A Positive Feedback Loop between Glycogen Synthase Kinase 3β and Protein Phosphatase 1 after Stimulation of NR2B NMDA Receptors in Forebrain Neurons*

N-methyl-d-aspartate receptors (NMDARs) are critical for neuronal plasticity and survival, whereas their excessive activation produces excitotoxicity and may accelerate neurodegeneration. Here, we report that stimulation of NMDARs in cultured rat hippocampal or cortical neurons and in the adult mouse brain in vivo disinhibited glycogen synthase kinase 3β (GSK3β) by protein phosphatase 1(PP1)-mediated dephosphorylation of GSK3β at the serine 9 residue. NMDA-triggered GSK3β activation was mediated by NMDAR that contained the NR2B subunit. Interestingly, GSK3β inhibition reduced inhibitory phosphorylation of the PP1 inhibitor 2 (I2) and attenuated serine 9 dephosphorylation by PP1. These data suggest existence of a feedback loop between GSK3β and PP1 that results in amplification of PP1 activation by GSK3β. In addition, GSK3β inhibition decreased PP1-mediated dephosphorylation of the cAMP-response element-binding protein (CREB) at the serine 133 residue in NMDA-stimulated neurons. Conversely, overexpression of GSK3β abolished non-NR2B-mediated activation of CRE-driven transcription. These data suggest that cross-talk between GSK3β and PP1 contributes to NR2B NMDAR-induced inhibition of CREB signaling by non-NR2B NMDAR. The excessive activation of NR2B-PP1-GSK3β-PP1 circuitry may contribute to the deficits of CREB-dependent neuronal plasticity in neurodegenerative diseases.

It has been recently proposed that NMDARs may produce diverse outcomes depending on their subunit composition. Whereas the NR2A-containing NMDARs stimulate CRE-driven transcription and enhance neuronal survival, the NR2B NMDARs may inhibit CRE-mediated transcription and trigger excitotoxic cell death (3). Decreases in CRE-mediated transcription may play a role in neurodegeneration. For instance, Huntington disease (HD)-associated mutant forms of huntingtin inhibit transcription of CRE-regulated genes, which are required for proper neuronal function and long term neuronal survival (4 -6).
As the mechanism of NR2B-mediated inhibition of CRE transcription is not fully understood, we investigated a possibility that CREB dephosphorylation following NR2B stimulation is regulated by GSK3␤. We report the NR2B NMDAR-triggered activation of GSK3␤ that was mediated by PP1. We also show that GSK3␤ further amplified PP1 activity and enhanced CREB dephosphorylation by PP1.
Cell Culture and Transfection-Cortical or hippocampal neurons were prepared from newborn Sprague-Dawley rats as described (29). Culture medium was Basal Medium Eagle (BME) supplemented with 10% heat-inactivated bovine calf serum (Hyclone, Logan, Ut), 35 mM glucose, 1 mM L-glutamine, 100 units/ml of penicillin, 0.1 mg/ml streptomycin. Cell-plating densities were 2 ϫ or 0.8 ϫ 10 6 per 35-mm plate for cortical or hippocampal neurons, respectively. Cytosine arabinoside (2.5 M) was added to cultures on the second day after seeding (DIV2) to inhibit the proliferation of non-neuronal cells. Additional glucose (4.5 mM) was added at DIV2 and then at DIV6. Cells were used for experiments at DIV6 -8. Transient transfections were performed on DIV4 using the Lipofectamine 2000 reagent (Invitrogen) (30).
Drug Treatment-Dizocilpine maleate (MK-801), ifenprodil, Ro-25-6981, NBQX, CNQX, FK506, tautomycin, SB216763, and okadaic acid were dissolved in dimethyl sulfoxide (Me 2 SO). The final concentration of Me 2 SO in the medium was 0.2-0.4%. NMDA and glutamate were dissolved in culture medium. For experiments with tautomycin and SB216763, cells were placed in serum-free culture medium (BME supplemented with 35 mM glucose, 1 mM L-glutamine, 100 units/ml of penicillin, 0.1 mg/ml streptomycin, and 2.5 M cytosine arabinoside). Tautomycin or SB216763 were added 1 or 2 h before NMDA, respectively. All other drug inhibitors were added 30 min before the stimulation of NMDAR with NMDA or glutamate.
