Resolution of the Nuclear Localization Mechanism of Glycogen Synthase Kinase-3

Mechanisms regulating the nuclear localization of glycogen synthase kinase-3β (GSK3β) remained enigmatic despite the crucial regulation by nuclear GSK3β of important cellular functions. These include regulation of gene expression, cell cycle progression, and apoptosis, achieved by the phosphorylation by GSK3 of nuclear substrates (e.g. numerous transcription factors). We resolved this mechanism by identifying a bipartite nuclear localization sequence (NLS) that is necessary for the nuclear accumulation of GSK3β and is sufficient to drive yellow fluorescent protein into the nucleus. Despite the NLS, most GSK3β is cytosolic, sequestered in protein complexes that, although still mobile in the cytosol, block the NLS. Conditions promoting nuclear translocation of GSK3β release it from cytosolic complexes, allowing the NLS to direct nuclear import. Using this information to prepare a nucleus-excluded active GSK3 construct, we found that the antiapoptotic effect of GSK3β in tumor necrosis factor-induced apoptosis is mediated by cytosolic, not nuclear, GSK3β. Identification of a GSK3β NLS allows new strategies to decipher and manipulate its subcellular actions regulating gene expression and apoptosis and its involvement in diseases.

Cells have the inherent ability to rapidly and concurrently modulate the actions of many proteins in response to environmental and biological cues. A large part of these coordinated cellular responses are due to posttranslational modifications of key regulatory proteins. This requires the appropriate localization of the modifying enzymes, particularly protein kinases, since phosphorylation is the predominant mechanism for protein posttranslational modification. Dynamic fluxes in the nuclear level of the serine/threonine kinase glycogen synthase kinase-3␤ (GSK3␤) 2 occur during key cellular events. Nuclear GSK3␤ levels markedly increase when cells enter the S-phase of the cell cycle, followed by a loss of nuclear GSK3␤ as the cell cycle progresses (1). Nuclear GSK3␤ levels decrease in response to stimulation by proliferative growth factors, whereas insults that induce apoptosis can cause accumulation of GSK3␤ in the nucleus prior to the activation of the caspase cascade (2). Additionally, GSK3␤ in the nucleus is highly active relative to cytosolic GSK3␤ (3). Fluxes in GSK3␤ in the nucleus at critical periods may be related to the well documented capacity of nuclear GSK3␤ to regulate many transcription factors, such as nuclear factor-B (NF-B), cyclic AMP response elementbinding protein (CREB), Snail, p53, AP-1, Myc, and others, that exert widespread effects on cellular functions by regulating the expression of many genes (4 -7). Thus, cellular phenotype and survival can be influenced by the level of GSK3␤ in the nucleus. Despite these crucial cellular events involving nuclear GSK3␤, a longstanding problem remains unresolved, deciphering the mechanism of its nuclear import.
There are two isoforms of GSK3, GSK3␤ and GSK3␣, and these are involved in modulating a large and diverse number of cellular functions, including metabolism, development, cellular architecture, gene expression, and survival (7)(8)(9). GSK3␤ is constitutively active, but several mechanisms contribute to controlling its actions. Inhibitory phosphorylation by Akt and other kinases occurs on serine 9 of GSK3␤. Conversely, phosphorylation of GSK3␤ on tyrosine 216 increases its activity (8,10). GSK3␤ is also regulated by protein complex formation, especially well known in the cytosolic Wnt signaling pathway, where GSK3␤ associated with a large protein complex phosphorylates ␤-catenin to promote its degradation (11). Phosphorylation of substrates by GSK3␤ is also usually controlled by the phosphorylation state of the substrate, because most substrates of GSK3␤ must be "primed," prephosphorylated at a residue four amino acids C-terminal to the GSK3␤ target site. A mutation in the primed substrate binding pocket of GSK3␤, R96A, that blocks its binding to primed substrates is widely used to identify primed substrates of GSK3␤ (12). In addition to these three regulatory mechanisms, the subcellular distribution of GSK3␤ regulates its actions by controlling its accessibility to substrates, such as those in the nucleus (7).
Nuclear levels of proteins are balanced by export and import mechanisms. The nuclear export of GSK3␤ is inhibited by lep-tomycin B, indicating that at least a portion of GSK3␤ nuclear export is CRM1-dependent (2). Nuclear export of GSK3␤ is partially mediated by FRAT-1, which binds GSK3␤ in the nucleus followed by export of the complex via the nuclear export sequence (NES) in FRAT-1 (13). To counter export, we hypothesized that GSK3␤ may contain an NLS; however, examination of the primary amino acid sequence did not reveal a consensus NLS, nor was one identified using the CUBIC NLS prediction algorithm (14). Therefore, we examined regions in GSK3␤ rich in basic amino acids (lysine and arginine), a general requirement for an NLS. This enabled us to identify an NLS motif in GSK3␤ that is necessary for nuclear import of GSK3␤ and is sufficient to drive the nuclear import of yellow fluorescent protein (YFP). Further experiments revealed that the subcellular distribution of GSK3␤ is regulated by the NLS, by the N-terminal tail of GSK3␤, and by cytosolic protein complexes that are capable of blocking the NLS function by sequestering GSK3␤ in the cytosol. Finally, the controversial intracellular site of action of GSK3␤ that inhibits tumor necrosis factor (TNF)-induced cytotoxicity was resolved by showing that this was mediated by NLS-deficient GSK3␤ in the cytosol.
