Multiple Effects of Aspartate Mutant Presenilin 1 on the Processing and Trafficking of Amyloid Precursor Protein*

PS1 deficiency and expression of PS1 with substitutions of two conserved transmembrane aspartate residues (“PS1 aspartate variants”) leads to the reduction of Aβ peptide secretion and the accumulation of amyloid precursor protein (APP) C-terminal fragments. To define the nature of the “dominant negative” effect of the PS1 aspartate variants, we stably expressed PS1 harboring aspartate to alanine substitutions at codons 257 (D257A) or 385 (D385A), singly or in combination (D257A/D385A), in mouse neuroblastoma, N2a cells. Expression of the PS1 aspartate variants resulted in marked accumulation of intracellular and cell surface APP C-terminal fragments. While expression of the D385A PS1 variant reduced the levels of secreted Aβ peptides, we now show that neither the PS1 D257A nor D257A/D385A variants impair Aβ production. Surprisingly, the stability of both immature and mature forms of APP is dramatically elevated in cells expressing PS1 aspartate variants, commensurate with an increase in the cell surface levels of APP. These findings lead us to conclude that the stability and trafficking of APP can be profoundly modulated by coexpression of PS1 with mutations at aspartate 257 and aspartate 385.

On the other hand, the identity of the ␥-secretase has been an enigma. Several lines of evidence, however, have lent strong support for the view that the PS1 and PS2 play an essential role in ␥-secretase processing of APP and the signaling receptor, Notch 1. First, cells with genetic ablations of PS1 and PS2 do not exhibit ␥-secretase activity (20,21); A␤ secretion is completely abolished, and intramembraneous processing of Notch 1 is fully abrogated (22)(23)(24), although very recent reports have shown that A␤ derived from endogenous APP appears not to be altered by PS deficiency (25) and that the generation of N-terminally truncated forms of A␤ (A␤-(2-42)) are not dependent on PS1 expression (26). Interestingly, cells expressing PS1 or PS2 with amino acid substitutions of two conserved aspartate residues within predicted transmembrane domains 6 and 7 largely mimic the PS-null state with respect to ␥-secretase cleavage of APP and Notch 1 (24,(27)(28)(29)(30). These data have been interpreted to suggest that PS are unusual diaspartyl proteases that catalyze intramembraneous proteolysis. Further support for the "PS are ␥-secretases" model emerged from membrane fractionation studies by Li et al. (10) showing that detergent-solubilized ␥-secretase activity coelutes with PS1 and that this activity is depleted by an anti-PS1 antibody. These data and the demonstration that transition state analogue inhibitors of ␥-secretase can specifically label PS1 and PS2 N-terminal fragment and CTF (31)(32)(33)(34) strongly suggest that PS are ␥-secretases. In this regard, Steiner et al. (35) have noted sequence similarities between a PS sequence that includes the "critical" aspartate 385 residue and a domain of the polytopic membrane proteases, termed bacterial prepilin peptidases (36); the aspartate residue in bacterial prepilin peptidases and PS are embedded within the sequences GXGD(F/L/I) and GLGDF, respectively.
Despite the strengths of these conclusions, cell biological studies have revealed that the subcellular distributions of PS and ␥-secretase activity are largely discordant. Although several reports have revealed that PS1 are resident in late compartments, including endosomes and plasma membranes in certain cell types (24,37,38), other reports have documented that PS1 and PS2 are localized predominantly in early compartments of the secretory apparatus (endoplasmic reticulum or endoplasmic reticulum to Golgi intermediate compartment) (39 -44). In contrast, the activities responsible for intramembraneous processing of APP and Notch occur in distal compartments including the Golgi, plasma membrane, and endosomes (45)(46)(47)(48)(49)(50). These observations have led to the suggestion that PS1 plays an indirect role in modulating ␥-secretase activity, perhaps by affecting folding or trafficking of ␥-secretase or its substrates. Supporting this notion are the observation that trafficking of APP, the tyrosine kinase receptor, TrkB, and the PS1-interacting protein, ICAM-5/telencephalin, are markedly affected in PS1-deficient neurons (51,52).
