Decorin Regulates Endothelial Cell Motility on Collagen I through Activation of Insulin-like Growth Factor I Receptor and Modulation of α2β1 Integrin Activity*

The proteoglycan decorin is expressed by sprouting but not quiescent endothelial cells, and angiogenesis is dysregulated in its absence. Previously, we have shown that decorin core protein can bind to and activate insulin-like growth factor-I receptor (IGF-IR) in endothelial cells. In this study, we show that decorin promotes α2β1 integrin-dependent endothelial cell adhesion and migration on fibrillar collagen type I. We provide evidence that decorin modulates cell-matrix interaction in this context by stimulating cytoskeletal and focal adhesion reorganization through activation of the IGF-IR and the small GTPase Rac. Further, the glycosaminoglycan moiety of decorin interacts with α2β1, but not α1β1 integrin, at a site distinct from the collagen I-binding A-domain, to allosterically modulate collagen I-binding activity of the integrin. We propose that induction of decorin expression in angiogenic, as opposed to quiescent, endothelial cells promotes a motile phenotype in an interstitial collagen I-rich environment by both signaling through IGF-IR and influencing α2β1 integrin activity.

Decorin is a component of the extracellular matrix, often found in association with collagen I-rich matrices (1,2). As a member of the small leucine-rich repeat proteoglycan family, decorin is composed of a leucine-rich repeat core protein, and a single covalently attached glycosaminoglycan chain of varying length and composition (3,4). Embryonic vasculogenesis and development is unaffected by the absence of decorin (5), which may be explained by compensation from other members of the small leucine-rich repeat proteoglycan family (6). However, the absence of decorin cannot be compensated for in the event of postnatal angiogenesis. Neoangiogenesis in wounded cornea is reduced in the absence of decorin (7). Conversely, neoangiogenesis is enhanced during dermal wound healing (8). These studies indicate a regulatory role for decorin in inflammationassociated angiogenesis. In accordance with these observations, decorin is not expressed by quiescent endothelium but is induced in nascent blood vessels formed under inflammatory conditions (7,9,10). Further, endothelial cells synthesize decorin only when undergoing angiogenic morphogenesis (11), and expression can be induced by the inflammatory mediators interleukin-6 and -10 (12).
However, investigation of whether decorin acts as a pro-or antiangiogenic factor has yielded conflicting results. In a collagen I environment, endogenous decorin enhances tube formation (10), but as a substrate, it inhibits this process (13). Exogenous decorin did not influence tube formation on matrigel (14) but inhibited vascular endothelial growth factor (VEGF) 4 -induced tube formation on this substrate (15). Suppression of VEGF activity through down-regulation of this growth factor has been previously described (14), although VEGF down-regulation could not be demonstrated in another study (7). These differences may in part depend on experimental design, including the manner in which decorin is presented, and may reflect a context-dependent response to decorin. Additionally, however, the use of denaturing agents and/or precipitation steps during isolation may compromise decorin activity, whereas differential post-translational modifications could also contribute. Of particular relevance, two studies have noted varying levels of activity of different decorin preparations in inhibiting tube formation (13,15). Endothelial tube formation is itself a complex process requiring extensive rearrangement of cell-matrix interactions. Therefore, it is essential to investigate the role of decorin in directly regulating endothelial cell adhesion and motility. Decorin has been shown to inhibit endothelial cell adhesion and migration on collagen I (15,16), which could involve inhibition of VEGF activity (15) or promotion of pericellular matrix organization (16). Alternatively, decorin could inhibit cell-matrix interactions by masking integrin binding sites (17). Although these mechanisms would certainly be expected to contribute to modulation of tube formation by decorin, direct interactions of decorin with cell surface receptors could play an essential role in this process. For example, we have shown that decorin core protein signals through the insulin-like growth factor-I receptor (IGF-IR) in endothelial cells (18). Moreover, IGF-IR activation modulates motility in endothelial and other cells by influencing integrin affinity for their matrix ligands (19 -21). Decorin has also been reported to support platelet adhesion via interaction with the collagen I binding integrin ␣2␤1 (22). This interaction could also be relevant for control of endothelial cell behavior by decorin, since ␣2␤1 integrin activity critically modulates endothelial cell motility and capillary morphogenesis in a collagen I environment (23).
Coordinated signals from the extracellular matrix via integrins and from growth factors to their receptors control endothelial cell motility and angiogenic morphogenesis. We therefore hypothesized that the mechanism(s) by which decorin regulates endothelial cell motility could involve direct activation of IGF-I receptor and/or ␣2␤1 integrin on endothelial cells. In this study, we investigated whether activation of these receptors is involved in modulation of endothelial cell motility by decorin.

