Differential Regulation of Glycogen Synthase Kinase 3β by Insulin and Wnt Signaling*

Glycogen synthase kinase 3β (GSK3β) is a key component in many biological processes including insulin and Wnt signaling. Since the activation of each signaling pathway results in a decrease in GSK3β activity, we examined the specificity of their downstream effects in the same cell type. Insulin induces an increased activity of glycogen synthase but has no influence on the protein level of β-catenin. In contrast, Wnt increases the cytosolic pool of β-catenin but not glycogen synthase activity. We found that, unlike insulin, neither the phosphorylation status of the serine9 residue of GSK3β nor the activity of protein kinase B is regulated by Wnt. Although the decrease in GSK3β activity is required, GSK3β may not be the limiting component for Wnt signaling in the cells that we examined. Our results suggest that the axin-conductin complexed GSK3β may be dedicated to Wnt rather than insulin signaling. Insulin and Wnt pathways regulate GSK3β through different mechanisms, and therefore lead to distinct downstream events.

Glycogen synthase kinase 3␤ (GSK3␤) is a key component in many biological processes including insulin and Wnt signaling. Since the activation of each signaling pathway results in a decrease in GSK3␤ activity, we examined the specificity of their downstream effects in the same cell type. Insulin induces an increased activity of glycogen synthase but has no influence on the protein level of ␤-catenin. In contrast, Wnt increases the cytosolic pool of ␤-catenin but not glycogen synthase activity. We found that, unlike insulin, neither the phosphorylation status of the serine9 residue of GSK3␤ nor the activity of protein kinase B is regulated by Wnt. Although the decrease in GSK3␤ activity is required, GSK3␤ may not be the limiting component for Wnt signaling in the cells that we examined. Our results suggest that the axin-conductin complexed GSK3␤ may be dedicated to Wnt rather than insulin signaling. Insulin and Wnt pathways regulate GSK3␤ through different mechanisms, and therefore lead to distinct downstream events.
Glycogen synthase kinase 3 (GSK3) 1 was originally identified for its ability to phosphorylate and inhibit glycogen synthase (GS) (1,2). It is a serine/threonine kinase that recognizes the target sequence SXXXS with the second serine prephosphorylated (3). Many proteins other than GS also contain GSK3 recognition sequences, some of which can be phosphorylated by GSK3 in vitro. These include ATP-citrate lyase, protein phosphatase 1, cAMP-dependent protein kinase, eIF2B, inhibitor-2, c-Jun, Myc, Myb, CREB, Tau, ␤-catenin, and IB (4 -6). GSK3 is also unusual in that its enzymatic activity remains high at resting state and decreases upon stimulation. GSK3 is conserved from yeast to mammals and has been implicated in strikingly versatile biological functions. However, how different signals regulate GSK3 is still unknown.
GSK3 plays an important role in the cellular response to insulin (7). The regulation of GSK3 by insulin has been shown to be mediated by protein kinase B (PKB). Upon insulin stim-ulation, threonine 308 (Thr-308) and serine 473 (Ser-473) residues of PKB are phosphorylated and PKB is activated (8). Subsequently, both GSK3 isotypes (GSK3␣ and GSK3␤) in mammalian cells are phosphorylated on a serine residue at the N terminus (serine 21 of GSK3␣ and serine 9 of GSK3␤) (9,10), which leads to a decrease in GSK3 activity. Although this has usually been detected as a 50 -70% drop, it is apparently sufficient to relieve the inhibition of GS and allow cells to complete glycogen synthesis.
