The role of the C-terminal domain of human collagenase-3 (MMP-13) in the activation of procollagenase-3, substrate specificity, and tissue inhibitor of metalloproteinase interaction.

Recombinant human procollagenase-3 and a C-terminal truncated form (Δ249-451 procollagenase-3) have been stably expressed in myeloma cells and purified. The truncated proenzyme could be processed by aminophenylmercuric acetate via a short-lived intermediate form (N-terminal Leu58) to the final active form (N-terminal Tyr85). The kinetics of activation were not affected by removal of the hemopexin-like C-terminal domain. The specific activities of both collagenase-3 and Δ249-451 collagenase-3 were found to be similar using two quenched fluorescent substrates, but Δ249-451 collagenase-3 failed to cleave native triple helical collagens (types I and II) into characteristic one- and three-quarter fragments. It was noted, however, that the β1,2(I) chains of type I collagen were susceptible to Δ249-451 collagenase-3, which indicates that the catalytic domain displays telopeptidase activity, thereby generating α1,2(I) chains that are slightly shorter than those in native type I collagen. It can be concluded that the C-terminal domain is only essential for the triple helicase activity of collagenase-3. Binding of procollagenase-3 and active collagenase-3 to type I collagen is mediated by the C-terminal domain. Both collagenase-3 and Δ249-451 collagenase-3 hydrolyzed the large tenascin C isoform, fibronectin, recombinant fibronectin fragments, and type IV, IX, X, and XIV collagens; thus, these events were independent from C-terminal domain interactions. In contrast, the minor cartilage type XI collagen was resistant to cleavage. Kinetic analysis of the mechanism of inhibition of wild-type and Δ249-451 collagenase-3 by wild-type and mutant tissue inhibitors of metalloproteinase (TIMPs) revealed that the association rates for complex formation were influenced by both N- and C-terminal domain interactions. The C-terminal domain of wild-type collagenase-3 promoted increased association rates with the full-length inhibitors TIMP-1 and TIMP-3 and the hybrid N.TIMP-2/C.TIMP-1 by a factor of up to 33. In contrast, the association rates for complex formation with TIMP-2 and N.TIMP-1/C.TIMP-2 were only marginally affected by C-terminal domain interactions.

The matrix metalloproteinases (MMPs) 1 are a family of zincdependent enzymes that have the capacity to degrade most protein components of the extracellular matrix. Their uncontrolled activity contributes to the tissue destruction that is observed during such diverse pathologies as arthritis and cancer. Four main subfamilies of MMPs have been defined according to their substrate specificity, primary structures, and cellular localization: the collagenases, stromelysins, gelatinases, and membrane-type matrix metalloproteinases.
Human procollagenase-3 (MMP-13) is a new member of the collagenase subfamily of MMPs, which consists of three members showing an overall sequence homology of 55% (1)(2)(3)(4). Human collagenase-3 expression has been demonstrated in breast tumors and in osteoarthritic cartilage (1,5), indicating that the enzyme plays a role in degradation of collagen during disease progression. Furthermore, its expression in cartilage was strongly induced at both the message and protein levels by interleukin-1␣. Biochemical characterization of human collagenase-3 has shown that it is a powerful collagenolytic enzyme, preferentially cleaving type II collagen, while it is five or six times less efficient at hydrolyzing type I or III collagen (5,6). Type II collagen was initially hydrolyzed at the same peptide bond as by MMP-1, but this was then followed by a secondary cleavage event, thereby removing three amino acids from the initial N terminus of the one-quarter fragment (5). Furthermore, denatured collagens and small synthetic peptide substrates were more efficiently cleaved than by the two other homologous human collagenases, MMP-1 (fibroblast collagenase) and MMP-8 (neutrophil collagenase). In addition, we recently demonstrated that collagenase-3 cleaved human cartilage aggrecan at three sites within the interglobular G1-G2 domain (7), which indicates that this enzyme may play a role in aggrecan loss during arthritis. Analysis of the activation mechanism of human procollagenase-3 revealed that APMA, stromelysin, trypsin, plasmin, gelatinase A, and MT1-MMP were able to activate the proenzyme, which led to the removal of the complete propeptide domain in each case (6,8). It was thereby established that the conserved PRCGVPD sequence motif of the propeptide domain was responsible for the latency of the proenzyme (9), which was earlier demonstrated for the two other human collagenases, MMP-1 and MMP-8 (10 -13). Cellular activation of procollagenase-3 could also be demonstrated using concanavalin A-stimulated fibroblast monolayers, which was shown to be dependent on the activity of MT1-MMP and/or gelatinase A. This mechanism is unique within the collagenase subfamily of MMPs since MMP-1 and MMP-8 were not affected (8).
