A Noncatalytic Domain of Glycogen Synthase Kinase-3 (GSK-3) Is Essential for Activity*

Glycogen synthase kinase-3 (GSK-3) isoforms, GSK-3α and GSK-3β, are serine/threonine kinases involved in numerous cellular processes and diverse diseases, including Alzheimer disease, cancer, and diabetes. GSK-3 isoforms function redundantly in some settings, while, in others, they exhibit distinct activities. Despite intensive investigation into the physiological roles of GSK-3 isoforms, the basis for their differential activities remains unresolved. A more comprehensive understanding of the mechanistic basis for GSK-3 isoform-specific functions could lead to the development of isoform-specific inhibitors. Here, we describe a structure-function analysis of GSK-3α and GSK-3β in mammalian cells. We deleted the noncatalytic N and C termini in both GSK-3 isoforms and generated point mutations of key regulatory residues. We examined the effect of these mutations on GSK-3 activity toward Tau, activity in Wnt signaling, interaction with Axin, and GSK-3α/β Tyr279/216 phosphorylation. We found that the N termini of both GSK-3 isoforms were dispensable, whereas progressive C-terminal deletions resulted in protein misfolding exhibited by deficient activity, impaired ability to interact with Axin, and a loss of Tyr279/216 phosphorylation. Our data predict that small molecules targeting the divergent C terminus may lead to isoform-specific GSK-3 inhibition through destabilization of the GSK-3 structure.

Glycogen synthase kinase-3 (GSK-3) 2 enzymes are serine/ threonine kinases originally identified based on their activity toward glycogen synthase (1). Two mammalian GSK-3 isoforms exist as products of distinct genes, GSK-3␣ and GSK-3␤ (2), which are highly homologous within their internal kinase domains but diverge in sequence outside this region. Aberrant regulation and subsequent hyperactivity of GSK-3 has been linked to diseases such as Alzheimer disease, cancer, and diabetes (3). Thus, tight regulation of ubiquitous (4) and constitu-tive (5) GSK-3 activity is important for maintaining normal cellular function. Control of GSK-3 activity occurs by a combination of mechanisms, including prerequisite phosphorylation of GSK-3 substrates (priming), phosphorylation of GSK-3 itself, and localization of GSK-3 activity through protein interactions.
With Ͼ40 putative substrates (6,7), GSK-3 isoforms are broadly influential, and their activity toward substrates is commonly influenced by a priming phosphorylation mechanism (8). Recognition of a primed substrate by GSK-3 requires a triad of conserved residues, Arg 159/96 , Arg 243/180 , and Lys 268/205 of GSK-3␣/␤, which form a phosphate-binding pocket and facilitate optimal alignment of GSK-3 into a catalytically active conformation (9 -11). Although priming enhances the efficiency of GSK-3 phosphorylation, it is not a strict requirement. For example, the microtubule-associated protein Tau has both primed and unprimed sites targeted by GSK-3 activity (12).
Phosphorylation of Tyr 279 in GSK-3␣ and Tyr 216 in GSK-3␤ has been reported to facilitate GSK-3 activity by inducing rotation of the tyrosine side chain, which moves out of the substrate-binding groove, promoting substrate accessibility (9,11,13). Though other mechanisms have been reported (7,14,15), phosphorylation of GSK-3␤ at Tyr 216 is believed to occur through a post-translational and intramolecular autophosphorylation event during protein folding in an Hsp90-dependent manner (5,16,17).
Recently, a new mechanism of GSK-3 regulation was reported. Persistent activation of p38 MAPK was shown to mediate inhibitory phosphorylation of Thr 390 in GSK-3␤, resulting in accumulation of ␤-catenin and increased expression of Wnt target genes (23). Interestingly, p38 MAPK specifically phosphorylated GSK-3␤ but not GSK-3␣ (23). Consistently, other evidence has also suggested that GSK-3 isoforms exhibit distinct biological roles (24 -29), providing rationale for the development of GSK-3 isoform-specific inhibitors.
