Targeted Mutagenesis of the Angiogenic Protein CCN1 (CYR61)

The matricellular protein CCN1 (CYR61) regulates multiple cellular processes and plays essential roles in embryonic vascular development. A ligand of several integrin receptors, CCN1 acts through integrin α6β1 and heparan sulfate proteoglycans (HSPGs) to promote specific functions in fibroblasts, smooth muscle cells, and endothelial cells. We have previously identified a novel α6β1 binding site, T1, in domain III of CCN1. Here we uncover two novel 16-residue sequences, H1 and H2, in domain IV that can support α6β1- and HSPGs-dependent cell adhesion, suggesting that these sequences contain closely juxtaposed or overlapping sites for interaction with α6β1 and HSPGs. Furthermore, fibroblast adhesion to the H1 and H2 peptides is sufficient to induce prolonged MAPK activation, whereas adhesion to T1 induces transient MAPK activation. To dissect the roles of these sites in CCN1 function, we have created mutants disrupted in T1, H1, and H2 or in all three sites in the context of full-length CCN1. We show that the T1 and H1/H2 sites are functionally non-equivalent, and disruption of these sites differentially affected cell adhesion, migration, mitogen-activated protein kinase activation, and regulation of gene expression. Disruption of all three sites completely abolished α6β1-HSPG-mediated cellular activities. All mutants disrupting T1, H1, and H2 fully retain αvβ3-mediated pro-angiogenic activities, indicating that these mutants are biologically active and are defective only in α6β1-HSPG-mediated functions. Together, these findings identify and dissect the differential roles of the three sites (T1, H1, H2) required for α6β1-HSPG-dependent CCN1 activities and provide a strategy to investigate these α6β1-HSPG-specific activities in vivo.

The matricellular protein CCN1 (CYR61) regulates multiple cellular processes and plays essential roles in embryonic vascular development. A ligand of several integrin receptors, CCN1 acts through integrin ␣ 6 ␤ 1 and heparan sulfate proteoglycans (HSPGs) to promote specific functions in fibroblasts, smooth muscle cells, and endothelial cells. We have previously identified a novel ␣ 6 ␤ 1 binding site, T1, in domain III of CCN1. Here we uncover two novel 16-residue sequences, H1 and H2, in domain IV that can support ␣ 6 ␤ 1 -and HSPGs-dependent cell adhesion, suggesting that these sequences contain closely juxtaposed or overlapping sites for interaction with ␣ 6 ␤ 1 and HSPGs. Furthermore, fibroblast adhesion to the H1 and H2 peptides is sufficient to induce prolonged MAPK activation, whereas adhesion to T1 induces transient MAPK activation. To dissect the roles of these sites in CCN1 function, we have created mutants disrupted in T1, H1, and H2 or in all three sites in the context of full-length CCN1. We show that the T1 and H1/H2 sites are functionally non-equivalent, and disruption of these sites differentially affected cell adhesion, migration, mitogen-activated protein kinase activation, and regulation of gene expression. Disruption of all three sites completely abolished ␣ 6 ␤ 1 -HSPGmediated cellular activities. All mutants disrupting T1, H1, and H2 fully retain ␣ v ␤ 3 -mediated pro-angiogenic activities, indicating that these mutants are biologically active and are defective only in ␣ 6 ␤ 1 -HSPG-mediated functions. Together, these findings identify and dissect the differential roles of the three sites (T1, H1, H2) required for ␣ 6 ␤ 1 -HSPG-dependent CCN1 activities and provide a strategy to investigate these ␣ 6 ␤ 1 -HSPG-specific activities in vivo.
The CCN 1 family of ECM-associated signaling proteins has recently emerged as developmentally important regulators with diverse functions (1)(2)(3). The significance of this protein family is underscored by the finding that targeted disruptions of Ccn1 (cysteine-rich 61, Cyr61) and Ccn2 (connective tissue growth factor, Ctgf) both lead to lethality in mice but with distinct phenotypes (4,5), and mutations in CCN6 cause progressive pseudorheumatoid dysplasia in humans (6). Both CCN1 and CCN2 promote a broad spectrum of cellular processes including cell adhesion, migration, proliferation, survival, and differentiation (1,2,7). On the organismal level, growing evidence now supports the hypothesis that the regulation of angiogenesis and matrix remodeling underlie their biological functions in several contexts, including vessel morphogenesis, skeletal development, wound healing, and tumor growth (4, 5, 8 -10).
CCN1 was first identified as a cysteine-rich protein of ϳ40 kDa encoded by a growth factor-inducible immediate-early gene (11). Upon synthesis, CCN1 is quantitatively secreted and associates with the ECM and the cell surface, in part because of its high affinity for HSPGs (12). As is characteristic of the protein family, CCN1 is composed of an N-terminal secretory signal followed by four distinct modular domains with sequence similarities to insulin-like growth factor binding protein (IG-FBP), von Willebrand factor type C repeat (VWC), thrombospondin type I repeat (TSP1), and regions of Slit and mucins in the C terminus (CT) (13) (see Fig. 2A). As a modular ECM protein with characteristics intermediate between growth factors, cytokines, and structural ECM molecules, CCN1 can be considered a "matricellular" protein that serves regulatory roles without functioning as an integral component of structural elements (14 -16).
