Collagen Phagocytosis by Fibroblasts Is Regulated by Decorin*

Decorin is a small, leucine-rich proteoglycan that binds to collagen and regulates fibrillogenesis. We hypothesized that decorin binding to collagen inhibits phagocytosis of collagen fibrils. To determine the effects of decorin on collagen degradation, we analyzed phagocytosis of collagen and collagen/decorin-coated fluorescent beads by Rat-2 and gingival fibroblasts. Collagen beads bound to gingival cells by α2β1 integrins. Binding and internalization of decorin/collagen-coated beads decreased dose-dependently with increasing decorin concentration (p < 0.001). Inhibition of binding was sustained over 5 h (p < 0.001) and was attributed to interactions between decorin and collagen and not to decorin-collagen receptor interactions. Both the non-glycosylated decorin core protein and the thermally denatured decorin significantly inhibited collagen bead binding (∼50 and 89%, respectively; p < 0.05). Mimetic peptides corresponding to leucine-rich repeats 1–3, encompassed by a collagen-binding ∼11-kDa cyanogen bromide fragment of decorin and leucine-rich repeats 4 and 5, previously shown to bind to collagen, were tested for their ability to inhibit collagen bead binding. Although the synthetic peptide 3 alone exhibited saturable binding to collagen, neither peptides 3 nor 1 and 2 markedly inhibited phagocytosis. Leucine-rich repeat 3 bound to a triple helical peptide containing the α2 integrin-binding site of collagen. When collagen beads were co-incubated with peptides 3 and 4, inhibition of collagen phagocytosis (55%) was equivalent to intact native/recombinant core protein. Thus a novel collagen binding domain in decorin acts cooperatively with leucine-rich repeat 4 to mask the α2β1 integrin-binding site on collagen, an important sequence for the phagocytosis of collagen fibrils.

The intracellular phagocytic pathway in fibroblasts contributes to the physiological remodeling of collagen by lysosomal degradation of internalized collagen fibrils (1)(2)(3)(4), but the mechanisms that regulate this pathway in vivo are poorly characterized. Previous morphological studies have shown that collagen fibrils are "decorated" by proteoglycans (5); consequently, decorin may affect the binding step of collagen phagocytosis (6).
Decorin (DCN) 1 is a matrix proteoglycan that belongs to the small leucine-rich proteoglycan family (7). The mature form of DCN (ϳ100 kDa) consists of an ϳ45-kDa core protein, a single dermatan or chondroitin sulfate glycosaminoglycan chain, cysteine loops near the N and C terminus, and either two or three asparagine-bound oligosaccharides. The central part of the core protein consists of 10 leucine-rich repeat (LRR) sequences in tandem array (8). Rotary shadowing-electron microscopy and molecular modeling studies suggest that the DCN core protein is horseshoe-shaped (9,10) and that the inner concavity accommodates and may provide a binding site for type I collagen (10). Indeed, DCN binds not only to type I but also to collagen types II, III, VI, and XIV (11)(12)(13)(14)(15). DCN binding to collagen molecules is thought to influence collagen fibrillogenesis and the final diameter of fibrils. The DCN core protein may mediate these interactions (16), but the impact of DCN-collagen binding interactions on collagen phagocytosis is not known.
Triple helical collagen type I possesses a specific DCN core protein-binding site at the d-band in each D-period (15,17) and a second DCN core protein-binding site located in a narrow region ϳ25 nm from the C terminus of type I collagen in a zone that coincides with the c 1 band of the collagen fibril D-period (18). However, the structural elements of DCN that mediate binding to collagen type I are not completely defined. Although analyses of the DCN core protein show that neither the Nterminal half nor the central LRR repeats can, by themselves, bind tightly to fibrillar collagen (19), more recent work indicates that LRRs 4 -6 may contain putative high affinity binding site(s) for collagen type I (20 -22).
Despite these reports describing putative structural determinants that determine the interactions between DCN and collagen type I, there is no consensus model for binding. Furthermore, previous studies describe interactions between DCN and collagen in the context of regulation of fibrillogenesis. Currently, there are no reports describing a role for DCN as a modulator of collagen degradation by phagocytosis. Here we examined the effect of DCN on the collagen binding and internalization steps of phagocytosis by using quantitative flow cytometry (23). Based on the data obtained with DCN digestions and DCN mimetic peptides, we have identified putative collagen-binding sites in DCN that may be involved in the regulation of collagen phagocytosis.
Cell Culture-Rat-2 cells (CRL 1764, ATCC, Manassas, VA) were cultured in Dulbecco's modified Eagle's medium containing 5% fetal bovine serum and antibiotics (0.017% penicillin G; 0.01% gentamycin sulfate). Rat-2 cells exhibit several phenotypic traits in common with gingival fibroblasts, including rapid collagen phagocytosis (3), and were used to model collagen binding and internalization processes. In some experiments, human gingival fibroblasts (HGF; passages 5-12) cultured in minimal essential medium with antibiotics and 10% (v/v) fetal bovine serum were used in bead binding experiments with cells grown to confluence. All bead incubations were conducted for 1 h at 37°C in serum-free conditions.