EXPERIMENTAL PROCEDURES
Protein Production-Decorin was purified from conditioned medium of human skin fibroblasts without denaturing and/or precipitation steps using anion exchange chromatography (DEAE-Sepharose) essentially as described (24). Decorin-containing fractions were then applied to a Resource Q-15 column (Amersham Biosciences), washed with 0.7 M NaCl in 20 mM Tris-HCl, pH 7.4, and eluted with 1 M NaCl in 20 mM Tris-HCl, pH 7.4. Decorin was dialyzed into PBS and concentrated to Ͼ0.5 mg/ml by ultrafiltration (Centriprep YM-10; Amicon) to preserve activity (25), aliquoted, and stored at Ϫ20°C. Decorin core protein was obtained by digestion with chondroitin ABC lyase (Seikagaku Fine Biochemicals), at 0.2 milliunits enzyme/g of decorin in 0.1 M Tris-HCl, 0.03 M sodium acetate, pH 8.0, at 37°C for 2 h, followed by purification on DEAE-Trisacryl M (Serva) as described (26). Decorin proteoglycan and core protein purity was checked by SDS-PAGE followed by silver staining. Immunoreactivity was confirmed using a polyclonal rabbit antiserum to human decorin, 1:500 (27). A 90-kDa core protein complex immunoreactive with decorin antibodies yielded exclusively sequences consistent with decorin as determined by matrix-assisted laser desorption/ionization-time-offlight mass spectrometry and tandem mass spectrometry (Cardiff University Services Facility). 5 Decorin activity was verified by induction of Akt phosphorylation in EA.hy926 cells as previously described (28). Soluble human integrins ␣2␤1 and ␣1␤1 and GST-␣2A-domain fusion protein were recombinantly produced and isolated as described (29,30). Rhodocetin was purified from the venom of the Malayan pit viper (29).
Dermatan sulfate (porcine skin, 90% L-iduronic acid, 10% chondroitin 4/6-sulfate) was obtained from Sigma and dissolved in double-distilled water at 1 mg/ml. Integrin-modulating monoclonal antibodies JA221 and 9EG7 were produced as described (30). The recombinant minicollagen FC3 contains the integrinbinding motif GFPGER within a guest collagen triple helix of 10 GPP triplets. 6 Cell Culture-The human endothelial cell line EA.hy926 (31) was propagated in MCDB 131 medium (Invitrogen) supplemented with 10% fetal calf serum (FCS), 100 M hypoxanthine, 0.4 M aminopterin, 16 M thymidine (Invitrogen), 100 units/ml penicillin, and 100 g/ml streptomycin. For experiments, cells were seeded on a fibrillar collagen I substrate and cultured in Waymouth MAB 87/3 medium supplemented with antibiotics and 0.5% heat-inactivated FCS for a minimum of 24 h. Collagen gels were prepared by coating hydrophilic culture dishes with acid-extracted rat tail tendon collagen immediately following neutralization with serum-free medium/ NaOH as described (10). Where indicated, decorin (0.1-0.6 M) was immediately mixed into the gels prior to setting. Where required, cells were treated with 10 M tyrphostin AG1024 (10 mM stock in DMSO; Alexis) for 1 h prior to stimulation.
Cell Migration Assay-A migration assay based on the work of Korff and Augustin (32) was developed. Briefly, uniform spheroids were generated by incubating 1,000 EA.hy926 cells/ well in hydrophobic 96-well plates (Greiner Bio-one) in Waymouth MAB 87/3 medium supplemented with 5% FCS, antibiotics, and 30% (v/v) methyl-cellulose (from filtered stock solution of 1.3 g/ml methyl-cellulose (Sigma) in Waymouth medium) for 48 h. Spheroids were removed using a stripette and collected by centrifugation at 725 ϫ g for 2 min, washed with serum-free medium, and resuspended in medium supplemented with 0.5% heat-inactivated FCS and antibiotics to a volume of 100 l/spheroid. A 100-l suspension was transferred per well of a 96-well plate containing 50 l of fibrillar collagen I or mixed decorin-collagen I gels. Wells were inspected microscopically, and those containing a single spheroid located close to the center of the well were selected for analysis. Plates were incubated at 37°C and 5% CO 2 , and migration was recorded by capturing phase-contrast images at time of plating and every 24 h thereafter for 5 days. Radial outgrowth was determined from diameter measurements (n ϭ 10).
Cell Attachment Assay-Cells were trypsinized and resuspended in serum-free medium, and 15,000 cells/well were seeded into a 96-well tissue culture plate containing collagen I or mixed decorin-collagen gels. Where required, cells were mixed with soluble decorin immediately prior to seeding. Cells were allowed to adhere for the indicated times at 37°C and washed in PBS. Adherent cells were quantified using crystal violet staining as previously described (33).
