Decorin Is a Zn2+ Metalloprotein*

Decorin is ubiquitously distributed in the extracellular matrix of mammals and a member of the proteoglycan family characterized by a core protein dominated by leucine-rich repeat motifs. We show here that decorin extracted from bovine tissues under denaturing conditions or produced in recombinant “native” form by cultured mammalian cells has a high affinity for Zn2+as demonstrated by equilibrium dialyses. The Zn2+-binding sites are localized to the N-terminal domain of the core protein that contains 4 Cys residues in a spacing reminiscent of a zinc finger. A recombinant 41-amino acid long peptide representing the N-terminal domain of decorin has full Zn2+ binding activity and binds two Zn2+ ions with an average K D of 3 × 10−7 m. Binding of Zn2+to this peptide results in a change in secondary structure as shown by circular dichroism spectroscopy. Biglycan, a proteoglycan that is structurally closely related to decorin contains a similar high affinity Zn2+-binding segment, whereas the structurally more distantly related proteoglycans, epiphycan and osteoglycin, do not bind Zn2+ with high affinity.

Decorin, a small chondroitin/dermatan sulfate proteoglycan, is found in the extracellular matrix of a variety of tissues such as skin (1)(2)(3), cartilage (4,5), and bone (6,7). This proteoglycan is composed of a 40-kDa core protein and one glycosaminoglycan chain attached to a serine residue in the N-terminal part of the protein. The decorin core protein is dominated by a central region composed of 10 leucine-rich repeat units. Each unit contains 21-26 amino acid residues and is proposed to adopt a characteristic ␣-helix/␤-sheet folding pattern (8,9). The C-and N-terminal regions of the core protein are believed to form globular structures stabilized by disulfide bonds between sets of cysteine residues. Several proteoglycans have core proteins of similar size and structural organization. These related molecules are considered to form a family called small leucine-rich proteoglycans (SLRP) 1 (9,10). The family of SLRPs include decorin, biglycan, and epiphycan, all of which contain chondroitin/dermatan sulfate chains attached to the N-terminal domain of the core protein and fibromodulin, lumican, keratocan, PRELP, and osteoglycin which often have keratan sulfate linked to asparagine residues in the central region of the core protein. The extracellular matrix glycoprotein chondroadherin has a structural organization similar to the core proteins of the SLRPs but has not been shown to be substituted with glycosaminoglycan chains (11). The biological importance of the different SLRPs is unclear. In vitro binding studies have shown that decorin, biglycan, and fibromodulin can interact with several types of collagen (12)(13)(14)(15)(16) and different SLRPs are believed to be important regulators of collagen fibrillogenesis. In support of this hypothesis, a decorin-deficient mouse was found to have fragile skin with an abnormal organization of collagen fibers (17). The phenotype appears to be largely restricted to the skin, perhaps suggesting that other SLRPs have similar functions and may fulfill this role in other collagenous tissues. In fact, a lumican-deficient mouse also exhibited abnormal collagen fibers both in the skin and cornea (18).
Decorin may also affect the production of other extracellular matrix components by regulating the activity of transforming growth factor-␤ (19,20). Additionally, decorin can modulate the interactions of matrix molecules such as fibronectin with cells (21)(22)(23). These observations suggest that decorin and perhaps other SLRPs regulate at several levels the production and assembly of the extracellular matrix and hence the remodeling of connective tissue.
Zinc, a divalent cation, is one of the essential trace elements for eukaryotic organisms. Zinc ions play a key role in biological processes by being directly involved in enzyme catalysis or by binding to specific sites in a protein to stabilize the conformations which are of importance to the function of the protein (24,25). In the extracellular matrix, zinc is required for the activity of matrix metalloproteases which are responsible for the degradation of structural extracellular matrix components (26). Breakdown and remodeling of the ECM occur in normal embryo development, wound healing, and many pathological processes such as cancer, arthritis, and osteoporosis (26,27). Structural extracellular matrix molecules such as laminin (29), link protein (30), nidogen (31), and COMP (32) have been shown to bind zinc ions. The Zn 2ϩ -binding sites have been located to subdomains of some of these proteins (31,33).
