Expression and Characterization of GSK-3 Mutants and Their Effect on β-Catenin Phosphorylation in Intact Cells*

Glycogen synthase kinase-3 (GSK-3) is a serine-threonine kinase that is involved in multiple cellular signaling pathways, including the Wnt signaling cascade where it phosphorylates β-catenin, thus targeting it for proteasome-mediated degradation. Unlike phosphorylation of glycogen synthase, phosphorylation of β-catenin by GSK-3 does not require primingin vitro, i.e. it is not dependent on the presence of a phosphoserine, four residues C-terminal to the GSK-3 phosphorylation site. Recently, a means of dissecting GSK-3 activity toward primed and non-primed substrates has been made possible by identification of the R96A mutant of GSK-3β. This mutant is unable to phosphorylate primed but can still phosphorylate unprimed substrates (Frame, S., Cohen, P., and Biondi R. M. (2001) Mol. Cell 7, 1321–1327). Here we have investigated whether phosphorylation of Ser33, Ser37, and Thr41 in β-catenin requires priming through prior phosphorylation at Ser45 in intact cells. We have shown that the Arg96 mutant does not induce β-catenin degradation but instead stabilizes β-catenin, indicating that it is unable to phosphorylate β-catenin in intact cells. Furthermore, if Ser45 in β-catenin is mutated to Ala, β-catenin is markedly stabilized, and phosphorylation of Ser33, Ser37, and Thr41 in β-catenin by wild type GSK-3β is prevented in intact cells. In addition, we have shown that the L128A mutant, which is deficient in phosphorylating Axin in vitro, is still able to phosphorylate β-catenin in intact cells although it has reduced activity. Mutation of Tyr216 to Phe markedly reduces the ability of GSK-3β to phosphorylate and down-regulate β-catenin. In conclusion, we have found that the Arg96 mutant has a dominant-negative effect on GSK-3β-dependent phosphorylation of β-catenin and that targeting of β-catenin for degradation requires prior priming through phosphorylation of Ser45.

Glycogen synthase kinase-3 (GSK-3) 1 is a serine-threonine kinase that is involved in both insulin and Wnt signaling. In both signaling pathways, GSK-3 is constitutively active and becomes inhibited upon binding of insulin or Wnt ligand to their respective receptors. As a result of insulin signaling, GSK-3 phosphorylates glycogen synthase, and inhibition of GSK-3 leads to accumulation of dephosphorylated, active glycogen synthase, thus leading to stimulation of glycogen synthesis. For Wnt signaling, the primary substrate of GSK-3 is ␤-catenin. Phosphorylation by GSK-3 targets ␤-catenin for ubiquitination and proteasome-mediated degradation. Inhibition of GSK-3 thus results in ␤-catenin stabilization. ␤-catenin then translocates into the nucleus, where it activates gene transcription in conjunction with transcription factors of the Lef/TCF family (1)(2)(3).
GSK-3 activity toward cellular substrates can be regulated at several levels, including serine and tyrosine phosphorylation of GSK-3 and interaction of GSK-3 with other proteins in multiprotein complexes, as well as priming of substrates through an independent phosphorylation event (4).
Upon binding of insulin to its receptor, Akt/protein kinase B is activated through a phosphatidylinositol 3-kinase-dependent cascade and phosphorylates GSK-3 at an N-terminal serine (Ser 21 in GSK-3␣ and Ser 9 in GSK-3␤), resulting in inhibition of its activity (5).It has been shown recently that the phosphorylated N terminus of GSK-3 acts as a competitive pseudosubstrate that occupies the substrate binding site of GSK-3 (6,7). In addition to the N-terminal serine, the activity of GSK-3 is modulated through phosphorylation of a tyrosine residue (Tyr 279 in GSK-3␣ and Tyr 216 in GSK-3␤). Tyrosine phosphorylation leads to increased GSK-3 activity, and mutation of the tyrosine to phenylalanine reduces its activity in vitro as well as in cells, as assessed by inhibition of c-Jun activity by GSK-3 (8). However, it is unclear whether tyrosine phosphorylation of GSK-3 plays a regulatory role under physiological conditions. For instance, tyrosine phosphorylation does not change in response to growth factors (9). The identity of the kinase responsible for the tyrosine phosphorylation event is not known, although it has been suggested that this may be an autophosphorylation event (4).
