Analysis of binding interfaces of the human scaffold protein AXIN1 by peptide microarrays

Intrinsically disordered regions (IDRs) are protein regions that lack persistent secondary or tertiary structure under native conditions. IDRs represent >40% of the eukaryotic proteome and play a crucial role in protein–protein interactions. The classical approach for identification of these interaction interfaces is based on mutagenesis combined with biochemical techniques such as coimmunoprecipitation or yeast two-hybrid screening. This approach either provides information of low resolution (large deletions) or very laboriously tries to precisely define the binding epitope via single amino acid substitutions. Here, we report the use of a peptide microarray based on the human scaffold protein AXIN1 for high-throughput and -resolution mapping of binding sites for several AXIN1 interaction partners in vitro. For each of the AXIN1-binding partners tested, i.e. casein kinase 1 ϵ (CK1ϵ); c-Myc; peptidyl-prolyl cis/trans isomerase, NIMA-interacting 1 (Pin1); and p53, we found at least three different epitopes, predominantly in the central IDR of AXIN1. We functionally validated the specific AXIN1–CK1ϵ interaction identified here with epitope-mimicking peptides and with AXIN1 variants having deletions of short binding epitopes. On the basis of these results, we propose a model in which AXIN1 competes with dishevelled (DVL) for CK1ϵ and regulates CK1ϵ-induced phosphorylation of DVL and activation of Wnt/β-catenin signaling.

Intrinsically disordered regions (IDRs) are protein regions that lack persistent secondary or tertiary structure under native conditions. IDRs represent >40% of the eukaryotic proteome and play a crucial role in protein-protein interactions. The classical approach for identification of these interaction interfaces is based on mutagenesis combined with biochemical techniques such as coimmunoprecipitation or yeast two-hybrid screening. This approach either provides information of low resolution (large deletions) or very laboriously tries to precisely define the binding epitope via single amino acid substitutions. Here, we report the use of a peptide microarray based on the human scaffold protein AXIN1 for high-throughput and -resolution mapping of binding sites for several AXIN1 interaction partners in vitro. For each of the AXIN1-binding partners tested, i.e. casein kinase 1 ⑀ (CK1⑀); c-Myc; peptidyl-prolyl cis/trans isomerase, NIMA-interacting 1 (Pin1); and p53, we found at least three different epitopes, predominantly in the central IDR of AXIN1. We functionally validated the specific AXIN1-CK1⑀ interaction identified here with epitope-mimicking peptides and with AXIN1 variants having deletions of short binding epitopes. On the basis of these results, we propose a model in which AXIN1 competes with dishevelled (DVL) for CK1⑀ and regulates CK1⑀induced phosphorylation of DVL and activation of Wnt/␤catenin signaling.
Proper spatial and temporal organization of molecules acting in the cell signaling pathways is crucial for specific and efficient information transfer. Scaffold proteins coordinate the action of the signaling components by bringing them together into multiprotein complexes, thereby increasing their proximity and effective concentration (1). To bind multiple partners, scaffold proteins usually contain several modular structured domains, often in combination with regions that do not adopt secondary structures under native conditions, termed intrinsically disordered regions (IDRs) 3 (2). IDRs significantly increase binding capacity of the scaffold due to their flexibility and ability to mediate protein-protein interactions via relatively short linear peptide motifs (3,4).
Tumor suppressor AXIN1 is a representative of intrinsically disordered scaffold proteins. It consists of N-terminal RGS and C-terminal DIX (Dishevelled, Axin) structured domains that are separated by the large "central" IDR (5). Besides its central role in Wnt signaling, AXIN1 has been identified as a component of several different pathways, including p53, c-Myc, transforming growth factor ␤, and c-Jun N-terminal kinase (for reviews, see Refs. 3 and 6). To gain mechanistic insight into the formation of the signaling complexes (i.e. which proteins can bind simultaneously/exclusively), it is essential to map binding epitopes for the interaction partners on the scaffold at high resolution, which is complicated to achieve with commonly used biochemical methods such as coimmunoprecipitation or yeast two-hybrid screening.
A high-throughput screening technique derived from oligonucleotide microarray (7,8), in which peptides were immobilized via a linker on a solid surface, has been used recently to achieve this goal and to profile the structural requirements of the binding of protein domains (9,10). For systematic mapping of protein-binding epitopes, IDRs of protein of interest are chopped into small (overlapping) peptides, which are immobilized via a linker on a solid surface, usually a glass or plastic chip. Posttranslational modifications, such as phosphorylation, can be incorporated in a site-specific manner by organic synthesis procedures into the peptides involved in the array. For the detection of protein-peptide interactions, the array can be incubated either with a purified recombinant protein or a crude cell lysate that can serve as the source of protein (11). Finally, the protein partner bound to the array can be easily detected using an immunofluorescence technique. The array format allows a high-throughput approach for determination of short peptide motifs within the (scaffold) protein of interest that are responsible for docking of the protein partners involved in the complex.
In this study, we present the use of a peptide microarray for detailed in vitro mapping of interaction epitopes within the IDRs of the scaffold protein AXIN1. Using this approach, we identified putative binding interfaces in the IDRs of human AXIN1 with several interaction partners (casein kinase 1⑀ (CK1⑀); c-Myc; p53; and peptidyl-prolyl cis/trans isomerase, NIMA-interacting 1 (Pin1)) at high resolution. Subsequent validation of these findings allowed us to propose a model for AXIN1-dependent regulation of Wnt/␤-catenin signaling by CK1⑀ via dishevelled (DVL).