Collagen and DCN Bead Coatings-Carboxylate-modified, yellowgreen polystyrene beads (2 m diameter; Molecular Probes, Eugene, OR) were coated with collagen, and fibril formation was induced as described previously (24). In some experiments, monomeric collagen was bound to beads by using nonpolymerizing conditions during the bead preparations. Nonspecific binding sites were blocked with 0.1% (w/v) BSA. For preparation of DCN/collagen beads, collagen beads were prepared, but instead of resuspending the bead pellet in PBS, the pellet was resuspended in varying concentrations of DCN (0.01, 0.1, and 1.0 M), sonicated, incubated at 37°C for 1 h in the DCN solution, and counted. For calculation of molar concentrations, the total molecular weight of either recombinant DCN (38 kDa) or of bovine articular cartilage DCN (100 kDa from manufacturer's specifications and our own estimates) or the various peptides (molecular weights provided by the supplier) were used to obtain the appropriate concentrations. For bead binding assays, cells were incubated at a ratio of 4 or 6 beads:cell (6) for 1 h at 37°C in serum-free conditions. For triple helical collagen peptide beads, carboxylate-modified, nonfluorescent beads were coupled by Arg residues using carbodiimide with coupling efficiencies of ϳ2-3%.
Immunostaining of DCN/collagen-coated beads and immunoblotting of bead-associated proteins were done to estimate binding of DCN to the collagen beads and quantified by flow cytometry and densitometry. Briefly, DCN/collagen-coated, nonfluorescent polystyrene beads were stained with mouse anti-bovine decorin antibody (1:10) at 37°C for 1 h, washed, counterstained with FITC-goat anti-mouse antibody (1:10 dilution; 1 h at 37°C), and analyzed by flow cytometry to quantify DCN on the collagen bead. Bead-associated proteins were analyzed by immunoblotting and quantified by comparison with standards.
Collagen Bead Binding-We optimized concentrations of DCN that affected collagen bead binding by dorsally loading cells with either collagen or DCN/collagen-coated beads for 1 h at 37°C. Bead binding to cells was quantified by flow cytometry (6,23,25,26). In some experiments, two-color flow cytometry was used to analyze simultaneously the binding of collagen-and DCN/collagen-coated beads to cells. Red and blue fluorescence beads (2.0 m; Molecular Probes; red beads, Ex max ϭ 580 nm and Em max ϭ 605 nm; blue beads, Ex max ϭ 365 nm and Em max ϭ 415 nm) were coated with collagen (COL-RED and COL-BLUE) or collagen and DCN (DCN/COL-RED and DCN/COL-BLUE) as described above.
Cell Attachment and Spreading-Fibrillar collagen substrates were created on 35-mm non-tissue culture dishes (0.3 mg collagen/ml; pH 7.4) and blocked with 0.1% BSA. In some experiments, collagen surfaces were further coated with DCN (1 M; 500 l) for 1 h at 37°C. Cells were plated at 1 ϫ 10 4 cells/cm 2 in serum-free medium, allowed to attach and spread for 1 or 3 h, washed, fixed with formaldehyde, permeabilized with 0.03% Triton X, and stained with rhodamine phalloidin to visualize actin filaments. Cell attachment and spreading of individual cells was imaged at the substrate-cell interface by confocal microscopy.
To determine whether DCN affects cell attachment to collagen, 8-well tissue culture surfaces were prepared with DCN coating as described above. Cells (4 ϫ 10 4 cells/well) were allowed to attach for 1 or 3 h in serum-free medium. At each time point, the medium was removed, and the cells were gently jet-washed in PBS to remove any loosely bound cells. Adherent cells were trypsinized, suspended in isotonic solution, and counted.
Cell-surface DCN-HGFs were incubated with DCN (10 M solution) for 15 min at 4°C, washed, incubated at 4°C for 1, 10, and 60 min in medium, and washed, trypsinized, and neutralized with growth medium. Cells were pelleted, resuspended in normal goat and mouse sera (1:1000 each), incubated on ice for 10 min, washed, sedimented, resuspended in 1:50 solution of mouse anti-DCN antibody in PBS, and incubated on ice for 1 h. Cells were washed, counterstained with FITCgoat anti-mouse antibody, incubated on ice for 1 h, washed, sedimented, resuspended in PBS, and analyzed by flow cytometry to assess cell surface staining of DCN.
DCN Cleavage-For preparation of DCN core protein (cpDCN), DCN was digested with chondroitinase ABC (0.2 units/ml) in 0.1 M Tris-HCl (pH 8.5) at 37°C for 1 h. Digestion was confirmed by Coomassie Bluestained 10% SDS-PAGE. The effect of cpDCN on inhibition of binding was examined using collagen beads coated with the cpDCN (as described above) at equimolar concentrations as used for intact DCN. Cells were incubated with these beads for 1 h at 37°C, and bead binding was assessed by flow cytometry.