Immunocytochemistry-Cells were seeded in serum-free medium at 100,000/well of a 24-well plate onto glass coverslips coated with fibrillar collagen I or mixed decorin-collagen gels. After incubation at 37°C and 5% CO 2 for the indicated time 5 G. Martin, and D. Aeschlimann, and E. Schö nherr, unpublished data. 6 S. Niland and J. A. Eble, manuscript in preparation.
GTPase Assays-Adherent cells (3 ϫ 10 6 cells/10-cm dish) cultured on fibrillar collagen I gels were serum-starved for 24 h prior to stimulation with 0.7 M decorin. At the indicated times, cells were washed with PBS and lysed in 500 l of lysis buffer supplied with the Rac activation assay kit (Cytoskeleton Inc.) per dish. The active form of the small GTPases was captured from 1 ml of cell lysate at 700 g/ml protein using GSTtagged PAK-PBD beads as specified by the manufacturer, and Rac-1 was detected by immunoblotting. Alternatively, active Rac-1 was quantified using the G-LISA TM Rac activation assay kit (Cytoskeleton Inc.) as specified by the manufacturer. Briefly, 3.5 ϫ 10 5 cells/35-mm dish were cultured on collagen I gels, serum-starved for 24 h, and pretreated with inhibitors for 1 h prior to stimulation with 0.7 M decorin for 15 min.
Integrin Binding Studies-Microtiter plates were coated overnight at 4°C with 30 g/ml decorin, 12.5 g/ml FC3 peptide in PBS, or 10 g/ml collagen I in 0.1 M acetic acid. After washing with TBS (50 mM Tris/HCl, pH 7.4, 150 mM NaCl) containing 2 mM MgCl 2 (TBS/Mg), nonspecific binding sites were blocked with 1% BSA in the same buffer (blocking buffer). Integrins were diluted in blocking buffer containing 1 mM MnCl 2 , unless otherwise indicated, and mixed with additives or competitors at specified concentrations immediately prior to plating. For analysis of pH dependence, ampholines (pH 4 -10; Amersham Biosciences) at 40 mg/ml were used as buffering agents. The integrin solution was incubated with immobilized substrates for 2 h at room temperature. After washing, bound integrin was chemically fixed and quantified by ELISA using a rabbit antiserum against the ␤1 subunit (1:600) as described (30). Experiments with GST-integrin ␣2A-domain followed the same procedure, and binding was detected using polyclonal rabbit antibodies to the GST moiety (Molecular Probes) (30).
Statistics-Statistical significance within matched groups was assessed using repeated measures analysis of variance and Dunnett multiple comparison test with p Ͻ 0.05 considered statistically significant (*, p Ͻ 0.05; **, p Ͻ 0.005).

Decorin Enhances ␣2␤1 Integrin-dependent Endothelial Cell
Adhesion and Migration on Fibrillar Collagen I-To investigate decorin effects on endothelial cell motility during sprout formation from preexisting vessels, we adapted a three-dimensional spheroid model. This system has established cell-cell contacts and mimics the quiescent endothelial cell phenotype (32). Transferring spheroids onto the surface of fibrillar collagen I gels induced radial outward migration of endothelial cells (Fig. 1A). Decorin, as a soluble medium supplement (as a model for endothelial cell-secreted decorin), enhanced this process and promoted formation of elongated, sproutlike extensions ( Fig. 1A), an effect that was dose-dependent ( Fig. 1B). Enhanced proliferation did not appear to contribute significantly to this observation, since migration was conducted in low serum (0.5% FCS) and reached a plateau at extended time points (data not shown), indicating the presence of a finite cell population. Further, decorin does not appear to modulate endothelial cell proliferation (15,16). To delineate the mechanism of endothelial cell motility in our system, we used rhodocetin, a specific inhibitor of the major collagen I-binding integrin, ␣2␤1. Rhodocetin reduced outward migration of endothelial cells from spheroids in a dose-dependent manner, and cells adopted a rounded morphology ( Fig. 1, A and C), indicating that motility was, at least in part, mediated by ␣2␤1 integrin. Since decorin deposited by fibroblasts regulates collagen I fibrillogenesis, it could affect ␣2␤1 integrin-mediated cell-collagen I interactions by altering the supramolecular structure of collagen aggregates. To test this hypothesis, collagen was polymerized in the presence of decorin. Under these conditions, endothelial cell migration was also enhanced (Fig. 1D). Alternatively, since decorin has been reported to support platelet adhesion by direct interactions with ␣2␤1 integrin (22), decorin-containing substrates could similarly support endothelial cell attachment and migration. By itself, however, decorin failed to support endothelial cell attachment ( Fig. 2A). Rhodocetin inhibited adhesion to collagen I ( Fig. 2A), indicating that adhesion was mediated, at least in part, by ␣2␤1 integrin. In contrast, decorin promoted endothelial cell attachment to collagen I when presented either as a medium supplement or in a collagen-bound form ( Fig. 2A). Enhancement of adhesion by both soluble and collagenbound decorin was dose-dependent and was apparent up to at least 3 h after seeding ( Fig. 2B and data not shown). To investigate which moiety of decorin was involved in enhancing adhesion, endothelial cells were seeded onto collagen gels in the presence of either decorin core protein or glycos-aminoglycan alone or onto collagen gels that had been prepared in the presence of these moieties. Neither moiety of decorin by itself had any similar effect, indicating that the intact proteoglycan is required to enhance adhesion to collagen I (supplemental Fig. 1). These data demonstrate that decorin, irrespective of its presentation as either a soluble factor or in association with collagen I fibrils, influences the cellular interactions with collagen and that ␣2␤1 integrin plays a central role in this system.
