Phosphorylation of Axin, a Wnt Signal Negative Regulator, by Glycogen Synthase Kinase-3β Regulates Its Stability*

Axin forms a complex with glycogen synthase kinase-3β (GSK-3β) and β-catenin and promotes GSK-3β-dependent phosphorylation of β-catenin, thereby stimulating the degradation of β-catenin. Because GSK-3β also phosphorylates Axin in the complex, the physiological significance of the phosphorylation of Axin was examined. Treatment of COS cells with LiCl, a GSK-3β inhibitor, and okadaic acid, a protein phosphatase inhibitor, decreased and increased, respectively, the cellular protein level of Axin. Pulse-chase analyses showed that the phosphorylated form of Axin was more stable than the unphosphorylated form and that an Axin mutant, in which the possible phosphorylation sites for GSK-3β were mutated, exhibited a shorter half-life than wild type Axin. Dvl-1, which was genetically shown to function upstream of GSK-3β, inhibited the phosphorylation of Axin by GSK-3β in vitro. Furthermore, Wnt-3a-containing conditioned medium down-regulated Axin and accumulated β-catenin in L cells and expression of Dvl-1ΔPDZ, in which the PDZ domain was deleted, suppressed this action of Wnt-3a. These results suggest that the phosphorylation of Axin is important for the regulation of its stability and that Wnt down-regulates Axin through Dvl.

Genetic and biochemical analyses have revealed that there are components that are structurally and functionally conserved in the Wnt signaling pathway among flies, frogs, and mammals (1)(2)(3). In mammals these include Wnt, frizzled, Dvl, GSK-3␤, 1 ␤-catenin, and Lef/Tcf, which are homologous to the Drosophila proteins Wg, Dfz2, Dsh (Dishevelled), Shaggy, Armadillo, and Pangolin, respectively. The current model for the Wnt signaling pathway proposes that in the absence of Wnt, GSK-3␤ phosphorylates ␤-catenin, resulting in the degradation of ␤-catenin. In response to Wnt, Dvl antagonizes GSK-3␤ activity through an as yet unknown mechanism. This leads to the stabilization and the accumulation of ␤-catenin. The accumulated ␤-catenin translocates to the nucleus, associates with the transcriptional enhancers of the Lef/Tcf family (4 -6), and stimulates gene expression such as Myc (7).
Axin was originally identified as a product of mouse fused gene (8). fused carries recessive mutations that are lethal and that cause a duplication of the embryonic axis (9,10). Injection of Axin into Xenopus embryos causes strong axis defects, and coexpression of Axin inhibits the Xwnt8-dependent axis duplication (8). Thus, Axin is a negative regulator of the Wnt signaling pathway and inhibits axis formation. We have identified rat Axin (rAxin) and its homolog, Axil (for Axin-like), as GSK-3␤-interacting proteins (11,12). Conductin has been identified as a ␤-catenin-binding protein (13) and is identical to Axil. We have found that both Axin and Axil bind not only to GSK-3␤ but also to ␤-catenin and that they promote GSK-3␤-dependent phosphorylation of ␤-catenin (11,12). We have also shown that the regulators of G protein signaling (RGS) domain of rAxin directly interacts with APC and that expression of rAxin in COS and SW480 cells stimulates the degradation of ␤-catenin (14,15). Other groups have reported similar results (13, 16 -19). Therefore, it appears that Axin family members downregulate ␤-catenin.
Axin enhances GSK-3␤-dependent phosphorylation of APC in addition to ␤-catenin in vitro (17), and the phosphorylation of APC increases its binding to ␤-catenin (20). Although Axin is also phosphorylated by GSK-3␤ directly, the phosphorylation of Axin does not affect its binding to GSK-3␤ and ␤-catenin in vitro (11). These results indicate that ␤-catenin, APC, and Axin form a complex with GSK-3␤ and that the phosphorylation occurs efficiently in the complex. However, the physiological significance of the phosphorylation of Axin is not known. Therefore, we examined a role of the phosphorylation of Axin in the Wnt signaling pathway. Here we demonstrate that the phosphorylation of Axin by GSK-3␤ regulates its stability, that Dvl inhibits the GSK-3␤-dependent phosphorylation of Axin, and that Wnt-3a down-regulates Axin through Dvl. thionine, and [ 35 S]cysteine were purchased from Amersham Pharmacia Biotech (Buckinghamshire, UK). Other materials were from commercial sources.