Another in vitro substrate of GSK3␤ is ␤-catenin, a protein involved in cell adhesion, oncogenesis and development (11)(12)(13). Together with axin-conductin and APC, GSK3␤ is one of the components of a protein complex that regulates the stability of ␤-catenin (14 -17). Phosphorylation of the GSK3␤ sites in the N terminus of ␤-catenin is believed to be a signal for degradation. When either APC or the GSK3␤ sites of ␤-catenin are mutated, as in 90% of colon cancer, levels of ␤-catenin are elevated (13). Excess ␤-catenin accumulates in the cytosol and nucleus, outside of cell adhesion complexes on cytoplasmic membrane where it normally resides. Nuclear ␤-catenin is capable of interacting with the LEF/TCF family DNA-binding proteins and activating transcription of genes containing LEF/ TCF binding sites (18,19). Increased ␤-catenin levels can also be achieved through the activation of Wnt/wingless signaling pathway (20). GSK3␤ has been placed between Dishevelled (Dvl in mammalian cells, Dsh in other organisms) and ␤-catenin in the Wnt pathway based on a combination of genetic and biochemical evidence (21)(22)(23). It is not clear how the extracellular Wnt signal is transduced from the membrane receptor Frizzled to Dsh/Dvl, and then to GSK3␤ resulting in increased ␤-catenin levels. Decreases in the activity of GSK3␤ have been observed in mouse fibroblasts and Drosophila cells responding to wingless and Dsh (24,25). Furthermore, inhibition of GSK3␤ activity by lithium salt or GSK3␤-binding protein (GBP/FRAT) mimics Wnt signaling (26,27). Recently, it is reported that Dvl and GBP/FRAT are able to associate with the axin-conductin-APC-␤-catenin complex (23,28). Moreover, this entire complex is believed to dissociate in response to Wnt signaling (20,25,28).
In this study, we investigated how different signals such as insulin and Wnt regulate GSK3␤. Using mammalian cells that respond to both signals, we found that the downstream effect is specific to each pathway, despite the indistinguishable decrease in GSK3␤ activity. We also generated the first inducible system to conditionally activate Dishevelled in mammalian cells as an independent method to turn on the Wnt signaling pathway. Serine 9 of GSK3␤ is not regulated in cells that are activated by Wnt or Dishevelled. Furthermore, we have evidence that the axin-conductin complexed GSK3␤ is not significantly phosphorylated at serine 9 upon insulin stimulation and, therefore, may be protected from insulin signaling.

EXPERIMENTAL PROCEDURES
Cell Culture and Transfections-Human embryonic kidney 293 cells were purchased from ATCC. C57MG, Rat2-MV7, and Rat2-Wnt1 cell lines were generous gifts from Dr. Anthony Brown. These cell lines were maintained in Dulbecco's modified Eagle's medium (DMEM, Life Technologies, Inc.). CHOIR cells were kindly given by Drs. Richard Roth (29) and Ira Goldfine, maintained in Ham's F12 medium (Life Technologies, Inc.). All cell culture media were supplemented with 10% fetal calf serum (Life Technologies, Inc.) and 1% penicillin/streptomycin (Life Technologies, Inc.).
For conditioned media, Rat2-MV7 and Rat2-Wnt1 were grown to 95% confluence. Cells were washed with phosphate-buffered saline and maintained in serum-free DMEM overnight. The conditioned media were filtered through 0.22-m filter units, aliquoted, and stored at Ϫ80°C until use.
For insulin or Wnt stimulation, cells were grown to 70 -80% confluence and serum-starved overnight, after which 5 g/ml insulin or 0.2 ml/cm 2 conditioned medium was added.
Transfections of plasmids were performed by using LipofectAMINE Plus (Life Technologies, Inc.) or FuGENE6 (Roche Molecular Biochemicals) according to instructions from the manufacturer.
To generate 293-D-ER cells, 293 cells were transfected with the Dvl-ER plasmid. Selection with 1 mg/ml Geneticin (Life Technologies, Inc.) started 24 h after transfection. Resistant cells were pooled together after two rounds of complete killing of the parental 293 cells. Once these cells were verified to have stable expression and inducible Dvl-ER, they were transfected with the His 6 -tagged conductin plasmid. 25 g/ml blasticidin (Invitrogen) was used to select for 293-D-ER-His 6conductin cells.
Plasmids-Mammalian expression plasmid encoding human Dishevelled 2 was a generous gift from Dr. Misha Semenov (30). The coding region of Dishevelled 2 was also epitope-tagged with Glu-Glu (EE) tag and fused to the hormone binding domain of a modified version of the murine estrogen receptor (Dvl-ER). Conductin expression plasmid was from Dr. Walter Birchmeier (16), and was then epitope-tagged with the EE tag in pCDNA3 or His 6 tag in pCDNA6. TOPTK reporter plasmid for TCF/LEF-dependent transcription was from Dr. Hans Clevers (31). Dominant negative form of TCF4 (DNTCF4) expression plasmid was kindly provided by Dr. Osamu Tetsu (32). cDNA encoding human GSK3␤ in pBlueScriptSKϩ was from Dr. James Woodgett (33). GSK3␤ was amplified from this template by polymerase chain reaction and cloned into a pCDNA3-based plasmid with an N-terminal HA tag. Mutants of GSK3 were created by using QuickChange site-directed mutagenesis kit (Stratagene). Three versions of kinase-dead GSK3␤ were made with amino acid substitutions at the ATP binding site: lysine 85 to alanine, lysine 85 and 86 to arginines, and lysine 85 to methionine plus lysine 86 to alanine.
