The Wnt Pool of Glycogen Synthase Kinase 3β Is Critical for Trophic-Deprivation-Induced Neuronal Death

Abstract

Glycogen synthase kinase 3 (GSK-3) is implicated in neuronal death through a causal role, and precise mechanisms have not been unambiguously defined. We show that short hairpin RNA (shRNA) knockdown of GSK-3β, but not GSK-3α, protects cerebellar granule neurons from trophic-deprivation-induced death. Using compartment-targeted inhibitors of the Wnt-regulated GSK-3 pool, NLS-FRAT1, NES-FRAT1, and axin-GSK-3-interacting domain (axin-GID), we locate proapoptotic GSK-3 action to the cytosol and regulation of Bim protein turnover despite constitutive cycling of GSK-3 between the cytosol and nucleus, revealed by leptomycin B. We examine the importance of Ser21/9 (GSK-3α/β) phosphorylation on proapoptotic GSK-3 function. Neurons isolated from GSK-3α/β^S21A/S9A knock-in mice survive normally and are fully sensitive to trophic-deprivation-induced death. Nonetheless, inhibition of GSK-3 catalytic activity with lithium or SB216763 protects GSK-3α/β^S21A/S9A neurons from death. This indicates that dephosphorylation of GSK-3β/Ser9 and GSK-3α/Ser21 is insufficient for GSK-3 proapoptotic function and that another level of regulation is required. Gel filtration reveals a stress-induced loss of neuronal GSK-3β from a high-molecular-mass complex with a concomitant decrease in axin-bound GSK-3β. These data imply that Wnt-regulated GSK-3β plays a nonredundant role in trophic-deprivation-induced death of neurons.

Glycogen synthase kinase (GSK) is a ubiquitously expressed serine/threonine kinase that has received considerable attention from drug companies because of its association with major diseases of the nervous system, for example, Alzheimer's disease, stroke, and mood disorders (22, 33, 37) as well as diabetes (48). The evidence for potential GSK-3 involvement in brain pathologies stems largely from studies in cell culture where inhibition of GSK-3 using lithium or small-molecule inhibitors protects against a range of insults, such as excitotoxicity, trophic factor withdrawal, and β-amyloid-induced death (reviewed in reference 23). GSK-3 is believed to influence Alzheimer's disease pathology at multiple levels; GSK-3 can phosphorylate tau on residues that contribute to paired helical filament formation, and GSK-3 is associated with presenilin 1 toxicity and with cleavage of amyloid precursor protein (APP), which leads to amyloid plaque formation (19, 33, 35, 40). Importantly, the therapeutic efficacy of lithium for treatment of mood disorders may result from inhibition of GSK-3 (34; reviewed in reference 22). Nonetheless, defining a causal role for GSK-3 in neuropathology has been confounded by the lack of selectivity of the inhibitors used. GSK-3β knockout mice are embryonic lethal (26), precluding their use for loss-of-function genetic studies. Gene silencing thus provides an alternative approach to verify whether GSK-3 plays a requisite role in neuronal death.

There are two genes for GSK-3, the GSK-3α and GSK-3β genes, which share 85% sequence identity and are both highly expressed in the brain (47). GSK-3α and GSK-3β show similar substrate specificity and are inhibited to a similar extent by lithium and by small-molecule GSK-3 inhibitors (12, 33). In spite of these similarities, they serve nonredundant functions during development, and GSK-3β-deficient mice are embryonic lethal due to severe liver degeneration (26). Whether GSK-3α and -3β display functional redundancy in regulating neuronal cell death has not been reported. GSK-3 activity can be negatively regulated by either insulin/growth factor signaling or by the Wnt pathway, both events leading to distinct functional outcomes. In response to insulin or growth factors, many protein kinases can phosphorylate the serine 9 of GSK-3β (serine 21 of GSK-3α), among them Akt, protein kinase A, and pp90Rsk (1). Phosphorylation of this N-terminal serine leads to autoinhibition of kinase activity via a pseudosubstrate mechanism (16). In the absence of growth factor signaling, the pseudosubstrate domain vacates the substrate docking site, thereby enabling GSK-3 to bind and phosphorylate targets (reviewed in reference 13). The second mechanism conferring negative regulation on GSK-3 is the Wnt cascade. Wnt negatively regulates GSK-3 activity by a poorly defined mechanism that involves a multiprotein complex (1). In the presence of Wnt stimulation, GSK-3 is unable to phosphorylate Wnt cascade targets, such as β-catenin. Upon removal of the Wnt ligand, GSK-3 activity is derepressed and phosphorylates β-catenin. Interestingly, the Wnt-regulated pool of GSK-3 is insulated from the insulin/growth factor-regulated pool, as it is independent of GSK-3β/serine 9 phosphorylation (17). Notably, small-molecule inhibitors of GSK-3 and lithium inhibit equally both Wnt- and insulin-regulated GSK-3β pools (12).

In addition to regulation by posttranslational phosphorylation and interactions with scaffold proteins, GSK-3 function can be regulated by subcellular localization. Although GSK-3 predominates in the cytosol in neurons, stress induces the accumulation of nuclear and mitochondrial GSK-3 activity in neuroblastoma cells (3, 4). This raises the possibility that GSK-3 may execute its proapoptotic function in a subcellular compartment that is distinct from the cytosol where GSK-3 predominates in unstressed neurons.

In this study we demonstrate using gene silencing that GSK-3β is a critical player in trophic-deprivation-induced death of freshly isolated cerebellar granule neurons. We explore the importance of Akt versus Wnt-regulated GSK-3β in death of neurons from homozygous knock-in mice in which serine 9 of GSK-3β and serine 21 of GSK-3α are mutated to alanine (32). Our data show that the GSK-3α/β Ser21/9 phosphorylation state is not critical for death. Instead we observe a displacement of GSK-3β from an axin-bound complex in response to trophic deprivation. Moreover, exogenous expression of inhibitors of the Wnt-regulated GSK-3 pool protects neurons from death. Together these data implicate that the Wnt-regulated pool of GSK-3β is instrumental in cerebellar granule neuron death upon withdrawal of trophic support.

MATERIALS AND METHODS

Reagents, antibodies, and plasmids.