CCN1 induces angiogenesis in vivo (8,17), consistent with the embryonic lethality of Ccn1-deficient mice because of vascular defects in the placenta and the embryo (4). In keeping with its structural roots in ECM proteins, CCN1 functions as a ligand of integrins, which are heterodimeric transmembrane signaling receptors that mediate diverse cellular responses to ECM proteins (18). CCN1 binds at least five distinct integrins, which mediate its functions in a cell type-and context-specific manner (19 -23). In fibroblasts, smooth muscle cells, and endothelial cells, integrin ␣ 6 ␤ 1 and HSPGs act as co-receptors to mediate cell adhesion, and these co-receptors also support smooth muscle cell migration (21,24,25). By contrast, the pro-angiogenic activities of CCN1 in activated endothelial cells are all mediated through integrin ␣ v ␤ 3 (25).
Focusing on the first three domains of CCN1, we have recently identified a novel integrin ␣ 6 ␤ 1 binding site, T1, in domain III (26). In this study, we have uncovered two addi-tional sequences, H1 and H2, that can support ␣ 6 ␤ 1 -and HSPGdependent cell adhesion. To dissect the functional roles of the three sites for ␣ 6 ␤ 1 -HSPG interaction, we have created CCN1 mutants disrupted in T1, H1, and H2 and in all three sites. Analysis of these mutants shows that CCN1-receptor interaction through T1 (binding only ␣ 6 ␤ 1 ) and H1/H2 (interacting with both ␣ 6 ␤ 1 and HSPGs) differentially regulate cell adhesion, migration, MAPK activation, and regulation of gene expression. In particular, T1 and H1/H2 induce transient and prolonged p42/p44 MAPK activation, respectively. Disruption of all three sites completely abolish ␣ 6 ␤ 1 -HSPG-mediated activities but retain ␣ v ␤ 3 -dependent pro-angiogenic activities. These findings identify and dissect the sites required for ␣ 6 ␤ 1 -HSPG-dependent CCN1 activities, show that they can be dissociated from ␣ v ␤ 3 -mediated functions, and provide a molecular strategy for dissecting integrin-specific CCN1 functions in an in vivo context.
Enzyme-linked Immunosorbent Assay-Microtiter wells were coated with GST fusion proteins or BSA (50 l/well) overnight at 4°C followed by blocking with 1% BSA for 2 h at room temperature. Protein-coating efficiency was examined by incubation with an anti-GST 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 absorbance measurements at 420 nm (Zymed Laboratories Inc., San Francisco, CA).
Preparation of GST-Peptide Fusions, CCN1, and Mutant Proteins-The coding sequences for various CT domain peptides (Table I) were amplified by PCR using the murine Ccn1 cDNA as template, essentially as described (26). Primers used corresponded to the appropriate coding sequence and contained the BamHI and EcoRI restriction sites for cloning. 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. GST-peptide fusion proteins were produced in Escherichia coli strain BL21 and purified by glutathione affinity chromatography (Amersham Biosciences) as described by the manufacturer followed by dialysis against phosphate-buffered saline overnight at 4°C. GST-T1 peptide fusion protein was prepared as described previously and GST-T1-(K239E) peptide fusion was prepared similarly (26).
Mouse CCN1 engineered with a C-terminal FLAG tag was constructed using the primers 5Ј-CGCAATTGGAAAAGGCAGCTCACTG-AAGAGGC-3Ј (F1) and 5Ј-CCGGAATTCCTACTTGTCATCGTCATCC-TTGTAGTCGTCCCTGAACTTGTGGATGTCATTG-3Ј (F2) upon the mouse Ccn1 cDNA, yielding a PCR product with the last codon of Ccn1 followed by the FLAG tag coding sequence and a stop codon. The PCR product was digested with NcoI and EcoR1 and used to replace the corresponding NcoI-EcoR1 fragment of the full-length mouse Ccn1 cDNA in pBlueBac4.5 vector (21). Using the same procedure, the previously constructed DM coding sequence was used to generate the DM construct with a FLAG tag. The SM (K239E) mutant was constructed by site-directed mutagenesis using a two-step PCR procedure as described (27). The internal primer sets were 5Ј-GTCTTGGTCCCAGTG-TTCCGAGAGCTGCGG-3Ј and 5Ј-CACTGGGACCAAGACGTGGTCTG-AACGATGC-3Ј. This construct also created a silent mutation at C237, thereby providing a screening marker by eliminating a BSP12861 restriction site. The outside primers used in PCR were F1 and F2 as described above. SM was constructed using the mouse Ccn1 cDNA with a FLAG tag as PCR template, and TM was created by using DM as a template to generate the K239E mutation. All constructs were confirmed by direct sequence analysis. To express CCN1 and mutant proteins, recombinant baculovirus stocks were prepared as previously described (28) and used to infect SF-9 cells maintained under serumfree conditions in EX-CELL 400 medium (JRH Biosciences) at a multiplicity of infection of 10. Media were collected 36 -40 h post-infection, cleared by centrifugation, and adjusted to 0.4 M salt before loading onto an affinity column with anti-FLAG M2 affinity gel (Sigma). The column was washed with 0.4 M NaCl, 50 mM Hepes at pH 7.4 before elution with the FLAG peptide, DYKDDDDK, in the washing buffer. Eluted fractions were analyzed by SDS-polyacrylamide gel electrophoresis followed by Coomassie Brilliant Blue staining and immunoblotting. Pooled fractions for each construct were dialyzed against 20 mM Hepes at pH 7.4, 150 mM NaCl overnight at 4°C to remove the FLAG peptide. Alternatively, the bound protein was eluted with 0.1 M glycine followed by neutralization with 1 M HEPES at pH 8.0. Proteins eluted by both methods exhibited the same activities. Construction and purification of the CCN1-H1 and CCN1-H2 mutants (Fig. 4B) were previously described (21).
