Integrin Cytoplasmic Domain-associated Protein 1α (ICAP-1α) Interacts Directly with the Metastasis Suppressor nm23-H2, and Both Proteins Are Targeted to Newly Formed Cell Adhesion Sites upon Integrin Engagement*

Cell adhesion-dependent signaling implicates cytoplasmic proteins interacting with the intracellular tails of integrins. Among those, the integrincytoplasmic domain-associatedprotein 1α (ICAP-1α) has been shown to interact specifically with the β1 integrin cytoplasmic domain. Although it is likely that this protein plays an important role in controlling cell adhesion and migration, little is known about its actual function. To search for potential ICAP-1α-binding proteins, we used a yeast two-hybrid screen and identified the human metastatic suppressor protein nm23-H2 as a new partner of ICAP-1α. This direct interaction was confirmed in vitro, using purified recombinant ICAP-1α and nm23-H2, and by co-immunoprecipitation from CHO cell lysates over-expressing ICAP-1α. The physiological relevance of this interaction is provided by confocal fluorescence microscopy, which shows that ICAP-1α and nm23-H2 are co-localized in lamellipodia during the early stages of cell spreading. These adhesion sites are enriched in occupied β1 integrins and precede the formation of focal adhesions devoid of ICAP-1α and nm23-H2, indicating the dynamic segregation of components of matrix adhesions. This peripheral staining of ICAP-1α and nm23-H2 is only observed in cells spreading on fibronectin and collagen and is absent in cells spreading on poly-l-lysine, vitronectin, or laminin. This is consistent with the fact that targeting of both ICAP-1α and nm23-H2 to the cell periphery is dependent on β1 integrin engagement rather than being a consequence of cell adhesion. This finding represents the first evidence that the tumor suppressor nm23-H2 could act on β1 integrin-mediated cell adhesion by interacting with one of the integrin partners, ICAP-1α.

Cell adhesion to the extracellular matrix is mediated mainly by integrin clusters organized in transient focal complexes and more stable focal adhesions (1)(2)(3). These structures are linked physically to the actin cytoskeleton. Besides being the mechanical anchors of the cells, focal adhesions participate in outside-in and inside-out signaling. Integrin cytoplasmic domains have no known catalytic function, but they play a key role in the control of cytoskeleton organization and in signal transduction by recruiting many structural and signaling proteins (4). Although it is well documented that the adhesive function of most members of the integrin family can be activated in a phenotypically similar fashion, it is unclear whether common or independent cellular pathways underlie this apparent uniformity. Several proteins interacting with specific integrin cytoplasmic tails have been identified recently, suggesting that although the cytoplasmic domains of integrin ␤ subunits are quite similar, they are coupled to distinct functional pathways. For instance, ␤ 3 -endonexin binds specifically to the ␤ 3 integrin cytoplasmic tail (5) and increases the affinity of the integrin ␣ IIb ␤ 3 (6). A ␤ 2 integrin cytoplasmic domain-binding protein, cytohesin-1, has been found to increase ␣ L ␤ 2 -mediated cell adhesion (7). TAP-20 interacts with the ␤ 5 integrin and negatively regulates ␣ v ␤ 5 -dependent adhesion and focal adhesion assembly (8). Finally, the integrin cytoplasmic domain-associated protein 1␣ (ICAP-1␣) 1 interacts specifically with the Cterminal NPXY motif of the ␤ 1 integrin cytoplasmic domain (9,10) and impairs cell spreading when expressed as the T38D mutant, which potentially mimics ICAP-1␣ phosphorylated on threonine 38 (11).
Although it is clear that ICAP-1␣ plays important roles in the regulation of cell adhesion, the mechanism of ICAP-1␣ function in the signaling pathways has not yet been completely understood, due in part to the lack of information of the protein-protein interactions involving ICAP-1␣.
