ADIP, a Novel Afadin- and α-Actinin-Binding Protein Localized at Cell-Cell Adherens Junctions*

Afadin is an actin filament (F-actin)-binding protein that is associated with the cytoplasmic tail of nectin, a Ca2+-independent immunoglobulin-like cell-cell adhesion molecule. Nectin and afadin strictly localize at cell-cell adherens junctions (AJs) undercoated with F-actin bundles and are involved in the formation of AJs in cooperation with E-cadherin in epithelial cells. In epithelial cells of afadin (−/−) mice and (−/−) embryoid bodies, the proper organization of AJs is markedly impaired. However, the molecular mechanism of how the nectin-afadin system is associated with the E-cadherin-catenin system or functions in the formation of AJs has not yet been fully understood. Here we identified a novel afadin-binding protein, named ADIP (afadin DIL domain-interactingprotein). ADIP consists of 615 amino acids with a calculated M r of 70,954 and has three coiled-coil domains. Northern and Western blot analyses in mouse tissues indicated that ADIP was widely distributed. Immunofluorescence and immunoelectron microscopy revealed that ADIP strictly localized at cell-cell AJs undercoated with F-actin bundles in small intestine absorptive epithelial cells. This localization pattern was the same as those of afadin and nectin. ADIP was undetectable at cell-matrix AJs. ADIP furthermore bound α-actinin, an F-actin-bundling protein known to be indirectly associated with E-cadherin through its direct binding to α-catenin. These results indicate that ADIP is an afadin- and α-actinin-binding protein that localizes at cell-cell AJs and may have two functions. ADIP may connect the nectin-afadin and E-cadherin-catenin systems through α-actinin, and ADIP may be involved in organization of the actin cytoskeleton at AJs through afadin and α-actinin.

Cells in multicellular organisms recognize their neighboring cells, adhere to them, and form intercellular junctions. Such junctions have essential roles in various cellular functions, including morphogenesis, differentiation, proliferation, and migration (for reviews, see Refs. [1][2][3][4][5][6]. In polarized epithelial cells, intercellular adhesion is mediated through a junctional complex composed of tight junctions (TJs), 1 adherens junctions (AJs), and desmosomes. These junctional structures are typically aligned from the apical to basal sides, although desmosomes are independently distributed in other areas.
AJs were originally defined using ultrastructural analysis as closely apposed plasma membrane domains reinforced by a dense cytoplasmic plaque, where actin filament (F-actin) bundles are undercoated (7). Molecular analysis shows that AJs are cell-cell adhesion sites where classic cadherins function as cell adhesion molecules and where the actin-based cytoskeleton and several cytoplasmic components are assembled (8 -10). E-cadherin, like other classical cadherins, is a single-pass transmembrane protein whose extracellular domain mediates homophilic recognition and adhesive binding in a Ca 2ϩ -dependent manner (10). E-cadherin associates with the actin cytoskeleton through peripheral membrane proteins, including ␣-, ␤-, and ␥-catenins, ␣-actinin, and vinculin (9,(11)(12)(13)(14). ␤-Catenin directly interacts with the cytoplasmic tail of E-cadherin and connects E-cadherin to ␣-catenin that directly binds to F-actin (15). ␣-Actinin and vinculin are also F-actin-binding proteins that directly bind to ␣-catenin (13,14,16). The association of E-cadherin with the actin cytoskeleton through these peripheral membrane proteins strengthens the cell-cell adhesion activity of E-cadherin (1,17).
