Cell Adhesion Kinase β Forms a Complex with a New Member, Hic-5, of Proteins Localized at Focal Adhesions*

Cell adhesion kinase β (CAKβ/PYK2) is the second protein-tyrosine kinase of the focal adhesion kinase subfamily. We identified a cDNA that encodes a CAKβ-binding protein. This cDNA clone encodes the human homologue of Hic-5, the cDNA of which was cloned in 1994 as transforming growth factor β1- and hydrogen peroxide-inducible mRNA. We found that Hic-5 exclusively localized at focal adhesions in a rat fibroblast line, WFB. This localization of Hic-5 was confirmed in WFB cells expressing Myc-tagged Hic-5. The amino acid sequence of Hic-5 is highly similar to that of paxillin in the four LD motifs as well as in the four contiguous LIM domains. The Hic-5 N-terminal domain directly associated in vitrowith the extreme C-terminal region (residue 801 to the end) of CAKβ. CAKβ was coimmunoprecipitated with Hic-5 from the WFB cell lysate. The coimmunoprecipitation of CAKβ with Hic-5 was markedly inhibited by the addition of the extreme C-terminal region of CAKβ. Coimmunoprecipitation of Hic-5 with CAKβ, which was shown in COS-7 cells doubly transfected with cDNA constructs of CAKβ and Myc-tagged Hic-5, was lost when the CAKβ amino acid residues 741–903 were deleted. Hic-5 was tyrosine-phosphorylated in Src-transformed 3Y1 cells and in cells treated with pervanadate. Hic-5 associated with CAKβ was selectively tyrosine-phosphorylated in WFB cells exposed to hypertonic osmotic stress. These results indicate that Hic-5 is a paxillin-related component of focal adhesions and binds to CAKβ, implying possible involvement of Hic-5 in the downstream signaling of CAKβ.

CAK␤ and FAK are closely related in their overall structures and have sequence similarity over their entire length except for the extreme N-terminal regions and 10 C-terminal residues. The 88 N-terminal residues of CAK␤ are markedly different from the corresponding 81 N-terminal residues of FAK (1). FAK is important as a docking protein. Four regions of the FAK sequence have been identified as the ligand sequences (8). All these ligand sequences are at least partly conserved in CAK␤. Tyrosine residue 397 of FAK and the corresponding residue 402 of CAK␤ are sites of autophosphorylation and also ligand sites to the SH2 domains of the Src family protein-tyrosine kinases with conserved ligand sequence, YAEI (9); this binding activates the Src family kinases (9,10). The second ligand sequence in FAK for SH2, Y 925 ENV of the mouse FAK, is known to be the ligand site for Grb2 (11) and is also functionally conserved in residues 881-884 of CAK␤, YHNV of rat CAK␤, and YLNV of human CAK␤. The third ligand sequence in FAK, EAPPKPSR, participates in the binding to the SH3 domains of pp130 cas and related proteins (12,13) and is functionally conserved in CAK␤ residues 712-719, EPPPKPSR. A weak coimmunoprecipitation of pp130 cas with CAK␤ was shown (14). There is one more proline-rich sequence in the C-terminal domain (C-domain) of FAK, PAAPPKKPPRPGAP (residues 869 -882). An SH3-containing GTPase-activating protein for Rho and Cdc42, named Graf by Hildebrand et al. (15), was identified as a protein with specific affinity to this sequence. CAK␤ also has a proline-rich sequence at the corresponding region, PPQKPPR (residues 855-861).
CAK␤ in cultured epithelial cells is mainly found in the perinuclear region and the cytoplasm in addition to the cell-tocell border. In rat tissues, CAK␤ is present in association with microvilli, cilia, and axons (16). FAK mediates signaling through integrins. A complex assembly of proteins is formed at focal adhesions in association with FAK (8,17). It has been shown that paxillin and talin bind to the C-domain of FAK (18,19). Two short stretches of 17 and 9 amino acid residues were identified as the FAK sequences participating in the binding to paxillin (18). These sequences are highly conserved in CAK␤. Paxillin binding to CAK␤ was recently shown (20). FAK is also activated by stimulation of receptors coupled to phospholipase C activation such as neuropeptide receptors and the plateletderived growth factor receptor (21). This second mode of activation is also found in CAK␤ (2). Moreover, the tyrosine phosphorylation of CAK␤ is markedly enhanced when the cytoplasmic free Ca 2ϩ concentration is increased (2) and when cells are stressed by osmotic shock (5,22). The differences in the subcellular and tissue distributions of CAK␤ and FAK indicate different functions for these two protein-tyrosine kinases.
In a study to elucidate the upstream and downstream signaling pathways of CAK␤, we used an expression cloning technique to identify binding partners for the C-domain of CAK␤. We report here the identification of a cDNA that encodes a CAK␤-binding protein. The predicted amino acid sequence of this cDNA indicated that the protein was the human homologue of Hic-5, the cDNA of which was described by Shibanuma et al. (23) as a transforming growth factor ␤1and hydrogen peroxide-inducible mRNA. Hic-5 is closely related to paxillin in its amino acid sequence. Immunocytochemical staining of rat fibroblast line WFB with a specific anti-Hic-5 antibody revealed that Hic-5 localized exclusively at focal adhesions as paxillin did, a result in disagreement with the original identification of Hic-5 as a nuclear protein (23,24). We further demonstrated coimmunoprecipitation as well as direct binding of Hic-5 and CAK␤. Hic-5 associated with CAK␤ was preferentially tyrosine-phosphorylated, implying a functional interplay between CAK␤ and Hic-5.

MATERIALS AND METHODS
Cloning of a cDNA Encoding a CAK␤-binding Protein-An oligo(dT)and random-primed human (normal female, 2 years old) hippocampus cDNA library constructed in ZAPII vector (Stratagene, La Jolla, CA; catalog number 936205) was screened by affinity binding to a glutathione S-transferase (GST)-CAK␤ fusion protein, GST-CAK␤-Cdom, that had been labeled with 32 P. Preparation of 32 P-labeled GST-CAK␤-Cdom was as follows. cDNA encoding the C-domain (amino acid residues 670 -1009) of rat CAK␤ was amplified from a cDNA clone 17N (1) by polymerase chain reaction and inserted into pGEX-2TK vector. The GST fusion protein was expressed in Escherichia coli strain BL21(DE3), affinity-purified by using glutathione-agarose, and phosphorylated in vitro using the catalytic subunit of cAMP-dependent protein kinase (Sigma) and [␥-32 P]ATP (ICN Biochemicals Inc., CA) as described by Hildebrand et al. (15) and Kaelin et al. (25).
Screening of ZAPII expression libraries was done as described below. Phage plaques were formed on culture plates at 37°C, and the protein expression was induced by overlaying nitrocellulose membranes (BA85; Schleicher & Schuell) that had been soaked in 10 mM isopropyl-␤-D-thiogalactopyranoside. After overnight incubation at 37°C, the membranes were removed and washed twice at 4°C in Hyb75 buffer (20 mM Hepes (pH 7.5), 75 mM KCl, 0.1 mM EDTA, 2.5 mM Mg 2 Cl, 1 mM dithiothreitol, 0.05% Nonidet P-40) (25). The membranes were soaked twice for 10 min in Hyb75 buffer supplemented with 6 M guanidine hydrochloride each time in fresh solution and then for 5 min each in Hyb75 buffers supplemented with 3 M, 1.5 M, and 0.75 M guanidine hydrochloride, in this order. After soaking in Hyb75 buffer containing 5% skim milk at 4°C for 1 h with constant shaking, which represents a blocking step, the membranes were then incubated with 10 6 cpm/ml of the 32 P-labeled GST-CAK␤-Cdom (50 ng/ml) in Hyb75 buffer, which contained 1% skim milk and 0.6 g/ml of GST prepared by expression from pGEX-2TK vector. The membranes were washed seven times for 15 min each in Hyb75 buffer containing 1% skim milk. Positive plaques were made visible by exposure of the membranes to x-ray films. The positive clone thus obtained, cbp-1, was subcloned into pBluescript and subjected to sequencing in both directions after the preparation of internal deletion mutants.