In contrast to insulin signaling, phosphorylation of the Nterminal serine is not involved in Wnt-mediated inhibition of GSK-3 (10). In the Wnt/␤-catenin pathway, GSK-3 activity is regulated through interaction with binding partners in multiprotein complexes. These complexes include ␤-catenin, the scaffold protein Axin, and APC, the product of the adenomatous polyposis coli gene.Both Axin and APC are required for GSK-3-mediated phosphorylation of ␤-catenin and are also substrates for GSK-3 (11,12). Phosphorylation of Axin by GSK-3 stabilizes the protein (13) and facilitates the binding between Axin and ␤-catenin (14,15). Phosphorylation of APC by GSK-3 also facilitates its interaction with ␤-catenin (12,16). Thus, phosphorylation of both Axin and APC by GSK-3 promotes the formation and stabilization of the ␤-catenin degradation complex. Wnt signaling is transmitted via Dishevelled, which is recruited into the ␤-catenin degradation complex, resulting in inhibition of GSK-3-mediated phosphorylation and stabilization of ␤-catenin. In addition, it has been suggested that the GSK-3-binding protein GBP/FRAT also transmits the Wnt signal (16,17). FRAT has been shown to compete with Axin for binding at the same or overlapping binding site in GSK-3 (18,19) and so would be expected to disrupt the interaction between GSK-3 and Axin that is required for the phosphorylation of ␤-catenin. GBP/FRAT inhibits GSK-3 selectively only toward specific substrates such as ␤-catenin and Tau, while not affecting phosphorylation of other substrates such as glycogen synthase, eukaryotic initiation factor 2B (eIF2B), and the cAMPresponse element-binding protein (CREB) peptide (18 -20).
Additionally, utilization of substrates by GSK-3 can be regulated by prior phosphorylation of the substrates. This occurs through phosphorylation at a serine or threonine residue four amino acids C-terminal to the GSK-3 phosphorylation site (4). For example, glycogen synthase can be primed by CK2 at Ser 656 (21), which facilitates GSK-3-mediated phosphorylation at Ser 652 , followed by phosphorylation of Ser 648 , Ser 644 , and then Ser 640 . A similar motif in ␤-catenin (Ser 29 -X 3 -Ser 33 -X 3 -Ser 37 -X 3 -Thr 41 -X 3 -Ser 45 ) is known to be phosphorylated by GSK-3. However, no priming kinase has been identified, and it is at present unclear whether phosphorylation of ␤-catenin by GSK-3 requires priming (22). Binding of the F-box protein ␤-TrCP, a component of the E3 ubiquitin ligase, which results in targeting of ␤-catenin for proteasome-mediated degradation, is dependent on the phosphorylation of Ser 33 and Ser 37 (23)(24)(25). Mutation of these Ser residues in cell lines and in human patients bearing colon and other carcinomas has been shown to result in ␤-catenin stabilization (26 -28). Likewise, mutations of Thr 41 or Ser 45 prevent ␤-catenin degradation (27)(28)(29). This suggests that phosphorylation of both Thr 41 and Ser 45 is required for phosphorylation of the two Ser residues within the ␤-TrCP recognition element and that, like glycogen synthase, ␤-catenin is phosphorylated by GSK-3 in a sequential manner beginning from the C terminus of the consensus sequence.
Recently, the crystal structure of GSK-3␤ was determined, and it was demonstrated that the phosphorylated priming site of the substrate interacts with three positive residues in GSK-3␤, Arg 96 , Arg 180 , and Lys 205 , which form a binding pocket for the priming phosphate (6,30,31). Mutation of one of those positive residues, Arg 96 to Ala, has been shown to prevent phosphorylation of primed substrates by GSK-3 (7). GSK-3 is also known to be able to phosphorylate unprimed substrates in vitro, albeit with lower activity. Interestingly, mutation of Arg 96 did not affect GSK-3-mediated phosphorylation of unprimed substrates in vitro (7). Thus, the Arg 96 mutant was unable to phosphorylate glycogen synthase or primed glycogen synthase-derived peptide in vitro but was able to phosphorylate non-primed glycogen synthase-derived peptide as efficiently as wild type GSK-3 (7). The Arg 96 mutant was shown to completely retain the ability to phosphorylate ␤-catenin and Axin in vitro (7). Furthermore, a phospho-octapeptide derived from the N terminus of GSK-3␤, which interacts with the binding site for the priming phosphate, did not impair the phosphorylation of ␤-catenin and Axin in vitro (7). This suggests that the binding site for the priming phosphate is not required for phosphorylation of ␤-catenin by GSK-3 in the Wnt signaling pathway. Frame et al. (7) also identified a residue, Leu 128 , that is required for the phosphorylation of Axin in vitro. Mutation of Leu 128 to Ala, however, did not impair the ability of GSK-3 to phosphorylate primed substrates.
