Binding of Fibromodulin and Decorin to Separate Sites on Fibrillar Collagens*

The small proteoglycans, decorin, fibromodulin, bi- glycan, and lumican, represent a family of structurally related but genetically distinct molecules present in many types of connective tissues. Fibromodulin and decorin interact with collagens I and I1 (Hedbom, E., and Heineglrd, D. (1989) J. Biol. Chem. 264, 6898-6905). These interactions have been characterized fur- ther by using native radiolabeled components from fibroblast cultures and nonlabeled proteoglycans pu- rified from guanidine hydrochloride extracts of bovine tendon. Binding of metabolically labeled macromole- cules to collagen I was measured in an assay based on precipitation of collagen fibrils formed in vitro. Among a large number of secreted fibroblast products, decorin and fibromodulin represented the vast majority of the collagen binding components. These molecules showed poor binding to denatured collagen, in contrast to fi- bronectin, which was also present in the medium. Decorin and fibromodulin bind to different sites on col- lagen I fibrils, since the binding of either radiolabeled component could be competed for only by the corre- sponding nonlabeled proteoglycan. Similarly, these proteoglycans showed binding to separate sites on col- lagen 11. Binding of isolated fibromodulin and decorin to collagens in solution was measured in a solid-phase inhibition assay. Each of the proteoglycans interacted with triple helical molecules, but not with denatured collagen chain constituents or fragments. For fibro- modulin, the data indicated an average of one binding site per collagen I molecule (& = 9.9 nM). The data on decorin indicated additional interactions, some apparently mediated by the dermatan sulfate side chain. The results suggest that the small proteoglycans bind to distinct triple helical sites, apparently differing from several other similar structures within each collagen molecule. cipitates from the collagen fibril binding assay, and gel beads from the collagen-agarose binding assay were incubated in electrophoresis sample buffer containing SDS and 8-mercaptoethanol (Laemmli, 1970). The samples were applied on polyacrylamide gradient gels. After electrophoresis, gels were stained with 0.1% Coomassie Brilliant Blue R-250. Radiolabeled components were detected by fluorography with sodium salicylate (Chamberlain, 1979).


6905).
These interactions have been characterized further by using native radiolabeled components from fibroblast cultures and nonlabeled proteoglycans purified from guanidine hydrochloride extracts of bovine tendon. Binding of metabolically labeled macromolecules to collagen I was measured in an assay based on precipitation of collagen fibrils formed in vitro. Among a large number of secreted fibroblast products, decorin and fibromodulin represented the vast majority of the collagen binding components. These molecules showed poor binding to denatured collagen, in contrast to fibronectin, which was also present in the medium. Decorin and fibromodulin bind to different sites on collagen I fibrils, since the binding of either radiolabeled component could be competed for only by the corresponding nonlabeled proteoglycan. Similarly, these proteoglycans showed binding to separate sites on collagen 11. Binding of isolated fibromodulin and decorin to collagens in solution was measured in a solid-phase inhibition assay. Each of the proteoglycans interacted with triple helical molecules, but not with denatured collagen chain constituents or fragments. For fibromodulin, the data indicated an average of one binding site per collagen I molecule (& = 9.9 nM). The data on decorin indicated additional interactions, some apparently mediated by the dermatan sulfate side chain. The results suggest that the small proteoglycans bind to distinct triple helical sites, apparently differing from several other similar structures within each collagen molecule.
Connective tissues contain collagen and proteoglycans as predominant components. Each of these is now recognized to represent large families of distinct extracellular matrix molecules. Appropriate assembly of these and other components into a well organized matrix is of crucial importance to the tissue function. Most likely, specific interactions between the individual macromolecules determine the organization at the supramolecular level.
