Disruption of Focal Adhesions by Integrin Cytoplasmic Domain-associated Protein-1α*

EXPERIMENTAL PROCEDURES

Antibodies

The anti-β₁ tail serum (anti cyto-β₁) was raised against a synthetic peptide corresponding to the cytoplasmic domain of the β₁ chain covalently coupled to keyhole limpet hemocyanin. Anti-talin monoclonal antibody (8d4) was purchased from Sigma (St. Louis, MO). The monoclonal antibody 9EG7 directed against the β₁ subunit was kindly supplied by Dr. D. Vestweber (Muenster, Germany). The monoclonal antibody 7E2 directed against the hamster β₁ subunit was a generous gift of Dr. R. Juliano (Chapel Hill, NC). Polyclonal antibody directed against the human ICAP-1α protein was previously described (13). Cyanin3-, Alexa-, or rhodamine-conjugated goat anti-mouse or anti-rabbit from Molecular Probes (Eugene, OR) or Immunotech (Marseille, France) were used as secondary antibodies.

Cells and Cell Culture

The murine NIH3T3, the hamster CHO, and the human HeLa cell lines were grown in α-minimal essential medium supplemented with 10% fetal calf serum, 100 units/ml penicillin, and 100 μg/ml streptomycin. The murine GD25, GD25-β_1A, and GD25-β_1D were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 100 units/ml penicillin, and 100 μg/ml streptomycin. GD25 cells do not express the β₁ integrin chain due to a null mutation in both alleles (29). GD25 cells transfected with either the murine β_1A, or the human β_1A and β_1D full-length cDNA are called GD25-β_1A and GD-β_1D, respectively, and have been described earlier (30, 31). All transfected cells were grown in complete medium supplemented with the appropriate antibiotics for the selection of the transfected cells.

Protein Purifications

ICAP-1α and ICAP-1α fragments fused to a polyhistidine tag at the N-terminal position were purified from the BL21(DE3) Escherichia coli strain containing the vector pET19b-ICAP-1α. Briefly, human ICAP-1α cDNA cloned in pBluescript was used as a template in a PCR reaction using primers with an XhoI site in the 5′ position. In the sense primer theXhoI site is in-frame with the first methionine of ICAP-1α. Then the XhoI-digested PCR product was cloned into the XhoI site of pET-19b vector (Novagen). Fragments were obtained by insertion of stop codons at different positions using the QuikChange mutagenesis kit (Stratagene). All constructs used in this study have been sequenced by the Eurogentec direct sequencing department (Belgium). Purification was carried out using the nickel-charged resin nickel-nitrilotriacetic acid from Qiagen. Inclusion bodies were solubilized in urea. Protein refolding was performed directly on the column by progressive removal of the chaotropic agent. The purity of the protein was checked by SDS-PAGE and Coomassie Blue staining and was greater than 90–95%. All experiments were carried out with freshly purified proteins. Before each experiment, the capacity of each batch of the purified protein to interact with the β₁ cytoplasmic domain was estimated in a solid-phase assay.

The polypeptide corresponding to the β₁integrin cytoplasmic domain was produced from the BLR(DE3)pLysSE. coli strain containing the vector pET19b-cytoβ1. This construct allows the production of the fragment 752–798 of the β₁ integrin cytoplasmic domain. This peptide was recognized by a polyclonal antibody raised against a synthetic β₁ cytoplasmic peptide coupled to keyhole limpet hemocyanin. Talin and α-actinin were purified as previously described (32), and fibronectin was purified according to a previous study (33).

Transfection in Mammalian Cells and Selection of Stable Clones

Full-length human ICAP-1α was excised from the pBS-ICAP-1α vector as an EcoRI/XbaI fragment and inserted into the pcDNA3.1(+) vector (Invitrogen, The Netherlands). Stable GD25-β_1A cell lines expressing ICAP-1α were obtained by electroporation of 4 × 10⁶cells in 400 μl of PBS at 280 V with 15 μg of pcDNA3.1(+)-ICAP-1α vector. Transfected cells were selected in complete medium with Zeocin (Invitrogen, The Netherlands) at a final concentration of 300 μg/ml. The expression of ICAP-1α was monitored by indirect immunofluorescence and Western blot analysis using the ICAP-1α polyclonal antibodies.

Immunofluorescence Microscopy

Immunofluorescence was carried out using standard procedures. Stained cells were analyzed with an inverted fluorescence microscope (Olympus Provis AX70) equipped with a Plan Apo ×63 oil immersion, numerical aperture 1.40 objective lens. For all double-staining experiments, the appropriate controls were performed to ensure that no undesired cross-reactivity occurred between the primary and secondary antibodies.