LY294002 was from Alexis, SB216763 was from Calbiochem, and cycloheximide was from Sigma-Aldrich. Phospho-specific antibodies to Akt Thr 308 (catalog no. 9275S) and phospho-GSK3-β Ser9 (catalog no. 93365) were from Cell Signaling Technologies (Beverly, MA). Monoclonal antibodies to GSK-3β (catalog no. G22320) and β-catenin (catalog no. C19220) were from Transduction Laboratories (Lexington, KY). Monoclonal antibodies to green fluorescent protein (GFP) (catalog no. 8371), Bim-EL (catalog no. AAP-330), GSK-3α (catalog no. 7389), and GSK-3α (clone H-12) were from Clontech, Stressgen Biotechnologies, Upstate Biotechnology Incorporated, and Santa Cruz Biotechnology, respectively. pEGFP-FRAT1 was described previously (27) and 3xNLS- and NES-targeted FRAT1 were obtained by excising human FRAT1 from pEGFP-FRAT1 and ligating into pEGFP-NLS or pEGFP-NES as described previously (5). Rat GSK-3α and -3β were obtained by PCR from rat cDNA using the following oligonucleotides: BglII-rGSK-3α (+) (AGATCTATGAGCGGCGGCGGGCCTTC), BglII-rGSK-3α (−) (TCAGGAAGAGTTAGTGAGGGTAGG), BglII-rGSK-3β (+) (AGATCTATGTCGGGGCGACCGAGAAC), and BglII-rGSK-3β (−) (TCAGGTAGAGTTGGAGGCTG). Rat GSK-3βi was obtained by PCR from pEGFP-rGSK-3β using the following oligonucleotides: GGACAAGCGATTTAAGAACCGAGAGCTCCAGATCATGAGAAAGCTAGATCACT on the plus strand and AGTGATCTAGCTTTCTCATGATCTGGAGCTCTCGGTTCTTAAATCGCTTGTCC on the minus strand. PCR products were ligated into BglII/SalI sites of pEGFP-C1, pEGFP-NES, and pEGFP-NLS as described previously (5). Rat axin-GID (nucleotides 1114 to 1705) was obtained by PCR-based methods from rat cDNA and ligated into pVenus-C3 or pGEX-6P2. GSK-3α and -3β shRNAs, cJun-Asp^{58, 62, 63, 73, 91, 93}, and leptomycin B were generous gifts from David Turner (Michigan), Dirk Bohmann (Rochester, NY), and Minory Yoshida (Tokyo, Japan).

Antibody generation and purification.

Polyclonal α-axin antibodies were raised against bacterially expressed GST-axin-GID (nucleotides 1114 to 1705). Rabbits were inoculated by PickCell Laboratories (Leiden, The Netherlands). Crude antiserum was affinity purified using recombinant GST-axin-GID coupled to Affigel-10 Sepharose (Bio-Rad) according to the manufacturer's instructions.

Cell culture and trophic deprivation treatment.

Cerebellar granule neurons were prepared from 7-day Sprague Dawley rats or from wild-type or GSK-3α/β^21A/9A knock-in mice (32) as previously described (10). Cells were plated (250,000 per cm²) onto poly-l-lysine-coated dishes or 10.5- by 10.5-mm coverslips as required. For trophic deprivation treatment, cerebellar granule neurons at 7 days in vitro (DIV) were changed from medium containing high KCl (25 mM) and 10% fetal calf serum (FCS) to medium containing low KCl (5 mM) without FCS. When inhibitors were used, they were added 30 min prior to trophic deprivation. To measure Bim induction, 7 DIV neurons (3.5-cm-diameter dishes) were switched to low-KCl medium containing 10% of dialyzed FCS for 8 h in the presence or absence of inhibitors as indicated. Cells were washed with phosphate-buffered saline (PBS) and lysed for 15 min in 150 μl lysis buffer (20 mM Tris [pH 7.6], 140 mM NaCl, 1% NP-40, 10% glycerol, 500 μM Na₃VO₄, 1 μg/ml each of aprotinin, leupeptin, and pepstatin A, and 100 μg/ml of phenylmethylsulfonyl fluoride [PMSF]) at 4°C as described previously (42). Lysates were homogenized with eight strokes of a 27-gauge syringe and centrifuged at 15,700 × g for 15 min at 4°C. The pellets were resuspended in 20 μl of Laemmli sample buffer and immunoblotted for Bim-EL.

Transfection and gene silencing.

Rat cerebellar granule neurons at 5 or 6 DIV were transfected using the calcium phosphate method as previously described (10). pEGFP-CAAX (0.2 μg) was used as a transfection marker together with 0.8 μg GSK-3β shRNA (GAUCUGGAGCUCUCGGUUCU) and GSK-3α shRNA (GUGGAUGUAGGCCAAGCUCC) as previously described (50) or with control, nontargeting shRNA (UAGCCUCUAUCUAGUCCAU). For add-back experiments, neurons were transfected with 0.8 μg GSK-3β or control shRNA together with 0.1 μg pEGFP-GSK-3β, or shRNA-insensitive mutants (pEGFP-GSK-3βi and pEGFP-NES-GSK-3βi) as shown in the figures. At 8 days posttransfection, cells were changed to low-KCl medium without FCS for 24 h after which they were fixed with 4% paraformaldehyde and stained with Hoechst 33342 (1:500). Transfected cells with pyknotic nuclei were scored as dead cells. For analysis of compartment-targeted GSK-3 inhibitors, neurons at 6 DIV were transfected with 0.4 μg pEGFP-CAAX and 1.6 μg of pEGFP-FRAT-1, pEGFP-NLS-FRAT1, or pEGFP-NES-FRAT1 as described in the figure legends. At 24 h posttransfection, cells were switched to low-KCl medium for 24 h. To evaluate AP-1-induced death, neurons at 6 DIV were transfected with 0.5 μg pEGFP-CAAX, 0.7 μg pMT161-HA (cJun-Asp^{58, 62, 63, 73, 91, 93}), 0.8 μg pEGFP, pEGFP-NES-FRAT1, pEGFP-NLS-FRAT1, or pVenus-axin-GID. After 18 h, cells were changed to serum-free medium containing 25 mM KCl for 24 h before the cells were fixed as previously described (27). SB216763 (3 μM) was added where indicated 30 min prior to medium change.

Immunoblotting.

To determine the half-life of GSK-3β, cerebellar granule neurons at 6 DIV were treated with cycloheximide (50 μM) for 2 to 48 h as indicated. Samples were lysed in Laemmli sample buffer and immunoblotted for GSK-3β. To evaluate the relative expression of GSK-3α and -3β in cerebellar granule neuron lysates, lysates of HEK-293 cells expressing GFP-labeled GSK-3α (GFP-GSK-3α) or -3β were used as standards. In order to normalize the concentration of GFP-GSK-3α and -3β, lysates were immunoblotted for GFP followed by densitometry. Equal levels of GFP-GSK-3α and -3β were then loaded alongside neuronal lysate and blotted with GSK-3α- and GSK-3β-specific antibodies.

Analysis of β-catenin.

