Interaction of Hic-5, A Senescence-related Protein, with Focal Adhesion Kinase*

Hydrogen peroxide-inducible clone (Hic)-5 is induced during the senescent process in human fibroblasts, and the overexpression of Hic-5 induces a senescence-like phenotype. Structurally, Hic-5 and paxillin, a 68-kDa cytoskeletal protein, share homology such as the LD motifs in the N-terminal half and the LIM domains in the C-terminal half. Here we show that Hic-5 binds to focal adhesion kinase (FAK) by its N-terminal domain, and is localized to focal adhesions by its C-terminal LIM domains. However, Hic-5 is not tyrosine phosphorylated either by the coexpressed FAK in COS cells or by integrin stimulation in 293T cells. Furthermore, overexpression of Hic-5 results in a decreased tyrosine phosphorylation of paxillin. These findings suggest that putative functions of Hic-5 are the recruitment of FAK to focal adhesions and a competitive inhibition of tyrosine phosphorylation of paxillin.

Integrin-mediated cell adhesion induces various biological events such as cell proliferation, survival, migration, cytokine production, and cytoskeleton reorganization (1)(2)(3)(4)(5)(6)(7). Despite the significance of these events, signal transduction pathways of integrins have not been fully understood. One of the approaches for integrin-mediated signal transduction is to study the proteins that are localized to focal adhesions.
Hydrogen peroxide-inducible clone (hic)-5 was initially identified as a gene which was induced by transforming growth factor ␤1 or by H 2 O 2 (19). The expression of hic-5 mRNA was augmented in the in vitro senescent process of human diploid fibroblasts (19). The forced overexpression of Hic-5 induced growth retardation, senescence-like morphology such as the enlarged and flattened morphology, and the increased expression of p21/WAF1/Cip1/sdi1 and extracellular matrix-related proteins such as fibronectin (FN), collagen, and collagenase (20). These findings suggest that Hic-5 is involved in very interesting function in the senescent process and in transforming growth factor ␤1-mediated signal transduction. However, the nature of Hic-5 function has not been clarified yet. Structurally, Hic-5 contains four LIM domains in the C-terminal half and LD domains in the N-terminal half (19,21). These domains are conserved in paxillin, and the four LIM domains of Hic-5 and paxillin share 62% homology (17, 19 -21). These structural homology with paxillin suggested that Hic-5 is localized to focal adhesions and involved in integrin-mediated signal transduction. Recently, Matsuya et al. (21) reported that Hic-5 bound to Cak ␤, a FAKrelated tyrosine kinase, and was localized to focal adhesions.
In this study, we demonstrate that Hic-5 is localized to focal adhesions by its LIM domains and binds to the FAK C-terminal domain by its N-terminal half. However, unlike paxillin, Hic-5 is not tyrosine phosphorylated by FAK and by integrin stimulation, suggesting that Hic-5 and paxillin have a different function in integrin-mediated protein tyrosine phosphorylation.

MATERIALS AND METHODS
Cell Culture and Transfection-COS1, 293T, Swiss 3T3, 3Y1, and SR-3Y1 cells were described previously (16,22,23). Bosc 23 was obtained from ATCC (24). Cells were cultured in RPMI 1640 or Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated fetal calf serum and 20 M gentamicin. Plasmid DNAs were transfected into cells using LipofectAMINE (Life Technologies, Inc.) following the manufacturer's protocol. Usually, 4 g of DNA were used for a 10-cm dish. In infection studies, Swiss 3T3 cells were incubated in the super-* This work was supported by National Institutes of Health Grants AR33713 and AI29530. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Plasmid DNAs-The oligonucleotides of the c-Myc epitope (MEQKLISEEDL) was jointed to the cDNAs of mouse hic-5 and human paxillin by polymerase chain reaction-mediated method (23). Human paxillin cDNA was kindly provided by Dr. Ravi Salgia and Dr. James D. Griffin (Dana-Faber Cancer Institute, Boston, MA). c-Myc-tagged hic-5 and paxillin were inserted into an expression vector, pMT3. Hemagglutinin (HA) epitope-tagged wild-type FAK and mutants were described elsewhere (23). The N-terminal half of the hic-5 cDNA was inserted into pGEX (Pharmacia LKB, Uppsala, Sweden). The other plasmids used to generate the glutathione S-transferase (GST) fusion proteins of FAK were described before (16). The DNA fragment coding enhanced green fluorescent protein (EGFP, CLONTECH, Palo Alto, CA) was jointed to the hic-5 cDNA and inserted into a retrovirus vector pBABE (25).
