Recombinant decorin glycoforms. Purification and structure.

The vaccinia virus/T7 bacteriophage expression system was used to express human decorin in HT-1080 cells by co-infection with vTF7-3, encoding T7 RNA polymerase, and vDCN1, encoding the decorin core protein fused to a polyhistidine-insulin signal sequence fusion-protein cassette. Overexpression using the vaccinia virus/T7 phage system resulted in secretion of approximately 30 mg of decorin/109 cells per 24 h which enabled purification and separation of multiple glycoforms under native conditions. Cells were cultured in the presence of [35S]methionine or a mixture of [3H]glucosamine and [35S]sulfate, and recombinant glycoprotein purified by metal affinity chromatography which resolved the secreted decorin into two classes, a proteoglycan form and a core protein form. About 25% of the recombinant protein was secreted into the culture medium as core protein devoid of glycosaminoglycan chains. The decorin core protein was resolved into two forms (~49 and ~53 kDa) that differed in the extent of N-linked oligosaccharide substitution (2 and 3 N-linked oligosaccharides, respectively). Deglycosylation of the recombinant proteoglycans and core proteins resulted in a single band migrating with an apparent molecular mass ~ 43 kDa when analyzed by SDS-polyacrylamide gel electrophoresis. Far-UV circular dichroism spectra of native decorin proteoglycan showed a minima at 218 nm, consistent with a secondary structure that is predominantly β-sheet. Circular dichroism spectra of bovine decorin extracted from articular cartilage and recombinant decorin similarly treated revealed a minima of 205 nm indicating a loss of secondary structure. The affinity of decorin proteoglycan and core protein for collagen-like molecules was demonstrated, with the complement component C1q exhibiting the most striking affinity for decorin, although adherence to collagen types I and V was also observed. The extensive secondary structure maintained in the purified recombinant protein is likely to be important for the biological function of decorin.

The vaccinia virus/T7 bacteriophage expression system was used to express human decorin in HT-1080 cells by co-infection with vTF7-3, encoding T7 RNA polymerase, and vDCN1, encoding the decorin core protein fused to a polyhistidine-insulin signal sequence fusion-protein cassette. Overexpression using the vaccinia virus/T7 phage system resulted in secretion of approximately 30 mg of decorin/10 9 cells per 24 h which enabled purification and separation of multiple glycoforms under native conditions. Cells were cultured in the presence of [ 35 S]methionine or a mixture of [ 3 H]glucosamine and [ 35 S]sulfate, and recombinant glycoprotein purified by metal affinity chromatography which resolved the secreted decorin into two classes, a proteoglycan form and a core protein form. About 25% of the recombinant protein was secreted into the culture medium as core protein devoid of glycosaminoglycan chains. The decorin core protein was resolved into two forms (ϳ49 and ϳ53 kDa) that differed in the extent of N-linked oligosaccharide substitution (2 and 3 N-linked oligosaccharides, respectively). Deglycosylation of the recombinant proteoglycans and core proteins resulted in a single band migrating with an apparent molecular mass ϳ 43 kDa when analyzed by SDS-polyacrylamide gel electrophoresis. Far-UV circular dichroism spectra of native decorin proteoglycan showed a minima at 218 nm, consistent with a secondary structure that is predominantly ␤-sheet. Circular dichroism spectra of bovine decorin extracted from articular cartilage and recombinant decorin similarly treated revealed a minima of 205 nm indicating a loss of secondary structure. The affinity of decorin proteoglycan and core protein for collagenlike molecules was demonstrated, with the complement component C1q exhibiting the most striking affinity for decorin, although adherence to collagen types I and V was also observed. The extensive secondary structure maintained in the purified recombinant protein is likely to be important for the biological function of decorin.
