Retrovirally mediated expression of decorin by macrovascular endothelial cells. Effects on cellular migration and fibronectin fibrillogenesis in vitro.

Decorin is a member of the widely expressed family of small leucine-rich proteoglycans. In addition to a primary role as a modulator of extracellular matrix protein fibrillogenesis, decorin can inhibit the cellular response to growth factors. Decorin expression is induced in endothelial cells during angiogenesis, but not when migration and proliferation are stimulated. Thus, decorin may support the formation of the fibrillar pericellular matrix that stabilizes the differentiated endothelial phenotype during the later stages of angiogenesis. Therefore, we tested whether constitutive decorin expression alone could modify endothelial cell migration and proliferation or affect pericellular matrix formation. To this end, replication-defective retroviral vectors were used to stably express bovine decorin, which was detected by Northern and Western blotting. The migration of endothelial cells that express decorin is significantly inhibited in both monolayer outgrowth and microchemotaxis chamber assays. The inhibition of cell migration by decorin was not accompanied by decreased proliferation. In addition, endothelial cells that express decorin assemble an extensive fibrillar fibronectin matrix more rapidly than control cells as assessed by immunocytochemical and fibronectin fibrillogenesis assays. These observations suggest that cell migration may be modulated by the influence of decorin on the assembly of the cell-associated extracellular matrix.

Migration and proliferation of endothelial cells and the subsequent establishment of a stable monolayer are critical events in the repair of injured vessels and in angiogenesis and vasculogenesis during development, tumor growth, and tissue repair. Cell migration and proliferation are controlled by growth factors and cell adhesive interactions, which are mediated by cell surface receptors, such as integrins (1) and syndecan-4 (2,3). Thus, control of cell migration may involve a change in cell-matrix interaction characterized by extracellular matrix remodeling and changes in signaling generated by receptors for both soluble and matrix ligands.
Proteoglycans (PGs) 1 are a heterogeneous group of protein families that bear anionic glycosaminoglycan (GAG) chains covalently bound to core proteins. These molecules are prominent constituents of both the extracellular matrix and the cell surface, where they are proposed to play roles in cell adhesion, growth factor interactions, and matrix assembly (4,5). Decorin is a member of the small leucine-rich PG family (6 -8). The association of decorin with collagen fibrils has long been recognized (9,10), and experimental studies indicated that decorin regulates collagen fibrillogenesis (11,12). In addition, decorin may affect cell migration and proliferation both by modulation of interactions of cell surface receptors with their matrix ligands, such as fibronectin and thrombospondin (13)(14)(15), or by an influence on the availability (16,17) and function (18,19) of growth factors that direct these cellular processes.
We have previously reported that the induction of migration in wounded endothelial cell monolayers is accompanied by increased small leucine-rich PG synthesis, which is associated with cells at the wound edge (20). The principal dermatan sulfate PG that is synthesized by confluent cultured aortic endothelial cells is the decorin-related PG, biglycan (21,22), consistent with immunochemical and in situ hybridization studies of endothelia in vessels (23,24). The induction of the synthesis and turnover of biglycan, but not of decorin, is regulated by the release of fibroblast growth factor-2 from endothelial cells in wounded monolayers (25). However, decorin expression, which is not detectable in monolayers of aortic endothelial cells (22), is induced in concert with type I collagen and biglycan when cells undergo sprouting and tube formation in vitro (26). However, the particular cellular processes and the mechanism(s) by which decorin may regulate cell behavior during capillary formation remain unclear. During angiogenesis, both in vivo and in vitro, cell proliferation and migration are first stimulated as cells sprout from the monolayer or existing vessel and then are inhibited as the cells establish a stable vascular structure in association with the newly formed extracellular matrix. Therefore, a role for decorin may involve either a stimulation of cell growth and migration during the induction of neovessel formation or an inhibition of growth or migration by an influence on matrix fibrillogenesis during the stabilization of the developing neovessel.
In these studies, we have induced stable constitutive expression of decorin in macrovascular bovine endothelial cells with the use a replication-defective retroviral vector to examine directly the role of decorin in the control of cell proliferation and migration. We find that cell migration, but not proliferation, is significantly inhibited by decorin expression, both in an assay of monolayer outgrowth and in microchemotaxis assays.
In addition, decorin expression is associated with an acceleration of fibronectin fibrillogenesis, which was assessed as a marker for the assembly of the endothelial pericellular matrix. The enhanced stabilization of a fibrillar pericellular matrix may provide an alternative mechanism by which decorin may inhibit endothelial cell migration. Moreover, these observations suggest that decorin expression may be associated with a later phase of angiogenesis in which the newly formed endothelial cords and tubes are stabilized by the association of cells with the extracellular matrix.

