Identification of a Novel Integrin α6β1 Binding Site in the Angiogenic Inducer CCN1 (CYR61)*

The angiogenic inducer CCN1 (cysteine-rich 61, CYR61), a secreted matricellular protein of the CCN family, is a ligand of multiple integrins, including α6β1. Previous studies have shown that CCN1 interaction with integrin α6β1 mediates adhesion of fibroblasts, endothelial cells, and smooth muscle cells, as well as migration of smooth muscle cells. Recently, we have reported that CCN1-induced tubule formation of unactivated endothelial cells is also mediated through integrin α6β1. In this study, we demonstrate that human skin fibroblasts adhere specifically to the T1 sequence (GQKCIVQTTSWSQCSKS) within domain III of CCN1, and this process is blocked by anti-α6 and anti-β1 monoclonal antibodies. Alanine substitution mutagenesis of the T1 sequence further defines the sequence TTSWSQCSKS as the critical determinant for mediating α6β1-dependent adhesion. Soluble T1 peptide specifically inhibits fibroblast adhesion to CCN1 in a dose-dependent manner. Furthermore, T1 also inhibits cell adhesion to other α6β1 ligands, including CCN2 (CTGF), CCN3 (NOV), and laminin, but not to ligands of other integrins. In addition, T1 specifically inhibits α6β1-dependent tubule formation of unactivated endothelial cells in a CCN1-containing collagen gel matrix. To confirm that T1 binds integrin α6β1 directly, we perform affinity chromatography and show that integrin α6β1 is isolated from an octylglucoside extract of fibroblasts on T1-coupled Affi-gel. Taken together, these findings define the T1 sequence in CCN1 as a novel binding motif for integrin α6β1, providing the basis for the development of peptide mimetics to examine the functional role of α6β1 in angiogenesis.

From the Departments of ‡Molecular Genetics and §Pharmacology, University of Illinois College of Medicine, Chicago, Illinois 60607-7170 The angiogenic inducer CCN1 (cysteine-rich 61, CYR61), a secreted matricellular protein of the CCN family, is a ligand of multiple integrins, including ␣ 6 ␤ 1 . Previous studies have shown that CCN1 interaction with integrin ␣ 6 ␤ 1 mediates adhesion of fibroblasts, endothelial cells, and smooth muscle cells, as well as migration of smooth muscle cells. Recently, we have reported that CCN1-induced tubule formation of unactivated endothelial cells is also mediated through integrin ␣ 6 ␤ 1 . In this study, we demonstrate that human skin fibroblasts adhere specifically to the T1 sequence (GQKCIVQTTSWSQCSKS) within domain III of CCN1, and this process is blocked by anti-␣ 6 and anti-␤ 1 monoclonal antibodies. Alanine substitution mutagenesis of the T1 sequence further defines the sequence TTSWSQC-SKS as the critical determinant for mediating ␣ 6 ␤ 1 -dependent adhesion. Soluble T1 peptide specifically inhibits fibroblast adhesion to CCN1 in a dose-dependent manner. Furthermore, T1 also inhibits cell adhesion to other ␣ 6 ␤ 1 ligands, including CCN2 (CTGF), CCN3 (NOV), and laminin, but not to ligands of other integrins. In addition, T1 specifically inhibits ␣ 6 ␤ 1 -dependent tubule formation of unactivated endothelial cells in a CCN1-containing collagen gel matrix. To confirm that T1 binds integrin ␣ 6 ␤ 1 directly, we perform affinity chromatography and show that integrin ␣ 6 ␤ 1 is isolated from an octylglucoside extract of fibroblasts on T1-coupled Affi-gel. Taken together, these findings define the T1 sequence in CCN1 as a novel binding motif for integrin ␣ 6 ␤ 1 , providing the basis for the development of peptide mimetics to examine the functional role of ␣ 6 ␤ 1 in angiogenesis.
CCN1 1 (cysteine-rich 61, CYR61), an angiogenic inducer encoded by a growth factor-inducible immediate-early gene, is a novel integrin ligand whose expression is essential for proper embryonic development. Recent studies by targeted disruption of the CCN1 gene in mice show that CCN1-null embryos suffer embryonic death due primarily to vascular defects in both the placenta and the embryo (1). In addition to embryonic angiogenesis, CCN1 may also promote pathological angiogenesis under such conditions as tumor growth and wound healing. Stable transfection of CCN1 in tumor cell lines that do not otherwise express CCN1 enhances tumorigenicity with an increased vascularization of the CCN1-expressing tumors (2)(3)(4). Furthermore, estrogen-induced CCN1 expression has been associated with advanced human breast cancer (5)(6)(7). Overexpression of CCN1 has also been observed in restenosed blood vessels and advanced atherosclerotic lesions, underscoring its pathological importance in vascular diseases (8 -11). In addition, the expression of CCN1 in cutaneous healing wounds, coupled with its ability to activate a genetic program for wound healing in human skin fibroblasts, suggests an important role for CCN1 in injury repair (12,13).