For CNBr cleavage, DCN was dissolved in trifluoroacetic acid; CNBr was added, and the mixture was incubated for 4 h at room temperature. Distilled water was added to each tube and evaporated to dryness, and the residue was rehydrated. Control samples confirmed that the cleavage occurred as a result of CNBr digestion and not as a consequence of acid hydrolysis. Collagen beads were prepared and coated with 30 l of the CNBr digestion mixture for 1 h at 37°C. The beads were washed with PBS, and bound DCN fragments were eluted with SDS sample buffer or acetic acid. The eluates were separated on 10 -20% Tris-Tricine gradient gels and silver-stained, and the relevant bands were analyzed by mass spectrometry.
Preparation of Gel Samples for Mass Spectrometry-In-gel tryptic digests were used to prepare eluate bands for mass spectrometry. Briefly, each silver-stained gel band was minced, destained, reduced, and alkylated prior to trypsin digestion. The gel pieces were vacuumdried and then digested with sequencing grade trypsin (Roche Applied Science) overnight at 37°C. After quenching with trifluoroacetic acid, the supernatant was retained, and the gel was twice extracted with acetonitrile. The combined supernatant and extracts were vacuumdried prior to MALDI-mass spectrometry protein identification and sequencing.
Mass Spectrometry-For mass spectrometry (Mass Spectrometry Facility, Faculty of Medicine, University of Toronto) to identify putative collagen-binding sequences within the DCN core protein, saturated ␣-cyano-4-hydroxycinnamic acid in 70% acetonitrile, 0.1% trifluoroacetic acid was used as the matrix solution. Protein enzymatic digestion mixtures (1 l) were spotted on the sample target, and saturated matrix solution (1 l) was added. After crystal formation, the sample target was analyzed by MALDI-mass spectrometry in linear mode on an Applied Biosystems Voyager-DE STR MALDI-time of flight mass spectrometer (337 nm laser). Acceleration voltage was 20 kV; grid voltage was 94%; guide wire was 0.05%; delay time was 175 ns, and low mass gate was 800 Da. The mass spectra were externally calibrated from molecular weights of a mixture of standard peptides. Masses of fragments from experimental samples that did not appear in the mass spectrometry analysis of controls were analyzed by MALDI-mass spectrometry to provide predicted sequences that were then probability matched for DCN using standard proteomics data bases and search engines.
Mimetic Peptides-Peptides were synthesized based on the LRR sequences 1-5 of DCN. Control peptides contained the same amino acids but with scrambled sequences (Advanced Protein Technology Centre Peptide Synthesis Facility, University of Toronto). The sequences for the LRR 1 and scrambled peptides were LGLEKVPKDLPPDTALLD-LQNNKI and GEVKLLDPTLKPNIQKLDDNLAPL, respectively. The sequences for the LRR 2 and scrambled peptides were TEIKDGDFKN-LHTLILINNK and GFELNIKLLNIDIITKHNDTK. The sequences for the LRR 3 and scrambled peptides were SKISPGAFAPLVKLERLYL-SKNQL and LKIAPALSGFPVRSELKYSNKLLQ. A portion of each of these LRR 1-3 peptides was biotinylated at the N terminus for use in matrix protein binding and competitive inhibition studies. The sequences for the LRR 4 and scrambled peptides were LQELRVHENEI and VNQIERLHELE, respectively. The sequences for the LRR 5 and scrambled peptides were VELGTNPLK and LTPGNKLVE. The amino acid sequences of the peptides were based on the sequencing results obtained from mass spectrometry indicating binding to LRR 1-3 and other studies, which suggest that putative collagen-binding motifs are located within the LRR 4 and LRR 5 of the DCN molecule (20,21). Nonfluorescent (2.0 m diameter) collagen beads were blocked with 0.1% BSA, washed in PBS, centrifuged, and resuspended in a 0.1 or 1 mM solution of each of the biotinylated LRR peptides and incubated for 1 h at 37°C. The beads were stained with FITC-streptavidin (1:500) and washed, and peptide binding was analyzed by flow cytometry.
Competitive Inhibition of Peptide Binding and Collagen Bead Binding-Collagen beads prepared as described above were incubated with increasing concentrations of biotinylated LRR 3 peptide (1-1000 M) and stained with FITC-streptavidin as described above. Bead fluorescence was analyzed by flow cytometry. In competitive inhibition assays, collagen beads were prepared and blocked with BSA, coated with biotinylated DCN peptides for 1 h at 37°C, pelleted, resuspended in a solution of 100ϫ excess of unlabeled peptides for 1 h at 37°C, stained with FITC-streptavidin, and analyzed by flow cytometry. Positive control beads were prepared in the same manner as the experimental beads but were not subjected to incubation with unlabeled peptides. For examination of the effects of peptides on collagen bead binding, collagen beads were incubated with either single peptides (0.1 or 1.0 mM solution; 30 min) or all possible peptide pair combinations of LRR 1-5 and then added to cells for flow cytometry analysis of bead binding as described above.