Preparation of Retroviral Vectors for Small Leucine-rich Proteoglycan
Expression-Retroviral vectors were used to prepare pools of stably transduced bovine aortic endothelial cells that constitutively express bovine decorin. Retroviral vectors typically give a very high transduction frequency (27). Therefore, this approach avoids the necessity of targeted cell cloning, which gives rise to experimental clonal variation and rapid senescence of nonimmortalized cells. Replication-defective retroviral expression was used to allow the stable expression of decorin or biglycan in low passage number normal cells. Macrovascular endothelial cells provide a particularly useful parental background to study the effects of decorin expression on cell proliferation and migration because decorin is not expressed in confluent monolayers in vitro (22,25). A retroviral vector for bovine decorin (LDSN) was prepared by ligating the complete coding sequence for bovine decorin (28) (clone Pg28, provided by Marian Young, NIDCR, National Institutes of Health, Bethesda, MD) into the multicloning site of the replicationdefective retroviral vector, LXSN (27) (provided by A. Dusty Miller, University of Washington). A retroviral vector for human biglycan (LBSN) was prepared similarly with the use of a cDNA that includes the complete coding sequence for human biglycan (P16), which was provided by Dr. L. Fisher, NIDCR, National Institutes of Health (29). Prepared vector was transfected into Escherichia coli (DH5␣), and clones in which the insert was oriented correctly were selected by restriction analysis. Analysis of the cDNA sequence through the coding regions of the decorin vector, LDSN, confirmed that the expected coding sequence was present. To prepare replication-defective virus, the mouse 3T3 cell ecotropic packaging line, PE501, was transfected with the vector cDNA by the calcium phosphate co-precipitation technique. Virus-containing supernatants were collected after 16 h and used to infect cells of the amphotropic cell line, PA317. Clones of transduced PA317 cells were grown under Geneticin (G418) selection, and virus was titered on NIH 3T3 TK Ϫ cells. Clones producing between 5 ϫ 10 5 and 5 ϫ 10 6 virus/ml were used to prepare viral supernatants for transduction of endothelial cells.
Cell Culture and Transduction of Cells with Retroviral Vectors for Small Leucine-rich Proteoglycan Expression-Cultures of endothelial cells were isolated from calf thoracic aortas and maintained as described previously (20,22,25). Pools of cells transduced with retroviral vectors were prepared from parental cells between the fourth and eighth passage after isolation and used between the first and sixth passage after selection. For transductions with retroviral vector, cells were plated at 2.5 ϫ 10 4 cells/cm 2 and grown overnight in Dulbecco's modified Eagle's medium with 25 mM glucose, supplemented with nonessential amino acids, pyruvate, and glutamine from 100ϫ stocks (Life Technologies, Inc.) and with 10% fetal bovine serum (Hyclone). Cloned virus-producing cells (PA317) were plated at 5 ϫ 10 5 cells per 6-cm tissue culture dish and grown overnight before the medium was replaced to collect viral supernatants. Sixteen-hour viral supernatants were collected and mixed with fresh medium with 4 g/ml Polybrene (Sigma), which was added to cell cultures and left for 24 h. The transduction of endothelial cells with fresh viral supernatants was repeated up to four times. After transduction, cells were trypsinized and split 1:4, before plating in Dulbecco's modified Eagle's medium with 10% fetal bovine serum and 800 g/ml Geneticin (G-418, Sigma), in parallel with cells that had been mock-transduced. Selection was carried out until all cells in the mock-transduced cultures were dead, a period ranging from 10 to 14 days, with medium changes every 3 days. Cells were then split once more and grown to confluence in the presence of G-418 to prepare cells for use in experiments.
Analysis of Decorin and Biglycan Expression-For analytical separation of radiolabeled PGs on SDS-polyacrylamide gel electrophoresis, samples were partially purified by application to an 0.5 ml DEAE-Sephacel column in 8 M urea with 0.5% Triton X-100, 0.1 M Tris-HCl, pH 7.5, and 0.25 M NaCl (urea buffer), washing with ϳ10 volumes of urea buffer, and then eluting bound macromolecules in urea buffer with 3 M NaCl. Portions of eluted material were precipitated twice by addition of 3.5 volumes of 95% ethanol containing 1.3% potassium acetate and dried. Dried samples were resuspended in deionized 8 M urea, either with or without prior digestion with 0.02 units of chondroitin ABC lyase (ICN Pharmaceuticals) in Tris-HCl buffer, pH 8, containing 15 mM sodium acetate for 3 h at 37°C. Digested and undigested samples in urea were then boiled in SDS-containing sample buffer and applied to 8% SDS-polyacrylamide gel electrophoresis minigels and electrophoresed using a Bio-Rad Transblot apparatus. Prestained protein molecular weight standards (Amersham Pharmacia Biotech) were used to estimate PG size. For Western blotting of core proteins, SDS-polyacrylamide gel electrophoresis gels were equilibrated in 50 mM Tris, 40 mM glycine, pH 9.2, transfer buffer with 20% methanol and 0.0375% SDS (30) and transferred to nitrocellulose (BA83, Schleicher and Schuell, Inc.) for 50 min with a electrophoretic transfer apparatus (Transblot, Bio-Rad). Nitrocellulose membranes were blocked with 2% bovine serum albumin (Fraction V, Roche Molecular Biochemicals) in Tris-buffered saline with 0.05% Tween-20 (TBS-Tween) and exposed to rabbit anti-bovine decorin peptide (LF-94) or rabbit anti-human biglycan peptide (LF-51), which were provided by Dr. Larry Fisher, NIDCR, National Institutes of Health (diluted 1:1000 in blocking solution) overnight at 4°C. After incubation of the blot with alkaline phosphataseconjugated goat anti-rabbit IgG (Jackson ImmunoResearch Laboratories), bands that bound primary antibodies were visualized by an enzyme-linked chemiluminescence procedure (Tropix, Bedford, MA).
