Axin, an inhibitor of the Wnt signalling pathway, interacts with β-catenin, GSK-3β and APC and reduces the β-catenin level

Abstract

Background:

The Wnt/Wingless signalling pathway plays an important role in both embryonic development and tumorigenesis. β-Catenin and Axin are positive and negative effectors of the Wnt signalling pathway, respectively.

Results:

We found that Axin interacts with β-catenin and glycogen synthase kinase-3β (GSK-3β). Furthermore, the regulation of the G-protein signalling (RGS) domain of Axin is associated with the colorectal tumour suppressor adenomatous polyposis coli (APC). Overexpression of Axin in the human colorectal cancer cell line SW480 induced a drastic reduction in the level of β-catenin. Interaction with β-catenin and GSK-3β was required for the Axin-mediated β-catenin reduction.

Conclusion:

Axin interacts with β-catenin, GSK-3β and APC, and negatively regulates the Wnt signalling pathway, presumably by regulating the level of β-catenin.

Introduction

The Wnt gene family consists of more than 15 closely related genes that encode secreted glycoproteins with important functions in a number of developmental processes, including body axis formation, development of the central nervous system, axial specification in limb development and mouse mammary gland development (Nusse & Varmus 1992; Cadigan & Nusse 1997). The Drosophila segment polarity gene wingless is the homologue of Wnt-1 and is important for correct cellular patterning within the embryo and imaginal discs (Klingensmith & Nusse 1994).

Recent studies have identified several components which transduce the Wnt signal to the nucleus (Miller & Moon 1996; Cadigan & Nusse 1997). The Wnt proteins bind to their receptors, the frizzled family of proteins, and elicit a signal which is transduced to a cytoplasmic protein, Dishevelled (Dsh). Once activated by the Wnt signal, Dsh negatively regulates the activity of glycogen synthase kinase-3β (GSK-3β). In the absence of the Wnt signal, GSK-3β is assumed to phosphorylate and consequently induce the degradation of β-catenin, a multifunctional protein originally identified as a cadherin-associated protein (Peifer et al. 1994a,b; Cook et al. 1996). Thus, the Wnt signal stabilizes β-catenin, which in turn associates with the high mobility group (HMG)-domain proteins, the Lef/Tcf family transcription factors, ultimately altering the expression of Wnt signalling target genes (Behrens et al. 1996; Huber et al. 1996a,b; Molenaaret al. 1996).

The level of β-catenin is also regulated by the adenomatous polyposis coli (APC) protein, mutation in which is responsible for familial adenomatous polyposis (FAP) and sporadic colorectal tumours (Groden et al. 1991; Kinzler et al. 1991; Nishisho et al. 1991; Miyoshi et al. 1992a). APC is a 300 kDa protein associated with β-catenin, GSK-3β and a homologue of the Drosophila Discs Large tumour suppressor protein (Rubinfeld et al. 1993, 1996; Su et al. 1993; Matsumine et al. 1996). Overexpression of APC in the human colorectal tumour cell line SW480, which is defective for APC, induces a dramatic down-regulation of β-catenin expression (Munemitsu et al. 1995). This down-regulation activity is localized to the central region of APC, which contains a region of seven repeated sequences of 20 amino acids each, which is frequently disrupted by mutations in colon cancers (Munemitsu et al. 1995; Rubinfeld et al. 1997a). Thus, the products of mutant APC genes present in most colon cancers are defective in this activity. Consistent with these findings, colon carcinoma cells were found to contain elevated levels of β-catenin, which forms a stable constitutive active complex with Tcf-4 (Korinek et al. 1997; Morin et al. 1997; Rubinfeld et al. 1997b). Furthermore, some colorectal tumours and melanoma cell lines with intact APC were found to contain mutations which activate ββ-catenin (Morin et al. 1997; Rubinfeld et al. 1997a). Therefore, the accumulation of β-catenin due to mutations in either APC or β-catenin may be important in colorectal tumorigenesis.

