Mutations in the Interglobular Domain of Aggrecan Alter Matrix Metalloproteinase and Aggrecanase Cleavage Patterns

We have expressed G1-G2 mutants with amino acid changes at the DIPEN341↓342FFGVG and ITEGE373↓374ARGSV cleavage sites, in order to investigate the relationship between matrix metalloproteinase (MMP) and aggrecanase activities in the interglobular domain (IGD) of aggrecan. The mutation DIPEN341 to DIGSA341 partially blocked cleavage by MMP-13 and MMP-8 at the MMP site, while the mutation 342FFGVG to342GTRVG completely blocked cleavage at this site by MMP-1, -2, -3, -7, -8, -9, -13, -14. Each of the MMP cleavage site mutants, including a four-amino acid deletion mutant lacking residues ENFF343, were efficiently cleaved by aggrecanase, suggesting that the primary sequence at the MMP site had no effect on aggrecanase activity in the IGD. The mutation 374ARGSV to374NVYSV completely blocked cleavage at the aggrecanase site by aggrecanase, MMP-8 and atrolysin C but had no effect on the ability of MMP-8 and MMP-13 to cleave at the Asn341↓Phe bond. Susceptibility to atrolysin C cleavage at the MMP site was conferred in the DIGSA341 mutant but absent in the wild-type,342GTRVG, 374NVYSV, and deletion mutants. To further explore the relationship between MMP and aggrecanase activities, sequential digest experiments were done in which MMP degradation products were subsequently digested with aggrecanase andvice versa. Aggrecanase-derived G1 domains with ITEGE373 C termini were viable substrates for MMPs; however, MMP-derived G2 fragments were resistant to cleavage by aggrecanase. A 10-mer peptide FVDIPENFFG, which is a substrate analogue for the MMP cleavage site, inhibited aggrecanase cleavage at the Glu373↓Ala bond. This study demonstrates that MMPs and aggrecanase have unique substrate recognition in the IGD of aggrecan and suggests that sequences at the C terminus of the DIPEN341 G1 domain may be important for regulating aggrecanase cleavage.

Aggrecanase is the major mediator of aggrecanolysis in vitro (29 -32); however, MMPs are also active and the relationship between these activities is not understood. Three potential models for cleavage in the IGD have been proposed (33,34). In the first model, aggrecan is cleaved initially by aggrecanase, and the G1 fragments remaining in the tissue are subsequently cleaved by MMPs. Model 1 therefore generates aggrecanase and MMP G1 fragments and a 32-amino acid Phe 342 -Glu 373 fragment but fails to account for the presence of 342 FFGVG fragments in which the aggrecanase site remains intact (19). In the second model, aggrecan is cleaved initially by MMPs, and the released fragment containing the 342 FFGVG neoepitope is subsequently cleaved by aggrecanase. Model 2 generates both aggrecanase (with 374 ARGSV N terminus) and MMP (with 342 FFGVG N terminus) fragments and the 32-amino acid Phe 342 -Glu 373 fragment but fails to account for the presence of aggrecanase G1 fragments present in human (21,22), bovine (35,29), pig (29), rat (35), and mouse (23,26) cartilage. Model 3, which proposes that aggrecanase and MMP activities are independent, is favored by us (19) and others (21,22). We favor this model, because in studies investigating the validity of each model (34) (see accompanying article (55)) we have observed that both ITEGE 373 G1 domains and large 342 FFGVG fragments are generated during IL-1␣ treatment of pig articular cartilage in vitro. Since the ITEGE 373 G1 domain and 342 FF-GVG fragments cannot be derived from the one aggrecan molecule, we concluded that in pig cartilage stimulated with IL-1␣, MMP, and aggrecanase activities were mutually exclusive (36).
The aim of this study was to resolve the mechanism by which MMP and aggrecanase activities could be mutually exclusive. We asked whether mutations at the MMP cleavage site would affect cleavage by aggrecanase or vice versa, and we asked whether the products of one activity were viable substrates for the other. The answers to these questions led us to test the hypothesis that sequences surrounding the MMP cleavage site may be important for aggrecanase activity.
