Identification of a Link between the SAMP Repeats of Adenomatous Polyposis Coli Tumor Suppressor and the Src Homology 3 Domain of DDEF*

The adenomatous polyposis coli (APC) tumor suppressor protein is a multifunctional protein with a well characterized role in the Wnt signal transduction pathway and in cytoskeletal regulation. The SAMP repeats region of APC, an Axin-binding site, is known to be important for tumor suppression and for the developmental function of APC. We performed a yeast two-hybrid screening using the first SAMP motif-containing region of Xenopus APC as bait and obtained several SAMP binding candidates including DDEF2 (development and differentiation enhancing factor 2), which is an ADP-ribosylation factor (Arf) GTPase-activating protein (GAP (ArfGAP)) involved in the regulation of focal adhesions. In vitro and in cells the Src homology 3 (SH3) domain of DDEF2 and its close homolog, DDEF1, are associated with the SAMP motif of APC competitively with Axin1. Moreover, NMR chemical shift perturbation experiments revealed that the SAMP motif interacts at the same surface of the SH3 domain of DDEF as the known SH3 binding motif, PXXP. When fluorescent protein-tagged APC and DDEF are expressed in Xenopus A6 cells, co-localization at microtubule ends is observed. Overexpression and RNA interference experiments indicate that APC and DDEFs cooperatively regulate the distributions of microtubules and focal adhesions. Our findings reveal that the SAMP motif of APC specifically binds to the SH3 domains of DDEFs, providing new insights into the functions of APC in cell migration.

To date ϩTIPs have been proven to play an important role in the polarized organization of MT networks. So far, some of ϩTIPs such as APC, ACF7, and CLASPs were reported to be associated with the cell cortex near the migrating edges of cells and to stabilize MTs (3, 8 -11). Without the activities of these cortical ϩTIPs, in most cases the MT stabilizing effect at the leading edges is lost, and the cells fail to maintain polarized, coordinated migration in response to monolayer wounding (9,12,13).
In the front of a motile fibroblast, the transient interaction of the plus-end of MT with focal adhesions (FAs) promotes adhesion disassembly and remodeling of the actin cytoskeleton (14). Indeed, MTs are involved in the disassembly of FAs via focal adhesion kinase (FAK) and dynamin (14,15). On the other hand, paxillin, a FA adaptor protein involved in FA dynamics, regulates MT behaviors by promoting the disassembly of MTs (16). However, the factor(s) and the precise molecular mechanism(s) connecting the MT plus end and FAs are largely unknown. In this context, it is interesting to examine the involvement of APC in the FA dynamics.
The APC protein is a large, 310-kDa protein with multiple structural domains and multiple binding partners, as shown in Fig. 1A (17). In the NH 2 -terminal armadillo repeat region, APC associates with the kinesin motor protein via KAP3 to transport the molecule to MT plus ends (18). In the COOH-terminal region, APC directly associates with MTs and with MT-binding proteins of the EB1 (end-binding 1) family (19 -22). APC stabilizes MTs (23) and promotes the net growth of MTs in cells (24). The middle part of APC protein, which contains 15-amino acid (aa) repeats and distinct 20-aa repeats, directly interacts with ␤or ␥-catenin (25,26). The 20-aa repeat region of APC contains 3 repeats of a Ser-Ala-Met-Pro (SAMP) motif, which binds to Axin1/conductin (27,28). These repeats, initially termed SAMP repeats because of the SAMP motifs found in human APC (27), contain a conserved sequence (I/L)XXXCIX-SXMX(K/R) (where X is any amino acid) where (I/L)XXXCI also appears to be invariant in Drosophila APC (29). Through this middle portion, APC forms a complex with catenins, Axin1/conductin, and glycogen synthase kinase-3␤, an important functional unit in the Wnt signaling pathway (25,30), which leads ␤-catenin to the proteasome-mediated degradation pathway (31).
The SAMP repeat region is thought to be critical in the developmental function and tumor-suppressing activity of APC. During the progression of sporadic colorectal tumors as well as in patients with familial adenomatous polyposis (FAP), mutations in the APC gene result in a roughly half-sized APC protein truncated upstream of the SAMP repeats region (for reviews, see Refs. 32 and 33). The importance of the first SAMP motif has been proven using mouse models carrying targeted mutations in APC (34). Truncations of APC upstream of the SAMP repeats (e.g. Apc Min and Apc1572T) have been shown to be associated with tumorigenesis in heterozygotes and embryonic lethal phenotype in homozygotes. In contrast and importantly, homozygosity for Apc 1638T , in which mutated APC produces a truncated APC protein retaining the first SAMP motif but missing all of the COOH-terminal domains, is compatible with postnatal life and Apc 1638T /Apc 1638T animals do not show any increased tumor susceptibility. To date, the function of the SAMP repeats has been discussed only in the terms of the involvement of Axin1/conductin. However, it is still possible that molecules other than Axin1/conductin bind to the SAMP repeat.