AXIN1 contains several binding epitopes for CK1⑀, c-Myc, p53, and Pin1
All Axin proteins have a conserved structure with two welldefined domains: an N-terminally located RGS domain (aa 88 -211 in human AXIN1) and a C-terminally located DIX domain (aa 780 -862 in human AXIN1). There are also two IDRs in the rest of protein: (i) the N-terminal IDR (aa 1-87) and (ii) the central IDR (aa 212-779) (Fig. 1A, top).
The binding regions for biologically very important AXIN1 interaction partners, namely CK1⑀, c-Myc, p53, and Pin1, have been mapped earlier by deletion mutagenesis (for schematics see Fig. 1A, bottom). Not surprisingly, these proteins regulating distinct biochemical processes bind to the central IDR of Figure 1. Domain structure of AXIN1 and its interaction partners. A, top, secondary structure prediction of human AXIN1 (hAXIN1) protein performed by PONDR-FIT software (33). Values Ͻ0.5 indicate structured protein regions (ordered; in gray color); values Ͼ0.5 indicate intrinsic parts lacking secondary structure (disordered; in purple color). RGS and DIX domains are indicated, and the central IDR is depicted in purple. Bottom, overview of binding regions of known selected AXIN1-interacting proteins from various signaling pathways reported in the literature: CK1⑀, c-Myc, Pin1, and p53, and DVL, respectively. B, the general workflow of a peptide microarray approach used to study protein-protein interactions mediated via IDRs: a glass slide containing immobilized peptides is incubated with the protein of interest followed by incubation with primary antibody against the recombinant protein. The positive spots are then visualized with a fluorescently labeled secondary antibody. See "Experimental procedures" for details.

