Structures of the APC–ARM domain in complexes with discrete Amer1/WTX fragments reveal that it uses a consensus mode to recognize its binding partners

The tumor suppressor APC employs its conserved armadillo repeat (ARM) domain to recognize many of its binding partners, including Amer1/WTX, which is mutated in Wilms' tumor and bone overgrowth syndrome. The APC–Amer1 complex has important roles in regulating Wnt signaling and cell adhesion. Three sites A1, A2, and A3 of Amer1 have been reported to mediate its interaction with APC-ARM. In this study, crystal structures of APC–ARM in complexes with Amer1-A1, -A2, and -A4, which is newly identified in this work, were determined. Combined with our GST pull-down, yeast two-hybrid, and isothermal titration calorimetry (ITC) assay results using mutants of APC and Amer1 interface residues, our structures demonstrate that Amer1-A1, -A2, and -A4, as well as other APC-binding proteins such as Asef and Sam68, all employ a common recognition pattern to associate with APC–ARM. In contrast, Amer1-A3 binds to the C-terminal side of APC–ARM through a bipartite interaction mode. Composite mutations on either APC or Amer1 disrupting all four interfaces abrogated their association in cultured cells and impaired the membrane recruitment of APC by Amer1. Our study thus comprehensively elucidated the recognition mechanism between APC and Amer1, and revealed a consensus recognition sequence employed by various APC–ARM binding partners.


Introduction
Adenomatous polyposis coli (APC) is an important human tumor suppressor protein which plays critical roles in diverse cellular processes such as Wnt/β-catenin signaling [1], cell adhesion [2,3], cell migration [4,5], mitosis and chromosomal instability [6], and so on. Mutations of the human APC gene are found in most of familial adenomatous polyposis (FAP) patients, as well as in a majority of sporadic colorectal cancer cases [7,8]. The N-terminal armadillo repeat (ARM) domain of APC is the most conserved region among its vertebrate and invertebrate homologs [9,10], and mediates its association with a variety of binding partners including APC membrane recruitment 1 (Amer1, also named as WTX for 'Wilms tumor gene on the X chromosome') [11,12], Asef [13], Sam68 [14], and IQGAP1 [15].
Amer1/WTX is another important human tumor suppressor whose gene is somatically inactivated in one-third of Wilms' tumors, the most common pediatric kidney cancers [12]. In contrast, germline mutations of the Amer1/WTX gene predispose to osteopathia striata congenita with cranial sclerosis, a bone overgrowth syndrome [16]. Amer1/WTX-knockout mice exhibited both kidney and bone defects due to aberrant specification of mesenchymal progenitor cell fate [17]. The Amer1 protein possesses two membrane targeting regions at its N-terminus, three APC-interacting sites named as A1 (residues 280-368), A2 (residues 380-531), and A3 (residues 717-834) [11], as well as a β-catenin-binding arginine-glutamate-alanine (REA) repeats [18] (Supplementary Figure S1). By recruiting APC to the plasma membrane, Amer1 regulates the APC-dependent maintenance of intercellular junctions [11]. In addition, together with APC and Axin, Amer1 promotes ubiquitination and degradation of β-catenin, thus negatively regulates the Wnt signal transduction pathway [19]. Moreover, Amer1 also has a variety of other functions including participating in the upstream activation of Wnt signaling by promoting the phosphorylation of the Wnt co-receptor LRP6 [20], shuttling to the nucleus and enhancing the transcriptional activity of Wilms tumor 1 (WT1) [21], and positively regulating the CBP/p300-mediated p53 acetylation [22].
The three APC-binding regions A1, A2, and A3 of Amer1 display no obvious sequence similarities among themselves nor exhibit any resemblance with other APC-ARM-binding motifs such as the APC-binding region (ABR) of Asef [10] and the YY motif of Sam68 [14]. These three sites have only been roughly mapped to regions spanning 50-120 residues, and it is completely unknown which residues at these three sites mediate specific recognitions with APC. In the case of the ARM domain of β-catenin, its interaction partners such as phospho-APC 20-amino-acid repeat [23,24], phospho-E-cadherin cytoplasmic domain [25], and TCF's β-catenin-binding domain [26] all reside in the same surface groove of β-catenin-ARM with a similar binding pattern despite their diverse sequences. It will be interesting to examine whether different APC-ARM-binding fragments including those from Amer1, Asef, and Sam68 also exhibit a common mode to associate with APC-ARM, although they have no apparent recognizable homology with each other.
