The Amyloidogenic Pathway of Amyloid Precursor Protein (APP) Is Independent of Its Cleavage by Caspases*

Amyloid b -protein (A b ) is the main constituent of senile plaques in Alzheimer’s disease and is derived by proteolysis from the amyloid precursor protein (APP). Generation and secretion of both A b 40 and A b 42 isoforms depend largely on internalization of APP and occurs mainly in the endocytic pathway. Evidence has also been presented (Gervais, F. G., Xu, D., Robertson, G. S., Vaillancourt, J. P., Zhu, Y., Huang, J., LeBlanc, A., Smith, D., Rigby, M., Shearman, M. S., Clarke, E. E., H., L. H. T., Ruffolo, S. C., Thornberry, N. A., Xanthoudakis, S., Zamboni, R. J., Roy, S., and Nicholson, D. W. (1999) Cell, 97, 395–406) that caspase cleavage of APP at its cytosolic tail affects its processing such that it is redirected to a more amyloidogenic pathway, result-ing in enhanced A b generation. However, caspase cleavage of APP also results in loss of its endocytosis signal (YENP), an event that would predict a decline in internalization and a concomitant decrease, not an increase, in A b generation. In the present work, we examined whether caspase cleavage of APP is relevant to amyloidogenesis. We found that

Alzheimer's disease is accompanied by deposition of the amyloid peptide A␤ in senile plaques and cerebral blood vessels. Two major species of amyloid ␤-protein (A␤), differing by two amino acids in length (A␤-(1-40) and A␤- ) at the C terminus, have been characterized. A␤-(1-42), the longer A␤ isoform, readily aggregates in vitro and appears to be the more amyloidogenic and hence pathogenic species (reviewed in Ref. 1). Although the precise mechanism by which A␤ is generated from its precursor, the amyloid precursor protein (APP), 1 is not well understood, internalization of APP from the cell surface with subsequent processing in the endocytic pathway is a major route for generation and secretion of both A␤ isoforms (2)(3)(4). Accordingly, deletion or site-directed mutagenesis of the endocytosis motif in APP abrogates secretion of A␤-(1-40) and A␤-(1-42) (3).
Two reports have shown that neurons undergoing apoptotic cell death secrete approximately 2-to 3-fold more A␤ than healthy neurons (5,6). Because A␤ is neurotoxic and contributes to apoptosis in a variety of cultured cells, it has been proposed that increased A␤ secretion, due to genetic predisposition or to other factors, causes increased cell death in susceptible neurons. This initiates a cycle in which dying neurons release more A␤, which in turn causes more cell death to account for death of neurons seen in Alzheimer's disease (6,7).
A number of laboratories have recently demonstrated that APP can be cleaved in the cytoplasmic domain by caspases after the aspartate residue at position 664, Val-Glu-Val-Asp 664 2Ala, (APP 695 numbering, or Asp 720 using APP 751 numbering) (7)(8)(9)(10)(11)(12). This cleavage would generate a C-terminal-truncated APP molecule that is ϳ3.5 kDa shorter (APP⌬C31) (7,12). The consequences of this cleavage event in the C-terminal domain, as well as those in the lumenal domain, are unclear. One hypothesis suggests that following cleavage in the cytoplasmic region, a toxic fragment, coined C31, is released and contributes to neuronal death (12). Alternatively, it has been proposed that caspase cleavage of APP directly contributes to the increased A␤ secretion seen in apoptotic cells (7). Evidence for this hypothesis was obtained from B103 neuroblastoma cells where expression of APP⌬C31 resulted in a ϳ5-fold increase of A␤ secretion. Because caspase activation and hence cleavage of APP follows an apoptotic signal, this model is consistent with the observation that more A␤ is secreted from neurons undergoing apoptosis. A corollary to this model is that increased A␤ secretion may induce further neuronal apoptosis, thereby propagating this cycle of caspase cleavage, A␤ release, and apoptosis. Although provocative, this hypothesis is inconsistent with previous observations that the APP cytoplasmic region contains a tetrapeptide motif ( 683 YENP 686 ) that functions as a signal for endocytosis (3). Accordingly, loss of this endocytic signal by deletion or alanine mutagenesis abrogates processing of cell surface APP in the endocytic pathway and severely impairs A␤ production and secretion. Because caspase-cleaved APP lacks the endocytic signal, the truncated APP⌬C31 should in theory release less, rather than more, A␤ unless cleavage at this particular position leads to an APP molecule that behaves differently from other internalization deficient mutants.