Immunoblotting-Cells were washed twice with phosphatebuffered saline (PBS) and were lysed with lysis buffer (20 mM Tris, pH 7.5, 150 mM NaCl, 2 mM EDTA, 2 mM EGTA, 0.5% Nonidet P-40, 1 mM sodium orthovanadate, 100 M phenylmethanesulfonyl fluoride, 0.1 M okadaic acid, 50 mM sodium fluoride, and 10 g/ml each of leupeptin, aprotinin, and pepstatin). The lysates were sonicated and centrifuged at 20,800 ϫ g for 15 min. Protein concentrations were determined by the bicinchoninic acid method (Pierce). Samples were mixed with Laemmli sample buffer (2% SDS) and placed in a boiling water bath for 5 min, and proteins were resolved in SDS-polyacrylamide gels, transferred to nitrocellulose, and incubated with antibodies to phospho-Ser 9 -GSK3␤ (Cell Signaling, Danvers, MA), total GSK3␤, total poly(ADP-ribose) polymerase (BD Phar-Mingen/Transduction Laboratories, San Diego, CA), total GSK3␣/␤ (Upstate Biotechnology, Inc., Lake Placid, NY), or HA tag (Covance, Berkeley, CA). Immunoblots were developed using horseradish peroxidase-conjugated goat anti-mouse or goat anti-rabbit IgG, followed by detection with enhanced chemiluminescence, and the protein bands were quantitated with a densitometer. Subcellular Fractionation-Nuclear and cytosolic fractions were prepared as previously described (2) with minor modifications. Cells were washed twice with PBS and then harvested in 200 l of lysis buffer (10 mM Tris, pH 7.5, 10 mM NaCl, 3 mM MgCl 2 , 0.05% Nonidet P-40, 1 mM EGTA, 1 mM sodium orthovanadate, 100 M phenylmethanesulfonyl fluoride, 0.1 M okadaic acid, 50 mM sodium fluoride, and 10 g/ml each of leupeptin, aprotinin, and pepstatin). Cells were centrifuged at 2700 ϫ g for 10 min at 4°C. The supernatant was centrifuged at 20,800 ϫ g for 15 min at 4°C to obtain the cytosolic fraction. The pellet containing nuclei was washed twice in 200 l of wash buffer (5 mM HEPES, pH 7.4, 3 mM MgCl 2 , 1 mM EGTA, 250 mM sucrose, 0.1% BSA, with protease and phosphatase inhibitors). The pellet was then resuspended in wash buffer and layered on top of 1 ml of 1 M sucrose (with protease and phosphatase inhibitors), and centrifuged at 2700 ϫ g for 10 min at 4°C. The nuclear pellet was washed in lysis buffer containing 0.05% Nonidet P-40. The nuclear proteins were extracted by resuspending the pellet in nuclear extraction buffer (20 mM HEPES, pH 7.9, 300 mM NaCl, 1.5 mM MgCl 2 , 0.2 mM EDTA, 0.1 mM ␤-glycerophosphate, 1 mM sodium orthovanadate, 100 M phenylmethanesulfonyl fluoride, 0.1 M okadaic acid, 50 mM sodium fluoride, and 10 g/ml each of leupeptin, aprotinin, and pepstatin) and incubating on ice for 30 min. After extraction, the nuclear samples were centrifuged at 20,800 ϫ g for 15 min at 4°C, and the supernatant was retained as the nuclear extract.
Cloning, Site-directed Mutagenesis, and Transfection-Sitedirected mutagenesis was performed on GSK3␤-HA in pcDNA3.1 ϩ (generously provided by Dr. J. M. Woodgett, University of Toronto) using the GeneEditor site-directed mutagenesis system (Promega, Madison, WI) following the manufacturer's protocol. Primers used for mutagenesis were as follows: for lysines 85 and 86 to alanine, GAGAACTGGTCG-CCATCGCGGCAGTATTGCAGGACAAGAG; for arginine 102 to glycine and lysine 103 to alanine, GAGCTCCAGATC-ATGGGAGCGCTAGATCACTGTAAC; for lysines 122 and 123 to alanine, CTACTCCAGTGGTGAGGCGGCAGATGA-GGTCTATC. The arginine 96 to alanine and serine 9 to alanine mutants were a kind gift from Dr. G. V. W. Johnson (University of Alabama at Birmingham). To generate ⌬9-GSK3␤-HA (deletion of the nine N-terminal amino acids), wild-type GSK3␤-HA was used as a PCR template with the following primers: 5Ј-TCTAGCTAGCATGTTTGCGGAGAGCTGCAAGC and 5Ј-TAAACTATGCGGCCGCTCCGTCGGGACTGTACTAG. The truncated product was inserted into pcDNA3.1 ϩ using NotI and NheI restriction enzymes. DNA encoding the GSK3␤ NLS was synthesized to contain a 5Ј EcoRV compatible end and a 3Ј overhang compatible with XbaI, a TAG stop codon, and a 5Ј phosphorylation, in addition to the coding region for the NLS (sequence available upon request). The eYFP vector was digested with EcoRV and XbaI, and the NLS was ligated in to generate the YFP-NLS fusion. All DNA was synthesized and PAGE-purified by Integrated DNA Technologies (Coralville, IA). The coding regions of all constructs were verified by sequencing at the University of Alabama at Birmingham center for AIDS research DNA sequencing core facility. Cells were transfected 24 h after plating using Fugene 6 (Roche Applied Science) at a ratio of 4 l of Fugene to 1 g of DNA following the manufacturer's protocol. To achieve equal expression levels, all constructs were used at equal amounts except K85A,K86A-GSK3␤-HA, which required twice as much DNA to reach an equal expression level.
Immunoprecipitation-Immunoprecipitations were performed using sheep anti-mouse IgG Dynabeads (Dynal Biotech, Oslo, Norway). The Dynabeads, 10 l of slurry per sample, were washed twice in PBS plus 0.1% BSA; all washes were performed using a magnetic particle separator. The beads were incubated with 1 g of anti-GSK3␤ antibody or 1 l of anti-HA antibody in PBS/BSA at 4°C with end-over-end mixing overnight. The beads were then washed twice in PBS/BSA followed by the addition of cell lysate (200 g of protein). The beads plus lysate were incubated at 4°C for 2 h with end-over-end mixing. The beads were then washed three times with PBS/BSA. GSK3␤ Activity Measurement-GSK3␤ was immunoprecipitated, and the beads were washed a final time in kinase buffer (20 mM Tris, pH 7.5, 5 mM MgCl 2 , 1 mM dithiothreitol). The beads were then resuspended in 30 l of kinase buffer containing 250 M ATP, 1.4 Ci of [␥-32 P]ATP, and 50 M phosphoglycogen synthase peptide (Upstate Biotechnology). The samples were incubated at 30°C for 30 min, the reaction tubes were centrifuged for 1 min, and triplicate 9-l aliquots of each sample were spotted on P81 filter paper circles. The filter paper was allowed to dry and then washed three times in 0.5% phosphoric acid for a total time of 1 h, followed by a 10-min wash in 95% ethanol. The filter paper was then air-dried and counted in a liquid scintillation counter. Activity was normalized to the amount of GSK3␤ immunoprecipitated, which was determined by immunoblotting.