Purification of Ventral Plasma Membranes

The purification of HeLa, GD25-β₁, or NIH3T3 ventral plasma membranes was performed as previously described by Cattelino et al. (34). The cells were grown overnight on fibronectin-coated coverslips in complete medium. After two washes in PBS, the cells were incubated with cold water for 2 min and then flushed with a 1000-μl tip. Cell disruption was confirmed by microscopy. Ventral plasma membranes were either immediately fixed with paraformaldehyde or were preincubated for 30 min at 4 °C with ICAP-1α or ICAP-1α fragments at the concentration of 5 μm in a VPM buffer containing 125 mm potassium acetate, 2.5 mm MgCl₂, 12 mm glucose, and 25 mm HEPES, pH 7.5, prior to fixation.

Solid-phase Assays

The interaction between ICAP-1α and the cyto-β₁ peptide or the whole α₅β₁ integrin was carried out using a solid-phase assay. Briefly, a 96-well tray (MaxiSorp, Nunc) was coated with the whole ICAP-1α protein or ICAP-1α fragments for 16 h at 4 °C and blocked with a 3% BSA/PBS solution for 1 h at room temperature. A Triton X-100 CHO cell lysate made in PBS supplemented with 1% Triton X-100 (w:v) or the cyto-β₁ peptide were incubated for 1 h at 37 °C. After three washes in PBS containing 3% BSA and 0.01% Tween-20, detection of the α₅β₁ integrin from the CHO cell lysate was performed using the 7E2 monoclonal antibody, whereas the detection of the cyto-β₁ peptide was achieved with a polyclonal antibody directed against a synthetic peptide corresponding to the β₁ tail.

Microinjection into NIH3T3 Cells

NIH3T3 cells were seeded onto fibronectin-coated glass coverslips overnight at 37 °C. All injections were carried out with the aid of a micromanipulator 5171 connected to an Eppendorf microinjector unit (Transjector 5246). The cells were microinjected with PBS containing a final concentration of 1 mg/ml of the freshly purified recombinant ICAP-1α protein, or the N-terminal (1–100) or C-terminal (101–200) fragments, in the presence of 100 μm tetramethylrhodamine-dextran amine (M _r 3000, Molecular Probes, Interchim, France) to view the injected cells. Three hours (whole ICAP-1α protein) or 30 min (ICAP-1α fragments) after microinjection, the cells were fixed with 3% paraformaldehyde and 2% sucrose in PBS for 10 min at 37 °C and then immunostained for vinculin localization.

RESULTS

ICAP-1α Does Not Localize in Focal Adhesions

The protein ICAP-1α was isolated as a β_1A-interacting protein in a yeast two-hybrid screen (23) and was shown to modulate CHO cell adhesion (13) and to promote cell motility (26). In epithelial cells or in cell lines derived from epithelial cells such as HeLa, ICAP-1α could be detected in a cell lysate by Western blot using a polyclonal antibody raised against the full-length recombinant protein (Fig.1 A). The endogenous human ICAP-1α protein in HeLa cells migrates in SDS-PAGE like the ectopically expressed protein in rodent fibroblast-like GD25 cells (Fig. 1 B). GD25, CHO, and NIH3T3 cells showed no detectable ICAP-1α expression as monitored by Western blot analysis. To determine the physiological relevance of the interaction between ICAP-1α and the β₁ integrin, we carried out immunomicroscopy experiments of ICAP-1α in different cell lines. In HeLa cells, ICAP-1α showed a diffuse expression pattern and often some nuclear localization (Fig. 1 C, panel a). Surprisingly, no accumulation of ICAP-1α was observed in focal adhesions visualized by vinculin staining (Fig. 1 C,panels a–c). Similarly, we reported previously that in the Hs68 cell line, ICAP-1α and β₁ colocalize in ruffles but not in focal adhesions (35). A direct competition of endogenous ICAP-1α with the purified recombinant protein revealed a dramatic decrease in ICAP-1α immunostaining and confirmed the specificity of the immunolabeling (Fig. 1 C, panels d–f). Despite the diffuse ICAP-1α localization, these cells were able to form well-organized focal adhesions connected to stress fibers as judged by double labeling using a monoclonal antibody directed against vinculin and phalloidin-rhodamine-stained stress fibers (not shown).