The collagens are proteins that contain three polypeptides forming at least one extended domain with a characteristic triple-stranded helix. As yet, some 15-16 different collagen types have been identified (van der Rest and Garrone, 1991;Pan et al., 1992). These can be classified on the basis of molecular shape and properties (Miller and Gay, 1987;Kielty et al., 1993). The major class, i.e. the fibrillar collagens, comprises molecules having a single helical domain that makes up more than 95% of the molecule. The rodlike shape and regular structure of these molecules offer exceptional capacity for lateral association and aggregation into fibers. The fibrillar collagens are type I, which is present in most tissues, type I1 in cartilage, type I11 in distensible connective tissues, and types V and XI, which appear associated with types I and 11, respectively. Another distinct class of collagens, including types IX, XII, and XIV, contains multiple short helical domains interrupted by nonhelical domains. These do not form fibrils by themselves, but they are associated with the major collagen fibrils. Collagen IX, for example, can be located to the surface of collagen I1 fibers in cartilage, periodically arranged, with globular domains protruding at regular intervals (Vaughan et al., 1988).
The proteoglycans are proteins that have a variable number of sulfated carbohydrate chains, glycosaminoglycans, covalently attached (reviewed in Kjell6n and Lindahl (1991)). These molecules, together with other glycoproteins and the glycosaminoglycan hyaluronan, form the gel in which the collagen fibrils are embedded. There are two major classes of proteoglycans present interstitially in mesenchymal connective tissues. One is represented by large molecules (>lo6 Da) having the capacity to form aggregates with hyaluronan (Heineglrd et al., 1985;Morgelin et al., 1988) (for review see Wight et al. (1991)). The other major class of interstitial proteoglycans has a low molecular mass . They are present in many tissues and predominate in fibrous connective tissues. Three members of this group have been studied more extensively, i.e. decorin (PG-S2, PGII, PG40), biglycan (PG-SI, PGI), and fibromodulin (reviewed in HeinegHd and Oldberg (1989Oldberg ( ,1993). A fourth member is a keratan sulfate proteoglycan (Axelsson and Heinegiird, 1978) that is referred to as lumican (Blochberger et al., 1992). These molecules display core proteins of M , of about 40,000 that are structurally related but genetically distinct. They contain some 10 homologous repeats of about 25 amino acids in their central domains and cysteine residues located at conserved positions. Decorin contains a single CS'/DS chain, attached to the N-terminal part of the protein. Biglycan carries two CS/DS chains similarly located. Fibromodulin contains keratan sulfate (Oldberg et al., 1989) with usually one or two such chains distributed The abbreviations used are: CS, chondroitin sulfate; DS, dermatan sulfate; PBS, phosphate-buffered saline; BSA, bovine serum albumin; TES, 2-([2-hydroxy-l,l-bis(hydroxymethyl)ethyl] aminolethanesulfonic acid.
27307 among four sites in the central domain (Plaas et al., 1990). There are several sulfated tyrosine residues in the N-terminal part of fibromodulin (Antonsson et al., 1991).
Proteoglycans and fibrillar collagens appear to interact, which is of potential importance in extracellular matrix assembly (for references see Scott (1988), Hedbom andHeinegbrd (1989), and Bidanset et al. (1992)). Collagen IX, which occurs associated with fibers of collagen 11, is actually also a proteoglycan containing a glycosaminoglycan side chain . Ultrastructural studies on several tissues suggest that the small CS/DS proteoglycans can be regularly associated with collagen fibers (Scott and Orford, 1981;Pringle and Dodd, 1990). Both decorin (Vogel et al., 1984) and fibromodulin (Hedbom and Heinegbrd, 1989) inhibit collagen fibril growth in uitro. These two proteoglycans show similar affinity for collagen I, with apparent dissociation constants of approximately lo-' M (Hedbom and HeinegArd, 1989). Biglycan, although very similar in structure to decorin and fibromodulin, does not appear to bind to fibril-forming collagen (Brown and Vogel, 1989).
The present study was undertaken to characterize the collagen structures recognized as binding sites by decorin and fibromodulin. The primary aim was to reveal whether these sites are different for the individual proteoglycans.