Epitope-tagged Hic-5 and Fusion Proteins-The plasmid construct encoding the N-terminally Myc-tagged Hic-5 was generated as follows. Using polymerase chain reaction, the cDNA encoding full-length (according to the open reading frame described by Shibanuma et al. (23)) human Hic-5 was amplified, and BamHI and EcoRI restriction sites were created at nucleotide positions Ϫ6 (immediate 5Ј-side of the presumed translational initiation codon ATG) and 1492, respectively. The amplified cDNA was ligated in frame to the BamHI and EcoRI sites of the pcDNA3Myc vector to obtain pHic5-Myc. The pcDNA3Myc vector was constructed by ligating the HindIII-BglII fragment of the pJ3M vector (26) into the HindIII and BamHI sites of pcDNA3 (Invitrogen). The 10 amino acid residues of the epitope tag are specifically recognized by the anti-Myc monoclonal antibody 9E10.
Deletion Mutants of CAK␤-The full-length CAK␤ cDNA clone, 17N, and the C-terminally epitope-tagged CAK␤ cDNA were subcloned into expression vector pSRE to obtain pCAK␤(S) and pCAK␤Tag as described previously (1). pCAK␤(S) and pCAK␤Tag were used for the generation of CAK␤ variants. Deletion (dl) mutations are designated by the amino acid residues deleted. The base pair (bp) designation corresponds to the nucleotide sequence of the CAK␤ cDNA counting from the translational initiation codon (1). Mutation dl 86 -321 was generated by digesting pCAK␤Tag with PvuII (which cleaves at bp 253, 529, 862, 961, and 2698) and isolating the largest fragment and the fragment of 1737 base pairs followed by rejoining of the PvuII termini; rejoining of the 1737-bp fragment in the right direction was confirmed. To construct dl 159 -552, pCAK␤(S) was digested with BspEI (which cleaves at bp 473, 680, and 1655) followed by religation of the BspEI termini. This dl 159 -552 mutant of pCAK␤(S) was digested with SacI (which cleaves at bp 2854), and then the SacI fragment of pCAK␤Tag containing the Tag cDNA sequences was ligated into the SacI site of the deletion mutant of pCAK␤(S) to generate the dl 159 -552 mutant of pCAK␤Tag. Mutation dl 741-903 was created by digesting pCAK␤Tag with Bsu36I (which cleaves at bp 2218 and 2707) followed by religation of the Bsu36I termini.
Production of Antiserum to Hic-5 and Affinity Purification of the Antibody-The anti-Hic-5 antibody was raised in rabbits against a GST fusion protein of full-length human Hic-5, GST-Hic5(fl). Anti-GST antibody was first removed from the serum by the use of a column of covalently bound GST. Anti-Hic-5 was then affinity-purified on a column of the immunogen covalently bound to cyanogen bromide-activated Sepharose 4B, from which the antibody was eluted with 0.5 M ammonium hydroxide containing 3 M sodium thiocyanate (pH 11.0). The anti-Hic-5 antibody immunoprecipitated Hic-5 of human and rat origin and was good for use in immunoblotting and immunocytochemistry.
Antibodies and Other Materials-The first anti-CAK␤ rabbit antibody used in this study, anti-CAK␤(C-a), was raised against a GST fusion protein of residues 670 -716 of rat CAK␤ and was affinitypurified on a column of the immunogen covalently bound to cyanogen bromide-activated Sepharose 4B. Anti-CAK␤(C-a) was found to be specific to CAK␤; the antibody did not immunoprecipitate or immunoblot FAK. The second anti-CAK␤ rabbit antibody, anti-CAK␤(N), was raised against a GST fusion protein of rat CAK␤ residues from Ϫ5 to 416, GST-CAK␤-Ndom; this antiserum was used either without purification or after removal of anti-GST followed by affinity purification on a column of immobilized immunogen. Anti-CAK␤(N) was also found to be specific to CAK␤. Anti-CAK␤(N)-mAb and anti-CAK␤(C)-mAb are mouse IgG 1 monoclonal antibodies raised against GST-CAK␤-Ndom and GST-CAK␤-Cdom, respectively.
Immunofluorescence Microscopy and Confocal Laser-scanning Microscopy-Cells grown on glass coverslips coated with rat tail collagen (33) were fixed with cold absolute ethanol unless otherwise stated and kept at Ϫ20°C until use. After being rinsed with phosphate-buffered saline (PBS), the cells were incubated with Block Ace (Dainippon Pharmaceutical Co., Tokyo, Japan) at room temperature for 30 min. Then a primary antibody was applied. The incubation with anti-Hic-5 antibodies was done for 1 h at room temperature, and the incubation with other antibodies was done for 30 min at room temperature. Fluorescein isothiocyanate-or rhodamine-conjugated antibodies were then applied for 30 min at room temperature. The cells were thoroughly washed in PBS and incubated with the secondary antibody. After being rinsed with PBS, the coverslips were mounted in a solution of 10% PBS and 90% glycerol containing 1 mg/ml p-phenylenediamine (Kanto Chemical Co., Tokyo, Japan). For double staining, the following combinations of primary antibodies or stains were used; anti-Hic-5 and anti-FAK antibodies, anti-Hic-5 antibody and rhodamine-conjugated phalloidin, anti-Myc antibody and rhodamine-conjugated phalloidin, anti-Hic-5 and antivinculin antibodies, and anti-CAK␤ and anti-vinculin antibodies. The samples were examined with an Olympus epifluorescence photomicroscope (Olympus, Tokyo, Japan). Some samples were imaged with a confocal laser-scanning microscope (model TCS NT, Leica, Heebrugg, Switzerland).