Cerebellar granule neurons from wild-type or GSK-3α/β^21A/9A knock-in mice were plated on 12-well plates and switched to low-KCl medium for 4 h. Cells were washed twice with PBS and lysed in 100 μl of 4°C homogenization buffer (20 mM Tris [pH 7.5], 1 mM EDTA, 25 mM NaF, 1 μg/ml each of aprotinin, leupeptin, and pepstatin A, and 100 μg/ml of PMSF). Lysates were homogenized using 15 strokes of a 27-gauge syringe, and membrane fractions were removed by centrifugation at 16,000 × g for 30 min at 4°C. Laemmli sample buffer was added to the membrane fractions and remaining supernatants (cytosol), and the samples were immunoblotted as shown in the figures.

Immunostaining and leptomycin B treatment.

Cerebellar granule neurons at 7 DIV were treated with leptomycin B (10 ng/ml) for the times indicated in the figures after which cells were fixed and immunostained with 1:100 anti-mouse GSK-3β and detected with 1:500 Alexa Fluor 555-labeled anti-mouse antibody. Staining with GSK-3α antibody was carried out using 1:400 mouse anti-GSK-3α (H-12) and detected with 1:700 Alexa Fluor 568-labeled anti-mouse antibody. Nonspecific background associated with the primary antibody was obtained on coverslips that underwent transfection. This was reduced sufficiently to visualize loss of signal in cells expressing GSK-3α shRNA, by blocking in goat serum and including multiple washes with PBS containing 2% Tween 20 at each step of the procedure. DNA was stained with Hoechst 33342 (1:500).

Immunoprecipitation.

Cerebellar granule neurons plated on 6-cm dishes were deprived of trophic support for 4 h, washed with ice cold PBS, and lysed in 1 ml immunoprecipitation buffer (20 mM HEPES [pH 7.4], 2 mM EGTA, 50 mM β-glycerophosphate, 1 mM dithiothreitol, 1 mM Na₃VO₄,1% Triton X-100, 10% glycerol, 50 mM NaF, 1 mM benzamadine,1 μg/ml each of aprotinin, leupeptin, and pepstatin A, and 100 μg/ml of PMSF), homogenized with six strokes of a 27-gauge syringe, and centrifuged at 15,700 × g for 15 min at 4°C. Supernatants were incubated overnight at 4°C with 10 μl of affinity-purified antiaxin antibody. Antibody complexes were sequestered using protein A-Sepharose and washed five times with PBS. Axin-bound GSK-3β was detected by immunoblotting for GSK-3β.

Gel filtration chromatography.

Seven DIV cerebellar granule neurons on 10-cm dishes were switched to low-KCl medium for 2 h, resuspended in ice cold PBS, and collected by centrifugation at 100 × g for 1 min at 4°C. The pellet was resuspended in 800 μl elution buffer (20 mM Na₂-β-glycerophosphate [pH 7.0], 30 mM NaF, 2 mM EDTA, 1 mM dithiothreitol, 2 mM Na₄P₂O₇, 0.5% Igepal, 1 μg/ml each of aprotinin, leupeptin, and pepstatin A, 0.2 mM Na₃VO₄, 100 μg/ml PMSF) and homogenized with a 27-gauge syringe (eight strokes). Gel filtration chromatography was performed using an ÄKTAbasic fast-performance liquid chromatography system (Amersham Biosciences). Precleared lysates were loaded on a Superose 6 column (1.6 by 40 cm) preequilibrated in elution buffer without Igepal. Proteins were eluted in a volume of 77 ml at a flow rate of 1 ml per minute. The first 20 ml (void volume) was discarded after which 55 1.5-ml fractions were collected and 250-μl aliquots were precipitated with 1.25 ml of −20°C acetone for 10 min and collected by centrifugation at 13,400 × g at 4°C. Precipitates were dried in a SpeedVac, resuspended in Laemmli sample buffer, and immunoblotted.

Isolation of mRNA and reverse transcription-PCR.

Cerebellar granule neurons at 7 DIV were treated with inhibitors as indicated in the figures and switched to low-KCl medium for 8 h. Total RNA was isolated using the Qiagen mRNAeasy kit according to the manufacturer's instructions. RNA was heat denatured for 3 min at 72°C, reverse transcribed using 200 U Moloney murine leukemia virus reverse transcriptase (Promega), Moloney murine leukemia virus buffer (provided by the manufacturer), 1.25 mM deoxynucleoside triphosphate (Finnzymes), and 30 U of prime RNase inhibitor (Eppendorf), and incubated at 42°C for 90 min followed by termination for 5 min at 95°C. PCR parameters were as follows: for bim, 1 min at 94°C and 30 cycles (1 min at 94°C, 1 min at 64°C, and 0.5 min at 72°C); and for β-actin, 1 min at 95°C and 25 cycles (1.5 min at 95°C, 1.5 min at 59°C, and 1.5 min at 72°C). Primers used were as follows: bim (+) (5′-CTACCAGATCCCCACTTT TC-3′), bim (−) (5′-GCCCTCCTCGTGTAAGTC TC-3′), β-actin (+) (5′-TCCGGAGACGGGGTCACCCA-3′), and β-actin (−) (5′-CTAGAAGCATTTGCGGTGCACG-3′) as previously described (11).

Statistical analysis.

Statistical analysis was done using Student's t test. Significance levels at P < 0.05, P < 0.01, and P < 0.001 were used.

RESULTS

Freshly isolated cerebellar granule neurons survive in culture if maintained in elevated KCl (25 mM). The trophic support conferred by high KCl is believed to mimic the innervation of these neurons by mossy fibers in vivo in the developing brain (6). Once the neurons have matured in culture, changing the culture medium to low KCl (5 mM; trophic deprivation) results in apoptotic death. The signaling events underlying this form of death are well documented. Among the molecules believed to be important is GSK-3, and treatment of cells with structurally independent inhibitors of GSK-3 lithium (10 mM) or SB216763 (3 μM) protects the cells from trophic-deprivation-induced death (Fig. 1A). However, small-molecule inhibitors of GSK-3 typically cross-react with members of the cyclin-dependent kinase family which are also implicated in nerve cell death (33). Moreover, the pharmacology of these inhibitors is not completely characterized, and they do not distinguish between GSK-3α and GSK-3β (15, 33). To clarify whether GSK-3α and -3β are critical players in neuronal death, we used gene silencing. In neurons, analysis of GSK-3β protein stability revealed a relatively long half-life of 48 h (Fig. 1B). We therefore expressed GSK-3α and -3β shRNAs for 8 days after which GSK-3β was undetectable by immunostaining (Fig. 1C). GSK-3α expression was consistently reduced in cells transfected with GSK-3α shRNA though it was not completely eradicated. The residual signal that remained following 8 days expression of GSK-3α shRNA most likely results from nonspecific signal associated with the primary antibody (see Materials and Methods for more details). We also examined the relative potency of GSK-3α and -3β shRNAs against rat GFP-GSK-3α and -3β expressed in HEK-293 cells. Both shRNAs showed comparable knockdown of the respective GSK-3 isoforms following 3 days expression (Fig. 1D). To examine the influence of GSK-3α/β knockdown on neuronal survival following trophic deprivation, silencing was carried out for 8 days. Silencing of endogenous GSK-3β expression provided complete protection from trophic-deprivation-induced death, while cells expressing GSK-3α shRNA or nontargeting (control) shRNA were fully sensitive (Fig. 1E). We postulated that this could be due to relatively low expression of GSK-3α compared to GSK-3β. We therefore measured the relative expression of GSK-3α and -3β in cerebellar granule neurons using normalized GFP-GSK-3α and -3β as internal standards (Fig. 1F). The levels of GSK-3α and GSK-3β were comparable with slightly more (1.6-fold) GSK-3α than GSK-3β. GSK-3α expression was unaltered in cells transfected with GSK-3β shRNA (not shown). These data establish that GSK-3β is a critical effector of neuronal death and that isoform-specific silencing of GSK-3β expression is sufficient to provide neuroprotection.