Cell Culture and Cell Adhesion-Primary human foreskin fibroblast (ATCC CRL-2076, passage 2) were kept in Iscove's modified Dulbecco's medium (Invitrogen) with 10% fetal bovine serum (Intergen) at 37°C with 5% CO 2 and used within 10 passages (21). Primary bovine aortic smooth muscle cells were obtained from Clonetics (San Diego, CA) and maintained in their smooth muscle basal medium supplemented with SmGm-2 (5% fetal bovine serum, 0.5 ng/ml human epidermal growth factor, 2 ng/ml human fibroblast growth factor, 50 g/ml gentamycin, 50 ng/ml amphotericin-B, 5 g/ml insulin) (24). Primary HUVEC were maintained at 37°C with 5% CO 2 in Medium 200 containing 2% serum and endothelial growth supplements according to vendor protocols (Cascade Biologics) and used within the fifth and tenth passages for all experiments (25). For fibroblast adhesion assays, test proteins were coated onto 96-well microtiter plates (BD Biosciences) in phosphatebuffered saline (50 l per well), and wells were blocked with 1% BSA at room temperature for 1 h. Fibroblasts were washed, detached in phosphate-buffered saline containing 2.5 mM EDTA, resuspended in serumfree basal medium, and allowed to adhere to wells at 5 ϫ 10 5 cells/ml (50 l/well). For HUVEC, detached cells were resuspended in serum-free basal medium containing 0.2% BSA, 10 mM HEPES, pH 7.2, at 7 ϫ 10 5 cells/ml with 0.5 mM Mn 2ϩ for 30 min to activate integrin receptors. Where indicated, cells were preincubated with EDTA, peptides, heparin, or function-blocking mAbs for 30 min (fibroblasts) or 60 min (endothelial cells) before plating. Adherent cells were fixed 20 min after plating and stained with methylene blue before quantification by dye extraction and measurement of absorbance at 620 nm as described (7). Where indicated, fibroblasts were treated with heparinase I (EC 4.2.2.7; 2 units/ml, Sigma) at 37°C for 30 min before plating.
MAPK Activation and Gene Expression-Primary human fibroblasts were serum-starved, resuspended in serum-free medium, and plated on 35-mm dishes pre-coated with CCN1 (10 g/ml), SM (10 g/ml), DM (250 g/ml), or laminin (10 g/ml) for various times. Clarified cell lysates were electrophoresed on 10% SDS-PAGE and immunoblotted with polyclonal anti-MAPK antibodies or antibodies specific for dual-phosphorylated p42/ p44 MAPKs. To test the effects of soluble CCN1 and mutants on gene expression, serum-starved cells were treated with 10 g/ml protein in serum-free condition for 24 h. Total cellular RNA was isolated, resolved on agarose-formaldehyde gel, blotted onto nylon membrane, and hybridized to [ 32 P]dCTP-labeled probes for human VEGF-A, MMP-1, and glyceraldehyde-3-phosphate dehydrogenase cDNAs as described (9). The blots were washed at high stringency (0.1 ϫ SSC (1ϫ SSC ϭ 0.15 M NaCl and 0.015 M sodium citrate), 0.1% SDS at 65°C) and analyzed by Phosphor-Imager (Amersham Biosciences).
VSMC Migration Assay-Cell migration assay was performed using a modified Boyden chamber as previously described (29). Test proteins were diluted in smooth muscle basal serum-free medium containing 0.5% BSA, loaded in the lower chamber in quadruplicate, and covered with a polycarbonate filter (5 M pore diameter, Nucleopore). VSMC were trypsinized, resuspended in serum-free smooth muscle basal medium containing 0.5% BSA, and applied to the upper chamber (5 ϫ 10 4 cells/well). After a 6-h incubation at 37°C, 5% CO 2 , the membrane was removed and stained with Diff-Quik (Dade-Behring). The chemotactic response was determined by counting the total number of cells migrated in 10 randomly selected microscope fields at 400ϫ magnification.

RESULTS
Identification of Two Novel Peptides That Support Integrin ␣ 6 ␤ 1 -and HSPG-dependent Cell Adhesion-Previous analysis of a C-terminal-truncated CCN1 mutant indicated the presence of an integrin ␣ 6 ␤ 1 binding site within the first three domains (24), leading to the identification of a novel ␣ 6 ␤ 1 binding site, T1, in domain III (26). More recent studies, however, suggested the potential presence of additional ␣ 6 ␤ 1 binding sites in the CT domain (see below). To evaluate this possibility, we constructed a series of six overlapping peptides encompassing the entire CT domain as fusion proteins linked to GST ( Table I). The fusion proteins were purified to apparent homogeneity and showed similar coating efficiency on microtiter wells as detected by enzyme-linked immunosorbent assay using an anti-GST antibody (data not shown). We tested the ability of these GST-peptide fusion proteins to support primary human fibroblast adhesion, which occurs through ␣ 6 ␤ 1 and HSPGs when CCN1 serves as substrate (21). As shown in Fig. 1A, the H1 and H2 peptide fusion proteins were able to support fibroblast adhesion, whereas GST alone and other peptide fusion proteins were inactive. Consistent with an integrin-mediated event, cell adhesion to H1 and H2 was inhibited by EDTA or Ca 2ϩ , whereas Mg 2ϩ restored cell adhesion in EDTA-treated cells (Fig. 1B). Further, preincubation of fibroblasts with functionblocking mAbs against integrin ␣ 6 (GoH3) or ␤ 1 (P4C10) abolished cell adhesion to H1, H2, or CCN1, whereas mAbs against ␣ v ␤ 3 (LM609) or control mouse IgG had no effect (Fig. 1C). These results show that fibroblast adhesion to H1 and H2 is mediated through integrin ␣ 6 ␤ 1 .