In this study, we investigated the molecular basis underlying the complex formation between Amer1 and APC. We identified core sequences of the Amer1-A1 and -A2 sites for APC binding, discovered a fourth APC-binding site on Amer1 (which we named as A4), and determined the crystal structures of APC-ARM in complexes with the A1, A2, and A4 peptides of Amer1. In addition, point mutations of key interface residues on Amer1-A1/A2/A4 or APC-ARM abrogated their interactions by glutathione sulfur transferase (GST) pull-down, yeast two-hybrid, and isothermal titration calorimetry (ITC) assays. In contrast to Amer1-A1/A2/ A4, Amer1-A3 was found to use a bipartite interaction mode to bind to the C-terminal side of APC-ARM. Composite mutations on APC or Amer1 disrupting all four interaction interfaces abolished their association in cultured cells and abrogated the membrane recruitment of APC by Amer1. Finally, structural superimposition reveals a common mode used by APC-ARM to recognize Amer1-A1/A2/A4, Asef-ABR, and Sam68-YY. By summarizing their sequences and interaction patterns, a consensus APC-ARM recognition motif employed by its binding partners is proposed.

Results
Identification of the core Amer1-A1 and -A2 sequences for APC-ARM binding Previous studies have roughly mapped the Amer1-A1 and -A2 sites to residues 280-368 and 380-531, respectively [11]. However, examination of the structure of APC-ARM shows that the ligand-binding groove on its surface is~50 Å long and~20 Å wide [9,10], which can only accommodate a peptide of 12-15 residues in a fully extended conformation or a longer peptide containing bends or secondary structures. Alignment of the human Amer1 protein sequence with those of its chicken, frog, and zebrafish orthologs as well as its paralogs Amer2/FAM123A and Amer3/FAM123C revealed conserved sequence blocks (Amer1 residues 315-335 and 496-508 for the A1 and A2 sites, respectively) flanked by non-conserved residues ( Figure 1a) [27]. Intriguingly, ITC experiments demonstrated that peptides of Amer1-A1 (residues 325-335) and -A2 (residues 496-508) sequences exhibited approximately equal binding affinities for APC-ARM (Figures 1b and c, Table 1) as full Amer1-A1 (residues 280-368) and -A2 (residues 380-531), respectively (Table 1, Supplementary Figures S2A and  S2B). In previous studies, the N507K point mutation of APC has been shown to abrogate its association with Amer1 [11] and Asef [10,28], and its corresponding mutation of N175K in Drosophila APC2 (also known as E-APC) resulted in developmental defects [2]. In accordance with these findings, both Amer1 (325-335) and Amer1 (496-508) displayed non-detectable interaction with the N507K point mutant protein of APC-ARM, as shown by the GST pull-down assay (Supplementary Figure S3). Therefore, Amer1 residues 325-335 and 496-508 are regarded to represent the core A1 and A2 fragments for APC binding, respectively, although other conserved Amer1 residues nearby such as residues 315-324 may also contribute to the interaction with APC.

Identification of A4, a fourth APC-binding site on Amer1
Amino-acid sequences of the three APC-binding sites of Amer1/WTX, especially A1 and A2, show little variance among different Amer1 orthologs and paralogs (Figures 1a and 1d), suggesting that the recognition of APC is a conserved function of Amer1 throughout evolution. When the percentage of conservation for each Amer1 residue was carefully examined, another highly invariant sequence block between the A1 and A2 sites, residues 365-375, was found (Figures 1d and e) [27]. Similar to the core A1 and A2 sequences, it also contains an acidic residue in the middle and several hydrophobic residues at its C-terminal part (Figure 1e). When the ITC assay was performed to examine the binding affinity between Amer1 (365-375) and APC-ARM, their dissociation constant (K d ) was measured to be 1.96 μM (Figure 1f), stronger than that of Amer1-A1 and similar to that of Amer1-A2 (Figures 1b and c). Hence, this sequence block represents a fourth motif on Amer1 for APC recognition, which is named as A4 after the three previously described APC-binding sites.