FIG. 1.

Suppression of GSK-3β expression using shRNA protects cerebellar granule neurons from trophic-deprivation-induced death. (A) Cerebellar granule neurons were cultured under control conditions or deprived of trophic support (trophic deprivation [TD]) for 24 h in the presence or absence of GSK-3 inhibitors LiCl (10 mM) or SB216763 (3 μM). Representative micrographs of nuclei stained with Hoechst 33342 are shown. Inhibitors of GSK3 prevent TD-induced pyknosis. (B) To determine the half-life of GSK-3 protein, cerebellar granule neurons (6 DIV) were treated with cycloheximide (50 μM) for the indicated times after which GSK-3β levels were analyzed by immunoblotting. The half-life (T_1/2) of GSK-3β in mature neurons was 48 h. (C) Cerebellar granule neurons (5 DIV) were transfected with shRNAs targeting GSK-3β and GSK-3α, respectively, together with GFP-CAAX to mark transfected cells (white arrows) and incubated for 8 days. Representative confocal sections of transfected neurons (green), GSK-3α and GSK-3β immunostaining (red), and Hoechst 33342 staining are shown. Expression of GSK-3 shRNAs led to considerable suppression of GSK-3α and GSK-3β levels. (D) HEK 293T cells were transfected with GFP-GSK-3α or GFP-GSK-3β together with control nontargeting, GSK-3α or -3β-targeted shRNA. After 3 days of expression, GFP-GSK-3α and GFP-GSK-3β expression was examined by immunoblotting. The percent knockdown is show at the bottom of the panel. (E) Cerebellar granule neurons transfected with GSK-3α shRNA, GSK-3β shRNA, or control shRNA together with GFP-CAAX as a transfection marker were deprived of trophic support for 24 h. Cells with pyknotic nuclei were scored as dead cells. The data shown are means plus standard errors of the means (error bars) for 12 or 13 replicates. Significance levels with respect to corresponding control shRNA-expressing cells are shown (***, P < 0.001; ns, not significant). The number of neurons counted for each condition is shown above the respective histogram bar. (F) To determine the relative expression of GSK-3α and -3β in neurons, 7 DIV cerebellar granule neuron (CGN) lysate was run alongside GFP-GSK-3α and GFP-GSK-3β standards (normalized for equal expression) obtained from expression in HEK-293 cells. Gels were immunoblotted using antibodies against GFP, GSK-3α, and GSK-3β as indicated. Densitometric analysis revealed that cerebellar granule neurons expressed on average 1.6 times more GSK-3α than GSK-3β. The positions of molecular mass standards (in kilodaltons) are shown to the left of the gels.

It is commonly accepted that compartmental localization of kinases can confer specificity for distinct functions. To understand better the proapoptotic mechanism of GSK-3β in neuronal death, we examined its subcellular localization. Confocal analysis revealed that GSK-3β localized predominantly to the cytosol in mature neurons. Interestingly, however, inhibition of carrier-mediated nuclear export with leptomycin B (49) revealed a rapid accumulation of GSK-3β in the nucleus. There was a clear increase in nuclear GSK-3β by 30 min following leptomycin B treatment, and by 4 h, GSK-3β had concentrated in the nucleus (Fig. 2A and B), indicating that GSK-3β cycles continuously between the cytosol and nucleus in these neurons. To test whether GSK-3β accumulated in the nucleus in response to physiological stress, we examined the localization of GSK-3β in neurons deprived of trophic support. There was no net accumulation of GSK-3β in the nucleus following stress (Fig. 2C). To test whether the rate of GSK-3β cytosolic-nuclear cycling was sensitive to stress, we repeated these experiments in the presence of leptomycin B. There was no detectable change in the rate of GSK-3β accumulation in the nucleus in stressed neurons (unpublished data). These data indicate that cytosol-nuclear cycling of GSK-3β is not altered following withdrawal of trophic support.

FIG. 2.

GSK-3β cycles between nuclear and cytoplasmic compartments and is predominantly cytosolic. (A) Confocal sections through nuclei of cerebellar granule neurons (7 DIV) treated with leptomycin B (10 ng/ml) are shown. GSK-3β accumulates in nuclei within 30 min (30′) of treatment. (B) Line profiles of fluorescence intensity through nuclei are shown before and after leptomycin B treatment (inset). (C) Cerebellar granule neurons (7 DIV) were plated in low-KCl medium as indicated (trophic deprivation). GSK-3β localization was assessed by confocal microscopy. −1°, staining in the absence of primary antibody. There was no net nuclear accumulation of GSK-3β following trophic deprivation.