Decorin Induces a Motile Phenotype by Activating IGF-IR and the Small GTPase Rac-To further delineate decorin effects on motility, the cellular phenotype induced by decorin during adhesion to fibrillar collagen I was investigated. In cells adhering to collagen I alone, mature focal adhesions associated with actin stress fibers were formed within 1 h (Fig. 3, A-C, Col I, arrows). In contrast, when cells were seeded onto collagen gels in the presence of soluble decorin or onto collagen-bound decorin, large vinculin-positive focal complexes (Fig. 3, A-C, Dcn and Dcn-Col I, arrows), ruffles, and microspike-like structures (Fig. 3, A-D, Dcn and Dcn-Col I, arrowheads) were frequently observed at the periphery of cells.
In a manner reminiscent of these morphological changes, activation of the small GTPase Rac-1 induces membrane ruf-  fling and formation of peripheral focal complexes (34). Therefore, we investigated whether decorin induces Rac activation. Indeed, Rac activation occurred within 15 min of incubation of decorin with adherent cells (Fig. 3E). Since we have shown that decorin can activate IGF-IR signaling pathways in these cells (18), we investigated whether Rac activation by decorin involved signaling through this receptor. To this end, serumstarved endothelial cells were treated with AG1024, an inhibitor of IGF-IR, or vehicle (DMSO) for 1 h prior to stimulation with decorin. Decorin enhanced levels of active Rac 1.3-fold but could not stimulate Rac activity when IGF-IR activation was blocked (Fig. 3F). To ascertain whether activation of IGF-IR by decorin was therefore responsible for enhanced motility, endothelial cell spheroids migrating on collagen I were treated with AG1024 prior to the addition of decorin. Under these conditions, decorin did not promote endothelial cell migration from spheroids (Fig. 4A). Similarly, blocking IGF-IR activation prevented collagen I-bound decorin from promoting endothelial cell migration (Fig. 4B), indicating that decorin promotes endothelial cell motility by signaling through IGF-IR.
Matrix metalloprotease activity is essential for sprout formation and invasion through interstitial matrices, and endothelial cells in contact with fibrillar collagen I up-regulate MMP2 (matrix metalloprotease 2) (35). Since decorin is a substrate of MMP2 (36), we tested conditioned media at the conclusion of the migration assay for decorin or proteolytically processed products. Since the endothelial cells did not produce endogenous decorin under our noninflammatory culture conditions (Fig. 4C, right, lane 3), in line with previous results (12), the detected decorin must have originated from exogenously added decorin. Only the intact proteoglycan could be detected where decorin had been added to the   -Dcn; B). Endothelial cell migration on collagen I was quantified as in Fig. 1. Data are shown as the mean of 10 spheroids Ϯ S.E. In C, decorin as an additive is not degraded (left); nor do endothelial cells produce endogenous decorin under the assay conditions (absence of inflammatory cytokines) (right). Conditioned medium was analyzed at the conclusion of migration experiments by digestion with chondroitin ABC lyase (ABCase) and immunoblotting with antibodies to decorin core protein, detected with anti-rabbit horseradish peroxidase (left) or anti-rabbit biotin followed by streptavidinhorseradish peroxidase (right). The left panel indicates that soluble decorin was not degraded over the time course of the experiment (Std, purified decorin standard (proteoglycan and core protein); Dcn, soluble decorinconditioned medium). The right panel indicates that decorin was released from decorin-containing collagen gels (1, chondroitin ABC lyase; 2, conditioned medium from Col I-Dcn gel; 3, conditioned medium from Col I gel). Migration of molecular mass markers is indicated on the right. Dcn PG, intact proteoglycan showing up as a characteristic "smear"; arrows, monomeric decorin core protein of 43 and 45 kDa, corresponding to two or three N-linked oligosaccharides, respectively (24); arrowhead, decorin complex of ϳ90 kDa (resistant to boiling in SDS and reduction), consistent with decorin core dimer; asterisk, contaminant originating from chondroitin ABC lyase that reacted with secondary antibodies. media (Fig. 4C, left), indicating that the intact proteoglycan rather than proteolytically processed products modulates endothelial cell motility in our system. However, since the collagen I-immobilized decorin was also found in the conditioned medium (Fig. 4C, right, lane 2), the effects of collagen-bound decorin cannot be easily distinguished from that of soluble decorin.
