Recombinant Human Aggrecan G1-G2 Exhibits Native Binding Properties and Substrate Specificity for Matrix Metalloproteinases and Aggrecanase*

A recombinant human aggrecan G1-G2 fragment comprising amino acids Val1-Arg656 has been expressed in Sf21 cells using a baculovirus expression system. The recombinant G1-G2 (rG1-G2) was purified to homogeneity by hyaluronan-Sepharose affinity chromatography followed by high performance liquid chromatography gel filtration, and gave a single band of M r 90,000–95,000 by silver stain or immunoblotting with monoclonal antibody 1-C-6. The expressed G1-G2 bound to both hyaluronan and link protein indicating that the immunoglobulin-fold motif and proteoglycan tandem repeat loops of the G1 domain were correctly folded. Further analysis of secondary structure by rotary shadowing electron microscopy confirmed a double globe appearance, but revealed that the rG1-G2 was more compact than its native counterpart. The size of rG1-G2 by SDS-polyacrylamide gel electorphoresis was unchanged following digestion with keratanase and keratanase II and reduced by only 2–5 kDa following digestion with either O-glycosidase or N-glycosidase F. Recombinant G1-G2 was digested with purified matrix metalloproteinases (MMP), isolated aggrecanase, purified atrolysin C, or proteinases present in conditioned medium from cartilage explant cultures, and the products analyzed on SDS gels by silver stain and immunoblotting. Neoepitope antibodies recognizing the N-terminal F342FGVG or C-terminal DIPEN341 sequences were used to confirm MMP cleavage at the Asn341 ↓ Phe bond, while neoepitope antibodies recognizing the N-terminal A374RGSV or C-terminal ITEGE373 sequences were used to confirm aggrecanase cleavage at the Glu373 ↓ Ala bond. Cleavage at the authentic MMP and aggrecanase sites revealed that these proteinases have the same specificity for rG1-G2 as for native aggrecan. Incubation of rG1-G2 with conditioned medium from porcine cartilage cultures revealed that active soluble aggrecanase but no active MMPs, was released following stimulation with interleukin-1α or retinoic acid. Atrolysin C, which cleaves native bovine aggrecan at both the aggrecanase and MMP sites, efficiently cleaved rG1-G2 at the aggrecanase site but failed to cleave at the MMP site. In contrast, native glycosylated G1-G2 with or without keratanase treatment was cleaved by atrolysin C at both the aggrecanase and MMP sites. The results suggest that the presence or absence per se of keratan sulfate on native G1-G2 does not affect the activity of atrolysin C toward the two sites.

rG1-G2 at the aggrecanase site but failed to cleave at the MMP site. In contrast, native glycosylated G1-G2 with or without keratanase treatment was cleaved by atrolysin C at both the aggrecanase and MMP sites. The results suggest that the presence or absence per se of keratan sulfate on native G1-G2 does not affect the activity of atrolysin C toward the two sites.
Aggrecan is a chondroitin sulfate and keratan sulfate-bearing proteoglycan. It is present in cartilage as large multimolecular aggregates bound noncovalently to hyaluronan (HA). 1 Aggregate formation is mediated via the proteoglycan tandem repeat loops in the N-terminal G1 domain which bind to decasaccharide units within the polymeric HA. This binding is further stabilized by link protein, which binds to form stable trimeric complexes. The formation of large aggregate structures provides a mechanism for trapping aggrecan in the tissue and preventing its loss by diffusion. A major feature of the pathology of arthritis is the gradual loss of aggrecan from the cartilage matrix. This loss involves proteolysis of the core protein and release into culture medium (1,2) or synovial fluid (3) of large fragments lacking G1 domains that are therefore unable to bind hyaluronan. The interglobular domain (IGD) between G1 and G2 is particularly sensitive to proteolysis although there is catalytic processing of more C-terminal regions of aggrecan as well.
Following cloning and sequencing of the full-length cDNA for human aggrecan (4), sequence analysis of aggrecan degradation products revealed the location of specific aggrecanase cleavage sites within the core protein (5)(6)(7). The Glu 373 2 Ala aggrecanase site within the IGD has been most widely studied since cleavage at this site releases the entire glycosaminoglycan-bearing portion of the molecule and therefore has the greatest biological consequence for weight bearing functions. The N-terminal sequence of the IGD-derived aggrecanase fragment found in cartilage explant medium was ARGSV, commencing at Ala 374 (5-7) (Fig. 1). Aggrecan fragments with the same A 374 RGSV N termini have since been identified in synovial fluids from osteoarthritis patients (8) and patients with inflammatory arthritides and joint injury (9) indicating that aggrecanase is a key enzyme in human aggrecan catabolism.
