Distinct isoforms of chicken decorin contain either one or two dermatan sulfate chains.

Decorin, a member of a family of proteins with leucine-rich repeat motifs, is a widely distributed extracellular matrix proteoglycan that is thought to be responsible for the structure, tissue organization, and surface properties of fibrils. In mammals, decorin carries one chondroitin/dermatan sulfate chain as a distinction from its homologue, biglycan, which contains two glycosaminoglycan chains. With the aim to study decorin-collagen interactions in chicken, where the fibrillar organization of cartilage collagens is best understood, we have isolated decorin-related proteoglycans from sternal cartilage of 40-day-old broiler chickens. Small chondroitin/dermatan sulfate proteoglycans were resolved by hydrophobic interaction chromatography into two fractions, DCN I and DCN II. Both forms contained dermatan sulfate and, in addition, keratan sulfate chains. Tryptic fingerprinting revealed that the core proteins of DCN I and DCN II were identical. The protein was identified as decorin by amino-terminal sequencing. DCN II was found to contain two dermatan sulfate chains, whereas DCN I had a single dermatan sulfate chain. The dermatan sulfate attachment sites are located near the NH2 terminus of the core protein, i.e. at Ser-4 and Ser-16 in DCN II and at Ser-4 in DCN I. The keratan sulfate attachment sites are located in the central portion of the core protein, at Asn-179 and Asn-230. The presence of two dermatan sulfate chains renders the chicken proteoglycan DCN II structurally similar to mammalian biglycan. Interestingly, biglycan has not been detected in chicken. Therefore, in birds, DCN II may function as a biglycan substitute.

The mechanical properties of a connective tissue are largely determined by the composition of its extracellular matrix and the interactions between the matrix macromolecules. Tensile strength is generated by a framework of insoluble fibrils that are rich in collagen. The fibrils are embedded in a hydrated gel formed by proteoglycans and other glycoproteins. In cartilage fibrils, collagens II, IX, and XI form a heterotypic aggregate which, presumably, is essential for the control of fibril dimensions and surface properties (1,2). Other participants in the fibril assemblages are collagen-binding glycoproteins, such as decorin and fibromodulin (3), but their functions are not well understood.
Decorin (4) is a small chondroitin sulfate (CS) 1 /dermatan sulfate (DS) proteoglycan present in most extracellular matrices (reviewed in Refs. 3 and 5). It belongs to a family of secreted glycoproteins also including the CS/DS proteoglycans biglycan (6) and PG-Lb (7) and the keratan sulfate (KS) proteoglycans fibromodulin (8), lumican (9), and keratocan (10), as well as the proteins chondroadherin (11) and PRELP (12). Within this group of macromolecules, the protein structures are composed mainly of repeated motifs of 20 -25 amino acid residues with abundant leucine residues in conserved positions (3). Leucinerich repeats are known to occur in more than 60 different proteins of eucaryotic or procaryotic origin, and in many cases they appear to be involved in protein-protein interactions (13,14). In the members of the decorin family, the leucine-rich repeat portion is located in the center of the protein and is flanked by structurally less conserved NH 2 -and COOH-terminal regions. The NH 2 -terminal region harbors four conserved cysteine residues, with a disulfide bond connecting the first with the fourth (15). Similarly, there is an intrachain disulfide bond between two conserved cysteines near the COOH terminus. There are significant structural differences between these molecules in the segment NH 2 -terminal to the first cysteine. Mammalian decorin contains a single CS/DS chain that is attached to a serine residue at position 4 (16). Biglycan usually has two CS/DS chains attached to serine residues at positions 5 and 11 (15). Fibromodulin and lumican do not have glycosaminoglycans in the NH 2 -terminal regions, but tyrosine sulfate residues may be present instead (17).
