Cooperation between Thrombospondin-1 Type 1 Repeat Peptides and αvβ3 Integrin Ligands to Promote Melanoma Cell Spreading and Focal Adhesion Kinase Phosphorylation*

CD47-binding sequences from the carboxyl-terminal domain of thrombospondin-1 (TSP1) are known to regulate activity of the αvβ3 integrin (Gao, G., Lindberg, F. P., Dimitry, J. M., Brown, E. J., and Frazier, W. A. (1996) J. Cell Biol. 135, 533–544). Here we show that peptides from the type 1 repeats of TSP1 also stimulate αvβ3 integrin function in melanoma cells. Addition of soluble peptide 246 (KRFKQDGGWSHWSPWSS) enhances spreading of A2058 melanoma cells on several αvβ3integrin ligands, including vitronectin, recombinant TSP1 fragments containing the Arg-Gly-Asp sequence, and native TSP1. This activity requires the Trp residues and is independent of CD36-binding sequences in the type 1 repeats. Recombinant type 1 repeats expressed as a glutathione S-transferase fusion protein also enhance spreading on vitronectin and TSP1. Activation of αvβ3 integrin by the soluble peptide 246 stimulates organization of F-actin and increases tyrosine phosphorylation of focal adhesion kinase. In contrast, direct adhesion of melanoma cells on immobilized peptide 246 inhibits tyrosine phosphorylation of focal adhesion kinase. Stimulation of αvβ3 integrin function by the type 1 repeat peptide differs from that induced by CD47-binding TSP1 peptides in that heparan sulfate proteoglycans are required and pertussis toxin does not inhibit the former activity. Thus, the type 1 repeats contain a second sequence of TSP1 that can enhance αvβ3integrin signaling, and these two sequences stimulate recognition of both vitronectin and TSP1 by the αvβ3integrin.

Thrombospondin-1 (TSP1) 1 is an extracellular matrix protein that directly mediates adhesion of some cell types. A more important function of TSP1, however, may be to regulate interactions of cells with other extracellular matrix molecules (reviewed in Ref. 1). This regulation includes stimulation or inhibition of adhesion to other extracellular matrix ligands (2,3), stimulation or inhibition of proteolytic degradation of extracellular matrix components (4,5), and stimulation or antagonism of growth or survival signals from the extracellular matrix (6 -8).
Domains or peptide sequences from TSP1 that express some of these activities have recently been defined, and receptors have been identified that interact with some of these TSP1 sequences. An RGD sequence in the last type 3 repeat module promotes cell adhesion and binds to the integrin ␣ v ␤ 3 (9). Two sequences from the carboxyl-terminal domain that contain a VVM motif bind to CD47 and regulate activity of the integrins ␣ IIb ␤ 3 , ␣ v ␤ 3 , and ␣ 2 ␤ 1 in specific cell types (3,10,11). Peptides from the type 1 repeats and procollagen domains that bind to CD36 inhibit endothelial cell motility (12,13). Heparin-binding peptides from the type I repeats promote cell adhesion and chemotaxis (14) but also inhibit growth, motility, and survival of endothelial cells (8,15). The RFK sequence in the second type 1 repeat activates latent transforming growth factor-␤1 (16). Additional peptide sequences have been identified that disrupt focal adhesion contacts in endothelial cells (2) or promote adhesion of breast carcinoma cells (17), but the cell receptors interacting with these sequences are not known.
While examining the ability of TSP1 to regulate adhesion of melanoma cells, we found that some type 1 repeat peptides have a similar ability to stimulate melanoma cell spreading on vitronectin as that reported for the CD47-binding sequences from the carboxyl terminus of TSP1 (3). We report here that recombinant type 1 repeats and soluble peptides from these modules containing the WSXW motif specifically stimulate ␣ v ␤ 3 integrin-mediated melanoma cell adhesion and tyrosine kinase signaling when added as soluble peptides but not when immobilized on the substratum. Stimulation of ␣ v ␤ 3 integrin is mediated by binding of this peptide to cell surface HSPGs.

