Complex Formation of Adenomatous Polyposis Coli Gene Product and Axin Facilitates Glycogen Synthase Kinase-3β-dependent Phosphorylation of β-Catenin and Down-regulates β-Catenin*

Adenomatous polyposis coli gene product (APC) functions as a tumor suppressor and its mutations in familial adenomatous polyposis and colorectal cancers lead to the accumulation of cytoplasmic β-catenin. The molecular mechanism by which APC regulates the stability of β-catenin was investigated. The central region of APC, APC-(1211–2075), has the β-catenin- and Axin-binding sites and down-regulates β-catenin. Glycogen synthase kinase-3β (GSK-3β) phosphorylated β-catenin slightly in the presence of either APC-(1211–2075) or AxinΔ β -catenin, in which the β-catenin-binding site is deleted, and greatly in the presence of both proteins. The enhancement of the GSK-3β-dependent phosphorylation of β-catenin was eliminated by the APC-binding site of Axin. Axin down-regulated β-catenin in SW480 cells, but not AxinΔ β -catenin. In L cells where APC is intact, AxinΔ β -catenin inhibited Wnt-dependent accumulation of β-catenin but not Axin-(298–832)Δ β -catenin in which the APC- and β-catenin-binding sites are deleted. These results indicate that the complex formation of APC and Axin enhances the phosphorylation of β-catenin by GSK-3β, leading to the down-regulation of β-catenin.

APC 1 is a tumor suppressor linked to FAP and to the initiation of sporadic human colorectal cancer (1). APC encodes a 300-kDa multifunctional protein with several structural domains. The N terminus contains an oligomerization domain followed by seven repeats of an armadillo motif. The middle portion of APC contains three successive 15-amino acid (15-aa) repeats followed by seven related but distinct 20-aa repeats, both of which are able to bind independently to ␤-catenin (2)(3)(4). Following the 20-aa repeats is a basic region, while the C terminus has an serine/threonine-X-valine motif that interacts with the PDZ domain-containing human Dlg homolog (5). In FAP and colorectal cancers, most patients carry APC mutations, which result in the expression of truncated proteins (1). Almost all mutant proteins lack the C-terminal half, including most of the 20-aa repeats, but retain the 15-aa repeats. Colorectal cancer cells with mutant APC contain a large amount of monomeric ␤-catenin (6). Indeed, APC down-regulates the level of cytoplasmic ␤-catenin when introduced into the SW480 colorectal cancer cell line (6). This activity of APC is localized to the central region of the protein, and the region containing 20-aa repeats are necessary for the degradation of ␤-catenin (1,7). However, the role of APC in the down-regulation of ␤-catenin is not clear.
␤-Catenin was originally identified as a binding protein of cadherin (8). In addition to having roles in cell-cell adhesion, ␤-catenin is a component of the Wnt signaling pathway (9 -11). Wnt is a secreted glycoprotein. The Wnt signaling pathway is essential for developmental decisions, regulating anterior-posterior and dorsal-ventral patterns in both vertebrates and flies. In vertebrates, the Wnt signaling pathway consists of an intracellular cascade that includes frizzled, Dvl, GSK-3␤, ␤-catenin, and Lef/Tcf. In the absence of Wnt, protein kinase GSK-3␤ is active and antagonizes downstream elements of the Wnt signaling pathway through changes in the level of cytoplasmic ␤-catenin (12,13). When Wnt acts on its receptor, frizzled, the signal is transmitted to Dvl by an unknown mechanism. Then, GSK-3␤ is inactivated probably through Dvl, there is a decrease in the phosphorylation of ␤-catenin and an increase in its stability, and ␤-catenin translocates to the nucleus where it associates with and activates transcription factor Lef/Tcf (14,15), thereby leading to the gene expression including Myc, Fra, Jun, cyclin D1, and peroxisome proliferator-activated receptor ␦ (16 -19). Thus, Wnt induces the accumulation of ␤-catenin and transmits the signal to nucleus.