Ultrastructural studies suggest that decorin is associated with collagen-containing fibrils in a number of connective tissues (18 -20). Decorin binds to collagens I and II and inhibits collagen fibril formation in vitro (21,22). The effects on collagen I fibrillogenesis in vitro include a delayed initial assembly of collagen molecules and a decreased final fibril diameter (23). The proteoglycan shows a high affinity for native collagen I molecules, with a dissociation constant in the order of 10 Ϫ8 M (22,24), or even 10 Ϫ9 M when the decorin has been isolated under strictly nondenaturing conditions (25,26). High affinity binding depends on the core protein, but the glycosaminoglycan chain can provide additional collagen binding sites (27). The latter interactions presumably are electrostatic since they are disrupted in the presence of phosphate or sulfate ions at Ͼ20 mM (28). Analogously to decorin, fibromodulin binds to collagens I and II and delays fibrillogenesis in vitro (22). However, the two proteoglycans bind to separate distinct sites on collagen fibrils (27). The core protein of biglycan, which is 50 -60% homologous to that of decorin, does not show a similar interaction with collagens (24,27,28) although a weaker affinity for collagen I (K d ϳ 10 Ϫ7 M) has been reported (25). Recently, it was shown that a recombinant proteoglycan chimera having the structure of biglycan except for leucine-rich repeats 4 -5, which were from decorin, was able to bind to collagen I with almost the same affinity as recombinant decorin (26). This suggests that the affinity for collagen is determined by the detail structure of a stretch of some 40 amino acid residues within the central portion of the core protein.
With a view to investigating the properties of cartilage fibrils, our attention has focused on small interstitial proteoglycans of chicken, because collagen structure in chicken cartilage is already well studied. In this article, we report on the isolation of decorin from chicken sternal cartilage. A novel variant with two DS chains is described. Since the avian counterpart to biglycan has not yet been identified, we speculate that this decorin isoform might partially substitute for biglycan functions.

EXPERIMENTAL PROCEDURES
Materials-Chondroitinase ABC was purchased from Seikagaku Kogyo (Tokyo, Japan). Endo-␤-galactosidase, endoproteinase Asp-N, and N-glycosidase F were from Boehringer Mannheim. Trypsin (L-1-tosylamido-2-phenylethyl chloromethyl ketone-treated), papain, chondroitin sulfate A, and dermatan sulfate were from Sigma. Octyl-Sepharose CL-4B, Sephacryl S-300 HR, and Sephadex G-25M were from Pharmacia Biotech Inc., and DEAE-cellulose was DE-52 from Whatman Ltd. Peroxidase-conjugated goat anti-(mouse IgG) antibodies were from Kirkegaard and Perry Laboratories (Gaithersburg, MD). Coomassie Brilliant Blue R-250 and dimethylmethylene blue were from Serva Feinbiochemica (Heidelberg, Germany). Guanidine HCl solutions were prepared from an 8 M stock solution, which had been treated overnight with activated charcoal and then filtered. Urea solutions were prepared from an 8 M stock solution, which was deionized with Serdolit MB-3 (Serva) prior to use.
Isolation of DCN I and DCN II-Sterna from 40-day-old broiler chickens were taken within 1 h after slaughter. Cartilage was freed from adherent tissues and homogenized in a 15-fold excess of 4 M guanidine HCl, 50 mM sodium acetate, pH 5.8, containing fresh protease inhibitors (100 mM 6-aminohexanoic acid, 10 mM N-ethylmaleimide, 10 mM benzamidine HCl, 50 mM EDTA, 0.1 mM phenylmethylsulfonyl fluoride). The suspension was stirred for 24 h at 4°C and then centrifuged at 17,700 ϫ g. The supernatant was collected, and its density was adjusted to 1.40 g/ml by addition of solid CsCl. A density gradient was formed by ultracentrifugation in a Beckman 60 Ti angle rotor at 108,000 ϫ g for 72 h at 14°C. The content of each tube was collected from the bottom as 36 fractions of 1.0 ml. A pool of fractions 10 -32, having densities between 1.45 and 1.30 g/ml, was transferred into 2 M guanidine HCl, 0.1 M