In a previous study, biglycan was shown to self-associate in the presence of zinc ions and shown to bind Zn 2ϩ , Ni 2ϩ , and Cu 2ϩ in metal chelate affinity chromatography studies (34). The binding affinity and binding domain, however, have not been reported. In this study, we demonstrate that decorin is a metalloprotein and binds two zinc ions per core protein with an average K D of ϳ1 M. The Zn 2ϩ -binding domain is mapped to the N-terminal region in the decorin core protein and a short (41 amino acid long) peptide is shown to have full Zn 2ϩ binding activity. Furthermore, the Zn 2ϩ binding activity of the corresponding segments in the structurally related core proteins of biglycan, epiphycan, and osteoglycin are analyzed.

EXPERIMENTAL PROCEDURES
Preparation of Decorin Proteoglycans-Decorin was extracted and purified from bovine skin under denaturing conditions as described previously (3). Intact recombinant decorin was produced in HT1080 cells using a vaccinia virus-based expression system. The recombinant virus construct contains a segment encoding a N-terminal polyhistidine tag, a Factor Xa cleavage site, and the mature human decorin core protein starting with an Asp residue found at position 31 of the fulllength human sequence (35). The polyhistidine tag in the recombinant protein allows for its efficient purification by Ni 2ϩ chelating chromatography. The resulting decorin preparation contains a mixture of proteoglycan and core protein forms. The details of this system have been described earlier (35,36). For the studies described here, the polyhistidine tag was removed from the recombinant decorin by digestion with Factor Xa (Pierce, Rockford, IL) followed by reapplication of the digestion mixture to a Ni 2ϩ charged column. The final product of cleaved recombinant decorin was collected in the "flow-through" fractions whereas the polyhistidine tag and uncleaved protein were retained on the column.
Recombinant Decorin Core Protein Fragments-A series of recombinant decorin core protein fragments were produced in Escherichia coli. A mouse embryo cDNA library (CLONTECH, Palo Alto, CA) was used as a template along with appropriate primers (Table I) to PCR amplify the decorin cDNA segments. Conditions for PCR were 94°C for 7 min, followed by 35 cycles involving 94°C for 1 min, 50°C for 2 min, and 72°C for 3 min. The individual segments were purified, cleaved with appropriate restriction enzymes, and ligated into the pMAL-p2 expression vector (New England Biolabs, Beverly, MA) which contain a segment encoding the maltose-binding protein (MBP) at the 5Ј end of the expression cassette. The cDNA fragments (MD and MD3) generated with the DCNR2 primer which contains a EcoRI site at 3Ј end were first subcloned into the pBluscript SKϮ vector (Stratagene, La Jolla, CA) through BamHI and EcoRI sites. Subsequently, the MD/pBluscript and MD3/pBluscript constructs were digested with BamHI and HindIII and then ligated into the pMAL-p2 expression vector. To produce MBP as a control protein without any segment of decorin fused to it, a modified pMAL-p2 vector containing a stop codon and a short polylinker was constructed. The complementary oligonucleotides, MBPF (5Ј-gatcctgatctagagcatgcctgca-3Ј) and MBPR (5Ј-ggcatgctctagatcag-3Ј) were obtained and allowed to form a double-stranded oligonucleotide containing a cohesive BamHI site at the 5Ј-end followed by a stop codon, XbaI, SphI, and PstI sites. The original pMAL-p2 vector digested with BamHI and PstI was then ligated with that double-stranded oligonucleotide. The newly introduced SphI site which is not present in the original pMAL-p2 vector was used in a restriction enzyme digestion screen to isolate clones carrying the engineered plasmid. The sequences of all decorin/pMAL-p2 constructs isolated from selected clones were confirmed by DNA sequencing (Department of Veterinary Pathobiology, Texas A&M University, College Station, TX).
Expression and Purification of Recombinant Decorin Core Protein Subdomains-Previously, a method developed for the production of full-length bovine decorin core protein using a pMAL expression system has been reported (37). This procedure involved solubilization and refolding of protein recovered from inclusion bodies. In our case, we obtained significant amounts of soluble recombinant protein from all constructs and we did not attempt to solubilize protein from inclusion bodies.
