B and C, depicts the production of APP carboxyl-terminal fragments in cell lysates and the appearance of APP s in the medium, respectively. Analysis of the carboxyl-terminal fragments generated in cells expressing APPwt revealed the nearly concomitant generation of both ϳ12and ϳ13.5-kDa fragments at 20 min into the chase period, species which are continually generated (or accumulate) over the next 25 min (Fig. 2B, left  panel). We suggest that the ϳ13.5-kDa fragment is the penul-timate precursor of A␤ and that the kinetics of appearance is consistent with previous studies which demonstrated A␤ generation from plasma membrane-bound APP after endosomal recycling (23). In parallel, APP s first appear in the conditioned medium of cells expressing APPwt APP at 30 min into the chase period (Fig. 2C, left panel). Taken together with the kinetics of appearance of the ϳ12-kDa carboxyl-terminal fragment, we argue that the APP s species are produced by endoproteolytic cleavage by ␣-secretase, an activity which occurs late in the secretory pathway, and predominantly on the plasma membrane (see below, Fig. 3). In striking contrast to the results obtained in cells expressing APPwt, cells expressing 2 S. Sisodia, T. Golde, and S. Younkin, unpublished observations.

FIG. 2. Kinetics of APP processing in N2a cells.
A, analysis of APP carboxyl-terminal fragments generated by proteolytic processing. Detergent lysates of transiently transfected N2a cells (Fig. 1A) were analyzed by immunoprecipitation with either Myc-I or 369 antiserum. B and C, kinetics of cleavage and secretion of APP molecules. Parallel dishes of N2a cells transfected with plasmids encoding APPwt (Wt) or APPswe (Swe) were pulse-labeled for 10 min with [ 35 S]methionine and chased for the times indicated. At each time point, medium was collected and lysates were prepared from cell monolayer. Carboxyl-terminal APP fragments were immunoprecipitated from lysates with 369 antiserum. Secreted APP s molecules were immunoprecipitated from conditioned medium using mAb P2-1. The ϳ13.5-kDa fragment whose levels were elevated in cells expressing APPswe is marked by an arrow. The ϳ12-kDa band (marked by an arrowhead) likely represents the carboxyl-terminal fragment generated after ␣-secretase cleavage of APP.

␤-Secretase Cleavage of Swedish APP Variant
APPswe contain a prominent ϳ13.5-kDa fragment which first appears within 10 min into the chase period (Fig. 2B, right  panel). Notably, a ϳ12-kDa carboxyl-terminal fragment is also present, albeit at somewhat lower levels, in cells expressing APPswe, and the kinetics of appearance of this species fragment is indistinguishable to that observed in cells expressing APPwt. Interestingly, APP s first appear in the conditioned medium within 20 min of chase (Fig. 2B, right panel), a rate considerably accelerated relative to the secretion of APP s␣ species from cells expressing APPwt. Moreover, the soluble derivatives in medium of cells expressing APPswe clearly exhibit accelerated migration relative to APP s␣ . Our interpretation of this result is that these APP s␤ species which are produced early in the secretory pathway are liberated from specific intracellular sorting/retention signals contained within the core protein and hence rapidly transit through the secretory pathway prior to secretion. Thus, the early production of the ϳ13.5-kDa carboxyl-terminal fragment and accelerated appearance of secreted APP s␤ supports our view that ␤-secretase cleavage of APPswe is initiated early in the secretory pathway. Interestingly, we have failed to recover APP carboxyl-terminal fragment(s) that encompasses sequences between Ala 637 (in APP-695) and the myc-epitope tag, a species which would be generated by the action of ␥-secretase (data not shown). The reason for this discrepancy is unclear but suggests that ␥-secretase-generated 56 -59 amino acid fragment is subject to rapid degradation.
