Lithium Blocks the c-Jun Stress Response and Protects Neurons via Its Action on Glycogen Synthase Kinase 3

Abstract

Lithium has been used as an effective mood-stabilizing drug for the treatment of manic episodes and depression for 50 years. More recently, lithium has been found to protect neurons from death induced by a wide array of neurotoxic insults. However, the molecular basis for the prophylactic effects of lithium have remained obscure. A target of lithium, glycogen synthase kinase 3 (GSK-3), is implicated in neuronal death after trophic deprivation. The mechanism whereby GSK-3 exerts its neurotoxic effects is also unknown. Here we show that lithium blocks the canonical c-Jun apoptotic pathway in cerebellar granule neurons deprived of trophic support. This effect is mimicked by the structurally independent inhibitors of GSK-3, FRAT1, and indirubin. Like lithium, these prevent the stress induced c-Jun protein increase and subsequent apoptosis. These events are downstream of c-Jun transactivation, since GSK-3 inhibitors block neuronal death induced by constitutively active c-Jun (Ser/Thr→Asp) and FRAT1 expression inhibits AP1 reporter activity. Consistent with this, AP1-dependent expression of proapoptotic Bim requires GSK-3-like activity. These data suggest that a GSK-3-like kinase acts in tandem with c-Jun N-terminal kinase to coordinate the full execution of the c-Jun stress response and neuronal death in response to trophic deprivation.

The transcription factor c-Jun is a key player mediating transcriptional responses to stress, a function that is conserved among Jun family members from yeast to mammals (46, 53). In the nervous system, transcriptionally active c-Jun is a pivotal trigger of apoptosis after neurotoxic insults such as excitotoxicity occurring during ischemia and epilepsy, in response to axotomy and upon withdrawal of trophic support, a model for developmental neuronal death (4, 11, 23, 55). Candidate targets for transcriptionally active c-Jun in neurons include the proapoptotic genes CD95-ligand and Bim (Bcl-2 interacting mediator of cell death (22, 29, 34, 56), the induction of which trigger caspase activation and apoptosis (45). Transactivation of c-Jun requires phosphorylation of its N-terminal serines 63 and 73 by the c-Jun N-terminal kinase (JNK) family (15, 25, 27). JNK comprises a family of stress-activated protein kinases that are implicated in a wide range of diseases (31). In the nervous system, JNK plays both pro- and antiapoptotic functions during development (30) and is also implicated in neuronal excitoxicity (57, 58).

Glycogen synthase kinase 3 (GSK-3) has emerged as a new regulator of neuronal death (12, 13, 16, 20, 24). GSK-3 is a serine/threonine protein kinase known for its role in glycogen metabolism, Wnt signaling (9) and now for its role in a number of neuropathological disorders (20, 43, 50). In response to insulin and growth factor stimulation, GSK-3 activity is negatively regulated by phosphorylation on serine 9 in the pseudosubstrate domain (18, 47, 49). This is mediated by the survival-promoting kinase Akt (among others [see references 9 and 36). GSK-3 activity toward its substrates can also be regulated by an entirely independent mechanism. The GSK-3-binding protein FRAT1 (for frequently rearranged in advanced T-cell lymphoma type 1), also known as GBP, binds to GSK-3 and prevents it from interacting with the scaffold protein axin (52). This inhibits GSK-3 phosphorylation of select targets; thus, β-catenin phosphorylation by GSK-3 is blocked by FRAT1 expression in vivo, whereas glycogen synthase phosphorylation is not (3, 13). GSK-3 is understood to be constitutively active in resting cells and subject to negative regulation in response to external stimuli. Consistent with this, neuronal GSK-3 is activated upon the withdrawal of trophic stimuli (12, 24), and the expression of dominant-negative GSK-3 or the addition of small molecule GSK-3 inhibitors prevents apoptosis. Lithium has been shown to selectively inhibit GSK-3 at concentrations within the therapeutic range (K_i = 2 mM) (28); however, the mechanism whereby GSK-3 mediates neuronal death and its role in the neuroprotective influence of lithium is unknown.

MATERIALS AND METHODS

Cell culture.

Cerebellar granule neurons were prepared from 7-day-old rats or from wild-type or JNK1^−/− mice as previously described (8). Cells were cultured in minimal essential medium (Life Technologies, Paisley, Scotland) supplemented with 10% (vol/vol) fetal calf serum (Life Technologies), 33 mM glucose, 2 mM glutamine, 50 U of penicillin/ml, 50 μM streptomycin, and 20 mM supplementary KCl (final 25 mM KCl). Cells were plated at 250,000/cm² onto culture surfaces coated with poly-l-lysine (30 μg/ml) in 35-mm dishes or wells of 12- or 24-well plates (Nalgene A/S, Roskilde, Denmark; Falcon, Becton Dickinson, Paramus, N.J.) for kinase assays and immunoblotting, and 10.5-by-10.5-mm coverslips for immunofluorescent staining. Culture medium was replaced after 24 h with the inclusion of 10 μM cytosine arabinofuranoside (Sigma, St. Louis, Mo.) to reduce nonneuronal cell proliferation. For trophic deprivation treatment, cells at 7 days in vitro (div) were changed to serum-free medium containing a low concentration (5 mM) of KCl. Cells were lysed at the indicated times after medium change. 293-HEK cells were cultured in Dulbecco modified Eagle medium (Life Technologies) containing 10% (vol/vol) fetal calf serum, 15 mM HEPES (pH 7.4), 16 mM glucose, and 2 mM glutamine, penicillin, and streptomycin as described above. All cells were cultured in a humidified 5% CO₂ atmosphere at 37°C.

