Alleviating the Suppression of Glycogen Synthase Kinase-3β by Akt Leads to the Phosphorylation of cAMP-response Element-binding Protein and Its Transactivation in Intact Cell Nuclei*

Glycogen synthase kinase-3β (GSK-3β) activity is suppressed when it becomes phosphorylated on serine 9 by protein kinase B (Akt). To determine how GSK-3β activity opposes Akt function we used various methods to alleviate GSK-3β suppression in prostate carcinoma cells. In some experiments, LY294002, a specific inhibitor of phosphatidylinositol 3-kinase (a kinase involved in activating Akt) and tumor necrosis factor-α (TNF-α) were used to activate GSK-3β. In other experiments mutant forms of GSK-3β, GSK-3βΔ9 (a constitutively active deletion mutant of GSK-3β) and GSK-3βY216F (an inactive point mutant of GSK-3β) were used to alter GSK-3β activity. LY294002, TNF-α, and overexpression of wild-type GSK-3β or of GSK-3βΔ9, but not GSK-3βY216F, alleviated the suppression of GSK-3β activity in prostate carcinoma cells and enhanced the turnover of β-catenin. Forced expression of wild-type GSK-3β or of GSK-3βΔ9, but not GSK-3βY216F, suppressed cell growth and showed that the phosphorylation status of GSK-3β can affect its intracellular distribution. When transcription factors activator protein-1 and cyclic AMP-response element (CRE)-binding protein were analyzed as targets of GSK-3β activity, overexpression of wild-type GSK-3β suppressed AP1-mediated transcription and activated CRE-mediated transcription. Overexpression of GSK-3βΔ9 caused an (80-fold) increase in CRE-mediated transcription, which was further amplified (up to 130-fold) by combining GSK-3βΔ9 overexpression with the suppression of Jun activity. This study also demonstrated for the first time that expression of constitutively active GSK-3βΔ9 results in the phosphorylation of CRE-binding protein on serine 129 and enhancement of CRE-mediated transcription in intact cell nuclei.

Several transcription factors are directly phosphorylated by GSK-3␤ (3, 4, 9 -11). The majority of these transcription factors are inhibited after being phosphorylated by GSK-3␤; for example, GSK-3␤ blocks Jun-DNA binding by phosphorylating the Jun DNA binding domain. Only two transcription factors, cyclic AMP response element (CRE) binding (CREB) protein and possibly microphthalmia-associated transcription factor are stimulated after phosphorylation by GSK-3␤ (3). In vitro studies have shown that CREB proteins and peptides are primed by phosphorylation at Ser 133 (12) followed by progressive phosphorylation by GSK-3␤ at Ser 129 , which fully activates CRE transcription (13,14). No studies to date have demonstrated that the phosphorylation of CREB at Ser 129 occurs in vivo or in intact cell nuclei (3).
We conducted the present study to 1) determine the effects of alleviating GSK-3␤ suppression by Akt and assess GSK-3␤ function in prostate cancer cells, 2) examine the phosphorylation of CREB on Ser 129 by GSK-3␤ in intact cell nuclei and its effects on CRE-mediated transcription, and 3) examine the intracellular distribution of GSK-3␤ mutants that were altered at critical phosphorylation sites.

EXPERIMENTAL PROCEDURES
Cells and Culture Conditions-PC-3 human prostate adenocarcinoma cells were obtained from the American Type Culture Collection (Manassas, VA). These cells were maintained in a 1:1 (vol:vol) mixture of Dulbecco's minimal essential medium and F-12 medium supplemented with 10% fetal bovine serum.
Preparation of Anti-CKRREILS 129 RRPS 133 YR-CREB Antibody-Rabbit polyclonal antibodies were raised against a diphosphorylated CREB peptide CKRREILS 129 RRPS 133 YR. The diphospho peptide CKRREILS 129 RRPS 133 YR was conjugated with a keyhole limpet hemocyanin carrier by generating a sulfide linkage using maleimide chemistry, and a diphospho peptide antibody was purified on a CKRREILS 129 RRPS 133 YR affinity column. A second affinity purification step was carried out using a monophospho peptide CKRREILSRRPS 133 YR affinity column to eliminate antibodies that interacted with the monophosphorylated peptide. The resulting passthrough fraction produced an anti-diphospho peptide antibody that was high titer (1:173,000) based on a peptide linked-enzyme immunosorbent assay that failed to bind CKRREILSRRPS 133 YR but could specifically recognize both CKRREILS 129 RRPSYR and CKRREILS 129 RRPS 133 YR.