Immunofluorescence-Cells were plated on poly-D-lysinecoated coverslips in 35-mm dishes. After transfection (24 h), cells were fixed in 2% paraformaldehyde at 37°C for 20 min. Cells were washed twice in PBS followed by permeabilization with 0.1% Triton X-100 in PBS for 5 min at room temperature. The cells were then blocked in 3% BSA, PBS for 1 h at room temperature. Cells were stained with Alexa-488-HA antibody (1:2500 in 3% BSA/PBS; Covance) and incubated with rocking at 4°C overnight. The coverslips were washed with PBS and then incubated with 100 ng/ml Hoechst 33342 for 15 min at room temperature. The coverslips were washed for several hours in PBS, with frequent changes of the PBS. The coverslips were mounted on slides and examined by fluorescence microscopy (Nikon) with a ϫ40 oil immersion objective. Images were captured using Image Pro-plus software.
Fluorescence Recovery after Photobleaching (FRAP) and Confocal Microscopy-FRAP analysis was performed in live cells expressing GSK3␤-eYFP using the FRAP module on a Leica SP2 laser-scanning confocal microscope with a Leica DMRXE upright microscope using a ϫ100 oil immersion objective. Prebleach and postbleach images were captured using a 514-nm laser at 20% power with image acquisition at 800 MHz. For bleaching, an area of 13 m 2 was illuminated with the 514-nm laser at 100% power for 6 frames (ϳ0.7 s/frame). The mobile fraction M f was calculated using the equation (15,16), where F ∞ is the fluorescence intensity in the bleached region after recovery, F 0 is the fluorescence in the bleached region immediately after the bleach, and F f is the intensity before the bleach. All values were corrected for loss of fluorescence during acquisition using a region outside of the bleached region. Standard confocal microscopy was performed on live cells using a Leica SP1 laser-scanning confocal with an inverted microscope and ϫ100 oil immersion objective. All confocal analyses were performed at the University of Alabama at Birmingham high resolution imaging core facility.
Flow Cytometry-Annexin-V-PE staining was performed according to the manufacturer's protocol (BD PharMingen). Stained cells were analyzed using a BD LSRII flow cytometer and BD FACSdiva software. Flow cytometry was performed at the University of Alabama at Birmingham center for AIDS research flow cytometry core facility.
Cell Death Assay-Cell death in MEFs was measured as described previously with minor modifications (4). Cells in 60-mm dishes were co-transfected with the indicated construct and LacZ at a ratio of 2:1. After transfection (24 h), cells were trypsinized and counted, and 20,000 cells/well were plated in a 48-well plate. Cells were grown an additional 24 h, followed by treatment with TNF. Following treatment, ␤-galactosidase activity was measured and compared with the untreated control for each construct as a measure of cell viability.
In Situ Cross-linking-Following the indicated treatments, cells were washed twice in 37°C PBS, followed by the addition of the cross-linker, 2 mM ethylene glycolbis(succinimidylsuccinate) (EGS) (Pierce) in PBS, pH 7.4. Cells were incubated at 37°C for 30 min, the reaction was then quenched by the addition of 50 mM Tris, pH 8.0, and samples were incubated at 37°C for 10 min. Cells were washed in PBS and harvested, and subcellular fractions were prepared. Nuclei were lysed using lysis buffer containing 0.5% Nonidet P-40.
Statistical Analysis-Statistical analysis was performed using one-way analysis of variance with Dunnett's post hoc multiple comparisons test or Student's t test using In Stat software.

Identification and Mutagenesis of the Putative NLS in
GSK3␤-GSK3␤ contains a region of 38 amino acids consisting of residues 85-123 that is rich in lysine and arginine (32%), which we hypothesized may encompass an NLS (Fig. 1A). An important feature of an NLS is that it must be externally exposed so that it can bind to nuclear import proteins. The crystal structure of GSK3␤ (17) shows that this putative NLS domain forms an external loop, indicating that it is easily accessible to proteins involved in the intracellular transport of GSK3␤ (Fig. 1B). To begin to test if this region contains an NLS, several of the basic amino acids were mutated to alanine or glycine. These mutated constructs of GSK3␤ containing HA tags included K85A,K86A-GSK3␤-HA, R96A-GSK3␤-HA, R102G,K103A-GSK3␤-HA, and K122A,K123A-GSK3␤-HA. Each GSK3␤ construct included an HA tag because this slows electrophoretic mobility of the protein on SDS-polyacrylamide gels, allowing the expressed GSK3␤ constructs to be distinguished from endogenous GSK3␤ on Western blots. Each GSK3␤ construct was expressed in HEK293 cells, and Western blot analysis demonstrated that expression levels were equivalent to expressed wild-type GSK3␤-HA and that the expression of GSK3␤ constructs did not alter the level of endogenous GSK3␤ (Fig. 1C). Additionally, the phosphorylation status at serine 9 and tyrosine 216 of each GSK3␤ construct was examined using site-specific phosphorylation-dependent antibodies. The serine 9 phosphorylation level of wild-type GSK3␤-HA and each of the mutants paralleled expression levels of each construct with the exception of K85A,K86A-GSK3␤-HA, which exhibited relatively less serine 9 phosphorylation (Fig.  1D). Similarly, the tyrosine 216 phosphorylation of wild-type GSK3␤-HA and each of the mutants was approximately equivalent with the exception of K85A,K86A-GSK3␤, which displayed no detectable phosphorylation at tyrosine 216 (Fig. 1D). This is in accordance with previous evidence that mutation of the Lys 85 residue results in a kinase-dead protein (18) and that phosphorylation of GSK3␤ at tyrosine 216 is due to autophosphorylation (19).
To test if these mutations affected GSK3␤ activity, the kinase activity of each mutant was measured using a phosphoglycogen synthase peptide as a primed substrate. Each construct was expressed in HEK293 cells and immunoprecipitated with an HA antibody. K85A,K86A-GSK3␤-HA had essentially no activity toward phosphoglycogen synthase (Fig. 1E), consistent with evidence that mutation of the Lys 85 residue eliminates kinase activity (18). Similarly, R96A-GSK3␤-HA had little activity toward phosphoglycogen synthase, consistent with reports that Arg 96 is required for GSK3␤ to bind and phosphorylate primed substrates (12). However, R96A-GSK3␤ retained catalytic activity toward the unprimed substrate recombinant Tau (not shown), as reported previously (20). Using phosphoglycogen synthase as the substrate, R102G,K103A-GSK3␤-HA and K122A,K123A-GSK3␤-HA retained ϳ70 and 85%, respectively, of the wild-type GSK3␤ activity.