The expression vector constructs were used to transform E. coli strain TB1. The procedure adopted for the expression and purification of the MBP-decorin fusion protein was based on the manual provided by New England Biolabs. In brief, a flask containing 950 ml of Lennox LB (Sigma) was inoculated with 50 ml of a TB1/MBP-decorin overnight culture and grown on a platform shaker for 3ϳ4 h at 37°C when the culture reached an OD 600 of 0.6 -0.8. Expression of fusion protein was induced by adding isopropyl-1-thio-␤-galactopyranoside (Life Technologies, Inc., Gaithersburg, MD) to the culture at a final concentration of 0.2 mM and incubation was continued for another 3 h. The bacterial cells were harvested by centrifugation at 4500 rpm for 20 min. The cell pellet was resuspended in Buffer A (20 mM Tris, 200 mM NaCl, pH 8.0) to a final volume of 10 ml/liter of culture and frozen at Ϫ80°C for a minimum of 18 h. To purify the fusion proteins, the cells were thawed and lysed using a French press. The lysate was cleared by centrifugation at 40,000 rpm for 20 min and the supernatant was applied to a 5-ml amylose affinity column (New England Biolabs, Beverly, MA) equilibrated with 25 ml of Buffer A. The column was washed with 50 ml of  a Asterisk, stop codon. b,c The nucleotide assigned number and amino acid numbers were based on the mouse decorin (PGS2) sequence search from data base with accession number P28654, and mouse bigycan (PGI), accession number X53928.
Buffer A and the MBP-decorin fusion protein was eluted by Buffer A containing 10 mM maltose. The purity of the fusion protein was Ͼ70% as judged by SDS-PAGE and the yield was 5-10 mg/liter of culture. To improve the purity of fusion proteins, an additional purification step was required. Ion exchange chromatography was used for further purifying MBP-MD2 and MBP-MD3 protein. The proteins were dialyzed into a buffer composed of 25 mM NaCl, 20 mM Tris, pH 8.0, and applied to a column of Q-Sepharose (Pharmacia Biotech, Piscataway, NJ) equilibrated with the same buffer. The protein was subsequently eluted with a NaCl gradient (25-500 mM) in 20 mM Tris, pH 8.0. Metal ion chelating chromatography on iminodiacetic acid immobilized Sepharose 6B resin (Sigma) charged with Zn 2ϩ was used to further purify MBP-MD, MBP-MD1, and MBP-MD4. The procedure used for charging the column with Zn 2ϩ ions is described in the following section. The purified recombinant proteins were analyzed on SDS-PAGE and Western blot using two polyclonal antibodies that were raised against synthetic peptides corresponding to amino acid residues 31-47 and 309 -326 of the mouse decorin core proteins, respectively. 2 The protein concentrations were determined based on the absorbance at A 280 and the calculated molar extinction coefficient of the different protein constructs (38).
Production and Purification of a N-terminal Decorin Peptide (MD4)-To produce a large quantity of MD4 peptide, we used a pGEX-2T expression vector (Pharmacia Biotech Inc., Piscataway, NJ). This MD4/pGEX-2T construct encodes glutathione S-transferase (GST) followed a thrombin cleavage site and the MD4 fragment. The cDNA fragment encoding MD4 was obtained by PCR using the mouse embryo cDNA library and primers DCNF9 and DCNR19 as described above (Table IA). The resulting PCR fragment was cleaved by BamHI and PstI, purified, and subcloned into pBluscript. The isolated MD4/pBluscript plasmid was digested with BamHI and EcoRI and ligated into the pGEX-2T expression vector. The recombinant MD4/pGEX-2T plasmid was then transformed into E. coli strain BL-21 (Stratagene, La Jolla, CA). An isolated clone containing the MD4/pGEX-2T plasmid was used after the plasmid had been analyzed by restriction enzyme digestion and sequencing. To express the GST-MD4 fusion protein, flasks containing 950 ml of Lennox LB were inoculated with 15 ml of a BL21/ GST-MD4 overnight culture and incubated for 3-4 h when the culture reached an A 600 of 0.4 -0.5. Recombinant protein expression was induced by adding isopropyl-1-thio-␤-galactopyranoside to the culture at a final concentration of 0.2 mM. Bacterial cells were harvested 3 h later by centrifugation at 4500 rpm for 20 min. The cell pellet was resuspended in phosphate-buffered saline (8.4 mM Na 2 HPO 4 , 1.9 mM NaH 2 PO 4 , 150 mM NaCl, pH 7.4), supplemented with 5 mM EDTA and adjusting the pH 7.5 to a final volume of 10 ml/liter of culture and frozen at Ϫ80°C for a minimum of 18 h. The cells were thawed and lysed using a French press. One ml of 10% Triton X-100 (Sigma) was added to the cell lysate and mixed until homogeneous. The cell homogenate was centrifuged and filtered through a 0.45-m membrane to remove cell debris.