It is noteworthy that the H1 and H2 sequences subsume multiple basic residues previously identified to be important for binding heparin (21). Therefore, we tested whether cell adhesion to GST-H1 and GST-H2 also involved heparin binding. Indeed, the presence of soluble heparin in the plating medium, thereby saturating the heparin binding sites of CCN1, obliterated cell adhesion to the GST-H1 or GST-H2 as well as to CCN1 (Fig. 1D). Likewise, treatment of fibroblasts with heparinase I to remove heparin moieties from cell surface HSPGs completely inhibited cell adhesion to H1, H2, and CCN1, whereas the heparinase-treated cells adhered to laminin or vitronectin indifferently. Thus, fibroblast adhesion to H1 and H2 also requires interaction with cell surface HSPGs, suggesting that H1 and H2 contain closely juxtaposed or overlapping sites for interaction with ␣ 6 ␤ 1 and HSPGs.
Construction of CCN1 Mutants Disrupted in Binding Sites for the ␣ 6 ␤ 1 -HSPG Coreceptors-Results above indicate that CCN1 contains three sequences important for interaction with ␣ 6 ␤ 1 and HSPGs: T1, H1, and H2. To dissect their relative contributions to CCN1 functions, we constructed mutants in the context of full-length CCN1 that are disrupted in T1 (SM, single mutant), H1 and H2 (DM, double mutant), and all three sites (TM, triple mutant) ( Fig. 2A).
To disrupt the T1 site (GQKCIVQTTSWSQCSKS), we turned to previous mutagenesis in which each conserved residue was individually altered (26). Alanine substitution in any of 4 residues (Trp-234, Ser-235, Ser-238, Lys-239) reduced ␣ 6 ␤ 1 -mediated cell adhesion by Ͼ90%, and mutation in both Thr-231 and Thr-232 completely abolished cell adhesion. However, our attempts to produce mutant CCN1 proteins with alanine substitutions in these critical T1 residues did not yield soluble proteins (data not shown). The expressed mutants aggregated and formed inclusion bodies inside the cell, suggesting that these mutants adopted detrimental conformations that prevented them from being secreted. To avoid mutations that may cause major structural changes, we applied two algorithms, PSIPRED and FRAGFOLD (30,31), to predict possible structural alterations that may result from mutations. Each of the critical residues in T1 was substituted with each of the other 19 amino acids to predict the protein structure using these algorithms. Consistent with the inability of alanine substitution mutants to be secreted, the T1 region of CCN1 appears highly sensitive to perturbation. Substitutions of the critical residues in T1 with nearly any amino acid resulted in drastic changes in the predicted protein structure. However, we found one mutation, namely K239E, that did not elicit a predicted conformational change. To test the efficacy of this mutation in disrupting interaction with ␣ 6 ␤ 1 , we constructed a GST fusion protein linked to the T1 peptide carrying the K239E mutation (GQKCIVATTSWSQCSES). Fibroblast adhesion to this mutant peptide was assessed. GST-T1 (wild type) and GST-T1-(K239E) have similar coating efficiency as determined by enzyme-linked immunosorbent assay (data not shown). As expected, fibroblasts adhered to the GST-T1 fusion but not to BSA or the GST control ( Fig. 2B) (26). GST-T1-(K239E) was completely unable to support cell adhesion, indicating that the single charge-reversed K239E mutation effectively abolished T1 binding to integrin ␣ 6 ␤ 1 . We, therefore, constructed the K239E mutation in the context of full-length CCN1, designated SM ( Fig. 2A).
We have previously shown that mutations in the basic residues within the H1 and H2 sequences obliterated heparin binding in CCN1 (21). We, therefore, incorporated these mutations, changing KGKKCSKTKKSPEPVR (H1) to AGAAC-SATAKSPEPVR and FTYAGCSSVKKYRPKY (H2) to FTY-AGCSSVAAYAPKY in CCN1 to create the mutant DM ( Fig.  2A). In addition, we also created a mutant (TM) that combined the K293E mutation in the DM background, thereby disrupting all three sites for ␣ 6 ␤ 1 -HSPG interaction. For consistency, WT CCN1 and the mutants SM, DM, and TM used in this study were similarly constructed with a C-terminal FLAG epitope tag. The recombinant proteins were expressed via a baculovirus vector and purified to apparent homogeneity from insect cell conditioned media (see "Materials and Methods"; Fig. 2C).
Cell Adhesive Properties of CCN1 Mutants-Fibroblast adhesion to CCN1 is mediated exclusively through integrin ␣ 6 ␤ 1 with HSPGs as coreceptors (21), thus providing an assay for the interaction of CCN1 with ␣ 6 ␤ 1 -HSPGs. Surprisingly, both SM and WT CCN1 supported fibroblast adhesion in a dose-dependent manner with similar efficiency, achieving maximal cell adhesion at a coating concentration of 1 g/ml (Fig. 3A). Thus, disruption of the T1 binding site for ␣ 6 ␤ 1 did not significantly affect the cell adhesive activity of CCN1, indicating that H1 and H2 are sufficient to support fibroblast adhesion through ␣ 6 ␤ 1 -HSPGs.