Upon synthesis, CCN1 is secreted and becomes associated with the cell surface or the extracellular matrix (14). Previous studies have shown that CCN1 supports cell adhesion, induces cell migration, enhances growth factor-induced mitogenesis, and promotes cell survival under apoptotic conditions (15,16). These cellular activities of CCN1 can be attributed in part to its ability to interact with integrin adhesion receptors. To date, five integrins, ␣ 6 ␤ 1 , ␣ v ␤ 3 , ␣ v ␤ 5 , ␣ IIb ␤ 3 , and ␣ M ␤ 2 , have been identified as CCN1 receptors in various cell types (11,(17)(18)(19)(20). In an earlier study, we have demonstrated that CCN1 induces neovascularization in the rat corneal micropocket assay (2). Consistent with these in vivo findings, CCN1 promotes tubule formation of HUVECs in a collagen gel assay, and this process is dependent on integrins ␣ 6 ␤ 1 and ␣ v ␤ 3 (16).
Integrin ␣ 6 ␤ 1 has been shown to mediate a number of CCN1 activities in several cell types. CCN1 supports fibroblast adhesion through interaction with integrin ␣ 6 ␤ 1 and cell surface heparan sulfate proteoglycans, leading to extensive formation of filopodia and lamellipodia with ␣ 6 ␤ 1 -containing focal complexes localized at leading edges of the pseudopods (21). Moreover, integrin-dependent outside-in signaling is induced resulting in the activation of focal adhesion kinase, paxillin, Rac, and mitogen-activated protein kinases, and up-regulation of several angiogenic regulators, including vascular epidermal growth factor (13,21). In addition to fibroblasts, CCN1 also interacts with integrin ␣ 6 ␤ 1 on vascular smooth muscle cells and vascular endothelial cells (10,16). Recently, we have shown that pro-angiogenic activities of CCN1 are differentially mediated through integrins ␣ 6 ␤ 1 and ␣ v ␤ 3 in unactivated and activated HUVECs, respectively (16).
In addition to CCN1, other members of the CCN family include CCN2 (connective tissue growth factor, CTGF), CCN3 (nephroblastoma-overexpressed, NOV), and the Wnt-inducible secreted proteins CCN4 (WISP-1), CCN5 (WISP-2), and CCN6 (WISP-3) (22)(23)(24). CCN proteins contain an N-terminal secre-tory signal, followed by four distinct modular domains: 1) an insulin-like growth factor binding protein homology domain, 2) a von Willebrand factor type C repeat domain, 3) a thrombospondin type I repeat (TSP1) domain, and 4) a C-terminal domain with heparin binding motifs and sequence similarities to the C termini of von Willebrand factor and mucins (see Fig.  1A). Several CCN proteins have been shown to interact with multiple integrins (11, 17-21, 25, 26), and therefore, localization of the integrin binding sites in CCN proteins will provide new insights into the structure-function relationship of this newly established family of matricellular proteins (27,28). We previously found that a truncated CCN1 lacking the C-terminal domain is capable of inducing smooth muscle cell migration through integrin ␣ 6 ␤ 1 (19). These findings suggest that the one or more integrin ␣ 6 ␤ 1 binding sites reside within the first three domains of CCN1. In the present study, we identify a novel 17-residue sequence, designated T1, in the CCN1 thrombospondin type I repeat domain that mediates ␣ 6 ␤ 1 -dependent cell adhesion. By affinity chromatography, we demonstrate direct interaction of ␣ 6 ␤ 1 with the T1 sequence. Inasmuch as synthetic peptides derived from the T1 sequence specifically block ␣ 6 ␤ 1 -dependent cell adhesion, our newly identified ␣ 6 ␤ 1 binding site in CCN1 may serve as a basis for the development of antagonists to integrin ␣ 6 ␤ 1 .

MATERIALS AND METHODS
Proteins, Peptides, Antibodies, and Reagents-Recombinant murine CCN1 protein was purified from serum-free insect cell conditioned media using the baculovirus expression system as described previously (15). Rat Type I collagen, vitronectin, laminin, and fibronectin were purchased from Collaborative Biomedical (Bedford, MA). GRGDSP and GRGESP peptides and function-blocking monoclonal antibody (mAb) against integrin ␤ 1 (P4C10) were from Invitrogen. Function-blocking mAbs against integrin ␣ v ␤ 3 (LM609) and integrin ␣ 6 (GoH3) were from Chemicon (Temecula, CA) and Immunotech (Marseille, France), respectively. Horseradish peroxidase-conjugated secondary antibodies were obtained from Amersham Biosciences (Piscataway, NJ). Anti-polyhistidine mAb was from Invitrogen (Carlsbad, CA). Synthetic peptides corresponding to partial sequences of domain III (TSP1 domain) of CCN1 were prepared by ResGen Inc. (Huntsville, AL), followed by purification on reverse-phase high performance liquid chromatography and analysis by mass spectroscopy.
