Multiple Mechanisms Underlie Neurotoxicity by Different Types of Alzheimer's Disease Mutations of Amyloid Precursor Protein*

We examined a neuronal cell system in which single-cell expression of either familial Alzheimer's disease (FAD) gene V642I-APP or K595N/M596L-APP (NL-APP) in an inducible plasmid was controlled without affecting transfection efficiency. This system revealed that (i) low expression of both mutants exerted toxicity sensitive to both Ac-DEVD-CHO (DEVD) and glutathione ethyl ester (GEE), whereas wild-type APP (wtAPP) only at higher expression levels caused GEE/DEVD-resistant death to lesser degrees; (ii) toxicity by the V642I mutation was entirely GEE/DEVD sensitive; and (iii) toxicity by higher expression of NL-APP was GEE/DEVD resistant. The GEE/DEVD-sensitive death was sensitive to pertussis toxin and was due to Go-interacting His657-Lys676 domain. The GEE/DEVD-resistant death was due to C-terminal Met677-Asn695. APP mutants lacking either domain unraveled elaborate intracellular cross-talk between these domains. E618Q-APP, responsible for non-AD type of a human disease, only exerted GEE/DEVD-resistant death at higher expression. Therefore, (i) different FAD mutations in APP cause neuronal cell death through different cytoplasmic domains via different sets of mechanisms; (ii) expression levels of FAD genes are critical in activating specific death mechanisms; and (iii) toxicity by low expression of both mutants most likely reflects the pathogenetic mechanism of FAD.

Alzheimer's disease (AD), 1 the most prevalent neurodegenerative disease, is characterized by neuronal loss and extracellular senile plaques, whose major constituent is A␤ amyloid, cleaved off from the transmembrane precursor APP (1). Genetic studies of early-onset FAD have demonstrated that structural alterations in APP cause AD. There are at least two different types of mutations reported in APP as established causes for FAD: Ile/Phe/Gly mutations at Val 642 or the Asn/Leu mutation at Lys 595 /Met 596 in APP 695 (the numbering follows Kang et al. (1)). Despite the fact that neuronal death is a central abnormality in AD, exactly how these FAD-linked mutants of APP cause neuronal death has been little understood. Multiple groups (2)(3)(4)(5)(6) have so far found that FAD-associated Val 642 mutants of APP induce death through intracellular signaling cascades in neuronal cells. We (2,7) also found that neuronal cell death by Val 642 -type FAD mutants of APP may be a potentially controllable process mediated by PTX-sensitive G protein G o and its ␤␥ subunit. Wolozin et al. (4) verified that the Val 642 -type mutant of APP induces death in PC12 cells in a PTX-sensitive manner, with the discovery that FAD-associated N141I presenilin-2 also causes PTX-sensitive death. Presenilin-2 is a member gene of the PS family, whose mutation is responsible for certain forms of FAD. One potential mechanism underlying the neurotoxicity of these FAD mutants is that neuronal death occurs by deposition of A␤, particularly A␤1-42/43, a longer version of A␤ polypeptides. Indeed, the deposition of A␤XϪ42/43 is the earliest abnormality observed in AD brains (8); A␤ polypeptides, including A␤1-42, kill neuronal cells in vitro (9,10); and cellular secretion of A␤XϪ42/43 increases by expression of FAD mutants of APP (2,(11)(12)(13). It has been reported by multiple research groups that intracellular signaling mechanisms, including oxidative stress-relevant pathways (14 -16), calpain-activated cdk5 pathways (17), and caspase-dependent pathways (18), mediate A␤ amyloid-induced neurotoxicity. Taken together, these observations suggest that countermeasures against neuronal cell death in these types of FAD and even sporadic AD might be feasible even after A␤ deposition if the countermeasures could suppress intracellular toxicity signals inside the neurons expressing the FAD genes or in the neurons exposed to A␤-related insults. Therefore, it is important to understand fully the entire body of intracellular death signals generated by the expression of ADcausative genes. The aforementioned studies also suggest that FAD mutants of both APP and presenilin-2 may cause neuronal cell death through a common mechanism.
On the other hand, it has not been determined whether the function of K595N/M596L-APP (NL-APP), another established cause of FAD, is relevant to neuronal cell death. In contrast, virus-mediated overexpression of wtAPP causes significant death in neuronal cells (19,20). The present study was thus conducted to investigate whether expression of NL-APP causes death in neuronal cells, like V642I-APP, and if so, whether both FAD mutants kill neuronal cells through the same mechanism, and in what relationship between the mutation-specific mechanisms and the toxicity of wtAPP. Here we report unexpectedly complicated potential of FAD mutants to cause neurotoxicity through different cytoplasmic domains via different sets of distinct mechanisms.