Disruption of H1 and H2 in DM, however, severely inactivated cell adhesive activity, indicating that H1/H2 are critical for mediating fibroblast adhesion (Fig. 3B). Whereas cell adhesion to CCN1 and SM plateaued when coated at 1 g/ml (Fig.  3A), DM failed to support cell adhesion at this concentration. Because DM still retained the T1 binding site for ␣ 6 ␤ 1 , we postulated that it may be able to support cell adhesion at very high coating concentrations, analogous to the GST-T1 peptide fusion protein (Fig. 2B). Indeed, DM can support fibroblast adhesion when coated at ϳ50 -100-fold higher concentration than WT CCN1, with maximal adhesion observed at 50 -100 g/ml (Fig. 3B). Consistent with this adhesion being mediated through the T1 sequence, further disruption of T1 in TM com- pletely abolished cell adhesion at any concentration, even when coated at 250 g/ml. Function-blocking mAbs against integrin ␣ 6 (GoH3) or ␤ 1 (P4C10) obliterated cell adhesion to CCN1, SM, or DM, whereas mAb against integrin ␣ v ␤ 3 (LM609) had no effect, thus establishing that cell adhesion to these mutants occurs through ␣ 6 ␤ 1 as in wild type CCN1 (Fig. 3C). Taken together, these results show that ␣ 6 ␤ 1 -HSPG-mediated fibroblast adhesion to CCN1 occurs primarily through H1 and H2, and mutations of these sites decimated this activity. Cell adhesion through the ␣ 6 ␤ 1 binding site T1 is evident only when H1 and H2 are mutated and at high protein concentrations. T1, H1, and H2 Peptides Induce Differential Activation of MAPKs-Transient versus sustained activation of p42/p44 MAPKs by extracellular signals are known to specify different biological outcomes (32,33). Whereas cell adhesion to most ECM proteins induces a transient activation of p42/p44 MAPKs, CCN1 has the unusual ability to induce sustained MAPK activation as an adhesive substrate (7). To test the possibility that specific CCN1 peptides supporting cell adhesion may contribute directly to activation of MAPKs, we examined fibroblast adhesion to CCN1 or to the GST-peptide fusion proteins. As expected, fibroblast adhesion to CCN1 resulted in rapid and sustained MAPK activation lasting for at least 9 h (Fig. 4A). Remarkably, cells adhered to the GST-H2 also showed similar kinetics of MAPK activation. Likewise, cell adhesion to GST-H1 gave similar results (data not shown). By contrast, cells adhered to GST-T1 or laminin induced rapid but transient MAPK activation that peaked 30 min to 1 h after plating and declined to basal level thereafter. These results reveal two distinct kinetics of MAPK activation by CCN1 peptides, one rapid and transient (when cells adhere to T1) and the other rapid and sustained (when cells adhere to H1 or H2).
To address the signaling contributions of the T1, H1, and H2 sites further, we examined MAPK activation in full-length CCN1 mutants disrupted in these sites. CCN1 mutants disrupted in the H1 or H2 sequence were prepared (21) (Fig. 4B). As shown in Fig. 4C, mutations in H1 or H2 did not abolish the ability to induce sustained MAPK activation, which was observed beyond 3 h after plating. In contrast, cell adhesion to laminin induced only transient MAPK activation lasting about 1 h. Thus, the H1 and H2 sites appear to be equivalent and redundant with respect to their ability to bind heparin (21), support ␣ 6 ␤ 1 -HSPGs-mediated cell adhesion (Fig. 1), and induce sustained p42/p44 MAPKs activation (Fig. 4).
Based on the above results, we postulated that mutations in both H1 and H2 may abolish sustained MAPK activation, FIG. 1. GST-H1 and GST-H2 fusion proteins support ␣ 6 ␤ 1 -dependent adhesion of primary human fibroblasts. A, 96-well microtiter plates were coated with various recombinant GST-peptide fusion proteins (200 g/ml) at 4°C overnight and blocked with 1% BSA. Primary human skin fibroblasts were allowed to adhere at 37°C for 20 min. Adherent cells were fixed, stained with methylene blue, and quantified by absorbance at 620 nm. B, human skin fibroblast adhesion was evaluated as above in microtiter wells coated with GST (50 g/ml), GST-H1 (50 g/ml), GST-H2 (50 g/ml), or WT CCN1 (1 g/ml) at 4°C overnight and blocked with 1% BSA. EDTA (5 mM), Mg 2ϩ (10 mM), or Ca 2ϩ (5 mM) was added as indicated. C, human fibroblast adhesion was assessed as described above. Where indicated, 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 before plating. D, fibroblast adhesion was assayed with either without treatment (control) or in the presence of 1 g/ml soluble heparin or with fibroblasts treated with 5 units/ml heparinase I for 30 min. prior to plating. Microtiter wells were coated with various proteins as described above or with laminin (LN, 10 g/ml) or vitronectin (VN, 1 g/ml). Data are shown as the means Ϯ S.D. of triplicate determinations and representative of three separate experiments. Each construct is similarly endowed with an N-terminal secretory signal and a Cterminal FLAG epitope tag. Wild type T1, H1, and H2 sequences and specific amino acid changes in the mutants are shown in bold italics. Recombinant proteins were expressed in insect cells via a baculovirus vector. IGFBP, insulin-like growth factor binding protein; VWC, von Willebrand factor type C repeat; TSP1, thrombospondin type I repeat. B, primary human skin fibroblasts were plated on microtiter wells coated with either GST, GST-T1, or GST-T1-(K239E) fusion proteins (50 g/ml each). Cells were allowed to adhere at 37°C for 20 min. After washing, adherent cells were fixed, stained with methylene blue, and quantified by absorbance at 620 nm. C, FLAG affinitypurified recombinant proteins (2 g each) were electrophoresed in 10% SDS-PAGE followed by staining with Coomassie Brilliant Blue. Molecular mass (kDa) of the markers are shown at the left. The gel was immunoblotted with polyclonal anti-CCN1 antibodies and shown in the lower panel.
