A Common Phosphate Binding Site Explains the Unique Substrate Specificity of GSK3 and Its Inactivation by Phosphorylation

Molecular Cell, Vol. 7, 1321–1327, June, 2001, Copyright 2001 by Cell Press

A Common Phosphate Binding Site Explains the Unique Substrate Specificity of GSK3 and Its Inactivation by Phosphorylation Sheelagh Frame1, Philip Cohen,1,2,3 and Ricardo M. Biondi1,3 1 Division of Signal Transduction Therapy and 2 MRC Protein Phosphorylation Unit School of Life Sciences University of Dundee Dundee DD1 5EH United Kingdom

Summary The inhibition of GSK3 is required for the stimulation of glycogen and protein synthesis by insulin and the specification of cell fate during development. Here, we demonstrate that the insulin-induced inhibition of GSK3 and its unique substrate specificity are explained by the existence of a phosphate binding site in which Arg-96 is critical. Thus, mutation of Arg-96 abolishes the phosphorylation of “primed” glycogen synthase as well as inhibition by PKB-mediated phosphorylation of Ser-9. Hence, the phosphorylated N terminus acts as a pseudosubstrate, occupying the same phosphate binding site used by primed substrates. Significantly, this mutation does not affect phosphorylation of “nonprimed” substrates in the Wnt-signaling pathway (Axin and ␤-catenin), suggesting new approaches to design more selective GSK3 inhibitors for the treatment of diabetes. Introduction Glycogen synthase kinase 3 (GSK3) is a remarkable protein kinase whose inhibition is essential for two distinct functions within cells, namely insulin signaling and the specification of cell fates during embryonic development. First, GSK3 becomes inhibited in response to insulin, leading to the dephosphorylation and activation of housekeeping proteins, such as glycogen synthase and eukaryotic protein synthesis initiation factor 2B (eIF2B), and hence to the stimulation of glycogen and protein synthesis (reviewed in Cohen et al., 1997). The inability of insulin to trigger these and other processes is at the heart of non-insulin-dependent or type II diabetes mellitus (NIDDM), the most common disorder of metabolism. Secondly, GSK3 is an essential component of the Wnt signaling pathway, which is essential for normal development (reviewed in Dale, 1998) and which regulates cell proliferation in adult tissues. In this pathway, GSK3 becomes inhibited in response to Wnts, causing the dephosphorylation of other substrates, including Axin (Willert et al., 1997; Ikeda et al., 1998), the adenomatous polyposis coli gene product (APC) (Rubinfeld et al., 1996), and ␤-catenin (Yost et al., 1996). Aberrant regulation of this pathway occurs in many human can3 Correspondence: [email protected] (P.C.), r.m.biondi@dundee. ac.uk (R.M.B.)

cers (reviewed in Polakis, 2000). For example, mutations in APC that disrupt its normal function are commonly found in colorectal cancers, while mutations in ␤-catenin that make it resistant to degradation are found in several different tumor types. These distinct functions of GSK3 highlight the need to control its specificity and regulation very tightly. During insulin signal transduction, GSK3 becomes inhibited through the PKB-catalyzed phosphorylation of an N-terminal serine residue (Ser-9 in GSK3␤ and Ser-21 in GSK3␣) (Cross et al., 1995). However, the mechanism by which phosphorylation inhibits GSK3 activity is currently not understood. Inhibition of GSK3 in response to Wnts occurs through a mechanism that is not yet completely understood but which does not involve phosphorylation of the N-terminal serine of GSK3 (Ruel et al., 1999; Ding et al., 2000). Protein phosphorylation is a key event in cellular regulation. However, the molecular basis for regulating the activity, the efficiency, and the fidelity of protein kinases remains poorly understood. Collectively, evidence suggests that the recognition between protein kinases and their substrates can involve “docking interactions” at sites in the kinase which are distinct from the enzymesubstrate interaction that occurs at the active center, thereby providing specificity (Tanoue et al., 2001; Frodin et al., 2000; Balendran et al., 2000). Nevertheless, these types of studies have been performed on a limited number of protein kinases (mostly on MAPK family members [Jacobs et al., 1999]), and the molecular mechanisms that operate and the extent to which they occur in nature are yet to be established. GSK3 is phylogenetically most closely related to the cyclin-dependent protein kinases (CDKs), such as CDK1 (also called cdc2) and CDK2. However, the specificity of GSK3 is unique in that it requires a priming phosphate located at n ⫹ 4 (where n is the site of phosphorylation) in order to phosphorylate many of its substrates, such as glycogen synthase (Fiol et al., 1987). In contrast, the phosphorylation of Axin and ␤-catenin in the Wnt signaling pathway is not known to require a priming phosphate and may rely on high-affinity interactions in a multiprotein complex with GSK3. Thus, Axin binds to both GSK3 and ␤-catenin, bringing these proteins into close proximity to facilitate their phosphorylation by GSK3 (Ikeda et al., 1998). In this study, we present evidence for a specific site of interaction between the phosphate of the primed substrate and Arg-96 of GSK3␤. In addition, we provide evidence that this same phosphate binding site is also occupied by Ser-9 once it becomes phosphorylated by PKB. The existence of this site helps to explain several features of GSK3, such as its unusual substrate specificity requirements and the mechanism by which it becomes inhibited in response to insulin and growth factors. These findings have important implications for drug development in this area.