Circular Dichroism-For assessment of the efficacy of thermal denaturation, CD spectra for bovine articular cartilage decorin were collected on an AVIV circular dichroism spectrometer (model 62A DS) at a concentration of 0.20 mg/ml in 10 mM Tris-HCl (pH 7.5). For each sample, five scans were collected from 200 to 260 nm in 1-nm increments in a 1-mm cuvette and averaged. The temperature of the sample was recorded at 25 and 70°C. The loss of ellipticity at 205 nm was recorded at each temperature as a measure of thermal stability as described earlier (27).
Statistical Analysis-For all comparisons utilizing continuous variables, Student's t tests were performed. Statistical significance was set at p Ͻ 0.05. Experiments with multiple groups were analyzed by ANOVA. Post hoc comparisons were done by Tukey's test. All experiments were repeated at least three times.

Effect of DCN on the Binding
Step of Collagen Phagocytosis-Quantification of collagen binding to beads (2 m diameter) by using biotinylated bovine type I collagen, detection with streptavidin, blotting and comparison with standards showed that when beads were incubated in a collagen solution (3 M; 500 l), there was 0.026 pg of collagen/bead. Binding to collagen or bare beads was proportional to DCN concentration in the incubation solution (200 l; Table I). The moles of DCN bound/bead and the number of moles of collagen bound/bead were similar for DCN incubation solutions (Table I). We also used immunostaining and flow cytometry to assess DCN binding to collagen beads. These experiments showed progressive increases of fluorescence with increasing DCN concentrations (ϳ20 -30-fold compared with uncoated bead controls; p Ͻ 0.01; Fig. 1A). Immunostaining of DCN on collagen-coated beads was specific as beads without DCN coating showed very low levels of fluorescence.
We assessed differences between collagen and DCN/collagen bead binding in HGF and Rat-2 cells to determine whether DCN affected collagen bead binding and to assess the degree of equivalence between the two cell types. Collagen beads coated with DCN (1 M) significantly inhibited collagen bead binding in both HGFs and Rat-2 cells compared with cells incubated with collagen beads (Table II; comparison of inhibition between HGF-Rat 2 cells, p Ͼ 0.2). There was also no significant difference in the inhibitory effect between recombinant human DCN and bovine DCN in HGFs (Table II;  DCN also inhibited the ability of cells to bind one or more beads per cell ( Fig. 1, C-H). The inhibitory effect was not likely due to the ability of DCN to bind to cells directly as DCN-coated beads showed very little attachment to cells (Fig. 1, E and F). Generally, the percentage of cells binding DCN/collagen-coated beads was slightly higher than DCN-coated beads alone. Indeed, we found it difficult to obtain images of DCN-coated beads bound to individual cells because the beads bound in such low numbers and were so loosely attached that they were easily dislodged by gentle washing. Time course experiments were performed to assess the effect of DCN on collagen bead binding (1, 3, 5 h; Fig. 1I). There was a time-dependent increase in the percentage of cells binding collagen beads over time (ANOVA, p Ͻ 0.001). In contrast, the near 10-fold inhibition of collagen bead binding by DCN was sustained over a 5-h period (p Ͻ 0.001, comparison of collagen with DCN/collagen at each time point). In some experiments we compared fibrillar collagen beads with monomeric collagen beads. With biotinylated bovine type I collagen and streptavidin-peroxidase assay, we found similar amounts of collagen binding if the collagen was polymerized by incubation at pH 7.4 prior to incubation with cells or in monomeric form (0.026 pg of collagen/bead). The degree of DCN-induced inhibition of binding was sharply reduced using monomeric collagen (only 20% inhibition at 0.1 M DCN on collagen beads compared with 90% inhibition with fibrillar collagen).
The ␣2␤1 integrin is the major collagen receptor on human gingival fibroblasts and mediates collagen phagocytosis by these cells; the other ␤1 integrins do not make a major contribution to phagocytosis of fibrillar collagen (24). We examined the impact of blocking collagen bead binding in HGF with the use of an antibody to the ␤1 integrin (clone 4B4; 1 g/ml) followed by incubation with collagen beads, a protocol that abrogates collagen binding to the ␣2␤1 integrin in these cells. Collagen bead binding was reduced by Ͼ4-fold with antibody, but binding of DCN-collagen beads was not reduced any additional amount with antibody treatment (Table III).
Internalization of Collagen-As DCN significantly inhibited collagen binding, we assessed whether this effect would impact on the subsequent internalization step (28). Cells were top-loaded with either FITC-collagen or DCN/FITCcollagen-coated nonfluorescent latex beads for 1, 3, and 5 h. At each specified time point, 0.2% trypan blue was added to the culture medium to quench extracellular FITC fluorescence, thereby allowing us to quantify the proportion of internalized beads for individual cells. For collagen bead con-  Competition Experiments-Dual color flow cytometry was used to assess whether the DCN-induced inhibition of collagen bead binding was because of competition between DCN and collagen for binding to collagen receptors on the same cell. Cell cultures were co-incubated with COL-RED beads and DCN/ COL-BLUE beads at saturating bead concentrations (8 beads: cell), and bivariate plots of bead binding were analyzed (Fig. 2,  A-C). Cells that were loaded simultaneously with COL-RED beads and DCN/COL-BLUE showed an ϳ5-fold lower percentage of bead binding for DCN/COL-BLUE beads compared with COL-RED beads (p Ͻ 0.005; Fig. 2D) but no effect on collagen bead binding. These data and the very low binding of DCN beads to cells (Fig. 1) show that DCN inhibits collagen bead binding and internalization.