Immunoprecipitation of Hic-5, CAK␤, and Other Proteins-Confluent monolayer cultures of cells in 10-cm dishes were washed twice with PBS and then lysed on ice in 0.5 ml per dish of a lysis buffer (20 mM Tris-HCl (pH 7.4), 150 mM NaCl, 2.5 mM EDTA, 1% Nonidet P-40, 10% glycerol, 10 g/ml each leupeptin and aprotinin, 1 mM phenylmethylsulfonyl fluoride, 50 mM NaF, 1 mM Na 3 VO 4 , 20 mM Na 4 P 2 O 7 ). The lysates were subjected to centrifugation at 12,000 ϫ g for 10 min at 4°C to obtain clarified lysates. Portions of the lysates were precleared by mixing for 2 h at 4°C with either normal rabbit IgG bound to protein A-Sepharose or mouse IgG bound to anti-mouse IgG agarose, depending on the antibody to be used in the immunoprecipitation. The cell lysates thus precleared were then incubated at 4°C for 4 h or overnight with antibody beads. The anti-Hic-5 beads and the anti-CAK␤ beads were prepared for each assay by mixing either 1 g of protein of affinity-purified anti-Hic-5, 3 g of protein of affinity-purified anti-CAK␤, or 4 l of anti-CAK␤ serum with 10 l (packed volume) of protein A-Sepharose and washing the Sepharose beads with the lysis buffer. The mouse antibody beads were prepared for each assay by mixing 1 g of protein of either anti-paxillin or anti-FAK monoclonal antibody or 3 g of protein of an anti-CAK␤ monoclonal antibody with 10 l (packed volume) of anti-mouse IgG-agarose and washing the agarose beads with the lysis buffer. Each immunoprecipitation was done from 1 mg of protein of the clarified lysates. As a control, rabbit immunoglobulin beads, mouse IgG beads, or preimmune rabbit serum beads were prepared and used in each assay. Immunoprecipitates were washed three times with the lysis buffer, and proteins were separated by SDS-PAGE according to the method of Laemmli and Favre (34). The separated proteins were blotted onto polyvinylidene difluoride (PVDF) membranes (Immobilon-P, Millipore Corp., Bedford, MA). The membranes were blocked with 3% bovine serum albumin in TBST (25 mM Tris-HCl (pH 7.5), 150 mM NaCl, and 0.05% Tween 20) for 20 min at 60°C and then probed with a primary antibody in TBST containing 1% bovine serum albumin for 1 h at room temperature. For immunoblotting, affinity-purified anti-Hic-5 and affinity-purified anti-CAK␤(C-a) antibodies were used at 1 g of protein/ml, and anti-CAK␤ serum was used at a 200-fold dilution. The membranes were washed with TBST three times and probed again in TBST for 1 h with a second antibody conjugated with alkaline phosphatase or, for enzyme-linked chemiluminescence, with a second antibody conjugated with horseradish peroxidase, followed by washing three times in TBST. Positive bands were detected either by enzyme-linked chemiluminescence according to the manufacturer's (Amersham) protocol or by incubation in nitro blue tetrazolium and 5-bromo-4-chloro-3-indolyl phosphate.
Blot Overlay Assay-Four to five g each of GST-Hic5(fl), GST-Hic5-Ndom, GST-Hic5-Cdom, and GST was subjected to SDS-PAGE and blotted to a PVDF membrane. The membranes were soaked in solutions of guanidine hydrochloride for denaturation and renaturation of the bound proteins as described above under "Cloning of a cDNA Encoding a CAK␤-binding Protein." Then the proteins on membranes were probed with 32 P-labeled probes as also described in that section. Probes used were GST fusion proteins of the N-and C-domains of CAK␤ and FAK and portions of them. The GST fusion proteins were expressed from pGEX-2TK vector as described above and thus contained a site of phosphorylation by cAMP-dependent protein kinase. 32 P labeling of these probes was as described above. One g of each probe was labeled at about 3 ϫ 10 7 cpm. The probes were added to each assay at a concentration of 3 ϫ 10 6 cpm/ml.

Isolation of a cDNA Clone Encoding a Binding Protein to the CAK␤ C-domain-
In an effort to identify proteins that bind CAK␤, we used the N-and C-domains of CAK␤, which are the regions contiguous to the kinase domain at the N-and Cterminal sides, to screen a ZAPII expression library derived from human hippocampus cDNA. One clone (clone cbp-1) was positively identified by screening of 2 ϫ 10 6 plaques with the C-domain, GST-CAK␤-Cdom (consisting of amino acid residues 670 -1009 of rat CAK␤), as a probe. A comparison of 1759 base pairs of the cbp-1 cDNA sequence with those in the GenBank TM data base by the BLASTx program (35) of NCBI revealed high similarity of the amino acid sequence translated from cbp-1 in one reading frame and the Hic-5 amino acid sequence (Gen-Bank TM L22482) over their entire length. This result indicated that cbp-1 is a cDNA clone encoding the full-length, human homologue of Hic-5, which had been cloned from a mouse cDNA library as a transforming growth factor ␤1and hydrogen peroxide-inducible mRNA by Shibanuma et al. (23). Human Hic-5 has the same number of amino acid residues as mouse Hic-5, and their amino acid sequences are 97.0% identical in their double zinc finger LIM domains (amino acid residues 211-444 (23)) (36, 37) and 84.3% identical in their N-domain (amino acid residues 1-210 (23)). The clone cbp-1 contained flanking 5Јand 3Ј-untranslated sequences of 49 and 376 base pairs, respectively. In the fusion protein encoded by clone cbp-1, the 49 base pairs at the presumed 5Ј-untranslated region (23) predicted 16-amino acid residues contiguous to ␤-galactosidase encoded by the phage vector used in the construction of the cDNA library.
A homology search in GenBank TM using the BLASTx program revealed that the amino acid sequence of Hic-5 had high similarity with that of paxillin both in its LIM domains and in its regions of LD motifs (38) (Fig. 1). Paxillin has four LD motifs. Three LD motifs are identified in Hic-5 lacking the LD motif corresponding to the first one in paxillin. However, the 16 amino acid residues encoded by the presumed 5Ј-untranslated region contiguous to ␤-galactosidase were highly similar to the first LD motif of paxillin (Fig. 1). This sequence similarity between Hic-5 and paxillin strongly suggests that the 16 amino acid residues encoded by the presumed 5Ј-untranslated region are, in fact, a part of the Hic-5 amino acid sequence and that the methionine at the residue 1 of Hic-5, according to the numbering by Shibanuma et al. (23), is not the true translational initiation site. The nucleotide sequences around this amino acid residue 1 fit poorly to the sequence context for translational initiation defined by Kozak (39), with neither purine at position Ϫ3 nor a G at position ϩ4 (see the nucleotide sequences submitted to data bases). The nucleotide sequence of human Hic-5 cDNA is highly similar to that of mouse Hic-5 cDNA up to the 39 base pairs upstream of the presumed translational initiation codon, but the remaining 5Ј-terminal 10 base pairs of human Hic-5 cDNA are totally different from those of mouse cDNA cloned by Shibanuma et al. (23). We tried to identify the true translational initiation site of the human Hic-5 cDNA. However, we found only one more Hic-5 cDNA clone in the cDNA library where cbp-1 was cloned after an extensive screening with a fragment of cbp-1 as a probe and by nested polymerase chain reactions. This cDNA clone was 9 base pairs shorter at the 5Ј-end than the clone cbp-1. For this reason, the amino acid residues of human Hic-5 were tentatively numbered according to the numbering by Shibanuma et al. (23) (Fig. 1).
Hic-5 and paxillin contain four contiguous double zinc finger LIM domains at their carboxyl-terminal halves, amino acid residues 211-444 of Hic-5, and 324 -557 of paxillin. In the four LIM domains, the amino acid sequences of the two proteins are 62% identical. In the Hic-5 N-domain, amino acid sequence similarity with paxillin is limited to the LD motifs. Hic-5 has a proline-rich sequence at amino acid residues 14 -20, which is a potential ligand to SH3 domains.
Hic-5 Localizes at Focal Adhesions-An anti-Hic-5 antibody was raised in rabbits against the GST fusion protein of human Hic-5 (GST-Hic5(fl)) and affinity-purified on a column of immobilized immunogen. The anti-Hic-5 antibody was found to be specific to Hic-5 as shown by immunoblotting and immunoprecipitation from the rat fibroblast WFB cell lysate, where a band of about 55 kDa was detected by the antibody (Fig. 2). The anti-paxillin monoclonal antibody obtained from Transduction Laboratories immunoprecipitated and immunoblotted not only paxillin but also Hic-5 (Fig. 2). The anti-Hic-5 antibody neither immunoprecipitated paxillin nor bound to paxillin blotted from gel after immunoprecipitation with anti-paxillin and separation by SDS-PAGE (Fig. 2).