The catalytic domain of human collagenase-3, which contains the catalytic zinc-binding site, has 55% sequence similarity to the catalytic domains of human MMP-1 and MMP-8. The catalytic domains of these enzymes are all known to consist of a five-stranded ␤-sheet structure containing three ␣-helices in a typical sequential order (14 -17). The C-terminal domain of MMPs shows homology to hemopexin and vitronectin and is of particular interest in the case of the collagenases MMP-1, MMP-8, and MMP-13 since this domain is essential for their specific triple helicase (collagenolytic) activity (6, 18 -23). In the case of MMP-1 (17) and MMP-13 (24), x-ray crystallographic data have shown that this domain displays a fourbladed ␤-propeller structure that is linked via a short hinge sequence motif to the catalytic domain.
Detailed analyses of fibroblast and neutrophil collagenases in relation to their domain organization are well advanced (18,21,22), but there are currently no data available regarding the functions of the collagenase-3 domains. We have expressed full-length procollagenase-3 and a C-terminal deletion mutant (⌬ 249 -451 procollagenase-3) and assessed the role of the Cterminal domain of collagenase-3 in relation to activation of the proenzyme, substrate specificity, and TIMP binding.

EXPERIMENTAL PROCEDURES
Materials-The digoxigenin antibody labeling kit was purchased from Boehringer (Mannheim, Germany). All other chemicals were obtained from Sigma and were the purest grade available. Collagens were generously donated by Drs. M. Barnes, K. Kü hn, E. Aubert-Foucher, J.-J. Wu, and D. R. Eyre.
Generation of ⌬ 249 -451 Procollagenase-3 cDNA, Expression and Purification of the Deletion Mutant and Wild-type Procollagenase-3, and Activation of Proenzymes-The procollagenase-3 cDNA in pSP64 was cleaved with XcmI and EcoRI, which removed the complete C-terminal domain. The purified cleaved plasmid was ligated to the oligonucleotides 5Ј-TCTATGGTTAAG-3Ј and 5Ј-AATTCTTAACCATAGAG-3Ј, thereby introducing a stop codon following the codon for glycine 248, which was followed by an EcoRI site. The ⌬ 249 -451 procollagenase-3 cDNA was sequenced using the dideoxy method, which conformed its identity, except for replacement of the C-terminal coding sequences with a C-terminal stop codon. The construct was subcloned into the HindIII and EcoRI sites of the mammalian expression vector pEE12. The NaeI-linearized ⌬ 249 -451 procollagenase-3 cDNA in pEE12 (50 g) was transfected into NSO mouse myeloma cells by electroporation. Stable clones were selected through growth in glutamine-free medium essentially as described for other members of the matrix metalloproteinase family (6,18). Clones producing high amounts of ⌬ 249 -451 procollagenase-3 were grown in bulk in serum-free defined medium. Cellconditioned culture medium containing ⌬ 249 -451 procollagenase-3 was dialyzed against 20 mM Tris-HCl (pH 7.3), 5 mM CaCl 2 , and 0.05% NaN 3 (buffer A) and applied to a DEAE-Sepharose fast flow column (10 ϫ 3 cm). ⌬ 249 -451 procollagenase-3 bound to the column under these conditions and was eluted in buffer A containing 100 mM NaCl. The partially purified enzyme solution was dialyzed against buffer A and applied to an S-Sepharose fast flow column (10 ϫ 1.4 cm), followed by gradient elution (0 -0.5 M NaCl). About 4.5 mg of pure ⌬ 249 -451 procollagenase-3 were obtained from 500 ml of culture medium.
The expression and purification of wild-type procollagenase-3 have been described in detail elsewhere (6). Activation of the proenzymes was achieved by treatment with 1 mM APMA at 37°C for 30 min (6).

Construction and Expression of Chimeric TIMPs and Purification of Recombinant Wild-type and Mutant
TIMPs-An N.TIMP-1/C.TIMP-2 chimeric cDNA was prepared by polymerase chain reaction using one mutagenic oligonucleotide to the sequence VGCEE 126/128 CKITR (5Ј-GCGCGTGATCTTGCATTCCTCACAGCCAAC-3Ј) and an N.TIMP-2/ C.TIMP-1 chimeric cDNA using oligonucleotides to the sequence LN-HRYQMGCE 127/127 CTVFPCLSIPC (5Ј-GCAGGGGATGGATAAACAG-GGAAACACTGTGCACTCGCAGCCCATCTGGTACCTGTGGTTCAG-GC-3Ј and 5Ј-GCCTGAACCACAGGTACCAGATGGGCTGCGAGTG-CACAGTGTTTCCCTGTTTATCCATCCCCTGC-3Ј) in the presence of vector oligonucleotides and the respective wild-type cDNAs in pEE12 using overlap extension mutagenesis. The sequences of the constructs were confirmed using the dideoxy method. The linearized cDNAs in pEE12 were transfected into NSO mouse myeloma cells by electroporation, and stable clones were selected through growth in glutaminefree medium as described above for ⌬ 249 -451 procollagenase-3.
Enzymatic Assays-The fluorescent substrate Mca-PLGL-Dpa-AR-NH 2 was used throughout. Routine assays were performed at 25°C at a substrate concentration of 1-3 M in assay buffer containing 0.1 M Tris-HCl, 10 mM CaCl 2 , 150 mM NaCl, and 0.05% (v/v) Brij 35 (pH 7.5) (28). Under these conditions, progress curves were first-order in substrate, fulfilling the requirements [S] Ͻ Ͻ K m and allowing the direct determination of k cat /K m . Collagenolytic and gelatinolytic activities were assayed using 14 C-labeled substrates as described (29).