Decorin Counteracts the Effects of Rhodocetin on ␣2␤1 Integrin Activity-Activation of ␣2␤1 integrin and/or IGF-IR can modulate activity of focal adhesion kinase (FAK), an intracellular integrin-associated protein involved in control of focal adhesion turnover and cell motility. To investigate whether decorin influences FAK activity, endothelial cells were cultured on fibrillar collagen I and incubated with either decorin or rhodocetin. Rhodocetin reduced FAK autophosphorylation at Tyr 397 (Fig. 5A), in accordance with its role as an antagonist of ␣2␤1 integrin. In contrast, decorin did not influence FAK Tyr 397 phosphorylation under the same conditions (Fig. 5B). To test whether decorin instead influenced FAK autophosphorylation during formation of new cell-extracellular matrix contacts, endothelial cells were seeded onto fibrillar collagen I in the presence of soluble decorin or onto mixed decorin-collagen I gels. Decorin did not induce any changes in FAK Tyr(P) 397 compared with cells adhering to collagen I alone (Fig. 5C). Thus, in contrast to rhodocetin, classical integrin-mediated signaling by collagen I is not altered by decorin. Further, decorinmediated enhancement of adhesion appears to be independent of alterations in classical integrin-mediated signaling.
Decorin Modulates ␣2␤1 Integrin Binding Activity to Collagen I-Since decorin promotes motility on collagen I, an ␣2␤1 integrin-dependent process, we investigated whether decorin could modulate ␣2␤1 integrin binding activity to collagen I. In accordance with its role in blocking ␣2␤1 integrin-collagen I interaction, rhodocetin induced a rounded morphology in endothelial cells cultured on collagen I (Fig. 6A). In contrast, decorin induced more subtle changes in cell morphology (similar to those shown in Fig. 3A), and cells remained spread (Fig.  6A), consistent with undisrupted ␣2␤1 integrin-collagen I Rhodocetin reduced FAK phosphorylation after 120 min by 70% (compared with control levels of FAK Tyr(P) 397 at this time), whereas decorin had no effect, as determined by densitometry. C, endothelial cells were seeded onto collagen gels in the absence (control) or presence of 0.5 M decorin, or onto mixed decorin (0.5 M)-collagen gels (Col I-Dcn). At the indicated times, cell lysates were analyzed by immunoblotting for FAK Tyr(P) 397 or ␤-tubulin to demonstrate equal sample loading. interaction and persistent FAK activation. However, when adherent cells were preincubated with decorin for 1 h prior to treatment with rhodocetin for a further 1 h, decorin counteracted the morphological changes induced by rhodocetin (Fig. 6A) and, correspondingly, prevented the reduction in FAK Tyr 397 phosphorylation induced by rhodocetin (Fig. 6B). Simultaneous co-incubation of decorin and rhodocetin showed similar results (Fig. 6C). Therefore, although decorin by itself did not alter collagen I-induced signaling through ␣2␤1 integrin, it interfered with inhibition of collagen I-␣2␤1 integrin interaction by rhodocetin. By ELISA and dot blot assays, we did not detect any direct interaction between rhodocetin and decorin (data not shown). These data therefore provide evidence that decorin and rhodocetin could target the same molecule on endothelial cells: ␣2␤1 integrin.