In the present study, we have addressed this issue by characterizing both A␤ secretion and internalization rates, from a mutant that is not recognized by caspases (APP(D664A)) and from APP⌬C31, in cells of both neural and non-neural origin. Our previous studies did not specifically examine APP cleaved at this position (APP 664 ), and it is possible that caspase cleavage, as suggested by the earlier report, alters APP processing differently. In addition, we explored whether caspase activation modulates A␤ secretion. We show, in agreement with the endocytosis-requirement hypothesis, that APP⌬C31 is not efficiently internalized and, accordingly, A␤ secretion decreases dramatically. Conversely, A␤ secretion from APP(D664A), the APP mutant that is not a caspase substrate, remains unaltered. Finally, we show that caspase cleavage of wild type APP does not result in increased A␤ secretion. Thus, caspases are unlikely to contribute to the generation and secretion of A␤ peptide under these culture conditions.

MATERIALS AND METHODS
Cell Lines-Chinese Hamster Ovary (CHO) and B103 (rat neuronalderived) cells were maintained in Dulbecco's modified Eagle's medium containing 10% fetal calf serum (Life Technologies, Inc.). CHO lines expressing wild type APP 751 or truncated forms lacking the cytosolic tail (APP⌬C) or the C-terminal 31 amino acids (APP⌬C31, equivalent to caspase-cleaved APP) have been described (2,3,12). APP 751 lacking the C-terminal 15 amino acids (APP⌬C15) was generated by standard polymerase chain reaction mutagenesis methods and transfected into CHO cells with Fugene-6 reagent (Roche Diagnostics Corp.). Clones with similar levels of transgene expression were used.
B103 cells expressing wild type APP 751 , APP 695 , or C-terminal-truncated forms APP⌬C, APP⌬C15 , and APP⌬C31 were generated by transfection with Fugene-6 reagent (Roche Diagnostics Corp.) and clones with similar levels of transgene expression were used.
Antibodies-The following antibodies used have been previously described: monoclonal antibodies 5A3 and 1G7 against the mid-region of APP (2) and 26D6 against amino acids 1-12 of A␤ (12). Polyclonal antibodies included R3134 generated against residues 22-41 of A␤ (13) and ␣664 directed against amino acids 657 to 664 (by 695 numbering) of APP. This latter antibody is end-specific and only recognizes APP when it has been cleaved after residue 664 (see Fig. 1). 2 Measurement of A␤ Secretion-CHO or B103 cells expressing comparable amounts of the indicated APP constructs were seeded in duplicate on six-well tissue culture plates at an initial confluence of ϳ80%. Cells were incubated for 24 h, medium (900 l) was harvested and immunoprecipitated with antibody R3134 (1:200 dilution), and samples were separated on 4 -12% Bis-Tris gradient gels (Novex, San Diego, CA). A␤ bands were visualized on immunoblots using antibody 26D6 in conjunction with ECL (Pierce) either on film or by phosphorimaging (Bio-Rad, Hercules, CA) for quantitation. Cell lysates were immunoblotted in parallel to monitor APP expression. Measurements of A␤ and fragment p3 levels from [ 35 S]methionine metabolically labeled cells were performed as described (2). ELISA measurements of A␤-(1-40) and A␤-(1-42) in N2a cells were performed as described (3).
Internalization Assay-Triplicate cultures of CHO or B103 cells were grown to confluence in six-well tissue culture plates. Internalization of cell surface APP was then performed as described (3). Briefly, iodinated whole 1G7 monoclonal antibody (of a specific activity of ϳ3-6 Ci/g) was diluted in binding medium (BM; RPMI 1640 supplemented with 20 mM Hepes ϩ 0.2% bovine serum albumin), applied to triplicate confluent layers of CHO cells and incubated at 37°C for 30 min. Cells were then chilled on ice and washed once with BM and extensively with ice-cold phosphate-buffered saline. 1G7 antibody bound to cell-surface APP was uncoupled by two 5-min washes with ice-cold phosphatebuffered saline, pH 2.5, followed by cell lysis with 0.2 M NaOH. Radioactivity from the pooled acid washes and the cell lysates was determined in a ␥ counter. The ratio of radioactivity of acid-resistant to acid-labile fractions represents a measure of internalized versus cellsurface pool of APP.