Cell Culture, Cell Adhesion, and Endothelial Tubule Formation-Primary human foreskin 1064SK fibroblasts (ATCC CRL-2076, passage 2) were kept in Iscove's modified Dulbecco's medium (Invitrogen) with 10% fetal bovine serum (Intergen, Purchase, NY) at 37°C with 5% CO 2 . Cells were used within the 5th to 20th passages for all experiments. Recombinant CCN1 domain fragments were produced as hexahistidine-tagged fusion proteins by a baculovirus-expression system and purified by chromatography on cobalt-agarose. A, a schematic representation of the structural domains of full-length CCN1 and the isolated domain fragments. The T1 sequence in domain III (TSP1 domain) is indicated by the shaded area. SP, signal peptide. B, recombinant CCN1 domain fragments and full-length CCN1 (2 g) were electrophoresed on 15% SDS-polyacrylamide gel and detected by Coomassie Blue staining. C, the resolved proteins were subjected to immunoblotting with polyclonal anti-CCN1 antibodies (15). Molecular mass markers are indicated in kDa on the left. D, maleic anhydride Reacti-Bind microtiter wells were coated with purified recombinant CCN1 domain fragments or BSA (50 g/ml and 50 l/well) overnight at 4°C and blocked with 1% BSA. Protein coating efficiency was detected by an ELISA using an anti-polyhistidine mAb. E, washed 1064SK fibroblasts, resuspended in serum-free medium, were plated onto wells (3 ϫ 10 4 cells per well) precoated with CCN1 (20 g/ml) or the indicated domain fragments (50 g/ml). Cells were allowed to adhere for 20 min at 37°C. Adherent cells were fixed, stained with methylene blue, and quantified by absorbance at 620 nm. Data are means Ϯ S.D. of triplicate determinations. D and E are representative of three experiments.
Test proteins were coated onto 96-well microtiter plates (BD Biosciences) in PBS (50 l per well), and wells were blocked with 1% BSA at room temperature for 1 h. To enhance coating efficiency, CCN1 domain polypeptides were covalently linked to maleic anhydride Reacti-Bind microtiter plates (Pierce, Rockford, IL) at 4°C overnight followed by blocking with 1% BSA at 37°C for 2 h. Cell adhesion was conducted using washed subconfluent cells resuspended in serum-free basal medium at 5 ϫ 10 5 cells/ml as described previously (21). Where indicated, cells were preincubated with EDTA, peptides, or function-blocking mAbs for 30 min prior to plating. To assay for endothelial cell tubule formation, human umbilical vein endothelial cells (HUVECs) were examined in a three-dimensional collagen gel in the presence or absence of test proteins or peptides as described (16).
Preparation of GST-peptide Fusion Proteins-The coding sequences for various peptides (Fig. 3) were amplified by PCRs upon the murine CCN1 cDNA as template. Primers used corresponded to the appropriate coding sequences and contained the BamHI and EcoRI restriction sites for cloning. For example, 5Ј-CGGGATCCGCGGGCCAGAAATGCATC-GTT-3Ј and 5Ј-CCGGAATTCCGCTCTTGGAGCACTGGGACC-3Ј were used to generate the T1 peptide coding sequence. PCR products were purified on polyacrylamide gels, digested with BamHI and EcoRI, and ligated into the pGEX-4T-2 vector (Amersham Biosciences). All cloning steps were confirmed by sequence analysis. To generate site-directed alanine substitutions for the T1 peptide ( Fig. 6), synthetic oligonucleotides were annealed to generate the appropriate coding sequences and cloned into pGEX-4T-2. The primers 5Ј-GATCCGGTCAAAAATGTAT-TGTTCAAACTACTTCTTGGTCTCAATGCTCTAAATCTGG-3Ј and 5Ј-AATTCCAGATTTAGAGCATTGAGACCAAGAAGTAGTAGTTTGAAC AATACATTTTTGACCG-3Ј were used to prepare the coding sequence for the T1 sequence and cloned into pGEX-4T-2. To create the mutant peptides, relevant codons were changed to either GCA or GCT for alanine. GST-peptide fusion proteins were produced in Escherichia coli strain BL21 and purified by glutathione affinity chromatography (Amersham Biosciences), followed by extensive dialysis against PBS overnight at 4°C.