Antibodies-Anti-GSK3␤ and anti-␤-catenin antibodies were from Transduction Laboratory. Phosphotyrosine antibody 4G10 and anti-PKB antibody were from Upstate Biotechnology. Phosphospecific antibody against Ser-9 of GSK3␤ was from Quality Controlled Biochemicals. Phosphospecific antibodies against PKB were generously provided by Dr. David Stokoe. Anti-EE antibody was from Harlan Bioproducts. Anti-HA antibody was from Santa Cruz Biotechnology.
Cytosolic Fractionation, Immunoprecipitations, and Western Blots-To prepare cytosolic fractions, cells were washed and collected in ice-cold phosphate-buffered saline. Cell pellets were resuspended in ice-cold hypotonic buffer (25 mM Tris, pH 7.5, 1 mM EDTA, 25 mM NaF, 1 mM dithiothreitol) with Complete protease inhibitor mixture (Roche Molecular Biochemicals). Cells were lysed after incubating on ice for 10 min (verified by microscope). The lysates were subjected to ultracentrifugation at 100,000 ϫ g for 30 min at 4°C, and the supernatant was collected.
For immunoprecipitation, cells were washed twice in ice-cold phosphate-buffered saline, then lysed in IP buffer (125 mM NaCl, 25 mM NaF, 25 mM Tris, pH 7.5, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 10 mM ␤-glycerol phosphate, 5 mM sodium pyrophosphate, 1 mM NaVO 3 , 200 nM okadaic acid, 1 mM dithiothreitol) with Complete protease inhibitor mixture. Anti-GSK3␤, anti-HA, or anti-EE antibody was added to clarified lysates for 1 h at 4°C, and then Protein G beads (Sigma) were added for another 1 h. Immunoprecipitates were washed three times with IP buffer. To coimmunopricipitate GSK3␤ with His 6conductin, Ni-IP buffer was used. Ni-IP buffer was IP buffer without EDTA, EGTA, or dithiothreitol and supplemented with EDTA-free Complete protease inhibitor mixture (Roche Molecular Biochemicals). Nickel beads (ProBond resin, Invitrogen) were first blocked with 2 mg/ml bovine serum albumin in Ni-IP buffer for 2 h. After incubating with cleared lysates, nickel beads were washed three times with Ni-IP buffer supplemented with 200 mM immidazole and then once with Ni-IP buffer. Western blotting was carried out following standard procedures. 10% Tris-glycine polyacrylamide gels were used.
Enzyme Assays-For GSK3␤ kinase assays, GSK3␤ immunoprecipitates were washed once with kinase buffer (25 mM Tris, pH 7.5, 10 mM MgCl 2 ) first. Kinase reactions were performed in kinase buffer with 100 M [␥-32 P]ATP and 100 M 2BSP peptide as the substrate (synthesized by the Biomedical Resource Center, University of California, San Francisco, CA). 2BSP is based on the GSK3 target site in eIF2B (34). After 20 min at 30°C, the reactions were spotted on phosphocellulose P81 paper (Whatman), washed four times with 100 mM phosphoric acid, and counted in scintillation counter.
Luciferase assays were performed by using dual luciferase reporter assay system (Promega) in a Microplate Luminometer (EG&G Berthold). Transfection efficiency was normalized to the expression of Renilla luciferase from the cotransfected pRL-TK plasmid.
GS assays were performed as described previously (35). Parallel assays were performed using low (0.1 mM) and high (10 mM) concentrations of glucose 6-phosphate to give active and total activities of GS. GS activity was calculated as the fraction of active from total activity.
Phosphopeptide Mapping-Cells in 10-cm dishes were grown to 70 -80% confluence, serum-starved for 8 h in regular DMEM, then metabolically labeled with 2 mCi of [ 32 P]orthophosphate (Amersham Pharmacia Biotech) in serum-free and phosphate-free DMEM for overnight. After 10 min of insulin or conditioned media stimulation, GSK3␤ was immunoprecipitated as described above and processed for phosphopeptide mapping as described previously (33).

