Necl-5/Poliovirus Receptor Interacts in cis with Integrin αVβ3 and Regulates Its Clustering and Focal Complex Formation*

Integrin αvβ3, which forms focal complexes at leading edges in moving cells, is up-regulated in cancer cells and so is implicated in their invasiveness. Necl-5, originally identified as a poliovirus receptor and also up-regulated in cancer cells, colocalizes with integrin αvβ3 at leading edges in moving cells and enhances growth factor-induced cell movement. Here, we show that Necl-5 interacts directly, in cis, with integrin αvβ3, and enhances integrin αvβ3 clustering and focal complex formation at leading edges in NIH3T3 cells. The extracellular region of Necl-5, but not the cytoplasmic region, is necessary for its interaction with integrin αvβ3; however, both regions are necessary for its action. An interaction between integrin αvβ3 and vitronectin and PDGF-induced activation of Rac are also necessary for integrin αvβ3 clustering. The interaction between Necl-5 and integrin αvβ3 enhances PDGF-induced Rac activation, facilitating integrin αvβ3 clustering presumably in a feedback amplification manner. Thus, Necl-5 has a critical role in integrin αvβ3 clustering and focal complex formation.

Integrins are key molecules for adhesion between cells and extracellular matrix (ECM) 2 proteins (1). They are heterodimers of ␣ and ␤ subunits, both of which have one transmembrane segment. The extracellular region of integrins binds to ECM proteins, whereas the cytoplasmic region is directly and indirectly associated with many F-actin-binding proteins, such as talin, vinculin, and ␣-actinin, and intracellular signaling molecules, such as FAK and c-Src (1,2). Integrins have at least two forms: conformations with either low or high affinity for their ECM binding partners (1,3,4). When talin binds to the low affinity form, this is converted to the high affinity form (4). Upon binding to ECM proteins, integrins transduce signals inside cells that then cause the reorganization of the actin cytoskeleton, eventually resulting in integrin clustering and the formation of focal complexes and focal adhesions (4). Focal complexes are formed at leading edges in moving cells, whereas focal adhesions, also called focal contacts, are formed at sites to the rear of the leading edges (5,6). Focal complexes are generally smaller than mature focal adhesions, being less than 0.5 m in diameter. Protrusions, such as filopodia and lamellipodia, are also formed at leading edges, while ruffles are formed on the dorsal surfaces of lamellipodia. Focal complexes are formed under these protrusions. Focal complexes, protrusions, and ruffles are formed by the actions of the small G proteins Rac and Cdc42; their formation is inhibited by the action of the small G protein RhoA (5,7). By contrast, focal adhesions are formed by the action of RhoA. Inactivation of Rac and Cdc42 and activation of RhoA lead to the transformation of focal complexes into focal adhesions. The dynamic formation of all of these structures is necessary for efficient cell movement. The dynamic activation and inactivation of Rac, Cdc42, and RhoA are regulated not only by the outside-in signaling of integrins but also by growth factor-induced signaling (8). Of the many integrins, integrin ␣ V ␤ 3 has a particularly important role in focal complex formation (9). This integrin is expressed in many cell types, such as fibroblasts and vascular endothelial cells (10,11), and is often up-regulated in many cancer cells, such as colon carcinoma, melanoma, and glioblastoma cells (12)(13)(14).
We found that an Ig-like molecule Necl-5/poliovirus receptor (PVR)/CD155/Tage4 co-localizes with integrin ␣ V ␤ 3 at leading edges in moving cells and enhances growth factor-induced cell movement (15). Human PVR/CD155 was originally identified as a poliovirus receptor (16,17), whereas rodent Tage4 was originally identified as the product of a gene that is overexpressed in rodent colon carcinoma (18). PVR/CD155 is also overexpressed in many human cancer cells, such as colon carcinoma, melanoma, and glioblastoma cells, in which integrin ␣ V ␤ 3 is up-regulated (19 -21). This molecule, with four nomenclatures, is also named nectin-like molecule-5, Necl-5 (22). Necl-5 is expressed in many cell types, such as fibroblasts and vascular endothelial cells, in which integrin ␣ V ␤ 3 is expressed (23,24). Necl-5 does not show homophilic cell-cell adhesion activity, but it heterophilically interacts in trans (i.e. present on a different cell) with nectin-3, a member of the Iglike nectin family that is a Ca 2ϩ -independent cell-cell adhesion molecule and forms adherens junctions cooperatively with cad-herins (22,25). When cells come into contact with other cells, Necl-5 is down-regulated from the cell surface by trans-interacting with nectin-3 (26); however, when there is no contact with other cells, Necl-5 is up-regulated at the transcriptional level by growth factor-induced signaling (27). Up-regulation of Necl-5 enhances growth factor-induced cell movement, FIGURE 1. Co-localization of Necl-5 with integrin ␣ V ␤ 3 at peripheral ruffles and focal complexes. A, immunofluorescence images of PDGF-stimulated NIH3T3 cells cultured on vitronectin-coated -slide dishes. Cells were double-or triple-stained with various combinations of the anti-Necl-5 mAb, the anti-integrin ␤ 3 mAb, and the anti-talin mAb. a, wild-type NIH3T3 cells; b, Necl-5-NIH3T3 cells; c, wild-type NIH3T3 cells co-transfected with siRNA vector and pEGFP-tub; c1, Necl-5-knockdown-NIH3T3 cells; c2, control-NIH3T3 cells. Arrowheads, leading edges; asterisks, siRNA vector-transfected cells; scale bars, 10 m. Inset boxes show the area of high magnification images. The high magnification images are indicated as intensity ratio images. Analysis of the co-localization of Necl-5 and integrin ␤ 3 and generation of the intensity ratio images were performed by ImageJ 1.34S software with RG2B colocalization plugin. B, quantitative analysis of various structures in NIH3T3 and Necl-5-NIH3T3 cells. a, quantification of peripheral ruffles at leading edges. To quantify peripheral ruffles, the signal for F-actin, which is a major component of peripheral ruffles, was observed. Peripheral ruffles at leading edges were measured and classified into two categories: small ruffles (less than 10 m in width) and large ruffles (10 m or more in width). The results are the means Ϯ S.E. of the three independent experiments. b, strength of peripheral ruffles at leading edges. To analyze the strength of peripheral ruffles, the maximum intensity of F-actin signal at the peripheral ruffles was measured using ImageJ 1.34S software and averaged (n ϭ 27). *, p Ͻ 0.0002. The results are the means Ϯ S.E. of the twenty-seven cells. c, quantification of focal complexes at leading edges. The patterns of the signal for integrin ␤ 3 at focal complexes of leading edges were measured and classified into three categories: scattered focal complexes (Scattered FX), line-like distributed focal complexes (Line-like FX), and belt-like distributed focal complexes, which were assembled more than two line-like FX (Belt-like FX). The results are the means Ϯ S.E. of the three independent experiments. Immunofluorescence images are the staining of integrin ␤ 3 and examples of three categories of focal complexes. d, quantification of focal adhesions per cell. The signal for integrin ␤ 3 at focal adhesions was counted and averaged per cell (n ϭ 27). **, p Ͻ 0.0001. The results are the means Ϯ S.E. of the 27 cells.