Binding interfaces of AXIN1
AXIN1. It is obvious, just simply for steric reasons, that only a very limited number of proteins can be associated with the AXIN1 scaffold at the same time, and it is likely that one AXIN1 complex is dedicated to only one biological function. To understand better how AXIN1 achieves this, we decided to map binding epitopes within the IDRs of AXIN1 with high resolution. We designed a peptide microarray containing 13-mer peptides overlapping by 10 residues covering the entire human AXIN1 sequence (Fig. S1) that were immobilized via a linker molecule on a microscopic slide. To perform a single peptide microarray experiment, the glass slide with immobilized peptides was incubated with the recombinant protein of interest followed by antibody staining and visualization using a fluorescence marker. The signal was subsequently detected by a high-resolution fluorescence scanner and analyzed using spot-recognition software (schematized in Fig. 1B).
To identify the peptide epitopes required for direct interaction with AXIN1, the following candidates were used for a peptide microarray experiment: CK1⑀, c-Myc, p53, and Pin1. These candidates were chosen as the representative AXIN1-binding partners from individual signaling pathways (Fig. 1A, bottom). Microarray screening analysis ( Fig. 2; for an example of raw data, see Fig. S2) revealed that each interaction partner of AXIN1 can efficiently interact with at least three linear peptide motifs. All tested proteins showed interaction with at least one of two linear peptides forming solvent-exposed ␣-helix (aa 133-152 and aa 154 -175 in human AXIN1) from the RGS domain; here, however, the interpretation is not so straightforward because the linear peptides may not truly recapitulate the situation in the folded RGS domain. However, this information can be still relevant because the epitopes identified within the RGS domain were located on the ␣-helices and exposed to the solvent. Peptides mapping to the DIX domain were not considered as putative binding sites of c-Myc (see Fig. 2B) as they were located predominantly neither on the surface nor on the "linear" ␣-helix (see "Experimental procedures" for more details about quantification). All the remaining binding sites of the aforementioned candidate proteins were located in the central IDR of AXIN1 (Fig. 2, A-D).
The identified epitopes narrow down the regions involved in the interaction from several hundred aa reported in the literature to the ϳ20-aa-long motifs. Specifically, for CK1⑀, two binding sites (located in the central IDR) previously mapped to (i) the N-terminal part of IDR, aa 198 -353 in mouse AXIN1 (corresponding to aa 198 -353 in human AXIN1) (12), and (ii) the C-terminal part of IDR in different studies attributed to aa 509 -832 in mouse AXIN1 (corresponding to aa 511-862 in human AXIN1) (12), aa 475-676 in human AXIN1 (13), or aa 495-684 in mouse AXIN1 (corresponding to aa 495-683 in human AXIN1) (14) could now be mapped with high resolution to (i) aa 274 -295 and (ii) aa 592-616 ( Fig. 2A). For c-Myc, the previously reported large region aa 331-777 in human AXIN1 (15) was narrowed down to aa 724 -743, and an additional epitope, aa 280 -304, in the central IDR was discovered (Fig.  2B). For p53, the initially reported interaction region aa 210 -337 in mouse AXIN1 (corresponding to aa 210 -337 in human AXIN1) (16) was reduced to aa 274 -304, and one novel binding epitope, aa 364 -382, in the C-terminal part of the central IDR was identified (Fig. 2C). Although Pin1 is known to interact with AXIN1, no information on binding interface has been available thus far. The microarray screening analysis indicated two binding sites for Pin1 in the C-terminal part of the central IDR of AXIN1 (Fig. 2D).
Despite that each analyzed interaction partner (with notable exception of Myc) of AXIN1 showed a unique interaction motif within the central IDR, aa 592-616 for CK1⑀, aa 637-658 for Pin1, and aa 364 -382 for p53, there was a significant overlap in the remaining sites. Namely, CK1⑀, c-Myc, and p53 partially overlapped in aa 274 -304, whereas c-Myc and Pin1 overlapped in aa 721-742. These results provide a rationale for the molecular mechanism that ensures mutually exclusive recruitment of AXIN1 to the distinct molecular complexes involved in Wnt signaling, c-Myc degradation, or p53-mediated transcription.

The binding epitopes aa 154 -175, 274 -295, and 592-616 are required for the recruitment of CK1⑀ to AXIN1 complex
The AXIN1 epitopes identified with the help of peptide arrays agreed very well with the previously reported interaction regions of AXIN1 obtained by deletion mapping. This suggests that our results are reliable and can be used as a basis for future studies and for the design of an intervention in a particular binding event. To provide a proof of principle that it is indeed the case, we decided to functionally validate the importance of epitopes involved in the interaction of AXIN1 with CK1⑀, one of the key kinases in the Wnt/␤-catenin pathway (17).
With the peptide microarray screen, we identified three CK1⑀-binding epitopes in AXIN1 comprising aa residues 154 -175, 274 -295, and 592-616, herein referred to as epitopes "1," "2," and "3" (Figs. 2A and 3A). To verify the results of the AXIN1-CK1⑀ interaction interfaces from the peptide microarray assay, we prepared seven AXIN1 deletion mutants: three variants, each lacking one of the CK1⑀ epitopes; three variants lacking two of the three CK1⑀ epitopes; and a variant lacking all three epitopes (Fig. 3B). Mutations of the epitopes within the central IDR of AXIN1 were designed as precise in-frame deletions of aa 274 -295 and 592-616 (⌬2 and ⌬3 constructs). Because the deletion of a peptide motif within the RGS domain might compromise its three-dimensional structure, the RGS domain was removed completely to obtain mutant ⌬1, which lacks the binding site aa 154 -175 (Fig. 3B).
The AXIN1 constructs were expressed as N-terminally Myctagged fusion proteins in transiently transfected HEK293 cells, and recruitment of endogenous CK1⑀ to the complexes formed on mutant AXIN1 scaffold was tested using coimmunoprecipitation. This approach revealed that deletion of binding sites gradually reduced interaction between AXIN1 and CK1⑀ but also showed that only the loss of all three epitopes abolished the binding completely (Fig. 3C). These results confirm the functional importance of the AXIN1 epitopes identified by peptide arrays for CK1⑀ binding and provide validation to the results from the AXIN1 peptide array.