EXPERIMENTAL PROCEDURES
Proteins and Peptides-Calcium-replete TSP1 was purified from human platelets as described previously (18). Recombinant fragments and GST or T7 fusion proteins expressing fragments of TSP1 were prepared as described (15,19,20). Synthetic peptides containing TSP1 sequences were prepared as described previously (2,14,(21)(22)(23)(24). Bovine type I collagen was obtained from Collaborative Research, and human vitronectin was from Sigma. Fibronectin was purified from human plasma (National Institutes of Health Blood Bank) as described (25).
Adhesion Assays-Adhesion was measured on polystyrene or glass substrates (for immunofluorescence) coated with peptides or proteins as described previously (14). Stimulation of melanoma cell spreading on vitronectin immobilized at limiting concentrations was used to detect activation of the ␣ v ␤ 3 integrin as described previously (3). For some experiments, cells were treated with 1 g/ml pertussis toxin (Sigma) as described previously (20). Nonpeptide antagonists of ␣ IIb ␤ 3 (SB208651) and ␣ v ␤ 3 integrins (SB223245) were provided by Dr. William H. Miller (SmithKline Beecham, King of Prussia, PA) (26,27). Cell spreading was assessed microscopically by counting Ͼ100 cells per experimental point using a calibrated reticle to determine the cells/unit area. Cell spread-ing on vitronectin or TSP1 was quantified by counting cells with a typical polygonal or spindle morphology. Cells spreading on peptides 246 and 7N3 typically did not organize actin filaments but were scored as spread if they increased their diameter more than 3-fold. Statistical significance of pairwise comparisons was assessed using a two-tailed Student's t test.
Inhibition of glycosaminoglycan sulfation was achieved by growth of the cells in a sulfate-free Ham's F12 medium containing sodium chlorate as described previously (28). Briefly, A2058 cells were plated in four dishes using complete RPMI medium without antibiotics and containing 10% dialyzed fetal calf serum. After 3 h, media in three of the dishes were changed to Ham's F12 containing 10% dialyzed fetal calf serum with no antibiotics. After 24 h, the medium in the control dish was changed to fresh RPMI medium with 10% dialyzed fetal calf serum, and the media in the remaining three dishes were replaced with Ham's F12, 10% dialyzed fetal calf serum, the same medium with 10 mM sodium chlorate, or with 10 mM sodium chlorate and 1 mM sodium sulfate. Cells were harvested 24 h later by dissociation with EDTA and used for adhesion assays.
Involvement of heparan sulfate and chondroitin sulfate proteoglycans in peptide responses was examined by pretreatment of melanoma cells with heparitinase or chondroitinase ABC (Seikagaku America, Ijamsville, MD) as described previously (29). Briefly, Falcon 1008 bacteriological dishes were coated overnight at 4 o C with 10 g/ml vitronectin or 5 g/ml type I collagen. The solution was aspirated, and the dishes were blocked using 1% BSA in Dulbecco's PBS for 1 h. A suspension of A2058 cells, detached using EDTA, was diluted to 2 ϫ 10 6 /ml in RPMI, 0.1% BSA and treated with 0.015 units/ml heparitinase, 0.1 units/ml chondroitinase ABC, or medium alone for 30 min at 25 o C with rocking. The cells were collected by centrifugation, washed twice with Dulbecco's PBS, and tested for adhesion to both proteins. Adhesion to collagen was used as a control for viable cell number and did not differ significantly among the samples.