MATERIALS AND METHODS
wtAPP gene, the full-length APP 695 cDNA (2), was subcloned into pIND plasmid (Invitrogen). V642I-APP cDNA was described previously (2), and E618Q-APP was constructed by a site-directed mutagenesis, using a QuikChange Site-directed Mutagenesis Kit (Stratagene), with confirmation of generated mutations by sequencing. The sense and antisense primers used for E618Q construction were 5Ј-CTGGTGT-TCTTTGCTCAAGATGTGGGTTCGAACAAAGGC-3Ј and 5Ј-GCCTTT-GTTCGAACCCACATCTTGAGCAAAGAACACCAG-3Ј, respectively. NL-APP cDNA was provided by Dr. T. Okamoto (RIKEN, Wako, Japan). These mutant APP cDNAs were subcloned to pIND with sequence confirmation. The pIND-encoded wtAPP, V642I-APP, or NL-APP was named as pIND-wtAPP, pIND-V642I-APP, or pIND-NL-APP, respectively. EGFP cDNA was purchased from CLONTECH (pEGFP-N1), and was also subcloned to pIND (pIND-EGFP). Glutathione ethyl ester (GEE) and PTX were from Sigma and Calbiochem-Novabiochem, respectively, and Ac-DEVD-CHO was from Peptide Institute Inc. Ponasterone (Invitrogen) was employed as EcD. Vitamin E and A␤1-43 were from Wako Pure Chemicals and BACHEM, respectively. F11 cells were grown in Ham's F-12 plus 18% FBS and antibiotics. F11 cells are the hybrid of a rat embryonic day 13 primary cultured neuron with a mouse neuroblastoma NTG18. These cells are one of the best models for primary cultured neurons, exhibiting without differentiation factor treatment, a number of characteristics for primary neurons, including generation of action potentials (21). F11 cells (F11/EcR cells) overexpressing both EcR and RXR were established using the co-expression vector pVgRXR and Zeocin selection (Invitrogen). For transient transfection of the pIND plasmids, F11/EcR cells were seeded at 7 ϫ 10 4 cells/well in a 6-well plate and cultured in Ham's F-12 plus 18% FBS for 12-16 h and transfected with EcD-inducible pIND plasmids (1 g of pIND plasmids, 2 l of LipofectAMINE, and 4 l of plus reagent) in the absence of serum for 3 h. After subsequent incubation with Ham's F-12 plus 18% FBS for 12-16 h, cells were cultured with or without inhibitors in Ham's F-12 plus 10% FBS for 2 h, and EcD was then added to the media. Cell mortality was measured by trypan blue exclusion assay at 72 h after the onset of EcD treatment. Transfection efficiency was assessed with pEGFP-N1 by fluorescence microscopy. F11/EcR cells were transfected with this plasmid (1 g of plasmid, 2 l of LipofectAMINE, and 4 l of plus reagent) in the absence of serum for 3 h. After subsequent incubation with Ham's F-12 plus 10% FBS for 48 h, transfection efficiency was assessed by: (i) dividing the number of green fluorescent cells by the total cell number in the same randomly chosen fields in each transfection; and (ii) calculating the mean Ϯ S. D. of these ratios for each transfection. The mean Ϯ S.E. of independent transfections was then calculated.
For the A␤ experiment, F11/EcR cells were seeded at 7 ϫ 10 4 cells/ well in a 6-well plate or a 35 mm-dish and cultured in Ham's F-12 plus 18% FBS for 12-16 h. Cells were transfected with EcD-inducible pIND plasmids (1 g of pIND plasmids, 2 l of LipofectAMINE, and 4 l of plus reagent) in the absence of serum for 3 h, then reseeded at 1.4 ϫ 10 4 cells/well in a 24-well plate, and cultured in Ham's F-12 plus 18% FBS for 18 h. Cells were then cultured with or without 25 M A␤1-43 in Ham's F-12 plus 10% FBS. Two hours after the onset of A␤ treatment, various concentrations of EcD or equivalent volumes of EtOH were added to the culture media and cells were cultured for an additional 72 h. Cell mortality was then measured by trypan blue exclusion assay.