To assess total decorin protein synthesis, secreted protein was collected 24 h after medium change (complete medium with 0.2% fetal bovine serum) from confluent LXSN and LDSN-transduced culture pools. Duplicate aliquots of a series of medium dilutions and a series of dilutions of purified bovine decorin (0 -1600 ng, provided by Dr. Kathyrn Vogel, University of New Mexico, Albuquerque, NM) were applied with the use of a slot-blot apparatus (Bio-Rad) directly to nitrocellulose membranes (BA83, Schleicher and Schuell, Inc.) that had been equilibrated in transfer buffer, as described above. Membranes were blocked with 2% bovine serum albumin in TBS with 0.1% Tween-20, pH 7.4, and then treated with 0.02 units/ml chondroitin ABC lyase for 1 h in enriched Tris-HCl buffer, pH 8, with 15 mM sodium acetate and 10 mg/ml bovine serum albumen. Blots were then probed with rabbit anti-bovine decorin (1:1000 LF-94) for 4 h, and bound antibody was visualized on film by enzyme-linked chemiluminescence after secondary antibody incubation as described above. The amounts of decorin secreted by pools of LDSN were calculated by comparison of signals captured on XAR2 film (Eastman Kodak) from medium samples with a plot of signal/amount of purified decorin. Scans of multiple exposures made for different times were quantified with the use of Scion Image (version beta 3B).
Assays of Migration, Monolayer Outgrowth, and Proliferation-Microchemotaxis (Boyden) chamber assays were conducted in a 48-well microchemotaxis chamber (Neuro Probe Inc., Cabin John, MD) using polycarbonate filters with 10-m-diameter pores (Poretics Products, Livermore, CA). The filters were precoated with basement membrane gel matrix by submerging the filter overnight in a solution of 130 g/ml Matrigel in Dulbecco's modified Eagle's medium. In some experiments, filters were precoated with Vitrogen (95-98% type I collagen, Collagen Corp., Fremont CA) by submerging the filter overnight in a solution of 0.1 mg/ml Vitrogen in culture medium. Fetal bovine serum (10%) or, in some experiments, fibroblast growth factor-2 (10 ng/ml, Intergen) was added to the bottom compartment of the microchemotaxis chamber. For conditioned medium experiments, cells were suspended in conditioned medium prepared by harvest after 48 h from transduced LXSN or LDSN cell cultures, either undiluted or diluted to 50% (v/v) with fresh medium. Cells were plated on the upper surface of the filter, and chambers were incubated at 37°C for 5 h. In some experiments, purified bovine decorin (a kind gift of Dr. Kathryn Vogel, University of New Mexico, Albuquerque, NM) that was prepared by dissociative extraction from tendons, which has been shown to be active in collagen fibrillogenesis assays (11), was added to the medium in the chambers. The cells that had migrated to the bottom surface of the filter were counted after fixation and staining, and the results were expressed as cells/mm 2 /filter. An in vitro model for endothelial monolayer outgrowth in vivo was developed to assess long-term monolayer migration of endothelial cells. The primary assay was modified from an under-agarose assay developed for studying the interaction of cells in co-culture (33). Briefly, 6-cm tissue culture dishes, some of which had been previously coated with fibrillar collagen (Vitrogen), that contained 3 ml of solidified 1.5% agarose in phosphate-buffered saline (PBS) were prepared with 4 -6 wells (7 mm) in the agarose. Test cells were plated at confluent density into wells in the dishes of agarose, which had been previously equilibrated in culture medium with 10% fetal bovine serum. Outgrowths areas beneath the agarose overlays were calculated 3 days after plating from the average of two measurements of diameter, taken at right angles to each other, with the initial well area subtracted. In some experiments, 100 g/ml 5-bromo-2Ј-deoxyuridine (BrdUrd) (Sigma) was added to the cultures 18 h prior to fixation. After fixation, cells that had incorporated BrdUrd were identified by indirect immunoperoxidase staining with a monoclonal antibody (6B6, developed by B. Gumbiner and obtained from the Developmental Studies Hybridoma Bank maintained by the Department of Biological Sciences, University of Iowa (Iowa City, IA)). The nuclear labeling index of cells was determined by counting, with the aid of an eyepiece grid reticle, at least 200 total nuclei at a magnification of ϫ 200, within random squares of a (10 ϫ 10) square grid (ϳ500 m 2 ), placed either at the edge or the center of a monolayer outgrowth.