Proteolytic Degradation of Extracellular Matrix Components-Several extracellular matrix proteins (10 g each) such as the large and small isoforms of tenascin and recombinant fibronectin fragments as well as intact fibronectin were incubated with active wild-type collagenase-3 (100 ng) and ⌬ 249 -451 collagenase-3 (45 ng) for 24 h at 37°C, followed by analysis of the degradation products by silver-stained SDS-PAGE. Three different type IV collagen preparations (10 g each) were cleaved with 200 ng of wild-type collagenase-3 or 90 ng of ⌬ 249 -451 collagenase-3 for 16 h at 25°C, followed by analysis of the reaction products under nonreducing and reducing conditions by SDS-PAGE. Bovine tendon type XIV collagen (4.2 g) was cleaved with 400 ng of wild-type collagenase-3 or 180 ng of ⌬ 249 -451 collagenase-3 for 16 h at 25°C. Bovine or rat type IX, X, and XI collagens (10 g each) were incubated with 400 ng of wild-type collagenase-3 or 180 ng of ⌬ 249 -451 collagenase-3 for 16 h at 25°C. The reaction products were analyzed by SDS-PAGE under nonreducing (types IX and X) or reducing conditions and stained with Coomassie Blue.
Determination of the Enzyme Concentrations by Active-site Titration and Kinetic Analysis of TIMP Binding-The concentrations of the active enzymes were estimated by titration against TIMP-1 of known concentration (25). The initial rate of substrate hydrolysis (y axis) was plotted versus TIMP-1 concentration (x axis), and the active enzyme concentration was determined from the x axis intercept. The kinetics of the inhibition of active full-length collagenase-3 and active ⌬ 249 -451 collagenase-3 by wild-type and mutant TIMPs were analyzed by evaluation of the second-order rate constant (k on ) (30).
Determination of Binding of Procollagenase-3, Active Collagenase-3, ⌬ 249 -451 Procollagenase-3, and Active ⌬ 249 -451 Collagenase-3 to Type I Collagen Films-Full-length procollagenase-3 and ⌬ 249 -451 procollagenase-3 were labeled with digoxigenin using a digoxigenin-3-O-succinyl-␥-aminocaproic acid-N-hydroxysuccinimide ester according to the manufacturer's instructions (DIG antibody labeling kit, Boehringer), except that phosphate-buffered saline was replaced with 50 mM HEPES (pH 7.6), 10 mM CaCl 2 , 150 mM NaCl, and 0.04% Brij 35. It was verified that labeling had no effect on latency and enzymatic activity following APMA activation using the quenched fluorescence substrate assay. The binding of these labeled enzymes to type I collagen was determined using a modification of the enzyme-linked immunosorbent assay method described by Murphy et al. (18). Rat skin type I collagen (0.5 mg/ml in phosphate-buffered saline) was plated at 50 l/well onto a 96-well microtiter plate and allowed to form fibrils before drying at 37°C. The films were washed (50 mM Tris-HCl (pH 7.0), 150 mM NaCl, and 0.02% Tween 20) and then blocked in 100 mM Tris (pH 7.4), 150 mM NaCl, 1% bovine serum albumin, 2% nonfat dry milk, and 0.1% Triton X-100 (blocking buffer) for 20 min at room temperature. Latent or active enzymes were applied in blocking buffer (100 l/well) and incubated at 4°C for 12 h, and then unbound enzymes were removed by washing. Bound enzymes were detected and quantified using anti-digoxigenin polyclonal antibody Fab fragments conjugated to alkaline phosphatase (Boehringer; 1:2000 in blocking buffer for 1 h at 15°C) followed by p-nitrophenyl phosphate substrate. The absorbance at 405 nm was measured after stopping the reaction with NaOH.

RESULTS
Autoproteolytic Fragmentation of Active Full-length Collagenase-3-Studies of the stability of active collagenase-3 at 37°C showed that fragmentation occurred, as observed for other members of the MMP family. The enzyme was cleaved into two major fragments, a doublet of M r 29,000, which is due to N-glycosylation, and M r 27,000 (Fig. 1). These represent the catalytic domain as revealed by gelatin zymography (data not shown; doublet of M r 29,000) and the C-terminal domain (M r 27,000) of collagenase-3. N-terminal sequence determination showed that the Ser 245 -Leu 246 peptide bond had been autoproteolytically hydrolyzed, thereby releasing the C-terminal domain (Fig. 2). The initial specific collagenolytic activity of 100 g/min/nmol was lost completely, indicating that the C-terminal domain was essential for the collagenolytic activity of the enzyme.