GSK3␤ Is Involved in Wnt Signaling Pathway in Mamma-
lian Cells-To compare the regulation of GSK3␤ by insulin and Wnt, we needed to choose cell lines that respond to both signals. In this study, we used conditioned media from a stable Rat2 cell line expressing mouse Wnt-1 as a source of Wnt protein (36). We first tested the effect of the Wnt conditioned media on human embryonic 293 cells as this epithelial cell line does respond to insulin (37). We observed a maximal decrease in GSK3␤ activity at 10 min after the addition of Wnt media to cells (Fig. 1A). Wnt media also caused an accumulation of the cytosolic fraction of ␤-catenin, which peaked at about 3 h after stimulation (Fig. 1B). These results again place mammalian GSK3 upstream of ␤-catenin and downstream of Wnt. Similar results were also seen with C57MG (C57) cells, an immortalized mouse mammary gland epithelial cell line (data not shown). We then examined the relationship between Dsh/Dvl and GSK3␤ in 293 cells. Overexpression of Dsh/Dvl is known to activate the Wnt pathway and give rise to elevated levels of cytosolic ␤-catenin (23,38). We utilized a luciferase reporter driven by TCF/LEF binding sites (TOPTK) to measure the activity of transient overexpression of human Dishevelled 2 (hDvl2) (Fig. 1C). Wild type GSK3␤ and a dominant negative form of the TCF4 transcription factor (DNTCF4) blocked hDvl2 activity (Fig. 1C). GSK3␤ mutants that retain kinase activity, including serine 9 mutated to alanine or glutamic acid, and tyrosine 216 mutated to phenylalanine or glutamic acid, also retained the ability to block hDvl2 activity (see below and data not shown). In contrast, coexpression of kinase-dead mutants of GSK3␤ did not have any effect on the activity of hDvl2, nor did coexpression of active MEKK, an irrelevant protein kinase (Fig. 1C). Similar results were also obtained from same experiments using CHOIR, a Chinese hamster ovary cell line stably expressing human insulin receptor (data not shown). Due to the low transfection efficiency of C57 cells, CHOIR and 293 cells were used for experiments involving transient transfections. These observations confirmed that a decrease in GSK3␤ activity is necessary to convey signals from Dvl to ␤-catenin and GSK3␤ is downstream of Dvl in mammalian cells.
Wnt and Insulin Signaling Lead to Distinct Downstream Events, although GSK3 Is Involved in Both Pathways-Since GSK3␤ is a major player in both insulin and Wnt signaling, we compared changes in GSK3␤ activity upon insulin and Wnt stimulation. A similar decrease in GSK3␤ activity was observed in 3 cell lines, 293, CHOIR and C57 ( Fig. 2A). We then examined the downstream events of activated insulin and Wnt pathways. The level of cytosolic ␤-catenin was increased in cells stimulated by Wnt but unchanged in insulin-treated cells (Fig.  2B). GS activity was analyzed in CHOIR and C57 cells. Insulinstimulated cells yielded higher GS activity, while Wnt conditioned media had no effect (Fig. 2C). 293 cells had high basal GS activity, and no significant activity increase was detected with insulin treatment (data not shown). These data represent an example of specificity in signaling, yet raised the question how different downstream effects were achieved through a seemingly indistinguishable change in the activity of GSK3␤, a common component of the two signaling pathways.
Wnt Regulates GSK3 Activity through Mechanisms Other than Serine 9 Phosphorylation-Serine 9 (Ser-9) is a key regulation site of GSK3␤ responding to insulin signaling (7). Using a phosphospecific antibody against phospho-Ser-9 in GSK3␤, we were able to detect a clear increase of phosphorylation on this residue upon insulin stimulation in CHOIR and C57 cells (Fig. 3A). However, we did not observe any obvious difference of phospho-Ser-9 reactivity in samples treated with Wnt media or control media. The phosphotyrosine content remained constant before and after either insulin or Wnt stimulation. Similar results were observed from 293 cells (data not shown). Since PKB is known to be the upstream regulator of GSK3 in insulin signaling, we analyzed the phosphorylation status of two key residues in PKB, Thr-308 and Ser-473, using phosphospecific antibodies. There was a strong increase in phosphorylation of both residues responding to insulin but not to Wnt (Fig. 3B).