To elucidate the molecular basis of the ICAP-1␣ signaling pathway, we have carried out a yeast two-hybrid screen to identify its binding partners. Here we report that ICAP-1␣ interacts with the human metastatic suppressor protein nm23-H2 (called also NDP kinase B) (12), a cellular protein belonging to a family of highly conserved proteins in eukaryotes. Nm23 family proteins possess a nucleoside diphosphate kinase activity (13)(14)(15). Eight different genes of this family have now been identified in humans and were named nm23-H1, nm23-H2, to nm23-H8 (16). Apart from their role in nucleotides metabolism, nm23 isoforms are reportedly involved in a variety of cellular functions (17). Nm23-H2 has been shown to bind to the nuclease hypersensitive element of the c-myc and PDGF-A (platelet-derived growth factor A) promoter (12,18). Interestingly, expression of the nm23 genes is linked to suppression of tumor metastasis, differentiation, apoptosis, proliferation, and DNA mutation (19 -21). Introduction of nm23-H1 or -H2 reduces the metastatic potential and in vitro cell motility of tumor cells (22,23). Kantor et al. (24) report that murine melanoma cell lines and human breast carcinoma cells stably transfected with nm23-H1 lose their ability to migrate in response to different factors. Zhu et al. (25) report that nm23-H1 interacts with the Ras-related GTPase member Rad and reversibly converts GDP-Rad to GTP-Rad, thus acting as an exchange factor and a GTPase-activating protein for Rad. More recently, an association between nm23-H1 and Tiam1, a product of an invasion and metastasis-inducing gene, has been shown. This interaction could lead to the down-regulation of Rac1 activity (26). The mechanism of tumor suppression by nm23 is still poorly understood, although some speculations about the role of the enzyme have been presented (19). In this report, we show that ICAP-1␣ and nm23-H2 interact directly, co-localize and concentrate in peripheral ruffles, and are recruited to ␤ 1 integrin-rich cell adhesion sites in cells spreading on fibronectin and collagen. This particular cell localization supports the view that this association is relevant to a physiological process during the early stages of cell adhesion. It is the first report linking the tumor suppressor protein nm23-H2 to the cell adhesion and migration machinery.
Generation of Anti-ICAP-1␣ Antibodies-Mouse monoclonal anti-ICAP-1␣ antibodies (4D1D6 and 9B10) were prepared using recombinant His-tagged ICAP-1␣ protein as antigen. Briefly, hybridoma supernatants were screened initially by ELISA and Western blotting using recombinant His-tagged ICAP-1␣ protein. The monoclonal antibody 4D1D6 was further selected for reactivity in immunofluorescence studies. Recombinant His-tagged ICAP-1␣ protein was also used for the production of rabbit polyclonal antibodies (Elevage des Dombes, Romans, France) that have been tested by Western blot using recombinant ICAP-1␣ and mammalian cell lysates.
Yeast Two-hybrid Assays-A cDNA fragment encoding the fulllength human ICAP1-␣ protein was inserted into the NdeI/BamHI sites of pAS2-1 vector (CLONTECH distributed by Ozyme, Montigny le Bretonneux, France). The sequence of the bait construct was verified by DNA sequencing, and the construct was introduced into Y190 yeast cells using a lithium acetate transformation protocol. The resulting construct (pAS2-1/ICAP-1␣) was used as bait to screen a human placenta MATCHMAKER cDNA library (6 ϫ 10 6 independent clones) according the manufacturer's protocol. Briefly, Y190 ( pAS2-1/ICAP1␣) cells transformed by the library plasmids were selected by plating on SD medium lacking tryptophan and leucine (SD-WL). Interaction of proteins encoded by pAS2-1/ICAP-1␣ and by the pACT2 library vectors was tested by growing the cells in the presence of 25 mM 3Ј-amino-1,2,4,-triazole (SD Ϫ WLH ϩ 3AT). Histidine-positive colonies were further tested for LacZ activation. The growth of blue colonies in the histidine-deficient medium indicated a positive interaction. 42 positive yeast colonies, as indicated by activation of both reporter genes (histidine and lacZ) were independently identified and isolated. Plasmids were isolated from positive yeast clones by a glass beads/phenol-chloroform extraction protocol provided by the manufacturer (CLON-TECH). Escherichia coli 1066 bacteria were then electroporated with purified plasmids according to the protocol provided by Qiagen. The pACT2 plasmids were isolated from E. coli 1066 and restriction (Hin-dIII)-mapped. Subsequently, the sequences of the inserts were determined by DNA sequencing (Genaxis, Nimes, France). The specificity of positive colonies with respect to the protein/protein interaction was further confirmed by testing ICAM-1 cytoplasmic domain inserted in pAS2-1, an irrelevant protein in this context and ␣-actinin-1 inserted in pACT2, a protein present in focal adhesions.