Amino Acids 85-103 Contain Residues Required for Nuclear Localization of GSK3␤-To test if the nuclear localization of GSK3␤ was dependent on the lysines and arginines within the 85-123 putative NLS region, each construct was expressed in HEK293 cells in parallel with wild-type GSK3␤-HA, and their distributions between the nucleus and cytosol were compared after subcellular fractionation. Cytosolic and nuclear fractions were immunoblotted for CREB as a nuclear marker and ␣-tubulin as a cytosolic marker, and no CREB was detected in the cytosolic fractions and no ␣-tubulin was detected in the nuclear fractions, verifying efficient isolation of these fractions ( Fig. 2A). Wild-type GSK3␤-HA was localized in both the nucleus and the cytosol (Fig. 2B). In marked contrast, K85A,K86A-GSK3␤-HA was completely excluded from the nucleus. Similarly, R96A-GSK3␤-HA and R102G,K103A-GSK3␤-HA also were excluded from the nucleus. However, K122A,K123A-GSK3␤-HA was distributed in both the nucleus and the cytosol to the same extent as wild-type GSK3␤-HA, indicating that Lys 122 and Lys 123 are not required for the nuclear import of GSK3␤. These results indicated that a sequence extending from Lys 85 to Lys 103 , but not to Lys 123 , may constitute an NLS for GSK3␤.
To test if the subcellular localization of GSK3␤ constructs was independent of cell type, immunofluorescent microscopy and immunoblotting were used to examine the localization of each mutant in HeLa cells. Consistent with the results from HEK293 cells, K85A,K86A-GSK3␤-HA, R96A-GSK3␤-HA, and R102G,K103A-GSK3␤-HA were largely excluded from FIGURE 1. Identification and mutagenesis of an NLS in GSK3␤. A, the sequence of GSK3␤ shows that the region labeled as the basic domain (BD), consisting of amino acids 85-123, contains 12 basic amino acids and is located within the kinase domain. B, an accessible external loop is contained in the basic domain based on the crystal structure of GSK3␤. The molecular rendering of a monomer of GSK3␤ was generated using the Cn3D software from the NCBI based on the protein data bank file 1H8F. C, the lysines and arginines (indicated by the asterisks in A) in the basic domain of GSK3␤ were mutated to alanine or glycine. HEK293 cells were transiently transfected with wild-type GSK3␤-HA or mutants of GSK3␤-HA, and 24 h after transfection the cells were harvested, and expression levels of endogenous and expressed GSK3␤ constructs were examined by Western blot. The expression level for each mutant is similar to wild-type GSK3␤-HA. D, the phosphorylation status of endogenous and expressed GSK3␤ was examined by immunoblotting for phospho-Ser 9 -GSK3␤ and phospho-Tyr 279/216 -GSK3␣/␤. E, the activity of each mutant was measured using a phosphoglycogen synthase peptide (P-GS). Cells were transfected with wild-type or mutant GSK3␤-HA, 24 h after transfection HA was immunoprecipitated, and the activity of GSK3␤ was measured. The immunoprecipitated GSK3␤ was measured by immunoblotting, and the activity of each GSK3␤ construct was normalized to the amount of GSK3␤. Values are means Ϯ S.E.; n ϭ 3.
the nucleus, whereas K122A,K123A-GSK3␤ and wild-type GSK3␤-HA were readily detectable in the nucleus (Fig. 2, C-E). These data confirm the conclusion that mutations of Lys 85 and Lys 86 , Arg 96 , or Arg 102 and Lys 103 impair the nuclear localization of GSK3␤.
To verify that these mutations in GSK3␤ impair nuclear import, rather than facilitate nuclear export, the subcellular localization of GSK3␤ was measured in HeLa cells treated with leptomycin B to inhibit CRM-1-dependent nuclear export, which has been reported to cause nuclear accumulation of endogenous GSK3␤ (2). This result was confirmed, since endogenous GSK3␤ accumulated in the nucleus of HeLa cells treated with leptomycin B for 5 h (Fig.  2F). However, little nuclear accumulation of expressed wild-type GSK3␤ or K122A,K123A-GSK3␤-HA occurred following treatment with leptomycin B, suggesting that overexpression may circumvent the CRM-1 export mechanism. Nevertheless, even with leptomycin B treatment, little, if any, K85A,K86A-GSK3␤-HA, R96A-GSK3␤-HA, or R102G,K103A-GSK3␤-HA was detected in the nucleus (Fig. 2G), indicating that these mutations impede nuclear import.
Amino Acids 85-103 Are Sufficient for Nuclear Localization-The previous findings demonstrated that the region of GSK3␤ consisting of amino acids 85-103 may be an NLS. In order to test if this domain meets the criteria for an NLS of being sufficient to induce nuclear localization, a YFP fusion protein was constructed containing residues 85-103 of GSK3␤ on the C terminus (YFP-NLS). Expressed wildtype YFP visualized in live cells was present diffusely throughout the cells (Fig. 3A). In contrast, YFP-NLS preferentially accumulated in the nucleus. Measurements of the nuclear and cytosolic distributions of YFP and YFP-NLS demonstrated that the NLS was sufficient to drive the majority of the YFP-NLS into the nucleus in virtually all cells (Fig.  3B). Taken together with the data shown in Fig. 2, these results confirm that the region of GSK3␤ consisting of amino acids 85-103 is a bona fide bipartite NLS.
GSK3 or its homologue is found in most eukaryotes. To determine if the NLS of human GSK3␤ is conserved, the sequences of several GSK3 homologues were compared. Alignment of the GSK3 sequences from several species demonstrated that the region containing the NLS, and particularly the identified critical basic residues, is highly conserved (Fig. 3C).
The N Terminus of GSK3 Contributes to Regulating Nuclear GSK3␤-The crystal structure and biochemical evidence indicate that the N terminus of GSK3␤ interacts with the primed substrate binding site through an intramolecular interaction (12,17). The primed binding site and the NLS overlap, indicating that the N terminus of GSK3␤ can interact with the NLS. Therefore, we considered the possibility that the N terminus of GSK3␤ might contribute to modulating the function of the NLS. To test this, two mutants of GSK3␤ were used: conversion of serine 9 to alanine (S9A-GSK3␤-HA) to test if serine 9 phosphorylation is regulatory and truncation of the first 9 amino acids (⌬9-GSK3␤-HA) to test if the N-terminal tail contributes to the function of the NLS. Wild-type GSK3␤ and each mutant were expressed in HeLa cells and immunoblots using anti-HA demonstrated equivalent expression levels of all three, and phospho-Ser 9 -GSK3␤ immunoblots verified the absence of serine 9 in the two mutant constructs (Fig. 4A).