Site-directed Mutagenesis of the MMP and Aggrecanase Cleavage Sites in the Aggrecan IGD-Mutations in the human G1-G2 construct (46) were produced using splicing by overlap extension PCR (48). The primer sets and their relative positions are shown in Table I. Primers Aggr S12 and Aggr S13 introduced a 12-base deletion of GAAAACT-TCTTT, resulting in deletion of four amino acids, ENFF 343 . Primers Aggr S8 and Aggr S9 were used to replace CCAGAAAAC with GGAT-CAGCC, changing amino acids PEN 341 to GSA 341 . Primers Aggr S14 and Aggr S15 were used to replace TTCTTTGGA with GGCACTAGA, changing amino acids 342 FFG to 342 GTR. Primers Aggr S10 and Aggr S11 were used to replace GCCCGAGGC with AACGTATAC, changing amino acids 374 ARG to 374 NVY. Each mutation resulted in the introduction or removal of restriction enzyme sites to facilitate screening.
The polymerase chain reactions were done using a PerkinElmer Life Sciences thermal cycler model 2400 or 9600, over 30 cycles. Cycle one was performed at 94°C for 120 s, annealing (Table I) for 90 s and 72°C for 90 s, followed by 30 cycles of 94°C for 30 s, annealing (Table I) for 45 s and 72°C for 30 s. The reactions were terminated at 72°C for 7 min. Two independent PCR products were first produced using Aggr 7 (sense) with primer A (antisense), and primer B (sense) with Aggr 6 (antisense) of the primer sets in the first round of PCR using 10 ng of pBsktG1-G2 (46) as a template. The amplified products were purified by agarose electrophoresis, recovered using Geneclean, and subjected to a second round of overlapping PCR with primers Aggr 6 and Aggr 7. The PCR fragments were purified and digested with SphI and BsmI restriction enzymes to release a 375-bp fragment, which was then subcloned

5Ј-AATGCGGCTGCCCAGGGACA-3Ј
Reverse primer 1421-1399 60 into pBsktG1-G2, replacing the normal 375-bp cassette. The pB-sktG1-G2 mutants were then digested with EcoRI and XbaI, and the G1-G2 mutant construct was subcloned into the pBacPAK8 transfer vector. Prior to production of recombinant virus, the mutant constructs were screened by restriction enzyme digestion and sequenced using Amplitaq Cycle sequencing. Wild-type and mutant rG1-G2 were expressed and purified as described (46). Proteinase Digestions of Wild-type and Mutant rG1-G2-Matrix metalloproteinase, aggrecanase, and atrolysin C digestions were done at 37°C in buffer containing 10 mM calcium chloride, 100 mM sodium chloride, 50 mM Tris-HCl, pH 7.5. Digests were stopped either by boiling or by adding EDTA and 1,10-phenanthroline to final concentrations of 10 and 2 mM, respectively. Denatured samples were analyzed by Western blotting or silver stain after SDS-polyacrylamide gel electrophoresis. In sequential digest experiments, 10 g of substrates digested with 3 g/ml MMP-13 for 2 h were incubated with a 2-fold molar excess of TIMP-1 for 30 min on ice to inhibit further MMP-13 action and then incubated overnight with 3 l of aggrecanase. Similarly, substrates digested overnight with 3 l of aggrecanase were subsequently digested for 2 h with 3 g/ml MMP-13.