Although many cell types are responsive to both the insulin and Wnt signals, there is generally no cross-talk between the two pathways. The specificity is likely to be maintained through the interaction with different binding partners and compartmentalization of GSK-3 within the cell (22). In addition, cross-talk may also be prevented as a result of different substrate characteristics in the two signaling pathways. Thus, primed and unprimed substrates may be affected differentially by inhibition of GSK-3 through phosphorylation of the N-terminal serine or through interaction with GBP/FRAT, as has been shown in vitro (18,20). However, it is not clear at present whether ␤-catenin is an unprimed or primed substrate in vivo and whether these mechanisms are likely to play a role in achieving specificity of the two signaling cascades.
In this study we have investigated the mechanism of the phosphorylation of ␤-catenin by GSK-3 in intact cells. Our data support a sequential phosphorylation of ␤-catenin beginning at the C terminus that is dependent on the prior phosphorylation of Ser 45 . The Arg 96 mutant GSK-3␤ (which is unable to phosphorylate primed substrates) but not kinase-inactive GSK-3␤ has a dominant-negative effect on ␤-catenin phosphorylation.

Plasmid Constructs and Transfection of HEK293T Cells-The
V5-␤-catenin-pcDNA3.1 expression vector was purchased from Invitrogen. To generate the Myc-GSK-3␤ expression construct, human GSK-3␤ was cloned from a cDNA library and incorporated into the pcDNA3.1 expression vector. The Myc epitope tag was added at the N terminus. S9A, K85M/K86I, R96A, L128A, and Y216F GSK-3␤ mutants and the S45A mutant of human ␤-catenin were prepared using the Stratagene site-directed mutagenesis kit. Wild type or mutant Myc-GSK-3␤-pcDNA3.1 and V5-␤-catenin-pcDNA3.1 expression vectors were transfected into subconfluent HEK293T cells using the FuGENE reagent (Roche) according to the instructions of the manufacturer. 48 h after transfection, the cells were washed with ice-cold phosphate-buffered saline and then lysed in Triton X-100 containing lysis buffer as previously described (19). The lysates were snap-frozen in liquid nitrogen and stored at Ϫ80°C until required. Lysates were precleared by centrifugation before use. For measurements of cytosolic ␤-catenin by immunoblotting, a hypotonic lysis buffer was used as described in Culbert et al. (19).
Immunoprecipitation of V5-␤-Catenin-10 l of protein G-Sepharose was coupled to 5 g of monoclonal anti-V5 antibody (Invitrogen), which was used to immunoprecipitate V5-␤-catenin from 0.5 ml of precleared hypotonic cell lysate. The pellets were washed twice with 1 ml of Buffer A (50 mM Tris/HCl, 0.1 mM EGTA, 0.1% (v/v) ␤-mercaptoethanol, pH 7.5) containing 0.5 M NaCl and then twice with 1 ml of Buffer A without NaCl. The pellets were then used for immunoblotting with the phospho-␤-catenin antibody.
␤-Catenin/Lef/TCF-regulated Gene Reporter Assay-Luciferase activity in cells transiently transfected with a Lef/TCF-regulated luciferase gene reporter construct was determined as previously described (32).

