Central role of glycogen synthase kinase-3beta in endoplasmic reticulum stress-induced caspase-3 activation.

Stress of the endoplasmic reticulum (ER), which is associated with many neurodegenerative conditions, can lead to the elimination of affected cells by apoptosis through only partially understood mechanisms. Thapsigargin, which causes ER stress by inhibiting the ER Ca(2+)-ATPase, was found to not only activate the apoptosis effector caspase-3 but also to cause a large and prolonged increase in the activity of glycogen synthase kinase-3beta (GSK3beta). Activation of GSK3beta was obligatory for thapsigargin-induced activation of caspase-3, because inhibition of GSK3beta by expression of dominant-negative GSK3beta or by the GSK3beta inhibitor lithium blocked caspase-3 activation. Thapsigargin treatment activated GSK3beta by inducing dephosphorylation of phospho-Ser-9 of GSK3beta, a phosphorylation that normally maintains GSK3beta inactivated. Caspase-3 activation induced by thapsigargin was blocked by increasing the phosphorylation of Ser-9-GSK3beta with insulin-like growth factor-1 or with the phosphatase inhibitors okadaic acid and calyculin A, but the calcineurin inhibitors FK506 and cyclosporin A were ineffective. Insulin-like growth factor-1, okadaic acid, calyculin A, and lithium also protected cells from two other inducers of ER stress, tunicamycin and brefeldin A. Thus, ER stress activates GSK3beta through dephosphorylation of phospho-Ser-9, a prerequisite for caspase-3 activation, and this process is amenable to pharmacological intervention.

Impaired function of the endoplasmic reticulum (ER), 1 commonly referred to as ER stress, is an important factor in the neuropathology of a wide variety of neurological disorders (reviewed in Refs. [1][2][3][4]. Exemplary among these is Alzheimer's disease (AD). Studies related to AD have shown that the neurotoxic effects of the amyloid ␤-peptide are at least partially targeted to the ER (e.g. Refs. 5 and 6). Also, AD-associated mutations of presenilin-1 disrupt calcium homeostasis and increase susceptibility to ER stress and apoptosis (7)(8)(9)(10)(11)(12), whereas wild-type presenilin-1 is necessary for cellular responses to ER stress (13). Furthermore, an AD-associated splice variant of presenilin-2 increases vulnerability to ER stress (14). Substantial evidence links ER stress to several other neurological disorders as well as to the declination in neuronal function associated with aging (4). Because the ER is a central site of protein folding, ER stress can lead to increased intracellular levels of misfolded proteins, and eventual cell death by apoptosis, processes that may contribute to neurodegenerative disorders.
Several agents can be used to induce ER stress experimentally, and likely the most widely applied is thapsigargin. Thapsigargin inhibits the Ca 2ϩ -ATPase in the ER (15), which blocks sequestration of calcium by the ER, causing increases in the intracellular concentration of calcium, accumulation of unfolded or misfolded proteins, and activation of caspase-3-mediated apoptosis (16 -18). The mechanisms mediating ER stressinduced activation of the apoptosis program remain incompletely elucidated, although both caspase-7 and caspase-12 have been implicated in addition to the crucial effector caspase-3 (6,19,20). We considered the possibility that glycogen synthase kinase-3␤ (GSK3␤) may be involved in this apoptotic program, because it recently has been shown to be a key intermediate in several apoptotic signaling pathways that lead to activation of caspase-3 (reviewed in Ref. 21). This was first shown by the findings that GSK3␤ antisense oligonucleotides blocked apoptosis induced by the Alzheimer's disease amyloid-␤-peptide (22) and that transient overexpression of GSK3␤ caused PC12 and Rat-1 cells to undergo apoptotic programmed cell death (23). Further studies have extended the known links between GSK3␤ and apoptosis. Moderate overexpression of GSK3␤ (3.5-fold), which was insufficient to induce apoptosis alone, facilitated apoptosis induced by stressors (24,25). The human immunodeficiency virus type 1 regulatory protein, Tat, induced neuronal apoptosis in a GSK3␤-dependent manner, a signal mediated by platelet-activating factor receptor-induced activation of GSK3␤ (26,27). GSK3␤ also has been implicated as contributing to neuronal cell death induced by ischemia (28,29), excitotoxicity induced by glutamate receptor activation (30,31), and models of Huntington's disease (32,33). Additionally, many studies of apoptotic conditions involving growth factor withdrawal or inhibition of the phosphatidylinositol 3-kinase/Akt signaling system, pathways that normally maintain GSK3␤ in an inhibited state through phosphorylation of Ser-9 (34), have shown that GSK3␤ promotes the subsequent apoptotic process (35)(36)(37)(38)(39). More that just neuronal apoptosis is promoted by GSK3␤ activity, because this relationship has been demonstrated in a wide variety of cell types, for example in vascular smooth muscle cells (40), fibroblasts (41), human erythroid progenitors (42), and cardiac cells (43). A number of these studies advantageously used lithium, along with other approaches, to identify the contributory effects of GSK3␤ to apoptosis. Lithium is useful in this regard, because it is a selective inhibitor of GSK3␤ (44, 45), a finding substantiated by an examination of 24 kinases, which showed that GSK3␤, and the closely related GSK3␣, to be the only kinases substantially inhibited by lithium (46). Extensive studies have clearly documented that several of lithium's prominent effects, such as inhibition of the phosphorylation of the microtubule-associated protein tau, increased levels of ␤-catenin, and protection from GSK3␤-facilitated apoptosis, are directly dependent on lithium's inhibition of GSK3␤ (24,(47)(48)(49)reviewed in Refs. 21 and 50). Based on the extensive evidence that GSK3␤ promotes apoptosis and that ER stress is involved in a variety of neurodegenerative disorders in which apoptosis may contribute to neuronal loss, we investigated whether there is an association between GSK3␤ activity and apoptotic signaling induced by ER stress.
