Familial Alzheimer Disease-linked Presenilin 1 Variants Enhance Production of Both Aβ1–40 and Aβ1–42 Peptides That Are Only Partially Sensitive to a Potent Aspartyl Protease Transition State Inhibitor of “γ-Secretase”*

Presenilin 1 (PS1) plays an essential role in intramembranous “γ-secretase” processing of several type I membrane proteins, including the β-amyloid precursor proteins (APP) and Notch1. In this report, we examine the activity of two familial Alzheimer's disease-linked PS1 variants on the production of secreted Aβ peptides and the effects of L-685,458, a potent γ-secretase inhibitor, on inhibition of Aβ peptides from cells expressing these PS1 variants. We now report that PS1 variants enhance the production and secretion of both Aβ1–42 and Aβ1–40 peptides. More surprisingly, whereas the IC50 for inhibition of Aβ1–40 peptide production from cells expressing wild-type PS1 is ∼1.5 μm, cells expressing the PS1ΔE9 mutant PS1 exhibit an IC50 of ∼4 μm. Immunoprecipitation and matrix-assisted laser desorption/ionization time-of-flight mass spectrometry reveal that the levels of Aβ1–43 peptides are elevated in medium of PS1ΔE9 cells treated with higher concentrations of inhibitor. The differential effects of wild-type and mutant PS1 on γ-secretase production of Aβ peptides and the disparity in sensitivity of these peptides to a potent γ-secretase suggest that PS may be necessary, but not sufficient, to catalyze hydrolysis at the scissile bonds that generate the termini of Aβ1–40 and Aβ1-42 peptides.

Presenilin 1 and 2 (PS1 and PS2) 1 are polytopic membrane proteins that are mutated in the majority of pedigrees with early onset familial Alzheimer's disease (FAD) (1)(2)(3). It is now established that PS play an essential role in intramembranous "␥-secretase" processing of type I membrane proteins, including the ␤-amyloid precursor proteins (APP) (4,5), the developmental signaling receptor, Notch1 (6 -8), the tyrosine kinase recep-tor ErbB4 (9,10), and N-and E-cadherins (11). For APP, ␥-secretase catalyzes proteolysis of a set of membrane-tethered APP derivatives, termed APP-CTFs, resulting in the production and secretion of ␤-amyloid (A␤) peptides. On the other hand, ␥-secretase-mediated processing of the membrane-tethered Notch1 derivative, termed S2/NEXT, releases the intracellular domain (S3/NICD) that subsequently translocates to the nucleus and activates transcription of target genes (7,8). The observation that A␤ and S3/NICD production are completely eliminated in cells derived from mouse blastocysts with compound deletions of PS1 and PS2, lends convincing support to the notion that PS are critical for intramembranous cleavage of APP and Notch1 (12,13).
Although the mechanism(s) by which PS facilitates ␥-secretase processing of APP and Notch1 have not been fully elucidated, the generation of A␤ (14 -16) and S3/NICD (17) have been show to be inhibited by highly potent and selective aspartyl protease transition state inhibitors that bind specifically to PS1 and PS2 (14,18). These data, taken with the description of a family of signal peptide peptidases with limited homology to PS (19), have led to the conclusion that PS are the elusive ␥-secretases (20).
While appealing, the "PS is ␥-secretase" model has several weaknesses. First, mutagenesis studies have revealed that ␥-secretase has relaxed substrate selectivity within the APP transmembrane domain and occurs at heterogeneous sites (21,22), while ␥-secretase cleavage of Notch1 is highly sequencespecific and appears to generate a single S3/NICD species (7). Second, whereas endocytosis and recycling of APP-CTFs are required for the generation of A␤ (23), S3/NICD production does not require endocytic trafficking of the Notch derivative, S2/NEXT (24). Third, the identification of several PS-interacting membrane proteins, including nicastrin (25), APH-1 (26), and PEN2 (27) that also modulate production of S3/NICD (25)(26)(27) and A␤ (25,28) suggests that a protein complex, comprised of PS and other factors are required for intramembranous proteolysis of APP and Notch1. Finally, PS1 harboring a substitution of aspartate 257 with alanine is capable of processing APP to A␤ peptides (29,30), but fails to generate S3/NICD from a truncated Notch1 molecule, termed Notch⌬E (31). Similarly, expression of several FAD-linked PS1 variants (30 -32) or the experimental L286E or L286R PS1 mutants (33) leads to exaggerated overproduction of highly fibrillogenic A␤42 peptides, but surprisingly, these PS1 variants fail to generate S3/NICD from Notch⌬E.