To have direct access to focal adhesion proteins, ventral plasma membranes were obtained from HeLa cells grown overnight on fibronectin. Double immunostaining was performed with an anti-vinculin antibody and anti-ICAP-1α polyclonal antibodies. In these membrane preparations, focal adhesions could be viewed by vinculin staining (Fig.1 D, panel b) or by talin or β₁staining (not shown), whereas anti-ICAP-1α antibodies showed a faint background staining that was barely detectable (Fig. 1 D,panel a). Altogether these results suggest that ICAP-1α is not present in focal adhesions.

Interaction of ICAP-1α with the α₅β₁Integrin

The absence of ICAP-1α in focal adhesions prompted us to study the interaction of ICAP-1α with β₁ integrins in more detail. ICAP-1α and the β₁ cytoplasmic domains were expressed in bacteria as polyhistidine fusion proteins. Fig.2 A shows that the purified ICAP-1α protein interacted specifically with the purified β₁ cytoplasmic domain in a solid-phase assay, which is consistent with previous reports (23, 26). As a control, we used a β₁ cytoplasmic domain bearing the point mutation Y to S in the NPXY membrane distal (cyto3) domain. In full agreement with a previous report (23), this mutation abolished the interaction between ICAP-1α and the β₁ cytoplasmic tail (Fig. 2 A).

Next, we tested whether ICAP-1α was able to interact with the whole α₅β₁ integrin from a CHO cell lysate. This was crucial, because beta subunits do not exist in isolation in cells, and therefore, two hybrid experiments with integrins may be prone to artifacts. Increasing amounts of the recombinant ICAP-1α protein were used to coat 96-well trays. The protein concentration during coating was maintained constant by adding BSA. An equal amount of a CHO cell lysate in Triton X-100 was subsequently incubated in each coated well. A dose-dependent and -specific binding of the β₁ integrin was detected by a specific antibody (Fig.2 B). These data indicate that ICAP-1α expressed in bacteria is able to interact with the β₁Α cytoplasmic domain, and that the cytoplasmic domain of the α subunit did not impair the interaction with ICAP-1α.

Finally, we expressed ICAP-1α fragments in bacteria and used them in a solid-phase binding assay to map the β₁ binding site. Only the C-terminal moiety (amino acids 100–200) of the protein was able to bind to the β₁ integrin (Fig. 2 C). But neither the fragment corresponding to amino acids 1–150 nor the fragment corresponding to amino acids 151–200 of ICAP-1α were found to interact strongly with the α₅β₁ integrin from cell lysate (Fig. 2 C).

ICAP-1α Disorganizes Focal Adhesions ex Vivo

Despite its specific and direct association with the β₁ integrin, ICAP-1α was not localized in focal adhesions. One possible explanation for these contradictory results could be that ICAP-1α might act as a negative regulator of the recruitment of focal adhesion components. To investigate this possibility we microinjected ICAP-1α recombinant protein into the cytoplasm of NIH3T3 cells and monitored focal adhesion organization by staining for vinculin. Although microinjection of dextran-coupled rhodamine alone had no significant effect on the localization of vinculin (Fig.3, A–C), talin, and α-actinin (not shown), microinjection of the full-length ICAP-1α in the dextran-coupled rhodamine buffer induced a rapid delocalization of vinculin (Fig. 3, D–F) or talin and α-actinin (not shown) observed in 70% of the cells. Microinjection of the C-terminal moiety of ICAP-1α (amino acids 101–200) that encompasses the β₁ binding site had similar effects (Fig. 3,J–L) in 77% of the injected cells. Because the N-terminal fragment (amino acids 1–100) does not bind the β₁integrin domain (Fig. 2 C), we made use of this recombinant fragment as a control. Indeed, the microinjection of this part of ICAP-1α did not interfere with vinculin staining (Fig. 3,G–I).

Finally, disruption of focal adhesions by ICAP-1α was also investigated in a cellular context after stable transfection into GD25-β_1A cells of a vector containing human ICAP-1α cDNA. This cell line expresses functional β₁integrins at the cell surface (19) that can be monitored by the 9EG7 monoclonal antibody, which recognizes a ligand-induced binding site epitope correlating with the occupied conformational state of β₁ integrins (36, 37). Under our experimental conditions, immunofluorescence microscopy did not reveal any detectable staining for endogenous ICAP-1α in GD25-β_1A cells (Fig.4 A). On the other hand, these cells exhibited surface expression of β_1A integrins confined to focal adhesions that could be monitored by the 9EG7 antibody (Fig. 4 B). In a non-clonal population of GD25-β_1A cells transfected with a cDNA encoding the human ICAP-1α, a positive immunofluorescence signal for ICAP-1α was diffusely present within the cytoplasm (Fig. 4 C). Simultaneously, a diminution of cell spreading and loss of 9EG7 monoclonal antibody staining was observed, suggesting that β₁ integrins were no longer occupied and involved in focal adhesions (Fig. 4 D).