Decorin, a small chondroitin sulfate proteoglycan is a ubiquitous component of the extracellular matrix of many tissues (1,2). The decorin mRNA encodes a prepro-core protein of 360 amino acids (3,4). This core protein can be divided into several distinct structural domains comprising: 1) a short signal sequence of 16 amino acids that targets the protein to the rough endoplasmic reticulum; 2) a propeptide of 14 amino acids of unknown function; 3) the glycosaminoglycan (GAG) 1 acceptor region with the chondroitin/dermatan sulfate chain substituted at the Ser-4 residue of the mature core protein (5); 4) a hypervariable cysteine globular domain; 5) a leucine-rich domain comprising a major portion of the core protein and containing three N-linked oligosaccharide attachment sites; and 6) a carboxyl-terminal globular domain. The mature decorin molecule that is recovered from tissues lack the propeptide domain. Decorin belongs to a growing family of proteoglycans with one or more GAG chains, that are characterized by core proteins of about 40 kDa containing 8 -12 homologous leucine-rich repeats (LRR) of 20 -29 residues with leucines in conserved positions (3,6). Similar LRR motifs are found in other proteoglycans like biglycan (7), fibromodulin (8), and lumican (9), and numerous other apparently unrelated molecules such as the porcine ribonuclease inhibitor (10,11). In the ribonuclease inhibitor, each LRR contains an alternating ␤ sheet-␣ helix structural unit, with the short ␤-strand and the ␣-helix approximately parallel to each other (12). The leucine-rich sequences of some of these proteins have been postulated to mediate protein-protein (13) and protein-membrane (14) interactions. The LRR repeats in the extracellular matrix proteoglycans are flanked by disulfide loops located at conserved positions near the NH 2 and COOH termini and thus may comprise a specialized subgroup within the LRR superfamily (3,6). Although biglycan, fibromodulin, lumican, and decorin are similar in their general structural features they differ markedly in their gene regulation, pattern of expression, and functional interactions.
Most of the biological functions ascribed to decorin are believed to relate to complex formation with other molecules via the GAG or core protein domains. Thus decorin has been demonstrated in vitro to bind to several proteins including fibronectin (15), transforming growth factor-␤ (16), and membrane receptors required for its endocytosis (17). In addition, decorin binds to certain types of fibrillar collagens including type I (18), type II (19), and type VI (20) and retards fibrillogenesis (19). Decorin also interacts with the collagen-like molecule C1q, a subcomponent of the C1 complex, the first component of the classical pathway of complement activation (21,22). However, without exception these studies have utilized decorin extracted from tissues under harsh denaturing conditions which are likely to alter the secondary structure of this molecule. In fact, we demonstrate in this study that decorin purified under native conditions has an extensive secondary structure that is severely and irreversibly disrupted when treated by standard extraction techniques (23,24). Binding of denatured decorin with other matrix molecules may not reflect biologically rele-vant interactions.
In the present study we have successfully utilized the vaccinia virus/T7 phage expression system to produce chemical amounts of differentially glycosylated forms of decorin that encompass the range of species described to date for tissue extracts and cell culture systems. We have purified the decorin glycoforms under native conditions and analyzed their glycosylation patterns. Preliminary investigation of the secondary structure of the decorin and biglycan extracted under denaturing and non-denaturing conditions is presented. The affinity of these recombinant native forms of decorin for collagenous extracellular matrix molecules is likely to better represent interactions that may occur in vivo.

EXPERIMENTAL PROCEDURES
All experimental procedures are as described in the accompanying paper (34), except for the following.
First-strand cDNA Synthesis and PCR Amplification-First-strand cDNA synthesis was done (Gene Amp kit, Perkin Elmer) with specific downstream oligonucleotide primers (0.5 g/reaction) and 1-2 g of RNA in a final reaction volume of 20 l. cDNA extension was at 42°C for 60 min, and the reaction terminated by incubation at 100°C for 5 min. PCR cDNA amplification was done using the entire 20-l firststrand cDNA reaction as a template in a final reaction volume of 100 l containing upstream oligonucleotide primer (0.5 g) and 2.5 units of Pfu DNA polymerase.