Intrigued by the apparent discordance between the activities of FAD-linked PS1 mutants on the production of A␤42 peptides and S3/NICD production, we examined the activity of these FAD-linked PS1 variants on the production of secreted A␤ peptides and the effects of a potent aspartyl protease transition state inhibitor of ␥-secretase, termed L-685,458 (15,16) on the production of these A␤ species. We now report that while PS1 variants enhance production of A␤42, as expected, there is an unexpected enhancement in levels of secreted A␤40 peptides. We also provide the first demonstration that in the conditioned medium of "pools" of stable cell lines that express individual FAD-linked mutant PS1, both A␤1-40 and 1-42 peptides accumulate to higher levels than the A␤ peptide variants in medium of cell pools that express wild-type PS1. More surprisingly, under conditions at which the ␥-secretase inhibitor completely eliminates production of all A␤-related peptides from cells expressing wild-type PS1, we now report that the inhibitor is not fully effective at lowering production of A␤ variants from cells expressing two independent FAD-linked PS1 mutants. Hence, we argue that production of A␤ peptides are differentially regulated by the expression of wild-type and FAD-linked PS1 variants.
␥-Secretase Inhibitor Assay-For ␥-secretase inhibitor assays, cells were incubated for 16 h in medium containing 2 M (or indicated concentrations) of the ␥-secretase inhibitor, L-685,458 (16), prepared in dimethyl sulfoxide (Me 2 SO) or an equivalent concentration of Me 2 SO as a vehicle control.
Conditioned media were collected and immediately adjusted to 0.5 mM phenylmethylsulfonyl fluoride. Cultured cells were lysed in 1ϫ immunoprecipitation (IP) buffer containing 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 5 mM EDTA, 0.5% Nonidet P-40, 0.5% sodium deoxycholate, and protease inhibitor mixture (Sigma). Nuclei and debris were removed by centrifugation and the protein concentration of detergentsoluble proteins in each lysate was determined using a bicinchoninic acid protein assay kit (Pierce).
Immunoprecipitation, Electrophoresis, and Immunoblotting-For immunoprecipitations, we used equivalent volumes of conditioned medium based on the calculation of the protein concentration in each plate of cells to avoid experimental bias because of variations in cell density. Normalized conditioned media were immunoprecipitated with 26D6, a monoclonal antibody raised against A␤1-12 (37), for 16 h at 4°C. The immune complexes were "bridged" by the addition of rabbit anti-mouse IgG (Pierce), collected with protein A-conjugated agarose beads (Pierce), and eluted by boiling for 5 min in Laemmli SDS sample buffer prior to fractionation on SDS-PAGE.
Aliquots of detergent lysate were fractionated on high percentage Tris-Tricine SDS-PAGE gels for detection of full-length APP and APP-CTFs, or Tris glycine SDS-PAGE for analysis of PS1. To detect secreted A␤40/42, immunoprecipitated samples were fractionated on Bicine/ urea gels (38). Fractionated proteins were electrophoretically transferred to polyvinylidene difluoride membranes (Bio-Rad), and the membranes were probed with appropriate primary antibodies. Full-length APP and APP-CTF were detected by CT15, an antisera that recognizes the carboxyl-terminal 15 amino acids of APP (39). A polyclonal antibody, PS1 NT , was used to detect full-length PS1 and PS1 NTF (40). Soluble APP␣ and A␤40/42 were detected by 26D6 (37). After incubation with horseradish peroxidase-coupled secondary antibodies (Pierce), bound antibodies were visualized using an enhanced chemiluminescence (ECL) detection system (Perkin-Elmer Life Sciences).
Metabolic Labeling and Immunoprecipitation-N2a cells were starved for 30 min in methionine-free Dulbecco's modified Eagle's medium (Invitrogen) and then labeled with 250 Ci/ml [ 35 S]methionine (PerkinElmer Life Sciences) in methionine-free Dulbecco's modified Eagle's medium supplemented with 1% dialyzed fetal bovine serum (Invitrogen) for 10 min (for pulse-labeling) or 2 h. Conditioned medium was collected and cells were lysed in IP buffer. For immunoprecipitations we used a volume of conditioned medium that was normalized to the calculated trichloroacetic acid-precipitable radioactive counts (cpm) in cell lysates. Soluble APP␣ and A␤40/42 were immunoprecipitated with monoclonal antibody, 26D6 (37). To examine APP synthesis, cells were pulse-labeled with [ 35 S]methionine for 10 min, and APP was immunoprecipitated with 369 antibody, raised against a peptide corresponding amino acids 649 -695 of APP 695 (41). Immunoprecipitates were fractionated by SDS-PAGE, and radioactive bands were visualized and quantified using a PhosphorImager (Amersham Biosciences).