We have found that another cell-cell adhesion molecule, nectin, and its associated F-actin-binding protein, afadin, strictly localize at AJs undercoated with F-actin bundles (18 -20). In contrast, E-cadherin is concentrated at AJs but is more widely distributed from the apical to basal sides of the lateral plasma membranes (2,21). Nectin is a Ca 2ϩ -independent immunoglobulin-like cell-cell adhesion molecule (19,(22)(23)(24)(25)(26). Nectin comprises a family, which, at present, consists of four members, nectin-1, -2, -3, and -4. All nectins have two or three splice variants (19,(25)(26)(27)(28)(29)(30)(31). Nectin-1 was originally identified as one of the poliovirus receptor-related proteins (PRR-1) (30). Nectin-2 was originally identified as the murine homolog of human poliovirus receptor protein (27) but turned out to be another poliovirus receptor-related protein (PRR-2) (29,30). Neither PRR-1 nor PRR-2 has thus far been shown to serve as a poliovirus receptor. PRR-1 and PRR-2 were later shown to serve as * The investigation at Osaka University Medical School was supported by grants-in-aid for Scientific Research and for Cancer Research from the Ministry of Education, Culture, Sports, Science, and Technology, Japan (2001 and 2002). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the DDBJ/GenBank TM  the receptors for ␣-herpes virus, facilitating their entry and intercellular spread and so were renamed HveC and HveB, respectively (31)(32)(33)(34)(35)(36). It remains unknown whether nectin-3 and nectin-4 serve as receptors for viruses. All of the members have an extracellular domain with three immunoglobulin-like loops, a single transmembrane region, and a cytoplasmic region. Furthermore, all of them, except nectin-1␤, -3␥, and -4, have a conserved motif of 4 amino acid (aa) residues (Glu/Ala-X-Tyr-Val) at their carboxyl termini, and this motif binds the PDZ domain of afadin (18,19,25,26).
Afadin has at least two splice variants, l-and s-afadin (18). l-Afadin, the larger splice variant, binds nectin and, through its F-actin binding domain, F-actin. l-afadin binds to the side of F-actin but does not cross-link it to form bundles. Afadin also has two Ras-associated domains, a forkhead-associated domain, a dilute (DIL) domain, a PDZ domain, and three prolinerich domains (see Fig. 1A). F-actin binds to the region containing the third proline-rich domain (18). s-afadin, the smaller splice variant, has two Ras-associated domains, a forkheadassociated domain, a DIL, a PDZ, and two proline-rich domains, but lacks the F-actin-binding domain. Human s-afadin is identical to the gene product of AF-6, a gene that has been identified as an ALL-1 fusion partner that is involved in acute myeloid leukemias (37). Unless otherwise specified, afadin refers to l-afadin in this paper.
Nectin has a potency to recruit the E-cadherin-␤-catenin complex to the nectin-based cell-cell adhesion sites through afadin and ␣-catenin in fibroblasts and epithelial cells (38 -40). Nectin has furthermore a potency to recruit the components of TJs including ZO-1, claudin, occludin, and junctional adhesion molecule (JAM) to the nectin-based cell-cell adhesion sites through afadin in fibroblasts and epithelial cells (39 -42). Claudin is a key cell-cell adhesion molecule that forms TJ strands (43)(44)(45)(46), and occludin and JAM are other transmembrane proteins at TJs (45)(46)(47). Claudin, occludin, and JAM interact with an F-actin-binding scaffold molecule, ZO-1 (48 -59). In epithelial cells of afadin (Ϫ/Ϫ) mice and (Ϫ/Ϫ) embryoid bodies, the proper organization of AJs and TJs is impaired (60). Nectin-1 has recently been determined by positional cloning to be responsible for cleft lip/palate-ectodermal dysplasia, which is characterized by cleft lip/palate, syndactyly, mental retardation, and ectodermal dysplasia (61). In addition, we have recently found that the nectin-afadin system is involved in the formation of synapses in cooperation with N-cadherin in neurons (62) and that the nectin-afadin system constitutes an important adhesion system in the organization of Sertoli cellspermatid junctions in the testis (63). Nectin and afadin are therefore important for the organization of a wide variety of intercellular junctions either with or independently of known cell adhesion molecules. However, the molecular mechanism of how the nectin-afadin system is associated with the E-cadherin-catenin system or functions in the formation of these intercellular junctions has not yet been fully understood.