The mechanism by which the stability of ␤-catenin is regulated has been largely clarified. Discovery and functional analyses of Axin have provided new clues as to how the stability of ␤-catenin is regulated (20 -22). Axin was originally identified as a product of the mouse Fused locus (22). The mouse mutant Fused is recessive lethal; mutants have a duplication of the embryonic axis (23,24). We have identified rat Axin (rAxin) and its homolog, Axil (for Axin like), as GSK-3␤-interacting proteins (25,26). Conductin has been identified as a ␤-catenin-* This work was supported by grants-in-aid for scientific research (B) and for scientific research on priority areas (A) from the Ministry of Education, Science, and Culture, Japan (1998,1999), and by grants from the Yamanouchi Foundation for Research on Metabolic Disorders (1998,1999), and the Uehara Memorial Foundation (1998). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ Contributed equally to the results of this work.
binding protein (27) and is identical to Axil. Both Axin and Axil bind not only to GSK-3␤ but also to ␤-catenin (25)(26)(27)(28)(29)(30)(31) and promote GSK-3␤-dependent phosphorylation of ␤-catenin (25,26,30,32). Phosphorylated ␤-catenin forms a complex with ␤TrCP/ FWD1, a member of the F-box protein family, resulting in the degradation of ␤-catenin by ubiquitin-proteasome pathway (33)(34)(35). Indeed, Axin inhibits Wnt-dependent ␤-catenin accumulation and Tcf activation (29,36). Furthermore, the RGS domains of Axin and conductin interact directly with the region containing the third to seventh 20-aa repeats of APC (27,28,30). In this region, it is likely that a SAMP (Ser-Ala-Met-Pro) sequence is important, since mutation of SAMP to AALP (Ala-Ala-Leu-Pro) in APC abolishes its binding to conductin (27). APC is truncated at amino acids 1337 and 1427 in SW480 and DLD-1 cells, respectively, which are derived from human colon cancers. The APC mutants in these cells do not bind to Axin (28). These results suggest that the interaction of APC with Axin is important for regulating the stability of ␤-catenin. However, the roles of the complex formation of APC and Axin in the degradation of ␤-catenin have not been demonstrated. Here we show that APC promotes GSK-3␤-dependent phosphorylation of ␤-catenin by interacting with Axin, thereby down-regulating ␤-catenin.
Cell Culture-COS and L cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% calf serum and fetal calf serum (ICN Biomedicals, Inc.), respectively, at 37°C. SW480 cells were grown in RPMI 1640 supplemented with 10% fetal calf serum. L cells stably expressing various deletion mutants of Myc-rAxin were obtained as described (36). When the accumulation of ␤-catenin in response to Wnt-3a was examined in L cells, confluent cells (in 35-mm diameter dishes) were washed with Dulbecco's modified Eagle's medium twice and 0.2 ml of Wnt-3a-conditioned medium which was adjusted to a total volume of 1 ml with Dulbecco's modified Eagle's medium was added to the cells. After stimulation for 2 h, the cells were lysed in 100 l of lysis buffer (40) and the lysates (20 l) were probed with the anti-␤-catenin antibody.
Ubiquitination of ␤-Catenin in Intact Cells-To detect the ubiquitination of ␤-catenin, COS, L, and SW480 cells were treated with 25 M ALLN or 10 M lactacystin for 8 h. The lysates (5-500 g of protein) were immunoprecipitated with the anti-␤-catenin antibody and the immunoprecipitates were probed with the anti-␤-catenin antibody. Where specified, the cells transfected with pCGN/ubiquitin to express HA-ubiquitin were treated with 25 M ALLN or 10 M lactacystin for 8 h, then the lysates were immunoprecipitated with the anti-␤-catenin antibody and the immunoprecipitates were probed with the anti-HA antibody.