Antibodies.

Monoclonal antibodies to c-Jun (J31920) and GSK-3 (G22320) were from Transduction Laboratories (Lexington, Ky.). Anti-dephospho-tau (Tau-1; clone PC1C6) was from Boehringer Mannheim. Polyclonal anti-Bim was from Stressgen Biotechnologies (Victoria, British Columbia, Canada). Phospho-specific antibodies to JNK and GSK-3 were from New England Biolabs (Beverly, Mass.).

Plasmids.

The preparation of pEBG-JNK1, pEBG-JNK2, pRL-EF1, and ΔMEKK1(801-1493) was as described previously (8). The coding sequence for human FRAT1 was obtained by PCR from SH-SY5Y cell cDNA. pEGFP-C1 and pEGFP-F were from Clontech. All other plasmids were generous gifts from Dirk Bohmann (Rochester, N.Y.), Sirpa Leppä (Helsinki, Finland), John Kyriakis (Massachusetts General Hospital [MGH]), Bruce Mayer (MGH), Michael Birrer (National Cancer Institute), Sander van den Heuvel (MGH), Ami Aronheim (Haifa, Israel), and Tuula Kallunki (Danish Cancer Society).

Kinase assays.

For in vitro assays of JNK activity, pEBG-JNK1 and pEBG-JNK2 were expressed in 293-HEK cells, together with 1% activating ΔMEKK1. Cells were harvested in lysis buffer (20 mM HEPES [pH 7.4]; 2 mM EGTA; 50 mM β-glycerophosphate, 1 mM dithiothreitol [DTT]; 1 mM Na₃VO₄; 1% Triton X-100; 10% glycerol; 1 mM benzamidine; 50 mM NaF; 1 μg of leupeptin, pepstatin, and aprotinin/ml; 100 μg/ml of phenylmethylsulfonyl fluoride), and precleared supernatants were incubated with S-hexylglutathione agarose overnight at 4°C. Kinases were eluted with 50 mM glutathione and dialyzed overnight into 25 mM Tris (pH 7.4)-5 mM EGTA-2 mM DTT-0.1% (wt/vol) Triton X-100-50% (vol/vol) glycerol. pGEX-p54-SAPKβ was expressed in BL21(DE3) cells and purified as previously described (8). Kinase activity measurements were carried out in kinase buffer (20 mM morpholinepropanesulfonic acid [pH 7.2], 2 mM EGTA, 1 mM DTT, 0.1% [vol/vol] Triton X-100) supplemented with 50 μM ATP, 5 μCi of [γ-³²P]ATP (NEN Life Science Products, Albany, Mass.), and 4 μg of glutathione S-transferase (GST)-c-Jun(5-89)/sample. Equivalent activity units of kinases were used. Reactions were carried out for 30 min at 30°C and stopped by addition of 4× Laemmli sample buffer. Samples were resolved on sodium dodecyl sulfate (SDS)-polyacrylamide gels and exposed to film, and autoradiographs quantitated by densitometry.

Neuronal transfection and c-Jun analysis.

For analysis of viability and c-Jun protein expression in transfected cells, cerebellar neurons at 6 div plated on 10.5-by-10.5-mm coverslips were transiently transfected as previously described (7). A ratio 1:1 of marker plasmid (pEGFP-F) to remaining DNA (pEGFP-FRAT1, TAM67-c-Jun, or pCMV) was used. At 20 h after transfection, growth medium was replaced with low-KCl (5 mM), serum-free minimal essential medium (trophic deprivation) for 24 h (survival assay) or 4 h (c-Jun analysis). For cell death assays, cells were stained with 4 μg of Hoechst 33342/ml, fixed, and scored as live or dead based on nuclear morphology and pyknotic nuclei, indicating apoptotic death. To analyze c-Jun levels in transfected neurons, cells were fixed with 4% paraformaldehyde for 20 min followed by permeablilization in phosphate-buffered saline (PBS)-Triton X-100 (1%) for 3 min. After a wash with PBS, cells were blocked with 10% serum-0.2% Tween 20-PBS and stained with 5 μg of c-Jun/ml, followed by the addition of 1:300 Alexa-546-F(ab′)₂ anti-mouse immunoglobulin G. Nuclear c-Jun expression was quantitated from multiple digitized fluorescent images by using a Leica DMRE microscope, a Hamamatsu Orca camera, and imaging software developed by the authors (10). The background fluorescence intensity was quantitated from cells stained without 1° antibody. For each separate experiment, two −1° coverslips were prepared. Averaged fluorescence was obtained from nuclei in six fields from each coverslip and subtracted from averaged values from positive c-Jun immunofluorescence. c-Jun fluorescence intensity values from trophic deprivation treated neurons was expressed as the percentage of control c-Jun levels. Six fields per coverslip were taken, and seven coverslips from separate experiments were used for each datum point.