Molecular Cell 1322

Figure 1. Prediction of Residues in GSK3 Involved in the Interaction with the n ⫹ 4 Phospho-Serine Present in Some GSK3 Substrates (A and B) The ribbon structures of the kinase domain of PKA (A) and CDK2 (B) bound to ATP are shown. Structures shown are 1ATP for PKA and 1QMZ for cdk2, available on the NCBI database. Both structures are in the active conformation, phosphorylated in the activation loop (not shown). The backbones of the specific protein kinase A inhibitor peptide (PKI) and the CDK2 peptide substrate (shown as sticks) are indicated. Based on the assumption that the substrate will bind in a similar manner in GSK3, the region within the box would contain the putative phosphate binding site. Major constituents of this region are the ␣1 helix (equivalent to ␣C in PKA) and the activation loop; residues N90 from PKA and R50 from CDK2, equivalent to R96 in GSK3␤, are also represented by sticks. The position of the PSTAIRE motif involved in CDK2 interaction with cyclins is marked. The inactive (nonphosphorylated) CDK2 structure shows important changes in the position of the activation loop as well as changes in the ␣1 helix from those shown here. (C) Alignment of the amino acid residues spanning the ␣1 helix of human GSK3␤, mouse PKA, and human CDK2. Identical residues are denoted by white letters on a black background, and similar residues are denoted by gray boxes. Arg-96 in GSK3␤, the residue mutated in this study, is marked with an asterisk.

Results Arg-96 Interacts with the Priming Phosphate of Substrates In order to predict residues involved in the putative binding site for the priming phosphate, we modeled the structure of GSK3 on the three-dimensional structures of cyclic AMP- and cyclin-dependent protein kinases. The structures of PKA and CDK2 (Figure 1) show the position occupied by peptide substrates. If the substrate is positioned similarly in GSK3, then the phosphate binding site for primed substrates should be restricted to residues within the boxed region. This region contains the ␣1 helix of the small lobe and residues of the activation loop. Residues within this region are involved in substrate interaction in the protein kinase CK2 (Sarno et al., 1997). We mutated several positively charged residues that were candidates to interact with the priming phosphate and tested whether the activity of these mutants was altered toward primed synthetic phospho-

peptides corresponding to the sites on glycogen synthase and eIF2B phosphorylated by GSK3. The mutation of Arg-96 to Ala in the small lobe of the kinase domain was found to severely impair the phosphorylation of these peptides as well as glycogen synthase itself (Figure 2). In contrast, the activity of the Arg-96-Ala mutant was similar to wild-type GSK3 when assayed against nonprimed substrates, such as Axin and ␤-catenin (Figure 2). The mutation of several other basic residues in the vicinity of Arg-96 (including Arg-92-Ala and Lys94-Ala) did not affect selectivity toward either class of substrate, suggesting that Arg-96 interacts directly with the priming phosphate. This is strongly supported by the finding that the replacement of Arg-96 by another basic residue (Lys) also abolished activity toward primed substrates but not nonprimed substrates (Figure 2). Lithium ions, which are used widely to inhibit GSK3 in cellbased assays, suppressed the activity of all the GSK3 mutants toward primed and nonprimed substrates similarly to the wild-type enzyme. Thus, lithium ions