Decorin Fusion Protein Expression Constructs-Human skin fibroblast RNA was used to generate a first-strand cDNA encoding the mature form of the decorin core protein (i.e. amino acids 31-359). First-strand cDNA was synthesized using the downstream primer, DCN-M2 5Ј-GCCAGGTTATAAAAATGAGGGGATGCTTGAGA-3Ј (base pairs 1163-1194). A cDNA encoding the mature decorin core protein was then PCR amplified using the upstream primer, DCN-M1 5Ј-GGCTTATTTGACTGCATGCTAGAAGATGAGGC-3Ј (base pairs 148 -179). An SphI site was created by site-specific changes in both oligonucleotide primers (bold). A shuttle vector, pAH1 containing a factor Xa protease cleavage sequence was digested with SphI. The decorin cDNA was cloned into pAH1 at the SphI site to yield pAH1-DCN. The pAH1-DCN plasmid was used as a template for PCR amplification using the following synthetic oligonucleotide primers: upstream primer, P3 5Ј-CGGGATCCGAATTCGCCATCGAGGGTAGGGGCATG-3Ј; and downstream primer, P4 5Ј-CGGAAACTATAAGTAATTCTCTG-CAGCCGCGGGGA-3Ј. An EcoRI site at the 5Ј end (primer 3) and a SacII site at the 3Ј end (primer 4) (bold) were created. These sites were used to clone the decorin cDNA into a polyhistidine vaccinia expression vector, pBGN4 (34). pBGN4 was digested with EcoRI and SacII to create sites for insertion of the decorin cDNA fragment. The resulting polyhistidine fusion vector, pDCN1 was sequenced to ensure no rearrangements occurred during PCR reactions.
Immunoprecipitation-A 41-mer synthetic peptide spanning amino acid residues 180 -220 of human decorin was coupled to KLH (Immunogen Conjugation kit, Pierce), and used to generate a rabbit polyclonal antiserum (under contract with Cocalico Inc., Reamstown, PA). The antiserum, PR2 demonstrated specificity toward decorin and in separate experiments (not shown) exhibited immunoreactivity similar to the widely used LF-30 antiserum (kindly provided by Dr. Larry Fisher, National Institutes of Health). Five l of the cell-free translation reaction mixture was preadsorbed with normal rabbit serum and protein A-Sepharose for 2 h at 4°C. The supernatant was incubated with PR2 coupled to protein A-Sepharose and bound antigen released from the beads by incubation with SDS-PAGE sample buffer (see below).
Vaccinia Virus-The decorin expression plasmid, pDCN1 was used to generate a recombinant vaccinia virus by homologous recombination. This recombinant virus containing the polyhistidine-decorin fusion protein sequence under control of the T7 promoter, vDCN1, was propagated, amplified, and titered by standard methods (25).
Circular Dichroism (CD) Spectroscopy-A Jasco J720 spectropolarimeter was calibrated with a 0.1% (w/v) 10-d-camphor sulfonic acid solution and CD spectra measured at room temperature in a 2-mm path-length quartz cell. Purified protein was exchanged into 10 mM Tris, 1 mM EDTA, pH 8.0, and protein concentration calculated by theoretical molar extinction coefficient (26). Estimation of secondary structure content was done according to Sreerama and Woody (27).
Competitive Inhibition Assay-Decorin proteoglycan and core protein were radiolabeled (28) and separated from free [ 125 I]iodine by chromatography on PD-10 (Pharmacia) columns equilibrated with phosphatebuffered saline containing 0.5 mM MgCl 2 , 0.9 mM CaCl 2 (PBS ϩ ), 0.1% BSA. Specific activity of the [ 125 I]iodine-labeled preparations were ϳ3 ϫ 10 7 cpm/g. Immulon 1 wells were incubated overnight with 2 g of C1q (20 g/ml) or 50 l of 1% BSA. Wells were washed with PBS ϩ and blocked with 1% BSA for 1 h at 37°C. Non-radiolabeled competitor (see text) was added at 0 -100 g/ml for 1 h and [ 125 I]decorin (5 ϫ 10 4 cpm/well) was added and allowed to incubate for 3 h at room temperature. Wells were washed 3 times and the amount of bound radiolabel measured. All determinations were done in triplicate (mean Ϯ S.D.).

