Localization of a binding site for the proteoglycan decorin on collagen XIV (undulin).

Through its ability to bind extracellular matrix constituents and growth factors the small leucine-rich chondroitin/dermatan sulfate proteoglycan decorin which is present in many types of connective tissues may play an important biological role in remodeling and maintenance of extracellular matrices during inflammation, fibrosis, and cancer growth. In this study we investigated the known binding of decorin to human collagen XIV. This binding was unaffected when the small collagenous moiety of collagen XIV was removed with collagenase. Therefore, fragments covering the large noncollagenous domain NC3 of collagen XIV were expressed in Escherichia coli, each fused to a 26-kDa fragment of glutathione S-transferase. Using radioiodinated decorin as ligand for the immobilized fusion proteins, a binding site that interacted with the decorin core protein could be assigned to the NH2-terminal fibronectin type III repeat of collagen XIV. In addition, an auxiliary binding site located COOH-terminal to this fibronectin type III repeat interacted with the glycosaminoglycan component of decorin.

Through its ability to bind extracellular matrix constituents and growth factors the small leucine-rich chondroitin/dermatan sulfate proteoglycan decorin which is present in many types of connective tissues may play an important biological role in remodeling and maintenance of extracellular matrices during inflammation, fibrosis, and cancer growth. In this study we investigated the known binding of decorin to human collagen XIV. This binding was unaffected when the small collagenous moiety of collagen XIV was removed with collagenase. Therefore, fragments covering the large noncollagenous domain NC3 of collagen XIV were expressed in Escherichia coli, each fused to a 26-kDa fragment of glutathione S-transferase. Using radioiodinated decorin as ligand for the immobilized fusion proteins, a binding site that interacted with the decorin core protein could be assigned to the NH 2 -terminal fibronectin type III repeat of collagen XIV. In addition, an auxiliary binding site located COOH-terminal to this fibronectin type III repeat interacted with the glycosaminoglycan component of decorin.
Collagen XIV is a large, non-fibrillar extracellular matrix protein with two short carboxyl-terminal collagenous sequences and three noncollagenous domains termed NC 1, NC 2, and NC 3. Together with the collagens IX and XII it belongs to the subfamily of fibril-associated collagens with interrupted triple-helices (1)(2)(3)(4)(5)(6)(7). Collagen XIV consists of functional and structural modules, such as fibronectin type III repeats and von Willebrand factor A domains, which may mediate interactions with cellular receptors and other collagens (1,(7)(8)(9)(10)(11)(12). Collagen XIV is abundant in cartilage and soft connective tissues that contain large amounts of fibrillar collagens (8,(13)(14)(15)(16)(17). It is mainly found in well differentiated mesenchymal tissues, but virtually absent from tumor stroma (8) and in early stages of embryonic development (17), indicating that it might play a role in differentiation. Using immunoelectron micros-copy collagen XIV was shown to be associated with the surface of banded collagen fibrils, possibly forming interfibrillar connections in a variety of tissues (8,18,19). As described for the similar collagen XII (20), collagen XIV might also influence fibril formation.
Several mesenchymal and epithelial cells adhere to immobilized collagen XIV, and we identified a chondroitin/dermatan sulfate form of CD44 as the prominent collagen XIV receptor on human skin fibroblasts (21). Heparin and decorin that inhibited cell binding to the immobilized collagen XIV have been suggested as physiological modulators of the interactions of cells with this matrix molecule (21). In addition, decorin inhibits attachment of fibroblasts to fibronectin (22,23), collagens I and II (24), and thrombospondin (25).
Decorin is a small leucine-rich proteoglycan that consists of a core protein with a single chondroitin/dermatan sulfate chain (26). It binds to fibronectin (27), thrombospondin (25), the collagens I (28,29), II (29), VI (30), and XIV (31), the complement component C1q (32) and the transforming growth factor-␤ (33). The binding of transforming growth factor-␤ to decorin can modulate the growth factor's biological availability and activity both in a positive and negative way (34 -36). In fibril-forming assays in the presence of decorin, collagen fibrillogenesis is delayed and results in a thinner final fibril diameter (37,38). The appearance of irregular collagen fibrils which indicate uncontrolled lateral fibril fusion in decorin-deficient mice (39) underlines a role of decorin in the organization of collagen fibrils.