Mass Spectrometric Analysis-Conditioned media from N2a swe.10 cell pools stably expressing wild-type PS1 or the PS1⌬E9 were immunoprecipitated with 4G8 antibody, specific for amino acids 17-24 of A␤, and collected with Protein A/G-coupled agarose beads prior to analysis by matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometric analysis, as described (42). Each mass spectrum was averaged from at least 500 measurements, and bovine insulin was included as an internal mass calibrant.

Effects of FAD-linked PS1 Variants on Production of A␤ Variants and Inhibition with a ␥-Secretase
Inhibitor-It is now well accepted that expression of FAD-linked PS1 variants elevate the levels of secreted A␤1-42 peptides and, in so doing, increase the calculated ratio of A␤40/A␤42 peptides. However, the absolute levels of A␤ peptide variants have rarely been reported, a reflection, in large part, of the variability in transgene-encoded APP between individual lines. With this caveat, we chose to transfect a neuroblastoma N2a cell line that constitutively expresses a carboxyl-terminal Myc epitope-tagged human APP 695 harboring the FAD-linked "Swedish" mutations (38) with human wild-type PS1 (wtPS1) or the FADlinked PS1 variants, PS1⌬E9, or E280A to generate stable pools that express human PS1 polypeptides. Western blot analysis of stable cell pools revealed the accumulation of human PS1 NTF and low levels of full-length precursor in cells expressing wild-type PS1 (Fig. 1A, lane 1), uncleaved ϳ43-kDa PS1⌬E9, and low levels of endogenous mouse PS1 NTF in cells expressing PS1⌬E9 (Fig. 1A, lane 2), and mutant human PS1 NTF and low levels of full-length precursor in cells expressing the E280A variant (Fig. 1A, lane 3). In these cell pools, "replacement" of the bulk of murine PS1 fragments has occurred (45), although residual levels of murine NTF (as seen in the PS1⌬E9 cells) are still present. This would be expected in a cell pool in which transgene-derived products are expressed at varying levels in independent clones. We examined the levels of secreted A␤-related species by immunoprecipitation with antibody 26D6 (39), specific for A␤ residues 1-12, fractionation of immune complexes on Bicine/urea gels, and analysis of immunoprecipitated A␤ peptides using 26D6 antibody and enhanced chemiluminescence detection. For these studies, we calculated the protein concentration in each plate of cells, and used normalized volumes of medium so that there would be no experimental bias because of differences in cell density. In Fig. 1, we show that constitutive expression of wild-type PS1 leads to robust secretion of A␤1-40 peptides, limited levels of secreted A␤1-37, A␤1-38, and A␤1-39 peptides and nearly undetectable levels of A␤1-42 peptides (Fig. 1B, lane 1). This level of secreted A␤ peptides is no higher than parental APPswe.10 cells (data not shown). On the other hand, we consistently observed that the levels of accumulated A␤1-40 and A␤1-42 peptides were elevated in medium of cells expressing either the PS1⌬E9 or A280E variants (Fig. 1B, lanes 3 and 5, respectively). Even more surprising was the observation that under conditions in which treatment of wtPS1 cells with a potent ␥-secretase inhibitor, L-685,458 (17) (2 M for 16 h), resulted in nearly complete inhibition of secreted A␤ peptides (Fig. 1B, lane 2), low levels of A␤1-40 and A␤1-42 peptides still remained in the medium of cells expressing either PS1 mutant (Fig. 1B, lanes 4 and 6).  2 and 4) of the ␥-secretase inhibitor. Conditioned medium was immunoprecipitated with 26D6 and separated by Tris-Tricine SDS-PAGE. APP s␣ generated by ␣-secretase and A␤1-40/A␤1-42 peptides are indicated. Note that the level of total secreted A␤ peptides in medium of cells expressing the PS1⌬E9 are elevated relative to the A␤ peptides in wild-type PS1 cells. E, to examine the identity of secreted A␤ species, A␤ radiolabeled from cells expressing wild-type PS1 (lanes 1 and 2) or the FAD-linked PS1⌬E9 (lanes 3 and 4) was subjected to immunoprecipitation with 26D6, and immune complexes were fractionated on Bicine/urea gels. Note that dramatic effects of the inhibitor are on the A␤1-40 species, with relative sparing of A␤1-42 species (lane 4). F, the absolute PhosphorImager units of the bands corresponding to A␤1-40 (left panel) and A␤1-42 (middle panel) were determined by phosphorimaging, and the ratio of counts/min in A␤42 to A␤40 was plotted (right panel). The asterisk in the middle and right panels reflects the undetectable levels of A␤42 in medium of wtPS1 cells in the presence of the inhibitor.