Protein Phosphorylation-1008 or 1029 Falcon bacteriological dishes were coated with vitronectin, fibronectin, T7-RGD fusion protein (expressing TSP1 amino acids 879 -947) or TSP1 peptides (388, 7N3, or 246) in Dulbecco's PBS with divalent cations, pH 7.4, overnight at 4°C. The dishes were aspirated and blocked with Dulbecco's PBS, 1% BSA for 60 min at room temperature. Cells were detached with Dulbecco's PBS containing 2.5 mM EDTA and added to coated dishes in RPMI, 0.1% BSA with or without TSP1 peptides 388, 7N3, or 246 at the indicated concentrations and incubated 20 to 60 min at 37°C with 5% CO 2 . The plates were chilled on ice for 5 min, and all cells were recovered for phosphotyrosine assays. The cell medium was pipetted into chilled tubes and centrifuged for 3 min to pellet any unbound cells. Modified RIPA buffer (50 mM Tris/HCl, pH 7.4, 0.15 M NaCl, 1% (w/v) Nonidet P-40, 0.5% (w/v) sodium deoxycholate, 1 mM EGTA, 1 mM Na 3 VO 4 , and protease inhibitors at 10 g/ml antipain, pepstatin A, chymostatin, leupeptin, soybean trypsin inhibitor, aprotinin, and 1 mM phenylmethanesulfonyl fluoride) was added to the pellets and returned to the cells bound to the dishes. After a 30-min incubation at 4°C with modified RIPA buffer, lysed cells were scraped from the dishes and centrifuged at 13,000 ϫ g for 30 min at 4°C. The supernatant was removed from the cell pellet, and 30 l of the sample was heated at 95°C with 2ϫ ␤-mercaptoethanol/SDS sample buffer and held for Western blotting. For immunoprecipitation, the lysates were incubated with monoclonal anti-FAK antibody (mouse IgG, Transduction Laboratories) for 3 h at 4°C. 2.5 l of goat anti-mouse IgG-agarose was added, and the samples were briefly vortexed and incubated overnight at 4°C. The beads were washed and centrifuged three times for 5 s in 1ϫ modified RIPA buffer before resuspending the pellets in 30 l of 2ϫ ␤-mercaptoethanol/SDS sample buffer and boiling. Samples were run on 4 -15% gradient Tris glycine SDS-polyacrylamide gels. Proteins were transferred to nitrocellulose membranes in a semi-dry Western blot apparatus. The membrane was blocked with 1% BSA in TBST (0.1% Tween 20, 10 mM Tris/HCl, pH 7.5, 0.10 M NaCl) for 1 h. Horseradish peroxidaseconjugated RC-20 anti-phosphotyrosine antibody was added and incubated in TBST-0.1% BSA for 1 h at room temperature. The membrane was washed in TBST, 0.1% BSA for 30 min, changing the TBST, 0.1% BSA every 5 min. Antibody binding was detected using ECL reagent (Amersham Pharmacia Biotech) as directed by the manufacturer.
Fluorescence Microscopy-Lab-Tek 177402 glass chamber slides were coated with human vitronectin, fibronectin, TSP1, or TSP1 peptides in Dulbecco's PBS with divalent cations and incubated overnight at 4°C. A2058 cells, grown to 65% confluence, were preincubated for 3 h at 37°C with complete RPMI (10% fetal bovine serum, penicillin/streptomycin, glutamine) with or without 1 g/ml pertussis toxin (Sigma). The slides were aspirated and blocked with Dulbecco's PBS, 1% BSA for 1 h at room temperature. Cells were detached with PBS containing 2.5 mM EDTA and added to the slides at 2 ϫ 10 5 cells per ml in 200 l of RPMI, 0.1% BSA, with or without 1 g/ml pertussis toxin, and with or without TSP1 peptides at 3 M final concentration. After incubation for 60 min at 37°C with 5% CO 2 , the slides were gently aspirated, washed twice in warm Dulbecco's PBS, and fixed in Dulbecco's PBS with 4% formaldehyde for 15 min at room temperature. After washing twice in Dulbecco's PBS, the slides were placed in Dulbecco's PBS with 0.2% Triton X-100 for 4 min at room temperature. The slides were washed twice in Dulbecco's PBS and stained with BODIPY TR-X Phallacidin (Molecular Probes) according to the manufacturer's protocol. Slides were photographed with a Zeiss fluorescence microscope using a ϫ 63 oil immersion objective.