To gain the insight into these issues, we attempted here to identify an afadin-binding protein using yeast two-hybrid screening. For this purpose, we used the DIL domain of afadin as a bait. The DIL domain has been found in afadin and type V myosins including dilute, Myo2, and Myo4, but its function remains unknown (64). The recent finding that the region of Myo4 containing its DIL domain binds to an adapter protein, She3 (65,66), implies that the DIL domain is involved in the protein-protein interaction. We have identified and characterized here a novel afadin-binding protein, named ADIP (afadin DIL domain-interacting protein), that binds to the DIL domain of afadin. ADIP furthermore binds ␣-actinin, an F-actin-bundling protein known to be indirectly associated with E-cadherin through its direct binding to ␣-catenin (9,16). These results indicate that ADIP is an afadin-and ␣-actinin-binding protein that localizes at cell-cell AJs and may have two functions. ADIP may connect the nectin-afadin and E-cadherin-catenin systems through ␣-actinin, and ADIP may be involved in organization of the actin cytoskeleton at AJs through afadin and ␣-actinin.
Construction of Expression Vectors-For full-length cDNAs of mA-DIP and rat ADIP (rADIP), BLAST searches were conducted against GenBank TM and EMBL databases. A selection of hits obtained by BLAST searches against the human subset of GenBank TM and EMBL sequences was used to assemble the cDNA sequence of the human homologue of KIAA0923 (AB023140; GenBank TM /EMBL/DDBJ). A cDNA of KIAA0923 was kindly supplied by Dr. T. Nagase (Kazusa DNA Research Institute). The full-length cDNAs of mADIP (accession number AF532969) and rADIP (accession number AF532970) were generated from mouse and rat brain cDNAs (Clontech), respectively, by reverse transcription-coupled PCR using the following primer sets: for mADIP cDNA, 5Ј-CGTAGGAGAGTGACAGGAGCTG-3Ј and 5Ј-GGTT-ATCGAGTTTTTCTACATGAC-3Ј; for rADIP cDNA, 5Ј-CGTAGGAGA-GTGACAGGAGCTG-3Ј and 5Ј-TTCCTGTTTTTGCACTGTAGCTG-3Ј.
Cell Culture and Transfection-MDCK cells were kindly supplied by Dr. W. Birchmeier (Max-Delbruck-Center for Molecular Medicine, Berlin, Germany). HEK293 and MDCK were cultured in Dulbecco's modified Eagle's medium with 10% fetal calf serum. HEK293 cells were transfected using the CalPhos mammalian transfection kit (Clontech).
Assay for Co-immunoprecipitation of ADIP with Afadin and ␣-Actinin-Co-immunoprecipitation experiments using HEK293 cells were performed as follows. HEK293 cells were transfected with the expression plasmids in various combinations. The cells were suspended in 1 ml of Buffer A (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Triton X-100, 1 mM EDTA, 10 M ␣-phenylmethanesulfonyl fluoride hydrochloride, and 10 g/ml aprotinin), sonicated for 10 s three times at 10-s intervals, and incubated on ice for 30 min. The cell extract (1.2 mg of protein) was obtained by centrifugation at 100,000 ϫ g for 15 min and then precleared by incubation with protein G-Sepharose 4 Fast Flow beads (Amersham Biosciences). The cell extract was incubated with 20 l of anti-FLAG M2 mAb-coated protein G-Sepharose 4 Fast Flow beads at 4°C for 18 h. After the beads were washed with Buffer A, the bound proteins were eluted by boiling the beads in an SDS sample buffer (60 mM Tris-HCl, pH 6.7, 3% SDS, 2% 2-mercaptoethanol, and 5% glycerol) for 10 min. The samples were then subjected to SDS-PAGE, followed by Western blotting with the anti-T7, anti-FLAG, anti-HA, and anti-␣actinin Abs.
Co-immunoprecipitation experiments using MDCK cells were performed as follows. MDCK cells on two 10-cm dishes were sonicated in 2 ml of Buffer A, followed by ultracentrifugation at 100,000 ϫ g for 15 min. The cell extract was precleared by incubation with protein A-Sepharose CL-4B beads (Amersham Biosciences) and then incubated with 20 l of anti-ADIP pAb (M05)-coated protein A-Sepharose CL-4B beads at 4°C for 18 h. After the beads were washed with Buffer A, the bound proteins were eluted by boiling the beads in the SDS sample buffer for 10 min. The samples were then subjected to SDS-PAGE, followed by Western blotting with the anti-ADIP, anti-afadin, and anti-␣-actinin Abs.
