Upstream Signaling Pathways Leading to the Activation of Double-stranded RNA-dependent Serine/Threonine Protein Kinase in β-Amyloid Peptide Neurotoxicity*

One of the hallmarks of Alzheimer's disease is extracellular accumulation of senile plaques composed primarily of aggregated β-amyloid (Aβ) peptide. Treatment of cultured neurons with Aβ peptide induces neuronal death in which apoptosis is suggested to be one of the mechanisms. We have demonstrated previously that Aβ peptide induces activation of double-stranded RNA-dependent serine/threonine protein kinase (PKR) and phosphorylation of eukaryotic initiation factor 2α (eIF2α) in neurons in vitro. Degenerating neurons in brain tissues from Alzheimer's disease patients also displayed high immunoreactivity for phosphorylated PKR and eIF2α. Our previous data have also indicated that PKR plays a significant role in mediating Aβ peptide-induced neuronal death, because neurons from PKR knockout mice and neuroblastoma SH-SY5Y cells stably transfected with dominant negative mutant of PKR are less susceptible to Aβ peptide toxicity. Therefore, it is important to understand how PKR is activated by Aβ peptide. We report here that inhibition of caspase-3 activity reduces phosphorylation of PKR and to a certain extent, cleavage of PKR and eIF2α in neurons exposed to Aβ peptide. Calcium release from the endoplasmic reticulum and activation of caspase-8 are the upstream signals modulating the caspase-3-mediated activation of PKR by Aβ peptide. Although in other systems HSP90 serves as a repressor for PKR, it is unlikely the candidate for caspase-3 to affect PKR activation in neurons after Aβ peptide exposure. Elucidation of the upstream pathways for PKR activation can help us to understand how this kinase participates in Aβ peptide neurotoxicity and to develop effective neuroprotective strategy.

One of the pathological hallmarks of Alzheimer's disease (AD) 1 is extracellular accumulation of senile plaques composed primarily of aggregated ␤-amyloid (A␤) peptide (1). A␤ peptide is a heterogeneous 39 -43-amino acid peptide generated by sequential cleavage of amyloid precursor protein by ␤-secretase and ␥-secretase (1). It is generally considered that A␤ peptide plays a pivotal role in the pathogenesis of AD (for review see Ref. 2). Treatment of cultured neurons with A␤ peptide has been shown to induce apoptosis (3) and necrosis (4). Neurons undergoing A␤ peptide-induced apoptosis are morphologically characterized by membrane blebbing, cell shrinkage, DNA fragmentation, and chromatin condensation and biochemically by an ordered activation of a conserved family of cysteine proteases called caspases (5)(6)(7). In the brains of AD patients, it has been demonstrated that numerous tangle-bearing neurons and non-tangle-bearing neurons display apoptotic features with DNA fragmentation (8,9).
Caspases have been considered to play important roles in coordinating apoptosis (10,11). A␤ peptide is able to induce selective activation of caspases including caspase-2, -3, -6, -8, -9, and -12 (12)(13)(14)(15)(16). Among all the caspases, caspase-3 has been shown to play important roles in the execution phase of apoptosis (17,18). Caspase-3 has been reported to be activated in brain tissues from AD patients (19) and in A␤ peptide-treated neurons (13,20). Although the significance of other caspases such as caspase-2 cannot be neglected (15), inhibition of caspase-3 activation has been shown to protect neurons from A␤ peptide-induced apoptosis, implying that caspase-3 might represent one of the key players in mediating A␤ peptideinduced apoptosis (20).