Human Decorin Expression
Vector-A recombinant expression vector, pDCN1 (Fig. 1a), was developed for optimum production and secretion of decorin in eukaryotic cells using the vaccinia virus/bacteriophage T7 polymerase expression system, essentially as described in the accompanying paper (34). The vector, pDCN1, contains a cDNA that encodes the mature decorin core protein and does not contain sequence encoding the endogenous signal sequence or the putative propeptide domain. pDCN1 was transcribed and translated in vitro in the presence of Trans 35 S-label (Fig. 1b). The radiolabeled products migrated as a single predominant band with an apparent molecular mass ϳ 46 kDa (lane 1) that was selectively immunoprecipitated with PR2 (lane 2) confirming the presence of the epitope in the translation product. Transcription/translation and immunoprecipitation of other fusion proteins did not crossreact with PR2 (data not shown).
Expression and Purification of Recombinant Decorin-The vector pDCN1 was used to generate a new recombinant vaccinia virus by homologous recombination utilizing thymidine kinase sequences that flank the decorin fusion protein construct shown in Fig. 1a. HT-1080 cells do not secrete endogenous decorin (data not shown) and were co-infected with vDCN1 and vTF7-3. Cultures were incubated in medium containing Trans 35 S-label, the medium harvested and recombinant decorin purified by nickel chelating chromatography (Fig.  2a). Several peaks were observed, including a large peak of radioactivity in the unbound fraction which did not contain immunoreactive decorin. The material eluted from the column by stepwise increases in imidazole concentration was moni-  2a). A mixed population containing decorin proteoglycan (median molecular mass ϳ 90 kDa) and decorin core protein (molecular mass ϳ 50 kDa) eluted at 80 mM (lane 46), with the majority of the radioactivity associated with the core protein form. Higher concentrations of imidazole, between 80 and 100 mM, were required to elute the remainder of the bound radioactivity (Fig. 2a, fractions 45-55) which was predominantly decorin core protein (Fig. 2b, fractions 46, 49, and 52). Identity of both the proteoglycan and core protein forms as decorin was confirmed by Western blot with PR2 antiserum (data not shown). The recombinant decorin was purified to greater than 90% by a single passage over the nickel column based on the absence of any other radiolabeled proteins co-purifying (Fig.  2b) and on the absence of any additional Coomassie staining bands (see below).
A sample containing both decorin proteoglycan and decorin core protein radiolabeled biosynthetically with Trans 35 S-label and purified by nickel affinity chromatography was incubated with factor Xa protease and reapplied to a nickel column. Decorin that was not incubated with the factor Xa protease bound to the resin and was eluted with 200 mM imidazole, whereas after factor Xa digestion about 80% of the decorin (both proteoglycan and core protein) no longer bound to the resin and eluted in the unbound fraction (data not shown). The reason for incomplete cleavage is not clear, but may relate to enzyme inactivation. There was no apparent preferential cleav-age of the proteoglycan versus the core protein forms indicating that the level of glycosylation of the fusion protein did not influence access of the protease to the cleavage site.
About 75% of the total recombinant protein purified from the culture supernatant was the proteoglycan form of decorin based on incorporation of 35 S-radiolabel (Cys and Met) before and after chondrotinase ABC digestion. In addition to the apparent high level of purity attained by the nickel column, we also noted that there was some resolution into subpopulations: the early eluting decorin proteoglycan was of larger molecular size compared to the later eluting proteoglycan (Fig. 2b, fractions 30 and 46, square brackets). The resolution of two subpopulations of proteoglycan was clearer when the proteins were labeled with [ 35 S]sulfate and [ 3 H]glucosamine (see below). Although separation into subpopulations was more evident using a step elution procedure, it was still apparent during gradient elution procedures (data not shown). The core protein form of decorin also migrated on SDS-PAGE as a doublet, molecular mass ϳ 53 and ϳ49 kDa, with the larger core protein eluting slightly ahead of the smaller form on the nickel column (Fig.  2b, fractions 46 -52).