Enhancement of Melanoma Cell Spreading by Soluble TSP1
Peptides-Peptides containing the two VVM sequences from the carboxyl-terminal domain of TSP1 that bind to CD47 stimulate spreading of melanoma cells on substrates coated with low densities of vitronectin (3). This stimulation was concluded to be specific based on lack of activity of the Mal III and Hep III peptides derived from two other domains of TSP1. Although the CD36-binding Mal III peptide from the type 1 repeats was inactive in the previous study (3), a second peptide from the type 1 repeats, peptide 246, strongly induced spreading of melanoma cells on vitronectin (Fig. 1A). The stimulation of spreading was comparable for peptide 246 and the CD47-binding peptide 7N3, and both peptides were active at low micromolar concentrations (Fig. 1B). The activity of peptide 246 was maximal at 3 M but reproducibly decreased at higher concentrations. The stimulation of spreading on vitronectin by peptide 246 required the Trp residues, as peptide 388, with the three Trp residues replaced by Ala residues, was inactive at all concentrations tested (Fig. 1, A and B). Two control peptides for 7N3 with the VVM sequence altered to GGM (604) or VAI (605) were also inactive.
The known cell surface ligands for the type 1 repeat peptides containing the WSXW motif are HSPGs and sulfated glycolipids (21). However, at least two adjacent sequences in the type 1 repeats have been reported to bind to CD36 (13,30,31). To verify that the activity of peptide 246 is independent of CD36 binding, we compared the response of A2058 melanoma cells to that of two other melanoma cell lines that are known to express (C32) or lack expression (G361) of CD36 (32,33). The VVM and WSXW peptides both stimulated spreading of the melanoma cells on vitronectin, regardless of their CD36 phenotype (Fig.  1C, p Ͻ 0.003 for each peptide treatment compared with control for all three cell lines).
The type 1 repeat peptide induced cells plated on a limiting concentration of vitronectin to adopt a similar morphology (Fig.  2C) as induced by the CD47-binding peptide 7N3 (Fig. 2B). Both peptides stimulated spreading on vitronectin and increased the organization of F-actin. The spreading response induced by CD47-binding peptides was reported to be pertussis toxin-sensitive (3). We observed a 63 Ϯ 10% inhibition of the A2058 cell spreading induced by the VVM peptide 7N3 in the presence of pertussis toxin ( Fig. 2E and data not shown). However, stimulation by the type 1 repeat peptide 246 was not inhibited (104 Ϯ 3% of control) in the presence of pertussis toxin ( Fig. 2F and data not shown).
The differential sensitivity to pertussis toxin of the responses to peptide 246 and 7N3 combined with the known ligand binding specificities of these peptides (14,24) suggested that different receptors mediate their effects on ␣ v ␤ 3 integrin function. Because 246 is a heparin-binding peptide (14) and TSP1 binds to HSPGs expressed by melanoma cells (29), we examined the role of glycosaminoglycans in mediating the stimulatory activity of peptide 246. Inhibition of sulfation by preincubating A2058 melanoma cells with the inhibitor of 3Ј-phosphoadenosine 5Ј-phosphosulfate synthesis sodium chlorate at 10 mM ablated the stimulatory activity of peptide 246 but not that of peptide 7N3 (Fig. 3A). This concentration of chlorate inhibited glycosaminoglycan sulfation by 95% in A2058 cells (28). A partial inhibition of the response to peptide 246 but not of the response to peptide 7N3 was also observed by simply growing the cells in a sulfate-free Ham's F12 medium (Fig. 3A). The inhibitory activity of chlorate was verified to be mediated through sulfation inhibition by restoring the stimulating activity of peptide 246 in the presence of both chlorate and 1 mM sulfate (Fig. 3A).