Disruption of Focal Adhesions by ICAP-1α Requires Direct Interaction with the β₁ Integrin Chain

The action of ICAP-1α on focal adhesions might be indirect, for instance due to the interference with some regulatory pathways. Therefore, the purified recombinant ICAP-1α was also tested for its ability to disassemble focal adhesions in vitro in a cytosol-free ventral plasma membrane preparation (VPM). These preparations are depleted in nucleotide triphosphate and soluble signaling enzymes. The cell membranes were incubated for 30 min at 4 °C with a solution of purified ICAP-1α in acetate buffer and glucose. Although buffer alone did not interfere with the detection of focal adhesion proteins such as vinculin (Fig. 5 A), the incubation with ICAP-1α efficiently displaced vinculin from focal adhesions (Fig. 5 B). A similar result was also observed for talin and α-actinin (not shown). The same result was obtained by the incubation of the C-terminal part (amino acids 101–200) of ICAP-1α (Fig. 5 D). Finally, incubation of these ventral membranes with the N-terminal purified fragment (amino acids 1–100) had no effect on focal adhesion organization (Fig. 5 C).

ICAP-1α was suggested to have a GDP dissociating inhibitor activity for Rac and Cdc42 (38), two monomeric G proteins of the Rho family involved in the regulation of cytoskeleton organization. This activity might account for ICAP-1α destabilizing action on focal adhesions of ventral plasma membranes. To assess whether ICAP-1α action on focal adhesions was due to its direct binding on β₁ integrin chains or to some interference with Rho signaling pathways, we performed similar experiment on VPM from GD-25β_1A and GD-25β_1D cells lines. The β_1D and β_1A isoforms are functionally similar with regard to integrin-mediated signaling (39), but the former strongly binds talin (31) and does not bind ICAP-1α (38). Upon addition of the ICAP-1α fragment 100–200, the dispersion of β_1A integrins initially clustered into focal adhesions was observed (Fig. 6, A andC), whereas β_1D-containing focal adhesions remained unaffected (Fig. 6, B and D). This result strongly suggests that a direct interaction between ICAP-1α and the β₁ chain is a prerequisite for focal adhesion disassembly.

Talin and ICAP-1α Compete for Binding on the Cytosolic Domain of the β₁ Integrin Chain

Because talin interacts directly with the β₁ integrin cytoplasmic domain and is crucial for focal adhesion assembly, one attractive hypothesis is that ICAP-1α is involved in the control of talin-integrin interaction. Therefore, we tested whether ICAP-1α could modulate the binding of talin to the integrin β₁ cytoplasmic domain. In anin vitro solid-phase assay, ICAP-1α could inhibit talin binding to the cytoplasmic tail of the β_1A chain in a dose-dependent manner (Fig.7 A). These data suggest that the displacement of talin from its binding site on β_1Amay be sufficient for focal adhesion disruption and, consequently, for a decrease in the integrin avidity. Moreover, the competition of ICAP-1α and talin for the binding to β₁ was specific, because it could not be observed either with α-actinin, another β₁ interacting protein (Fig. 7 B), or with the 1–100 ICAP-1α moiety (Fig. 7 C).

DISCUSSION

We examined the cellular localization of the endogenous ICAP-1α protein. Surprisingly, this protein was never detected in focal adhesions, but instead, exhibited a diffuse pattern within the cell, although a significant amount of the protein was associated within the Triton X-100-insoluble fraction (not shown) and often, a nuclear staining was observed. Using purified ventral membrane preparation from HeLa cells, we never observed ICAP-1α colocalized with vinculin or talin, which were used as markers of focal adhesions.