GSK-3␤ Assays-PC-3 cells were incubated overnight in serum-free Dulbecco's minimal essential medium/F-12 medium and then incubated with or without 10 ng/ml TNF-␣ for 20 min. The cells were harvested, and GSK-3␤ was immunoprecipitated from detergent lysates using an anti-GSK-3␤ antibody (Transduction Laboratories, Los Angeles, CA). GSK-3␤ kinase assays were then carried out as described by Pap and Cooper (17) using a Ser(P) 133 -CREB peptide (amino acids 123-135; New England Biolabs, Beverly, MA) in the presence of [␥-32 P]ATP (specific activity, 3000 Ci/mM; Amersham Biosciences). Cells were treated with 20 M LY294002 (Sigma), a PI3K-specific inhibitor, to inhibit Akt and isolate GSK-3␤ in an active form. The results were normalized as a relative cpm incorporated into a CREB peptide, compared with untreated control samples, and analyzed for statistical significance using the StatView software program (SAS Institute, Inc., Cary, NC).
Nuclear Extracts-Nuclear extracts were prepared as described by Schaefer et al. (31). Briefly, after cells were washed with cold phosphate-buffered saline wash solution (phosphate-buffered saline containing 0.1ϫ BM complete protease mixture tablet (Roche Molecular Biochemicals) and 10 M Na 2 VO 4 ), they were scrape-harvested into 1.5-ml conical tubes. Cell suspensions were centrifuged at 3000 revolutions per min (rpm) at 4°C for 5 min, and pelleted cells were resuspended in 100 l of hypotonic lysis buffer (10 mM Hepes, pH 8.0, 1.5 mM MgCl 2 , 10 mM KCl, 1 mM dithiothreitol, 25 mM NaF, 1 mM Na 2 VO 4 , and 1ϫ BM complete protease mixture inhibitors) and incubated on wet ice for 10 min. The resuspended lysate was centrifuged at 7500 rpm at 4°C for 5 min. The nuclear suspension as washed 2ϫ and centrifuged at 12,000 rpm for 5 min. Nuclear preparations were examined by light microscopy for the presence of cytosolic components, which were not present. The nuclear pellets were lysed with sodium dodecyl sulfate sample buffer and DNA-sheared through a 20-gauge needle. These lysates were then compared with total cell lysates for the presence of actin. Only trace amounts of actin were present in the nuclear lysate preparations, whereas the actin content was very high in the total cell lysate (data not shown).
Western Blot Analysis-PC-3 cells were left untreated or treated with 10 ng/ml TNF-␣, 1 ng/ml IGF-1, 10 ng/ml EGF, or 1 M insulin for 16 h at 37°C and subsequently analyzed for expression and activation of various proteins. Whole-cell lysates were made from various PC-3 cell samples according to a previously described method (31). Proteins were separated on a NuPAGE 6 -12% gradient gel (Novex, Carlsbad, CA), electrotransferred to a nitrocellulose membrane, and immunoblotted with a primary antibody overnight at 4°C. Immune complexes were visualized via Super Signal chemiluminescence (Pierce).
To normalize LUC activity either pCRE-LUC or pAP1-LUC was transfected in excess of 6-fold (0.86 g/ml) greater than Renilla-LUC (0.13 g/ml; pTK-LUC, Promega, Madison, WI). The transfected cells were harvested and analyzed using the Firelight system (a gift from Corinn Rich and Packard Instrument Co., Naperville, IL) in accordance with the manufacturer's instructions. Quantification was performed in a 96-well black plate using a TOPCOUNT multiwell plate scintillation detector (Packard Instrument Co.) in the photon-counting mode. All pCRE-LUC or pAP1-LUC data were normalized to Renilla-LUC that was co-transfected to determine the effect on the relative promoter activity. To establish if there was any change in the level of pCRE-LUC and pAP1-LUC DNA in PC-3 cells or those cotransfected with GSK-3␤ LUC, DNA template levels were determined by PCR using primers designed to amplify LUC. The levels of LUC DNA were the same in cells transfected with either pCRE-LUC or pAP1-LUC DNA alone or when LUC reporter constructs were cotransfected with any of the GSK-3␤ expression vectors (data not shown).
Densitometry-These experiments were performed multiple times and quantified by densitometry, and these data were subjected to statistical analysis. All densitometric analyses were performed using a Personal Densitometer SI (Molecular Dynamics, Sunnyvale, CA) and the corresponding software program, ImageQuant (Molecular Dynamics). Images were quantified using NIH Image 1.62 (National Institutes of Health), and statistical analysis was performed using StatView 5.01 (SAS Institute Inc.). The density of the proteins analyzed was normalized to the appropriate controls (e.g. actin). Student's t test was used to determine the significant differences between the mean relative densities of the various experimental groups, represented as p values.