Solid Phase-based Binding Assays-The interaction between recombinant ICAP-1␣ and recombinant nm23-H2 was analyzed using a solid phase assay. Briefly, a 96-well tray (MaxiSorp, Nunc) was coated with either ICAP-1␣ proteins (40 g/ml) or NDPK proteins (nm23-H2 or NDPK from D. discoideum; 10 g/ml), for 16 h at 4°C and blocked with a PBS, 3% BSA solution for 1 h at room temperature. Increasing concentrations of soluble nm23-H2 or ICAP-1␣ were incubated for 1 h. After three washes in PBS, 0.1% Tween 20, detection of bound nm23-H2 or ICAP-1␣ was performed using the affinity-purified polyclonal antibodies directed against nm23-H2 or monoclonal antibody 9B10 directed against ICAP-1␣. Nonspecific binding to BSA-coated wells was subtracted from the results as background.
Coimmunoprecipitation Experiments-CHO cells were transiently transfected with pcDNA3.1/ICAP-1␣ or pcDNA3.1 vector using Exgen (Euromedex, Souffelweyersheim, France). Twenty-four hours after the transfection, the cells were lysed in 1% Nonidet P-40/glycerol buffer containing protease and phosphatase inhibitors for 45 min. The cell lysates (500 g of proteins) were incubated with 20 l of 9B10 ascites containing anti-ICAP-1␣ monoclonal antibody for 2 h. Subsequently, the samples were mixed with 60 l of immobilized protein G (Sigma-Aldrich). After incubation for 1 h, the beads were washed four times with the lysis buffer, and the bound proteins were released from the beads by boiling in 20 l of SDS-PAGE Laemmli sample buffer for 5 min. The samples were analyzed by Western blotting with either rabbit polyclonal anti-ICAP-1␣ antibodies (to check the immunoprecipitation of ICAP-1␣) or affinity-purified rabbit polyclonal anti-nm23-H2 antibodies (to evaluate the interaction between ICAP-1␣ and endogenous nm23-H2). Immunological detection was achieved with horseradish peroxidase-conjugated secondary antibody, and the staining was carried out with ECL according to the manufacturer's instructions (Amersham Biosciences).
Immunofluorescence Staining of Cells-Hs68 cells were cultured as a monolayer in Dulbecco's modified Eagle's medium containing 10% fetal calf serum and harvested with trypsin/EDTA. The cells were plated on coverslips that were precoated with 25 g/ml human plasma fibronectin and incubated for different lengths of times (as specified for each experiment) in a 37°C incubator under a 5% CO 2 , 95% air atmosphere to obtain cells at different stages of spreading. Within the first hour of plating, extensive membrane ruffling was observed in many of the cells that were spreading on fibronectin. Under these experimental conditions, most of the cells were fully spread within 4 h. The cells were fixed with 3% paraformaldehyde in PBS and permeabilized with 0.2% Triton X-100 in PBS. Nonspecific sites were blocked in 10% goat serum for 1 h at room temperature. Cells were stained for 1 h with either monoclonal or polyclonal antibodies in a moist chamber. Anti-nm23-H2 monoclonal (Seigakaku) and polyclonal antibodies were used at a final concentration of 1 g/ml. Anti ICAP-1␣ 4D1D6 monoclonal from hybridoma supernatant was used at a ratio of 1:3 and anti ICAP-1␣ polyclonal antibodies were used at 1:500. The 4B7R monoclonal antibody specific for human ␤ 1 integrin was used at 5 g/ml, anti-Rac1 was used at 2.5 g/ml, and anti-tubulin at 1:100. After rinsing, coverslips were incubated with appropriate Alexa-conjugated secondary antibodies (Molecular Probes, distributed by Interchim) for 30 min. For actin staining, coverslips were incubated with TRITC-phalloidin. The cells were mounted in Mowiol solution and viewed using a confocal laser scanning microscope (Zeiss LSM 410).