Phosphorylation of Axin in Intact Cells-COS cells expressing Myc-rAxin or Myc-rAxin 322/326/330A (35-mm-diameter dish) were metabolically labeled with 32 P i (100 Ci/ml) in phosphate-free RPMI for 12 h in the presence or absence of 30 mM LiCl or 100 nM okadaic acid. The cells were lysed, and the lysates were immunoprecipitated with the anti-Myc antibody (11). The immunoprecipitates were probed with the anti-Myc antibody and subjected to autoradiography.
Pulse-Chase Analysis-COS cells (35-mm-diameter dish) were transfected with pBJ-Myc/rAxin or pBJ/Myc-rAxin 322/326/330A . After 48 h, pulsechase analysis was performed as described (14,25). Briefly, the cells were pulse-labeled with [ 35 S]methionine and [ 35 S]cysteine (50 Ci/ml) for 1 h at 37°C. Then the cells were lysed immediately or at the indicated times following incubation with excess unlabeled methionine and cysteine in the presence or absence of 30 mM LiCl or 100 nM okadaic acid. The lysates were immunoprecipitated with the anti-Myc antibody, the precipitates were subjected to autoradiography, and then the densities of the labeled proteins were analyzed with a Fuji BAS 2000 image analyzer.
Down-regulation of Axin by Wnt-3a-Confluent wild type L cells or L cells expressing HA-Dvl-1 ⌬PDZ (35-mm-diameter dish) were washed with Dulbecco's modified essential medium twice, and the indicated volume of Wnt-3a-conditioned medium, which was adjusted to a total volume of 700 l with Dulbecco's modified essential medium, was added to the cells. After stimulation for 6 h, the cells were lysed in 100 l of lysis buffer (11), and the lysates (20 g of protein) were probed with the anti-Axin and anti-␤-catenin antibodies.

RESULTS AND DISCUSSION
Stabilization of Axin by Phosphorylation-rAxin was phosphorylated by GSK-3␤ directly in vitro, and SANDSEQQS 330 of rAxin was one of the phosphorylation sites for GSK-3␤ (11). First we examined whether rAxin is phosphorylated by GSK-3␤ in intact cells. Myc-rAxin was phosphorylated when COS cells were metabolically labeled with 32 P i (Fig. 1A). We tried to express Myc-rAxin 322/326/330A , in which Ser 322 , Ser 326 , and Ser 330 were mutated to Ala, in COS cells, but its protein level was lower than that of Myc-rAxin (wild type) (Fig. 1A, lanes 3 and 4). Consistent with the protein level, the phosphorylation and apparent molecular weight of Myc-rAxin 322/326/330A were reduced in comparison with Myc-rAxin (Fig. 1A, lanes 1 and 2). Therefore, we used LiCl, which is known to be an inhibitor of GSK-3␤ (26,27). It appeared that treatment of COS cells with LiCl decreased the phosphorylation of Myc-rAxin, whereas okadaic acid, a protein phosphatase 1 or 2A inhibitor, increased it (Fig. 1A, lanes 5-7). However, these changes by LiCl and okadaic acid were also correlated with the protein level of Myc-rAxin (Fig. 1A, lanes 8 -10). LiCl decreased the protein level of Myc-rAxin in a dose-dependent manner (Fig. 1B). Consistent with the previous observations (28), treatment of COS cells with LiCl resulted in the cytoplasmic accumulation of ␤-catenin (Fig. 1B). Okadaic acid prevented the decrease of Myc-rAxin by LiCl (Fig. 1B). LiCl did not affect the protein level of transfected Myc-RalBP1, an effector protein of small GTP-binding protein Ral (29) or endogenous GSK-3␤ (Fig. 1C). Therefore, the effect of LiCl that reduces rAxin is not nonspecific. These results suggest that the phosphorylation of rAxin is correlated with its stability.