Melanoma cells express both chondroitin sulfate proteoglycans and HSPGs. To determine which mediated the response to peptide 246, A2058 melanoma cells were treated with heparitinase or chondroitinase ABC, and their spreading on vitronectin was assessed in the absence and presence of peptides (Fig. 3B). Chondroitinase treatment did not significantly alter responses to peptide 246 or 7N3 (p Ͼ 0.1 for both peptides versus treatment control), whereas heparitinase strongly and selectively inhibited the enhancement of spreading simulated by peptide 246 (p ϭ 0.0004 versus treatment control). Therefore, HSPGs are required for the response to peptide 246 but not for the response to the CD47-binding peptide 7N3.
Because TSP1 is also an ␣ v ␤ 3 integrin ligand (9), we examined whether these TSP1 peptides modulated A2058 cell adhesion to TSP1. A2058 melanoma cells express ␣ v ␤ 3 (34), and their adhesion on TSP1, a recombinant T7 fusion protein expressing the type 3 repeat modules of TSP1 containing the RGD sequence, or vitronectin was inhibited by an ␣ v ␤ 3 integrin antagonist SB223245 but not by the ␣ IIb ␤ 3 integrin antagonist SB208651 (Fig. 4A). The partial (32%) resistance of TSP1 adhesion to the ␣ v ␤ 3 antagonist suggested that additional TSP1 receptors may contribute to adhesion to the intact protein. We observed induction of spreading on TSP1 by both the VVM peptide 7N3 and the type 1 repeat peptide 246 (Fig. 4B). This induction was specific for the ␣ v ␤ 3 integrin ligands in A2058 melanoma cells, however, because spreading on the ␣ 2 ␤ 1 integrin ligand type I collagen was not stimulated by either peptide 246 or 7N3 (Fig. 4B). Enhancement of spreading on TSP1 was also specific for peptides 246 and 7N3 in that the TSP1 peptides Mal II from the second type 1 repeat, peptide Hep1 from the amino-terminal heparin-binding domain, and two analogs of peptide 7N3 with the VVM sequence disrupted (peptides 604 and 605) failed to stimulate spreading on TSP1 (Fig. 4B). Peptide 246 was more potent than peptide 7N3 for inducing spreading of A2058 melanoma cells on TSP1 (Fig. 4C). The activity of peptide 246 to enhance melanoma cell spreading on TSP1 also required the Trp residues, as peptide 388 was inactive at all concentrations examined.
Based on the activities of recombinant TSP1 fragments, the RGD sequence in the type 3 repeats of TSP1 are primarily responsible for adhesion of A2058 melanoma cells on this protein (Fig. 5A). The entire type 3 repeats are not required, as a GST fusion protein expressing only the last type 3 repeat and the carboxyl-terminal domain (residues 877-1152) promoted strong adhesion, whereas a GST fusion protein expressing a part of the type 3 repeats without the RGD sequence (residues 784 -824) was inactive. The RGD sequence in the type 3 repeats also mediated enhancement of spreading induced by the TSP1 peptides (Fig. 5B).