Assay for the Binding of ADIP to Afadin and ␣-Actinin-Affinity chromatography using MDCK cells were done as follows: MDCK cells on two 10-cm dishes were sonicated in 2 ml of Buffer A, followed by ultracentrifugation at 100,000 ϫ g for 15 min. The supernatant was incubated with MBP or MBP-mADIP (200 pmol each) immobilized on 20 l (wet volume) of amylose resin beads (New England Biolabs) at 4°C for 18 h. After the beads were extensively washed with Buffer A, the bound proteins were eluted by Buffer A containing 20 mM maltose. The eluates were boiled in the SDS sample buffer. The samples were then subjected to SDS-PAGE, followed by Western blotting with the anti-afadin mAb.
Direct binding of ADIP to the DIL domain of afadin or the C-terminal region of ␣-actinin was done as follows: GST-DIL or GST-␣-actinin-1-C was incubated with MBP or MBP-mADIP-F (200 pmol each) immobilized on 20 l (wet volume) of amylose resin beads (New England Biolabs) at 4°C for 1 h. After the beads were extensively washed with PBS, the bound proteins were eluted with PBS containing 20 mM maltose. The eluates were boiled in the SDS sample buffer. The samples were then subjected to SDS-PAGE, followed by Western blotting with the anti-GST mAb.
Other Procedures-Northern blotting was performed as described (72). The mADIP cDNA fragment was radiolabeled with ␣-32 P by a standard random priming method and used to probe a mouse multiple tissue Northern blot (Clontech). Subcellular fractionation of rat liver was performed as described (73). Immunofluorescence microscopy of cultured cells and frozen sections of mouse tissues was performed as described (18,19). Immunoelectron microscopy of mouse intestine ab-sorptive epithelial cells using the ultrathin cryosection technique was performed as described (18). Protein concentrations were determined with bovine serum albumin as a reference protein (74). SDS-PAGE was performed as described by Laemmli (75).

Identification of a Novel
Afadin-binding Protein-To identify an afadin-binding protein, we performed the yeast two-hybrid screening using the region of afadin containing the DIL domain (aa 606 -983) as a bait (Fig. 1A). We screened 2 ϫ 10 5 clones of a mouse embryo library and 2 ϫ 10 5 clones of a rat brain library and obtained 31 and 25 positive clones, respectively. Four mouse clones and one rat clone encoded proteins similar to the carboxyl-terminal portion of human KIAA0923 (AB023140; GenBank TM /EMBL/DDBJ). The full-length clones of these mouse and rat cDNAs were isolated, and they encoded proteins composed of 615 aa with a calculated M r of 70,954 and 613 aa with a calculated M r of 70,684, respectively (Fig. 1B). We named this protein ADIP ((afadin DIL domain-interacting protein)). ADIP has three coiled-coil domains (Figs. 1B and 2A). The aa sequences of mADIP and rADIP were 92% identical to each other and 88 and 87% identical to that of human KIAA0923, respectively (Fig. 1B).
To confirm whether the isolated cDNAs encode full-length ADIP, HEK293 cells were transfected with pCMV-HA-mADIP, which expressed HA-tagged full-length mADIP. The cell extract was subjected to SDS-PAGE, followed by Western blotting with the anti-ADIP pAbs. Three anti-ADIP pAbs, M57 against the N-terminal portion of mADIP (aa 1-226), M01 against the C-terminal portion of mADIP (aa 339 -615), and M05 against the C-terminal portion of rADIP (aa 159 -613), were generated. A protein with a molecular mass of about 78 kDa was detected in the extracts of HEK293 cells expressing HA-mADIP by these three pAbs (Fig. 1C and data not shown). When the extracts of MDCK and HEK293 cells were subjected to SDS-PAGE, followed by Western blotting with the three anti-ADIP pAbs, a protein with a similar molecular mass to that of HA-tagged ADIP was detected ( Fig. 1C and data not shown). Therefore, we concluded that the isolated cDNA encodes full-length ADIP.