AXIN1-derived binding epitope peptides inhibit interaction of CK1⑀ and AXIN1
Identification of the interaction interfaces at high resolution in principle provides the possibility to use an approach based on

Binding interfaces of AXIN1
competitive blocking of the interaction using epitope-mimicking peptides. We synthesized three peptides corresponding to the CK1⑀-binding sites on AXIN1. We named these peptides corresponding to aa 154 -175, 274 -295, and 592-616 in human AXIN1 as 1, 2, and 3 ( Fig. 4A). To investigate whether these peptide reagents can be used to study the functional importance of CK1⑀-AXIN1 interaction, we aimed to analyze the capacity of these peptide "competitors" to interfere with the CK1⑀-AXIN1 binding.
First, we investigated whether the peptide competitors impair CK1⑀ kinase activity. Not surprisingly, in vitro kinase assays showed that, in the presence of AXIN1-derived peptides, CK1⑀ was still able to phosphorylate its substrate, casein, as well as undergo autophosphorylation (Fig. 4B).
Second, we tested the capacity of AXIN1-derived peptides to block the interaction of CK1⑀ and AXIN1. Binding of CK1⑀ and AXIN1 in HEK293 cell lysates was tested by a modified coimmunoprecipitation approach where a mixture of the three AXIN1 peptides, 1ϩ2ϩ3, stabilized by PEGylation at their C termini was added to the fresh HEK293 cell lysate (Fig. 4C). Indeed, addition of the peptides decreased the amount of endogenous AXIN1 coprecipitated with either endogenous or overexpressed CK1⑀ from the cell lysate (Fig. 4D). These results demonstrate that the peptide blockers can compete with AXIN1 for CK1⑀ binding and that they are able to disrupt AXIN1-CK1⑀ interaction in cell lysates.

Dissociation of CK1⑀ from AXIN1 complex promotes its binding to DVL
CK1⑀ is known to interact with several other cytoplasmic components of the Wnt pathway. Among them, DVL protein, the key Wnt signal mediator, is probably the best defined CK1⑀ target (18,19). Interestingly, AXIN1 and DVL exhibit opposite functions in the Wnt/␤-catenin pathway: AXIN1 serves as a key component of the ␤-catenin destruction complex, whereas DVL mediates dissolution/inhibition of the destruction complex upon pathway activation.
We thus hypothesized that release of CK1⑀ from AXIN1 may be accompanied by changes in its interaction with DVL and decided to test how the presence of WT AXIN1 or AXIN1 ⌬2ϩ3 variant, with decreased capacity to bind CK1⑀, affects the interaction of CK1⑀ with DVL. The RGS domain in AXIN1 was kept intact to prevent the interaction with another component of the destruction complex, APC, that binds via this domain (20). In HEK293 cells, EGFP-tagged AXIN1 constructs were coexpressed either with FLAG-tagged DVL3 or FLAG-tagged Shp2, an unrelated protein, as a control. The endogenous CK1⑀ was precipitated from the lysates with anti-CK1⑀ antibody, and the interacting DVL3 and AXIN1 were detected on an immunoblot using anti-FLAG and anti-GFP antibodies, respectively (Fig. 5A). The analysis confirmed the previous observation that CK1⑀ binds to AXIN1 ⌬2ϩ3 construct with lower affinity than to full-length AXIN1. Moreover, upon coexpression of FLAG-DVL3, overall binding of CK1⑀ to AXIN1 constructs decreased, which suggests that indeed DVL and AXIN1 compete for CK1⑀. However, at the same time, the analysis revealed that decreased binding of CK1⑀ to AXIN1 ⌬2ϩ3 was accompanied with increased CK1⑀ binding to DVL3 (Fig. 5A). Similar data, i.e. increased binding of CK1⑀ to DVL3, were reproduced when we interfered with the CK1⑀-AXIN1 interaction in HEK293 cell lysates by the AXIN1 peptide competitors 1ϩ2ϩ3 (Fig. 5B).
These data suggest that AXIN1 sequesters CK1⑀ and limits its availability for DVL, but the functional consequences of this competition are unclear. To address this question, we decided to test how CK1⑀ binding-compromised AXIN1 ⌬2ϩ3 variant differs from WT AXIN1. For example, can the presence of AXIN1 affect the phosphorylation of DVL3 by CK1⑀? We took advantage of the phosphospecific antibody against pSer 643 -DVL3 that serves as a good readout of DVL3 phosphorylated by CK1⑀ (21). When we coexpressed CK1⑀ with AXIN1 WT and ⌬2ϩ3 variant, pSer 643 -DVL3 phosphorylation signal was decreased by AXIN1 WT but not by AXIN1 ⌬2ϩ3 variant (Fig.  5C). This is in line with the observations in Fig. 4, A and B, and suggests that AXIN1 can, via control of CK1⑀ availability, regulate phosphorylation of DVL and activation of the Wnt/␤catenin pathway.
To further validate this assumption, we compared AXIN1 WT and AXIN1 ⌬2ϩ3 mutant behavior in the Dual-Luciferase TopFlash/Renilla reporter gene assay (21,22) that monitors the ␤-catenin/TCF-dependent transcription. In these assays, AXIN1 ⌬2ϩ3 variant behaved differently than AXIN1 and (i) promoted ␤-catenin/TCF-dependent transcription even in the absence of the external stimuli (Fig. 5D, left) and (ii) behaved as a less potent inhibitor of Wnt-3a-driven transcription when compared with AXIN1 WT (Fig. 5D, right). These events relate to the cytoplasmic Wnt pathway complexes because neither WT AXIN1 nor AXIN1 ⌬2ϩ3 mutant was able to interfere with the TopFlash signal triggered by constitutively activated ␤-catenin S33A (Fig. 5E). In summary, our data demonstrate that AXIN1 sequesters CK1⑀ and via this mechanism controls its capacity to stimulate the Wnt pathway via DVL.

