Independent Generation of Aβ42 and Aβ38 Peptide Species by γ-Secretase*

Proteolytic processing of the amyloid precursor protein by β- and γ-secretase generates the amyloid-β (Aβ) peptides, which are principal drug targets in Alzheimer disease therapeutics. γ-Secretase has imprecise cleavage specificity and generates themostabundant Aβ40 and Aβ42 species together with longer and shorter peptides such as Aβ38. Several mechanisms could explain the production of multiple Aβ peptides by γ-secretase, including sequential processing of longer into shorter Aβ peptides. A novel class of γ-secretase modulators (GSMs) that includes some non-steroidal anti-inflammatory drugs has been shown to selectively lower Aβ42 levels without a change in Aβ40 levels. A signature of GSMs is the concomitant increase in shorter Aβ peptides, such as Aβ38, leading to the suggestion that generation of Aβ42 and Aβ38 peptide species by γ-secretase is coordinately regulated. However, no evidence for or against such a precursor-product relationship has been provided. We have previously shown that stable overexpression of aggressive presenilin-1 (PS1) mutations associated with early-onset familial Alzheimer disease attenuated the cellular response to GSMs, resulting in greatly diminished Aβ42 reductions as compared with wild type PS1. We have now used this model system to investigate whether Aβ38 production would be similarly affected indicating coupled generation of Aβ42 and Aβ38 peptides. Surprisingly, treatment with the GSM sulindac sulfide increased Aβ38 production to similar levels in four different PS1 mutant cell lines as compared with wild type PS1 cells. This was confirmed with the structurally divergent GSMs ibuprofen and indomethacin. Mass spectrometry analysis and high resolution urea gel electrophoresis further demonstrated that sulindac sulfide did not induce detectable compensatory changes in levels of other Aβ peptide species. These data provide evidence that Aβ42 and Aβ38 species can be independently generated by γ-secretase and argue against a precursor-product relationship between these peptides.

Proteolytic processing of the amyloid precursor protein by ␤and ␥-secretase generates the amyloid-␤ (A␤) peptides, which are principal drug targets in Alzheimer disease therapeutics. ␥-Secretase has imprecise cleavage specificity and generates the most abundant A␤40 and A␤42 species together with longer and shorter peptides such as A␤38. Several mechanisms could explain the production of multiple A␤ peptides by ␥-secretase, including sequential processing of longer into shorter A␤ peptides. A novel class of ␥-secretase modulators (GSMs) that includes some non-steroidal anti-inflammatory drugs has been shown to selectively lower A␤42 levels without a change in A␤40 levels. A signature of GSMs is the concomitant increase in shorter A␤ peptides, such as A␤38, leading to the suggestion that generation of A␤42 and A␤38 peptide species by ␥-secretase is coordinately regulated. However, no evidence for or against such a precursor-product relationship has been provided. We have previously shown that stable overexpression of aggressive presenilin-1 (PS1) mutations associated with early-onset familial Alzheimer disease attenuated the cellular response to GSMs, resulting in greatly diminished A␤42 reductions as compared with wild type PS1. We have now used this model system to investigate whether A␤38 production would be similarly affected indicating coupled generation of A␤42 and A␤38 peptides. Surprisingly, treatment with the GSM sulindac sulfide increased A␤38 production to similar levels in four different PS1 mutant cell lines as compared with wild type PS1 cells. This was confirmed with the structurally divergent GSMs ibuprofen and indomethacin. Mass spectrometry analysis and high resolution urea gel electrophoresis further demonstrated that sulindac sulfide did not induce detectable compensatory changes in levels of other A␤ peptide species. These data provide evidence that A␤42 and A␤38 species can be independently generated by ␥-secretase and argue against a precursor-product relationship between these peptides.