FIG. 1. Wnt signaling pathway is intact in 293 cells. A, GSK3␤
activity decreased when 293 cells were stimulated with Wnt. 293 cells in 6-cm dishes were incubated with control media or Wnt conditioned media for the indicated lengths of time. Endogenous GSK3␤ was immunoprecipitated from each sample and assayed for kinase activity. Activities were expressed as the percentage of that of the untreated sample. Error bars represent standard deviations from at least three independent experiments with duplicates in each experiment. B, ␤-catenin accumulated in 293 cells that were stimulated with Wnt. 293 cells in 6-cm dishes were incubated with Wnt conditioned media for the indicated lengths of time. Cytosolic fractions were prepared from each sample and immunoblotted for ␤-catenin and actin. Actin served as the loading control. C, wild type GSK3␤ blocked hDvl2 activated ␤-catenin/ TCF-dependent transcription. 293 cells in 24-well plates were transfected with FuGENE6 with a total of 0.5 g of plasmids including 50 ng of reporter plasmid TOPTK and 1 ng of internal control plasmid pRLTK. 75 ng of hDvl2 was used in the indicated transfections. Luciferase assays were performed 48 h after transfection. Luciferase activities were expressed as -fold increase compared with the vector control, which is set at 1. Wild type (wt) GSK3␤, kinase-dead (kd) GSK3␤, DNTCF4, or MEKK alone do not activate ␤-catenin/TCF-dependent transcription. Error bars represent standard deviations from at least three independent experiments with duplicates in each experiment.

FIG. 2. Insulin and Wnt lead to decreased GSK3 activity but distinct downstream events.
A, 293, C57, and CHOIR cells were incubated with insulin, control media or Wnt conditioned media for 10 min. Endogenous GSK3␤ was immunoprecipitated from each sample and assayed for kinase activity. Activities were expressed as the percentage of that of the untreated sample. Error bars represent standard deviations from at least three independent experiments with duplicates in each experiment. B, 293, C57, and CHOIR cells were incubated with insulin, control media, or Wnt conditioned media for 2 h. Cytosolic fractions were prepared from each sample and immunoblotted for ␤-catenin and actin. C, C57 and CHOIR cells were incubated with insulin, control media, or Wnt conditioned media for 2 h and glycogen synthase activities were assayed. The activities were expressed as -fold increase compared with the untreated sample, which is set at 1. Error bars represent standard deviations from at least three independent experiments with duplicates in each experiment.
We also metabolically labeled cells in vivo with [ 32 P]orthophosphate before treatment with insulin or Wnt. We did not detect any significant changes in the total level of phosphorylation or phosphoamino acid analysis of GSK3␤, although it confirmed that the majority of phosphorylation was on serine residues (data not shown). Phosphopeptide mapping was also performed (Fig. 3C). The phosphopeptide pattern for GSK3␤ from C57 cells treated with control or Wnt media appeared to be essentially the same. Nevertheless, GSK3␤ from insulin-treated cells elicited a distinctive increase in phosphorylation at the positions corresponding to peptides containing Ser-9 (33). Similar patterns were obtained from 293 and CHOIR cells (data not shown).
We then examined the response of a Ser-9 to alanine mutant of GSK3␤ (S9A-GSK3␤) to insulin and Wnt signals. 293 or CHOIR cells transiently expressing an HA-epitope-tagged wild type or S9A mutant of GSK3␤ were stimulated with insulin or Wnt conditioned media. The kinase activity of S9A-GSK3␤ no longer decreased in response to insulin (Fig. 4A), similar to what was reported previously (37). Upon Wnt stimulation, the S9A mutant exhibited an activity drop similar to that for the wild type kinase (Fig. 4A). We also tested the ability of S9A to block Wnt signal by hDvl2 (Fig. 4B). At higher expression level, both S9A mutant or wild type GSK3␤ efficiently blocked hDvl2activated TCF/LEF-driven luciferase activity. Interestingly, at lower expression level, S9A was able to block hDvl2 activity roughly 2-fold better than the wild type GSK3␤. This is probably due to the higher intrinsic kinase activity of S9A mutant that we and others have observed (35). Similar effects were observed using CHOIR cells (data not shown).