whereas down-regulation of Necl-5 reduces it (26). Cultured cells continue to move until they contact other cells and become confluent. They then undergo cell-cell adhesion and stop moving. This phenomenon has been known for more than fifty years as contact inhibition of cell movement (28,29). We have proposed that Necl-5 has a role, at least in part, in this contact inhibition of cell movement (26). However, we did not previously examine which structure of leading edges Necl-5 localizes to or whether, and if so how, Necl-5 regulates integrin ␣ V ␤ 3 -based focal complex formation. We attempt to address these issues here using NIH3T3 cells as a model cell type, which express both integrin ␣ V ␤ 3 and Necl-5.
Directional Stimulation by PDGF-To generate a concentration gradient of PDGF, a -Slide VI flow (uncoated; Ibidi) was used. The -Slide VI flow has six parallel channels, which were coated with 5 g/ml vitronectin, 25 g/ml fibronectin, or 80 g/ml laminin. Cells were seeded at a density of 5 ϫ 10 3 cells per square centimeter, cultured for 16 h, and starved of serum with DMEM containing 0.5% BSA for 1 h. The concentration gradient of PDGF was made using DMEM containing 0.5% BSA and 30 ng/ml PDGF according to manufacturer's protocol. After 30 min, cells were fixed with acetone/methanol (1:1), incubated with 1% BSA in PBS, and then incubated with 20% BlockAce in PBS, prior to immunofluorescence microscopy (26). The samples were analyzed by confocal laser-scanning microscope systems, digital eclipse C1si-ready (NIKON), and LSM510 META (Carl Zeiss MicroImaging).
Co-immunoprecipitation Assay-HEK293 cells were co-transfected with various combinations of plasmids, cultured for 24 h, detached with 0.05% trypsin and 0.53 mM EDTA, and treated with a trypsin inhibitor. To make predominantly high affinity integrin ␣ V ␤ 3 , cells were cultured in suspension with Ca 2ϩ -free DMEM (Invitrogen) containing 0.5% BSA, 1 mM MnCl 2 , and 50 g/ml cyclo-RGDfV for 1 h, collected by centrifugation, washed with Wash buffer (20 mM Tris-HCl, pH 8.0, 150 mM NaCl, and 1 mM Na 3 VO 4 ), and lysed with Buffer A (20 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1 mM MnCl 2 , 10% glycerol, 1% Nonidet P-40, 10 mM NaF, 1 mM Na 3 VO 4 , 10 g/ml leupeptin, 2 g/ml aprotinin, and 10 M APMSF). To make predominantly low affinity integrin ␣ V ␤ 3 , cells were cultured in suspension with DMEM containing 0.5% BSA for 1 h, collected by centrifugation, washed with Wash buffer and lysed with Buffer B (20 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 10% glycerol, 1% Nonidet P-40, 10 mM NaF, 1 mM Na 3 VO 4 , 10 g/ml leupeptin, 2 g/ml aprotinin, and 10 M APMSF). The lysates were rotated for 30 min and subjected to centrifugation at 12,000 ϫ g for 20 min. The supernatant was precleared with protein G-Sepharose 4 Fast Flow beads (Amersham Biosciences) at 4°C for 1 h, incubated with the anti-FLAG FIGURE 2. Physical interaction of Necl-5 with integrin ␣ V ␤ 3 . A, co-immunoprecipitation of integrin ␣ V ␤ 3 with Necl-5. a and b, HEK293 cells were transfected with various combinations of the indicated vectors and cultured in suspension under the indicated conditions. FLAG-tagged Necl-5, Necl-5-⌬CP, or Necl-5-⌬EC was immunoprecipitated using the anti-FLAG mAb and samples were assessed by Western blotting using the anti-FLAG mAb, the anti-integrin ␣ V mAb, the anti-integrin ␤ 3 mAb, the anti-integrin ␣ 5 mAb, and the anti-integrin ␤ 1 mAb. a, integrin ␣ V ␤ 3 ; b, integrin ␣ 5 ␤ 1 . c, NIH3T3 cells were cultured on vitronectincoated dishes, starved of serum, and cultured in the presence or absence of PDGF. Cells were treated with DTSSP and then subjected to the co-immunoprecipitation assay using the anti-integrin ␣ V pAb. Samples were assessed by Western blotting using the anti-Necl-5 mAb, the anti-integrin ␣ V mAb, and the anti-integrin ␤ 3 pAb. The results shown are representative of three independent experiments. B, in vitro binding of Necl-5 to integrin ␣ V ␤ 3 . Necl-5 EC (60 pmol) or buffer alone was incubated with integrin ␣ V ␤ 3 EC (6 pmol)-immobilized Ni-Sepharose beads or Ni-Sepharose beads alone. After the beads were extensively washed, bound proteins were eluted with Elution buffer. The samples were assessed by Western blotting using the anti-Necl-5 mAb, the anti-integrin ␣ V mAb, and the anti-integrin ␤ 3 mAb. The results shown are representative of three independent experiments. mAb at 4°C for 4 h, and then incubated with protein G-Sepharose beads at 4°C for 2 h. After the beads were extensively washed with Buffer A or B, bound proteins were eluted by boiling the beads in SDS Sample Buffer (60 mM Tris-HCl, pH 6.7, 3% SDS, 2% 2-mercaptoethanol, and 5% glycerol) for 5 min and subjected to SDS-PAGE followed by Western blotting.