Decorin Interacts with ␣2␤1 Integrin via Its Glycosaminoglycan Chain at a Site Distinct from That of the Collagen I-and
Rhodocetin-binding A-domain-To further study direct interaction between ␣2␤1 integrin and decorin, we used an established solid-phase assay (29). The ␣2␤1 integrin bound to immobilized decorin in a dose-dependent manner, approaching saturation (Fig. 7A). Similar to collagen I binding of ␣2␤1 integrin, the interaction was divalent cation-dependent, since it was abolished in the presence of EDTA (Fig. 7A). For binding in the presence of Mn 2ϩ ions and the integrin-activating antibody 9EG7, the apparent K D value for ␣2␤1 integrin-decorin interaction was estimated according to Heyn and Weischet (37) to be in the range of 30 -35 nM. The ␣2␤1 integrin recognizes collagen I and rhodocetin via the classical ligand-binding A-domain within the ␣2 subunit. To test whether decorin similarly interacts with ␣2␤1 integrin, we employed a previously characterized GST-␣2A fusion protein (30). No binding of ␣2A-domain to decorin could be detected up to 9 M ␣2A-domain, conditions under which binding to collagen had reached saturation (Fig. 7B). Further, this interaction of decorin is specific for ␣2␤1 integrin, since no interaction of decorin with the related collagen I-binding integrin, ␣1␤1, could be detected (Fig. 7C). This could also indicate that the binding site for decorin is located at least partially within the ␣2 subunit.
The interaction of ␣2␤1 integrin with decorin was then further analyzed under conditions that enhance or inhibit binding to collagen type I (30). Binding of ␣2␤1 to either immobilized collagen or decorin was similarly enhanced in the presence of the activating mAb 9EG7 and Mn 2ϩ over Mn 2ϩ alone (Fig. 7D). However, another activating mAb, JA221, that enhanced integrin binding to collagen did not influence binding to decorin (Fig. 7D). Further, Ca 2ϩ ions, which decrease the affinity of ␣2␤1 integrin for collagen I, could not inhibit interaction of the integrin with decorin (Fig. 7D). Interestingly, the inhibitor rhodocetin, which competitively blocks collagen I-A-domain interaction (30), abolished binding of ␣2␤1 integrin to collagen type I but enhanced integrin binding to decorin (Fig. 7D). We also tested whether rhodocetin could similarly promote binding of the isolated ␣2 A-domain to decorin but did not observe any interaction under these conditions (data not shown). These data suggest that decorin interacts with a site on the ␣2 subunit distinct from the classical ligand-binding A-domain.
To delineate the decorin moiety responsible for interaction with ␣2␤1 integrin, binding of soluble ␣2␤1 integrin to immobilized decorin was assessed in the presence of increasing concentrations of either the intact proteoglycan or its moieties, glycosaminoglycan side chains or purified core protein. Both dermatan sulfate (Fig. 7E) and chondroitin 4/6-sulfate (data not shown) dose-dependently and completely blocked integrin binding to decorin. In contrast to the glycosaminoglycan-bearing entire decorin molecule, decorin core protein did not consistently or strongly interfere with this interaction (Fig. 7F). These data indicate that the glycosaminoglycan moiety of or collagen I (10 g/ml; Col I) was incubated with different concentrations of GST-␣2A-domain fusion protein in binding buffer, and binding was detected as above using antibodies to GST. In C, immobilized decorin, collagen I, or BSA was incubated with 100 nM ␣1␤1 integrin ectodomain in binding buffer (Control) and additional supplements as indicated. Values for BSA-coated wells were not subtracted as background in this instance and are shown (BSA). D, immobilized decorin or collagen I was incubated with 100 nM ␣2␤1 integrin ectodomain in binding buffer (Control) and additional supplements as indicated (200 nM mAb JA221, 2 mM Ca 2ϩ , 2 M rhodocetin). E and F, immobilized decorin was incubated with 100 nM ␣2␤1 integrin ectodomain in binding buffer in the presence of increasing concentrations of dermatan sulfate (DS) (E) or purified decorin core protein (core protein) or the entire decorin molecule (Dcn PG) (F).
decorin plays a key role in the interaction of ␣2␤1 integrin with decorin.
To test whether decorin binding to ␣2␤1 integrin influences the interaction with collagen I, a recombinant minicollagen, FC3, was employed. FC3 contains an ␣2␤1 integrin binding site but does not interact with decorin. ␣2␤1 integrin bound dosedependently to FC3 in a cation-dependent manner (Fig. 8B). However, premixing of ␣2␤1 integrin with decorin reduced integrin interaction with FC3 (Fig. 8B). Since decorin does not interact with either FC3 or the collagen I-binding A-domain of the integrin ␣2 subunit, a competitive mode of inhibition does not appear likely. However, the consistently reduced binding signals indicate that decorin could compromise the collagenbinding capability of ␣2␤1 integrin by inducing/stabilizing a less active conformation.