The GST-MD4 fusion protein present in the filtered supernatant was purified by affinity chromatography on a 10-ml column of glutathioneagarose (Sigma) equilibrated with 50 ml of Buffer B (phosphate-buffered saline with 1 mM EDTA, pH 7.5). The column was washed with 75 ml of Buffer B and the GST-MD4 protein was eluted from the column with 20 ml of Buffer C (50 mM Tris, 10 mM glutathione, pH 7.5). The purity of GST-MD4 in the eluate was Ͼ90%, as judged by SDS-PAGE and the yield was 25-30 mg/liter of culture.
To isolate the MD4 peptide, the fusion protein was cleaved with thrombin. The pH and the concentration of NaCl of the eluant containing GST-MD4 was adjusted to pH 8.3, 0.15 M NaCl using 1 M NaOH and 3 M NaCl stock solutions. Bovine thrombin (Sigma) was added to give a 1/40 (w/w) enzyme/substrate ratio. The digestion was allowed to proceed overnight at room temperature. Subsequently, ␤-mercaptoethanol was added to the incubation mixture to a final concentration of 1% (v/v) and the solution was adjusted to pH 9.5 by addition of 1 M NaOH. The solution was incubated at 37°C for 30 min and filtered through a 0.45-m membrane to remove any particulate material. The MD4 peptide was purified on a Waters 25 ϫ 200-mm RCM semi-preparative C18 HPLC column and eluted with a gradient of 28 to 36% solvent B (95% acetonitrile, 5% H 2 O, 0.1% trifluoroacetic acid) in solvent A (95% H 2 O, 5% acetonitrile, 0.1% trifluoroacetic acid) over 10 min at a flow rate of 25 ml min Ϫ1 .
The purity of the peptide was monitored by running a three-layer Tricine-SDS-PAGE gel consisting of a 15% acrylamide slab, 10% acryl-amide spacer, and a 3.5% acrylamide stacker (39). The identity of the peptide was confirmed by matrix-assisted laser desorption ionizationmass spectroscopy (The Center for Analytical Chemistry, University of Texas in Houston, Houston, TX).
Biglycan Preparations-Biglycan isolated from bovine articular cartilage was extracted and purified under denaturing conditions as described previously (3). Recombinant biglycan containing a N-terminal polyhistidine tag was produced in HT1080 cells using the vaccinia virus expression system as described (36). The polyhistidine tag was removed and the recombinant proteoglycan was re-isolated as described above. The preparation of recombinant biglycan contains a mixture of proteoglycan and core protein forms (36). The N-terminal segment of the biglycan core protein was produced as a recombinant MBP fusion protein in E. coli. The encoded fusion protein MBP-MB-N is composed of the maltose-binding protein followed by a biglycan peptide, corresponding to amino acid residues 38 to 77. The primers BGNF11 and BGNR13 (Table I) and the mouse embryo cDNA library were used to PCR amplify an appropriate cDNA fragment. The resulting PCR product was cleaved, purified, and ligated into the pMAL-p2 vector. Protein expression and purification protocols were the same as those described above in the preparation of MBP-decorin fusion proteins.
Dialysis of the Protein-In preparation for Zn 2ϩ binding experiments, all proteins were dialyzed against Buffer D (20 mM Tris-HCl, 150 mM NaCl, pH 7.0) supplemented with 5 mM EDTA (Sigma) to remove possibly contaminating divalent metal ions and subsequently dialyzed against Buffer D (without EDTA) with three changes. Buffer D was shown to contain less than 5 ϫ 10 Ϫ8 M Zn 2ϩ ions as analyzed by 4-(2-pyridylazo)resorcinol (PAR) (Sigma) complex formation (40,41).
Zn 2ϩ -chelating Affinity Column Chromatography-A resin composed of iminodiacetic acid-immobilized Sepharose 6B (Sigma) was charged with 10 volumes of 2 mg/ml ZnCl 2 (Sigma) in deionized water. The resin was then washed with 5 volumes of Buffer E (20 mM Tris-HCl, 150 mM NaCl, pH 8). The charged Sepharose 6B was packed on top of an equal amount of uncharged iminodiacetic acid-Sepharose 6B that has been pre-equilibrated in Buffer E. The purpose of having the uncharged resin on the bottom of the column is to trap zinc ions released from the charged resin during the experiment. The dialyzed test sample was applied to the column. The column was washed with excess amount of Buffer E and the bound protein was eluted by Buffer F in which the pH of Buffer E was titered to 4.0 by addition of glacial acetic acid.