This anti-Hic-5 antibody specifically immunostained focal adhesions in WFB cells (Fig. 3). Mixing the antibody with a 10-fold molar excess of immunogen before immunostaining prevented the specific staining (data not shown). Confocal laserscanning microscopy revealed exact colocalization of Hic-5 and FAK at focal adhesions in WFB cells (Fig. 3a). Hic-5 in WFB cells localized exactly at the sites extending along the stress fibers to the cell surface (Fig. 3b). Thick bundles of microfilaments (stress fibers) crossing the cytoplasm were made visible by decoration with rhodamine-conjugated phalloidin.
To examine whether our anti-Hic-5 antibody stained the protein in WFB cells encoded by the cloned Hic-5 cDNA, a Hic-5 cDNA construct with an N-terminal Myc tag was prepared, and The sequence of human Hic-5 was deduced from the nucleotide sequence of clone cbp-1. The numbers on the right indicate the amino acid residue numbers of human Hic-5 (Hic) and human paxillin (Pax) (40). The amino acid residues of human Hic-5 were tentatively numbered according to the numbering by Shibanuma et al. (23). The amino acid sequence encoded by the presumed (23) "5Ј-untranslated region" of human Hic-5 cDNA are indicated by italic type. Amino acid residues of human paxillin identical with those of human Hic-5 are indicated by dashes. Dots represent gaps introduced to improve the alignment. The metal-liganding residues that define the LIM consensus sequence are boxed. The LD motifs (38) are underlined.

FIG. 2. Immunoprecipitation and immunoblotting of Hic-5 in WFB cells with anti-Hic-5.
The WFB cell lysate was prepared in a lysis buffer containing 0.5% sodium deoxycholate and 0.1% SDS in addition to 1% Nonidet P-40. Hic-5 was immunoprecipitated from 1.2 mg of protein of the lysate with 1 g of protein of affinity-purified anti-Hic-5 bound to protein A-Sepharose (10 l of packed volume). Paxillin was immunoprecipitated from 0.4 mg of protein of the lysate with 1 g of protein of anti-paxillin monoclonal antibody bound to anti-mouse IgG-agarose (7.5 l of packed volume). The immunoprecipitates and 60 g of protein of the WFB cell lysate (total lysate) were subjected to SDS-PAGE in a 10% gel. The separated proteins were blotted onto a PVDF membrane. Anti-Hic5 antibody and anti-paxillin monoclonal antibody were used for immunoblotting, which is indicated at the bottom of each lane. Positions of molecular mass markers (Sigma SDS-7B) are indicated on the right. i.p., immunoprecipitation. the tagged Hic-5 was transiently expressed in WFB cells. The presence of the tagged Hic-5 at focal adhesions in transfected WFB cells is shown in Fig. 4 by immunostaining with anti-Myc. The specificity of the immunostaining with anti-Myc is obvious because the WFB cells expressing the tagged Hic-5 are surrounded by more than a 100-fold excess in number of the untransfected cells. In the cells expressing tagged Hic-5, the ends of stress fibers terminate at the sites of tagged Hic-5 (Fig.  4B).
Direct Binding of the Hic-5 N-domain to the C-terminal Region of CAK␤-Direct binding of Hic-5 to CAK␤ was shown by blot overlay assays, and the specificity of this binding was examined. GST fusion proteins of almost full-length Hic-5 (amino acid residues Ϫ16 to 444; GST-Hic5(fl)), the Hic-5 N-domain (amino acid residues 1-223; GST-Hic5-Ndom), the Hic-5 Cdomain (amino acid residues 203-444 containing the LIM domains; GST-Hic5-Cdom), and, as a control, GST were subjected to SDS-PAGE and immobilized on a PVDF membrane. Hic-5 in the WFB cell lysate was immunoprecipitated with anti-Hic-5 and was also immobilized on the membrane after the electrophoretic separation. After procedures for denaturation and renaturation of the proteins on the membrane, the fusion proteins and Hic-5 were probed for binding to CAK␤ and FAK. The probes used for the binding assay were 32 P-labeled GST fusion proteins of the N-and C-domains of CAK␤, GST-CAK␤-Ndom and GST-CAK␤-Cdom; fragments of the C-domain, GST-CAK␤-CdomA and GST-CAK␤-CdomB; and the C-domain of FAK, GST-FAK-Cdom, and its fragment, GST-FAK-CdomB. The CAK␤ C-domain, GST-CAK␤-Cdom, was bound by GST-Hic5(fl), GST-Hic5-Ndom, and Hic-5 from WFB cells but not by GST-Hic5-Cdom (Fig. 5B). The CAK␤ N-domain was not bound by these GST fusion proteins or by Hic-5 from WFB cells (data not shown). The FAK C-domain, GST-FAK-Cdom, was also bound by GST-Hic5(fl) and GST-Hic5-Ndom (data not shown). When the two regions of the divided CAK␤ C-domain were tested for the binding, only the extreme C-terminal region, GST-CAK␤-CdomB, was bound by GST-Hic5(fl) and GST-Hic5-Ndom (Fig. 5, C and E). These results are consistent with the results obtained by dot blots, in which the ZAPII phage plaques of the original cbp-1 clone induced to produce the fusion protein were probed with 32 P-labeled GST-CAK␤-CdomA and GST-CAK␤-CdomB; the positive signal was obtained only with GST-CAK␤-CdomB (data not shown). The same region of FAK contained the binding site; only the extreme C-terminal region of the FAK C-domain, GST-FAK-CdomB, was bound by GST-Hic5(fl) and GST-Hic5-Ndom (Fig.  5F). It was noted that the extreme C-terminal halves of the CAK␤ and FAK C-domains gave better signals than the whole C-domains in the blot overlay assays. These results indicate that a common structure in the extreme C-terminal halves of the CAK␤ and FAK C-domains has specific affinity to the N-domain of Hic-5.
GST Fusion Protein of the CAK␤ C-terminal Region Bound in Vitro Hic-5 from the WFB Cell Lysate-Hic-5 in the WFB cell lysate was bound to the GST fusion protein of full-length CAK␤, GST-CAK␤(fl), and to the fusion protein of the C-terminal portion of CAK␤, GST-CAK␤-CdomB (Fig. 6, lanes 4 and  10). Paxillin was also bound to GST-CAK␤-CdomB (Fig. 6, lane 10) but was not significantly bound to GST-CAK␤(fl) (Fig. 6,  lane 4). The anti-paxillin monoclonal antibody used in this study immunostained Hic-5 in addition to paxillin as shown in Fig. 2 (Fig. 6, middle). The C-terminal portion of FAK, GST-FAK-CdomB, also bound Hic-5 and paxillin (Fig. 6, lane 12). The GST fusion proteins of the CAK␤ N-domain, GST-CAK␤-Ndom, and the region of the CAK␤ C-domain proximal to the kinase domain, GST-CAK␤-CdomA, did not bind Hic-5 or paxillin (Fig. 6, lanes 6 and 8). These results further prove that Hic-5 and paxillin bind to the extreme C-terminal halves of CAK␤ and FAK C-domains.