To localize the binding site(s) of decorin on collagen XIV, we designed recombinant fragments of collagen XIV. Binding assays demonstrated that decorin binds to the NH 2 -terminal fibronectin type III-repeat in collagen XIV. The interaction was enhanced by a COOH-terminally adjoining region that contains a potential heparin-binding site.

MATERIALS AND METHODS
Reagents-Unless stated otherwise, all reagents were obtained from Sigma (Munich, Germany). Na 125 I was from Amersham Buchler (Braunschweig, Germany). Decorin was purified from the secretions of cultured human skin fibroblasts (40). Highly purified collagen XIV was isolated from human placenta as described (21). Escherichia coli NM522 was purchased from Invitrogen (Leek, The Netherlands). Plasmid pGEX-2T and glutathione-Sepharose 4B were obtained from Pharmacia Biotech GmbH (Freiburg, Germany).
The amino acid sequence reported in this paper was deduced from the human cDNA sequence submitted to the EMBL nucleotide sequence data base with accession numbers M64108, Y11709, Y11710, and Y11711.
Expression and Purification of Fusion Proteins-In 500-ml cultures of transformed E. coli NM522 expression of fusion proteins was induced with 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside at A 600 ϭ 0.2 followed by incubation at room temperature for an additional 12-16 h. Cells were spun down and resuspended in 10 ml of buffer (50 mM Tris-HCl, 150 mM sodium chloride, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, pH 8.0) at 4°C followed by addition of 5 mg of lysozyme. After 20 min at 4°C cells were lysed on ice by mild sonication and subjected to centrifugation at 10,000 ϫ g for 5 min at 4°C. The supernatants containing the fusion proteins were incubated with 1 ml of glutathione-Sepharose for 30 min at room temperature. Beads were collected by centrifugation at 1,000 ϫ g for 5 min and washed twice with 30 ml of 150 mM NaCl, 16 mM Na 2 HPO 4 , 4 mM NaH 2 PO 4 , pH 7.3 (PBS). 1 Fusion proteins were eluted with 5 mM reduced glutathione (Sigma) in 50 mM Tris-HCl, pH 8.0.
Gel Electrophoresis, Radiolabeling, and Ligand Blotting-Proteins were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) according to the method of Laemmli (42). Apparent molecular masses were assessed by use of a globular protein standard. Disulfides were reduced by heating proteins to 90°C in 37 mM Tris, 1.14 M urea, 2% dithiothreitol, 2.14% SDS, 0.5% glycine, and 0.007% bromphenol blue, pH 6.8, for 25 min. Gels were stained with Coomassie Brilliant Blue R-250, destained with 7% acetic acid, and dried at 60°C. 125 I-Labeled proteins were detected by exposure to Kodak X-Omat® film with an intensifying screen at Ϫ80°C for 1 day.
Decorin was radiolabeled with 125 I by the chloramine-T method (43). For ligand blots 1 g of each fusion protein was separated by SDS-PAGE on a 12.5% gel and blotted to nitrocellulose. Unspecific binding sites were blocked with 1% BSA in PBS before incubating the blots with PBS containing 0.05% Tween 20, 0.1% BSA, and 390,000 cpm of 125 Idecorin (approximately 15 ng) for 2.5 h at 22°C. The blots were washed thoroughly with PBS, air dried, and exposed to Kodak X-Omat® film with an intensifying screen at Ϫ80°C for 1 day.
Binding Assays-Adsorbtion to polystyrene microtiter plates (Immulon 2 Removawells, Dynatech, Germany) was performed by incubating each well with 0.5 g of collagen XIV, 0.5 g of its noncollagenous domains NC 1-3, 1 g of the various fusion proteins, or 1 g of the 26-kDa fragment of glutathione S-transferase in 100 l of PBS for 4 h at 37°C. Remaining binding sites were blocked with 1% BSA in PBS at 4°C overnight. Coated wells were washed and preincubated for 2 h at 22°C with no, 0.02, 0.2, or 5 g of heparin, or chondroitin sulfates A, B, or C in 100 l of PBS, 0.05% Tween 20, 1% BSA per well. After thorough washing, wells were incubated with 100 l of PBS, 0.05% Tween 20, 1% BSA, containing 26,000 -30,000 cpm of 125 I-labeled decorin (approximately 1-1.2 ng) or the decorin core protein and increasing amounts of unlabeled decorin, 5 g of heparin, or no additives, for 2.5 h at 22°C. Unbound decorin was removed by suction, followed by three washes with PBS, 0.05% Tween 20, 1% BSA and counting of bound radioactivity in a ␥-counter (MultiPrias, Canberra-Packard, Dreieich, Germany).

