A Novel Integrin-Linked Kinase–Binding Protein, Affixin, Is Involved in the Early Stage of Cell–Substrate Interaction

Embryogenesis, metamorphosis, and wound healing depend critically on the ability of cells to form specific contacts with the extracellular matrix (ECM) (Jockusch et al. 1995). In addition to guaranteeing the mechanical adhesion of cells, these cell–ECM interactions also trigger signal transduction pathways that lead to cell proliferation, differentiation, apoptosis, migration, and specific gene expressions (Hughes and Pfaff 1998). The contact sites are defined by morphologically discrete structures that mediate between the intra- and extracellular regions. Focal adhesions (FAs) represent specialized structures observed at sites of tight juxtaposition between the cell surface and ECM in many types of cultured cells. Although typical FAs are only rarely found in organisms, they have been proven to be an attractive model for the analysis of the architecture and regulation of cell–ECM contact in general. At these sites, actin filaments are bound to transmembrane glycoproteins, integrins, through a complex of structural “plaque” proteins that include vinculin, talin, and α-actinin. In addition to these components, signal transduction molecules such as FA kinase (FAK) and Src are also concentrated at FAs (Miyamoto et al. 1995). Indeed, in extremely integrated cell functions such as cell migration the integrin signals are activated at FAs by ECM stimulation and transmitted to intracellular components, producing a dynamic regulation of cytoskeletal organization. Importantly, although sufficiently rigid to provide strong adhesion, FAs are highly dynamic and can be reversibly assembled and disassembled in response to both internal and external signals. For example, the process of cell migration of cultured fibroblasts consists of (a) extending membrane protrusions called lamellipodia or filopodia to form initial cell–substrate attachments, (b) de novo formation of initial FAs at the tip of these membrane protrusions, (c) development of the mature forms of the FA complex and actin stress fibers (SFs) through the organization of several cytoskeletal proteins, and (d) disruption of FA complexes at the rear of the cell and retraction of the tail (Lauffenburger and Horwitz 1996). However, how matrix molecules, integrins, cytoskeletal molecules, and signaling molecules collaborate to promote such integrated events remains to be determined.

Integrins, the major transmembrane glycoproteins in FAs, consist of two different subunits (α and β chains) that form a heterodimer (Hynes 1992). Although their extracellular domains together form a ligand-linking site, their cytoplasmic domains are considered to provide attachment sites for cytoplasmic structural and signaling molecules. So far, to understand the molecular basis of integrin signaling, many molecules that can bind directly to the cytoplasmic tails of integrins have been successfully identified. One of these is integrin-linked kinase (ILK), which is a ubiquitously expressed serine–threonine protein kinase capable of interacting with the integrin β cytoplasmic domain (Hannigan et al. 1996). Although the first report on ILK demonstrated that its activity is regulated by cell–ECM interaction and that it plays a role in cell adhesion, the multifunctional aspects of this kinase being involved in signal transduction pathways including insulin and Wnt signaling were demonstrated in subsequent works (Hannigan et al. 1996; Delcommenne et al. 1998; Novak et al. 1998). On the other hand, recent studies have suggested the possibility that ILK is involved in the process of FA formation: the overexpression of ILK promotes the colocalization of α5β1 integrin and fibronectin with vinculin (Wu et al. 1998) and ILK colocalizes with α5β1 integrin and FAK in FAs (Li et al. 1999). However, the underlying molecular mechanism by which ILK is involved in the regulation of FA formation is still unclear.

In this study, we identified a novel calponin homology (CH) domain–containing protein, affixin, which interacts specifically with the kinase domain of ILK. Immunocytochemical analyses demonstrate that affixin colocalizes with ILK at FAs and at the tip of the leading edge. Furthermore, in cells replated on fibronectin-coated coverslips affixin shows codistribution with ILK in bleb-like initial membrane protrusions at a very early stage of cell spreading. The introduction of the COOH-terminal half of affixin, which binds ILK and is phosphorylated by ILK in vitro, inhibits the development of FAs and SFs and blocks the cell spreading process at a very early stage. Interestingly, when overexpressed in well-spread CHO cells, the COOH-terminal half of affixin also disrupts preformed FAs and SFs, but sufficient activity is observed only when ILK is coexpressed. These results suggest that affixin may be one of the downstream targets of ILK, which works at a very early stage of cell–substrate adhesion to allow the formation of FAs and SFs.