To examine the role of PS1 aspartate variants in ␥-secretase processing of APP, we stably expressed PS1 harboring aspartate to alanine substitutions at codon 257 (D257A) or 385 (D385A), singly or in combination (D257A/D385A) in mouse neuroblastoma, N2a cells. While expression of the D385A PS1 variant reduced the levels of secreted A␤ peptides, we now show that neither the PS1 D257A nor D257A/D385A variants impair A␤ production. Surprisingly, the stability of both immature and mature forms of APP is dramatically elevated in cells expressing PS1 aspartate variants, commensurate with an increase in the cell surface levels of APP. These studies reveal an unpredicted activity of presenilin aspartate variants in the stability and trafficking of APP.
Western Blot-Cells were lysed in immunoprecipitation buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 0.5% Nonidet P-40, and 0.5% sodium deoxycholate) supplemented with protease inhibitor mixture (a mixture of 4-(2aminoethyl)-benzenesulfonyl fluoride, pepstatin A, E-64, bestain, leupeptin, and aprotinin; Sigma). Conditioned medium was collected and immediately adjusted to 0.5 mM phenylmethylsulfonyl fluoride. Cell lysates or conditioned medium was subject to Tris/glycine or 16.5% Tris/ Tricine SDS-PAGE (57), and fractionated proteins were transferred to nitrocellulose membrane (Schleicher and Schuell). The membranes were blocked by incubation with 10% nonfat dry milk and 0.1% bovine serum albumin in phosphate-buffered saline containing 0.15% Tween 20 (PBS-T) and probed with various primary antibodies. After incubation for 2 h with gentle agitation at room temperature, the blots were reacted with either horseradish peroxidase-coupled goat anti-rabbit or anti-mouse IgG secondary antibodies (Pierce) for 1 h. Between steps, the blots were extensively washed with PBS-T for 30 min. The immunoreactive bands were visualized using an enhanced chemiluminescence detection (ECL) system (PerkinElmer Life Sciences).
Metabolic Labeling, Pulse-Chase, and Immunoprecipitation-Cells were labeled with 200 Ci/ml [ 35 S]methionine (PerkinElmer Life Sciences) in methionine-free DMEM supplemented with 1% dialyzed fetal bovine serum (Life Technologies) for 1 or 4 h. Conditioned medium was collected and adjusted to 0.5 mM phenylmethylsulfonyl fluoride. Labeled cells were harvested as described above. Immunoprecipitation was carried out with respective antibodies at 4°C overnight. The immune complexes were collected with protein A-conjugated agarose beads (Pierce) and eluted by boiling for 5 min in Laemmli SDS sample buffer.
For pulse-chase experiments, N2a cells were starved for 30 min in methionine-free DMEM and then labeled with 200 Ci/ml [ 35 S]methionine for 10 min. At the end of labeling period, one dish of each cell line was harvested in immunoprecipitation buffer, as described above, while the remaining dishes were washed once, and incubated for various periods of time in DMEM containing 1% dialyzed fetal bovine serum and 0.5 mM L-methionine (Life Technologies). At each time point, the conditioned medium was collected, and the cells were lysed in immunoprecipitation buffer and immunoprecipitated, as described above. After separation on SDS-PAGE, radioactive bands were visualized and quantified by phosphorimaging (PhosphorImager; Molecular Dynamics, Inc., Sunnyvale, CA).
Surface Biotinylation-Cells were grown to near confluence in a 10-cm tissue culture dish, washed twice with ice-cold PBS-CM (PBS containing 1 mM CaCl 2 and 1 mM MgSO 4 ). The cells were incubated in 2 ml of ice-cold 0.5 mg/ml sulfosuccinimidobiotin (Pierce) in a borate buffer solution (10 mM sodium borate, pH 9.0, 154 mM NaCl, 12 mM KCl, 2.25 mM CaCl 2 ) for a total of 30 min. The biotin solution was changed twice during the 30-min incubation. The cells were then washed four times with PBS-CM containing 50 mM NH 4 Cl to quench any unconjugated biotin. The cells were then lysed in immunoprecipitation buffer containing protease inhibitors. Each lysate was incubated with 50 l of streptavidin-agarose beads (Pierce) at 4°C, and captured proteins were eluted by boiling in Laemmli SDS sample buffer.