Expression and Purification of CCN1 Domain Fragments-To enhance protein secretion, we employed a modified pBlueBac4.5/V5-His vector (Invitrogen) with the insect honeybee melittin secretory signal peptide engineered into the N terminus of the expressed protein. To produce the coding sequences for domain I (IGFBP), domain II (VWC), and domain III (TSP1), we used the primer sets 5Ј-CGCGGATCCGG-CGCTCTCCACCTGC-3Ј and 5Ј-GGAATTCCCTCTGCAGATCCCTTT-CAGAGCGG-3Ј, 5Ј-CGCGGATCCGGCTCAGTCAGAAGGCAGAC-3Ј and 5Ј-GGAATTCCCAGGAAGCCTCTTCAGTGAGCTGCC-3Ј, and 5Ј-CGCGGATCCGGTCTTTGGCACC-3Ј and 5Ј-GGAATTCCCTTTTAGG-CTGCTGTACACTGGTTGTC-3Ј, respectively, for PCR upon the CCN1 cDNA. The PCR products were digested with BamH1 and EcoR1, and ligated into the vector. Each expressed recombinant polypeptide contained the V5 epitope and a polyhistidine tag at the C terminus, and was purified from Sf9 cells using a serum-free baculovirus expression system as described (19). Briefly, cells were maintained under serumfree conditions in EX-CELL 400 medium (JRH Biosciences, Lenexa, KS), infected at a multiplicity of infection of 10, and collected 42-46 h post-infection. The collected medium was cleared by centrifuge and subsequently concentrated by 10-to 15-fold using the Biomax-5 centrifugal filter (Millipore, Billerica, MA) and dialyzed against native buffer (50 mM sodium phosphate and 10 mM Hepes at pH 7.4, 0.5 M NaCl) overnight at 4°C and then applied to a Talon cobalt-agarose column (Clontech, Palo Alto, CA). The column was washed with native binding buffer at pH 7.0, before being eluted in 50 mM phosphate at pH 7.0, 0.3 M NaCl, and 150 mM imidazole. Products were analyzed by SDS-PAGE followed by Coomassie Brilliant Blue staining and immunoblotting. The pooled fractions for each domain fragment were dialyzed against 20 mM Hepes at pH 7.4, 150 mM NaCl overnight at 4°C to remove imidazole.
ELISA-Microtiter wells were coated with hexahistidine-tagged proteins or BSA (50 l/well) overnight at 4°C, followed by blocking with 1% Fibroblast adhesion to microtiter wells coated with full-length CCN1 (20 g/ml) or domain III fragment (50 g/ml) was performed as described in the legend of Fig. 1. A, where indicated, cells were suspended in serum-free medium containing EDTA (2.5 mM), Mg 2ϩ (5 mM), Ca 2ϩ (5 mM), or Mn 2ϩ (0.5 mM) before plating. B, cells were preincubated with vehicle buffer (No Add), normal mouse IgG (100 g/ml), anti-␣ v ␤ 3 mAb LM609 (50 g/ml), anti-␣ 6 mAb GoH3 (50 g/ml), or anti-␤ 1 mAb P4C10 (1:50 ascites) for 60 min prior to plating. Data are means Ϯ S.D. of triplicate determinations and are representative of three experiments. BSA for 2 h at room temperature. Protein coating efficiency was examined by incubation with an anti-polyhistidine mAb (2 h at 37°C) followed by a horseradish peroxidase-conjugated secondary antibody (1 h at 37°C). The color reaction was developed and quantified by A 420 measurements (Zymed Laboratories Inc., San Francisco, CA).
Cell Surface Radioiodination of Fibroblasts and Affinity Chromatography on T1-coupled Agarose-Subconfluent 1064SK fibroblasts were detached with 2 mM EDTA and 0.05% BSA in PBS, washed twice and resuspended in PBS containing 20 mM glucose at 2 ϫ 10 7 cells/ml. For surfacing labeling (29,30), the cell suspension was incubated with 100 millunits/ml glucose oxidase, 200 g/ml lactoperoxidase (Calbiochem-Novabiochem, La Jolla, CA), and ϳ400 Ci/ml carrier-free Na 125 I (Amersham Biosciences) for 30 -60 min at 4°C with gentle rotation. To terminate labeling, 10 ml of cell culture medium was added. The labeled cells were washed and solubilized in 1 ml of lysis buffer (50 mM Hepes, pH 7.4, 200 mM octyl-␤-D-glucopyranoside, proteinase inhibitor mixture, and 0.5 mM Mn 2ϩ ). For affinity chromatography, GST-T1 or GSTscrambled T1 protein was coupled to Affi-Gel 10 (Bio-Rad Laboratories, Hercules, CA) at 10 mg/ml gel suspension. The labeled cell lysates were applied onto the affinity matrices (3 ml/ml gel) and incubated for 2 h at 4°C. The columns were washed with 30 column volumes of lysis buffer followed by elution with 0.35 M NaCl in the lysis buffer. The labeled proteins in the eluted fractions were analyzed by electrophoresis on 7% polyacrylamide gels under non-reducing conditions followed by autoradiography. In immunoprecipitation analyses, labeled proteins were incubated with 5 g of anti-␣ 6 (GoH3) or anti-␣ v (P3G8) mAbs (Chemicon, Temecula, CA) as indicated. The immunoprecipitated proteins were collected on protein G-Sepharose and resolved on 7% polyacrylamide gels under non-reducing conditions.

RESULTS
Domain III (TSP1 Homology Domain) of CCN1 Supports ␣ 6 ␤ 1 -dependent Cell Adhesion-Previous studies have established that primary human skin fibroblasts adhere to CCN1 through integrin ␣ 6 ␤ 1 and heparan sulfate proteoglycans, inducing the formation of ␣ 6 ␤ 1 -containing focal complexes and the activation of focal adhesion kinase, paxillin, and Rac (20,21). Deletion analysis showed that a C-terminal truncated CCN1 mutant containing only the first three domains retains the ability to induce chemotaxis in smooth muscle cells through integrin ␣ 6 ␤ 1 , thus localizing an integrin ␣ 6 ␤ 1 binding site within the first three domains (10). To define the CCN1 structural domain that interacts with integrin ␣ 6 ␤ 1 , we expressed each of these three domains in insect cells via a baculovirus vector (Fig. 1A). Each polypeptide is endowed with an N-terminal secretory signal and a C-terminal polyhistidine tag. The expressed polypeptides were purified to apparent homogeneity from conditioned insect cell media by cobalt-agarose affinity chromatography. Each domain fragment had the expected molecular mass (ϳ11 kDa, 18 kDa, and 9 kDa for domains I, II, and III, respectively) and were immunoreactive with an anti-CCN1 polyclonal antibody (Fig. 1, B and C).