Recently, we found that activation of double-stranded RNA (dsRNA)-dependent protein kinase (PKR) and phosphorylation of eukaryotic initiation factor 2␣ (eIF2␣) are observed in the degenerating neurons of the brains from AD patients (21) and in cultured neuronal cells challenged with A␤ peptide (22). We have also reported that PKR plays a crucial role in mediating A␤ peptide-induced neuronal death, because primary cortical neurons from PKR knockout mice and neuroblastoma SH-SY5Y cells stably transfected with dominant negative of mutant PKR are less susceptible to A␤ peptide toxicity (22). Human PKR consists of 551 amino acids that form two functional domains: an N-terminal dsRNA binding regulatory domain and a C-terminal kinase catalytic domain (23). Following virus infection, dsRNAs produced by viruses activate PKR after binding to its regulatory domain. As a result, PKR becomes auto-phosphorylated to further promote its activity. Increasing lines of compelling evidence have shown that PKR can be activated without the presence of dsRNA. Instead, non-dsRNA molecules including polyanions (23), PACT (PKR-activating protein) (24), and cellular stresses such as serum-deprivation and calcium depletion from the endoplasmic reticulum (ER) (25) can activate PKR. Activated PKR can phosphorylate eIF2␣ at serine 51. The phosphorylation of eIF2␣ inhibits the guanine nucleotide exchange factor eIF2␤ by preventing the exchange of GDP for GTP on eIF2 so that global protein translation of mRNAs using 5Ј-cap initiation codon is inhibited leading to apoptosis (26). In addition to the translational regulation, PKR also mediates apoptosis by controlling the activation of several transcriptional factors (25,(27)(28)(29), regulation of selective pro-apoptotic molecules such as Bax or Fas (22,25,30), and selective activation of caspases (25,31,32).
Although many studies have been focusing on downstream pathways of PKR (25,31,32), how PKR is activated by A␤ peptide in neurons is still unclear. In the present study, we aim to examine the upstream signaling pathways of PKR triggered by A␤ peptide in neurons. Our previous report (22) has shown that activation of caspase-3 occurs within a short period of time after exposure of A␤ peptide. Therefore, we further investigated whether this early activation of caspase-3 is required for the activation of PKR in A␤ peptide-induced neuronal apoptosis. We found that inhibition of caspase-3 activity attenuated phosphorylation of PKR in neurons exposed to A␤ peptide. We also provided novel information that upstream signals for caspase-3 activation including calcium release from the ER and the activation of caspase-8 could mediate A␤ peptide-triggered PKR activation. Our results suggest that early activation of caspase-3 could function as pro-apoptotic signaling machinery rather than its role as an executioner in apoptosis. Taken together, we elucidate the molecular mechanism of how PKR is activated in A␤ peptide neurotoxicity. Understanding the signaling pathway of PKR activation may help to develop therapeutic intervention against neuronal apoptosis in A␤ peptide toxicity.
Co-immunoprecipitation Assay-Immunoprecipitation of PKR was done as described elsewhere (34) with slight modifications. After treatments, neurons were scratched and lysed in ice-cold E1A buffer containing HEPES (pH 7.6, 50 mM), NaCl (250 mM), Nonidet P-40 (0.1%), EDTA (5 mM), protease inhibitor mixture (Sigma), and phosphatase inhibitor mixture (Sigma). The lysate was then centrifuged at 14,000 ϫ g for 30 min at 4°C. Quantity of protein content in the supernatant was measured by using a protein assay kit (Bio-Rad). 150 g of cellular proteins were pre-cleared with Protein G-Sepharose beads (Protein G-Sepharose 4 Fast Flow; Amersham Biosciences). The unbounded proteins were then incubated with a monoclonal PKR antibody (100 g protein/1.5 g antibody; BD Biosciences) overnight at 4°C. Protein G-Sepharose was then added and mixed for 2 h at 4°C. After washing the beads four times with E1A buffer, the immunoprecipitated samples were subject to Western blot analysis according to the method mentioned before.
Caspase Activity Assay-After treatments, neurons were scratched and lysed in lysis buffer without protease and phosphatase inhibitors for caspase activity assays (BIOSOURCE). The lysate was then centrifuged at 14,000 ϫ g for 30 min at 4°C. 50 g of cellular proteins from the supernatant were used for different caspase activity assays. For colorimetric activity assays, substrates for caspase-3 (Ac-Asp-Glu-Val-Asp-p-nitroanilide, Ac-DEVD-pNA; Calbiochem) or caspase-8 (Ac-Ile-Glu-Thr-Asp-p-nitroanilide, Ac-IETD-pNA; Calbiochem) were incubated with the protein extracts for 2 h at 37°C to yield a yellow-brown product p-nitroanilide, which was measured by a spectrophotometric reader at 405 nm. For fluorogenic activity assays, fluorogenic signals from the cleaved product of the caspase-7 substrate (MCA-Val-Asp-Gln-Val-Asp-Gly-Trp-Lys-(DNP)-NH 2 , MCA-VDQVDGWK-(DNP)-NH 2 ; Calbiochem) were measured with excitation and emission wavelengths at 320 and 405 nm, respectively, after 2 h of incubation at 37°C.