Response of GSK3␤ to Insulin and Wnt in Axin-Conductin Complex-GSK3␤ has been found in the axin-conductin-APC-␤-catenin complex (16,17). Because of the multitude of biological processes that GSK3␤ is involved in, it is reasonable to hypothesize that only a fraction of the total cellular GSK3␤ is FIG. 3. Serine 9 is not modified when cells are stimulated by Wnt. A, CHOIR and C57 cells were incubated with insulin, control media, or Wnt conditioned media for 10 min. Endogenous GSK3␤ was immunoprecipitated and immunoblotted with phosphospecific antibody against phospho-Ser-9 of GSK3␤. The same membrane was reprobed with 4G10 (anti-phosphotyrosine) and finally reprobed with antibody against GSK3. B, CHOIR cells were incubated with insulin, control media, or Wnt conditioned media for 10 min. Cells were lysed in RIPA buffer, and 150 g of total protein from each sample was used for the immunoblots. The same membrane was probed first with phosphospecific antibody against Thr-308 of PKB, then reprobed with phosphospecific antibody against Ser-473 of PKB, and finally with anti-PKB. C, C57 cells were labeled with [ 32 P]orthophosphate overnight before being stimulated with insulin, control media or Wnt conditioned media for 10 min. Phosphopeptide mapping was performed on immunoprecipitated endogenous GSK3␤.
FIG. 4. Serine 9 to alanine mutant of GSK3␤ and wild type GSK3␤ are regulated by Wnt signaling similarly. A, 293 cells are transfected with HA-tagged wild type or S9A-GSK3␤ plasmids. 24 h later, they are changed to serum-free media overnight. Cells were incubated with insulin, control media, or Wnt conditioned media for 10 min. HA-tagged GSK3␤ was immunoprecipitated with anti-HA from each sample and assayed for kinase activity. Activities were expressed as the percentage of wild type GSK3␤ in untreated cells. Error bars represent standard deviations from at least three independent experiments with duplicates in each experiment. B, 293 cells in 24-well plates were transfected with a total of 0.5 g of plasmids including 50 ng of reporter plasmid TOPTK and 1 ng of internal control plasmid pRLTK. 75 ng of hDvl2 and 375, 150, 75, and 15 ng of each GSK3␤ plasmid were used in the indicated transfections. Luciferase assays were performed 48 h after transfection. Luciferase activities were expressed as percentage of the activity of hDvl2 alone. Error bars represent standard deviations from at least three independent experiments with duplicates in each experiment.
in the axin-conductin complex. We investigated the response of GSK3␤ in the axin-conductin complex to insulin or Wnt. From CHOIR cells with insulin or Wnt treatment, endogenous GSK3␤ was coimmunoprecipitated with ectopically expressed EE-tagged conductin (EE-conductin). Only a small fraction of GSK3␤ was coimmunoprecipitated with EE-conductin and Ser-9 phosphorylation status of these GSK3␤ was not altered in response to Wnt media (Fig. 5A). After insulin treatment, we observed a much less significant increase in phosphorylation on Ser-9 of the GSK3␤ that was coimmunoprecipitated with EEconductin. Phosphorylation on tyrosine residues were constant (data not shown). In vitro kinase assays were also performed on the pool of GSK3␤ that coimmunoprecipitated with EE-conductin (Fig. 5B). We observed decreased kinase activity of GSK3␤ responding to Wnt. Furthermore, in response to insulin, the decrease in kinase activity of this subpool of GSK3␤ seemed to be less prominent than the total pool of GSK3␤.