sodium acetate, pH 6.3, by diaflow over a PM-10 membrane (Amicon, Beverly, MA). This material was chromatographed on a column of Octyl-Sepharose CL-4B (2.5 ϫ 40 cm) eluted with three bed volumes of the 2 M guanidine HCl buffer, followed by a linear gradient of 2-6 M guanidine HCl in 0.1 M sodium acetate, pH 6.3 (29). Fractions containing two CS/DS proteoglycans were pooled. Next, the proteoglycans were purified by anion-exchange chromatography. The sample was first equilibrated with 0.15 M NaCl, 7 M urea, 50 mM Tris-HCl, pH 7.4, by diaflow, and then passed through a column of DEAE-cellulose (1.6 ϫ 6 cm). Glycoprotein contaminants were recovered as unbound material by elution with 0.15 M NaCl, 7 M urea, 50 mM Tris HCl, pH 7.4. Proteoglycans were collected by a subsequent singlestep elution with 7 M urea buffer containing 1 M NaCl. This eluate was directly transferred back into 2 M guanidine HCl, 0.1 M sodium acetate, pH 6.3. Decorin was finally resolved into two fractions, DCN I and DCN II, by chromatography on a column of Octyl-Sepharose CL-4B (1.6 ϫ 10 cm) in 2 M guanidine HCl, 0.1 M sodium acetate, pH 6.3, at 20°C. A sample of proteoglycans, representing about 20 mg in 7 ml of buffer, was applied at a flow rate of 1 ml/h. The column was then washed with 60 ml of the 2 M guanidine HCl buffer. Elution followed at a flow rate of 5 ml/h with a linear gradient of 2-6 M guanidine HCl, formed using a GM-1 gradient mixer (Pharmacia Biotech Inc.), which contained 100 ml of 2 M guanidine HCl buffer and 100 ml of 6 M guanidine HCl, 0.1 M sodium acetate, pH 6.3. Fractions of 2 ml were collected and analyzed by SDS-PAGE. Fractions containing DCN I and DCN II, respectively, were pooled and stored at Ϫ20°C.
Polyacrylamide Gel Electrophoresis-Samples to be analyzed by SDS-PAGE were precipitated by adding nine volumes of ice-cold etha-nol and centrifuged at 4°C for 30 min at 17,000 ϫ g. The precipitates were dissolved in sample buffer containing 2% ␤-mercaptoethanol, heated to 95°C for 3 min, and then electrophoresed in 3.5-12% polyacrylamide gradient gels using the buffer system of Laemmli (30). After completion of electrophoresis, gels were stained with Coomassie Brilliant Blue.
Tryptic Peptide Mapping-Samples of proteoglycans were reduced with 20 mM dithioerythritol in 4 M guanidine HCl for 2 h at 37°C, followed by alkylation in the presence of a 2.5-fold molar excess of N-ethylmaleimide for 2 h at 20°C. The proteoglycans were recovered by ethanol precipitation and digested with trypsin (10 g/mg of proteoglycan) for 12 h at 37°C in 1% NH 4 HCO 3 , pH 8. After lyophilization, the peptides were dissolved in 0.1% (v/v) trifluoroacetic acid and injected onto an ODS Ultrasphere reversed phase HPLC column, 0.46 ϫ 15 cm, of 5-m 80-Å silica with a C-18 phase (Beckman, Palo Alto, CA), using a 20-l sample loop. The peptides were eluted at a flow rate of 1 ml/min with a gradient made from 0.1% trifluoroacetic acid (A) and acetonitrile containing 0.08% trifluoroacetic acid (B), as follows: 0 -2 min, 0% B; 2-42 min, 0 -50% B; 42-47 min, 50 -100% B.
Enzymatic Deglycosylation-Samples of proteoglycans were deglycosylated to generate core proteins and fragmented glycosaminoglycan chains. Chondroitinase ABC was added to proteoglycans (0.1 milliunit/g of proteoglycan) in 0.15 M NaCl, 50 mM Tris-HCl, pH 7.4, containing 1 mM ovomucoid. These samples were incubated for 4 h at 37°C. Digestion with endo-␤-galactosidase (0.2 milliunit/g of proteoglycan) was for 2 h at 37°C in 0.15 M NaCl, 50 mM sodium acetate, pH 5.8. Treatment with N-glycosidase F was carried out with samples in 0.15 M NaCl, 50 mM Tris-HCl, pH 7.4, that had been preheated to 60°C for 10 min. After cooling to 37°C, enzyme (0.05 unit/g of proteoglycan) was added and digestion was allowed to proceed for 12 h.
Disaccharide Analysis-The proportion of chondroitin 4-sulfate to chondroitin 6-sulfate was determined after digestion with chondroitinase ABC according to the method of Zebrower et al. (32). Chondroitin sulfate A, a mixture of 70% chondroitin 4-sulfate and 30% chondroitin 6-sulfate, was used as standard.