At all three interfaces between APC-ARM and Amer1-A1/A2/A4, four highly conserved asparagine residues, N507/N550/N594/N641 from the H3 helices of armadillo repeats 2/3/4/5 of APC-ARM function as Figure 1 Identification of a fourth APC-binding fragment A4 of Amer1/WTX, and crystal structures of APC-ARM in complex with the A1, A2, and A4 fragments of Amer1. (a) Sequence alignment of the APC-binding A1 and A2 fragments of human Amer1 (hAmer1). hAmer1-A1 (residues 280-368) and -A2 (residues 380-531) are aligned with orthologs and paralogs. The A1 and A2 fragments used in the crystallization experiments are marked with red underlines. Residues identical in all the Amer1 homologs are highlighted in yellow. Residue numbers for hAmer1 are indicated above the sequences. (b, c) The binding affinities of hAmer1 (residues 325-335, (b) and hAmer1 (residues 496-508, (c) for APC-ARM. (d) Sequence comparison of human, chicken, frog, and zebrafish Amer1, human Amer2 and Amer3 reveals a highly conserved fragment, which is named as A4. hAmer1 residue numbers are indicated, and the percentage of conservation for each residue is shown as a red, orange, yellow, green, cyan, and blue bar from high to low conservation, respectively. The membrane-binding regions M1 and M2, as well as the other APCbinding sites A1, A2, and A3, are marked. (e) Sequence alignment of the A4 site (residues 365-375) of hAmer1 and its homologs. Residues identical in at least 5 out of 6 homologs are highlighted in yellow. (f) The dissociation constant (K d ) between APC-ARM and Amer1-A4 was measured to be 1.96 μM by the ITC assay. (g) Overall crystal structures of APC-ARM in complexes with hAmer1-A1 (325-335), -A2 (496-508), and -A4 (365-375).

Mutations of key residues of APC-ARM and Amer1-A1/A2/A4 disrupt their associations
To corroborate our structural observations, we performed site-directed mutagenesis on critical interface residues of APC-ARM and Amer1-A1/A2/A4, and examined the effects of these point mutations by yeast two-hybrid, GST pull-down, and ITC assays.
Consistent with their crucial roles observed in the cocrystal structures, replacing the three key asparagines of APC, N507/N550/N594, by lysines disrupted or diminished the association between APC-ARM and Amer1-A1/A2/A4 (   Different from Amer1-A1/A2/A4, Amer1-A3 employs a bipartite binding mode to interact with the C-terminal side of APC-ARM The APC-ARM-binding site A3 of Amer1 was only roughly mapped to a region of about 120-amino acids from 717 to 834 [11]. Sequence comparison of the human Amer1-A3 with its orthologs shows that residues 767-819 are highly conserved, whereas residues outside this region possess little homology (Figure 4a). Indeed, when we compared the binding of APC-ARM with Amer1 (716-834) or Amer1 (766-823) by the ITC and GST pull-down assays, these two constructs of Amer1 exhibited comparable interaction affinities (Figures 4b and c, Table 1, Supplementary Figure S9A), suggesting that Amer1 (residues 766-823) might represent the core A3 fragment for associating with APC.
Within Amer1 (766-823), residues 767-777 and 796-819 are more conserved, while the amino acids 778-795 between them display much more diversity (Figure 4a). When residues 778-794 were deleted from Amer1  Table 1). Our results showed that Amer1 (766-777) or Amer1 (786-823) displayed no detectable interactions with APC-ARM. Residues 778-794 probably form a bulging loop when Amer1-A3 is bound to APC-ARM, and thus do not provide much contribution to the recognition between APC and Amer1. Therefore, Amer1-A3 employs residues 766-777 and 795-823 to associate with APC-ARM by a bipartite binding mechanism, and neither of these two fragments is able to interact with APC-ARM separately on its own. When we tested the set of APC-ARM point mutants defective in binding Amer1-A1/A2/A4 such as N507K/K516E/R549E/N550K for association with Amer1-A3, we surprisingly found that these mutations did not affect their interaction (Figure 4g). In contrast, a point mutation at the C-terminal side of APC-ARM, M717K, destroyed the APC-ARM/Amer1-A3 complex formation (Figure 4g). M717 was mutated because it was shown to be a key residue in mediating the interaction of APC-ARM with Sam68 [14] and lies outside of the residues mediating interaction of APC-ARM with Amer1-A1/A2/A4, which were not relevant for binding to Amer1-A3. Hence, Amer1-A3 interacts with the C-terminal portion of APC-ARM using a bipartite binding mode, which is different from the way that Amer1-A1/ A2/A4 bind.