To examine the interaction between endogenous Necl-5 and endogenous integrin ␣ V ␤ 3 , a co-immunoprecipitation assay was performed using NIH3T3 cells. Cells were plated at a density of 5 ϫ 10 3 cells per square centimeter on vitronectin-coated dish, cultured for 16 h, starved of serum with DMEM containing 0.5% BSA for 1 h, and then stimulated with DMEM containing 0.5% BSA and/or 3 ng/ml PDGF. After 30 min, cells were washed with ice-cold PBS and incubated with 2 mM DTSSP in PBS at 4°C for 2 h. To quench the cross-linking reaction, 1 M Tris-HCl (pH 7.5) was added at a final concentration of 50 mM. Cells were washed with PBS and lysed with Buffer B. The lysates were rotated for 30 min and subjected to centrifugation at 12,000 ϫ g for 20 min. The supernatant was precleared with protein A-Sepharose (Amersham Biosciences) at 4°C for 1 h, incubated with the anti-integrin ␣ V pAb at 4°C for 16 h, and then incubated with protein A-Sepharose beads at 4°C for 4 h. After the beads were extensively washed with Buffer B, bound proteins were eluted by boiling the beads in SDS sample buffer for 5 min and subjected to SDS-PAGE followed by Western blotting.
In Vitro Binding of Necl-5 to Integrin ␣ V ␤ 3 -Necl-5 EC was prepared as described (25). To obtain integrin ␣ V ␤ 3 EC, CHO-lec 3.2.8.1 cells expressing integrin ␣ V ␤ 3 EC were cultured, and the culture supernatant containing soluble integrin ␣ V ␤ 3 EC was collected (3). This cell line was kindly supplied by Dr. J. Takagi (Osaka University, Suita, Japan). The culture supernatant was applied to Ni-Sepharose 6 Fast Flow beads (Amersham Biosciences) and equilibrated with Buffer C (25 mM Tris-HCl at pH 8.0, 200 mM NaCl, 1 mM CaCl 2 , 1 mM MgCl 2 , and 20 mM imidazole at pH 8.0). After the beads were extensively washed with Buffer C and then Buffer C containing 0.6 M NaCl, bound integrin ␣ V ␤ 3 EC was eluted with Elution buffer (25 mM Tris-HCl at pH 8.0, 200 mM NaCl, 1 mM CaCl 2 , 1 mM MgCl 2 , and 500 mM imidazole at pH 8.0) and the eluate was dialyzed with Buffer C. The protein concentration of integrin ␣ V ␤ 3 EC was determined with BSA as a reference protein by SDS-PAGE.
To examine the binding of Necl-5 EC to integrin ␣ V ␤ 3 EC, integrin ␣ V ␤ 3 EC (6 pmol) was immobilized on Ni-Sepharose beads and soluble Necl-5 EC (60 pmol) was incubated with integrin ␣ V ␤ 3 EC-immobilized Ni-Sepharose beads or Ni-Sepharose beads alone in 0.3 ml of Buffer C containing 1% BSA. After the beads were extensively washed with Buffer C, bound proteins were eluted with Elution buffer. The eluate was then subjected to SDS-PAGE, followed by Western blotting.
Assay for Bead-Cell Contact-Latex-sulfate microbeads (4.55 ϫ 10 6 ; 10-m diameter; Polyscience Inc.) were washed, resuspended in 0.2 ml of PBS, and incubated with 10 g of vitronectin, 50 g of fibronectin, or 160 g of laminin with gentle mixing at room temperature. After the incubation, the beads were washed three times with 0.5 ml of PBS and resuspended in 0.1 ml of PBS containing 1% BSA. ConA-coated beads were prepared as described (32) and used as control beads. Cells were starved of serum with DMEM containing 0.5% BSA for 18 h. After serum starvation, cells were detached with 0.05% trypsin and 0.53 mM EDTA and then treated with a trypsin inhibitor (Sigma). Cells were plated at a density of 1 ϫ 10 4 cells per square centimeter on laminin-coated coverglass, cultured for 3 h, and incubated with vitronectin-, fibronectin-, laminin-, or ConA-coated beads for 1 h. Cells were fixed with acetone/methanol (1:1), incubated with 1% BSA in PBS, and then incubated with 20% BlockAce in PBS, followed by immunofluorescence microscopy (26).
Assay for Rac Activation-The assay for Rac activation using GST-PAK CRIB was performed as described (33). Briefly, NIH3T3, Necl-5-NIH3T3, or Necl-5-⌬CP-NIH3T3 cells were cultured on vitronectin-coated dishes for 16 h, starved of serum with DMEM containing 0.5% BSA for 1 h, stimulated with 3 ng/ml PDGF, and then subjected to the assay for Rac activation. For analysis of Necl-5-knockdown cells, NIH3T3 cells were twice transfected with siRNA oligo every 24 h, cultured for 24 h, and subjected to the assay.

RESULTS
Co-localization of Necl-5 with Integrin ␣ V ␤ 3 at Peripheral Ruffles and Focal Complexes-NIH3T3 cells were sparsely plated on -slide VI flow dishes precoated with vitronectin, an ECM protein that binds to integrin ␣ V ␤ 3 (1), starved of serum, and directionally stimulated by PDGF. Most cells became polarized and formed lamellipodia with peripheral ruffles at leading edges, in the direction of higher concentrations of PDGF. The immunofluorescence signals for Necl-5 and integrin ␤ 3 were concentrated and co-localized at the peripheral ruffles of the leading edges in the middle section of the cells (Fig.  1Aa), consistent with earlier observations (15). In the basal section of the cells, the signal for integrin ␤ 3 was observed as dotlike structures under the peripheral ruffles. The signal for Necl-5 was also observed as fuzzy dot-like structures and mostly overlapped with the signal for integrin ␤ 3 . In addition, the signal for integrin ␤ 3 , but not that for Necl-5, was observed as dot-like structures at sites to the rear of the leading edges. Essentially the same results were obtained when integrin ␣ V was stained instead of integrin ␤ 3 . 3 Indeed, unless otherwise specified, essentially the same results were obtained for both integrin ␣ V and integrin ␤ 3 in the experiments that follow. The dot-like structures under the peripheral ruffles (which were immunopositive for Necl-5, integrin ␣ V , and integrin ␤ 3 ) were smaller in size than those at sites to the rear of the leading edges (which were immunopositive for integrin ␣ V and integrin ␤ 3 , but not for Necl-5). The signal for talin co-localized with both types of dot-like structure, consistent with earlier observations (6). In contrast to these signals, the signal for N-cadherin was not concentrated at any site on the entire plasma membrane (supplemental Fig. S1A). Here we determine that the dot-like structures, which are immunopositive for Necl-5, integrin ␣ V , integrin ␤ 3 , and talin and locate under the peripheral ruffles, correspond to focal complexes, whereas the dot-like structures, which are immunopositive for integrin ␣ V , integrin ␤ 3 , and talin, but not for Necl-5, correspond to focal adhesions. These results indicate that Necl-5 co-localizes with integrin ␣ V ␤ 3 both at peripheral ruffles over the lamellipodia and at focal complexes under the peripheral ruffles of leading edges, but not at focal adhesions, in NIH3T3 cells. Integrin ␣ V ␤ 3 at these structures includes at least the high affinity form, because talin co-localizes with it (4).