Trypan blue exclusion assay was performed as follows. At the termination of experiments, cells were suspended by pipetting gently, and 50 l of 0.4% trypan blue solution (Sigma) was mixed with 200 l of the cell suspension (final concentration 0.08%) at room temperature. Stained cells were counted within 3 min after the mixture with trypan blue solution. The mortality of cells was then determined as a percentage of trypan blue-stained cells in total cells. The cell mortality assessed by this method thus represents the population of dead cells in total cells, including both adhesive and floating cells at the termination of experiments. The basal death rates with or without pIND vector transfection with or without EcD treatment indicated the actual fraction of dead cells, but not artificial cell death occurred after detaching cells, as in situ staining of trypan blue-positive cells indicated the presence of similar fractions of dead cells. In all experiments shown in each figure presented in this study, we performed the experiments examining cell mortality (i) in the presence or absence of 40 M EcD without transfection and (ii) in the presence or absence of 40 M EcD with empty pIND transfection, both of which were constantly as low as the basal cell mortality in the absence of EcD with pIND-APP construct transfection (data not shown).
Immunoblot analysis of expressed APP constructs and endogenous tubulin was performed as follows. Cell lysates (20 g/lane) were submitted to SDS-polyacrylamide gel electrophoresis and separated proteins were transferred onto polyvinylidene difluoride sheets. After blocking, the blots were probed with the primary antibody (2.5 g/ml anti-APP monoclonal antibody 22C11 (Roche Diagnostics) or 1/3000 dilution of anti-␣tubulin monoclonal antibody TU-02 (Santa Cruz Biotechnology)) and 1/5000 dilution of the secondary antibody horseradish peroxidase-conjugated anti-mouse IgG antibody (Bio-Rad), followed by visualization of the immunoreactive bands by ECL (Amersham Pharmacia Biotech). The densities of the 120-kDa APP immunoreactive band and the 50-kDa tubulin band were, respectively, measured by densitometrical analysis.
All of the experiments described in this study were repeated at least three times with independent transfections and treatments, each of which yielded essentially the same result. Statistical analysis was performed with Student's t test.

EcD-dependent Expression of APP Constructs in the F11/
EcR/pIND System-We transfected F11/EcR cells with pINDencoded mutant APP cDNA driven by an integrated EcD-responsive promoter (22); about 1 day after transfection, we treated cells with EcD. In this context, the cDNA-encoding protein should express in a single cell in an EcD dose-dependent manner, uninfluenced by the variation in transfection efficiency. In addition, the transfection efficiency in F11/EcR cells with the present lipofection method was appreciably high and stable. We performed three independent transfections of F11/ EcR cells with (constitutively active promoter-driven) pEGFP-N1 plasmid, which revealed the transfection efficiency to be 68.0 Ϯ 3.1% as the mean Ϯ S.E. (Fig. 1A). As expected, when F11/EcR cells were transfected with pIND-EGFP and Seventy-two hours after EcD treatment, the cell lysate samples (20 g/lane) were submitted to immunoblot analysis using both 2.5 g/ml 22C11 (anti-APP monoclonal antibody) and 1/3000 TU-02 (anti-␣ tubulin monoclonal antibody). For each EcD concentration, the densities of the bands for ϳ120-kDa APP and ϳ50-kDa tubulin in the same lane were measured and the ratio of the APP density over the tubulin density was calculated. The fold expression was assessed by dividing the ratio in the presence of each EcD concentration with the ratio in the absence of EcD. Values indicate mean Ϯ S.D. of at least three measurements. Similar experiments were performed more than three times, each with similar results. treated with increasing concentrations of EcD, the expression of EGFP in a single cell was unidirectionally augmented, as the concentration of EcD became higher (Fig. 1B), suggesting that a single-cell expression of pIND-encoded cDNA can be controlled by altering the concentrations of EcD in this system. We also confirmed that the expression of EGFP by Յ40 M EcD was not affected by 100 M Ac-DEVD-CHO, 1 mM GEE, or 1 g/ml PTX in cells transfected with pIND-EGFP (data not shown).
We next examined whether wtAPP, V642I-APP, or NL-APP was expressed in proportion to the concentration of treated EcD (Fig. 1C). Cells were transfected with pIND-encoded APP genes and treated with EcD. The result revealed that expression of either V642I-APP or NL-APP was not linearly augmented by EcD, and reached saturation by Ͼ10 M EcD. In contrast, wtAPP was expressed proportionally to the EcD concentration under the same condition.
