Glycogen Synthase Kinase-3β Is Complexed with Tau Protein in Brain Microtubules*

In Alzheimer's disease, microtubule-associated protein tau is hyperphosphorylated by an unknown mechanism and is aggregated into paired helical filaments. Hyperphosphorylation causes loss of tau function, microtubule instability, and neurodegeneration. Glycogen synthase kinase-3β (GSK3β) has been implicated in the phosphorylation of tau in normal and Alzheimer's disease brain. The molecular mechanism of GSK3β-tau interaction has not been clarified. In this study, we find that when microtubules are disassembled, microtubule-associated GSK3β dissociates from microtubules. From a gel filtration column, the dissociated GSK3β elutes as an ∼400-kDa complex. When fractions containing the ∼400-kDa complex are chromatographed through an anti-GSK3β immunoaffinity column, tau co-elutes with GSK3β. From fractions containing the ∼400-kDa complex, both tau and GSK3β co-immunoprecipitate with each other. GSK3β binds to nonphosphorylated tau, and the GSK3β-binding region is located within the N-terminal projection domain of tau. In vitro, GSK3β associates with microtubules only in the presence of tau. From brain extract, ∼6-fold more GSK3β co-immunoprecipitates with tau than GSK3α. These data indicate that, in brain, GSK3β is bound to tau within a ∼400-kDa microtubule-associated complex, and GSK3β associates with microtubules via tau.

In Alzheimer's disease (AD), 1 microtubule-associated protein tau becomes hyperphosphorylated and aggregates into paired helical filaments (PHFs) (reviewed in Refs. 1 and 2). PHF-tau (tau isolated from PHFs) is highly insoluble, displays a retarded mobility on an SDS gel, is incapable of binding to microtubules and promoting microtubule assembly, and is abnor-mally hyperphosphorylated (2). Upon dephosphorylation, PHFtau regains its normal mobility on an SDS gel, binds to microtubules, and promotes microtubule assembly (3,4). Abnormal phosphorylation has been suggested to cause the loss of tau function, microtubule instability, and neurodegeneration in AD brain (1,2). Thus, abnormal phosphorylation of tau is an important pathological process during AD ontogeny. The identification of tau kinases and the elucidation of the mechanism of tau phosphorylation are essential to determine how tau becomes abnormally phosphorylated in AD brain.
There are six tau isoforms generated through mRNA alternative splicing (5). On an SDS gel, tau isolated from brain extract displays several bands with sizes 50 -68 kDa representing various tau isoforms with different phosphorylation states (5,6). In normal brain, tau binds to microtubules and stabilizes microtubule structure. Tau phosphorylation reduces the affinity of tau for microtubules, destabilizes microtubules, and regulates microtubule dynamics (1,7). Normal adult tau is phosphorylated on four sites, whereas juvenile tau is phosphorylated on 10 sites (8). PHF-tau on the other hand, is phosphorylated on 19 sites (9). Most of these sites are situated at an (S/T)P motif recognized by a proline-directed kinase. Among proline-directed kinases, MAP kinase, Cdc2 kinase, Cdk5, and glycogen synthase kinase-3␤ (GSK3␤) phosphorylate tau in vitro (2). In addition, many non-proline-directed kinases such as cAMP-dependent protein kinase (PKA), protein kinase C, calmodulin-dependent protein kinase II, and phosphorylase kinase also phosphorylate tau in vitro (for a list of tau kinases, see Ref. 10).
Ishiguro et al. isolated tau kinase 1 from bovine brain microtubules (30). Biochemical and molecular cloning studies later determined that tau kinase 1 is GSK3␤ (31). Subsequent studies have confirmed that GSK3␤ is tightly associated with neuronal microtubules (32). As discussed above, tau is a microtubule-associated protein and regulates microtubule dynamics by binding to microtubules and stabilizing microtubule structure (7). These studies therefore suggest that GSK3␤-tau interaction may occur within microtubules, and this interaction may be important in the regulation of microtubule dynamics. Elucidation of the biochemical nature of tau-GSK3␤ interaction may also provide a clue as to the cause of abnormal tau phosphorylation in AD brain. However, very little information is available about microtubule-associated GSK3␤. In this study, we examined GSK3␤ within bovine brain microtubules. We now demonstrate that microtubule-associated GSK3␤ is a ϳ400-kDa species complexed with tau. We also provide evidence for the binding of GSK3␤ to microtubules via tau.