We have analyzed the localization and function of full-length and truncated forms of APC using the A6 cell line, which was established from a normal Xenopus laevis kidney and which can overexpress exogenous APC (11,35). This trait of A6 cells is advantageous for functional analysis of the APC protein, because in most cultured mammalian cells the expression of exogenous APC is very difficult due to the induction of cell death through apoptosis (36). In the present study we found that the SAMP repeat region has a MT-related function in A6 cells. By a yeast two-hybrid screening system, we identified DDEF1 (also known as AMAP1, ASAP1, or PAG2) and DDEF2 (also known as AMAP2 or PAG3), paxillin-and FAK-binding proteins, as novel SAMP repeat-interacting proteins, the interaction being mediated through their Src homology 3 (SH3) domains. DDEFs co-localized with APC at the cell edges in a MT-dependent manner when expressed in A6 cells. Through nuclear magnetic resonance (NMR) chemical shift perturbation experiments, we further found that the SAMP motif interacts with the same region of the SH3 domains of DDEFs as the known SH3 binding motif PXXP. Finally, we found that APC and DDEFs cooperatively regulate microtubules and FAs through the SAMP repeats region.

EXPERIMENTAL PROCEDURES
Plasmid Construction and Recombinant Proteins-Enhanced green fluorescent protein expression vectors (pEGFP-C series) were purchased from Clontech (TaKaRa). To generate mRFP1 expression vectors, the GFP coding sequences of pEGFP-C were substituted with mRFP1 (a gift from Dr. R. Tsien). The GFP-fused X. laevis APC protein constructs, fAPC-GFP, GFP-fAPC, nAPC-GFP, ⌬cAPC-GFP, and GFP-cAPC, are described previously (11,35). Other expression plasmids encoding a series of GFP-fused mutated APC proteins were generated using the GFP-fAPC plasmid. The details are provided in the supplemental materials.
A rat Axin1 cDNA (rAxin1) was a gift from Dr. A. Kikuchi (Hiroshima University). A X. laevis Axin1 cDNA (BC089276, IMAGE:3379167) and a human vinculin cDNA were obtained from Open Biosystems and OriGene, respectively. To generate GFP or mRFP fusion proteins, Axin1, rAxin1, vinculin, and candidates for SAMP-binding proteins were subcloned into pEGFP or pmRFP expression vectors. The details are provided in the supplemental materials.
In Vitro Binding Assay, Immunoprecipitation, and Western Blot Analysis-FLAG-SAMP#1 peptides were immobilized on anti-FLAG M2-conjugated agarose beads (Sigma) in a buffer consisting of 20 mM Tris-HCl (pH 7.5), 1 mM EDTA, 100 mM NaCl, 0.01% Triton X-100, Complete Protease Inhibitor Cocktail (Roche), and 1% bovine serum albumin. Trx-S-tagged recombinant proteins were incubated with the beads in the same buffer for 1 h at a room temperature. After washing with the same buffer, the bound proteins were eluted from the beads by boiling in SDS sample buffer and subjected to SDS-PAGE followed by Western blot analysis.
For Western blot analysis, the samples were separated by SDS-PAGE using gels with appropriate concentrations of acrylamide and transferred to polyvinylidene difluoride or to nitrocellulose membranes. After incubation with appropriate antibodies, the bound antibodies were detected with the ECL plus Western blotting detection system (Amersham Biosciences).
MT Co-sedimentation Assay-Purified bovine tubulin proteins (Cytoskeleton) were polymerized into MTs and stabilized with taxol in BRB80 tubulin polymerization buffer using a standard protocol. Cells were cultured on 10-cm dishes, briefly washed once with phosphate-buffered saline, and lysed with 1 ml of buffer A. The cell lysates were clarified by centrifugation at 100,000 rpm using a TLA-100.3 rotor (Beckman) for 30 min at 4°C in a Beckman TLA100 ultracentrifuge. The clarified cell lysates were mixed with 10 M taxol and 0.2 mg of MTs, layered onto a prewarmed 1-ml glycerol cushion (buffer A containing 40% (v/v) glycerol and 10 M taxol) in open-top thick-walled polycarbonate tubes and centrifuged at 70,000 rpm for 10 min at 37°C in a TLA-100.3 rotor to pellet the MTs. After complete aspiration of the supernatants and cushions, the MT pellets were resuspended in 100 l of SDS sample buffer.
