Biochemical Characterization of Human Collagenase-3*

The cDNA of a novel matrix metalloproteinase, colla- genase-3 (MMP-13) has been isolated from a breast tumor library M. M., M., J., and C. J. Biol. Chem. 269, 16766–16773), and a potential role in tumor progression has been proposed for this enzyme. In order to establish the possible role of colla- genase-3 in connective tissue turnover, we have expressed and purified recombinant human procolla- genase-3 and characterized the enzyme biochemically. The purified procollagenase-3 was shown to be glycosyl- ated and displayed a M r of 60,000, the N-terminal sequence being LPLPSGGD, which is consistent with the cDNA-predicted sequence. The proenzyme was activated by p -aminophenylmercuric acetate or stromely- sin, yielding an intermediate form of M r 50,000, which displayed the N-terminal sequence L 58 EVTGK. Further processing resulted in cleavage of the Glu 84 –Tyr 85 peptide bond to the final active enzyme ( M r 48,000). Trypsin activation of procollagenase-3 also 2 hydrolysis (Willen- The K i for both were determined using the following equations: k off (cid:53) k obs v s / v 0 and K i (cid:53) k off / k on . N-terminal Amino Acid Sequencing— N-terminal sequence determi-nations of purified procollagenase-3 or active collagenase-3 were per- formed by automated Edman degradation using an Applied Biosystems 470A protein sequencer with on-line 190A HPLC for phenylthiohydan- toin-derivative analysis.

The human matrix metalloproteinases (MMPs) 1 comprise a family of at least 11 homologous zinc-dependent endopeptidases that degrade the macromolecular components of extracellular matrices. They have been implicated in matrix remodeling processes associated with normal mammalian development and growth and in the degradative processes accompanying arthritis and tumor invasion. The MMPs can be divided into three main subfamilies, collagenases, stromelysins, and gelatinases, and other enzymes that do not belong to these groupings. Three highly homologous human collagenases, fibroblast (MMP-1), neutrophil , and collagenase-3 (MMP-13) have been identified by analysis of their respective cDNAs (Goldberg et al., 1986;Whitham et al., 1986;Hasty et al., 1990;Freije et al., 1994). Sequence comparison revealed that they share more than 50% sequence identity and three functionally important domains, namely the propeptide, catalytic, and C-terminal domains. Procollagenase latency is due to the propeptide domain, which consists of about 80 amino acids including a free cysteine residue within the highly conserved PRCGVPD sequence motif. The catalytic domain of about 180 amino acids contains two or one calcium and two zinc binding sites as revealed by x-ray crystallographic analysis of the catalytic domains of fibroblast and neutrophil collagenases in the presence of synthetic inhibitors (Borkakoti et al., 1994;Bode et al., 1994;Lovejoy et al., 1994). The structure comprises a five-stranded ␤-sheet, two bridging loops, and two ␣-helices. The C-terminal domain is linked via a short hinge sequence motif to the catalytic domain and shares sequence homology with vitronectin, being essential for the triple helicase activity of fibroblast and neutrophil collagenases Clark and Cawston, 1989;Sanchez-Lopez et al., 1993;Hirose et al., 1993;Knä uper et al., 1993a). The active enzymes form tight binding noncovalent complexes with their natural inhibitors, referred to as tissue inhibitors of metalloproteinases (TIMPs), in a 1:1 stoichiometric fashion. The interaction of the collagenases with TIMPs is mainly regulated by the catalytic domain , but C-terminal domain interactions increase the association rates of complex formation.
Biochemical studies on fibroblast and neutrophil collagenases describing their activation mechanism, substrate specificity, and inhibitor interaction in relation to their domain organization are well advanced (Murphy et al., 1987Clark and Cawston, 1989;Hirose et al., 1993;Sanchez-Lopez et al., 1993;Knä uper et al., 1990aKnä uper et al., , 1990bKnä uper et al., , 1993aKnä uper et al., , 1993b, but there are currently no data available regarding the activation * This work was supported in part by the Arthritis and Rheumatism Council, United Kingdom, by a Wellcome Trust Travelling Fellowship (to V. K.) and by Comision Interministerial de Ciencia y Tecnologia Spain Project . The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Activation of Procollagenases and Determination of Their Concentration-Procollagenase-3 and neutrophil procollagenase were routinely activated by incubation with 1 mM APMA at 37°C. Procollagenase-3 (20 g) was activated with either 1.4 or 2.8 g of active stromelysin at 37°C for up to 7 h. In order to achieve "superactivation" of both neutrophil and fibroblast procollagenase (purified according to Knä uper et al. (1990a) and Murphy et al. (1992)), the enzymes were activated by combined treatment with trypsin and stromelysin (Knä uper et al., 1993b;Murphy et al., 1992). The concentrations of the three active collagenases were determined by titration against the synthetic hydroxamic acid based metalloproteinase inhibitors CT1399 and CT1847 (kindly provided by Celltech Ltd., Slough, United Kingdom).