We employed human 1064SK fibroblasts to address the ability of each domain to support cell adhesion. Whereas all three domains were coated onto microtiter wells with similar efficiency, only Domain III was able to support fibroblast adhesion (Fig. 1, D and E). Fibroblast adhesion to Domain III was inhibited by EDTA (2.5 mM), and this inhibition was relieved by the addition of Mg 2ϩ (5 mM) in the assay media ( Fig. 2A). Cell adhesion was also inhibited by Ca 2ϩ (5 mM) and promoted by Mn 2ϩ (0.5 mM). This divalent cation sensitivity profile is similar to that of full-length CCN1 and is consistent with cell adhesion through integrin ␣ 6 ␤ 1 (20). To ascertain which specific integrin receptor mediated cell adhesion to Domain III, we tested the inhibitory attributes of function-blocking mAbs. Preincubation of fibroblasts with mAbs against ␣ 6 (GoH3) or ␤ 1 (P4C10) obliterated cell adhesion to Domain III as well as full-length CCN1, whereas mAb against integrin ␣ v ␤ 3 (LM609) or control mouse IgG had no effect (Fig. 2B). Together, these results show that human skin fibroblasts adhesion to the isolated Domain III of CCN1, like adhesion to full-length CCN1, is mediated through integrin ␣ 6 ␤ 1 .
The T1 Sequence in Domain III of CCN1 Contains an Integrin ␣ 6 ␤ 1 Binding Site-We employed another systematic screening strategy to pinpoint the integrin ␣ 6 ␤ 1 binding site in CCN1. A series of overlapping peptides (Table I) that covers the entire first three domains of CCN1 was prepared by expression of the peptides as fusion proteins linked to GST. These fusion proteins were purified to near homogeneity and had similar levels of coating efficiency in microtiter wells as detected by ELISA using an anti-GST antibody (data not shown). The ability of each peptide-GST fusion protein to support fibroblast adhesion was assessed. Only one peptide fusion protein, namely T1 from Domain III, was able to support cell adhesion  Table I. Protein coating was performed overnight at 4°C followed by blocking with 1% BSA. Fibroblast adhesion was assessed as described in the legend of Fig. 1. B, cells were suspended in serum-free medium containing EDTA (2.5 mM), Mg 2ϩ (5 mM), Ca 2ϩ (5 mM), or Mn 2ϩ (0.5 mM) and plated onto the microtiter wells coated with GST (50 g/ml), GST-T1 (50 g/ml), or CCN1 (1 g/ml). C, cells were preincubated with vehicle buffer (No Add), normal mouse IgG (100 g/ml), anti-␣ v ␤ 3 mAb LM609 (50 g/ml), anti-␣ 6 mAb GoH3 (50 g/ml), or anti-␤ 1 mAb clone P4C10 (1:50 ascites) for 60 min prior to plating. Data are means Ϯ S.D. of triplicate determinations and are representative of three experiments. (Fig. 3A). Again, fibroblast adhesion to T1-GST was inhibited by EDTA and Ca 2ϩ and promoted by Mn 2ϩ in the assay media (Fig. 3B). Also, cell adhesion to T1-GST was blocked by preincubation of cells with anti-␣ 6 (GoH3) or anti-␤ 1 (P4C10) mAb but unaffected by other integrin-disrupting agents such as GRGDSP peptide or anti-␣ v ␤ 3 (LM609) (Fig. 3C), indicating that the T1-GST fusion protein supports the adhesion of fibroblasts through integrin ␣ 6 ␤ 1 . Similar results were obtained using a synthetic T1 peptide in place of the T1-GST fusion protein as the adhesive substrate (data not shown). Additionally, T1-GST also supported ␣ 6 ␤ 1 -mediated cell adhesion in other cell types, including endothelial cells, smooth muscle cells, and PC3 prostate cancer cells (data not shown).
Soluble T1 Peptide Inhibits ␣ 6 ␤ 1 -dependent Cell Adhesion-To establish further that the T1 sequence contains a binding site for integrin ␣ 6 ␤ 1 , we synthesized four peptides spanning the CCN1 domain III (Table I) and tested their abilities to inhibit cell adhesion mediated through integrin ␣ 6 ␤ 1 . As shown in Fig. 4A, addition of 0.2 mM T1 to the cell suspension effectively blocked fibroblast adhesion to CCN1, whereas T2, T3, or T4 had no effect. The inhibitory effect of T1 on cell adhesion to CCN1 was dose-dependent, achieving maximal inhibition at 100:9 (Fig. 4C). Other members of the CCN protein family, CCN2 (CTGF) and CCN3 (NOV), have also been shown to support fibroblast adhesion through integrin ␣ 6 ␤ 1 (21,26), and a high degree of homology exists among the corresponding T1 sequences in these CCN proteins. Fig. 4A shows that T1 also specifically inhibited cell adhesion to CCN2 and CCN3, suggesting that the T1 sequence in CCN proteins is a common binding site for integrin ␣ 6 ␤ 1 .