Immunoblot Analysis-For immunoblotting, cells were washed twice with phosphate-buffered saline and lysed with 100 l of lysis buffer (20 mM Tris, pH 7.5, 150 mM NaCl, 2 mM EDTA, 2 mM EGTA, 1 mM sodium orthovanadate, 50 mM sodium fluoride, 100 M phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, 10 g/ml aprotinin, 5 g/ml pepstatin A, 1 nM okadaic acid, and 0.2% Nonidet P-40). The lysates were collected in centrifuge tubes, sonicated, and centrifuged at 16,000 ϫ g for 10 min at 4°C. Protein concentrations were determined using the BCA method (Pierce). Where indicated, cells were fractionated as described previously (54). For subcellular fractionation, lysed cells were collected in microcentrifuge tubes, and centrifuged at 2,700 ϫ g for 10 min at 4°C. The supernatant containing the cytosol was further centrifuged at 20,800 ϫ g for 15 min at 4°C to obtain the cytosolic fraction. The nuclei in the pellet were washed three times by gently resuspending the nuclei in 200 l of wash buffer (10 mM PIPES, pH 6.8, 300 mM sucrose, 3 mM MgCl 2 , 1 mM EGTA, 25 mM NaCl, 1 mM sodium orthovanadate, 50 mM sodium fluoride, 100 M phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, 10 g/ml aprotinin, and 5 g/ml pepstatin A) and centrifuging at 2,700 ϫ g for 5 min at 4°C. For a final wash, the nuclei were resuspended in 100 l of wash buffer, layered over a cushion of 1 ml of sucrose buffer (1 M sucrose, 1 mM sodium orthovanadate, 50 mM sodium fluoride, 100 M phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, 10 g/ml aprotinin, and 5 g/ml pepstatin A), and centrifuged at 2,700 ϫ g for 10 min. The sucrose buffer containing non-sedimented cellular debris was discarded, and the pellet containing nuclei was washed in 100 l of lysis buffer and centrifuged at 2,700 ϫ g for 5 min at 4°C to remove residual sucrose buffer. Extracts were mixed with Laemmli sample buffer (2% SDS) and placed in a boiling water bath for 5 min. Proteins were resolved in SDS-polyacrylamide gels, transferred to nitrocellulose, and incubated with anti-GSK3␤, anti-poly(ADP-ribose) polymerase (PARP), anti-proteolyzed PARP 85-kDa fragment (Pharmingen/Transduction Laboratories, San Diego, CA), anti-phospho-Ser-9-GSK3␤, anti-phospho-Ser-473-Akt, anti-total Akt, anti-␤-catenin, and anti-active casapse-3 (Cell Signaling, Beverly, MA) antibodies. Immunoblots were developed using horseradish peroxidase-conjugated goat anti-mouse or goat anti-rabbit IgG, followed by detection with enhanced chemiluminescence, and the protein bands were quantitated with a densitometer.