Wild Type and Mutant Myc-GSK-3␤
Are Expressed at Similar Levels-Four mutant cDNA expression vectors, Myc-GSK-3␤ S9A, Y216F, L128A, and R96A, were generated by site-directed mutagenesis and expressed in HEK293T cells. All mutant proteins were expressed at similar levels compared with wild type Myc-GSK-3␤ (Fig. 1). The transfected proteins were much more abundant compared with endogenous GSK-3␤ protein, thus allowing for functional characterization of the mutant proteins. To assess the activity of the various mutants toward ␤-catenin, V5-␤-catenin was cotransfected with wild type or mutant Myc-GSK-3␤. GSK-3-dependent phosphoryla-tion of V5-␤-catenin was measured directly using a phosphospecific antibody or indirectly by determining GSK-3-dependent stabilization of V5-␤-catenin and transcriptional activation of a recombinant ␤-catenin/Lef/TCF-regulated luciferase reporter construct.
Activity of the S9A, Y216F, and L128A Mutants-Expression of S9A, Y216F, and L128A Myc-GSK-3␤ resulted in only small changes in total levels of cotransfected V5-␤-catenin ( Fig. 2A). Wild type Myc-GSK-3␤ and all three mutants significantly increased the phosphorylation of V5-␤-catenin at Ser 33 , Ser 37 , and Thr 41 (Fig. 2, A and B). However, the degree of phosphorylation of V5-␤-catenin varied between the different mutants. Although both wild type and S9A Myc-GSK-3␤ increased the phosphorylation of ␤-catenin 4-to 5-fold, the increase with the L128A mutant was ϳ3-fold and with the Y216F mutant ϳ2fold. However, only the difference between the effects of the Y216F mutant compared with wild type and S9A Myc-GSK-3␤ reached statistical significance (p Ͻ 0.05).
Wild type Myc-GSK-3␤ and the different mutants also led to a significant decrease in ␤-catenin/Lef/TCF-dependent reporter gene activity (Fig. 2C). The relative inhibitory effects of wild type Myc-GSK-3␤ and the different mutants were qualitatively similar to their effects on V5-␤-catenin phosphorylation. However, as with Myc-GSK-3␤-dependent phosphorylation of V5-␤catenin, only the difference between the effects of the Y216F mutant compared with wild type and S9A Myc-GSK-3␤ were statistically significant (p Ͻ 0.05).
Expression of R96A Myc-GSK-3␤ Markedly Increases V5-␤-Catenin Levels-We next determined whether the R96A mutant of Myc-GSK-3␤ was still able to phosphorylate V5-␤-catenin and target it for proteasome-dependent degradation. Surprisingly, we found that R96A Myc-GSK-3␤ markedly increased steady state levels of V5-␤-catenin (Fig. 3A). This increase in the protein concentration was not due to a change in the mRNA level of transfected V5-␤-catenin when the R96A mutant was cotransfected, as determined using an reverse transcriptase-PCR assay (not shown). Similarly, a marked increase in gene reporter activity was observed in the presence of R96A Myc-GSK3-␤ compared with the empty vector control (Fig. 3B), indicating that V5-␤-catenin was stabilized and presumably would translocate to the nucleus to activate Lef/TCFdependent transcription.
The R96A Mutant of GSK-3␤, but Not Kinase-inactive GSK-3␤, Acts as Dominant-negative-Our results suggested that the R96A mutant is unable to phosphorylate ␤-catenin and initiate its degradation and that it increases ␤-catenin by competing with endogenous GSK-3␤. To test this further, we measured luciferase gene reporter activity in the presence of the GSK-3 inhibitor SB-415286 (32). As reported previously (32), inhibition of endogenous GSK-3 in untransfected cells markedly in-creased gene reporter activity (Fig. 4). When cells were transfected with only V5-␤-catenin, basal gene reporter activity was highly elevated; addition of SB-415286 increased transcriptional activity of the ␤-catenin/Lef/TCF-dependent reporter construct further. Recombinant overexpression of Myc-GSK-3␤ decreased basal gene reporter activity, which was reversed in the presence of the GSK-3 inhibitor. In contrast, R96A Myc-GSK-3␤ markedly increased basal gene reporter activity. Addition of SB-415286 led to only a small further increase in transcriptional activity, suggesting that GSK-3-dependent phosphorylation and degradation of ␤-catenin was already inhibited because R96A Myc-GSK-3␤ acted as dominantnegative.