cDNAs corresponding to full-length human ILK were amplified from a human bone marrow cDNA library (CLONTECH Laboratories, Inc.) by PCR. These fragments were subcloned into pAS2-1, and the resultant ILK–pAS2-1 was used to screen human bone marrow and fetal liver cDNA libraries (CLONTECH Laboratories, Inc.) in Y190(a) yeast strain. Cotransformants of the bait and library plasmids were grown for 3–7 d at 30°C on minimum essential plates lacking histidine, tryptophan, leucine, and uracil and containing 35 mM 3-aminotriazole (Sigma-Aldrich). Positive colonies were further screened for β-galactosidase activity according to the manufacturer's instructions (CLONTECH Laboratories, Inc.). Y187(a) yeast strain was also used only for β-galactosidase assay to verify the two-hybrid interactions. Approximately 1 × 10⁶ and 4 × 10⁶ clones were screened, respectively, and seven independent clones positive for growth on plates lacking histidine and β-galactosidase activity were identified. Sequencing analysis revealed that five of these clones contained cDNA inserts encoding a 570-bp overlapping sequence. Subsequent backscreening against phage cDNA libraries from NEC and Jurkat cell lines to obtain the corresponding full cDNA resulted in the identification of several affixin cDNAs with the same ORF but containing different 5′ or 3′ noncoding regions.

To analyze the tissue distribution of mRNA expression, multiple tissue Northern blot membrane (CLONTECH Laboratories, Inc.) was probed with a ³²P-labeled human affixin cDNA probe, corresponding to amino acid residues 206–481 of the l-affixin prepared using a random-primed DNA labeling kit (Amersham Pharmacia Biotech). The hybridization was performed according to the manufacturer's instructions (CLONTECH Laboratories, Inc.), and an x-ray film was exposed at −80°C for 10 d with an intensifying screen.

Affixin and ILK deletion mutants were generated by PCR using appropriate primers. Point mutations in ILK mutants (E359K, K220M, K220A) were introduced using a QuickChange site-directed mutagenesis kit (Stratagene). The fidelities of the amplified sequences were all verified by DNA sequencing.

CHO-K1 cells were maintained at 37°C in a humidified atmosphere of 5% CO₂ in F-12 medium containing 10% FCS (Cell Culture Technologies, Inc.), 100 U/ml penicillin, and 100 μg/ml streptomycin. COS-7 cells were cultured under the same conditions as those for CHO-K1, except for the use of DME instead of F-12 medium. cDNA transfection was performed by either electroporation for the immunoprecipitation assay or lipofection using a Fugene6 transfection reagent (Roche) for immunofluorescence analysis. When performing the replating assay, CHO-K1 cells were transfected with the appropriate expression plasmids, harvested 48 h later by incubating in 0.05% trypsin in PBS containing 0.02% (wt/wt) EDTA, washed two times with PBS, and replated on fibronectin-coated coverslips.

The antibodies used in this study were anti-Flag and antivinculin monoclonal antibodies (Sigma-Aldrich), anti-FAK monoclonal antibodies (Transduction Laboratories), anti-ILK monoclonal antibody (Upstate Biotechnology), anti–α-actinin monoclonal antibody (provided by Yukiko Hayashi, National Institute of Neuroscience, NCNP, Japan), anti-T7 monoclonal antibody (Novagen), and FITC-phalloidin and rhodamine-phalloidin (Molecular Probes). Antiaffixin antibodies were generated in rabbits using glutathione S-transferase (GST)–ss-affixin as an antigen and affinity purified with the antigen before use.

For analysis of affixin expression in various rat tissues, each organ was excised from 10-wk-old Sprague Dawley rats deeply anesthetized with diethyl ether, washed with ice-cold PBS, and frozen immediately in liquid nitrogen. The tissue blocks were crushed using a Cryo-Press (Diatron) precooled in liquid nitrogen, and the resultant powder was suspended in 10 vol (vol/wt) of SDS sample buffer, homogenized with a Polytron homogenizer (Kinematica), and sonicated. 15-μg aliquots of these samples were loaded in each lane. Electrophoresis was carried out by one-dimensional SDS-PAGE (10 or 12% polyacrylamide). The separated proteins were transferred onto PVDF membranes, which were subsequently blocked with 5% skimmed milk. The membranes were treated with appropriate antibodies, and antibody reactions were visualized by a chemiluminescence ECL system (Amersham Pharmacia Biotech).