Preparation of Bovine Nasal Aggrecanase-Aggrecanase was purified from bovine nasal cartilage in a similar fashion to that previously reported (4). In brief, nasal cartilage was dissected from bovine noses obtained within 4 h after slaughter. The tissue was cut into 1.2-mm cubed pieces and cultured for 2 days in Dulbecco's modified Eagle's medium containing 5% fetal calf serum, followed by 1 day in Dulbecco's modified Eagle's medium containing 2.5% fetal calf serum and finally 1 day in Dulbecco's modified Eagle's medium without fetal calf serum. After the 1 day in serum-free medium, IL-1␣ was added at a concentration of 0.15 nM. The conditioned medium was collected 48 h later, and aggrecanase was purified by ion exchange, gel filtration, and wheat germ agglutinin chromatography. Several silver-stained bands were present on SDS gels; however, the preparation was "proteolytically pure," since it contained no MMP activity, and other classes of proteinase inhibitors failed to reduce the amount of ITEGE 373 products formed. The aggrecanase preparation was not inhibitable by TIMP-1 when assessed using bovine aggrecan as substrate and anti-ITEGE 373 neoepitope as the read-out. TIMP-1 at 2.2 M failed to inhibit the aggrecanase activity but gave an IC 50 against MMP-1 and MMP-8 of 4 nM and Ͻ2 nM, respectively, using a synthetic substrate assay. Since TIMP-1 has an IC 50 of 210 nM against ADAMTS-4, 2 it is unlikely that the aggrecanase preparation used in the present work is identical to AD-AMTS-4 (3).
Inhibition of Aggrecanase Activity by Synthetic Peptides-The 7-mer peptides IPENFFG and TEGEARG were from Charing Cross Hospital (London), and the 10-mer peptide FVDIPENFFG was from Auspep. IPENFFG and FVDIPENFFG were dissolved in distilled water at 25 mg/ml, and TEGEARG was dissolved in 75 mM NaCO 3 , pH 8.0, at the same concentration. The peptides were present at a 3000-fold molar excess over rG1-G2 substrate in aggrecanase digests. The addition of 75 mM NaCO 3 buffer alone to aggrecanase digests did not alter the pH, nor did it affect the generation of ITEGE 373 epitope.

RESULTS
Mutations at the DIPEN 341 2FFGVG and ITEGE 373 2ARGSV cleavage sites in the aggrecan IGD were made (Fig. 1), and the mutant substrates were tested for their susceptibility to digestion by MMPs, aggrecanase, and atrolysin C.
MMP-13 Digestion of Wild-type and Mutant rG1-G2-Wild type and mutant G1-G2 substrates were digested with MMP-13, since this MMP is abundantly expressed in arthritic cartilage. Silver staining (Fig. 2, a, d, g, j, and m) was used to monitor 1) the extent of digestion and the approximate ratio of undigested:digested material and 2) uniform loading of samples on gels. Silver staining was not useful for specifically identifying G1 and G2 domains, since the fragments often migrated together. Western botting with neoepitope antibodies was used to determine whether cleavage was occurring at the MMP site. Mutation of the sequence DIPEN 341 to DIGSA 341 retarded, but did not block, cleavage at the MMP site, since a higher concentration of MMP-13 (10 g/ml) was required to produce the maximum amount of 342 FFGVG epitope in digests of the DIGSA 341 mutant (Fig. 2e) compared with the wild type (Fig. 2b). However, mutation of 342 FFGVG to 342 GTRVG completely abolished cleavage at the MMP site, since no DIPEN 341 epitope was detected in digests of the 342 GTRVG mutant (Fig.  2i). This result is consistent with previous reports that most substitutions at the PЈ 1 position are detrimental for collagenase activity (49). Silver staining showed that the 342 GTRVG mutant was cleaved by MMP-13 elsewhere in the IGD, most likely at the minor sites Pro 384 2Val 385 and Asp 441 2Leu 442 , as shown for native G1-G2 (11). Generation in the 374 NVYSV mutant of DIPEN 341 and 342 FFGVG neoepitopes by MMP-13 (Fig. 2, m-o) was similar to the wild type (Fig. 2, a-c), indicating that the 374 NVYSV mutation at the aggrecanase site had no affect on MMP-13 activity against rG1-G2. MMP-13 at 100 g/ml appeared to "overdigest" the substrate, leading to loss of 342 FF-GVG (Fig. 2, b, e, and n) and DIPEN 341 (Fig. 2, c and o) neoepitopes as well as G1 domain epitopes detected with polyclonal antisera (data not shown). This is consistent with our previous observation that rG1-G2 is more sensitive to proteolysis than native glycosylated pig G1-G2 (46).