The catalytic domain of collagenase-3 was also further processed internally, releasing 79 amino acid residues from the initial N-terminal Tyr 85 . This second fragmentation product displayed the N-terminal sequence LLAHAFPPG, which was the result of the hydrolysis of the Gly 164 -Leu 165 peptide bond (Fig. 2). During autoproteolysis, the enzymatic activity versus Mca-PLGL-Dpa-AR-NH 2 decreased to 20% of the initial value (after 16 h) and further declined upon prolonged incubation (Ͼ24 h) until no proteolytic activity was measurable, which we deduced was due to cleavage of the Gly 164 -Leu 165 peptide bond within the catalytic domain. It was therefore not possible to use the original fragmentation products for analysis of the enzymatic and TIMP binding properties of the catalytic domain of collagenase-3; hence, we produced an intact C-terminal deletion mutant (⌬ 249 -451 procollagenase-3) by protein engineering.
Purification of ⌬ 249 -451 Procollagenase-3, Activation by APMA, and Comparison of the Proteolytic Activity with Wildtype Collagenase-3 against Interstitial Collagens-⌬ 249 -451 procollagenase-3 was expressed by NSO cells and purified from the resulting conditioned medium. The recombinant deletion mutant (⌬ 249 -451 procollagenase-3) was analyzed by SDS-PAGE and shown to occur as a single band of M r 35,000 (Fig.  3A, lane 1) relative to the M r 60,000 (Fig. 3B, lane 1) for full-length procollagenase-3. The discrepancy between the calculated M r of ⌬ 249 -451 procollagenase-3 of 25,000 and the observed M r of 35,000 is due to N-glycosylation as earlier demonstrated for the full-length proenzyme (6). No proteolytic activity was detectable prior to ⌬ 249 -451 procollagenase-3 activation, indicating that the proenzyme was correctly folded. Furthermore, N-terminal sequence analysis confirmed that ⌬ 249 -451 procollagenase-3 displayed the sequence LPLPSGGD, which was earlier demonstrated for the wild-type proenzyme (6).
Activation by APMA treatment at 37°C for 30 min resulted in autoproteolytic processing of ⌬ 249 -451 procollagenase-3 and wild-type procollagenase-3 through one detectable intermediate of M r 28,000 ( . N-terminal sequence data confirmed that ⌬ 249 -451 procollagenase-3 was processed via a short-lived intermediate form displaying the sequence 58 LEVTGK that was converted to the final active form through hydrolysis of the Glu 84 -Tyr 85 peptide bond. These data confirm that ⌬ 249 -451 procollagenase-3 was processed in an identical manner as earlier described for wild-type procollagenase-3 (6). Our sequence analysis of active ⌬ 249 -451 collagenase-3 furthermore confirmed that the enzyme was N-glycosylated at Asn 98 due to a lack of a signal during amino acid sequencing.
Activity measurements using the synthetic peptide substrate Mca-PLGL-Dpa-AR-NH 2 demonstrated the rapid generation of enzymatic activity, which reached a plateau after 30 min and declined to 20% of the initial maximal activity after 16 h at 37°C (data not shown). This suggests that the secondary cleavage at Gly 164 -Leu 165 described for wild-type collagenase-3 also occurred in the case of active ⌬ 249 -451 collagenase-3. The k cat /K m values of optimally activated enzymes (30 min at 37°C) were determined after the enzyme concentrations had been determined by active-site titration using TIMP-1 and gave similar values for both enzymes (Table I) when incubated with type I or II collagen (Fig. 4A, lanes 2 and  5). It is noteworthy, however, that ⌬ 249 -451 collagenase-3 cleaved the ␤1,2(I) chains of type I collagen, generating ␣1,2(I) chains that were somewhat smaller than the ␣1,2(I) chains of the native substrate (Fig. 4B, lane 1). This indicates that the catalytic domain of collagenase-3 is an efficient telopeptidase that is not dependent on the presence of the C-terminal domain and does not require binding of the enzyme to the substrate (see below). This is a unique feature of collagenase-3 since MMP-1 and MMP-8 do not display any significant telopeptidase activity. In the case of type III collagen, partial hydrolysis (10%) by ⌬ 249 -451 collagenase-3 was observed, which generated fragments corresponding to one-and three-quarters in size (data not shown). This was due to nonspecific susceptibility of a part of the type III collagen preparation to hydrolysis since identical products were generated by gelatinase B (data not shown). However, if the collagens were heat-denatured prior to cleavage, ⌬ 249 -451 collagenase-3 degraded these with the same efficiency as the full-length enzyme (data not shown).
Collagen binding experiments revealed that labeled procollagenase-3 and active collagenase-3 bound to type I collagen films, while ⌬ 249 -451 procollagenase-3 and active ⌬ 249 -451 collagenase-3 did not bind (Fig. 5). Thus, binding was clearly promoted by the C-terminal domain of the enzyme. The binding experiments were performed over a wide range of concentrations, and saturation was reached at ϳ200 nM enzyme binding to the substrate. The assay revealed that the active enzyme bound marginally better than procollagenase-3. The binding of the proform of collagenase-3 to type I collagen was somewhat unexpected since the two other human collagenases bind only as active enzymes to type I collagen. To confirm these data, the amount of bound enzymes was also quantitated by activity measurements using the quenched fluorescence substrate Mca-PLGL-Dpa-AR-NH 2 after the proenzyme had been activated by APMA treatment (data not shown), and the results showed that both the active enzyme and the proenzyme bound with comparable efficacy. However, the amount of bound active enzyme was marginally higher compared with the proenzyme, and this was consistent over a wide range of bound enzyme concentrations (50 -400 nM[R)) (SEE FIG. 5).