To determine in which compartment GSK-3β exhibited proapoptotic function, we utilized the GSK-3β binding proteins FRAT1 and axin-GID (16). These proteins localize in the cytoplasm where they bind to and disrupt GSK-3β dimer formation (16). In addition, binding of FRAT1 to GSK-3β selectively inhibits GSK-3-catalyzed phosphorylation of Wnt pathway targets (41). Enhanced GFP (EGFP)-FRAT1 was artificially directed to the nucleus or cytoplasm by insertion of three nuclear localization sequences (NLS) or three nuclear export sequences (NES), upstream of the FRAT1 coding sequence. The localization of EGFP-FRAT1, EGFP-NLS-FRAT1, EGFP-NES-FRAT1, and Venus-axin-GID was examined. As expected, wild-type EGFP-FRAT1 localized to the cytoplasm (21) as did Venus-axin-GID, while EGFP-NLS-FRAT1 concentrated in the nucleus (Fig. 3A). Expression of GFP-NLS-FRAT1 caused a partial sequestration of endogenous GSK-3β to the nucleus, indicating that these molecules interact stably (Fig. 3A). We next examined the effects of these compartment-targeted inhibitors on neuronal survival in response to trophic deprivation. Overexpression of GSK-3β binding proteins in the cytosol provided significant protection from death, whereas the nucleus-targeted FRAT1 was significantly less protective (Fig. 3B). These results suggest one of two mechanisms of protection by FRAT1; either NES-FRAT1 isolates GSK-3β from nuclear death effectors by tethering it in the cytosol, or it prevents GSK-3 from phosphorylating its cytosolic targets as has been reported for axin and β-catenin (41). With these two possible mechanisms in mind, it is notable that the accumulation of GSK-3β in the nucleus per se was not toxic, as no death was observed in neurons expressing NLS-FRAT1 in the absence of stress (Fig. 3B). We would speculate that the low level of protection that did occur with NLS-FRAT1 is likely due to the presence of residual NLS-FRAT1 in the cytoplasm (Fig. 3B, inset). We propose that proapoptotic GSK-3β acts in the cytoplasm, as targeting of FRAT1 to the nucleus results in significantly less protection. In support of this, overexpression of axin-GID, which like NES-FRAT1 resides in the cytosol, is also neuroprotective (Fig. 3B). Notably, the levels of expression of GFP-NES-FRAT1 and GFP-NLS-FRAT1 in neurons were equal (Fig. 3C).

FIG. 3.

Cytosolic-targeted GSK-3β inhibitors protect from trophic-deprivation-induced death. (A) The compartmental localization of GFP-FRAT1, GFP-NLS-FRAT1, and GFP-NES-FRAT1 was assessed by confocal microscopy. Corresponding sections of endogenous GSK-3β and Hoechst 33342 staining are shown. (B) Cerebellar granule neurons (6 DIV) were transfected with compartment-targeted inhibitors of the Wnt-regulated GSK-3 pool and then deprived of trophic support for 24 h (tropic deprivation) as shown. The cells were stained with Hoechst 33342, and cells with pyknotic nuclei were scored as dead cells. GFP-NES-FRAT1 and wild-type GFP-FRAT1 provide significantly greater protection than GFP-NLS-FRAT1. Data are means of the percent survival of transfected cell population plus standard errors of the means. Values that were significantly different (P < 0.05) from the corresponding control values are indicated by an asterisk. The number of cells counted is indicated above each histogram bar. A representative image of a cell expressing GFP-NLS-FRAT1 is shown (inset) to illustrate the presence of low-level GFP-NLS-FRAT1 in the cytoplasm. (C) The relative expression of GFP-NLS-FRAT1, GFP-FRAT1, or GFP-NES-FRAT1 was evaluated by expressing in cerebellar granule neurons. Lysates were immunoblotted with GFP antibody. The addition of targeting sequences NES or 3xNLS induced a retarded mobility on sodium dodecyl sulfate-polyacrylamide gels consistent with the increasing size of these proteins. Nonspecific bands are labeled with an asterisk. (D) To assess the contribution of GSK-3 pools to AP-1-induced death, neurons were cotransfected with cJun-Asp in the presence or absence of compartment-targeted GSK-3 inhibitors or treated with SB216763 (3 μM) as shown. GFP-NES-FRAT1 provided significantly more protection than GFP-NLS-FRAT1. Means plus standard errors of the means (error bars) are shown. Values that are significantly different from the control value are indicated (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

Trophic-deprivation-induced death of cerebellar granule neurons is protein synthesis dependent, and AP-1-dependent transcription is thought to be pivotal (43, 45). We have previously shown that inhibition of GSK-3 activity protects neurons from death resulting from expression of a constitutively active c-Jun (cJun-Asp) (27). To determine whether the cytosolic or nuclear pool of GSK-3 was involved in cJun-Asp-induced death, cells were cotransfected with NLS-FRAT1 to block nuclear GSK-3 action and with NES-FRAT1 or axin-GID to block cytosolic GSK-3 action. Expression of NES-FRAT1 or axin-GID, but not NLS-FRAT1, provided robust protection, suggesting that as with trophic deprivation-induced stress, the cytosol was the location of proapoptotic GSK-3 action in response to cJun-Asp expression (Fig. 3D).

As it was difficult to distinguish unambiguously between the two modes of action of FRAT1, sequestration and substrate inhibition, we took an independent approach to assess which pool of GSK-3β mediated neuronal death. We tested whether an NES-targeted GSK-3β mutant that was insensitive to GSK-3β shRNA could resensitize neurons lacking endogenous GSK-3β to trophic-deprivation-induced death. The sequences of the silent mutations leading to shRNA insensitivity in the GSK-3β mutant (hereafter referred to as GSK3βi), are shown in Fig. 4A. The mutant was cloned into pEGFP-NES, and subcellular localization was examined by confocal microscopy. GFP-NES-GSK-3βi localized exclusively in the cytoplasm and was not detectable in the nucleus (Fig. 4B). Both nontargeted and nucleus-targeted GSK-3bi showed equal expression levels in cerebellar granule neurons (Fig. 4C). Expression of nontargeted GFP-GSK-3βi resensitized GSK-3β-depleted neurons to trophic-deprivation-induced death (Fig. 4D), indicating that the neuroprotection evoked by GSK-3β shRNA (Fig. 1E) was due to GSK-3β knockdown, rather than off-target effects. Moreover, addition of a NES targeting sequence upstream of GSK-3βi resulted in a significant increase in apoptosis compared to that of nontargeted GSK-3βi (Fig. 4D). Thus, increased cytoplasmic localization of GSK-3βi leads to increased apoptotic potency. These data reveal that GSK-3βi targeted to the cytoplasm is sufficient to “rescue” the apoptotic response in cerebellar granule neurons and is consistent with a cytoplasmic localization of GSK-3-dependent apoptotic events.

FIG. 4.

Adding back NES-GSK-3β sensitizes neurons to apoptosis in GSK-3β-silenced cells. (A) GSK-3β shRNA-insensitive mutants (GSK-3βi) were prepared by introducing silent mutations within the shRNA-targeted sequence as shown. aa, amino acids. (B) Cerebellar granule neurons were transfected with GFP-GSK-3βi or GFP-NES-GSK-3βi, and subcellular localization was examined by confocal microscopy. Sections through nuclei are shown. (C) To determine the relative expression levels of nontargeted and NES-targeted GFP-GSK-3βi, cerebellar granule neurons were transfected as indicated, and lysates were immunoblotted for GFP. GFP-GSK-3βi and GFP-NES-GSK-3βi expression levels were equal. Nonspecific bands are indicated by an asterisk. (D) Cerebellar granule neurons (5 DIV) were cotransfected (+) with control shRNA or GSK-3β shRNA with (+) or without (−) targeted GFP-GSK-3βi or GFP-NES-GSK-3βi as shown. After 8 days of expression, neurons were deprived of trophic support for 20 h. Cells were stained with Hoechst 33342, and cells with pyknotic nuclei were scored as dead cells. Data are means plus standard errors of the means (error bars) from 5 to 10 replicates. Values that are significantly different from those of the corresponding control values are indicated (*, P < 0.05; ***, P < 0.001).