EXPERIMENTAL PROCEDURES
Preparation of Collagen-Collagen I was prepared by pepsin digestion or by acid extraction from the fibrous proximal part of bovine flexor tendon (Vogel et al., 1984;Hedbom and HeinegHrd, 1989). To avoid proteolysis, acid extraction was done with 0.5 M acetic acid containing the proteinase inhibitors phenylmethylsulfonyl fluoride (0.5 mM), pepstatin (5 p~) , N-ethylmaleimide (2 mM), and EDTA (5 mM). Collagen I1 was prepared from pepsin extracts of bovine nasal cartilage using the method of Miller (1972) with previously described modifications (Vogel et al., 1984). The preparations were stored freeze-dried.
Samples were redissolved at 4 or 5 mg/ml in 0.5 M acetic acid, then dialyzed into 0.1 M acetic acid and kept at +4 "C until used in the binding experiments. The collagen I1 preparation was further processed to enrich molecules having the capacity to form fibrils. The acidic solution was diluted to 1 mg/ml in PBS (0.14 M NaCl, 30 mM sodium phosphate, pH 7.3) and incubated at 37 "C for 24 h. Precipitated collagen was collected by centrifugation at 4000 X g for 30 min. This material was again dissolved in 0.5 M acetic acid at +4 'C and dialyzed into 0.1 M acetic acid. Finally, the solution was centrifuged at 10,000 X g for 30 min to remove a minor insoluble component. The collagen concentrations were determined by quantitation of hydroxyproline in samples after hydrolysis in 6 M HCl for 24 h at 100 "C (Stegeman and Stalder, 1967).
Denatured collagen chain constituents were obtained by heating a sample, diluted into 10 mM acetic acid, at 60 "C for 10 min. Peptides derived from the collagen chains were prepared by cleavage at methionyl residues with 2% (w/v) CNBr in 70% (v/v) formic acid for 20 h at 20 "C (Bornstein and Piez, 1965).
Isolation of Proteoglycam-Decorin and fibromodulin were isolated from a 4 M guanidine hydrochloride extract of adult bovine tendon and purified essentially as described elsewhere Heinegird et al., 1986). The proteoglycans were pooled and stored in 4 M guanidine HCl, 0.05 M sodium acetate, pH 5.8, frozen as aliquots. Before use in the collagen binding experiments, proteoglycans in thawed samples were precipitated by adding 10 volumes of ethanol, pelleted by centrifugation, and briefly vacuum-dried. These samples were dissolved in the buffer used for the binding assay and added to collagen within 10 min.
Preparation of Radiolabeled Components-Fibroblasts from the deep flexor tendon of a 2-year-old calf were allowed to colonize 145-cm2 plastic dishes (Nunclon, A/S Nunc). The cells were cultured in Ham's F-12 supplemented with 10% fetal calf serum and utilized at passages 5-15. Confluent cultures were radiolabeled for 20 h in serumfree medium containing 20 pCi/ml [3H]leucine. To the harvested culture medium was added phenylmethylsulfonyl fluoride (0.5 mM), N-ethylmaleimide (2 mM), EDTA (2.5 mM), and BSA (50 pg/ml). This medium, containing a mixture of secreted fibroblast products, was prepared for binding studies by desalting on a 10-ml column of Sephadex G-50 (Pharmacia LKB Biotechnology Inc.) eluted with PBS diluted with 3 volumes of water. Samples containing the macromolecules were pooled, freeze-dried, and then dissolved in the appropriate volume of water to reconstitute the PBS.