RESULTS
Comparable quantities of human collagen XIV or its noncollagenous domains NC 1-3, generated by digestion with collagenase, were immobilized on microtiter wells and incubated with 125 I-labeled decorin. Since decorin bound to both substrates with similar efficiency (collagen XIV, 46.2 Ϯ 3.5%; NC 1-3, 50.3 Ϯ 1.1%), we reasoned that decorin binds to the noncollagenous part of the molecule. To further localize the binding site, we generated a panel of recombinant fusion proteins that spanned the whole noncollagenous fragment NC 3 of collagen XIV and used them in ligand blots and microtiter binding assays. The pattern of the NC 3 sequences, each of which linked by its NH 2 terminus to a 26-kDa fragment of glutathione Stransferase from Schistosoma japonicum, is illustrated in Fig.  1. Fig. 2 shows the characterization of the fusion proteins by SDS-PAGE.
When 1 g of the fusion proteins were run on SDS-PAGE, blotted to nitrocellulose, and incubated with 125 I-labeled decorin, only bands comprising amino acids 29 -115 of the amino-terminal fibronectin type III repeat of collagen XIV bound decorin, as judged by autoradiography (Fig. 3). The higher intensity of the band resulting from binding of 125 Ilabeled decorin to fusion protein 29 -154 compared with that to fusion protein 29 -115 indicated a stronger binding to fusion protein 29 -154. This finding was plausible, since the extension from residue 116 -154 contained the potential heparin binding sequence EKRKDPKP (46) at positions 121-128 (Fig. 4).
Binding of decorin to the amino-terminal fibronectin type III repeat of collagen XIV was confirmed by microtiter binding assays. Fusion proteins were adsorbed to microtiter wells and incubated with 125 I-labeled decorin after blocking of free binding sites. Again decorin bound to the fusion proteins that contained the amino-terminal fibronectin type III repeat of colla- 1 The abbreviations used are: PBS, phosphate-buffered saline; PAGE, polyacrylamide gel electrophoresis; BSA, bovine serum albumin. gen XIV (Fig. 5), whereas binding to all other fusion proteins or to glutathione S-transferase alone was insignificant (0.1-2.1%). As indicated by ligand blotting the binding strongly increased when the fusion protein contained not only the fibronectin type III repeat but also the COOH-terminally adjoining segment that includes the potential heparin binding sequence: EKRKDPKP (46) (Figs. 4 and 5). Heparin strongly inhibited the binding of decorin to collagen XIV or to the fusion proteins containing the amino-terminal fibronectin type III repeat (Fig. 5).