Reporter assays.

For AP1 reporter assays, 293-HEK cells were transfected with 0.6 μg of 4× AP1-luciferase, 0.05 μg of pRL-EF1 expressing Renilla luciferase as an internal standard against which signals were normalized, and 0.25 μg of pEGFP-FRAT1 and 0.05 μg of HA-Jun-Asp (c-Jun-Asp^{58/62/63/73/89/91/93}). Empty vector pCMV was added to normalize DNA between wells. pEGFP-C1 (0.05 μg) was used as a transfection efficiency indicator. At 40 h after transfection, cells were lysed, and firefly (reporter) and Renilla (internal standard) luciferase activities were assayed as previously described (7, 8).

[³⁵S]methionine metabolic labeling.

Cerebellar granule neurons were washed twice with methionine-free minimal essential medium (Gibco), followed by incubation in methionine-free medium for 15 min in a humidified 5% CO₂ atmosphere at 37°C. Neurons were then labeled for 3 h with 20 μCi [³⁵S]methionine per ml of culture medium in the presence or absence of 10 mM lithium. Lysates were separated by SDS-polyacrylamide gel electrophoresis and [³⁵S]methionine incorporation measured by phosphorimager analysis.

RESULTS

Lithium prevents c-Jun upregulation after trophic deprivation in cerebellar neurons.

Cerebellar granule neurons depend on chronic depolarization for continued survival in culture (10). This dependence is proposed to mimic a trophic requirement of cells in the cerebellar granule layer for innervation by the mossy fibers (2, 51). The removal of trophic support results in upregulation of transcription factor c-Jun expression, and this is considered a requisite, early event in the cascade of signals leading to apoptosis in neurons (4, 8, 55, 56). A screen of neuroprotective compounds of unknown mechanism identified that simultaneous lithium treatment inhibited the stress-induced increase in c-Jun protein in cerebellar neurons (Fig. 1a and b). To determine the effective dose of lithium required for neuroprotection from trophic withdrawal induced death, cerebellar neurons at 7 div were deprived of serum and switched to a low-KCl medium (5 mM final concentration) for the times shown, in the presence or absence of increasing concentrations of lithium. Acute treatment with lithium provided dose-dependent, long-lasting protection from trophic deprivation-induced death, as assessed by chromatin shrinkage (50% inhibitory concentration [IC₅₀] = 2.7 mM; Fig. 1c to e).

FIG. 1.

Lithium prevents the c-Jun stress response in neurons deprived of trophic support and dose dependently protects from apoptosis. Induction of c-Jun protein is a critical early event in the onset of neuronal cell death. (a) Cerebellar granule neurons at 7 div were deprived of trophic support (trophic deprivation) in the presence or absence of simultaneous 10 mM lithium treatment. Samples were lysed at the indicated times and immunoblotted for c-Jun protein. (b) Quantitative data normalized to 4 h of trophic deprivation (n ≥ 7). (c) Cerebellar neurons were deprived of trophic support for 28 h in the presence or absence of increasing concentrations of lithium. Nuclear morphology was revealed by staining with Hoechst 33342, and apoptotic cells were distinguished by the shrunken and bright appearance of their nuclei and scored dead. (d) Lithium is most protective in the 10 to 20 mM range with a calculated IC₅₀ of 2.7 mM, whereas 50 mM lithium is highly toxic. (e) Neuroprotection by lithium is long-lasting, continuing for up to 36 h after stress. Quantitated data (means ± the standard errors of the mean [SEM], n = 5) are shown.

Lithium does not prevent JNK activation in response to stress in neurons.

JNK activation is required for c-Jun protein induction and subsequent death of cerebellar granule neurons deprived of trophic support (8). To determine whether lithium blocked c-Jun induction by interfering with upstream JNK regulators, we examined neuronal JNK activity in the presence or absence of lithium. We know from earlier studies that the stress-responsive JNK pool in neurons is largely masked by a constitutively active pool of JNK1 that does not respond to stress (7, 8) and, consistent with this, no significant changes in total JNK activity were detected on withdrawal of trophic support (Fig. 2a). To avoid this background interfering JNK1 activity, we prepared neurons from JNK1^−/− mice that show reduced basal JNK activity (8). JNK1^−/− neurons at 7 div were deprived of trophic support for 1 h in the presence or absence of 10 mM lithium. These neurons displayed a clear increase in nuclear P-JNK immunoreactivity after trophic deprivation that was not prevented by lithium (Fig. 2b). This indicated that the site of lithium action in blocking the c-Jun stress response is downstream of JNK activation. Lithium ions are considered specific inhibitors of GSK-3β (IC₅₀ = 2 mM) among a range of kinases tested and lithium (10 mM) is reported not to inhibit JNK1 activity (14). Nonetheless, to exclude the possibility that lithium acted directly on other JNK isoforms, recombinant active JNK1, JNK2, and JNK3 were tested for sensitivity to lithium by in vitro kinase assay. Lithium did not inhibit the activity of JNK1, JNK2, or JNK3 under standard kinase assay conditions (Fig. 2c), indicating that the lithium target is downstream of JNK.