Injections of Quinolinic Acid-Adult male FVB mice (6 -8 weeks old, weighing 20 g) (n ϭ 12) were deeply anesthetized with freshly prepared 4.0 mg of Avertin (tribromoethanol) (31) per 10 g of body weight dissolved in 0.2 ml of 1.25% (v/v) 2-methyl-2-butanol (Sigma-Aldrich) in saline. These mice were stereotactically injected using a 10-l Hamilton syringe with 1 l of saline containing 0 or 30 nmol of quinolinic acid (cat. P63204, Sigma-Aldrich) at 4 coordinates (all from Bregma): for striatal injections 0.7 mm rostral, 1.9 mm lateral (left and right), 2.5 mm ventral, and for hippocampal injections 2.0 mm caudal, 1.5 mm lateral (left and right), 1.6 mm ventral. The injection of quinolinic acid was performed over 2 min, and the injection needle was retained in position for 2 min before and after injection. Mice were decapitated 1 h after injection and dissected on ice. The striata, hippocampi, and cerebellum were isolated, frozen on dry ice, and kept at Ϫ80°C for future protein extraction.
Western Analysis and Immunostaining-Western blot analysis were performed as described (29,33,34). To detect the phosphorylation shift of I2 by Western blotting, 15% polyacrylamide gels were used. For all other epitopes, proteins were separated on 10% gels.
Promoter Assays-Promoter assays were performed as described (35). Briefly, 0.2 ϫ 10 6 cells were plated onto each well of a 24-well plate coated with poly-D-lysine. At DIV4-DIV5, neurons were co-transfected with the CRE-Luc (0.8 g/4 wells) and EF1␣LacZ DNA (0.55 g/4 wells) with or without GSK3␤ (0.12-0.2 g/4 wells) expression plasmid. Three days after transfection neurons were treated with NMDA for 20 h. The luciferase activity was determined using a luciferase assay kit (Promega) and normalized to ␤-galactosidase activity that was assayed using a kit from Promega.
Statistical Analysis-Statistical analysis of the data was performed using one-way analysis of variance (ANOVA).

NMDAR Stimulation
Activates GSK3␤-To test the possibility that GSK3␤ participates in signaling activated by NMDARs, we evaluated the effects of NMDA on GSK3␤ activity. As others have shown that reduced levels of the inhibitory GSK3␤ phosphorylation at serine 9 correlate with its activation (22,23), we measured the levels of phospho-Ser 9 (pSer 9 ) in NMDA-stimulated neurons. Western blot analysis with an antibody specific for pSer 9 revealed decreased levels of this phosphorylation in cultured neurons that were stimulated with NMDA ( Fig. 1, A  and B). A 60% decrease was evident in both hippocampal and cortical neurons 20 min after addition of 100 M NMDA (Fig. 1, A and B). The reduced pSer 9 levels were observed up to 6 h after initiation of the treatment (Fig. 1, A and B).
Dose response experiments showed that the decrease in pSer 9 was triggered by NMDA concentrations as low as 10 M (Fig. 1C). Also, glutamate, a physiological ligand of NMDAR, reduced pSer 9 in cortical neurons (Fig. 1D). The glutamate effect was mediated by NMDAR as NMDAR, but not non-NMDA receptor antagonists, prevented the pSer 9 decline (Fig. 1D). These data suggest that NMDAR stimulation in cultured cortical or hippocampal neurons reduced the pSer 9 -dependent inhibition of GSK3␤.
As it was reported that GSK3␤ may be activated through the phosphorylation at Tyr 216 residue (36), we evaluated NMDA effects on pTyr 216 levels. Phosphorylation of Tyr 216 was not affected by NMDA in cortical or hippocampal neurons (Fig. 1E). This suggests that NMDAR signaling does not involve modulation of pTyr 216 .
To test whether NMDAR stimulation in the adult brain in vivo can also produce a decrease in GSK3␤ pSer 9 levels, we injected a NMDAR agonist, quinolinic acid (QA) into neostriatum and hippocampus of adult mice. Intrastriatal injections of QA produce a pathology with a striking resemblance to HD (37). Intrahippocampal QA injections result in seizures and loss of pyramidal neurons (38). As a negative control we used cerebellum that is far from the forebrain injection sites. One hour after QA administration, there was a significant decline of pSer 9 levels that reached 39 or 51% of control values in striata or hippocampi, respectively (Fig. 2). The cerebellar pSer 9 levels in QA-treated mice were similar to those in control animals (Fig. 2). These data indicate that NMDAR stimulation in the adult mouse brain rapidly activates GSK3␤ by dephosphorylation of pSer 9 .