Laser Confocal Immunofluorescence Microscopy-Cells grown on poly-L-lysine-coated glass coverslips were fixed in methanol for 10 min at Ϫ20°C and washed with PBS. Fixed cells were blocked with 5% BSA in PBS and then incubated with primary antibodies in PBS containing 1% BSA at 4°C overnight. After incubation with primary antibodies, double staining was carried out with FITC-labeled anti-mouse and Texas Redlabeled anti-rabbit secondary antibodies (Vector Laboratories, Burlingame, CA). After washing with PBS, cells were mounted onto glass slides with Aqua Poly/Mount (Polysciences Inc., Warrington, PA). Samples were examined under an inverted microscope using a ϫ 100 lens, and confocal images were acquired on a confocal laser-scanning microscope. For lectin staining, FITC-conjugated Vicia villosa agglutinin (VVA; Vector Laboratories; 10 g/ml) were directly added during the secondary antibody incubation. To label endosomal/lysosomal compartments, N2a cells were incubated with DMEM containing 1 mg/ml FITC-BSA (Sigma) for 15 min at 37°C. The plates were washed in ice-cold PBS and incubated in complete medium at 37°C for 90 min. After a 90-min chase, cells were labeled with affinity-purified ␣PS1Loop antibody followed by Texas Redconjugated anti-rabbit secondary antibody.
For live cell surface staining, N2a cells stably expressing wild type or aspartate mutant PS1 were incubated with either 3D6 or P2-1 antibodies or FITC-VVA at 10°C for 45 min without fixation, essentially as described (58). After fixing the cells with 4% paraformaldehyde, cells were incubated with secondary antibody at room temperature for 1 h. Laser confocal images were acquired under identical settings for direct comparison between cell lines.
Electron Microscopy-For electron microscopy, N2a cells expressing D385A PS1 mutant were fixed for 30 min with 4% paraformaldehyde and 0.1% glutaraldehyde in 0.1 M phosphate buffer, pH 7.3. After several rinses in PBS with 0.05% saponin, cells were blocked for 1 h with incubation buffer (PBS with 0.05% saponin, 2% normal goat serum, 1% BSA) and incubated overnight at 4°C with 3D6 antibody (2 g/ml). Following incubation with biotinylated goat anti-mouse IgG (Vector Laboratories) and avidin-biotinylated horseradish peroxidase complex (Vectastain Elite; Vector Laboratories), immunoreactivity was visualized with 0.05% 3,3Ј-diaminobenzidine tetrahydrochloride and 0.01% H 2 O 2 in 50 mM Tris (pH 7.6). Cells were postfixed with 1% OsO 4 and then dehydrated through graded alcohol and propylene oxide before embedding in epon. Ultrathin sections were cut on a Reichert Ultracut S ultramicrotome (Leica, Deerfield, IL), grid-stained with uranyl acetate and lead citrate and examined on a Hitachi H2400 transmission electron microscope.
To determine the level of replacement of endogenous mouse PS1 fragments, the same amount of detergent lysates from nontransfected N2a cells were analyzed with lysates from WT.7 and D385A.16 cells (right panel, lanes 1Ј-3Ј). An asterisk indicates human PS1-CTF, and an arrow indicates endogenous mouse PS1-CTF. B, the same cell lines were metabolically labeled with [ 35 S]methionine for 1 h, and the detergent lysates were immunoprecipitated with Ab369. C, the intensity of the bands corresponding to APP CTFs and full-length APP were quantified by phosphorimaging, and relative levels of APP CTFs normalized to full-length APP were calculated and expressed as percentages of average value of all wild type lines. In A and B, molecular mass markers are in kDa. thetically labeled these cultures with [ 35 S]methionine for 4 h. A␤ peptides were immunoprecipitated with 26D6, an antibody specific for residues 1-12 of A␤. Relative to cells expressing wild type PS1, the level of A␤ in the conditioned medium is reduced by ϳ60 -70% in cells expressing D385A mutant. Surprisingly, in a medium of cells expressing D257A or double aspartate mutant PS1, we failed to detect any changes in the levels of secreted A␤ relative to A␤ in the medium of wild type PS1 cells (Fig. 2, A and B), despite elevated levels of accumulated APP CTFs.