Antibodies, Precipitations, and Immunoblotting-The GST fusion protein of the Hic-5 N-terminal half (residues 2 to 249) was immunized to a rabbit to generate anti-Hic-5 polyclonal antibody (pAb). Rabbit anti-Hic-5 pAb was purified from anti-sera with protein A-conjugated beads (Pharmacia LKB) for immunoblotting and immunoprecipitation, and was affinity purified with the GST-Hic-5 fusion protein for immunohistochemistry. Cells were lyzed in 1% Nonidet P-40 lysis buffer as described (16). Proteins were precipitated with Ab plus protein-A conjugated beads or with glutathione beads (Pharmacia LKB) conjugated with GST-FAK fusion protein. Cellular lysates and precipitates were fractionated by SDS-polyacrylamide gel electrophoresis and electrotransfered onto nitrocellulose membranes. Immunoblotting was performed with a primary Ab, horseradish peroxidase-conjugated secondary Ab (Amersham Corp.), and chemiluminescence reagent (NEN Life Science Products Inc.), or with 125 iodine-labeled anti-phosphotyrosine mAb (4G10, Upstate Biotechnology, Inc.) as described (22,23). The relative intensity of each band was digitalized by scanning films with Alpha Imager 2000 (Alpha Innotech Corp., San Leangro, CA).
Immunohistochemistry-3Y1 cells were incubated on glass coverslips for 2 days. Cells were washed with phosphate-buffered saline (PBS), immersed, and fixed in the fixation solution (3.7% paraformaldehyde, 0.1% Triton X-100, in phosphate-buffered saline), and immunostained with primary Ab and then fluorescein-conjugated anti-mouse or antirabbit Ab (KPL, Gaithersburg, MD). After infection of retroviruses, Swiss 3T3 cells were cultured on glass coverslips for 2 days, and then washed and fixed. Subcellular localization of proteins were analyzed with fluoromicroscopy (Axioskop; Carl Zeiss Inc., Thornwood, NY) as described (16).

RESULTS
We reported FAK-paxillin interaction and the paxillin-mediated focal adhesion targeting of FAK (16). Because Hic-5 has a similar structure as paxillin, we studied FAK-Hic-5 interaction and the subcellular localization of Hic-5. To define the FAK-Hic-5 interaction, HA epitope-tagged FAK was coexpressed with c-Myc epitope-tagged Hic-5 or paxillin in COS1 cells. Then, Hic-5 or paxillin was immunoprecipitated with anti-c-Myc mAb, and the coprecipitation of FAK was investigated by immunoblotting with anti-HA mAb. As shown in Fig. 1A, HA- tagged wild-type FAK was coimmunoprecipitated with Hic-5 and paxillin. A similar result was obtained when HA-tagged FAK C-terminal domain (FAK-CT) was coexpressed with Hic-5 or paxillin. Therefore, like paxillin, Hic-5 can bind to the Cterminal domain of FAK.
Paxillin binds to the C-terminal domain of FAK by its Nterminal half. To determine the FAK-binding site of Hic-5, the c-Myc-tagged N-terminal and C-terminal halves of Hic-5 were expressed in COS1 cells and precipitated with the GST fusion protein of the FAK C-terminal domain (GST-FAK-CT). As shown in Fig. 1B, the wild-type and N-terminal half of Hic-5 were precipitated with GST-FAK-CT, whereas the C-terminal half of Hic-5 was not precipitated with GST-FAK-CT. Taken together, the N-terminal domain of Hic-5 can bind to the FAK-CT. We next defined the binding specificity of FAK to Hic-5 and paxillin. c-Myc-tagged Hic-5 and paxillin were expressed in COS1 cells, precipitated with various mutants of GST-FAK, and detected with immunoblotting with anti-c-Myc mAb. As shown in Fig. 1C, both Hic-5 and paxillin were precipitated with GST-FAK (903-1052) and its substitution mutants, K933E and Q1040G. However, neither protein were precipitated with GST and GST-FAK mutants, V928G and L1034S. These results indicate that Hic-5 and paxillin share a very similar FAK binding specificity.