whereas transient MAPK activation, an activity supported by T1, might be preserved. Indeed, we observed that MAPK activation in cells adhered to CCN1 and SM was sustained for at least 5 h (Fig. 4C). DM, however, lost the ability to induce sustained MAPK activation but instead supported transient activation similar to laminin, with MAPK activation returning
to background level within 3 h. In these experiments, we used a high coating concentration of DM to ensure that similar number of cells were adherent to DM compared with SM and CCN1. Together, these results show that H1 or H2 is sufficient to induce prolonged activation of p42/p44 MAPK, and the presence of either site is necessary for this activity, whereas T1 is sufficient to induce transient MAPK activation.
Mutations in H1 and H2 Impair Regulation of Gene Expression-CCN1 activates a genetic program in fibroblasts, leading to up-regulation of genes implicated in angiogenesis and matrix metabolism several hours after treatment (9). Although the involvement of integrin ␣ 6 ␤ 1 in gene regulation by CCN1 has not been established, up-regulation of Vegf-A and MMP-1 is inhibited by PD98059, suggesting a dependence on MAPKs (9). To test whether the loss of sustained p42/p44 MAPK activation in DM and TM may alter CCN1-regulated gene expression, serum-starved fibroblasts were treated with soluble CCN1 proteins and analyzed for gene expression. Because activation of gene expression is observed both when CCN1 is used as an adhesion substrate or in a soluble form (7,9), this assay does not preclude analysis of TM. As shown in Fig. 5, SM upregulates Vegf-A and MMP-1 expression similar to WT CCN1, whereas DM and TM were defective in this activity. Thus, MAPK-dependent up-regulation of Vegf-A and MMP-1 by CCN1 requires the presence of the H1/H2 sites for ␣ 6 ␤ 1 -HSPGs interaction and correlates with sustained MAPK activation through these sites.
Defect in Stimulation of Smooth Muscle Cell Migration-Another CCN1 activity known to be mediated through ␣ 6 ␤ 1 -HSPGs is the induction of smooth muscle cell chemotaxis (24). To test the effect of CCN1 mutants, we employed a modified Boyden chamber assay to quantify VSMC migration. Cells were loaded on the upper chamber, and those that migrated through a membrane to the lower chamber where the chemoattractant was placed were counted. As shown in Fig. 6, VSMC migrated to soluble CCN1 or SM similarly and in a dosedependent manner, with maximal migration occurring at 2-4 g/ml. By contrast, TM was completely inactive at any concentration tested. These results show that the T1 site for ␣ 6 ␤ 1 binding is dispensable for inducing VSMC migration, whereas inactivation of all three sites for ␣ 6 ␤ 1 -HSPG interaction completely abolished this activity.
CCN1 Mutants Retain ␣ v ␤ 3 -mediated Angiogenic Activities-Having lost all three sites for ␣ 6 ␤ 1 -HSPG interaction, TM was completely inactive in ␣ 6 ␤ 1 -HSPG-mediated activities described above. It may be questioned whether TM is inactive due to possible conformational changes resulting from mutations or due to a specific loss of ␣ 6 ␤ 1 -HSPG interaction. Moreover, whether the ␣ v ␤ 3 -mediated pro-angiogenic activities of CCN1 require interaction with HSPGs is not known (25). To address these questions, we examined the pro-angiogenic functions of CCN1 mutants in activated endothelial cells. HUVEC adhesion to CCN1 occurs through both ␣ v ␤ 3 and ␣ 6 ␤ 1 -HSPGs (25), and SM supported HUVEC adhesion similar to WT CCN1 (Fig. 7A). Cell adhesion was abolished by EDTA and partially inhibited by GRGDSP peptide or the anti-␣ v ␤ 3 mAb LM609, indicating that cell adhesion to CCN1 and SM was partly mediated through integrin ␣ v ␤ 3 . Cell adhesion was also partially inhibited by the anti-␣ 6 mAb GoH3, consistent with involvement of the ␣ 6 ␤ 1 -HSPG coreceptors. In contrast, HUVEC adhesion to DM and TM was completely inhibited by EDTA, GRGDSP peptide, and LM609 but not at all by GoH3. These results indicate that DM and TM support HUVEC adhesion solely through ␣ v ␤ 3 , having lost ␣ 6 ␤ 1 -HSPG interactions.