Identification of nm23-H2 as a Binding Partner of ICAP-
1␣-To identify the proteins directly involved in ICAP-1␣-mediated transduction signals, we used the full-length ICAP-1␣ cDN〈 fused to the GAL4 DNA-binding domain as bait in a yeast two-hybrid system to screen a human placenta cDNA library. Forty-two positive clones were obtained and sequenced. BLAST searches in cDNA data bases revealed that four inserts overlap with the cytoplasmic domain of ␤ 1 integrin, confirming the reported ICAP-1␣/␤ 1 integrin interaction already described (9). Four additional inserts coded for the human protein named nm23-H2. Introduction of only pAS-2/ICAP-1␣ or pACT2/ nm23-H2 construction did not result in activation of both reporter genes, indicating that neither ICAP-1␣ nor nm23-H2 can activate the reporter genes in the absence of the other binding partner (Fig. 1). In additional control experiments, another "bait," the cytoplasmic domain of ICAM-1, and another "prey," ␣-actinin-1, were tested for interaction with nm23-H2 and ICAP-1␣, respectively. In all of these cases, no detectable ␤-galactosidase activity was observed, characterizing the specificity of our screen.
Nm23-H2 Binds to ICAP-1␣ in Vitro and ex Vivo-To confirm the direct interaction of nm23-H2 with ICAP-1␣, we carried out an ELISA-based solid phase binding assay. These experiments revealed a saturable binding of ICAP-1␣ to nm23-H2 and vice versa ( Fig. 2A). In contrast, NDPK from D. discoideum did not bind to ICAP-1␣. In an independent approach to corroborate these results, we incubated recombinant His-tagged ICAP-1␣ bound to a cobalt chelating resin with a solution of purified recombinant nm23-H2 protein or with HeLa cell lysates containing endogenous nm23-H2 (Fig. 2B). In both cases, nm23-H2 bound to ICAP-1␣ protein as revealed by Western blot analysis. The results obtained with these pull-down assays indicate that recombinant ICAP-1␣ interacts with recombinant nm23-H2 protein detected as a monomer and an SDS-resistant dimer. When endogenous nm23-H2 from HeLa cells was used instead, only the dimeric form of nm23-H2 was retained by ICAP-1␣. We presume that this dimer is due to oxidative conditions in our experimental procedure. 2 The interaction with nm23-H2 was also tested with recombinant protein containing the N-terminal (1-99) or C-terminal (100 -200) half of ICAP-1␣ protein by pull-down assay (Fig. 2B) and solid phase assay (Fig.  2C). Only the C-terminal polypeptide was able to interact strongly with recombinant or cellular nm23-H2, supporting the idea that the nm23-H2 binding site is localized at the ICAP-1␣ C-terminal half. To determine whether the interaction between ICAP-1␣ and nm23-H2 also occurred ex vivo, we expressed ICAP-1␣ by transient transfection in CHO cells. Soluble extracts were prepared as described under "Experimental Procedures." Only in ICAP-1␣-transfected cells, immunoprecipitation of ICAP-1␣ using the anti-ICAP-1␣ 9B10 monoclonal antibody resulted in a co-immunoprecipitation of endogenous nm23-H2 as detected by Western blot analysis using an affinity-purified polyclonal antibody (Fig. 2D). Thus, consistent with ICAP-1␣/nm23-H2 interaction detected in yeast cells and in vitro, ICAP-1␣ and nm23-H2 form a complex in mammalian cells.