We next compared the effect of the R96A mutant with that of kinase-inactive Myc-GSK-3␤. We found that although the R96A mutant markedly increased the total level of cotransfected V5-␤-catenin, expression of kinase-inactive K85M/K86I Myc-GSK-3␤ did not result in any increase in V5-␤-catenin compared with cells transfected with empty vector and V5-␤catenin (Fig. 5). V5-␤-catenin and wild type (wt) or mutant Myc-GSK-3␤ were cotransfected at a vector DNA ratio of 2:1 as described under "Materials and Methods." A, total V5-␤-catenin was determined by immunoblotting. Cell lysates were then immunoprecipitated, using V5 antiserum. The immunoprecipitates were subjected to SDS-PAGE and immunoblotting using a phosphospecific antibody that detects ␤-catenin when phosphorylated at Ser 33 , Ser 37 , and Thr 41 . B, summary of four independent immunoblots using the phospho-␤-catenin-specific antibody. C, ␤-catenin/Lef/TCF-dependent luciferase reporter activity was determined as previously described (32) and expressed as percent decrease of the V5-␤-catenin-induced activity (n ϭ 5). 33 , Ser 37 , and Thr 41 by GSK-3␤-The R96A mutant of GSK-3␤ has been shown to be unable to phosphorylate primed substrates, while retaining the ability to phosphorylate unprimed substrates in vitro (7). Thus, our finding that it inhibits the GSK-3-mediated degradation of ␤-catenin suggests that phosphorylation of the two serine residues (Ser 33 and Ser 37 ) in the ␤-TrCP recognition motif of ␤-catenin requires priming by means of prior phosphorylation of Ser 45 and subsequently of Thr 41 . To determine whether this is the case, Ser 45 in V5-␤-catenin was mutated to Ala. This resulted in a marked stabilization of V5-␤-catenin (Fig. 6A), as had been previously observed when the residue was deleted (27). Coexpression of wild type Myc-GSK-3␤ decreased levels of wild type V5-␤catenin but did not affect S45A V5-␤-catenin levels (Fig. 6A).

Mutation of Ser 45 in ␤-Catenin to Ala Prevents Phosphorylation of Ser
Similarly, ␤-catenin/Lef/TCF-dependent reporter activity was increased about 3-fold with the S45A mutant compared with wild type V5-␤-catenin (Fig. 6B). Coexpression of wild type Myc-GSK-3␤ decreased the reporter activity with wild type V5-␤-catenin by 43.1 Ϯ 9.5% (p Ͻ 0.001). In contrast, wild type Myc-GSK-3␤ decreased the reporter activity in the presence of the S45A mutant by only 3.8 Ϯ 1.4% (not significant). These results suggest that GSK-3␤ is unable to phosphorylate V5-␤catenin and target it for proteasome-dependent degradation if Ser 45 is mutated to Ala. Consistent with this, Fig. 7 shows that the phosphorylation of V5-␤-catenin at residues Ser 33 , Ser 37 , and Thr 41 was prevented in the Ser 45 mutant, suggesting that priming of Ser 45 is required for the phosphorylation event. DISCUSSION The aim of this study was to gain insight into the regulation of the GSK-3␤-mediated phosphorylation of ␤-catenin in cells. In particular, conflicting data have been reported with respect to the requirement of priming phosphorylation on ␤-catenin to facilitate its utilization as a GSK-3 substrate. In glycogen synthase, GSK-3 phosphorylates the four N-terminal serines in the motif Ser 640 -X 3 -Ser 644 -X 3 -Ser 648 -X 3 -Ser 652 -X 3 -Ser 656 (21). Phosphorylation of these serines by GSK-3 is sequential from the C to the N terminus and is dependent on prior phosphorylation of Ser 656 by CK2. GSK-3␤ is known to phosphorylate a similar motif in ␤-catenin: Ser 33 -X 3 -Ser 37 -X 3 -Thr 41 -X 3 -Ser 45 . Although only Ser 33 and Ser 37 are part of the recognition motif for binding of ␤-TrCP to ␤-catenin, mutations of all of these serine and threonine residues, including Thr 41 and Ser 45 , are known to stabilize ␤-catenin and to be associated with various cancers (27)(28)(29). This suggests that, like glycogen synthase, ␤-catenin is phosphorylated in a sequential manner beginning at the C terminus. However, it has been shown in vitro that phosphorylation of ␤-catenin by GSK-3 does not require priming. Consistent with this, the R96A mutant of GSK-3␤ was shown in vitro to phosphorylate ␤-catenin as efficiently as the wild type enzyme (7).