Discussion
In this study, we applied peptide microarrays to identify the interaction epitopes for AXIN1-binding partners CK1⑀, c-Myc, p53, and Pin1. Our peptide microarray-based results revealed that many of the binding sites overlap: at least one "common" binding interface was found to be shared, and one or more binding interfaces were unique for each binding partner. This observation opens the possibility that AXIN1 exploits just few dedicated regions in its sequence for the binding to other protein partners. After binding of a single partner protein/protein . The charts show normalized signal values from single peptide spots, and the dashed line represents the threshold signal intensity. AXIN1 is schematically depicted above the charts with the numbers and sequences of amino acids defining the particular binding sites. The signals from control glass (antibodies only) were subtracted from the experimental values, and the four strongest signals of at least four consecutive (i.e. neighboring and overlapping) peptides higher than 10,000 LUs were considered as candidate binding sites; see "Experimental procedures" and main text for more details. For each protein of interest, the average from two to three replicates is shown. hAXIN1, human AXIN1.

Binding interfaces of AXIN1
complex, likely to appear by combination of simultaneous interaction with three binding interfaces to the scaffold, the structure of scaffold protein (here human AXIN1) can adopt the unique 3D fold leading to one particular "cell signaling message." This explanation would help in understanding how AXIN1 distinctively influences multiple signaling pathways and events in a cell.
For CK1⑀, we validated the results from peptide microarray screen using deletion mutagenesis and small peptide competitors. These data provide a proof of principle that relatively minor deletions in AXIN1 can generate mutants that will be biased or even restricted in function for a certain signaling cascade. The same outcome can be achieved, at least in cell lysates, with the combination of small peptides corresponding to the binding epitopes that, in a specific manner, interact with AXIN1-binding partners and decrease their abundance in the AXIN1-based complexes.
We observed that the functional consequence of weakening the interaction between CK1⑀ and AXIN1 is the increased recruitment of CK1⑀ to DVL3. It is an important observation because the phosphorylation of DVL proteins induced by Wnt ligands and mediated via action of CK1⑀ is the key event that mediates signal transduction in the Wnt/␤-catenin pathway (19,21,23). Local availability of CK1⑀ thus represents an important regulatory mechanism that can control activation of the Wnt pathway.
The biological importance of our observations remains to be defined. The binding epitopes identified on AXIN1 and subsequently deleted in AXIN1 ⌬2ϩ3 variant, which has lowered ability to bind CK1⑀, do not obviously interfere with other known functions of AXIN1 in the Wnt signaling pathway, such as associations with APC (mediated via the RGS domain (20)), ␤-catenin, or GSK3 that were mapped to the central IDR of AXIN1 (aa 438 -508 in human AXIN1 for the binding of ␤-catenin (24) and aa 383-401 in human AXIN1 for the binding of GSK3␤ (25,26)). The deleted regions are also distinct from the parts of AXIN1 that were reported to be involved in the formation of "closed" and "open" conformations of AXIN1 (27). Open/closed dynamics of AXIN1 is mediated via interaction of the DIX domain with the ␤-cateninbinding domain (aa 435-498 in human AXIN1) (24) and regulated through phosphorylation of four conserved serine residues (including Ser 497 and Ser 500 in mouse Axin1, which correspond to Ser 493 and Ser 496 in human AXIN1). Phosphorylation induced by GSK3␤ promotes the open (i.e. active) AXIN1 conformation (27). The three identified CK1⑀-binding epitopes in AXIN1 are distinct.
Nevertheless, AXIN1 ⌬2ϩ3 variant (compared with WT AXIN1) was able to induce Wnt/␤-catenindependent transcription but failed to reduce phosphorylation of Ser 643 in DVL3 (21) or to efficiently inhibit Wnt-3a-induced transcription (Fig. 5C). This suggests that interaction of AXIN1 and CK1⑀ is required for the efficient regulation of the inhibitory   Input panels represent 10% of the initial material. Numbers on the right correspond to the molecular mass marker in kDa. TCL, total cell lysate.