Proteoglycans from radiolabeled culture medium were partially purified by ion-exchange chromatography. The medium, obtained as described above, was diluted with an equal volume of water and chromatographed on a column (1.0 X 6.4 cm) of DEAE-Sepharose (Pharmacia). The column was eluted with 0.1 M LiC1, 25 mM Tris/ HCl, pH 7.3, followed by a linear gradient of 0.1-1.1 M LiCl in 100 ml of the Tris buffer. Fractions of 1 ml were collected and analyzed by B-scintillation counting and SDS-polyacrylamide gel electrophoresis (see below). Fractions containing fibromodulin and decorin were pooled separately, dialyzed against water, and freeze-dried.
Collagen Fibril Binding Assay-Samples containing radiolabeled molecules, and sometimes nonlabeled proteoglycans, were combined with double-strength PBS, water, BSA, and finally 20 pg of acidextracted collagen I or 100 pg of pepsin-extracted collagen 11. When ready, each sample was in 200 p1 of 140 mM NaCl, 30 mM sodium phosphate, pH 7.3, and contained -100 pg of BSA. The samples were incubated at 37 "C for 20 h to permit the formation of insoluble collagen fibrils. These fibrils were pelleted by centrifugation at 10,000 X g for 10 min at 37 "C, and the supernatant was removed. The precipitate was washed once with 0.2 ml of PBS at 37 "C and then dissolved in 0.1 M HCl. In the initial experiments, with crude mixtures of radiolabeled components, these samples were directly subjected to SDS-gel electrophoresis. However, large amounts of collagen affected the appearance of the radiolabeled molecules. In the following experiments, only one-tenth of each sample was electrophoresed as such. The remaining nine-tenths was digested for 6 h at 37 "C, with 0.4 units of high purity collagenase (type VII, Sigma) in 0.15 M NaCl, 5 mM CaC12, 50 mM Tris/HCl, pH 7.4, containing 10 pg/ml ovomucoid trypsin inhibitor (type 11-0, Sigma) and electrophoresed separately. The relative amounts of collagen in the nondigested samples were determined by densitometric scanning of the Coomassie Blue-stained gels.
Collagen-Agarose Binding Assay-Pepsin-extracted collagen I at 2.5 mg/ml in PBS was coupled to CNBr-activated Sepharose 4B (Pharmacia) at 4 "C overnight. Similarly, gel beads were incubated in PBS containing heat-denatured collagen I or no additives. Remaining reactive groups were blocked with 0.2 M ethanolamine, followed by several rinses with 0.1 M acetic acid, PBS, and finally PBS containing 0.2% (v/v) of Tween 20. Samples of the collagencontaining solutions, taken before and after the coupling, were used to determine the degree of coupling. Typically, there was 3.4-4.0 mg of collagen/ml of gel.
Samples containing radiolabeled components in 200 pl of PBS were mixed with approximately 25 pl of gel beads and incubated under gentle agitation for 20 h at 37 "C. The beads were collected by centrifugation, washed once in PBS, and incubated in electrophoresis sample buffer prior to analysis by SDS-polyacrylamide gel electrophoresis.