After the cells were stained with secondary antibodies, counterstaining was performed using 500 nM 4Ј,6-diamidino-2-phenylindole dilactate (Molecular Probes) to identify the nuclei and 1 unit/ml Alexa 594-phalloidin (Molecular Probes) in TBST at 4°C to stain the cytoplasmic actin. The samples were incubated overnight and then washed and mounted on glass slides using Prolong antifade solution (Molecular Probes). The slides were then analyzed via epifluorescence microscopy, and data were acquired using digital image analysis as described previously (35).
Cell Growth-Cell growth was measured using the vital dye calcein AM (CAM) ester (Molecular Probes), which is membrane-permeable and nonfluorescent before activation by nonspecific esterases within viable cells. The cleavage product CAM emits green fluorescence. Cells transfected with either GSK-3␤ expression vectors or control vector were plated in 96-well plates and incubated with CAM ester in HEPESbuffered saline solution for 15 min at 25°C. Fluorescence was quantified using a Biolumin 9600 plate reader at 488 nm, and data were analyzed using the Statview software program (SAS Institute, Inc.).
Immunostaining of Paraffin Sections for Ser(P) 129 -S(P) 133 -CREB-Immunohistochemistry was carried out in paraffin-embedded sections (5 m) after deparaffinization and rehydration in xylene, graded alcohol, and phosphate-buffered saline, respectively. The endogenous peroxidase activity was quenched by incubating the sections in 0.3% hydrogen peroxide for 20 min at room temperature. After blocking the sections with 1.5% normal horse serum in phosphate-buffered saline for 1 h, anti-Ser(P) 129 -Ser(P) 133 -CREB was applied to the sections and incubated overnight at 4°C. Then the sections were incubated with the appropriate biotinylated secondary antibody and ABC-AP reagent according to the manufacturer's instructions (Vectastain ABC Elite kit, Vector Laboratories, Burlingame, CA). Peroxidase activity was detected with by applying 3,3Ј-diaminobenzidine tetrahydrochloride containing 0.02% of hydrogen peroxide for 10 min. The sections were counterstained with hematoxylin and mounted with coverslips.
Molecular Modeling-Structural data obtained from Protein Data Bank code 1KDX (36) were used to generate a model of CREB-kinase inducible domain (KID) and CBP-KID interaction domain interactions. BiotechniX 3d software (Gentech, Parc Sophia Antipolis, France) was used for three-dimensional rendering of CREB on a G4 Titanium Powerbook Computer (Apple Computer Corp., Cupertino, CA).

GSK-3␤ Activity Is Suppressed in PC-3 Cells-Endogenous
GSK-3␤ enzymatic activity was low in PC-3 cells (Fig. 1A, bar 1) as determined by the immune complex kinase assay using CREB peptide as the substrate. This low GSK-3␤ activity in untreated cells correlated with a relatively high level of phosphorylation on Ser 9 (Fig. 1B, lane 1). Because PI3K is involved in the activation of Akt, which in turn phosphorylates GSK-3␤ at Ser 9 and, thus, inactivates it, we treated cells with LY294002, a PI3K inhibitor, to determine whether GSK-3␤ is suppressed through this pathway. LY294002 increased GSK-3␤ activity ϳ2-fold (Fig. 1A, bar 2), and this effect was enhanced by combining LY294002 with TNF-␣ (Fig. 1A, bar 4). This increase in enzyme activity correlated with a decrease in the phosphorylation of Ser 9 with LY294002 alone (Fig. 1B, lane  2), which decreased further when LY294002 treatment was combined with TNF-␣ ( Fig 1B, lane 8). Furthermore, PC-3 cells were treated with increasing concentrations of LY294002, and then enzymatic activity was determined by CREB peptide phosphorylation along with measuring the phosphorylation of Ser 9 by Western analysis. After densitometric quantification, regression analysis was performed using the Statview program to compare the relative GSK-3␤ enzymatic activity in relation to the inverse value of the mean phospho-Ser 9 density. An R 2 value approximately equal to 0.9 indicated that the relative activity of GSK-3␤ decreased in conjunction with increasing the phosphorylation of GSK-3␤ on Ser 9 . In addition, two PI3K stimulators, EGF and IGF-1 (Fig. 1B, lanes 3 and 4, respectively), did not further suppress GSK-3␤ activity by increasing the phosphorylation of GSK-3␤ at Ser 9 beyond control levels. Similar studies showed that these treatments caused a decrease in ␤-catenin levels (Fig. 2). Furthermore, results similar to those observed for EGF and IGF were obtained using insulin (data not shown). These results indicate that endogenous GSK-3␤ activity is largely suppressed by the PI3K pathway in proliferating PC-3 cells and that activation of GSK-3␤ enhanced the breakdown of ␤-catenin.