Binding was also inhibited, albeit to a lower degree, by preincubation of decorin with chondroitinase ABC (Fig. 5) which removes the chondroitin/dermatan sulfate side chain of decorin as judged by SDS-PAGE and autoradiography (Fig. 6). It is notable that the remaining core protein of decorin bound equally well to the fusion proteins 29 -115 (12.7 Ϯ 0.3%) and 29 -154 (13.1 Ϯ 0.3%). This suggests that the core protein of decorin mediates an interaction with the NH 2 -terminal fibronectin type III repeat (amino acids 29 -115) whereas its chondroitin/dermatan sulfate side chain interacts with the COOH-terminally adjoining segment comprising the amino acids 116 -154, resulting in a stronger binding of the complete proteoglycan to fusion protein 29 -154. The presence of the potential heparin binding sequence EKRKDPKP (46) in the COOH-terminal extension is well in line with this interpretation. However, the complete proteoglycan also bound more efficiently to fusion protein 29 -115 (24.2 Ϯ 1.3%) than the decorin core protein (12.7 Ϯ 0.3%), indicating a role of the chondroitin/dermatan sulfate side chain also for the binding to the NH 2 -terminal fibronectin type III repeat (Fig. 5). This is supported by inhibition of the binding of decorin and its core protein by preincubation of the immobilized fusion protein 29 -115 with heparin and chondroitin sulfates A and B, demonstrating a glycosaminoglycan-binding site on the amino-terminal fibronectin type III repeat (Fig. 7). Fig. 7 also shows that bound glycosaminoglycans do not only compete with the chondroitin/dermatan sulfate chain of decorin, but also interfere with the protein-protein interaction, since they inhibited binding of the decorin core protein to collagen XIV and to the fusion proteins. All tested glycosaminoglycans inhibited the binding of  1 and 3) and 29 -154 (lanes 2 and 4) were separated by 12.5% SDS-PAGE and blotted to nitrocellulose. The blots were either stained with Amido Black (lanes 1 and 2), or incubated with 125 I-decorin for 2.5 h after blocking of unspecific binding sites with 1% BSA in PBS. The blots were washed, air-dried, and subjected to autoradiography (lanes 3  and 4). 125  decorin and its core protein to collagen XIV but they blocked the binding to the fusion proteins to different degrees, with heparin as the most potent inhibitor. Thus, preincubation with 0.02 g of heparin per well reduced binding of decorin and the decorin core protein to collagen XIV as well as to fusion proteins 29 -154 and 29 -115 maximally, except for binding of decorin to collagen XIV which was less completely inhibited at this concentration (Fig. 7). Preincubation with chondroitin sulfate C did not or only slightly affect the binding of decorin and its core protein to the fusion proteins, whereas chondroitin sulfates A and B were intermediate inhibitors (Fig. 7).
To further investigate the specifity of the interaction, binding of 125 I-decorin to collagen XIV or to fusion protein 29 -154 was performed in the presence of increasing amounts of unlabeled decorin. The experiments showed that unlabeled decorin can compete efficiently with 125 I-decorin for the binding. Excess unlabeled decorin displaced 125 I-decorin from collagen XIV and the fusion protein 29 -154 demonstrating the binding to be saturatable and thereby specific (Fig. 8). DISCUSSION We demonstrated that the large NH 2 -terminal noncollagenous fragment NC 3 of human collagen XIV harbors a binding site for the small chondroitin/dermatan sulfate proteoglycan decorin. Ligand blots and microtiter binding assays performed with recombinant fusion proteins that covered the whole noncollagenous fragment NC 3 clearly assigned a binding site for decorin to the amino-terminal fibronectin type III repeat of collagen XIV. Binding of decorin to this domain was enhanced when the fusion protein additionally contained the COOHterminal extension with the potential heparin binding sequence: EKRKDPKP (Fig. 4). This sequence accords to the heparin binding consensus sequence (-XBBBXXBX-), where B is a basic and X is a nonbasic amino acid residue (46). The specifity of the binding of decorin to this fusion protein or to total collagen XIV was demonstrated by the displacement of 125 I-decorin with an excess of unlabeled decorin.