Phosphorylation of ␤-Catenin by GSK-3␤ in Vitro-One hundred and fifty nM GST-GSK-3␤ was incubated with 1.4 M His 6 -␤-catenin in the presence of given concentrations of APC and Axin mutants in 30 l of a kinase reaction mixture (50 mM Tris/HCl, pH 7.5, 10 mM MgCl 2 , 1 mM dithiothreitol, 10 M [␥-32 P]ATP (1,000 -3,000 cpm/pmol)) for 30 min at 30°C. After the incubation, the mixture was further incubated with cobalt beads for 1 h at 4°C, then His 6 -␤-catenin was precipitated by centrifugation. The samples were subjected to SDS-PAGE followed by autoradiography, and then the radioactivity of the phosphorylated ␤-catenin was measured.
Immunofluorescence Microscopy-SW480 cells grown on coverslips were fixed for 20 min in PBS containing 4% paraformaldehyde. The cells were washed with PBS three times, and then permeabilized with PBS containing 0.1% Triton X-100 and 2 mg/ml bovine serum albumin for 12 h. The cells were washed and incubated for 1 h with the anti-Myc and the anti-␤-catenin antibodies. After being washed with PBS, they were further incubated for 1 h with Alexa 594-labeled anti-mouse and Alexa 488-labeled anti-rabbit IgG. The nuclei were counterstained with 4Ј,6diamidine-2-phenylindole. The coverslips were washed with PBS, mounted on glass slides, and viewed with a confocal laser-scanning microscope (TCS-NT TM , Leica-laser-technik GmbH, Heidelberg, Germany).
Others-Luciferase assay was performed as described (36). Protein concentrations were determined with bovine serum albumin as a standard (41).

Failure of Ubiquitination of ␤-Catenin in SW480 Cells-It
has been demonstrated that the ubiquitin-proteasome pathway degrades ␤-catenin and that the phosphorylation of ␤-catenin by GSK-3␤ is required for its ubiquitination (33,34). Since ␤-catenin is accumulated in SW480 cells, human colon cancer cell lines where the C-terminal half of APC is truncated, we first examined whether the ubiquitination of ␤-catenin occurs in the cells. On treatment with proteasome inhibitors such as ALLN or lactacystin, cells exhibit slowly migrating forms of ␤-catenin and the accumulation of the protein (33). Consistent with these observations, treatment of COS and L cells with ALLN or lactacystin, respectively, resulted in the appearance of slowly migrating forms of ␤-catenin and the accumulation of the protein (Fig. 1A, lanes 1-4). When ␤-catenin was immunoprecipitated from COS cells expressing HA-ubiquitin and treated with ALLN, HA-positive high molecular weight proteins were detected in the ␤-catenin immunoprecipitates (Fig.  1B, lanes 1 and 2). Thus, ␤-catenin is polyubiquitinated in COS and L cells. However, SW480 cells exhibited neither slowly migrated forms of ␤-catenin nor HA-positive high molecular weight proteins when treated with ALLN or lactacystin (Fig.  1A, lanes 5-7, and 1B, lanes 3-5). Ubiquitination of various proteins in SW480 cells was observed when the cells expressing HA-ubiquitin were treated with lactacystin (Fig. 1C). These results suggest that ␤-catenin is not ubiquitinated in SW480 cells although as other proteins. Since the ubiquitination of ␤-catenin requires its phosphorylation by GSK-3␤ (33, 34), we speculated that APC plays an important role in the phosphorylation of ␤-catenin and that the truncated form of APC loses its activity.