Induction of endothelial cell migration is a pro-angiogenic activity that correlates with induction of neovascularization in vivo. Using a modified Boyden chamber assay, we found that CCN1, SM, DM, and TM were all able to stimulate HUVEC migration (Fig. 7B). HUVEC migration to CCN1 and all mutants tested was completely inhibited by the anti-␣ v ␤ 3 mAb LM609, whereas the anti-␣ 6 mAb GoH3 had no effect. Thus, all CCN1 mutants with disrupted ␣ 6 ␤ 1 -HSPG binding sites are still capable of inducing ␣ v ␤ 3 -dependent HUVEC migration, similar to WT. Consistent with this angiogenic activity, both WT CCN1 and TM were able to induce tubule formation in phorbol 12-myristate 13-acetate-activated HUVEC cultured in collagen gel (Fig. 7C). Preincubation of cells with LM609 inhibited tubule formation, whereas the addition of GoH3 had no effect, showing that tubule formation is ␣ v ␤ 3 -dependent. Similar results were obtained with SM and DM (data not shown). These findings show that CCN1 mutants defective in ␣ 6 ␤ 1 -HSPG interactions are functionally intact and are fully capable of inducing ␣ v ␤ 3 -dependent angiogenic activities in activated endothelial cells. DISCUSSION CCN1 is an angiogenic inducer that plays essential roles in embryonic development (4). Acting through integrins and HSPGs, CCN1 regulates multiple cellular processes. In fibroblasts, smooth muscle cells, and endothelial cells, CCN1 supports cell adhesion through the integrin ␣ 6 ␤ 1 -HSPG coreceptors, which also mediate CCN1-stimulated cell migration in smooth muscle cells (21,24,25). This study has identified two novel sequences, H1 and H2, that are responsible for ␣ 6 ␤ 1 -HSPG-dependent cell adhesion in CCN1. Functional and mutational analyses demonstrate that the H1/H2 sequences and the previously identified ␣ 6 ␤ 1 binding site T1 contribute differentially to CCN1 activities including cell adhesion, migration, MAPK activation, and regulation of gene expression. Furthermore, we show that mutation of all three sites completely eliminate ␣ 6 ␤ 1 -HSPG-mediated activities, which are fully dispensable for ␣ v ␤ 3 -dependent CCN1 functions. Together, these findings identify and dissect the three sites relevant for ␣ 6 ␤ 1 -HSPG-mediated activities and provide a strategy for analyzing integrin-specific functions of CCN1 in vivo.
Previous studies pointed to the presence of an ␣ 6 ␤ 1 binding site within the first three domains, leading to the identification of the T1 sequence in domain III as a novel binding site for ␣ 6 ␤ 1 (24,26). Further studies indicated the presence of additional sites, culminating in the identification of the H1 and H2 sequences as being critical for ␣ 6 ␤ 1 -HSPGs-dependent CCN1 activities based on the following observations (Fig. 1). 1) Fibroblasts adhere to H1 and H2 peptides, and this adhesion is inhibited by EDTA and Ca 2ϩ but supported by Mg 2ϩ . 2) Anti-␣ 6 and ␤ 1 mAbs specifically blocked cell adhesion to H1 and H2. 3) Fibroblast adhesion to H1 and H2 was abolished by the presence of soluble heparin in the medium or by heparinase treatment of cells. Consistent with these observations is the finding that H1 and H2 contain multiple basic residues, mutations of which eliminate heparin binding activity (21). Analysis of targeted mutations of these sites provided further evidence that H1 and H2 mediate ␣ 6 ␤ 1 -HSPG-dependent CCN1 activities (Figs. [2][3][4][5]. These results suggest that H1 and H2 contain closely juxtaposed or overlapping sites for interaction with ␣ 6 ␤ 1 and HSPGs. Compelling evidence demonstrates that the T1 sequence in the TSP1 domain is a bona fide binding site for integrin ␣ 6 ␤ 1 (26). Synthetic T1 peptide supports ␣ 6 ␤ 1 -dependent cell adhesion as an immobilized substrate, blocks ␣ 6 ␤ 1 -mediated activities as a soluble inhibitor, and purifies ␣ 6 ␤ 1 from octylglucoside extracts of fibroblasts as an affinity matrix. The difference in cell adhesive properties between DM and TM further establishes T1 as an ␣ 6 ␤ 1 binding site (Fig. 3B). It is interesting to note that crystallographic analysis revealed the thrombospondin-1 type 1 repeat as an antiparallel, threestranded fold capped by three disulfide bonds (34). Because alanine substitutions in the T1 sequence had a propensity to adopt apparently detrimental conformations, it is likely that mutations in most residues within T1 may destabilize this highly compact and structured domain. However, the physiological role of the T1 binding site is currently unclear, since disruption of T1 (SM) does not result in any apparent loss of activity. It appears that T1 is not the primary ␣ 6 ␤ 1 binding site in CCN1; rather, ␣ 6 ␤ 1 -HSPGs-dependent CCN1 activities are mediated mainly through H1 and H2. Nevertheless, the possibility that T1 might be critical for ␣ 6 ␤ 1 -dependent functions in specialized contexts cannot be excluded. For example, the T1 site may become more available or exposed for ␣ 6 ␤ 1 binding if the CT domain of CCN1 interacts with another protein or is removed by proteolysis. Existing evidence for proteolytic processing in CCN2 (35,36) and protein-protein interaction through the CT domain in CCN3 (37)(38)(39) are consistent with these possibilities.