MATERIALS AND METHODS
cDNA Cloning-Pfu DNA polymerase-catalyzed PCR was carried out in a reaction mixture supplemented with 10% Me 2 SO by using a forward primer (5Ј-TCC CCC GGG ATG GGG CGG-3Ј) containing an SmaI site (italicized), a reverse primer (5Ј-CTC GAG TCA GGT GGA GTT GGA AGC-3Ј) containing an XhoI site (italicized), and a pBluescript plasmid template containing human brain GSK3␤ cDNA (a gift from Dr. J. R. Woodgett, University of Toronto). PCR conditions were as follows: one cycle of 94°C for one min; 25 cycles of 94°C for 1 min, 57°C for 1 min, and 72°C for 2 min; and one cycle of 72°C for 10 min. Adenine overhangs were added to the PCR product by incubating the product with 1 unit of Taq DNA polymerase at 72°C for 10 min. The final PCR product was purified and ligated into a pGEX-T Easy TA vector (Promega, Madison, WI) and amplified. GSK3␤ cDNA was excised by SmaI/XhoI from the pGEX-T Easy TA vector and ligated into the SmaI/XhoI site of a pGEX-6P-2 vector (Amersham Biosciences). The recombinant plasmid was transfected into Escherichia coli BL21(DE3) cells. The cloning was verified by DNA sequencing. Construction of various tau deletion mutant plasmids has been described previously (33).
Proteins-Glutathione S-transferase (GST)-GSK3␤ was purified from bacterial lysates. The overnight bacterial culture was diluted 20-fold in a fresh medium and incubated at 37°C with vigorous shaking. After ϳ4 h, protein expression was induced by adding isopropyl-␤-D-thiogalactoside to 1 mM, and the incubation and shaking was continued for another 4 h. The incubated medium was centrifuged at 16,000 ϫ g for 30 min, and the pellet was suspended in ice-cold lysis buffer (50 mM Tris-HCl (pH 7.5), 0.2 M NaCl, 1 mM DTT, and 1 mM EDTA) containing 1% Tween 20 and protease inhibitor mixture (1 mM phenylmethylsulfonyl fluoride, 5 mg/ml of benzamidine, and 1 g/ml each of leupeptin, aprotinin, and pepstatin). The suspended pellet was sonicated twice for 30 s each and centrifuged at 27,000 ϫ g for 30 min at 4°C. The supernatant was dialyzed against lysis buffer for 4 h and then mixed with glutathione-agarose beads (Sigma) previously equilibrated with lysis buffer. The beads were shaken end-over-end overnight and then packed in a glass column. The column was washed with ϳ30 ml of lysis buffer and eluted with 10 mM reduced glutathione in lysis buffer. Fractions containing GST-GSK3␤ were combined, concentrated to ϳ1 ml by dialysis against Aquacide III (Calbiochem), and dialyzed against lysis buffer at 4°C. After dialyzing for 2 h, 10 units of precision protease (Sigma) were added, and the sample was incubated with end-over-end shaking at 4°C for 16 h to remove the GST tag. The incubated sample was loaded onto a glutathione-agarose column preequilibrated in lysis buffer. GST remained bound to the column, whereas flow-through fractions containing GSK3␤ were combined and dialyzed against PEM buffer (0.1 M PIPES (pH 6.8), 1 mM EGTA, 1 mM MgSO 4 , and 1 mM ␤-mercaptoethanol) for 4 h at 4°C and used to generate Fig. 7.
Protein and Peptide Concentrations-The concentration of R-tau was determined spectrophotometrically (35). Concentrations of R-tau deletion mutants and phosphorylated R-tau were determined by a Bio-Rad protein assay using R-tau as the standard. All other protein concentrations were determined by the Bio-Rad protein assay using bovine serum albumin (BSA) as the standard. The concentration of GSK3 peptide substrate was based on amino acid analysis.
Anti-GSK3␤ Immunoaffinity Column-Anti-GSK3␤ rabbit serum (ϳ0.5 ml) was diluted with an equal volume of coupling buffer (0.1 M NaHCO 3 (pH 8.3), 0.5 M NaCl) and mixed with ϳ3 ml of CNBr-activated Sepharose 4B gel (Amersham Biosciences) previously washed with 1 mM HCl and equilibrated in coupling buffer. The mixture was shaken end-over-end overnight at 4°C. After shaking, the gel was recovered by centrifugation and washed with ϳ20 ml of coupling buffer. GSK3 Activity Assay-A synthetic peptide (KRREILSRRPSYR), derived from the cAMP response element-binding protein that becomes a specific substrate for GSK3 upon PKA phosphorylation (18), was synthesized at the peptide synthesis facility of the University of Calgary. This peptide was phosphorylated by PKA, and the phosphopeptide was purified by high pressure liquid chromatography. Purified phosphopeptide was used to assay GSK3 activity. The assay was initiated by adding 10 l of the sample to 20 l of the reaction mixture containing the rest of the assay components. The final concentrations of various assay components were 50 mM HEPES (pH 7.2), 0.1 mM EDTA, 0.1 mM DTT, 0.3 mM [␥ 32 P]ATP, 10 mM MgCl 2 , and 50 M PKA-phosphorylated peptide substrate. After 30 min at 30°C, 10 l of trichloroacetic acid was added to the assay mixture to stop the reaction, and the mixture was incubated at 4°C for 10 min. Following incubation, the assay mixture was centrifuged for 5 min using a bench top centrifuge, and 20 l of the supernatant was analyzed by phosphocellulose filter paper assay to determine the amount of radioactivity incorporated into the substrate peptide (34). One unit of GSK3 activity transfers 1 pmol of phosphate/min to substrate peptide under the above described standard assay conditions.