Microtiter Plate Assay-The wells of microtiter plates were coated overnight with pepsin-extracted collagen I at 10 pg/ml in 0.15 M NaC1, 10 mM Tris/HCl, pH 7.4 (Hedbom and Heinegird, 1989). Control wells were coated with BSA at 10 pg/ml. The adsorbed protein was cross-linked by treatment with 0.25% glutaraldehyde in 0.15 M NaC1, 0.5 mM NaCNBH4, 20 mM TES, pH 7.4, for 2.5 h at 37 "C, followed by blocking of remaining reactive groups with 0.2 M ethanolamine in the same buffer for 0.5 h at 37 "C. To prevent nonspecific binding of components, all wells were incubated with 10 mg/ml BSA in 0.15 M NaC1, 10 mM Tris/HCl, pH 7.4, for 4 h at 20 "c. In some cases, the wells were additionally incubated for 1 h with 0.2 mg/ml CS, prepared as described in Antonopoulos et al. (1967). Isolated proteoglycans, dissolved in 0.15 M NaCl, 10 mM Tris/HCl, pH 7.4, containing 0.05% (v/v) Tween 20 were combined with soluble collagen at various concentrations and preincubated for 5 h at 20 "C. These samples were transferred to the coated wells and incubated overnight at 20 "C. Proteoglycans bound to the wells were detected by using an enzyme-linked immunosorbent assay procedure (Hedbom and Heinegird, 1989). In some experiments, the dissolved decorin was first digested with 0.05 units of chondroitinase ABC (Sigma) per mg of protein for 30 min at 37 "C. These samples were diluted with Tris buffer containing 0.05% (v/v) Tween 20 and immediately used for the binding assay. SDS-Polyacrylamide Gel Electrophoresis-Ethanol-precipitated samples of fractions from the described chromatography steps, pre-cipitates from the collagen fibril binding assay, and gel beads from the collagen-agarose binding assay were incubated in electrophoresis sample buffer containing SDS and 8-mercaptoethanol (Laemmli, 1970). The samples were applied on polyacrylamide gradient gels. After electrophoresis, gels were stained with 0.1% Coomassie Brilliant Blue R-250. Radiolabeled components were detected by fluorography with sodium salicylate (Chamberlain, 1979).

Binding of Metabolically Radiolabeled Components to Colla-
gen-Bovine tendon fibroblasts in confluent cultures were labeled with [3H]leucine. The macromolecules in the culture medium were separated from low molecular weight components and transferred into PBS by gel filtration. The proteins were analyzed by SDS-polyacrylamide gel electrophoresis. As shown in Fig. 1, a sample representing the total fibroblast medium contained several radioactive components. Among these, fibronectin and procollagen I were predominant.
The mixture of components was tested for binding to collagen I, using two different methods. In one of these, the sample was added to a solution of collagen at the onset of in uitro fibrillogenesis. The two small proteoglycans, decorin and fibromodulin, were the only radiolabeled molecules that showed pronounced binding to the precipitated collagen fibrils (Fig. 1). Some fibronectin and trace amounts of procollagen were also associated with the collagen. The other detectable components remained entirely in the supernatant. Before fluorography, the gels were stained with Coomassie Blue in order to check that appropriate amounts of collagen I (>95%) had been precipitated. A typical result is shown in Fig. 1, lane   4. Due to the presence of large amounts of collagen quenching the radioactivity, pale zones were seen on the fluorography plates at the positions of the collagen bands affecting the appearance of decorin and fibronectin.
Binding of radiolabeled components to collagen was also tested by using Sepharose beads to which collagen molecules had been covalently linked. If the coupled collagen was in the form of denatured a-chains, large amounts of fibronectin were bound (Fig. 2, lane 2). The small proteoglycans, however, were not associated with the beads. If instead the collagen was coupled under conditions promoting the retention of mainly triple helical monomers, binding of decorin and fibromodulin was observed in addition to that of fibronectin (Fig. 2, lane  3). Thus, the binding of decorin and fibromodulin is depend- ent on the conformation of the collagen. Furthermore, fibronectin apparently does not mediate the association between collagen and the small proteoglycans.

The Effect of Isolated Decorin and F i b r o d d i n on the Binding of Radiolabeled Proteoglycans to Collagen Fibrils-To
examine whether decorin and fibromodulin bind to the same sites on fibrillated collagen, competition experiments were performed. For this purpose, decorin and fibromodulin from the [3H]leucine-labeled fibroblast culture medium, were partially purified by ion-exchange chromatography on DEAE-Sepharose eluted with a LiCl gradient (data not shown). In order to minimize changes in protein conformation, a buffer was used that did not contain detergents or chaotropic agents.