Another negative regulator of the Wnt signalling pathway is the product of the fused locus, Axin (Gluecksohn-Schoenheimer 1949; Jacobs-Cohen et al. 1984; Perry et al. 1995; Zeng et al. 1997). The most remarkable abnormality of embryos homozygous for fused is the formation of axial duplications. In Xenopus, it is well established that the Wnt pathway plays a crucial role in the development of the embryonic axis, as demonstrated by the observation that injection of mRNA encoding various components of the Wnt pathway into Xenopus embryos induces the formation of an ectopic axis (Miller & Moon 1996; Cadigan & Nusse 1997). Injection of Axin mRNA into Xenopus embryos inhibits dorsal axis formation, while injection of mutant Axin mRNA induces an ectopic axis, apparently through a dominant negative mechanism (Zeng et al. 1997). Co-injection of Axin mRNA with components of the Wnt pathway suggests that Axin somehow interferes with the Wnt pathway, although the precise nature of this effect is not understood.

To further elucidate the regulation of the Wnt signalling pathway, we attempted to isolate cDNAs encoding β-catenin-associated proteins using the yeast two-hybrid system. We identified Axin as a β-catenin-associated protein and also found that it interacts with GSK-3β and APC. We further showed that overexpression of Axin induces a reduction of β-catenin level in SW480 cells.

Results

β-catenin binds to Axin

We screened a human brain library using the armadillo repeat domain of β-catenin as ‘bait’, and obtained two positive clones from 1 × 10⁶ transformants. Sequencing of these two clones revealed that they are overlapping cDNA fragments encoding Axin.

To confirm that β-catenin and Axin associate directly, we examined the ability of β-catenin fused to glutathione-S-transferase (GST) to interact with Axin produced by in vitro translation. We found that an in vitro-translated Axin associated specifically with GST-β-catenin, but not with GST alone (Fig. 1). The armadillo repeat domain of β-catenin (amino acids 128–683) was responsible for binding to Axin, whereas the amino-terminal (amino acids 1–127) and carboxyl-terminal (amino acids 684–781) domains did not show an affinity to Axin.

Figure 1. Association of Axin with β-catenin, APC and GSK-3βin vitro. In vitro-translated ³⁵S-labelled Axin (IVT-mAxin) was incubated with GST-fusion proteins indicated in the figure that were immobilized to glutathione-Sepharose. The absorbed proteins were separated by 7.5% SDS-PAGE. Codon numbers that define each construct are indicated in parenthesis.

When a lysate from human fibrosarcoma HT1080 cells transiently transfected to overexpress epitope-tagged Axin (Axin-Flag) was subjected to immunoprecipitation with anti-Flag antibody followed by immunoblotting with anti-β-catenin antibody, β-catenin was found to co-immunoprecipitate with Axin-Flag (Fig. 2). Pre-incubation of the anti-Flag antibody with the antigen prevented co-precipitation of both Axin and β-catenin. These results suggest that β-catenin interacts with Axin both in vitro and in vivo.

Figure 2. Association of Axin with β-catenin and APC in vivo. Lysates prepared from HT1080 cells transfected with Axin-Flag were subjected to immunoprecipitation with antibodies to Flag, fractionated by SDS/PAGE, and immunoblotted with the antibodies indicated. Pep. + , antibodies were preincubated with antigen before use in immunoprecipitation.

Amino acids 592–616 of Axin is responsible for binding to β-catenin

For identification of the region of Axin responsible for its interaction with β-catenin, we performed two-hybrid assays using various deletion fragments of Axin. While mutants lacking amino acids 490–641 were negative for interaction with β-catenin, a fragment containing amino acids 581–627 showed β-catenin-binding activity (Fig. 3A). Furthermore, a smaller fragment containing amino acids 592–616 was positive for the interaction, suggesting that this 25 amino acid segment of Axin is responsible for binding to the armadillo repeats of β-catenin.

Figure 3. Mapping of regions in Axin required for binding to β-catenin (A) and GSK-3β (B). Deletion constructs of Axin were analysed for their ability to interact with β-catenin or GSK-3β, respectively, in the two-hybrid system. Left: schematic representation of Axin deletion mutants. Right: corresponding binding activities. (+) detectable activity; (–) no detectable activity.