Microtubule Sedimentation Assay-Microtubule sedimentation assay was carried out as described previously (33) with some modifications. Microtubules were assembled at 37°C in an assay mixture (50 l each) containing 0.1 M PIPES (pH 7.0), 1 mM EGTA, 1 mM MgSO 4 , 1 mM ␤-mercaptoethanol, 0.4 mg/ml tubulin, 24 M microtubule-stabilizing drug taxol (Sigma), and 1 mM GTP. When included, the concentrations of GSK3␤, R-tau, and BSA were 10, 40, and 40 g/ml, respectively. After mixing all of the assay components in various combinations, each mixture was incubated at 37°C for 30 min and centrifuged at 50,000 ϫ g for 5 min at 37°C. The supernatant was withdrawn, and the pellet was dissolved in 30 l of sample buffer for SDS-PAGE. The pellet and the supernatant (15 l each) were subjected to SDS-PAGE. The gels were either immunoblotted using anti-GSK3␤ antibody or stained with Coomassie Brilliant Blue (Bio-Rad).
Purification of GSK3␤ from the Microtubule Fraction-All procedures were performed at 4°C. Microtubules were purified from a fresh bovine brain extract by three cycles of temperature-induced microtubule assembly and disassembly as described previously (33). The purified microtubule preparation containing ϳ5 mg/ml protein was incubated at 0°C for 30 min to disassemble microtubules. The incubated sample was centrifuged at 27,000 ϫ g for 30 min. The supernatant (ϳ25 ml) was loaded onto a phosphocellulose (Whatman, Fairfield, NJ) column (30 ϫ 5 cm) preequilibrated in PEM buffer containing 0.1 mM GTP. Flow-through fractions contained tubulin, whereas GSK3 activity remained bound to the column. The column was washed with ϳ75 ml of PEM buffer and eluted with 200 ml of a NaCl gradient (0 -0.8 M) in PEM buffer. Fractions containing GSK3 activity were combined (ϳ50 ml) and dialyzed against Mops buffer (25 mM Mops (pH 7.4), 50 mM ␤-glycerophosphate, 0.2 M NaCl, 10 mM NaF, 15 mM MgCl 2 , 1 mM EDTA, and 1 mM DTT). The dialyzed sample (labeled MAP-fraction in Figs. 6A and 9) was concentrated to ϳ7 ml by dialysis against Aquacide III. The concentrated sample was dialyzed against Mops buffer for 2 h and centrifuged at 27,000 ϫ g for 30 min. The supernatant was chro-matographed through an FPLC Superose 12 (Amersham Biosciences) column (50 ϫ 1.6 cm) equilibrated and eluted with Mops buffer. Effluent fractions containing GSK3 activity were combined (designated as the gel filtration fraction) and loaded onto an anti-GSK3␤ immunoaffinity column (1 ϫ 10 cm) preequilibrated in 25 mM Tris-HCl (pH 7.5), 0.2 M NaCl, 1 mM EDTA, and 1 mM DTT. The column was washed with 75 ml of equilibration buffer and eluted with 0.2 M glycine (pH 2.2). Fractions (0.5 ml each) were collected.
GST Pull-down Assay-The GST pull-down assay was performed essentially as described previously (33). Glutathione-agarose beads (50 l each) coated with GST or GST-GSK3␤ were mixed with 500 l of tau solution (50 mM Tris-HCl (pH 7.5), 50 mM NaCl, 5 mM EDTA, 0.3 M sucrose, 0.05% Tween 20, 0.3% BSA, and 50 g/ml of the indicated R-tau species) and incubated at 4°C overnight with end-over-end shaking. Incubated beads were recovered by centrifugation and washed four times with 50 mM Tris-HCl (pH 7.5), 0.5% Nonidet P-40, 1 mM EDTA, and 1 mM DTT. Washed beads were dissolved in 50 l of sample buffer for SDS-PAGE, boiled, and centrifuged, and 20 l of the supernatant was immunoblotted using an anti-tau antibody.