In the cytoplasm, GSK-3β activity can in theory be negatively regulated by several kinases which phosphorylate Ser9 (Akt, pp90RSK, and protein kinase A), or by Wnt signaling (18, 29). To determine to what extent phosphatidylinositol 3-kinase (PI3K)-Akt signaling regulates GSK-3β Ser-9 phosphorylation, we treated cerebellar granule neurons with the PI3K inhibitor LY294002 to block signaling to Akt. Concentrations of LY294002 that effectively blocked Akt activation resulted in a robust reduction in GSK-3β Ser9 phosphorylation, consistent with Akt being largely responsible for phosphorylation on this site in differentiating cerebellar granule neurons (Fig. 5A).

FIG. 5.

Mutation of GSK-3β serine 9 and GSK-3α serine 21 to alanine does not induce neuronal death. GSK-3α/β^S21A/S9A knock-in neurons are sensitive to trophic deprivation death and protected by GSK-3 inhibitors. (A) To determine the contribution of Akt to GSK-3β serine 9 phosphorylation, cerebellar granule neurons (7 DIV) were treated with the PI3-kinase inhibitor LY294002 (50 μM) for 30 minutes (30′) to 360 minutes (360′) as shown. Lysates were immunoblotted for phosphothreonine 308-Akt (P-Akt) and phosphoserine 9-GSK-3β (P-GSK3β) as shown. Inhibition of Akt substantially reduced GSK-3β serine 9 phosphorylation, indicating that Akt is the dominant kinase for this site. (B) Cerebellar neurons (7 DIV) isolated from wild-type and GSK-3α/β^(S21A/S9A) knock-in mice were either untreated (−) or deprived of trophic support (deprived [+]). Lysates were immunoblotted for phosphoserine 9-GSK-3β. (C) Cells isolated from wild-type and GSK-3α/β^S21A/S9A knock-in mice were treated with LiCl (10 mM) or SB216763 (3 μM) for 30 min prior to trophic deprivation. Cells were fixed after 24 h, and nuclei were stained with Hoechst 33342. Cells with pyknotic nuclei were scored as dead cells. (D) Quantitative data are shown. The number of cells counted is indicated above the histogram bars. (E) To investigate β-catenin signaling in wild-type and GSK-3α/β^S21A/S9A mice, neurons were treated as described above for panel C, except that cells were lysed after 4 h of trophic deprivation. Cytosolic fractions were isolated and immunoblotted for β-catenin. (F) Quantitative data for cytosolic β-catenin levels are shown (for the wild type, means ± standard errors of the means [SEM] [error bars] from four replicates; for the knock-in, means ± SEM from two replicates). (G) The influence of GSK-3 phosphorylation on AP-1- induced neuronal death was assessed by transfecting wild-type or GSK-3α/β^S21A/S9A neurons with cJun-Asp in the presence or absence of Venus-Axin-GID. Basal survival of GSK-3α/β^S21A/S9A cells was moderately reduced following transfection, and these cells were partially sensitized to AP-1-induced death. The number of transfected cells counted is shown in each case. Means ± SEM are shown. Values that are significantly different from the control values are indicated (*, P < 0.05; **, P < 0.01).

It has been proposed that dephosphorylation of GSK-3α/β on Ser21/Ser9 leads to full activation of the kinase (20), and loss of GSK-3β Ser9 phosphorylation is observed subsequent to neuronal death (1, 25, 27). To test the importance of Ser21/Ser9 phosphorylation in neuronal death, we isolated neurons from GSK-3α/β^S21A/S9A knock-in mice (32). In these mice, GSK-3α and -3β cannot be negatively regulated by Ser21/Ser9 phosphorylation (Fig. 5B). To our surprise, cerebellar granule neurons from GSK-3α/β^S21A/S9A mice survived normally and were fully sensitive to trophic-deprivation-induced death (Fig. 5C and D). Moreover, these neurons were protected by treatment with molecularly distinct inhibitors of GSK-3 (lithium or SB216763). These findings indicate that dephosphorylation of GSK-3β Ser9 alone is insufficient to induce death and that another level of GSK-3β regulation is required.

We therefore switched our attention to the Wnt pathway which is known to regulate GSK-3β by an ill-defined mechanism involving interactions with a multiprotein complex, including axin and adenomatosis polyposis coli. When bound to this multiprotein complex, GSK-3 is able to phosphorylate β-catenin, targeting it for ubiquitin-mediated degradation. In the presence of Wnt, GSK-3 is unable to phosphorylate β-catenin, leading to accumulation of β-catenin in the cytoplasm, subsequent nuclear translocation, and regulation of gene transcription (1, 7). Thus, soluble β-catenin levels can provide a direct read-out of Wnt-regulated GSK-3 signaling. We found that both in wild-type and GSK-3α/β^21A/9A neurons, trophic deprivation moderately but significantly reduced levels of β-catenin in the cytoplasm (Fig. 5E and F). This is consistent with activation of the Wnt pool of GSK-3. There was no difference in the resting levels of β-catenin in wild-type and knock-in neurons, while treatment with the GSK-3α/β inhibitor SB216763 (12) prevented the downregulation of β-catenin following trophic deprivation.

We next tested the significance of GSK-3α/β Ser21/Ser9 dephosphorylation on neuronal death induced upon expression of cJun-Asp. Interestingly, following transfection with GFP-CAAX alone, neurons derived from GSK-3α/β^S21A/S9A mice survived less well, 63% surviving compared to 71% in cells from wild-type mice (Fig. 5G). Moreover, GSK-3α/β^S21A/S9A cells were slightly sensitized to cJun-Asp-induced death compared to wild-type. However, even in this artificially induced model of transcription-dependent neuronal death, overexpression of the cytosolic GSK-3 scaffold axin-GID provided significant protection, also implicating the Wnt-regulated GSK-3 pool in this death mechanism.