It is now abundantly clear that both aggrecanase, which is a member of the ADAMTS family (25), and MMPs have a role in human disease (12,13). The recent development of neoepitope antibodies (13, 26 -30) enable aggrecanase and MMP activities to be distinguished from each other and compared. However, the relative contribution of aggrecanase and MMPs to aggrecan turnover has not been quantitated, nor is it clear whether one activity is more predominant in normal catabolism and the other more so in pathology. Most studies have been done with animal tissue but the relative involvement of MMPs and aggrecanase appears to differ between species (31), making interspecies comparisons difficult, and adding little to our understanding of human aggrecanolysis. In this paper we describe production of a human rG1-G2 fragment. The rG1-G2 retains its functional properties of binding to hyaluronan and link protein and exhibits a typical double globe structure. The rG1-G2 is cleaved in vitro by aggrecanase and MMPs at the authentic Glu 373 2 Ala and Asn 341 2 Phe bonds, respectively, and therefore represents a valuable substrate for studying the role of these enzymes in cell-mediated aggrecanolysis. (IL-1␣) was from Genzyme Diagnostics. All-trans-retinoic acid was from ICN Biochemicals. Keratanase II was from Seikagaku, Japan. The Biosep-SEC S4000, 300 ϫ 7.8-mm analytical column was from Phenomenex and the BioSil SEC-400, 600 ϫ 21.5-mm preparative column was from Bio-Rad. The 10-mer synthetic peptides for competition assays, EDFVDIPENF, GEDFVDIPEN, TGEDFVDIPE, YTGEDFVDIP, LPRNITEGEA, PLPRNITEGE, LPLPRNITEG, and ELPLPRNITE, were from AUSPEP, Australia. Hyaluronan-coupled-Sepharose (HA-Sepharose) was kindly provided by Professor T. Hardingham, University of Manchester, United Kingdom. The following enzymes were generously provided by Professor G. Murphy, University of East Anglia, Norwich, UK: recombinant human MMP-1 (32), recombinant human MMP-3 (33), recombinant human MMP-7 (34), MMP-9 purified from human gingival fibroblast conditioned media (35), and recombinant human MMP-10 (34). Recombinant human pro-MMP-13 (36) and recombinant human pro-MMP-8 were generously provided by Dr. V. Knä uper and Professor G. Murphy, University of East Anglia, Norwich, UK. The snake venom hemorrhagic toxin HT-d (atrolysin C) was purified from rattlesnake venom (37) and kindly provided by Professor J. Fox, University of Virginia, Charlottesville, VA. The MMP inhibitor XS309 ([3S-[3R*,2-[2R*,2-(R*,S*)]-hexahydro-2-[2-[2-(hydroxyamino)-1-methyl-2-oxoethyl]-4-methyl-1-oxopentyl]-N-methyl-3-pyridazinecarboxamide) (38) was synthesized at DuPont Pharmaceuticals Co. (Wilmington, DE). All other reagents were of analytical grade.

Materials-
Construction of the Human Aggrecan G1-G2 Plasmid-A cDNA clone pSA005 (4) encoding the signal peptide sequence, the G1 and G2 globular domains, and the IGD of human aggrecan (nucleotides  was modified by polymerase chain reaction to include a polyhistidine tag, a stop codon, and an EcoRI restriction site at the 3Ј end. Primers Aggr1 (5Ј-TCTTCGCCACACGCC-3Ј) and HisAgg (5Ј-CGGAATTCCT-TAGTGATGATGGTGATGATGTCGGAAGCAGAAGGC-3Ј) were used to amplify a 312-bp fragment which was digested with restriction endonucleases SacII and EcoRI to generate two bands, 202 and 110 bp. The 202-bp band containing the stop codon, polyhistidine tag, and EcoRI site was purified and used to replace a 174-bp cassette excised from pSA005. The G1-G2 construct was subcloned into the pBacPAK8 transfer vector between the polyhedrin promoter and polyadenylation signal, generating pBacPAK8-G1-G2, then sequenced using Amplitaq Cycle sequencing.
Production of Recombinant Baculovirus-In the initial experiments, IPLB-Sf21 (Sf21) cells were maintained in culture at 27°C in TNM-FH medium (Grace's insect medium supplemented with 0.33% yeastolate and 0.33% lactalbumin hydrolysate) containing 10% (v/v) fetal calf serum, 50 units/ml penicillin, and 50 g/ml streptomycin. In later experiments, Sf21 cells were cultured in Sf900II serum-free medium. Recombinant virus was produced by co-transfecting Sf21 cells with pBacPAK8-G1-G2 transfer vector and non-viable BacPak6 viral DNA. Homologous sequences within the BacPak6 viral expression vector, and the pBacPAK8 transfer vector allow double recombination to occur. The recombination event transfers the aggrecan G1-G2 cDNA, the polyhedrin promoter, and the polyadenylation sequences to the viral ge- The polyhistidine tail present on rG1-G2, and the pattern of disulfide bonds within the globular domains are shown. The amino acid sequence flanking and bridging the major MMP cleavage site and the aggrecanase cleavage site in the IGD, and the neoepitope sequences generated by proteolytic cleavage are included. The specific antibodies used to detect neoepitopes in this study are boxed. Asterisks denote residues reported to be substituted in other species (59,63). nome, which renders only the recombinant virus viable. Polymerase chain reaction using the Aggr 1 and HisAgg primers to amplify a 312-bp fragment confirmed the identity of the recombinant virus. Plaque assays showed that the recombinant virus stock contained 5 ϫ 10 8 plaque forming units. Subsequently, Sf21 cells (12 ϫ 10 6 /175-cm 2 flask) were infected with 2.5 plaque forming units of recombinant virus per cell for 1 h at room temperature. The medium was replaced and the cells incubated for a further 96 h. rG1-G2 secreted into the medium was detected by silver stain and immunoblotting with monoclonal antibody 1-C-6 (39).