HeLa (Fig. 4B) or HEK293 (Fig. 4C) cells expressing wild-type GSK3␤-HA, S9A-GSK3␤-HA, or ⌬9-GSK3␤-HA were separated into cytosolic and nuclear fractions, and these were immunoblotted for GSK3␤-HA. The levels of wild-type GSK3␤-HA and S9A-GSK3␤-HA were equivalent in the nucleus, but the level of ⌬9-GSK3␤-HA in the nucleus was 40% lower than wild-type GSK3␤-HA. Localization in HeLa cells was also examined by immunocytochemistry, and this corroborated the biochemical finding of impaired nuclear localization of ⌬9-GSK3␤-HA (Fig. 4D). This finding indicates that the nine N-terminal amino acids are not absolutely required for nuclear localization, but they contribute to the nuclear localization via a serine 9 phosphorylation-independent interaction with the NLS.
Nuclear GSK3␤ Is Not Required for GSK3␤-mediated Protection from TNF-induced Apoptosis-An antiapoptotic role for GSK3␤ in the context of TNF-mediated toxicity was first identified in a study of GSK3␤ knock-out mice (4) and subsequently was confirmed in a number of reports and expanded to include apoptosis induced by other members of the death receptor family that induce extrinsic apoptosis (21)(22)(23)(24)(25). Despite these extensive studies, the important question of whether this action of GSK3␤ occurs in the nucleus or the cytosol remains unresolved. This is because nuclear GSK3␤ promoted the cell-protective transcriptional activity of NF-B, but other data implicate a cytosolic protective effect of GSK3␤, because it inhibits activation of the initiator caspase, caspase-8, which occurs directly after receptor activation in the cytosol (reviewed in Ref. 26).
In order to clarify whether cytosolic GSK3␤ or nuclear GSK3␤ promotes survival following exposure to TNF, we used wild-type GSK3␤ and an active, but NLS-deficient, mutant of GSK3␤ to test if cytosolic GSK3␤ was capable of protecting GSK3␤-null MEFs (GSK3␤ Ϫ/Ϫ ) from TNF-mediated cell death. Hoeflich et al. (4) designed an elegant strategy to identify the antiapoptotic effect of GSK3␤, demonstrating that TNFinduced cytotoxicity was enhanced in GSK3␤ Ϫ/Ϫ -MEFs from GSK3␤ knock-out mice and that this was reversed by expression of wild-type GSK3␤. Additionally, they demonstrated that endogenous GSK3␣ was unable to compensate for the loss of GSK3␤ in TNF-induced apoptosis in GSK3␤ Ϫ/Ϫ -MEFs (4). Consistent with their findings, treating GSK3␤ Ϫ/Ϫ -MEFs with TNF for 16 h resulted in nearly half of the cells undergoing apoptosis, as indicated by positive staining for annexin-V, whereas no increase in apoptosis was detected in wild-type GSK3␤ ϩ/ϩ -MEFs treated with TNF (Fig. 5A). Examination of the subcellular distribution of endogenous GSK3␣ and GSK3␤ demonstrated that their nuclear and cytosolic levels did not change after TNF treatment in wild-type MEFs (Fig. 5B). In order to clarify the site of action of the protective effect of GSK3␤ in TNF-induced apoptosis, we compared the protective capacities of wild-type GSK3␤-HA and R102G,K103A-GSK3␤-HA expressed in GSK3␤ Ϫ/Ϫ -MEFs treated with TNF. First, we confirmed the nuclear and cytosolic localization of wild-type GSK3␤-HA and cytosolic localization of R102G,K103A-GSK3␤-HA expressed in GSK3␤ Ϫ/Ϫ -MEFs (Fig. 5C). Furthermore, TNF treatment did not alter the subcellular distribution of the expressed GSK3␤ constructs (Fig. 5D). Transfection with vector alone followed by treatment with TNF (5 ng/ml; 16 h) resulted in 44 Ϯ 4% cell death of GSK3␤ Ϫ/Ϫ -MEFs (Fig. 5E). As expected, GSK3␤ Ϫ/Ϫ -MEFs transfected with wild-type GSK3␤-HA and treated with TNF showed a reduction to only 25 Ϯ 4% cell death. Similarly, GSK3␤ Ϫ/Ϫ -MEFs expressing R102G,K103A-GSK3␤-HA and treated with TNF also displayed reduced cell death of only 19 Ϯ 3%. As an additional control, GSK3␤ Ϫ/Ϫ -MEFs expressing the kinasedead K85A,K86A-GSK3␤-HA mutant were treated with TNF and, similar to cells transfected with vector alone, this resulted in 51 Ϯ 3% cell death, demonstrating that cytosolic GSK3␤ C, sequence alignment of the GSK3 NLS from human, mouse, Caenorhabditis elegans, and Schizosaccharomyces. Basic residues are shown in boldface type, and residues identified as required for nuclear import are marked with boxes.
must be active to provide protection from TNF-induced toxicity. These findings demonstrate that GSK3␤ attenuates TNFinduced toxicity through a cytoplasmic target.

Mechanisms Controlling Nuclear Localization of GSK3␤-
The predominantly cytosolic distribution of wild-type GSK3␤ was distinctly different from the predominant nuclear accumulation of YFP-NLS, although both proteins contained the same NLS. This indicated that additional mechanisms must contribute to regulating the nuclear localization of GSK3␤. Therefore, we assessed several potential mechanisms that might control the function of the NLS in GSK3␤.
Many proteins, including kinases, such as mitogen-activated protein kinases, are known to enter the nucleus by active transport as dimers (27), and the crystal structure of GSK3␤ raised the possibility that it also may exist as a dimer (17). Therefore, homodimerization of GSK3␤ or its heterodimerization with GSK3␣ may represent an additional level of regulation for the subcellular localization of GSK3. However, GSK3 has not been shown to exist as a dimer in situ. To test if wild-type GSK3 forms dimers, GSK3␤-YFP was expressed and immunoprecipitated with an antibody recognizing YFP, followed by immunoblotting for GSK3␤ to test if endogenous GSK3␤ was associated with the expressed GSK3␤-YFP (Fig. 6A). However, no endogenous GSK3␤ was detected in the YFP-immunoprecipitants. To test for the existence of heterodimers of the two isoforms of GSK3, GSK3␣ was immunoprecipitated, followed by immunoblotting for GSK3␤ and vice versa (Fig. 6B). No detectable co-immunoprecipitation occurred between GSK3␣ and GSK3␤. Furthermore, expression of K85A,K86A-GSK3␤-HA did not alter the nuclear level of endogenous GSK3␤, although K85A,K86A-GSK3␤-HA was completely absent from the nucleus, indicating that K85A,K86A-GSK3␤-HA did not act as a dominant negative regulator of the nuclear import of endogenous GSK3␤ to prevent its nuclear localization (Fig. 6C). Taken together, these findings suggest that dimerization of GSK3␤ is unlikely to regulate its distribution to the nucleus.