Response of GSK3␤ to Inducible Dishevelled-As the most upstream intracellular component of the Wnt pathway, Dvl is capable of activating downstream molecules independent of the extracellular ligand Wnt (23,25,38). To further substantiate our findings described in the earlier sections, we also investigated the response of GSK3␤ to activated Dvl. We chose to generate a fusion protein between Dvl and a modified version of the hormone binding domain of the murine estrogen receptor (Dvl-ER) and use this as our inducible system to achieve rapid activation of the Wnt pathway (39). Expression of Dvl-ER activates TOPTK reporter activity in a hormone (4-hydroxyta-moxifen (4-HT))-dependent manner (data not shown). A 293 cell line was then made to stably express Dvl-ER (293-D-ER). As shown in Fig. 6A, cytosolic ␤-catenin accumulated upon 4-HT treatment in 293-D-ER cells. We then examined the phosphorylation status of the Ser-9 residue of GSK3␤. In this cell line, there remains a strong increase of phosphorylation of Ser-9 of GSK3␤ in response to insulin (Fig. 6B). 4-HT treatment did not cause any significant change in Ser-9 phosphorylation (Fig. 6B). Furthermore, a His 6 -tagged conductin was integrated into 293-D-ER cells (293-D-ER-His 6 -conductin) so that the conductin-bound pool of GSK3␤ can be coimmunoprecipitated using nickel beads. Upon either insulin or 4-HT stimulation, the phosphorylation of Ser-9 on this pool of GSK3␤ was not significantly altered (Fig. 6C). DISCUSSION GSK3␤ has been implicated in mediating many diverse signals in various cell types. It is believed that many signals can down-regulate the kinase activity of GSK3␤. How different FIG. 5. The regulation of GSK3␤ complexed with axin-conductin. A, CHOIR cells were transfected with EE-conductin using Lipo-fectAMINE Plus and changed to serum-free media 24 h later. Following overnight incubation, serum-starved cells were stimulated with insulin, control media, or Wnt conditioned media for 10 min. Cells were then lysed and immunoprecipitated with anti-EE antibody. The supernatant after the anti-EE immunoprecipitation was immunoprecipitated with anti-GSK3␤. All of the anti-EE immunoprecipitates and one sixth of the anti-GSK3␤ immunoprecipitates were loaded onto the gel. The immunoprecipitates were probed with phosphospecific antibody against phospho-Ser-9 of GSK3␤. Control experiments showed that the upper band is cross-reactivity with IgG. The same membranes were reprobed with antibody against GSK3␤. B, immunoprecipitates prepared as in A were assayed for kinase activity. Activities were expressed as the percentage of that of the untreated sample. Error bars represent standard deviations from at least three independent experiments with duplicates in each experiment. Cells were then lysed and incubated with nickel beads. The supernatant after nickel bead binding was immunoprecipitated with anti-GSK3␤. All of the nickel beads bound and one third of the anti-GSK3␤ immunoprecipitates were loaded onto the gel. The immunoprecipitates were probed with phosphospecific antibody against phospho-Ser-9 of GSK3␤. The same membranes were reprobed with antibody against GSK3␤ .
signaling pathways achieve specificity through GSK3␤ remains unclear. In this study, we investigated the differential regulation of GSK3␤ by two extracellular signals, insulin and Wnt. Although both signals decrease GSK3␤ activity to a similar extent, we found that insulin and Wnt lead to very distinct downstream events. Furthermore, unlike in insulin signaling, Ser-9 of GSK3␤ is not phosphorylated by the Wnt signaling pathway.
In any given organism, many cells will receive and respond to multiple extracellular signals. The cell lines that we chose for this study, for example, are able to respond to both insulin and Wnt. Insulin stimulation leads to increased glycogen synthesis, whereas Wnt causes accumulation of ␤-catenin. As expected, we observed a decrease in GSK3␤ activity in response to both insulin and Wnt. Decrease in GSK3␤ activity is sometimes sufficient to lead to downstream events. For example, lithium as a cell-permeable, noncompetitive inhibitor of GSK3␤ is able to mimic insulin in adipocytes (40) and Wnt signaling in Drosophila cells (26). HGF was shown to down-regulate GSK3␤ activity and up-regulate ␤-catenin level (41). However, we found that insulin did not cause ␤-catenin accumulation and Wnt did not increase glycogen synthase activity. Therefore, the effect of each signaling pathway is highly specific. A similar finding was reported by Staal et al. (42) that T cell activation causes decrease in GSK3␤ activity but no change in ␤-catenin accumulation.