Isolation and Purification of Recombinant Human G1-G2-Recombinant G1-G2 present in the culture medium was purified by a two-step procedure involving affinity chromatography on HA-Sepharose followed by size exclusion chromatography. Up to 60 ml of harvested medium was applied to a 0.6-ml column of HA-Sepharose (with a binding capacity of 2 mg/ml G1 domain) equilibrated in PBS. The sample was allowed to cycle through the column twice at a flow rate of 0.4 ml/min, followed by washing with 15 column volumes of PBS. Proteins that specifically bound to the HA-Sepharose were then eluted with 2 ml of 4 M GdnHCl, 50 mM sodium acetate, pH 5.8. The bulk of the serum proteins were removed by this affinity chromatography step. rG1-G2 eluted from HA-Sepharose was further purified by HPLC size exclusion chromatography using either BioSil SEC-400 or BioSep-SEC S4000 columns eluted in buffer containing 4 M GdnHCl, 50 mM sodium acetate, pH 5.8. Fractions were analyzed by SDS-PAGE, silver stain, and Western blotting. Those fractions containing rG1-G2 were pooled, concentrated by centrifugal filtration, and rechromatographed on the same column to yield a single band of purified protein, as assessed by silver staining.
Rotary Shadowing of rG1-G2-Lyophilized preparations of purified human rG1-G2 and native pig G1-G2 (40) were dissolved in 4 M GdnHCl at a concentration of 100 g/ml, then dialyzed overnight against 0.2 M ammonium bicarbonate, 0.2 mM phenylmethylsulfonyl fluoride at 4°C. Samples were diluted into 70% glycerol and sprayed onto mica for rotary shadowing with platinum, and transmission electron microscopy on a Philips 410LS as described previously (41).
Proteinase Digestions of rG1-G2-Matrix metalloproteinase and atrolysin C digestions were done at 37°C in buffer containing 50 mM Tris-HCl, pH 7.5, 100 mM sodium chloride, and 10 mM calcium chloride. The digests were stopped by adding EDTA and 1,10-phenanthroline to final concentrations of 10 and 2 mM, respectively, or boiling. Aggrecanase was generated in conditioned media from IL-1␤-stimulated bovine nasal cartilage cultures as described (42). rG1-G2 was digested at 37°C with 10 l of aggrecanase-containing conditioned media in 50 mM Tris-HCl, pH 7.5, 100 mM sodium chloride, and 10 mM calcium chloride in the presence of 1 M XS309. XS309, which is a nanomolar inhibitor of multiple MMPs (MMP-1 K i Ͻ 1 nM; MMP-2 K i ϭ 1.4 nM; MMP-3 K i ϭ 3.0 nM; MMP-8 K i Ͻ 1 nM; MMP-9 K i Ͻ 1 nM) but is ineffective in inhibiting aggrecanase (38,42), was used to block potential cleavage by any active MMPs present in the conditioned media. Digestion was quenched with 25 mM EDTA, pH 8.0.
Deglycosylation of rG1-G2-Native and recombinant G1-G2 (10 g/20 l) were digested overnight at 37°C with either 0.01 units of keratanase or 0.001 units of keratanase II in 50 mM Tris acetate, pH 7.5. Keratanase-digested samples were analyzed by SDS-PAGE on Fairbanks gels (43) rather than Laemmli gels (44) since they are better suited to resolving highly glycosylated proteins and the native G1-G2 migrates more true to size (compare Fig. 6a, lane 4, with b, lane 4). For deglycosylation of native G1-G2 prior to digestion with atrolysin C, 20 g of G1-G2 was digested for 4 h at 37°C with 0.02 units of keratanase in 50 mM Tris acetate, pH 7.5, in a total volume of 15 l. For Nglycosidase F treatment, native and recombinant G1-G2 were denatured by boiling in 0.125 M Tris-HCl, pH 6.8, containing 1% SDS and 0.6% (w/v) Nonidet P-40. The samples were cooled to room temperature, then digested overnight at 37°C with 1 unit of N-glycosidase F. For treatment with O-glycosidase, native and recombinant G1-G2 were denatured by boiling in 20 mM sodium phosphate buffer, pH 7.2, containing 1% SDS and 0.5% (w/v) Nonidet P-40. The cooled samples were digested overnight at 37°C with 1 milliunit of O-glycosidase.