Comparison of the Proteolytic Activities of Collagenase-3 and ⌬ 249 -451 Collagenase-3 Versus Different Extracellular Matrix
Components-The large TN-C isoform was very susceptible to proteolytic cleavage by both enzymes and was degraded into identical fragments displaying M r values of 190,000 and 120,000, respectively (Fig. 6, lanes 1 and 2). In contrast, the small TN-C isoform was resistant to proteolytic attack (data not shown), indicating that the major collagenase-3-sensitive sites are located within the alternatively spliced FN type III repeats. The C-terminal domain of collagenase-3 had no influence on the rate of TN-C hydrolysis, indicating that specificity is mediated by the catalytic domain alone.
Purified human plasma FN and recombinant FN fragments were hydrolyzed by both enzymes with equal efficiency (data not shown). Intact plasma FN was cleaved into four main fragments with M r values of 100,000, 43,000, 35,000, and 29,000. In contrast, the recombinant M r 120,000 FN fragment was hydrolyzed into smaller fragments with M r values of 48,000, 46,000, 35,000, 32,000, and 28,000, while the M r 110,000 FN fragment was cleaved into fragments with M r values of 100,000, 38,000, and 36,000.
Wild-type and ⌬ 249 -451 collagenase-3 were able to hydrolyze type IV collagen into fragments, and hydrolysis even proceeded at 25°C. Both the ␣2(IV) chain (M r 247,000) and the ␣1(IV) chain (M r 217,500) were cleaved as observed by analysis of the reaction products under reducing conditions. The major cleavage products displayed M r values of 240,500 and 80,000, respectively (Fig. 7). Under nonreducing conditions, a single cleavage product of M r 76,000 was visualized (data not shown). In comparison, human gelatinase A was not able to hydrolyze type IV collagen under identical conditions and cleaved only at elevated temperatures (data not shown and Ref. 31). These data indicate that collagenase-3 may play a considerable role in the dissolution of type IV collagen, a major component of the basement membrane.
Furthermore, collagenase-3 and ⌬ 249 -451 collagenase-3 cleaved native type XIV collagen with comparable kinetics (data not shown). When the cleavage products were compared with those generated by gelatinase B, a different cleavage pattern was observed. The major cleavage product generated by collagenase-3 or ⌬ 249 -451 collagenase-3 showed a M r of 165,000, being considerably smaller than the NC3 domain of type XIV collagen (Fig. 8). A minor smaller cleavage product of M r 110,000 was also generated; and furthermore, the type XIV collagen dimer was partially degraded, yielding a product that retained the reduction-resistant disulfide bridge of the NC2 domain.
The minor cartilage collagens (types IX, X, and XI) were also analyzed in cleavage experiments using collagenase-3 or ⌬ 249 -451 collagenase-3. Type X collagen was susceptible to collagenase-3 and ⌬ 249 -451 collagenase-3, generating a M r 48,000 fragment (Fig. 9). Type IX collagen was also susceptible to cleavage by collagenase-3 and ⌬ 249 -451 collagenase-3, although the differences between cleaved and noncleaved type IX collagens were only very subtle and were dependent on the species source of the substrate under investigation. The intact ␣3(IX) and ␣1(IX) chains of bovine type IX collagen were reduced from M r 175,000 to 170,000, while the low molecular weight Col1 domain, which is a result of pepsin extraction, was resistant. In contrast, rat type IX collagen showed reverse susceptibility, i.e. the high molecular weight ␣1(IX) chain was resistant, while the low molecular weight Col1 domain was cleaved (M r 52,000 compared with 53,000 for the noncleaved material). It is furthermore noteworthy that the higher molecular weight material (Ͼ175,000) in both type IX collagen preparations was cleaved (Fig. 9). In contrast, the ␣1(XI), ␣2(XI), and ␣3(XI) chains were completely resistant to both enzymes.

Kinetic Analysis of the Inhibition of Wild-type Collagenase-3 and ⌬ 249 -451 Collagenase-3 by Wild-type and Mutant TIMPs-
Active-site titrations of collagenase-3 and ⌬ 249 -451 collagenase-3 with wild-type and mutant TIMPs revealed a stoichiometry of 1:1 (data not shown). It is currently not possible to determine accurate K i values for collagenase-3 and ⌬ 249 -451 collagenase-3 interactions with wild-type and mutant TIMPs due to the tight binding nature of their interaction, with appropriate K i values well below 200 pm. This is due to the limitations in assay sensitivity, substrate solubility, and fluorescence quenching at high substrate concentrations. Furthermore, the enzymes are unstable at low concentrations, which does not allow analyses during long-term assays. It is possible, however, to determine the apparent first-order rate constant (k obs ) at low reagent concentrations (50 pM enzymes, 0.06 -2 nM inhibitors) from the analysis of the curvature in the progress of substrate cleavage. At these low concentrations, inhibition can be treated as "slow binding" as recently discussed (30). The second-order rate constant (k on ) can be calculated for the interaction of TIMPs with collagenase-3 and ⌬ 249 -451 collagenase-3; the data are summarized in Table II (30).