Size Exclusion Chromatography of Glycosaminoglycans-O-Linked glycosaminoglycans were liberated from the proteoglycans by treatment with 0.5 M NaBH 4 , 0.1 M NaOH, at 48°C for 48 h (33). The samples were neutralized with acetic acid and chromatographed on a column of Sephacryl S-300 HR (100 ϫ 0.6 cm) eluted with 2 M guanidine HCl, 0.25 M sodium acetate, pH 6.3. In a separate experiment, a sample of proteoglycan was digested with papain (200 g/mg of proteoglycan) in 50 mM KH 2 PO 4 , 10 mM EDTA, 2 mM dithioerythritol, pH 6.8, at 40°C for 24 h, and chromatographed on the Sephacryl S-300 column.
Isolation of Peptides with Attached Glycosaminoglycans-A sample of 2 mg of DCN II was reduced, alkylated, and digested with trypsin as described above. The peptides were dissolved in 300 l of 2 M guanidine HCl, 0.25 M sodium acetate, pH 6.3, and separated by size-exclusion chromatography on a column of Sephacryl S-300 HR (110 ϫ 1.0 cm) in the 2 M guanidine HCl buffer. Fractions of 1.0 ml were collected. These were analyzed for protein by monitoring the absorbance at 280 nm and for glycosaminoglycans as described (34). Fractions containing KS were pooled and lyophilized, followed by desalting on a column of Sephadex G-25M (11.5 ϫ 1.5 cm) in 1% NH 4 HCO 3 . A separate pool, which contained the CS/DS, was dialyzed against water and lyophilized. This material was digested with 2 g of endoproteinase Asp-N in 10 mM Tris-HCl, 50 mM sodium phosphate, pH 8.0, for 14 h at 37°C. It was then re-chromatographed on Sephacryl S-300 HR. Fractions containing CS/DS were pooled, treated with 10 milliunits of chondroitinase ABC, and subjected to reversed phase HPLC as described under "Tryptic Peptide Mapping." Peptide fractions were collected and vacuum-dried.
Detection of Glycosaminoglycans-Concentrations of sulfated glycosaminoglycans were determined by using the reagent 1,9-dimethylmethylene blue as described by Farndale et al. (34).
KS was detected using an immunoassay with the monoclonal antibody 5D4 (35). Aliquots of column fractions, diluted in 0.15 M NaCl, 50 mM Tris-HCl, pH 7.4, were kept in 96-well plates for 18 h at room temperature to allow adsorption of molecules. The plates were then rinsed with buffer containing 0.05% (v/v) Tween 20. The wells were incubated for 4 h with 5D4 at a dilution of 1:10,000, followed by incubation for 2 h with peroxidase-conjugated goat anti-(mouse IgG) antibodies at a dilution of 1:5000. Enzyme activity was measured after addition of 200 l/well of a substrate solution prepared by combining o-phenylenediamine (0.4 mg/ml) and 30% H 2 O 2 (0.4 l/ml) in 0.05 M sodium citrate, 0.1 M sodium phosphate. The reaction was stopped by addition of 50 l of 2.5 M H 2 SO 4 . Absorbance was read at 490 nm.
KS was also detected by immunoblotting (36) using the antibody 5D4 at a 1:5000 dilution and the peroxidase-conjugated goat anti-(mouse IgG) antibodies at a 1:5000 dilution.
Amino Acid Sequence Analysis-Samples of peptides were sequenced by Edman degradation in an automatic protein sequencer (Applied Biosystems model 477A, Foster City, CA) using the trifluoroacetic acidconversion program provided by the manufacturer.

RESULTS
Isolation of Proteoglycans-Extracts of sternal cartilage of 40-day-old chickens were subjected to CsCl-density gradient centrifugation in 4 M guanidine hydrochloride, and fractions containing low buoyant density proteoglycans (1.30 -1.45 g/ml) were pooled. The material was chromatographed on a column of Octyl-Sepharose CL-4B eluted with a linear gradient of 2-6 M guanidine HCl. Two CS/DS proteoglycans were recovered within a pool collected at guanidine HCl concentrations between 2.2 and 3.8 M. Contaminating glycoproteins were removed from this material by chromatography on DEAE-cellulose. Finally, the two proteoglycans were separated by hydrophobic interaction chromatography on Octyl-Sepharose (Fig. 1). Material that was eluted at the beginning of the gradient, between 2.2 and 2.7 M guanidine HCl, migrated as a single diffuse band upon SDS-PAGE. The position of the band indicated a molecular mass of more than 200 kDa. Fractions eluted toward the end of the gradient contained a small proteoglycan with an apparent molecular mass of 80 -150 kDa. The latter proteoglycan was tentatively designated as DCN I and the Ͼ200-kDa form as DCN II.