Composite mutations on Amer1 (2-700) disrupting the A1/A2/A4 sites abrogated its association with FL APC and compromised the recruitment of APC to the plasma membrane Having examined the interaction mechanism between APC-ARM and each of Amer1-A1/A2/A3/ A4 sites, we next examined what the effects would be when mutations targeting different sites were combined. To this end, we first prepared a triple mutant construct of Amer1 (2-700) in which the A1/A2/A4 sites were destroyed by the D330K/D503K/E370K Structures of the APC-Amer1 complex 6 triple mutation. Amer1 (2-700) does not contain the A3 site, thus this triple mutant construct effectively has all of the APC-binding sites removed. Indeed, coimmunoprecipitation assays verified that Amer1 (2-700, D330K/D503K/E370K) was not able to assemble with either APC-ARM (Figures 5a and b) or FL APC (Figure 5c) anymore, although the Amer1 (2-700) double mutant D330K/D503K with only the A1/A2 sites disrupted still associated with APC (Figures 5a and c).
One important function of Amer1 is to recruit APC to the plasma membrane, thus regulating the formation of intercellular junctions and Wnt signaling (Figures 6a and aʹ) [11,18]. Similarly, Amer1 (2-700) could also bring APC to the plasma membrane (Figures 6b and bʹ). When the double mutation of D330K/D503K with only the A1/A2 sites destroyed was introduced to Amer1 (2-700), APC was still recruited to the plasma membrane by Amer1 (Figures 6c and cʹ). In contrast, the D330K/D503K/E370K triple mutation on Amer1 (2-700), which eliminated all four APC-binding sites, abrogated the membrane co-localization of APC with Amer1 (Figures 6d  and dʹ). Therefore, our co-immunoprecipitation and subcellular co-localization experiments demonstrate that each of the four APC-binding motifs of Amer1 is functional in recruiting APC in cultured cells and translocating APC to the plasma membrane.
Composite mutations of APC-ARM destroying all four Amer1-binding interfaces reduced its complex formation with FL Amer1 As described above, the K516E mutation on APC-ARM impaired its association with Amer1-A1/ A2/A4 (Figures 2a and b). In contrast, the M717K mutation on APC-ARM diminished its binding to Amer1-A3 (Figure 3g). However, single-point mutants of APC-ARM-K516E or -M717K were still co-immunoprecipitated with FL Amer1/WTX, presumably due to interactions mediated by the remaining binding interface(s) (Figure 5d). In contrast, the double mutation of K516E/M717K on APC-ARM strongly reduced its complex formation with FL Amer1 (Figure 5d). Therefore, disruption of all four binding interfaces A1/A2/A3/A4 of Amer1 is necessary and sufficient to dissociate its complex formation with APC.