Regulation by Necl-5 of Integrin ␣ V ␤ 3 Clustering at Peripheral Ruffles and Focal Complexes-We previously showed that the amount of cell surface Necl-5 is regulated by cell density (26). Therefore, we next examined the effect of the amount of cell surface Necl-5 on integrin ␣ V ␤ 3 clustering at peripheral ruffles and focal complexes in NIH3T3 cells. When NIH3T3 cells stably overexpressing Necl-5 (Necl-5-NIH3T3 cells) were cultured on -slide dishes as described above and directionally stimulated by PDGF, most cells showed morphologies similar 3 Y. Takai, unpublished data. to those of wild-type NIH3T3 cells, except that they formed more marked peripheral ruffles than wild-type cells (Fig. 1, Ab, Ba, and Bb, see also Aa). The immunofluorescence signals for Necl-5, integrin ␤ 3 , and talin, but not that for N-cadherin, were highly concentrated at peripheral ruffles and focal complexes ( Fig. 1Ab and supplemental Fig. S1B). The number of focal complexes in the basal sections of these cells was increased compared with that in wild-type cells, but their diameters were unchanged (Fig. 1, Ab and Bc, see also Aa). In addition, the immunofluorescence signals for integrin ␤ 3 and talin were markedly decreased at focal adhesions in Necl-5-NIH3T3 cells compared with those in wild-type cells (Fig. 1, Ab and Bd, see also Aa). The signal for Necl-5 was observed as belt-like structures at leading edges and partly overlapped with the signals for integrin ␤ 3 and talin at focal complexes. The Necl-5-positive belt-like structures might be due to the increased number of dot-like structures caused by overexpression of Necl-5. When similar experiments were performed using NIH3T3 cells in which Necl-5 was knocked down (Necl-5-knockdown-NIH3T3 cells) by siRNA vector (about 80 -90% decrease of Necl-5; supplemental Fig. S2A), most cells did not form definitely leading edges, but randomly formed lamellipodia at various sites; the signal for Necl-5 was negligible (Fig. 1Ac, 1 and 2). The signal for integrin ␤ 3 was not concentrated at any region. The amounts of cell surface integrin ␣ V and integrin ␤ 3 were indistinguishable among wild-type NIH3T3, Necl-5-NIH3T3, Necl-5-knockdown-NIH3T3, and control-NIH3T3 cells as estimated by FACS analysis (supplemental Fig. S3, A and B).
These results indicate that Necl-5 enhances integrin ␣ V ␤ 3 clustering and formation of lamellipodia, peripheral ruffles, and focal complexes at leading edges in NIH3T3 cells, and suggest that Necl-5 physically interacts with integrin Physical Interaction of Necl-5 with Integrin ␣ V ␤ 3 -We therefore examined the physical interaction between Necl-5 and integrin ␣ V ␤ 3 by three methods: immunoprecipitation, in vitro binding, and beadcell contact assays. In the immunoprecipitation assay, we prepared low and high affinity forms of integrin ␣ V ␤ 3 to determine with which form Necl-5 interacts. To make predominantly high affinity integrin ␣ V ␤ 3 , human integrin ␣ V , human integrin ␤ 3 , and FLAG-Necl-5 were co-expressed in HEK293 cells, and the cells were cultured in suspension in the presence of Mn 2ϩ and cyclo-RGDfV (3). Alternatively, to make predominantly low affinity integrin ␣ V ␤ 3 , human integrin ␣ V , human integrin ␤ 3 T329C/ A347C, and FLAG-Necl-5 were co-expressed in HEK293 cells, and the cells were cultured in suspension in the presence of Ca 2ϩ and Mg 2ϩ . Integrin ␤ 3 T329C/A347C is stabilized in the low affinity form and cannot convert to the high affinity form with integrin ␣ V (34). When FLAG-Necl-5 was immunoprecipitated by an anti-FLAG mAb from cell lysates, both human integrin ␣ V and human integrin ␤ 3 or human integrin ␤ 3 T329C/A347C were co-immunoprecipitated (Fig. 2Aa). These results indicate that Necl-5 has the ability to physically interact with both the low and high affinity forms of integrin ␣ V ␤ 3 . The interaction of Necl-5 with integrin ␣ V ␤ 3 was noted as being specific, because integrin ␣ 5 ␤ 1 was not co-immunoprecipitated with FLAG-Necl-5 under comparable conditions (Fig.  2Ab).
The interaction of endogenous Necl-5 with endogenous integrin ␣ V ␤ 3 was then confirmed by a co-immunoprecipitation assay using lysates of NIH3T3 cells. NIH3T3 cells were cultured on vitronectin-coated dishes in the presence or absence of PDGF, and then treated with or without a chemical cross-linker DTSSP, which is membrane insoluble and has a disulfide bond. When endogenous integrin ␣ V was immunoprecipitated from lysates of cells cultured in the presence or absence of PDGF and treated with DTSSP, integrin ␤ 3 and Necl-5 were co-immunoprecipitated with it from both cell lysates irrespective of the presence and absence of PDGF, although the amount of coimmunoprecipitated Necl-5 in cells cultured in the presence of PDGF was slightly less than that in cells cultured in the absence of PDGF (Fig. 2Ac). The amount of integrin ␤ 3 co-immunoprecipitated with integrin ␣ V was not affected by PDGF stimulation. Integrin ␤ 3 , but not Necl-5, was co-immunoprecipitated with integrin ␣ V from lysates of cells which were not pretreated with DTSSP. 3 The exact reason why Necl-5 was not co-immunoprecipitated with integrin ␣ V in the absence of the cross-linker is currently unknown, but may be due to low amounts of these proteins in NIH3T3 cells and low affinities of these proteins.