A variety of therapeutic strategies in clinical development for Alzheimer disease (AD), 2 the most common neurodegenerative disorder, target the amyloid-␤ (A␤) peptides that are generated through proteolytic processing of the transmembrane amyloid precursor protein (APP) (1). In the A␤-producing pathway, APP is cleaved by two aspartyl proteases, first by ␤-secretase within its ectodomain and subsequently by ␥-secretase, which cleaves APP within its transmembrane domain (TMD) (2). ␥-Secretase is a multiprotein complex with the presenilin (PS) proteins at its enzymatic core (2). Because of its imprecise cleavage specificity, ␥-secretase generates A␤ peptides of variable length at the carboxyl terminus, with the highly amyloidogenic A␤42 isoform thought to be the key pathogenic species (3). A central role of A␤42 developed largely from genetic research demonstrating that mutations in the APP and PS genes associated with early-onset familial AD (FAD) invariably increase the A␤42/A␤40 ratio in primary fibroblasts and plasma of affected individuals, in transfected cells, and in transgenic animals (3).
In contrast to pan ␥-secretase inhibitors that prevent all ␥-secretase-mediated cleavage events in APP and other substrates, ␥-secretase modulators (GSMs) have been shown to selectively lower production of the A␤42 species without affecting other ␥-secretase substrates (3)(4)(5)(6)(7)(8). GSMs such as sulindac sulfide and ibuprofen were first discovered in the class of nonsteroidal anti-inflammatory drugs (7), and recent derivatives have shown promise as therapeutic agents in AD animal models and clinical trials (9,10). Several characteristics indicate that GSMs act directly on the ␥-secretase complex or its substrate APP, including their activity in cell-free ␥-secretase assays (4 -6, 11, 12), their ability to affect conformation of presenilin-1 (PS1) (13), and the observation that overexpression of FAD PS1 mutations altered the cellular response to GSMs, resulting in enhanced or diminished A␤42 reductions (12,14).
In addition, mass spectrometry analysis has shown that A␤42-lowering GSMs induced a concomitant increase in shorter A␤ species, primarily A␤38 (5, 7). Conversely, inverse GSMs such as fenofibrate selectively increased A␤42 production with a concomitant decrease in A␤38 and other shorter A␤ species (15). Interestingly, certain pan ␥-secretase inhibitors elevated A␤42 levels at subinhibitory concentrations, and this also correlated with a decrease in A␤38 levels (16). Finally, mutations in a GXXXG motif in the APP TMD decreased A␤42 generation and increased A␤38 levels from the mutant substrate (17). Taken together, these findings strongly indicated coordinated production of A␤42 and shorter species such as A␤38 by ␥-secretase, but no confirmation for or against interdependence between these peptides has been provided so far. Importantly, this issue is of significance not only for the mode of action of GSMs but also to understand the molecular mechanism of A␤ generation by ␥-secretase. In this respect, it has been proposed that A␤ peptides are generated by sequential cleavage of longer into shorter peptide species (16 -21). Further, FAD PS1 mutations might lower the catalytic activity of ␥-secretase, thereby reducing the turnover of A␤42 into shorter species (18,20,22,23). We have recently shown that several FAD PS1 mutations, characterized by their aggressive nature with disease onset in the second to fourth decade in life, rendered cells non-responsive to the ability of GSMs to lower A␤42 (14). Now, we have used this tissue culture model to investigate a potential precursor-product relationship between A␤42 and A␤38 peptide species. Unexpectedly, treatment with structurally divergent GSMs increased A␤38 production to similar levels in PS1 mutant cell lines as compared with wild type PS1 control cells despite the insensitivity of the mutants to A␤42 reduction. These data are inconsistent with strictly coordinated cleavages and support independent generation of A␤42 and the shorter A␤38 peptides by ␥-secretase.

EXPERIMENTAL PROCEDURES
Drugs, Antibodies, Cell Lines, and Cell Culture-The nonsteroidal anti-inflammatory drugs sulindac sulfide, ibuprofen, and indomethacin were purchased from BIOMOL (Plymouth Meeting, PA). All other chemicals were from Sigma-Aldrich. Monoclonal antibody Ab9 against amino acids 1-16 of human A␤ and the carboxyl terminus-specific A␤ antibodies BAP24, BAP15, and BAP29 have been described (24,25). Biotinylated monoclonal antibody 6E10 recognizing amino acids 1-17 of human A␤ was purchased from Signet (Dedham, MA). Generation of Chinese hamster ovary (CHO) cells with stable coexpression of wild type APP751 and wild type PS1 or the PS1 mutations PS1-P117L, PS1-L166P, PS1-G384A, and PS1-⌬Exon9 has been described previously, and comparable PS1 and APP expression in all cell lines has been demonstrated (14). All cell lines were maintained in ␣-minimum essential medium supplemented with 10% fetal bovine serum, 1 mM sodiumpyruvate, 2 mM L-glutamine, and 100 units/ml penicillin/streptomycin (Invitrogen).