Enzyme Activity Measurements-Fluorometric assays of caspase-3 activity using the substrate Ac-DEVD-AMC (Alexis) were carried out as described previously (24). For this, fluorometric assays were conducted in 96-well clear bottom plates, and all measurements were carried out in triplicate wells. To each well 200 l of assay buffer (20 mM HEPES, pH 7.5, 10% glycerol, 2 mM dithiothreitol) was added. Peptide substrates for caspase-3 (Ac-DEVD-AMC) (Alexis Biochemicals, San Diego, CA) were added to each well to a final concentration of 25 ng/l. When the caspase-3 inhibitor (Ac-DEVD-CHO) was used, it was added at a concentration of 2.5 ng/l immediately before addition of the caspase-3 substrate. Cell lysates (20 g of protein) were added to start the reaction. Background fluorescence was measured in wells containing assay buffer, substrate, and lysis buffer without the cell lysate. Assay plates were incubated at 37°C for 1 h for measurement of caspase-3, and fluorescence was measured on a fluorescence plate reader (Bio-Tek, Winooski, VT) set at 360-nm excitation and 460-nm emission. Caspase activity was calculated as [(mean AMC fluorescence from triplicate wells) Ϫ (background fluorescence)]/g of protein.
The activity of GSK3␤ was measured as described previously (54). For this, to immunoprecipitate GSK3␤, 100 g of protein was incubated with 0.75 g of monoclonal GSK3␤ antibody overnight at 4°C with gentle agitation. Extracts were incubated with 30 l of protein G-Sepharose for 1 h at 4°C. The immobilized immune complexes were washed twice with immunoprecipitation lysis buffer and twice with kinase buffer (20 mM Tris, pH 7.5, 5 mM MgCl 2 , and 1 mM dithiothreitol). Kinase activity was measured by mixing immunoprecipitated GSK3␤ with 25 l of kinase buffer containing 20 mM Tris, pH 7.5, 5 mM MgCl 2 , 1 mM dithiothreitol, 250 M ATP, 1.4 Ci of [␥-32 P]ATP (Amersham Biosciences, Arlington Heights, IL), and 0.1 g/l recombinant tau protein (Panvera, Madison, WI). The GSK3␤ inhibitor lithium (20 mM (44)) was added in vitro to confirm that phosphorylation was mediated by GSK3␤. The samples were incubated at 30°C for 15 min, and 25 l of Laemmli sample buffer (2% SDS) was added to each sample to stop the reaction. Samples were placed in a boiling water bath for 5 min, and proteins were separated in 7.5% SDS-polyacrylamide gels. The gels were vacuum-dried, exposed to a phosphoscreen overnight, and quantitated using a PhosphorImager (Molecular Dynamics, Sunnyvale, CA). The efficiency of GSK3␤ immunoprecipitation was determined by immunoblotting for GSK3␤.

Thapsigargin Treatment Induces Apoptosis and Activates
GSK3␤-Characteristics associated with apoptosis were assessed in human neuroblastoma SH-SY5Y cells after the induction of ER stress with thapsigargin. These parameters included measurements of the activity of the effector caspase, caspase-3, proteolysis of PARP (a classical substrate cleaved by caspase-3), immunoblot detection of the cleavage of procaspase-3 to active caspase-3 fragments, and changes in morphology. Caspase-3 activity, measured by fluorogenic substrate cleavage (24), increased between 2 and 4 h after treatment with 2 M thapsigargin ( Fig. 1A). Similar time courses after thapsigargin treatment were observed in Western blot analyses of the proteolysis of PARP from an intact 116-kDa protein to a stable 85-kDa breakdown product, and the production of 17-and 19-kDa activated caspase-3 (Fig. 1A). Examination of cells treated with thapsigargin and stained with Hoechst 33342 (24) revealed the characteristic morphology associated with apoptosis, including nuclear condensation and cell shrinkage (data not shown). These results confirm previous reports that thapsigargin causes SH-SY5Y cells to undergo caspase-3-mediated apoptosis (55)(56)(57).
To test if GSK3␤ is involved in the apoptotic response to ER stress, the activity of GSK3␤ was assessed in SH-SY5Y cells after thapsigargin treatment. As described previously (54), GSK3␤ activity was measured by immunoprecipitating GSK3␤, measuring its catalysis of the phosphorylation of recombinant tau protein, a well-characterized substrate of GSK3␤ (reviewed in Ref. 58), and confirming that phosphorylation was mediated by immunoprecipitated GSK3␤ by inclusion of the GSK3␤ inhibitor lithium (20 mM (44)) in the kinase assay. These measurements revealed that there was a large and prolonged increase in GSK3␤ activity after thapsigargin treatment (Fig. 1B). GSK3␤ activity increased within 2 h after exposure of cells to 2 M thapsigargin and was 277 Ϯ 27% (n ϭ 3) of control levels after 4 h of thapsigargin treatment (Fig. 1B).