From the literature (2-6, 23-25), we reasoned that in this system, V642I-APP and NL-APP may be induced by EcD similarly to the induction of wtAPP, but degraded through mechanisms involving caspase activation by the FAD mutants themselves, resulting in certain balanced expression. In accord with this idea, in the presence of 100 M Ac-DEVD-CHO, an established cell-permeable inhibitor of caspases, expression of either FAD mutant became linear in relation to EcD concentrations (Fig. 1C, right). Under the same conditions, expression of wtAPP was not affected by Ac-DEVD-CHO. These data suggest that the two FAD genes in pIND, like wtAPP in the same vector, were induced in proportion to the EcD concentration, and that both FAD mutants, but not wtAPP, were degraded by activating DEVD-sensitive mechanisms. The transfection of either pIND-encoded FAD gene in the presence of Ac-DEVD-CHO or in the transfection of pIND-wtAPP in its presence or absence, treatment with 10, 20, and 40 M EcD resulted in the expression of APP immunoreactivity ϳ2.5-, ϳ4-, and ϳ7-fold, respectively, of the basal expression, indicating that 10, 20, and 40 M EcD caused ϳ1.5-, ϳ3-, and ϳ6-fold expression of the transfected APP constructs, respectively.
EcD-dependent Death in Cells Transfected with pIND-encoded FAD Genes-In F11/EcR cells, we examined whether and how robustly FAD mutants in pIND (pIND-V642I-APP and pIND-NL-APP) cause cell death by various concentrations of EcD. Fig. 2A indicates that in cells transfected with either pIND-V642I-APP or pIND-NL-APP, EcD augmented cell mortality dose dependently. In contrast, the vehicle ethanol caused no increase in cell mortality in either case. Treatment of nontransfected F11/EcR cells or vector-transfected F11/EcR cells with or without EcD resulted in low cell mortality around 10%

TABLE I
Toxic effect of A␤ amyloid on the basal death rates F11/EcR cells were transfected with (pIND transfection) or without (no transfection) empty pIND and treated with (Abeta (ϩ)) or without (Abeta (Ϫ)) 25 M A␤1-43 with 40 M EcD (EcD (ϩ)) or an equivalent volume of EtOH (EcD (Ϫ)). Cell mortality was measured by trypan blue exclusion assay 72 h after the onset of EcD treatment. The values indicate mean Ϯ S.D. of three independent experiments (in % dead cells of total cells). No  for 72 h, the basal death rate of these cells ( Fig. 2A, upper  panel). Considering that transfection efficiency was 60 -70%, it followed that induction of either V642I-APP or NL-APP by Ն10 M EcD caused death in most of the transfected cells after 72 h. Expression of wtAPP resulted in a different dose-response curve for death (Fig. 2B, red closed circles). EcD caused little death in pIND-wtAPP-transfected cells at Ͻ20 M, and stimulated death dose-dependently only at Ն20 M, and to lesser degrees than the toxicity by the same concentrations of EcD in cells transfected with either pIND-FAD mutant. These data indicate that expression of V642I-APP or NL-APP as low as endogenous APP effectively killed neuronal cells and that wtAPP was toxic at only higher levels of expression. DEVD and GEE Sensitivity of Cell Death by V642I APP or wtAPP-We next investigated whether Ac-DEVD-CHO affects cell death by either FAD gene. EcD-induced death of pIND-V642I-APP-transfected cells was greatly suppressed by 100 M Ac-DEVD-CHO (Fig. 2B, blue closed circles). Surprisingly, the dose-response curve of V642I-APP-induced death in the presence of 100 M Ac-DEVD-CHO was virtually identical to the dose-response curve of wtAPP-induced death. This was also the case with GEE, a cell-permeable antioxidant (Fig. 2C, green  closed circles). The dose-response curve of V642I-APP-induced death in the presence of 1 mM GEE was again virtually identical to its dose-response curve in the presence of 100 M Ac-DEVD-CHO and the curve of wtAPP-induced death. The EcD dependence of pIND-V642I-APP expression in the presence of 100 M Ac-DEVD-CHO was linear, equivalent to that in the presence of 1 mM GEE (data not shown), and virtually identical to that of pIND-wtAPP expression (Fig. 1C, right).
These data indicated that V642I turned on a specific death mechanism sensitive to both Ac-DEVD-CHO and GEE, and also suggested that death by wtAPP was resistant to both.
In fact, toxicity by wtAPP was totally resistant to 100 M Ac-DEVD-CHO (Fig. 2D, blue closed circles) and 1 mM GEE (Fig. 2D, green closed circles). Neither Ac-DEVD-CHO (Fig. 1C) nor GEE (data not shown) affected the EcD-dependent expression of wtAPP. These data indicate that toxicity by wtAPP was through a mechanism completely different from the toxicity stimulated by the V642I mutation.