Aspartate PS1 Variants Inhibit the Generation of NICD-Next, we examined the effects of aspartate PS1 variants on the intramembraneous ␥-cleavage of Notch. We transiently cotransfected cDNA encoding the truncated form of Notch 1, termed N⌬E (49), and either wild type PS1 or aspartate PS1 variants into PS1-deficient fibroblasts. In PS1-deficient fibroblasts, N⌬E is not proteolytically processed to generate NICD (Fig. 3, lane 1). While expression of wild type PS1 restores NICD production (Fig. 3, lane 2), none of the PS1 aspartate variants (D257A, D385A, or D257A/D385A) could restore NICD production (Fig. 3, lanes 3-5, respectively). These results are consistent with previous studies that demonstrate that expression of the D257A PS1 or D385A PS1 variants fails to generate NICD and is unable to rescue the phenotype associated with the sel-12 deficiency in Caenorhabditis elegans (59).
PS1 Aspartate Mutant Stabilizes Full-length APP-Realizing that our failure to replicate earlier findings that expression of the D257A PS1 variant leads to diminished A␤ production (27) might be a reflection of selecting clonal lines, we chose to generate pools of stable N2a cells expressing the PS1 aspartate variants. The PS1 aspartate variants were stably introduced into a stable N2a line expressing Swedish variant APP 695 , and pools of between 100 -200 independent clones were isolated and subject to pulse-chase analysis (Fig. 4). Cells were pulse-labeled for 10 min and then chased for the indicated time periods. The synthetic rate of APP after 10 min of pulse labeling is comparable in these pools (Fig. 4A, compare lanes 1, 4, 7,  and 10). After a 30-min chase, both immature N-glycosylated APP and mature N-and O-glycosylated forms of APP were detected in all pools (Fig. 4A, lanes 2, 5, 8, and 11). The levels of full-length APP were markedly reduced in wild type PS1expressing cells at the 60-min time point (Fig. 4A, lane 3), consistent with the reported t1 ⁄2 of APP of ϳ45 min (60). In contrast, full-length APP glycoforms appeared to be stabilized in all N2a pools expressing aspartate mutant PS1 at the end of the 60-min chase period (Fig. 4A, lanes 6, 9, and 12). Consistent with these observations, in a 3-h continuous metabolic labeling experiment, the level of accumulated APP is consistently higher in aspartate mutant PS1 pools (Fig. 4A, lanes 14 -16), compared with the wild type PS1 pool (Fig. 4A, lane 13). The levels of soluble form of APP (APP s ) are also slightly elevated in medium of the aspartate mutant cell pools (Fig. 4C, lanes 3, 6,  9, and 12), consistent with an increase in the stability and cell surface levels of APP (see below; Figs. 6 and 7). In addition, APP CTFs accumulated to high levels in PS1 aspartate cell lysates (Fig. 4B, left and right panels). Again, and as we had observed in the cloned lines, the level of A␤ in the conditioned medium was only seen to be decreased in D385A mutant PS1expressing pools (to ϳ55% less than the levels of A␤ in medium of wild type PS1 pools (Fig. 4D, lanes 13-16; quantified in Fig.  4E). Here, the levels of secreted A␤ were quantified relative to the synthetic levels of APP in each pool (10-min pulse; Fig. 3A,  lanes 1, 4, 7, and 10). These data demonstrate that expression of either the D257A, D385A, or double mutant PS1 molecules increases the stability of full-length APP and APP CTFs but that the D385A mutant is the only species that reduces A␤ secretion.  1, 4, 7, and 10), 30 (lanes 2, 5, 8, and 11), or 60 min (lanes 3, 6, 9, and 12) (left panels). Parallel dishes were continuously labeled for 3 h (right panels; lanes [13][14][15][16]. Detergent lysates were immunoprecipitated with either P2-1 (for full-length APP) (A) or Ab 369 (for APP CTFs) (B) antibodies, and conditioned medium was immunoprecipitated with either P2-1 (for APP s ) (C) or 26D6 (for A␤) (D) antibodies and resolved by 7% SDS-PAGE (A and C) or 16.5% Tris/Tricine SDS-PAGE (B and D). E, levels of secreted A␤ in a 3-h continuous labeling experiment (D, lanes [13][14][15][16] were quantified by phosphorimaging and normalized to the synthesis level of APP at a 10-min pulse (A, lanes 1, 4, 7, and 10). F, to determine the level of PS1 expression in these stable pools, parallel dishes were directly lysed and analyzed by immunoblot with ␣PS1Loop antibody. An asterisk indicates human PS1-CTF, and an arrow indicates endogenous mouse PS1-CTF. In A-D and F, molecular mass markers are in kDa.