As the Wnt pool of GSK-3 is regulated by an ill-defined mechanism involving multiprotein complexes, we used size exclusion chromatography to examine whether GSK-3β associated with high-molecular-weight (HMW) complexes under resting conditions and following trophic deprivation stress. Although the majority of GSK-3 was soluble or in low-molecular-weight (LMW) complexes in neurons, a fraction resided in HMW complexes (Fig. 6A and B). Following 2 h of trophic deprivation, there was a decrease in the proportion of GSK-3β in HMW protein complexes. A simultaneous increase in GSK-3β was measured in a LMW pool of approximately 50 to 100 kDa which could conceivably consist of GSK-3 monomers. We speculated that this may reflect a translocation of GSK-3β from a multiprotein complex containing axin and adenomatosis polyposis coli. To test whether trophic deprivation interfered with GSK-3β binding to axin, we generated polyclonal antibodies against axin. These antibodies effectively immunoprecipitated Venus-tagged rat axin (Fig. 6C) and were subsequently used to isolate axin from trophic factor-deprived cerebellar granule neurons (Fig. 6D). Trophic deprivation led to a 35% decrease in the amount of GSK-3β that coprecipitated with axin. This is consistent with the loss of GSK-3β from a HMW complex observed by gel filtration (Fig. 6A and B) and suggests that stress induces a subfraction of GSK-3β to translocate from an axin-bound complex to an unbound, soluble pool.

FIG. 6.

GSK-3β is released from an axin-bound complex following trophic deprivation. (A) To determine whether GSK-3β changed interaction partners in response to stress, cerebellar granule neurons (7 DIV) were deprived of trophic support for 2 h, and protein complexes were separated by gel filtration. Fractions (1.5 ml each) (fractions 1 to 55) were collected after the void volume (20 ml) was eluted. Fractions 9 to 49 (7.5%) were loaded alongside equal proportion (7.5%) of total input (Input) and insoluble pellet (Pellet) and immunoblotted as shown. P-GSK, phosphorylated GSK. (B) GSK-3β intensity was quantified from nonsaturated exposures and expressed as a percentage of the total expression. Following trophic deprivation, there is a reduction in the amount of HMW GSK-3β and an increase in the LMW GSK-3β (∼50 kDa). MM, molecular mass markers. (C) Lysates from HEK-293 cells expressing Venus-axin-GID were used to characterize antibodies generated against axin. Antiaxin antibodies were used to immunoprecipitate Venus-axin (IP: Axin). Whole-cell lysate is shown in the right panel. (D) To determine whether GSK-3β-axin interaction was altered in response to stress, 7 DIV cerebellar granule neurons were deprived of trophic support for 4 h. Axin-bound GSK-3β was isolated by immunoprecipitating endogenous axin. A representative exposure is shown. Quantified data from three replicates (means plus standard errors of the means [error bars]) show that there is a significant decrease in axin-bound GSK-3β following trophic deprivation (*, P < 0.05). α-Axin, antiaxin antibody.

The effectors of GSK-3 mediating neuronal apoptosis remain unknown. However, rapid induction of the BH3 only protein Bim is considered a critical effector of neuronal death resulting from removal of trophic support (2), and we have previously shown that inhibition of GSK-3 interferes with induction of Bim protein expression in stressed neurons (27). Exactly how GSK-3 contributes to Bim induction is not known, but one possibility would be by inducing bim gene expression. To test this, we measured bim mRNA levels in cerebellar granule neurons in response to trophic deprivation. Inhibition of GSK-3β with SB216763 or lithium did not prevent the induction of bim mRNA following trophic deprivation (Fig. 7A and B); however, both of these inhibitors completely blocked the stress-induced Bim protein elevation (Fig. 7C). These data suggest that the locus of GSK-3β action in neuronal death is downstream of bim gene expression where GSK-3β activity governs Bim protein levels. To determine whether GSK-3 inhibitors regulated either the translation or stability of Bim protein, we examined Bim regulation in the presence of 50 μM cycloheximide. We have previously shown that cycloheximide blocks de novo protein synthesis in cerebellar granule neurons with an 50% inhibitory concentration of 0.5 μM and 50 μM blocks over 90% of protein synthesis (14). Neurons were deprived of trophic support to induce Bim levels above baseline. Following 8 h of trophic deprivation, cells were treated with cycloheximide for a further 4 h (Fig. 7D). In the presence of cycloheximide, application of a GSK-3 inhibitor (SB216763) reduced Bim to basal levels. These data suggest that GSK-3 acts downstream of translation and stabilizes Bim protein turnover and is consistent with a cytosol-localized proapoptotic mechanism of GSK-3 action.

FIG. 7.

GSK-3β acts downstream of Bim mRNA induction and translation to stabilize Bim leading to increased proapoptotic protein expression. (A) Cerebellar granule neurons were deprived of trophic support for 8 h in the presence or absence (−) of LiCl (10 mM) to inhibit GSK-3, SB216763 (3 μM) to inhibit GSK-3, or SP600125 (3 μM) to inhibit JNK. bim-EL and actin mRNA levels were assessed by reverse transcription-PCR. Representative gel images are shown. Neither inhibition of GSK-3 or JNK prevents the stress-induced increase in bim-EL expression. (B) Quantitated data (means plus standard errors of the means [SEM] [error bars] for 3 to 12 replicates) are shown. Values that are significantly from the control values are indicated (*, P < 0.05; **, P < 0.01; ***, P < 0.001). (C) To examine the effect of GSK-3 or JNK inhibition on stress-induced Bim protein increase, cerebellar granule neurons (7 DIV) were treated as described above for panel A, except that cells were lysed in Laemmli sample buffer and immunoblotted with a Bim antibody. A representative gel (inset) and quantitative data from 5 to 10 experiments are shown. GSK-3, but not JNK, activity is critical for Bim protein induction in response to stress. (D) To measure the influence of GSK-3 on Bim stability cerebellar granule neurons (7 DIV) were deprived of trophic support (trophic deprivation [TD]) for 8 h after which trophic deprivation was continued for 4 h in the presence (+) or absence (−) of cycloheximide (50 μΜ) to inhibit translation and SB216763 (3 μM) to inhibit GSK-3. Blocking of translation allowed us to monitor the influence of GSK-3 on Bim protein turnover. GSK-3 inhibition significantly reduced Bim levels in the presence of cycloheximide, suggesting that GSK-3 activity blocks Bim degradation following trophic deprivation. Values that were significantly different (P < 0.05) from the control values are indicated by an asterisk.