Activity Assays-The specific activities of the active collagenases were determined using 14 C-labeled type I collagen in a diffuse fibril assay at 35°C essentially as described by Cawston et al. (1981). Correspondingly, the gelatinolytic activity was determined at 37°C using [ 14 C]gelatin as the substrate (Cawston et al., 1981). The degradation of acid-soluble type II and III collagen was quantitated using the gel scanning protocol described by Welgus et al. (1981). The activity of collagenase-3 versus the synthetic quenched fluorescent peptide substrates Mca-Pro-Leu-Gly-Leu-Dpa-Ala-Arg-NH 2 and Mca-Pro-Cha-Gly-Nva-His-Ala-Dpa-NH 2 was assessed using a Perkin Elmer spectrofluorometer (LS 50B) (Knight et al., 1992). The three serpins, ␣ 1antichymotrypsin, antithrombin III, and plasminogen activator inhibitor 2 (5 g each) were incubated with 86 ng of active collagenase-3 at 37°C for 5 h. The reaction was terminated by the addition of reducing sample buffer prior to electrophoretic analysis. In the case of ␣ 1antichymotrypsin, the reaction products were purified by reverse-phase HPLC on a Vydac 218TP54 column (4.6 ϫ 250 mm) at a constant flow rate of 1 ml/min using a linear gradient of 5-95% acetonitrile and further analyzed by amino acid sequencing.
Purification of Recombinant Tissue Inhibitors of Metalloproteinases TIMP-1, TIMP-2, and TIMP-3 and Determination of Their Concentrations-Recombinant forms of the three human TIMPs were purified from the relevant transfected NSO mouse myeloma clones (Murphy et al., 1991;Willenbrock et al., 1993;Apte et al., 1995). TIMP concentrations were determined by titration against human recombinant stromelysin, the concentration of which had been determined by titration with a standard preparation of TIMP-1 (concentration determined by amino acid analysis) (Murphy and Willenbrock, 1995).
Kinetic Studies Using Synthetic Hydroxamic Acid-based Inhibitors CT1399 and CT1847-The rate constants for the association of 50 pM active collagenase-3 with the synthetic inhibitors CT1847 (2-8 nM) and CT1399 (100 -300 pM) were determined by analysis of the progress curves of Mca-Pro-Leu-Gly-Leu-Dpa-Ala-Arg-NH 2 hydrolysis (Willenbrock et al., 1993). The apparent K i values for both inhibitors were determined using the following equations: k off ϭ k obs (v s /v 0 ) and K i ϭ k off /k on .
N-terminal Amino Acid Sequencing-N-terminal sequence determinations of purified procollagenase-3 or active collagenase-3 were performed by automated Edman degradation using an Applied Biosystems 470A protein sequencer with on-line 190A HPLC for phenylthiohydantoin-derivative analysis.

Expression and Purification of Human Procollagenase-3-
Human procollagenase-3 was expressed by stable transfected NSO mouse myeloma cells and purified using S-Sepharose fast flow and Sephacryl S-200. The procollagenase-3 preparation was free of other matrix metalloproteinases as assessed by gelatin and casein zymographic analysis (results not shown). The final purified procollagenase-3 displayed a M r of 60,000 when analyzed by SDS-PAGE under reducing conditions (Fig. 1, lane 1). The proenzyme was shown to be glycosylated as demonstrated by N-glycosidase F treatment (Fig. 1, lane 2). This reduced the M r to 53,600, which is in excellent agreement with the M r predicted from the cDNA sequence. Thus 10% of the M r of the proenzyme corresponds to N-linked sugars. N-terminal amino acid sequencing of procollagenase-3 revealed the sequence LPLPSGGD, which is consistent with the cDNA predicted sequence (Fig. 2). A minor portion of the secreted procollagenase-3 displayed the N-terminal amino acid sequence PLPSGGD. The loss of Leu 1 in a part of the enzyme preparation may be due to the activity of a leucine aminopeptidase produced by the NSO mouse myeloma cells during the average culture period of 2 weeks. The procollagenase-3 preparation was Ͼ98% latent and displayed barely detectable levels of enzymatic activity prior to activation, and it can be concluded that the loss of Leu 1 did not effect the latency of the proenzyme.