We next examined whether the extracellular and/or cytoplasmic regions of Necl-5 are necessary for the interaction with integrin ␣ V ␤ 3 . We expressed human integrin ␣ V , human integrin ␤ 3 , or integrin ␤ 3 T329C/A347C and FLAG-Necl-5 with the cytoplasmic or extracellular region deleted (FLAG-Necl-5-⌬CP or FLAG-Necl-5-⌬EC, respectively), in HEK293 cells, and performed similar experiments to those described above. Both human integrin ␣ V and human integrin ␤ 3 or integrin ␤ 3 T329C/ A347C were co-immunoprecipitated with FLAG-Necl-5-⌬CP, whereas neither of them was co-immunoprecipitated with FLAG-Necl-5-⌬EC (Fig. 2Aa). These results indicate that the extracellular region of Necl-5, but not the cytoplasmic region, is essential for the interaction with both low and high affinity forms of integrin ␣ V ␤ 3 .
We then confirmed this result by use of purified recombinant protein samples. We prepared recombinant proteins of the extracellular region of Necl-5 fused to human IgG Fc portion (Necl-5 EC) and a heterodimer containing the extracellular regions of human integrin ␣ V and human integrin ␤ 3 with a hexahistidine-tag fused to the integrin ␤ 3 COOH terminus (integrin ␣ V ␤ 3 EC) (3). When Necl-5 EC was applied to a column of integrin ␣ V ␤ 3 EC-immobilized Ni-Sepharose beads, Necl-5 EC bound to integrin ␣ V ␤ 3 EC (Fig. 2B).
We furthermore confirmed the physical interaction between Necl-5 and integrin ␣ V ␤ 3 by the bead-cell contact assay. We first prepared microbeads coated with vitronectin and placed them on the surface of NIH3T3 cells that were starved of serum and sparsely plated on glass coverslips precoated with laminin. Consistently, the immunofluorescence signals for integrin ␤ 3 and talin were concentrated at the bead-cell contact sites (Fig.  3Aa1). The signal for Necl-5 was also concentrated there. These signals were markedly enhanced by overexpression of Necl-5, and reduced by its knockdown (Fig. 3A,  b1, c1, and c2). These results were statistically significant (Fig. 3B, a  and b). When ConA-coated beads were used as a control, none of these proteins were concentrated at the bead-cell contact sites (Fig. 3A, a2  and b2). These results support the cis-interaction between Necl-5 and integrin ␣ V ␤ 3 .
Regulation of Integrin ␣ V ␤ 3 Clustering by Necl-5-Integrin ␣ V ␤ 3 Interaction-We then examined whether the interaction between Necl-5 and integrin ␣ V ␤ 3 indeed regulates integrin ␣ V ␤ 3 clustering and the formation of lamellipodia, peripheral ruffles, and focal complexes at leading edges in NIH3T3 cells. For this purpose, we took advantage of a Necl-5 construct in which the cytoplasmic region was deleted (Necl-5-⌬CP), because we previously showed that this deletion mutant reduces the stimulatory effect of full-length Necl-5 on movement of L cells stably overexpressing Necl-5 and NIH3T3 and V12-Ki-Ras-NIH3T3 cells (15) and showed above that the extracellular region of Necl-5 interacts in cis (i.e. in the same cell membrane) with the extracellular region of integrin ␣ V ␤ 3 . Consistently, Necl-5-⌬CP inhibited the interaction between Necl-5 and integrin ␣ V ␤ 3 as detected by the immunoprecipitation assay (Fig. 4A). When Necl-5-⌬CP was stably expressed in NIH3T3 cells that were cultured on -slide dishes as described above and directionally stimulated by PDGF, most cells did not form leading edges but randomly formed lamellipodia at various sites (Fig.  4B). The immunofluorescence signal for Necl-5-⌬CP was diffusely distributed. The signal for integrin ␤ 3 was markedly decreased at focal complexes compared with that in wild-type cells. These results indicate that Necl-5-⌬CP shows a dominant-negative effect on intact Necl-5 through the inhibition of the interaction between Necl-5 and integrin ␣ V ␤ 3 and that the interac-tion between Necl-5 and integrin ␣ V ␤ 3 regulates integrin ␣ V ␤ 3 clustering and formation of lamellipodia, peripheral ruffles, and focal complexes at leading edges in NIH3T3 cells.
Necessity of Both the Extracellular and Cytoplasmic Regions of Necl-5 for Integrin ␣ V ␤ 3 Clustering-We examined whether integrin ␣ V ␤ 3 clustering needs both the extracellular and cytoplasmic regions of Necl-5. For this purpose, various Necl-5 mutants were expressed in Necl-5-knockdown-NIH3T3 cells. When fulllength FLAG-Necl-5 R , which is resistant to Necl-5 siRNA, was expressed in Necl-5-knockdown-NIH3T3 cells (supplemental Fig.  S2B, a and b), and the cells were cultured on -slide dishes precoated with vitronectin and directionally stimulated by PDGF, they showed morphologies similar to those of Necl-5-NIH3T3 cells (Fig. 5A, see also Fig. 1Ab). They had a polarized structure with lamellipodia and peripheral ruffles and the immunofluorescence signals for Necl-5 and integrin ␤ 3 were concentrated at peripheral ruffles over the lamellipodia and focal complexes under the peripheral ruffles. When FLAG-Necl-5-⌬CP R , which is also resistant to Necl-5 siRNA, or FLAG-Necl-5-⌬EC was expressed in Necl-5-knockdown-NIH3T3 cells (supplemental Fig. S2B, a and  b), the phenotypes of most Necl-5-NIH3T3 cells were not restored (Fig. 5, B and C). Morphologically, most cells did not form leading edges but randomly formed lamellipodia at various sites. The signals for FLAG-Necl-5-⌬CP R and FLAG-Necl-5-⌬EC were diffusely distributed. The signal for integrin ␤ 3 was not concentrated at any region, although the intensity of its signal was higher in the  cells expressing FLAG-Necl-5-⌬CP R than in the cells expressing FLAG-Necl-5-⌬EC. These results indicate that both the extracellular and cytoplasmic regions of Necl-5 are necessary for integrin ␣ V ␤ 3 clustering and formation of lamellipodia, peripheral ruffles, and focal complexes at leading edges in NIH3T3 cells.