Identification of DCN I and DCN II-DCN I and DCN II were identified as isoforms of chicken decorin by sequence analysis of 20 NH 2 -terminal residues (Fig. 2). Blank cycles were observed in positions 4 and 10 of DCN I and in positions 4, 10, and 16 of DCN II. Comparison with an amino acid sequence derived from the published cDNA sequence of chicken decorin (37) revealed that these blank positions were occupied by Ser, Thr, and Ser, respectively. Moreover, the identity between the core proteins of DCN I and DCN II was confirmed by tryptic peptide mapping. DCN I and DCN II gave rise to identical patterns (Fig. 3).
Treatment of either DCN I or DCN II with chondroitinase ABC yielded a more compact band in a subsequent SDS-PAGE, corresponding to a molecular mass of ϳ45-55 kDa (Fig. 4, lanes  1 and 4). Digestion of the chondroitinase-treated proteoglycans with endo-␤-galactosidase did not substantially increase the electrophoretic mobility but resulted in a further sharpening of the bands (Fig. 4, lanes 2 and 5). Separately, chondroitinasetreated samples were digested with N-glycosidase F, an enzyme removing N-linked oligosaccharides including N-linked keratan sulfate. For DCN I as well as DCN II, a single core protein with an apparent molecular mass of ϳ40 kDa was observed (Fig. 4, lanes 3 and 6). The susceptibility to endo-␤-galactosidase suggested the presence of KS chain substituents. This was confirmed by immunoblots showing binding of the monoclonal antibody 5D4 to DCN I and DCN II. This antibody, which recognizes highly sulfated regions in KS (35), did not bind to the core proteins after digestion with N-glycosidase F (results not shown). In conclusion, both decorin isoforms DCN I and DCN II contained two types of glycosaminoglycan substituents, i.e. CS/DS as well as N-linked KS.

DCN I and DCN II Contain Dermatan 4-Sulfate-
The glycosaminoglycans of DCN I and DCN II were cleaved from the core proteins by alkaline borohydride treatment and hydrolyzed in 2 M trifluoroacetic acid. The uronic acid moieties were quantified by reversed phase HPLC of the perbenzoyl derivatives. Glucuronic acid constituted 95% of the total, the remainder being iduronic acid (data not shown). Hence, both isoforms of decorin were substituted with low iduronate dermatan sulfate chains. Furthermore, the DS chains of DCN I or DCN II were digested with chondroitinase ABC. HPLC analysis revealed that 90 -95% of the resulting disaccharides were 4-sulfated and 5-10% 6-sulfated. Thus, the glycosaminoglycan chains of DCN I and DCN II from chicken sternal cartilage largely consisted of dermatan 4-sulfate. Similar results have been reported previously for the decorin in chick cornea (38).
Dermatan Sulfate Chain Size of DCN II-DS chains from DCN II and bovine tendon decorin were prepared by ␤-elimination using treatment with alkaline borohydride. By this procedure, O-glycosidically linked saccharides are extensively hydrolyzed, while glycosaminoglycan chains remain intact. Consistent with this fact, papain treatment of the decorin molecules gave virtually identical results as observed for alkaline borohydride treatment (see below). Upon gel filtration on Sephacryl S-300 HR, the glycosaminoglycan chains of DCN II eluted on the same position as DS chains from bovine decorin (K AV ϭ 0.16), but significantly later than intact decorin. This suggests that the glycosaminoglycan chains of DCN II and bovine decorin are of similar size, i.e. 37 kDa (39).