In situ activation of GSK3␤ by thapsigargin treatment was further confirmed by measuring the level of ␤-catenin. Phosphorylation of cytosolic ␤-catenin by GSK3␤ promotes its degradation, whereas inhibition of GSK3␤ allows the stabilization, accumulation, and nuclear translocation of ␤-catenin (59). An additional pool of ␤-catenin is sequestered at the plasma membrane, and its stability is unaffected by GSK3␤. Treatment of SH-SY5Y cells with thapsigargin caused depletion of cytosolic ␤-catenin (Fig. 1C), and a modest reduction of nuclear ␤-catenin, whereas membrane-bound ␤-catenin was unaltered. Treatment with lithium to inhibit GSK3␤ attenuated thapsigargin-induced depletions of cytosolic and nuclear ␤-catenin. These results are indicative of GSK3␤ activation by thapsigargin and inhibition by lithium. Thus, thapsigargin treatment caused the activation of GSK3␤, a previously unknown response to ER stress, concomitantly with the initiation of apoptotic signaling, which raised the possibility that GSK3␤ may be involved in the signaling pathway linking ER stress to caspase-3 activation.
GSK3␤ Is an Obligate Intermediate in Thapsigargin-induced Apoptosis-Considering recent findings that GSK3␤ can promote apoptosis (reviewed in Ref. 21), the thapsigargin-induced activation of GSK3␤ raised the question of whether this is an essential component of the apoptosis signaling cascade that is induced by ER stress leading to activation of caspase-3. To test this, the effects of thapsigargin on caspase-3 activation were examined under conditions where the activity of GSK3␤ was modified. To test if increased GSK3␤ activity is stimulatory, thapsigargin-induced PARP proteolysis was compared in control SH-SY5Y cells, vector-transfected cells, and two different stable lines of SH-SY5Y cells that overexpress active GSK3␤ 3to 4-fold above the endogenous level of GSK3␤, which have been described previously (24). Thapsigargin-induced PARP proteolysis was similar in wild-type and vector-transfected SH-SY5Y cells but was much greater in cells overexpressing GSK3␤ ( Fig. 2A), indicating that increased active GSK3␤ promotes thapsigargin-induced caspase activation. In opposition to overexpression of GSK3␤, thapsigargin-induced PARP proteolysis was examined in SH-SY5Y cells stably expressing a dominant-negative mutant of GSK3␤ (60). Although incubation with 2 M thapsigargin caused a time-dependent increase in PARP proteolysis in control cells, there was little PARP proteolysis in cells expressing dominant-negative GSK3␤ (Fig. 2B).
To test further if GSK3␤ is a necessary intermediate in thapsigargin-induced caspase-3 activation, cells were pretreated with lithium, a selective inhibitor of GSK3␤ (44 -46). Thapsigargin-induced PARP proteolysis was concentration-dependently inhibited in cells pretreated with 1-20 mM lithium (Fig. 2C). Pretreatment with 20 mM lithium, which we have shown inhibits GSK3␤ in vitro by ϳ80% (49), reduced thapsigargin-induced PARP proteolysis by 60 -80%, indicating a close correspondence between inhibition of GSK3␤ activity and of caspase-3 activation. Furthermore, examination of the lithium concentration-dependent attenuation of thapsigargin-induced PARP proteolysis revealed less protection in cells overexpressing GSK3␤ than in wild-type or vector-transfected SH-SY5Y cells due to the greater activity of GSK3␤ (Fig. 2C), substantiating the conclusion that lithium's protective action is due to inhibition of GSK3␤. Taken together, these results indicate that GSK3␤ is a necessary and regulatory component of thapsigargin-induced signaling leading to activation of caspase-3.
Mechanism of Thapsigargin-induced GSK3␤ Activation-Although GSK3␤ is a constitutively active enzyme, the activity of GSK3␤ is modulated by phosphorylation, with phosphorylation of Ser-9 decreasing activity and phosphorylation of Tyr-216 increasing activity (reviewed in Ref. 21). To examine if either of these post-translational modifications of GSK3␤ was altered by thapsigargin treatment to account for the thapsigargininduced activation of GSK3␤, the phosphorylation state of GSK3␤ in SH-SY5Y cells was examined by immunoblot analyses. These measurements revealed a time-dependent decrease in phospho-Ser-9-GSK3␤ immunoreactivity after thapsigargin treatment (Fig. 3A), whereas phospho-Tyr-216-GSK3␤ immunoreactivity was unaltered (data not shown). Hence, thapsigargin treatment activated GSK3␤ by reducing the inhibitory Ser-9 phosphorylation of GSK3␤. Because Akt (also known as protein kinase B) is a primary kinase responsible for phosphorylating Ser-9 of GSK3␤ (34), we tested if thapsigargin affected the activation-associated phosphorylation of Ser-473 of Akt. These measurements demonstrated that treatment with thapsigargin greatly decreased phospho-Ser-473-Akt immunoreactivity (Fig. 3B). Thus, thapsigargin treatment reduced both the inhibitory phosphorylation of Ser-9 on GSK3␤ and the activating phosphorylation of Ser-473 on Akt, leading to increased GSK3␤ activity.