The RGS domain of Axin interacts with APC

Axin contains an RGS domain and a region of homology with Dsh at its amino-terminal and carboxy-terminal regions, respectively (Zeng et al. 1997). The RGS domain is reported to be important for the ability of Axin to negatively regulate the Wnt signalling pathway. Therefore we tried to identify proteins that associate with the RGS domain of Axin by using it as ‘bait’ to screen a human brain. The sequencing of one clone isolated from 1 × 10⁶ transformants revealed that it encodes amino acids 1491–1700 of APC which contains the third and fourth 20-amino acid repeats. When produced in E. coli as a GST-fusion protein, this region was shown to interact with Axin generated by in vitro translation. While neither the third nor fourth 20-amino acid repeat was responsible for Axin binding, a small segment between these two repeats (amino acids 1561–1630) efficiently associated with Axin (Fig. 1). In addition, the fragment containing the second and third β-catenin-binding sites (amino acids 1136–1169) did not exhibit Axin-binding activity. When a lysate from HT1080 cells overexpressing Axin-Flag were subjected to immunoprecipitation with anti-Flag antibody, APC was found to co-precipitate with Axin-Flag (Fig. 2). These results suggest that amino acids 1561–1630 of APC directly interacts with the RGS domain of Axin.

Axin is associated with GSK-3β

Since Axin is associated with β-catenin and APC, we also examined whether it is associated with another component of the Wnt/Wingless signalling pathway, GSK-3β. An in vitro pull-down assay showed that GSK-3β which is fused to GST interacts with in vitro-translated Axin (Fig. 1). Two-hybrid experiments with a series of deletion mutants of Axin delineated the region of GSK-3β binding to amino acids 444–543 of Axin (Fig. 3).

Overexpression of Axin reduces β-catenin levels

Overexpression of APC in human colorectal tumour SW480 cells is known to induce the degradation of β-catenin (Munemitsu et al. 1995). We therefore examined the effect of Axin on β-catenin expression levels. As shown in Fig. 4, overexpression of Axin-Flag in SW480 cells induced a drastic reduction in the level of ββ-catenin. To determine which domain of Axin is responsible for this effect on β-catenin, we generated a series of mutant Axins which lack the RGS domain, the ββ-catenin-binding domain, the GSK-3β-binding domain, or the Dsh-like domain, respectively. Transient transfection of these constructs into SW480 cells revealed that mutants lacking the RGS domain or the Dsh-like domain are as competent as wild-type Axin in causing a reduction in β-catenin levels, whereas mutants lacking the β-catenin-binding or the GSK-3β-binding domain are not capable of reducing β-catenin levels. These results suggest that Axin negatively regulates the Wnt signalling pathway by regulating the level of β-catenin.

Figure 4. Effect of Axin expression on β-catenin level. (A) Schematic representation of Axin and its deletion mutants. Effects of the Axin constructs on β-catenin level in SW480 cells are summarized on the right side of the schema; (+) detectable activity; (–) no detectable activity. (B) SW480 cells (A and H) were transiently transfected with GFP cDNA along with control vector (B and I), wild-type Axin-Flag (C and J) or mutant Axin-Flag lacking the RGS domain (D and K), β-catenin-binding domain (E and L), GSK-3β-binding domain (F and M) or Dsh-like domain (G and N), respectively. Cells were stained with antibodies to GFP followed by FITC-labelled anti-rabbit secondary antibodies (A–G) or antibody to β-catenin followed by RITC-labelled anti-mouse secondary antibodies (B–N). (C) The histogram shows the percentage of β-catenin-negative cells. Fifty cells were scored for each experiment.

Discussion

The fact that mouse embryos homozygous for the Axin mutation develop ectopic axial structures suggested that Axin normally plays a negative role in the formation of the embryonic axis (Gluecksohn-Schoenheimer 1949; Jacobs-Cohen et al. 1984; Perry et al. 1995). Consistent with this notion, it was shown that a dorsal injection of wild-type Axin mRNA into the Xenopus embryo blocks axis formation, whereas a ventral injection of mutant Axin induces an ectopic axis (Zeng et al. 1997). Moreover, Axin abrogates the axis-inducing activity of Xwnt8, Xdsh and a dominant-negative GSK-3β mutant, suggesting that Axin is a negative regulator of the Wnt signalling pathway. However, the molecular mechanisms by which Axin exerts its negative effects had not been elucidated. In the present study, we demonstrate that Axin interacts with β-catenin, GSK-3β and APC. Furthermore, overexpression of Axin in SW480 cells induced a dramatic reduction in the levels of β-catenin. These findings suggest that Axin may function as a docking station facilitating the interaction of β-catenin and its negative regulators GSK-3β and APC, thereby negatively regulating the Wnt signalling pathway by inducing the down-regulation of β-catenin.