Attachment Sites of Dermatan Sulfate Chains in DCN II-
The apparent molecular weight of DCN II was unusually high, although neither the core protein nor the DS chains were larger than in other decorin molecules. Therefore, it was suspected that DCN II harbors more than one DS chain. The most likely attachment sites are serine residues in positions 4 and 16, both of which gave blank cycles in the amino acid sequencing. Ser-16 is part of the sequence Ser-Gly-Xaa-Gly, which is recognized as the attachment site for the single DS chain in mammalian decorin. Ser-4 represents the previously described attachment site in chick decorin (37). To identify the glycosaminoglycan attachment sites in DCN II, oligopeptides were prepared by proteolytic cleavage and fragments carrying DS or KS chains were purified by size-exclusion chromatography on Sephacryl S-300 HR. DS chains released by digestion of DCN II with papain eluted as a homogeneous peak with K AV ϭ 0.16 (Fig.  5a). In contrast, a single tryptic fragment of DCN II containing all of the DS chains was recovered in the void volume (Fig. 5b). This result was in agreement with the view of two DS chains attached near the NH 2 terminus of the core protein.
The isolated tryptic fragment (Fig. 5b, pool I) was digested with endoproteinase Asp-N. This enzyme is known to cleave peptidyl-aspartate bonds, but not between two adjacent aspartate residues (40). The amino-terminal tryptic peptide of chicken decorin that comprises residues 1-25 harbors two cleavage sites for endoproteinase Asp-N, i.e. between residues 5 and 6 and between residues 10 and 11. Both sites are located between the presumptive DS attachment sites at residues 4 and 16 (Fig. 2). Peptides generated by cleavage with endoproteinase Asp-N were re-chromatographed on Sephacryl S-300 HR (Fig. 5c). The glycosaminoglycan carrying peptides eluted in fractions corresponding to K AV ϭ 0.16, similarly as after cleavage with papain. This material was further resolved into three fractions by reversed phase HPLC after removal of the DS chains by chondroitinase ABC (Fig. 6). One peptide was recovered from the breakthrough fraction, together with the digestion products of the glycosaminoglycan (Fig. 6, pool I). This peptide was separated from disaccharides by gel filtration on Sephadex G-25M (Fig. 7). The peptide was recovered from fractions 20 -25 and brought to sequence analysis. It was identified as a pentapeptide consisting of the five NH 2 -terminal amino acids of decorin (Fig. 2, solid line). Again, Ser-4 was not detected by automated Edman degradation because of the glycosylation of this amino acid residue.
The identified DS-carrying peptides were recovered from the fractions corresponding to the two major peaks in the reversed phase chromatogram (Fig. 6, pools I and II). The third peak was much smaller (Fig. 6, pool III), and no peptide material was found in the corresponding fractions. Therefore, the described sequence analyses confirmed that O-linked glycosaminoglycans were attached to Ser-4 and Ser-16 only.
Attachment Sites of Keratan Sulfate Chains-Peptides containing KS eluted in fractions with K AV values between 0.12 and 0.67 (Fig. 5b). Amino acid sequencing of this material showed two peptides in similar amounts, representing tryptic peptide-(175-185) and peptide-(218 -240). Each peptide contained one of the two Asn-Xaa-Ser/Thr sequences, which are FIG. 5. Gel chromatography of peptides generated by partial proteolysis of DCN II. Reduced and alkylated DCN II was cleaved with proteases followed by chromatography on Sephacryl S-300 HR columns. The eluent was 2 M guanidine hydrochloride, 0.25 M sodium acetate, pH 6.3. The column effluents were monitored for absorbance at 280 nm (---). Contents of CS/DS were estimated by using the 1,9dimethylmethylene blue assay (OO), KS was detected by immunoassay with the monoclonal antibody 5D4 (---). The measured signal is shown to an arbitrary scale. a, 0.25 mg of DCN II was digested with papain and applied to a 100 ϫ 0.6-cm column. b, 2 mg of DCN II was digested with trypsin and chromatographed on a 100 ϫ 1.0-cm column. Two pools were collected as indicated by the bars, pool I containing the CS/DS and pool II containing the KS. c, the material in pool I was digested with endoproteinase Asp-N and re-chromatographed on the 100 ϫ 1.0-cm column. Fractions containing CS/DS were pooled as indicated.  Fig. 6) was chromatographed on a column of Sephadex G-25M. The eluent was 1% ammonium bicarbonate. Fractions of 0.5 ml were collected and absorbances at 206 nm (q) and 232 nm (E) were measured, the latter wavelength in order to detect glycosaminoglycan digestion products. Fractions 20 -25 were pooled and taken to amino acid sequencing, which revealed a pentapeptide corresponding to the NH 2 terminus of decorin. potential N-glycosylation sites. Hence, the KS attachment sites were Asn-179 and Asn-230. DISCUSSION In this study, we find that the predominant isoform of decorin in chicken cartilage differs from its mammalian counterpart in that it contains two CS/DS chain substituents. The two CS/DS attachment sites are located close to each other near the NH 2 terminus of the core protein, i.e. at the residues Ser-4 and Ser-16. In addition to the biglycanated form, DCN II, we also find a monoglycanated form, DCN I, with the characteristics published previously for chick corneal decorin (37,38,42).