GSK-3␤ Stimulated CRE-mediated Transcription and Suppressed Activator Protein-1 (AP1)-mediated Transcription-CREB and Jun family members can take part in binding to and activating CRE-containing promoters; Jun family members can also activate AP1 transcription response elements. To determine how CRE and AP1 transcriptional activities were affected by GSK-3␤, PC-3 cells were cotransfected with GSK-3␤, with either the CRE or AP1 response element-containing luciferase reporter constructs (CRE-LUC and AP1-LUC, respectively). The CRE-LUC activity in PC-3 cells cotransfected with GSK-3␤ was 6-fold higher than that in cells transfected with CRE-LUC alone (Fig. 3). The LUC vector without the CRE elements (pCISCK) showed no LUC activity, and the expression vector pCMV4 without the GSK-3 cDNA did not affect LUC activity (data not shown). We also observed that in the absence of cotransfected GSK-3␤, AP1-LUC activity was about 20 times higher than for CRE-LUC. In contrast to its effect on CRE-LUC activity, GSK-3␤ strongly inhibited AP1-LUC activity. These data indicate that overexpression of GSK-3␤ can differentially regulate CREB-mediated and AP1-mediated activity in PC-3 cells.
Phosphorylation of GSK-3␤ Variants-We next investigated the overexpression of GSK-3␤ proteins that contained various point mutations or deletions to determine how these molecular changes affected the protein phosphorylation patterns and the intracellular distribution of GSK-3␤ (Fig. 4). Deletion of the first nine amino acids of GSK-3␤ results in the GSK-3␤ ⌬9 protein, which is constitutively active and cannot be inactivated by PI3K/Akt. Ser 9 phosphorylation was present on all forms of GSK-3␤ except on the GSK-3␤ ⌬9 transfectants (Fig.  4A, ⌬9 lane lower band is absent). In contrast, GSK-3␤ Y216F contains a mutation of tyrosine Tyr 216 to a phenylalanine, resulting in a dominant-negative-acting protein that cannot activate CREB. Tyr 216 was phosphorylated on endogenous GSK-3␤ in PC-3 cells (Fig. 4A, lane C) and on the wild-type (wt) and ⌬9-transfected forms of GSK-3␤. The total Tyr 216 phosphorylation was reduced in the GSK-3␤ Y216F transfectants, which included some Tyr 216 phosphorylation on endogenous GSK-3␤, since the phenylalanine on GSK-3␤ Y216F cannot be phosphorylated (Fig. 4A, lane Y216F).
Intracellular Distribution of GSK-3␤ Variants-Examination of the intracellular distribution of the different GSK-3␤ proteins by immunofluorescence using a GSK-3␤-specific antibody showed that untransfected and empty vector-transfected control cells contained low levels of endogenous GSK-3␤ in both the cytoplasm and nucleus of PC-3 cells (Fig. 4B, control (C)). Exogenous, wt GSK-3␤ (when overexpressed) was found in both the cytoplasm and the nucleus of PC-3 cells (Fig. 4B, wt). The GSK-3␤ Y216F protein was predominantly found in the cytoplasm, and the GSK-3␤ ⌬9 protein was almost exclusively nuclear (Fig. 4B, Y216F and ⌬9).
Active GSK-3␤ Expression Suppresses Cell Growth-Forced expression of wt GSK-3␤ suppressed growth by 12% and the GSK-3␤ ⌬9 by 33%, whereas the expression of empty vector and GSK-3␤ Y216F had no effect (Fig. 6). These data indicate that the overexpression of enzymatically active GSK-3␤ proteins suppresses the prostate cell growth.
Effects of Totally Suppressing Jun Transactivation on CRE-LUC Activity-Jun can increase CRE promoter activity but to a lesser extent than CREB (37). Because CREB and Jun can both activate CRE, we examined GSK-3␤ effects on CRE transactivation in the absence of Jun activity. To achieve this condition we attempted to eliminate Jun transactivation in PC-3 cells. The phosphorylation of Jun at Ser 63 by Jun N-terminal kinase (JNK) controls the transactivation state of Jun (38,39). By limiting Jun phosphorylation at Ser 63 , it should be possible to limit the effects of Jun on CRE. The JNK binding domain (JBD) fragment of the scaffold protein Jun-interacting protein-1 competitively binds JNK and prevents its activation in cells (32,40). JBD was expressed in PC-3 cells to sequester JNK and, thereby, minimize its ability to transactivate Jun (Fig. 7B). When JBD was coexpressed with wt GSK-3␤ in PC-3 cells, it doubled the level of CRE-LUC activity. GSK-3␤ Y216F sity of Ser(P) 9 using a Student's t test. Statistical analysis showed a significant difference between control samples and TNF-␣ treated samples (a, p Ͻ 0.004) and between control samples and all of the LY294002-treated samples (b, p Ͻ 0.001).

FIG. 1. The activity of GSK-3␤ is suppressed in proliferating PC-3 cells.