Coimmunoprecipitation of CAK␤ with Hic-5 from WFB Cell Lysate-When Hic-5 was immunoprecipitated with anti-Hic-5 from the lysate of WFB cells, CAK␤ was coimmunoprecipitated with Hic-5 (Fig. 7, lane 4 of top). This association of CAK␤ with Hic-5 was found when the lysate was prepared in a lysis buffer containing 1% Nonidet P-40 as the detergent. The addition of sodium deoxycholate at 0.5% to the lysis buffer prevented successful demonstration of the coimmunoprecipitation. As shown in Fig. 2, the anti-paxillin monoclonal antibody immunoprecipitated Hic-5 (Fig. 7, lane 12 of middle) in addition to paxillin, which migrated as a broad band behind Hic-5; thus, CAK␤ was also found in the immunoprecipitate with this anti-paxillin FIG. 4. Localization at focal adhesions of Myc-tagged Hic-5 expressed in WFB cells. pHic5-Myc plasmid for expression of Myctagged Hic-5 was transfected into WFB cells at 2.2 g/3.5-cm dish by the use of Tfx -50 (Promega, Madison, WI). The transfected cells grown on glass coverslips coated with rat tail collagen were cultured for 2 days. Cells were fixed with 4% paraformaldehyde and permeabilized with 0.1% Triton X-100. The cells were doubly stained with anti-Myc (A) and with rhodamine-conjugated phalloidin (B) and were viewed under a fluorescence microscope. Pairs of photographs, A and B, were taken at the same height from the dish surface. Magnification was ϫ 1361.

FIG. 5. Specific in vitro binding of the C-terminal regions of CAK␤ and FAK to Hic-5 and the Hic-5 N-domain.
Four g each of purified GST-Hic5-Ndom and GST proteins and 5 and 20 g of protein of total bacterial lysates expressing GST-Hic5(fl) and GST-Hic5-Cdom, respectively, were subjected in triplicate to SDS-PAGE in 7.5% gels and blotted onto PVDF membranes. Hic-5 immunoprecipitated from 1.5 mg of protein of the WFB cell lysate with anti-Hic-5 was also subjected to the procedures. After denaturation and renaturation of the GST fusion proteins, the proteins on the membrane were probed for binding by 32 P-labeled GST-CAK␤-Ndom (data not shown), GST-CAK␤-Cdom (GSTCAK␤Cdom) (B), GST-CAK␤-CdomA (GSTCAK␤CdomA) (E), GST-CAK␤-CdomB (GSTCAK␤CdomB) (C), GST-FAK-CdomA (data not shown), and GST-FAK-CdomB (GSTFAKCdomB) (F). These probes had 3 ϫ 10 7 cpm/g and were used at 3 ϫ 10 6 cpm/ml. Autoradiograms are shown in B, C, E, and F. The stained protein patterns shown in A and D were obtained by staining of the gels with Coomassie Blue after electroblotting the separated proteins onto PVDF membranes. GST fusion proteins subjected to SDS-PAGE are indicated across the top. 32 P-Labeled probes used in the blot overlay assay are indicated at the bottom of each panel. (Fig. 7, lane 12 of top). Hic-5 was resolved by SDS-PAGE into double bands just above immunoglobulin heavy chains. When the association of Hic-5 and CAK␤ was examined in the reverse direction by immunoprecipitating CAK␤ from the WFB cell lysate, it was hard to find Hic-5 in the anti-CAK␤ immunoprecipitate by blotting with anti-Hic-5 (Fig. 6, lane 2 and 3 of middle). In an attempt to show coimmunoprecipitation of Hic-5 with CAK␤, we immunoprecipitated CAK␤ with different anti-CAK␤ antibodies: anti-CAK␤(C-a), anti-CAK␤(N), anti-CAK␤(N)-mAb, and anti-CAK␤(C)-mAb. In the immunoprecipitates from the WFB cell lysate with any of these anti-CAK␤ antibodies, no significant amount of Hic-5 was demonstrated by blotting with anti-Hic-5 (Fig. 7). However, as shown in Fig.  11, coimmunoprecipitation with CAK␤ of a tyrosine-phosphorylated protein migrating at the position of Hic-5 was found by blotting with anti-phosphotyrosine. Moreover, when the A-431 cell lysate was used, a small amount of Hic-5 was found by blotting with anti-Hic-5 in the immunoprecipitates with anti-CAK␤(C-a) and anti-CAK␤(C)-mAb (data not shown). A small amount of paxillin, but no Hic-5, was found coimmunoprecipitated with FAK from the WFB cell lysate (Fig. 7, lane 8) (the blotting with anti-paxillin is not shown). Paxillin was faintly detected in the immunoprecipitates with anti-CAK␤(N), anti-CAK␤(N)-mAb, and anti-CAK␤(C)-mAb by extended blotting with anti-paxillin (data not shown). As shown below, coimmunoprecipitation of Hic-5 with CAK␤ was clearly demonstrated in COS-7 cells expressing these proteins from transfected cDNA constructs.
Inhibition of CAK␤ Coimmunoprecipitation with Hic-5 by GST Fusion Proteins of the Extreme C-terminal Portions of CAK␤ and FAK-The association of the C-terminal region of CAK␤ with Hic-5, shown in Fig. 5 by blot overlay assay and in Fig. 6 by pull-down assay, was confirmed by showing an inhibition of CAK␤ coimmunoprecipitation with Hic-5 from the WFB lysate. The addition of the extreme C-terminal region of CAK␤ fused to GST, GST-CAK␤-CdomB, to the WFB cell lysate prior to the immunoprecipitation with anti-Hic-5 markedly interfered with the CAK␤ coimmunoprecipitation with Hic-5 (Fig. 7, lane 5 of top). This inhibition of the CAK␤ association with Hic-5 was also found when the corresponding C-terminal region of FAK, GST-FAK-CdomB, was added to the lysate, but the inhibition was not found when the N-domain of CAK␤, GST-CAK␤-Ndom, was added (Fig. 7, lanes 6 and 7 of top). These results are consistent with the in vitro binding data shown in Figs. 5 and 6.

Analysis of the Association of CAK␤ and Hic-5 by the Use of Deletion Mutants of CAK␤ Expressed in COS-7 Cells from
Transfected cDNA Constructs-COS-7 cells endogenously express a small amount of Hic-5 but almost no CAK␤. In the experiments shown in Fig. 8, CAK␤ and Hic-5 were expressed in COS-7 cells from transfected cDNA constructs. In this analysis of the CAK␤ association with Hic-5 by immunoprecipitation with anti-Hic-5 and anti-CAK␤ from the lysates of

FIG. 7. Coimmunoprecipitation of CAK␤ with Hic-5 from the WFB cell lysate and inhibition of this coimmunoprecipitation by GST fusion proteins of the C-domains of CAK␤ and FAK.