Cells cultured in 10-cm dishes were suspended in 200 μl lysis buffer containing 20 mM Hepes, pH 7.5, 150 mM NaCl, 1 mM EDTA, 10 μg/ml leupeptin, 1 mM PMSF, 1% Triton X-100, 0.1% deoxycholate, and 0.1% SDS. After a 30-min incubation on ice, the lysates were clarified by centrifugation at 14,000 rpm for 30 min. 15 μl of protein G–Sepharose (Amersham Pharmacia Biotech) conjugated with 2 μg of affinity-purified antiaffixin, anti-Flag antibodies, or control normal rabbit IgG were incubated with the cell lysates for 1 h at 4°C. After washing with lysis buffer, the immunocomplex was solubilized by adding SDS sample buffer to the resin.

CHO-K1 cells or those transfected with expression plasmids were cultured on fibronectin-coated coverslips for 48 h and, after washing with PBS, fixed with 1 or 2% formaldehyde in PBS for 15 min and then permeabilized with 0.1% Triton X-100 in PBS for 15 min at room temperature. In some experiments, cells were fixed with 100% methanol. The cells were blocked with PBS containing 10% calf serum for 1 h at room temperature and then treated with appropriate primary antibodies for 45 min at 37°C in a moist chamber. After washing with PBS containing 0.05% Tween 20, the cells were incubated with secondary antibodies (Cy3-conjugated goat anti–rabbit [Amersham Pharmacia Biotech] and Alexa488-conjugated goat anti–mouse antibodies [Molecular Probes]) at 37°C for 45 min. After washing, samples were observed under a fluorescence microscope (BX50; Olympus) equipped with a cooled CCD camera (Photometrics). Human skeletal muscle sample was obtained by biopsy for diagnostic purpose with informed consent. Thin sections of 6-μm thickness were fixed for 10 min with 100% acetone at −20°C, blocked for 15 min at 37°C with 2% BSA and 0.5% goat serum in PBS, and then processed for immunofluorescence analysis. Confocal microscopic analysis was performed using a Bio-Rad Laboratories Radiance 2000 scan head mounted on a Nikon Eclipse E600 microscope.

GST–ss-affixin and GST–RP2 fusion proteins were induced in E. coli with isopropyl β-d-thio-galactopyranoside (Amersham Pharmacia Biotech), and the proteins were purified with glutathione–Sepharose 4B beads (Amersham Pharmacia Biotech). The GST linker sites of these fusion proteins were digested with PreScission™ protease (Amersham Pharmacia Biotech) according to the manufacturer's protocol, and the excised recombinant proteins eluted from the resin were dialyzed against the appropriate buffers before use.

COS-7 cells transfected with expression vectors encoding Flag-tagged ILK or its mutants were lysed in 20 mM Hepes, pH 7.0, 150 mM NaCl, 1 mM EDTA, 10 μg/ml leupeptin, 1 mM PMSF, 1% Triton X-100, and 0.1% deoxycholate. Immunoprecipitates by anti-Flag antibody were extensively washed with lysis buffer and then kinase reaction buffer (50 mM Hepes, pH 7.0, 10 mM MnCl₂, 10 mM MgCl₂, 2 mM NaF, 1 mM Na₃VO₄) and subjected to protein kinase assays in 20 μl kinase reaction buffer containing 10 μCi [γ-³²P]ATP and an appropriate substrate (myelin basic protein [MBP] or recombinant affixin). After incubation for 60 min at 30°C, the reaction mixture was resolved by 10% SDS-PAGE, and bands were visualized by a Bio-imaging analyzer system (BAS2000; Fuji).

Two-hybrid screening against human bone marrow and fetal liver cDNA libraries using full-length human ILK as bait resulted in the identification of a cDNA sequence encoding a novel protein of 364 amino acids (predicted molecular mass is 41.71 kD) as the longest ORF (Fig. 1; described in Materials and Methods). Another candidate for initiation codon with the preceding Kozak consensus sequence (Fig. 1, arrowhead) predicts a shorter product of 350 amino acids (40.13 kD). We named these products l- and s-affixin, respectively, from the Latin word “affixa,” meaning fixtures, because they are involved in cell–substrate adhesion (see below). A similarity search revealed that there are affixin homologues in Caenorhabditis elegans and Drosophila melanogaster (Fig. 1; see below). During this analysis, a cDNA clone, CGI-56, containing a sequence identical to affixin except for the 5′ sequence, was identified as one of the human genes conserved in C. elegans (Lai et al. 2000). The remaining two positive clones obtained in the two-hybrid screening also encoded novel proteins containing sequences homologous to the ILK-binding region (see below) of affixin, suggesting that affixin comprises a novel protein family (data not shown).