Production and Characterization of Polyclonal Anti-DIPEN and Anti-ITEGE Antisera-Polyclonal rabbit sera specific for the C-terminal neoepitope sequences DIPEN and ITEGE were prepared and shown to be similar to those described previously (13,28,30). The synthetic peptide immunogens CGGNITEGE and CGGFVDIPEN which were generous gifts from Drs. P. Roughley and J. Mort, Shriners Hospital for Children, Montreal Canada, were conjugated to ovalbumin using bromoacetic acid N-hydroxysuccinimide ester (45). Four-month-old female rabbits were primed subcutaneously with 400 g of conjugate in 500 l of a 1:3 mixture of peptide conjugate in PBS:Freund's complete adju-vant. Booster injections at days 14, 28, 42, and 99 contained 400 g of conjugate in 300 l of a 1:2 mixture of peptide conjugate in PBS: Freund's incomplete adjuvant, and the final bleed was on day 128. Peptides conjugated to bovine serum albumin rather than ovalbumin were used to titrate antisera in enzyme-linked immunosorbent assays, and also in competition assays with truncated and extended peptides to demonstrate specificity.
Solid Phase Binding of rG1-G2 to Matrix Ligands-The binding of rG1-G2 to link protein, G1 domain, and hyaluronan was demonstrated by incubation of 125 I-labeled rG1-G2 with nitrocellulose membrane containing dots of purified ligands. rG1-G2 was iodinated using the chloramine-T method (46). Dilutions (10 -1,000 g/ml) of link protein in 4 M GdnHCl and G1 domain, both purified from pig laryngeal aggrecan, hyaluronan, bovine serum albumin, and undiluted 4 M GdnHCl were dotted onto a nitrocellulose membrane using a dot-blotting apparatus and washed through with 2 ml of PBS. The membrane was blocked with 5% skim milk/PBS for 3 h at room temperature, followed by overnight incubation with 125 I-rG1-G2 (2 ϫ 10 6 cpm/5 ml) at 4°C. The membrane was washed in 0.1% Tween/PBS for 45 min, then exposed to a phosphorscreen for 24 h. The phosphorscreen was scanned on a Storm 840 PhosphorImager (Molecular Dynamics) and the dots analyzed by densitometry.
Antibodies-The neoepitope monoclonal antibodies AF-28 (29) and BC-3 (26) recognizing the N-terminal sequences F 342 FGVG and A 374 RGSV, respectively, have been described. Monoclonal antibody 1-C-6 (Developmental Studies Hybridoma Bank, University of Iowa, Dept. of Biological Sciences, Ames, IA) recognizes an epitope within the sequence SPEQLQAAYG in the G1 and G2 domains (39,47). Monoclonal antibody 5-D-4 recognizing a highly sulfated 5 disaccharide unit of keratan sulfate (48,49) was a gift from Professor B. Caterson, University of Wales, Cardiff, UK. Polyclonal anti-human G1 domain antisera was a gift from Professor T. Hardingham, University of Manchester, United Kingdom. Aggrecanase-digested samples were analyzed on 8 -16% Tris glycine gradient gels. BC-3 Western blots were done with goat anti-mouse IgG linked to alkaline phosphatase and the blots were developed with a Western blue phosphatase system. AF-28, anti-DIPEN 341 , 1-C-6, and 5-D-4 Western blots were developed with horseradish peroxidase-conjugated secondary antibodies followed by enhanced chemiluminescence. Anti-ITEGE 373 blots were developed with either alkaline phosphatase or horseradish peroxidase-conjugated secondary antibodies.
Incubation of rG1-G2 with Conditioned Medium-Conditioned medium was collected from cartilage explant cultures and used as a source of unpurified proteinases. Articular cartilage dissected from the metacarpophalangeal joints of adult pigs was placed into sterile containers containing approximately 1 g of cartilage/10 ml of Dulbecco's modified Eagle's medium. The tissue was cultured in a humidified incubator at 37°C with 5% CO 2 for 5 days in serum-free Dulbecco's modified Eagle's medium containing 100 units/ml penicillin and 100 g/ml streptomycin antibiotics, 2 mM glutamine, 20 mM Hepes. The tissue was untreated for the first 2 days, then stimulated with 10 ng/ml IL-1␣ or 1 M retinoic acid for the remaining 3 days of culture. Control cultures were unstimulated for 5 days. The medium was changed daily and stored frozen at Ϫ20°C. Thawed medium was concentrated ϫ15 by centrifugation in an Ultrafree centrifugal device (Millipore) with a 10,000 dalton molecular size cutoff. rG1-G2 (20 g) was then incubated overnight at 37°C with 10 l of concentrated day 5 medium from control, IL-1␣, or retinoic acid-stimulated cultures. Samples containing 2 g of rG1-G2 were analyzed by SDS-PAGE and Western blotting. Separate samples containing the same volume of conditioned medium, but without rG1-G2 were analyzed in parallel.