Recombinant decorin was isolated from a large number of co-infected HT-1080 cells (2.5 ϫ 10 9 ) by a preparative scale nickel column. The concentration of purified samples were estimated by theoretical molar extinction coefficient and analyzed by 7.5% SDS-PAGE stained with Coomassie Blue (Fig. 3). Decorin proteoglycan was predominantly eluted in the 60 mM imidazole fraction, and 10, 5, and 1 g were applied to the gel (Fig. 3, lanes 1-3, respectively). Decorin core protein was enriched in the 80 mM imidazole elution and an estimated 5 g was loaded (Fig. 3, lane 4) which migrated as a doublet, with the larger form (molecular mass ϳ 53 kDa) more abundant. No other contaminating protein bands were visualized on the gel, indicating purity of the sample. Based on these data, we are able to estimate that expression and purification of recombinant decorin under native conditions (i.e. in the absence of chaotropic solvents or detergents) yields about 10 mg of proteoglycan and 3.3 mg of core protein/10 9 cells per 24 h. Recombinant decorin proteoglycan was also isolated and purified under denaturing conditions from HT-1080 culture supernatant in the presence of the chaotropic solvent 4 M GdnHCl and the detergent 0.5% Triton X-100. The use of this solvent throughout the extraction and purification protocol increased the final yield of proteoglycan 2-3-fold compared with the yield under native conditions. The significant increase in yield of recombinant protein in the presence of detergent was most likely a consequence of the hydrophobicity of native material. 2 Differential Glycosylation of Recombinant Decorin-Samples of Trans 35 S-labeled decorin proteoglycan and core protein were subjected to differential digestion with glycosidases to assess the nature of the carbohydrate substitutions. Undigested sample (Fig. 4, lane 1) shows the presence of the proteoglycan smear and a doublet core protein. Following digestion with chondroitinase ABC (Fig. 4, lane 2), the proteoglycan smear is no longer visible and all the radiolabel is coincident with the core protein doublet. Separate experiments showed that the proteoglycan was equally susceptible to chondroitinase ACII digestion indicating that the proteoglycan contained predominantly chondroitin sulfate. Chondroitinase ABC-treated sample was further incubated with N-glycosidase F (Fig. 4, lane 3) which resulted in the doublet core protein reducing to a single band with an apparent molecular mass ϳ 43 kDa. Chondroitinase ABC digestion of proteoglycan generated a doublet core protein, indicating that both core protein forms contain GAG (data not shown). The core protein doublet appears to reflect differential N-glycosylation of the decorin core protein, most likely representing 3 and 2 N-linked oligosaccharides based on the shift in migration position (ϳ10 and ϳ6 kDa, respectively). There was no evidence for secretion of unglycosylated decorin, or a proteoglycan form without N-linked oligosaccharides.
Proteoglycan Characterization-HT-1080 cells co-infected with vDCN1 and vTF7-3 were incubated in the presence of [ 35 S]sulfate and [ 3 H]glucosamine. Radiolabeled macromolecules were applied to a nickel column, bound material eluted using the protocol described for Fig. 2a, and radioactivity of collected fractions monitored. Two distinct peaks of radioactivity were eluted (data not shown), one at 60 mM imidazole and the second at 70 -80 mM imidazole, and were distinguished by the differential ratio of 35 S: 3 H which reflected the composition of the two pools. The early eluting peak (pool 1) and the later eluting peak (pool 2) were collected and aliquots subjected to gel filtration chromatography on Superose 6. Pool 1 eluted predominantly as a single broad peak (Fig. 5a) centered at K d 0.40 (mass ϳ 94 kDa, compared to globular protein standards), with the 35 S and 3 H co-incident. Pool 2 resolved into a single broad 35 S-labeled peak centered at K d 0.45 (mass ϳ 80 kDa); and two distinct 3 H-labeled peaks, one coincident with the 35 S label, and a second, sharp peak at K d 0.60 (mass ϳ 50 kDa) which consisted exclusively of 3 H label (Fig. 5b). Based on the 2 Separate experiments investigating the hydrophobicity of native preparations of recombinant decorin were done by elution on Q-Sepharose anion exchange resin in the presence of a variety of solvents. Solvents containing 10 M formamide and 0.5% Triton X-100 allowed for greater than 90% recovery of both proteoglycan and core protein forms, however, the absence of either a denaturing solvent (formamide) or detergent (Triton X-100) reduced total recovery to about 60%. Chromatography in physiological solvents (50 mM Tris-HCl, pH 6.5) resulted in 60% loss of the proteoglycan and almost 80% loss of core protein. Most of the lost material could be recovered by extraction of the resin with 4 M GdnHCl, 0.5% Triton X-100. predicted composition of the two pools (see Fig. 2b), pool 1 represents decorin proteoglycan alone and pool 2 comprises a mixture of proteoglycan and core protein. The core protein from pool 2 is radiolabeled via 3 H incorporation into N-linked oligosaccharides. The broad, high molecular weight peaks in both pools represent proteoglycan containing 35 S-and 3 H-labeled chondroitin sulfate chains. This was confirmed by chondroitinase ABC digestion of pool 1 and pool 2 and elution on Superose 6 (Fig. 5, c and d). More than 95% of the 35 S label and 80% of the 3 H label in pool 1 shifted to the V t of the Superose 6 column after chondrotinase ABC digestion, leaving a small, sharp peak containing 3 H label at K d 0.60 (Fig. 5c, asterisk). Chondroitinase ABC digestion of pool 2 shifted more than 95% of the 35 S and 60% of the 3 H label to the V t , leaving a single, sharp peak containing primarily 3 H label at K d 0.60 (Fig. 5d). After chondroitinase ABC digestion, the Superose 6 resolves a core protein pool of molecular mass ϳ 50 kDa, consistent with the core protein size determined on SDS-PAGE (ϳ49 and ϳ53 kDa).