The data presented clearly demonstrate that the C-terminal domain of collagenase-3 contributed significantly to the binding of wild-type TIMP-1 and TIMP-3 and the chimeric inhibitor N.TIMP-2/C.TIMP-1 since the association rates of TIMP-1, TIMP-3, and N.TIMP-2/C.TIMP-1 with ⌬ 249 -451 collagenase-3 were significantly reduced (17-33 times slower). In contrast, TIMP-2, N.TIMP-1/C.TIMP-2, and ⌬ 187-194 TIMP-2 showed k on values of 0.3-1.8 ϫ 10 6 M Ϫ1 s Ϫ1 with either full-length collagenase-3 or ⌬ 249 -451 collagenase-3 and thus are not significantly affected by C-terminal domain interactions with the enzyme. The increase in the association rate of collagenase-3⅐TIMP complexes was also shown to be mediated by the Cterminal domain of the inhibitor since the k on values for the interaction of full-length collagenase-3 with ⌬ 127-184 TIMP-1 and ⌬ 128 -194 TIMP-2 were in the range of those obtained with ⌬ 249 -451 collagenase-3 and full-length TIMPs. In contrast, the C-terminal domain of TIMP-2 does not contribute significantly to complex formation between enzyme and inhibitor, which is applicable to the charged tail deletion mutant ⌬ 187-194 TIMP-2. The N-terminal domains of both enzyme and inhibitor lead to association rates of 3-8.1 ϫ 10 5 M Ϫ1 s Ϫ1 ; and it can therefore be concluded that the N-terminal domain interactions are the major contributors of complex formation between collagenase-3 and TIMPs. DISCUSSION We have determined the autoproteolytic fragmentation sites in collagenase-3 and assessed the function of the C-terminal domain of the molecule in activation, substrate specificity, and inhibitor interaction. Since collagenase-3 has been implicated in the pathologies of breast cancer and osteoarthritis, we have also examined the ability of collagenase-3 and ⌬ 249 -451 collagenase-3 to degrade different components of the extracellular matrix in order to evaluate the possible function of the enzyme in vivo.
Active collagenase-3 is not stable, and autoproteolytic cleavage was observed at the Ser 245 -Leu 246 peptide bond, which is localized at the end of the catalytic domain, thereby releasing the C-terminal domain including the complete hinge sequence motif. Similar autoproteolytic cleavages have been observed earlier for the two other human collagenases (MMP-1 and MMP-8), although these underwent fragmentation farther downstream from the catalytic domain within the hinge sequence motif (19,23). MMP-1 is hydrolyzed at the Pro 250 -Ile 251 locus and MMP-8 at the Gly 242 -Leu 243 or Pro 247 -Ile 248 locus, indicating similarities between the three collagenases, but clearly demonstrating differences in their precise autoproteolytic processing. The peptide bonds hydrolyzed by all three enzymes resemble typical collagenase cleavage sites, and autoproteolytic fragmentation could be inhibited by complex for-  The catalytic domain of collagenase-3 generated by autoproteolytic fragmentation was itself unstable, and further autoproteolysis was observed by cleavage of the Gly 164 -Leu 165 peptide bond, which results in a decrease in the peptidolytic activity to 20 or 0% of the initial value, which was dependent on the incubation time employed. Similar processing has been observed in the case of gelatinase A, although it is not clear whether the activity of the enzyme was affected (32). We have earlier drawn attention to the similarities between collagenase-3 and the gelatinases, and the similarities in autoproteolytic fragmentation of the active enzymes underline our earlier results (6).
To assess the function of the C-terminal domain of collagenase-3, we prepared a recombinant form lacking this domain. Comparison of the activation of wild-type and ⌬ 249 -451 procollagenase-3 by APMA revealed that the corresponding propeptides were processed in an indistinguishable way. In addition, activation by the catalytic domain of MT1-MMP was observed for both full-length and ⌬ 249 -451 procollagenase-3, indicating that these events were not influenced by the C-terminal domain (data not shown). We have previously shown that progelatinase A and a C-terminal deletion mutant were equally well processed by the catalytic domain of MT1-MMP in vitro, indicating that the biochemistry of the reaction is also not influenced by C-terminal domain interactions in solution at high concentrations (33).