For further analysis, we developed rabbit anti-Hic-5 polyclonal Ab (pAb) and used this antibody to detect endogenous Hic-5 protein. As shown in Fig. 2A, a 54-kDa peptide was detected by this anti-Hic-5 pAb in the lysate of rat fibroblast, 3Y1. Since this peptide was migrated to the same size as c-Myc-tagged mouse Hic-5 that was expressed in COS-1 cells by transfection, this 54-kDa peptide represented endogenous Hic-5 in 3Y1 cells. A protein with a similar molecular weight was detected in mouse fibroblasts, Swiss 3T3 cells, although the expression level of this protein was lower than in 3Y1. In Src-transformed 3Y1 cells, a slightly slower migrated protein in addition to this 54-kDa protein was detected. In contrast, a 56-kDa peptide was detected by this anti-Hic-5 pAb in the lysate of human kidney-derived 293T cells, and in African green monkey kidney-derived COS1 cells. p68 paxillin, a Hic-5 related protein, was not detected by this anti-Hic-5 pAb, although paxillin was expressed in these cells as shown by reblotting with anti-paxillin mAb.
Using this anti-Hic-5 pAb, we studied the in vivo association of Hic-5 and FAK. Endogenous Hic-5 was immunoprecipitated by anti-Hic-5 pAb from the lysate of 293T cells. Then, precipitates were analyzed by immunoblotting with anti-FAK mAb. As shown in Fig. 2B, endogenous FAK was coprecipiated with endogenous Hic-5, indicating the FAK-Hic-5 interaction in vivo.
For further analysis, we studied the subcellular localization of Hic-5. Since FAK is localized to focal adhesions, Hic-5 should be localized to focal adhesions if Hic-5 binds to FAK in vivo. We determined the subcellular localization of Hic-5 in 3Y1 cells by immunohistochemical analysis with anti-Hic-5 pAb. As shown in Fig. 3A, Hic-5 was demonstrated as rod-shaped staining in cytoplasm, mostly in the periphery of a cell, with a direction of peripheral to center of a cell (arrows). This staining of Hic-5 was similar as the staining of vinculin which was shown as typical focal adhesions, and was located at the identical sites with vinculin in double staining, indicating that Hic-5 is localized to focal adhesions in 3Y1 cells. Next, we wished to determine the Hic-5 domain which is responsible for the focal adhesion targeting. Since Brown et al. (17) reported that paxillin is localized to focal adhesions by its C-terminal LIM domains, it is likely that the Hic-5 C-terminal LIM domains are essential for the focal adhesion targeting of Hic-5. To determine the focal adhesion targeting domain of Hic-5, we expressed EGFPtagged wild-type Hic-5 and its deletion mutants in Swiss 3T3 cells and analyzed their subcellular localization under fluorescent microscopy. As shown in Fig. 3B, the EGFP-tagged wildtype and C-terminal half of Hic-5 were detected at focal adhesions, whereas EGFP alone and the EGFP-tagged Hic-5 N-terminal half were localized diffusely and were not detected at focal adhesions. Thus, Hic-5 is localized to focal adhesions by its C-terminal LIM domains. However, the EGFP-tagged wildtype and N-terminal half of Hic-5 were localized outside of nuclei, whereas EGFP alone and the EGFP-tagged C-terminal domain were localized both inside and outside of nuclei. Therefore, the N-terminal domain of Hic-5 appears to be involved in the cytoplasmic localization of Hic-5.
Thus, Hic-5 and paxillin share not only structural homology, but also several characteristics such as binding activity to FAK and subcellular localization to focal adhesions. However, Hic-5 does not contain several tyrosine residues in paxillin, which are phosphorylated by integrin stimulation and putative binding sites for the Crk SH2 domain. Therefore, we next studied whether Hic-5 is tyrosine phosphorylated by integrin stimulation or not. c-Myc-tagged Hic-5 and paxillin were coexpressed with HA-tagged FAK, immunoprecipitated, and analyzed by immunoblotting with anti-phosphotyrosine mAb (anti-Tyr(P)). As shown in Fig. 4A, paxillin was tyrosine phosphorylated when coexpressed with FAK wild-type or Y397F mutant. Since paxillin was not tyrosine phosphorylated when coexpressed with FAK-kinase negative, this phosphorylation of paxillin was dependent on the kinase activity of the coexpressed FAK. Ty- rosine phosphorylation of paxillin by FAK-Y397F was less compared with that by wild-type FAK. These results were consistent with the reports of Schaller and Parsons (15) and Bellis et al. (18), and indicated tyrosine phosphorylation of paxillin by FAK and a FAK-associated kinase. Unlike paxillin, Hic-5 was not tyrosine phosphorylated when coexpressed with FAK. These data strongly suggest that Hic-5 is not a substrate of FAK, whereas paxillin is a substrate of both FAK and a tyrosine kinase(s) that binds to phosphorylated FAK-Y397.