In contrast, autoactivated collagenase-3 displayed a M r of 48,000 when analyzed by SDS-PAGE, and its proteolytic activity could not be enhanced by APMA treatment. N-terminal amino acid analysis revealed the sequence YNVFPRTLKWSK-MXL demonstrating the complete loss of the propeptide domain and assigning Tyr 85 as the first amino acid of the active enzyme. The Asn 98 residue was clearly glycosylated due to the lack of a signal during amino acid sequencing.
Activation of Procollagenase-3 by APMA-Purified procollagenase-3 was activated by treatment with 1 mM APMA in a time-dependent fashion (Fig. 3A). The activity generated was monitored using Mca-Pro-Leu-Gly-Leu-Dpa-Ala-Arg-NH 2 described by Knight et al. (1992). Full activation was achieved after a time interval of 30 min. Parallel analysis of the M r of the enzyme by SDS-PAGE revealed that the M r of the proenzyme was reduced through at least one short-lived intermediate form (M r 50,000) to the final active collagenase-3 displaying a M r of 48,000 (Fig. 1, lanes 3-5). Preincubation of procollagenase-3 with a two molar excess of recombinant TIMP-1 prior to APMA activation prevented the formation of the low M r final active enzyme form (M r 48,000), clearly demonstrating that this process was autoproteolytic (Fig. 1, lane 7). Under these conditions, two intermediate enzyme forms were demonstrated displaying apparent M r values of 56,000 and 50,000. The M r 56,000 intermediate was not detectable in the absence of TIMP-1, which might indicate that it was extremely unstable being rapidly converted to the M r 50,000 species. This might indicate that procollagenase-3 activation by APMA is a three-step process.
N-terminal amino acid sequencing showed the initial generation of the sequence LEVTGKL after a 4-min activation of procollagenase-3 by APMA, which is due to the cleavage of the Gly 57 -Leu 58 peptide bond (Fig. 2). During the progress of activation, the initial intermediate form (M r 50,000) was converted to the final fully active enzyme by hydrolysis of the Glu 84 -Tyr 85 peptide bond leading to the release of the complete propeptide domain (Fig. 2).
Activation of Procollagenase-3 by Stromelysin-1-The ability of stromelysin to activate procollagenase-3 was monitored by following the increase in the rate of hydrolysis of Mca-Pro-Leu-Gly-Dpa-Ala-Arg-NH 2 and the loss of the procollagenase-3 propeptide by SDS-PAGE ( Fig. 3B and Fig. 1, lanes 9 -11). At the concentration of procollagenase-3 tested, there was no detectable activation by autoproteolysis, but the addition of trypsin-activated stromelysin led to an increase in collagenase-3 activity. Activation of procollagenase-3 by stromelysin was dependent on the stromelysin concentration employed during the incubation period (Fig. 3B). Analysis of the reaction products by SDS-PAGE revealed that the proenzyme was converted to a doublet of M r 50,000 and 48,000, respectively (Fig. 1, lane 10). N-terminal sequence analysis of the stromelysin-activated collagenase-3 showed that the Gly 57 -Leu 58 and Glu 84 -Tyr 85 peptide bonds were cleaved. Thus, stromelysin activation of procollagenase-3 proceeds in a two-step mechanism via intermediates also obtained during APMA-induced autoproteolytic activation (Fig. 2).
Activation of Procollagenase-3 by TPCK-treated Trypsin-Procollagenase-3 was rapidly activated by TPCK-trypsin through a very transient intermediate form of M r 52,000, which was only faintly visible after mixing of the proenzyme with trypsin during the start of the reaction. During the activation progress, the proenzyme was converted to the active collagenase-3 displaying a molecular mass of 48,000, which showed the new N terminus Tyr 85 as a result of the hydrolysis of the Glu 84 -Tyr 85 peptide bond. Thus it was not the result of tryptic cleavage that should occur after Lys or Arg residues in position P 1 . It has therefore to be concluded that the initial tryptic cleavage in the propeptide domain lead to the autoproteolytic loss of the rest of the propeptide by an autoproteolytic event. Furthermore, the active enzyme was not stable in the presence of TPCK-treated trypsin and was further hydrolyzed into smaller sized fragments, which might be due to tryptic cleavage at the Lys 257 -His 258 or Lys 260 -Thr 261 peptide bonds within the hinge region of collagenase-3. Identical results were obtained when a mixture of TPCK-treated trypsin, and stromelysin was used to activate procollagenase-3. As the collagenolytic activity of collagenase-3 is dependent on the presence of the C-terminal domain, 2 we did not determine the specific activity of trypsin activated collagenase-3 since high amounts of the catalytic and C-terminal domain were present in the reaction mixture even after only short incubation times (Fig. 1,  lane 13). In addition, the activity versus the peptide substrate declined during prolonged incubation such that after 3 h only 60% of the initial maximal activity was retained, while the active enzyme was completely converted to the catalytic and C-terminal domain respectively (Fig. 1, lane 14).