Immunofluorescence Staining and Fluorescence Microscopy-Cells cultured on coverslips were fixed with 3.7% formaldehyde for 15 min or in methanol-acetone (1:1) at Ϫ20°C for 2 min. After washing with phosphate-buffered saline, cells were permeabilized with 0.1 or 0.5% Triton X-100 and incubated with 10% fetal bovine serum. Samples were then processed for indirect immunostaining using appropriate primary and secondary antibodies. Samples were washed several times and mounted in ProLong anti-fade reagent (Molecular Probes). Images of cells were acquired and analyzed as described previously (11). For live cell imaging, cells were cultured on glass-based dishes (IWAKI) and observed using a DeltaVision Core microscope system (Applied Precision Inc.) equipped with an Olympus IX71 microscope (PlanApo 100ϫ/1.40 NA oil immersion objective. Out-of-focus signals were removed using the deconvolution technique of the DeltaVision system. NMR Spectroscopy-NMR experiments were performed on a Bruker AV-400 M or on a DRX-500 spectrometer. The sample conditions were 20 mM sodium phosphate (pH 7.4), 50 mM NaCl, 0.02% NaN 3 , and 10% D 2 O for the chemical shift perturbation analyses of the proteins and the SAMP peptides (exper-FIGURE 2. Identification of SAMP motif-binding proteins. A, the Xenopus APC fragment used for yeast two-hybrid screening and the corresponding regions of human and mouse APC. The SAMP (Ser-Ala-Met-Pro) motif is indicated with dots. B, the binding site to the prey proteins was narrowed down by yeast two-hybrid ␤-galactosidase activity detection using APC fragments containing the first SAMP motif (Frag.1) and the last half of APC fragment used for library screening (Frag.2). The results are indicated to the right. C, coprecipitation (IP) of mRFP-fused SAMP binding candidates with GFP-fAPC immobilized to protein A-Sepharose using anti GFP mAb. Axin1 (positive control), DDEF1, DDEF2, and AP-2 mu1 subunit were significantly precipitated with GFP-fAPC. WB, Western blot. Input lanes contain 10% of the cell lysate. NOVEMBER 21, 2008 • VOLUME 283 • NUMBER 47 iments shown in Fig. 4) or 20 mM sodium phosphate (pH 6.2), 50 mM NaCl, 0.02% NaN 3 , and 10% D 2 O for the main-chain assignment of DDEF1-SH3 and for the titration experiments of FLAG-SAMP#1 and FLAG-FAK-SII on DDEF1-SH3 (experiments shown in Fig. 5). To analyze chemical shift perturbations, weighted averages of chemical shift changes were summed for all the resonances in each 1 H, 15 N heteronuclear single quantum correlation (HSQC) spectrum of a 15 N-labeled protein (0.07 mM) as the result of a peptide addition (see Fig. 4D). The weighted averages were calculated according to the function [(⌬␦ H ) 2 ϩ (0.14 ϫ⌬␦ N ) 2 ] 1/2 (ppm), where ␦ H and ␦ N are, respectively, 1 H and 15 N chemical shifts, and ⌬␦ is the deviation of the observed chemical shift in the presence of a peptide from that in the free-form. The value of 0.14 was used as a scaling factor to normalize the 15 N chemical shift range to that of 1 H. For the peak that broadened beyond detection upon the peptide addition, the chemical shift change was assumed to be 0.1.

Association of APC Tumor Suppressor with DDEFs
NMR spectra for the main-chain resonance assignment of X. laevis DDEF1-SH3 were acquired at 298 K using 1.2 mM 13 C/ 15 N-labeled DDEF1-SH3 with a Bruker DRX-500 NMR spectrometer equipped with a triple resonance ( 1 H, 15 N, and 13 C) probe head that had self-shielded triple-axis gradient coils. To assign the 1 H N , 15 N, and 13 C resonances, a series of twoand three-dimensional experiments was performed. These were two-dimensional 1 H, 15 N HSQC three-dimensional HNCO, HN(CA)CO, CBCA(CO)NH, and HNCACB. All the 15 N and 13 C indirect dimensions were acquired by the time proportional phase incrementation (TPPI)states method, except for the 15 N dimension of CBCA(CO)NH, which was acquired by the sensitivity enhancement and gradient-echo method. All the experiments except for CBCA(CO)NH used the WATERGATE and the water-flipback methods for suppression of the large water signal. For all experiments, the spectral width of the 15 N dimension was set to 14.5 ppm. The 1 H carrier was set at the frequency of the residual water resonance (4.773 ppm), and the 15 N, 13 13 CO, and 13 C␣␤ dimensions, respectively, and the corresponding numbers of complex sampling points are described in parentheses. All data were processed with the program NMRPipe (39). The resolutions in the 15 N and 13 C dimensions were doubled by linear-prediction before Fourier transformation. The peaks were analyzed with the program Sparky (developed by T. D. Goddard and D. G. Kneller at University of California at San Francisco).

RESULTS
Subcellular Distribution of Truncated APC Proteins-It has been reported that endogenous and GFP-fused full-length APC are localized at the edges of cells in an MT-dependent manner in a wide variety of cell types (Fig. 1B) (24,35,40). The kinesinbased active transport via KAP3/KIF3 complex associating with the armadillo region has been shown to be involved in the localization of APC to the cell ends (18). To gain further insight into the APC domains involved in the MT-related function of APC, we made a series of GFP-fused APC mutants and analyzed their localization in A6 cells as summarized in Fig. 1 and supplemental Fig. 1.