MMP-8 Digestion of Wild-type and Mutant rG1-G2-MMP-8 cleaves bovine aggrecan (50) and native pig G1-G2 (51) at both the MMP and aggrecanase sites. MMP-8 cleaves its substrate in a sequential manner, at the Asn 341 2Phe bond initially, and the Glu 373 2Ala bond subsequently. Wild-type rG1-G2 was also cleaved in a sequential manner by MMP-8. 342 FFGVG epitope was maximal at low concentrations of enzyme but decreased at higher concentrations (Fig. 3b), and the decrease in epitope was concomitant with an increase in 374 ARGSV epitope (Fig. 3d). DIPEN 341 epitope on the other hand was unchanged (Fig. 3c). As with MMP-13, the DIGSA 341 mutant was less susceptible to MMP-8 cleavage at the MMP site. Higher concentrations of MMP-8 were required to achieve maximum 342 FFGVG epitope in the DIGSA 341 digests, and epitope levels were not noticeably 2 E. Arner, personal communication. decreased at the highest concentration of enzyme (200 g/ml) (Fig. 3f), although a small amount of 374 ARGSV epitope was detected (Fig. 3h). MMP-8 digestion of the 342 GTRVG mutant showed that hydrolysis of the Asn 341 2Phe bond was not an essential prerequisite for cleavage at Glu 373 2Ala. 374 ARGSV epitope was detected in digested 342 GTRVG mutant at enzyme concentrations of 30 g/ml and higher (Fig. 3l), in the absence of MMP site cleavage (Fig. 3k). The 374 NVYSV mutation did not interfere with MMP-8 cleavage at DIPEN 341 2FFGVG but did block cleavage at ITEGE 373 2ARGSV, since there was no loss of 342 FFGVG epitope at high concentrations of enzyme (Fig. 3r).
Aggrecanase Digestion of Wild-type and Mutant rG1-G2-Wild-type and mutant rG1-G2 were incubated with increasing amounts of bovine aggrecanase, and the products were detected by Western blotting with anti-ITEGE 373 and anti-374 ARGSV antibodies. All of the rG1-G2 substrates except the 374 NVYSV mutant were efficiently cleaved by aggrecanase at the ITEGE 373 2ARGSV site (Fig. 5). Silver staining showed that no 374 NVYSV degradation products were produced by aggrecanase (Fig. 5m), suggesting that, unlike the MMPs, aggrecanase had specificity for only one site in the IGD. Atrolysin C Digestion of Wild-type and Mutant rG1-G2-Native glycosylated bovine aggrecan (52) and pig G1-G2 (46) are cleaved by atrolysin C at both the DIPEN 341 2FFGVG and ITEGE 373 2ARGSV bonds, in an independent rather than sequential manner. However, we have recently found that wildtype rG1-G2, which is largely unglycosylated, is cleaved by atrolysin C only at ITEGE 373 2ARGSV and not DIPEN 341 2FFGVG (46). In the present experiments, all of the rG1-G2 substrates except the 374 NVYSV mutant (Fig. 6n) were cleaved by atrolysin C at ITEGE 373 2ARGSV, as detected by anti-ITEGE 373 immunoreactivity (Fig. 6, b, e, h, and k). No DIPEN 341 epitope was detected in any of the digests (data not shown), suggesting that atrolysin C was not able to cleave at DIPEN 341 2FFGVG in the wild-type, 342 GTRVG, or 374 NVYSV substrates. Surprisingly, 342 FFGVG immunoreactivity was de- tected in digested DIGSA 341 mutant (Fig. 6f). These results suggest that the conformational shape of the substrate surrounding the sequence DIPEN 341 may dictate atrolysin C specificity for cleavage in the IGD.
The Western blots revealed that DIPEN 341 epitope was indeed increased (Fig. 7e, lanes 3 and 4), while ITEGE 373 epitope was completely destroyed following digestion of aggrecanase-G1 with MMP-13 (Fig. 7f, lanes 3 and 4). These results therefore confirm that the aggrecanase-derived G1 domain is a viable substrate for MMPs.