Binding interfaces of AXIN1
activity of AXIN1 in the Wnt/␤-catenin pathway. This may seem surprising given the previous reports that described formation of a trimeric AXIN1-DVL-CK1⑀ complex (18) and recruitment of endogenous AXIN1 to DVL3 upon CK1⑀ overexpression (19). These data can be reconciled by the existence of (at least) two distinct complexes containing both AXIN1 and CK1⑀. Direct interaction of AXIN1 with CK1⑀ can be important for the proper function and regulation of the destruction complex but is dispensable for recruitment of AXIN1 to the DVLbased "signalosome" complex. CK1⑀ binds DVL directly (18), and CK1⑀-mediated phosphorylation of DVL results in the more efficient recruitment of AXIN1 to the signalosome (19). Recruitment of AXIN1 to the signalosome, connected with the inhibition/inactivation of the destruction complex, is mediated and critically dependent on the direct interaction between DIX domains of AXIN1 and DVL (28), and as such, AXIN1 in the signalosome does not need to directly bind CK1. Future work must elucidate whether the proposed mechanism is unique to CK1⑀ and the closely related CK1␦ (connected primarily with DVL phosphorylation) or whether it applies also to CK1␣ that seems to a principal CK1 component in the destruction complex in vivo (29). Taken together, our AXIN1 peptide microarray-based results (specifically those from CK1⑀-AXIN1 interaction) open up the possibility to design and generate AXIN1 variants and specific peptide competitors that can direct signaling activities of AXIN1. Using this approach, we describe a novel regulatory mechanism limiting the availability of CK1 for DVL phosphorylation mediated by direct interaction of AXIN1 and CK1⑀.

Peptide array
Generation of the peptide arrays-The peptide library comprised 13-mer peptides overlapping with 10 residues that cover the whole sequence of human AXIN1 (Fig. S1). The designed peptide library was generated by JPT Peptide Technologies GmbH (Berlin, Germany) in an array format wherein the peptides were immobilized on a glass slide (25 ϫ 75-mm slides). Briefly, all peptides were synthesized in a stepwise manner (SPOT-synthesis) on a cellulose membrane with a JPT-developed fully automated robotic system, resulting in a defined arrangement (3 ϫ 284 peptide spots per slide). By coupling a reactivity tag (tag ϩ linker-Ttds-linker molecule, where Ttds stands for trioxatridecan-succinamic acid) on the N terminus of the peptides (truncated side products are capped by acetylation steps), all target peptides could be immobilized chemoselectively and purified by reaction of the peptides with the modified glass surface. An N-terminal reactivity tag for immobilizing the peptides onto microarray slides ensures that only full-length peptides are bound to the final chip after cleavage from the cellulose membrane. The peptides were transferred onto the slide by a contact printing technique. The resulting information of a covalent bond between the target peptide and the chip surface allows removal of all truncated (and acetylated) sequences by subsequent washing steps. The peptide microarray was printed in three identical subarrays. This enables efficient intrachip-reproducibility tests. Each peptide subarray was printed in individual blocks. Each spot in the microarray repre-