To verify that the NMDAR-triggered decrease of pSer 9 was accompanied by an elevation of GSK3␤ activity, we performed a GSK3␤ immunoprecipitation kinase assay. In cultured hippocampal neurons, the kinase activity of GSK3␤ increased as early as 5 min after 100 M NMDA treatment (159% increase, p Ͻ 0.01, Fig. 3A). The maximal increase of 172% was seen at 20 min after addition of 100 M NMDA (p Ͻ 0.001, Fig. 3A). The elevated GSK activity was still present at 60 min (128%, p Ͻ 0.05, Fig. 3A) and declined to control values 3 h after treatment. The reduction of GSK3␤ activity was despite the persistent dephosphorylation of pSer 9 (Fig. 1A). This discrepancy is likely because of decreased levels of the total GSK3␤ protein observed at 1 and 3 h after NMDA stimulation (Fig. 1A). The reduction of GSK3␤ levels may be caused by the degradation of this protein during excitotoxicity, which is induced by 100 M NMDA. 3 To determine if stimulation of NMDAR increased GSK3␤ activity in intact neurons, we evaluated the effects of NMDA on Tau phosphorylation at the Ser 202 residue. This phosphorylation has been shown to be carried out by GSK3␤ (39). In cortical neurons, phospho-Tau levels increased by 42% at 5 min after treatment with 100 M NMDA (p Ͻ 0.01) suggesting GSK3␤ activation. At later time points, phospho-Tau levels decreased (data not shown), which is consistent with reports that PP2B or PP1/PP2A dephosphorylate Tau after NMDAR stimulation (40,41). Therefore, activation of Tau phosphatases may antagonize the effects of GSK3␤ on Ser 202 . Together, these data indicate that activation of NMDAR can increase the kinase activity of GSK3␤.
GSK3␤ Activation Is a Specific Response to Stimulation of NR2B NMDARs-NMDARs are heterotetramers of two NR1 and two NR2 subunits (1,42). There is one gene for NR1 subunit and four genes for NR2 subunits including NR2A, -B, -C, and -D. We found that NR2A, NR2B, and NR2C were expressed in cultured cortical or hippocampal neurons (Fig. 4A), whereas NR2D expression was undetectable (data not shown). This suggests that cultured neurons possess diverse NMDARs that may trigger distinct responses to NMDA. Therefore, we determined which NMDARs may mediate GSK3␤ activation by NMDA in neurons. The NMDA-induced decrease in GSK3␤ pSer 9 levels was abolished by an NR2B-selective NMDAR blockers, ifenprodil or Ro-25-6981 (Fig. 4, B and C). The effect was present in both hippocampal and cortical neurons and was identical to that of APV, which is a non- Cultured rat cortical or hippocampal neurons were isolated from P1 pups and treated as indicated, 6 -8 days after plating (DIV6 -7). The phosphorylation of GSK3␤ at Ser 9 (pGSK3␤) or Tyr 216 (p(Tyr216)GSK3␤) was analyzed by Western blotting and quantified after normalization against the total GSK3␤ content. A and B, NMDA at 100 M decreased pSer 9 in both hippocampal and cortical neurons. Data in the graph represent four independent experiments. Error bars are S.E. The decline in pSer 9 was statistically significant (p Ͻ 0.001, cortical neurons; p Ͻ 0.001, hippocampal neurons, ANOVA). C, Ser 9 dephosphorylation by NMDA was dose-dependent as measured at either 20 min (hippocampal) or 60 (cortical) minutes after treatment. D, glutamate-induced pSer 9 dephosphorylation was mediated by NMDAR. Cortical neurons were exposed for 60 min to glutamate in the presence or absence of NMDAR blocker (MK, 10 M MK801) or non-NMDA glutamate receptor blockers (CN, 10 M CNQX; NB, 10 M NBQX). E, phosphorylation of Tyr 216 of GSK3␤ or Tyr 279 of GSK3␣ was not affected by the NMDA treatment.
The numbers under the blots in C, D, and E are relative pSer 9 or pTyr 216 levels. Results shown in C-E were replicated in three independent experiments. FIGURE 2. The NMDAR agonist quinolinic acid reduces pSer 9 levels in the adult mouse brain. 30 nmol of quinolinic acid (QA) or vehicle (saline, Veh) were injected into the striata and hippocampi of male FVB mice. Animals were sacrificed 60 min after the treatment. The phosphorylation of Ser 9 of GSK3␤ (pGSK3␤) was analyzed by Western blotting. A significant dephosphorylation of pSer 9 was found in the striatum and in the hippocampus (Hip). The pSer 9 levels were unaffected in cerebellum (Cb), which was used as a negative control. Data on the graph represent the mean of 5-6 animals from two independent experiments Ϯ S.E.; **, p Ͻ 0.01; ***, p Ͻ 0.001, ANOVA.
selective NMDAR antagonist (Fig. 4, B and C). This suggests that pSer 9 dephosphorylation in NMDA-treated neurons is mediated by NMDARs that contain the NR2B subunit.
NMDARs act as ion channels, which upon activation increase intracellular Ca 2ϩ concentration ([Ca 2ϩ ] i ) (42). Because the rising [Ca 2ϩ ] i is critical for mobilization of NMDAR-activated signaling pathways (42), we tested whether Ca 2ϩ influx by routes other than NMDAR can reduce pSer 9 levels. Treatment with 55 mM KCl in the presence of a NMDAR antagonist, APV (100 M) increases [Ca 2ϩ ] i through the voltage-gated calcium channels (VGCC) (43). Indeed, hippocampal neurons treated for 60 min with KCl plus AVP showed activation of ERK1/2 signaling, indicating a rise in [Ca 2ϩ ] i (Fig. 4D). However, this treatment did not significantly affect pSer 9 levels (Fig. 4D), suggesting that the NMDAmediated decrease in pSer 9 is a specific response to stimulation of NR2B NMDAR and not a general Ca 2ϩ -mediated response.