In Vitro and in Vivo Binding of ADIP to Afadin-Two-hybrid analysis revealed that ADIP specifically bound to the DIL domain of afadin but not to the DIL domain of yeast Myo4 (Fig.  2B). The region containing the third coiled-coil domain (aa 339 -436) of ADIP was required for the binding to the DIL domain of afadin ( Fig. 2A). We further confirmed the in vitro and in vivo binding of ADIP to afadin. First, we performed the immunoprecipitation analysis. HEK293 cells were co-transfected with the T7-tagged DIL domain of afadin (T7-DIL; aa 606 -983) and the FLAG-tagged C-terminal portion of ADIP containing the third coiled-coil domain (FLAG-mADIP-C; aa 339 -615), which was isolated in the two-hybrid screening. When FLAG-mADIP-C was immunoprecipitated from the cell extract with the anti-FLAG mAb, T7-DIL was co-immunoprecipitated, as detected by Western blotting with the T7 mAb (Fig. 3A). Second, we performed the affinity chromatography using MDCK cells. The extract of MDCK cells expressing endogenous afadin was incubated with an MBP-fusion protein of full-length mADIP (aa 1-615) immobilized on amylose-resin beads. After the beads were washed with the lysis buffer, the bound proteins were eluted, and the eluate was subjected to SDS-PAGE, followed by Western blotting with the anti-afadin mAb. Afadin indeed bound to MBP-mADIP, but not to MBP alone (Fig. 3B). Third, we examined whether ADIP directly interacted with afadin. A GST fusion protein of the DIL domain of afadin (GST-DIL) bound to an MBP-fusion protein of the region of ADIP containing the all three coiled-coil domains (aa 121-436) (MBP-mADIP-F) immobilized on amylose resin beads. GST-DIL bound to MBP-mADIP-F, but not to MBP alone (Fig. 3C). Finally, to confirm that ADIP binds to afadin in vivo, we examined whether endogenous afadin was co-immunoprecipitated with endogenous ADIP from the extract of MDCK cells. When endogenous ADIP was immunoprecipitated from the extract of MDCK cells with anti-ADIP pAb, endogenous afadin was co-immunoprecipitated (Fig. 3D). Afadin was not co-immunoprecipitated with control IgG. These results together with the two-hybrid analysis indicate that ADIP directly binds to afadin both in vitro and in vivo.
Tissue and Subcellular Distribution of ADIP-Northern blot analysis using the fragment of mADIP cDNA (bp 552-3194) as a probe detected an ϳ4.3-kb mRNA in all the mouse tissues examined, including heart, brain, spleen, lung, liver, skeletal muscle, kidney, and testis (Fig. 4A). The smaller band (ϳ3.0 kb) was also detected in the liver and the testis (Fig. 4A, lanes  5 and 8). Western blot analysis using the anti-ADIP pAb, M57, detected an ϳ78-kDa protein, which was the same size of ADIP detected in HEK293 and MDCK cells and in mouse spleen, lung, and kidney (Fig. 4B, lanes 3, 4, 7, and 9). The smaller bands (ϳ60 kDa in heart, ϳ76 and ϳ40 kDa in testis) were also detected, suggesting that smaller splice variants of ADIP may be expressed (Fig. 4B, lanes 1 and 8). Subcellular fractionation analysis of ADIP in rat liver indicated that it was enriched in the fraction rich in AJs and TJs, where afadin and E-cadherin were also enriched (Fig. 4C, lanes 4 and 5). The reason why the Western blot analysis of tissue distribution did not detect the ϳ78-kDa protein in the liver (Fig. 4B, lanes 5) was probably its low expression level. These results indicate that ADIP is widely expressed, although its expression levels vary depending on tissues.