APC is intact and binds to Axin in L cells (28). Wnt-3aconditioned medium induced the accumulation of ␤-catenin in L cells and expression of Myc-rAxin in this cell line inhibited it (36) (Fig. 9A, lanes 1-4). Expression of Myc-rAxin ⌬GSK-3␤ induced the accumulation of ␤-catenin in the absence of Wnt-3a and did not suppress the accumulation of ␤-catenin by Wnt-3a (Fig. 9A, lanes 5 and 6). This indicates that Axin ⌬GSK-3␤ acts as a dominant negative form. Interestingly, in contrast to SW480 cells, expression of Myc-rAxin ⌬␤-catenin in L cells suppressed the Wnt-3a-dependent accumulation of ␤-catenin (Fig. 9A, lanes 7  and 8). However, on deletion of the region containing the RGS domain (1-297 amino acids) from rAxin ⌬␤-catenin (rAxin-(298 -832) ⌬␤-catenin ), the activity was abolished (Fig. 9A, lanes 9 and  10) although Myc-rAxin-(298 -832) still kept the ability to suppress the Wnt-3a-dependent accumulation of ␤-catenin (Fig.  9A, lanes 11 and 12). These rAxin mutants were expressed with the similar levels as full-length rAxin (data not shown). The endogenous levels of GSK-3␤ were not changed in these L cells (Fig. 9A). Similar observations were obtained with three independent clones of each cell line. Consistent with these observations, basal activity of Tcf-4 was increased in L cells expressing Myc-rAxin ⌬GSK-3␤ (Fig. 9B, lanes 5 and 6). The activity of Tcf-4 in this L cell mutant was higher than that in wild-type L cells stimulated with Wnt-3a although the expression levels of ␤-catenin were similar. The reason is not known at present. Wnt-3a-dependent activation of Tcf-4 was suppressed in L cells expressing Myc-rAxin ⌬␤-catenin as well as in the cells expressing Myc-rAxin (full length) and Myc-rAxin-(298 -832) but not in the cells expressing rAxin-(298 -832) ⌬␤-catenin (Fig. 9B, lanes  7-12). These results suggest that the binding of ␤-catenin to Axin is not essential for the down-regulation of ␤-catenin and that the interaction of APC and GSK-3␤ with Axin is necessary. The arrows indicate the cells expressing Myc-rAxin and its mutants. Myc-rAxin and ␤-catenin were observed in red and green, respectively. The nuclei were shown as colored in blue by 4Ј,6-diamidine-2-phenylindole (DAPI) (G). Note that cytoplasmic and nuclear ␤-catenin disappear in SW480 cells expressing full-length rAxin (D), in contrast, the staining pattern of ␤-catenin in the cells expressing rAxin ⌬␤-catenin is essentially the same as that in the nontransfected cells (E). The results shown are representative of three independent experiments. DISCUSSION We have shown that APC promotes the GSK-3␤-dependent phosphorylation of ␤-catenin by interacting with Axin. To clarify the role of APC in the phosphorylation of ␤-catenin, we have made two Axin mutants, Axin ⌬␤-catenin and Axin ⌬GSK-3␤ . Neither of them enhances the phosphorylation of ␤-catenin by GSK-3␤ as compared with full-length Axin. These results indicate that the binding of both proteins to Axin is necessary for the efficient phosphorylation of ␤-catenin. APC-(1211-2075) is able to down-regulate ␤-catenin in intact cells but not enhance the phosphorylation of ␤-catenin by GSK-3␤ in vitro. The combination of APC-(1211-2075) and Axin ⌬␤-catenin but not Axin ⌬GSK-3␤ enhances the phosphorylation of ␤-catenin greatly. Since APC-(1211-2075) binds to both Axin and ␤-catenin but not to GSK-3␤, these results indicate that GSK-3␤ bound to Axin can phosphorylate ␤-catenin bound to APC in the Axin⅐APC complex. These findings are supported by the observations that the combination of APC-(959 -1338), which has the ␤-catenin-binding site but not the Axin-binding site, and Axin ⌬␤-catenin does not enhance the phosphorylation of ␤-catenin and that the RGS domain of Axin inhibits the phosphorylation of ␤-catenin enhanced by the combination of APC-(1211-2075) and Axin ⌬␤-catenin . Furthermore, APC forms a complex with GSK-3␤ when Axin is present. When Axin ⌬GSK-3␤ is present, APC does not. Therefore, APC can form a complex with GSK-3␤ through Axin. It is likely that APC exerts its role in the GSK-3␤-dependent phosphorylation of ␤-catenin by interacting with Axin.