We next considered the possibility that factors external to GSK3␤ itself may be important for regulating the nuclear localization of GSK3␤. Specifically, since GSK3␤ is known to bind many cytosolic proteins (7), we hypothesized that these may anchor GSK3␤ in the cytoplasm. If GSK3␤ is sequestered in the cytosol by protein interactions, it should be possible to saturate the interactions by overexpression of GSK3␤-HA, which would lead to an increase in the percentage of GSK3␤-HA in the nucleus. Consistent with this, overexpression of GSK3␤-HA resulted in a 50% greater nuclear/cytosol distribution of GSK3␤-HA compared with endogenous GSK3␤ in cells without overexpressed GSK3␤-HA (Fig. 7A). This supports the hypothesis that cytoplasmic anchoring proteins were saturated by overexpressed GSK3␤-HA to allow greater import into the nucleus.
ing strategy was used to verify that cytosolic binding partners are saturated by overexpression of GSK3␤ and to estimate the proportion of cytosolic GSK3␤ that was sequestered in protein complexes. Treatment of cells with the membrane-permeable cross-linker EGS to stabilize GSK3␤ bound in protein complexes showed that very little of the endogenous GSK3␤ in the cytosol was unbound (not cross-linked to other proteins by EGS). This is indicated by the greatly reduced amount of endog- enous cytosolic GSK3␤ that migrated at the native molecular weight in SDS-polyacrylamide gels, indicating that it exits predominantly in protein complexes (Fig. 7B, first two lanes). In contrast, in cells overexpressing wild-type GSK3␤-HA, R96A-GSK3␤-HA, or R102G,K103A-GSK3␤-HA, cross-linking with EGS caused only a minor decrease of the expressed constructs in the cytosol, indicating that they are mostly unbound (not cross-linked to other proteins), consistent with saturation of binding partners (Fig. 7B). Interestingly, K85A,K86A-GSK3␤-HA was highly cross-linked following treatment with EGS. The reason for this anomalous behavior is not known, but it may be because this kinase-dead mutant is associated with substrates of GSK3 that it is unable to phosphorylate and release, thereby causing the high level of protein interactions stabilized by the cross-linker. The same cross-linking strategy was used to test if endogenous GSK3␤ was released from cytosolic protein complexes by conditions that increase nuclear GSK3. As reported previously (2), serum withdrawal or staurosporine treatment increased the nuclear level of endogenous GSK3 (Fig. 7C). In parallel, cells were treated, followed by incubation with or without EGS to stabilize endogenous GSK3 bound in protein complexes. The cytosolic and nuclear proteins were isolated and immunoblotted for the remaining unbound GSK3␤ or GSK3␣. Fig. 7D shows that only ϳ15% of endogenous GSK3 in the cytosol was unbound (not crosslinked to other proteins by EGS) in control cells maintained in serum-containing medium compared with total cytosolic GSK3 in the uncross-linked control, indicating that there is little unbound GSK3 in the cytosol. However, there were dramatic increases in unbound cytosolic GSK3 induced by treatments that cause nuclear accumulation of GSK3, since the amount of free cytosolic GSK3␤ was increased almost 3-fold following serum withdrawal and nearly 5-fold following staurosporine treatment compared with cells maintained in serum (Fig. 7, D and E). Thus, most of the cytosolic endogenous GSK3␤ is present in protein complexes, and the release of cytosolic GSK3 from protein complexes correlated with increased nuclear import of GSK3. Compared with the cytosol, the amount of free GSK3 in the nucleus was relatively high, consistent with the notion that free GSK3 may preferentially localize to the nucleus (Fig. 7, D and E). In contrast to the cytosolic pool of GSK3, the level of free uncross-linked nuclear GSK3 mirrored the increase in total nuclear GSK3 caused by serum withdrawal or staurosporine treatment, indicating that the percentage of free GSK3 in the nucleus remains relatively constant, independent of its absolute nuclear concentration (Fig. 7, C and D). This finding supports the hypothesis that GSK3 is predominantly proteinbound in the cytosol and is released in a stimulant-dependent manner, and release is related to increases in nuclear GSK3. Furthermore, this mechanism provides a rationale as to why most of the GSK3 is cytoplasmic despite the presence of an NLS. Having found that GSK3 is largely protein-bound in the cytosol raised the possibility that it may be in immobile structures. To test if cytosolic GSK3␤ is mobile, FRAP analysis was used. GSK3␤ Ϫ/Ϫ MEFs were used to reduce the possibility of saturating the GSK3␤-binding proteins when exogenous GSK3␤ was expressed. GSK3␤ Ϫ/Ϫ MEFs were transfected with GSK3␤-YFP, and FRAP analysis was performed in live cells (Fig. 7F). From this analysis, it was determined that nearly all of the cytosolic GSK3␤, 84 Ϯ 3%, was freely mobile. Overall, these data demonstrate that cytosolic GSK3␤ is predominantly bound in freely mobile protein complexes, which upon stimulation can disassociate and allow nuclear localization of GSK3␤.

DISCUSSION
Identifying the central mechanism controlling the nuclear localization of GSK3␤ has long been a crucial objective needed to clarify the regulatory effects of GSK3␤ on apoptosis, gene expression, and other nuclear events. We report here the previously unexpected existence of a bipartite NLS in GSK3␤ consisting of residues 85-103 that was identified by assessing the subcellular localization of GSK3␤ mutants created by site-directed mutagenesis. The data also provide insight into mechanisms controlling the function of the NLS, including maintenance of a predominantly cytosolic localization of GSK3␤ while allowing nuclear accumulation in response to demand.
