Crystal structure of a beta-catenin/axin complex suggests a mechanism for the beta-catenin destruction complex

Abstract

The "beta-catenin destruction complex" is central to canonical Wnt/beta-catenin signaling. The scaffolding protein Axin and the tumor suppressor adenomatous polyposis coli protein (APC) are critical components of this complex, required for rapid beta-catenin turnover. We determined the crystal structure of a complex between beta-catenin and the beta-catenin-binding domain of Axin (Axin-CBD). The Axin-CBD forms a helix that occupies the groove formed by the third and fourth armadillo repeats of beta-catenin and thus precludes the simultaneous binding of other beta-catenin partners in this region. Our biochemical studies demonstrate that, when phosphorylated, the 20-amino acid repeat region of APC competes with Axin for binding to beta-catenin. We propose that a key function of APC in the beta-catenin destruction complex is to remove phosphorylated beta-catenin product from the active site.

Figure 1.

Overall structure of the β-catenin/XAxin-CBD complex. (A) β-catenin/XAxin-CBD complex structure with the β-catenin molecular surface outlined. Each armadillo repeat of β-catenin, except repeat 7, is composed of three helices that are shown as blue, green, and yellow cylinders, whereas the XAxin-CBD is shown as a red ribbon. The N and C termini are marked for each protein. (B) Primary structure and sequence alignment of Axin. Different functional units of Axin are designated as rectangles. The red helix cartoon represents the XAxin-CBD residues visible in the crystal structure. Residues not visible in the structure are marked in black dashes. Residues highlighted in red are required for the interaction with β-catenin in the β-catenin/XAxin-CBD structure, and other residues conserved among all vertebrate Axin sequences are highlighted in yellow.

Figure 2.

Critical interactions on the β-catenin/Axin interface. In each case, β-catenin is in yellow and the XAxin-CBD is in red, except in panels B and C. β-Catenin and XAxin-CBD are labeled in black and red, respectively, except in panel C. (A) Stereoview of the 2Fo-Fc electron density map of the XAxin-CBD bound to β-catenin. The map is contoured at 1σ. (B) β-catenin electrostatic surface map of the XAxin-CBD-binding site. The surface of β-catenin is colored according to its relative electrostatic potential, with red representing negative charge and blue representing positive charge. The XAxin-CBD is shown as a stick model. (C) β-Catenin/XAxin-CBD bonding diagram. β-Catenin residues are labeled in red boxes with red text. The backbone and side chains of the XAxin-CBD are shown in blue, and XAxin-CBD residues are labeled in black circles with black text. Hydrogen bonds and charge–charge interactions are designated with green broken lines. A red star-burst together with a red broken line represents hydrophobic interactions. (D) Critical contacts in the β-catenin/XAxin-CBD interface. Residues making critical contacts are shown with side chains in ball-and-stick format.

Figure 3.

L473/D474 and H476 are required for the interaction between Axin and β-catenin. The armadillo repeat region of β-catenin tagged with GST (GST-Arm) was tested for its ability to coprecipitate wild-type XAxin (wt) or XAxin containing point mutations in residues predicted to be important for interacting with β-catenin (L473A/D474A and H476A). (Lanes 1–7) Levels of input proteins prior to pull-down. GST-Arm specifically coprecipitates XAxin (lane 12), but XAxin with a double-point mutation in L473 and D474 (lane 13) or with a single-point mutation in H476 (lane 14) shows greatly decreased interaction with GST-Arm.

Figure 4.