FIG. 2.

Lithium treatment does not block stress-induced JNK activity in cerebellar granule neurons. Stress-induced JNK activation is a key step in the c-Jun stress response and subsequent neuronal death. To rule out the possibility that lithium acted upstream of c-Jun, we analyzed the effect of lithium on stress induced JNK activation in neurons. (a) Cerebellar granule neurons deprived of trophic support in the presence or absence of 10 mM lithium (Li) for the times shown were immunoblotted to determine phospho-JNK (P-JNK) immunoreactivity. Lithium did not inhibit total JNK activity. (b) Neurons from JNK1^−/− mice showed lower basal JNK activities and heightened responses to stress (15). JNK1^−/− neurons were deprived of trophic support (i.e., trophic deprivation [TD]) for 1 h in the presence or absence of 10 mM lithium and then immunostained for P-JNK. Immunofluorescence micrographs revealed an increase in nuclear JNK activity after trophic deprivation that was not blocked by lithium. (c) To evaluate whether lithium inhibited JNK activity directly, the activity of recombinant JNK1, JNK2, and JNK3 toward GST-c-Jun was assessed by in vitro kinase assay with or without lithium. The addition of 10 mM lithium did not inhibit JNK activity. Quantitated data (means ± the SEM, n = 3) are shown.

Inositol monophosphatases do not mediate the effects of lithium on the c-Jun stress response.

We next considered whether the known lithium targets, inositol monophosphatases, could mediate the action of lithium on c-Jun regulation. Inhibition of inositol monophosphatases by lithium (K_i = 1 mM) results in the depletion of cytoplasmic free inositol pools (21, 40). To investigate whether perturbation of inositol signaling mediated the effects of lithium on c-Jun regulation, neurons were deprived of trophic support combined with myoinositol (20 mM) to maintain cytosolic inositol levels. Inositol addition did not reverse the effects of lithium on c-Jun protein levels (Fig. 3a and b) or survival (Fig. 3c). We conclude that inositol monophosphatases are not the targets of lithium in the c-Jun stress response.

FIG. 3.

Lithium-mediated neuroprotection and c-Jun regulation are not mediated by inositol monophosphatase inhibition. Lithium inhibits inositol monophosphatases with a K_i of ∼1 mM, leading to a depletion of cellular inositol (25). We examined whether replenishing inositol pools by exogenous addition of myoinositol would reverse the effects of lithium on c-Jun induction. (a) At 7 div, granule neurons were deprived of trophic support for 2 h in the presence or absence of 10 mM lithium and 20 mM myoinositol as indicated. Cells were lysed and immunoblotted for c-Jun. (b) Quantitative data for c-Jun expression levels (means ± the SEM, n = 3) are shown. Refilling of inositol pools did not reverse the inhibitory effect of lithium on the c-Jun stress response. (c) To test whether the neuroprotective effects of lithium depended on downregulation of cellular inositol pools, granule neurons at 7 div were deprived of trophic support for 24 h in the presence of 10 mM lithium and 20 mM exogenously added myoinositol. Cell survival was assessed by analysis of nuclear morphology after Hoechst 33342 staining as described for Fig. 1. Quantitative data (means ± the SEM, n = 3) are shown. Refilling of inositol pools did not reverse the effects of lithium on neuronal survival.

Stress-activated GSK-3 is dose dependently inhibited by lithium.

Finally, we tested GSK-3 as a candidate mediator of lithium regulation of c-Jun. If GSK-3 activation plays a critical role in regulating the neuronal c-Jun stress response, it should be activated prior to c-Jun protein increase. Immunoblotting with antibodies directed against phospho-Ser-9, a negative regulatory site on GSK-3, reveals the level of inactive GSK-3 in neuronal lysates (13). Using these antibodies we determined that GSK-3 Ser-9 phosphorylation is significantly decreased after 15 min of trophic deprivation (P < 0.02), indicating a rapid activation of the kinase that precedes the upregulation of c-Jun protein at 2 h (Fig. 4a). The microtubule-binding protein tau is an in vivo substrate for GSK-3, the phosphorylation sites of which have been well characterized (20). To evaluate whether the concentration of lithium that protected from death (Fig. 1c) was sufficient to reverse stress-induced GSK-3 activity in neurons, we utilized antibodies that recognize dephosphorylated serines 189 to 206 of tau, sites specifically phosphorylated by GSK-3 (13). Trophic deprivation stress induced a decrease in dephospho-tau levels, indicating an activation of GSK-3. This was reversed to control levels by the addition of 5 mM lithium but not by the addition of 1 mM lithium, a finding correlating well with the neuroprotective concentration (Fig. 4b and 1c).