␤-Secretase Cleavage Does Not Occur on the Plasma Membrane-To assess whether ␤-secretase could exert its activity at the plasma membrane, we performed a cell surface biotinylation and release assay (9,27). N2a cells transiently transfected with cDNA encoding APPwt or APPswe APP were labeled continuously for 3 h with [ 35 S]methionine. As we observed earlier (Fig. 1A), elevated levels of soluble APP derivatives accumulated in medium from cells expressing APPswe (Fig. 3,  lane 2). Cells were subsequently washed, and molecules residing on the plasma membrane were biotinylated at 4°C with NHS-SS-biotin. Surface-modified cells were either held at 4°C or placed at 37°C for 10 min. To assess the level of surfacebound transgene-encoded polypeptides, detergent-soluble ly-sates were subject to immunoprecipitation with APP COOHterminal antibody, Ab369, and the recovered material was subject to a second round of purification with immobilized streptavidin. We reproducibly demonstrated that the level of surface-resident full-length APPwt (Fig. 3, lane 3) is elevated relative to full-length surface-bound APPswe (Fig. 3, lane 5), a result that would be anticipated if a fraction of APPswe are cleaved by ␤-secretase prior to appearance on the plasma membrane. After shifting cells to 37°C for 10 min, we also observed a slight diminution in the levels of biotinylated APPwt or APPswe (Fig. 3, lanes 4 and 6, respectively) which represents both surface-bound or internalized full-length species. Concomitantly, we detected the appearance of biotinylated APP s in the conditioned medium of cells expressing APPwt or APPswe (Fig.  3, lanes 7 and 8, respectively) using the APP NH 2 -terminal mAb 22C11. Although phosphorimaging analysis revealed that the absolute level of biotinylated APP s released from cells expressing APPswe was ϳ70% of the level of biotinylated APP s released from cells expressing APPwt APP, this result is fully expected given the difference in steady-state levels of surfaceresident APPwt and APPswe (Fig. 3, compare lanes 3 and 5,  respectively). Furthermore, the electrophoretic migration of biotinylated APP s generated from cells expressing APPwt or APPswe was indistinguishable. To discriminate between cleavage of cell surface molecules at the ␣or ␤-secretase sites, we immunopurified biotinylated APP s from medium of cells expressing APPwt or APPswe with A␤ NH 2 -terminal-specific antibody, 6E10 (Fig. 3, lanes 9 and 10, respectively). Phosphorimaging analysis revealed that the level of released biotinylated APP s from cells expressing APPswe was ϳ60% of the level of released biotinylated APP s from cells expressing APPwt, a result which parallels the result with the APP-specific antibody, 22C11. Although the fraction of surface-labeled APPwt and APPswe APP which are cleaved by ␣-secretase is not fully established, the relative levels of total (i.e. 22C11-immunoprecipitable) or 6E10-immunoreactive APP s forms are essentially identical irrespective of the transgene, leading us to argue that both APPwt and APPswe substrates are equally susceptible to ␣-secretase cleavage. In summary, our results provide strong support for the notion that biotinylated APP s released from the surface of cells expressing APPswe are generated following endoproteolytic cleavage by ␣-secretase, not ␤-secretase.
␤-Secretase Cleavage of APPswe Is Initiated in the Medial Golgi Compartment-In order to identify potential intracellular compartments in which ␤-secretase exerts its activity, we examined the rate of appearance of specific APP carboxylterminal fragments and the extent of oligosaccharide modification of APP or other glycoproteins. For these analyses, we transiently transfected cDNA encoding the G protein of the vescicular stomatitis virus (VSVG) (28) into parallel dishes of N2a cells stably transfected with cDNA encoding APPswe (line Swe.24). Cells were pulse-labeled for 5 min, then chased for various periods of time at 37°C. Detergent soluble lysates prepared at each time point were subject to immunoprecipitation analysis with APP-specific COOH-terminal antibody Ab369 (Fig. 3A). Clearly, and consistent with our data from transiently transfected N2a cells (Fig. 2B), a ϳ13.5-kDa fragment appeared within 12.5 min into the chase period in Swe.24 cells. In parallel, we assessed the rate of appearance of endoglycosidase H-resistant, N-linked mixed oligosaccharide-modified forms of APP. Full-length APPswe were immunoprecipitated from lysates prepared at selected time points, and one-half of the recovered material was reacted with endoglycosidase H. Studies of a variety of glycoproteins in a host of mammalian cells have established that the generation of endoglycosidase H-resistant complex-type N-linked oligosaccharides occurs in the medial Golgi compartment (41). As expected, newly synthesized ϳ105-kDa APP was sensitive to digestion by endoglycosidase H (Fig. 4C, t ϭ 0 min), similar to observations in transfected CHO cells (42); the residual ϳ100-kDa species represents APPswe lacking high mannose oligosaccharides. However, within 7.5-12.5 min into the chase period, an ϳ115-kDa APP form appeared that was resistant to digestion with endoglycosidase