To further examine the relationship between the activities of Akt, GSK3␤, and caspase-3, we tested if receptor-mediated activation of Akt affected the changes in phosphorylation of Akt and GSK3␤ and caspase-3 activity induced by thapsigargin in SH-SY5Y cells. Administration of insulin-like growth factor-1 (IGF-1), a growth factor, which activates receptors endogenously expressed in SH-SY5Y cells known to activate Akt (61), counteracted the inhibitory effect of thapsigargin on Akt, causing an increase in phospho-Ser-473-Akt (Fig. 3C), blocked the  (24), and values are expressed as a percent of caspase activity in untreated cells. Means Ϯ S.E., n ϭ 3 experiments. PARP proteolysis (intact PARP is indicated by the band at 116 kDa, and proteolyzed PARP is indicated by the band at 85 kDa) and cleavage of procaspase-3 to active caspase-3 (ϳ17-and 19-kDa bands) were measured by immunoblot analyses. B, GSK3␤ activity was measured by immunoprecipitating GSK3␤ from cells 1-6 h after administration of 2 M thapsigargin, and measuring the phosphorylation of recombinant tau using [ 32 P]ATP as described previously (54). Quantitative values are expressed as a percentage of GSK3␤ activity in untreated cells. Means Ϯ S.E., n ϭ 3. C, ␤-catenin levels were measured in the indicated fractions following incubation with 2 M thapsigargin (Thp) for 3 h, 20 mM lithium for 4 h, or pretreatment with lithium for 1 h followed by treatment with thapsigargin for 3 h.
In contrast to the lack of effects of inhibitors of PP2B, treatment with the phosphatase inhibitors okadaic acid or calyculin A provided protection from the effects of thapsigargin at concentrations reported to be selective (65) for protein phosphatase 2A (although additional inhibition of protein phosphatase 1 cannot entirely be ruled out). Pretreatment with 1 M okadaic acid or 1 M calyculin A blocked thapsigargin-induced PARP proteolysis and increased the levels of phospho-Ser-9-GSK3␤ and of phospho-Ser-473-Akt (Fig. 3E). This is in accordance with previous reports that PP2A, but not PP1, dephosphorylates both GSK3␤ and Akt (66 -68). Thus, activation of GSK3␤ via dephosphorylation of phospho-Ser-9 is mediated by a PP2Adependent mechanism and activation of GSK3␤ promotes caspase-3 activation following thapsigargin treatment.
In addition to directly inhibiting GSK3␤ (44), lithium also has been reported to increase the inhibitory serine phosphorylation of GSK3 (29,69), indicating that lithium has dual mechanisms for inhibiting GSK3␤. Examination of phospho-Ser-9-GSK3␤ levels revealed that, although thapsigargin treatment caused a reduction, the level of phospho-Ser-9-GSK3␤ was increased in SH-SY5Y cells pretreated with lithium (Fig. 3F). This finding is in accordance with emerging evidence that lithium counteracts the effects of PP2A (70,71), to increase levels of phospho-Ser-9-GSK3␤, and suggests that lithium may protect cells from ER stress-induced caspase-3 activation both by direct inhibition of GSK3␤ and by increasing the inhibitory phosphorylation of Ser-9 of GSK3␤.
To determine if similar signaling activities are generated by ER stress in another neuronal model system, and particularly in non-proliferating cells, the responses to treatment with thapsigargin were examined in immortalized hippocampal cells that had been differentiated for 6 days. Treatment with 2 M thapsigargin resulted in activation of caspase-3 and proteolysis of PARP, although with a somewhat delayed time course compared with SH-SY5Y cells, and pretreatment with 20 mM lithium effectively blocked these effects, consistent with a facilitating effect of GSK3␤ on apoptotic signaling (Fig. 3G). A slightly higher concentration of thapsigargin (4 M) activated caspase-3 and caused proteolysis of PARP in differentiated hippocampal cells (Fig. 3H) in a similar time frame as was obtained with 2 M thapsigargin in SH-SY5Y cells. Thapsigargin treatment also caused a profound dephosphorylation of phospho-Ser-9-GSK and of phospho-Ser-473-Akt in differentiated hippocampal cells. Caspase-3 activation, PARP proteolysis, and dephosphorylation of GSK3␤ and Akt were blocked by the PP2A/PP1 inhibitors okadaic acid and calyculin A, but not by the PP2B inhibitors cyclosporin A and FK506, in differentiated hippocampal cells. Thus, although differentiated hippocampal cells displayed slightly less vulnerability to thapsigargin-induced caspase-3 activation than did SH-SY5Y cells, similar dependences on GSK3␤ and PP2A/PP1 were evident.