FIG. 4.

Mechanistically independent GSK-3 inhibitors lithium and indirubin dose dependently block GSK-3 activation after stress. (a) To analyze whether the kinetics of GSK-3 activation conformed with the c-Jun stress response, cerebellar granule neurons at 7 div were deprived of trophic support for the times shown in the presence or absence of 10 mM lithium. Cells were lysed, samples were separated by SDS-polyacrylamide gel electrophoresis, and immunoreactivity to GSK-3 phospho-Ser-9 was detected. After trophic deprivation there is a rapid (within 15 min) decrease in GSK-3 phosphorylation indicative of increased kinase activity. GSK-3 expression did not change (lower panel of inset). Quantitated data (means ± the SEM, n = 6) were normalized to control values. (✽, P < 0.02; ✽✽, P < 0.001 [as determined by paired Student t test). (b) To evaluate the lithium dose required to reverse stress-induced GSK-3 activation, neurons were deprived of trophic support for 2 h in the presence of increasing concentrations of lithium. GSK-3 activity was detected by immunoblotting with an antibody specific for GSK-3 targeted Ser-189-206 of tau (DephosphoTau-1). (c) Cerebellar neurons were deprived of trophic support for 2 h in the presence or absence of increasing concentrations of a mechanistically distinct inhibitor of GSK-3; 5-iodoindirubin-3′-monoxime (indirubin). To evaluate endogenous GSK-3 activity, lysates were blotted with dephospho-Tau-1. Stress induces GSK-3-dependent tau phosphorylation that is blocked dose dependently by indirubin. (d) Recombinant JNK1, JNK2, and JNK3 were used to phosphorylate GST-c-Jun(5-89) in vitro in the absence or presence of 0.1, 1.0, or 10 μM indirubin. A concentration of 1 μM indirubin did not inhibit JNK activity, although it strongly reduced induction of c-Jun protein (see panel f). (e) Indirubin does not block endogenous JNK activity in neurons. Cerebellar neurons (at 7 div) were deprived of trophic support for 2 h in the presence or absence of increasing concentrations of the GSK-3 inhibitor indirubin. Inhibitor-treated cells were preincubated for 30 min with indirubin before being changed to a low-KCl medium. Cells were lysed and immunoblotted for active JNK (P-JNK) and JNK. (f) Cerebellar neurons treated as in panel c were lysed and immunoblotted for c-Jun. The indirubin dose dependently blocked c-Jun protein induction. (g) Quantitative data for c-Jun expression (means ± the SEM, n = 4) and de-phospho tau (means ± the standard deviation, n = 2) are shown.

Doses of 5-iodoindirubin-3′-monoxime that block stress-induced GSK-3 activity prevent the c-Jun stress response.

The possibility that GSK-3 activity may be required for the c-Jun stress response in neurons was unexpected since there was no precedence for GSK-3 facilitation of c-Jun stress responses in the literature. Although we had excluded the other well-characterized lithium target, inositol monophosphatase (21) from this mechanism (Fig. 3), lithium pharmacology remains complex. To test whether GSK-3 was indeed the target for lithium in this mechanism, we evaluated the effects of structurally independent GSK-3 inhibitors on c-Jun induction and neuronal survival upon trophic deprivation. Treatment with 5-iodoindirubin-3′-monoxime (indirubin), a potent inhibitor of GSK-3β (33), dose dependently inhibited cerebellar granule neuron GSK-3 activity as visualized by measuring dephospho tau (Ser-189-206) immunoreactivity (Fig. 4c). Treatment with 1 μM indirubin reversed GSK-3 dependent tau phosphorylation after trophic deprivation, indicating that this concentration was sufficient to inhibit GSK-3 activation in response to stress (Fig. 4c). Since the specificity of this inhibitor had not been tested against JNK, we evaluated the effect of indirubin on the activity of JNK by in vitro kinase assay (Fig. 4d). Indirubin (1 μM) did not inhibit the activity of JNKs (JNK1, JNK2, or JNK3) in vitro, nor did it inhibit upstream JNK regulation in vivo (Fig. 4e). Furthermore, this concentration of indirubin significantly attenuated the induction of c-Jun protein after 2 h of trophic deprivation (Fig. 4f and g). This supports the proposal that GSK-3 activity is required for c-Jun protein induction in neurons.

A physiological inhibitor of GSK-3, FRAT1, inhibits c-Jun upregulation and neuronal death after trophic deprivation.