The molecular size difference between the decorin proteoglycan in pools 1 and 2 is illustrated by their different migration positions on SDS-PAGE (Fig. 6a). This size difference is largely due to differences in chondroitin sulfate chain size as determined from Superose 6 chromatography after alkaline hydrolysis. However, anomolous migration of proteoglycans on SDS-PAGE may be influenced by microheterogeneity in disaccharide composition which could affect the total charge density of the molecule. Disaccharide analysis of the chondroitin sulfate chains (summarized in Fig. 6b) show ratios of ⌬di-0S:⌬di-4S:⌬di-6S in the larger chondroitin sulfate chains (pool 1) of 2.7:1.0:10.6, markedly different for that observed for the shorter chains (pool 2) of 2.0:1.0:3.7. These two subpopulations partially resolve by nickel chelating chromatography and this may relate to both the size and different disaccharide composition of the chondroitin sulfate chains.
Circular Dichroism-Recombinant decorin proteoglycan purified under native conditions was analyzed by CD spectroscopy. The far-UV spectra of the native proteoglycan showed greater than 70% secondary structure which was predominantly ␤-sheet as indicated by a sharp minima at 218 nm (Fig.  7a, closed circles). Analysis of the CD spectra based on computer modeling (27) predicted the native proteoglycan to comprise 54% ␤-sheet, 14% ␤-turn, 12% ␣-helix, and 20% random coil. Recombinant decorin proteoglycan was also isolated and purified under denaturing conditions from HT-1080 culture supernatant in the presence of the chaotropic solvent 4 M GdnHCl and 0.5% Triton X-100. After exchange back into physiological solvent, the CD spectra obtained for this material (Fig.  7a, open circles) revealed a sharp minima around 205 nm indicating an unordered conformation with no detectable secondary structure. Decorin extracted from bovine articular car-tilage, representative of material utilized in numerous binding studies (29,30), was also devoid of any secondary structure (Fig. 7a, open squares), and was similar to the spectra obtained for native decorin after thermal denaturation. The CD spectra for similarly produced recombinant biglycan (34) was identical to that for recombinant decorin; and exposure to denaturants resulted in disruption of the secondary structure. Consistent with these observations of significant secondary structure in the recombinant decorin proteoglycan, the PR2 antiserum selectively immunoprecipitated guanidine HCl-extracted decorin core protein (Fig. 7b, lane 1) but not the native recombinant decorin (Fig. 7b, lane 2), however, both denatured and native preparations of recombinant decorin were recognized by Western blot analysis (Fig. 7b, lanes 3 and 4). This is consistent with the epitope being inaccessible in the native conformation (i.e. conditions for immunoprecipitation), but available upon unfolding of the molecule (i.e. SDS treatment and transfer to nitrocellulose membrane).