Dose-response Experiments and Statistical Analysis-A␤ secretion of individual cell lines after GSM treatments was compared in dose-response experiments as described (14). All cell lines intended for comparison were cultured and treated in parallel at similar cell densities. Cells were cultured in serumcontaining medium and treated for 24 h with indicated concentrations of GSMs or Me 2 SO vehicle. A␤40, A␤42, and A␤38 levels in conditioned medium were then analyzed by sandwich immunoassay. Duplicate measurements from each drug concentration were averaged and normalized to Me 2 SO control condition. These experiments were repeated five times, and results were analyzed by one-way analysis of variance (ANOVA) with DunnettЈs post tests using GraphPad Prism software (San Diego, CA).
A␤ Liquid Phase Electrochemiluminescence Assay-A␤ levels were analyzed by sandwich immunoassay as described (14,24). In brief, the biotinylated antibody 6E10 was used as capture antibody, and carboxyl terminus-specific A␤ antibodies BAP24, BAP15, and BAP29 were labeled with TAG electrochemiluminescent label (Bioveris, Gaithersburg, MD) and used for detection. Culture media were collected following conditioning for 24 h, cell debris was removed, and complete protease inhibitor mixture (Roche Diagnostics) was added. For liquid phase electrochemiluminescence assay analysis, conditioned media were incubated for 3 h with M-280 paramagnetic beads (Invitrogen) and antibodies 6E10 and BAP24-TAG (for A␤40), BAP15-TAG (A␤42), or BAP29-TAG (A␤38). Electrochemiluminescence was quantified using an M-Series M8 analyzer (Bioveris).
Mass Spectrometry Analysis-Matrix-assisted laser desorption/ionization time of flight (MALDI-TOF) mass spectrometry of A␤ peptides was performed on a 4800 MALDI-TOF-TOF (Applied Biosystems/MDS-Sceix, Foster City, CA). A␤ peptides were immunoprecipitated from conditioned medium with Ab9 antibody covalently coupled to Seize TM beads (Pierce). Peptides were eluted from the beads with 25% 0.1% trifluoroacetic acid:75% acetonitrile. Samples were mixed 1:1 with ␣-cyano-4-hydroxycinnamic acid matrix in methanol:acetonitrile:water (36%:56%:8%) (Agilent, Santa Clara, CA) and spotted on the MALDI target. Mass spectra were acquired from m/z 3500 -5000 Da in reflector positive mode at 10,000 shots/ spectrum using single shot protection and a delayed extraction time of 420 ns. The area of the isotope pattern (isotopic cluster area) was used as a measure of apparent relative abundance and expressed as a % of total. Results from three independent experiments were averaged for statistical analysis.
Urea Gel Electrophoresis-Cells were lysed in radioimmune precipitation buffer (1% Nonidet P-40, 1% sodium deoxycholate, 0.1% SDS, 25 mM Tris-HCl, pH 7.6, 150 mM NaCl), and A␤ peptides from 1.25 mg of total protein were immunoprecipitated using 25 l of Dynabeads (Dynal Biotech, Hamburg, Germany) coated with 1E8 monoclonal antibody directed against the amino terminus of A␤ (Bayer-Schering AG). Immunoprecipitated material or conditioned supernatants were separated on 10% Bicine/Tris gels containing 8 M urea (22), and A␤ peptide species were revealed by Western blotting using antibody 1E8. Synthetic A␤ peptides of different size were run in parallel in the same gel system and under the same conditions for the identification and quantification of A␤ peptides by densitometry. Immunoreactive band intensities were quantified with the Quantity One v4.1 software (Bio-Rad). All samples were run as duplicates, and each gel carried a five-step dilution series of a synthetic A␤ peptide mix. Bands were quantified relative to this dilution series. The inter-and intra-assay coefficients were below 10%. Mean values were used for subsequent calculations.