ER Stress Induced by Tunicamycin and Brefeldin-A-Similar, but not identical, results were obtained in SH-SY5Y cells using two other agents that cause ER stress, tunicamycin, which causes ER stress by inhibiting N-linked glycosylation and protein folding in the ER, and brefeldin-A, which perturbs ER-Golgi protein trafficking. Changes in SH-SY5Y cells caused by tunicamycin, probably the second most widely used agent to

FIG. 2. Regulation of GSK3␤ modulates thapsigargin-induced apoptosis. A, thapsigargin (2 M, 3 h)-induced PARP proteolysis was compared in control SH-SY5Y cells (WT), vector-transfected cells (V), and two different stable lines of SH-SY5Y cells that overexpress GSK3␤
3-to 4-fold above the endogenous level of GSK3␤ (GSK3 and GSK7). Light exposures are shown to minimize overexposure of the proteolyzed PARP band in the GSK3␤-overexpressing cells. Basal proteolyzed PARP was equivalent in all cell lines in the absence of treatment with thapsigargin (data not shown). B, 2 M thapsigargin-induced PARP proteolysis was nearly completely blocked in SH-SY5Y cells expressing a dominant-negative mutant of GSK3␤. C, wild-type SH-SY5Y cells, GSK3␤-overexpressing SH-SY5Y cells, and vector-transfected SH-SY5Y cells were pretreated for 1 h with 0, 1, 2, 5, 10, or 20 mM lithium and then incubated with 2 M thapsigargin for 3 h, and PARP proteolysis was measured. Lithium concentration-dependently attenuated thapsigargin-induced PARP proteolysis in control cells, and lithium was a more potent protector in control cells than in cells overexpressing GSK3␤.
induce ER stress after thapsigargin, were similar to those caused by thapsigargin. Treatment with tunicamycin (1-6 g/ ml) caused concentration-dependent increases in PARP cleavage and in the appearance of active caspase-3, and it reduced the level of phospho-Ser-9-GSK3␤ (Fig. 4A). Pretreatment with lithium completely blocked tunicamycin-induced PARP cleavage, activation of caspase-3, and dephosphorylation of GSK3␤ (Fig. 4A). The protection from the deleterious effects of tunicamycin afforded by lithium was compared with the effects of IGF-1 and phosphatase inhibitors. Fig. 4B shows that, as was found with thapsigargin, in SH-SY5Y cells treated with tunicamycin (2 g/ml), both lithium and IGF-1 attenuated PARP proteolysis, caspase-3 activation, and dephosphorylation of phospho-Ser-9-GSK3␤. Furthermore, the PP2A/PP1 inhibitors okadaic acid and calyculin A, but not the PP2B inhibitors FK506 and cyclosporin A, were similarly protective. Somewhat similar effects were obtained with brefeldin A. Treatment of SH-SY5Y cells with 10 g/ml brefeldin A for 4 h also activated caspase-3 and increased the proteolysis of PARP (Fig. 4C). As with thapsigargin and tunicamycin, apoptotic signaling induced by brefeldin A was blocked by pretreatment with lithium, IGF-1, and calyculin A and was unaffected by cyclosporin A. Unlike thapsigargin and tunicamycin, there was not an evident dephosphorylation of phospho-Ser-9-GSK3␤ 4 h after brefeldin A treatment, perhaps indicating that endogenous GSK3␤ need not be further activated to contribute to brefeldin A-induced apoptotic signaling. However, examination of early times after treatment with brefeldin A revealed a rapid, but transient, decrease in phospho-Ser-9-GSK3␤, indicating that all three treatments that induce ER stress cause dephosphorylation of phospho-Ser-9-GSK3␤ but with differences in the duration of the decrease. DISCUSSION Impaired ER function can cause accumulation of unfolded and misfolded proteins, actions that can initiate the apoptotic signaling cascade, and indications of neuronal ER stress have been identified in aging and a number of neurodegenerative conditions (1)(2)(3)(4). Furthermore, the ER appears to have a key role in apoptosis initiated from other cellular sites, because trafficking of members of the bcl-2 family of apoptosis regulators to the ER is a critical action modulating many types of apoptosis (72)(73)(74). Thapsigargin is one of the most useful agents available to identify cellular responses to ER stress because of its specific and potent action of inhibiting the Ca 2ϩ -ATPase in the ER (15), which leads to caspase-3-mediated apoptosis (18). The present investigation revealed a previously unrecognized and obligatory early step in thapsigargin-induced apoptosis, because thapsigargin treatment activated GSK3␤. This response was found to be due to dephosphorylation of phospho-Ser-9-GSK3␤, and blockade of caspase-3 activation by inhibition of GSK3␤ directly or by phosphatase inhibition demonstrated that this is a critical intermediate step coupling ER stress to caspase-3 activation.