Although these experiments indicated that DCN inhibits collagen binding to cells by interfering with the ability of collagen to interact with collagen receptors, it is also possible that DCN may bind to cell surface receptors directly and thereby interfere with collagen binding. Accordingly, we examined the ability of DCN to bind to HGFs as the collagen receptors on these cells have been examined extensively in the context of collagen phagocytosis (24). Cells were incubated with 10 M DCN in solution for 1, 10, or 60 min at 4°C to block internalization. Cell surface binding was measured by immunostaining for DCN and quantified by flow cytometry. Cell surface DCN staining for HGFs was not significantly higher than controls ( Fig. 2E; p ϭ 0.6 by ANOVA). This effect was not because of trypsin-induced degradation of DCN during cell dissociation because trypsin did not significantly reduce immunostaining of DCN-coated beads (DCN-collagen beads without trypsin treatment ( Fig. 2E; p Ͼ 0.2). Furthermore, the lack of binding of soluble DCN was not because of internalization because cells at 4°C that were incubated with DCN, fixed, and immunostained for surface DCN showed no fluorescence above background levels. Thus, binding of DCN to human gingival fibroblasts was experimentally insignificant.
Cell Attachment and Spreading-As cell attachment and spreading are critical early steps in collagen phagocytosis (6), we asked if DCN affects these processes when cells were plated on collagen substrates. Cell attachment to DCN/collagencoated substrates was reduced with increasing DCN concentration. Compared with cells plated on collagen, there were significant reductions in cell attachment for cells plated on DCN/ collagen substrates with 0.1 and 1.0 M DCN (ϳ2-fold, p Ͻ 0.05; Fig. 3A). For spreading studies, cells were plated on collagen-coated surfaces and allowed to attach and spread for 1 or 3 h in the absence of serum followed by staining with rhodamine phalloidin to visualize actin filaments (Fig. 3, B-E). Cells that attached and spread in the absence of DCN showed abundant actin filaments and exhibited spreading within 1 h. At 3 h, there was further enhancement of actin filament staining and increased cell spreading. Cells that were plated on DCN/collagen-coated substrate ([DCN] in solution,1.0 M) were able to adhere to the substrate; however, spreading and actin filament formation were minimal.
Effect of DCN Structure on Collagen Binding-We assessed if the DCN core protein mediates DCN-induced inhibition of collagen binding. DCN core protein was produced by digestion with chondroitinase ABC, and the efficacy of this treatment was examined by SDS-PAGE. Whole, undigested DCN exhibited a molecular mass of ϳ100 kDa, whereas the digested sample exhibited bands at ϳ45 and 48 kDa, corresponding to the DCN core protein (Fig. 4A). With this digestion protocol, we assessed whether the DCN core protein could inhibit bead binding by human gingival fibroblasts. DCN core protein (1 M) significantly inhibited collagen bead binding (ϳ50% of controls, p Ͻ 0.01; Fig. 4B). This reduction was not because of large reductions of DCN core protein binding to the collagen beads as SDS-PAGE showed little difference in the abundance of whole DCN or DCN core protein that was eluted from the beads. Compared with DCN core protein, recombinant DCN (1 M) inhibited collagen bead binding by ϳ30% more (Table II). At coating concentrations of 10 M core protein, there was more inhibition of collagen bead binding (75 Ϯ 7% reduction).
We also considered the importance of the glycosaminoglycan chain of DCN in the inhibition of collagen bead binding. Cells incubated with collagen beads coated with either 1 M dermatan sulfate or 1 M chondroitin sulfate showed no inhibition of bead binding compared with collagen bead controls (dermatan sulfate 98 Ϯ 4% of controls, p Ͼ 0.2; chondroitin sulfate, 99 Ϯ 3% of controls, p Ͼ 0.2). Incubation of beads with Alcian blue showed increased staining when collagen beads were coated with glycosaminoglycans, indicating that dermatan sulfate and DCN/collagen-coated bead binding. Cells bound significantly fewer DCN/collagen-coated beads than collagen-coated beads for each time point (p Ͻ 0.01). J, time course and dose dependence of collagen and DCN/collagen-coated bead internalization using trypan blue to quench fluorescence from bound, extracellular beads. With increasing concentration of DCN in the incubation solution, internalization was significantly reduced (p Ͻ 0.01; ANOVA). Data in all graphs are presented as mean Ϯ S.E.