Assay of Fibronectin Fibrillogenesis-An differential extraction assay to quantify fibronectin fibrillogenesis was modified from a method of Sechler et al. (34). Briefly, cells plated in 48-well tissue culture plates (Costar) were washed twice with serum-free culture medium at 37°C and extracted first with 200 l of 2% deoxycholate in 0.02 M Tris-HCl, pH 8.8, at 4°C with protease inhibitors (2 mM phenylmethylsulfonyl fluoride, 2 mM ethylenediaminetetraacetic acid, 2 mM iodoacetic acid, and 2 mM N-ethylmaleimide) and then with the same volume of 10% SDS with 6 M urea, also in the presence of protease inhibitors. The nonionic detergent deoxycholate extracts only cell layer nonfibrillar fibronectin, whereas the second sequential extract solubilizes the remaining fibrillar fibronectin. Samples of the extracts, when Western blotted as described above with the use of 1:1000 rabbit anti-bovine fibronectin antiserum (Chemicon) gave only the expected bands at a molecular mass of ϳ220 kDa. The percent fibrillar fibronectin, defined as the deoxycholate-insoluble fraction of cell layer-associated fibronectin, was quantified by dot blotting (Bio-Rad) duplicate aliquots of samples from the differential extracts of triplicate cultures to nitrocellulose (BA 83) and, after blocking 4 h in 2% bovine serum albumin in Trisbuffered saline with 0.1% Tween-20, immunostaining with 1:1000 rabbit anti-bovine fibronectin antibody, which was detected with an alkaline phosphatase-conjugated goat-anti rabbit secondary antibody (1: 10,000, Jackson ImmunoResearch Laboratories) and enzyme-linked chemiluminescence (Tropix). Signal from the procedure was captured on XAR2 film (Kodak). Scans of multiple exposures made for different times were quantified with the use of Scion Image (version beta 3B).
Immunocytochemisty-Cells were plated on coverslips at various densities and cultured in medium with 10% fetal bovine serum for 3-48 h. Coverslips with cells were washed twice with PBS at 37°C and fixed for 15 min in freshly prepared 3% paraformaldehyde in PBS, pH 7.4, at room temperature. After washing in PBS, cells were blocked in 10% calf serum and 2% goat serum (Sigma Immunochemicals) in PBS. Fixed cells were exposed to rabbit anti-bovine fibronectin (Chemicon, diluted 1:1000 in PBS) at room temperature for 2 h, washed in PBS, and treated with fluorescein isothiocyanate-conjugated goat anti-rabbit IgG (Zymed Laboratories Inc.), diluted 1:200 in PBS, for 1 h, washed and mounted in Gel/Mount (Biomeda Corp.) for fluorescent photomicroscopy (Zeiss).

Analysis of Decorin Expression in LDSN-transduced Endo-
thelial Cells-Cultured bovine aortic endothelial cells that had been transduced with an empty retroviral vector (LXSN) or the same vector containing full-length bovine decorin cDNA (LDSN) were examined by Northern blotting and Western blotting for expression of the transgene (Fig. 1). A vector containing a full-length human biglycan cDNA (LBSN) was prepared as a control and analyzed similarly. Confluent and migrating macrovascular endothelial cells synthesize little endogenous decorin in vitro (22,25), consistent with the lack of detectable mRNA transcripts by Northern blotting (Fig. 1A) and with their very low expression of decorin core protein, as determined by enzyme-linked chemiluminescence assay of Western blots (Fig. 1B) in pools of cells transduced with LXSN vector alone. In contrast, decorin mRNA transcripts (Fig. 1A) and core protein (Fig. 1B) were detected in extracts of cultures that had been transduced with LDSN vector. Secretion of both human biglycan and endogenous bovine biglycan into the medium of cells transduced with LBSN vector is detectable by Western blotting with species-specific antibodies (Fig. 1C,  lanes 1 and 2), whereas vector control cells synthesize only bovine biglycan (Fig. 1C, lanes 4 and 5). Expression of decorin at very high levels in cells transfected with expression vectors can result in the synthesis of some decorin core protein lacking glycosaminoglycan chains (35), although this observation is not universal (36) and is probably related directly to level of decorin overexpression. Aliquots of conditioned medium from  1-4) were selected by growth in G418, replated, and grown to confluence. Medium conditioned overnight by the cells was harvested and processed for Western blotting as described under "Experimental Procedures." Chondroitin ABC lyase-digested aliquots were transferred to nitrocellulose, and decorin core proteins (arrows) were detected with a bovine decorinspecific antibody (LF-94), followed by an enzyme-linked chemiluminescence. Decorin expression was not detectable in LXSN-transduced pools of cells but was present in pools of cells transduced with the LDSN vector. C, culture medium conditioned overnight by either cells transduced with a retroviral vector containing human biglycan cDNA (lanes 1 and 2) or the retroviral vector alone (lanes 3-5) was harvested and processed for Western blotting. Chondroitin ABC lyase digested aliquots were transferred to nitrocellulose. Endogenous biglycan core protein was detected with a bovine biglycan-specific antibody (LF-96, lanes  1 and 4), whereas the product of the transduced human biglycan gene was detected with a human biglycan-specific antibody (LF-51, lanes 2 and 5), followed by an enzyme-linked chemiluminescence. Normal rabbit serum served as a primary antibody control (lane 3). Endogenous bovine biglycan expression was detectable in both human biglycantransduced and vector control cells, but core protein that bound the anti-human biglycan antibody was present only in medium from cells transduced with the LBSN vector.