Necessity of Binding of Integrin ␣ V ␤ 3 to Specific ECM Proteins for Necl-5-enhanced Integrin ␣ V ␤ 3 Clustering-It has been shown that the clustering of integrin requires its binding to ECM proteins and that integrin ␣ V ␤ 3 binds both vitronectin and fibronectin, but does not bind laminin (1). We examined whether the binding of integrin ␣ V ␤ 3 to its ECM proteins is necessary for PDGF-induced, Necl-5-enhanced integrin ␣ V ␤ 3 clustering and formation of lamellipodia, peripheral ruffles, and focal complexes at leading edges in NIH3T3 cells. When NIH3T3 cells were sparsely plated on -slide dishes precoated with fibronectin, starved of serum, and directionally stimulated by PDGF, essentially the same results were obtained as when dishes were precoated with vitronectin, in regards to the immunofluorescence signals for Necl-5, integrin ␤ 3 , and talin at leading edges (Fig. 6Aa1, see also Fig. 1Aa). However, the signal for integrin ␤ 3 was not observed at focal adhesions at sites to the rear of the leading edges, where the signal for integrin ␤ 3 was concentrated when cells were cultured on vitronectin. Instead of focal adhesions, the signal for talin was observed at fibrillar adhesions, which were formed by fibrillogenesis of fibronectin (Fig. 6Aa1). Fibrillar adhesions might be formed by other fibronectin receptors, such as integrin ␣ 5 ␤ 1 , which is expressed in NIH3T3 cells (1,35). When similar experiments were performed using Necl-5-NIH3T3 cells, essentially the same results as those obtained for vitronectin were obtained except that the signal for talin was observed at fibrillar adhesions (Fig. 6Aa2, see also Fig. 1Ab). In contrast, when similar experiments were performed using NIH3T3 or Necl-5-NIH3T3 cells that were cultured on dishes precoated with laminin instead of vitronectin, most cells spread but did not form leading edges in response to PDGF (Fig. 6A, b1 and b2). The signal for Necl-5 was diffusely observed on the plasma membrane and the signals for integrin ␤ 3 and talin were negligible at peripheral regions. However, the signal for talin was weakly observed at fibrillar adhesion-like structures. Laminin receptor, integrin ␣ 6 ␤ 1 , is expressed in NIH3T3 cells, and this result is consistent with earlier observations (36,37).
Essentially the same results were obtained in the bead-cell contact assay. When fibronectin-coated beads were placed on the surface of NIH3T3 cells that were starved of serum and sparsely plated on the dishes precoated with laminin, the signals, although faint, for Necl-5, integrin ␤ 3 , and talin were concentrated at the bead-cell contact sites (Fig. 6Ba1). Furthermore, the concentration of these signals was enhanced at the bead-cell contact sites when Necl-5-NIH3T3 cells were used (Fig. 6Ba2). These results were statistically significant (Fig.  6Bc). In contrast, when laminin-coated beads were used instead of vitronectin-coated ones, the signal for Necl-5, integrin ␤ 3 , or talin was not concentrated at the contact sites between the beads and NIH3T3 or Necl-5-NIH3T3 cells (Fig. 6B, b1, b2, and  c). Taken together, these results indicate that the binding of integrin ␣ V ␤ 3 to its ECM proteins is necessary for Necl-5-en-hanced integrin ␣ V ␤ 3 clustering and formation of lamellipodia, peripheral ruffles, and focal complexes at leading edges in NIH3T3 cells.
Necessity of PDGF Stimulation for Necl-5-enhanced Integrin ␣ V ␤ 3 Clustering-It has been shown that cell movement requires the synergistic activity of PDGF-and integrin-induced signaling (10). We therefore examined the effect of PDGF on Necl-5-enhanced integrin ␣ V ␤ 3 clustering and formation of lamellipodia, peripheral ruffles, and focal complexes at leading edges in NIH3T3 cells. When NIH3T3 cells were cultured on -slide dishes precoated with vitronectin and not stimulated by PDGF, most cells failed to form leading edges (Fig. 7A). The immunofluorescence signals for Necl-5, integrin ␤ 3 , and talin were negligible at peripheral regions. However, the large dotlike signals for integrin ␤ 3 and talin were concentrated at focal adhesions and the number of focal adhesions was increased compared with those observed in the presence of PDGF (see also Fig. 1Aa). When similar experiments were performed using Necl-5-NIH3T3 cells, most cells did not form leading edges (Fig. 7B). The signal for Necl-5 was diffusely observed on the plasma membrane but the signals for integrin ␤ 3 and talin were again negligible at peripheral regions. However, the signals for integrin ␤ 3 and talin were observed at focal adhesions, although they were not observed in Necl-5-NIH3T3 cells stimulated by PDGF (see also Fig. 1Ab). When similar experiments were performed using Necl-5-knockdown-NIH3T3 cells, most cells did not form leading edges (Fig. 7C). The signal for Necl-5 was essentially nil. The signal for integrin ␤ 3 was not concentrated anywhere, and the phenotype was different to those observed in NIH3T3 and Necl-5-NIH3T3 cells (see also Fig. 7,  A and B). These results indicate that PDGF stimulation is necessary for Necl-5-enhanced integrin ␣ V ␤ 3 clustering and formation of lamellipodia, peripheral ruffles, and focal complexes at leading edges in NIH3T3 cells.
Involvement of Rac Activation in PDGF-induced, Necl-5-enhanced Integrin ␣ V ␤ 3 Clustering-It has been shown that Rac, which is activated by the action of PDGF, regulates these morphological changes (6,8,38). We previously showed that overexpression of Necl-5 enhances the serum-induced activation of Rac in an integrin ␣ V ␤ 3 -dependent manner in L fibroblasts (15). We therefore examined whether Necl-5-integrin ␣ V ␤ 3 interaction enhances PDGF-induced Rac activation in NIH3T3 cells. PDGFinduced Rac activation was estimated by a pull-down assay. This form of Rac activation was enhanced and sustained by overexpression of Necl-5 and reduced by its knockdown (about 70% decrease of Necl-5) (Fig. 8A, a and b). It was also reduced by overexpression of Necl-5-⌬CP (Fig. 8Aa). These results indicate that Necl-5-integrin ␣ V ␤ 3 interaction enhances PDGF-induced Rac activation in NIH3T3 cells.