SDS-PAGE, Immunoblotting, and Immunoprecipitation-SDS-PAGE was performed by the method of Laemmli (36). After electrophoresis, gels were stained with either Coomassie Brilliant Blue or silver stain (Bio-Rad). Immunoblottings were carried out as described (34) by resolving samples on 10% SDS gels and using the indicated antibodies. For immunoprecipitation, samples (0.5 ml each) were precleared with 25 l of protein A-agarose beads (Sigma) preequilibrated with 25 mM Tris-HCl (pH 7.5) containing 0.2 M NaCl and divided into two halves. To each half, 30 l of either preimmune or indicated primary antiserum was added. Both halves were shaken end-over-end for 5 h at 4°C. After shaking, 25 l of protein A-agarose beads were added to each half, and shaking continued for another 2 h. Beads were collected by centrifugation and washed three times for 30 min each with equilibration buffer. Washed beads were dissolved in 50 l of sample buffer for SDS-PAGE, boiled, and centrifuged, and 20 l of the supernatant was immunoblotted using the indicated antibody. Band intensities were quantitated by scanning blots as described (35).

Gel Filtration Analysis of Microtubule-associated GSK3␤-
In a previous study, ammonium sulfate precipitation was utilized during purification of GSK3␤ from microtubule fraction (30). This treatment can dissociate a protein complex (33). We therefore excluded the ammonium sulfate precipitation step in our purification procedure and dissembled purified microtubules by incubating them at 0°C for 30 min. Disassembled microtubules were then chromatographed through a phosphocellulose column. Tubulin was recovered within the flowthrough fractions, and column-bound GSK3␤ was eluted with a NaCl gradient (data not shown). Effluent MAP fraction-containing GSK3␤ was analyzed by an FPLC Superose 12 gel filtration column (Fig. 1A). The activity assay (Fig. 1A) and immunoblot analysis (Fig. 1B) showed that GSK3␤ elutes as a relatively large species from the gel filtration column. The size of GSK3␤ in Fig. 1A (fraction 52) was estimated to be ϳ400 kDa. Since GSK3␤ is a globular protein with size ϳ47 kDa (11), our data indicate that microtubule-associated GSK3␤ is part of a ϳ400-kDa complex.
Identification of Protein Complexed with Microtubule-associated GSK3␤-To identify the protein complexed with microtubule-associated GSK3␤, we analyzed various gel filtration column fractions in Fig. 1A by immunoblot analysis using antibodies against MAP1, MAP2, tau, and tubulin (data not shown). MAP1 and MAP2 eluted within fractions 32-52. Tau was present within fractions 34 -72, whereas tubulin was undetectable in all of the column fractions analyzed. A proteinstained SDS gel (Fig. 1C) showed many protein bands with various sizes in fractions enriched with GSK3␤ (fractions 44 -64).
Since the above described SDS gel (Fig. 1C) and immunoblots did not give us any indication as to the identification of protein(s) complexed with GSK3␤ in the microtubule fraction, we combined fractions 44 -64 from Fig. 1A (designated as the gel filtration fraction) and chromatographed the gel filtration fraction through an anti-GSK3␤ immunoaffinity column. A silver-stained SDS gel showed a prominent band and three or four faint bands ( Fig. 2A) in fraction 5 (the peak effluent fraction). The major ϳ47-kDa band was identified as GSK3␤ by immunoblot analysis using anti-GSK3␤ antibody (data not shown). Faint protein bands migrated with sizes ϳ50 -68 kDa on an SDS gel. Since the various tau isoforms also migrate as ϳ50 -68-kDa bands on a SDS gel (5, 6), we immunoblotted various anti-GSK3␤ immunoaffinity effluent column fractions with anti-tau antibody (Fig. 2B). Tau was indeed present in anti-GSK3␤ immunoaffinity column effluent fractions. These data demonstrate that tau co-elutes with GSK3␤ from an anti-GSK3␤ immunoaffinity column and suggest that GSK3␤ is complexed with tau in the gel filtration fraction.