LXSN and decorin-overexpressing cells were subjected to Western blotting without chondroitin ABC lyase digestion to determine whether decorin core protein that lacked GAG chains was synthesized. No decorin core protein lacking the GAG chain was detected by Western blotting (not shown), indicating that the decorin expressed in these cultures is fully glycosylated. Semiquantitative analysis (see under "Experimental Procedures") of decorin core protein indicated that secretion of decorin by LDSN cultures is approximately 6 g of decorin/10 7 cells/24 h, or approximately 50-fold less than is synthesized by HT-1080 cells transfected with the high expression vaccinia virus/T7 bacteriophage system used by Ramamurthy et al. (35).

Decorin Expression Inhibits Endothelial Cell Monolayer Outgrowth on Plastic or Fibrillar Collagen by a Mechanism That Is
Independent of Inhibition of Cell Proliferation-In a model developed to mimic macrovascular monolayer outgrowth, LDSN and LXSN-transduced cells were plated at confluent density in wells made in agarose inlays either in tissue culture plastic dishes or in dishes previously coated with fibrillar collagen. Although LDSN and LXSN cells in confluent cultures were not obviously different morphologically, LDSN cells at the wound edge of monolayer outgrowths at 72 h after plating appeared more dispersed than LXSN cells (not shown). The area of 72-h outgrowths was significantly less for LDSN cells than LXSN or LBSN cells (Fig. 2), suggesting that decorin expression inhibited monolayer expansion. This relative inhibition of monolayer outgrowth by LDSN cells was maintained on fibrillar collagen, although control (LXSN) cell monolayer outgrowth was also reduced on fibrillar collagen. Cells that overexpressed biglycan (LBSN) did not show a decrease in migration. This observation suggests that the decrease in monolayer outgrowth by decorin-expressing cells may be specific to decorin and is not caused simply by the effects that overproduction of proteoglycan may have on generalized GAG chain synthesis (35).
Because the extent of monolayer outgrowth in the agarose undergrowth assay, as well as in vivo (37,38), is a reflection both of cell migration and cell proliferation and because exog-enous decorin has been shown to inhibit the proliferation of transformed cells (7, 39), we examined cell proliferative index in outgrowths of LDSN and LXSN cells during monolayer outgrowth in vitro (Fig. 3). To assess the proliferative indices of cells in monolayer outgrowths, cultures were treated with BrdUrd 18 h prior to fixation, and cells were immunostained to identify proliferating cells. Few cells in the center of either LDSN or LXSN outgrowths incorporated BrdUrd (ϳ4 -10% labeling index, not shown). However, cells within 500 m (ϳ20 -25 cell diameters) of the outgrowth edge of both LDSN and LXSN monolayers had high labeling indices (91-97%), which did not differ significantly between the two groups ( Fig.  3). As in previous experiments, the monolayer outgrowth area of LDSN cells at 72 h was substantially reduced when compared with that of LXSN cells. In order to assess the contribution of cell proliferation to the extent of monolayer outgrowth in these cultures, cell proliferation was inhibited by the addition of 200 g/ml hydroxyurea 2 h after plating. Interestingly, whereas hydroxyurea reduced the labeling index of cells at the wound edge dramatically for both LDSN and LXSN cells (labeling index ϳ10 -20%), only the monolayer outgrowth area of LXSN cells was significantly affected (Fig. 3). These observations suggest that although cell proliferation makes a significant contribution to monolayer outgrowth area in this assay, it is not a limiting factor for LDSN monolayer outgrowth.
Cell growth curves were used to compare the growth rates of sparsely plated LXSN and LDSN cells directly to confirm that overexpression of decorin in endothelial cells had no overall effect on cell proliferation (Fig. 4). The experiments indicate that for the two independent pairs of transduced cell pools that were examined, cell growth rates during the log phase were identical and reached virtually identical saturation densities at 8 days after plating.