DEVD and GEE Sensitivity of Cell Death by NL APP-Therefore, a novel possibility was raised that K595N/M596L might induce neuronal cell death through further different mechanisms. Both Ac-DEVD-CHO and GEE inhibited NL-APP-induced death, but not in the same curves as V642I-APPinduced death in the presence of either reagent (Fig. 2E). In the presence of 100 M Ac-DEVD-CHO or 1 mM GEE, death by NL-APP occurred in mutually equivalent dose-response curves. The dose-response curve of NL-APP-induced death revealed that death by low expression of NL-APP was completely suppressed by Ac-DEVD-CHO and GEE, whereas death by higher induction of NL-APP was resistant to both reagents. EcD dependence of pIND-NL-APP expression was linear and equivalent to that of pIND-wtAPP in the presence of 100 M Ac-DEVD-CHO (Fig. 1C, right) or 1 mM GEE (data not shown). These data indicate that (i) low expression of both FAD mutants caused neuronal cell death GEE/DEVD sensitively, whereas wtAPP only at higher expression caused GEE/DEVD resistant death to lesser degrees; (ii) toxicity given by V642I was entirely GEE/DEVD-sensitive; and (iii) toxicity by higher induction of NL-APP was GEE/DEVD-resistant.
Effects of Vitamin E and A␤ on Cell Death by pIND-FAD Genes-To confirm that GEE acts on reactive oxygen species to block cell death by low induction of both FAD genes, we examined the effects of vitamin E and A␤. Vitamin E acts as an antioxidant at 10 -100 M, and A␤ induces or enhances oxidative stress in neuronal cells (26 -28). As shown in Fig. 3A, 100 M vitamin E suppressed cell death by V642I-APP and NL-APP in dose-response curves virtually identical to those of V642I-APP-and NL-APP-induced death in the presence of 1 mM GEE, respectively. These data indicate that vitamin E precisely mimics the inhibitory effect of GEE. In contrast, A␤ potentiated the toxic actions of both V642I-APP and NL-APP. In cells with no transfection or with empty-pIND transfection, 25 M A␤1-43 had no effect on cell death (Table I). However, in the presence of 25 M A␤1-43, both V642I-APP and NL-APP exerted cytotoxicity at lower expression than that in the absence of A␤, and caused death at 5-10% higher rates (Fig. 3B). These data suggest that A␤ enhances oxidative stress induced by both APP mutants, consistent with the study of Lockhart et al. (29) reporting that A␤ peptide renders hippocampal neurons more sensitive to free radical attack without direct action on free radical generation.
The APP Domain Responsible for GEE/DEVD-sensitive Death-We next investigated the domains in APP responsible for each mechanism for death by V642I-APP, NL-APP, or wtAPP. A clue was that the His 657 -Lys 676 domain (Domain 20) in APP interacts directly and specifically with the PTX-sensitive G protein G o in vitro (30 -32). As shown in Fig. 4A, 1 g/ml PTX completely suppressed death stimulated by the V642I mutation (but not completely by V642I-APP), indicating that the GEE/DEVD-sensitive mechanism for death by V642I-APP is entirely PTX-sensitive. This was also the case with NL-APP (Fig. 4B). The dose-response curve of NL-APP-induced death indicated that (i) the GEE/DEVD-sensitive mechanism for death by low expression of NL-APP was totally sensitive to PTX; and (ii) the GEE/DEVD-resistant death mechanism by higher induction of NL-APP was resistant to PTX. In contrast, PTX did not affect the dose-response curve of wtAPP-induced death (Fig. 2C), indicating that (i) the observed PTX effects on death by FAD mutants were not artifacts; and (ii) wtAPPinduced death was PTX/GEE/DEVD-resistant. In the presence of 1 g/ml PTX, EcD dependence of the expression of pIND-V642I-APP or pIND-NL-APP was similar to that of pIND-wtAPP (data not shown).
We next examined the function of APP constructs lacking Domain 20: V642I-APP⌬20, NL-APP⌬20, or wtAPP⌬20 (V642I-APP⌬20 is V642I-APP lacking Domain 20, and others are similarly named). Either V642I-APP⌬20 or NL-APP⌬20 was expressed in transfected F11/EcR cells by EcD in the presence of 100 M Ac-DEVD-CHO in a manner similar to the expression of V642I-APP or NL-APP in the presence of 100 M Ac-DEVD-CHO and to the expression of wtAPP (data not shown). The dose-response curve of V642I-APP⌬20-induced death was equivalent to that of wtAPP-induced death, indicating that death stimulated by V642I was abolished by the deletion of Domain 20. A similar observation was obtained from NL-APP⌬20. The dose-response curve of NL-APP⌬20-induced death was almost identical to the dose-response curve of NL- APP-induced death in the presence of PTX (Fig. 4B). These data indicate that Domain 20 mediates PTX/GEE/DEVD-sensitive death by low expression of both FAD mutants.