PP1 Mediates NMDAR-induced GSK Activation-The NMDA-triggered decrease in pSer 9 levels may be caused either by inhibition of Ser 9 kinases or by stimulation of Ser 9 phosphatases. As a first step to differentiate between these two possibilities, we examined the activity of Akt, the main GSK3␤ Ser 9 kinase in cultured neurons (30). Because in some cell types ERK1/2 signaling may increase GSK3␤ Ser 9 phosphorylation (25, 44 -46), we also studied the status of ERK1/2 after NMDA treatment. The activities of Akt and ERK1/2 were monitored by Western blotting with antibodies specifically recognizing the phosphorylated, activated forms of these kinases. NMDAR stimulation did not affect Akt whereas it activated ERK1/2 at 5, 20, or 60 min after treatment (Fig. 5A). However, pSer 9 levels were significantly reduced at these time points (Fig. 1A). Therefore, inhibition of Ser 9 kinases including Akt or ERK1/ 2-activated p90RSK is not a likely explanation for GSK3␤ activation by NMDA.
NMDARs activate several protein phosphatases including PP1, PP2A, and PP2B (47,48). Therefore, we determined the effects of protein phosphatase inhibitors OA, tautomycin, and FK506 on the NMDAtriggered pSer 9 decline. If applied to cultured cells, 1 M OA inhibits PP2A and PP1, whereas at the lower concentrations it acts selectively on PP2A (49). Indeed, OA at concentrations Ն10 nM significantly increased phosphorylation of a recognized PP2A target, ERK1/2 (Ref. 50 and Fig. 5B). Similarly, OA applied at the concentrations Ն10 nM increased basal pSer 9 levels (Fig. 5, C and D). The increase was concentration-dependent reaching maximal values at 1 M (10.9-or 5.7-fold of control levels in cortical or hippocampal neurons, respectively; Fig. 5, C  and D). These data suggest that under basal conditions, Ser 9 dephosphorylation may be carried out by both PP2A and PP1. On the other hand, the NMDA-induced reduction in pSer 9 levels was abolished by 1 M OA but not by lower OA concentrations (Fig. 5, C and D). In addition, a PP1-specific inhibitor, tautomycin (51) did not affect the basal pSer 9 or pERK1/2 levels (Fig. 5E). In contrast, tautomycin prevented pSer 9 decline in NMDA-treated neurons (Fig. 5E). PP2B inhibition by FK506 did not affect pSer 9 levels in basal conditions or after NMDAR activation (Fig. 5F). These data suggest that a PP1-like phosphatase activity mediates the NMDA-induced dephosphorylation of pSer 9 whereas a PP2A-like phosphatase dephosphorylates pSer 9 under basal conditions.
GSK3␤ Enhances PP1 Activation in NMDAR-stimulated Neurons-GSK3␤ can stimulate PP1 activity by phosphorylation of I2 (11). Therefore, we tested if NMDAR-mediated GSK3␤ activation by PP1 can lead to a further increase of PP1 activity. To address this issue, we inhibited GSK3␤ with LiCl or a specific pharmacological inhibitor, SB216763 (52)(53)(54). In basal conditions and after NMDA treatment, 20 mM LiCl effectively reduced Tau phosphorylation at Ser 202 , indicating GSK3␤ inhibition (Fig. 6A). Similar effects were observed if GSK3␤ was inhibited with 5 M SB216763 (Fig. 6B). Interestingly, LiCl or SB216763 reduced the NMDA-induced decline of pSer 9 suggesting that GSK contributes to the NMDA-triggered increase of PP1 activity that targets GSK3␤ itself (Fig. 6, C and D). GSK3␤ activates PP1 by inhibitory phosphorylation of I2 at the Thr 72 (55). This phosphorylation can be detected by a mobility shift of phospho-I2 visualized by SDS/PAGE and Western blotting (16). Therefore, we used this technique to determine if GSK may phosphorylate I2 in primary neurons. We found that in cortical neurons, inhibition of PP1 with OA produced accumulation of phospho-I2 (Fig. 6E). This indicates that in neurons, I2 phosphorylation is transient because of rapid dephosphorylation and reactivation by PP1. Indeed, we did not detect pI2 in either NMDA-stimulated or unstimulated neurons if OA was not included in the media (data not shown). OA-induced accumulation of pI2 was blocked by LiCl, suggesting that GSK3␤ phosphorylates I2 in intact neurons (Fig. 6E). Together, these data indicate that in NMDA-treated neurons, PP1 activates GSK3␤ which further stimulates PP1. That effect may be mediated at least in part by the transient phosphorylation of I2.