Co-localization of ADIP with Afadin at AJs in Epithelial Cells-Since afadin has been shown to strictly localize at AJs undercoated with F-actin bundles (18), we examined by immunofluorescence microscopy whether ADIP co-localized with afadin at AJs in MDCK cells. ADIP and afadin co-localized at the sites of cell-cell contacts (Fig. 5A). The cross-sectional analysis using a confocal microscopy revealed that ADIP co-localized with afadin at the junctional complex region (data not shown). Thus, ADIP and afadin co-localized at the cell-cell junctions. Essentially the same results were obtained with three anti-ADIP pAbs, M57, M01, and M05 (data not shown). ADIP and afadin were also co-stained at the perinuclear regions, most presumably the Golgi complex, as estimated by the co-staining with Golgi 58-kDa protein, a marker for the Golgi complex ( Fig.  5A and data not shown), but the physiological significance of this staining is currently unknown.
To examine the precise localization of ADIP at the junctional complex region, the frozen sections of small intestine were triple-stained with the anti-ADIP pAb, the anti-afadin mAb, and the anti-ZO-1 mAb, since TJs and AJs are well separated in this cell type (49). ZO-1 is known to be a marker for TJs (48,49). Immunofluorescence microscopy revealed that ADIP colocalized with afadin but localized at the slightly more basal side than ZO-1 in the absorptive epithelia (Fig. 5, B and C,  arrows). A signal for ADIP was also observed in the microvilli, but its significance is not clear. It may be nonspecific staining. Finally, immunoelectron microscopy revealed that ADIP was exclusively localized at AJs undercoated with F-actin bundles (Fig. 6). The signal for ADIP was rarely observed at TJs and desmosomes. This localization pattern of ADIP was the same as those of afadin and nectin but was different from that of Ecadherin, which is concentrated at AJs but is more widely distributed from the apical to basal sides of the lateral plasma membranes (18,19). These results indicate that ADIP strictly co-localizes with afadin and nectin at cell-cell AJs undercoated with F-actin bundles.
The ADIP signal was not detected at focal adhesions where the vinculin signal was detected in MDCK cells (Fig. 7A). In the heart, focal adhesions, named costameres, are well developed and periodically located along the lateral borders of cardiac muscle cells (76). Vinculin localizes at costameres, whereas nectin and afadin do not (18,19). The ADIP signal was not detected at costameres where the vinculin signal was detected ( Fig. 7B, arrows). Both the ADIP and vinculin signals were detected at the intercalated discs, corresponding to cell-cell AJs (Fig. 7B, arrowheads). In addition to the signals for ADIP and vinculin at intercalated discs, these signals were also observed in stripes within the cardiac muscle cells, but their significance is not clear. They may be nonspecific staining. These results indicate that ADIP does not localize at cell-matrix junctions.
Binding of ADIP to ␣-Actinin-To gain insight into the function of ADIP, we attempted to identify an ADIP-binding protein(s). As described above, ADIP has three coiled-coil domains, and the region containing the third coiled-coil domain (aa 339 -436) of ADIP is required for the binding to the DIL domain of afadin ( Fig. 2A). We performed the yeast two-hybrid screening using the region of ADIP containing all three coiled-coil domains (aa 152-436) as a bait ( Fig. 2A). We screened 7 ϫ 10 5 clones of a rat lung library and obtained 18 positive clones. Nine clones encoded the C-terminal portion of ␣-actinin-1 (aa 406 -892; human, BC015766; GenBank TM ). ␣-Actinin is a well characterized protein that shows F-actin-cross-linking activity (77). Four isoforms of human ␣-actinin have been identified: nonmuscle actinin-1 and actinin-4 and muscle actinin-2 and actinin-3 (78 -81). ADIP bound ␣-actinin-2 in addition to ␣-actinin-1 as estimated by yeast two-hybrid analysis (data not shown), indicating that the binding of ADIP to ␣-actinin is not specific for ␣-actinin-1. Two-hybrid analysis revealed that the region containing only the first coiled-coil domain (aa 1-226) of ADIP bound ␣-actinin-1 ( Fig. 2A) and that the two C-terminal EF-hand motifs of ␣-actinin-1 (aa 740 -892) were required for its binding to ADIP (data not shown). These results indicate that the first coiled-coil domain of ADIP binds to the EF-hand motifs of ␣-actinin-1.