Hic-5, CAK␤, FAK, and paxillin were immunoprecipitated from 1.0 mg of protein of the WFB cell lysate with anti-Hic-5, anti-CAK␤(C-a), anti-CAK␤(N), and polyclonal anti-FAK bound to protein A-Sepharose and anti-paxillin, anti-CAK␤(N)-mAb, and anti-CAK␤(C)-mAb bound to anti-mouse IgG-agarose as indicated at i.p. As controls, normal rabbit Ig and normal mouse IgG1 bound to these beads were used for immunoprecipitation from the lysate (lanes 1 and 9). In the immunoprecipitation shown in lanes 5, 6, and 7, 20 g of protein of GST-CAK␤-CdomB, GST-FAK-CdomB, or GST-CAK␤-Ndom was mixed with the cell lysate before immunoprecipitation as indicated (addition). The immunoprecipitates were subjected to SDS-PAGE in a 7.5% gel, and the separated proteins were blotted onto a PVDF membrane. The membrane was cut at the 89-kDa prestained marker protein into high and low molecular weight regions. The proteins in the high molecular weight region were first probed with anti-CAK␤(C-a) (anti-CAK␤) to obtain the data shown in the top, deprobed, and then reprobed with monoclonal anti-FAK to obtain the data shown at the bottom. The proteins in the low molecular weight region were first probed with anti-Hic-5 to obtain the data shown in the middle, deprobed, and then reprobed with anti-paxillin (data not shown). Binding of antibody probes was visualized either by alkaline phosphatase (bottom) or by peroxidase (ECL) (top and middle). The position of oligomeric GST-CAK␤-CdomB (CdomB) is indicated by an arrow above the position of CAK␤ in the top. The positions of prestained molecular mass markers (Sigma SDS-7B) are indicated on the left. i.p., immunoprecipitation. blot, immunoblot. Ig, heavy chains of immunoglobulins clearly found when rabbit sera were used as antibody preparations (lanes 1, 3, and 8). mAb, monoclonal antibody. transfected COS-7 cells, we were able to show the coimmunoprecipitation of CAK␤ with Hic-5 and Hic-5 with CAK␤ (Fig. 8,  lanes 7 and 8). In these experiments, CAK␤ was expressed as C-terminally HSV-tagged CAK␤ and Hic-5 was expressed as N-terminally Myc-tagged Hic-5. There may be at least two reasons for this successful demonstration of Hic-5 coimmunoprecipitation with CAK␤. We found that the anti-Myc monoclonal antibody detected Hic-5 at a sensitivity significantly higher than that with anti-Hic-5 (the data of immunoblotting with anti-Hic-5 are not shown in Fig. 8). It was also noted that Hic-5 with an N-terminal Myc-tag was expressed at a significantly high level in COS-7 cells.
To determine the region of CAK␤ responsible for the intracellular association with Hic-5, three deletion mutants of CAK␤ were coexpressed with Hic-5 in COS-7 cells and examined for coimmunoprecipitation with Hic-5. Immunoprecipitation with anti-Hic-5 and with anti-CAK␤ gave the same results for the association of Myc-tagged Hic-5 and the CAK␤ mutants (Fig.  8). The association was not affected in deletion mutants dl 86 -321 and dl 159 -552 of CAK␤ (Fig. 8, lanes 11-14). No association with Hic-5 was found in deletion mutant dl 741-903 of CAK␤ (Fig. 8, lanes 9 and 10). The results indicated that a region of the CAK␤ C-terminal domain was responsible for the intracellular association with Hic-5.
Tyrosine Phosphorylation of Hic-5-The results shown above indicated that Hic-5 was a protein highly related to paxillin in its structure and function. It is known that paxillin is markedly tyrosine-phosphorylated upon activation of FAK (41). Therefore, we examined whether Hic-5 was tyrosine-phosphorylated upon activation of CAK␤ and in Src-transformed cells. Hic-5 was strongly tyrosine-phosphorylated in 3Y1 and WFB cells treated with pervanadate (Fig. 9). Mobility retardation was observed in Hic-5 when the protein was heavily phosphorylated. Hic-5 in Src-transformed 3Y1 cells, SR-3Y1, was also significantly tyrosine-phosphorylated as compared with the protein in 3Y1 cells (Fig. 9). In accordance with the data reported by Shibanuma et al. (23,24), SR-3Y1 cells, a transformed cell line, contained much smaller amounts of Hic-5 than the untransformed counterpart, 3Y1 cells (Fig. 9A, lanes  1 and 3).
When CAK␤ in serum-starved WFB cells was activated by stimulation with serum, lysophosphatidic acid, or endothelin or by exposure to hypertonic osmotic stress, the levels of Hic-5 tyrosine phosphorylation were enhanced parallel with the levels of CAK␤ tyrosine phosphorylation (Fig. 10). These results indicated that Hic-5 was tyrosine-phosphorylated in Src-transformed cells and also when CAK␤ was activated.
CAK␤ and Hic-5 Were Tyrosine-phosphorylated in Parallel in WFB Cells either Exposed to Hypertonic Osmotic Stress or Stimulated with Lysophosphatidic Acid-The association of Hic-5 with CAK␤ was examined under conditions where the tyrosine phosphorylation of CAK␤ was enhanced. The level of CAK␤ tyrosine phosphorylation decreased upon detachment of WFB cells from culture dishes by trypsinization (Fig. 11, bottom, lane 3 as compared with lane 5). The tyrosine phosphorylation of CAK␤ was enhanced by stimulation of WFB cells with lysophosphatidic acid (Fig. 11, lane 7) and by exposing the cells to hypertonic osmotic stress (Fig. 11, lane 9). The amounts of CAK␤ coimmunoprecipitated with Hic-5 did not significantly change under these various conditions of WFB cells where the levels of CAK␤ tyrosine phosphorylation varied (Fig. 11). However, blotting with anti-phosphotyrosine revealed that the tyrosine-phosphorylated CAK␤ present in anti-Hic-5 immunoprecipitates decreased upon detachment of WFB cells from culture dishes by trypsinization (Fig. 11, lane 4) and increased upon stimulation of the cells with lysophosphatidic acid and exposure of the cells to osmotic stress (Fig. 11, lanes 8 and 10). A tyrosine-phosphorylated band was found above CAK␤ in the anti-Hic-5 immunoprecipitates; this band was most prominent when cells were stimulated with lysophosphatidic acid (Fig. 11,  lane 8) but was also found in cells adhering on dishes and when cells were exposed to osmotic stress (Fig. 11, lanes 6 and 10). In  1 and 2) and SR-3Y1 cells (lanes 3 and 4). These lysates were prepared in a lysis buffer containing 0.5% sodium deoxycholate and 0.1% SDS in addition to 1% Nonidet P-40. Cells used for lane 2 were treated with 1 mM pervanadate (P. Van.) for 10 min; pervanadate was prepared by mixing 200 l of 100 mM sodium orthovanadate with 2 l of 35% hydrogen peroxide, and, after being allowed to stand at room temperature for 20 min, the mixture was added to the medium at 1% volume. The immunoprecipitates were subjected to SDS-PAGE in a 9% gel. After transfer onto a PVDF membrane, the separated proteins were probed with anti-Hic-5 (A) and anti-phosphotyrosine 4G10 (B). The lower band seen at Hic-5 in lane 3 of A is the heavy chains of immunoglobulins used in the immunoprecipitation. blot, immunoblotting. another experiment (data not shown), this band above CAK␤ revealed with anti-phosphotyrosine was resolved into double bands, an upper major band and a lower minor band. The upper major band had a mobility slower than FAK; possible candidates for these bands are phosphorylated FAK, vinculin, and pp130 cas .