We have extended our previous analysis of the substrate specificity of collagenase-3 (6) using a wide range of extracellular matrix proteins and compared the full-length enzyme with the C-terminal deletion mutant in order to determine the role of the C-terminal domain in substrate specificity. Comparison of the enzymatic activities of full-length and ⌬ 249 -451 collagenase-3 revealed that the collagenolytic activity (triple helicase activity) versus interstitial collagens was mediated by the C-terminal domain, while the gelatinolytic and peptidolytic activities were unchanged. These results confirm earlier observations that the triple helicase activities of the two other human collagenases are dependent on the C-terminal domain (18,19,23). We have furthermore established that human collagenase-3 is an efficient telopeptidase and demonstrated that this activity is dependent on its catalytic domain and independent from C-terminal domain interactions. This feature distinguishes collagenase-3 from the two other human collagenases (MMP-1 and MMP-8) since these show no detectable telopeptidase activity. However, an N-telopeptidase activity has been described for the highly homologous rat collagenase (34) using a mutant type I collagen that cannot be cleaved into one-and three-quarter fragments. It was suggested that the N-telopeptidase activity might be sufficient for resorption of type I collagen during embryonic and early adult life, while triple helicase activity is necessary during intense tissue resorption such as observed in the postpartum uterus and in the dermis later in life (34). The telopeptidase activity of collagenase-3 does not require binding of the enzyme to the substrate since ⌬ 249 -451 collagenase-3 does not bind to triple helical type I collagen (see below), and this activity might be important during bone resorption and cartilage turnover since collagenase-3 is expressed at high levels in these tissues in the human.
We have also demonstrated that the C-terminal domain promotes binding of both procollagenase-3 and active collagenase-3 to type I collagen. Procollagenase-3 and active collagenase-3 bound nearly equally well to triple helical type I collagen, with the amount of active enzyme being marginally higher. Our data are in agreement with earlier data published on the highly homologous mouse collagenase-3 (35). In this case, the proen-zyme could be eluted from a type I collagen-Sepharose column at lower salt concentrations than the active form (35), which might indicate that the interaction between the active enzyme and substrate is tighter and in part of ionic nature. The association and dissociation rates for proenzyme and active enzyme binding to triple helical collagen await further detailed analysis. Binding is essential for the triple helicase activity of active collagenase-3, but is not necessary for the telopeptidase activity of the enzyme. In contrast, the two other human collagenases bind only as active enzymes to collagen, while the latent counterparts do not bind to the triple helical substrates (18,19). From binding data of chimeric N-terminal stromelysin-C-terminal collagenase and N-terminal collagenase-C-terminal stromelysin molecules to fibrillar type I collagen, it is known that binding is promoted by the C-terminal domain of both collagenase and stromelysin (18). However, stromelysin and the chimeric N-terminal collagenase-C-terminal stromelysin mutant also bound as proenzymes, which indicates that the binding motif within the respective C-terminal domain is unmasked in the respective proenzymes, which is also true for procollagenase-3. Furthermore, active stromelysin and the active chimeric proteinases were not able to cleave the triple helical substrates. From further data using a mutant neutrophil collagenase that contained the hinge sequence motif of stromelysin, which is nine amino acid residues longer than the collagenase hinge sequence, it became clear that collagenolysis only proceeds if the connecting "hinge" between the catalytic and C-terminal domains contains the correct number of amino acid residues (22). Thus, the capacity of the collagenases to cleave triple helical substrates obviously depends on the correct interplay between catalytic and C-terminal domains, which appear to be separately folded entities. Our recent x-ray crystallographic analysis of the C-terminal domain of collagenase-3 revealed that the overall structure of this domain shows more similarity to the C-terminal domain of fibroblast collagenase than to gelatinase A, which indicates that those structural features important for efficient triple helicase activity are conserved within the collagenases (24). The triple helicase activity of the collagenases might be coordinated by the hinge sequence motif, but it is not clear how this is mediated specifically. We have noted that none of the residues conserved within the C-terminal domains of the three human collagenases is unique to this subfamily. Recently, Bode (36) suggested that triple helical collagen might be bound between the catalytic and C-terminal domains in such a way that the substrate is bound like a waffle in a waffle iron. The triple helical substrate could be destabilized by this interaction, causing unwinding or relaxing around the cleavable bond. The unwound "single" strands would then be able to fit into the active-site cleft of the molecule, where hydrolysis of each strand would proceed. This hypothesis has been underlined by modeling experiments performed by De Souza et al. (37).
The analysis of the cleavage of the large TN-C isoform, FN, FN fragments, and type IV, IX, X, and XIV collagens by wildtype collagenase-3 and ⌬ 249 -451 collagenase-3 revealed that these substrates were equally well hydrolyzed, indicating that the C-terminal domain had no affect on specificity. The cleavage products of the large TN-C isoform revealed fragment sizes that were identical to those recently identified for gelatinase A hydrolysis of this extracellular matrix component (38). This indicates that collagenase-3 shows very similar specificity versus large TN-C, with the major cleavage sites being located within the alternatively spliced FN type III repeats, which is confirmed by the resistance of the small TN-C isoform to proteolysis. The increased sensitivity of the large TN-C isoform to degradation could modify its biological activities by unmasking or abolishing specific functional sites. Thus, a more suitable environment for cellular proliferation and migration may well be established.