GSK3␤ Contributes to the PP1-mediated Inactivation of CREB-In mature neurons, activation of CREB by NMDAR is mediated through the transient increase in phosphorylation of Ser 133 (3,9). Also in our hands, a 5-min treatment of cortical neurons with 100 M NMDA increased pSer 133 levels 6.3-fold above controls (Fig. 7A). The increase was followed by a decline reaching 0.32-or 0.29-fold of control levels 20 or 60 min after treatment, respectively (Fig. 7A). This suggests that NMDA triggers phosphorylation and then dephosphorylation of CREB Ser 133 . The reduction of pSer 133 levels at 60 min was blocked by OA at 1 M but not at the lower concentrations that affected PP2A (Figs. 7B and  5B). Also a PP1-specific inhibitor, tautomycin abolished pSer 133  NOVEMBER 11, 2005 • VOLUME 280 • NUMBER 45 dephosphorylation after 60 min of NMDA exposure (Fig. 7C). These results indicate a PP1 role in CREB dephosphorylation after NMDA treatment.

GSK3␤ and PP1 in NMDA Receptor Signaling
Because NR2B NMDARs appear to mediate both GSK3␤-dependent amplification of PP1 activity (Fig. 6) and pSer 133 dephosphorylation (3), we tested if GSK3␤ may contribute to CREB dephosphorylation at Ser 133 . First, we determined if GSK3␤ inhibition may affect the increase in pSer 133 that was present at 5 min of NMDAR stimulation (Fig. 7, A and D) and that reflects the increased activity of CREB Ser 133 kinase/ kinases (3,9,56). GSK3␤ inhibition with LiCl or SB216763 did not affect the early increase of pSer 133 levels (Fig. 7D). These suggest that GSK3␤ does not regulate the phosphorylation of CREB Ser 133 . To determine GSK effects on CREB pSer 133 dephosphorylation, neurons were treated for 60 min with 100 M NMDA in the presence or absence of LiCl or SB216763. For instance, NMDA reduced pSer 133 levels to 0.3-fold of controls (Fig. 7E). In NMDA-stimulated neurons with GSK3␤ inhibited by LiCl, pSer 133 decline was reduced to 1.4-fold of controls (Fig. 7E). Similar results were obtained with a GSK3␤ drug inhibitor, SB216763 (Fig. 7F). These data suggest that GSK3␤ enhances PP1-mediated inhibition of CREB.
To test if NR2B-activated GSK3␤ inhibits nonNR2B-mediated stimulation of CRE transcription, we determined the effects of GSK3␤ overexpression on the activity of a CRE-driven luciferase reporter gene. Hippocampal neurons were transfected with this plasmid and 72 h later were stimulated with 20 M NMDA. After 20 h, luciferase activity was determined. We have used NMDA at 20 M because this concentration did not significantly reduce neuronal survival (data not shown). Therefore, changes in luciferase activity in neurons stimulated with 20 M NMDA are because of transcriptional regulation but not cell death. NMDA increased CRE-driven transcription to 350% of control levels (Fig. 8A). The increase was abolished by the non-selective NMDAR antagonists, MK801 or APV, but not by the NR2B-specific blocker, ifenprodil, supporting the involvement of non-NR2B NMDARs in the activation of CRE transcription (Fig. 8A). Noteworthy, neurons co-transfected with wild-type GSK3␤, but not the empty vector, failed to activate CRE transcription in response to NMDA (Fig. 8B). These data support the notion that NR2B-mediated activation of the GSK3␤-PP1 signaling antagonizes non-NR2B-mediated activation of CREB.

DISCUSSION
Our data demonstrate that: (i) stimulation of the NR2B NMDAR causes a PP1-mediated disinhibition of GSK3␤ in cultured primary neurons and in the adult mouse brain, (ii) GSK3␤ provided a stimulatory feedback to PP1 further enhancing the activation of this phosphatase by NMDA, and (iii) GSK3␤ increased PP1-mediated inhibition of CREB in NR2B NMDAR-stimulated neurons. Collectively, these data identify a novel role for GSK3␤ in the regulation of phosphorylation-dependent signaling activated by NMDARs. GSK3␤ appears to serve as an amplifier of NR2B NMDAR-mediated activation of PP1. As a consequence, GSK3␤ antagonizes the NMDA-triggered signaling events that involve increased phosphorylation of PP1 targets including non-NR2B-dependent CRE-driven transcription.