ICAP-1␣ and nm23-H2 Co-localize in Peripheral Ruffles and Are Recruited to ␤ 1 Integrin-rich Cell Adhesion Sites in Cell
Spreading on Fibronectin-To ascribe a physiological role to the association between ICAP-1␣ and nm23-H2, we examined their co-localization in vivo. To analyze the subcellular localization of ICAP-1␣, we generated a monoclonal ICAP-1␣ antibody that recognizes both recombinant and endogenous human ICAP-1␣ in immunofluorescence studies. By immunoblotting with His-tagged fusion proteins containing different domains of the ICAP-1␣ protein, we showed that this antibody recognizes an epitope located within the N-terminal 100 amino acid residues (not shown). This antibody was specific because it reacted neither with a His-tagged fusion protein containing the Cterminal region of ICAP-1␣ nor with other irrelevant Histagged fusion proteins (data not shown).
Despite the observed association of ICAP-1␣ with the cytoplasmic domain of ␤ 1 integrin, we obtained no evidence for ICAP-1␣ accumulation at ␤ 1 integrin-or vinculin-rich focal adhesion sites in fully spread cells. ICAP-1␣ was found primarily in the cytosol, with some concentrations in the perinuclear or nuclear region. We therefore analyzed the subcellular localization of ICAP-1␣ in cells during the early stages of spreading. Hs68 cells newly plated on fibronectin-coated coverslips were stained with either polyclonal or monoclonal anti-ICAP-1␣ antibodies. ICAP-1␣ was observed to be highly concentrated at the edge or at peripheral ruffles of spreading cells (Fig. 3). A similar localization of nm23-H2 was observed with the monoclonal as well as the specific polyclonal antibodies. Noticeably, until 45 min of adhesion, ICAP-1␣ co-localized with nm23-H2 in many cell adhesion sites resembling ruffles or lamellipodia at the cell periphery, suggesting that ICAP-1␣ and nm23-H2 are involved in integrin-mediated cell spreading (Fig. 3). As cells spread further (1 h after seeding), both ICAP-1␣ and nm23-H2 staining at the cell edges decreased. Thus, high concentrations of ICAP-1␣ and nm23-H2 appear transiently at the cell periphery during the process of spreading. We noted that immunostaining with anti-ICAP-1␣ antibodies showed a labeling similar to that of stress fibers in fully spread cells.
Previous studies have shown that ␣ 5 ␤ 1 integrins accumulate in the peripheral ruffles of cells spreading on fibronectin (29). We confirmed such a localization of ␤ 1 integrin at the edge of the spreading cells and showed in addition co-localization of ␤ 1 integrins with ICAP-1␣ in the peripheral ruffles by co-staining the cells with a monoclonal anti-␤ 1 integrin antibody (Fig. 4). When Hs68 fibroblasts adhered to the extracellular matrix protein fibronectin, F-actin-containing membrane ruffling was stimulated as the initial response upon Rac1 activation as described by others (1). Indeed, additional co-staining in the early state of spreading (30 min of spreading) showed the peripheral co-localization of ICAP-1␣ with actin and Rac1, but not with tubulin or vimentin, allowing a better characterization of these areas (Fig. 4).
ICAP-1␣ and nm23-H2 Are Recruited to Areas Enriched in Occupied ␤ 1 Integrins-Like other integrins, ␤ 1 integrins can exist in different functional states with respect to ligand binding. These changes involve both affinity modulation, by which conformational changes in the integrin heterodimer govern affinity for individual extracellular matrix proteins, and avidity modulation, by which changes in lateral mobility and integrin clustering affect the binding of cells to multivalent matrices. Here we used the monoclonal antibody 12G10, which recognizes a ligand-induced binding site (30), to investigate the functional state of ␤ 1 integrins co-localized with ICAP-1␣ and nm23-H2. During initial cell spreading, the 12G10 monoclonal antibody recognized engaged integrins at the cell edge after 30 min of spreading (early spreading) and in focal adhesions after 4 h of spreading (late spreading). Fig. 5 shows that as cells spread further, 12G10 staining decreased at the cell edges, indicating that localization of occupied integrins at the cell edges precedes the formation of focal adhesions. In contrast, although ICAP-1␣ or nm23-H2 co-localized with engaged ␤ 1 integrins at the edges of the cells during initial spreading, they were never detected in focal adhesions (Fig. 5).