We then employed a physiological protein inhibitor of GSK-3 action, FRAT1, to further scrutinize the specificity of GSK-3 in the c-Jun stress response. FRAT1 binds to and prevents GSK-3 activity toward a subgroup of targets (3, 13, 52). Transfected cerebellar neurons responded to trophic deprivation with an average 2.2-fold increase in endogenous c-Jun immunoreactivity. This response was antagonized by coexpression of FRAT1 or by treatment with lithium or indirubin, indicating that GSK-3 contributes to stress-induced c-Jun expression in neurons (Fig. 5a and b). Furthermore, expression of FRAT1 or a dominant-negative c-Jun transactivation domain mutant (TAM67-Jun) protected cerebellar granule neurons from trophic deprivation-induced apoptosis (Fig. 5c). These data suggest that GSK-3 activity may be required to permit the full execution of the c-Jun stress response and subsequent neuronal death.

FIG. 5.

Overexpression of the GSK-3-binding protein FRAT1 represses c-Jun induction and protects neurons from stress induced apoptosis. (a) Cerebellar neurons transfected with farnesylated GFP (GFP-F; green), together with pCMV empty vector or GFP-FRAT1, were deprived of trophic support for 4 h in the presence or absence of lithium (10 mM) or indirubin (1 μM) as shown. Cells were fixed and immunostained for endogenous c-Jun (red in upper panels and white in lower panels; transfected cells are indicated by arrows). (b) Digitized fluorescence images from multiple fields were collected, and the nuclear c-Jun immunoreactivity was quantitated and expressed as a percentage of the control levels (see Materials and Methods). Mean data ± the SEM (n = 7) are shown. (c) Neurons were transfected with either pCMV (control), FRAT1, or a transactivating-domain mutant of c-Jun (TAM67-Jun) for 19 h, after which neurons were changed to low-KCl (5 mM) medium for a further 24 h. Neuronal survival was assessed as described for Fig. 1. Transfected neurons from five coverslips for each condition were counted. Expression of either FRAT1 or TAM67-Jun protected neurons from stress-induced death. Mean data ± the SEM (n = 5) are shown. Significance levels as assessed by using the Student t test are indicated (✽, P < 0.05; ✽✽✽, P < 0.002).

FRAT1 expression inhibits AP1 promoter reporter activity and c-Jun-Asp-induced death in cerebellar neurons.

To determine where GSK-3 intervenes in the sequence of events leading to increased c-Jun expression, we examined responses directly downstream of JNK activity and c-Jun phosphorylation. The ability of transactivated c-Jun to bind AP1 promoter elements and induce the expression of AP1 responsive death genes is well documented (15, 25, 27, 34, 56). We therefore analyzed the effect of FRAT1 expression on AP1 promoter activity induced by constitutively active c-Jun (hemagglutinin tagged [HA-Jun-Asp]) in 293-HEK cells (Fig. 6a). Under these artificial conditions, coexpression of FRAT1 significantly inhibited HA-Jun-Asp-induced AP1-driven luciferase activity without affecting cytomegalovirus (CMV)-driven HA-Jun-Asp expression. Taken together with the observation that lithium does not regulate JNK signaling, these data suggest that GSK-3 acts downstream of c-Jun transactivation to facilitate AP1 promoter activation. To test this hypothesis in neurons, we investigated whether GSK-3 inhibitors FRAT1 or lithium protected from HA-Jun-Asp-induced death (Fig. 6b). Cerebellar neurons were transfected with HA-Jun-Asp and allowed to express for 39 h, after which cell death was measured as the percentage of transfected cells displaying pyknosis, typical of apoptotic cells. This extended expression time resulted in lower basal survival among transfected neurons than in earlier experiments (Fig. 5c), nonetheless, HA-Jun-Asp potently induced apoptosis and coexpression of FRAT1 and treatment with lithium or indirubin prevented the death (Fig. 6b and c). Importantly, FRAT1 expression did not prevent HA-Jun-Asp expression (94% ± 4% cells coexpressing FRAT1 also expressed HA-Jun-Asp). These data suggested that GSK-3 activity may be required for c-Jun-dependent induction of AP1-responsive genes.

FIG. 6.