Close inspection of the sequence of affixin revealed an internal sequence repeat that shows a significant sequence similarity to the actin-binding regions of several actin cross-linking proteins, such as β-spectrin, α-actinin, and dystrophin (Fig. 1 and Fig. 2 A). In these actin cross-linking proteins, the sequences are composed of 240 amino acid residues containing a tandem repeat of two CH domains, followed by an extended rod domain with a coiled coil structure (spectrin repeats; Fig. 2 B). However, the affixin sequence exhibits several unique features. First, as shown in Fig. 2 B, most of the affixin molecule corresponds to only two repeated CH domains with short flanking NH₂- and COOH-terminal sequences and lacks the rod domain. Second, the two tandem CH domains in actin cross-linking proteins, CH1 (NH₂-terminal) and CH2 (COOH-terminal), show weaker homology to each other in contrast to their intermolecular homology. On the other hand, both CH domains of affixin show higher homology to the CH1 domains of actin cross-linking proteins (Fig. 2A and Fig. B). Third, the homology between actin cross-linking proteins and affixin is restricted to the NH₂-terminal half of the CH domains. The sequence of the COOH-terminal half of affixin is highly diverged, although the COOH-terminal hydrophobic moieties that have been suggested to be important for the actin binding of these actin cross-linking proteins are partially conserved (Fig. 2 A, underline; Carugo et al. 1997). Additionally, the COOH-terminal regions in the affixin CH domains show divergence from the consensus residues that are essentially conserved in CH domain–containing proteins such as calponin (Fig. 2 A, asterisks). Interestingly, recent progress in genome projects has revealed that affixin has counterparts in other species. C. elegans T21D12.4 shows 64% overall similarity and 45% identity, whereas D. melanogaster AAF49016 shows 71% similarity and 58% identity (Fig. 1 and Fig. 2 B). Taken together, we conclude that affixin is a unique member of the CH domain–containing proteins, which is well conserved evolutionary from worms to humans.

Tissue distribution of affixin was examined by Northern and Western blot analysis. Predicted affixin transcripts (1.4 kb) were detected in heart, skeletal muscle, spleen, kidney, liver, small intestine, placenta, lung, and peripheral blood leukocyte, with the highest expression in skeletal muscle and heart (Fig. 3 A). On the other hand, four additional larger transcripts (2.8, 3.5, 5.0, and 6.5 kb) were also observed in several human tissues, especially in skeletal muscle. Affinity-purified polyclonal antibodies raised against the full-length amino acid sequence of ss-affixin (see below) specifically detected two bands (47.4 and 50.5 kD, respectively) in CHO and NIH3T3 cells, each of which shows the same migration rate as l- and s-affixin overexpressed in COS-7 cells (Fig. 3 B, arrows). It should be noted that the expression vector encoding l-affixin also induced the expression of a protein with lower molecular weight corresponding to s-affixin. This may indicate that the identified two initiation codons (Fig. 1) are simultaneously active in this construct, although the possibility of specific proteolysis of l-affixin cannot be formally excluded. Consistent with Northern blot analysis, affixin is detected ubiquitously, with higher expression in heart and skeletal muscle (Fig. 3 C). In the spleen, the antibody also detected a smaller band strongly showing the same migration rate as overexpressed ss-affixin, which is considered to be expressed using a more downstream initiation codon with the preceding Kozak consensus sequence (Fig. 1, asterisk; data not shown). In fact, we have obtained an affixin cDNA species from the T cell lymphoma–derived Jurkat cell line library with an alternatively spliced 5′ sequence in which a stop codon appears in front of the third initiation codon mentioned above. Taken together, these results suggest that ss-affixin, another affixin isoform, is highly expressed and functions in spleen.

Fig. 4 A demonstrates that affixin is coprecipitated with ILK by antiaffixin antibody from CHO cell extract, indicating the endogenous interaction of these proteins in vivo. The following experimental results shown in Fig. 4 further reveal the binding sites on each molecule in detail. Since the cDNA clones obtained by the first two-hybrid screening contained overlapping sequences corresponding to 242–364 amino acids of affixin, the ILK-binding site on affixin was considered to be around the second CH domain. The subsequent two-hybrid analysis using a series of affixin deletion mutants revealed that the amino acid sequence 258–364, corresponding to the second CH domain, is sufficient for ILK binding, whereas mutants (l-AFX 292–364) lacking the NH₂-terminal half of the second CH domain lose ILK-binding activity (Fig. 4 B). This interaction was further confirmed by coimmunoprecipitation assays performed in COS-7 cells: full-length and an NH₂-terminal–truncated mutant of l-affixin, RP2, were coimmunoprecipitated with Flag-tagged full-length ILK by anti-Flag antibody, whereas a COOH-terminal truncated mutant, RP1, was not (Fig. 4 C). These results indicate that the second but not the first CH domain is required and sufficient for the interaction with ILK.