Consistent with previous observations, the armadillo repeat domain was indispensable for the MT-plus end accumulation of APC; GFP-APC(⌬arm) lacking only the armadillo repeats distributed along the entire MTs. Using the series of APC proteins shown in Fig. 1B, both the middle region and the MT binding basic region were shown to regulate APC localization to the cell ends. Therefore, we further dissected the middle portion as shown in Fig. 1C. In cells, GFP-APC(⌬1060), GFP-APC(⌬1314), and GFP-APC(⌬1574) distributed throughout the cytoplasm similarly to nAPC-GFP (supplemental Fig. 1). In contrast, GFP-APC(⌬1641) containing the first SAMP motif restored the ability to accumulate at the cell ends (see also Fig. 6). The SAMP repeats of APC are known to be an Axin binding region; however, using available anti-Axin antibodies, we could not detect endogenous Axin in the clusters of GFP-APC(⌬1641) at the cell ends using immunofluorescence. Therefore, we hypothesized that molecules other than Axin are associated with the SAMP motif to regulate APC function.
Identification of SAMP Motif-interacting Proteins-To identify SAMP motif-interacting proteins, we performed yeast two-hybrid screening as described previously (41)(42)(43). An X. laevis APC fragment containing the first SAMP motif (aa 1565-1644, Fig. 2A) fused to the GAL4 DNA binding domain was used as bait in two-hybrid screens of a mouse brain cDNA library. Approximately 110,000 transformants were screened. Among the colonies that appeared, we isolated 30 clones showing ␤-galactosidase activity, recovered the plasmids, and analyzed the insert sequences. We identified one clone containing the Axin1 cDNA, seven clones containing the DDEF2 (mKIAA400) cDNA (SH3 domain), one clone containing the ARHGAP26 (mKIAA0621) cDNA (SH3 domain), nine clones containing the FIP200 (mKIAA0203) cDNA, and six clones containing the adaptor protein complex AP-2 mu1 subunit cDNA. Other clones contained inserts encoding untranslated sequences. The endogenous expressions of all these molecules in A6 cells were confirmed by reverse transcription-PCR (data not shown). In the yeast two-hybrid system, the binding sites on APC were narrowed down to the first half of the fragment used for the library screening, which contains the SAMP motif (Fig. 2B).
To examine the physiological interactions with APC in cells, we performed pulldown assays using A6 cell lysates expressing a large amount of GFP-fAPC (Fig. 2C). GFP-fAPC was immobilized to protein A-Sepharose using an anti-GFP mAb. The SAMP binding candidates were fused to mRFP and expressed in HEK293 cells, mixed with GFP-fAPC-immobilized beads, and precipitated by centrifugation. As shown by Western blot analysis, Axin1, DDEF2, and AP-2 mu1 subunit significantly precipitated GFP-fAPC (Fig. 2C,  asterisks). Because vertebrates express two DDEF proteins, DDEF2 and its close homolog DDEF1, the interaction of DDEF1 with APC was also examined. mRFP-DDEF1 generated in HEK293 cells precipitated GFP-fAPC from cell lysates similarly to DDEF2 and Axin1. In the following studies we further analyzed only DDEFs.
In Vitro Binding Analysis of APC-SAMP Motif and SH3 Domains-DDEFs are multidomain proteins containing a pleckstrin homology (PH) domain, an ArfGAP domain, ankyrin repeats (Ank), a prolinerich region, and a COOH-terminal SH3 domain (Fig. 3A). Because in the yeast two-hybrid screen we isolated the SH3 domain of DDEF2 as a SAMP-interacting region, we examined in vitro the direct interactions of the first SAMP motif-containing polypeptide (SAMP#1) with the RGS domain of Axin1 (known as an APC binding region, positive control), the SH3 domains of the DDEFs, and the SH3 domains of ARHGAP26 and ARHGAP10, a close homolog of ARHGAP26, using a synthetic FLAGtagged first SAMP motif peptide (FLAG-SAMP#1, Fig. 3B) and purified recombinant proteins fused to thioredoxin and S (Trx-S) tags. In vitro pulldown assays revealed that the FLAG-SAMP#1 peptide significantly bound to Axin1 and to the SH3 domains of DDEFs but not to the SH3 domains of ARHGAPs (Fig. 3C). These results demonstrate specific associations between the SAMP motif and the SH3 domains of the DDEFs. Furthermore, the interaction between FLAG-SAMP#1 and Trx-S-Axin1-RGS was disrupted by the addition of GST-DDEF-SH3 in a concentrationdependent manner, showing that the DDEF-SH3 domain and the Axin1-RGS domain competitively bind to SAMP#1 (Fig. 3D).