To demonstrate further the specificity of T1 inhibition, we examined its ability to block cell adhesion to substrates that bind other integrins. In contrast to its dose-dependent inhibitory effect of cell adhesion to CCN1 (Fig. 4C), T1 had no significant effect on the adhesion of fibroblasts to fibronectin (ligand of integrin ␣ 5 ␤ 1 ), vitronectin (ligand of ␣ v integrins), and collagen (ligand of ␤ 1 integrins) (Fig. 4B). Cell adhesion to laminin, a known ligand for integrin ␣ 6 ␤ 1 , was partially inhibited by the T1 peptide (ϳ15%). This partial inhibition was similar to that achieved by the anti-␣ 6 mAb GoH3 (data not shown). Incomplete inhibition by T1 and GoH3 was likely due to the presence of other integrins, such as ␣ 2 ␤ 1 , that also serve as adhesion receptors for laminin. Together, these results show that the soluble T1 peptide specifically inhibits ␣ 6 ␤ 1 -dependent cell adhesion, thus providing further support that the T1 sequence contains a binding site for integrin ␣ 6 ␤ 1 .
Effect of Alanine Substitutions in the T1 Sequence on Cell Adhesion-To determine which residues within the T1 sequence are critical determinants for ␣ 6 ␤ 1 -dependent cell adhesion, we prepared a series of GST-peptide fusions that carries the T1 backbone with single or double alanine substitutions at residues conserved among CCN1, CCN2, and CCN3 and tested their abilities to support cell adhesion. As shown in Fig. 5, alanine substitutions at residues Lys-226, Ile-228, or Gln-230 did not affect the peptide's ability to support cell adhesion. Although single mutation at either Thr-231 or Thr-232 resulted in partial reduction of cell adhesion, combined alanine substitutions of Thr-231 and Thr-232 completely abolished the ability of T1 to support cell adhesion. In addition, single substitutions in Trp-234, Ser-235, Ser-238, or Lys-239 resulted in Ͼ90% loss of T1 activity. When mutations in Trp-234 and Lys-239 were combined, cell adhesion was completely obliterated. These results indicate that TTSWSQCSKS is the core sequence in T1 for mediating ␣ 6 ␤ 1 binding. These data also explain the inabil- ity of the T2 peptide, which overlaps with the T1 peptide but lacks the TT residues of the core sequence, to inhibit ␣ 6 ␤ 1 -dependent cell adhesion.
Affinity Purification of Integrin ␣ 6 ␤ 1 using a T1-coupled Affinity Matrix-To confirm that the T1 peptide binds directly to integrin ␣ 6 ␤ 1 , we performed affinity chromatography to isolate integrin ␣ 6 ␤ 1 on a T1-coupled affinity column. In these studies, cell surface proteins on fibroblasts were radioiodinated and solubilized in octyl-␤-D-glucopyranoside in the presence of Mn 2ϩ to enhance integrin-ligand interaction. The cell lysate was applied to an affinity column comprised of GST-peptide fusion protein conjugated to Affi-gel. Unbound proteins were removed by washing, and bound proteins were eluted by increasing the ionic strength of the washing buffer. Fractions were collected and analyzed by SDS-PAGE under non-reducing conditions. A control column was prepared using GST fused to a scrambled T1 sequence, and no labeled protein band was eluted from the scrambled T1-GST column (Fig. 6A). By contrast, from the T1-GST affinity column, two protein bands with apparent molecular masses corresponding to integrin ␣ 6 (ϳ150 kDa) and ␤ 1 (ϳ130 kDa) subunits were eluted at 0.35 M NaCl ( lanes 5-7, Fig. 6B). To confirm that the bound labeled proteins was indeed the integrin ␣ 6 ␤ 1 complex, the eluates were subjected to immunoprecipitation using GoH3 (anti-␣ 6 ) or P3G8 (anti-␣ v ) as a control. Fig. 6C shows that GoH3 immunoprecipitated the labeled protein bands from the eluate, whereas P3G8 failed to pull down the protein complex in the control sample. Collectively, we conclude that integrin ␣ 6 ␤ 1 binds directly to the T1 sequence in CCN1.