A, GSK-3␤ enzymatic activity was assessed by immune complex kinase assay using Ser(P) 133 -CREB peptide in the presence of [␥-32 P]ATP in PC-3 cells. Low levels of GSK-3␤ enzyme activity were observed in untreated cells (bar 3). Enzyme activity increased after treating cells with 20 M LY294002 (lane 2, a PI3K-specific inhibitor) and increased further after treatment with LY294002 plus TNF-␣ (bar 4). GSK-3␤ activity was represented as relative cpm incorporated into a CREB peptide normalized to untreated control samples from two independent experiments. B, after PC-3 cells were either left untreated (Ϫ) or treated (ϩ) with TNF-␣ (10 ng/ml), EGF (10 ng/ml), or IGF-1 (1 ng/ml) in the absence (lanes 1-4) or presence (lanes 5-8) of PI3Kspecific inhibitor LY294002 (20 M) for 16 h. Western analysis was performed using primary antibodies for Ser(P) 9 -GSK-3␤, Tyr(P) 216 -GSK-3␤, or total GSK-3␤ followed by chemiluminescence. Ser(P) 9 levels varied according to treatment, whereas Tyr(P) 216 and total GSK-3␤ levels did not change. Three independent experiments were quantified by densitometry, and the density of Ser(P) 9 was normalized to total GSK-3␤. Statistical analysis was performed on the mean relative den-  1-4) or presence (lanes 5-8) of PI3K-specific inhibitor LY294002 (20 M) for 16 h, and whole-cell lysates were prepared and analyzed by Western blot. Treatment with PI3K inhibitor LY294002 caused a significant decrease in ␤-catenin levels. Statistical analysis showed a significant difference (a, p Ͻ 0.003) between control samples and TNF-␣or LY294002-treated samples.

FIG. 3. GSK-3␤ effects on Jun/AP1-and CREB/CRE-mediated transcriptional activity in PC-3 cells. PC-3 cells were transiently
transfected with 1 g each of either pAP1-LUC or pCRE-LUC (firefly LUC) alone or in combination with 1 g of pGSK-3␤. All samples were cotransfected with 1 g of pTK-LUC (Renilla-LUC) for normalization purposes. Firefly LUC activity results are presented as mean relative LUC activity units (ϮS.E.) that were normalized to Renilla LUC activity from two independent experiments. alone or when combined with JBD did not stimulate CRE. However, the combination of GSK-3␤ ⌬9 and JBD hyperactivated CRE-LUC activity to the highest level that we observed (130-fold). These data suggest that Jun, which is a weaker activator of CRE-driven transcription than CREB itself, when fully suppressed (by combining GSK-3␤-induced phosphorylation on the Jun DNA binding domain with the elimination of Jun transactivation by JNK), allows a stronger transactivation by GSK-3␤-activated CREB.
Phosphorylation Patterns of CREB after Transfecting PC-3 Cells with GSK-3␤ ⌬9 Alone or in Combination with JBD-Examination of CREB and Jun phosphorylation patterns in PC-3 cells using Western blot analysis (Fig. 8A) showed that CREB was phosphorylated on Ser(P) 133 and remained unchanged in PC-3 cells under all conditions examined (data not shown). In contrast to Ser(P) 133 -CREB, the level of Ser(P) 129 -Ser(P) 133 -CREB was low in control nuclear lysates but significantly increased in the presence of GSK-3␤ ⌬9 alone or combined with JBD; these phosphorylation levels were amplified further after TNF-␣ treatment. In contrast to the effects on CREB phosphorylation patterns in PC-3 cells, phosphorylation of Jun on Ser 63 was not strongly suppressed by transfection with GSK-3␤ ⌬9 alone but was completely eliminated by JBD, consistent with the idea that expression of JBD inactivates Jun (data not shown).

Effects of GSK-3␤ ⌬9 Alone or in Combination with JBD on Intracellular Distribution of Ser(P) 133 -CREB and Ser(P) 129 -Ser(P) 133 -CREB-
We used our affinity-purified antibodies to localize and determine the phosphorylation status of Ser(P) 129 -Ser(P) 133 -CREB and Ser(P) 133 -CREB in cells transfected with GSK-3␤ ⌬9 alone or combined with JBD (Fig. 8B). The level of Ser(P) 133 -CREB in the nuclei of PC-3 cells did not change significantly under any conditions examined. In contrast to Ser(P) 133 -CREB, Ser(P) 129 -Ser(P) 133 -CREB labeling was absent from control PC-3 cell samples, but nuclear labeling of Ser(P) 129 -Ser(P) 133 -CREB increased when GSK-3␤ ⌬9 alone was transfected into PC-3 cells or combined with JBD. These data indicate that overexpression of the constitutively active GSK-3␤ ⌬9 protein in PC-3 cells leads to phosphorylation of Ser 129 on CREB in the cell nuclei after CREB has already been prephosphorylated at Ser 133 .