Analysis of SAMP-SH3 Interaction by NMR Chemical Shift
Perturbation-To further confirm the interactions between the SH3 domains and the SAMP motif in vitro, we acquired a series of twodimensional 1 H, 15 N HSQC NMR spectra of 15 N-labeled SH3 domains of DDEFs, ARHGAPs, and human Axin1-RGS upon the addition of a non-labeled SAMP peptide. The sequence of the human Axin-1-RGS was the same as that described in a previous structural analysis of APC-SAMP and Axin1-RGS complexes (37). We used a bacterially expressed recombinant peptide as well as a chemically synthesized FLAG-tagged peptide for each of the three SAMP motifs. Two kinds of peptides for each SAMP motif were examined to exclude the possibility of any nonspecific interaction that might be caused by the extrinsic sequence regions. Fig. 4A shows a series of two-dimensional and GFP-APC(⌬1641). Note the possible contribution of endogenous APC associating with GFP fusion proteins through their NH 2 -terminal coiled-coli region. Input lanes contain 100 and 2% of the cell lysates for GFP and for other protein detection, respectively. In B and C, to detect endogenous DDEF1, anti-AMAP1 pAb was used. E, co-sedimentation of DDEF1 and DDEF2 with MTs from A6 cell lysates. The blot of p150 glued , a MT-binding protein, is shown as a positive control. Input lanes contain 10% of the cell lysates. WB, Western blot. HSQC spectra of 15 N-labeled human Axin1-RGS with various molar ratios of FLAG-SAMP#1 peptide. As expected, significant chemical shift changes and peak broadening were observed upon the addition of the peptide, confirming that FLAG-SAMP#1 binds to hAxin1-RGS. Likewise, 1 H, 15 N HSQC spectra of 15 N-labeled DDEF1-SH3 revealed dramatic changes in the presence of FLAG-SAMP#1 compared with spectra in the absence of the peptide (Fig. 4B). In contrast, we observed only marginal differences in the spectra of the 15 N-labeled SH3 domain of Xenopus ARHGAP10 in the absence or in the presence of FLAG-SAMP#1 peptide (Fig. 4C). When FLAG-SAMP#2 or -#3 instead of FLAG-SAMP#1 was added to 15 Nlabeled ARHGAP10-SH3, almost no change was observed in the spectra (not shown). To quantitatively estimate the chemical shift perturbations, total chemical shift changes observed when one of the six SAMP peptides was added to one of the 15 N-labeled proteins at an equimolar ratio were calculated (Fig. 4D). We judged that a total chemical shift change of more than 1.0 corresponds to large spectral alteration and can be categorized as a specific interaction. Accordingly, in Fig. 4D, values more than 1.0 are colored blue, and those less than or equal to 1.0 are colored red. Although slight inconsistencies were found in the results in the combinations involving SAMP#3, this is probably due to a problem inherent to FLAG-SAMP#3 rather than to nonspecific binding of recombinant SAMP#3 because we found it difficult to handle FLAG-SAMP#3 peptide in additional experiments. Therefore, the differences in the peptide sequences, including which tag was fused and which SAMP motif was examined, did not affect the corresponding affinities. These results further support the existence of associations between the SH3 domains of DDEFs and APC-SAMP motifs, with affinities that are comparable with those between Axin1-RGS and the repeats. It is noteworthy that in all the combinations of the SH3 domains of DDEFs and the SAMP peptides, the spectra showed similar changes (not shown), indicating that the three SAMP motifs interact on the same region of DDEFs-SH3.
Interaction Site within the DDEF1-SH3 Domain for the APC-SAMP Motif-It is well known that the SH3 domain can bind to a proline-rich motif, PXXP; however, in the SAMP repeats, no such typical proline-rich motif is conserved. Therefore, to characterize the interaction between the DDEF1-SH3 domain and the SAMP motif, we analyzed the region of the SH3 domain that binds to FLAG-SAMP#1 and compared that with binding to the second proline-rich motif of FAK (referred to as FAK-SII hereafter), which is known to interact with DDEF1-SH3 and contains the PXXP motif (38). Initially, the backbone amide resonances of DDEF1-SH3 (consisting of 82 amino acids, as shown in Fig. 5A) were assigned except for seven residues including four prolines and two vector-derived NH 2 -terminal residues and Gly-64, whose peak broadened beyond detection probably because of a local conformational exchange. Then FLAG-SAMP#1 was gradually added to the 15 N-labeled DDEF1-SH3 solution (Fig. 5B). As the molar ratio of SAMP#1 increased, a subset of peaks including those derived from Thr-27, Gln-32, Ala-33, Asn-35, Thr-40, Gly-51, Glu-53, Asp-54, Gln-55, Trp-58, Gly-70, Val-71, Phe-72, Val-74, and Val-77 significantly shifted. Next, we tested the effects of FAK-SII peptide on DDEF1-SH3 (Fig. 5C). Surprisingly, spectra showed chemical shift changes that were very similar in both their directions and degrees to those observed when FLAG-SAMP#1 was added. This comparison suggests that SAMP#1 and FAK-SII bind to DDEF1-SH3 in a similar manner in terms of their interaction sites on the domain, their affinities, the van der Waals interactions, and the hydrogen-bond networks. The chemical shift changes observed when FLAG-SAMP#1 or FLAG-FAK-SII was added to the solution of 15 N-labeled DDEF1-SH3 are mapped onto the surface of the tertiary structure of the fifth SH3 domain of human Intersectin2, which among available structures exhibits the highest sequence similarity to Xenopus DDEF1-SH3 (Fig. 5, D-F). These figures further demonstrate that SAMP#1 interacts in almost the same region of DDEF1-SH3 as FAK-SII, which contains a typical SH3 binding motif. Therefore, DDEF-SH3s are suggested to have a distinct structural feature in the proline-rich motif binding site that can also specifically recognize the SAMP motif.