Basis for Specificity and Inhibition of GSK3 1323

Figure 2. Mutation of Arg-96 to Ala or Lys in GSK3␤ Abolishes Activity toward Primed Substrates but Not Nonprimed Substrates Nonprimed peptide substrates were phosphorylated much less efficiently than the corresponding primed peptide substrates (i.e., specific activities for the wild-type enzyme toward the primed pGS and peIF2B peptides were 415 U/mg and 925 U/mg, respectively, and for the nonprimed GS and eIF2B peptides, 1.33 U/mg and 0.3 U/mg, respectively). The pGS (YRRAAVPPSPSLSRHSSPHQSpEDEEE, where Sp is phospho-serine) and GS (YRRAAVPPSPSLSRHSSPHQAEDEEE) peptides were based on the sites in glycogen synthase phosphorylated by GSK3, and the peIF2B (RRAAEELDSRAGSpPQL) and eIF2B (RRAAEELDSRAGSPQL) peptides were derived from the GSK3 site in eIF2B. The activity of the Arg-96-Ala (R96A) and Arg-96-Lys (R96K) mutants was expressed as a percentage of wild-type (WT) activity, which was set at 100%.

do not appear to rely on the phosphate binding site of GSK3. The Phosphorylated N Terminus Competes with the Priming Phosphate of Substrates for Binding to the Phosphate Binding Site We then hypothesized that the insulin-induced phosphorylation of the N terminus of GSK3 might result in the displacement of the priming phosphate by the phosphorylated N terminus by competing for binding to the same phosphate binding site. If this were the case, then the Arg-96-Ala mutant should be resistant to inhibition by PKB. In the experiment shown in Figure 3A, wildtype GSK3␤ was inhibited about 90% by preincubation with PKB. In contrast, the Arg-96-Ala (Figure 3A) and Arg-96-Lys mutants were completely resistant to inhibition despite Ser-9 becoming phosphorylated to the same extent and at a similar rate to wild-type GSK3 (data not shown). This was confirmed by studies in 293 cells transfected with wild-type and mutant forms of GSK3. In these cells, IGF-1 triggered the phosphorylation of wild-type GSK3 and the Arg-96-Ala and Arg-96Lys mutants at Ser-9, but only the wild-type GSK3 was inhibited (data not shown). A further prediction is that the inhibition of GSK3 by PKB should be lost as the concentration of the primed substrate is increased, and the data presented in Figure 3B show that this is indeed the case. The Phosphorylated N Terminus Acts as a Pseudosubstrate Further evidence that the N terminus of GSK3 competes for substrate binding when it is phosphorylated was obtained through the use of an 11-residue synthetic phospho-peptide termed NTptide-11 (RPRTTSpFAESC, where

Sp is phospho-serine), corresponding to residues 4–14 of GSK3␤. This peptide inhibited the activity of wild-type GSK3␤ toward the standard primed substrate, and this effect was specific for NTptide-11, since neither the unphosphorylated version of this peptide nor a control phospho-peptide of the same length caused any inhibition (Figure 3C). In addition, it was found that substitution of the phospho-serine by a phospho-threonine in the context of the NTptide-11 peptide (RPRTTTpFAESC) also inhibited the phosphorylation of the primed substrate equally well (data not shown). However, replacement of the phosphoserine with either glutamic acid or aspartic acid, which is thought to mimic the charge of a phosphorylated residue, had no effect on phosphorylation of the pGS peptide (data not shown). This implies that a phosphorylated residue and the particular amino acid sequence surrounding it are both important for inhibition. Furthermore, the activity of the Arg-96-Ala mutant toward the standard primed substrate peptide was unaffected by NTptide-11 (data not shown), implying that Arg-96 interacts specifically with the phosphorylated N terminus. Thus, in effect, NTptide-11 and the phosphorylated N terminus of GSK3 exert their inhibitory effects by acting as pseudosubstrates. This was confirmed by the finding that NTptide-11 became a substrate for GSK3 when the proline (equivalent to Pro-5 of GSK3␤) was changed to serine (RSRTTSpFAESC), phosphorylation occurring at the first serine (data not shown). An 8-mer Phospho-Peptide Derived from the N Terminus of GSK3␤ Selectively Inhibits the Phosphorylation of Primed Substrates In order to obtain a peptide that interacts with the phosphate binding site but does not extend into the catalytic