TABLE II Decorin inhibition of collagen binding
Collagen beads were incubated with vehicle or the indicated decorin preparations in solution and then added to Rat-2 cells or HGF for 1 h. The % of cells with bound beads was determined by flow cytometry. The data are the % reduction of bead binding compared with vehicle-treated collagen bead controls. There were no significant differences between inhibition of collagen bead binding of Rat-2 cells and human gingival fibroblasts (p Ͼ 0.2) or between bovine articular cartilage decorin ( Fig. 4B). Because the native structure of DCN did not appear to be required for collagen binding, we used CNBr to digest the DCN core protein into smaller peptides. Collagen beads were coated with CNBr-digested DCN (1 M), and bead binding to cells was compared with collagen beads. The CNBr DCN fragments inhibited bead binding by 88 Ϯ 3% (n ϭ 4 independent samples) compared with collagen beads, indicat- ing that CNBr-derived peptide sequences within the DCN core protein can inhibit collagen binding.
CNBr digestion generated distinct DCN fragments that were resolved by silver staining of Tris-Tricine gradient gels (Fig.  5A). Based on the predicted cleavages from the primary amino acid sequence of DCN, we found a mixture of complete and incompletely digested DCN fragments ranging in size from 2.6 to 11 kDa (Fig. 5A). We determined which CNBr-digested fragments bound to collagen-coated beads by incubating CNBrdigested DCN with collagen beads. An 11-kDa fragment eluted from the collagen beads was analyzed by mass spectrometry and sequencing. The sequences from the tryptic digest that were positively matched to the bovine decorin sequences were VVQCSDLGLEK and ISPGAFAPLVK which correspond to LRR regions 1-3 in the ϳ11-kDa CNBr-digested fragment.
Peptide Binding Experiments-To assess the collagen binding ability of the LRR 1-3 peptides, we synthesized biotinylated peptides corresponding to each LRR and detected peptide binding to collagen beads with FITC-streptavidin. Beads were incubated in peptide solutions of 10 M. The LRR 3 peptide bound to collagen much more abundantly than did the LRR 1 and 2 peptides as shown by an ϳ18-fold increase in mean fluorescence/bead compared with negative controls (p Ͻ 0.005; Fig. 5B). Although the LRR 2 peptide also showed some binding to collagen, the mean fluorescence/bead was at least 3-fold lower than that observed with LRR 3. Evidently, LRR 1 did not demonstrate any appreciable binding to collagen, as the mean fluorescence/bead was comparable with that seen for negative controls. Negative control samples consisted of collagen beads incubated with FITC-streptavidin alone, scrambled peptides, and beads that were preincubated with whole DCN. These samples showed no fluorescence above background.
Competitive Inhibition of Peptide Binding-As the LRR 3 peptide bound more abundantly to collagen than the LRR 1 and 2 peptides, it was used to assess the specificity of binding to collagen in more detailed binding and competition experiments. Dose-response experiments demonstrated large and significant increases in biotinylated LRR 3 peptide binding with increasing concentrations (1-100 M peptide), after which a plateau was observed ( Fig. 6A; difference in mean fluorescence/ bead between 1 and 10 M peptide was ϳ8-fold; p Ͻ 0.005). For competitive inhibition of peptide binding, beads were prepared, blocked with BSA, coated with biotinylated peptides (10 M), and incubated with 100ϫ excess of unlabeled peptide. The difference in mean fluorescence/bead between the positive control and competition group was ϳ3.4-fold (p Ͻ 0.0005; Fig. 6B), indicating that the biotinylated peptide was released and that the unlabeled peptide effectively competed binding by the labeled peptide.
We measured binding of LRR 3 and scrambled sequence LRR 3 peptide to type I collagen-derived triple helical peptides from type I collagen as well as to BSA. The first triple helical peptide included residues ␣1(I)-(496 -507) (high affinity ligand binding to the ␣2 integrin subunit; K D ϭ 1.1 ϫ 10 Ϫ 6 M), and the second peptide included residues ␣1(I)-(772-786) which may mimic a DCN-binding site. Peptides (at 100 M) were attached to 2 M nonfluorescent, carboxylate-modified beads by carbodiimide coupling, and elution experiments followed by amino acid analysis showed that both triple helical peptides were bound to beads. With the use of biotinylated LRR 3 and the scrambled LRR 3 peptide, followed by streptavidin-FITC staining and measurement by flow cytometry, we found nearly 3-fold higher binding of LRR 3 to the triple helical collagen peptide (residues ␣1(I)-(496 -507) containing the binding to the ␣2 integrin subunit) than that exhibited by the scrambled LRR 3 sequence (Table IV).