To investigate the stability of Axin by phosphorylation further, pulse-chase analysis was performed. Pulse-labeled Myc-rAxin in COS cells migrated slowly on SDS-polyacrylamide gel electrophoresis in a time-dependent manner ( Fig. 2A), suggesting that Myc-rAxin was phosphorylated. Pulse-labeled Myc-rAxin did not exhibit a gel band shift and disappeared at 12 h in COS cells treated with LiCl ( Fig. 2A). In contrast, okadaic acid enhanced the band shift and prevented the decay of pulselabeled Myc-rAxin at 12 h ( Fig. 2A). Pulse-labeled Myc-rAxin decreased gradually with a half-life of approximately 8 h, and pulse-labeled Myc-rAxin 322/326/330A exhibited a shorter half-life (Fig. 2B). These results indicate that Axin is phosphorylated by GSK-3␤ in intact cells and that the phosphorylated form is more stable than the unphosphorylated form.
Inhibition of GSK-3␤-dependent Phosphorylation of Axin by Dvl-Drosophila Dsh encodes a cytoplasmic protein of unknown biochemical function in the Wg signaling pathway (1-3). In mammals, dvl-1, -2, and -3 genes have been isolated as homologs of Dsh (21,30,31). It has been shown that Dsh antagonizes shaggy, a fly homolog of GSK-3␤, in the Wg signaling pathway (2,3), and that overexpression of Dvl-1 in Chinese hamster ovary cells inhibits GSK-3 activity as measured by the GSK-3-mediated phosphorylation of tau proteins (32). However, little is known about the biochemical pathway leading from Dvl to GSK-3␤. Therefore, we examined whether Dvl-1 affects the phosphorylation of rAxin by GSK-3␤ in vitro. MBP-Dvl-1 itself was not phosphorylated by GST-GSK-3␤ (data not shown). GST-GSK-3␤ phosphorylated MBP-rAxin in a time-dependent manner (11) (Fig. 3A). MBP-Dvl-1 inhibited this phosphorylation of MBP-rAxin (Fig. 3A). This inhibitory activity of MBP-Dvl-1 was dose-dependent, and MBP alone did not inhibit the GST-GSK-3␤-dependent phosphorylation of MBP-rAxin (Fig. 3B). Dvl has the PDZ domain, and disruption of the PDZ domain abolishes its activity in the Wg-Armadillo pathway and in the Xenopus axis induction assay (33,34). Deletion of the PDZ domain from Dvl-1 (MBP-Dvl-1 ⌬PDZ ) greatly reduced its activity to inhibit the phosphorylation of MBP-rAxin by GST-GSK-3␤ (Fig. 3B). Inhibition of the phosphorylation of MBP-rAxin by MBP-Dvl-1 was not recovered even though the amounts of MBP-rAxin increased (Fig. 3C). Lineweaver-Burk plots indicated that the K m and V max values of MBP-rAxin for GST-GSK-3␤ in the absence of MBP-Dvl-1 were 131 nM and 4.3 nmol/min/mg, respectively, and that those in the presence of MBP-Dvl-1 were 129 nM and 2.5 nmol/min/mg (Fig.   3C). These results suggest that Dvl-1 inhibits the GSK-3␤-dependent phosphorylation of Axin in a noncompetitive manner. This is the first demonstration showing that Dvl inhibits the function of GSK-3␤ directly. However, it is not likely that Dvl-1 inhibits GSK-3␤ activity itself, because MBP-Dvl-1 did not affect the phosphorylation of synthetic peptide substrate, which is designed from glycogen synthase, by GST-GSK-3␤ (data not shown). We have recently found that Dvl-1 directly binds to Axin and that the binding of Dvl-1 to Axin does not affect the interaction of GSK-3␤ with Axin. 2 It is possible that the binding of Dvl to Axin induces the structural change of the Axin complex; therefore GSK-3␤ does not effectively phosphorylate Axin. However, higher concentrations (M order) of Dvl-1 are required to inhibit the GSK-3␤-dependent phosphorylation of Axin in our in vitro experiments. Therefore, modification of Dvl such as phosphorylation could be necessary to act on the Axin complex in intact cells. These results suggest that Dvl may regulate the stability of Axin.