DDEFs Associate with the First SAMP Motif in A6 Cells-We then examined the interaction between APC-SAMP and DDEFs in cells. When expressed in the cells, both mRFP-fused DDEF1 and DDEF2 colocalized with GFP-APC(⌬1641) at the cell edges (Fig. 6, A and B). Moreover, co-localization of endogenous DDEF2 and GFP-fAPC was also detectable (Fig. 6C). Furthermore, when either GFP-APC(⌬1641) or GFP-APC(⌬1574) was immunoprecipitated from A6 cell transfectants, we found that endogenous DDEFs and Axin1 had bound mainly to GFP-APC(⌬1641) as expected (Fig. 6D). The precipitated fraction including DDEF1 was notably more abundant than that including DDEF2. On the other hand, KAP3 and ␤-catenin, known APC-binding proteins used as positive controls, interacted equally with both APC fragments.
Based on the above observations, the SAMP-interacting proteins were expected to interact with MT systems. Therefore, the association of DDEFs with MTs was examined by a MT cosedimentation assay. As shown in Fig.  6E, both endogenous DDEF1 and DDEF2 were precipitated with MTs.
Mutual Regulation of FAs and MTs by APC, DDEFs, and Axin1-DDEFs have been shown to bind to focal adhesion proteins such as paxillin and FAK and regulate FA dynamics (44,45). Transient overexpression of DDEF1 retarded cell  NOVEMBER 21, 2008 • VOLUME 283 • NUMBER 47 spreading by regulating the function of a subset of focal adhesion proteins including paxillin (38,46). Consistent in A6 cells, endogenous DDEF1 and DDEF2 are localized at the FAs as well as plasma membranes (data not shown). We examined whether APC affects the focal adhesions through its SAMP repeat. For this purpose we examined the phosphorylation status of paxillin (Tyr-31), which is highly increased during cell adhesion and migration (47) in A6 transfectants overexpressing GFP-fAPC, GFP-APC (⌬SAMP), GFP-rAxin, or GFP-DDEF2. As shown in Fig. 7B, the phosphorylation of paxillin at Tyr-31 was significantly reduced in cells expressing GFP-APC (⌬SAMP), GFP-rAxin, or GFP-DDEF2, whereas the total expression levels of paxillin were not reduced in these cells. In addition, the acetylation of tubulin, an indicator of stable MTs, was reduced in GFP-rAxin-and GFP-DDEF2-expressing cells (Fig. 7C).

Association of APC Tumor Suppressor with DDEFs
To understand these phenomena, we observed the distribution of FAs and MTs in these transfectants by immunofluorescence (Fig. 8). The FAs were visualized with vinculin, a core adhesion component that remains in the adhesions even after paxillin displacement (16). Endogenous APC and GFP-fAPC are distributed along the MTs extending to the extreme edges of cells beyond the FA-reach areas (arrows). The expression of GFP-fAPC produced large number of FAs along the entire basal cell cortex. In contrast, in GFP-APC(⌬SAMP)-expressing cells, limited numbers of FAs were distributed only at the cell margin, and MTs decorated with GFP-APC(⌬SAMP) are tangled behind the FA-rich areas.
The processes of organization of FAs and MTs in these transfectants were visualized by a time-lapse imaging study shortly after cell plating (Fig. 9, supplemental Movie 1). Although GFP-fAPC-positive MTs dynamically extend toward the cell ends (arrows in Fig. 9A), most MTs decorated with GFP-APC(⌬SAMP) turn behind the FAs and cannot get over the FAs. It is striking that the remodeling of FAs is retarded in GFP-APC(⌬SAMP)-expressing cells (Fig. 9B), consistent with the Western blot result showing the significant reduction of the phosphorylation of paxillin (Tyr-31), an indicator of FA dynamics (Fig. 7B). These observations demonstrate the function of the SAMP repeats in MT organization and FA regulation.