Structural comparison reveals that APC-ARM employs the same surface groove to recognize a consensus motif from Amer1-A1, -A2, -A4, Asef-ABR, and Sam68-YY Amer1-A1, A2, and A4 have no apparent sequence similarity among each other, nor do they bear any Structures of the APC-Amer1 complex 8 Figure 5 Composite mutations in Amer1 or APC abolished the Amer1-APC interaction in cultured cells. (a, b) Coimmunoprecipitation assays demonstrated that triple mutation of A1-D330K/A4-E370K/A2-D503K in Flag-tagged Amer1 (residues 2-700) attenuated its association with EGFP-tagged APC-ARM, using either anti-GFP (a) or anti-Flag (b) immunoprecitation followed by western blot analysis. Note that transfection of the Amer1 cDNA generates two protein bands that are due to alternative splicing (arrows [18]). (c) Triple mutation of D330K/E370K/D503K in Amer1 (2-700) destroyed its interaction with FL APC. (d) Double mutation of K516E (which abrogated A1/A2/A4-binding) and M717K (which impaired A3-binding) in APC-ARM reduced its association with FL Amer1 by 60% as determined by densitometric analysis. resemblance with the APC-binding regions of Asef and Sam68. Yet, by superimposing the crystal structures of APC-ARM in complexes with Amer1-A1/A2/A4 as well as Asef-ABR and Sam68-YY [10,14], these APCinteracting ligands are all found to reside in the same surface groove of APC-ARM, all in anti-parallel manners with respect to the armadillo repeats of APC ( Figure 7a). As these APC-binding partners occupy roughly the same position, we wonder whether they would compete with each other in a mutually exclusive manner. Indeed, addition of the Asef-ABR-SH3 protein to the pre-assembled APC-ARM/Amer1-A2 complex progressively dissociated APC-ARM from Amer1-A2 in a dose-dependent manner (Figure 7b), which is consistent with our surmise. Although these APC-binding ligands look apparently very different at the first glance, a concealed consensus pattern XGGGD/EXΦΦ (X stands for any residue and Φ represents a hydrophobic residue) among these motifs was surprisingly found when we carefully compared their sequences (Figure 7c). The first residue varies among different APC-interacting motifs (T326 in Amer1-A1, S499 in Amer1-A2, Q366 in Amer1-A4, and Y380 in Sam68-YY). Its backbone is generally recognized by the formation of hydrogen bonds with the side chain amide group of APC-N641 (Q366 in Amer1-A4 is the exception), while its side chain normally forms polar contacts with APC-N679 and -W593 (Q366 in Amer1-A4 and Y380 in Sam68-YY are the exceptions).
The second, third, and fourth positions are generally glycines, which are contacted by the bulky side chains of W552 and W593, and hydrogen bonded with R549 and N594 of APC. Notably, the glycine residue at the fourth position of Amer-A1, -A2, and -A4 peptides (G529, G502, and G369, respectively) all adopt special dihedral angles in our crystal structures that enable them to access regions of the Ramachandran plot disallowed for the other nineteen kinds of amino acids (Supplementary Table S1). This is consistent with the fact that the fourth glycine is the most conserved one among the three glycine residues of the consensus motif (Figure 7c).
The fifth amino acid of the APC-ARM-binding consensus is an acidic one (D330 in Amer1-A1, D503 in Amer1-A2, E370 in Amer1-A4, E183 in Asef-ABR, and E384 in Sam68-YY), whose side chain carboxylate group is recognized by G511 and K516 of APC by hydrogen bonds and salt bridges. In contrast, the seventh and eighth sites of the consensus motif are hydrophobic residues (I332/A333 in Amer1-A1, L505/Y506 in Amer1-A2, M372/A373 in Amer1-A4, L185/A186 in Asef-ABR, and T386/T387 in Sam68-YY), which are surrounded by non-polar APC residues such as F458, M503, and F510 (Supplementary Figure S10). Therefore, these APC-binding partners all possess a common consensus motif to be recognized by the same assembly of APC-ARM surface groove residues.

Discussion
APC is a key regulatory molecule in the Wnt signaling pathway under both physiological and pathological conditions. It also has a crucial role in regulating the cytoskeleton formation and cell Figure 7 Comparison of the binding patterns of Amer1-A1, -A2, -A4, Asef-ABR, and Sam68-YY reveals a common recognition motif for APC-ARM association. (a) Structural superimposition of APC-ARM in complexes with its binding partners (Amer1-A1: cyan, Amer1-A2: green, Amer1-A4: magenta, Asef-ABR: yellow, and Sam68-YY: red). (b) Competition between Asef-ABR-SH3 and Amer1-A2 for binding to APC-ARM using the GST pull-down assay. (c) Structure-based alignment of the APC-binding sequences of human Amer1-A1, -A2, -A4, Asef-ABR, and Sam68-YY. The consensus APC-ARM binding motif XGGGD/EXΦΦ (X stands for any residue, and Φ represents a hydrophobic residue) is shown below the sequences.