Molecular Cell 1324

Figure 3. The Phosphorylated N terminus of GSK3 Binds to the Phosphate Binding Site and Inhibits Its Kinase Activity by Acting as a Pseudosubstrate (A) The GSK3 Arg-96-Ala mutant is resistant to inhibition by PKB. Wild-type (WT) or mutant GSK3 (R96A) enzymes, either unphosphorylated or phosphorylated with PKB, were assayed for activity toward the primed pGS peptide. Activities are expressed as a percentage of the activity of unphosphorylated GSK3 (100%). Note that activity of GSK3[R96A] was over 100-fold lower than the wild-type enzyme toward the primed pGS peptide. (B) The inhibition of GSK3␤ by PKB is diminished as the concentration of primed pGS peptide substrate is increased. Activity is expressed as percentage inhibition of the activity of unphosphorylated GSK3. (C) Effect of peptides derived from the N terminus of GSK3␤ on the activity of wild-type GSK3␤. The activity of wild-type GSK3 toward the pGS peptide was measured in the absence or presence of NTptide-11 (RPRTTSpFAESC, where Sp is phospho-serine), NTtide-11 (RPRTTSFAESC), or an unrelated control phospho-peptide (QGDLMTpPQFTP, where Tp is phospho-threonine). Duplicate determinations varied less than 1%. (D) NTptide-8 selectively inhibits the activity of wild-type GSK3␤ toward primed substrates. The activity of wild-type GSK3 toward primed and nonprimed substrates was measured in the absence or presence of NTptide-11 or NTptide-8 (TTSpFAESC) at 2 mM. Activity was expressed as a percentage of that obtained in the absence of either peptide.

site, we synthesized an N-terminally truncated version of NTptide-11 termed NTptide-8 (TTSpFAESC). Strikingly, NTptide-8, unlike NTptide-11, was able to discriminate between nonprimed and primed substrates (Figure 3D). NTptide-8 was almost as potent as NTptide-11 in inhibiting the phosphorylation of several primed substrates but did not inhibit the phosphorylation of nonprimed substrates, including Axin. In addition, two shorter peptides that lacked the C-terminal serine and/or cysteine of NTptide-8 (NTptide-7, TTSpFAES; and NTptide-6, TTSpFAE) behaved similarly to NTptide-8 but were slightly less potent (data not shown). As with NTptide-11, the activity of the Arg-96-Ala mutant toward the standard primed substrate peptide was unaffected by NTptide-8 (data not shown), indicating that NTptide-8 also interacts specifically with Arg-96 of GSK3␤.

ing site. Leu-155 is a key residue in 3-phosphoinositidedependent protein kinase-1 (PDK1) involved in the activation of the enzyme and in specific interactions with several of its substrates (Biondi et al., 2000). We therefore mutated the equivalent leucine residue in GSK3 (Leu-128) to Ala and tested the activity of the mutant enzyme toward several substrates. These studies demonstrated that the mutation of this residue had no effect on the phosphorylation of the standard primed substrate (Figure 4A), and the mutant enzyme was indistinguishable from wild-type GSK3␤ in every regard (data not shown), except that it had greatly reduced activity toward Axin (Figure 4B).

Mutation of Leu-128 Impairs Axin Phosphorylation The results presented above suggested that Axin must bind to GSK3 at a site distinct from the phosphate bind-

GSK3 is an unusual protein kinase in two respects. First, it has a unique specificity in requiring a C-terminal priming phosphate for efficient phosphorylation of several

Discussion

Basis for Specificity and Inhibition of GSK3 1325

Figure 4. Mutation of Leu-128 to Ala in GSK3␤ Abolishes Activity toward Axin Selectively Wild-type GSK3 and the L128A mutant were tested for their ability to phosphorylate the primed pGS peptide substrate (A) or Axin (B). The activity of the Leu-128-Ala (L128A) mutant was expressed as a percentage of wild-type (WT) activity, which was set at 100%. For the Axin assays, the reactions were loaded onto SDS-polyacrylamide gels, exposed to autoradiographic film ([B], lower panel), and quantified by phosphorimager analysis ([B], upper panel).