Soluble T1 Peptide Disrupts CCN1-induced Endothelial Tubule Formation-Several CCN proteins, including CCN1, CCN2, and CCN3, are potent angiogenic inducers (2,25,26). Furthermore, when formulated into collagen gel, CCN1 is capable of inducing tubule formation of unactivated human umbilical vein endothelial cells (HUVECs), and this process is blocked by the anti-␣ 6 mAb GoH3 (16). Because the T1 sequence represents a major binding site for integrin ␣ 6 ␤ 1 in CCN1, we examined whether soluble T1 peptide would inhibit CCN1-induced tubule formation of unactivated HUVECs. As shown in Fig. 7, when collagen gels are formulated with CCN1, human umbilical vein endothelial cells are induced to form tubules. Preincubation of HUVECs with T1 (0.2 mM) for 30 min prior to plating completely inhibited CCN1-induced tubule formation. By contrast, the control T2, T3, and T4 peptides had no effect. Together, these results indicate that T1 inhibits CCN1induced tubule formation by blocking the interaction of CCN1 with integrin ␣ 6 ␤ 1 on unactivated HUVECs. Proteins in the eluted fractions were resolved on 7% SDS-polyacrylamide gels under non-reducing conditions and detected by autoradiography. In C, the high salt eluates from the GST-T1 column were pooled and subjected to immunoprecipitation with anti-␣ 6 (GoH3), or anti-␣ v (P3G8) mAb. The immunoprecipitated proteins were analyzed under non-reducing conditions. Molecular mass markers are indicated in kDa on the left. Results are representative of two experiments.

DISCUSSION
CCN1 is an angiogenic inducer that plays an essential role in normal vascular development during embryogenesis (1). We have recently shown that the pro-angiogenic activities of CCN1 are mediated through integrins ␣ 6 ␤ 1 and ␣ v ␤ 3 in unactivated and activated HUVECs, respectively (16). In this study, we have employed functional and biochemical analyses to define a 17-residue T1 sequence (GQKCIVQTTSWSQCSKS) in the CCN1 domain III as a novel integrin ␣ 6 ␤ 1 binding site. These findings provide a basis for the development of ␣ 6 ␤ 1 antagonists and a target for mutational analyses to examine the role of integrin ␣ 6 ␤ 1 -CCN1 interaction in angiogenesis.
Consistent with our earlier findings that a truncated mutant encompassing domains I-III of CCN1 is capable of inducing ␣ 6 ␤ 1 -dependent smooth muscle cell migration (10), we find that a recombinant fragment corresponding to the isolated domain III (TSP1 domain) of CCN1 is sufficient to support ␣ 6 ␤ 1 -dependent fibroblast adhesion. The specificity of ␣ 6 ␤ 1 interaction with the CCN1 domain III is confirmed by the failure of the CCN1 domain I and domain II fragments to support cell adhesion and by the observation that anti-␣ 6 and anti-␤ 1 mAbs specifically block cell adhesion to the CCN1 domain III (Figs. 1 and 2). Within domain III, we have further pinpointed the T1 sequence as an integrin ␣ 6 ␤ 1 binding site in CCN1 based on the following observations: 1) a T1-GST fusion protein and a synthetic T1 peptide specifically support ␣ 6 ␤ 1 -dependent cell adhesion ( Fig. 3 and data not shown); 2) integrin ␣ 6 ␤ 1 is purified from a detergent lysate of fibroblasts on a T1-GST affinity matrix, demonstrating direct interaction between integrin ␣ 6 ␤ 1 and the T1 sequence (Fig. 6); 3) soluble T1 peptide inhibits cell adhesion to ␣ 6 ␤ 1 ligands, including CCN1, CCN2, CCN3, and laminin, but not to other integrin ligands such as fibronectin, vitronectin, and collagen (Fig. 4); and 4) T1 peptide also blocks ␣ 6 ␤ 1 -dependent tubule formation of unactivated HUVECs in a collagen matrix containing CCN1 (Fig. 7). It is noteworthy that soluble T1 peptide is an effective inhibitor on ␣ 6 ␤ 1 -dependent cellular activities. Half-maximal inhibition of cell adhesion occurs at a peptide concentration of 25-50 M (Fig. 4C). Thus, the inhibitory potency of T1 is comparable to linear RGD peptides that inhibit adhesive functions of other integrins, such as ␣ v ␤ 3 , also at the micromolar range (31).
By alanine substitution mutagenesis of the T1-GST fusion protein, we showed that the C-terminal portion of T1 (TTSWSQCSKS) contains critical determinants for ␣ 6 ␤ 1 -dependent cell adhesion. Of note are the double T231A/T232A and W234A/K239A substitutions that result in complete loss of its capacity to support cell adhesion. This 10-residue segment is highly conserved among other CCN family members with only two non-conserved substitutions among CCN1, CCN2, and CCN3. Therefore, it is conceivable that ␣ 6 ␤ 1 also binds to the corresponding T1 sequences in other CCN proteins. Consistent with this notion, soluble T1 peptide also inhibits ␣ 6 ␤ 1 -dependent fibroblast adhesion to CCN2 and CCN3. These results led us to conclude that the conserved TTXWSXCSKS sequence (X represents a non-conserved residue) in CCN proteins defines a novel recognition motif for integrin ␣ 6 ␤ 1 . An important feature of this sequence is that any single alanine substitution of the conserved residues (i.e. T232A, W234A, S235A, S238A, and K239A) results in a drastic loss of ␣ 6 ␤ 1 binding activity, suggesting that it requires multiple coordination interaction with the ligand binding pocket in integrin ␣ 6 ␤ 1 .