A novel and intriguing finding in this study is that cell adhesion to the T1 peptide induces transient activation of p42/ p44 MAPKs, whereas cell adhesion to H1 or H2 peptide induces sustained MAPK activation (Fig. 4A). Furthermore, sustained MAPK activation is abolished by mutations in both H1 and H2 (DM) in the context of full-length CCN1 but not by mutations in either H1 or H2 alone (Fig. 4C). Thus, the H1 and H2 sequences appear to be functionally redundant with respect to their ability to bind heparin (21), support ␣ 6 ␤ 1 -HSPGs-mediated cell adhesion (Fig. 1), and induce prolonged p42/p44 MAPK activation (Fig. 4). In contrast, the presence of the T1 sequence alone, either as a GST fusion protein, or in the context of full-length CCN1 (DM) is sufficient to induce transient but not prolonged MAPK activation. These findings suggest that the T1 and H1/H2 sequences are capable of inducing different cellular responses, since transient versus sustained MAPK activation are known to specify distinct biological outcomes (32,33). For example, the H1/H2 sequences may play a critical role in CCN1 regulated gene expression, consistent with the observation that DM and TM are concomitantly defective for sustained induction of MAPK and up-regulation of Vegf-A and MMP-1 (9) (Fig. 4).
Integrin ␣ 6 ␤ 1 is known to bind a number of substrates, including laminin, meltrin, thrombospondin, CCN proteins, tumstatin, invasin, fertilin, and human papillomavirus-16 (21, 40 -45). Interestingly, these ligands are structurally diverse, and no consensus binding site for integrin ␣ 6 ␤ 1 has emerged to date. Among the ligands of ␣ 6 ␤ 1 , CCN1 has an unusual dependence on HSPGs as coreceptors (21,24,25). Given the closely juxtaposed or overlapping sites for ␣ 6 ␤ 1 -HSPGs interaction in H1 and H2, it is tempting to speculate that these sites may simultaneously associate with ␣ 6 ␤ 1 and HSPGs, and this association may contribute to the unique signaling capabilities of H1/H2 as compared with T1 (Fig. 4). It is noteworthy that cell adhesion to fibronectin is known to require binding of HSPGs for ␣ 5 ␤ 1 -mediated functions but not for ␣ 4 ␤ 1 -dependent activities (47). Thus, integrin-specific association with HSPGs may confer signaling specificity of ligands. In this regard, the ␣ v ␤ 3mediated pro-angiogenic activities of CCN1 are independent of HSPGs based on functional and mutational studies (Fig. 7) (25).
The mutants SM, DM, and TM retain all pro-angiogenic activities mediated through integrin ␣ v ␤ 3 in activated HUVEC. Thus, these CCN1 mutants support endothelial cell adhesion, induce cell migration, and induce tubule formation (Fig. 6). In addition, these mutants are also able to enhance HUVEC DNA synthesis induced by VEGF and to promote endothelial cell survival under conditions of growth factor deprivation similar to WT CCN1, both ␣ v ␤ 3 -dependent activities (25). These findings demonstrate that the CCN1 mutants are biologically active, and their functional defects are indeed due to mutation of the specific receptor binding sites rather than structural perturbations. Furthermore, ␣ v ␤ 3 -mediated CCN1 pro-angiogenic activities are fully independent of interaction with ␣ 6 ␤ 1 and HSPGs.
Recently, we have identified a novel ␣ v ␤ 3 binding site within a 20-amino acid sequence (V2) in CCN1 (48). A single amino acid mutation in this V2 sequence in the context of full-length CCN1 impaired both ␣ v ␤ 3 binding and ␣ v ␤ 3 -mediated activities such as cell migration and DNA synthesis in endothelial cells. It is noteworthy that CCN1 activities previously associated with ␣ v ␤ 3 are specifically abolished in a targeted mutation of the ␣ v ␤ 3 binding site (48), and activities associated with ␣ 6 ␤ 1 -HSPGs are specifically lost in mutants targeting binding sites for these receptors (Figs. 3-7). Thus, there is a complete concordance of results from the current mutational analyses and previous studies that assigned specific integrins to activities using mAbs, agonists, and antagonists (21,24,25), thereby providing compelling evidence that the cellular activities of CCN1 are mediated through specific integrins and HSPGs.
The ability of CCN1 to up-regulate Vegf expression is rele-vant in the context of angiogenesis. Although CCN1 exerts pro-angiogenic activities in activated endothelial cells through binding to ␣ v ␤ 3 (19,25,48) (Fig. 7), it may nevertheless act in part through up-regulation of VEGF to induce angiogenesis in vivo (8). This study shows that DM and TM are fully capable of inducing in vitro angiogenesis but are unable to up-regulate Vegf; thus, these events can be dissociated. Future studies examining the angiogenic activities of DM and TM in vivo will likely yield new insight into the angiogenic mechanism of CCN1 and help to dissect its direct action on endothelial cells from potential indirect action through VEGF. CCN1 is a matricellular protein involved in multiple biological processes. Aside from being essential for successful embryonic vascular development (4), CCN1 expression is also associated with skeletal development (49,50), cutaneous wound healing (9,51), restenosed blood vessels (24,52), advanced atherosclerotic lesions (23,53), and advanced human breast cancer (10). Although a number of CCN1 activities have been characterized in cell culture systems, it is likely that many of its biological functions may be understood only in the context of an organ or tissue involving the interplay among multiple cell types. Characterization of CCN1 mutants that dissociate ␣ v ␤ 3and ␣ 6 ␤ 1 -HSPGs-mediated activities will provide a strategy for the in vivo dissection of integrin-specific CCN1 functions using transgenic or gene replacement approaches.