of its key substrates. Second, it is one of relatively few protein kinases that are inactivated by phosphorylation. In this paper, we demonstrate that these apparently distinct properties are actually connected through the utilization of a common phosphate binding site. Thus, the mutation of a single residue (Arg-96) simultaneously abolishes the ability of GSK3␤ to phosphorylate primed substrates (Figure 2) and to be inhibited by the phosphorylation of Ser-9 (Figure 3A). In contrast, the nonprimed substrates we tested were utilized with equal efficiency by either wild-type GSK3␤ or the Arg-96-Ala mutant enzyme (Figure 2). This suggests that Arg-96 plays a key role in interacting with both the priming phosphate of substrates and phosphorylated Ser-9. The existence of a phosphate binding site involving Arg-96 is supported by the three-dimensional structure of GSK3 reported by Dajani et al. (2001). Further evidence that the same binding site on GSK3␤ is used by the priming phosphate of substrates and phosphorylated Ser-9 was obtained by the finding that the degree of inhibition resulting from phosphorylation by PKB depended on the substrate concentration. The extent of inhibition decreased as the concentration of the primed substrate increased (Figure 3B). Thus, the degree of inhibition of GSK3 toward any physiological substrate is likely to depend on the local concentration of the substrate and its mode and affinity of interaction with GSK3. Ser-9 and the PKB consensus sequence surrounding it are conserved in GSK3 homologs from mammals, Xenopus, and Drosophila, but not in yeast, higher plants, Dictyostelium, or C. elegans. However, Arg-96 is conserved in all GSK3 homologs identified to date, suggesting conservation of the phosphate binding site in all organisms. Additional evidence that supports the ideas advanced above comes from the finding that a phospho-peptide corresponding to the sequence surrounding Ser-9 (NTptide-11) inhibited GSK3␤ activity toward the primed pGS substrate, whereas the nonphosphorylated peptide

(NTtide-11) did not (Figure 3C). This is not only consistent with our hypothesis, but implies that the phosphorylated N terminus acts as a pseudosubstrate, with the phosphorylated Ser-9 occupying the same position as the priming phosphate of substrates. Indeed, the fact that a serine-substituted version of the phospho-peptide ([P/S]NTptide-11) became a substrate for GSK3 confirms that this is the case. Furthermore, an N-terminally truncated version of this peptide (NTptide-8) can inhibit the phosphorylation of primed GSK3 substrates in a very specific and selective manner due to the fact that it no longer occupies the catalytic site and only competes for binding of primed substrates to the phosphate binding site on GSK3 (Figure 3D). NTptide-8 had no effect on the phosphorylation of nonprimed substrates, such as Axin. GSK3, in common with a number of other protein kinases, is reported to be phosphorylated in the activation loop (Y216 of GSK3␤). In other protein kinases whose structures have been solved, this phosphate group interacts with residues in the ␣1 helix of the small lobe; for example, Arg-50 of CDK2 interacts with phosphorylated Thr-160, and His-88 in the equivalent (␣C) helix of PKA interacts with phosphorylated Thr-198 (Figure 1). Thus, the area in close proximity to Arg-96 could potentially interact with the phosphate from the activation loop. It therefore remains a possibility that there is only one phosphate binding site that is used by primed substrates, phosphorylated Ser-9 and phosphorylated Tyr-216. However, phosphorylation of Tyr-216 is reported to increase GSK3 activity, in contrast to Ser-9 phosphorylaytion, which results in inhibition. Therefore, if phosphorylated Tyr-216 occupies the same binding site, it must be easily displaceable by either primed substrates or the phosphorylated Ser-9. The phosphate binding site described in this study does not affect the intrinsic specific activity of GSK3, since the activity toward nonprimed substrates (Axin, ␤-catenin, and the GS peptide) is unaffected by the mutation of Arg-96 to Ala or Lys. Hence, the phosphate binding site appears to be required as a “docking site” for interaction with prephosphorylated substrates. In this respect, it shows some similarities to docking sites found in the MAP kinases ERK1 and ERK2 (the CD and ED domains) that are required for specific interaction with their substrates (Tanoue et al., 2001). However, there are two important distinctions between the MAP kinase docking sites that have been described and the phosphate binding site in GSK3. First, the site in GSK3 is a phosphate-dependent docking interaction, which can be regulated by phosphorylation of the priming site in the substrate. This is initiated by CK2 in the case of glycogen synthase (Picton et al., 1982). Second, it provides an example of a specific site on a protein kinase that is required for the phosphorylation of a particular subset of substrates and not of others. If this is a more general phenomenon, it raises the possibility of developing compounds directed toward protein kinases that would affect the phosphorylation of selected downstream substrates without affecting the phosphorylation of others. Indeed, we have evidence that a hydrophobic pocket in PDK1 that is required for the phosphorylation