Determination of the Substrate Specificity of Collagenase-3: Physiologically Relevant Substrates-Active collagenase-3 degraded the interstitial collagens (I, II, III) at 25°C into typical 3/4 and 1/4 fragments. Collagenase-3 cleaved type II collagen about 5 times faster than type I and 6 times faster than type III collagen. Attempts to quantitate the cleavage of soluble type I, II, and III collagen using the SDS-gel scanning protocol (Welgus et al., 1981) were performed at varying enzyme to substrate ratios, but it proved impossible to establish linearity. In addition, we found that the 3/4 fragments were stained more efficiently than intact collagen, which made it impossible to accurately quantify collagenolysis. Quantitative comparison of the activity of collagenase-3 relative to those of fibroblast or neutrophil collagenase was therefore only possible by generating data simultaneously using the 14 C-labeled type I collagen diffuse fibril assay, and these are summarized in Table I. Collagenase-3 displayed a specific activity of 100 g/min/nmol, which was comparable with those values obtained for "super-2 V. Knä uper and G. Murphy, unpublished results. activated" fibroblast or APMA-activated neutrophil collagenase. In contrast, "superactivated" neutrophil collagenase was 3 times as active and can be assigned as the most efficient type I collagenolytic enzyme in the human.
The gelatinolytic activity of collagenase-3 and its homologous counterparts were determined using [ 14 C]gelatin (Table I). Collagenase-3 displayed the highest specific activity, 90.7 g/min/ nmol, respectively. Thus the enzyme was 44 times more efficient than fibroblast and 3-8 times better than neutrophil collagenase.
The rapid proteolytic degradation of two different serpins (␣ 1 -antichymotrypsin and plasminogen activator inhibitor 2) by highly purified active collagenase-3 was demonstrated by SDS-PAGE, while antithrombin III was resistant to degradation (not shown). Further analysis of the ␣ 1 -antichymotrypsin cleavage products by N-terminal amino acid sequence determination revealed that collagenase-3 hydrolyzed the Ala 362 -Leu 363 peptide bond within the extended reactive site loop of the serpin, two amino acid residues downstream from the reactive site center. The cleavage of the Ala 362 -Leu 363 peptide bond of ␣ 1 -antichymotrypsin coincides with its inactivation as recently demonstrated by Mast et al. (1991) for collagenase (MMP-1) and stromelysin (MMP-3).
Quenched Fluorescent Peptide Substrates-Active collagenase-3 cleaved the peptide substrates Mca-Pro-Leu-Gly-Leu-Dpa-Ala-Arg-NH 2 and Mca-Pro-Cha-Gly-Nva-His-Ala-Dpa-NH 2 at the Gly-Leu and Gly-Nva peptide bonds as revealed by amino acid analysis of the HPLC-purified reaction products. Active site titrations of fully APMA-activated collagenase-3 were performed using the synthetic inhibitor CT1399 to determine the enzyme concentration. The initial rate of substrate hydrolysis showed linear dependence on substrate concentration in the concentration range 0.7-8 M, demonstrating that K m Ϸ 8 M. At substrate concentrations greater than 8 M, estimates could not be made due to insolubility of the sub-strates and absorptive quenching. Therefore, individual values of k cat and K m could not be determined. The values of k cat /K m for the hydrolysis of both substrates were estimated at substrate concentrations of 0.7 and 1.4 M, which fulfilled the requirements of [S] Ͻ Ͻ K m allowing direct determination of k cat /K m . Simultaneously, k cat /K m values for fibroblast and neutrophil collagenase were determined under identical conditions and compared with the values obtained for collagenase-3 (Table II). Collagenase-3 hydrolyzed both synthetic peptide substrates 70 -100 or 7-10 times more efficiently than fibroblast or neutrophil collagenase. Thus collagenase-3 is the most potent peptidolytic enzyme of all three homologous collagenases.