Currently, small molecule inhibitors of GSK-3 largely target the ATP-binding pocket, which contains two conserved lysine residues, Lys 148/149 in GSK-3␣ and Lys 85/86 in GSK-3␤, that are responsible for binding ATP and transferring the ␥-phosphate (3,15). Such inhibitors lack isoform specificity due to high homology of GSK-3 isoforms in this region. Thus, the development of isoform-specific GSK-3 inhibitors necessitates a greater understanding of the importance of the divergent Nand C-terminal regions of each GSK-3 isoform. Thus, we performed a structure-function analysis of GSK-3 isoforms in HEK 293T cells by examining the effect of GSK-3␣ and GSK-3␤ deletion and point mutants on 1) activity toward Tau, 2) ability to suppress Wnt signaling, 3) interaction with an Axin fragment, and 4) Tyr 279/216 phosphorylation. Our data demonstrate that the C termini of GSK-3 isoforms are important for activity, protein interactions, and Tyr 279/216 phosphorylation and predict that the development of therapeutic modulators targeting the C terminus may result in isoform-specific GSK-3 inhibition.
Cloning and Site-directed Mutagenesis-Full-length and deletion mutant human GSK-3␣ (Origene, accession number NM_019884) and human GSK-3␤ (obtained from Peter Klein, University of Pennsylvania) were PCR-amplified and TA-cloned into the Gateway entry vector pCR8GW TOPO (Invitrogen). Subsequent directional cloning into the Gateway destination vector pDEST27 (Invitrogen) generated N-terminal GST-tagged proteins. Point mutant GSK-3 constructs were generated using site-directed mutagenesis (Stratagene) (supplemental Fig. S1). DNA integrity and mutations were confirmed by sequence analysis. The GSK-3 interaction domain (GID, amino acids 321-429) of Xenopus Axin (obtained from Peter Klein, University of Pennsylvania) was PCR-amplified and TA-cloned into Gateway entry vector pCR8GW TOPO (Invitrogen). Subsequent directional cloning into Gateway destination vector, pDEST-Myc (obtained from the Belgian Coordinated Collections of Microorganisms/LMBP plasmid collection), generated N-terminal 6ϫ Myc-tagged protein. DNA integrity was confirmed by sequence analysis.
Plasmids-FLAG-tagged human Tau was obtained from Hemant Paudel, McGill University. Super 8ϫ TOPFlash and Super 8ϫ FOPFlash were obtained from Randall Moon, University of Washington. pMAX EGFP was obtained from Amaxa. Renilla luciferase plasmid pRL SV40 was obtained from Promega. The empty vector pcDNA3.1 was obtained from Invitrogen.
GST Pulldown Assay-Cleared cell lysate was incubated with glutathione-Sepharose beads (GE Healthcare) for 2 h at 4°C. Subsequently, reactions were centrifuged to pellet beads, which were then washed four times with 2 volumes lysis buffer supplemented with protease inhibitor mixture. After the final wash, protein was eluted by boiling in sample buffer.
Immunoblotting and Antibodies-Approximately equal amounts of protein lysate were separated by Tricine-SDS-PAGE, transferred to nitrocellulose, and immunoblotted. The following primary antibodies were used: rabbit polyclonal anti-GST (Cell Signaling), mouse monoclonal anti-FLAG M2 (Sigma), mouse monoclonal PHF1 (obtained from Peter Davies, Albert Einstein College of Medicine), mouse monoclonal 9E10 (anti-Myc, developed by J. Michael Bishop and obtained from the Developmental Studies Hybridoma Bank, The University of Iowa) and mouse monoclonal anti-GSK-3␤ pY216 (BD Transduction Laboratories), which also recognized GSK-3␣ pY279. Corresponding secondary antibodies used were either horseradish peroxidase-linked anti-mouse or horseradish peroxidase-linked anti-rabbit (GE Healthcare). Detection was facilitated using ECL Western blotting Substrate (GE Healthcare).
Lef/TCF Luciferase Reporter Assay-Lef/TCF reporter constructs containing eight optimal (Super 8ϫ TOPFlash) or eight mutant (Super 8ϫ FOPFlash) Lef/TCF binding sites (30) driving firefly luciferase, were co-transfected into HEK 293T cells in conjunction with pRL SV40, and either empty vector, wild-type (WT) GSK-3␣ or GSK-3␤, or mutant GSK-3␣ or GSK-3␤. Twenty-four h later, each transfection was equally transferred to a 24-well plate. Twenty-four h after transfer, cells were treated in triplicate for 6 h with either 50% Wnt3a-conditioned media, to stimulate canonical Wnt signaling, or 50% L cellconditioned media for control. Cells were then lysed with 1ϫ passive lysis buffer and Firefly and Renilla luciferase activities were measured using Firefly and Renilla Luciferase Assay kit (Biotium) in a Veritas microplate luminometer (Turner Biosystems).