Molecular Cell 1326

of several of its physiological substrates (Biondi et al., 2000; Balendran et al., 2000) is not required for the phosphorylation of PKB (R.M.B. et al., submitted). In the present study, we found that the mutation of Leu-128 to Ala, which is equivalent to Leu-155 in the hydrophobic pocket of PDK1, inhibits the phosphorylation of Axin but not the phosphorylation of primed substrates (Figure 4). However, it is not clear why Leu-128 is required for Axin phosphorylation. It might be involved in binding Axin directly, but preliminary experiments have indicated that the mutation of Leu-128 to Ala did not abolish the interaction of GSK3 with Axin. It is therefore possible that Leu-128 plays a more subtle role in promoting GSK3 activity toward Axin. The activity of GSK3 is regulated by different mechanisms in the Wnt and insulin signaling pathways. The main mechanism leading to inhibition of GSK3 in response to Wnts does not involve the phosphorylation of Ser-9 (Ruel et al., 1999; Ding et al., 2000), but instead may involve displacement of Axin from its specific binding site on GSK3 as a result of the binding of other proteins, such as GBP/FRAT and Dishevelled (Li et al., 1999). Since a GSK3 binding peptide derived from FRAT 1 inhibits the phosphorylation of Axin and ␤-catenin but does not affect the phosphorylation of primed substrates, such as glycogen synthase or eIF2B (Thomas et al., 1999), this may provide a mechanism for restricting the effects of Wnts to the specific subset of GSK3 substrates required to transduce this developmental signal. In the insulin signaling pathway, the activity of GSK3 toward primed substrates, such as glycogen synthase or eIF2B, is suppressed by phosphorylation of the N terminus through competition for the same phosphate binding site. The common feature between the effects of Wnts and insulin is that in each case the inhibition of GSK3 is intimately linked to the specific requirements of particular substrates for their phosphorylation. Furthermore, it now appears that within the cell the population of GSK3␤ bound to Axin does not become phosphorylated at Ser-9 in response to insulin (Ding et al., 2000). This provides a mechanism for restricting the effects of insulin to the GSK3 substrates that control glycogen and protein synthesis, thereby preventing insulin from inhibiting the GSK3-catalyzed phosphorylation of components of the Wnt signaling network. Since insulin inhibits GSK3, compounds that inhibit GSK3 may mimic some of the actions of this hormone and have therapeutic potential for the treatment of type II (non-insulin-dependent) diabetes mellitus. Relatively specific cell permeant inhibitors of GSK3 have recently been described that in cell based assays mimic the ability of insulin to stimulate the conversion of glucose to glycogen (Coghlan et al., 2000). However, since these compounds are ATP competitive, they inhibit the phosphorylation of every GSK3 substrate tested to a similar extent. For this reason, they also stimulate ␤-catenindependent gene transcription (Coghlan et al., 2000), and therefore prolonged use of such drugs may have the potential to be oncogenic. The development of the next generation of protein kinase inhibitors may require, in some cases, a more subtle approach involving the selec-