Serum Withdrawal and APP Cleavage-B103 cells stably expressing APP, APP(D664A), and APP ⌬C31 were cultured to 70% confluence in six-well plates. The medium was then replaced with 500 l/well of Dulbecco's modified Eagle's medium without serum for 18 h. The cells were then lysed and analyzed for APP levels and APP cleavage by caspases, and medium was collected and analyzed for A␤ secretion.

Caspase-mediated Proteolysis of APP Abrogates
Internalization from the Cell Surface-Endocytosis of APP truncated after amino acid Asp 664 , APP⌬C31, corresponding to the product of caspase cleavage, was assessed using a well established protocol in which internalization of APP is monitored by uptake of radioiodinated 1G7 antibody (against the mid-region of APP; see diagram in Fig. 1) (2,20). In agreement with previous reports, APP mutants lacking either the complete cytosolic tail (APP⌬C) or the C-terminal 15-amino acids (APP⌬C15), both of which lack the YENP endocytosis motif (3), were unable to internalize efficiently in CHO cells (Fig. 2, CHO panel). The YENP motif is also absent in the APP form truncated at the caspase cleavage site (APP⌬C31; see Fig. 1). Accordingly, its internalization is also severely impaired (Fig. 2). Identical experiments performed using B103 cells (Fig. 2, B103 panel) showed that neither APP⌬C15 nor APP⌬C31 could be efficiently endocytosed in these cells. The decrease in internalization of APP endocytosis mutants was less pronounced in B103 than in CHO cells. It should be noted, however, that constitutive APP internalization from B103 cells is less efficient (59.3% Ϯ 4.05 of total APP) than from CHO cells (82.3% Ϯ 1.26%), indicating that the level of internalization of APP is different between cell lines. Nevertheless, our results clearly show that deletion of the YENP motif in APP significantly impairs its internalization, both in CHO and B103 cells.
A␤ Secretion Fully Depends on the Presence of the YENP Motif in APP-It has been shown that secretion of both A␤-(1-40) and A␤-(1-42) tightly correlates with internalization of APP (3). Therefore, we anticipated that secretion of A␤ would be impaired in cells expressing APP⌬C31 because endocytosis is impaired by loss of the internalization signal. To test this hypothesis, we measured total A␤ levels in conditioned medium from CHO cells expressing wild type or endocytosisdeficient, C-terminally truncated APP mutants. As expected, A␤ secretion from APP⌬C and APP⌬C15 (Fig. 3A) was reduced dramatically. Significantly, APP⌬C31 showed levels of secreted A␤ comparable with those from APP⌬C and APP⌬C15 (Fig.  3A), indicating that, as with other C-terminally truncated APP forms, absence of the YENP motif in APP⌬C31 resulted in a drastic reduction in A␤ secretion. Quantitation analysis revealed a decline of ϳ80% in cells expressing all endocytosis mutants, consistent with previous reports (3) (Fig. 3B). Diagram is not to scale. Caspases recognize and cleave the sequence 661 VEVD2A 665 after the aspartic acid residue to generate an Nterminal APP molecule of 664 amino acids (APP⌬C31) and a C-terminal fragment of 31 residues (C31). Represented are the epitopes recognized by the antibodies used in the present study: R3134, A␤-(22-41); 26D6, A␤-(1-12); 1G7, epitope in the extracellular region of APP. Note that the 682 YENP 686 endocytosis motif is absent from the caspasecleaved APP.
In contrast with these results, a recent report showed that B103 neuroblastoma cells secreted significantly higher amounts of A␤ when expressing APP⌬C31 as compared with cells expressing wild type APP (7). One obvious explanation for the apparent discrepancy is the use of different cell lines. To determine whether this was the case, we also measured A␤ secretion from B103 cells expressing either wild type or endocytosis-deficient APP 695 forms. As shown (Fig. 4, lanes 1-3), the level of A␤ from medium of B103 cells expressing APP⌬C31 was severely impaired, again demonstrating that cleavage of APP at the Asp 664 caspase site behaved similarly to other internalization deficient mutants (2, 3, 18).