We found that stimulation of the NR2B subtype of glutamate NMDA receptors activates GSK3␤ by PP1-mediated decrease in pSer 9 levels. In addition to glutamatergic neurotransmission, GSK3␤ is also modulated by dopamine and serotonin. For instance, drugs enhancing serotonergic neurotransmission increase pSer 9 levels and inhibit GSK3␤ activity in mouse brain (57). This effect was reported to involve 5HT1A receptors. On the other hand, activation of 5HT2 receptors activated GSK3␤ by decreasing the phosphorylation at Ser 9 . Also, stimulation of dopamin- ergic system in mouse brain was shown to activate the striatal GSK3␤ (58). The mechanism of activation involved D2 receptor-mediated decrease in activity of the Ser 9 kinase, Akt. Thus, GSK3␤ may be regulated in the brain by several neurotransmitters including glutamate, serotonin, and dopamine.
Protein phosphorylation is normally regulated by the opposing activities of phosphorylating kinases and dephosphorylating phosphatases (11). However, the reported cases of decreased pSer 9 in neurons were usually associated with reduced activity of Akt rather than increased activity of pSer 9 phosphatases (30,58,59). Others have reported a possible involvement of PP1/PP2A in the pSer 9 decrease seen in hippocampal neurons exposed to endoplasmic reticulum (ER) stress (60). However, because in that model Akt is also dephosphorylated, GSK activation and/or dephosphorylation could have been caused by both decreased activity of Ser 9 kinase (including Akt) and increased phosphatase-mediated dephosphorylation (60). In our hands, NMDA-mediated pSer 9 dephosphorylation was inhibited by the phosphatase blockers, OA, and tautomycin and was not accompanied by inactivation of Ser 9 kinase pathways, Akt, or ERK1/2. Therefore, our results imply that a selective increase in activity of a Ser 9 phosphatase, PP1 may serve as an important mechanism to control GSK3␤ activity in the nervous system. This is supported by the recent report that GSK3␤ activation by PP1mediated pSer 9 dephosphorylation is involved in the inhibitory action of a cyclin-dependent kinase 5 blocker, olomucin on axonal transport in cortical neurons (61). Also, PP1 was shown to dephosphorylate pSer 9 in several cell lines, including mouse neuroblastoma N2A (26).
In addition to PP1, pSer 9 may be dephosphorylated by PP2A (24,25). However, in the present study, the low OA concentrations that selectively inhibit PP2A failed to affect NMDA-stimulated dephosphorylation of pSer 9 . Conversely, the NMDA effect on pSer9 was blocked by 1 M OA, a concentration that inhibits both PP1 and PP2A (49). Furthermore, 1 M OA, but not the lower concentrations, reduced NMDA-induced dephosphorylation of CREB, indicating activation of PP1 following NMDAR stimulation. Similarly, a PP1-specific inhibitor, tautomycin blocked both NMDA-induced pSer 9 and pSer 133 dephosphorylations without affecting pERK1/2-targeted PP2A activity. Noteworthy, the low OA concentrations increased basal pSer 9 levels in cultured neurons, suggesting that PP2A may FIGURE 5. NMDAR stimulation reduces GSK3␤ pSer 9 levels by PP1-mediated dephosphorylation. A, cortical neurons were treated as indicated. Activity of ERK and Akt was monitored by Western blotting with phosphospecific antibodies (ERK1/2: phospho-Thr 183 and Tyr 185, Akt: phospho-Ser 473 ). NMDA stimulated ERK activity, whereas Akt was unaffected. Therefore, the pSer 9 decline does not appear to be mediated by the decreased activity of the Ser 9 kinase pathways, ERK1/2, and/or Akt. B, cortical neurons were treated with OA as indicated. After 90 min, cells were lysed, and ERK1/2 phosphorylation was determined as described in A. The graph represents the mean of three independent experiments Ϯ S.E.; statistical significance is indicated (***, p Ͻ 0.001; **, p Ͻ 0.05; ns, not significant, ANOVA). The data for the graph were obtained after normalizing pERK1/2 levels against total ERK1/2. OA at Ն10 nM increases basal pERK1/2 levels suggesting PP2A inhibition. C and D, cortical or hippocampal neurons were treated with 100 M NMDA for 60 or 20 min, respectively, in the presence or absence of OA. Note that the NMDA-mediated dephosphorylation of GSK3␤ pSer 9 was inhibited by 1000 nM but not by the lower OA concentrations suggesting PP1 involvement. The increase in basal phosphorylation of Ser 9 in neurons treated with OA at Ն10 nM indicates a possible role of PP2A in regulation of pSer 9 levels in unstimulated cells. The graphs represent the mean of four independent experiments Ϯ S.E.; statistical significance is indicated (***, p Ͻ 0.001; *, p Ͻ 0.05; ns, not significant, ANOVA). In D the numbers under the blot indicate relative pSer 9 levels. E, cortical neurons were treated with 100 M NMDA in the presence or absence of 500 nM tautomycin for 60 min. The NMDA-induced decline of pSer 9 was abolished by tautomycin suggesting PP1 mediated dephosphorylation of pSer 9 . Tautomycin did not affect pERK1/2 levels, indicating that it does not inhibit ERK1/2-targeted PP2A activity. The graph depicts the mean of three independent experiments Ϯ S.E.; statistical significance is indicated (***, p Ͻ 0.001; ns, not significant, ANOVA). F, cortical neurons were treated for 60 min with 100 M NMDA in the presence or absence of the PP2B inhibitor, FK506. Inhibition of PP2B did not affect NMDA-mediated pSer 9 dephosphorylation. participate in pSer 9 dephosphorylation under basal, non-NMDAstimulated conditions.