Subcellular Distribution of APP ␤CTF in D385A Mutant
Cells-To characterize the subcellular compartments where APP ␤CTF accumulates in cells expressing the D385A mutant PS1, we performed laser confocal immunofluorescence studies using 3D6, an antibody that specifically recognizes the N terminus of A␤ and the ␤-secretase-generated membrane-retained stub of APP (␤CTF). The 3D6 antibody recognizes neither the full-length APP nor the CTFs generated by ␣-secretase (Fig.  5A, compare lanes 1 and 2 and lanes 3 and 4). In N2a cells expressing D385A PS1, 3D6 immunoreactivity is detected on the plasma membrane as well as in perinuclear vesicular structures and peripheral small vesicles (Fig. 5B, left panel). On the other hand, in wild type PS1-expressing cells, 3D6 immunoreactivity is much weaker and mainly detected as peripheral dot-like structures (Fig. 5B, right panel).
To identify the compartments in which ␤CTF accumulates in D385A PS1 cells, we stained cells with FITC-VVA (Fig. 5C), a lectin that binds selectively to serine-or threonine-linked, Nacetyl-D-galactosamine moieties and labels glycoproteins found mainly in the Golgi, trans-Golgi network, and plasma membrane; the endoplasmic reticulum and transitional elements are not labeled by this lectin (61). 3D6 immunoreactivity overlaps with FITC-VVA binding in the perinuclear vesicular structures (Fig. 5C). To label endosomes and late endosomes/lysosomes, we incubated cells with FITC-conjugated BSA for 15 min followed by a 90-min chase at 37°C. 3D6 immunoreactivity partially overlaps with internalized FITC-BSA (Fig. 5D). To confirm these observations, we performed immunoelectron microscopy using 3D6 antibody. These studies revealed that APP ␤CTF is not only localized in the cell surface (Fig. 5E, arrowheads) but also in the large vesicular structures, reminiscent of multivesicular bodies (Fig. 5E, arrows). These results indicated that in D385A PS1 cells, APP ␤CTF accumulates in late compartments of the secretory apparatus, including the Golgi, endosomes, and plasma membranes.
Surface Accumulation of APP ␤CTF and Full-length APP-To support the immunofluorescence studies of APP ␤CTF in aspartate mutant cell lines, we performed cell surface biotinylation experiments. Stable N2a pools coexpressing Swedish APP and either wild type or aspartate mutant PS1 were reacted with the membrane-impermeant biotinylation reagent, sulfosuccinimidobiotin, for 30 min. Biotinylated proteins were recovered with streptavidin-conjugated agarose, and APP-related molecules were detected by Western blot analysis with Ab 369. These studies revealed markedly higher accumulated APP CTFs on the surface of aspartate mutant PS1-ex- FIG. 5. Immunofluorescence localization of APP ␤CTF. A, detergent lysates from N2a cells expressing wild type (lanes 1 and 3) or D385A PS1 (lanes 2 and 4) were fractionated on 16.5% Tris/Tricine SDS-PAGE and immunoblotted with either Ab 369 (lanes 1 and 2) or 3D6 (lanes 3 and 4) antibody. Note that 3D6 antibody specifically reacts with ␤CTF but not with full-length APP or ␣CTF. B, N2a D385A.21 (left panel) or WT.7 (right panel) cells were labeled with 3D6 antibody followed by FITC-conjugated anti-mouse IgG secondary antibody to localize ␤CTF in these cells. C, N2aD385.21 cells were stained with 3D6 antibody followed by Texas Red-labeled anti-mouse IgG antibody and FITC-labeled VVA, a lectin that binds to glycoproteins in the Golgi, trans-Golgi network, and plasma membrane. D, N2aD385.21 cells were treated with 1 mg/ml FITC-BSA for 15 min, chased for 90 min, and labeled with 3D6 antibody coupled with Texas Red-conjugated antimouse IgG secondary antibody. In B-D, images were acquired on a laser-scanning confocal microscope. In C and D, red (left), green (middle), and composite (right) images are shown. E, immunoelectron microscopy staining of APP ␤CTF with 3D6 antibody in D385A PS1expressing cells. The arrowheads indicate 3D6 immunoreactivity detected on the cell surface, and the arrows indicate multivesicular bodies stained with 3D6 antibody. The scale bars on the left images are 1.0 m, and that on the right image is 0.5 m.  1 and 5), D257A (lanes 2 and 6), D385A (lanes 3 and 7), or double aspartate mutant PS1 (lanes 4 and 8) were surface-biotinylated with sulfosuccinimidobiotin for 30 min at 4°C. Biotinylated surface proteins were captured with streptavidin-agarose and eluted by boiling in Laemmli SDS sample buffer. After resolving the proteins by 16.5% Tris/Tricine