Axin ⌬GSK-3␤ does not down-regulate ␤-catenin in SW480 or L cells. This indicates that the binding of GSK-3␤ to Axin is essential for its activity to degrade ␤-catenin, consistent with the observation that Axin ⌬GSK-3␤ does not enhance the phosphorylation of ␤-catenin even in the presence of APC-(1211-2075). Although Axin ⌬␤-catenin does not down-regulate ␤catenin in SW480 cells, it suppresses Wnt-3a-dependent accumulation of ␤-catenin in L cells. Interestingly, Axin-(298 -832) ⌬␤-catenin does not suppress Wnt-3a-dependent accumulation of ␤-catenin in L cells. Since APC is intact and binds to Axin in L cells (28), Axin ⌬␤-catenin forms a complex with ␤-catenin probably through APC in the cells. These results suggest that Axin ⌬␤-catenin can down-regulate ␤-catenin by associating with APC in L cells and that the binding of ␤-catenin to Axin may not be essential for its activity to degrade ␤-catenin when wild-type APC is present. Thus, it is likely that GSK-3␤ phosphorylates ␤-catenin efficiently in the Axin⅐APC complex, and ␤-catenin is thereby degraded in the cells. It has been demonstrated that APC enhances the binding of ␤-catenin to Axin in Xenopus extracts and that the RGS domain inhibits the Axin activity to degrade ␤-catenin in a cell-free system (47). Furthermore, it has been also shown that disruption of either the Axinor ␤-catenin-binding site of APC abolishes its activity to degrade ␤-catenin in SW480 cells (42). These results suggest that APC requires the interaction with Axin and ␤-catenin to downregulate ␤-catenin and are consistent with our findings. As observed in L cells, Axin ⌬GSK-3␤ acts as a dominant negative form. Therefore, the binding to APC and GSK-3␤ is important for the action of Axin. These are consistent with the observation that the Axin mutants containing both the RGS domain and the GSK-3␤-binding site can decrease the expression level of ␤-catenin in Xenopus embryos (45). Since Axin may work as a tumor suppressor gene product, the sequence analyses of the RGS domain and the GSK-3␤-binding site of Axin in human cancers would be necessary to clarify the role of Axin in tumorgenesis. Indeed, mutations in these two regions of Axin have been found in human hepatocarcinoma (48).
From the observations that expression of Axin and Axin-(298 -832) in SW480 cells down-regulates ␤-catenin (30,36), it has been thought that Axin does not require APC for the degradation of ␤-catenin. However, mRNA of Axin was intact and its protein expression was observed in SW480 cells (data not shown). Therefore, endogenous Axin alone is not sufficient for the down-regulation of ␤-catenin, but overexpressed Axin may be able to down-regulate it. Three binding sites of Axin on APC are identified (27) and these sites are lost in almost all FAP and colorectal cancers (1,49). Our results together with the previous observations provide evidence that the complex  lanes 1 and 2), and L cells stably expressing Myc-rAxin (full-length) (lanes 3 and 4), Myc-rAxin ⌬GSK-3␤ (lanes 5 and 6), Myc-rAxin ⌬␤-catenin (lanes 7 and 8), Myc-rAxin-(298 -832) ⌬␤-catenin (lanes 9 and 10), or Myc-rAxin-(298 -832) (lanes 11 and 12) were treated with Wnt-3a-conditioned medium (lanes 2, 4, 6, 8, 10, and 12) or control medium (lanes 1, 3, 5, 7, 9, and 11) for 2 h. An aliquot of the lysates was probed with the anti-␤-catenin and anti-GSK-3␤ antibodies. The amounts of ␤-catenin were analyzed with NIH image and expressed as the fold increase as compared with the level observed in wild-type L cells treated with control medium. The results shown in the upper panel are the mean Ϯ S.E. of three independent experiments. The results shown in the middle and lower panels are representative pictures of three independent