On an intact TSP1 substrate, enhancement of spreading by soluble TSP1 peptides was observed only at limiting concentrations of TSP1. At higher concentrations of TSP1, additional TSP1 receptors may permit ␣ v ␤ 3 integrin-independent spreading. The amino-terminal heparin binding domain may mediate this peptide-independent spreading, because melanoma cells spread on a proteolytic 140-kDa fragment of TSP1 only in the presence of the activating peptides. The CD47-binding site in the carboxyl terminus may not contribute to this response, because the GST-type 3 fusion protein (residues 674 -925) and the RGD ϩ carboxyl-terminal domain fusion protein (residues 877-1152) both promoted spreading only in the presence of the activating peptides. Although the RGD ϩ carboxyl-terminal FIG. 3. An HSPG mediates the stimulatory activity of peptide 246 on ␣ v ␤ 3 integrin. A, sulfation is required for modulation of ␣ v ␤ 3 integrin function by peptide 246. A2058 melanoma cells were passed into sulfate-free Ham's F12 medium containing dialyzed fetal calf serum. After 24 h, the medium was replaced with the same medium containing 10% dialyzed fetal calf serum or supplemented with 10 mM sodium chlorate (F12 ϩ chlorate) or with 10 mM sodium chlorate and 1 mM sodium sulfate (F12 ϩ chlorate ϩ SO 4 ) and grown for an additional 24 h. Control cells were grown in RPMI 1640 medium with 10% dialyzed fetal calf serum, with replacement of the medium at the same times as the treated cells. The cells were dissociated using EDTA, and adhesion to substrates coated using 10 g/ml (140 nM) vitronectin was determined in F12 medium containing 1 mg/ml BSA (solid bars) or the same medium supplemented with 3 M peptide 246 (striped bars) or 10 M peptide 7N3 (open bars). Cell spreading was determined after 60 min, and the spread cells/mm 2 are presented as mean Ϯ S.D., n ϭ 3. B, heparan sulfate is required for stimulation by peptide 246. A2058 cells were pretreated with 0.015 units/ml heparitinase (striped bars), 0.1 unit/ml chondroitinase (open bars), or medium alone for 30 min at room temperature (solid bars). Cell spreading was determined on substrates coated with 10 g/ml vitronectin and is presented as mean Ϯ S.D., n ϭ 3.
domain fusion protein contains both VVM sequences, these immobilized VVM sequences did not enhance recognition of the RGD in the same protein.
Enhancement of spreading on TSP1 and vitronectin are mediated by ␣ v ␤ 3 integrin, because responses to both peptides were reversed by the ␣ v ␤ 3 antagonist SB223245 (Fig. 6, p Ͻ 0.05 for all conditions treated with SB223245 versus the respec-tive controls). Thus, both the type 1 repeat peptide containing the WSXW motif and the CD47-binding peptide from the carboxyl-terminal domains of TSP1 can synergize with the RGD domain of TSP1 to promote ␣ v ␤ 3 -dependent spreading of melanoma cells.
A soluble GST fusion protein that expressed the type 1 repeat modules containing the TSP1 peptide 246 sequence (residues 385-522) also significantly stimulated ␣ v ␤ 3 activity (Fig.  7). The stimulating activity of the recombinant type 1 repeats was observed using either vitronectin (Fig. 7A) or TSP1 as substrates (Fig. 7B) and was specific in that GST alone or a GST fusion protein expressing the type 2 repeats of TSP1 (residues 559 -669) at the same concentration was inactive. Therefore, the activity of the peptide 246 sequence is replicated by a recombinant fusion protein containing this TSP1 sequence.
As reported previously for the CD47-binding sequences of TSP1 using C32 melanoma cells (3), addition of soluble peptide 246 induced tyrosine phosphorylation of a protein comigrating with FAK in A2058 melanoma cells plated on a vitronectin substrate (Fig. 8A). The control peptide 388 did not significantly enhance phosphorylation. The response to peptide 246 was maximal at 15 M and decreased at higher concentrations in parallel with the decrease in spreading (Fig. 1B). Soluble peptide 246 stimulated a similar level of tyrosine phosphorylation in A2058 cells on a TSP1 substrate (Fig. 8B) that was maximal after 40 -60 min. The phosphorylated protein was confirmed to be FAK by immunoprecipitation using a specific antibody (Fig. 8B).
Because the type 1 repeat peptides displayed a strong adhesive activity for melanoma cells (14,21), we considered the possibility that the peptide 246 response results from a proadhesive activity of the soluble peptide following its adsorption onto the substrate. If so, then direct coating of the peptides should induce the same cell spreading and tyrosine phosphorylation responses as the soluble peptide. Co-immobilization of the peptides with vitronectin was compared with addition of the same peptides in solution for induction of FAK phosphorylation (Fig. 8C). Immobilized peptide 7N3 was more active than the soluble peptide for inducing FAK phosphorylation, but peptide 246 was more active when added in solution.