Measurement of Intracellular Free Calcium Levels-Intracellular free calcium levels ([Ca 2ϩ ] i ) were determined by fluorescence imaging with acetoxymethyl-fura 2 (fura 2-AM; Calbiochem) using the methods described previously (35). Briefly, after pretreatment with XeC, cultured neurons were incubated with 5 M fura 2-AM for 30 min at 37°C for dye loading, followed by washing with Hanks' balanced saline supplemented with HEPES (10 mM) twice and glucose (10 mM) and a 30-min pre-incubation prior to calcium imaging. [Ca 2ϩ ] i in 20 -30 neuronal cell bodies per microscopic field was monitored under an inverted microscope prior to and after exposure of neurons to A␤ [25][26][27][28][29][30][31][32][33][34][35] peptide at 25 M. The mean of [Ca 2ϩ ] i from at least four separate cultures was determined from the ratio of the fluorescence emissions using two different excitation wavelengths (340 and 380 nm) according to the Statistical Analysis-Data for multiple variable comparisons were analyzed by one-way analysis of variance (ANOVA). For the comparison of significance, Tukey's test was used as an ad hoc test according to a statistical program SigmaStat® (Jandel Scientific, Chicago, IL). The level of significance was p Ͻ 0.05. For the comparison of significance between two groups, Student's t test was conducted at the level of significance of p Ͻ 0.05 or Ͻ 0.001 with SigmaStat® (Jandel Scientific). Results are expressed as the means Ϯ S.E. from at least three independent experiments.

A␤ Peptide Induced Early Activation of Caspase-3-We have
shown previously (22) that PKR and eIF2␣ play significant roles in A␤ peptide-induced neurodegeneration in vitro and in degenerating neurons of the brains from AD patients (21). However, how A␤ peptide activates PKR in neurons is still unclear. In the present study, we aim to examine the upstream signaling pathways leading to the activation of PKR in A␤ peptide neurotoxicity. Because we have shown that PKR is activated less than 2 h upon the treatment with A␤ peptide in primary cortical neurons (22), we search for any possible candidates that may start to participate in the A␤ peptide-triggered apoptotic process in a short moment of time. A time course study of caspase-3 specific activity indicated that caspase-3 was significantly (p Ͻ 0.05) activated in neurons 1 h after the treatment with A␤ 25-35 peptide at 25 M (0.29 ϩ 0.01 pmol/min/g; see Fig. 1a) when compared with the corresponding control (0.23 ϩ 0.01 pmol/min/g; see Fig. 1a). The A␤ [25][26][27][28][29][30][31][32][33][34][35] peptide-induced caspase-3 activity progressively augmented in a time-dependent manner from 1 to 16 h (Fig. 1a).
To study the role of caspase-3 in the activation of PKR in A␤ peptide neurotoxicity, one of the possible ways is to inhibit caspase-3 by a specific caspase-3 inhibitor (DEVD-CHO). We at first verified the effectiveness of this caspase-3 inhibitor in our experimental model. Caspase-3 activity induced by A␤ [25][26][27][28][29][30][31][32][33][34][35] peptide in cultured neurons at 2 and 7 h was significantly blocked by 100 M DEVD-CHO, which was added 1 h prior to the exposure of A␤ 25-35 peptide (Fig. 1b). DEVD-CHO at 100 M did not show any toxicity to the neurons (data not shown). In addition to the blocking effect on caspase-3, DEVD-CHO might also inhibit caspase-7. Despite the cross-inhibitory effects, our results demonstrated that caspase-7 was not significantly (p Ͻ 0.05) activated 2 and 7 h after the treatment with A␤ 25-35 peptide at 25 M (Fig. 1c). Therefore, the inhibitory effect of DEVD-CHO on caspase-7 could be neglected in our present study.