We then examined whether Rac is necessary for PDGF-induced, Necl-5-enhanced integrin ␣ V ␤ 3 clustering in NIH3T3 cells. Transient expression of a dominant-negative mutant of Rac1 (Rac1DN) in NIH3T3 cells or Necl-5-NIH3T3 cells cultured on -slide dishes precoated with vitronectin and directionally stimulated by PDGF, inhibited the formation of lamellipodia and ruffles at the peripheral regions of both cell lines (Fig. 8B, a and b). The immunofluorescence signals for Necl-5, integrin ␤ 3 , and Rac1DN were not concentrated at peripheral regions, but were concentrated at currently unknown structures on the plasma membrane and inside the cells. These results indicate that Rac is necessary for PDGF-induced, Necl-5-enhanced integrin ␣ V ␤ 3 clustering and formation of lamellipodia, peripheral ruffles, and focal complexes at leading edges in NIH3T3 cells.
Necessity of Both Rac Activation and Necl-5-Integrin ␣ V ␤ 3 Interaction for PDGF-induced Integrin ␣ V ␤ 3 Clustering-We next examined whether Rac activation alone is sufficient for integrin ␣ V ␤ 3 clustering and focal complex formation in NIH3T3 cells. For this purpose, we expressed a constitutively active mutant of Rac1 (Rac1CA) in the absence of PDGF stimulation. When Rac1CA was transiently expressed in NIH3T3 cells cultured on -slide dishes precoated with vitronectin in the absence of PDGF, most cells showed round structures with lamellipodia and peripheral ruffles over all peripheral regions (Fig. 9Aa). The immunofluorescence signals for Necl-5, integrin ␤ 3 , and Rac1CA were concentrated at peripheral ruffles and focal complexes. In addition, the signal for integrin ␤ 3 , but not that for Necl-5 or Rac1CA, was concentrated at focal adhesions. The number of focal complexes was markedly increased by Rac1CA compared with that in wild-type cells in the absence of PDGF (see also Fig, 7A). The reason why lamellipodia, peripheral ruffles, and focal complexes were formed randomly over all peripheral regions in the cells expressing Rac1CA might be due to the random localization of exogenously expressed Rac1CA to the plasma membrane. When NIH3T3 cells overexpressing Rac1CA were cultured on -slide dishes precoated with laminin instead of vitronectin, the cells showed morphologies different from those observed in the wildtype cells cultured on vitronectin, and formed lamellipodia unevenly along the plasma membrane but did not show round shapes (Fig. 9Ab). In these cells, the immunofluorescence signal for Necl-5 or integrin ␤ 3 was not concentrated at any region.
Necl-5-NIH3T3 cells overexpressing Rac1CA showed morphologies and staining patterns of Necl-5, integrin ␤ 3 , and Rac1CA similar to those of wild-type cells, except that they formed more marked peripheral ruffles and showed more marked staining of these markers than wild-type cells (Fig. 9Ba, see also Aa). When Necl-5-NIH3T3 cells overexpressing Rac1CA were cultured on -slide dishes precoated with laminin instead of vitronectin, the cells showed morphologies and staining patterns of Necl-5, integrin ␤ 3 , and Rac1CA similar to those of wildtype cells cultured on laminin (Fig.  9Bb, see also Ab). When NIH3T3 or Necl-5-NIH3T3 cells overexpressing Rac1CA were cultured on -slide dishes precoated with vitronectin and directionally stimulated by PDGF, PDGF showed no additional effect. 3 In contrast to wild-type and Necl-5-NIH3T3 cells, Necl-5-knockdown-NIH3T3 cells overexpressing Rac1CA showed morphologies different from those observed in wild-type cells: they formed lamellipodia unevenly along the plasma membrane, but did not show round shapes (Fig. 9C, see also Aa). The signal for Necl-5 was negligible. The signal for integrin ␤ 3 was not concentrated at any region, similar to the observation of Necl-5-knockdown-NIH3T3 cells in the absence of PDGF (see also Fig. 7C). Necl-5-⌬CP-NIH3T3 cells overexpressing Rac1CA showed phenotypes similar to those of Necl-5-knockdown-NIH3T3 cells overexpressing Rac1CA (Fig. 9D). Taken together, these results indicate that Rac activation alone or Necl-5-integrin ␣ V ␤ 3 interaction alone is not sufficient for integrin ␣ V ␤ 3 clustering and formation of lamellipodia, peripheral ruffles, and focal complexes in NIH3T3 cells. Rather, Rac activation, Necl-5-integrin ␣ V ␤ 3 interaction, and integrin ␣ V ␤ 3 -vitronectin interaction are all necessary for integrin ␣ V ␤ 3 clustering and morphological changes; however, Rac activation and integrin ␣ V ␤ 3 -vitronectin interaction are sufficient, without Necl-5-integrin ␣ V ␤ 3 interaction, for lamellipodium formation.

DISCUSSION
We first showed here that Necl-5 and integrin ␣ V ␤ 3 localized at leading edges in moving NIH3T3 cells predominantly where lamellipodia and peripheral ruffles were formed, and that they colocalized at peripheral ruffles and focal complexes that were formed over and under lamellipodia, respectively. We showed that Necl-5 did not localize at focal adhesions where integrin ␣ V ␤ 3 also localized. The immunofluorescence signal for Necl-5 was observed as fuzzy dot-like structures, which mostly, but not perfectly, overlapped with the dot-like signal for integrin ␣ V ␤ 3 at focal complexes. Because Necl-5 did not localize at focal adhesions and it was reported that focal complexes are transformed into focal adhesions by inactivation of Rac and Cdc42 and activation of RhoA (5, 7), Necl-5 might dissociate from integrin ␣ V ␤ 3 during this transformation, causing the formation of the fuzzy dot-like structures. The mechanism of dissociation of Necl-5 from integrin ␣ V ␤ 3 at focal complexes remains unknown.