The NLS of GSK3␤ was found to be both necessary and sufficient for nuclear localization. These properties were shown by the cytosolic localization of GSK3␤ containing mutations of key residues in the NLS and by the nuclear localization of YFP fused with the NLS of GSK3␤. Furthermore, the key residues in the GSK3␤ NLS are highly conserved in GSK3 homologues throughout the eukaryotic kingdom. The identified NLS in GSK3␤ is similar to the canonical bipartite NLS in which the amino acids are organized such that there are 2 basic amino acids plus any 10 amino acids plus 3 basic of 5 amino acids. The prototypical bipartite NLS from nucleoplasmin has the sequence KRPAATKKAGQAKKKK, where both the N-terminal and C-terminal basic clusters (underlined) are required for nuclear localization (28). The NLS from GSK3␤ has the sequence KKVLQDKRFKNRELQIMRK, which is similar in overall organization except that the C-terminal basic cluster is expanded to 8 amino acids. Thus, the NLS from GSK3␤ has an FIGURE 6. GSK3 dimers are not detectable and do not contribute to nuclear localization. A, GSK3␤-YFP was expressed and then immunoprecipitated (IP), and the immunoprecipitate was immunoblotted (IB) for GSK3␤. No endogenous GSK3␤ was co-immunoprecipitated with GSK3␤-YFP. B, GSK3␣ was immunoprecipitated from HEK293 cells and immunoblotted for GSK3␤, and the reciprocal immunoprecipitation is also shown. No coimmunoprecipitation of the two GSK3 isoforms was detected. Input, 5 g of protein of cell lysate; negative (Neg) control is isotype-matched nonimmune IgG. C, nuclear extract from HEK293 cells expressing wild-type GSK3␤-HA or K85A,K86A-GSK3␤ was immunoblotted for GSK3␤. Shown is a long exposure to demonstrate that the nuclear localization of endogenous GSK3␤ was not affected by expression of an NLS-deficient GSK3␤ mutant.
organization of 2 basic amino acids plus 9 amino acids plus 3 basic of 8 amino acids. Notably, this NLS of GSK3␤ overlaps with residues required for its catalytic activity, as well as for substrate binding. Kinase-dead mutations of Lys 85 in GSK3␤ are often used as a dominant negative construct of GSK3␤, but this mutation probably impedes entry to the nucleus, suggesting that its use is restricted to studying functional effects in the cytosol. The mutant R96A-GSK3␤-HA blocks binding to primed substrates and is commonly used to identify substrates of GSK3␤ that must be primed for subsequent phosphorylation by GSK3␤. The nuclear exclusion of R96A-GSK3␤-HA restricts to the cytosol its use to distinguish primed from unprimed substrates of GSK3␤. Thus, the identification of the NLS in GSK3␤ raises important limitations in the interpretation and design of studies of the functional effects of expressed mutants of GSK3␤.
Although GSK3␤ contains an NLS sufficient to drive YFP into the nucleus, cells maintain only a small portion of GSK3␤ within the nucleus. Thus, GSK3␤ is predominantly located in the cytosol without evidence of an NES, indicating that normally the NLS function is largely masked. Since crystallographic studies had suggested that GSK3␤ may exist as a dimer (17,29), although this had not been tested in situ, we hypothesized that dimerization might play a role in the nuclear localization of GSK3␤. However, we demonstrated that dimerization of GSK3␤ was unlikely to regulate nuclear localization, because expression of NLS-deficient mutants did not impede the nuclear localization of endogenous GSK3␤, and dimers were not detected in co-immunoprecipitation experiments.
Although dimerization of GSK3␤ was ruled out as a regulatory factor, our data and the data of others indicate that three mechanisms conspire to regulate the NLS: cytosolic protein complexes, the N-terminal tail of GSK3␤, and the phosphorylation state of tyrosine 216. Cross-linking analysis demonstrated that cytosolic GSK3␤ was predominantly bound in protein complexes, and we speculate that these sequester the majority of GSK3␤ in the cytosol. However, these protein complexes did not immobilize GSK3␤ in the cytosol, since FRAP analysis demonstrated that ϳ85% of cytosolic GSK3␤ was mobile. Perhaps most importantly for nuclear transport, a significant amount of GSK3␤ was freed from cytosolic protein complexes under conditions that promote nuclear accumulation of GSK3␤, suggesting that sequestration in cytosolic protein com-plexes masks the NLS to limit the capability of GSK3␤ to traverse the nuclear membrane. This conclusion is further supported by the recent demonstration that expression of the GSK3-binding protein Axin2 sequestered GSK3␤ in the cytosol, thereby blocking its access to its nuclear substrate Snail (30). Thus, despite the NLS, GSK3␤ is maintained localized predominantly in the cytosol by GSK3␤-binding proteins.
In addition to the NLS, the N-terminal tail of GSK3␤ was found to be capable of modulating nuclear localization of GSK3␤, because truncation of the first 9 amino acids, to construct ⌬9-GSK3␤-HA, reduced its level in the nucleus. This result is in contrast to the previous finding that ⌬9-GSK3␤-HA is predominantly nuclear in prostate cancer cell lines (31,32). This discrepancy may be due to cell type-specific effects, but we consistently found that ⌬9-GSK3␤-HA was not predominantly nuclear compared with wild-type GSK3␤-HA, as did another recent report (30). It is also possible that expression of a high level of unregulated, active ⌬9-GSK3␤-HA could cause toxicity, since overexpressed GSK3␤ can cause apoptosis with cell shrinkage and nuclear condensation (33), which may contribute to an apparent nuclear accumulation of ⌬9-GSK3␤-HA in some cell types. We speculate that reduced nuclear localization caused by deletion of the N-terminal tail indicates that the N-terminal tail interacts with the NLS and that the basic amino acids in the N-terminal tail may contribute to the affinity of GSK3␤ for import molecules. It is already well established that when serine 9 of GSK3␤ is phosphorylated, the N-terminal tail of GSK3␤ folds over the region encompassing part of the NLS with proline 5 occupying the phospho-accepting site (12,17). Furthermore, analysis of peptides corresponding to the sequence of the N-terminal amino acids of GSK3␤ demonstrated that binding of the peptides to GSK3␤ was not simply mediated by a phosphate group but was dependent on the sequence context of the phosphoserine, indicating that other residues within this sequence contribute to the intramolecular interaction between the N-terminal tail and its binding site in GSK3␤ (12,17). Moreover, merely dephosphorylating serine 9 of GSK3␤ is not sufficient to influence nuclear localization, because the distribution of S9A-GSK3␤ is predominantly cytosolic, similar to wild-type GSK3␤. Therefore, we suggest that the N-terminal tail interacts with the NLS of GSK3␤ via an intramolecular interaction to promote nuclear transport, so its elimination reduces the effectiveness of the NLS.  (n ϭ 3); *, p Ͻ 0.05 compared with endogenous GSK3␤. B, HeLa cells were either untransfected or transfected with the indicated GSK3␤ plasmid and treated with the membrane-permeable cross-linker EGS, where indicated, prior to lysis of the cells. Cytosolic fractions were prepared followed by immunoblotting for GSK3␤ to determine the amount of free, unbound GSK3␤. A decrease of the unbound, free GSK3 in the EGS-treated samples indicates an increase in the amount that is cross-linked (i.e. bound) in protein complexes. The fraction of GSK3␤ that is free is indicated under the immunoblot and was determined by dividing the densitometry value of the EGS-treated sample by the value of the untreated sample in each pair. C, HeLa cells were left in serum (S), placed in serum-free medium for 2 h (SF), or treated with staurosporine (1 M) for 1 h (STS), followed by preparation of cytosolic and nuclear fractions, which were immunoblotted for GSK3␣/␤. D, HeLa cells were left in serum, placed in serum-free media for 2 h, or treated with staurosporine (1 M) for 1 h, followed by the addition of the membrane-permeable cross-linker EGS where indicated, prior to lysis of the cells. Cells were fractionated into nuclear and cytosolic fractions, and the level of unbound (noncross-linked) GSK3 was examined. Lane 1 shows an equal amount of noncross-linked cytosolic lysate, representing total GSK3 in the cytosol, and lane 5 is an equal amount of noncross-linked nuclear lysate, representing total GSK3 in the nucleus. E, quantitation of the increases in unbound, free cytosolic GSK3␤, and lack of changes of unbound nuclear GSK3␤, caused by treatments with serum-free medium or staurosporine treatment. The results shown in E were normalized to the amount of GSK3␤ in each subcellular fraction with each condition shown in C. Values are expressed as a percentage of the noncross-linked controls, shown in lane 1 or lane 5 in D. F, cytosolic GSK3␤ is mobile as measured by FRAP. GSK3␤ Ϫ/Ϫ MEFs were transfected with GSK3␤-YFP 48 h before FRAP analysis. A region in the cytoplasm was photobleached with the 488-nm laser and allowed to recover for ϳ90 s. The graph represents the intensity in the region prior to the photobleaching, immediately following the photobleach, and the recovery. Each data point is the mean from 15 cells in two independent experiments. Values are means Ϯ S.E.; scale bar, 25 m. WT, wild type.
Recent findings suggest that phosphorylation of tyrosine 216 of GSK3␤ may contribute to regulating the nuclear distribution of GSK3␤. Upon mutation of tyrosine 216 to phenylalanine, which cannot be phosphorylated, less GSK3␤ is found in the nucleus (30,31), whereas there is evidence that GSK3␤ phosphorylated on tyrosine 216 is the preferred conformation for nuclear accumulation (34). The crystal structure of GSK3␤ predicts that when tyrosine 216 is not phosphorylated, the activation loop of GSK3␤ containing tyrosine 216 obstructs the primed substrate binding site in GSK3␤ (17). Our data show that this binding site is part of the NLS, and therefore the NLS is probably masked by the activation loop containing unphosphorylated tyrosine 216. However, whether or not this mechanism comes into play under physiological conditions is complicated by the controversy about how GSK3␤ is tyrosine-phosphorylated. Some evidence indicates that GSK3␤ is nearly completely phosphorylated on tyrosine 216 (19), whereas other evidence shows changes in tyrosine 216 phosphorylation under a variety of experimental conditions, especially notably within the nucleus (34).
GSK3␤ is well established to be an antiapoptotic regulator of death receptor-induced extrinsic apoptosis (26). This was first identified with TNF-induced apoptosis (4,21) and subsequently was expanded to include all other apoptosis-inducing members of the death receptor family (22)(23)(24). Despite this important regulatory effect, the site of action of GSK3␤ responsible for attenuating death receptor-induced apoptotic signaling was not clear. Several studies reported that GSK3␤ inhibits apoptotic signaling proximal to activation of the initiating caspase, caspase-8, at the death receptor (21,22,35), indicating a cytosolic action of GSK3␤. However, much evidence has shown that GSK3␤ in the nucleus promotes the transcriptional activity of NF-B (4,25), which is antiapoptotic in this context, providing circumstantial support for the idea that this action of GSK3␤ in the nucleus mediates its inhibitory effect on death receptor-induced apoptosis. We were able to address this issue by utilizing a nucleus-excluded, but active, GSK3␤ construct (R102G,K103A-GSK3␤-HA) with GSK3␤ Ϫ/Ϫ MEFs, which can be rescued from TNF-induced cell death by expression of GSK3␤ (4). These experiments demonstrated that GSK3␤ restricted to the cytosol was as efficient as wild-type GSK3␤ in rescuing GSK3␤ Ϫ/Ϫ MEFs from TNF-induced cytotoxicity. These results do not negate the clearly established regulatory effect of GSK3␤ on NF-B, but they show that the protective action of GSK3␤ is cytosolic, in agreement with findings that GSK3␤ inhibits an early step in the death receptor-induced apoptotic process (21,22,35).
In summary, the identification of an NLS in GSK3␤, in conjunction with previous reports, provides considerable insight into the mechanisms controlling the dynamic flux of nuclear GSK3␤. However, the identification of an NLS in GSK3␤, but no NES, raised the question of why GSK3␤ remains predominantly cytosolic despite the NLS. Under basal conditions, GSK3␤ is mostly cytosolic where it is anchored to cytoplasmic binding proteins, since GSK3␤ is well known to be associated in multiprotein complexes that serve as a mechanism to direct the actions of GSK3␤ toward specific substrates, an important task, since over 50 substrates of GSK3␤ have been identified (7). We speculate that these GSK3␤-binding proteins limit the pool of GSK3␤ that is free to translocate into the nucleus. Following stimulation, GSK3␤ is released from the cytoplasmic anchors, and GSK3␤ phosphorylated on tyrosine 216 is then available to be imported into the nucleus via the NLS in conjunction with support by the N-terminal tail. Inside the nucleus, GSK3␤ is highly active (3) and phosphorylates numerous substrates (7). Within the nucleus, GSK3␤ can bind FRAT and be exported from the nucleus in a CRM1-dependent fashion (2,13). Taken together, these mechanisms may conspire to largely block the NLS of GSK3␤ to maintain it predominantly in the cytosol but allow nuclear localization in the appropriate cellular conditions. Keeping in mind that the NLS encompasses residues required for kinase activity and substrate binding, these interactions provide an elegant mechanism in which GSK3␤ activity and nuclear localization can be tightly coupled and regulated.