These results contrast to those of Font et al. (31) who did not find an interaction of the NC 3 domain of collagen XIV with decorin. We have no simple explanation for these divergent results. Possible reasons might be differences between the human placental collagen XIV used by us and the bovine collagen XIV isolated from tendon that was used by Font et al. (31), or "nonspecific" proteases often present in collagenase preparations that could have removed a sensitive part of the isolated NC 3 domain. Heparin and chondroitin sulfates A and B inhibited the binding of decorin to collagen XIV and to the fusion proteins. As described previously by Font et al. (31) for the binding of decorin to the whole collagen XIV molecule, we observed reduced binding to the fusion proteins when the chondroitin/ dermatan sulfate chain of decorin was removed by digestion with chondroitinase ABC. Interestingly, the residual binding of the decorin core protein was independent of the presence of the COOH-terminal extension of the NH 2 -terminal fibronectin type III repeat that contains the potential heparin binding sequence: EKRKDPKP. These results strongly suggest that the proteoglycan binds to collagen XIV in at least two ways: 1) decorin binds to the amino-terminal fibronectin type III repeat via its core protein; 2) the chondroitin/dermatan sulfate chain of decorin is involved in the binding to the COOH-terminal extension of the NH 2 -terminal fibronectin type III repeat that contains the potential heparin binding sequence. However, binding of decorin to the amino-terminal fibronectin type III repeat of collagen XIV is increased when compared with its core protein. Furthermore, the binding of the core protein to this domain is inhibited by preincubation with several glycosaminoglycans. These findings imply that the amino-terminal fibronectin type III repeat contains a glycosaminoglycan-binding site the occupation of which inhibits binding of the core protein and that the chondroitin/dermatan sulfate chain of decorin can also interact with this site. Divergent inhibition patterns were obtained for the binding of decorin or decorin core protein to the fusion proteins compared with intact collagen XIV in the presence of prebound glycosaminoglycans (Fig. 7). This could imply additional binding sites for glycosaminoglycans, especially chondroitin sulfate C, in domains of collagen XIV not covered by fusion proteins 29 -115 and 29 -154. Thus preincubation with chondroitin sulfate C resulted in a similar inhibition of binding to intact collagen XIV as preincubation with other glycosaminoglycans, but had only little effect on the binding to the fusion proteins. However, in the absence of glycosaminoglycan inhibitors binding of decorin to intact collagen XIV and to fusion protein 29 -154 is highly similar as judged by the parallel displacement curves of radiolabeled decorin by increasing quantities of unlabeled decorin (Fig. 8). These parallel displacement curves indicate that the identified binding region for decorin may be the sole relevant site on collagen XIV. However, since fusion proteins examined in our study covered only the NC 3 domain of collagen XIV and were not glycosylated and possibly folded incorrectly due to their generation in E. coli, we cannot definitely exclude that other binding sites might have escaped detection.
Decorin is suggested to connect neighboring collagen fibrils (47-49). Rotary shadowing-electron microscopy (48) and molecular modeling (49) imply the decorin core protein to be horseshoe-shaped. It is located at the d-band (49) of collagen I fibrils and could bind within the gap between the collagen molecules in a fibril, thereby preventing incorrect addition of collagens in the gap and promoting the correct formation of fibrils (49). The interaction of the highly anionic chondroitin/ dermatan sulfate chain of decorin with fibrillar collagens (50) could occur directly via a cationic region of a neighboring collagen molecule (48), or indirectly by the formation of a duplex with the glycan chain of another collagen bound proteoglycan (47). Collagen XIV could modulate the interactions between decorin and collagen fibrils. How these interactions are influenced by the binding of collagen XIV and the collagen XIVbinding site on the decorin core protein remain to be elucidated.
The expected functional role of decorin in regulating collagen fibrillogenesis was underlined by observations in mice with a targeted disruption of the decorin gene (39). These animals have a fragile skin with reduced tensile strength and an abnormal collagen morphology in skin and tendon. The fibrils have coarser and irregular outlines, indicating uncontrolled lateral fusion of collagen fibrils and providing an explanation for the reduced tensile strength (39).
Collagen XIV is located primarily at the surface of highly ordered, regular collagen fibrils with uniform diameter (8,19). As the similar collagen XII, it may be involved in the formation of collagen fibrils (20), or together with decorin in their supramolecular organization. Since collagen XIV has been observed occasionally at the fibril surface at a repeat distance close to that of the collagen fibrillar D-period (18,19), its association with decorin which is found regularly at the d-band (49) is possible in vivo. Accordingly, collagen XIV and decorin are both located in tendon, dermis, and connective tissue septa of skeletal and cardiac muscle, produced by fibroblasts (8,17,40,51) and liver fat-storing cells (52,53), and up-regulated in their expression in liver fibrosis (52,53). The growth inhibition of mammalian cells and the suppression of the malignant phenotype in colon carcinoma cells by the expression of decorin (54), as well as the predominant expression of collagen XIV in well differentiated mesenchymal tissues and its virtual absence in tumor stroma (8), point to a further common role of both molecules in differentiation and cellular quiescence. Further investigations are required to determine the physiological role of the collagen XIV-decorin interaction.