Next, we determined the affixin-binding site on ILK using a similar approach. Two-hybrid assays revealed that the COOH-terminal kinase domain but not the preceding ankyrin repeats or pleckstrin homology domain is necessary and sufficient for the interaction with affixin (data not shown). Coimmunoprecipitation assays in COS-7 cells, the results of which are shown in Fig. 4 D, provide further insight into the requirement for the interaction. We constructed a Flag-tagged kinase-deficient mutant of ILK (K220M) by substituting the conserved lysine residue in the ATP-binding site of the kinase domain with methionine and found that the mutant coimmunoprecipitates with T7-tagged ss-affixin. Thus, we concluded that the kinase activity of ILK is not necessary for this interaction. However, another ILK mutant, E359K, which has been suggested to be kinase deficient (Novak et al. 1998; see below), showed greatly reduced interaction with affixin. This indicates that the glutamic acid residue located in the activation loop of the kinase domain of ILK is critical for the interaction.

Since the kinase domain of ILK is sufficient to associate with affixin, we next tested whether affixin can be phosphorylated by ILK in vitro. In Fig. 5 A, Flag-tagged wild-type or kinase-deficient ILK was overexpressed in COS-7 cells, and the kinase activity in the anti-Flag antibody immunoprecipitates was assayed using MBP as a substrate. The immunoprecipitates from cells expressing wild-type ILK showed enhanced kinase activity, whereas those from ILK (K220A)- and ILK(K220M)-expressing cells showed the background level of the activity. Surprisingly, the immunoprecipitate of ILK(E359K) reproducibly showed kinase activity comparable to wild-type ILK, suggesting that this mutant is not kinase deficient, although we do not know the reason for the discrepancy with the previous data (Novak et al. 1998). When we used recombinant affixin purified from E. coli as a substrate, the wild-type ILK immunoprecipitates but not the K220M immunoprecipitates phosphorylated affixin to the same extent as MBP, suggesting that affixin is a good substrate for ILK in vitro (Fig. 5 B). Consistent with the results on the narrowed binding site of the ILK–affixin interaction, the deletion mutant of affixin, RP2, was also phosphorylated by the ILK immunoprecipitates to a similar extent (Fig. 5 B). We also examined the possibility that affixin affects the kinase activity of ILK, since as shown in Fig. 4 D the activation loop of the kinase domain of ILK is involved in the interaction with affixin. However, the addition of recombinant affixin to the ILK immunoprecipitates did not affect ³²P incorporation into MBP (data not shown), suggesting that affixin can be a substrate but not a regulator of ILK kinase activity.

To investigate the physiological significance of the ILK–affixin interaction, we next examined the cellular localization of affixin and ILK in well-spread CHO cells cultured on fibronectin-coated coverslips. As shown in Fig. 6, affixin localizes at FAs identified by staining with antivinculin and antipaxillin antibody (Fig. 6A and Fig. B; data not shown). Furthermore, affixin staining is also detected at the tip of the leading edge, which is negative for vinculin staining (Fig. 6A and Fig. B, arrowheads). Double staining of affixin and ILK revealed that these proteins are colocalized at both FAs and at the tip of the leading edge. The colocalization with ILK was also observed for overexpressed RP2 but not for RP1 (data not shown), consistent with the molecular interaction between ILK and affixin detected in the yeast two-hybrid and immunoprecipitation assays described above (Fig. 4B and Fig. C). As can be seen in Fig. 6 A, paraformaldehyde fixation but not methanol fixation (Fig. 6 C) of cells also revealed filamentous staining of affixin outside dot-like FA structures, which seems to correspond to actin SFs. Immunostaining of overexpressed T7-tagged ss-affixin with anti-T7 antibody finally confirmed this observation: affixin localizes at FAs (Fig. 6E and Fig. F, arrowheads), at the leading edge (data not shown) and along actin SFs linking separate FAs (Fig. 6, E–H). Since affixin is highly expressed in muscle cells, we next examined the localization of affixin and ILK in skeletal muscle cells. Consistent with the localization of the muscle-specific integrin β1 isoform, β1D, reported to date (Vachon et al. 1997), immnohistochemical analysis of human skeletal muscle cells revealed that affixin and ILK colocalize at the sarcolemma, although affixin but not ILK staining is also detected in cytoplasm (Fig. 7, A–F). Close inspection of the sarcolemma staining of these proteins using confocal microscopy revealed their striated distributions on the membrane, to which Z-bands identified by α-actinin–staining anchor (Fig. 7, G–I, arrowheads). These results indicate that affixin and ILK are concentrated to the region on the sarcolemma where Z-bands attach, which are the corresponding structures of FAs in cultured cells.