To confirm the binding of ADIP to ␣-actinin-1 in vivo, we performed the immunoprecipitation analysis. HEK293 cells were co-transfected with the HA-tagged C-terminal portion of ␣-actinin-1 (HA-␣-actinin-1-C; aa 406 -892) and FLAG-tagged ADIP (FLAG-mADIP-M; aa 152-436). When FLAG-tagged ADIP was immunoprecipitated from the cell extract with the anti-FLAG mAb, HA-␣-actinin-1-C was co-immunoprecipitated, as detected by Western blotting with the HA mAb (Fig.  8A). When FLAG-mADIP-M was immunoprecipitated with the anti-FLAG Ab from the extract of HEK293 cells transiently expressing the FLAG-tagged mADIP (FLAG-mADIP-M; aa 152-436) alone, endogenous ␣-actinin was co-immunoprecipitated (Fig. 8B). We next examined whether ␣-actinin-1 directly interacted with ADIP in vitro. A GST fusion protein of the C-terminal region of ␣-actinin-1 (GST-␣-actinin-1-C) bound to an MBP fusion protein of the region of ADIP containing all three coiled-coil domains (aa 121-436) (MBP-mADIP-F) immobilized on amylose resin beads. GST-␣-actinin-1-C bound to MBP-mADIP-F but not to MBP alone (Fig. 8C). Finally, when endogenous ADIP was immunoprecipitated from the extract of MDCK cells with the anti-ADIP pAb, endogenous ␣-actinin was co-immunoprecipitated (Fig. 8D). ␣-Actinin was not co-immunoprecipitated with control IgG. These results indicate that ADIP directly binds ␣-actinin in vivo and in vitro. DISCUSSION We have isolated here a novel afadin-binding protein and named it ADIP. Several lines of evidence suggest that ADIP binds to afadin in vivo at cell-cell AJs: 1) ADIP binds to afadin, as estimated by the yeast two-hybrid analysis and co-immunoprecipitation from the extracts of cells exogenously expressing the fragments of ADIP; 2) recombinant ADIP directly binds to the DIL domain of recombinant afadin in a cell-free system; 3) endogenous ADIP and afadin are co-immunoprecipitated from the extracts of MDCK cells; and 4) ADIP co-localizes with afadin and nectin at cell-cell AJs. Several proteins, including E-cadherin, ␣-catenin, ␤-catenin, and ␣-actinin, are concentrated at AJs but more widely distributed from the apical to basal sides of the lateral plasma membranes (2,21). Four proteins (vinculin, afadin, nectin, and ponsin) strictly localize at AJs, which are defined using ultrastructural analysis as closely apposed plasma membrane domains reinforced by a dense cytoplasmic plaque to which F-actin bundles attach (7,18,19,82), whereas vinculin and ponsin furthermore localize at cell-matrix junctions (82). Ponsin is the afadin-and vinculinbinding protein containing three Src homology 3 domains (82). ADIP is the fifth protein that strictly localizes at AJs.
We have furthermore shown that ADIP binds ␣-actinin. Our results suggest that ADIP binds to ␣-actinin in vivo. 1) ADIP binds to ␣-actinin as estimated by the yeast two-hybrid analysis and co-immunoprecipitation from the extracts of HEK293 cells exogenously expressing the full-length ADIP or fragments; 2) recombinant ADIP directly binds to recombinant ␣-actinin in a cell-free system; and 3) endogenous ADIP and ␣-actinin are co-immunoprecipitated from the extracts of MDCK cells.