Down-regulation of Axin by Wnt-3a-Finally we examined whether Wnt signal regulates the stability of endogenous Axin in intact cells. Although Wnt proteins are secretory, they predominantly bind to the cell surface or extracellular matrix. Small amounts of biologically active Wnt-1 or Wg can be found in culture medium conditioned by cells expressing these proteins (35,36). The Wg-conditioned medium from Schneider cells increases the level of Armadillo in Drosophila disc cells and inactivates GSK-3 in 10T1/2 fibroblasts (35,37). Based on assays carried out with mammalian cell lines and Xenopus embryos, the Wnt proteins can be classified into two groups, Wnt-1 and Wnt-5a classes (38 -40). The Wnt-1 class includes Wnt-1, Wnt-2, Wnt-3, Wnt-3a, and Wnt-8, which have activities to transform the cells and to accumulate cytoplasmic ␤-catenin, whereas the Wnt-5a class includes Wnt-4, Wnt-5a, Wnt-5b, Wnt-7b, and Wnt-11, which do not exhibit the transformation and ␤-catenin accumulation activities. Because Wnt-3a displays characteristics similar to those of Wnt-1, we prepared Wnt-3a-containing conditioned medium. In these experiments we used mouse fibroblast L cells, because the changes in the expression level of ␤-catenin by Wnt are easily observed due to little expression of cadherin in the cells (15,23,41). Furthermore, Western blot analyses with the anti-Axin antibody demonstrated that Axin is most abundant in L cells among various cell lines including SW480, NIH3T3, COS, and Chinese hamster ovary cells (data not shown). Wnt-3a conditioned medium induced the accumulation of ␤-catenin in L cells in a dose-dependent manner (Fig. 4A). In contrast, Wnt-3a decreased Axin (Fig. 4A). Control conditioned medium did not affect the amounts of Axin and ␤-catenin (data not shown). To examine whether Dvl is involved in this action of Wnt-3a, we established L cells, which express HA-Dvl-1 ⌬PDZ stably. Wnt-3a-induced increase of ␤-catenin and decrease of Axin were suppressed in L cells expressing HA-Dvl-1 ⌬PDZ (Fig. 4B). These results indicate that Wnt not only accumulates ␤-catenin but also down-regulates Axin through Dvl.
We have recently found that in COS cells Axin interacts with GSK-3␤, ␤-catenin, and APC in a high molecular mass complex with a molecular mass of more than 10 3 kDa on gel filtration column chromatography (15). In L cells ␤-catenin is present in the high molecular mass complex in the absence of Wnt-3a, whereas addition of Wnt-3a to L cells increases ␤-catenin in a low molecular mass complex with a molecular mass of 200 -300 kDa (15). In L cells expressing Axin, Wnt-3a-induced increase of ␤-catenin in the low molecular mass complex is not observed (15). These results suggest that a balance between the high and low molecular mass complexes containing ␤-catenin is closely regulated and that Axin plays a role in limiting the accumulation of ␤-catenin in the low molecular mass complex. Wnt may regulate the assembly of the complex consisting of Axin, APC, ␤-catenin, and GSK-3␤ and induce the dissociation of ␤-catenin from the complex. It is possible that ␤-catenin free from the complex is accumulated, binds to different partners such as Lef/Tcf, and transmits the signals. Our results suggest that the Wnt signal could act on the Axin complex through Dvl, resulting in the inhibition of the GSK-3␤-dependent phosphorylation of Axin and the degradation of Axin. Degradation of Axin due to hypophosphorylation may induce the dissociation of ␤-catenin from the complex by decreasing the binding of ␤-catenin to Axin. Studies to clarify the mechanism of proteolysis of Axin are under way.