Both the proteoglycan and core protein exhibit a high degree of hydrophobicity when purified under native conditions, 2 with this being more pronounced for the core protein preparation. This hydrophobicity may be a consequence of the secondary structure, since similar behavior was not apparent for GdnHCltreated material. The presence of a polyanionic chondroitin sulfate chain on the decorin proteoglycan may play a role in maintaining a stable secondary structure in solution.
Decorin Binds Extracellular Matrix Collagens and C1q-The binding of 35 S-labeled decorin proteoglycan and core proteins to collagen types I, II, III, V, VI, C1q or BSA was examined, and showed an identical pattern to that of biglycan (34). The binding of decorin to the fibrillar collagens was overshadowed by the much greater binding observed with C1q. Interaction with C1q appears to be mediated by the core protein of decorin, with binding up to 18-fold over the BSA control compared to 9-fold binding of intact proteoglycan. Krumdieck et al. (21) have shown that decorin proteoglycan and chondroitinase ABC-generated core protein from bovine articular cartilage also bind C1q.
The binding of 125 I-labeled proteoglycan and core protein to C1q was subjected to competition by the addition of unlabeled decorin proteoglycan and core protein (Fig. 8a, open and closed  circles, respectively). Both decorin proteoglycan and decorin core protein inhibited binding of 125 I-labeled decorin to C1q. The IC 50 (concentration of competitor required for half-maximal inhibition) for the core protein was ϳ28 g/ml (0.62 M) and for intact proteoglycan the IC 50 value was ϳ18 g/ml (0.2 M; Fig. 8a). Saturability of the decorin-C1q interaction was  1 and 2) or probed by Western blot (lanes 3 and 4) using the antibody PR2. investigated by adding increasing amounts of radiolabeled decorin (specific activity 200,000 cpm/g) to wells coated with C1q or 1% BSA. The inset in Fig. 8, b and c, demonstrates that specific binding of decorin proteoglycan and core to C1q reaches a plateau at 0.6 g of proteoglycan and 0.4 g of core protein added. Assuming a single binding site on the C1q molecule, Scatchard analysis (31) yielded a dissociation constant for the decorin-C1q interaction of 50 and 26 nM for the binding of the recombinant decorin proteoglycan and core proteins, respectively.

DISCUSSION
In this study, differentially glycosylated forms of decorin have been overexpressed in a eukaryotic cell line with a con-nective tissue origin (HT-1080). These decorin glycoforms were separated from other secreted molecules to greater than 95% purity, and further partially resolved by metal chelating affinity chromatography into two distinct populations: proteoglycan forms and core protein forms. The proteoglycan population was further partially resolved into two species distinguished by different sized chondroitin sulfate chains with distinct disaccharide composition; the core protein subpopulation was also represented by two distinct forms, distinguished by the extent of N-linked oligosaccharide substitution. The core protein preparations generated by chondroitinase ABC digestion of the proteoglycan pool comprised both the 2 and 3 N-linked oligosaccharide forms; we cannot exclude the possibility that the different sized GAG chains are on different core proteins (i.e. the 2 and 3 N-linked oligosaccharide forms, respectively), however, this would seem unlikely given a similar bi-modal population of GAG chains was observed for recombinant biglycan (34) where there is no evidence for expression of two core proteins.
The two non-proteoglycan forms of recombinant decorin had an apparent mobility of ϳ49 and ϳ53 kDa on SDS-PAGE. Differential N-glycosylation of decorin synthesized by human skin fibroblasts in vitro has been demonstrated previously (32), whereby chondroitinase ABC-digested decorin migrated as a doublet core protein preparation which further reduced to a single band when synthesized in the presence of tunicamycin. Although these investigators were able to induce N-glycan-free decorin proteoglycan synthesis through exogenous addition of tunicamycin, we saw no evidence for expression of similarly N-glycan-free decorin core protein expression, nor is there any evidence that this is a biosynthetic product in vivo. In fact, it is likely that core protein devoid of both GAG and N-linked oligosaccharides may be insoluble, as evidenced by the difficulties associated with bacterial expression of these molecules (33).