What is the function of ADIP? Our study indicates that ADIP binds two F-actin-binding proteins, afadin and ␣-actinin. Afadin binds along the sides of F-actin but does not have a marked F-actin-cross-linking activity (18). ␣-Actinin has an F-actincross-linking activity (77). Thus, ADIP links afadin to ␣-actinin and functions in the formation of the specialized actin structure at cell-cell AJs. Recently, we have shown that the heterotypic trans-interaction between nectin-2 in Sertoli cells and nectin-3 in spermatids is formed at Sertoli cell-spermatid junctions, heterotypic AJs in the testis, and that each nectin-based adhesive membrane domain exhibits one-to-one co-localization with each actin bundle underlying Sertoli cell-spermatid junctions (63). Afadin also co-localizes with nectin at Sertoli cell-spermatid junctions. Our present finding that ADIP binds both afadin and ␣-actinin suggests that afadin, together with ADIP and ␣-actinin, functions in the formation of the actin bundle at the nectin-based cell-cell adhesion sites. Since it has been reported that E-cadherin also associates with ␣-actinin through ␣-catenin (9,16), two cell-cell adhesion systems at AJs, the nectinafadin and cadherin-catenin systems, are connected to actin cytoskeleton through ␣-actinin. We have previously shown that nectin has a potency to recruit the E-cadherin-␤-catenin complex to the nectin-based cell-cell adhesion sites through afadin and ␣-catenin in fibroblasts and epithelial cells (38 -40). Our finding that ADIP binds both afadin and ␣-actinin suggests that ADIP serves as a linker between the nectin-afadin and cadherin-catenin systems.
We originally isolated afadin as an F-actin-binding protein (18). We subsequently found that afadin binds nectin and ponsin (18,19) and furthermore isolated here another afadinbinding protein, ADIP, which bound ␣-actinin in this study (Fig. 9). Afadin binds these proteins through different regions; nectin binds to the PDZ domain (19), ponsin binds to the third proline-rich domain (82), and ADIP binds to the DIL domain (Fig. 9). It has furthermore been reported that afadin directly binds ␣-catenin, although its binding is not strong (38,83). In addition, the splice variant of afadin, AF-6, binds Rap1 and profilin through the Ras-associated domain and the carboxylterminal region, respectively (84), suggesting that afadin also binds them. Thus, the F-actin-binding protein, afadin, may serve as a scaffold molecule to organize various proteins including other actin-binding proteins, ␣-catenin, ␣-actinin, and profilin, at the nectin-based cell-cell adhesion sites (Fig. 9). The finding that the proper organization of AJs and TJs is impaired in epithelial cells of afadin (Ϫ/Ϫ)-mice and (Ϫ/Ϫ)-embryoid bodies (60), suggests that this afadin-based organization of the various proteins is important for the formation of AJs and TJs. It is of crucial importance to clarify the molecular mechanism FIG. 8. In vivo binding of ADIP to ␣-actinin. A, co-immunoprecipitation (IP) of the HA-tagged ␣-actinin-1-C with the FLAG-tagged mADIP-M. Expression vectors were transfected into HEK293 cells as indicated. The HA-tagged ␣-actinin-1-C was specifically co-immunoprecipitated with the FLAG-tagged mADIP-M, as is shown by Western blotting with the anti-HA and anti-FLAG mAbs. B, co-immunoprecipitation of endogenous ␣-actinin with the FLAG-tagged mADIP-M. Expression vectors were transfected into HEK293 cells as indicated. Endogenous ␣-actinin-1 was specifically immunoprecipitated with FLAGtagged mADIP-M, as is shown by Western blotting (IB) with the anti-␣-actinin pAb and the anti-FLAG mAb. C, in vitro binding of ␣-actinin-1 to mADIP. The purified protein of GST-␣-actinin-1-C (aa 406 -892) was incubated with either MBP or MBP-mADIP-F (aa 121-436) immobilized on amylose resin beads. The eluates were then subjected to SDS-PAGE (10% polyacrylamide gel), followed by Western blotting with the anti-GST mAb. D, co-immunoprecipitation of endogenous ␣-actinin and endogenous afadin with endogenous ADIP from MDCK cells. The extracts of MDCK cells were immunoprecipitated with the anti-ADIP pAb (M05) and analyzed by Western blotting with the anti-ADIP (M05), anti-␣-actinin, and anti-afadin Abs. The results are representative of three independent experiments.