In contrast, we found that the MMP-G2 domain was not digested by aggrecanase. We predicted that if MMP-G2 fragments were digested by aggrecanase (Fig. 8c), we would create 374 ARGSV epitope, lose 342 FFGVG epitope (present on the 3-kDa fragment), and observe no change in DIPEN 341 epitope (Fig. 8, d and e). However, Western blot analysis showed that 342 FFGVG epitope was not destroyed (Fig. 8f, lanes 2 and 4) and that there was no gain in 374 ARGSV epitope (Fig. 8i, lane  4). The small amount of anti-374 ARGSV epitope present in Fig.  8i, lane 4, is most likely derived from cleavage of intact substrate rather than MMP-G2 domain, since some undigested substrate survives digestion with MMP-13 for 2 h at 3 g/ml (Fig. 2a). Furthermore, pretreatment of rG1-G2 with MMP-13 (Fig. 8i, lane 4) markedly reduced the yield of 374 ARGSV epitope by aggrecanase (compare Fig. 8i, lane 4, with Fig. 8i,  lane 3). Aggrecanase was active in the presence of TIMP-inhibited MMP-13, since 342 GTRVG mutants digested under the same conditions gave 374 ARGSV epitope (Fig. 8j, lane 4). The results show that the MMP-G2 domain is not a viable substrate for aggrecanase.
Sequences in the MMP-G1 Domain Are Required for Aggrecanase Activity-A previous study in our laboratory showed that a 7-mer peptide IPENFFG, a substrate analogue for the MMP cleavage site, was able to inhibit the release of 374 ARGSV fragments from cartilage cultured with and without interleukin-1 (51). The results of this experiment suggested that the IPENFFG peptide was able to inhibit aggrecanase and now, in conjunction with the results presented in Fig. 8, raise the possibility that sequences present at the C terminus of the MMP-G1 domain may be necessary for aggrecanase activity. To test this hypothesis, we used several synthetic peptides as competitive substrates in aggrecanase digests of rG1-G2 (Fig.  9). In these experiments, the IPENFFG 7-mer peptide was unable to block aggrecanase cleavage at Glu 373 2Ala (Fig. 9, b,  lane 3, and c); however, cleavage was inhibited by 50% in the presence of a 10-mer peptide with sequence FVDIPENFFG (Fig. 9, b, lane 4, and c). A 7-mer peptide TEGEARG, a substrate analogue for the aggrecanase site, completely blocked aggrecanase cleavage.

DISCUSSION
Our finding that MMP-derived G2 fragments are resistant to aggrecanase cleavage in the IGD is novel and provides an explanation for our earlier observation (34,36) that IL-1␣induced loss of aggrecan from pig articular cartilage by MMPs and aggrecanase appeared to be mutually exclusive. The present findings are not limited to digestion of rG1-G2, since native glycosylated 342 FFGVG fragments (see Fig. 7 of accompanying article (55)) and native deglycosylated MMP-1-digested aggrecan (29) also resist digestion by aggrecanase. To revisit the models outlined in the Introduction, model 1, in which aggrecan is cleaved initially by aggrecanase and the G1 fragments remaining in the tissue are subsequently cleaved by MMPs, appears viable. However, the subsequent MMP cleavage in this model represents processing of the G1 domain and has no effect on further loss of aggrecan from the matrix. Model 2, in which aggrecan is cleaved initially by MMPs and the released fragment containing the 342 FFGVG neoepitope is subsequently cleaved by aggrecanase, seems unlikely. Our finding that recombinant 342 FFGVG fragments and native pig 342 FFGVG fragments (see accompanying article (55)) resist aggrecanase cleavage in the IGD thus extends the independent model, model 3. The results show that, in terms of aggrecan loss from tissue (as opposed to ITEGE 373 G1 domain processing), MMP and aggrecanase activities appear to be mutually exclusive in IL-1␣-stimulated aggrecan release from pig cartilage and aggrecanolysis of rG1-G2 in vitro. The results are also consistent with our finding that small 342 FFGVG fragments detected in synovial fluids of osteoarthritis and inflammatory arthritis patients do not contain an ITEGE 373 C terminus (19).