GSK-3 activity is required for c-Jun-induced death and downstream Bim expression in stressed neurons. To identify the point in the c-Jun stress response at which GSK-3 activity is required, we utilized an artificial reporter system to monitor the effect of FRAT1 on AP1 promoter activity. (a) 293-HEK cells were transfected with or without constitutively active HA-Jun-Asp (c-Jun-Asp^{58/62/63/73/89/91/93}), together with an AP1-responsive firefly luciferase reporter and a Renilla luciferase internal control. Jun-Asp potently induces AP1 reporter activity and this is significantly reduced by coexpression of FRAT1. Data are normalized to a Renilla luciferase internal standard. Quantitated data ± the SEM are shown. The statistical significance, determined by using the Student t test (n = 6), was P < 0.001 (✽✽✽). Identical lysates were immunoblotted for GFP or HA to detect GFP-FRAT1 and HA-Jun-Asp expression. (b) If GSK-3 activity is required for AP1 gene induction, GSK-3 inhibitors should block AP1-induced neuronal death. To test this, cerebellar neurons were transfected with pCMV or HA-Jun-Asp in the presence of GFP-FRAT1, lithium (10 mM), or indirubin (1 μM). Nuclei were counterstained with Hoechst 33342. Fluorescence micrographs of corresponding fields showing a GFP-F transfection marker (left panel), HA immunoreactivity (middle panel), and Hoechst-stained nuclei (right panel) are shown. HA-Jun-Asp-expressing cells are pyknotic, whereas coexpression of FRAT1 and treatment with lithium or indirubin is protective (see arrows). (c) Quantitated data (means ± the SEM, n = 6) where survival is indicated by the proportion of nonpyknotic nuclei. The total numbers of cells counted for each condition are shown above the respective bars. (d) Inhibition of GSK-3 blocks expression of the c-Jun-induced death protein Bim. Cerebellar neurons were deprived of trophic support for 8 h in the presence or absence of lithium (10 mM) or indirubin (3 μM). Lysates were immunoblotted for Bim expression (see inset). Quantitative data from 2 separate experiments (means ± the standard deviations) are shown. (e) Lithium does not inhibit overall de novo protein synthesis. To determine whether lithium had a general effect on protein metabolism, cerebellar neurons were labeled with [³⁵S]methionine for 3 h in the presence or absence of lithium. [³⁵S]methionine incorporation was quantified and expressed as a percentage of the control. Quantitated data (means ± the SEM) are shown.

GSK-3 inhibitors lithium and indirubin prevent stress-induced Bim induction in cerebellar neurons.

Among the stress-responsive AP1 target genes in neurons is the proapoptotic Bcl-2 family member bim (22, 56). Induction of Bim expression after trophic factor withdrawal is critical for an intrinsic pathway of neuronal apoptosis (43). If GSK-3 activity is relevant to c-Jun function during neuronal stress, inhibition of GSK-3 should be detectable as a negative regulation of known AP1 targets. We therefore evaluated the influence of GSK-3 inhibitors lithium and indirubin on the induction of Bim protein after trophic deprivation. Inhibition of GSK-3 by treatment with either lithium or indirubin specifically blocked the upregulation of the 25-kDa Bim protein (Fig. 6d). Notably, lithium did not interfere with overall de novo protein synthesis in neurons (Fig. 6e) or with HA-Jun-Asp expression (Fig. 6b), implying that GSK-3 does not act to stabilize c-Jun protein. Instead GSK-3 may be necessary for the induction of AP1-responsive targets such as c-Jun and Bim (Fig. 7).

FIG. 7.

Model depicting the cooperative action of JNK and GSK-3 in regulating the c-Jun stress response in neurons. After the withdrawal of trophic support, the activities of JNK and GSK-3 are rapidly induced. Stress-activated JNK phosphorylates c-Jun leading to increased transcriptional activity and induction of AP1-responsive proapoptotic proteins c-Jun and Bim. Coordinate activation of GSK-3 is required for increased c-Jun/Bim expression and subsequently for AP1-induced death. This suggests that GSK-3 acts as a checkpoint on this apoptotic pathway. Active AP1 alone is not sufficient to drive the full death program. In addition, GSK-3 activity is essential for the full execution of c-Jun-mediated death.

DISCUSSION

Lithium is a widely used drug for treatment of mood disorders, and yet the mechanism of action of lithium in neuronal disorders is unclear (35). We have shown that acute treatment of cultured neurons with lithium prevents stress-induced c-Jun protein increase and downstream apoptotic events. These effects are replicated by using mechanistically independent inhibitors, suggesting that GSK-3 facilitates the c-Jun stress response in neurons. Cooperation of GSK-3 with JNK to facilitate c-Jun stress signaling is unexpected. The possibility that these kinases converge on the same pathway has not been previously reported; however, a number of disease models have been described in which both kinases appear to play a role. For example both JNK and GSK-3 are implicated in Alzheimer's disease. Cortical neurons lacking JNK3 or cells expressing dominant-negative c-Jun are protected from amyloid-β peptide (a key player in Alzheimer's pathology)-induced death (39), and elevated JNK activity and c-Jun protein levels were detected in Alzheimer's disease brains (37, 48, 61). Similarly, GSK-3 activity is elevated in the brain in Alzheimer's patients, and in cultured hippocampal neurons GSK-3 activity is induced by amyloid-β peptide (43, 50). It would therefore be of interest to examine the frequency of such tandem regulation of stress responses by JNK and GSK-3 and investigate the effects of GSK-3 inhibition on c-Jun stress responses and survival in models of neurodegenerative disease.