We also examined the dose-response curve of wtAPP⌬20induced death, and found that wtAPP⌬20 caused DEVD-resistant death in a dose-response curve similar to that of wtAPPinduced death (Fig. 4C). wtAPP⌬20 was expressed by EcD similarly to the expression of wtAPP (Fig. 5A, inset). These data demonstrate that Domain 20 was not involved in death by wtAPP.
The APP Domain Responsible for GEE/DEVD-resistant Death-We next investigated the involvement of the extreme C-terminal M677-N695 (Domain 19) in cell death by wtAPP and FAD mutants. The dose-response curve of death by wtAPP lacking Domain 19 (wtAPP⌬C) was completely flat, and 40 M EcD did not increase mortality of pIND-wtAPP⌬C-transfected cells at all (Fig. 4A), despite the fact that 40 M EcD induced expression of wtAPP⌬C to degrees similar to those of wtAPP or wtAPP⌬20 (Fig. 5A, inset). Forty M EcD did enhance mortality of cells transfected with pIND-wtAPP or pIND-wtAPP⌬20, as observed in the prior experiments. These data demonstrate that Domain 19 mediates PTX/GEE/DEVD-resistant death by higher expression of wtAPP.
Using V642I-APP⌬C and NL-APP⌬C, we next investigated the involvement of Domain 19 in death by FAD mutants. The results, shown in Fig. 5B, revealed that (i) the dose-response curve of V642I-APP⌬C-induced death was virtually identical to the curve of V642I-APP-induced death; and (ii) V642I-APP⌬Cinduced death was totally suppressed by Ac-DEVD-CHO, even below wtAPP-induced death. These findings were also the case with NL-APP⌬C. The data thus demonstrate that (i) Domain 19 was not required for the function of Domain 20 activated by either FAD mutation; and (ii) the DEVD-resistant death by higher induction of NL-APP was due to Domain 19, because higher induction of NL-APP⌬C did not cause DEVD-resistant death at all. Combined with the data that death by higher induction of wtAPP was due to Domain 19 (see above), these findings suggest that the function of Domain 19 was activated by the NL mutation. In contrast, the dose-response curve of V642I-APP⌬20 was virtually identical to the dose-response curve of wtAPP or wtAPP⌬20 (Fig. 4, A and C), indicating that V642I little affects the function of Domain 19.
Cross-talk between Domain 20 and Domain 19 -Our next question was whether the function of Domain 19 requires Domain 20. There were two different components in the action of Domain 19: one was the basal action in wtAPP to constitutively cause low levels of death, and another was the action activated by the NL mutation, both of which were GEE/DEVD resistant. The aforementioned experiments, shown in Fig. 4C, have revealed that Domain 20 was not required for the basal function of Domain 19, because wtAPP lacking Domain 20 (wtAPP⌬20) caused death as much as wtAPP.
Unexpected results were obtained with NL-APP⌬20 (Fig. 6). As opposed to a simple speculation that death by NL-APP⌬20 may be GEE/DEVD resistant, either GEE or DEVD could completely suppress NL-APP⌬20-induced death to the level of wtAPP-or wtAPP⌬20-induced death. Treatment with 100 M Ac-DEVD-CHO did not suppress expression of NL-APP⌬20, as was the case with NL-APP (data not shown). Activation of GEE/DEVD-sensitive death by the NL mutation was thus observed at higher levels of construct induction, when NL-APP lacked Domain 20, suggesting that activation by the NL mutation of Domain 19-mediated GEE/DEVD-sensitive death did not require Domain 20. The result also indicates that in the absence of Domain 20, the NL mutation could not enhance GEE/DEVD-resistant death through Domain 19.
We further analyzed whether GEE/DEVD-sensitive death by the NL mutation through Domain 19 was sensitive to PTX. For this purpose, we examined the effect of PTX on death by NL-APP⌬20, and found that NL-APP⌬20-induced death was totally insensitive to PTX (Fig. 6). These data demonstrate that (i) the basal activity of Domain 19 to cause GEE/DEVD-resistant death was potentiated by the NL mutation, only in the presence of Domain 20; and (iii) the NL mutation also allowed Domain 19 to cause GEE/DEVD-sensitive death not through PTX-sensitive G proteins, at least in the absence of Domain 20.