A number of recent findings have linked GSK3␤ to apoptosis, but the present results represent the first to find that GSK3␤ is involved in the response to ER stress. For example, overexpression GSK3␤ was shown to be sufficient to induce apoptosis (23), moderately overexpressed GSK3␤ facilitated apoptosis induced by staurosporine or heat shock (24) or mitochondrial complex 1 inhibitors (25), and inhibition of the phosphatidylinositol 3-ki- nase pathway, which normally maintains phosphorylation of Ser-9-GSK3␤, led to GSK3␤-dependent apoptosis (35,36). These and other recent findings (reviewed in Ref. 21) support the concept that disinhibition of GSK3␤ promotes apoptotic signaling. The present findings expand the conditions in which GSK3␤ promotes apoptosis to those involving ER stress as activation of GSK3␤ was found to be a critical process in thapsigargin-induced apoptosis based on measurements of the effects of both increasing and decreasing GSK3␤ activity. Increased GSK3␤ activity achieved by its moderate overexpression facilitated thapsigargin-induced caspase-3 activation, whereas inhibition of GSK3␤ by expression of dominantnegative GSK3␤ greatly attenuated thapsigargin-induced caspase-3 activation. Furthermore, inhibition of GSK3␤ with lithium reduced caspase-3 activation to a level comparable with its inhibition of GSK3␤, suggesting that, although lithium has other targets, its inhibitory effect on GSK3␤ was likely to be the basis of its protective effect. This conclusion was further substantiated by the finding that higher concentrations of lithium were necessary for it to provide protection in cells overexpressing GSK3␤. These findings demonstrate that GSK3␤ is an important step in the initiation of apoptosis caused by thapsigargin treatment.
ER stress induced by thapsigargin treatment activated a signaling pathway that included dephosphorylation of phospho-Ser-473-Akt and activation of GSK3␤ through dephosphorylation of phospho-Ser-9-GSK3␤, and PP2A/PP1 inhibitors blocked these dephosphorylation actions and caspase-3 activation. PP2A is known to dephosphorylate phospho-Ser-9-GSK3␤ (66, 67), and PP2A but not PP1 was shown to dephosphorylate Akt (68,75). With thapsigargin treatment, PP2A/PP1 inhibitors blocked dephosphorylation of both Akt and GSK3␤, thus, it is not possible to distinguish between two possible PP2A-dependent mechanisms causing GSK3␤ dephosphorylation and activation: this could result from a direct effect of PP2A on phospho-Ser-9-GSK3␤ or as an indirect effect of PP2A resulting from the inactivation of Akt. Presently, it appears that both mechanisms may contribute to GSK3␤ activation following thapsigargin treatment. Protein phosphatases constitute a critical component of many signaling systems that initiate the apoptotic program. For example, both PP2B (64) and PP2A (76) dephosphorylate phospho-BAD, thereby converting it to its proapoptotic dephosphorylated form, as well as regulating other proteins involved in apoptosis. However, in SH-SY5Y cells treated with thapsigargin, PP2B activation occurs after caspase activation in a caspase-dependent manner, suggesting that PP2B contributes to late events in cell death (57). In contrast to the lack of effects of inhibitors of PP2B on caspase-3 activation induced by thapsigargin, PP2A inhibitors blocked FIG. 4. Comparison of the protective effects of lithium, IGF-1, and phosphatase inhibitors on ER stress signaling induced by tunicamycin and brefeldin A. A, SH-SY5Y cells were pretreated with 20 mM lithium for 1 h, followed by incubation with 1, 2, 4, or 6 g/ml tunicamycin for 3 h, and proteolyzed PARP, active caspase-3, and phospho-Ser-9-GSK3␤ were measured in immunoblots. B, SH-SY5Y cells were pretreated with 20 mM lithium for 1 h, or for 30 min with 50 ng/ml IGF-1, 1 M okadaic acid (OA), 1 M calyculin A (Cal), 1 M FK506 (FK), or 1 M cyclosporin A (Cyc), followed by incubation with 2 g/ml tunicamycin for 3 h, and proteolyzed PARP, active caspase-3, phospho-Ser-9-GSK3␤, and phospho-Akt were measured in immunoblots. Quantitative values are expressed as a percentage of PARP proteolytic fragment in untreated cells. Means Ϯ S.E., n ϭ 3. C, SH-SY5Y cells were pretreated