In the present study the effects of acute lithium treatment on neuronal stress signaling are described. In contrast, neuroprotective effects requiring chronic lithium pretreatment have also been reported. For example, long-term but not acute lithium treatment protects cerebellar granule neurons from N-methyl-d-aspartate-induced excitotoxicity (42). However, glutamate toxicity in cerebellar neurons is independent of transcriptional regulation, and therefore c-Jun does not play an immediate role (1). Consequently, there is no conflict between this previously published data and the model presented here. Moreover, the long-term ameliorative effects of lithium have been attributed to Bcl-2, p53, and Bax regulation, events that may also be dependent on c-Jun (6). Trophic deprivation of cerebellar neurons activates an intrinsic apoptotic pathway involving a JNK/c-Jun-dependent induction of the BH3-only proapoptotic protein Bim and downstream caspase-3 activation (44). Previous findings have shown that lithium blocks caspase-3 activation in this model (38). Consistent with this, we demonstrate that independent GSK-3 inhibitors prevent Bim increase after stress. This places lithium action upstream of caspase activation and Bim protein induction and suggests that GSK-3 activity is absolutely required for this event. Furthermore, lithium, indirubin, and FRAT1 protect from c-Jun-Asp-induced death in neurons. Taken together, these findings imply that the likely common target of these inhibitors, GSK-3 facilitates events downstream of the initial c-Jun transactivation (Fig. 7).

GSK-3 is involved in diverse cellular processes, including glycogen synthesis, proliferation, apoptosis, and development (9). Specificity of GSK-3 signaling is regulated by the scaffold protein axin that directs GSK-3 toward Wnt pathway proteins and may separate this pool of GSK-3 from the insulin-regulated pool. The existence of physiological protein inhibitors of GSK-3 such as FRAT1 confers a further level of regulation to this kinase. FRAT1 selectively inhibits GSK-3 phosphorylation of specific targets (3, 13, 51). FRAT1 is a proto-oncogene first characterized as a promoter of lymphogenesis (32) and subsequently shown to play a role in development (59). FRAT1 is also expressed in the brain (19), where nothing is known about its physiological function. We show that ectopically expressed FRAT1 can block AP1 gene activation, as well as Jun-Asp-induced death, in cerebellar neurons. The protective influence of FRAT1 on neuronal survival, described here and by others (13), suggests that FRAT1 deserves attention as a putative survival protein in neurons.

The effector through which GSK-3 positively regulates the c-Jun stress response is currently unknown and inspection of the accepted GSK-3 targets fails to reveal probable candidates. For example, GSK-3 phosphorylation of eIF2B (eukaryotic initiation factor 2B), a positive regulator of translation, inhibits its activity and is inconsistent with GSK-3-mediated c-Jun upregulation (54). Known transcriptional regulator targets of GSK-3 include CREB (cyclic AMP response element-binding protein) and β-catenin (17, 60). CREB is phosphorylated on Ser-129 by GSK-3 after a priming phosphorylation on Ser-133 by cyclic AMP-activated protein kinase A. Secondary phosphorylation of Ser-129 by GSK-3 increases CREB transcriptional activity. However, after trophic factor deprivation, cerebellar neuron CREB Ser-133 phosphorylation falls to undetectable levels (V. Hongisto and E. T. Coffey, unpublished observations), suggesting that under conditions in which GSK-3 is activated, CREB is in an unprimed state and will not be targeted by GSK-3. β-Catenin belongs to the subgroup of FRAT1-sensitive GSK-3 substrates (52). Active GSK-3 phosphorylates β-catenin and targets it for proteolytic degradation. Consistent with this, we observe a downregulation of soluble β-catenin in cerebellar neurons deprived of trophic support (data not shown). Reduced levels of β-catenin are associated with sensitization of hippocampal neurons in culture to β-amyloid-induced death and decreased β-catenin levels are reported in the brains of subjects with Alzheimer's disease (60). However, overexpression of β-catenin does not protect cortical neurons from trophic factor deprivation, suggesting that an alternative target for GSK-3 mediates this form of death (24). A functionally homologous mechanism to that described here for GSK-3 action during stress has been described in Saccharomyces cerevisiae. Msn2p is a stress responsive transcription factor that requires GSK-3 for binding to stress response elements and implementation of a proper stress response (26). Since Msn2p is not phosphorylated by GSK-3, an indirect mechanism is proposed. GSK-3 regulation of the c-Jun stress response in neurons may also be indirect since direct phosphorylation of c-Jun by GSK-3 decreases c-Jun DNA binding (5, 41). In summary, we find no obvious candidates from the list of known GSK-3 targets that may mediate the actions described here. It may be that known targets are not involved, and the identification of new binding partners and substrates may be required to explain the molecular mechanism in full.

Lithium is not an ideal drug, being both teratogenic and toxic at doses close to those required for therapeutic benefit. This small molecule profoundly effects morphogenesis and cell fate determination in developing organisms, effects attributed to direct inhibition of GSK-3. Understanding the mechanism of action of lithium, which is already approved as a drug, may benefit development of neuroprotectants. In the present study we have shown that lithium inhibits the canonical JNK/c-Jun stress pathway in neurons. The data presented here support our proposal that the target of lithium in this response is GSK-3 or a GSK-3-like kinase. Moreover, we report that GSK-3 acts downstream of JNK activation to facilitate AP1-induced Bim expression and subsequent apoptosis (Fig. 7). These data reveal a novel mechanism both for neuroprotective lithium and proapoptotic GSK-3 action during neuronal death. This may have important implications for neuronal disease mechanisms and possibly for development of future neuroprotective agents.