Decorin Expression Inhibits Endothelial Cell Migration in a Microchemotaxis Chamber Assay-A Boyden chamber microchemotaxis apparatus was used to assess inherent differences in LDSN and LXSN cell migration. Chemotaxis of LDSN cells toward 10% serum in the bottom chamber was decreased in 6-h assays when compared with LXSN cells, but in the absence of serum, chemokinesis by cells expressing decorin was also reduced (Fig. 5A). Because decorin binds to collagen, the effect on cell migration of varying the substrate from a basement membrane extract (Matrigel) to fibrillar collagen was also determined (Fig. 5B). In two experiments using matched pairs of LXSN and LDSN-transduced cells, the inhibition of cell migration in decorin-expressing cells was unaffected by migration substrate.
Two different approaches were used to test whether the decrease in migration by decorin-expressing cells was directly dependent on secreted decorin. First, 48-h conditioned medium was collected from decorin-expressing cells and control cells and used either alone or mixed at 50% with fresh medium, in Boyden chamber assays of control cell migration (Fig. 6A). Medium collected from decorin-producing cells inhibited the migration of cells transduced with the vector alone, compared with the addition of conditioned medium from the control cultures. This experiment indicates that a secreted product of LDSN cultures can inhibit cell migration but does not clearly demonstrate that secreted decorin is responsible for the effect. Therefore, exogenous purified decorin was added to the culture medium during the assay of chemotactic response by LXSN cells to determine whether decorin could inhibit cell migration directly (Fig. 6B). Exogenous decorin, added at concentrations ranging from 1 to 100 g/ml to both chambers of the Boyden apparatus, also inhibited LXSN cell migration, with a concentration of between 10 and 100 g/ml causing a level of inhibition comparable to that seen in LDSN cells. Taken together, these experiments demonstrate that secretion of decorin by LDSN cells can cause decreased cell migration in a microchemotaxis assay.
Decorin Expression Enhances Fibronectin Fibrillogenesis by Endothelial Cells-Because decorin binds to several elements of the fibrillar extracellular matrix, and is known to modify collagen fibrillogenesis, fibronectin fibrillogenesis was assayed as a marker for the effect of decorin on pericellular matrix assembly. In the monolayer outgrowth assay, dense mats of fibrils that stained with an anti-fibronectin antibody were present in the centers of the outgrowths, but little fibrillar fibronec-tin was assembled on migrating cells near the outgrowth edge (not shown). When confluent monolayers of LDSN and LXSN cells were examined 16 h after plating, immunofluorescent staining for fibronectin detected a well defined fibrillar fibronectin matrix associated with most LDSN cells (Fig. 7). In contrast, in LXSN cultures, a few discrete foci of cells with well formed fibrillar fibronectin matrices were widely distributed on the coverslip, but most cells had little associated fibrillar fibronectin. Fibronectin mRNA expression was not altered in LDSN cultures as assessed by Northern blotting (not shown).
A differential extraction assay (see under "Experimental Procedures") was used to quantify fibronectin fibrillogenesis in decorin-expressing and control cell cultures, in order to examine more closely the qualitative immunocytochemical observations of enhanced fibronectin fibrillogenesis in cultures in which decorin was expressed. In a representative experiment (Fig. 8), sparsely plated decorin-expressing and vector control cells were assayed for the proportion of cell layer fibronectin that was present as detergent-insoluble fibrils at 6 and 24 h after plating. At both times, decorin-expressing cells had higher proportions of detergent-insoluble fibrillar fibronectin. In the same experiment, some cultures of vector control (LXSN) cells were also plated in the presence of 5 or 50 g/ml of exogenous decorin. LXSN cell cultures at 6 h after plating had an increased proportion of fibrillar fibronectin only at the higher concentration of added decorin. However, the proportion of fibrillar fibronectin in LXSN cultures grown in the presence of exogenous decorin for 24 h was increased significantly over both LDSN and LXSN cell levels, at both added concentrations. These results are consistent with the interpretation that secretion of decorin in LDSN cell cultures results in enhanced or accelerated fibronectin fibrillogenesis. This novel effect of decorin on cell-mediated fibronectin fibrillogenesis may provide a mechanism independent of growth factor-mediated effects for the control of cell migratory behavior. DISCUSSION In this study, we have compared the migration and proliferation of retrovirally transduced decorin-expressing bovine aortic endothelial cells to that of vector control cells, in which endogenous decorin synthesis was not detected. Our results indicate that either the local expression or the addition of purified, exogenous decorin results in the inhibition of endothelial cell monolayer outgrowth. This inhibition appears to be the direct result of the inhibition of cell migration, but not cell proliferation, of decorin-expressing, endothelial cells. Moreover, pericellular matrix assembly, as assessed by assays of cell-associated fibronectin fibrillogenesis, occurs concurrently with increased decorin secretion. The migration of cells that overexpress biglycan is not inhibited, although it should be noted that endothelial cells constitutively synthesize biglycan (21,22,25,26), and therefore, control cells in those experiments have significant levels of endogenous biglycan.