tive inhibition of a discrete subset of substrates. The identification of a specific phosphate binding site in GSK3 has opened up a new opportunity to develop drugs that selectively inhibit the phosphorylation of key proteins involved in mediating the metabolic actions of insulin that are suitable for the long term treatment of diabetes, without the potential to be oncogenic. They may also be suitable for the treatment of other diseases in which GSK3 has been implicated (Hetman et al., 2000; Sperber et al., 1995). Experimental Procedures Purified Proteins Human GSK3␤ (wild-type and mutant) proteins were expressed in 293 cells with an N-terminal GST tag from the pEBG-2T vector and affinity purified on glutathione-Sepharose as described previously (Biondi et al., 2000). The results presented were virtually identical using GSK3 from two to four independent purifications. Human Axin [275–510] was expressed in bacteria and purified as described previously (Thomas et al., 1999). Glycogen synthase was purified from rabbit skeletal muscle as described previously (Nimmo et al., 1976). PKB␣ was expressed as a hexahistidine-tagged protein in insect cells, purified on nickel/nitriloacetate-agarose, and then activated to a specific activity of 310 U/mg by incubation with PDK1 and MAPKAP-K2. Molecular biology techniques, cell culture, and transfections were performed as previously described (Biondi et al., 2000). Measurement of GSK3 Activity The ability of wild-type and mutant GSK3␤ to phosphorylate various substrates was determined in 20 ␮l assays containing 50 mM TrisHCl (pH 7.5), 0.1% 2-mercaptoethanol, 10 mM MgCl2, 100 ␮M [␥-32P] ATP (500 cpm/pmol), GSK3␤, and either Axin [275–510] (1.8 ␮M), ␤-catenin (1.5 ␮M), glycogen synthase (0.6 ␮M), the primed pGS peptide (20 ␮M), the primed peIF2B peptide (20 ␮M), the nonprimed GS peptide (1.5 mM), the nonprimed eIF2B peptide (3 mM), or the [P/S]NTptide-11 peptide (20 ␮M), unless otherwise stated. After incubation for 15 min at 30⬚C, the reaction was stopped. For the pGS, peIF2B, and [P/S]NTptide-11 peptide assays, this was done by addition of 20 ␮l of 150 mM phosphoric acid, then 35 ␮l of the mixture was spotted onto P81 phosphocellulose paper, and the papers were washed and analyzed as described previously for assays of MAP kinase (Biondi et al., 2000). For the nonprimed GS and eIF2B peptide assays and assays using Axin [275–510], ␤-catenin, or glycogen synthase, the reactions were stopped by addition of SDS, heated at 90⬚C for 5 min, and loaded onto SDS-polyacrylamide gels. ␤-catenin assays were performed in the presence of Axin to facilitate complex formation. GSK3 activity was then quantified by phosphorimager analysis. Control assays were carried out in parallel in which either GSK3␤ or the substrate was omitted; these values were always less than 5% of the activity measured in the presence of these reagents. Inactivation of GSK3␤ was performed by phosphorylating GSK3 with PKB (86 nM) in a 20 ␮l assay consisting of 50 mM Tris-HCl (pH 7.5), 0.1% 2-mercaptoethanol, 10 mM MgCl2, 100 ␮M cold ATP, GSK3␤ (0.1 ␮M) for 10 min at 30⬚C. Then 20 ␮l assay buffer (50 mM Tris-HCl [pH 7.5], 0.1% 2-mercaptoethanol, 10 mM MgCl2, 100 ␮M [␥-32P]ATP (500 cpm./pmol), and the pGS peptide at the concentrations indicated) was added to this reaction and incubated for a further 10 min at 30⬚C. Acknowledgments We thank Andrew Paterson (Division of Signal Transduction Therapy) for providing the human GSK3␤ expression plasmid and active PKB␣ and Gareth Browne (MRC Protein Phosphorylation Unit) for glycogen synthase. We acknowledge the help of Agnieszka Kieloch and Neil Quinney for culture of 293 cells and the support of members of the Division of Signal Transduction Therapy and MRC Protein Phosphorylation Unit. The Division of Signal Transduction Therapy