with 20 mM lithium for 1 h, or for 30 min with 50 ng/ml IGF-1, 1 M calyculin A (Cal), or 1 M cyclosporin A (Cyc), followed by incubation with 10 g/ml brefeldin A for 4 h, and proteolyzed PARP, active caspase-3, phospho-Ser-9-GSK3␤, and phospho-Ser-473-Akt were measured in immunoblots. Phospho-Ser-9-GSK3␤ was measured after treatment with 10 g/ml brefeldin A for 15, 30, 60, or 120 min. Quantitative values are expressed as a percentage of PARP proteolytic fragment or active caspase-3 in untreated cells. Means Ϯ S.E., n ϭ 3. caspase-3 activation, indicating that PP2A contributes to an early event in the apoptotic program initiated by thapsigargin. It is notable that another cell stress, hyperosmotic stress, inactivated Akt via dephosphorylation mediated by PP2A without changing the cellular activity of PP2A (75). Thus, the action of PP2A appears to be important in the apoptotic signaling cascades following both hyperosmotic stress and ER stress induced by thapsigargin, perhaps through subcellular re-localization of the enzyme. PP2A and GSK3␤ also were implicated in the signaling mechanisms leading to caspase-3 activation following ER stress induced by tunicamycin and by brefeldin A. With both agents, caspase-3 activation was attenuated by lithium, IGF-1, and inhibition of PP2A, but not inhibition of PP2B. Although the mechanism whereby thapsigargin treatment leads to PP2A-mediated dephosphorylation of Akt and GSK3␤ is unknown, it is evident that dephosphorylation of the inhibitory site on GSK3␤ contributes to early events in apoptotic signaling. There are many potential targets whereby GSK3␤ could promote apoptosis. For example, GSK3␤ directly phosphorylates, and thereby regulates, at least eight transcription factors, consequentially impacting the expression of many genes (reviewed in Ref. 21). Among these are several survivalpromoting transcription factors, such as cAMP response element-binding protein, of which inhibition by GSK3␤ (49) may contribute to facilitation of apoptosis. Also, activation of GSK3␤ inhibits protein synthesis (77), a logical response to ER stress to reduce the further production of proteins, and this action has been linked to the promotion of apoptosis by GSK3␤ (78). However, the one or more precise proapoptotic targets of GSK3␤ during ER stress remain to be identified. Considering the well-known antiapoptotic actions of Akt (79 -81) along with the proapoptotic actions of GSK3␤ (reviewed in Ref. 21), it is evident that inactivation of Akt and activation of GSK3␤ following thapsigargin-induced ER stress provides a strong stimulus for activation of the apoptotic program.
ER stress appears to be an early event contributing to neuronal dysfunction and death in AD and likely also in aging and in other neurodegenerative conditions (1)(2)(3)(4). Many pathways involving multiple kinases have been identified that inhibit GSK3␤ activity, but fewer mechanisms are known that increase the activity of GSK3␤. Activation of GSK3␤ primarily requires dephosphorylation of phospho-Ser-9-GSK3␤, an action previously shown to be mediated by PP2A and here shown to be caused by ER stress. This raises the possibility that GSK3␤ may be activated by ER stress in neurodegenerative conditions where this occurs, such as in affected neurons in AD. In this regard, it is relevant that GSK3␤ is a prime candidate kinase for causing the hyperphosphorylation of tau that is associated with neurofibrillary tangles in AD (reviewed in Ref. 21). In the few studies that have assessed GSK3␤ in AD, increased levels of GSK3␤ were found in AD, compared with non-diseased, human brain; immunohistochemical measurements found GSK3␤ associated with neurofibrillary tangles in AD brain (82)(83)(84)(85)(86); and active GSK3␤ was found to be accumulated in pretangle neurons (86). Furthermore, the AD-associated A␤ peptide is known to activate GSK3␤ (22), and the GSK3␤ inhibitor lithium provides protection from A␤ toxicity (87,88). Taken together, these studies suggest that alterations in the control of GSK3␤ may result from ER stress, which could contribute to the neuropathology of AD as well as other neurodegenerative conditions in which ER stress occurs.