Effect of E618Q Mutation on APP-induced Death-Finally, we investigated whether and how the E618Q mutation affects APP-induced death. It has been established (33-35) that E618Q-APP causes HCHWA-D, that associates with the secondary microvascular degeneration different from the AD type of neurodegeneration. As shown in Fig. 6, the dose-response curve of E618Q-APP-induced death was virtually identical to that of wtAPP-induced death. Expression of pIND-E618Q-APP was similar to that of pIND-wtAPP (Fig. 7, upper panel). In addition, E618Q-APP-induced death was totally insensitive to either 100 M Ac-DEVD-CHO or 1 mM GEE. Therefore, the toxicity of E618Q-APP was totally attributed to the DEVD/ GEE-resistant toxicity retained by wtAPP. DISCUSSION The present study indicates that NL-APP causes neuronal cell death as significantly as V642I-APP. In addition, expression of V642I-APP, NL-APP, or wtAPP by 10 M EcD was ϳ1.5-fold of endogenously expressed APP in the presence or absence of Ac-DEVD-CHO. Therefore, physiologically low expression of V642I-APP and NL-APP effectively induced cell death. Although Val 642 -type FAD mutants of APP have been reported to cause neuronal death in various systems (2)(3)(4)(5)(6), it remained unclear how high expression of FAD mutant is required for its neurotoxicity. The primary importance of this study is thus the finding that both V642I-APP and NL-APP exert neurotoxicity at physiologically low levels of expression. This study also indicates that low expression of wtAPP caused little toxicity and that only higher expression of wtAPP elicited toxicity, but to lesser degrees than those by FAD mutants. Multiple reports have shown that virus-mediated overexpression of wtAPP induces neurotoxicity (19,20). In comparison with the ability of wtAPP, quantitative evaluation of the ability of FAD mutants to induce neuronal cell death was thus necessitated.
We next examined whether the two different FAD mutations in the same APP cause cell death via the same mechanism and what relationship lies between the mutation-specific mechanisms and the basal toxicity of wtAPP. It was found that death by V642I-APP (all ranges of expression examined) and by low FIG. 8. Illustration of the suggested mechanisms for death by wtAPP, V642I-APP, or NL-APP. a, in wtAPP, Domain 20 (D20) is inert, and only Domain 19 (D19) exerts weak constitutive action to cause GEE/DEVD-resistant death at higher expression of wtAPP. b, the V642I mutation activates Domain 20, leading to GEE/DEVD-sensitive death through PTX-sensitive G proteins (PG). This mechanism occurs at low expression of V642I-APP. The basal action of Domain 19 is unaffected in V642I-APP. c, the NL mutation potently activates Domain 20, leading to GEE/DEVD-sensitive death through PTX-sensitive G proteins, as is the case with the V642I mutation. In addition, the NL mutation enhances the basal activity of Domain 19 that causes GEE/DEVD-resistant death. This activation requires higher expression of NL-APP and PG-mediated assistance of Domain 20. Higher expression of NL-APP also allows Domain 19 to cause GEE/DEVD-sensitive death not through PG, at least in the absence of Domain 20 or its function. d, the E618Q mutation neither stimulates GEE/DEVD-sensitive death at low expression of the construct nor affects basal GEE/DEVD-resistant death at higher expression. See text for details. expression of NL-APP were through a GEE/DEVD-sensitive mechanism and that death by higher induction of NL-APP was via a GEE/DEVD-resistant mechanism. While wtAPP could cause cell death only at higher expression, the toxicity was GEE/DEVD-resistant, suggesting that higher induction of NL-APP potentiates the basal toxicity of wtAPP. This notion was verified by the finding that both GEE/DEVD-resistant mechanisms were through Domain 19. In contrast to the NL mutation, the V642I mutation only activated the GEE/DEVD-sensitive death mechanism independently of the wtAPP toxicity. This study has thus clarified that (i) different FAD mutations in the same APP cause neuronal cell death through distinct sets of different mechanisms; (ii) the expression level is the critical determinant for FAD mutants to trigger specific mechanisms for death induction; and (iii) the NL mutation enhances the toxicity of wtAPP at higher expression, whereas the V642I mutation does not.
The data also revealed that GEE/DEVD-sensitive death by low expression of both FAD mutants was totally sensitive to PTX and occurred through the cytoplasmic Domain 20. In contrast, GEE/DEVD-resistant death by higher induction of NL-APP and wtAPP was due to the C-terminal Domain 19. PTX exhibited exactly the same effects as the deletion of Domain 20 from either FAD mutant, suggesting that all observed functions of Domain 20 in both V642I-APP and NL-APP are mediated by PTX-sensitive G proteins, probably G o , as suggested by prior studies (30 -32). This is consistent with the reports that PTX-sensitive G proteins mediate cell death in various systems (2, 3, 36 -43). It is also intriguing to examine whether death induced through Domain 19 is mediated by Fe65, X11, or mDab1, the known Domain 19-interacting adapters (44).