RESULTS
The construct made for expression of human G1-G2 contained the first 656 amino acids of the mature aggrecan core protein and a C-terminal polyhistidine tag. The construct was sequenced and found to be identical with the published cDNA sequence (4), except for two modest changes reported previously (50); one that changed Pro 462 -Gly 463 to Leu 462 -Arg 463 , and insertion of CTG (coding for leucine) between nucleotides 1872 and 1873 of the published cDNA sequence. The construct was cloned into pGemZ11(fϩ) and analyzed by cell-free transcription and translation. The [ 35 S]methionine-labeled translation product migrated on SDS-PAGE with an apparent molecular mass of 90,000, slightly higher than the predicted size of 73,290 daltons (data not shown). The size of the translation product confirmed that the human rG1-G2 was full-length. rG1-G2 of the same size was also present in the medium and cells of Sf21 cells infected with recombinant virus (data not shown).
Purification of rG1-G2-The specific and high affinity binding of the aggrecan G1 domain to HA (51) suggested that HA binding could be exploited for the purification of rG1-G2, provided the expressed protein was correctly folded. Sepharoselinked HA efficiently bound rG1-G2 under associative conditions (Fig. 2). No rG1-G2 was recovered in the column washes (Fig. 2, lanes 3 and 4) and the rG1-G2 was eluted by dissociation from HA with 4 M GdnHCl, 50 mM sodium acetate, pH 5.8 (Fig. 2, lanes 5, 6, and 9). Silver stain analysis showed that the majority of contaminating proteins present in the fetal calf serum, and other proteins secreted from the insect cells, did not bind to HA-Sepharose and were recovered in the unbound fraction (Fig. 2, lanes 2 and 8). Thus, HA-Sepharose was a highly efficient first step in purification of the rG1-G2. Subsequent size exclusion chromatography in acetate-buffered 4 M GdnHCl was used to purify the rG1-G2 to homogeneity. Fractionation of the sample on a Bio-Sep SEC 400 column showed one major peak in the included volume of the column that was purified to homogeneity by refractionation on the same column (Fig. 3). The maximum yield of rG1-G2 obtained from 240 ml of serum-free culture medium and 8 ϫ 10 8 cells was approximately 7.7 mg.
Rotary Shadowing Electron Microscopy of rG1-G2-Native G1-G2 purified from pig laryngeal aggrecan (40) and human rG1-G2 were analyzed by rotary shadowing electron microscopy. Several fields of each were observed at comparable magnifications and representative structures are shown in Fig. 4. Both samples showed a similar "double-globe" structure as the predominant feature in the field. The native pig material (Fig.  4B) appeared slightly larger, with the separation between the globular domains longer. The native pig G1-G2 measured approximately 39 nm globe-to-globe, in excellent agreement with previous reports of the length of G1-G2 measured by similar means (52). The human rG1-G2 preparation (Fig. 4A) appeared less extended, with larger globular regions and the approximate length of these structures was 29 nm.
Interaction of rG1-G2 with Matrix Ligands-Link protein and the aggrecan G1 domain participate in aggregate forma-tion by their noncovalent binding to HA and each other. The ability of rG1-G2 to interact with link protein, G1 domain, and HA was investigated in a solid phase binding experiment, where ligands were immobilized on a nitrocellulose membrane and allowed to interact with 125 I-rG1-G2 (Fig. 5). The order for the specific binding of 125 I-rG1-G2 was HA Ͼ Ͼ link protein Ͼ G1 domain, and there was no binding of rG1-G2 to bovine serum albumin at any concentration tested. The results of these interaction studies are comparable with earlier interaction studies between isolated native aggrecan components (40,53).
Digestion of rG1-G2 with Aggrecanase-If rG1-G2 is to be a suitable in vitro substrate for investigating aggrecan cleavage by purified proteinases or cultured cells, it must show the same cleavage site specificity for aggrecanase (Glu 373 2 Ala) and MMPs (Asn 341 2 Phe) as the native molecule. Aggrecanase activity was generated in conditioned media from IL-1-stimulated bovine nasal cartilage (42). rG1-G2 was digested with aggrecanase in the presence of a potent MMP inhibitor XS309, and the products detected by Western blotting with monoclonal antibody BC-3 specific for the A 374 RGSV N-terminal sequence and polyclonal anti-ITEGE specific for the C-terminal sequence (Fig. 1). A single band of M r 45,000 was detected with BC-3 (Fig. 7a) and a single G1 species of M r 56,000 was detected with anti-ITEGE (Fig. 7b), confirming that aggrecanase cleaved the rG1-G2 at the Glu 373 2 Ala bond. No bands were detected with AF-28 or anti-DIPEN antibodies (results not shown). Competitive enzyme-linked immunosorbent assay experiments confirmed the specificity of the anti-ITEGE antisera since the 10-mer peptide with an extension of 1 amino acid (LPRNITE-GEA) was 235 times less competitive than the true neoepitope peptide (PLPRNITEGE). 10-Mer peptides truncated by one (LPLPRNITEG) or two (ELPLPRNITE) amino acids gave no competition at all (results not shown).