Even though ICAP-1α was not detected in focal adhesions, the purified recombinant protein interacted strongly with the cytoplasmic domain of the β_1A integrin chain as reported previously (23, 26). Additionally, this interaction also occurred with the whole integrin receptors purified from a cell lysate. The strong binding of ICAP-1α to the cytoplasmic domain of the β₁ integrin and its complete absence from focal adhesions suggested that this interaction may disrupt focal adhesion structures. To confirm this hypothesis we microinjected ICAP-1α in NIH3T3 cells, and we indeed observed a rapid disorganization of focal adhesions. In addition, recombinant ICAP-1α was able to disaggregate focal adhesions when added to purified ventral plasma membranes from NIH3T3 and GD-25β_1A cells. Conversely, the β_1D-containing integrins were resistant to ICAP-1α. This latter experiment strongly suggests that the disassembly of focal adhesions is due to a direct interaction with the β_1A integrin subunit and is independent of a cellular signaling pathway. Furthermore, the focal adhesion disruption mediated by ICAP-1α is in good correlation with our previous data, which have shown that ectopic expression of ICAP-1α-regulated CHO cell spreading (13).

Several reports have shown that talin is crucial for the formation of focal adhesions (27, 28, 40). A simple explanation for the negative effect of ICAP-1α on focal adhesion structure could be its ability to disrupt the direct association between the integrin and talin. To investigate this hypothesis we performed an in vitro assay and found that talin and ICAP-1α compete for binding to the β_1A cytoplasmic domain. On the other hand, we found that the interaction between α-actinin and the β₁ integrin is not inhibited by the presence of ICAP-1α. This shows that ICAP-1α inhibits the interaction between β_1A integrins and talin in a specific manner and confirms previous reports showing that the interaction of α-actinin with the β₁cytoplasmic domain is not sufficient to stabilize focal adhesion sites (40). The lack of effect of ICAP-1α on β_1D localization suggests that, under our experimental conditions, this action is direct and not dependent on the GDP dissociating inhibitor activity recently suggested (38). Based on these findings we propose that ICAP-1α and talin compete for integrin β_1A binding and thereby modulate focal adhesion assembly and/or dynamic. How ICAP-1α interferes with talin binding on the β₁ integrin needs further investigation. The talin binding site is not unambiguously defined. Recent reports have demonstrated that the talin N-terminal head binds to the β₃, β_1A, and β_1D cytoplasmic domains (41, 42). Some data indicated that the binding site of the talin head could be located on the proximal membrane region of the integrin β chain (41). Conversely, other reports indicate that a phosphotyrosine binding-like subdomain of the FERM domain of talin head is the major binding site that triggers the activation of the α_IIbβ₃ integrin (43). This finding is very interesting, because it offers some molecular basis of ICAP-1α and talin competition. Indeed, sequence homology and molecular modeling favor the view that ICAP-1α is a phosphotyrosine binding domain protein. It was suggested that the interaction specificity with the β_1A cytosolic tail was due to the interaction of Val-787 on the integrin and an hydrophobic pocket created by Leu-82 and Tyr-144 of ICAP-1α (25). This is fairly consistent with the lack of interaction of ICAP-1α with the β_1D isoform that do not have a valine at this position. This latter residue is very close to the tyrosine 783 on the human β_1A chain. The tyrosine at this position on the β₁ chain or on the homologous position 747 on the β₃ chain seems to be crucial for integrin conformational switch and talin head binding. Moreover, talin C-terminal rod domain contains another binding site located within the residues 1984–2541 (44). Because the talin-active form is an anti-parallel homodimer (32, 45), the head and tail integrin binding sites in the adjacent talin molecules would be in close proximity with each other. Therefore, it is likely that talin and ICAP-1α binding sites on the integrin β_1A tail overlap.

The distribution of ICAP-1α in ruffles and its absence from focal adhesions suggest that the interaction between ICAP-1α and the β₁ integrin cytoplasmic domain is regulated. It is possible that ICAP-1α is sequestered inside the cell and that the interaction between a sequestering protein and ICAP-1α may be the regulated event. Alternatively, the interaction of ICAP-1α with the cytoplasmic domain of the β₁ integrin may be modulated by post-translational modifications (like phosphorylation). Indeed we have previously shown that a point mutation into the CaMKII putative phosphorylation site dramatically affected cell spreading (13). Moreover, pull-down assays showed that only a small fraction of ICAP-1α was able to interact with β_1A (26). How the interaction of ICAP-1α and the integrin is regulated is not yet understood and requires further investigations.

Recently, a 20-kDa protein named TAP-20 (with marked homology with β₃-endonexin) was shown to interact specifically with the β₅ cytosolic domain of the α_vβ₅ integrin (46). Overexpression of this protein leads to decreased adhesion and focal adhesion formation, and enhances migration. These properties are quite reminiscent of those of ICAP-1α, suggesting that a family of negative regulators may control specific integrin classes in a similar fashion.