Because the leading edge is a site where the formation and growth of de novo cell–substrate adhesions actively occur, the above results on the distribution of affixin and ILK at the tip of the leading edge suggest the possibility that their interaction plays a role in the initial phase of FA formation. This notion was supported by analyzing their localization in reseeded CHO cells that are actively spreading. When CHO cells are harvested in trypsin/EDTA solution and reseeded on fibronectin-coated coverslips, their cell–substrate adhesions are gradually restored, and their shape changes from round to spread within 4 h (Bauer et al. 1993). During the early stages of this spreading process (1 h after replating) when the cells attach to the substrate with a limited central area of round cell body, most cells transiently develop many spherical out-pouchings of the plasma membrane, blebs, which are ultimately replaced by flat ruffles or small lamellipodia. As shown in Fig. 8A–C, high concentrations of affixin and ILK are observed in these blebs during the early stages of cell spreading when the cells still show irregular dot-like staining of vinculin in their inner area (Fig. 8 D). Overexpressed affixin and ILK also show similar distribution in blebs (Fig. 8G and Fig. H). On the other hand, FAK was not observed in these peripheral membrane protrusions even when blebs have developed into more flattened lammelipodia-like structures (Fig. 8E and Fig. F). Actin filament identified by rhodamine-phalloidin staining was observed only in restricted numbers of blebs (data not shown), but at later stages when cells have started to spread, short actin bundles started to appear in the peripheral lamellae from the dot-like structures at which affixin localizes (Fig. 8I and Fig. J). These results indicate that affixin and ILK are recruited into nascent cell–substrate adhesion structures at a very early stage of cell spreading faster than FAK and vinculin from which mature FAs and SFs develop.

Next, to obtain further insight into the roles of affixin and ILK in the initial phase of cell–substrate adhesion, the effects of the overexpression of ss-affixin and its mutants, RP1 and RP2, were investigated in this replate assay of CHO cells (Fig. 9 A). 1 h after reseeding onto fibronectin-coated coverslips, >70% of the cells overexpressing each affixin construct still displayed a round morphology, although cells expressing RP1 tended to show flatter shape (Fig. 9B and Fig. C, and Fig. 1 H). On the other hand, 4 h after cell replating a drastic deleterious effect of RP2 overexpression on cell spreading was observed: in contrast to cells overexpressing ss-affixin or RP1, many of which exhibited a well-spread flat shape (round cells are <40%), >70% of RP2-expressing cells still remained in circular or oval shape with poor membrane extensions (Fig. 9B and Fig. C, and Fig. 4 H). Close inspection of the RP2-overexpressing cells revealed that they were arrested and unable to proceed from the early stage of cell spreading with peripheral blebs (Fig. 9 D). In these cells, vinculin weakly localized at inner FA-like structures, whereas filamentous actin formed a weak cortical ring with a high concentration within limited numbers of blebs (Fig. 9 D). On the other hand, cells overexpressing RP1 again looked more actively spread compared with cells expressing wild-type affixin (ss-affixin), suggesting that RP1 and RP2 have opposite effects on the establishment of cell–substrate adhesion (Fig. 9 B and Fig. 4 H).