To gain more evidence in support of the above suggestion, we immunoprecipitated GSK3␤ from the gel filtration fraction by using an anti-GSK3␤ antibody and then immunoblotted the anti-GSK3␤ immune complex with an anti-tau antibody. In a similar manner, we immunoprecipitated tau from the gel filtration fraction and immunoblotted the anti-tau immune complex with an anti-GSK3␤ antibody. Tau co-immunoprecipitated with GSK3␤ (Fig. 3A) and GSK3␤ co-immunoprecipitated with tau (Fig. 3B). Based on these data, we conclude that tau is complexed with GSK3␤ in the gel filtration fraction. The effluent microtubule fraction containing GSK3␤ was analyzed by gel filtration chromatography using an FPLC Superose 12 gel filtration column (50 ϫ 1.6 cm) previously calibrated with blue dextran 2000, thyroglobulin (669 kDa), ferritin (440 kDa), catalase (232 kDa), aldolase (158 kDa), and BSA (67 kDa). Gel filtration was carried out at 4°C using an Amersham Biosciences FPLC system at a 1 ml/min flow rate. Fractions (1 ml each) were collected, and aliquots (10 l each) from the indicated fractions were used for GSK3 activity assay or SDS-PAGE. After electrophoresis, gels were either stained with Coomassie Brilliant Blue for proteins or immunoblotted using anti-GSK3␤ antibody. A, the column profile; B, immunoblot; C, protein-stained gel. Similar results were obtained in three independent experiments.
GSK3␤ Directly Binds to Tau-GSK3␤ may directly bind to tau within a GSK3␤⅐tau complex. Alternatively, GSK3␤ may associate with tau via another adapter molecule within the complex. To test these possibilities, we performed a GST pulldown assay. Glutathione-agarose beads coated with GST or GST-GSK3␤ were mixed with R-tau, washed, and then immunoblotted with anti-tau antibody. As shown in Fig. 4, R-tau bound to GST-GSK3␤ but not GST. These data indicate that GSK3␤ directly binds to tau.
Effect of Phosphorylation on the Binding of GSK3␤ to Tau-Tau is a phosphoprotein and binds to many proteins in a phosphorylation-dependent manner (33,37). To examine if tau phosphorylation influences GSK3␤-tau interaction, we examined the phosphorylation state of tau that eluted from an anti-GSK3␤ immunoaffinity column by immunoblot analysis using various tau phosphorylation-sensitive monoclonal antibodies. Tau-1 cross-reacts with nonphosphorylated tau, whereas 12E8, PHF-1, and AD-2 recognize phosphorylated tau (10, 38 -41). As shown in Fig. 6A, we detected Tau-1-, 12E8-, PHF-1-, and AD-2-immunoreactive tau in brain extract, indicating that the brain contains a mixture of phosphorylated and nonphosphorylated tau. However, tau within the MAP fraction cross-reacted with Tau-1 and AD-2 but not with PHF-1 or 12E8, which indicated that AD-2-specific sites are phosphorylated in microtubule-associated tau. Tau in the anti-GSK3␤ immunoaffinity fraction cross-reacted with Tau-1 but not with 12E8, PHF-1, and AD-2 (Fig. 6A), demonstrating that tau within the GSK3␤⅐tau complex is not phosphorylated on sites recognized by the aforementioned antibodies. These data suggest that GSK3␤ may bind to tau that has not been phosphorylated on Tau-1-, 12E8-, PHF-1-, and AD-2-specific sites.
To substantiate the above suggestion, we performed a GST pull-down assay. Glutathione-agarose beads coated with GSK3␤ were incubated with R-tau or Cdk5-phosphorylated R-tau under identical conditions. Incubated beads were washed and immunoblotted with an anti-tau antibody that equally cross-reacts with both tau and phosphorylated tau. GST-GSK3␤ bound to R-tau and completely failed to bind to phosphorylated R-tau (Fig. 6B). These data indicate that GSK3␤ binds to nonphosphorylated tau and does not bind to Cdk5phosphorylated tau.
GSK3␤ Associates with Microtubules through Tau-To determine the functional significance of GSK3␤Ϫtau binding, we performed a microtubule sedimentation assay. Microtubules were assembled from tubulin, GTP, and taxol. GSK3␤ was incubated with microtubules alone or in the presence of R-tau or BSA. Assembled microtubules were recovered as the pellet by centrifugation. The pellet containing microtubules and microtubule-bound proteins as well as the supernatant containing proteins that did not bind to microtubules were separated by SDS-PAGE. Gels were either stained for proteins (Fig. 7A) or immunoblotted using an anti-GSK3␤ antibody (Fig. 7B). All GSK3␤ was present in the supernatant when GSK3␤ was incubated with microtubules alone (Fig. 7B, lane 3) or in the presence of BSA (data not shown) and was undetectable in the microtubule pellet (Fig. 7B, lane 4). These data indicated that GSK3␤ does not bind to microtubules by itself. However, when incubated with microtubules in the presence of R-tau, ϳ30% of GSK3␤ was recovered in the microtubule pellet (Fig. 7B, lane 6) while ϳ70% remained in the supernatant (Fig. 7B, lane 5). Based on these observations, we concluded that GSK3␤ associates with microtubules only in the presence of R-tau.