Our finding that two FAD-linked PS1 variants enhance secretion of the principal A␤ variant, A␤1-40, is novel and we felt it important to fully validate this finding. We chose to focus on the PS1⌬E9 pool. First, to establish that the differences in accumulated A␤ peptides between wtPS1 and PS1⌬E9 cell pools was not simply a reflection of differences in synthetic levels of transgene-encoded APPswe, we pulse-labeled cells with [ 35 S]methionine for 10 min and analyzed newly synthesized APP in cell lysates by subjecting equivalent detergentsoluble, trichloroacetic acid-precipitable, radioactivity (cpm) to immunoprecipitation with antibody 369, raised against a peptide corresponding to amino acids 649 -695 of APP (43), fractionation of immune complexes on SDS-PAGE, and phosphorimaging. In Fig. 1C, we show that the synthesis of full-length APP is indistinguishable between the cell pools that express either human wtPS1 or PS1⌬E9. To further quantify the absolute increase in both A␤1-40 and A␤1-42 peptides in medium of mutant PS1-expressing cells relative to cells expressing wtPS1, we incubated cell pools with [ 35 S]methionine for 2 h and quantified the levels of secreted A␤ peptides in medium by immunoprecipitation with antibody 26D6, fractionation of immune complexes on Tris-Tricine gels, and phosphorimaging. For these analyses, we quantified total counts/min in detergent-solubilized cell lysates and used normalized volumes of radiolabeled conditioned medium for immunoprecipitations. As we have shown by Western blot analysis (Fig. 1B), quantitative phosphorimaging analysis revealed an elevation in total A␤ peptides in medium of cells expressing the PS1⌬E9 mutant (Fig. 1D, lane 3) by 2.8-fold relative to A␤ peptides secreted from cells expressing wtPS1 (Fig. 1D, lane 1). Notably, the 26D6 antibody, specific for A␤ residues 1-12, also detects soluble derivatives generated by ␣-secretase, termed APP sa , quantitative phosphorimaging revealed a 1.5-fold increase in levels of APP sa in medium of PS1⌬E9 cells relative to cells expressing wild-type PS1 (Fig. 1D, compare lanes 3 and 1, respectively), this despite identical synthetic rates of the APPswe precursor between cell pools (Fig. 1C). These findings offer the suggestion that expression of the PS1⌬E9 variant leads to enhanced trafficking (or processing) of full-length APPswe to cellular compartments in which ␣-secretase is active, but further studies will be necessary to address this issue. In any event, quantitative phosphorimaging revealed that while treatment of wtPS1 cells with 2 M inhibitor reduced production of newly synthesized A␤ peptides to ϳ0.4% of untreated controls (Fig.  1D, lane 2), the inhibitor diminished the levels of total A␤ peptide species in medium of PS1⌬E9 cells to ϳ12% of the untreated control (Fig. 1D, lane 4). Further examination of the complexity of radiolabeled A␤ peptide variants by Bicine/ urea gels (Fig. 1E) revealed that the absolute levels of both A␤40 and A␤42 variants were elevated in medium of PS1⌬E9 cell medium (Fig. 1E, lane 3; quantified in Fig. 1F, left panel) relative to the levels in medium of cells expressing wild-type PS1 (Fig. 1E, lane 1; quantified in Fig. 1F, left panel). Furthermore, under conditions where the ␥-secretase inhibitor almost completely eliminated A␤1-40 species in medium of cells expressing wild-type PS1, this compound diminished A␤1-40 peptide levels to ϳ13% in medium of cells expressing PS1⌬E9 ( Fig. 1E; quantified in Fig. 1F, left panel). Similarly, the ␥-secretase inhibitor fully eliminates secreted A␤1-42 peptides from cells expressing wild-type PS1, but L-685,458 treatment only reduced the level of A␤1-42 peptides to ϳ58% in medium of cells that express PS1⌬E9 (Fig. 1E, lane 4; quantified in Fig. 1F, middle panel). Notably, the levels of A␤1-40 and A␤1-42 peptides are nearly identical in PS1⌬E9 cell medium after treatment with L-685,458 (Fig. 1F, right  panel).