RESULTS AND DISCUSSION
We previously demonstrated that aggressive PS1 mutations associated with early onset FAD displayed a dramatic increase in the A␤42/A␤40 ratio (14). Further, these mutants were partially non-responsive to pan ␥-secretase inhibitors in vitro and in vivo, indicating that certain AD tissue culture and animal models harboring aggressive PS1 mutations might not be suitable to assess the potency and efficacy of ␥-secretase inhibitors. In the same study, we showed that these PS1 mutations were also insensitive to the A␤42-lowering activity of GSMs (14). For these experiments, CHO cells with stable co-expression of APP and wild type PS1 (PS1-WT) or the PS1 mutants PS1-P117L, PS1-L166P, PS1-G384A, and PS1-⌬Exon9 were treated with 30 -60 M of the GSM sulindac sulfide (SS), A␤42 levels in corresponding cell culture supernatants were determined, and A␤42 secretion of individual cell lines was assessed. All cell lines expressing mutant PS1 were either completely refractory to SS (PS1-⌬Exon9, PS1-L166P) with no A␤42 reductions or showed only minor A␤42 reductions (10% reduction at 60 M, PS1-P117L, PS1-G384A), whereas PS1-WT control cells displayed robust dose-dependent reductions in A␤42 levels (40% at 60 M) (14). We now took advantage of this culture system to investigate a potential precursor-product relationship between A␤42 and A␤38 peptides by measuring A␤38 levels in culture supernatants from the same experiments. If A␤38 production by ␥-secretase were strictly, inversely coupled to A␤42 production or were dependent on A␤42 turnover, then treatment of PS1 mutant cell lines with GSMs should also result in greatly diminished A␤38 elevations as compared with cells expressing PS1-WT. However, we found that sulindac sulfide treatment robustly increased A␤38 levels in a dose-dependent manner from all PS1 mutant cell lines without any significant difference as compared with PS1-WT control cells ( Fig. 1 and Table 1). To exclude a compound-specific effect, we repeated these experiments with the structurally divergent GSM ibuprofen, which belongs to the aryl propionic acid class of non-steroidal anti-inflammatory drugs (Fig. 2). Cell lines expressing PS1-WT or mutant PS1 were treated with 250 M ibuprofen, and A␤40, A␤42, and A␤38 levels in culture supernatants were measured and compared by one-way ANOVA. At this concentration, ibuprofen did not cause toxicity in CHO cells (7) and no significant reductions in A␤40 levels were observed (data not shown). In PS1-WT control cells, ibuprofen induced a strong reduction in A␤42 levels whereas all cell lines expressing PS1 mutants displayed a significantly attenuated response ( Fig. 2A and Table 2). In contrast, comparable with our findings with sulindac sulfide, A␤38 levels were elevated by ibuprofen treatment in PS1 mutant cell lines indistinguishable to PS1-WT control cells (Fig. 2B and Table 2). This was further confirmed with the GSM indomethacin (data not shown).