SDS-PAGE, biotinylated APP related species were visualized by Western blot analysis with 369 antibody. One-twentieth of the lysates used for precipitation were also loaded for comparison (right panel, lanes [5][6][7][8]. The bands corresponding to APP, APP ␤CTF, and APP ␣CTF are indicated. pressing cells (Fig. 6, lanes 2-4). Notably, it appears that the levels of 369-precipitable biotinylated ␤CTFs are somewhat higher than the more rapidly migrating ␣CTFs, despite the fact that the levels of ␣CTFs are higher in straight lysates. We speculate that the differences observed in 369-precipitable biotinylated species versus straight lysates can be accounted by differences in the number of lysine residues in ␤CTF (n ϭ 2) versus ␣CTF (n ϭ 1) to which the sulfosuccinimidobiotin reagent can react. Virtually no APP CTFs were detected in the streptavidin precipitates in wild type cells (Fig. 6, lane 1), despite the presence of accumulated APP CTFs in cell lysates (Fig. 6, lane 5). Thus, despite increased levels of accumulated APP CTFs in cell lysates of aspartate PS1-expressing cells (Fig.  6, lanes 6 -8), it is clear that a higher proportion of APP CTFs are present on the cell surface of PS1 aspartate cells compared with cells expressing wild type PS1. Surprisingly, these studies also revealed that the levels of full-length APP on the cell surface of PS1 aspartate cells (Fig. 6, lanes 2-4) are increased relative to the levels in cells expressing wild type PS1 (Fig. 6,  lane 1). To further confirm the increased surface localization of APP and APP ␤CTF, we stained unfixed, live cells with either 3D6 or P2-1 antibodies or FITC-VVA at 10°C for 45 min. As shown in Fig. 7, surface immunoreactivities of 3D6, specific for ␤CTF and P2-1, specific for full-length APP were markedly increased in cells expressing aspartate PS1 variants compared with cells expressing wild type PS1 (Fig. 7, left and middle panels). In contrast, surface labeling by the lectin, FITC-VVA, was not significantly different between the cell types (Fig. 7, right panels), indicating that PS1 aspartate variants have selective effects on the surface accumulation of APP and its CTFs rather than general effects on all plasma membrane glycoproteins. DISCUSSION A variety of genetic and biochemical studies have supported the conclusion that presenilins are ␥-secretase, an activity responsible for intramembraneous processing of APP, Notch 1, and other transmembrane proteins (for a review, see Ref. 62).
An important contribution to these latter conclusions was the demonstration that expression of PS1 harboring alanine substitutions of two highly conserved aspartate residues at position 257 or 385 in Chinese hamster ovary cells led to reduced A␤ secretion and accumulation of APP CTFs (27). Because these aspartate residues are thought to reside within TM6 and TM7, respectively, it has been proposed that PS1 and PS2 are unusual transmembrane diaspartyl proteases that catalyze intramembraneous proteolysis of APP (27). While the notion that PS are novel aspartyl proteases is attractive, there is currently no evidence that PS1 or PS2 exhibits proteolytic activity. Moreover, it does not seem obvious how the transmembrane aspartate 257 and aspartate 385 could catalyze proteolysis of APP near the middle of the bilayer (to generate A␤40/42) and also process Notch 1 near the C terminus of the transmembrane segment. In this regard, intramembraneous processing of Notch occurs at a single site and has strict sequence requirements surrounding the scissile bond (50), whereas ␥-secretase generates A␤ peptides with heterogenous C terminus, and the production of these species are largely insensitive to the sequence of the transmembrane domain (63). Interestingly, recent reports indicate that a peptide corresponding to "APP CTF␥" that is generated in vitro in cytosol-free membranes and detected in vivo is generated by proteolysis between the 21st (leucine 645) and 22nd (valine 646) amino acids of the 24-amino acid APP TM domain (64 -67), a site corresponding to the scissile bond at which Notch 1 is processed to generate NICD (50). Finally, and in view of the nonoverlapping subcellular distributions of PS and the sites of A␤/NICD production (termed the "spatial paradox" (42,44)), it is conceivable that PS plays an indirect role in facilitating intramembraneous processing by influencing the trafficking of ␥-secretase or its substrates to compartments in which proteolysis occurs.