Inhibition of Caspase-3 Activity Attenuated A␤ Peptide-induced Phosphorylation of PKR and eIF2␣-Having shown that caspase-3 was early activated by A␤ peptide, and the caspase-3 activity can be significantly inhibited by DEVD-CHO, we then study whether modulation of caspase-3 activity can mediate the activation of PKR-eIF2␣ pathway. Western blot analysis demonstrated that 25 M A␤ 25-35 peptide induced an increased phosphorylation of PKR at threonine 446 and 451 in neurons when compared with the corresponding controls (Fig. 2a). Pretreatment of neurons with 100 M DEVD-CHO for 1 h markedly reduced the A␤ 25-35 peptide-induced PKR phosphorylation (Fig. 2a). 100 M DEVD-CHO per se did not induce PKR phosphorylation (Fig. 2a). There was no other band detected near the region of 38 to 48 kDa (the kinase fragment), indicating that no cleaved PKR was phosphorylated at threonine 446 and 451 (data not shown). Fig. 2b shows that 25 M A␤ 25-35 peptide induced phosphorylation of eIF2␣ at serine 51 in neurons, and pretreatment of neurons with 100 M DEVD-CHO for 1 h markedly reduced the eIF2␣ phosphorylation. In controls, there was also a mild expression of the phosphorylated PKR (Fig. 2a) and phosphorylated eIF2␣ (Fig. 2b), which was possibly because of the effects of serum deprivation during experimental treatments.
Inhibition of Caspase-3 Activity Blocked A␤ Peptide-induced Cleavage of eIF2␣ and PKR-Previous reports have shown that PKR and eIF2␣ can be cleaved by caspase-3. To further dem- onstrate that caspase-3 is actively involved in neurons exposed to A␤ peptide, we study whether caspase-3-mediated cleavage of PKR and eIF2␣ occurs in our model. There is an additional form of eIF2␣, migrating with a slightly faster electrophoretic mobility than the full-length eIF2␣ in Western blot analysis (Fig. 3, a and b). This additional form of eIF2␣ is its cleaved form. The difference in the molecular mass between the fulllength and cleaved form of eIF2␣ proteins shown in Fig. 3, a and b is about 2-3 kDa, which is in agreement with the previous findings (36 -38). In controls, in addition to the appearance of the full-length form of eIF2␣, the cleaved bands were also observed ( Fig. 3, a and b), which is possibly related to the effects of serum deprivation during experimental treatments. This is because some proteases including caspases and calpain could be activated in cultured neurons under the serumfree condition (39 -41). Because the complete inhibition of caspase-3 activity cannot fully prevent the cleavage of eIF2␣ in the controls, there might be some other proteases responsible for this cleavage under the serum-free condition. Neurons exposed to A␤ [25][26][27][28][29][30][31][32][33][34][35] peptide appeared to have the cleaved form of eIF2␣ only, indicating that the full-length form of the protein was totally cleaved (Fig. 3, a and b). A␤  peptide, a reverse peptide of A␤ 25-35 as a negative control, did not induce cleavage of eIF2␣ when compared with the corresponding control (Fig.  3b). A␤ [25][26][27][28][29][30][31][32][33][34][35] peptide-induced cleavage of eIF2␣ was blocked by pretreating neurons with 100 M DEVD-CHO for 1 h (Fig. 3a). Fig. 3, c and d indicates that treatment of neurons with A␤ 25-35 peptide at 25 M also induced a cleavage of PKR. Western blot analysis using an antibody against the C-terminal kinase domain of PKR demonstrated that the molecular mass of cleaved PKR was nearly 43 kDa (Fig. 3, c and d), which is in agreement with the recent findings (36). A mild expression of cleaved PKR was visualized in some controls, which might be because of the effects of serum deprivation (Fig. 3, c and d). The cleavage of PKR induced by A␤ 25-35 peptide did not appear as complete cleavage as seen in eIF2␣, because a large portion of the full-length form of PKR was detected (Fig. 3, c and d). A␤  peptide, a reverse peptide of A␤ 25-35 as a negative control in this experiment, did not induce proteolysis of PKR when compared with the corresponding control (Fig. 3d). Inhibition of caspase-3 activity by using 50 or 100 M DEVD-CHO blocked A␤ [25][26][27][28][29][30][31][32][33][34][35] peptide-induced cleavage of PKR in neurons, indicating that caspase-3 is responsible for this cleavage (Fig.  3c).