Integrin ␣ 6 ␤ 1 has a limited ligand spectrum that includes laminin, CCN proteins, invasin, fertilin, and a collagen fragment known as tumstatin (32)(33)(34)(35). These diverse ␣ 6 ␤ 1 ligands that are involved in various biological processes are not structurally related. Several ␣ 6 ␤ 1 binding sequences have been iden-tified by screening synthetic peptides derived from some of these ␣ 6 ␤ 1 ligands. These include the NPWHSIYITRFG and TWYKIAFQRNRK sequences from the laminin ␣1 chain (32,36,37). In addition, TDE-containing peptides from the disintegrin domain of the fertilin ␤ subunit disrupt sperm-egg fusion presumably by blocking integrin ␣ 6 ␤ 1 -fertilin interaction (38). Several other ␣ 6 ␤ 1 binding peptides have also been isolated by screening phage display and synthetic peptide combinatorial libraries; however, these sequences are not present in any known ␣ 6 ␤ 1 ligand (39 -41). A comparison of the ␣ 6 ␤ 1 binding sequences reported to date reveals no consensus sequence that acts as an ␣ 6 ␤ 1 binding motif. Furthermore, our newly identified T1 sequence in CCN1 does not exhibit any sequence similarity to these ␣ 6 ␤ 1 binding peptides. Thus, integrin ␣ 6 ␤ 1 , like ␣ M ␤ 2 , is capable of recognizing a broad range of binding sequences. At present, whether these vastly different peptide sequences bind to the same or different sites in ␣ 6 ␤ 1 remains to be determined. Nonetheless, given that integrin ␣ 6 ␤ 1 has been implicated in a multitude of biological processes, it is tempting to speculate that different ␣ 6 ␤ 1 binding sequences may interact with distinct coordination sites within the ␣ 6 ␤ 1 ligand binding pocket to induce different signaling pathways that mediate disparate biological activities.
To date, three CCN proteins have been shown to induce neovascularization in vivo (2,25,26,42,43). Endothelial cell migration, proliferation, and differentiation into tubule structures are essential for the formation of new blood vessels. CCN1 is an activation-independent ligand of integrin ␣ 6 ␤ 1 in non-stimulated endothelial cells, mediating both cell adhesion and tubule formation through this integrin receptor (16). Whereas intact CCN1 is an angiogenic inducer, the T1 peptide acts as an ␣ 6 ␤ 1 antagonist to block CCN1-induced tubule formation of unactivated endothelial cells. Interestingly, the T1 FIG. 7. T1 peptide blocks CCN1-induced endothelial tubule formation in a collagen gel matrix. Unstimulated HUVECs were plated on 24-well plates precoated with type I collagen gels (2 mg/ml) in the absence (No Add) or presence of 50 g/ml CCN1, and a second layer of gels was overlaid on the attached cells as described under "Materials and Methods." Where indicated, cell suspension was incubated with the tested peptides for 30 min prior to plating. Tubule formation was assessed 16 -20 h thereafter. Results are representative of three separate experiments (magnification, ϫ100). sequence resides within the thrombospondin type 1 repeat homology domain of CCN1, and thrombospondin is an inhibitor of angiogenesis with its anti-angiogenic activity being localized to the procollagen homology region and the properdin-like type 1 repeat (44). A number of anti-angiogenic peptides have been derived from thrombospondin type I repeat, including the CS-VTCG-containing peptides that interact with CD36 on endothelial cells (45,46). Interaction of CD36 with the TSP1 domain of CCN proteins has not been demonstrated; however, CD36 has been shown to associate with integrin ␣ 6 ␤ 1 on human platelets and melanoma cells (47,48). If the CD36-␣ 6 ␤ 1 complex also exists on endothelial cells, it is an intriguing possibility that these two cell surface receptors may act in concert to regulate angiogenesis through interaction with proximal recognition sequences in the thrombospondin type 1 repeat of matricellular proteins.
In addition to integrin ␣ 6 ␤ 1 interaction with the T1 sequence in the TSP1 domain of CCN1, adhesion of fibroblasts and unactivated endothelial cells to CCN1 also requires heparan sulfate proteoglycans to act as co-receptors, which interact with the heparin binding motifs in the CCN1 C-terminal domain (20). Furthermore, CCN1-induced adhesive signaling and gene expression in fibroblasts is blocked by soluble heparin, suggesting the importance of the heparin binding motifs in this process (13). Angiogenesis is a complex biological process that likely requires the involvement of multiple cell surface receptors. It is unclear whether the TSP1 domain of CCN1, which contains the ␣ 6 ␤ 1 binding site, is sufficient to induce angiogenesis. An alternative possibility is that other functional domains such as the integrin ␣ v ␤ 3 binding site and the heparin binding motifs are also involved in this process. The identification of the T1 sequence as a major ␣ 6 ␤ 1 binding site in CCN1 will therefore aid in further mutational studies to define the critical elements in CCN1 required for its angiogenic activities.