Then, to investigate whether APC affects FA dynamics through DDEF or Axin, we transiently overexpressed RFP-Axin1 or RFP-DDEF2 in A6 transfectants expressing GFP-fAPC and analyzed the effect on the FA distribution (Fig. 8, B and C). Although the distribution of GFP-fAPC along MTs was not affected by the overexpression of RFP-DDEFs, which was clearly co-localized with GFP-fAPC (data not shown), the number of FAs was significantly decreased, especially from the central region of cells, similarly to the phenotype observed in the GFP-APC(⌬SAMP)-expressing cells. When RFP-Axin1 was overexpressed, the FA number was also decreased. However, unlike the effect of RFP-DDEF1 overexpression, the distribution of GFP-fAPC was seriously disturbed and removed from MTs by RFP-Axin1. A similar effect of overexpression of Axin1 on the distribution of endogenous APC and FAs was observed in the parental A6 cells (data not shown). Therefore, in the overexpression study, both DDEFs and Axin1 affected MT and FA dynamics probably via APC, but their mechanisms of action are distinct.

Effect of RNA Interference of APC or DDEFs in Non-cancer
Cell Line MCF10A-Finally, we confirmed the function of APC and DDEFs on the distribution of microtubules and FAs using an RNA interference strategy in the non-cancer human mammary epithelial cell line MCF10A (Fig. 10A). In APC knocked-down cells, FAs became enlarged at the cell periphery, whereas DDEFs knocked-down cells formed thinner FAs (Fig. 10B). At the peripheral regions of cells, MTs are radially aligned facing their plus ends toward the edges. However, in the DDEF1-concentrated lamellipodia (Fig.  10C, arrowheads), only few MTs decollated with APC extended toward the cell edges, whereas many MTs are terminated behind the DDEF1-rich areas. In APC knockeddown cells, the number of MTs invaded into the DDEF1-rich areas was slightly decreased (Fig. 10D). On the contrary, after DDEF knockdown, large numbers of MTs reached the cell edges in most lamellipodia. These findings suggest that DDEFs suppress the invasion of MTs, whereas APC promotes the growth of MTs, consistent with the observation from the overexpression experiments described above.

DISCUSSION
In the present study, based on the subcellular distribution of a series of engineered APC proteins in Xenopus A6 cells, we found that, in addition to the armadillo repeats associating with kinesin motor protein and the basic region binding to MTs directly, the SAMP repeats region has a MT-related function, although the molecular mechanism that targets the APC fragment retaining the SAMP repeats is not yet clear. The SAMP repeats of APC have been known to be an Axin binding region, which is involved in the tumor suppressing activity and in the developmental function of APC. Here, we explored the possibility that the SAMP motif binds to molecules other than Axin and found that DDEF1 and DDEF2 interacted with this region through their SH3 domains. In vitro binding assays demonstrated that DDEFs bind to the SAMP motif competitively with Axin1. This interaction was also verified by the NMR chemical shift perturbation study. Although the SH3 domain of ARH-GAP26 was also identified as a SAMP binding candidate in yeast two-hybrid screening, in vitro binding assays and NMR studies revealed that the SAMP motif binds selectively to DDEFs-SH3.
The SH3 domains of both DDEF1 and ARHGAP26 have been reported to bind to the SII region of FAK, which contains the typical SH3 binding motif PXXP (38,48), suggesting that the SH3 domains of these molecules have relatively similar characteristics in the context of protein-protein interaction. Nevertheless, we observed preferred associations of the SAMP motif with the SH3 domains of DDEFs. Thus, these observations suggest a specific association of the SAMP motif with DDEFs, relative to all other SH3containing proteins, although the possibility of its interaction with other SH3 proteins cannot be excluded.
The SH3 domain is well known to bind to the proline-rich motif PXXP. However, this motif is not observed in any of the three SAMP motifs of APC. Therefore, the mode of interaction between APC-SAMP and DDEFs-SH3 is structurally interesting. A proline-rich sequence encompassing the PXXP motif forms a lefthanded polyproline type II (PPII) helical conformation (49).
Because, despite its name, such a PPII helix can also be formed by amino acid sequences containing no proline residues, the SAMP motif, containing one proline residue, may also adopt a PPII helix. The PPII helix has a triangular cross-section, and NOVEMBER 21, 2008 • VOLUME 283 • NUMBER 47 residues along two of the three edges interact with the SH3 domain. Because the two prolines in the PXXP motif and their respective preceding residues occupy each of the two edges, these XP pairs usually contribute to the specificity of the interaction (49). Comparing the sequences of the SAMP motif and FAK-SII, both have at least one pair of basic residues and a proline. Therefore, these amino acids arrangements may be involved in binding to the SH3 domains of DDEFs by forming a PPII helix. However, these residues alone could not explain the selectivity of the SAMP motifs for DDEFs over ARHGAPs, because FAK-SII can associate with both of them.