To examine the possibility that PS1 mutants would cause overall alterations in the pattern of A␤ peptides or that GSM treatment would induce compensatory changes in other species not detectable with our carboxyl terminus-specific A␤ antibodies, we analyzed the full spectrum of A␤ peptides FIGURE 1. Independent generation of A␤42 and A␤38 peptides by ␥-secretase in cell lines expressing aggressive PS1 mutants. CHO cells with stable co-expression of wild type APP and wild type PS1 (PS1-WT) or FAD PS1 mutants were treated with increasing concentrations of the ␥-secretase modulator sulindac sulfide or Me 2 SO vehicle, and A␤38 levels in conditioned medium were quantified by sandwich immunoassay. We have previously shown that cell lines expressing the FAD mutants PS1-⌬Exon9, PS1-P117L, PS1-L166P, and PS1-G384A were non-responsive to the A␤42-lowering activity of sulindac sulfide and did not show reductions in A␤42 levels, whereas A␤42 levels were significantly reduced in PS1-WT control cells (14). However, A␤38 levels were increased in a dose-dependent manner in PS1 mutant cell lines with no significant difference as compared with PS1-WT control cells (Table 1), indicating that A␤42 and A␤38 peptides can be generated independently by ␥-secretase. Dose-response experiments (n ϭ 5) were analyzed by one-way ANOVA with PS1-WT cells as control group. FIGURE 2. PS1 mutants are non-responsive to the A␤42-lowering activity of ibuprofen but display similar increase of A␤38 levels as compared with wild type PS1. CHO cells with stable co-expression of wild type APP and wild type PS1 (PS1-WT) or FAD PS1 mutants were treated with 250 M of the structurally divergent ␥-secretase modulator ibuprofen or Me 2 SO vehicle. A␤42 and A␤38 levels were quantified in conditioned medium, and statistical analysis was performed as described in Fig. 1. A, cells overexpressing FAD mutants displayed a strongly diminished A␤42 reduction as compared with PS1-WT control cells. B, on the contrary, A␤38 levels were increased to the same level in cells expressing PS1 mutants and PS1-WT control cells (Table 2). n ϭ 5; one-way ANOVA. **, p Ͻ 0.01, ***, p Ͻ 0.001 Dunnett's post tests.

TABLE 1 A␤38 levels are similarly increased in cell lines expressing wild type or mutant PS1 after treatment with the ␥-secretase modulator sulindac sulfide
Dose-response experiments were performed as described under "Experimental Procedures." and analyzed by one-way ANOVA with PS1-WT cells as control group. n ϭ 5. secreted by PS1-WT control cells or cells expressing PS1-L166P or PS1-⌬Exon9 by mass spectrometry. Cells were treated with 60 M SS, and tissue culture supernatants were immunoprecipitated with antibody Ab9 recognizing amino acids 1-16 of the human A␤ sequence. The immunoprecipi-tated material was then analyzed by MALDI-TOF-TOF. To determine the relative abundance of individual A␤ peptides, the area of the isotopic cluster for each peptide was analyzed (Fig. 3). We observed that this tended to be more sensitive than measuring peak height alone (data not shown). The profiles of A␤ peptides produced by cells expressing PS1 mutants have not been examined in depth previously. Interestingly, although the full range of A␤ peptides from A␤1-33 to A␤1-40 and A␤42 was produced, we observed that A␤34 levels appeared slightly higher in PS1 mutant cells as compared with PS1-WT control cells (Fig. 3A). Following SS treatment, the MALDI-TOF results also confirmed our sandwich immunoassay data, namely, cells expressing PS1 mutants displayed an attenuated response to SS with diminished A␤42 reductions as compared with PS1-WT control cells, whereas the levels of A␤38 were increased in all cell lines. More importantly, no substantial additional changes in the pattern of A␤ peptides were observed after SS treatment in PS1 mutant cells or PS1-WT control cells (Fig. 3). However, it remains plausible that A␤38 peptides could be generated by trimming of other longer species such as A␤45/A␤46 in PS1 mutant cells. In support of this possibility, replacement of the A␤40 and A␤42 cleavage sites in the APP TMD by tryptophan mutagenesis abolished A␤40 and A␤42 secretion and caused accumulation of longer A␤ peptides in cell lysates but still allowed secretion of A␤38 (26). The highly hydrophobic A␤ peptides longer than A␤42 are inefficiently secreted and cannot be analyzed by mass spectrometry (19). Consequently, we used high resolution urea gel electrophoresis to examine longer A␤ species in cell lysates. In cell supernatants, we were able to detect A␤ peptides ranging from A␤37 to A␤40 and A␤42 by urea gel electrophoresis as previously shown (22). Confirming our sandwich immunoassay and mass spectrometry results, treatment with sulindac sulfide reduced A␤42 and increased A␤38 levels in PS1-WT cells, whereas in cells expressing PS1 mutants L166P or G384A A␤38 levels were increased despite unchanged A␤42 levels (Fig. 4A). In corresponding cell lysates from the same experiments (Fig. 4B), we further detected minute amounts of longer A␤ species such as A␤44, A␤45, and A␤46, but these peptides were not differentially affected by GSM treatment. In fact, careful quantitative analysis of several independent experiments showed that  the intracellular pool of all detectable A␤ peptides, including A␤42 and A␤38, was not affected by GSM treatment in either PS1-WT or PS1 mutant cells (Fig. 4C). We have reported previously that sulindac sulfide treatment was able to lower intracellular A␤42 levels in CHO cells with stable co-expression of APP and the PS1 mutant PS1-M146L (12). However, in the same study, we showed that the PS1-M146L mutation strongly enhances the cellular response to A␤42-lowering GSMs as compared with PS1-WT (12). Furthermore, at the time, our analysis of intracellular A␤ species using less developed protocols for immunoprecipitation and Western blot detection was evidently close to the detection limit.