DISCUSSION
The role played by decorin in angiogenesis is controversial and is further complicated by the large number of potential interactions of decorin with both matrix components and cell surface receptors. Angiogenesis is mediated by growth factors acting in concert with extracellular matrix-derived signals; hence, decorin could contribute to angiogenesis in a contextual manner, as a matrix organizer and/or through direct signaling. In this study, we sought to examine the role of decorin by using an easily manipulated in vitro model in order to minimize inter-ference from the complexity of the physiological angiogenic environment.
Using this model, we present evidence that decorin regulates angiogenesis through the following two molecular mechanisms. 1) Decorin-IGF-IR interaction on endothelial cells triggers Rac activation, which is in line with actin cytoskeleton rearrangement and enhanced motility. 2) Decorin influences activity of ␣2␤1 integrin, one of the major collagen I receptors on endothelial cells.
At first glance, the concentrations of decorin used in this study (100 -700 nM) may seem rather high, although they are consistent with those found in collagen I-rich connective tissues (38,39). Albeit unknown, the local concentration of soluble decorin synthesized by sprouting endothelial cells could conceivably be high and might also locally accumulate by association with collagen I. Further, since decorin may be dimeric in solution (4), the concentration of a biologically relevant form of decorin may in fact be substantially less than that calculated based on a monomeric molecular mass. It is also to be expected that greater concentrations of decorin might be required to elicit effects similar to those of IGF-I. First, the affinity of decorin for the IGF-I receptor is 10-fold less than that of IGF-I (18), and second, the amount of decorin available to interact with the IGF-I receptor would be less than that added, due to competing interactions of decorin with both collagen I matrices and with alternative cell surface receptors (such as ␣2␤1 integrin). For comparison, maximal stimulation of endothelial cells is achieved with 50 -200 ng/ml (7-28 nM) IGF-I, whereas 100 nM decorin already stimulated adhesion and motility (although higher concentrations are required for statistical significance). Given these considerations, the concentrations employed in this study do not seem unreasonable.
Decorin on its own did not support endothelial cell adhesion but induced a motile phenotype and promoted adhesion and migration on fibrillar collagen I, suggesting that it acts as a modulator of integrin-extracellular matrix interactions. In contrast, decorin previously inhibited endothelial cell adhesion and migration on collagen I (15,16). Different sources of decorin or use of natively purified decorin, as in the current study, could have contributed to these contrasting results. Further, we studied motility of endothelial cells from three-dimensional spheroids rather than single cell suspensions or monolayers. Integration of endothelial cells into spheroids prevents apoptosis, induces quiescence, and results in established cell-cell contacts (32), which may result in differential cell responses, and more closely reflect the in vivo situation during endothelial cell sprouting from existing vessels.
Additional to direct activation by extracellular ligands, integrin activity and cell motility are controlled by inside-out signaling, whereby growth factor receptor activation disrupts intracellular integrin subunit interactions to influence ligand binding at the ectodomain (40). Therefore, inside-out signaling by decorin through IGF-IR could be responsible for modulation of ␣2␤1 integrin-mediated motility, in accordance with our observation that decorin-mediated enhanced motility was dependent on IGF-IR activity. It could be argued that the IGF-IR inhibitor AG1024 had effects other than blocking IGF-IR signaling, since treatment with inhibitor alone already FIGURE 8. Decorin recognizes a less active conformation of ␣2␤1 integrin and modulates ␣2␤1 integrin-collagen I interaction. A, binding of ␣2␤1 integrin ectodomain (100 nM) to immobilized decorin (300 nM) or collagen I (10 g/ml) was performed in ampholine-buffered saline adjusted to the indicated pH. B, different concentrations of ␣2␤1 integrin ectodomain were incubated with immobilized collagen peptide FC3 (12.5 g/ml) in the absence or presence of 200 nM decorin. EDTA-containing controls that represent divalent-cation independent, nonspecific interactions are also shown. reduced migration. However, at the concentration employed, AG1024 has been shown to be highly specific for the IGF-I receptor (41). Additionally, EA.hy926 cells may endogenously produce IGF-I, as has been shown for the parental HUVECs (42). Thus, autocrine stimulation could have induced a basal level of IGF-R-I/PI3K/Rac activation that was blocked by AG1024. Moreover, our data are in accordance with a previous study, where decorin promoted fibroblast motility by activating small GTPases (43). In addition, IGF-IR activation has been shown to promote ␣2␤1 integrin-dependent motility and membrane ruffling (21), and Rac activation has been shown to promote ␣2␤1 integrin-mediated cell spreading on collagen I (44).
Of physiological relevance, endothelial cells undergoing vessel morphogenesis in an inflammatory environment initiate expression of decorin (7, 9 -11). Additionally, increased IGF-IR expression was observed in sprouting vessels in the decorin null mouse model, whereas long term exposure of decorin to endothelial cells in vitro resulted in receptor down-regulation (18). Thus, control of IGF-IR signaling and availability on endothelial cells could be an important mechanism by which decorin regulates endothelial cell responses during angiogenesis.