The functions of the small interstitial proteoglycans are incompletely understood. However, the broad tissue distribution of decorin and the high degree of structural conservation of the decorin protein sequence indicate that the function of this proteoglycan is of general importance. It is assumed that decorin influences matrix assembly and supramolecular organization through interactions with other macromolecules. A number of studies have concerned decorin interactions in vitro. In most cases the binding capacity is provided by the core protein, as for example with collagens I, II, and VI, fibronectin, and transforming growth factor-␤ (see Ref. 3

for references).
There also exist some glycosaminoglycan-mediated decorin interactions, for example with collagen XIV (43) and collagen I (28). Interactions of the latter type could be strongly influenced by the number of CS/DS chains.
A major function of collagen-associated decorin might be to confer negative charges to the fibril surface. In tissues like cartilage, interactions between fibrils and perifibrillar components are necessary to prevent fusion of the fibrils and segregation of the extrafibrillar substance. It is reasonable to assume that these interactions show variation between and within tissues, to meet the need of biomechanical diversity. Since the fibril surfaces are rich in glycosaminoglycans, provided by decorin as well as collagen IX, it is likely that electrostatic forces modulate the interactions at the fibril surface. Therefore, the variable glycosylation of decorin may have a role in tuning of the mechanical properties of cartilage tissue.
The biosynthesis of O-linked glycosaminoglycans follows upon the transfer of xyloside to serine residues in a polypeptide. It has been suggested that the responsible enzyme, xylosyltransferase, recognizes the tetrapeptide Ser-Gly-Xaa-Gly preceded by a few acidic residues (44). In chicken, however, the glycosaminoglycan attachment sites are Gly-Ser in DCN I, in collagen IX (45), and probably in PG-Lb (7). Hence, it has been supposed that avian xylosyltransferase activity differs from that in mammals (37). Now we find that the second CS/DS substitution site in chicken decorin corresponds to a classical acceptor sequence. None of the five other Ser-Gly or Gly-Ser sequences of the core protein receives a glycosaminoglycan substituent. These observations support the notion that xylosyltransferase recognition is not specific for a certain peptide sequence, but may involve conformational determinants (46).
The finding of both mono-and biglycanated forms of decorin raises the question as to whether the substitution with glycosaminoglycan chains is regulated in a tissue-specific manner. If so, biglycanated decorin could be rather restricted in its tissue distribution. It is not specific for cartilage, however, as it is found in chicken tendon as well. 2 In addition, close inspection of the published SDS-PAGE of intact corneal decorin (Fig. 6 in Ref. 38) reveals that both DCN II and DCN I may be present in cornea. However, in an immunoblot of chick embryonic skeletal muscle proteoglycans, only monoglycanated decorin was detected (47).
Keratan sulfate was detected as a constituent of decorin from adult chicken cornea, but it was not found in the embryonic proteoglycan (42). Our present data show that such KS substituents are not specific for corneal decorin, but occur also in both DCN I and DCN II from adult chicken cartilage. Their functional significance is unclear, but they may play a role in morphogenesis of corneal fibrils as a prerequisite of the transparency of that tissue (42).
The mammalian decorin forms that have been studied were never biglycanated. However, biglycan exists in mammals as a separate gene product with a core protein that is extensively homologous to that of decorin. Biglycan and decorin show differences in expression and tissue distribution (3). The biglycan counterpart in avian species has not yet been identified, and its existence as a distinct protein is uncertain. On the basis of our present data, it is tempting to speculate that the decorin with two glycosaminoglycan chains represents the chicken equivalent of biglycan. This would imply that substitution with glycosaminoglycans occurs via strictly regulated processes that are important in determining the function of the molecules.