Densitometry-Chemiluminescence was quantified by measuring area density using UltraLum documentation system and software (UltraLum Inc., Claremont, CA).

RESULTS
To investigate the functional significance of divergent regions and regulatory residues of GSK-3 isoforms, mammalian expression constructs containing N-terminal deletions, C-terminal deletions, or point mutations of human GSK-3␣ (Fig. 1A) and human GSK-3␤ (Fig. 1B) were generated. Deletion and point mutant GSK-3 expression constructs were tagged at the N terminus with GST.
Deletion of GSK-3 C Termini Impairs Tau Phosphorylation-To assay the effects of mutations on the activity of GSK-3␣ ( Fig.  2A) and GSK-3␤ (Fig. 2B), we examined the ability of WT and mutant enzymes to phosphorylate the well characterized GSK-3 substrate Tau in HEK 293T cells. Immunoblot analysis of total protein lysate revealed comparable expression levels of FLAG-Tau. Coexpression of WT GSK-3␣ and WT GSK-3␤ enhanced Tau phosphorylation at Ser 396 and Ser 404 as recognized by the phosphorylation-specific antibody PHF1 (31). Deletion of the N terminus did not affect the activity of either isoform toward Tau (␣⌬NT and ␤⌬NT). However, when the C terminus was deleted (␣⌬CT-4 and ␤⌬CT-4), GSK-3 isoforms lost nearly all activity, thus indicating amino acids 417-430 of GSK-3␣ and 345-367 of GSK-3␤ are essential for activity toward Tau. As anticipated, the point mutants resulting in GSK-3 inactivation (␣K148R and ␤K85R) demonstrated little activity toward Tau. Surprisingly, we saw no impairment of activity when we mutated the tyrosine residues implicated in potentiating GSK-3 activity (␣Y279F and ␤Y216F). Additionally, we observed a novel GSK-3␣-specific requirement for priming exhibited by the impaired activity of ␣R159A activity, whereas ␤R96A activity was comparable with WT.
Deletion of GSK-3 C Termini Abolishes Suppression of Wnt Signaling-Next, we quantified the ability of GSK-3 mutants to regulate canonical Wnt signaling by measuring transcriptional activation of a ␤-catenin-inducible Lef/TCF luciferase reporter, Super 8ϫ TOPFlash (30). HEK 293T cells were transfected with GST-GSK-3 constructs, Super 8ϫ TOPFlash, and Renilla luciferase for normalizing, followed by treatment with either Wnt3a-conditioned media or control media (L cell-conditioned media) (32). Addition of Wnt3a ligand-enhanced luciferase reporter activity of Super 8ϫ TOPFlash, without affecting the activity of a negative control, Super 8ϫ FOPFlash. As expected, overexpression of WT GSK-3␣ (Fig. 3A) and WT GSK-3␤ (Fig. 3B) suppressed luciferase reporter activity. N-terminal deletion mutants (␣⌬NT and ␤⌬NT) retained activity in the Wnt signaling pathway, whereas C-terminal deletion (␣⌬CT-4 and ␤⌬CT-4) completely abolished GSK-3 activity, as noted by the lack of luciferase reporter suppression. Gene expression analysis of the Wnt target Axin2 by real-time quantitative PCR confirmed that ␣⌬CT-4 and ␤⌬CT-4 lack activity in the Wnt signaling pathway (supplemental Fig. S1). These   MARCH 12, 2010 • VOLUME 285 • NUMBER 11 JOURNAL OF BIOLOGICAL CHEMISTRY 7959 data further argue that amino acids 417-430 in GSK-3␣ and 345-420 in GSK-3␤ are essential for GSK-3 activity. Point mutants ␣K148R and ␤K85R function as dominant-negatives in Wnt signaling by interfering with endogenous GSK-3 activities and activating luciferase reporter activity in the absence of the Wnt3a ligand, with the ␤K85R mutant producing a much more potent dominant-negative effect than the corresponding ␣K148R mutation. Expectedly, both ␣R159A and ␤R96A were impaired in their ability suppress the luciferase reporter, confirming that ␤-catenin is a primed GSK-3 substrate.