Microtubule-associated GSK3␣ and GSK3␤-Mammalian brain contains two GSK3 isoforms: 51-kDa GSK3␣ and 47-kDa GSK3␤ (11). Both GSK3␣ and GSK3␤ associate with microtubules (32). To compare the relative amounts of the two GSK3 isoforms in microtubules, we subjected microtubules in a brain extract to four cycles of temperature-induced microtubule assembly and disassembly. Assembled microtubules and proteins that associated with microtubules after each cycle were recovered as a pellet by centrifugation, while unassembled microtubules and proteins that did not associate with microtubules remained in the supernatant. We immunoblotted various microtubule fractions by using two different monoclonal anti-GSK3␣/␤ antibodies that equally cross-react with GSK3␣ and GSK3␤: one from Santa Cruz Biotechnology and another from Chemicon International (Temecula, CA). Both Santa Cruz Biotechnology (Fig. 8A) and Chemicon International (data not shown) antibodies detected two immunoreactive bands with sizes ϳ51 and ϳ47 kDa. We confirmed the ϳ51and the 47-kDa bands as GSK3␣ and GSK3␤, respectively, by immunoblot analyses of various microtubule fractions by using antibodies directed to GSK3␣ and GSK3␤ (data not shown). Blot band intensity quantitation indicated that the ratio of GSK3␤/ GSK3␣ was ϳ0.95, ϳ0.97, ϳ0.99, ϳ0.97, and ϳ1.01 in brain extract (Ex), microtubule pellet after the first cycle (P1), microtubule pellet after the second cycle (P2), microtubule pellet after the third cycle (P3), and microtubule pellet after the fourth cycle (P4), respectively (Fig. 8B). These data indicate that GSK3␣ and GSK3␤ are present in brain extract and various microtubule pellets in almost equal amounts.
In Figs. 6 and 7, we demonstrated that GSK3␤ binds to tau. Since we found that GSK3␣ is also present in microtubule fractions, we wished to examine how tau interacts with GSK3␣ as compared with GSK3␤. To do this, we immunoprecipitated tau from brain extract and MAP fraction using an anti-tau antibody. The anti-tau immune complex was then immunoblotted by using an anti-GSK3␣/␤ antibody. As shown in Fig.  9A, GSK3␣ and GSK3␤ were detected in the anti-tau immune complex from both the MAP fraction and brain extract. Blot band intensity quantitation revealed that the GSK3␤/GSK3␣ ratio in MAP fraction and brain extract was 1.05 and 0.85, respectively (Fig. 9B). However, the GSK3␤/GSK3␣ ratio within the anti-tau immune complex from MAP fraction and brain extract was 6.3 and 5.5, respectively. These data indicate that GSK3␤ co-immunoprecipitates with tau from the MAP fraction and brain extract ϳ6-fold more than GSK3␣. Thus ϳ6-fold more tau is complexed with GSK3␤ than with GSK3␣ in MAP fraction and brain extract. DISCUSSION To investigate the biochemical nature of microtubule-associated GSK3␤ and to study its interaction with tau, we dissociated microtubules by cold incubation and removed tubulin from the MAP fraction containing GSK3␤ by phosphocellulose chromatography. When the MAP fraction was subjected to an FPLC gel filtration analysis, GSK3␤ eluted as a ϳ400-kDa complex FIG. 6. Immunoblot analysis with phosphorylation-dependent anti-tau monoclonal antibodies and GST pull-down assay. A, immunoblot analysis. Indicated fractions (10 l each) were immunoblotted with the indicated monoclonal antibodies. B, GST pull-down assay. Glutathione-agarose beads coated with GST or GST-GSK3␤ were mixed with either R-tau or phosphorylated R-tau, washed, and immunoblotted by using an anti-tau polyclonal antibody. Phosphorylated R-tau was prepared as described (35)  ( Fig. 1). When the ϳ400-kDa complex was chromatographed through an anti-GSK3␤ immunoaffinity column, tau co-eluted with GSK3␤ (Fig. 2). Similarly, tau and GSK3␤ co-immunoprecipitated with each other from fractions containing the ϳ400-kDa complex (Fig. 3). In vitro, GSK3␤ bound to R-tau (Fig. 4). Taken together, these data indicate that GSK3␤ and tau are the components of a ϳ400-kDa microtubule-associated complex, and these two proteins are directly bound to each other within the complex.