This work was undertaken in light of previous studies that indicated that decorin expression is induced during the process of in vitro angiogenesis (26) but not during cell migration after monolayer wounding (25). Moreover, the presence of decorin in neovessels within granulation tissue (40) and in atheromas (41) is consistent with a role for this small leucine-rich PG in the phenotypic modulation and differentiation of endothelial cells that is required during angiogenesis. However, because decorin synthesis is not induced during cell migration alone, the possibility remains that decorin up-regulation during angiogenesis might be either inhibitory to cell migration or involved in matrix assembly processes that are required during stabilization of neovessels. Our work suggests that decorin may function in both of these capacities.
Several mechanisms by which decorin might regulate cell proliferation or migration have been proposed in other cellular systems. The several models predict that decorin may modulate cellular growth and migration through effects on growth factors that regulate these cellular processes. Observations initially made with Chinese hamster ovary cells that were transfected to cause decorin overexpression indicated that decorin could form complexes with transforming growth factor-␤ (TGF-␤) and inhibit the interaction of this growth factor with its high affinity receptor on the cell (17). The analysis of an interaction between TGF-␤ and decorin by binding analysis in vitro supports this hypothesis (42,43), and an expressed domain within the decorin core protein also binds TGF-␤ in in vitro assays (43). Latent TGF-␤ is synthesized by macrovascular endothelial cells and, when activated, dramatically inhibits proliferation (44,45) and migration (46,47) of these cells. In our experiments, endothelial cells that express decorin migrate less than control cells that do not express decorin. This finding suggests that any potential endogenous inhibition of migration due to TGF-␤ activation in control cultures was not relieved by decorin expression or addition, because that effect would result in an enhancement of cell migration over control. Recent reports that decorin is a ligand for the epidermal growth factor (EGF) receptor (48) provide an alternative growth factor-de-pendent mechanism by which decorin might modify endothelial cell migration. Stimulation of this receptor by decorin has been shown to be the mechanism involved in the down-regulation of cell growth of some tumor cell lines (19). EGF and heparinbinding EGF, which both bind to the EGF receptor, only slightly stimulate the proliferation of endothelial cells but dramatically stimulate endothelial cell migration (49,50). In our cultures, activation of the EGF receptor by decorin would be expected to enhance cell migration, in contrast to our observation. The expression of several matrix proteins, such as fibronectin, is also closely tied to growth factor receptor stimulation. The lack of effect of decorin synthesis on fibronectin expression in LDSN cells is consistent with a mechanism distinct from modulation of growth factor activity. Our results, then, suggest that modification of endothelial cell response to growth factors that interact with decorin and alter growth and migration is not the mechanism by which decorin influences cell migration in our experiments.
In addition to evidence that decorin can directly influence cell growth and migration by interaction with growth factors or their receptors, decorin binds to several matrix proteins and influences extracellular structure (6 -8). In turn, interaction of cells with their pericellular matrix is important in the regulation of cell growth, migration, and differentiation. The core protein of decorin interacts with several matrix proteins, such as collagens (51)(52)(53)(54)(55), fibronectin (14,56), and thrombospondin (15). Decorin has been visualized at the ultrastructural level bound at the d band of fibrillar collagen (10) and in association with fibronectin fibrils at the surface of cells (57). Studies of in vitro fibrillogenesis have shown that the interaction of decorin with collagen regulates the rate of collagen fibrillogenesis as well as fibril size and interfibrillar spacing (11,12,58,59). Collagen fibril bundles in decorin null mice are irregular in shape and size (60), consistent with conclusions derived from in vitro studies. In recent experiments, adenovirus-mediated decorin expression was sufficient to induce phenotypic changes associated with in vitro angiogenesis only when the cells were grown in collagen gels (40). These results are consistent with our earlier study in which decorin expression occurred in concert with collagen I expression in long-term cultures undergoing angiogenesis in vitro (26). Although the effect of decorin on the macromolecular assembly of other matrix macromolecules to which decorin binds has not been previously examined, several studies have characterized the domains of fibronectin that bind decorin and determined that decorin and biglycan can both inhibit adhesion of fibroblasts to fibronectin (13,14). Interestingly, biglycan failed to inhibit monolayer outgrowth in our studies, suggesting that inhibition of fibronectin-dependent cell adhesion does not wholly explain the effect of decorin on cell migration in our studies. In these studies (13,14) and in the EGF receptor studies discussed above (19), the response of cells to exogenously added decorin was found to be dose-dependent, with effective levels of decorin ranging from approximately 10 to 60 g/ml or greater, consistent with the amount of exogenous decorin that was added in our assays. Differences in effective decorin concentrations may be dependent upon the nature of the decorin added. Decorin extracted from tissue and purified with the use of denaturing solvents (13), such as we have used in this study, may be less active in some assays than decorin prepared without denaturation (14,19), or when cellular decorin is supplied directly by endogenous or induced synthesis. Therefore, different experimental dosages of exogenous decorin may be required to produce similar effects as those induced by endogenous cellular synthesis in this study.