GSK-3 Structure-Function Analysis
GSK-3 C-terminal Deletions Fail to Interact with Axin GID-In the canonical Wnt signaling pathway, GSK-3 phosphorylation of ␤-catenin is regulated via interactions with the scaffold protein Axin (18,19). To better understand why deletion of the C termini of GSK-3 isoforms resulted in a loss of activity toward ␤-catenin, we examined the ability of GSK-3 mutants to interact with the GID of Axin (33). We predicted that the inability of ␣⌬CT-4 and ␤⌬CT-4 to suppress Lef/TCF luciferase reporter activity might result from an impaired interaction with Axin GID. Following purification on glutathione-Sepharose, the interaction of GST-GSK-3 mutants with Myc-tagged Axin GID (residues 321-429) was assayed by immunoblotting. WT GSK-3␣ (Fig. 4A) and WT GSK-3␤ (Fig. 4B) strongly interacted with Axin GID, whereas ␣⌬CT-4 and ␤⌬CT-4 were greatly impaired. Unexpectedly, ␣⌬CT-2, ␣⌬CT-3, ␤⌬CT-2, and ␤⌬CT-3, which exhibited activity in Wnt signaling, were also greatly impaired in their ability to interact with Axin GID. Furthermore, despite their dominant-negative activity in Wnt signaling, ␣K148R and ␤K85R exhibited impaired ability to interact with Axin GID. These data suggest a potential Axinindependent mechanism of Wnt signaling modulation by GSK-3 isoforms.
Mutating Conserved C-terminal Proline Residues Does Not Affect GSK-3 Activity-GSK-3␤ was crystallized in association with a minimal peptide derived from the Axin GID (13). Based on this crystal structure, the minimal Axin peptide occupies a hydrophobic groove in GSK-3␤ that is identical in sequence to GSK-3␣ (325-336 and 348 -362 in GSK-3␣ and 262-273 and 285-299 in GSK-3␤). This region was not deleted from our GSK-3 C-terminal deletion mutants. Thus, we speculated that GSK-3␣ and GSK-3␤ C-terminal deletion mutants were unable to bind Axin GID due to compromised structurally integrity. In silico sequence analysis revealed that two C-terminal proline residues were evolutionarily conserved (data not shown), Pro 442/443 of GSK-3␣ and Pro 379/380 of GSK-3␤, suggesting a functional importance. Therefore, we sought to determine the structural important of these proline residues were by converting them to alanine in the context of full-length GSK-3. We found no difference between the proline mutants and WT GSK-3 in Tau phosphorylation ( Fig. 2A, B) or Wnt suppression (Fig. 3A, B). A reduction in Axin GID interaction was noted with ␣P442/443A but not with ␤P379/380A (Fig. 4A, B). Taken together, the loss of activity observed with C-terminal deletion likely resulted from a general effect on protein structure, such as misfolding, and not from deletion of any particular residue(s).

DISCUSSION
Deletion of the C terminus in both GSK-3 isoforms resulted in a loss of activity, impaired ability to interact with Axin GID, and a concomitant lack of Tyr 279/216 phosphorylation. Additionally, C-terminal deletion mutants exhibited perinuclear aggregated localization in HeLa cells (supplemental Fig. S2) and also demonstrated reduced protein half-lives (supplemental Fig. S3). These data strongly suggest that C-terminal deletion of GSK-3 isoforms resulted in protein misfolding.
Furthermore, GSK-3␣ appeared to have greater sensitivity to C-terminal deletion, becoming impaired in activity beginning with ␣⌬CT-2 and progressing to complete loss of activity with ␣⌬CT-4. GSK-3␤, however, did not display impairment of activity until ␤⌬CT-4. Thus, GSK-3␣ activity may require unique intramolecular interactions involving the C terminus that are inconsequential for GSK-3␤ activity.
Intriguingly, GSK-3␣ and GSK-3␤ deletion mutants, ␣/␤⌬CT-2 and ␣/␤⌬CT-3, possessed the ability to suppress Wnt signaling, yet failed to bind Axin GID. A similar observation was reported by Fraser and colleagues, who described a GSK-3␤ point mutant, V267G/E268R, which failed to interact with Axin GID, but retained the ability to repress TCF luciferase reporter activity (34). These data imply that an Axin-independent mechanism of GSK-3 modulation of Wnt signaling may exist but this requires further validation.