Binding interfaces of AXIN1
sents a single individual peptide. The positive controls were represented by the spots in the bottom row with three adjacent spots for each control (e.g. human IgG or mouse IgG). Screening of the peptide arrays for binding to protein of interest-In each experiment, two glass slides were used: control, incubated with antibodies only, and experimental, incubated with the recombinant protein of interest. Each individual slide was assembled in a sandwich-like format with a dummy/ blank slide and two plastic spacers to create a simple incubation chamber with a reaction volume of Ϸ300 l. The slides were first blocked with 300 l of SmartBlock solution (CANDOR Bioscience, 113125) for 1 h at 30°C. Then the experimental slide was incubated overnight with 10 g/ml recombinant protein, CK1⑀ (OriGene Technologies, TP302436), p53 (OriGene Technologies, TP300003) Pin1 (OriGene Technologies, TP302543), or c-Myc (Abnova, H00004609-P01), in Smart-Block solution (final volume, 300 l) at 4°C. The next day, both slides were washed in 300 l of TBS buffer (4 ϫ 10 min, room temperature) and then incubated with 1 g/ml primary antibody in SmartBlock solution (final volume, 300 l) for 1 h at 30°C. Then the slides were washed again in 300 l of TBS buffer (4 ϫ 10 min, room temperature) followed by incubation with 1 g/ml secondary fluorescent antibodies in SmartBlock solution (final volume, 300 l) for 45 min at 30°C and subsequently washed in 300 l of TBS buffer (4 ϫ 10 min, room temperature) and in 300 l of water (4 ϫ 10 min, room temperature). To remove excess water, the slides were dried by centrifugation (1200 relative centrifugal force for 2 min at room temperature with slides placed in 50-ml Falcon tubes). The bound recombinant protein was detected by reading the fluorescence intensity of the peptide spots. In these experiments, the following primary antibodies were used (1 g of antibody/ml): anti-CSNK1E (Abnova, 2276), anti-p53 (Abcam, 17869), anti-Pin1 (Abnova, 1080), and anti-glutathione S-transferase (Amersham Biosci-

Binding interfaces of AXIN1
ences, PA92002). For the secondary antibody (1 g of antibody/ ml), DyLight 649 anti-mouse-IgG (Pierce, 35515) was used. The slide data were analyzed using PepSlide Analyzer software (Sicasys). Representative images of the peptide array are in Fig. S2.

Quantification of the signal
Intensities of corresponding spots on the control glass (i.e. incubated with antibodies only) were subtracted from the experimental values. The experiment was repeated three times (two to three technical replicates each), and averages of the signal values were plotted in the graph (light units (LU) on y axis).
There are no strictly given criteria that define the binding epitope in a peptide array. Our parameters included (i) three to four consecutive peptides with signal exceeding 10,000 and (ii) a maximum of four of the strongest sites per protein. These parameters were set arbitrarily, and the final decision, which site to consider as a binding epitope for further testing, was made upon visual inspection of the data. The peptide array approach in principle enables searching for binding epitopes within linear (intrinsically disordered) protein regions. Therefore, the signals within structured domains were not taken into consideration except for the two sites in the RGS domain that have been mapped to the linear ␣-helices exposed to the solvent.
The following ad hoc exceptions in the central IDR were made. CK1⑀-binding site aa 274 -295 in human AXIN1 (although possessing one central peptide below 10,000 LU) was considered as putative, according to previously publishedok;1 results (12). c-Myc-binding site aa 724 -743 in human AXIN1 was also considered as putative, according to previously published results (15).

Plasmid preparation
Full-length AXIN1 coding sequence was amplified from human cDNA using forward primer 5Ј-CACCATGAATATC-CAAGAGCAGG-3Ј and reverse primer 5Ј-TCAGTCCACCT-TCTCCACTT-3Ј (where bold type represents the start codon in the forward primer and the stop codon in the reverse primer). For the truncated version of AXIN1 lacking the RGS domain, forward primer 5Ј-CACCGGCAGAGACGCTGCTC-CCC-3Ј was used. The PCR products were inserted into pENTR/TEV/D-TOPO vector. For in-frame deletions of the CK1⑀-binding sites within the AXIN1 IDR, a PCR-based mutagenesis approach was used with the following pairs of primers: 5Ј-CTCCCTCAGAAGCTGCTCTCCTGGCGGGA-GCCAGTCA-3Ј and 5Ј-TGACTGGCTCCCGCCAGGAGA-GCAGCTTCTGAGGGAG-3Ј for site 2 and 5Ј-GCCCC-CAACGCCAGTGATACCGAGGTGCCAGGTGCCT-3Ј and 5Ј-AGGCACCTGGCACCTCGGTATCACTGGCGTTGGG-GGC-3Ј for site 3. For expression in mammalian cells, the AXIN1 constructs were inserted from pENTR/TEV/D-TOPO vector into pDEST-Myc and pDEST-EGFP vectors using the Gateway cloning system. All mutations described in this study were verified by sequencing.