HSP90 Was Not Involved in the Activation of PKR in A␤ Peptide Neurotoxicity-Apart from the cleavage of PKR, we demonstrated that phosphorylation of PKR was also mediated by caspase-3. Therefore, we attempted to further investigate how caspase-3 affects the phosphorylation of PKR. Because the biological nature of caspase-3 is a protease, it cannot function as a kinase to activate PKR by phosphorylating it. We speculate that caspase-3 might mediate PKR phosphorylation through a PKR repressor. HSP90 has been reported recently (47) to interact with PKR and to repress the activity of PKR, because an inhibitor of HSP90 geldanamycin has been shown to induce the activation of PKR by triggering dissociation of HSP90 from PKR. Using immunoprecipitation, we found that HSP90 was co-immunoprecipitated with PKR (Fig. 7). Treatment with A␤ 25-35 peptide in neurons did not decrease the association of HSP90 from PKR (Fig. 7). Therefore, HSP90 may not be the candidate for caspase-3 to modulate neuronal PKR activity following the exposure of A␤ peptide. DISCUSSION We have demonstrated that PKR plays significant roles in A␤ peptide neurotoxicity (22) and in AD (21). Inhibition of caspase-3 activity reduces phosphorylation of PKR in neurons exposed to A␤ peptide. Early and mild activation of caspase-3 is triggered by an increase in [Ca 2ϩ ] i released from the ER and the activation of caspase-8. Although activation of caspase-3 is usually followed by the cleavage of its substrates, a PKR repressor HSP90 is unlikely to be the substrate for caspase-3 in A␤ peptide-induced apoptosis in neurons.
Both the synthetic A␤ 1-42 and A␤ [25][26][27][28][29][30][31][32][33][34][35] peptides are able to form fibrils in vitro (50). Fibril formation has been suggested to be required for A␤ peptide neurotoxicity (1, 2, 50 -52). Treatment of neurons with A␤ peptide can induce activation of caspase-3 (20). Recent studies have indicated that other amyloidogenic peptides such as human amylin can also trigger activation of caspase-3 in RINm5F cells (53), suggesting that activation of caspase-3 may not be specific for A␤ peptide toxicity. It is also known that human amylin acts in a similar way as A␤ peptide in inducing calcium changes and that antagonists of human amylin can block A␤ peptide toxicity, implying that A␤ peptide and human amylin may have a combined site of action (54,55). Yet, human amylin does not accumulate in AD brains, and nothing is known about PKR and human amy- lin. More importantly, human amylin induces a relatively late activation of caspase-3 (12 h after the treatment) (53) whereas our results show that A␤ peptide significantly induces caspase-3 activation as early as 1 h after the treatment. Early activation of caspase-3 is important for the activation of PKR in neurons (as discussed below).
A Novel Role of Caspase-3: Activator of a Cellular Signaling Pathway-Caspase-3 has long been considered as an executioner in apoptosis (10,11,56,57). Our present results may add a new concept on the functions of caspase-3; early activation of caspase-3 is an important event in mediating cellular signaling pathways such as PKR rather than only participating in the execution phase of apoptosis. There are a few recent studies indicating the importance of early activation of caspase-3 in physiological responses; caspase-3 is activated 2 min after a learning process in adult male zebra finch songbirds (58), and early activation of caspases including caspase-3 is involved in the early steps of lymphocyte activation (59). In the present study, caspase-3 is activated significantly in neurons 1 h after the treatment with A␤ peptide. Blockade of this early activation of caspase-3 can reduce phosphorylation of PKR and eIF2␣, implying that caspase-3 is an important candidate in mediating early cellular signaling events associated with A␤ peptide neurotoxicity.
Apart from mediating phosphorylation of PKR, we demonstrate here that caspase-3 can also cleave PKR and eIF2␣. For PKR cleavage, Western blot analysis using an antibody against the C-terminal domain of PKR reveals that the molecular mass of cleaved PKR was nearly 43 kDa, which is similar to the previous findings (36). For eIF2␣ cleavage, Western blot analysis reveals that the difference in molecular mass between the full-length and cleaved forms of eIF2␣ proteins is 2-3 kDa, which is also in agreement with the previous reports (36 -38).