Samples containing radiolabeled decorin and fibromodulin were added to collagen I in the fibril binding assay. As much as 80-90% of each of the proteoglycans bound to the insoluble fibrils. This binding could be competed for by including purified nonlabeled proteoglycans. Isolated fibromodulin effectively inhibited the binding of radioactive fibromodulin but appeared not to affect the binding of labeled decorin (Fig.  3a). Isolated decorin, on the other hand, did not decrease the binding of radiolabeled fibromodulin but significantly decreased the binding of labeled decorin (Fig. 3 a ) . Thus, decorin and fibromodulin bind to separate sites on collagen I.
In analogous experiment!! it was shown that fibromodulin and decorin bind to separate sites on collagen I1 as well (Fig.   3b). Only approximately 50% of the total collagen I1 was precipitated in this assay, even though the collagen used had been isolated from in vitro formed fibers. Notably, 75-8.5% of the radiolabeled proteoglycans could still be bound to the pelleted collagen fibrils. The relative amount of collagen precipitated as fibrils was determined for each sample in a separate electrophoretic analysis (not shown). The degree of precipitation was constant for each collagen preparation and was not significantly affected by the addition of proteoglycans. could be assessed. It was found to be essential to prevent interactions between the soluble and the coated collagen molecules. For this reason, the assay was performed in Tris buffer instead of the previously used PBS. Furthermore, primary amino groups in the coated protein were modified by cross-linking with 0.25% glutaraldehyde. This treatment apparently blocks the aggregation of collagen molecules, whereas proteoglycan binding still occurs. The binding of isolated fibromodulin to the collagen-coated wells displayed an apparent linear relationship to the fibromodulin concentration over 0-1 pg/ml and reached saturation above 40 pg/ml (data not shown).

Binding of Isolated Fibromodulin to Collagen Molecules in Solution-To
Preincubation of the proteoglycan with triple helical collagen I in solution resulted in inhibition of the solid-phase binding (Fig. 4, a and b ) . Triple helical collagen I1 also showed inhibition, but a t slightly higher concentrations (Fig. 4b).
Collagen I in the form of denatured a-chains showed no such activity and neither did the peptides obtained after cleavage with CNBr (Fig. 46). At higher concentrations of these (>32 pglml), there was a dramatic increase in binding of fibromodulin to the wells. This was probably due to nonspecific BSA only in Tris-buffered saline at 20 "C. The coated wells were incubated with 0.25% glutaraldehyde to stabilize the protein by crosslinking and to prevent lysine-dependent collagen-collagen interactions. Fibromodulin ( F M ) a t 1 pg/ml was preincuhated for 5 h at 20 "C with 0-128 pg/ml of different collagen preparations in Trisbuffered saline containing O.OS' % (v/v) of Tween 20. These samples were transferred to the microtiter plate wells and incubated overnight. Fibromodulin, bound to the coated wells, was quantitated by subsequent incubations with specific rabbit polyclonal antiserum, anti-(rabbit I&) antibodies conjugated to alkaline phosphatase, and pnitrophenyl phosphate. Formation of product was measured as an increase of absorbance a t 405 nm. a, fibromodulin, preincubated with pepsin-extracted collagen I, was tested for binding to wells coated with collagen I and BSA (solid line) or BSA only (dashed line). b. the percentage of inhibition by triple helical collagen I (0). heat-denatured collagen I (W), CNBr-cleaved collagen I (A), and triple helical collagen I1 (V) was calculated from absorbance values reflecting the binding of fibromodulin to coated collagen, after compensation for the binding to control wells without collagen. c. the interaction of fibromodulin with collagen I in solution, measured as inhibition of solid-phase binding, was analyzed in a Scatchard plot. According to a linear curve fit (RZ = 0.83), the dissociation constant was 9.9 nM and the number of fibromodulin binding sites per collagen molecule was 1.0.
interactions. Large amounts of proteoglycan were also bound to the control wells not coated with collagen. Likewise, the inhibitory effect of collagen I1 was abolished if the conformation was altered by cleavage or denaturation (not shown). Thus, the binding of fibromodulin requires triple helicity but not fibrillar assembly.