The sum of the molecular masses of GSK3␤ and tau is ϳ97-kDa, whereas the gel filtration size of the complex containing GSK3␤ and tau is ϳ400 kDa (Fig. 1A). The proteinstained SDS gel ( Fig. 2A) indicates that the fractions eluting from the anti-GSK3␤ immunoaffinity column contain ϳ2-3fold more GSK3␤ than tau. It is possible that there may be multiple GSK3␤-binding regions within the projection domain of tau, and a single tau molecule may bind 2-3 GSK3␤ molecules. The 400-kDa complex may therefore contain a tau molecule bound to several GSK3␤ molecules. Alternatively, from a gel filtration column, GSK3␤ elutes within fractions 44 -64 (Fig. 1A), with sizes ranging from 50 to 500 kDa. These data indicate that bovine brain extract contains free GSK3␤ as well as GSK3␤ complexed with tau. Since we used gel filtration fractions 44 -64 to generate Fig. 2, the relatively higher amount of GSK3␤ compared with tau in fractions eluting from an anti-GSK3␤ immunoaffinity column may be due to the presence of the GSK3␤⅐tau complex along with free GSK3␤ in these fractions. The 400-kDa complex may therefore contain tau and GSK3␤ with a 1:1 stoichiometry. The 400-kDa size of the GSK3␤⅐tau complex may be attributed to the asymmetric nature of tau within the complex, which is known to elute as a relatively large size species from a gel filtration column (35). Finally, a third possibility that cannot be excluded is that the 400-kDa complex may also contain protein(s) other than tau and GSK3␤ that may be below the detection limit of silver staining as shown in Fig. 2A. It has been reported that Cdk5 and protein phosphatase 1, which are also present in neuronal microtubules, bind to tau and associate with microtubules via tau in a manner similar to GSK3␤ (33,42). Therefore, besides tau and GSK3␤, the 400-kDa complex may also contain Cdk5, protein phosphatase 1, and/or any other tau-binding protein(s).
GSK3␤ co-purifies with microtubules during temperatureinduced microtubule assembly/disassembly and is an integral part of neuronal microtubules (30,32). The biochemical basis of the association of GSK3␤ to microtubules was unknown. In this study by microtubule sedimentation assay, we demonstrated that GSK3␤ binds to microtubules through tau. We also determined that GSK3␤ binds to the projection domain of tau. Since the microtubule-binding region of tau does not overlap with the projection domain (43), it is likely that tau can simultaneously binds to microtubules and GSK3␤ in vitro (Fig. 7) and perhaps in vivo.
Tau isolated from brain extract is heterogeneous due to the presence of several tau isoforms, which are endogenously phosphorylated on various sites (5,6). We find that tau within the 400-kDa complex is also heterogeneous (Figs. 2, 3A, and 6A). We analyzed the phosphorylation state of tau within the 400-kDa complex by using four anti-tau monoclonal antibodies sensitive to tau phosphorylation: Tau-1, 12E8, PHF-1, and AD-2. Tau-1 cross-reacts with tau that is not phosphorylated on residues Ser 199 , Ser 198 , and Ser 202 (38). 12E8 recognizes tau phosphorylated on Ser 262 (39). PHF-1 cross-reacts with tau phosphorylated on Ser 398 and/or Ser 400 (40), whereas AD-2 recognizes tau phosphorylated on Ser 396 and/or Ser 404 (41). We find that although brain extract contains tau that is immunoreactive to all of the above antibodies, only Tau-1 cross-reacts with the tau that elutes from the anti-GSK3␤ immunoaffinity column (Fig. 6A). These data indicate that tau isoforms within GSK3␤⅐tau complex are not phosphorylated on Ser 198 , Ser 199 , Ser 202 , Ser 262 , Ser 398 , Ser 400 , and Ser 404 and suggest that FIG. 8. Immunoblotting of various microtubule fractions with anti-GSK3␣/␤ antibody. Microtubules in a fresh bovine brain extract were subjected to four cycles of temperature-induced assembly and disassembly as described (31). After each cycle, the assembled microtubules were recovered as the pellet by centrifugation, and the supernatant containing unassembled microtubules and proteins that did not associate with microtubules were present in supernatant. A, immunoblot. Various microtubule fractions (3 l each) were immunoblotted by using a monoclonal anti-GSK3␣/␤ antibody that cross-reacts equally with GSK3␣ and GSK3␤. B, GSK3␤/GSK3␣ ratio. Blot A was scanned, and band intensity values were used to calculate the GSK3␤/GSK3␣ ratio. The ratio for the indicated fraction was determined by dividing the GSK3␤ band intensity value in a lane containing the indicated fraction by the GSK3␣ band intensity value in the same lane. Values are the average of three independent determinations Ϯ S.E. Ex, brain extract; P1, microtubule pellet after first cycle; S1, supernatant after first cycle; P2, microtubule pellet after second cycle; S2, supernatant after second cycle; P3, microtubule pellet after third cycle; S3, supernatant after third cycle; P4, microtubule pellet after fourth cycle.