Other studies have shown that PKR and eIF2␣ can be cleaved by caspases in other cell types during different apoptotic insults including treatment with poly(I)⅐poly(C) (38), tumor necrosis factor-␣ plus cycloheximide (36 -38), cisplatin (37,38), etoposide (37), and anti-Fas (36). Cleaved PKR is fully capable of phosphorylating eIF2␣ to the same extent as the full-length form of PKR, because the proteolysis of PKR by caspases releases the kinase domain from the N-terminal regulatory Nterminal domain (36). The N-terminal part of PKR fragment has been proposed to be the repressor of the PKR itself (60). Although PKR can be activated by cleavage, our results show that only a small portion of cleaved PKR was observed. This implies that the activation mode of PKR is not solely dependent on the cleavage of PKR. For the cleavage of eIF2␣, it has been suggested that its cleavage causes functional changes of the eIF2␣ complex, which can no longer stimulate global protein translation (37). There was a large portion of cleaved eIF2␣ observed in A␤ peptide-treated neurons, suggesting that cleaved eIF2␣ can be an additional way to inhibit global protein translation. Taken together, although cleavage of eIF2␣ can augment the effects of its phosphorylation in protein translation, cleavage of PKR by activated caspase-3 is not the major mode of activation for PKR. Phosphorylation of PKR remains to be the major way for the activation of PKR.
Activation of caspase-3 has been reported previously (20) in neurons exposed to A␤ peptide. Cleavage of poly(ADP-ribose) polymerase and ␣-fodrin have also been found to be endogenous substrates for caspase-3 (20). In the present study, eIF2␣ and PKR are not only the substrates for caspase-3, their activation is also mediated by this caspase. As mentioned before, our results show that A␤ peptide induces early activation of caspase-3 in neurons (as early as 1 h after the treatment of A␤ peptide) whereas Harada and Sugimoto (20) focused on the late activation of caspase-3. Because early and mild activation of caspase-3 plays an important role in mediating PKR activation, we further show that calcium release from the ER and caspase-8 activation are upstream of caspase-3-mediated PKR activation in A␤ peptide neurotoxicity. Therefore, our present results provide novel information on how A␤ peptide mediates toxicity in neurons through caspase-3 and PKR.
Caspase-3-mediated PKR Activation by A␤ Peptide Is Partially Regulated by ER Calcium Release-Caspase-3 is early activated and plays important roles in mediating the activation of PKR in A␤ peptide neurotoxicity. We further study the upstream signals leading to the caspase-3-mediated activation of PKR. A␤ peptide has been shown to induce release of calcium from the ER to the cytosol, which may result in disturbance in intracellular calcium homeostasis (61). We have reported recently (49) that both the A␤ 1-42 and A␤ [25][26][27][28][29][30][31][32][33][34][35] peptides can induce calcium release from the ER, which is in consistent with the previous findings showing that both the peptides can trigger an elevation in [Ca 2ϩ ] i (61,62). Therefore, we used XeC to modulate A␤ peptide-triggered ER calcium release possibly via inositol 1,4,5-trisphosphate receptor (42). Our results show that attenuation of A␤ peptide-triggered ER calcium release by XeC significantly reduces caspase-3 activity, indicating that ER calcium release is likely to be a source for the caspase-3 activation. Indeed, activation of caspase-3 by an increase in calcium concentration has been reported (63). Therefore, A␤ peptide-triggered calcium release from the ER may account for the early activation of caspase-3 and the caspase-3-mediated activation of PKR. We have shown previously (22) that an intracellular calcium chelator BAPTA-AM can reduce A␤ peptide-triggered PKR activation in neurons, suggesting that changes in [Ca 2ϩ ] i can stimulate PKR. We therefore cannot neglect the direct activation of PKR by calcium in A␤ peptide neurotoxicity. Mutations in PS-1 (presenilin-1) gene enhance calcium release from the ER (64), sensitizing neurons to calcium insults triggered by A␤ peptide (65,66). Recently, the activity of PERK (PKR-like ER kinase), which is activated by ER stress, has been found not to be altered by a PS-1 knock-in mutation (67). We speculate that PKR might play a more important role in PS-1-mediated ER calcium signaling as calcium release from the ER can activate PKR.