Basis for Specificity and Inhibition of GSK3 1327

is funded by Astra-Zeneca, Boehringer Ingelheim, GlaxoSmithKline, Novo Nordisk, and Pfizer. The work was also supported by the UK Medical Research Council and The Royal Society. Received February 28, 2001; revised April 25, 2001. References Balendran, A., Biondi, R.M., Cheung, P.C.F., Casamayor, A., Deak, M., and Alessi, D.R.A. (2000). A 3-phosphoinositide-dependent protein kinase-1 (PDK1) docking site is required for the phosphorylation of protein kinase C␨ (PKC␨) and PKC-related kinase 2 by PDK1. J. Biol. Chem. 275, 20806–20813. Biondi, R.M., Cheung, P.C.F., Casamayor, A., Deak, M., Currie, R.A., and Alessi, D.A. (2000). Identification of a pocket in the PDK1 kinase domain that interacts with PIF and the C-terminal residues of PKA. EMBO J. 19, 979–988. Coghlan, M.P., Culbert, A.A., Cross, D.A., Corcoran, S.L., Yates, J.W., Pearce, N.J., Rausch, O.L., Murphy, G.J., Carter, P.S., Roxbee Cox, L., et al. (2000). Selective small molecule inhibitors of glycogen synthase kinase-3 modulate glycogen metabolism and gene transcription. Chem. Biol. 7, 793–803. Cohen, P., Alessi, D.R., and Cross, D.A. (1997). PDK1, one of the missing links in insulin signal transduction? FEBS Lett. 410, 3–10. Cross, D.A., Alessi, D.R., Cohen, P., Andjelkovich, M., and Hemmings, B.A. (1995). Inhibition of glycogen synthase kinase-3 by insulin mediated by protein kinase B. Nature 378, 785–789. Dajani, R., Fraser, E., Roe, S.M., Young, N., Good, V., Dale, T.C., and Pearl, L.H. (2001). Crystal structure of glycogen synthase kinase 3␤: structural basis for phosphate-primed substrate specificity and autoinhibition. Cell 105, 721–732. Dale, T.C. (1998). Signal transduction by the Wnt family of ligands. Biochem.J. 329, 209–223. Ding, V.W., Chen, R.H., and McCormick, F. (2000). Differential regulation of glycogen synthase kinase 3␤ by insulin and wnt signaling. J. Biol. Chem. 275, 32475–32481. Fiol, C.J., Mahrenholz, A.M., Wang, Y., Roeske, R.W., and Roach, P.J. (1987). Formation of protein kinase recognition sites by covalent modification of the substrate. Molecular mechanism for the synergistic action of casein kinase II and glycogen synthase kinase 3. J. Biol. Chem. 262, 14042–14048. Frodin, M., Jensen, C.J., Merienne, K., and Gammeltoft, S. (2000). A phosphoserine-regulated docking site in the protein kinase RSK2 that recruits and activates PDK1. EMBO J. 19, 2924–2934. Hetman, M., Cavanaugh, J.E., Kimelman, D., and Xia, Z. (2000). Role of glycogen synthase kinase-3beta in neuronal apoptosis induced by trophic withdrawal. J. Neurosci. 20, 2567–2574. Ikeda, S., Kishida, S., Yamamoto, H., Murai, H., Koyama, S., and Kikuchi, A. (1998). Axin, a negative regulator of the Wnt signaling pathway, forms a complex with GSK-3␤ and ␤-catenin and promotes GSK-3␤-dependent phosphorylation of ␤-catenin. EMBO J. 17, 1371–1384. Jacobs, D., Glossip, D., Xing, H., Muslin, A.J., and Kornfeld, K. (1999). Multiple docking sites on substrate proteins form a modular system that mediates recognition by ERK MAP kinase. Genes Dev. 13, 163–175. Li, L., Yuan, H., Weaver, C.D., Mao, J., Farr, G.H., III, Sussman, D.J., Jonkers, J., Kimelman, D., and Wu, D. (1999). Axin and Frat1 interact with dvl and GSK, bridging Dvl to GSK in Wnt-mediated regulation of LEF-1. EMBO J. 18, 4233–4240. Nimmo, H.G., Proud, C.G., and Cohen, P. (1976). The purification and properties of rabbit skeletal muscle glycogen synthase. Eur. J. Biochem. 68, 21–30. Picton, C., Woodgett, J., Hemmings, B., and Cohen, P. (1982). Multisite phosphorylation of glycogen synthase from rabbit skeletal muscle. Phosphorylation of site 5 by glycogen synthase kinase-5 (casein kinase-II) is a prerequisite for phosphorylation of sites 3 by glycogen synthase kinase-3. FEBS Lett. 150, 191–196. Polakis, P. (2000). Wnt signaling and cancer. Genes Dev. 14, 1837– 1851.

Rubinfeld, B., Albert, I., Porfiri, E., Fiol, C., Munemitsu, S., and Polakis, P. (1996). Binding of GSK3␤ to the APC-␤-catenin complex and regulation of complex asembly. Science 272, 1023–1026. Ruel, L., Stambolic, V., Ali, A., Manoukian, A.S., and Woodgett, J.R. (1999). Regulation of the protein kinase activity of Shaggy(Zestewhite3) by components of the wingless pathway in Drosophila cells and embryos. J. Biol. Chem. 274, 21790–21796. Sarno, S., Vaglio, P., Marin, O., Issinger, O.G., Ruffato, K., and Pinna, L.A. (1997). Mutational analysis of residues implicated in the interaction between protein kinase CK2 and peptide substrates. Biochem. 36, 11717–11724. Sperber, B.R., Leight, S., Goedert, M., and Lee, V.M. (1995). Glycogen synthase kinase-3␤ phosphorylates tau protein at multiple sites in intact cells. Neurosci. Lett. 197, 149–153. Tanoue, T., Maeda, R., Adachi, M., and Nishida, E. (2001). Identification of a docking groove on ERK and p38 MAP kinases that regulates the specificity of docking interactions. EMBO J. 20, 466–479. Thomas, G.M., Frame, S., Goedert, M., Nathke, I., Polakis, P., and Cohen, P. (1999). A GSK3-binding peptide from FRAT1 selectively inhibits the GSK3-catalysed phosphorylation of Axin and ␤-catenin. FEBS Lett. 458, 247–251. Willert, K., Shibamoto, S., and Nusse, R. (1997). Wnt-induced dephosphorylation of axin releases beta-catenin from the axin complex. Genes Dev. 13, 1768–1773. Yost, C., Torres, M., Miller, J.R., Huang, E., Kimelman, D., and Moon, R.T. (1996). The axis-inducing activity, stability, and subcellular distribution of ␤-catenin is regulated in Xenopus embryos by glycogen synthase kinase 3. Genes Dev. 10, 1443–1454.