Effect of DCN Mimetic Peptides on Collagen Bead Binding-We assessed the effect of the LRR peptides on collagen binding by Rat-2 cells. In addition to the peptides corresponding to LRRs 1-3 as described above, small peptides corresponding to the ␤-pleated sheet regions of LRRs 4 and 5 were also synthesized, based on the possibility that they may impact on collagen binding (20,21). When collagen beads were incubated with single peptides (0.1 or 1 mM; from LRRs 1-5), there was significant peptide binding to collagen beads, but there was only marginal reduction of collagen bead binding to cells (LRRs 1-3, p Ͼ 0.2; data not shown; LRR 4 ϳ10%, p Ͼ 0.1; LRR 5 ϳ27% p Ͻ 0.05; see Fig. 6C). We also examined inhibition of binding using all possible combinations of pairs of peptides from LRR 1-5. A combination of LRR 3 ϩ 4 peptides (0.1 mM each) reduced collagen bead binding by ϳ55%; p Ͻ 0.01; Fig.   6C), but none of the other combinations caused significant reduction of binding (p Ͼ 0.2). Collagen bead binding was not affected by the corresponding scrambled control peptides. DISCUSSION The intracellular phagocytic pathway in fibroblasts is an important step in the degradation of collagen under physiological conditions (1,4). Collagen fibrils are coated by proteoglycans in vivo (5,29), and decorin-deficient mice exhibit enhanced collagen degradation (30). Consequently, we considered that decorin may affect collagen phagocytosis. Our central finding is that DCN, at densities as low as 6 ϫ 10 Ϫ22 mol/m 2 of DCN on bead surfaces, can inhibit internalization of collagen. These findings may have important implications for physiological connective tissue remodeling and wound healing because they point to an important role for decorin, and possibly other proteoglycans that bind to collagen, as critically important inhibitors of the phagocytosis step in the collagen degradation pathway. This proteoglycan-mediated inhibitory mechanism provides a novel corollary to the matrix metalloproteinase pathway of collagen degradation and its inhibition by tissue inhibitors of matrix metalloproteinases.
The biological significance of our data depends on the validity of the DCN/collagen bead model system for phagocytosis. As immunostaining demonstrated that DCN bound to collagencoated beads, this model system mimics the in vivo situation in which proteoglycans, possibly including DCN, are thought to "decorate" collagen fibrils (5,31). Furthermore, data obtained with the in vitro collagen phagocytosis bead model likely reflect authentic regulatory systems in vivo as previous studies (23,26,32) have demonstrated remarkable parallels between reduced collagen bead binding and drug-induced inhibition of collagen degradation in vivo. Notably, the amount of DCN on beads that was required to produce inhibition of collagen binding by cells was very low (3 ϫ 10 Ϫ20 mol of DCN bound/bead) and approached an equimolar ratio (0.34) with the amount of collagen on beads (8.7 ϫ 10 Ϫ20 mol of collagen bound/bead).
DCN and Collagen Binding-Previous studies on DCN-collagen interactions have largely focused on modulation of collagen fibrillogenesis (9,10,20,(32)(33)(34). The novelty of the current report is that DCN inhibits the collagen binding step of phagocytosis as demonstrated by a consistent and marked reduction in collagen bead binding. Although previous studies (20 -22) examined how DCN binding to collagen fibrils regulates fibrillogenesis, these results may also relate to the inhibitory effect of DCN on the collagen binding step of phagocytosis because DCN regulation of both fibrillogenesis and phagocytic degradation requires binding to collagen. Indeed, a three-dimensional model of decorin (10), as well as observations that several small leucine-rich proteins may lie within the concavity of the putative DCN structure (9), predicts close interactions between DCN and fibrillar collagen. Whereas these interactions may stabilize and orient fibrils during fibrillogenesis (35), in the context of collagen phagocytosis they may also inhibit the binding of collagen to integrins. The inner surface of the putative arch-shaped DCN molecule contains a series of charged residues that may facilitate adhesive interactions with triple helical type I collagen (10,20) and possibly with collagen molecules. Notably, the bead surface in our model system contains a mixture of both fibrils and collagen molecules so there are likely heterogeneous interactions between DCN and the collagen on the bead, an issue that has become more complex with recent reports showing that DCN may exist as dimers (36).
We considered that decorin may bind to cells independent of collagen, and subsequently inhibit collagen binding as a result of DCN receptor-mediated processes. The observation that very low numbers of HGF or Rat-2 cells bound beads coated with DCN alone was consistent with the finding that surface staining of HGF incubated with soluble DCN was negligible. Thus under our experimental conditions, any observed binding of DCN to cells was likely nonspecific and was of very low affinity. The apparent absence of DCN receptors in HGF is in contrast to other reports demonstrating receptor-mediated endocytosis of DCN by human skin fibroblasts (8) and of tumor cells (37,38). In contrast, using gingival fibroblasts, DCN receptors are evidently not expressed or they are not functional.