It has previously been shown (2) that cells expressing endo-cytosis-deficient APP secrete higher levels of the peptide generated by ␣ and ␥ secretases, known as the p3 fragment. Thus, it is possible that the apparent increase in secreted A␤ levels from APP⌬C31 seen by Gervais et al. (7) may have corre- Internalization of wild type and C-terminally truncated APP forms. Loss of the endocytosis motif YENP in APP⌬C, APP⌬C15, and APP⌬C31 results in poor rates of internalization in B103 and CHO cells. Uptake of radioiodinated 1G7 antibody against the middle portion of APP was measured as described under "Materials and Methods", and internalization rates were expressed as the ratio of intracellular to surface radioactivity (average of three experiments performed in triplicate Ϯ S.E.).
FIG. 3. Decreased A␤ secretion from caspase-cleaved APP. A␤ release was monitored from wild type APP and from endocytosis-deficient APP⌬C, APP⌬C15, and APP⌬C31 in CHO cells. A, cells were cultured in collection medium for 24 h and A␤ levels were assessed by immunoprecipitation using antibody R3134 followed by Western blotting using antibody 26D6 and ECL detection. Secretion of A␤ decreased similarly from all endocytosis-deficient APP forms. Note that the truncated mutants separated slightly farther than wild type APP. A representative experiment is shown. B, quantitation of A␤ levels from two experiments. Shown are averages Ϯ S.D.

FIG. 4. Decreased A␤ secretion from endocytosis-deficient APP forms in B103 cells.
A␤ secretion is impaired from cells expressing APP⌬C15 and APP⌬C31 forms but not from cells expressing APP(D664A), a form that cannot be cleaved by caspases. Cells were cultured for 36 h in collection medium, and A␤ levels were assessed by immunoprecipitation with polyclonal antibody R3134 followed by Western blotting with monoclonal antibody 26D6. A representative experiment is shown. sponded instead to an increase in p3 levels. The antibody used by these authors was directed against the sequence A␤-(x-40) thus potentially recognizing p3 although further details were not provided (7). Furthermore, these authors tested APP⌬C31 derived from APP 751 (7), whereas we tested APP⌬C31 derived from APP 695 in the same cells (Fig. 4), thus precluding a direct comparison between both results. We addressed both issues by measuring directly the levels of A␤ and p3 secreted from cells metabolically labeled with [ 35 S]methionine. Fig. 5 shows that both B103 and CHO cells expressing APP⌬C31 secreted higher levels of p3 concomitantly with a decrease in A␤, similarly to other endocytosis mutants including the point mutation Y738A, which disrupts the YENP motif and results in decreased internalization (2,3,18). Furthermore, the pattern of A␤ and p3 secretion from APP⌬C31 in B103 cells was identical whether derived from APP 695 or from APP 751 (Fig. 5, B103  panel), further confirming that the Kunitz protease inhibitor domain of APP is not involved in the regulation of APP endocytosis and therefore has little influence on secretion of A␤ and p3 fragments. Thus, our results showed that caspase cleavage of APP resulted in elevated levels of p3, as well as in reduced secretion of A␤, independently of either the APP isoforms or the cell line.
Removal of the Asp 664 Caspase Site in APP Does Not Affect A␤ Secretion-Our results so far demonstrated that caspase cleavage at the Asp 664 site did not result in a more amyloidogenic processing of APP, as has been hypothesized (7). Rather, it appeared that A␤ secretion from caspase-truncated APP follows the well established model in which A␤ secretion is closely correlated with the rate of internalization of APP from the cell surface. To confirm the preceding finding, we examined the role of caspase cleavage of APP in A␤ secretion using a mutated form of APP that cannot be cleaved by caspases (APP(D664A)). We reasoned that, if cleavage at the Asp 664 site is involved in the generation of A␤, removal of the caspase site should result in decreased A␤ levels. As shown in Fig. 4, this was not the case. No difference in A␤ secretion was found between cells expressing wild type APP and those expressing APP(D664A) in B103 (Fig. 4, lanes 1, 4). Moreover, ratios of A␤42/total A␤ measured in N2a cells were not altered after expression of the APP(D664A) mutant (Fig. 6) whether derived from wild type APP or the familial Alzheimer's disease APPV642F mutation, the latter selectively increasing the level of A␤42.