Digestion of rG1-G2 with MMPs-Recombinant human MMP-13 (36) was used to digest rG1-G2 since this MMP is relatively abundant in cartilage (54 -56) and the products of native pig G1-G2 digestion by MMP-13 have been characterized previously (19). Following MMP-13 digestion, three rG1-G2 fragments were detected by silver stain (Fig. 8a, lane  2). The fastest migrating band with M r 44,000 (Fig. 8a, fragment 3) was shown by Western blotting with anti-human G1 antiserum to be the G1 domain (Fig. 8b, lane 2). Based on the results of MMP-13 digestion of native pig G1-G2 (19) we predicted that MMP-13 digestion of the human rG1-G2 would yield two G2 bands by cleaving at both the major and minor MMP sites. A G2 product of M r 55,000 with F 342 FGVG N terminus was detected with monoclonal antibody AF-28 (Fig.  8c, lane 2), and corresponded in size to fragment 1 in the silver stain gels. Fragment 2 was not analyzed further, however, it is likely that it is the smaller rG2 fragment resulting from hydrolysis of the Pro 384 2 Val bond in the IGD (19). These results indicate that MMP-13 has the same specificity for human rG1-G2 as it does for native pig G1-G2. No bands were detected with BC-3 or anti-ITEGE antibodies (Fig. 1), confirming previous work (19) that showed MMP-13 does not have aggrecanase activity (results not shown).
Digestion of rG1-G2 with Atrolysin C-We next digested rG1-G2 with enzymes known to cleave at both the MMP and aggrecanase sites, atrolysin C and MMP-8. Digestion of native pig G1-G2 (22) or bovine aggrecan (23) with MMP-8 showed that the enzyme cleaved its substrates in a sequential manner, hydrolyzing the Asn 341 2 Phe bond first, and the Glu 373 2 Ala bond second. rG1-G2 was also cleaved sequentially by MMP-8. Digestion of rG1-G2 with increasing concentrations of MMP-8 showed that the AF-28 epitope was present at all concentrations, but was decreased at the higher concentrations when  1 and 4). b, samples digested with keratanase (lanes 2 and 5), keratanase II (lanes 3 and 6), or undigested (lanes 1 and 4). BC-3 epitope began to appear. DIPEN 341 epitope in contrast remained constant and did not decrease (data not shown). The snake venom proteinase atrolysin C cleaves bovine aggrecan at the Asn 341 2 Phe and Glu 373 2 Ala sites, but in an independent rather than sequential manner (24). When rG1-G2 and native pig G1-G2 were digested with atrolysin C at 0.1 M (data not shown) and 1.0 M, digestion products were detected by silver stain (Fig. 10a) and ITEGE 373 immunoreactivity (Fig.  10b). AF-28 fragments were detected for native G1-G2 (Fig.  10c, lane 4) but no AF-28 fragments were found in atrolysin C digests of rG1-G2 (Fig. 10c, lane 2), indicating that atrolysin C was unable to cleave rG1-G2 at the MMP site. To determine whether the absence of keratan sulfate chains on rG1-G2 conferred resistance to atrolysin C cleavage at the MMP site, we analyzed the susceptibility of native G1-G2 to atrolysin C following removal of keratan sulfate chains by keratanase (Fig.  10, d-f). We found that native G1-G2 was readily cleaved at the MMP site, irrespective of whether or not keratan sulfate chains had been enzymatically removed (Fig. 10, e and f). Anti-ITEGE 373 immunoreactive products were also present following keratanase digestion of native G1-G2 (data not shown). These results suggest that the presence or absence of keratan sulfate chains per se does not affect the specificity of action of atrolysin C.
Incubation of rG1-G2 with Conditioned Explant Culture Medium-Studies by Hughes et al. (31,57) have shown that aggrecanase present in chondrocyte-conditioned medium can be monitored using a chimeric recombinant substrate. We have tested whether aggrecanase present in conditioned media from cartilage explant cultures can be monitored with rG1-G2. Western blotting with anti-ITEGE (Fig. 1) rabbit sera revealed the presence of aggrecanase-derived G1 fragments following incubation of rG1-G2 with media from IL-1␣ or retinoate-stimulated cultures (Fig. 11, lanes 2 and 3). No aggrecanase-derived ITEGE fragments were found in samples incubated with media FIG. 8. MMP-13 digestion of rG1-G2. rG1-G2 (7 g) was digested for 1 h at 37°C with MMP-13 at a final concentration of 10 g/ml in a total volume of 20 l (lane 2). Aliquots (1 g) were electrophoresed on 7.5% SDS gels and bands detected by silver stain (a) or Western blotting with anti-G1 antiserum (b) or monoclonal AF-28 (c). from control cultures (Fig. 11, lane 1), nor was anti-ITEGE signal present in lanes containing conditioned medium alone (data not shown). No AF-28 or DIPEN fragments (Fig. 1) were found in any samples, suggesting that there were no active MMPs present in conditioned media from control, IL-1␣, or retinoic acid-treated cultures (data not shown). These data demonstrate the utility of rG1-G2 for the study of cell-mediated aggrecanolysis.