This study also unraveled the elaborate intracellular crosstalk between these two domains. The independent function of Domain 20 from Domain 19 was clearly indicated by the results that both V642I-APP⌬C and NL-APP⌬C caused DEVD-sensitive death in the same manner as observed in death by V642I-APP and NL-APP, respectively. On the contrary, the function of Domain 19 was influenced by Domain 20. In the absence of Domain 20, the NL mutation could not enhance the basal toxicity of Domain 19. Therefore, in order for the NL mutation to potentiate GEE/DEVD-resistant death through Domain 19, Domain 20 was necessary. This necessity was attributable not to a conformational defect that the Domain 20 deletion may potentially cause, but to functional interference, because PTX treatment reproduced the effect of Domain 20 deletion. Instead, when Domain 20 was absent or its function was inhibited by PTX, we observed that the NL mutation allowed Domain 19 to activate GEE/DEVD-sensitive death in a PTX-resistant manner. There are two possibilities about the occurrence of this activation mechanism. One is that higher expression of NL-APP activates this mechanism even in the presence of Domain 20; but this mechanism is concealed by Domain 20-mediated toxicity saturated by lower expression of NL-APP. Alternatively, only in the absence of Domain 20, stimulation of GEE/ DEVD-sensitive toxicity by the NL mutation may occur through Domain 19, suggesting, in this case, that Domain 20 constrains the Domain 19 function that links the NL mutation to the GEE/DEVD-sensitive death mechanism. It was impossible to determine whether Domain 19-mediated activation of GEE/DEVD-sensitive death by the NL mutation also occurs in the presence of Domain 20, because Domain 20-mediated activation of GEE/DEVD-sensitive death robustly occurred by lower expression of NL-APP (carrying Domain 20). Fig. 8 summarizes the suggested mechanisms. In wtAPP, Domain 20 appears to be inert, and only Domain 19 exerts weak constitutive action to cause GEE/DEVD-resistant death (Fig. 8a). While this basal activity of Domain 19 is unaltered in V642I-APP, the V642I mutation activates Domain 20 at low expression of V642I-APP, leading cells to GEE/DEVD-sensitive death through PTX-sensitive G proteins (Fig. 8b). Therefore, the toxicity of V642I-APP consists of two different components: GEE/DEVD-resistant toxicity of wtAPP and V642I-stimulated GEE/DEVD-sensitive toxicity. The NL mutation also promotes the function of Domain 20 leading to PTX/GEE/DEVD-sensitive death, for which low expression of NL-APP is sufficient. Concomitantly, the NL mutation enhances the basal activity of Domain 19 that causes GEE/DEVD-resistant death. This activation requires higher expression of NL-APP and the G protein-mediated assistance of Domain 20. Higher expression of NL-APP also allows Domain 19 to potentially cause GEE/ DEVD-sensitive death, but not through PTX-sensitive G proteins (Fig. 8c). The NL-induced potentiation of GEE/DEVDresistant death through Domain 19 should be the reason why neither DEVD nor GEE could suppress death by higher induction of NL-APP. Also, the NL-induced activation of GEE/ DEVD-sensitive death through Domain 19 should explain the result that PTX could not inhibit death by higher induction of NL-APP. As G protein-mediated assistance by Domain 20 was required for the NL mutation to stimulate GEE/DEVD-resistant toxicity of Domain 19, we examined and found that concomitant Ac-DEVD-CHO and PTX completely suppressed NLmediated stimulation of death, but did not affect basal death attributable to higher expression of wtAPP (Fig. 9). This lends additional credence to the present hypothesis.
Among the mechanisms activated by the two FAD mutations, GEE/DEVD-sensitive death by low expression of both FAD mutants and basal GEE/DEVD-insensitive death by their higher induction were common, suggesting that either or both could be involved in the mechanism causing FAD. We examined E618Q-APP, because this APP mutant is responsible for HCHWA-D, which does not associate AD type of neurodegeneration in most cases (33)(34)(35)45). The result indicates that (i) the GEE/DEVD-sensitive mechanism for neuronal cell death, which was commonly activated by FAD mutations, was not activated by the E618Q mutation; and (ii) higher expression of E618Q-APP only exerted GEE/DEVD-resistant death attributable to the toxicity of wtAPP. It is thus reasonable to assume that PTX/GEE/DEVD-sensitive death by low expression of both FAD genes is linked specifically to the FAD mutations and that toxicity by higher expression of wtAPP (or FAD mutants) may less reflect the mechanism essential for FAD. If so, it should also be cautioned that neurotoxicity induced by more than severalfold overexpression of wtAPP or these FAD mutants in cell culture or transgenic mice may not reflect the mechanism underlying neurodegeneration in AD patients. Detailed analysis of the involved molecules, particularly in the PTX/GEE/ DEVD-sensitive death by low expression of both FAD genes, must be undertaken.