A␤ Secretion Is Independent of Caspase Cleavage of APP-To further test a possible role of caspase activation in the generation of A␤, we induced apoptosis in B103 cells expressing wild type APP, APP(D664A), or APP⌬C31 by serum withdrawal. In B103 cells, serum withdrawal led to caspase activation and caspase cleavage of APP to generate APP⌬C31 (Fig. 7, middle panel). Polyclonal antibody ␣664 directed against amino acids 657-664 (by 695 numbering; 713-720 by 751 numbering) of APP was used to monitor caspase cleavage. This antibody constitutes an ideal tool to monitor levels of caspase-generated APP because it is end-specific and only recognizes APP when it has been cleaved after residue 664 (see Fig. 1). 2 As expected, serum withdrawal led to caspase cleavage of wild type APP but not of APP(D664A), indicating that caspases were activated in response to the treatment and that APP(D664A) mutant was not a substrate for caspase cleavage. Levels of A␤ were not significantly different in cells expressing APP or APP(D664A), whereas levels from APP⌬C31-expressing cells were dramatically reduced (Fig. 7, lower panel, lanes 1, 3, 5), as expected from inefficient APP internalization (Fig. 2). Upon serum starvation, the level of A␤ in medium was unchanged as compared with basal conditions (Fig. 7, lower panel, lanes 2, 4, 6), regardless of whether APP was a substrate for caspases (i.e. wild type APP) or not, i.e. APP(D664A). Thus, the levels of A␤ mirrored those of the APP forms susceptible to internalization, i.e. steady state levels of wild type APP and APP(D664A) (Fig. 7,  upper panel), and not those of caspase cleavage of APP. DISCUSSION A wide variety of cell lines, including neurons, undergo apoptosis when exposed to A␤ peptides, whereas apoptotic cells in turn have been shown in some cases to generate enhanced levels of A␤. This has led to the hypothesis that elevated levels of A␤ seen in both sporadic and familial Alzheimer's disease contribute to cell death, which in turn secretes more A␤ leading to a propagating cycle of cell death (6,7). Whether enhanced A␤ secretion in response to cell insult is a common phenomenon or occurs in vivo is unknown. It was recently proposed that caspase-3 cleavage of APP at amino acid Asp 664 generates a truncated APP molecule (APP⌬C31) that is then processed in a more amyloidogenic pathway, an intriguing notion that feeds into the amyloid/cell death hypothesis (7). This idea is supported by the findings that B103 cells expressing APP⌬C31 secrete significantly higher amounts of A␤ than control cells expressing wild type APP. However, these observations are in apparent contradiction with the hypothesis that the endocytic compartment is involved in generation and secretion of both A␤-(1-40) and A␤- (1-42) (2, 3). Because caspase-cleaved APP FIG. 5. Increased p3 secretion from endocytosis-deficient APP forms. B103 panel, expression of APP⌬C31 derived from APP 695 or APP 751 in B103 cells results in up-regulated p3 levels concomitantly with a decrease in A␤ secretion. CHO panel, similarly, p3 levels increase, and A␤ levels decrease in CHO cells expressing endocytosisdeficient APP⌬C31 or APPY738A forms but not wild type APP or caspase cleavage-deficient APP(D664A). Immunoprecipitated A␤ and p3 fragments from [ 35 S]methionine-labeled cells were visualized by phosphorimaging after electrophoresis separation as described under "Materials and Methods". lacks the endocytosis motif YENP, one would predict that caspase cleavage of APP would result in impaired internalization of APP and therefore in significantly lower levels of A␤ secretion from the truncated APP precursors.
In this study, we explored the role of caspase cleavage of APP on A␤ generation by analyzing internalization of cell surface APP and levels of A␤ in both caspase cleaved APP and by the non-cleavable mutant APP(D664A). We showed that 1) caspase cleavage of APP resulted in impaired internalization of the C-terminal-truncated APP molecule, similar to other YENPTY trafficking APP mutants; 2) as with other endocytosis-deficient APP forms, A␤ secretion from caspase-cleaved APP was drastically reduced, whereas secretion of the p3 fragment, generated by ␣ and ␥ secretases, was increased; 3) masking of the caspase site in APP did not affect A␤ levels; and 4) caspase activation in cells did not increase A␤ secretion. Importantly, these results were observed in cell types of both neural and non-neural origin.