The data on fibromodulin acting as a ligand to triple helical Bincling of Fibromodulin and Decorin to Fibrillar Collagens 27311 collagen I were used for Scatchard (1949) analysis. It was assumed that the measured solid-phase binding was a linear function of the concentration of free ligand. The proportion of ligand molecules that were bound would then be equal to the percentage of inhibition. Even though the scatter of data points was a limiting factor, the analysis gave an indication that the major, if not the only, class of binding sites had an apparent Kd of 9.9 nM and was represented by one site per collagen molecule on the average. The data on fibromodulin binding to collagen 11 did not display a similar linearity (not shown). Clearly, there were fewer and/or weaker binding sites in the collagen molecules of this preparation, as compared with collagen I.
Binding of Isolated Decorin to Collagen Molecules in Solution-The interaction of decorin with collagen, when measured as inhibition in a solid-phase assay, appeared different from that of fibromodulin. Triple helical collagen I, at concentrations of 2-8 pg/ml in the solution, produced increased binding of decorin to the coated collagen (Fig. 5a). At higher concentrations, the effect of the collagen was inhibitory. If the DS chain of decorin was first removed by digestion with chondroitinase ABC, the binding enhancement was less pronounced (Fig. 5a). Alternatively, the enhancing effect could be suppressed by addition of CS to the coated wells before these were used in the assay (Fig. 5 b ) . These observations indicate that not only the core protein of decorin but also the glycosaminoglycan chain is involved in collagen interactions. Microtiter plates were coated with collagen I and BSA or BSA only followed by incubation with 0.25% glutaraldehyde. Decorin at 2 pg/ ml was preincubated for 5 h at 20 "C with 0-64 pg/ml different collagen preparations. These samples were transferred to the microtiter plate wells, and proteoglycan binding to the coated wells was quantitated as described in the legend to Fig. 4. a, decorin ( The core protein-mediated solid-phase binding was inhibited by triple helical collagens I and I1 (Fig. 5b). Neither was inhibition shown by denatured or CNBr-cleaved collagen I ( Fig. 5b) nor with collagen I1 (not shown). At concentrations above 8 pg/ml of such competitors, there was a nonspecific increase in the binding of decorin to the coated surfaces.

DISCUSSION
The binding sites for decorin and fibromodulin on collagens I and I1 have been examined. The interactions between these proteoglycans and collagen in the form of fibrils, monomers, denatured chain constituents, and fragments have been evaluated. The proteoglycans used were either native as taken directly from cell culture medium or pure as isolated from tissue extracts.
Binding of metabolically radiolabeled components was tested in a collagen fibrillogenesis assay. Among the secreted fibroblast macromolecules, decorin and fibromodulin were those predominantly bound to the precipitated collagen fibrils. The two proteoglycans bound to separate sites on the fibrils, as revealed by competition experiments. Saturation of the binding sites occurred at relatively low concentrations of the added proteoglycan, representing 1-3 times the molar concentration of the collagen. Binding saturation of one of the proteoglycans apparently did not affect the binding of the other or the amount of collagen that was precipitated as fibrils. These observations taken together indicate that the binding sites for the two proteoglycans are distinct, specific, and limited in number.
One advantage of the fibril binding assay was that the radiolabeled proteoglycans never had been exposed to denaturing conditions or even been incorporated in a matrix. This contrasts with previous studies where proteoglycans have been extracted and purified in solutions of guanidine hydrochloride and urea. It is possible that such preparations only partially renature upon removal of the denaturing solvent and, therefore, show protein conformation heterogeneity and binding artifacts. Self-interaction of the molecules, whether artificial or genuine, may complicate the binding studies. The isolated small proteoglycans are indeed known to self-interact (Ward et al., 1987;Morgelin et al., 1989). This may at least partially explain some apparently conflicting data on the binding of decorin to fibrillar collagens (Bidanset et al., 1992) and its effect on collagen fibrillogenesis (Uldbjerg and Danielsen, 1988). When the isolated proteoglycans were used in our study, care was taken to perform the experiments immediately after the proteoglycans had been redissolved. Then, the isolated and renatured fibromodulin was able to bind to all of the sites that were available for the radiolabeled native molecules, shown as complete inhibition in the competition experiments. Thus the binding sites of the proteoglycans appear properly renatured. In the competition between decorin molecules for binding to collagen, a slightly larger excess of the isolated proteoglycan was required.