Zn 2ϩ Equilibrium Dialysis-The equilibrium dialysis experiments were carried out in a double acrylic microdialysis module (Hoffer, San Francisco, CA). A dialysis membrane was assembled between the two modules to separate each of the eight chambers into two compartments. The molecular weight cut-off of the dialysis membrane was chosen depending upon the molecular size of the test protein. The 12,000 -14,000 Da cut-off membrane (Hoffer) was used for the proteoglycan preparations and the MBP fusion proteins. The 2,000 Da cut-off membrane (Spectra/Pro, Spectrum Medical Industries, Inc., Houston, TX) was used for the MD4 peptide. Aliquots of 150 l of test protein in Buffer D were added to the inner compartments. The same volume of Buffer D containing increasing concentrations of ZnCl 2 (0, 2, 4, 8, 16, 24, 32, and 40 M) was added to the outer compartments. The assembled modules were fitted on a rotating axis and incubated at 4°C for at least 40 h to reach equilibrium.
After incubation, the concentration of Zn 2ϩ in the outer compartments was determined using the metallochromic indicator PAR (41). An aliquot of 120 l from each outer compartment was added to glass tubes containing 2.4 l of a 5 mM PAR. The components were mixed and incubated for 30 min at room temperature to allow the red PAR⅐Zn 2ϩ complex to form. The absorbance at A 500 from each sample was determined using a DU-70 spectrophotometer (Beckman, Fullerton, CA). A standard curve of ZnCl 2 (0 -40 M) in Buffer D was generated which showed a linear relationship by plotting A 500 versus the increasing concentration of Zn 2ϩ . The concentration of free Zn 2ϩ in solution from each outer compartment was determined by inserting the observed absorbance into the standard curve equation.
The concentration of Zn 2ϩ bound to protein was determined using Equation Total protein is the concentration of protein used in the equilibrium dialysis experiment, n is the number of Zn 2ϩ molecules binding to each protein molecule, and K D is the observed dissociation constant for the Zn 2ϩ -protein complex. The reported K D Ϯ S.E. and Zn 2ϩ binding curves represent the average of at least three independent experiments. Circular Dichroism Studies-CD spectra were collected on a Jasco J-720 spectropolarimeter calibrated with d-10-camphorsulfonic acid, using a round, 0.5-mm quartz cell. The CD data were collected at a scan speed of 20 nm/min at 0.5-nm intervals, a time constant of 1 s, and a band width of 1 nm, averaging 4 scans for the final data. All spectra were corrected for buffer contributions and are presented in units of mean residue ellipticity, [⌰] MRW , calculated with Equation 3, where ⌰ is the ellipticity in millidegrees, MRW is the mean residue weight of MD4, c is the concentration of MD4 in mg/ml, and d is the cell path length in centimeters. The samples were prepared by first reducing the MD4 peptide in 100 mM EDTA and 1% ␤-mercaptoethanol at 100°C for 1 min. The reduced sample was cooled to room temperature and the buffer exchanged to 50 mM Tricine, 0.1 mM EDTA at pH 6.6 on a PD-10 de-salting column (Pharmacia, Piscataway, NJ). The concentration of fractions from the PD-10 column were calibrated at 280 nm using ⑀ 280 ϭ 2600. Samples for CD spectroscopy were prepared by diluting the stock MD4 samples with 50 mM Tricine and 0.1 mM EDTA for the sample without zinc salt or adding an aliquot of a 100 mM ZnCl 2 or ZnSO 4 stock to the dilution buffer before the addition of peptide. All samples were prepared to a final volume of 200 l at concentrations of 200 g/ml MD4 peptide and 2 mM zinc salt where indicated.

RESULTS
The Zn 2ϩ Binding Activity of Decorin Is Located to the Core Protein-We recently reported that a Zn 2ϩ charged column of iminodiacetate-Sepharose could be used to separate decorin and epiphycan isolated from fetal bovine epiphyseal cartilage (43). Decorin was retained on the column suggesting that the proteoglycan can bind Zn 2ϩ . Equilibrium dialysis has now been used to demonstrate a saturable binding of Zn 2ϩ to decorin (Fig. 1). The proteoglycan isolated from bovine skin (BDCN) and recombinant human decorin produced in mammalian cells (vvHDCN) bound a maximum approaching two Zn 2ϩ ions per decorin molecule with an average K D of 1.0 Ϯ 0.3 M and 3.9 Ϯ 1.8 M, respectively (Fig. 1, A and B).