The different glycoforms of decorin expressed in this system are similar to the variety of decorin glycanated and non-glycanated species already identified in tissues. The proteoglycan form of the recombinant decorin is characteristic of material extracted from tissues in terms of chondroitin sulfate chain length and pattern of sulfation, as well as differential N-glycosylation. The recombinant decorin secreted devoid of GAG chains is likely a product of overexpression of core protein exceeding the capacity of one or more glycosyltransferases. Purification and quantitation of the recombinant decorin indicates that the HT-1080 cells secreted up to 30 -40 mg of decorin/10 9 cells per 24 h, about 75% as a proteoglycan form and the remainder as core glycoprotein. Under native conditions we experienced significant losses of material, but are nevertheless able to purify chemical amounts that are correctly folded and biologically active. Circular dichroism spectra and secondary structure prediction clearly demonstrate that about 70% of the native recombinant intact decorin assumes a secondary structure with a high proportion of ␤-sheet content. This is not surprising given that about 70% of the core protein is comprised of the LRR domain which is predicted to assume a complex structure in solution. However, both recombinant decorin subjected to denaturing conditions and decorin extracted from cartilage by standard techniques (29,30) demonstrated a loss of secondary structure in their CD spectra. This strongly argues that techniques commonly employed for isolation of decorin and biglycan from tissues (23,24,29) have a permanent effect on the conformation of the core protein. These changes in secondary structure may have implications on the interpretation of data reported by others related to interactions between decorin and biglycan with other extracellular matrix molecules. A recent FIG. 8. Decorin-C1q interaction. a, 125 I-labeled decorin proteoglycan (E) and 125 I-labeled core protein (q) binding to C1q-coated microtiter wells was inhibited with increasing amounts of unlabeled recombinant intact proteoglycan (E) and core protein (q). All determinations were done in triplicate (mean Ϯ S.D.). b and c, Scatchard analysis. Microtiter wells coated with 2 g of C1q or 1% BSA were incubated with increasing concentrations of 125 I-labeled decorin proteoglycan or 125 Ilabeled core protein (specific activity 200,000 cpm/g). The amount of bound and free ligand was quantitated for decorin proteoglycan (E) and core protein (q). The insets show that binding was saturable. study (23) reported that decorin extracted from cartilage with GdnHCl was not an effective competitor of native decorin binding to type I collagen. Given many previous reports of GdnHClextracted decorin interacting with type I collagen in a variety of assays, it is possible that there are both secondary structuredependent and -independent interactions. It is likely that interactions requiring correctly folded decorin are important in vivo so it is important to investigate these interactions using native preparations.
Decorin bound very effectively to the complement component C1q, showing an 18-fold increase in binding with core protein compared to BSA, and a 9-fold increase in binding with the proteoglycan form. Since the presence of a GAG chain appears to have attenuated the extent of binding it is likely that this interaction is mediated via the core protein, and the GAG chain may interfere with the ability of C1q to access the core protein domain. C1q is a subcomponent of the C1 complex, the first component of the classical pathway of complement activation (21,22). The C1q molecule consists of two domains, a collagenlike triple helical domain at the N terminus and a globular configuration at the C terminus and circulates in the plasma as a Ca 2ϩ -dependent complex with C1r and C1s (22). Although the biological significance of C1q-decorin interaction is not known, decorin may modulate complement-mediated inflammation at sites where extracellular matrices containing decorin are exposed to complement components (21). The C1q-decorin interaction was competitively inhibited by both core protein and proteoglycan preparations, however, an intact proteoglycan (i.e. substituted with a GAG chain) is required for inhibition of the complement cascade (21). Although both native and denatured decorin bind to C1q, it remains to be determined if the two forms interact via the same domain or whether the interaction with denatured decorin is simply nonspecific binding. The specificity and affinities of native decorin glycoforms for collagen-like molecules is the subject of further study.
This study, and the accompanying paper (34), have demonstrated the high level expression and purification under native conditions of the small interstitial proteoglycans, decorin and biglycan. These glycoconjugates possess a high degree of secondary structure not apparent in similar molecules extracted from hard tissues by standard techniques, in addition to maintaining biological activity in several binding assays. Furthermore, several glycoforms have been isolated that are similar to subpopulations present in a variety of tissues in vivo. These recombinant decorin proteoglycan and core protein forms isolated in chemical amounts will allow further study on the biological role and molecular interactions of this class of LRRcontaining molecules.