DISCUSSION

The large number of papers in the literature reporting proapoptotic roles for GSK-3 in a variety of neuronal death models has drawn attention to this molecule as a drug target for certain brain disorders, namely, neurodegenerative diseases, strokes, and mood disorders. Nonetheless, the evidence for GSK-3 involvement relies principally on the use of pharmacological compounds with incompletely defined specificity and often with well-characterized cross-reactivity with cyclin-dependent kinases (33). Genetic deletion of GSK-3 is embryonic lethal around embryonic day 14 (26), before substantial development of neurons has occurred. This precludes their use for analysis of neuronal death mechanisms. Gene silencing therefore provides an obvious alternative to unambiguously establish the importance of GSK-3 in neuronal death. Efficient silencing of GSK-3β expression in cerebellar granule neurons required lengthy exposure to shRNA. This is consistent with the relative stability of GSK-3β protein, which displayed a half-life of 48 h in freshly isolated neurons from the rat cerebellum. The complete protection observed upon selective silencing of GSK-3β, but not GSK-3α, indicates that GSK-3β plays a nonredundant role in regulating neuronal death following trophic deprivation. This is in contrast with the reported GSK-3 isoform dependence for APP cleavage where silencing of GSK-3α, but not GSK-3β, inhibits production of amyloid-β peptides (35). This difference may reflect the cell types studied; APP cleavage was analyzed in CHO cells, whereas our analysis of neuronal death was performed in freshly isolated rat neurons. Furthermore, it is worth noting that GSK-3α and -3β were expressed in both systems studied, indeed GSK-3α expression in neurons was 1.6-fold that of GSK-3β (Fig. 1F). Thus, our data indicate that there is a specific requirement for GSK-3β in neuronal death and that the neuroprotection observed with GSK-3β shRNA is not due to a mere reduction in GSK-3α/β titer.

GSK-3 function can be regulated at many levels. One level that has received attention in the neuronal death context is regulation by intracellular distribution (3, 4). GSK-3β localizes predominantly in the cytoplasm in neurons, although low amounts of GSK-3β with high catalytic activity have been detected in nuclear and mitochondrial fractions isolated from the cortex and hippocampus (4). However, in neuroblastoma cells, GSK-3β translocates to the nucleus following stress (3). It was therefore of interest to understand the spatial dynamics of GSK-3β in primary cultured neurons. Inhibition of NES-dependent nuclear export with leptomycin B (28) revealed that GSK-3β cycled between cytosolic and nuclear compartments in neurons at rest, even though at steady state, the kinase concentrated in punctate structures in the cytoplasm. It is likely that the constant traffic of GSK-3β to and from the nucleus is physiologically significant, especially given that the activity of nuclear GSK-3 is elevated (4). Consistent with this, GSK-3 regulates gene transcription in neurons (30). However, no net accumulation of nuclear GSK-3β was observed following trophic deprivation stress, and inhibition of cytosolic GSK-3 activity provided significantly greater protection than nuclear GSK-3 inhibitors in response to trophic deprivation and AP-1-induced death. Moreover, adding back cytosol-targeted GSK-3β (GFP-NES-GSK-3βi) was sufficient to rescue a death response in neurons where endogenous GSK-3β expression had been silenced. Together, these data suggest that execution of GSK-3β-dependent death may occur in the cytoplasm in neurons deprived of trophic support.

An extensive list of GSK-3 substrates has been reported (18), yet the mechanism of GSK-3 action in neuronal death is not known and is likely multifaceted. Our model which suggests that proapoptotic GSK-3 action is in the cytoplasm invokes a need for cytosolic effectors. One candidate mediator is the proapoptotic Bcl-2 family protein Bax. GSK-3β phosphorylates Bax, and GSK-3 activity is required for the conformational activation of Bax, detected using a conformation-specific Bax antibody in cerebellar granule neurons (31). However, we detected no conformational change in Bax using the same antibody, nor could we detect a translocation of GFP-Bax to mitochondria following trophic deprivation stress as did Linseman and colleagues (not shown). The reasons for these differences are not clear at present. We instead switched our attention to Bim, another proapoptotic Bcl-2 protein. bim mRNA and protein expression are induced upon removal of trophic support, and the mere expression of Bim leads to apoptotic death in neurons and in hematopoietic cells (36, 38). We have previously shown that inhibition of GSK-3 activity with lithium or indirubin prevents induction of Bim protein (27). Here we show that GSK-3 activity is not required for bim mRNA induction but for subsequent elevation in Bim protein levels. We propose that GSK-3 intervenes downstream of translation and stabilizes Bim protein turnover in stressed neurons. Indeed, a precedent for GSK-3 regulation of protein stability exists, and GSK-3 protects estrogen receptor α from proteolytic degradation (24). Our data therefore suggest that Bim may represent another protein the stability of which is regulated by GSK-3β.

The block of Bim induction by structurally distinct GSK-3 inhibitors was in marked contrast to the effect of Jun N-terminal protein kinase (JNK) inhibition. In agreement with Shi et al. (39), we find that treatment of cerebellar granule neurons with concentrations of SP600125 that prevent JNK activation (monitored by loss of c-Jun phosphorylation [8]) has no effect on bim mRNA or protein induction. This contrasts with the JNK requirement for Bim induction in sympathetic neurons deprived of nerve growth factor and cortical neurons treated with arsenite (45, 46). Although JNK is also a critical player in trophic-deprivation-induced neuronal death (11), we find that it is GSK-3β that plays a major role in facilitating proapoptotic Bim signaling in cerebellar granule neurons deprived of trophic support.

There is evidence that Wnt and insulin-regulated GSK-3 pools are insulated from each other (17). In this study we report that the Wnt-regulated pool of GSK-3β is important for neuronal death following removal of trophic support. The evidence for this is as follows: mutation of the Akt-regulated site on GSK-3 from serine to alanine does not potentiate or even sensitize to trophic-deprivation-induced death, yet inhibitors of GSK-3 are neuroprotective. Molecules that specifically inhibit the Wnt pool of GSK-3, FRAT1, and axin-GID protect neurons from death. Consistent with this, we observe a loss of GSK-3β from HMW protein complexes and a corresponding increase in soluble GSK-3β following trophic deprivation. Moreover, there is a decrease in axin-bound GSK-3 in neurons following withdrawal of trophic support. The Wnt pathway has also been implicated in Alzheimer's disease where β-amyloid upregulates the secreted protein Dickkopf-1, leading to inhibition of Wnt signaling and subsequent activation of GSK-3 (9). We would speculate that translocation of GSK-3 from an axin-bound pool to the cytoplasm following stress results in the phosphorylation of cytosolic GSK-3β effectors, which leads to Bim stabilization and apoptosis. What these effectors are is unknown. A classical target of Wnt-regulated GSK-3 is β-catenin. Phosphorylation of β-catenin by GSK-3β leads to proteosomal downregulation. However, β-catenin is not thought to be a critical player in this kind of neuronal death, as exogenous expression of β-catenin in neurons does not protect from trophic-deprivation-induced death (25).

Removal of growth factor support results in apoptosis in a broad range of cellular systems. The Akt pathway receives significant attention in this regard, as Akt phosphorylation of GSK-3 inhibits its activity. Dephosphorylation of this site is therefore taken as a convenient reporter of GSK-3 activity. We show here that loss of Akt site phosphorylation on GSK-3α and -3β does not evoke neuronal death. Instead, Wnt pathway regulation of GSK-3β is pivotal. This finding invites a reevaluation of existing models of neuronal death and urges caution in equating GSK-3 N-terminal phosphorylation loss with GSK-3 activation state and proapoptotic signaling in neurons.