Sulindac sulfide
Our new findings now clearly demonstrate that sulindac sulfide treatment exclusively affects the secreted pool of A␤42 peptides. This observation in conjunction with the very low abundance of longer A␤ peptides in our cell lines argues against the possibility that the increased A␤38 levels in PS1 mutant cells after GSM treatment can be explained by enhanced turnover of intracellular A␤ peptides longer than A␤42. In conclusion, these data demonstrate that A␤42 and A␤38 peptides can be generated independently by ␥-secretase and that the production of these peptides is not constrained by a stringent precursorproduct relationship. How do our observations conform to the current models of A␤ generation by ␥-secretase? After ectodomain shedding of APP by ␤-secretase, the remaining membrane-bound fragment becomes a substrate for ␥-secretase and is cleaved at multiple sites within its TMD (Fig. 5). The abundance of proteolytic products indicates that predominant ␥-secretase cleavages occur after Val-40 (generating A␤40), after Ala-42 (A␤42), and after Leu-49 (⑀-cleavage), the latter producing the APP intracellular domain, a cytosolic fragment with potential signaling functions (2). Less abundant peptides are generated by cleavage after various other residues and have been identified in cell supernatants (A␤37, A␤38, A␤39) or cell lysates (A␤43, A␤45, A␤46, A␤48) (19). Evidence indicates that the APP TMD has ␣-helical conformation with 3.6 residues forming one complete turn (27,28), which would align the cleavage sites for A␤40, A␤43, A␤46, and ⑀-cleavage on one surface of the helix and cleavage sites for A␤38, A␤42, A␤45, and A␤48 on the opposite surface (Fig. 5). It remains to be clarified whether these cleavages occur simultaneously and independently of each other or FIGURE 4. Urea gel electrophoresis analysis of A␤ peptide species after treatment with the ␥-secretase modulator sulindac sulfide. A, CHO cells with stable co-expression of wild type APP and wild type PS1 (PS1-WT) or FAD mutants PS1-L166P or PS1-G384A were treated with 60 M of the ␥-secretase modulator sulindac sulfide (SS) or Me 2 SO vehicle. Cell supernatants were separated on urea SDS-gels, and A␤ peptides were detected by Western blotting with antibody 1E8. The first three lanes (M) show serial dilutions of synthetic marker peptides for the identification of matching A␤ peptide species in the cell supernatants. Peptides ranging from A␤1-37 to A␤1-40 and A␤1-42 were detected. In accordance with the sandwich immunoassay and mass spectrometry results, cells expressing PS1 mutants displayed diminished A␤42 reductions in cell supernatants after SS treatment as compared with PS1-WT cells, whereas A␤38 levels were increased in cell supernatants of both PS1-WT and PS1 mutant cells. Other detectable A␤ peptide species, A␤37 and A␤39, were unchanged by SS treatment. One of three representative experiments is shown. B, analysis of A␤ peptide species in corresponding cell lysates of the same experiment. A␤ peptides were immunoprecipitated from cell lysates with antibody 1E8, the immunoprecipitated material was separated by urea gel electrophoresis, and A␤ peptides were detected by Western blotting with antibody 1E8. In addition to the A␤ peptide species detected in cell supernatants, cell lysates contained minute amounts of peptides longer than A␤42, such as A␤44, A␤45, and A␤46. Levels of these peptides were not affected by SS treatment in either PS1-WT or PS1 mutant cells. The lane on the right side shows separation of a mixture of synthetic peptides ranging from A␤41 to A␤49. The urea gel system did not allow clear separation of A␤41, A␤42, and A␤43 or peptides A␤46, A␤47, A␤48, and A␤49. The lower panel shows a longer exposure of the same gel for the area containing peptides ϾA␤42. C, densitometric quantification of A␤ peptide species from three independent experiments demonstrated that the intracellular pool of all