Dual-Luciferase TopFlash/Renilla reporter assay
Cells (seeded at 60,000/well in 24-well plates a day before; 500 l of DMEM/well) were transfected with 0.1 g of Super8X TopFlash construct and 0.1 g of Renilla luciferase construct/ well in a 24-well plate 24 h after seeding. For performing this assay, the Promega Dual-Luciferase assay kit (Promega, E1910) was used according to the manufacturer's instructions. Luminescence was measured by a Hidex Bioscan Plate Chameleon luminometer. Results are depicted as the ratio of TopFlash and Renilla signal (TopFlash -fold induction), which was normalized in Fig. 5E to the control column with ␤-catenin S33A only (normalized TopFlash -fold induction). Data were analyzed by MS Excel 2007 and GraphPad Prism 6, and results are shown as means Ϯ S.D. (number of experiments indicated ad hoc).

Coimmunoprecipitation
The coimmunoprecipitation protocol was used as described previously in Bryja and co-workers (21) The antibodies used for immunoprecipitation were as follows (1.0 g of antibody/sample): goat anti-CK1⑀ (Santa Cruz Biotechnology, sc-6471) and mouse anti-Myc (Santa Cruz Biotechnology, sc-40).
Modified coimmunoprecipitation with the competitor peptides was performed as follows. DMEM was removed 24 h after transfection, cells were washed with PBS, and 1 ml of cold NP-40 -based lysis buffer supplemented with 1 ϫ protease inhibitors (from 1000ϫ concentrated original stock; Roche Applied Science, 11836145001) and 1ϫ phosphatase inhibitors (from 200ϫ concentrated original stock; Calbiochem, 524625) was used per 10-cm dish. Lysate was collected after 15 min of lysis at 4°C and cleared by centrifugation at 16.1 relative centrifugal force for 30 min. 60 l of supernatant were then acquired and mixed with 15 l of 5ϫ Laemmli buffer, representing total cell lysate samples. The remaining supernatant was then divided into separate Eppendorf tubes (300 l/sample) and incubated at room temperature for 30 min on a carou-

Binding interfaces of AXIN1
sel with the PEGylated peptides added according to the scheme (final concentration of each peptide, 10 M; diluted in complete NP-40 -based lysis buffer at a stock concentration of 1 mM). Samples were then incubated with the antibody (1.0 g of antibody/sample) for 1 h at 4°C on a carousel. Then 30 l of protein G-Sepharose beads (GE Healthcare, 17-0618-05) equilibrated in BSA-free complete NP-40 -based lysis buffer were added to each sample (final volume, ϳ300 ϩ (3 ϫ 1) ϩ 1 ϩ 30 Ϸ 350 l). After overnight incubation on a carousel (4°C), samples were washed four times in NP-40 -based lysis buffer (800 l/wash), and 40 l of 2ϫ Laemmli buffer were added. All samples were boiled and processed for Western blotting (WB).

In vitro kinase assay
2 g of recombinant CK1⑀ (OriGene Technologies) were mixed with 2 g of dephosphorylated casein in reaction buffer containing 1ϫ magnesium/ATP mixture (Millipore, 20-113), 50 mM NaCl, 0.5 mM DTT, 100 g/ml BSA, 30 mM HEPES, pH 7.5, and [␥-32 P]ATP (final activity, 5 Ci) either in presence or absence of 4 M each AXIN1 peptide (1ϩ2ϩ3; aa sequences GIVSRQTKPATKSFIKGCIMKQ, ETAAPRVSSSRRYSEGRE-FRYG, and LAHSGKVGVACKRNAKKAESGKSAS). After 60-min incubation at 37°C, the reaction (final volume, 20 l) was stopped by addition of SDS sample buffer (5 l of 5ϫ SDS buffer) and heating at 95°C for 5 min. The samples were loaded on a 12% polyacrylamide gel and separated by SDS-PAGE. The gel was fixed and stained with Coomassie Blue, and phosphorylation was visualized using autoradiography.

Multiple sequence alignment
The multiple sequence alignment of selected sequences was performed using the ClustalW algorithm. The output alignment was refined manually using the BioEdit v7.0.1 sequence editor.