H. The ϳ115-kDa form likely represents APPswe with complex-type N-linked oligosaccharides. Thus, in view of the similar rates of appearance of the ϳ13.5-kDa carboxyl-terminal fragment and ϳ115-kDa complex-type oligosac-charide modified forms of APP, we suggest that ␤-secretase cleavage is initiated in the Golgi apparatus, perhaps as early as the medial compartment. However, since the bulk of immature ϳ105-kDa APP is not converted to the mature ϳ115-kDa species in N2a cells, it is not presently certain that the ϳ13.5-kDa carboxyl-terminal fragment is uniquely generated from the ϳ115-kDa precursor. In any event, we derived additional support for our view that ␤-secretase cleavage of APPswe occurs in the Golgi compartment by examining the maturation of the VSVG protein using the same detergent-soluble lysates used for the analysis shown in Fig. 4, A and B. Newly synthesized VSVG of ϳ65 kDa (0 min) matures to ϳ67 kDa within 12.5-15 min (Fig. 4C), resulting from the enzymatic conversion of high mannose oligosaccharides to complex-type oligosaccharides (43,44); we have confirmed that the ϳ67-kDa species is resistant to digestion by endoglycosidase H (data not shown). While it is highly conceivable that the rates of maturation and/or trafficking of VSVG protein and APP in N2a cells are different, the nearly concomitant appearance of the ϳ13.5-kDa carboxylterminal fragment (Fig. 4A), the mature ϳ115-kDa APP species (Fig. 4B), and mature ϳ67-kDa VSVG protein (Fig. 4C) provides strong support for our view that ␤-secretase cleavage of APPswe APP is initiated in the Golgi apparatus. DISCUSSION A␤, the principal component of parenchymal amyloid deposits in Alzheimer's disease, is derived from integral membrane glycoproteins, APP. Although A␤ is normally secreted by a variety of cultured cells, the molecular mechanisms involved in A␤ production have not been fully clarified. The present report provides several insights into the cellular compartments involved in ␤-secretase cleavage and the production of A␤. We have assessed the metabolism of a APP (APPswe) harboring a double mutation at codons 670 and 671 (of APP-770) in N2a cells. Using surface biotinylation and release approaches, we demonstrate that relative to the steady-state levels of surfacebound wild-type APP, the levels of APPswe are diminished. The middle panel depicts selected cellular compartments of the central vacuolar pathway involved in APP trafficking/processing. For APPwt, the majority of secreted APP s␣ are generated by cleavage of full-length molecules by ␣-secretase on the plasma membrane. Low levels of secreted A␤ presumably occurs via endocytic recycling of plasma membrane-bound APP (23). For APPswe, ␤-secretase cleaves full-length membrane bound molecules in the medial Golgi and compartments proximal to the plasma membrane. Resulting soluble APP s␤ traffic through the secretory pathway and are released into the medium. As a consequence of early cleavage by ␤-secretase, the levels of full-length APPswe that arrive on the plasma membrane is considerably diminished. Finally, the membrane-bound, ␤-secretase-generated 100-amino acid carboxyl-terminal fragment is cleaved by ␥-secretase at, or near the plasma membrane to release high levels of soluble A␤. Asterisks indicate the position of the Swedish mutation in APPswe; solid box, A␤ domain.

␤-Secretase Cleavage of Swedish APP Variant
Moreover, surface-bound APPswe and APPwt molecules are released from the plasma membrane after cleavage by ␣-secretase, but not by ␤-secretase. Finally, using kinetic approaches, we provide compelling evidence that ␤-secretase cleavage of APPswe occurs early in the secretory pathway and unequivocally demonstrate that a population of APPswe molecules are cleaved at the ␤-secretase site within the Golgi apparatus, as early as the medial compartment. Moreover, and consistent with the cell surface labeling and release studies, cleavage of both the APPwt and APPswe substrates occurs several minutes later, concomitant with the appearance of soluble APP s␣ derivatives in the conditioned medium. The principal conclusions of this work are summarized in Fig. 5.
Our finding that ␤-secretase cleavage of APP harboring the Swedish mutations is initiated in the Golgi apparatus provides a conceptual framework for developing a model to explain the increase in A␤ secretion by APPswe cells. We propose that the production of the membrane-bound ϳ13.5-kDa A␤-containing carboxyl-terminal fragment early in the secretory pathway liberates the fragment from potential sorting/retention signals in the APP lumenal domain. Support for a role for APP lumenal sequences in intracellular trafficking has recently emerged from studies of soluble APP molecules expressed in MDCK cells (45,46). Thus, in sharp contrast to APP or APP with a deletion of the cytoplasmic sequence, YENPTY, in which only ϳ20 or ϳ60%, respectively, of newly synthesized molecules reach the cell surface and are secreted (9), we suggest that the ϳ13.5-kDa fragment is efficiently shunted through the secretory pathway. These molecules would subsequently encounter ␥-secretase near, or at, the plasma membrane immediately prior to A␤ release.