Interestingly, collagenase-3 and ⌬ 249 -451 collagenase-3 were able to hydrolyze type IV collagen at 25°C, and these data strongly indicate that collagenase-3 is an efficient type IV collagenolytic enzyme. This observation may have important implications for the pathophysiological role of collagenase-3 during breast cancer pathology, allowing rapid dissolution of the basement membrane, leading to increased ability of tumor cells to extravasate and metastasize. Taking into account that gelatinase A cleaves type IV collagen only at elevated temperatures (Ͼ Ͼ25°C) (31), our findings might indeed represent important data for the proteolytic turnover of type IV collagen in vivo.
Type XIV collagen consists of two triple helical domains (Col1 and Col2) interspersed by non-triple helical domains (NC1, NC2, and NC3). Degradation by collagenase-3 and ⌬ 249 -451 collagenase-3 leads mainly to the accumulation of two fragments of M r 165,000 and 110,000, with minor products that retain the reduction-resistant disulfide bridge of the NC2 domain. In contrast, gelatinase B, which initially attacks the Col1 domain in a region of helix imperfections, mainly generates the NC3 domain (39).
The minor cartilage collagens (types IX, X, and XI) showed different susceptibility to collagenase-3 and ⌬ 249 -451 collagenase-3. While type XI was resistant to hydrolysis, types IX and X were cleaved. Within intact cartilage, type IX and XI collagens are cross-linked in a very specific manner to type II collagen either via lysyl oxidase-generated aldehyde residues or via pyridinoline-cross-linking residues; therefore,stabilizing the tissue and degradation of these minor collagens may be crucial in arthritic disease. Until now, it was thought that stromelysin-1 (MMP-3) was the only matrix metalloproteinase able to degrade type IX collagen into two main triple helical fragments, Col1 and Col2,3, with the cleavage sites being located within the NC2 domain of all three chains (40,41). Our results suggest that collagenase-3 and ⌬ 249 -451 collagenase-3 act as "telopeptidases" on type IX collagen as well as on type I collagen (see above) and type II collagen. 2 Taking into account that collagenase-3 is expressed at high levels in arthritic cartilage and the high specific activity of collagenase-3 on type II collagen, it may indeed be concluded that collagenase-3 contributes considerably to cartilage damage during arthritis. This is furthermore underlined by the fact that aggrecan is cleaved with four times the efficiency of stromelysin-1, which demonstrates that collagenase-3 can hydrolyze the two major cartilage proteins very effectively. 3 The assessment of the association rates for complex formation of full-length and ⌬ 249 -451 collagenase-3 with wild-type and mutant TIMPs revealed that these were affected by N-and C-terminal domain interactions. C-terminal domain interactions were most pronounced in the case of the full-length inhibitors TIMP-3, TIMP-1, and N.TIMP-2/C.TIMP-1, which reacted up to 33 times faster with full-length collagenase-3 than with ⌬ 249 -451 collagenase-3. We have recently reported that the association rate constants for complex formation between wildtype gelatinase A and TIMP-1 were increased by a factor of 130 relative to the interaction with a C-terminal deletion mutant of gelatinase A (30). The importance of C-terminal domain interactions in complex formation was underlined using ⌬ 127-194 TIMP-1, which reacted much more slowly than wild-type TIMP-1 with wild-type collagenase-3. These results show re-markable similarities to binding data obtained with wild-type TIMP-1 and gelatinase B (42), which were strongly affected by C-terminal domain interactions. However, the association rates between gelatinase B and wild-type TIMP-1 revealed a biphasic mechanism containing a fast and slow binding phase. In the case of complex formation of TIMP-1 with gelatinase B, the C-terminal domain of TIMP-1 increased the association rate by a factor of 132. Inhibition of a gelatinase B C-terminal deletion mutant (⌬ 426 -688 gelatinase B) by wild-type TIMP-1 revealed that the fast binding interaction was significantly compromised, and in addition, the association rate constant for the slow binding phase was further decreased by a factor of 17.
In contrast, the association rates for complex formation of TIMP-2 and N.TIMP-1/C.TIMP-2 with collagenase-3 and ⌬ 249 -451 collagenase-3 were only marginally affected by the C-terminal domain of TIMP-2 or collagenase-3, which has been earlier demonstrated for the interaction between gelatinase B and TIMP-2 (42). This is in marked contrast to the results described earlier for the interaction between gelatinase A and TIMP-2, where C-terminal domain interactions increase the association rates for complex formation by a factor of 187 (30). In addition, the charged tail of TIMP-2 (⌬ 187-194 TIMP-2) contributed significantly to the rate of complex formation with gelatinase A, which is not observed during complex formation with collagenase-3. It has been shown that the C-terminal domain and especially the charged tail of TIMP-2 (residues 187-194) are responsible for the formation of a progelatinase A⅐TIMP-2 complex (30), and this interaction is thought to be vital in the cellular activation of progelatinase A by MT-MMPs. We have recently described the activation of procollagenase-3 in the same system and can only conclude that TIMP-2 is unlikely to be involved in the interaction of procollagenase-3 with the cell surface during activation.