The recombinant mammalian decorin appears as a mixture of molecular species including proteoglycans and core proteins without glycosaminoglycan chains. The observation that the recombinant proteoglycan/core protein mixture could bind similar amounts of Zn 2ϩ suggests that the primary Zn 2ϩ -binding sites in decorin are located in the core protein. This hypothesis was further examined by analyzing the Zn 2ϩ binding properties of prokaryotic recombinants. Recombinant mouse decorin (MD) core protein produced in E. coli as a fusion protein where the core protein is linked at its N terminus to the MBP also bound Zn 2ϩ when examined by equilibrium dialyses whereas no saturable binding of Zn 2ϩ to MBP alone could be detected (Fig. 1, C and D). A maximum of 1.8 Zn 2ϩ ions could bind per molecule of MBP-MD fusion protein with a K D of 3 M under these conditions (Fig. 1C). Isolated dermatan sulfate or chondroitin sulfate polysaccharides were not retained on a Zn 2ϩ charged iminodiacetate-Sepharose (data not shown), suggesting that the core protein alone is responsible for Zn 2ϩ binding activity of decorin.
The Zn 2ϩ Binding Activity Is Located to the N-terminal Domain of the Decorin Core Protein-To further locate the domain responsible for the Zn 2ϩ binding activity in the core protein, we analyzed recombinant segments of the mouse decorin core protein produced in E. coli as fusion proteins linked to the C terminus of MBP. The different constructs made are shown in Fig. 2. All the fusion proteins which were at least partially soluble when produced as described under "Experimental Procedures," were purified by a combination of affinity chromatography on amylose-Sepharose and ion exchange chromatography. In the presence of reducing agent, a major component in the different preparations of core protein fragments migrated as expected when analyzed by SDS-PAGE (data not shown).
Chromatography of the different fusion proteins on a Zn 2ϩ charged matrix demonstrated that MBP-MD, MBP-MD1, and MBP-MD4 bound Zn 2ϩ whereas MBP-MD2 and MBP-MD3 did not bind the metal ions (summarized in Fig. 2). The Zn 2ϩbinding recombinant proteins all contained the N-terminal segment of the decorin core protein whereas this segment was not FIG. 1. Zn 2؉ binding to decorin; equilibrium dialysis. Bovine decorin extracted from fetal skin (BDCN), human decorin produced as a recombinant using a vaccinia virus vector (vvHDCN), mouse decorin core protein expressed as a prokaryotic recombinant as a fusion protein with maltose-binding proteins (MBP-MD), or MBP were examined for their Zn 2ϩ binding activity using equilibrium dialysis. For further details see "Experimental Procedures." The K D and number of binding sites (n) Ϯ S.E. represents the average of three experiments. present in the non-binders. Equilibrium dialyses showed that MBP-MD4 could bind Zn 2ϩ in a process that exhibits saturation kinetics approaching a maximum of two Zn 2ϩ ions bound per protein molecule (Fig. 3A). The average K D for these interactions was 2.4 Ϯ 0.6 M. Binding of Zn 2ϩ to MBP-MD2 and MBP-MD3, respectively, could not be demonstrated using equilibrium dialyses (data not shown). These results suggest that the 41-amino acid long N terminus of decorin retain full zinc binding activity.
Characterization of Zn 2ϩ Binding to the MD4 Peptide-We decided to examine the Zn 2ϩ binding behavior of the MD4 peptide in the absence of a fusion protein partner. A new MD4 expressing plasmid using the pGEX-2T vector was constructed. The resulting recombinant fusion protein is composed of GST followed by a linker containing a thrombin cleavage site and MD4. This fusion protein was expressed, purified, and cleaved. The released MD4 peptide, which contains 4 extra amino acid residues (Gly-Ser-Asn-Gly) at the N terminus followed by residue 31-71 of mouse decorin, was purified by RP-HPLC. The expected size of the peptide was 4878.56 Da as calculated from the amino acid sequence and the molecular mass of the isolated peptide was 4878.37 Da as determined by matrix-assisted laser desorption ionization-mass spectroscopy.