We demonstrate that in primary neurons, I2 undergoes phosphorylation that becomes evident if PP1 activity is blocked with okadaic acid. The phosphorylation produced a mobility shift that was removed if GSK3␤ was antagonized. Because GSK3␤ is known to directly phosphorylate I2 at Thr 72 , GSK3␤ is a likely I2 kinase in neurons. However, we cannot exclude an indirect regulation of I2 phosphorylation by GSK3␤. For instance, GSK3␤ could stimulate alternative Thr 72 kinases including ERK1/2 or CDK5 (15,16) or it could enhance casein kinase 2 (CK2)mediated phosphorylation of I2 at serines 86, 120, and/or 121 (62).
We did not detect pI2 in either NMDA-stimulated or unstimulated neurons if OA was not included in the medium (Fig. 6E and data not shown). This indicates that in neurons, I2 phosphorylation is transient because of rapid dephosphorylation and reactivation by PP1. As PP1 contributes to NMDAR-mediated activation of GSK3␤, it is expected that OA will perturb GSK3␤ activation by NMDA. Indeed, previous studies showed that OA inhibits GSK3␤ activity in brain slices (63). Consequently, it is difficult to directly test that NMDA-activated GSK3␤ phosphorylates I2. Nevertheless, the presence of I2 in neurons as well as its ability to undergo transient phosphorylation that is regulated by GSK3␤ indicate that I2 is a reasonable candidate mediator of the feedback between GSK3␤ and PP1 in NMDA-stimulated neurons.
We show that PP1-mediated activation of GSK3␤ further amplifies PP1 activity. This may be at least in part via phosphorylation and inhibition of I2. A similar positive feedback loop between PP1 and GSK3␤ was recently demonstrated in several cell lines (26). A transient character of phosphorylation-mediated inhibition of I2 (14,64) suggests that I2 may affect the activity of pre-existing I2/PP1 complexes at specific subcellular sites. In neurons, I2 was found in dendritic spines as well as in the cell bodies (65). As accumulation of I2 in the nucleus occurs only during the S and M phases of the cell cycle (66), postmitotic cells including neurons may have low levels of nuclear I2. Therefore, GSK3␤/I2/ PP1 signaling may modify neuronal PP1 activity locally, in the perikarial cytosol and/or in the dendritic spines.
Our data suggest that in NMDA-stimulated neurons, inhibition of GSK3␤ reduces PP1-mediated dephosphorylation of a nuclear protein CREB at the Ser 133 residue. Therefore, it appears that in addition to increasing PP1 activity in the cytosol and/or dendritic spines, GSK3␤ may also stimulate nuclear PP1. It is unclear what may be the mechanisms underlying GSK3␤ regulation of PP1 in the nucleus. The possibilities include (i) existence of a nuclear GSK3␤/I2/PP1 complex that targets CREB Ser 133 and/or (ii) a potential stimulatory effect of GSK3␤mediated phosphorylation of CREB Ser 129 (67) on CREB Ser 133 dephosphorylation.
An alternative explanation for increased pSer 133 levels in GSK3␤inhibited neurons is that GSK3␤ inhibits CREB phosphorylation by the CREB kinases rather than stimulates PP1-mediated CREB dephosphorylation. The increased CREB kinase activity is clearly noticeable at early time points after NMDAR stimulation when pSer 133 levels are elevated (3,9,56). As the early increase of pSer 133 levels is not enhanced if GSK3␤ is inhibited, CREB phosphorylation at Ser 133 does not seem to be regulated by GSK3␤. In addition, in NMDA-stimulated neurons, GSK inhibition did not affect activities of several kinases implicated in Ser 133 phosphorylation including PKA, Akt, or MSK1 (68,69). 4 Therefore, it seems more likely that GSK3␤ regulates pSer 133 levels by modulating its dephosphorylation rather than phosphorylation. LiCl. Phosphorylated I2 (pI2) was present in OA but not vehicle-treated neurons indicating that pI2 is rapidly dephosphorylated by the basal PP1 activity. The OA-induced accumulation of pI2 was abolished when GSK3␤ was inhibited with LiCl suggesting that in intact neurons, GSK3␤ phosphorylates I2.