Caspase-8 Is an Upstream Modulator of Caspase-3-mediated PKR Activation-Apart from intracellular calcium, our results also indicate that caspase-8 can modulate caspase-3 activation in our experimental model, because inhibition of caspase-8 activity can significantly reduce A␤ peptide-triggered caspase-3 activation. In fact, caspase-3 can be activated or processed in vitro by caspase-8 (44,45). We show that reduction of caspase-3 activity by inhibiting caspase-8 attenuates A␤ peptide-induced phosphorylation of PKR and eIF2␣, suggesting that caspase-8 is most likely upstream of the caspase-3-mediated activation of PKR. Activation of caspase-8 has been shown in A␤ peptide-induced neuronal apoptosis, because inhibition of caspase-8 activity and expression of dominant negative Fasassociated death domain protect neurons from A␤ peptide neurotoxicity (12). Fas and Fas ligand have also been shown to be involved in A␤ peptide toxicity in cultured hippocampal neurons (43). In addition, neutralization of TRAIL (tumor necrosis factor-related apoptosis-inducing ligand) by neutralizing antibody can protect human neuroblastoma SH-SY5Y cells from A␤ peptide neurotoxicity (68). Besides, caspase-8 can be cleaved by calcium-activated protease calpain (69), which has been shown to be activated in A␤-peptide treated neurons (70). Although a recent study using fluorescent resonance energy transfer in COS-7 and NIH3T3 cells did not find any activation of caspase-8 stimulated by A␤ peptide (71), ours, along with those from other laboratories, demonstrate the activation of caspase-8. In fact, our results show that caspase-8 activity in primary cultured neurons exposed to A␤ peptide was significantly increased even 1 h after the treatment. The discrepancy between different findings may be because of different types of cells used in the studies. For the activation of caspase-8 in neurons by A␤ peptide, it has been indicated to be linked with death receptors of the Fas/tumor necrosis factor receptor family (12,43). More importantly, the present results demonstrate that caspase-3-mediated activation of PKR can be mediated by caspase-8.
HSP90 Is Unlikely to Modulate Caspase-3-mediated PKR Activation in A␤ Peptide-induced Neuronal Death-Inhibition of caspase-3 activity by DEVD-CHO can block both the A␤ peptide-induced phosphorylation and cleavage of PKR and eIF2␣. It appears that application of this caspase-3 inhibitor may alter the activity of multiple downstream pathways of caspase-3. Yet, the relationship between caspase-3 and PKR-eIF2␣ pathway is still evident, because attenuation of caspase-3 activity by either reduction of ER calcium release or inhibition of caspase-8 activity can modulate the PKR-eIF2␣ pathway. As caspase-3 is not a kinase, its action on activating PKR would not through direct phosphorylation. Blockade of caspase-3 activity may therefore alter some PKR repressors or activators for PKR activation.
Having shown that early and mild activation of caspase-3 plays significant roles in PKR activation, we then attempt to investigate how caspase-3 mediates PKR activity in response to A␤ peptide. As stated above, our results show that cleavage of PKR by caspase-3 is not the major mode for the activation of PKR. Therefore, we searched for any PKR repressor that may be affected by caspase-3. Because the biological nature of caspase-3 is to cleave proteins but not to phosphorylate proteins, a PKR-activating protein, PACT, which is activated by phosphorylation (72), is unlikely to be the target for caspase-3-mediated PKR activation. HSP90 has been reported recently (47) to interact with PKR and to repress the activity of PKR, because geldanamycin, an inhibitor of HSP90, has been shown to induce activation of PKR by triggering dissociation of HSP90 from PKR. Our results indicate that HSP90 interacts with PKR in neurons as HSP90 is co-immunoprecipitated with PKR. However, A␤ peptide does not induce dissociation of HSP90 from PKR, suggesting that A␤ peptide-induced PKR activation is not mediated by the dissociation of HSP90 from PKR. Therefore, HSP90 might not be the target for caspase-3-mediated PKR activation in A␤ peptide neurotoxicity.
There have been some reports showing how PKR could regulate downstream apoptotic pathways, such as the activation of caspase-8 and caspase-9 (25,31,32). In the present study, we highlight how caspase-3, caspase-8, and ER calcium release at an upstream position modulate the PKR pathway in A␤ peptide-treated neurons (see the summarized diagram in Fig. 8). Taken together, our present results demonstrate a new mode of PKR activation and explore a new upstream signaling pathway of A␤ peptide neurotoxicity. Because phosphorylated PKR and eIF2␣ are observed in degenerating neurons of AD patients (21), understanding the molecular signaling mechanisms of neuronal apoptosis may pave the way for future therapeutic intervention against neuronal death in AD.