detectable A␤ peptides, including A␤42 and A␤38, was unchanged by SS treatment in either PS1-WT or PS1 mutant cells. After ectodomain shedding of APP by ␤-secretase, which generates the amino terminus of A␤ (Asp-1), ␥-secretase cleaves at multiple sites within the APP transmembrane domain (light gray box). Major cleavages occur after Val-40 (generating A␤40), after Ala-42 (A␤42), and after Leu-49 (⑀-cleavage), generating the APP intracellular domain. Less abundant peptides have been identified in cell supernatants (A␤37, A␤38, A␤39) or cell lysates (A␤43, A␤45, A␤46, A␤48). ␣-Helical models of the APP TMD align the cleavage sites for A␤40, A␤43, A␤46, and ⑀-cleavage on one surface of the helix and cleavage sites for A␤38, A␤42, A␤45, and A␤48 on the opposite surface. Cleavages may occur simultaneously and independently of each other. Alternatively, cleavages may occur sequentially at every 3-4 residues along the ␣-helical surface. Our data indicate that A␤42 and A␤38 peptides can be generated independently by ␥-secretase and that the production of these peptides is not defined by a precursor-product relationship.
whether A␤ peptides are generated by sequential trimming of longer into shorter species (18,20). If the cleavages occur independently of each other, then FAD-associated mutations in PS or APP or GSM treatment might simply result in subtle changes in substrate presentation to the active site of ␥-secretase (3). In case of a ␥-secretase complex containing PS1-WT, GSM treatment might favor exposure of the peptide bond at the A␤38 cleavage site at the expense of production of the longer A␤42 peptide. In contrast, aggressive PS1 mutations that are nonresponsive to GSMs might confer drastic conformational changes on the ␥-secretase that are only partially reversible by GSMs, resulting in increased A␤38 production in the absence of corresponding A␤42 reductions. On the other hand, several groups have provided strong evidence that A␤ peptides might be generated by sequential proteolytic cleavage, with ⑀-cleavage occurring first and subsequent cleavages taking place at every 3-4 residues along the ␣-helical surface (16,17,19,21,26). In this model, and in accordance with findings that many FAD PS mutations seemed to reduce the catalytic activity of ␥-secretase (22,23), ␥-secretase complexes containing mutant PS might be more prone to release A␤42 from the active site before further trimming to A␤38 (18,20). Conversely, GSMs may strengthen the substrate-enzyme interaction, thereby enhancing the turnover from A␤42 to A␤38. Our findings that production of A␤42 and A␤38 peptides can be uncoupled favors the first model of independent cleavages by ␥-secretase. However, our results do not refute the concept of A␤ generation by sequential ␥-secretase cleavage. Our tissue culture model using PS1 mutants and GSM treatment did only allow examining the relationship between A␤42 and A␤38 species. Furthermore, we cannot exclude that the presence of mutant PS1 in the ␥-secretase complex somehow uncouples A␤42 and A␤38 production. During preparation of this report, we learned that another laboratory had also observed uncoupling of A␤42 and A␤38 generation after GSM treatment in cell lines expressing mutant PS1 and in an AD transgenic mouse model expressing mutant PS2 (29). Taken together, these and our findings suggest that a pure sequential cleavage model could be an oversimplification and that the mechanism of A␤ generation by ␥-secretase might be even more complex than previously assumed. Moreover, it raises the possibility that selective modulators of A␤42 production might exist that do not cause induction of shorter A␤ peptides. This is also supported by the observation that the inverse GSM celecoxib did not affect A␤38 levels (15). As a consequence, reliance on A␤38 elevation as more easily detectable, surrogate readout could lead to false negative results in drug screenings for A␤42-lowering GSMs.