We have not assessed the effect of decorin expression on collagen fibrillogenesis, as fibrillar collagen expression and deposition is low in confluent monolayers of endothelial cells (61), and have instead used fibronectin fibrillogenesis as a marker for pericellular matrix assembly in our cultures. We cannot exclude the possibility that the effect of decorin on fibronectin fibrillogenesis in our cultures is indirect or that other changes in the content and structure of the pericellular matrix may occur. However, the enhancement of fibronectin fibrillogenesis by decorin-expressing endothelial cells, as shown in this work, suggests that an effect on pericellular matrix assembly may represent an additional novel mechanism by which decorin may alter cellular behavior and phenotype.
Fibronectin matrix assembly is a dynamic cellular process that results in the generation of an insoluble fibrillar fibronectin matrix from soluble fibronectin dimers. Fibronectin fibrillogenesis requires intracellular signaling, involvement of cytoskeletal elements, and specific fibronectin receptors, which include the ␣ 5 ␤ 1 and ␣ v ␤ 3 integrins that are important to endothelial cell angiogenesis and survival (62,63). Recent work indicates that blockade of fibronectin matrix assembly inhibits smooth muscle cell growth (64). Although engagement of integrins to their matrix ligands is necessary to allow cell attachment, growth, migration, and matrix assembly, there is ample evidence that fibronectin-dependent matrix signaling through integrins can also lead ultimately to inhibition of cell growth and migration. Our observation of dense fibronectin fibrils associated with stationary cells with a low proliferative index at the center of outgrowths rather than with migrating and rapidly dividing cells near the outgrowth edge is consistent with a report that induction of high levels of fibronectin matrix synthesis is associated with reduced growth and an increase in focal adhesion and actin stress fiber formation in endothelial cells (65). Moreover, some of the growth and migration-inhibitory effects of TGF-␤, which induces the expression of fibronectin (66), among other matrix proteins, are exerted through the modulation of extracellular matrix component synthesis (67). TGF-␤ also induces the expression of the fibronectin-binding integrin ␣ 5 ␤ 1 (68), the overexpression of which has been associated with growth inhibition of tumor cells (69). Moreover, the induced growth arrest of some tumor cells is associated with an increased expression of ␣ v integrins (70), including ␣ v ␤ 3 , which has been implicated in fibronectin fibrillogenesis and in endothelial cell survival and rescue from apoptosis (71). Interestingly, decorin expression has also recently been shown to have an anti-apoptotic effect on endothelial cells (40), perhaps consistent with an unreported effect on fibronectin matrix assembly in these cultures. Therefore, induction of matrix assembly, including fibronectin fibrillogenesis, or cell surface receptors that mediate that assembly could result in the down-regulation of cellular migration observed in cultures of decorin-expressing cells, perhaps in association with changes in endothelial cell phenotype, such as those associated with angiogenesis.
Although decorin has long been known to interact with fibronectin, the mechanism by which fibronectin fibrillogenesis is enhanced in cultures of endothelial cells that express decorin is currently unclear. There are several domains within fibronectin that are required for matrix assembly, with the RGDcontaining cell binding domain required for fibrillogenesis initiation, and the N-terminal and C-terminal heparin-binding domains required for de novo assembly of fibronectin into the pericellular matrix (62,72,73). Because chondroitin/dermatan sulfates compete for binding to fibronectin with heparin (74), decorin GAG chains may influence matrix assembly by altering the involvement of the heparin-binding domains of fibronectin that are required for fibrillogenesis. In addition, the core protein of decorin is known to bind via a repeated motif within its internal leucine-rich repeats (56), both to the cell binding domain and the heparin-binding fibronectin domains (13,14), which are required for fibronectin fibrillogenesis. Alternatively, the effect of decorin on fibronectin fibrillogenesis may be just one aspect of a generalized enhancement of assembly and stabilization of pericellular matrix proteins, including other extracellular matrix proteins with which decorin interacts and that are involved in endothelial cell migration and phenotypic changes, such as fibrillar collagen and thrombospondin. For example, fibronectin fibrillogenesis is also modified by interaction of type I collagen (75), which interacts with decorin and is co-expressed with decorin during in vitro angiogenesis (26). Thus, the effect of decorin on endothelial migration may be mediated through complex effects on matrix protein assembly, which now may include, along with more studied effects on collagen fibrillogenesis, a novel effect on fibronectin matrix assembly.