Structures of the APC-Amer1 complex migration. The ARM domain of APC is critical for these diverse activities by mediating interactions with Amer1, Asef, IQGAP1, and so on to regulate these diverse functions. It is therefore of high appealing interest to determine the molecular basis of how APC-ARM accommodates these various binding partners. Intriguingly, although there is no apparent sequence similarity among the APC-binding regions A1, A2, and A4 of Amer1 and those of Asef and Sam68, our structural and mutational analysis revealed a common interaction mode shared by them and allowed us to deduce a consensus sequence for APC-ARM recognition (Figure 7c). We propose, and showed for the Asef and Amer1-A2 pair (Figure 7b), that these APC-ARM-binding partners are competitive with each other for APC-binding, similar to the case for β-catenin.
There are four APC-binding sites in Amer1, the previously described A1, A2, A3 [11] and the newly identified A4 in this study. These four sites are the most highly conserved regions of Amer1 (Figure 4d), implicating that the interaction between APC and Amer1 at these four sites is specially retained by the selection pressure throughout evolution [27]. Singlepoint mutations of key interface residues on APC or Amer1 attenuated their associations mediated through the four individual motifs (Figure 3), while composite mutations disrupting all four binding sites abrogated interactions of full-length proteins ( Figure 5) and abolished the recruitment of APC to the plasma membrane by Amer1 ( Figure 6).
By comparing the interaction modes of Amer1-A1, -A2, -A4, Asef-ABR, and Sam68-YY, a consensus APC-ARM-binding motif is proposed (Figure 7c). Side chain amide groups of four asparagine residues, N507/N550/N594/N641 from the H3 helices of APC armadillo repeats 2 through 4, hydrogen bond with the main-chain NH and CO groups of the APC-ARMbinding motif in a zipper-like manner. In addition, three glycines, one acidic residue, and two hydrophobic residues of the consensus APC-ARM-interacting motif are recognized by W553/W593, G511/K516, and F458/ M503/F510 of APC, respectively (Supplementary Figure S10). It would be intriguing to investigate whether other APC-ARM-associating proteins such as IQGAP1 [15], PP2A-B56α [29], KAP3 [30], and Striatin [31] also use similar mechanisms to recognize APC. However, it would not be surprising if any of them employs a different recognition mode, as Amer1-A3 has already provided such a scenario.
Amer1 is a multifunctional protein whose gene knockout in mice resulted in defective specification of mesenchymal progenitor cell fate, mainly through aberrant β-catenin activation [17]. As the Amer1-APC complex is also involved in regulating β-catenin protein stability [19,20], it would be worthwhile to investigate what effect specific disruption of the Amer1-APC complex would have on the Wnt/β-catenin signaling pathway. Both APC and Amer1 are scaffold proteins containing binding sites for both axin/conductin and β-catenin. Besides, there are four APC-binding sites in Amer1. Therefore, there might exist a gigantic multimeric APC-Amer1-axin-β-catenin complex inside the cell to collaboratively facilitate the degradation of β-catenin.
There are two paralogs of Amer1 in vertebrates, Amer2/FAM123A and Amer3/FAM123C [11,27], which were suggested to be also involved in embryonic development [32]. Similar to Amer1, they possess the APC-binding sequences A1, A2, and A4 as well (Figures 1a and d). Amer2 was reported to collaborate with APC to function in neuroectodermal patterning through regulating Wnt/β-catenin signaling [33], microtubule stability, and cell migration [34]. In contrast, Amer3 was found to be a positive regulator of Wnt/β-catenin signaling [35]. It would be interesting to look into what tissue-specific cellular or developmental defects would result if complex formation of APC-Amer2 or APC-Amer3 is specifically disrupted.