Dose Dependence of ␥-Secretase Inhibitor on Secretion of A␤ Peptide Variants-Having established that treatment with 2 M L-685,458 for 16 h completely eliminates A␤ peptides production from cells expressing wtPS1 cells, but fails to fully inhibit production of secreted A␤ peptide from cells expressing PS1⌬E9 cells, we asked whether the individual A␤ peptide variants may be differentially sensitive to the dose of ␥-secretase inhibitor. We treated parallel dishes of N2a swe.10 cell pools expressing wild-type PS1 or the PS1⌬E9 variant with increasing concentrations of L-685,458 for 16 h. The conditioned medium was collected and the protein concentration in cell lysates was determined. Normalized volumes of medium, relative to protein concentration in cell lysates, were subjected to immunoprecipitation and Western blot analysis. Consistent with the data shown in Fig. 1, the constitutive levels of A␤1-40 and A␤1-42 peptides in medium of cells expressing the PS1⌬E9 variant are elevated compared with cells expressing wild-type PS1 (Fig. 2A, compare lanes 6 and  1, respectively). At 2 M, L-685,458 lowered the levels of secreted A␤1-40 peptides in medium of wtPS1 cells to ϳ20% of the levels in vehicle-treated cells (Fig. 2A, lane 3). However, at this concentration of inhibitor, the levels of A␤1-40 peptides in medium of cells expressing PS1⌬E9 were reduced to ϳ75% of the levels in vehicle-treated cells (Fig. 2A, lane 8).
Notably, treatment with 8 M inhibitor virtually eliminates all A␤ peptides in medium of cells expressing wild-type PS1 (Fig. 2A, lane 5), but this concentration of inhibitor diminished the levels of A␤1-40 peptides in medium of PS1⌬E9 cells to ϳ20% of the level in vehicle-treated cells (Fig. 2A,  lane 10).
Densitometric quantification of the data in Fig. 2A reveals that the production of A␤1-40 and A␤1-42 in cells expressing wild-type PS1 are equally sensitive to the ␥-secretase inhibitor, with an IC 50 of ϳ1.5 M (Fig. 2B, left panel). On the other hand, the production of A␤1-40 is inhibited with an IC 50 of over 4 M in cells expressing PS1⌬E9 (Fig. 2B, right panel). Moreover, in medium of PS1⌬E9 cells, 50% inhibition of A␤1-42 peptides occurs with 2 M inhibitor (also apparent by immunoprecipitation of radiolabeled A␤1-42 peptides; Fig. 1, E and F), but at higher concentrations of compound, there is a paradoxical plateau in levels of these peptides. However, MALDI-TOF analysis (see below, Fig. 3) of the peptides generated after treatment of PS1⌬E9 cells with high concentrations of compound reveal that A␤1-43 peptides now become prominent. We do not know where A␤1-43 peptides migrate in Bicine/ urea gels, but it is likely that these overlap with the A␤1-42 species and hence, we argue that the Western blot signals in lanes 9 and 10 of Fig. 2A represent a combination of A␤1-42 and A␤1-43 variants.
In parallel to the examination of secreted A␤ peptides, we examined the accumulation of APP-CTFs in detergent-solubilized lysates from cells treated with increasing concentrations of the ␥-inhibitor. In cells that express wild-type PS1, both endogenous ␣-CTF, and APPswe-derived ␣and ␤-CTFs accumulate even after treatment with 1 M inhibitor (Fig. 2C, lane  2), and the level of these fragments seems not to increase as a function of inhibitor concentration (Fig. 2C, lanes 3-5). In contrast, treatment of PS1⌬E9 cells with 1 M inhibitor leads to the accumulation of APP-CTFs (Fig. 2C, lane 7), but at considerably lower levels than that observed in lysates of wtPS1 cells treated with the same concentration of inhibitor. Indeed, the most robust increase in accumulated APP-CTFs in PS1⌬E9expressing cells occurs after treatment with inhibitor at 4 and 8 M (Fig. 2C, lanes 9 and 10, respectively), concentrations that also appear to have the most pronounced effect on A␤ production ( Fig. 2A, lanes 9 and 10).