DISCUSSION
Properties of rG1-G2-In this paper we report the utility of rG1-G2 for the study of aggrecanolysis in vitro, by demonstrating that the intrinsic properties of MMP and aggrecanase cleavage site specificity, as well as HA, link protein, and G1 binding, are retained. MMP digestion of rG1-G2 yields G1 and G2 products of M r 44,000 and 55,000, respectively, while aggrecanase digestion yields G1 and G2 products of M r 56,000 and 45,000, respectively. These products are readily identified with neoepitope antibodies in vitro. Recombinant fusion proteins comprising various domains of aggrecan G1-G2 have been reported previously. An artificial aggrecan substrate (rAgg1) comprising the IGD of human aggrecan flanked by the marker sequences FLAG TM at the N terminus and the human immunoglobulin G1 constant region at the C terminus has been expressed in COS cells (57) and shown to be processed by aggrecanase and MMPs (31). A G1-G2 fusion protein expressed in an embryonal kidney cell line, fused with the laminin ␥1 chain at the C terminus and an interleukin-2 signal peptide and FLAG TM sequence at the N terminus has also been produced and shown to bind hyaluronan (58). This paper describes for the first time the production of a natural rG1-G2 that has been expressed in insect cells, giving significantly greater yields of recombinant protein compared with mammalian expression, but without full glycosylation.
The rG1-G2 has been produced in high yield and purified by HA-affinity chromatography. We observed that the Sf21 cells maintained in serum-free medium produced rG1-G2 that migrated as a broader band on gels than the rG1-G2 secreted by cells maintained in fetal calf serum. The broader band presumably reflects a greater degree of heterogeneity, which may result from more extensive glycosylation and differential processing by the signal peptidase (59,60). During early experiments to develop and test methods for purification we found that freezing unpurified rG1-G2 in the conditioned culture medium reduced the final yield of purified protein, but that purified rG1-G2 was stable to freeze-thawing. We also found that HA-Sepharose affinity chromatography was far superior compared with metal affinity chromatography utilizing the polyhistidine tag.
Measurements by rotary shadowing electron microscopy (52) for the length of the G1-IGD-G2 region (double-globe) have been derived for aggrecan purified from bovine nasal cartilage, pig laryngeal cartilage, and the swarm rat chondrosarcoma.
The native pig G1-G2 analyzed in the present work measured approximately 39 nm globe-to-globe, in excellent agreement with previous reports (52,61). The enzymatic removal from bovine aggrecan of chondroitin sulfate chains caused a shortening of the distance between the G2 and G3 domains from 405 Ϯ 37 to 263 Ϯ 27 nm, which is approximately 35% (61). The authors suggested that charge repulsion or steric effects of chondroitin sulfate chains were responsible for the extended length of the E2 domain between G2 and G3. In the present work, the human rG1-G2 measured 29 nm, which is approximately 29% shorter than native G1-G2. One possible explanation is that in rG1-G2, the lack of keratan sulfate substitution in the IGD contributes to the shortened length of the E1 domain. However, detailed analyses of double globe structures from rat chondrosarcoma, bovine nasal, and pig laryngeal cartilage have shown that the length of the E1 segment between G1 and G2 is remarkably constant, at about 25 nm (52). Since there is no evidence for keratan sulfate substitution on rat chondrosarcoma aggrecan, this would suggest that keratan sulfate is not a key factor in determining the length of the E1 domain, and that perhaps an altered pattern of O-or N-linked glycosylation may contribute to shortening of E1 in rG1-G2.
rG1-G2 Glycosylation-The expression of keratan sulfatecontaining proteins by insect cells has not been reported. In our hands, there was no substitution of keratan sulfate on rG1-G2, suggesting that Sf21 cells lack the specialized glycosylation machinery necessary for keratan sulfate biosynthesis. Similarly, the chondroitin/dermatan sulfate proteoglycan, decorin, was not substituted with glycosaminoglycan chains by Sf21 cells (62). N-Glycosidase F releases asparagine-linked glycan chains from mammalian glycoproteins including N-linked keratan sulfate. Since keratanase treatment reduced the size of the native G1-G2 by about 30 kDa and N-glycosidase F by only about 15 kDa, it appears that about 50% of the keratan sulfate on native pig G1-G2 is N-linked to the protein core. Whether this N-linked keratan sulfate is present on the same asparagine residues suggested by Barry et al. (63) for bovine aggrecan HABR is under investigation.
N-Linked glycosylation is common in insect cells and a small amount was present on rG1-G2, albeit at 50 -85% less than that on native G1-G2. The N-linked glycans present on rG1-G2 are likely to comprise high mannose type (Man 9 -5 GlcNAc 2 ) or short truncated structures (Man 3-2 GlcNAc 2 ) since these are most commonly observed on proteins expressed in cells derived from S. frugiperda (64 -70) and are released by treatment with N-glycosidase F. However, we cannot exclude the possibility that the N-glycans may be of the complex type with or without terminal sialic acid residues, which are uncommonly reported for these cells (66,71). Some studies have shown that the N-linked glycans on recombinant proteins secreted from insect cells are important for functional properties. For example, ligand binding of the thromboxane A 2 receptor (72), secretion of decorin (62), and the ability of human plasminogen to be activated by urokinase (73) are all influenced by N-glycan substitution. At present we do not know whether differences in Nlinked glycosylation may contribute to the inability of atrolysin C to cleave the Asn 341 2 Phe bond in rG1-G2.