Targeting of Both ICAP-1␣ and nm23-H2 to the Cell Periphery Depends on the Integrins Engaged with the Extracellular Matrix Substrate-Our observations presented above imply that the targeting of both ICAP-1␣ and nm23-H2 proteins is spatially and temporally linked to initial cell spreading. As ICAP-1␣ interacts specifically with ␤ 1 integrins, we hypothesized that ICAP-1␣ and nm23-H2 targeting to peripheral cell membranes during initial cell spreading should be observed on typical ␤ 1 integrin substrates and not on substrates specific for other integrins. In other terms, the composition of the extracellular matrix substrate should control the localization of both ICAP-1␣ and nm23-H2 proteins at the cell periphery. Indeed, we observed peripheral staining of both ICAP-1␣ and nm23-H2 in cells spreading on fibronectin and collagen, typical ligands of ␤ 1 integrins. However this localization was not observed when the cells were spread on poly-L-lysine, laminin 1, or vitronectin (Fig. 6). The involvement of ␣ 6 ␤ 1 integrin in early spreading on laminin was ruled out because fluorescence-activated cell sorter analysis and immunofluorescence studies showed, on one hand, a very low level of ␣ 6 subunit in Hs68 cells, and on the other hand, the absence of ␣ 6 and ␤ 1 subunits in ruffles induced by laminin (data not schown). This observation indicates that the targeting of both ICAP-1␣ and nm23-H2 to the cell periphery is dependent on an engagement of ␤ 1 integrins interacting with fibronectin or collagen and is not just a consequence of cell adhesion. DISCUSSION Recent studies suggest that individual integrin ␣/␤ heterodimers can play unique roles in the regulation of cell migration, growth, survival, and differentiation (31)(32)(33)(34)(35)(36). These regulatory functions of integrins involve specific interactions between the cytoplasmic domains of individual integrins and intracellular proteins involved in signal transduction or other aspects of cell regulation. The protein ICAP-1␣ is particularly interesting in this regard because it has been identified as a specific partner of the cytoplasmic domain of ␤ 1 integrin (9) controlling cell adhesion (11) and cell migration (37). Using two-hybrid analysis, in vitro interaction studies, and co-immunoprecipitation of expressed proteins in cells, we demonstrate that ICAP-1␣ interacts directly with nm23-H2 through its C terminus and that this novel interaction can occur under physiological conditions. As endogenous or recombinant nm23-H2 exists as in a hexameric form in solution and as this oligomerization is necessary for its function (38 -40), the form interacting with ICAP-1␣ should be hexameric.
Confocal fluorescence microscopy revealed unambiguously the subcellular co-localization of both proteins in lamellipodia and ruffles during the early stages of cell spreading. Moreover, the specificity and physiological relevance of the peripheral staining of ICAP-1␣ and nm23-H2 during the early stages of cell spreading is underlined by the fact that it was observed only when cells were plated on fibronectin and collagen, both of these matrices that engage ␤ 1 integrins. Indeed, this is consistent with the known specificity of ICAP-1␣ for ␤ 1 integrins and strongly suggests that nm23-H2 targeting to specific occupied ␤ 1 integrins at the cell periphery is mediated by ICAP-1␣. Co-localization of ICAP-1␣ and nm23-H2 at the cell edges precedes the formation of focal adhesions devoid of both proteins. Both ICAP-1␣ and nm23-H2 are recruited only into these nascent substrate adhesion sites. This points out the molecular diversity of cell-matrix adhesions, indicating dynamic changes in the morphology, molecular composition and locations of cell matrix adhesions depending on spreading time. Therefore the recruitment of ICAP-1␣ and nm23-H2 is spatially and tempo-rally linked to the formation of newly formed adhesion sites and may play a role in regulating focal adhesion assembly and/or downstream events initiated at integrin-dependent focal contacts, such as altered cytoskeletal organization or intracellular signaling. Complementing this idea, we have recently shown that ICAP-1␣ quickly disassembles focal adhesions, probably because of a competition with talin for binding to the ␤ 1 integrin tail. 3 At the leading edge of the migrating or spreading cell, ICAP-1␣ could thus prevent focal adhesion assembly, contribute to lamellipodia extension, and promote integrin functions not requiring focal adhesion formation. This hypothesis is strengthened by the observations of Reddy et al. (41), who show that conversely to ICAP-1␣ and nm23-H2, talin colocalizes with integrins in focal adhesions but is absent from cell periphery at 30 min of spreading.