The 20-amino acid repeat region of APC competes with Axin for binding to β-catenin when phosphorylated. (A) Domain structure of APC. A schematic representation of the APC primary structure shows, from N to C terminus, the oligomerization domain (olig.), armadillo repeats (arm), 15-amino acid β-catenin-binding repeats (A–C), 20-amino acid β-catenin-binding repeats (1–7), Axin binding repeats (SAMP1-3), basic region (basic), and discs large interaction domain (dlg). The human APC (hAPC) peptide fragments used in the competition assay are depicted below. APC-B,C includes the second and third 15-amino acid repeats (amino acids 1133–1189). APC-2,3 includes the second and third 20-amino acid repeats (amino acids 1362–1540). (B) Phosphorylation of APC by CKI. SDS-PAGE was used to monitor the process of the phosphorylation reaction. (Lane 1) Molecular weight marker labeled in kilodaltons. APC-2,3 (unphosphorylated, lane 2) displays multiple band shifts after 1.5-h reaction (lane 3), and a further band shift after another 3-h reaction (lane 4). (C) Competition assay. GST-Arm was tested for its ability to coprecipitate Axin in the presence of various competitors. The 20-amino acid repeat fragment APC-2,3 once phosphorylated (lanes 7–9, 0.3 μg pAPC-2,3, 10- and 100-fold dilutions) can specifically and dose-dependently block coprecipitation of Axin by GST-Arm. The unphosphorylated APC-2,3 (lane 6, 0.3 μg APC-2,3), the 15-amino acid repeat fragment APC-B,C whether unphosphorylated (lane 4, 0.3 μg APC-B,C) or kinase-treated (lane 5, 0.3 μg pAPC-B,C), and a mock phosphorylation reaction containing no peptide (lane 3), all do not block Axin binding to β-catenin.

Figure 5.

The phosphorylation of β-catenin by CKI and GSK-3β does not affect its binding to Axin. (A) Native electrophoresis and Western blot of full-length β-catenin (βcat; 85kD) and full-length β-catenin treated with CKI and GSK-3β (pβcat). (Left) Coomassie blue-stained native gel. (Right) Western blot probed with a phospho-β-catenin (Ser 33/37/Thr 41) antibody, with relative molecular weight marked in kilodaltons on its right. (B) GST-pull-down binding assay. βcat and pβcat were compared in their ability to bind GST-tagged XAxin-CBD (GST-Axin). GST was used as a control. (Lanes 2–5) Input proteins. (Lanes 8,9) GST-Axin pulls down similar amounts of βcat and pβcat. (Lane 1) A molecular weight marker is labeled in kilodaltons on the left.

Figure 6.

Structural comparison. (A) Binding mode comparison between the C-terminal helix of the hTcf4-CBD and the XAxin-CBD. The armadillo repeat regions from β-catenin/XAxin-CBD and β-catenin/hTcf4 complexes were superimposed. Note that these two helices bind to β-catenin in almost completely opposite directions. The hTcf4 helix is colored in cyan and labeled in blue. Important residues are shown in ball-and-stick format. The hydrogen bonding and charge–charge interactions are designated with black broken lines. (B) Structural comparison of the XAxin-CBD and phospho-E-cadherin bound to β-catenin. β-Catenin is shown in solid cylinders colored as in Figure 1A with the residues labeled in black. XAxin-CBD is shown as a red ribbon, and E-cadherin is shown as a green ball-and-stick figure superimposed onto the β-catenin/XAxin-CBD complex. The side chain of β-catenin residue Lys 335 has been modified to the conformation in the β-catenin/phospho-E-cadherin complex to show the charge–charge interaction with the phosphoserine.

Figure 7.

A working model for the β-catenin destruction complex. The components in the β-catenin destruction complex are colored as follows: Axin (red); β-catenin (yellow); GSK-3β (green); APC (purple); PP2A (pink). Axin's binding regions for β-catenin, GSK-3β, and APC are shown as red rectangles. The 20-amino acid and 15-amino acid repeats of APC are labeled. Phosphorylation at the β-catenin N terminus or the APC 20-amino acid repeats is designated by red stars. Step 1 illustrates the recruitment of β-catenin with the help of APC. Step 2 shows phosphorylation of β-catenin by GSK-3β (after priming phosphorylation by CKI, which binds Axin but is not shown in the model). Step 3 illustrates the release of phosphorylated β-catenin product from Axin by the phosphorylated APC 20-amino acid repeats. Step 4 illustrates the dephosphorylation of the APC 20-amino acid repeats by PP2A and the release of phosphorylated β-catenin, which is subsequently degraded via the ubiquitin ligase–proteasome system. For simplicity, CKI and other regulatory proteins are not shown in the model.