Interestingly, if the overexpression of the affixin mutants was induced in cells that have already established well-developed cell–substrate adhesions, the effect of RP2 overexpression would be strongly restricted (Fig. 10). Although the overexpression of neither ss-affixin nor ILK affects cell morphology (data not shown), limited numbers (10%) of cells expressing RP2 but not RP1 (<3%) show an aberrant spherical morphology instead of the typical well-spread shape (Fig. 10 B, arrowheads). In these cells, FAs and SFs are disrupted, exhibiting similar vinculin and F-actin staining as observed in reseeded cells (data not shown; Fig. 9 D). Again, cells expressing RP1 tended to show an enhanced well-spread shape compared with control CHO cells. Interestingly, the coexpression of ILK with RP2 but not kinase-deficient ILK(K220M) or ILK(E359K), that is kinase active but incompetent to bind affixin (Fig. 4 and Fig. 5), increased the population of aberrant round-shaped cells (≤50%; Fig. 10E and Fig. F). This effect of ILK was not observed in cells expressing RP1 (Fig. 10 C). Therefore, these results indicate that the second CH domain of affixin requires the interaction with active ILK to exert the deleterious effects on cell–substrate adhesion, whereas it does not in cells that are actively spreading (Fig. 9 B compared with Fig. 10B and Fig. D).

Here, we report a novel CH domain–containing protein named affixin as a binding protein for ILK, a serine–threonine protein kinase that interacts with the cytoplasmic region of β1 integrin. Affixin and ILK associate in vivo and colocalize at FAs in well-spread cultured cells, suggesting that affixin is a novel component of FAs that may be involved in integrin–ILK signaling at this specific structure. Immunofluorescence analysis of cells in the spreading process after replating on fibronectin indicates that endogenous and exogenously expressed affixin and ILK accumulate in cell surface blebs observed during the very early phase of the spreading process. These blebs have been shown to be spherical out-pouchings of the plasma membrane that are commonly observed at the periphery of eukaryotic cells as they spread on a substrate or at the leading edge of moving cells (Bereiter-Hahn et al. 1990). They are considered to be transient structures that are driven by hydrodynamic forces and finally result in other types of protrusions such as membrane ruffles on lamellipodia (Cunningham 1995). They also correspond to the membrane ruffles where endogenous β-integrin and another ILK-binding protein, PINCH, have been shown to concentrate during cell spreading (Tu et al. 1999). These results suggest that integrin–ILK signaling plays a role in the early phase of cell spreading, and affixin and PINCH also participate in this process. Importantly, vinculin and FAK do not accumulate in these regions at this stage (Fig. 8D and Fig. F), suggesting that ILK and affixin are recruited into these nascent substrate adhesion sites in advance of these FA components. We further demonstrate that during the later stage of cell spreading, affixin shows punctuate staining at lammelipodia from which short actin bundles emanate (Fig. 8I and Fig. J) and finally colocalizes with vinculin and FAK at FAs (Fig. 6A and Fig. E; data not shown). In well-spread cells, ILK and affixin are also detected at the tip of the leading edge of lammelipodia where formation and growth of de novo cell–substrate adhesions actively occur. These results strongly indicate that affixin is involved in integrin–ILK signaling required for the nascent cell–substrate adhesion structures, which will develop into mature FAs.

FAs form and disappear continuously during cell locomotion and the cell spreading process, and the underlying molecular basis of this dynamic nature of FAs has been extensively investigated. FAK, one of the major protein kinases accumulating at FAs, was thought to function in the initial step of FA formation, partially because many FA components are tyrosine phosphorylated during the FA formation. However, recent results revealed that FAs can be formed even in the absence or inhibition of FAK activity, implying the presence of more essential signaling molecules regulating the formation of the FA complex (Ilic et al. 1995). The recent finding of ILK, which is activated within 30–45 min after plating on fibronectin (Delcommenne et al. 1998) and is localized to FAs in CHO cells (Li et al. 1999), suggests the new possibility that this serine–threonine kinase is involved in FA formation. In this context, the present results are important in not only supporting these notions about the physiological function of ILK signaling but also suggesting the involvement of affixin in this ILK function.