There was also some binding of radiolabeled fibronectin to the collagen fibrils, but this glycoprotein preferentially interacts with denatured collagen (Engvall et al., 1978). Interestingly, fibromodulin and decorin showed no affinity for denatured collagen I linked to Sepharose beads. Obviously, the proteoglycans bind to collagen by a different mechanism than fibronectin. The lack of binding of decorin and fibromodulin to the fibronectin-denatured collagen-Sepharose complex also indicates that these proteoglycans do not occur as complexes with fibronectin under the conditions described.
The interaction sites within the collagens were further examined in a solid-phase binding inhibition assay using the Binding of Fibromodulin and Decorin to Fibrillar Collagens isolated and renatured proteoglycans. It was found that each of the proteoglycans could interact with native collagen molecules in solution but not with denatured collagen chain constituents or fragments. The collagen I1 preparation was less effective than collagen I with regard to binding inhibition, albeit the fact that this preparation contained collagen I1 molecules that had been selected on the basis of their ability to form fibrils. From the fibromodulin binding inhibition data on collagen I, a dissociation constant of 9.9 nM was calculated.
This value agrees well with the relative Kd value of 35 nM previously obtained in a direct binding assay (Hedbom and Heinegird, 1989). The analysis clearly indicated the presence of one fibromodulin binding site per collagen molecule. This result does not exclude the possible existence of more than one proteoglycan binding site within a collagen molecule, but it brings further support to the view that the sites are limited in number and highly specific. The interaction of decorin with collagen, as observed in the solid-phase assay, was more complex than that of fibromodulin. Higher concentrations of collagen in the solution were required for inhibition of the binding to coated molecules. At low concentrations the soluble collagen instead increased the binding. Since this effect was reduced after removal of the DS chains, it appears that either DS-collagen interactions (Obrink, 1973) or DS-DS interactions (Fransson et al., 1982) were involved. In support, the addition of isolated CS chains reduced the binding. A possible explanation of the increased binding is that decorin binds to collagen in solution by protein-protein interaction, and then the dermatan sulfate side chain provides interaction with the coated collagen. This may then indicate that the glycosaminoglycan chain preferentially binds to a collagen molecule other than the one to which the core is bound, consistent with previous data showing little alterations of binding upon removal of the DS chain (Hedbom and HeinegPrd, 1989). It is thus possible that one function of decorin is to connect neighboring collagen fibrils.
Decorin has recently been shown to bind collagen VI by a core protein-mediated interaction (Bidanset et al., 1992). In that study, binding to collagen I-IV was poor, perhaps suggesting that the interaction with collagen VI is differently mediated. An interaction between decorin and fibronectin has been studied (Lewandowska et al., 1987;Schmidt et al., 1987Schmidt et al., , 1991 suggesting a role for the small proteoglycans in the regulation of cell attachment and modulation of the extracellular matrix. The binding of native decorin or fibromodulin to fibrillar collagens does not depend on any mediating molecules according to the present data. However, the exact function of these matrix components in vivo may require additional interactions. In continued studies, the combined effects of the multiple interactions and the possible interlinking of matrix constituents should be considered. Since each of the small proteoglycans shows distinct collagen binding characteristics, these molecules should be studied individually and not regarded as functionally identical.
The present study shows that the small proteoglycans bind to distinct triple helical sites, presumably selected among a large number of very similar structures in each collagen molecule. The limited structural differences between these proteoglycans determine the exact binding specificity and probably the function of the molecule.