It was found that Hic-5 immunoprecipitated from the cells exposed to osmotic stress was made visible by blotting with anti-phosphotyrosine (Fig. 11, lane 10). Blotting with antiphosphotyrosine also revealed a band at the location of Hic-5 in the lane of SDS-PAGE where the anti-CAK␤ immunoprecipitate from cells exposed to osmotic stress was run (Fig. 11, lane  9). These results suggested that Hic-5 coimmunoprecipitated with CAK␤ was selectively tyrosine-phosphorylated in cells exposed to osmotic stress as compared with total cellular Hic-5. This is because the intensity with anti-phosphotyrosine blotting of Hic-5 coimmunoprecipitated with CAK␤ was far stronger than expected from the intensity with anti-phosphotyrosine blotting of Hic-5 immunoprecipitated with anti-Hic-5 from the same cells exposed to osmotic stress (Fig. 11, lanes 9 and 10 of  bottom). The latter Hic-5 (lane 10) was clearly immunostained with anti-Hic-5, whereas it was not possible to make the former Hic-5 (lane 9) visible with anti-Hic-5 blotting (Fig. 11, bottom). Careful examination indicated that Hic-5 immunoprecipitated with anti-Hic-5 from cells on dishes and from cells stimulated with lysophosphatidic acid was also stained faintly with anti-phosphotyrosine.

A Small Portion of CAK␤ Is Present in WFB Cells at the Site of Focal
Adhesions-It was shown that the localization of FAK at focal adhesions is dependent on the association of FAK with paxillin (18), which is targeted to focal adhesions by itself (38). Since CAK␤ bound Hic-5 and paxillin and a fraction of CAK␤ coimmunoprecipitated with Hic-5 from the WFB cell lysate, we examined WFB cells by immunostaining to find CAK␤ at focal adhesions. The major portion of CAK␤ in WFB cells was found in the perinuclear region and in the cytoplasm (Fig. 12A). In the same cell line, FAK localized at focal adhesions (Fig. 3a). At the cell periphery of well spread WFB cells, where focal adhesions were seen by staining with anti-paxillin and anti-vinculin, CAK␤ was faintly immunostained at focal adhesions (Fig.  12). Thus, a small amount of CAK␤ was found at the sites of focal adhesions in WFB cells. However, CAK␤ was immuno- FIG. 11. CAK␤ and Hic-5 were simultaneously tyrosine-phosphorylated in WFB cells either exposed to hypertonic osmotic stress or stimulated with lysophosphatidic acid. Sixteen 10-cm dishes of confluent and quiescent WFB cells of the same culture lot were prepared and divided into four groups. Cells on the dishes of the first group were stimulated in Iscove's medium with 2 M lysophosphatidic acid (LPA) for 10 min at 37°C (lanes 7 and 8). Cells on the dishes of the second group were exposed to hypertonic osmotic stress (Osm) for 20 min at 37°C by replacing the culture medium with 2 ml of Iscove's medium containing 0.3 M sorbitol (lanes 9 and 10). Cells on the dishes of the third group were trypsinized (off) after washing with ATV solution (20); the cells were detached from dishes and allowed to stand at 37°C for 20 min after the addition of trypsin inhibitor (lanes 3 and 4). Cells on the dishes of the fourth group were directly subjected to analysis without treatment (on) (lanes 5 and 6). Cell lysates were prepared from these four groups of cells. Hic-5, CAK␤, FAK, and paxillin (Pax) were immunoprecipitated from 1.2 mg of protein of these WFB cell lysates with anti-Hic-5 and anti-CAK␤(C-a) (anti-CAK␤) bound to protein A-Sepharose and anti-FAK and anti-paxillin monoclonal antibodies bound to anti-mouse IgG-agarose as indicated at i.p. As controls, normal rabbit Ig and normal mouse IgG1 bound to these beads were used for immunoprecipitation from the on cell lysate ( lanes  1 and 12). In lane 2, anti-CAK␤(C-a) itself bound to protein A-Sepharose was run as a control. FAK and paxillin were immunoprecipitated from the lysate of on cells. The immunoprecipitates were subjected to SDS-PAGE in a 7.5% gel, and the separated proteins were blotted onto a PVDF membrane. The membrane was cut at the 89-kDa prestained marker protein into high and low molecular weight regions. The proteins in the high molecular weight region were first probed with anti-CAK␤(C-a) to obtain the data shown at the top. The proteins in the low molecular weight region were first probed with anti-Hic-5 to obtain the data shown at the top. The proteins in both regions were deprobed and then reprobed with anti-phosphotyrosine, 4G10, to obtain the data shown at the bottom; the heavy chains of the mouse immunoglobulins (Ig) used in the immunoprecipitation were stained at the position below Hic-5 in lanes 11, 12 and 13 of the bottom. Binding of antibody probes was visualized either by alkaline phosphatase (bottom) or by peroxidase (ECL) (top). Pax, paxillin; i.p., immunoprecipitation; blot, immunoblot. stained diffusely at the focal adhesions, not in the well confined, rodlike patchy structure of focal adhesions, an image obtained by immunostaining vinculin, paxillin, and Hic-5. When WFB cells were doubly stained for Hic-5 and vinculin, the localization of Hic-5 at the cell periphery overlapped with that of vinculin (data not shown).

Analysis of Tyrosine Phosphorylation Levels of CAK␤ and Hic-5 in WFB Cells: Effects of Trypsinization and Replating the Cells onto Dishes Coated with Fibronectin or Poly-L-lysine-
The changes in the tyrosine phosphorylation levels of FAK, CAK␤, and Hic-5 were compared by trypsinization of WFB cells and replating the cells on dishes coated with either fibronectin or poly-L-lysine. The tyrosine phosphorylation levels of FAK decreased upon detachment of well spread WFB cells from culture dishes by trypsinization, and the levels were recovered by incubating the cells at 37°C for 45 and 90 min after plating them onto dishes coated with fibronectin (data not shown). Plating the cells on poly-L-lysine dishes was not effective to recover the tyrosine phosphorylation level of FAK. These results on FAK obtained in WFB cells are consistent with those reported on FAK in BALB3T3 cells and NIH3T3 cells (11,27). In parallel experiments, it was found that the tyrosine phosphorylation levels of CAK␤ in WFB cells decreased on trypsinization and recovered on incubation at 37°C for 45 min after plating the cells, but this recovery was observed on plating the cells not only onto fibronectin dishes but also onto poly-L-lysine dishes, although the effect of fibronectin was somewhat stronger than the effect of poly-L-lysine (data not shown). Moreover, a longer incubation of the cells for 90 min after replating did not enhance the tyrosine phosphorylation levels of CAK␤.
The tyrosine phosphorylation levels of Hic-5 were also examined in the same WFB cells trypsinized and replated as described above. Hic-5 was tyrosine-phosphorylated only to a limited degree in well spread WFB cells, and the levels of tyrosine phosphorylation did not consistently change on trypsinization and replating of the cells (data not shown). DISCUSSION We identified Hic-5 in a search for CAK␤-binding proteins by cDNA expression cloning. Hic-5 had an amino acid sequence highly similar to paxillin. The results shown in this paper demonstrated that Hic-5 was a component of focal adhesions. This finding disagrees with the original identification by Shibanuma et al. (23,24) that Hic-5 is localized in nuclei. A binding of the Hic-5 LIM domain to certain DNA sequences was postulated (24). We consider that their assumption that Hic-5 is a nuclear protein was drawn from weak evidence obtained by the use of a less specific anti-Hic-5 antibody (23), although a possible function of Hic-5 in the nucleus cannot be excluded at present. In this paper, we have confirmed the Hic-5 localization at focal adhesions by immunostaining with anti-Myc in WFB cells expressing the Myc-tagged Hic-5 from a cDNA construct. The cell biological functions of Hic-5 should be reevaluated from the standpoint that Hic-5 is a component of focal adhesions. Recently, Brown et al. (38) showed that LIM domains are responsible for targeting paxillin to focal adhesion. They identified the LIM3 domain as the principal determinant of paxillin localization at focal adhesions. The four contiguous LIM domains of Hic-5 are most similar to those of paxillin among the known LIM domains (42); the highest similarity is found in each corresponding domain. Among the corresponding four LIM domains of Hic-5 and paxillin, the LIM3 domains of these proteins are most highly conserved with an identity of 70.6% in the amino acid sequences. In addition to a novel role of the LIM domain as protein dimerization motifs (36), it has been proposed that the LIM domain recognizes a tyrosine-containing tight turn (37). The identity of the protein recognized by the LIM3 domains of paxillin and Hic-5 is most interesting because Hic-5 may also be targeted to focal adhesions through its LIM3 domain.