Here we have used this mutant to investigate the mechanism of GSK-3-mediated phosphorylation of ␤-catenin in intact cells. In contrast to in vitro studies, we have found that the R96A mutant prevents GSK-3-mediated phosphorylation and degradation of ␤-catenin. Furthermore, mutation of Ser 45 in ␤-catenin to Ala prevented phosphorylation of the Thr 41 , Ser 37 , and Ser 33 residues in ␤-catenin. These results indicate that like glycogen synthase ␤-catenin is phosphorylated in a sequential manner, beginning at the C terminus, in cellular systems. If Ser 45 is phosphorylated, R96A GSK-3␤ is unable to catalyze the phosphorylation of Thr 41 because the substrate is primed. Consequently the mutant enzyme acts as a dominant-negative. It is presently unclear whether priming of ␤-catenin through phosphorylation of Ser 45 is mediated by GSK-3 itself or by a different kinase.
We have also found that the L128A mutant of GSK-3␤ was able to phosphorylate ␤-catenin in intact cells, although its activity was reduced compared with wild type enzyme. The Leu 128 mutant has been reported to be deficient in phosphorylating, but not in binding, Axin in vitro while retaining the ability to phosphorylate primed substrates (7). The ability of the Leu 128 mutant to phosphorylate ␤-catenin in intact cells would thus be in agreement with the mechanism of ␤-catenin phosphorylation by GSK-3. The reduced activity of L128A mutant GSK-3␤ may be due to impaired GSK-3-dependent phosphorylation of Axin and/or APC, which has been shown to promote the formation and stabilization of the ␤-catenin degradation complex (12, 14 -16). Thus, reduced phosphorylation of ␤-catenin by the L128A mutant may result from destabilization of the ␤-catenin signaling complex.
In contrast to the R96A mutant, we have found that kinase-inactive GSK-3␤ was unable to act as a dominant-negative in signaling to ␤-catenin. Although there are several reports showing dominant-negative effects of kinase-inactive GSK-3 in Wnt signaling in Xenopus (26,(33)(34)(35), activation of the Wnt signaling pathway by kinase-inactive GSK-3 in a mammalian cell system has been reported only in a study by Staal et al. (36). That study found that kinase-inactive GSK-3␤ and lithium activate ␤-catenin-dependent Tcf-1-mediated transcription in C57MG but not in Jurkat T cells. GSK-3 does not bind ␤-catenin directly but requires interaction with Axin and APC to phosphorylate ␤-catenin. It is possible that the dominantnegative activity is dependent on the interaction with Axin, because kinase-inactive GSK-3 has been shown to be unable to interact with Axin (11,37). This implies that the interaction of GSK-3 with Axin is dependent on its kinase activity. Because Axin may be an unprimed substrate for GSK-3 (7), it is possible that the R96A mutant retains the ability to phosphorylate and bind Axin. This would allow access for the R96A mutant, but not kinase-inactive, GSK-3 into the ␤-catenin degradation complex. Thus, only R96A mutant, but not kinase-inactive, GSK-3, would be able to compete with endogenous GSK-3. Finally, as expected, we have shown that mutation of Ser 9 to Ala did not affect GSK-3␤-mediated phosphorylation of ␤-catenin, whereas mutation of Tyr 216 to Phe markedly reduced the phosphorylation of ␤-catenin in intact cells. These findings are consistent with other studies, in which tyrosine phosphorylation has been shown to increase the activity of GSK-3␤ toward various substrates in vitro and in intact cells, including c-Jun (8) and Tau (38). Our results suggest that tyrosine phosphorylation is also required for GSK-3␤-dependent phosphorylation of ␤-catenin, although it is not clear whether it may serve a regulatory role in the Wnt/␤-catenin pathway.
In conclusion, this study shows that phosphorylation of ␤-catenin at Ser 33 , Ser 37 , and Thr 41 in intact cells requires priming through phosphorylation of Ser 45 . The Arg 96 mutant, which is unable to phosphorylate primed substrates, but not kinase-inactive GSK3␤, has a dominant-negative effect on the phosphorylation of ␤-catenin.