FIG. 9. Co-immunoprecipitation of GSK3␣ and GSK3␤ with tau. Tau was immunoprecipitated (IP) from either MAP fraction or brain extract. Each resulting anti-tau immune complex was immunoblotted with an anti-GSK3␣/␤ monoclonal antibody. A, immunoblot; B, GSK3␤/GSK3␣ ratio. Blot A was scanned, and band intensity values were used to calculate the GSK3␤/GSK3␣ ratio in the indicated samples as in Fig. 8B. Values are an average of three determinations Ϯ S.E. GSK3␤-bound tau isoforms may be in nonphosphorylated states. However in vivo, tau is also phosphorylated on several other sites (8) for which antibodies are not available, precluding more detailed analysis. Thus, we conclude that within the GSK3␤⅐tau complex, tau isoforms are not phosphorylated on above indicated sites but may or may not be phosphorylated on any additional sites.
Interestingly, some of the target sites of GSK3 are generated by phosphorylation through the action of another kinase. For example, the phosphorylation of glycogen synthase by GSK3 requires prior phosphorylation of glycogen synthase by casein kinase 2 (44,45). The phosphorylation of glycogen synthase by casein kinase 2 generates a recognition sequence for GSK3. Subsequently, GSK3 phosphorylates and inactivates glycogen synthase (44,45). Likewise, tau purified from bovine brain extract containing endogenous phosphate is efficiently phosphorylated by GSK3␤. However, purified brain extract tau that has been dephosphorylated by a phosphatase is a poor GSK3␤ substrate (46). This dephosphorylated tau is then robustly phosphorylated by GSK3␤ after being previously phosphorylated by Cdk5 (46). It has been suggested that Cdk5 primes tau for GSK3␤ action (23,46).
The presence of tau not phosphorylated on Ser 198 , Ser 199 , Ser 202 , Ser 262 , Ser 398 , Ser 400 , and Ser 404 within the tau⅐GSK3␤ complex (Fig. 6A) suggests that GSK3␤-tau interaction may be regulated by phosphorylation of tau on these sites. Since Cdk5 phosphorylates Ser 199 , Ser 202 , Ser 235 , Ser 398 , and Ser 404 in vitro (34,47), we evaluated the binding of Cdk5-phosphorylated R-tau with GSK3␤ and find that GSK3␤ binds to R-tau but not Cdk5-phosphorylated R-tau (Fig. 6B). Previous studies have shown that Cdk5 is also a component of neuronal microtubules, binds to tau, and associates with microtubules via tau (33) in a manner similar to GSK3␤ (this study). Taken together, these observations suggest that GSK3␤-tau binding may be regulated by phosphorylation of tau by Cdk5. This in turn may suggest that Cdk5 may not only prime tau for GSK3␤ action (23,46) but may also regulate tau-GSK3␤ binding in vitro and possibly in vivo. It should also be noted that Cdk5-phosphorylated tau does not bind to microtubules (37). GSK3␤-tau and tau-microtubule interactions may therefore be regulated by phosphorylation on the same tau sites. It is possible that, in the brain, nonphosphorylated tau simultaneously binds to GSK3␤ and microtubules, bridges GSK3␤ to microtubules, and stabilizes microtubule structure. Upon phosphorylation, tau dissociates from both microtubules and GSK3␤.
As reported previously (32), we observed that GSK3␣ and GSK3␤ are stably associated with microtubules in bovine brain extract (Fig. 8A). Despite the amounts of GSK3␣ and GSK3␤ in brain extract and MAP fraction being almost equal, ϳ6-fold more GSK3␤ co-immunoprecipitates with tau from brain extract and MAP fraction than GSK3␣ (Fig. 9B). These data indicate that ϳ6-fold more tau is complexed with GSK3␤ than with GSK3␣ in brain. With such a profound difference between the amounts of each kinase complexed with tau, it is very likely that most of tau in the brain will be phosphorylated by GSK3␤ rather than by GSK3␣.
Full elucidation of the biochemical basis and physiological significance of the presence of a significantly higher amount of tau complexed with GSK3␤ compared with GSK3␣ in the brain extract will require further investigation. It is known that both GSK3␤ and GSK3␣ contain a N-terminal regulatory region, a central kinase domain and a C-terminal tail (44). The central kinase domains of these two kinases are ϳ98% identical, whereas N-terminal regulatory regions and C-terminal tails are quite different (11). It is possible that the tau-binding region may be located within either the N-terminal regulatory region or the C-terminal tail of GSK3␤ and therefore tau may bind to GSK3␤ with greater affinity than to GSK3␣.