MALDI-TOF Analysis of A␤ Variants in Medium of Cells
Expressing Human Wild-type PS1 or PS1⌬E9 following Treatment with a ␥-Secretase Inhibitor-To characterize the A␤related species that accumulate in medium of cells expressing human wild-type PS1 and PS1⌬E9 cells, and in medium of these cell pools treated with differing concentrations of the inhibitor, we used antibody 4G8, specific for epitopes 17-24 of A␤, to immunoprecipitate A␤-related peptides from several of the samples shown in Fig. 2; resulting immune complexes were analyzed by MALDI-TOF mass spectrometry (42). In Fig. 3, panel A, we show that in medium of wild-type PS1 cells treated with vehicle (Me 2 SO), A␤1-40 (M r 4330) is the prominent species, with minor species of 1-34, 1-37, 1-38, and 1-39 also present. Upon treatment of these cells with 1 M L-685,458, A␤1-40 and the minor species are still present, and a small peak, representing A␤1-42 (M r 4514) also appears (Fig. 3,   panel B); low concentrations of ␥-secretase inhibitors have been shown to have a paradoxical effect on elevating A␤1-42 peptides (44). However, the levels of all peptides are dramatically reduced after treatment at 4 M (Fig. 3, panel C), and virtually nonexistent after treatment with 8 M L-685,458 (Fig. 3, panel  D). In medium of untreated PS1⌬E9 cells, we observed the presence of both A␤1-40 and A␤1-42 peptides and trace levels of A␤1-38, A␤1-39, and A␤1-43 peptides (Fig. 3, panel E). After treatment with 1 M L-685,458, we failed to see an appreciable difference in the levels of any of the aforementioned A␤ species (Fig. 3, panel F). However, and in sharp contrast to our observations of A␤ peptides in wtPS1 cell medium, treatment of PS1⌬E9 cells with 4 M L-685,458 lead to a reduction in levels of A␤40 peptides, with no appreciable change in the levels of A␤1-42 and A␤1-43 species (Fig. 3,  panel G). Most interestingly, we now observe a pronounced FIG. 2. Effects of FAD-linked ⌬E9 on inhibition by a ␥-secretase inhibitor. A, N2a swe.10 cells stably expressing wild-type PS1 (lanes [1][2][3][4][5] or the PS1⌬E9 (lanes 6 -10) were incubated for 16 h in serum-free medium containing the indicated concentrations of ␥-secretase inhibitor, L-685,458. Note that production of secreted A␤ in cells expressing wild-type PS1 was largely inhibited by treatment between 2 and 4 M (lanes 3 and 4), whereas a significant amount of A␤ peptides was detected at these concentrations of the inhibitor in cells expressing PS1⌬E9 (lanes 8 and  9). Notably, treatment of the inhibitor at 8 M fails to diminish the level of A␤ peptides in medium of the PS1⌬E9 cells (lane 10), whereas treatment at this concentration of the inhibitor virtually eliminates A␤ peptides in wild-type PS1 cells (lane 5). B, the band intensity corresponding to A␤1-40 (open circle) and A␤1-42 (closed circle) were quantified using densitometry (Molecular Dynamics) and plotted. C, accumulation of APP-CTFs in detergent lysates prepared from cells expressing wild-type PS1 (lanes 1-5) or the PS1⌬E9 (lanes 6 -10) were analyzed by immunoblotting with CT15 antibody. Note that both endogenous APP-CTF␣ and APPswe-derived ␣and ␤-CTFs accumulate even at 1 M inhibitor in cells expressing wild-type PS1 (lane 2), and that the level of these species does not change as the concentration of the inhibitor is increased (lanes 3-5). Treatment of the inhibitor at 1 M resulted in lower levels of accumulated APP-CTFs (lane 7) in cells expressing the PS1⌬E9 compared with those observed at the same concentration in wild-type PS1-expressing cells. Peaks that cannot be identified as A␤-related are designated as "n." elevation in levels of A␤1-43 peptides. This latter result becomes much more apparent in PS1⌬E9 cells treated with 8 M L-685,458, where the levels of A␤1-43 peptide now exceeds the levels of A␤1-42 peptides (Fig. 3, panel H); significant levels of the A␤1-40 peptide are still apparent under these conditions. Furthermore, and in sharp contrast to cells expressing wildtype PS1 (Fig. 3, panels B-D), the levels of A␤1-34, A␤1-37, A␤1-38, and A␤1-29 variants in medium of PS1⌬E9 cells appear unchanged no matter what concentration of inhibitor is employed.