We considered that DCN regulates collagen phagocytosis by masking the ligand-binding sites on collagen required for cell attachment and ultimately internalization. As we found that inhibition of monomeric collagen binding to cells by DCN was minimal, DCN interactions with collagen that impact on phagocytosis evidently require fibrillar collagen. Furthermore, as the ␣2␤1 integrin is required for collagen phagocytosis by human gingival fibroblasts (6,23,24,26), a likely site for DCNinduced masking of fibrillar collagen binding is the ␣2␤1 integrin-binding site within collagen (39). Accordingly, we used DCN peptides (see below) that bind optimally to collagen, and we examined their interaction with triple helical collagen peptides mimicking the ␣2␤1 integrin-binding site of collagen (40) and a second DCN-binding sequence within collagen (41,42). Based on binding of the LRR 3 DCN peptide (but not the scrambled sequence peptide) to both of these triple helical collagen peptides, short collagen-binding sequences within DCN may mask the ␣2␤1 integrin-binding site on collagen, which is required for the phagocytosis of collagen fibrils. There appear to be multiple binding sites within collagen for DCN (42), which include sequences modeled by the collagen peptide FIG. 5. Binding of LRR mimetic peptides to collagen. A, Tris-Tricine 10 -20% gradient gel shows CNBr digestion of bovine DCN (lane 1), flow-through fraction of CNBr DCN digest that did not bind to collagen-coated latex beads (lane 2), and SDS and acetic acid eluates of bead-associated CNBr digest fragments from collagen beads (lanes 3 and 4, respectively). Note that the indicated band corresponding to ϳ11 kDa appears particularly enriched in the SDS eluate; also, the band at ϳ9 kDa is enriched in all lanes. M are molecular mass standards. The numbers to the right (lane T) are the theoretically predicted masses of the CNBr-digested fragments. B, biotinylated LRR 3 peptide binds to collagen. Mean fluorescence/bead was 18-fold higher compared with negative controls (p Ͻ 0.005; n ϭ 10,000 beads/analysis). ␣1(I)-(772-786) (41). Whereas the LRR 3 DCN peptide apparently bound to this sequence, it is not known to mediate cell attachment via integrins and consequently may not be important in regulating collagen phagocytosis.
Importance of DCN Structure on Inhibition-In agreement with previous studies, we found that the DCN core protein is critical for binding to collagen fibrils (43,44). Although dermatan sulfate and chondroitin sulfate bound to collagen beads, they did not, by themselves, influence collagen bead binding. Heat disruption of the tertiary structure of DCN did not substantially affect binding to collagen, indicating that binding domains within the DCN core protein (10), which are not reliant on the native structure of DCN, may be responsible for inhibition of collagen binding.
Several studies have attempted to define collagen binding domains in DCN (20 -22), including putative regions in the LRRs 4 -6. Currently, there are no definitive reports of binding inhibition by short peptides. Although recombinant truncated forms of DCN core protein expressed in bacteria have been used to demonstrate two putative sites that bind fibrillar collagen (20), solubilization and folding have complicated the interpretation of these studies. Accordingly, we synthesized FIG. 6. LRR peptide binding inhibition studies. A, dose-response of LRR 3 peptide binding to collagen. Significant increases in LRR 3 peptide binding to collagen is observed between 0.001 and 0.1 mM (n ϭ 10,000 beads/analysis). B, competitive inhibition assay of LRR 3 peptide with a 100-fold increased concentration of unlabeled versus labeled peptide causes an ϳ3.4-fold reduction in fluorescence between competition and positive control groups (p Ͻ 0.0005). C, inhibition of collagen bead binding by mimetic peptides showing collagen beads, collagen beads coated with 0.1 M DCN, or collagen beads coated with peptides corresponding to ␤-sheet regions in LRR 3 ϩ 4 (100 M), LRR 4 (1 mM), and LRR 5 (2 mM) in DCN. The LRR 3 ϩ 4 peptide inhibits collagen bead binding by ϳ50%. LRR4A/collagenand LRR5A/collagen-coated beads were bound ϳ10 (p Ͼ 0.1) and ϳ27% (p Ͻ 0.05) less, respectively, compared with collagen-and scrambled peptide (LRR4B, 1 mM; LRR5B, 2 mM)-coated controls (n ϭ 10,000 cells per analysis). Data are mean % cells binding collagen beads Ϯ S.E. small peptides that correspond to sequences within LRRs 1-5. A 9-residue peptide based on the LRR 5 containing glutamate 180 weakly inhibited collagen bead binding by cells, whereas the corresponding scrambled sequence did not affect collagen binding. These results are consistent with earlier data (22) demonstrating that glutamate 180 was critical for DCN/collagen binding. Our peptide inhibition experiments showed that single, short peptide sequences are unlikely to affect collagen binding, whereas a combination of peptides (LRR 3 and LRR 4) were strongly inhibitory, suggesting that cooperativity between LRRs is required for collagen binding.
The inhibition of collagen binding and subsequent internalization indicates that DCN may block collagen phagocytosis in vivo. Thus DCN may be an important determinant of connective tissue homeostasis because DCN coatings on collagen fibrils may prevent, for example, inappropriate collagen degradation in early stages of wound healing. In addition to its collagen binding properties, DCN can bind and neutralize significant amounts of transforming growth factor-␤, a potent, pro-fibrotic cytokine (45,46). Thus in wound healing, DCN may serve a dual role by first neutralizing the effects of transforming growth factor-␤ and second, based on its ability to inhibit collagen phagocytosis, prevent excessive resorption and disorganized deposition of collagen, a characteristic of scar tissue (31). Conceivably, DCN may have therapeutic potential for enhancing wound healing based on its role in inhibiting collagen degradation in kidneys of decorin knock-out mice (30).