Association of APC Tumor Suppressor with DDEFs
As a possible cause for this selectivity, we have focused on a notable feature of the SAMP motifs; their COOH-terminal parts are rich in arginine and lysine residues. These basic residues may also be involved in favorable electrostatic interactions between the SAMP motif and DDEFs-SH3. This notion is consistent with the presence of basic residues proximal to the PXXP core element, which often govern the orientation on a certain SH3 domain of the proline-rich polypeptide and which can align itself in either one of the two opposite directions because of the pseudo 2-fold symmetry of a PPII helix (50,51). Another possibility is that the cluster of basic residues is the key determinant of the specific interactions between APC-SAMP and DDEFs-SH3. Indeed, several SH3 domains recognize consensus sequences comprised of arginine and lysine residues, such as (R/K)XX(R/K) (52) and (K/R)XXXXKX(K/R)(K/R) (53). Considering that three SAMP motifs interact with DDEFs, it is plausible that both the positively charged residues and the conserved residues including Ser-Ala-Met-Pro are necessary for the specific binding to the SH3 domains. Although the crystal structure of Axin1-SAMP complex has already revealed that the NH 2 -terminal part of the SAMP motif has propensity to form an ␣-helix (37), the conformation of the basic residueabundant COOH terminus, which we expect to be important for the binding to DDEFs-SH3 for the above reason, is not obvious. To elucidate the structural basis of the specific interaction between APC-SAMP and DDEFs-SH3, we are determining the detailed structure of the complex.
DDEFs have been reported to bind, not only to FAK, but also to other focal adhesion related proteins, including paxillin (44,45,54) and to endocytic proteins such as amphiphysin II (53). DDEFs are phosphatidylinositol 4,5-bisphosphate-dependent ArfGAPs and are reported to exhibit GAP activity toward Arf1 and Arf6 (55,56). Through their Arf GTPase-activating ability, DDEFs regulate membrane trafficking and the actin-based structures such as FAs and thereby regulate cell movement (46,57). The NMR study demonstrated that SAMP#1 interacts in almost the same region of DDEF1-SH3 as FAK-SII, suggesting that the binding of APC to DDEFs could compete with FAK. Considering the fact that APC is involved in cell migration at the leading edges of cells (11,58,59), APC was expected to affect FAs through a complex with DDEFs. Indeed, we found that the overexpression of APC enhanced FA turnover relying on the SAMP repeats region, and this effect was suppressed by the overexpression of DDEFs. These results suggest that the balance between the activities of APC and DDEFs cooperatively regulate the FA dynamics, although it is still possible that DDEFs compete with other SAMP-binding proteins having a function in FA regulation. Moreover, because the GFP-APC(⌬SAMP) construct lacks a large portion including the SAMP repeats and 6 of 7 20-aa repeats, it is possible that other molecules contribute to APC function by interacting with deleted regions outside of the SAMP motifs. Nevertheless, the alteration of FA-regulating function of APC caused by the loss of DDEF binding ability could be explained by the cell spreading activating and suppressing functions of APC and DDEFs, respectively. This hypothesis fits with the model that a subset of MT plus ends decorated with APC promotes FA remodeling at the migrating edges of cells (14,16). Therefore, future studies must investigate the detailed molecular mechanisms of cooperation between APC and DDEFs, possibly through the Arf signaling.
At the same time we found an unexpected effect of the overexpression of Axin1 on APC distribution and function; when overexpressed, RFP-Axin1 formed granular clusters with APC in the cytoplasm and removed APC from MT ends. Under these conditions, FA assembly was restrained, consistent with the notion that APC could accelerate the formation of FAs. However, in cells under normal conditions, the expression level of Axin is much lower than that of APC (60). In most cells expressing Axin endogenously, APC can be localized at MT ends; therefore, at the appropriate expression levels Axin may not exert such a drastic effect on APC localization. It is possible that Axin could moderately accelerate the turnover of APC to remodel the MT networks.
Previous studies focused on the involvement of Axin to explain the importance of the SAMP repeats of APC in its developmental and in its tumor suppressing function. However, this study has shown that molecules other than Axin, such as DDEFs, can contribute to regulation of the activity of APC. In this study we could not clarify the meaning of the competition between DDEFs and Axin for APC binding solely from the assays using cultured cells in vitro. An alternative explanation is that DDEFs do not directly compete with Axin, and both molecules simultaneously bind to the SAMP motifs that repeat three times at the middle portion of APC under physiological conditions. To gain insight into the physiological meaning of the association of DDEFs and Axin with APC, we need to perform more refined experiments in mammalian cells and on animals. Moreover, although in the present study we have not examined a possible involvement of DDEFs in the Wnt signaling pathway, undoubtedly this is of great importance for future study. In this connection it is interesting that the inhibition of ArfGAP activities in cells by a small molecule compound modulated Wnt/␤-catenin signaling (61). We expect that the evaluation of these possibilities will provide us with further understanding of the molecular mechanisms of the tumor suppressing function of APC.