We compared the morphology and F-actin organization in cells directly attached on the TSP1 peptides to that of cells attached on vitronectin or TSP1 substrates in the presence or absence of the same peptides. Although the cells spread on peptides 246 and 7N3 (Fig. 9, A and B), their morphologies were different from that induced by the soluble peptides added to cells on vitronectin or TSP1 substrates (compare Fig. 9 with Fig. 2, B and C) or by direct adhesion on fibronectin (Fig. 9C). Actin organized only in the periphery of cells on the peptides (Fig. 9, A and B). Although both the 7N3 and 246 substrates promoted some cell spreading, adhesion of cells on peptides 7N3 or 246 rapidly decreased FAK phosphorylation (Fig. 9D). Thus, the direct response of A2058 melanoma cells to these immobilized peptides in the absence of an ␣ v ␤ 3 integrin ligand was inhibition rather than stimulation of FAK phosphorylation. DISCUSSION We have identified the WSXW peptides in the type 1 repeats as a second TSP1 motif that stimulates ␣ v ␤ 3 integrin function in melanoma cells. Stimulation of ␣ v ␤ 3 integrin by both peptides results in similar enhancement of FAK phosphorylation. This response is sulfation-dependent and is mediated by binding of peptide 246 to HSPGs, which are expressed by the melanoma cells and were shown previously to bind TSP1 (29). This activity is expressed both by the synthetic peptide and a recombinant fusion protein containing the type 1 repeats of TSP1. In contrast to the CD47-mediated enhancement by the TSP1 carboxyl-terminal peptides (3), this signal does not re- quire pertussis toxin-sensitive G proteins. Thus, TSP1 contains two independent sequences that can stimulate function of the ␣ v ␤ 3 integrin to interact with ligands including TSP1 itself. This suggests that both the VVM and WSXW motifs may function as synergy sites (35) for ␣ v ␤ 3 -mediated melanoma cell adhesion to TSP1 (Fig. 10). Although soluble TSP1 shows some stimulatory activity for melanoma cell adhesion to immobilized ␣ v ␤ 3 integrin ligands, it is not known which of these two sequences is/are responsible for the activity of the soluble intact protein. Interactions with other extracellular matrix components may also alter the accessibility of both sites in TSP1.
The manner in which the type 1 sequences are presented to the melanoma cells appears to be critical for the resulting signal. Soluble peptide 246 stimulates tyrosine phosphorylation of FAK induced by ␣ v ␤ 3 ligands, but direct adhesion to the immobilized peptide inhibits the same response. Immobilized peptide 246 stimulates melanoma cell adhesion, but the resulting cell morphology differs from that induced on immobilized ␣ v ␤ 3 ligands by addition of the same peptide in solution. These differential responses may have biological significance, because TSP1 is found both as a soluble protein and immobilized in extracellular matrix. TSP1 effects on melanoma behavior in vivo show a similar dichotomy. Addition of soluble TSP1 to melanoma cells prior to injection in an experimental metastasis assay stimulated lung colonization (36), whereas subclones of a murine melanoma cell line with increased TSP1 expression, which may accumulate in an immobilized state in the matrix, were less metastatic in a spontaneous metastasis assay (37). These opposing effects on melanoma tumor progression could also arise from differences in the animal models used (38) but are consistent with the differential activities of soluble versus immobilized TSP1 peptides in vitro to modulate adhesive responses of melanoma cells.