Consistent with other reports (35)(36)(37), ␣K148R and ␤K85R conferred dominant-negative activities in Wnt signaling but, surprisingly, both mutants exhibited considerably reduced affinity for Axin GID compared with WT GSK-3␣ and WT GSK-3␤. Thus, the dominant-negative activity of ␣K148R and ␤K85R may not be due to displacement of endogenous GSK-3 from the Axin-mediated destruction complex as anticipated but rather, through an Axin-independent mechanism, perhaps similar to that of ␣/␤⌬CT-2 and ␣/␤⌬CT-3. However, we did  MARCH 12, 2010 • VOLUME 285 • NUMBER 11 not directly measure ␤-catenin levels, and we cannot discount ␤-catenin-independent activation of the Super 8X TOPFlash luciferase reporter.

GSK-3 Structure-Function Analysis
As demonstrated previously (38), disruption of the primed phosphate-binding pocket in ␣R159A and ␤R96A point mutants conferred dominant-negative activity to GSK-3␣ and GSK-3␤ in the luciferase reporter assay, though to a much lesser extent than ␣K148R and ␤K85R. Further, ␣R159A and ␤R96A retained strong interaction with Axin GID, which may account for their ability to repress luciferase reporter activity upon activation of Wnt signaling, albeit to a slightly lesser extent than WT GSK-3␣ and WT GSK-3␤.
The major difference we detected between GSK-3 isoforms was the distinct activities of ␣R159A and ␤R96A toward Tau phosphorylation at Ser 396 and Ser 404 . Although ␣R159A and ␤R96A functioned similarly toward ␤-catenin, ␣R159A was greatly compromised in its ability to phosphorylate Tau, suggesting GSK-3␣ phosphorylates Ser 396 and Ser 404 by a primed mechanism. GSK-3␤ R96A, however, retained full ability to phosphorylate Tau, suggesting GSK-3␤ does not require prerequisite priming. Our data has extended a previous observation (12) and by comparison, has revealed a unique difference between GSK-3␣ and GSK-3␤ activity toward Ser 396 and Ser 404 in Tau. Phosphorylation of Ser 396 has been implicated in the conversion of Tau to a pathological state in Alzheimer disease (39) and GSK-3 hyperactivity is believed to play an important role (reviewed in Ref. 40). Thus, GSK-3 serves as a target candidate for therapeutic intervention in Alzheimer disease, and based on our observation, targeting GSK-3 in an isoform-specific manner may be beneficial in preventing Ser 396 phosphorylation, and the resulting pathological conversion of Tau and destabilization of microtubules as exhibited in Alzheimer disease.
Several reports have suggested that phosphorylation at Tyr 216 enhances GSK-3␤ activity (5,7,14,15). However, our data suggest that tyrosine phosphorylation is not essential for activity of GSK-3 isoforms as demonstrated by the retained activity of point mutants ␣Y279F and ␤Y216F, which were comparable with WT GSK-3␣ and WT GSK-3␤, respectively. Our data correlates with crystal structure reports describing an active conformation of GSK-3␤ despite the lack of phosphorylation at Tyr 216 (9,11,41). Thus, the importance of GSK-3␣/␤ Tyr 279/216 phosphorylation is unclear, and the functional significance of this modification requires further investigation.
Our data were collected under conditions that did not stimulate Ser 21/9 phosphorylation or Thr 390 phosphorylation and therefore, point mutants ␣S21A, ␤S9A, and ␤T390A did not exhibit any significant change in activity or Tyr 279/216 phosphorylation compared with WT GSK-3␣ and GSK-3␤. Therefore, the effects of ␣S21A, ␤S9A, and ␤T390A were not obvious in these assays.
The role of GSK-3 isoforms in numerous and diverse disease processes together with evidence suggesting GSK-3 isoforms exhibit distinct activities and control discrete biological processes advocates for the development of isoform-specific GSK-3 inhibitors for use in therapeutic intervention. Currently, isoform-specific GSK-3 inhibitors are unavailable, and their development requires a more comprehensive understanding of the variable N and C termini in GSK-3. As a step in this direction, the aim of this study was to gain insight into the functional significance of the divergent N and C termini of GSK-3 isoforms and to evaluate the importance of specific regulatory residues. Here, we have identified a novel deletion mutation in the C terminus of GSK-3 isoforms that abolishes activity of both GSK-3␣ and GSK-3␤. Our data predict that targeting the divergent C terminus with modulators of GSK-3 activity may mimic C-terminal deletion and destabilize GSK-3 structure, resulting in isoform-specific inhibition. Future work directed at confirming and validating our hypothesis, particularly in an in vivo model system, may pave the way for new therapeutic approaches.