(P)S 129 -(P)S 133 -CREB Is Present in the Nuclei of Prostate Tissue Samples-Paraffin-embedded prostate tissue samples were analyzed by immunohistochemical staining for Ser(P) 129 -

FIG. 4. Phosphorylation and intracellular distribution of GSK-3␤ in PC-3 cells. PC-3 cells were transfected
with control (C), empty vector (lane 1) or various GSK-3␤ constructs (wt, ⌬9, and Y216F (lanes 2-4)) and examined for phosphorylation status using Ser(P)-9-GSK-3␤, Tyr(P)-216-GSK-3␤, or total GSK-3␤ antibodies. A, the expression products of the Y216F point-mutant construct (lane 4) or ⌬9-deletion construct (lane 3, arrows) failed to label missing amino acids with the appropriate anti-Tyr(P) 216 and anti-Ser(P) 9 antibodies, respectively. Total GSK-3␤ protein expression was detected using a GSK-3␤-specific antibody. Three independent experiments were quantified by densitometry, and the density of Ser(P) 9 was normalized to total GSK-3␤. Statistical analysis was performed using Student's t test for the mean relative density of Ser(P) 9 . Statistical analysis showed a significant difference between control samples and wt-transfected samples (a, p Ͻ 0.04) or control samples and the upper band (b, p Ͻ 0.01) or lower missing band (arrow; c, p Ͻ 0.0005) in the ⌬9 deletion constructtransfected samples. B, the intracellular distribution of GSK-3␤ was examined after transfection using GSK-3␤-specific antibody followed by an Alexa 488-conjugated secondary antibody. To achieve sufficient resolution of the point light source of interest (cells overexpressing GSK-3␤), the exposure times in our experiments were reduced, which caused decreased resolution of the remaining cells (endogenous GSK-3␤). These fluorescence images were taken at a set time exposure for all of the samples to make valid comparisons between these conditions and to achieve optimal resolution of the cells overexpressing GSK-3␤. The appearance of endogenous GSK-3␤ was low in the empty vector-transfected control samples; the control and wt-transfected cells showed both cytoplasmic and nuclear distribution of GSK-3␤. Y216F protein was predominantly found in the cell cytoplasm, whereas the ⌬9 protein was predominantly located in the cell nucleus.
Ser(P) 133 -CREB. Ser(P) 129 -Ser(P) 133 -CREB expression was high in the nuclei of low grade prostate cancer (well differentiated) but decreased in high grade (poorly differentiated) prostate cancer (Fig. 9). DISCUSSION GSK-3␤ is a primary target of Akt, which inhibits GSK-3␤ function by phosphorylating it on Ser 9 in proliferating cells (16 -19). Although many studies indicate that Akt is highly active in cancer (15), it is mainly by inference that GSK-3␤ activity has been thought to oppose Akt function in cancer cells. Our approach was to elevate GSK-3␤ activity in prostate cancer cells and examine the consequences of this activation. Our data show that basal levels of GSK-3␤ kinase activity were low in association with the high levels of GSK-3␤ phosphorylation on Ser 9 . We successfully increased the enzymatic activity of GSK-3␤ in PC-3 cells by decreasing the phosphorylation of GSK-3␤ on Ser 9 through various methods, including the inhibition of PI3K by LY294002 treatment (Fig. 1, A and B) or TNF-␣ treatment (Fig. 1, A and B) or the forced expression of a Ser 9 deletion mutant form of GSK-3␤ ⌬9 that was constitutively active (Fig. 4A). Because GSK-3␤ phosphorylates and thereby regulates the functions of numerous proteins involved in signaling, metabolism, and protein synthesis and structural proteins (1, 3, 4), we examined certain transcription factors as downstream targets of activated GSK-3␤.