Gao et al. (3) reported that stimulation of adhesion by VVM Immunoblotting with anti-phosphotyrosine antibody was performed using the total lysate in the upper panel or after immunoprecipitation with an anti-FAK antibody in the lower panel. Without soluble peptides, 5% of the cells were spread at each time. Dishes with peptides 7N3 and 246 showed 5% spread cells at 20 min and 20% at 40 min. At 60 min, dishes with peptide 7N3 had 75% spread cells, whereas those with peptide 246 had 85%. C, substrates were coated with vitronectin alone (1st and last two lanes) or mixed with 3 M peptide 7N3 or 246 (2nd and 3rd lanes). Adhesion of A2058 cells to vitronectin alone (VN) or vitronectin co-immobilized with the peptides was determined without further additions (immobilized) or, using a substrate coated with vitronectin alone, with 3 M of the indicated peptides added in the medium (soluble). Equal amounts of cell lysate were immunoprecipitated with anti-FAK and blotted with anti-phosphotyrosine as in B. peptides from TSP1 was not limited to vitronectin. TSP1 is another ␣ v ␤ 3 ligand for which recognition is enhanced by 246 and the VVM peptides. However, the ␣ 2 ␤ 1 integrin is not stimulated by either the WSXW or the CD47-binding peptides in A2058 melanoma cells. This result contrasts with smooth muscle cells, in which VVM peptides enhanced ␣ 2 ␤ 1 integrin function (39).
Several results indicate that adhesive activity is not sufficient for a peptide to stimulate ␣ v ␤ 3 integrin activity in melanoma cells. The Mal II peptide promotes adhesion of these cells (23), but it does not activate ␣ v ␤ 3 integrin. Direct adhesion to peptides 246 and 7N3 in the absence of an ␣ v ␤ 3 integrin ligand induces a signal that opposes that induced by the same peptide when added in solution. Adhesion to immobilized peptides results in inhibition rather than stimulation of FAK phosphorylation, and co-immobilization of peptide 246 with vitronectin did not stimulate FAK phosphorylation as well as addition of the same peptide in solution.
Although the effects of pertussis toxin indicate that a different signaling pathway links peptide 246 binding to integrin activation than that utilized by the CD47-binding peptides, the details of this pathway remain unknown. Lack of pertussis toxin sensitivity for the ␣ v ␤ 3 integrin response to peptide 246 is consistent with our previous report (20) that chemotactic and antiproliferative activities of the same peptide for A2058 cells were not inhibited by pertussis toxin.
Several previous studies (40 -44) have provided precedent for modulation of integrin functions by proteoglycans. A similar mechanism for cooperation between integrins and proteoglycans has been proposed to promote interactions of melanoma cells with both fibronectin and type IV collagen (45,46). Although some of these effects of proteoglycan binding may require only extracellular interactions, studies in melanoma (43,45,46) and other cell types (47) suggest that cell surface proteoglycans can also induce intracellular signals that modulate integrin function. Protein kinase C is implicated in signaling through syndecan 4 (48), regulates stimulation of melanoma cell chemotaxis by TSP1 (20), and can increase integrin activation (reviewed in Ref. 49). Activation of protein kinase C using phorbol ester also induced spreading of A2058 melanoma cells on vitronectin, low concentrations of TSP1, and recombinant fragments of TSP1 containing the RGD sequence. 2 Thus, the stimulatory effect of peptide 246 on ␣ v ␤ 3 may be mediated by protein kinase C.
TSP1 interacts with syndecans as well as other HSPGs on melanoma cells (29,50). Interactions of TSP1 with chondroitin sulfate proteoglycans may also promote cell spreading or formation of microspikes (51), although TSP1 bound to HSPGs but not chondroitin sulfate proteoglycans extracted from melanoma cells (29), and our data demonstrate that an HSPG is required for stimulation of ␣ v ␤ 3 integrin function by peptide 246. Based on the activity of this heparin-binding peptide, simultaneous engagement of the heparan sulfate and ␣ v ␤ 3 integrin-binding sites of intact TSP1 may induce synergistic signaling responses in cells that express both classes of receptors.