It should be noted that while the IP-MS paradigm is useful for the identification of A␤ peptides, the technique only serves a qualitative tool for assessment of peptide levels. For example, the levels of A␤1-40 peptides in lanes 7 and 9 of Fig. 2 with the analogous samples analyzed by IP-MS in Fig. 3, panels F and G, respectively, are not comparable. It is highly likely that under the conditions used in the IP-MS experiment, the antibody is not in excess, and hence, the antibodies are fully saturated with captured A␤-related peptides. Moreover, it is our experience, and those of others, that unusual biophysical properties (including aggregation) of A␤1-42 peptides greatly hinder desorption of these species from the matrix and thus, the peak heights observed in MALDI-TOF analysis do not accurately reflect steady-state levels. This artifact is likely exaggerated for A␤1-43 peptides. Hence, the signals observed in Western blot analysis of fractionated A␤ peptides on Bicine/ urea gels provide a more accurate reflection of the steady-state levels of A␤ species in the medium. DISCUSSION Taken together with the finding that intramembranous ␥-secretase processing of APP and Notch1 is abolished in cells with genetic ablations of PS1 and PS2 (12,13), and the demonstration that high affinity ␥-secretase inhibitors specifically bind to PS (14,18), it has been concluded that PS are the elusive ␥-secretases (20). However, several lines of evidence have emerged that have questioned the veracity of the PS are ␥-secretase model. First, intramembranous processing of APP and Notch1 can be discriminated by a JLK family of nonpeptidic inhibitors (45); the JLK inhibitors block A␤ peptide production, but have very little, if any, effect on production of the Notch1 derivative, S3/NICD (45). Second, expression of the FAD-linked L166P PS1 variant (32) or the experimental L286E or L286R PS1 mutants (34) leads to exaggerated overproduction of A␤42 peptides, but these PS1 variants fail to generate S3/NICD (32,34). Supporting these studies, St. George-Hyslop and colleagues (33) recently reported that expression of the L392V, G206A, or ⌬E9 variants results in compromised S3/ NICD production and "⑀" cleavage within the APP transmembrane domain. Thus, it would appear that while the production of A␤ peptides, cleavage at the Notch S3 and APP ⑀ sites are presenilin-dependent, the catalytic activities may be distinct (33). Finally, recent studies have shown that the generation of intracellular A␤1-42 peptides is PS-independent, suggesting that intramembranous processing of APP may be mediated by distinct ␥-secretase activities (46).
In an attempt to clarify the effects of FAD-linked mutant PS1 on the generation of A␤ peptides, we generated pools of stable cell lines expressing human PS1 or several independent FAD-linked PS1 variants and analyzed accumulated A␤ peptide variants in the medium of these cells. In addition, we assessed the effects of a potent and highly selective transition state inhibitor of ␥-secretase on FAD-linked PS1-mediated production of A␤ peptides.
In this report, we offer several novel insights relevant to PS-dependent ␥-secretase processing of APP. First, we provide the first demonstration that in the conditioned medium of pools of stable cell lines that express individual FAD-linked mutant PS1, both A␤1-40 and 1-42 peptides accumulate to higher levels than the A␤ peptide variants in medium of cell pools that express wild-type PS1. Despite years of investigation describing the effects of FAD-linked PS1 on elevation in the relative ratio of A␤42:A␤40 peptides, it comes as a surprise that the absolute levels of secreted A␤ species have not been compared in a systematic fashion. In large part, we suspect that in the analysis of individual stable cell lines, the levels of coexpressed human APP are highly variable, whether the APP transgene is coselected, or when the PS1 transgenes are stably expressed in a "parental" cell line that constitutively expresses a human APP transgene. In the present study, we utilize cell pools to "normalize" the level of APP expression across the entire population of between 100 and 200 individual lines. The mechanism(s) involved in the exaggerated production of A␤1-40 peptides are not presently known, but may involve PS1⌬E9 enhancement of cleavage at the ␤-secretase site of full-length APPswe. However, using radiolabeling studies, we have not observed any significant differences in the production, or accumulation of CTF␤ in cell pools expressing either wild-type PS1 or PS1⌬E9 (data not shown). Thus, we would offer the tentative conclusion that in addition to enhancing processing at sites within the APP transmembrane domain to generate A␤1-42, the PS1⌬E9 variant (and E280A variant) also enhances processing at the scissile bond between amino acids 636 and 637 (of APP695) to elevate production of A␤1-40 peptides. In support of this suggestion, we observe that in titration studies using L-685,458 (Fig. 2C), APP-CTFs only accumulate at higher concentrations of inhibitor in cells expressing PS1⌬E9. Thus, it would appear that PS⌬E9-mediated intramembranous processing of APP-CTFs is more efficient, thus enhancing production of A␤1-40 and 1-42 peptides, and that these reactions are less sensitive to concentrations of inhibitor that would otherwise block A␤ production in cells expressing wild-type PS1.