To determine the downstream consequence of increasing GSK-3␤ activity, we examined the phosphorylation of CREB and its mediation of CRE transcriptional activity in PC-3 cells. We chose to examine CREB because the phosphorylation of CREB on Ser 129 by GSK-3␤ and activation of CRE transcription have been demonstrated in vitro but not in intact cells (3). Our data show that the ability of GSK-3␤ to phosphorylate the CREB nuclear transcription factor in PC-3 cells depends not only on increasing GSK-3␤ activity but also on the presence of GSK-3␤ in the nucleus. We showed that CREB is fully activated by serial phosphorylation beginning at Ser 133 and then at Ser 129 by constitutively active GSK-3␤ ⌬9 in cell nuclei (Fig. 8B).  2-4)), subjected to detergent lysis, and examined by Western analysis for ␤-catenin and actin. Two independent experiments were quantified by densitometry, and the density of ␤-catenin was normalized to actin. Statistical analysis was performed using Student's t test for the mean relative density of ␤-catenin. Statistical analysis showed a significant difference between control samples and wt-transfected samples (a, p Ͻ 0.006) or control samples and cells transfected with the ⌬9 deletion construct (b, p Ͻ 0.006). The greatest increases in CRE-mediated LUC reporter activity occurred when active forms of GSK-3␤ were present in PC-3 cell nuclei. GSK-3␤ ⌬9 was the strongest stimulator of CREdriven LUC reporter activity (80-fold, Fig. 7A), and GSK-3␤ ⌬9 was found almost exclusively in the nucleus (Fig. 4B). In con-trast, GSK-3␤ Y216F failed to activate the CRE-LUC reporter (Fig. 7A) and did not enter the nucleus (Fig. 4B). The presence of GSK-3␤ in the nucleus was influenced by phosphorylation on Tyr 216 , and the activation state of the enzyme was controlled by phosphorylation on Ser 9 . Phosphorylation on Tyr 216 and Ser 9 FIG. 8. Effects of GSK-3␤ construct expression on CREB and Jun phosphorylation. PC-3 cells were transfected with GSK-3␤ ⌬9 alone or with JBD, treated with TNF-␣ for 4 h or left untreated, and harvested. Then nuclear lysates were isolated. A, nuclear lysates were separated on a gradient gel and immunoblotted using primary antibodies against Ser(P) 129 -Ser(P) 133 -CREB, Ser(P) 133 -CREB, total CREB; immunoblotting was followed by chemiluminescence detection. Two independent experiments were quantified by densitometry and the density of Ser(P) 129 -Ser(P) 133 -CREB that was normalized to total CREB. Statistical analysis of mean density levels was performed using Student's t test. A significant difference was observed between control samples and ⌬9 deletion construct alone (a, p Ͻ 0.006) or when treated with TNF-␣ (b, p Ͻ 0.002). Control samples were also statistically different from ⌬9 deletion construct cotransfected with JBD (c, p Ͻ 0.03) or when these cotransfected cells were treated with TNF-␣ (d, p Ͻ 0.03). B, PC-3 cells were transfected with GSK-3␤ ⌬9 alone or with JBD and then stained for immunofluorescence with a Ser(P) 129 -Ser(P) 133 -CREB or a Ser(P) 133 -CREB primary antibody followed by an Alexa 488 secondary antibody detection (green). Counterstaining was performed to detect actin (Alexa 594-phalloidin, red) and DNA (4,6-diamidino-2-phenylindole, blue). was important for determining whether GSK-3␤ could phosphorylate CREB in the cell nucleus.
Although we determined that constitutively activated GSK-3␤ ⌬9 was available in the nucleus to interact with preactivated Ser(P) 133 -CREB, the critical question remains as to whether Ser 129 is available as an exposed substrate for GSK-3␤ ⌬9 . Based on structural analysis, it is likely that Ser 129 on CREB is available for GSK-3␤ ⌬9 phosphorylation during CREB interactions with CREB-binding protein in transcriptional complexes. Ser(P) 129 -CREB occurs in the exposed peptide loop at the junction of the KID ␣A helix (Fig. 10, A and B) (36). This position is adjacent to the Ser(P) 133 -CREB priming site, which is at the ␣B helix of the KID domain. The presence of GSK-3␤ in the nucleus and the availability of Ser 129 -CREB as a substrate support a role for GSK-3␤ in catalyzing the phosphorylation of Ser 129 -CREB in cells.
GSK-3␤ inhibits the activity of many nuclear proteins, including Jun. The prime exception is CREB, which we show is activated in intact cells by GSK-3␤ through the phosphorylation of Ser 129 . In PC-3 cells, GSK-3␤ appears to control the balance between Jun and CREB-mediated transcriptional activity. Transfecting PC-3 cells with constitutively active GSK-3␤ ⌬9 plus JBD shifted the balance of CRE-LUC-mediated transcriptional activity from Jun-to CREB-dominated transcription. JBD has been shown to selectively inhibit JNK activity and prevent Jun and/or ATF-2 transactivation (32,40). When we cotransfected PC-3 cells with JBD and GSK-3␤ ⌬9 , basal-level CRE activity increased 130-fold (Fig. 7B). In this experiment, Jun was expected to simultaneously lose both DNA binding activity (because of phosphorylation by GSK-3␤ ⌬9 ) and transcriptional transactivating activity (because of JNK suppression by JBD). Our observation of suppressed Jun phosphorylation in the presence of JBD was consistent with this expectation and indicated that suppressed JNK activity plus elevated GSK-3␤ activity minimizes Jun activity in PC-3 cells.