In line with the concept of lamellipodia extension, this view could provide the functional significance of nm23-H2 association with ICAP-1␣. A previous report suggested that ICAP-1␣ interactions with the ␤ 1 integrin tail may support cell migration (37). Indeed, in these experiments, over-expression of ICAP-1␣ in COS-7 cells was associated with increased ␤ 1 integrin-dependent cell migration on fibronectin. Furthermore, mutations of the ICAP-1␣ binding sites localized on ␤ 1 integrin cytoplasmic tail abolished adhesion, invasion, and metastasis (42). On the other hand, numerous observations suggested that the nm23/NDPK protein family may perform more sophisticated roles in the cell physiology than the mere catalysis of a nonspecific exchange of phosphoryl groups between nucleotides (17,19,20,43). Notably, nm23-H2 has been described as a metastasis suppressor in tumor cell lines (44). For example, the S122P and H118Y mutations were identified in melanoma of high metastatic potential as tested by cell inoculation into mice or cell transfection (45)(46)(47). More recently, results obtained from Otsuki et al. (26) showed that the related family member nm23-H1 is able to associate with a Rac1-specific nucleotide exchange factor, Tiam1, involved in control of metastatic potential. These authors suggest that nm23-H1 negatively regulates Tiam1 and therefore inhibits Rac1 activation in vivo. Because nm23-H2 is able to form heterohexamers with other nm23 isoforms and because Rac is co-localized with ICAP-1␣ and controls lamellipodia extension (for review see Ridley (3)), one can speculate that the interaction of ICAP-1␣ with nm23-H2 may contribute to the overall regulation of Rac activity at the cell periphery. We can not rule out the possibility that the interaction between nm23-H2 and ICAP-1␣ might also counterbalance the interaction between ICAP-1␣ and ␤ 1 inte-grin, given the possibility of the dynamic of ruffles during cell spreading. Because the phosphorylation state of ICAP-1␣ could control cell adhesion, one can speculate that NDPK in the vicinity could somehow control phosphate donor availability.
In conclusion, the interaction between ICAP-1␣ and nm23-H2 may drastically change the understanding of the metastasis suppressor function of the nm23 protein family and will provide an alternative interpretation of the implication of these proteins in tumor invasion and metastasis. Focal adhesions form and disappear continuously during cell migration, and the cell spreading process as well as the interaction between ICAP-1␣ and nm23-H2 provide novel insight into the molecular basis of the dynamic nature of focal adhesion.
FIG. 6. Effects of matrix composition on targeting of ICAP-1␣ and nm23-H2 to the cell edges. Cells were plated on coverslips coated with 25 g/ml of collagen I (Co 1), collagen IV (Co 4), fibronectin (FN), vitronectin (VN), polylysine (PL), or laminin (LM), fixed after 30 min, and stained with polyclonal anti-ICAP-1␣ and monoclonal anti-nm23-H2. We observed a peripheral staining in Hs68 fibroblasts spread on collagen and fibronectin, which was not evident when they were spread on vitronectin, laminin, or polylysine. Visualization of a section and image capture were done with a confocal microscope. The bar represents 10 m in all cases.