Next, we investigated tyrosine phosphorylation of Hic-5 by integrin stimulation. c-Myc-tagged Hic-5 or paxillin were expressed in 293T cells by transfection. These 293T cells were detached from substrata, incubated on the plates coated with FN or poly-L-lysine (Sigma), and then lyzed in 1% Nonidet P-40 lysis buffer. Hic-5 and paxillin were immunoprecipitated from cellular lysates with anti-c-Myc mAb and analyzed by immunoblotting with anti-phosphotyrosine mAb. As shown in Fig.  4B, paxillin was tyrosine phosphorylated following integrin stimulation, whereas no increased tyrosine phosphorylation of Hic-5 was observed upon FN stimulation. We also analyzed integrin-mediated tyrosine phosphorylation of endogenous Hic-5 and paxillin in 293T cells. As shown in Fig. 4C, paxillin was tyrosine phosphorylated following FN stimulation. In contrast, we observed no increased tyrosine phosphorylation of endogenous Hic-5, further indicating that Hic-5 is not tyrosine phosphorylated following integrin stimulation.
The expression of Hic-5 is augmented in a senescent process in human fibroblasts. To elucidate a putative function of Hic-5, we studied integrin-mediated tyrosine phosphorylation of paxillin in the presence or absence of Hic-5. c-Myc-tagged paxillin and Hic-5 were coexpressed in 293T cells. Then, cells were detached and stimulated by the plates coated with FN. Paxillin and Hic-5 were immunoprecipitated with anti-c-Myc mAb and analyzed by immunoblotting with anti-phosphotyrosine mAb. As shown in Fig. 4D, FN-induced tyrosine phosphorylation of paxillin was decreased by the coexpression of Hic-5 in a dosedependent manner. This result suggests that overexpression of Hic-5 inhibits the signal from tyrosine-phosphorylated paxillin. DISCUSSION In this article, we have demonstrated that Hic-5, a senescence-related protein, is a cytoskeletal protein localized to focal adhesions. Similar to a related protein, paxillin, Hic-5 is localized to focal adhesions by its C-terminal LIM domains and binds to FAK by its N-terminal domain. Therefore, Hic-5 appears to be involved in the recruitment of FAK to focal adhesions and involved in integrin signal transduction.
However, unlike paxillin, Hic-5 is not tyrosine phosphorylated by integrin stimulation, or by the coexpression of FAK. Furthermore, overexpression of Hic-5 inhibited integrin-mediated tyrosine phosphorylation of paxillin. Gilmore and Romer (26) reported that microinjection of GST-FAK-CT resulted in reduced focal adhesion phosphotyrosine and reduced DNA synthesis. GST-FAK-CT contains the focal adhesion targeting domain, but does not contain a kinase domain. Therefore, GST-FAK-CT inhibits tyrosine phosphorylation of focal adhesion proteins by preventing endogenous FAK localization to focal adhesions and the binding of FAK to focal adhesion proteins. These findings strongly suggest that prevention of tyrosine phosphorylation of focal adhesion proteins, such as paxillin, inhibits cell proliferation.
The phosphotransfer sites of paxillin are putative binding sites for the SH2 domain of Crk, an oncogenic adapter protein (15,18,27). Crk is composed of SH2 and SH3 domains (27). Crk binds to tyrosine-phosphorylated proteins by its SH2 domain and binds to the effector proteins by its SH3 domains (28,29). Thus, Crk is involved in the recruitment of signaling molecules, such as C3G and SOS, to tyrosine-phosphorylated docking proteins (28,29). Since paxillin is localized at focal adhesions, the binding of Crk to tyrosine-phosphorylated paxillin results in the recruitment of these signaling molecules to focal adhesions that are localized close to cytoplasmic membrane. C3G and SOS are guanine nucleotide exchange factors for Rap1A/k-Rev1 and Ras, respectively (30,31). These small GTPases are involved in the regulation of mitogen-activated protein (MAP) kinases (32,33). Inhibition studies using Crk mutants showed that Crk is involved in the regulation of MAP kinase (34,35). v-Crk and C3G are also involved in the activation of c-Jun N-terminal kinase, another MAP kinase family kinase (36). Taken together, tyrosine phosphorylation of paxillin may be involved in the activation of MAP kinase family kinases through the recruitment of Crk. Since activation of MAP kinase family kinases regulates gene expression and cell proliferation, the overexpression of Hic-5 may induce growth retardation by blocking activation of MAP kinases.