The purified MD4 peptide bound to a Zn 2ϩ charged iminodiacetate-Sepharose column (data not shown) and equilibrium dialyses experiments showed that the MD4 peptide bound Zn 2ϩ ions in a concentration dependent manner that approached a maximum of two Zn 2ϩ ions bound per peptide molecule (Fig.  3B). Analyses of the binding data assuming the presence of two independent Zn 2ϩ -binding sites in the MD4 peptide suggest an average K D of 0.28 Ϯ 0.03 M for these sites. Analyses of the binding data in a Hill plot did not reveal any pronounced cooperativity between the two binding sites.
Zn 2ϩ Binding Induces a Conformational Change in the MD4 Peptide-Analyses by circular dichroism spectroscopy (Fig. 4) showed that the MD4 peptide has a different conformation in the presence of Zn 2ϩ ions. In the presence of EDTA (but absence of Zn 2ϩ ion), the spectrums exhibit a minimum of 200 nm. When Zn 2ϩ ions are added either in the form of ZnSO 4 or ZnCl 2 , the minima has a reduced amplitude and is shifted to a slightly lower wavelength. In addition, a shoulder appears at 215-220 mm. These observations demonstrate a change in secondary structure in the peptide on Zn 2ϩ binding.
The Zn 2ϩ Binding Activity of Biglycan-The amino acid se-quence of the Zn 2ϩ binding MD4 decorin peptide is partially conserved in biglycan but less conserved in epiphycan and osteoglycin. In previous studies, tissue extracted biglycan from bovine articular cartilage have been shown to form multimers in the presence of Zn 2ϩ (34). We therefore examined the Zn 2ϩ binding activity of different forms of biglycan. Binding to a Zn 2ϩ charged column was demonstrated for biglycan isolated under denaturing conditions from bovine articular cartilage (BBGN) or produced as a recombinant proteoglycan in HT1080  (Fig. 5, A and B). The N-terminal segment of mouse biglycan corresponding to amino acid residues 38 -77 was expressed in E. coli as a recombinant MBP fusion protein (MBP-MB-N) and shown to bind to a Zn 2ϩ charged column of iminodiacetate-Sepharose (data not shown). Analyses by equilibrium dialyses showed that MBP-MB-N could bind a maximum of two molecules of Zn 2ϩ with an average K D of 2.7 Ϯ 1.6 M (Fig. 5C). These results suggest that biglycan is also a Zn 2ϩ binding molecule and that the N terminus of its core protein has zinc binding activity.
Segments of mouse epiphycan and osteoglycin core proteins containing the N-terminal domains were expressed as fusion proteins (MBP-ME-N and MBP-MO-N, respectively), purified, and analyzed for Zn 2ϩ binding activity. Both fusion proteins bound to the Zn 2ϩ charged column (data not shown). In equilibrium dialysis experiments, however, the binding affinity of Zn 2ϩ to MBP-ME-N and MBP-MO-N proteins was too weak (K D Ͼ10 Ϫ5 M) to be detected. These results suggest that residues specifically found in decorin and biglycan but not in epiphycan or osteoglycin are required for forming the high affinity Zn 2ϩ -binding sites.

DISCUSSION
In this article, we report that decorin and biglycan are Zn 2ϩ metalloproteins. These macromolecules are capable of binding Zn 2ϩ regardless of whether they are purified under denaturing conditions from tissues or are produced and isolated under nondenaturing conditions from cultured cells infected by recombinant viruses. Furthermore, the ability to bind Zn 2ϩ is not restricted to decorin from one particular species since in this study, we demonstrate a binding of Zn 2ϩ to decorin derived from human, mouse, or bovine tissue. Decorin and biglycan are members of a growing family of ECM proteins that exhibit high affinity for Zn 2ϩ . In addition to matrix metalloproteases (28), this family includes, laminin (29), link protein (30), nidogen (31), and COMP (32).