Second, we provide unequivocal evidence that the production of A␤1-42 peptides from cells expressing FAD-linked PS1⌬E9 and E280A mutations are largely insensitive to a potent ␥-secretase inhibitor, L-685,458, at concentrations that would otherwise inhibit A␤ production from cells that express wildtype PS1. In addition, the FAD-linked PS1-dependent production of A␤1-40 is somewhat refractory to inhibition by this compound. Similar results have been obtained using a structurally unrelated ␥-secretase inhibitor, compound E (47). Of significant interest is our finding that while the production of A␤1-42 peptides are largely resistant to inhibition by L-685,458, the production of A␤1-43 peptides by cells expressing PS1⌬E9 are elevated in parallel with increasing concentrations of L-685,458. The molecular mechanisms underlying the curious observation that A␤1-42 peptide production is highly refractory to potent ␥-secretase inhibitors is perplexing. Earlier studies showed that a peptidomimetic inhibitor, termed compound 1, could block production of both A␤40 and 42 peptides with an IC 50 of ϳ16 M in Chinese hamster ovary cell lines that stably express human wild-type PS1, and that the IC 50 was slightly increased (to ϳ22 and ϳ20 M for inhibition of total A␤ and A␤42, respectively) in cell lines expressing the FAD-linked M146L PS1 variant (47). The authors argue that PS1 contains the active site of ␥-secretase and that FAD-linked mutations induce subtle changes in PS1 conformation within the proteolytic complex, thus requiring a higher concentration of inhibitor to block A␤ production. In contrast to these earlier studies, we now show that at the IC 50 for inhibition of A␤1-40 production from cells expressing wild-type PS1 (ϳ1.5 M), L-685,458 only reduces A␤1-40 levels by ϳ25% from cells expressing PS1⌬E9. Moreover, the production of A␤1-42 peptide produc-tion from cells expressing PS1⌬E9 is highly resistant to the inhibitor, even at the highest concentrations of compound tested. At a mechanistic level, a series of recent elegant enzymological studies of solubilized ␥-secretase activity that employed a variety of aspartyl protease transition state analogs (including L-685,458) and nontransition state analogs have led to the conclusion that these compounds act in a noncompetitive fashion (48). The model proposed is that the substrate binding site is distinct from the active site, but once "docked," the substrate is subsequently displaced to the catalytic site of the enzyme (48). The model that PS1 harbors the active site of ␥-secretase, and that FAD mutants subtly alter PS1 conformation so as to alter the IC 50 of ␥-secretase inhibitors (48) is tempting. However, it remains extremely perplexing that the FAD-linked PS missense mutations occur widely throughout the molecule, including in hydrophilic loop domains that are predicted to be quite remote from the lipid environment in which intramembranous cleavage takes place. Nevertheless, all of the FAD-linked PS variants have the unique property of enhancing cleavage at a single site in the APP transmembrane domain. It is difficult to reconcile this with a simple effect on an active site. Furthermore, expression of experimental PS1 variants harboring a P434A mutation in the PS1 carboxyl-terminal domain (49) or a deletion of the first two transmembrane domains (50), sequences that are remote from two transmembrane aspartate residues that are proposed to serve as the catalytic center (20), also reduces the production of A␤ peptides, arguing that multiple domains of PS1 are required for ␥-secretase activity (49,50). Finally, the observation that in cells expressing FAD-linked mutant PS1, the production of A␤1-40 versus A␤1-42 peptides are differentially sensitive to L-685,458 (this study) makes it difficult to conceive of a model in which hydrolysis at the scissile bonds that generate the termini of A␤1-40 and A␤1-42 peptides could be mediated by PS, alone.
Finally, it is now firmly established that PS1 is present in a high molecular weight complex, and that molecules, termed nicastrin (25), APH-1␣/APH-1␤ (26), and PEN2 (27), are components of this complex. Moreover, nicastrin and PS levels are coregulated (51)(52)(53), and RNA interference (RNAi)-mediated reduction of nicastrin, APH-1, or PEN2 levels results in compromised secretion of A␤1-40 and A␤1-42 peptides (27). Hence, each component of the complex appears to exert differential effects of A␤ production. Hence, it is conceivable that the ␥-secretase inhibitors block A␤ production by influencing the interactions and biochemical properties of individual subunits within the complex. Future efforts will require the development of reconstitution systems to elucidate the biochemical interactions of PS with nicastrin, APH-1, and PEN2 and the effects of these components on modulation of PS-and FADlinked PS-mediated ␥-secretase processing of APP-CTFs.