Intrigued with the observation that mutant PS1 with substitution of two aspartate residues affect APP processing in a manner that reduces A␤ peptides, we asked whether these variants might influence the trafficking of APP or CTFs. In this report, we offer several important insights into the role of the PS1 Asp 3 Ala variants on APP trafficking and A␤ production.
First, we reproduce the observation of Wolfe et al. (27) by showing that PS1 with a D385A substitution impairs secretion of A␤ peptides. However, we show that expression of the D257A aspartate mutant PS1 leads to the accumulation of APP CTFs, but this is not commensurate with a decrease in levels of A␤ peptides. While the latter observation is inconsistent with the report by Wolfe et al. (27), Capell et al. (59) have previously reported that human embryonic kidney (HEK) 293 cells expressing D257A PS1 or a naturally occurring splice variant PS1⌬E8 in which aspartate 257 is substituted with alanine does not impair A␤ secretion. These findings have been confirmed by Morihara et al. (68) in N2a cells. At present, the discrepancies between the data obtained by Wolfe et al. (27) and others (see Refs. 59 and 68; this report) have not been resolved. More importantly, we demonstrate that PS1 variants with substitutions of both aspartates 257 and 385 also fail to impair production of A␤ peptides despite the accumulation of APP CTFs. Thus, we argue that the conclusion that the transmembrane aspartates of PS1 are critical for catalysis within the APP transmembrane domain should be revised. Second, we now provide the entirely unexpected observation that PS1 aspartate variants have profound effects on APP trafficking. We show that in comparison with cells expressing wild type PS1, the half-life of full-length APP and the steady-state levels of APP on the plasma membrane of cells expressing any of the PS1 aspartate variants are markedly elevated. Moreover, the APP CTFs also accumulate on the plasma membrane of these cells, as has been previously reported (59). While the mechanism(s) underlying these fascinating observations remain to be determined, these studies support our earlier demonstration that the trafficking of a limited set of membrane proteins is significantly affected in PS1-deficient neurons. In these cells, the rate of secretion of APP s␣ is quantitatively increased, and APLP1 CTFs also accumulate. Moreover, the rate of acquisition and steady-state levels of complex oligosaccharide modifications of TrkB, the tyrosine kinase receptor for brain-derived neurotrophic factor, are diminished in PS1-deficient neurons (51). In addition, the trafficking of a cell adhesion molecule, termed ICAM-5 or telencephalin (69), to the plasma membrane is abolished in PS1-deficient neurons (42). It is also noteworthy that SPE-4, a divergent member of the presenilin family in C. elegans, is required for proper localization of macromolecules that are subject to asymmetric partitioning during spermatogenesis (70).
The mechanism(s) by which the aspartate 257 and aspartate 385 residues of PS1 regulate intramembraneous proteolysis and trafficking of APP and APP CTFs has not been determined. However, expression of PS1 with a deletion of the first two transmembrane domains (⌬TM1,2) also inhibits A␤ production (71), indicating that alterations of PS1 structure at sites relatively distant from the hypothesized "active" site aspartate residues can have an apparently dominant negative effect on ␥-secretase activity. Moreover, the observation that expression of the PS1 aspartate mutants reduces the size of the PS1 complex (72) suggests that the aspartates are critical for assembly of the functional ␥-secretase complex rather than for promoting catalysis, per se. Collectively, these data suggest that PS1 has a multiplicity of roles in regulating membrane trafficking and mediating intramembraneous proteolysis. Clarification of these enigmatic features of PS1 function will require the isolation and characterization of the constituents of PS1 "complex," reconstitution of the ␥-secretase reaction in vitro, and the development of cell biological strategies to assess the precise role of PS1 in the trafficking of selected membrane proteins.