The Zn 2ϩ binding activity of decorin is localized to a segment in the N-terminal part of the core protein. A GAG attachment site is also located in this region and involves a Ser residue at position 34 in the mouse decorin sequence. The carbohydrate components of the proteoglycans, however, are not directly involved in Zn 2ϩ binding. In fact, a recombinant peptide corresponding to the decorin N-terminal domain and produced in E. coli has full Zn 2ϩ binding activity. The amino acid sequence of this peptide which contains four Cys residues is reminiscent of a zinc finger but the spacing of the Cys residues is different from previously reported Zn 2ϩ -binding motifs (24,25). This apparently novel Zn 2ϩ binding sequence must adopt a conformation that allows the binding of two Zn 2ϩ ions per peptide. Cys residues appear to be involved in coordinating the Zn 2ϩ ions since in preliminary experiments we have shown that reduction and alkylation of the MBP-MD4 protein resulted in loss of Zn 2ϩ binding activity. Earlier analyses of tryptic peptides obtained from bovine biglycan showed that the first and the fourth Cys residues in the N-terminal domain are linked through a disulfide bond (44). This observation leaves the second and third Cys residues as potential Zn 2ϩ coordinators. If in fact the Cys residues are involved in Zn 2ϩ binding, this could help explain the different affinities we see for Zn 2ϩ among the different forms of decorin tested. The isolated MD4 peptide which was isolated as a homogenous form by RP-HPLC has the highest apparent affinity for Zn 2ϩ .
The Zn 2ϩ binding MBP fusion proteins all occurred as mixtures of several molecular forms where some Cys residues appear to be engaged in forming disulfied linked multimers. The measured K D values for Zn 2ϩ binding to these proteins were higher compared with that recorded for the isolated MD4 peptide. Perhaps Cys residues engaged in coordinating Zn 2ϩ in the MD4 peptides in the multimers are involved in disulfide linkages. Alternatively, this observation may suggest that structures outside the actual binding site may influence the Zn 2ϩ binding activity of this domain. Thus the isolated sequence present in the peptide appears to have a higher affinity for Zn 2ϩ than when the sequence is part of a larger structure as in an intact core protein or a proteoglycan. Decorin contains in addition to a traditional N-terminal signal, a 13-amino acid long propeptide, which also often has been removed from the proteoglycan isolated from tissues. The decorin MD4 peptide and the proteoglycan forms here shown to bind Zn 2ϩ with a high affinity do not include the propeptide sequence. It is unclear if the presence of a propeptide may affect Zn 2ϩ binding to decorin or biglycan.
Recombinant N-terminal segments of epiphycan and osteoglycin core protein made as MBP fusions, which contain sequences similar to the Zn 2ϩ binding sequences present in decorin and biglycan, are also retained on a Zn 2ϩ -charged iminodiacetate-Sepharose. However, the affinity of the MBPfusion proteins for Zn 2ϩ is too low to measure reliable binding constants by equilibrium dialyses. In an earlier study, we found that Zn 2ϩ chelating chromatography could be used to fractionate tissue-extracted epiphycan (which did not bind to the column) from decorin (which bound to the column). This result is in contrast to the chromatography data obtained in the current study and could be explained if the full-length epiphycan proteoglycan has a lower affinity for Zn 2ϩ compared with the recombinant N-terminal peptide. Such a pattern was established for decorin. In the N-terminal segments of epiphycan and osteoglycin, the Cys residues are spaced differently than in the Zn 2ϩ -binding peptides of decorin and biglycan; there are two compared with 3 amino acid residues between the first and second Cys in epiphycan osteoglycin, and decorin/biglycan, respectively. It is possible but not yet demonstrated that the spacing of Cys residues determines the Zn 2ϩ binding activity of these segments.
The concentration of Zn 2ϩ in body fluids such as blood plasma is approximately 15 M, which suggests that both decorin and biglycan in vivo occur in complex with Zn 2ϩ . The significance of this complex is unclear. Decorin and biglycan which are abundant molecules in the tissues could serve as Zn 2ϩ storage pools where the metal ions could be released to proteins which have a higher affinity for and need Zn 2ϩ ions for their activity. In addition, Zn 2ϩ may stabilize a conformation in the proteoglycan core proteins which are important for their functions. In fact, analyses of the MD4 peptides using circular dichroism spectroscopy demonstrate that the conformation of the peptide is altered in the presence of Zn 2ϩ . This observation is consistent with the observation of Liu et al. (34) who found that biglycan had a tendency to aggregate in the presence of Zn 2ϩ , which could be caused by a conformational change in the protein induced by Zn 2ϩ binding. Future studies of the biology of decorin and biglycan must take into account the fact that these molecules are Zn 2ϩ metalloproteins.