We have additional support for the involvement of CREB phosphorylation by GSK-3␤ in prostate cancer from the examination of prostate cancer tissue samples (Fig. 9). These data show that the phosphorylation of Ser(P) 129 -Ser(P) 133 -CREB is high in low grade (well differentiated) prostate cancer but decreases in high grade (poorly differentiated) cancer, corresponding to a decrease in GSK-3␤ activity. These findings suggest that Ser(P) 129 -Ser(P) 133 -CREB may be associated with differentiation status in these tissues. Furthermore, CREBdriven elevation of CRE transcriptional activity is known to enhance cell differentiation, which often involves the suppression of cell growth (41)(42)(43). The growth of PC-3 cells was significantly suppressed when enzymatically active GSK-3␤ was overexpressed in these cells (Fig. 6). Little is known about the genetic markers that are associated with prostate cell differentiation, but the usefulness of differentiating agents in prostate cancer therapy is well recognized (44). Although CREB activation may induce PC-3 cells to differentiate, further investigation will be necessary to determine the gene expression patterns that are involved in this process.
GSK-3␤ inhibits the stability of many proteins, including ␤-catenin levels, through the influence of the Wnt pathway in FIG. 9. Ser(P) 129 -Ser(P) 133 -CREB in the nuclei of prostate tissue samples. Paraffin-embedded prostate tissue was analyzed by immunohistochemical staining for Ser(P) 129 -Ser(P) 133 -CREB. Ser(P) 129 -Ser(P) 133 -CREB expression was high in the nuclei of low grade prostate cancer (well-differentiated) (A) but decreased in high grade (poorly differentiated) prostate cancer (B), corresponding to a decrease in GSK-3␤ activity that is expected to occur in advanced stage prostate cancer.
FIG. 10. Molecular models of Ser 129 -CREB availability. A and B illustrate that Ser 129 is available for phosphorylation by GSK-3␤ at the junction of the ␣B helix and the exposed peptide loop of the CREB-KID domain. The Ser 129 structure at the ␣A helix junction of the CREB kinase-inducible domain (KID) domain is exposed for phosphorylation after priming by Ser(P) 133  cancer cells. GSK-3␤ phosphorylates ␤-catenin on Ser-33 and -37 and Thr-41. This primes ␤-catenin for ubiquitination and lysis by proteasomes (45,46). Mutations of these Ser/Thr phosphorylation sites in ␤-catenin prevent proteolysis and cause the accumulation of ␤-catenin/TCF-4 complexes in nuclei, which activates transcription of a variety of genes (45,46). These mutations in ␤-catenin are reported to occur in 5% of prostate cancers (47,48). However, many prostate cell lines, including the PC-3 line, do not contain these mutations (47,48). The overexpression of wt GSK-3␤ or the constitutively active GSK-3␤ ⌬9 or treatment with PI3K inhibitor LY294002 all caused ␤-catenin levels to decrease (Figs. 3 and 5). These data indicate that ␤-catenin was normally regulated in PC-3 cells and not likely to affect transcriptional activity in these studies.
Our data also help establish a link between TNF-␣ responsiveness and PI3K/3-phosphoinositide-dependent kinase 1/Akt activation during the regulation of GSK-3␤ function. In studies by Hoeflich and coworkers (30), the loss of GSK-3␤ caused embryonic lethality in knockout mice and involved liver degeneration that correlated with hypersensitivity to TNF-␣. TNF-␣ caused the transcriptional activation of NFB through a GSK-3␤-dependent mechanism, but the GSK-3␤ Ϫ/Ϫ defect did not involve pathway activation through either IB inhibitor complex degradation or NFB nuclear import, because these activities were normal in GSK-3␤ Ϫ/Ϫ cells (30). These findings suggested that GSK-3␤ regulation of NFB-driven survival occurs at transcriptional complexes or involves other factors such as those of the PI3K/3-phosphoinositide-dependent kinase 1/Akt pathway since the activation of this pathway is known to regulate prostate cancer cell sensitivity to TNF-␣ (27)(28)(29). We show that treating PC-3 cells with TNF-␣ increases GSK-3␤ enzymatic activity, which synergizes with LY294002 treatment (Fig. 1A). Consistent with increases in GSK-3␤ enzymatic activity, TNF-␣ treatment alone or in combination with LY294002 treatment decreased the phosphorylation of GSK-3␤ on Ser 9 (Fig. 1B). TNF-␣ also increased Ser(P) 129 -Ser(P) 133 -CREB formation in the presence of GSK-3␤ ⌬9 alone or when combined with JBD, a suppressor of Jun activity (Fig. 8A). Our data suggest that activating GSK-3␤ in cells that normally suppress this enzyme will overcome cancer cell resistance to TNF-␣. Further study of GSK-3␤ regulation on these and other essential regulatory proteins is needed to increase our understanding of the precise mechanisms by which GSK-3␤ modulates the behavior of prostate cancer cells.