Yeast two-hybrid assays and immunoprecipitation assays in COS-7 cells revealed that the kinase domain of ILK interacts specifically with the second CH domain of affixin. The glutamic acid residue located in the activation loop of the ILK kinase domain is further demonstrated to be critical for the interaction. Consistently, in vitro kinase assay showed that the second CH domain can be a substrate for ILK, suggesting the possibility that affixin is a novel in vivo substrate for ILK, which links integrin–ILK signaling to the initial FA formation. This was further supported by showing the deleterious effects of an affixin deletion mutant, RP2, corresponding to the second CH domain, on FA formation: when CHO cells overexpressing RP2 were reseeded on fibronectin, >75% of the cells retained their round form and could not develop FAs and SFs. Closer inspection further suggested that the cells were completely blocked at the initial phase of the spreading process with many surface blebs. Interestingly, this effect of RP2 on FA formation was not significantly observed if it was expressed in cells which have already established FAs. Only a limited number (<10%) of cells expressing RP2 alone showed a round shape, and the number was dramatically increased to levels comparable to those of spreading cells when wild-type ILK was cotransfected. Because this enhancement was not observed by cotransfection of kinase-deficient ILK K220M or an affixin-binding–deficient ILK mutant, ILK(E359K), these results indicate that the effect of RP2 is dependent on the interaction with active ILK, probably on the phosphorylation by ILK. The apparent discrepancy of the effects of RP2 on well-spread cells and on actively spreading cells can be reasonably explained by assuming that integrin–ILK signaling evoked by initial cell–substrate interaction (Delcommenne et al. 1998) mimics the effect of ILK overexpressing in actively spreading cells. Although further studies will be needed, the present result about the positive effect of RP1, the first CH domain, on cell spreading may suggest that this domain is also involved in the interaction with the downstream target for FA formation in a different way from RP2.

Amino acid sequence analysis of affixin revealed it to represent a novel member of the CH domain–containing protein family that is conserved from worms to mammals. Considering that most CH domain–containing proteins have been shown to be related to the actin cytoskeleton (Stradal et al. 1998), the putative molecular target of affixin may be the actin cytoskeleton. This notion is further supported by the fact that both CH domains in affixin exhibit the closest homology to those of the actin-binding regions of actin cross-linking proteins such as β-spectrin and α-actinin. Hence, it is quite possible that affixin also binds directly to F-actin. In fact, we observed that although weakly, endogenous affixin localizes to SFs in cultured cells (Fig. 6A, Fig. E, and Fig. G). However, despite extensive efforts, we have not succeeded in detecting the direct interactions of affixin or RP2 with F-actin and cannot yet discuss the molecular mechanism by which affixin plays critical roles in FA development in detail. It may be due to the unique divergence of affixin from the sequence of actin-linking proteins. Of course, it is possible that the phosphorylation of affixin by ILK is required for its interaction with F-actin. We are now conducting further experiments to address these issues as one of the major extensions of the present study.

Northern and Western blot analyses showed that affixin is highly expressed in skeletal muscle and heart, suggesting its role in cell–substrate interaction in muscle cells. Interestingly, the mutation/depletion of the C. elegans homologue of β-integrin (pat-3) or ILK-binding protein, PINCH (unc-97), from the embryo was reported to result in a similar phenotype called “pat” showing paralysis and elongation arrest at the twofold stage due to defects in the integrity of myofibril structures in the body wall muscle (Williams and Waterston 1994; Gettner et al. 1995; Dedhar et al. 1999; Hobert et al. 1999). Since the body wall muscles of these mutants show disorganized dense bodies (structural analogues of FAs in cultured cells) to which β-integrin/PAT-3, vinculin/DEB-1, and PINCH/UNC-97 are localized, it has been suggested that defects in the development of dense bodies are the primary cause of the phenotype (Gettner et al. 1995; Hobert et al. 1999). Interestingly, we recently observed that the C. elegans embryo from which the expression of ILK or affixin homologues (CAB77052 or T21D12.4, respectively) was deleted by the RNA interference method shows pat phenotype characterized by an arrest during its development at twofold stage and paralysis (Sugiyama, Y., unpublished results). Although we have not confirmed that these are primarily caused by defects in muscle attachments, our present results showing ILK and affixin play important roles in FA development, and it might be reasonable to speculate that these embryo also have defects in dense body formation. Consistently, immunohistochemical analysis of human skeletal muscle demonstrated that affixin and ILK colocalize to the sarcolemma showing a striated pattern matching cytoplasmic Z-band striation, suggesting that these proteins accumulate at sites where the Z-band attaches to the sarcolemma, which correspond to dense bodies in C. elegans body wall muscle cells. Taken together, these results imply the possibility that affixin and ILK are also important for muscle development and function.

During the revision of this manuscript, a novel CH domain protein similar to affixin, actopaxin, was reported as a paxillin-binding protein. One of the clones we obtained in the two-hybrid screening using ILK as bait, FL29, was identical to actopaxin, indicating that this protein also binds ILK. Affixin and actopaxin are products of distinct genes with different expression patterns but share common features such as localization to focal contact.

We thank Y. Hayashi (National Institute of Neuroscience, NCNP, Japan) for helping us perform immunohistochemical analysis of human skeletal muscle cells.

This work was supported by grants from the Yokohama Foundation for Advancement of Medical Science (to S. Yamaji).