Thus, loss of the C-terminal 31 residues, which contain the internalization signal in the APP cytosolic domain, inhibits internalization and subsequent A␤ generation. In other words, caspase cleavage at position 664, generating APP⌬C31, does not result in an APP molecule that behaves differently from the other endocytic mutants with regards to either APP processing or A␤ and p3 production. Furthermore, one form of cell injury by serum withdrawal did not correspond to increased levels of A␤ in B103 cells (Fig. 7) despite the fact that APP was cleaved by activated caspases (Fig. 7) (12). Thus, our results clearly showed that generation and secretion of A␤ can be uncoupled from caspase activation and caspase cleavage of APP. This interpretation is also consistent with the findings from the APP caspase mutant, APP(D664A), where loss of the caspase cleav-age site did not alter the amount of A␤ secretion (Figs. 4 and 7). Thus, our results do not support the hypothesis that caspase cleavage of APP results in more A␤ secretion, leading to further cell death and even more A␤ secretion.
In this study, we were careful to examine not only CHO cells, as we have analyzed in the past, but also the same B103 neuroblastoma cell line used in both studies as well as N2a cells. We used APP 695 and APP 751 and both deletion (⌬C, ⌬C15, and ⌬C31) and single point (Y738A) mutants to demonstrate a correlation between internalization of APP and A␤ secretion. At present we cannot explain the differences between our findings and those reported by Gervais et al. (7). One possibility, however, may reside in the method chosen to measure A␤ levels. While we directly visualized A␤ as well as the fragment generated by ␣ and ␥ secretases, known as the p3 fragment (2) (see diagram in Fig. 1 and Fig. 5), Gervais et al. made use of an antibody directed against the sequence A␤-(x-40), which would potentially recognize p3. Because the method used (ELISA) may not be able to differentiate between p3 and A␤ fragments, and no evidence was presented to the contrary, it is possible that what Gervais and colleagues interpreted as an increase in A␤-(1-40) from APP⌬C31 corresponded in fact to an increase in p3 generation, as we have shown in the present work.
In summary, our studies provided evidence that caspases are unlikely to be involved in the amyloidogenic processing of APP. In this context, it is noteworthy that caspase-6 was proposed to represent the ␤-secretase activity, especially in the case of the Swedish APP mutation, where the mutant APP was shown to be an improved caspase substrate in in vitro assays (7). This is one possible explanation for the increased A␤ production associated with this mutation, a finding that places caspases as potentially important in amyloidogenesis under pathological conditions. However, cleavage by caspase-6 generates an A␤ peptide that is truncated of the N-terminal aspartate residue because caspase cleavage occurs after rather than before the aspartate at the P1 position (14). The most compelling evidence that caspases do not contribute to A␤ generation through ␤ secretase activity is the recent reports that mice lacking BACE1 (␤-site APP cleaving enzyme-1) do not produce any measurable amounts of A␤ (15,16). Thus, the concept that caspases play an important role in amyloidogenesis, though provocative, is not substantiated by current evidence, including our present work.
Finally, it is important to note that this study did not address the question of whether some apoptotic signals may exacerbate A␤ secretion under certain circumstances, as has been reported by two laboratories (5,6). Although we did not see an increase in A␤ secretion in response to serum withdrawal and caspase activation, differences in mode of cell death induction and in the cell lines used may explain the outcomes. Clearly, it re-FIG. 7. A␤ secretion in B103 cells is independent of caspase cleavage of APP. B103 cells expressing wild type APP, APP(D664A), or APP⌬C31 were incubated in the presence or absence of serum and caspase cleavage of APP, and A␤ levels were measured as described under "Materials and Methods". Note that A␤ levels do not depend on caspase cleavage of APP. Rather, they mirror the internalization rates of the APP forms as seen in Fig. 2.   FIG. 6. Deletion of the caspase cleavage site in wild type or FAD(V642F) mutant APP does not modify A␤-(1-40) or A␤-(1-42) levels in N2a cells. Total levels of A␤, as well as the A␤42/total A␤ ratio remained unaffected after introducing the mutation D664A in both APP forms. ELISA measurements were performed as described under "Materials and Methods". Results from two experiments performed in duplicate (mean Ϯ S.D) are shown. mains possible that apoptosis may eventually lead to enhanced A␤ secretion in some cell types. However, it is also clear from our present work that caspase-cleaved APP per se is not a major contributor to A␤ generation, and therefore a scenario in which caspase-cleaved APP actively contributes to A␤ secretion remains unlikely. Thus, the possibility remains that cytotoxic effects derived from caspase cleavage of APP may result from the generation of new APP-derived C-terminal peptides (10,17,19) or C31, as we have proposed (12).