Another mechanism by which decorin could contribute to control of angiogenesis is through direct interaction with and modulation of ␣2␤1 integrin activity. Of relevance, a recent study indicated that decorin competed with ␣2␤1 integrin for binding to collagen I (17). However, this would not explain enhancement of endothelial cell adhesion to collagen I by decorin in our system. Further, decorin interacts with collagen peptides that do not contain the ␣2␤1 integrin interaction site (45)(46)(47)(48), whereas ␣2␤1 integrin, but not decorin, interacted with the minicollagen FC3. Therefore, although decorin and ␣2␤1 integrin binding site(s) on collagen I could overlap, alternative, independent binding sites also appear to exist.
Our data suggest that unlike collagen I (49) and rhodocetin (30), decorin does not interact with the classical ligand-binding site, the A-domain of the ␣2 subunit (50). Previously, the monoclonal antibody P1E6, directed against this site (51), inhibited decorin binding to ␣2␤1 integrin (22), indicating a similar mode of interaction. However, since decorin does not interact with recombinant ␣2 A-domain, these findings cannot be easily explained by an overlap of the P1E6 epitope and decorin binding site. Like rhodocetin, binding of P1E6 to the A-domain may lead to a conformational change that induces and attenuates, respectively, a decorin binding-competent conformation of the integrin. In turn, it may be speculated that decorin binding to ␣2␤1 integrin outside of the A-domain induces an integrin conformation with lower collagen I binding activity. Such an allosteric mechanism would explain the reduced binding signals of ␣2␤1 integrin to the minicollagen FC3 even at saturating concentrations of the collagen fragment. Since decorin did not interact with the minicollagen, indirect effects of decorin on the collagen suprastructure were excluded. Cell-based assays also indicated that decorin can act independently of its role in regulating collagen fibrillogenesis, since soluble decorin (added after fibril formation) and collagen gels formed in the presence of decorin elicited similar endothelial cell responses.
In contrast to the antagonistic ␣2␤1 integrin inhibitor rhodocetin, decorin alone could not alter collagen I-␣2␤1 integrin-triggered activation of FAK. This may seem surprising, since peptides derived from decorin core protein altered VEGFinduced FAK phosphorylation (15). However, it was not shown whether these peptides alone modulated FAK phosphorylation, and in agreement with our data, IGF-I-induced, FAK-independent motility has been previously observed (52,53). Further, decorin alone could not alter FAK phosphorylation; nor as a substrate could it support integrin activation required to mediate endothelial cell adhesion. These observations are in accordance with a lack of interaction of decorin with the A-domain of the ␣2 subunit, since interactions with this domain would be expected to result in transduction of ligand occupancy to intracellularly associated FAK. Hence, we consider decorin neither as an antagonist nor as an agonist of ␣2␤1 integrin but rather as a modulator of its collagen I binding activity.
In the cell-free system, the glycosaminoglycan moiety, but not the core protein of decorin, was principally responsible for interaction with ␣2␤1 integrin. Nevertheless, the intact proteoglycan was required to modulate cell adhesion to collagen I. This could indicate that a simultaneous interaction of decorin with collagen I (via the core protein) and ␣2␤1 integrin (via the glycosaminoglycan) could act as a bridge and thereby modulate collagen I-␣2␤1 integrin interaction, potentially increasing the number of ␣2␤1 integrin-collagen contacts of a cell. However, it could also be speculated that decorin glycosaminoglycan-␣2␤1 integrin interaction in conjunction with decorin core protein-IGF-IR interaction (18) could influence stability and cross-talk within cell surface receptor complexes. Indeed, the large number of potential interactions of decorin with multiple matrix components and cell surface receptors makes it difficult to clearly ascertain the contribution of individual interactions to a complex process such as angiogenesis. Despite this, direct interaction of decorin with cell surface receptors of endothelial cells has been highlighted in this study as likely to play an important role in regulating endothelial cell motility. Further, the activities of both ␣2␤1 integrin and small Rho GTPases, which have been shown in this study to be modulated by decorin in endothelial cells, are critically involved in vessel formation and maturation (23,54,55), providing potential mechanisms for decorin in modulating capillary morphogenesis.
In conclusion, we suggest that decorin modulates endothelial cell behavior independent of masking integrin binding sites and of its role in collagen fibrillogenesis. We also provide evidence that decorin influences ␣2␤1 integrin-mediated endothelial cell responses by signaling through IGF-IR and/or by allosteric modulation of ␣2␤1 integrin activity.