A large body of evidence suggests that the regulation of GSK3␤ by insulin is a phosphorylation event at Ser-9 via activated PKB (7, 10). One way to achieve specificity via a common intermediate in different pathways could be through different posttranslational modifications. GSK3␤ activity can be down-regulated independent of Ser-9 phosphorylation or PKB in exercised muscle (43). We demonstrated that Wnt signaling did not cause Thr-308 or Ser-473 phosphorylation on PKB, nor was Ser-9 of GSK3␤ modified. Supporting our data, Yuan et al. (44) reported that activation of PKB alone is not sufficient to mimic Wnt signaling. However, in their report, exogenous PKB had a synergistic effect on ␤-catenin with exogenously expressed Wnt1 or Frat1. We did not find any synergistic effect if we stimulated cells with insulin and Wnt simultaneously (data not shown). It is possible that, although activating Wnt signaling does not rely on the activity of PKB, addition of active PKB in higher amounts than ordinary insulin stimulation can still act on one of its substrates, GSK3␤. This further decreased GSK3␤ activity may then contribute to the synergistic ␤-catenin accumulation effect. Our results are also in agreement with the report by Cook et al. (24), in which they showed that the decreased GSK3␤ activity in 10T1/2 cells responding to Drosophila wingless was insensitive to wortmannin. It was proposed that a phorbol ester-sensitive PKC may be the signaling molecule to GSK3␤ in the Wnt pathway (24). Certain isoforms of PKC are able to phosphorylate GSK3␤ in vitro (45). Nevertheless, PKC may not be the sole signaling molecule since inhibitors of PKC do not completely block Wnt effects (46). Integrin-linked kinase is another kinase that was able to induce nuclear ␤-catenin accumulation (47), activate PKB in vivo and phosphorylate GSK3␤ in vitro (48).
It is unclear whether GSK3␤ is regulated by phosphorylation during Wnt signaling. We were not able to detect any change in phosphorylation by in vivo labeling, phosphoamino acid analysis, and phosphopeptide mapping. Phosphorylation as a mode of regulating GSK3␤ in Wnt pathway cannot be ruled out although Ser-9 is not involved. Ruel et al. (25) observed an increased phosphorylation on serine residues of Drosophila GSK3 (shaggy) in inducible Wnt or Dsh cells. The hormoneinducible Dvl-ER cells we generated will be useful for revisiting this issue and also probing for other biological activities of Dsh/Dvl. In addition, there are two forms of GSK3 in mammalian cells, GSK3␣ and GSK3␤. The role of GSK3␣ in Wnt signaling is worthy of future investigation.
Specificity in signaling could be achieved by specific complex formation and subcellular translocation. Ras is an example of such regulation (49). At the focal point of a multitude of signals and mitogens, it has many downstream effectors. Ras interacts with different effectors via different groups of residues. Upon activation, Ras binds Raf and relocalizes Raf to the plasma membrane for further activation. GSK3␤ has been shown to form complexes with APC and axin-conductin (15)(16)(17). Recently, ectopically expressed Dvl, GBP/FRAT, and PP2A were found in the axin complex (23,28,50). Dvl was able to relocalize axin to the cell membrane (23). Peptides derived from Frat1 specifically inhibit kinase activity of GSK3 on axin and ␤-catenin but not GS (51). Several groups suggested that Wnt signaling dissociated this complex (20,25,28). Recently, overexpressed casein kinase I was found to interact with Dsh and to mimic Wnt signaling (52). Are there different pools of GSK3␤ complexes in different signaling pathways?
Several findings by us and others imply that GSK3␤ may not be limiting in cells. First, kinase-dead GSK3␤ mutants failed to elicit any effect on Wnt signaling in 293 or CHOIR cell lines based on the unaltered ␤-catenin-dependent luciferase activity (data not shown and 23). Although in Xenopus overexpression of kinase-dead GSK3␤ did cause axis duplication, mimicking activated Wnt signaling pathway (53), this effect has not been observed consistently in mammalian cell systems. Second, in cells that contain high amounts of ␤-catenin, such as SW480 cell line or 293 cells overexpressing ␤-catenin, exogenous GSK3␤ was not able to decrease the level of ␤-catenin-dependent transcription, whereas axin-conductin or APC did so efficiently (data not shown and Ref. 16). It is not certain whether the levels of overexpression of GSK3␤ and axin-conductin are comparable. Nevertheless, one explanation is that GSK3 is not the limiting factor, thus supporting the idea that a subpopulation of GSK3␤ is dedicated to form complexes with axin-conductin and only this pool of GSK3␤ participates in Wnt signaling. Furthermore, we observed that the GSK3␤ complexed to transiently or stably expressed conductin was significantly protected from Ser-9 phosphorylation by insulin. Further analysis through the investigation of this complex will shed light on our understanding of how GSK3 is regulated in the Wnt signaling pathway.