Inhibition of tyrosine phosphorylation of focal adhesion proteins causes alterations in cell shape and a decrease in cell migration. FAK knock-out or the overexpression of FAK-CT resulted in enlarged focal adhesions and reduced cell motility (37,38). More directly, Klemke et al. (39) reported that the binding of Crk to another focal adhesion protein, Crk-associated substrate (p130 Cas ), induces cell migration and membrane ruffling (39). We have observed a similar phenomenon in Cas-L transfected cells. 2 These findings suggest that the recruitment of Crk to paxillin can induce membrane ruffling and cell migration. Therefore, parts of senescent phenotypes, such as the enlarged and flattened morphology of cells, may be caused by the inhibition of tyrosine phosphorylation of paxillin. FIG. 4. Analysis of tyrosine phosphorylation of Hic-5. A, Hic-5 was not tyrosine phosphorylated by the coexpression of FAK. c-Myc-tagged Hic-5 and paxillin were coexpressed with HA-tagged wild-type, Y397F, or K454R (kinase negative) FAK in COS1 cells. Tyrosine phosphorylation of Hic-5 and paxillin was analyzed by immunoprecipitation with anti-c-Myc mAb and immunoblotting with anti-phosphotyrosine mAb (anti-Tyr(P)). The amounts of the precipitated Hic-5, paxillin, and coprecipitated FAK were demonstrated by immunoblotting. B, Hic-5 was not tyrosine phosphorylated following integrin stimulation. c-Myc-tagged Hic-5 and paxillin were expressed in 293T cells. Cells were detached, and then reattached to plates coated with FN or poly-L-lysine (PLL, Sigma). Hic-5 and paxillin were isolated by immunoprecipitation with anti-c-Myc mAb and analyzed by immunoblotting with anti-Tyr(P). C, endogenous Hic-5 was not tyrosine phosphorylated following integrin stimulation. Endogenous Hic-5 or paxillin were immunoprecipitated with anti-Hic-5 pAb or anti-paxillin mAb from the lysates of FN-or poly-L-lysine-stimulated 293T cells, and analyzed by immunoblotting with anti-Tyr(P). Both blots were from one membrane, and the anti-Tyr(P) blots were of the same exposure. D, inhibition of integrin-mediated tyrosine phosphorylation of paxillin by the co-expression of Hic-5. pMT3-c-Myc-tagged paxillin (0.25 g/10 cm-plate) was co-transfected with various amounts (0 -10 g) of pMT3-c-Myc-tagged hic-5 into 293T cells. Transfected cells were detached, incubated on FN-or poly-L-lysine-coated plates, and lyzed. c-Myc-tagged paxillin was immunoprecipitated and subjected to the analysis with anti-Tyr(P).
In summary, our study, along with the report of Matsuya et al. (21), addressed that Hic-5 has two distinct functions in integrin-mediated signal transduction, the recruitment of FAK and Cak ␤ to focal adhesions and the inhibition of tyrosine phosphorylation of paxillin. Since integrin-mediated cell adhesion induces cell proliferation and survival, overexpression of Hic-5 may induce senescence-like phenotypes in human diploid fibroblasts by modulating integrin-mediated signal transduction.
We also found that paxillin was tyrosine phosphorylated by the coexpression of FAK-Y397F in COS1 cells. Since Src family tyrosine kinases do not bind to FAK-Y397F (40), this result indicates that FAK can directly phosphorylate paxillin. We reported previously that Crk-associated substrate (p130 Cas ) and Cas-L are substrates of FAK (23). Tyrosine-phosphorylated Cas family proteins also bind to Crk (22,41). Based on these findings, we propose the following as one of integrin signaling pathways: upon integrin-mediated cell adhesion, focal adhesion proteins such as paxillin and Cas family proteins are tyrosine phosphorylated in a FAK-dependent manner. This leads to the recruitment of Crk to tyrosine-phosphorylated focal adhesion proteins, alternately resulting in integrin-mediated biological signals.