We also showed that Necl-5 enhanced integrin ␣ V ␤ 3 clustering and formation of lamellipodia, peripheral ruffles, and focal complexes at leading edges in NIH3T3 cells. Studies on the mode of action of Necl-5 revealed that Necl-5 directly inter-acted with the extracellular regions of integrin ␣ V ␤ 3 to form a binary complex. There are theoretically cis-and trans-interactions between Necl-5 and integrin ␣ V ␤ 3 ; however, the results of the immunoprecipitation assay using suspended HEK293 cells and the bead-cell contact assay support an interaction in cis. The result that this interaction was not affected by EDTA, which would inhibit the interaction between integrin ␣ V ␤ 3 and a typical ligand, as judged by the co-immunoprecipitation assay using suspended HEK293 cells, 3 also supports this conclusion. Necl-5 interacted with both low and high affinity forms of integrin ␣ V ␤ 3 . Neither PDGF nor vitronectin were necessary for the Necl-5-integrin ␣ V ␤ 3 interaction as judged by the co-immunoprecipitation assay using suspended HEK293 cells. 3 In general, the low affinity integrin does not bind talin and barely clusters, whereas the high affinity form binds talin and interacts with its specific ECM proteins, resulting in the clustering of the integrin owing to both outside-in and inside-out signals. Therefore, the low affinity form of integrin ␣ V ␤ 3 , which interacts with Necl-5, might be diffusely present along the plasma membrane, excluding focal adhesions. By contrast, most integrin ␣ V ␤ 3 in clusters is likely to be the high affinity form that interacts with Necl-5 at focal complexes, but not at focal adhesions. It was previously shown that cell movement requires the synergistic activity of PDGF-and integrin-induced signaling (10). It was also shown that the PDGF receptor and integrin ␣ V ␤ 3 interact with each other, and that PDGF-induced signaling requires integrin ␣ V ␤ 3 activation by binding to its ECM proteins (10). The binding of talin to integrin increases the affinity of integrin for its specific ECM protein, and the binding of integrin to its specific ECM protein transduces signals inside cells leading to the reorganization of the actin cytoskeleton as necessary for integrin clustering (1, 4). These inside-out and outside-in signaling pathways from the growth factor receptor and integrin bring about the formation of focal complexes and focal adhesions. Consistently, we confirmed that both the binding of PDGF to its receptor and the binding of integrin ␣ V ␤ 3 to its ECM proteins, vitronectin and fibronectin, were necessary for integrin ␣ V ␤ 3 clustering and the formation of lamellipodia, peripheral ruffles, and focal complexes at leading edges in NIH3T3 cells. We also showed that Necl-5-integrin ␣ V ␤ 3 inter- A, effect of Necl-5 on PDGF-induced Rac activation. Cells were cultured on vitronectin-coated dishes, starved of serum, stimulated with PDGF, and then subjected to the assay for Rac activation. a, wild-type NIH3T3, Necl-5-NIH3T3, and Necl-5-⌬CP-NIH3T3 cells. b, control-and Necl-5-knockdown-NIH3T3 cells. Wild-type NIH3T3 cells were transfected with siRNA oligo. Expression level of endogenous Necl-5 in Necl-5-knockdown-NIH3T3 cells was less than 30% of that in control-NIH3T3 cells. B, immunofluorescence images of PDGF-stimulated NIH3T3 cells expressing Rac1DN, cultured on vitronectin-coated -slide dishes. Cells were transfected with Myctagged Rac1DN and triple-stained with the anti-Necl-5 mAb, the anti-integrin ␤ 3 mAb, and the anti-Myc mAb. a, wild-type NIH3T3 cells; b, Necl-5-NIH3T3 cells. Scale bars, 10 m. The results shown are representative of three independent experiments. action is additionally necessary for integrin ␣ V ␤ 3 clustering and morphological changes. Thus, three transmembrane proteins, the PDGF receptor, integrin ␣ V ␤ 3 , and Necl-5, physically and functionally associate to regulate integrin ␣ V ␤ 3 clustering and formation of lamellipodia, peripheral ruffles, and focal complexes at leading edges in NIH3T3 cells. We further showed that the cytoplasmic region of Necl-5 was not necessary for Necl-5-integrin ␣ V ␤ 3 interaction, but that both the extracellular and cytoplasmic regions of Necl-5 were necessary for integrin ␣ V ␤ 3 clustering and formation of these structures.
Rac was previously shown to have a critical role in the dynamic reorganization of the actin cytoskeleton necessary for the formation of focal complexes (6,8,38). We previously showed that Necl-5 enhances the serum-induced, integrin ␣ V ␤ 3dependent activation of Rac (15). Consistently, we showed here that PDGF-induced Rac activation was necessary for PDGF-induced integrin ␣ V ␤ 3 clustering and morphological changes in NIH3T3 cells. Rac activation alone or Necl-5-integrin ␣ V ␤ 3 interaction alone was insufficient to induce these events; indeed, Rac activation, Necl-5-integrin ␣ V ␤ 3 interaction, and integrin ␣ V ␤ 3 -vitronectin interaction were all necessary for integrin ␣ V ␤ 3 clustering and changes in morphology, although the Rac activation and integrin ␣ V ␤ 3 -vitronectin interaction were sufficient, without Necl-5-integrin ␣ V ␤ 3 interaction, for lamellipodium formation. We noted that these structures were randomly formed at peripheral regions when Rac1CA was expressed instead of being directionally stimulated by PDGF. These results indicate that PDGF-induced Rac activation is necessary for cell polarization and specific localization of lamellipodia, peripheral ruffles, and focal complexes in NIH3T3 cells.
Considering all the evidence thus far available, we propose the following mechanism of integrin ␣ V ␤ 3 clustering and formation of lamellipodia, peripheral ruffles, and focal complexes at leading edges in NIH3T3 cells: PDGF reaches and binds to its receptor and induces Rac activation; Rac activated in this way induces reorganization of the actin cytoskeleton, which induces clustering of the Necl-5-integrin ␣ V ␤ 3 complex; the clustered complex then enhances PDGFinduced Rac activation in a positive feedback manner. Such a feedback amplification mechanism might efficiently form leading edges in the direction of higher concentrations of PDGF. We used here only NIH3T3 cells as a model cell, but both Necl-5 and integrin ␣ V ␤ 3 are expressed in the same types of cells, such as other fibroblasts, vascular endothelial cells, colon carcinoma cells, melanoma cells, and glioblastoma cells, this mechanism may be applied to these cell types. Further studies are necessary for generalization of this mechanism.
The exact mechanism of how Necl-5 enhances integrin ␣ V ␤ 3 clustering remains unknown, but one possible explanation is that the Necl-5-integrin ␣ V ␤ 3 interaction induces a conformational change in the cytoplasmic region of integrin ␣ V ␤ 3 , making it sensitive to binding talin, which converts the low affinity integrin ␣ V ␤ 3 to the high affinity form. This process and/or the subsequent clustering process might be facilitated by Rac. Another possible explanation is that the cytoplasmic region of Necl-5 makes talin sensitive to binding integrin ␣ V ␤ 3 by an unknown mechanism. It remains unknown how the clustered Necl-5-integrin ␣ V ␤ 3 complex enhances PDGF-induced Rac activation, but signals downstream of the complex might enhance the signaling pathway of the PDGF receptor at a step upstream of Rac. Further studies are necessary to elucidate these unresolved issues.