In principle, these results have implications for therapeutic strategies designed to limit aggrecan loss, since it appears that inhibition of both activities may be required. Our results with the mutant G1-G2 substrates also suggest that abrogation of one activity has no consequence for the other activity. A spectrum of sizes of degraded aggrecan fragments is present in synovial fluids of arthritis patients. The large, high buoyant density fragments that can be recovered from synovial fluids fractionated on cesium chloride density gradients are predominantly aggrecanase-derived and do not contain any fragments with 342 FFGVG N termini (16,17). Low buoyant density fragments fractionated on cesium chloride density gradients do contain small 342 FFGVG fragments (19), and since these fragments do not carry the ITEGE 373 C terminus, it is possible they are the products of more extensive MMP processing. Studies of aggrecanolysis in chondrocyte and cartilage explant cultures show that aggrecanase is the predominant activity in vitro (35,29) (see accompanying article (55)); however, the detection (45) and quantitation (19) of 342 FFGVG fragments in human synovial fluids or released from IL-1␣-treated human OA cartilage (29) suggests that MMPs may play a greater role in aggrecanolysis in human disease than in in vitro animal models. The relative involvement of MMPs and aggrecanase in arthritis remains unclear.
The processing of ITEGE 373 G1 to DIPEN 341 G1 is likely to occur in vivo. Immunolocalization studies in mice with experimentally induced arthritis have shown that ITEGE 373 neoepitopes were less prominent in areas showing advanced cartilage damage, compared with DIPEN 341 epitope, and that when intense DIPEN 341 staining appeared, ITEGE 373 epitope disappeared (26). These results suggest that either ITEGE 373 G1 is cleared quickly from the tissue or the G1 fragment is rapidly cleaved by MMPs or other proteinases, destroying the epitope. Under conditions where the tissue pH is acidic, the increased DIPEN 341 epitope and concomitant loss of ITEGE 373 epitope could arise from the endo-and exopeptidase activity of cathepsin B (53).
The peptide experiment (Fig. 9), together with our previous observation that the 7-mer peptide IPENFFG was able to inhibit the release of 374 ARGSV fragments from cartilage in culture (51), suggests that sequences present at the C terminus of the FVDIPEN 341 G1 domain may be important for aggrecanase activity. One possibility is that these sequences provide a docking site for the enzyme. If the Pro 339 , Glu 340 , or Asn 341 residues were critical for aggrecanase docking or activity, we would expect that cleavage at the ITEGE 373 2ARGSV site may have been reduced in the DIGSA 341 and deletion mutants. This was not the case, suggesting therefore that sequences other than Pro-Glu-Asn are involved. This is consistent with our observation that a longer peptide with sequence FVDIPENFFG was effective in inhibiting aggrecanase activity by 50%, while the shorter IPENFFG peptide had no effect in the present style of experiment. An alternative explanation is that the longer sequence allows the formation of a peptide with secondary or tertiary structure and that it is the conformational shape rather than the peptide sequence that is important. This possibility could be addressed by determining whether a peptidic hydroxamate inhibitor based on the Pro-Glu-Asn residues could inhibit aggrecanase.
Three MMP cleavage site mutants were made, and these partially or totally blocked cleavage at the DIPEN 341 2FFGVG site; however, none of the mutations conferred complete protection from degradation by MMP-8 or -13, as seen in the fragmentation pattern by silver stain. The enzymes were clearly able to cleave elsewhere in the IGD, possibly at the minor sites (10,11). The 342 GTRVG mutant resisted cleavage by stromelysin-1, two gelatinases, three collagenases, matrilysin, and MT1-MMP at the mutated MMP site. The DIGSA 341 mutant was partially resistant to some of these MMPs. In the future, it will be interesting to determine whether different glycosylation affects MMP or aggrecanase specificity for cleavage in the IGD.
In contrast to the MMP cleavage site mutants, the 374 NVYSV mutant was not cleaved at all by aggrecanase, showing that aggrecanase has specificity for only a single site in the IGD. Our ongoing studies generating an 374 NVYSV knock-in mouse will enable us to further explore the role of aggrecanase in normal growth and development and also in the initiation and progression of arthritis.