experiments. The arrow and arrowhead indicate ␤-catenin and GSK-3␤, respectively. B, effects of the Axin mutants on Wnt-3a-induced activation of Tcf-4. After wild-type L cells (WT) (lanes 1 and 2), and L cells stably expressing Myc-rAxin (full-length) (lanes 3 and 4), Myc-rAxin ⌬GSK-3␤ (lanes 5 and 6), Myc-rAxin ⌬␤-catenin (lanes 7 and 8), Myc-rAxin-(298 -832) ⌬␤-catenin (lanes 9 and 10), or Myc-rAxin-(298 -832) (lanes 11 and 12) were transfected with pEF-BOS-HA/hTcf-4E and pTOPFLASH, they were treated with Wnt-3a-conditioned medium (lanes 2, 4, 6, 8, 10, and 12) or control medium (lanes 1, 3, 5, 7, 9, and 11). The luciferase activity was assayed and expressed as the fold increase as compared with the level observed in wild-type L cells expressing pEF-BOS-HA/hTcf-4E and pTOPFLASH treated with control medium. The results shown are the mean Ϯ S.E. of four independent experiments. formation of APC and Axin is important for the regulation of the stability of ␤-catenin. APC has multiple binding sites of ␤-catenin in 15-and 20-aa repeats, while Axin has a single binding site (1,20,21). APC does not bind to GSK-3␤, while Axin does. GSK-3␤ also phosphorylates APC directly (50), and the binding of APC to Axin enhances the GSK-3␤-dependent phosphorylation (30,37). The phosphorylation of APC by GSK-3␤ increases its binding activity to ␤-catenin (50). This might allow proper presentation of ␤-catenin to GSK-3␤, resulting in that GSK-3␤ bound to Axin may phosphorylate multiple ␤-catenins bound to APC efficiently in the Axin⅐APC complex. Indeed, APC promotes the GSK-3␤-dependent phosphorylation of ␤-catenin in the presence of low but not high concentrations of Axin. The reason why endogenous Axin alone in SW480 cells does not down-regulate ␤-catenin may be that its binding to endogenous Axin is too small to reduce cytoplasmic ␤-catenin. Since it appears that both intact APC and Axin are necessary for keeping the level of cytoplasmic ␤-catenin low in physiological conditions, it is possible that mutations of either APC or Axin induce the accumulation of ␤-catenin, thereby promoting cancer.
We have proposed that Axin acts as a scaffold protein since it forms a complex with protein kinase, its substrates, and protein phosphatase (20,21). Axin itself is phosphorylated by GSK-3␤ and the phosphorylation stabilizes Axin in contrast to ␤-catenin (51). It has been shown that the phosphorylation of Axin increases its affinity for ␤-catenin (52,53). Thus, GSK-3␤ and its substrates, ␤-catenin, APC, and Axin itself, are simultaneously present in the Axin complex and their phosphorylation occurs efficiently in the complex. Two additional molecules, Dvl and PP2A, have been shown to inhibit the phosphorylation in the Axin complex. Dvl interacts with Axin (39,45,46) and inhibits GSK-3␤-dependent phosphorylation of ␤-catenin, APC, and Axin in the Axin complex (39,51). Xenopus Dishevelled forms a complex with Xenopus Axin-related protein, thereby displacing GSK-3 from the complex (54). These may explain the molecular mechanism by which Dvl antagonizes Axin and induces the accumulation of ␤-catenin (22,39,55,56). The C and A subunits of PP2A bind to Axin (37,57) and dephosphorylate APC and Axin (37). Furthermore, APC binds to the B56 subunit of PP2A and expression of the B56 subunit in HEK293 cells reduces the level of ␤-catenin (58). Thus, in the Axin⅐APC complex, the phosphorylation of ␤-catenin, APC, and Axin is regulated by GSK-3␤, Dvl, and PP2A, and their protein stabilities and functions are regulated by their phosphorylation. However, the mechanism of the assembly and disassembly of this Axin⅐APC complex remains to be clarified. Further study is necessary to understand the regulation of the Axin⅐APC complex by Wnt.