Zhenyi Zhang et al. 11 Crystallization and structure determination Crystallization experiments were performed at 14°C by the hanging-drop vapor-diffusion method. Crystals of the APC-ARM/Amer1-A1 complex were grown in 0.1 M MES, pH 6.5, and 12% PEG 20 000. Crystals of the APC-ARM/Amer1-A2 complex were obtained at 25% ethyleneglycol only. The APC-ARM/ Amer1-A4 complex was crystallized under the condition of 0.1 M sodium/potassium phosphate, pH 6.2, 0.2 M NaCl, and 10% PEG 8000. Crystal diffraction data sets were all collected at the beamline BL17U1 at Shanghai Synchrotron Radiation Facility (China), and processed using the HKL2000 software [36].
Crystals of the APC-ARM/Amer1-A1 complex belonged to the P2 1 2 1 2 1 space group, with one set of the complex in each asymmetric unit. The structure was determined at 1.90 Å by the method of molecular replacement with the CCP4 program Phaser [37], using the structure of APC-ARM by itself (PDB code: 3NMW) [10] as the searching model. After model-building by Coot [38] and refinement by the CCP4 program REFMAC [39,40], the final model has an R/R free factor of 18.06%/22.13%. In the Ramachandran plot, 99.4 and 0.6% of residues are in the most favored and allowed regions, respectively.
Crystals of the APC-ARM/Amer1-A2 complex belonged to the P1 space group, with six sets of complexes in the asymmetry unit. The structure was determined at 2.10 Å using the same method as above. After refinement, the model has an R/R free factor of 19.34%/21.27%. In the Ramachandran plot, 99.1 and 0.9% of residues are in the most favored and allowed regions, respectively.
Crystals of the APC-ARM/Amer1-A4 complex belonged to the P2 1 2 1 2 1 space group, with one set of the complex in each asymmetry unit. The structure was determined at 1.70 Å using the same method as above. After refinement, the model has an R/R free factor of 19.67%/23.98%. In the Ramachandran plot, 99.1 and 0.9% of residues are in the most favored and allowed regions, respectively.
The model qualities were all checked with the PROCHECK program [40].
GST pull-down assays GST pull-down assays between wild type (WT) or mutant APC-ARM proteins and WT/mutant Amer1 fragments were performed according to standard procedures as described previously [41]. The APC-ARM protein was added to preimmobilized GST-Amer1 fragments on the GST affinity column at 4°C, and then washed extensively using the GST column binding buffer (25 mM Tris-HCl, pH 8.0, 250 mM NaCl, 1 mM EDTA, and 14 mM β-mercaptoethanol). The bound proteins were eluted with the GST column elution buffer (50 mM Tris-HCl, pH 8.0, 300 mM NaCl, and 7 mM glutathione), and then analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and Coomassie Blue staining.

ITC assays
ITC experiments were performed using an ITC200 system (GE Healthcare) at 25°C as described previously [42]. The buffer contained 50 mM HEPES, pH 7.5, 300 mM NaCl, and 1 mM EDTA. Typically, 200 μM APC-ARM protein was injected 20 times in 2 μl aliquots into a 200 μl sample cell containing GST-Amer1-A1/A2/A3/A4 protein at a concentration of 20 μM, or 1 mM GST-Amer1-A1/A2/A3/A4 protein was injected 20 times in 2 μl aliquots into a 200 μl sample cell containing APC-ARM protein at a concentration of 80 μM. Data were fit with a nonlinear least-square routine using a single-site binding model with Origin for ITC version 7.0 (MicroCal, Worcestershire, UK), varying the stoichiometry (n), the enthalpy of the reaction (ΔH), and the association constant (K a ).

In vitro binding competition assay
The competition assay between Asef-ABR-SH3 and Amer1-A2 for binding to APC-ARM was performed by the GST pulldown method. Four milliliters of the GST-Amer1-A2/APC-ARM complex, with the concentration of 0.5 mg ml − 1 , was preloaded onto the GST affinity column (1 ml) at 4°C. 0.1 mg ml − 1 , 0.2 mg ml − 1 , 0.5 mg ml − 1 , and 1 mg ml − 1 of the ABR-SH3 domain (residues 170-271) of Asef in a volume of 0.5 ml were then added to the reactions as competitors. After extensive washing by the GST column binding buffer, the bound proteins were eluted and analyzed by SDS-PAGE and Coomassie blue staining.

Structures of the APC-Amer1 complex
Preparation of protein lysates, immunoprecipitation, and western blotting