We found that the Hic-5 N-domain directly associated in vitro with the C-terminal regions of CAK␤ and FAK (Figs. 5 and 6). Moreover, the association of Hic-5 with CAK␤ was shown in the lysates of WFB cells and of those COS-7 cells that were doubly transfected with CAK␤ and Hic-5 cDNA constructs. Brown et al. (38) localized on paxillin the binding sites of FAK and vinculin to small stretches of amino acid sequences, which they named LD motifs. They pointed out high sequence similarity between paxillin and Hic-5 in the LD motifs. The paxillin-binding sites on FAK and vinculin have also been identified by Tachibana et al. (18) and named paxillin-binding subdomains 1 and 2 (PBS1 and PBS2). These subdomains are also found in CAK␤ at amino acid residues 875-891 (PBS1) and 990 -998 (PBS2). The PBS1 and PBS2 of FAK and CAK␤ are highly conserved, being 58.8% identical in PBS1 and 77.8% identical in PBS2. The Hic-5 N-domain bound to the CAK␤ C-domain and its fragment containing both PBS1 and PBS2 (Figs. 5 and 6).
The presence of Hic-5 in association with CAK␤ in the WFB and COS-7 cell lysates (Figs. 7, 8, and 11) suggested that Hic-5 was a component in signaling pathways downstream of CAK␤. The results shown in Figs. 7 and 8 are consistent with the prediction that the CAK␤ amino acid residues containing both PBS1 and PBS2 are essential for the binding of CAK␤ to Hic-5. In the deletion mutant of CAK␤, dl 741-903, PBS2 was present, but PBS1 was deleted. In addition to the association of Hic-5 with CAK␤, the association of paxillin with CAK␤ and that of Hic-5 with FAK were also found. It is possible that there are differences in affinities of the associations between FAK/ CAK␤ and paxillin/Hic-5. In this relation, we noted that the GST fusion protein of the full-length CAK␤, GST-CAK␤(fl), had much less affinity to paxillin than to Hic-5 (Fig. 6, lane 4). Moreover, the results in Fig. 6 suggest less binding of Hic-5 to GST-FAK-CdomB than to GST-CAK␤-CdomB.
A high sequence similarity was found between the first LD motif of paxillin and the Hic-5 amino acid sequence translated from the "5Ј-untranslated region," which was presumed on mouse Hic-5 cDNA by Shibanuma et al. (23), of human Hic-5 cDNA, cbp-1 (Fig. 1). The published mouse Hic-5 amino acid sequence (23) contains only three LD motifs lacking the one corresponding to the first LD motif of paxillin. The translational initiation site of the Hic-5 cDNA presumed by Shibanuma et al. (23) does not fit to the sequence context for translational initiation (39) and may not be the true translational initiation site of the Hic-5 mRNA. The level of Hic-5 protein expression in the cDNA transfected cells has not been shown by these authors (23,24). The expression of Hic-5 from constructs of the original human Hic-5 cDNA, cbp-1, was quite low in the transfected cells (data not shown). We will continue our efforts to identify the N-terminal methionine and to sequence further 5Ј-upstream region of Hic-5 cDNA by isolating other Hic-5 cDNA clones.
Although Hic-5 is almost exclusively localized at focal adhesions in WFB cells (Fig. 3) and CAK␤ has specific affinity to Hic-5 and paxillin, only a limited, small fraction of CAK␤ localized at focal adhesions in WFB cells (Fig. 12). This subcellular localization of CAK␤ was different from that of FAK, a large portion of which was found at focal adhesions in WFB cells (Fig. 3). The major portion of CAK␤ in WFB cells is present in the perinuclear region and cytoplasm (Fig. 12). In cultured epithelial cells, CAK␤ was also found at the cell-to-cell borders in addition to the perinuclear regions and cytoplasm. A small amount of CAK␤ present at focal adhesions in WFB cells was found only by careful examination and was not immunostained in images of typical components of focal adhesions such as paxillin, vinculin, Hic-5, and FAK (Fig. 12). This intracellular localization of CAK␤ was compatible with our finding that CAK␤ is present in association with microvilli, cilia, and axons in rat tissues (16). Tachibana et al. (18) showed that FAK localizes at focal adhesions through its binding to paxillin. It was shown that the C-domain of CAK␤ exclusively localized at focal adhesions when this domain was singly expressed in chicken embryo fibroblasts from a designed recombinant cDNA construct (43). It is obvious that CAK␤ has an intrinsic property of localizing at focal adhesions via its C-domain. In most cells, however, the major portion of CAK␤ does not localize at focal adhesions. The reason for this subcellular localization of CAK␤ at sites other than focal adhesions is a focus of our current study. The regulatory mechanism of CAK␤ localization might be important in understanding the unique functions of CAK␤ different from FAK. We (1) previously showed that the tyrosine phosphorylation of CAK␤ is not enhanced in response to plating rat 3Y1 fibroblasts onto fibronectin. Thereafter, different results have been reported on the changes of the CAK␤ tyrosine phosphorylation in response to integrin signaling (14,44,45). The subcellular localization of CAK␤ shown in Fig. 12 may explain our finding that the tyrosine-phosphorylation was not regulated by cell-to-substratum adhesion at focal adhesions.
Although the amino acid sequences of Hic-5 and paxillin are similar at the LIM domains and also at the LD motifs, the other portions of the Hic-5 N-domain are not similar to those of the paxillin N-domain. Paxillin has a role as a signal-integrating protein or a docking protein. It was shown that the Src family protein kinases associate with paxillin by their SH3 domains (46). It was also shown that the transforming protein v-Crk binds with its SH2 domain to the phosphorylated sequences at tyrosine residues 31 and 118 of the paxillin N-domain (47)(48)(49). Although Hic-5 does not share these ligand sequences, the proline-rich sequence found at amino acid residues 14 -20 and some tyrosine-phosphorylated sequences of Hic-5 may also have ligand specificities to some SH3 and SH2 domains. A tyrosine residue at position 43 of Hic-5, which is not found in paxillin, is in a sequence context indicative of a possible phos-phorylation site. The Hic-5 coimmunoprecipitated with CAK␤ from the lysate of WFB cells stimulated by osmotic stress was selectively tyrosine-phosphorylated as compared with the total cellular Hic-5 (Fig. 11). This result indicates functional coupling of CAK␤ and Hic-5. However, it remains unanswered which tyrosine kinase, CAK␤, Src, or another one, is responsible for the phosphorylation of Hic-5. Thus, the portion of the Hic-5 N-domain outside the LD motifs is probably the site important in signal transduction. It seems important to study whether the cell biological effects of Hic-5 reported by Shibanuma et al. (24) are related to unique downstream signals, which are possibly generated from Hic-5 but not from paxillin. Because Hic-5 and paxillin have several properties in common, related but different roles are expected for their functions.