In the absence of the Wnt signal, GSK-3β induces the degradation of β-catenin (Peifer et al. 1994a,b; Cook et al. 1996). The amino-terminal portion of β-catenin contains four consensus motifs for phosphorylation by GSK-3β, and mutation of these sites results in ββ-catenin stabilization (Munemitsu et al. 1996; Yost et al. 1996; Zecca et al. 1996; Pai et al. 1997). The simplest interpretation of these findings is that GSK-3β directly phosphorylates β-catenin and induces its degradation, although β-catenin is reported to be a poor substrate for GSK-3βin vitro (Rubinfeld et al. 1996; Stambolic et al. 1996; Pai et al. 1997). We found that β-catenin and GSK-3β bind adjacently to the central region of Axin; amino acids 444–543 and 592–616, respectively. Thus, Axin may facilitate the interaction between GSK-3β and β-catenin by tethering these two molecules side-by-side. The importance of this finding was demonstrated by showing that mutant Axins lacking β-catenin- or GSK-3β-binding region, when overexpressed in SW480 cells, fail to induce the downregulation of β-catenin expression. It remains to be examined whether GSK-3β-mediated phosphorylation of β-catenin is indeed induced by the presence of Axin in vivo. It would also be interesting to examine whether the formation of the Axin complex is regulated by the Wnt signal or not.

Some proteins which possess an RGS domain bind to Gα subunits and function as GTPase-activating proteins (GAPs) for the Gi subfamily of Gα subunits (De Vries et al. 1995; Dohlman & Thorner 1997; Watson et al. 1996). However, the amino acid sequence of the Axin RGS domain differs at several key residues conserved in other RGS proteins (Zeng et al. 1997). Indeed, the RGS domain of Axin was found to bind to APC. The fragment of APC bound in our two-hybrid screening assay contains repeats 3 and 4 of the 20-amino acid repeat region. Whereas β-catenin binds to the 20-amino acid region (Munemitsu et al. 1995), Axin was found to bind to a small region between repeats 3 and 4, but not to repeats 3 and 4 themselves. The 20-amino acid repeat region is frequently mutated in colorectal cancers and is important for the degradation of β-catenin (Miyoshi et al. 1992b; Miyaki et al. 1994; Munemitsu et al. 1995). Furthermore, Rubinfeld et al. previously showed that the region containing repeats 2–4 partially promotes the down-regulation of ββ-catenin, although four or more repeats are required for complete effectiveness (Rubinfeld et al. 1997a). These results suggest that Axin association may be important for the APC-mediated degradation of β-catenin. On the other hand, overexpression of mutant Axin lacking the RGS domain has been shown to induce down-regulation of β-catenin as efficiently as wild-type Axin, showing that Axin can down-regulate β-catenin without binding to APC; we therefore imagine that Axin functions downstream of APC in the β-catenin degradation-inducing pathway.

The ubiquitin-proteasome system plays a crucial role in the rapid turnover of key regulatory molecules such as cyclins, IκB-NFκB and p53 (Coux et al. 1996). β-Catenin is also reported to be turned over by the ubiquitin-dependent proteolysis system (Aberle et al. 1997). Mutations in the GSK-3β phosphorylation consensus motif of β-catenin inhibit its ubiquitination and results in its stabilization (Munemitsu et al. 1996; Yost et al. 1996; Zecca et al. 1996; Pai et al. 1997). Thus it is possible that β-catenin is efficiently phosphorylated by GSK-3β in the Axin complex, and then subjected to the ubiquitin-dependent degradation. It is furthermore tempting to speculate that there is a molecule(s) associated with Axin which targets β-catenin for the ubiquitin-proteasome pathway. In this regard, it is interesting to note that the Drosophila gene slimb, loss of whose function results in an accumulation of high levels of Armadillo (Drosophilaβ-catenin) and Cubitus, encodes a conserved F-box-WD40-repeat protein related to Cdc4p, a protein in budding yeast that targets cell cycle regulators for degradation by the ubiquitin-proteasome pathway (Jiang & Struhl 1998). Hence, Slimb would be a reasonable candidate as a protein which targets β-catenin for degradation by the ubiquitin-proteasome pathway. Identification of a molecule(s) which links the Axin complex to the ubiquitin-proteasome system should greatly extend our understanding of the mechanisms regulating the Wnt signalling pathway.