Studies by Arner's laboratory 2 have suggested that keratan sulfate substitution on aggrecan may facilitate aggrecanase activity. Treatment of bovine nasal cartilage aggrecan with keratanases I and II generated a product which was not cleaved by crude aggrecanase, whereas treatment with chondroitinase ABC did not markedly alter the substrate cleavage. However, the present data clearly show that keratan sulfate chains are not strictly required for aggrecanase activity; this is in agreement with the demonstrated aggrecanase-mediated cleavage of other aggrecan substrates, such as rat chondrosarcoma aggrecan (27) and recombinant IGD, rAgg (57) which are devoid of keratan sulfate. With respect to MMP cleavage, rG1-G2 was a much better substrate then glycosylated G1-G2 since lower enzyme concentrations and shorter digestion times were required for the maximum yield of products. The role of keratan sulfate in the control of aggrecanolysis is an emerging theme and future studies with human rG1-G2 expressed in a system able to elaborate keratan sulfate will allow us to address this issue directly.
The susceptibility of rG1-G2 to atrolysin C was of interest because this reprolysin is able to cleave bovine aggrecan at both the MMP and aggrecanase sites (24). rG1-G2 was not cleaved at the MMP site by atrolysin C even though it was efficiently cleaved at the aggrecanase site. When we investigated whether native glycosylated G1-G2 was cleaved at the MMP site by atrolysin C we found not only that it was, but that native G1-G2 digested with keratanase was also cleaved at the MMP site. This result suggests that keratan sulfate substitution in the IGD has no influence on the specificity of atrolysin C for cleavage at the MMP site, and that factors other than lack of keratan sulfate chains on rG1-G2 abolishes atrolysin C cleavage at Asn 341 2 Phe. Using circular dichroism and fluorescent spectroscopy, Krishnan et al. (74) have observed that the conformation of biglycan is variably influenced by whether it is synthesized with or without a glycosaminoglycan chain, and that removal of the chain after secretion has no appreciable influence on its conformation. Although biglycan and aggrecan core proteins are not at all similar and the extent of glycosylation on each is vastly different, it is possible to speculate that the addition of keratan sulfate chains to the nascent aggrecan core protein during biosynthesis may result in a structural conformation that is permissive for atrolysin C cleavage at the Asn 341 2 Phe site. In such a scenario, subsequent removal of keratan sulfate would not alter protein conformation or abolish atrolysin C cleavage at Asn 341 2 Phe. However, rG1-G2 synthesized without keratan sulfate chains would possibly adopt a different conformational structure that allows binding of the enzyme (since the substrate is cleaved at the Glu 373 2 Ala bond), but is inhibitory for cleavage at the Asn 341 2 Phe site.
In Vitro Aggrecanolysis-Soluble active aggrecanase has been generated in conditioned media from interleukin-1-stimulated bovine nasal cartilage (42), and we now show that it can also be harvested from porcine articular cartilage following stimulation with IL-1␣ or retinoic acid. We show that in contrast to cell culture (31), conditioned medium from explant cultures does not contain active MMPs, since DIPEN 341 and F 342 FGVG neoepitopes (Fig. 1) were not detected in digests of rG1-G2 with either control or stimulated culture medium. There are several possible reasons why active MMPs are not released in cartilage cultures, but are released in chondrocyte cultures, and why active aggrecanase is released. One is that there are significant levels of TIMPs present in the cartilage matrix. MMP cleavage in tissue slices most likely occurs transiently, prior to inhibition by TIMPs. The IC 50 for inhibition of aggrecanase by TIMP-1 is 210 nM, whereas the IC 50 for inhibition of MMPs by TIMP-1 is about 100-fold lower (42). Furthermore, aggrecanase is not inhibited by TIMP-2. Thus, MMPs in tissue are much less likely to escape inhibition by TIMP than aggrecanase. Second, MMPs bind to extracellular matrix, and indeed there are studies to suggest that active stromelysin may be trapped in the cartilage matrix through tight binding, perhaps to its natural substrate (75). At present there is no infor-mation regarding binding of aggrecanase to extracellular matrix or endogenous inhibitors. Finally, the use of rG1-G2 as an exogenous substrate for chondrocyte-derived proteinases indicates that aggrecanase (from at least two species) can cleave at Glu 373 2 Ala irrespective of whether it is bound to HA, link protein, or other matrix proteins, confirming other work (31,42,57,76).
In summary, we have produced a natural recombinant G1-G2 substrate that has a typical, but compact, double-globe structure and which exhibits native properties of binding to link protein and HA. The substrate is cleaved at the authentic Asn 341 2 Phe and Glu 373 2 Ala sites by MMPs and aggrecanase, respectively, and has been used as an exogenous substrate to monitor release of MMPs and aggrecanase activities from cartilage cultures. The rG1-G2 can substitute for native substrates in studies of aggrecanase and MMPs which are of major interest in terms of aggrecanolysis.