In concert with the observations, that GSK3␤ reduces pSer 133 levels we found that overexpression of wild-type GSK3␤ abolished the stimulation of CRE transcription by non-NR2B NMDARs. The non-NR2Bmediated increase of CRE transcription depends on Ser 133 phosphorylation, while its inhibition by NR2B is via dephosphorylation of pSer 133 (3). Therefore, GSK3␤ may be a part of the circuitry that sets the balance between phosphorylation-and dephosphorylation-dependent events in NMDAR signaling.
As Ser 133 phosphorylation may be only one of several mechanisms controlling CRE-mediated transcription, one cannot exclude that GSK3␤ reduces CRE-driven transcription by affecting alternative determinants of CREB activity including such as (i) CREB phosphorylation/ dephosphorylation at other sites than Ser 133 and/or (ii) the activity of CREB-binding protein (CBP) (8,67).
We demonstrate that inhibition of GSK3␤ with 20 mM LiCl reduced NMDA-induced decline of pSer 133 . LiCl produced rapid inhibition of Tau phosphorylation and also blocked accumulation of phospho-I2 in PP1-inhibited neurons. These data suggest that in cultured neurons, lithium effectively inhibited GSK3␤. In NMDAR-stimulated cerebellar granule neurons, PP1 was shown to participate in CREB dephosphorylation, and prolonged but not acute lithium treatment affected this proc-ess (10). The concentration of LiCl used in that study was significantly lower than in our experiments (3 mM versus 20 mM). Therefore, it is possible that longer treatment was needed to reach intracellular concentrations of Li ϩ that will be sufficient to inhibit PP1-GSK-PP1 signaling.
Inhibition of CRE transcription induces neuronal death and impairs learning and memory (5,70,71). On the other hand, stimulation of CRE by neurotrophins protects against trophic deprivation induced apoptosis (72,73). Interestingly, excessive activation of GSK3␤ contributes to neuronal apoptosis following trophic deprivation (32, 74 -77). In addition, GSK inhibition attenuates cell death in a cellular model of HD (21). These data indicate that GSK may affect neuronal survival by inhibition of CRE transcription. However, in our hands, overexpression of GSK3␤ did not affect the basal rate of CRE transcription. Moreover, it did not affect stimulation of CRE by a neurotrophin, BDNF or by KCl suggesting a specific effect on non-NR2B-mediated increase of CRE. 4 These results argue against but do not exclude the possibility that GSK3␤ induces neuronal death by affecting CRE transcription.
In summary, we have identified a NR2B NMDAR-stimulated positive feedback loop that amplifies PP1 activation in neurons. In this circuitry, the PP1-mediated increase of GSK3␤ activity leads to further PP1 acti- FIGURE 7. Inhibition of GSK3␤ reduces the PP1mediated dephosphorylation of CREB at Ser 133 . A, cultured rat cortical neurons were treated as indicated. After NMDA treatment, the phosphorylation of Ser 133 (pCREB) increased at 5 min, followed by a decrease at 20 and 60 min. Data represent the mean of three independent experiments Ϯ S.E.; **, p Ͻ 0.01; ***, p Ͻ 0.001, ANOVA. B and C, cortical neurons were treated with 100 M NMDA for 60 min in the presence or absence of OA (B) or tautomycin (C). The NMDA-mediated dephosphorylation of CREB at the Ser 133 residue was inhibited by 1000 nM but not by the lower OA concentrations. Also, tautomycin reduced pSer 133 dephosphorylation. These data suggest PP1 involvement in pSer 133 dephosphorylation. The graphs depict the mean of three independent experiments Ϯ S.E.; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; ns, non-significant, ANOVA. D-F, cortical neurons were treated with NMDA in the presence or absence of LiCl (20 mM) or SB216763 (5 M, SB). At 5 min (D), the NMDA-mediated phosphorylation of pSer 133 was unaffected by LiCl or SB216763. At 60 min (E and F) the NMDA-mediated dephosphorylation of pSer 133 was reduced when GSK3␤ was inhibited by LiCl or SB216763. In D, numbers under the blot indicate relative levels of pSer 133 . Similar results were obtained in three independent experiments. In E and F, graphs represent the mean of three independent experiments Ϯ S.E.; *, p Ͻ 0.05; **, p Ͻ 0.01, ANOVA.
vation. In addition, in NMDA-treated neurons, GSK3␤ increased PP1mediated dephosphorylation of CREB and antagonized CRE-driven transcription. Therefore, GSK3␤ contributes to the control of the balance between phosphorylation-and dephosphorylation-mediated signaling in NMDAR-stimulated neurons. If GSK3␤ levels are increased by a neurodegenerative process, the up-regulation of GSK3␤-PP1 amplifier may produce excessive activation of PP1 and inhibit NMDAR-dependent neuronal plasticity.