During the preparation of this manuscript, two groups also reported that Axin binds to β-catenin and GSK-3β (Ikeda et al. 1998; Sakanaka et al. 1998). Furthermore, Behrens et al. have reported that conductin, which shows 45% amino acid sequence identity to Axin, interacts with β-catenin, GSK-3β and APC (Behrens et al. 1998). Most interestingly, Ikeda et al. showed that Axin promotes the phosphorylation of ββ-catenin by GSK-3βin vitro (Ikeda et al. 1998). Behrens et al. showed that the interaction between APC and conductin occurs through a previously unidentified repeat element, SAMP (Ser-Ala-Met-Pro), which resides between the 20-amino acid repeats 3 and 4, 4 and 5, and downstream of the repeat 7, respectively (Behrens et al. 1998).

Experimental procedures

Yeast two-hybrid cloning

The plasmids GAL4-β-catenin and GAL4-RGS, which encode the GAL4 DNA-binding domain fused to the armadillo repeat domain of β-catenin (amino acids 128–683) and the RGS domain of Axin (amino acids 213–339), respectively, were used as bait in two-hybrid screens of a human brain cDNA library (Clontech).

Antibodies

Antibodies to APC were prepared as described (Baeg et al. 1995). Antibodies were purified by affinity chromatography using columns to which peptides used for immunization had been linked. Mouse monoclonal antibodies to β-catenin and Flag were purchased from Transduction laboratory and Eastman Kodak, respectively. Rabbit polyclonal antibody to green fluorescent protein was obtained from Promega.

In vitro binding assays

³⁵S-labelled full-length Axin was synthesized by in vitro transcription-translation in the presence of [³⁵S]methionine using the TNT^TM-coupled reticulocyte lysate system (Promega). GST fusion proteins were generated by subcloning the respective cDNAs into pGEX-5X-1 (Pharmacia), expressing them in E. coli, and isolating them by absorption to glutathione-Sepharose. GST fusion proteins immobilized to glutathione-Sepharose were mixed with in vitro translated proteins in buffer A [10 mm Tris-HCl (pH 8.0), 140 mm NaCl, 1 mm EDTA, 5 mg/mL leupeptin, 5 mg/mL aprotinin] containing 0.1% Triton X-100 for 1 h at 4 °C and then washed extensively with buffer A. Proteins adhering to the beads were analysed by SDS/PAGE followed by autoradiography.

Immunoprecipitation and immunoblotting

Cells were lysed in buffer A containing 1% Triton X-100 and the lysates were incubated with antibodies for 1 h at 4 °C. The immunocomplexes were adsorbed to protein A-Sepharose 4B and washed extensively with buffer A containing 0.1% Triton X-100. Samples were resolved by SDS/PAGE and transferred to a poly(vinylidene difluoride) membrane filter (Immobilon P) (Millipore). The blot was subjected to immunoblotting analysis using alkaline phosphatase-conjugated goat anti-mouse or goat anti-rabbit IgG (Promega) as a secondary antibody.

Cell culture, transfections and immunofluorescence analysis

The human fibrosarcoma cell line HT1080 and human colorectal cancer cell line SW480 were cultured in Eagle's MEM medium and Leibovitz's L-15 medium, respectively, supplemented with 10% foetal calf serum. The axin cDNA was subcloned into the mammalian expression vector pMKITneo carrying the SRα promoter and then transfected into HT1080 cells transiently by use of LipofectAMINE (Life Technologies). Forty-eight hours after transfection, HT1080 cells were subjected to immunoprecipitation–immunoblotting as described above. The axin cDNA subcloned into pMKITneo was co-transfected with pCX-GFP into SW480 cells by the use of LipofectAMINE. SW480 cells were subjected to immunofluorescence analysis 48 h after transfection as described previously (Satoh et al. 1997). Staining patterns obtained with anti-β-catenin antibody were visualized with RITC-labelled anti-mouse secondary antibodies (Cappel); those obtained with anti-GFP were visualized with FITC-labelled anti-rabbit secondary antibodies (Cappel). Stained samples were viewed with an Olympus AH2-FL fluorescent microscope.

Acknowledgements

We thank A. Suzuki, Y. Katoh, A. Ogai and A. Furui for their helpful assistance. This work was supported by Grants-in-Aid for Scientific Research on Priority Areas.