Inhibition of Active Collagenase-3 by TIMPs-Inhibition studies of active collagenase-3 with TIMP-1, TIMP-2, and TIMP-3 were performed by preincubation of collagenase-3 with TIMP concentrations (determined by active site titration with active stromelysin) up to 2 times the enzyme concentration (determined by active site titration with CT1399) using 2-h preincubations. Residual enzymic activities were determined by hydrolysis of Mca-Pro-Leu-Gly-Leu-Dpa-Ala-Arg-NH 2 and plotted versus TIMP concentration. Analysis of the data revealed that all three TIMPs inhibited the enzyme in a 1:1 stoichiometric fashion (Fig. 4). Initial kinetic analysis of collagenase-3 TIMP interaction demonstrated that TIMP-1 showed association rate constants in the region of ϳ8 ϫ 10 6 M Ϫ1 s Ϫ1 and TIMP-3 ϳ10 ϫ 10 6 M Ϫ1 s Ϫ1 , while the value for TIMP-2 was ϳ1.8 ϫ 10 6 M Ϫ1 s Ϫ1 . 3 Thus TIMP-3 reacted 1.2 times faster than TIMP-1 and 5.5 times faster than TIMP-2.

Inhibition of Active Collagenase-3 by Hydroxamic Acid-based Inhibitors and Kinetic Analysis of Their Interaction-
The collagenase-3 concentration was determined by active site titration using the synthetic hydroxamic acid-based peptide inhibitors CT1399 and CT1847. These inhibitors are competitive and react with 1:1 stoichiometry as revealed from x-ray crystallographic analyses of structurally related inhibitors with the catalytic domains of fibroblast and neutrophil collagenase (Lovejoy et al., 1994;Borkakoti et al., 1994;Bode et al., 1994). Apparent k on values for their interaction with collagenase-3 were determined as described under "Experimental Procedures." The enzyme (50 pM) was added to the reaction mixture containing 0.7 M substrate and 100 -300 pM CT1399 or 2-8 nM CT1847. Inhibition was observed as curvature in the progress of substrate hydrolysis and analyzed according to Willenbrock et al. (1993). Equivalent assays in the absence of inhibitor FIG. 3. A, activation of procollagenase-3 by APMA. Procollagenase-3 was incubated at a concentration of 626 nM in the presence of 1 mM APMA at 37°C. At the indicated time points, aliquots were removed and assayed using Mca-Pro-Leu-Gly-Leu-Dpa-Ala-Arg-NH 2 . Results are presented as rates of substrate hydrolysis (M Ϫ1 s Ϫ1 ). B, activation of procollagenase-3 by stromelysin. Procollagenase-3 was incubated with 1.4 or 2.8 g of active stromelysin at 37°C. At the indicated time intervals, aliquots were removed and assayed for activity using Mca-Pro-Leu-Gly-Leu-Dpa-Ala-Arg-NH 2 . Results are presented as rates of substrate hydrolysis (M Ϫ1 s Ϫ1 ). f, procollagenase-3 activated by 1.4 g of stromelysin; q, procollagenase-3 activated by 2.8 g of stromelysin; Ç, procollagenase-3 in the presence of buffer.

TABLE I Comparison of the collagenolytic and gelatinolytic activities of collagenase-3 (MMP-13), fibroblast collagenase (MMP-1), and neutrophil collagenase (MMP-8)
The activated collagenases were incubated with 14 C-labeled type I collagen in a diffuse fibril assay at 35°C or with [ 14 C]gelatin at 37°C, essentially as described by Cawston et al. (1981). revealed that curvature was due only to inhibition by CT1399 or CT1847 and was not due to enzyme instability or substrate depletion. The initial velocities v 0 were independent of inhibitor concentration, and k obs showed linear dependence on inhibitor concentration. Thus it can be concluded that inhibition of collagenase-3 by CT1399 and CT1847 proceeds via a simple bimolecular collision. The second order rate constants k on were in the range of 1.4 ϫ 10 6 M Ϫ1 s Ϫ1 for CT1847 and 17.0 ϫ 10 6 M Ϫ1 s Ϫ1 for CT1399. This revealed that CT1399 reacted 12.1 times faster than CT1847. CT1399 showed an apparent K i value of 4 pM and CT1847 a value of 540 pM. The K i value for CT1399 of 4 pM can be regarded only as an upper estimate, since analysis at enzyme concentrations below K i could not be performed due to the limitations in assay sensitivity and the lack of enzyme stability at these low concentrations.

DISCUSSION
Human collagenase-3 is a novel member of the matrix metalloproteinase superfamily and has been cloned from a breast tumor cDNA library (Freije et al., 1994). The enzyme is expressed in the surrounding endothelia of the tumor and may be involved in tumor progression and metastasis. Consequently, biochemical analysis of the activation mechanism, substrate specificity, and inhibition profile of collagenase-3 is of vital importance in order to understand its possible role in vivo. We have, therefore, expressed and purified recombinant human procollagenase-3 and analyzed its biochemical properties in detail and compared these with the homologous human collagenases and gelatinase A.
Procollagenase-3 showed a high degree of N-linked glycosylation as demonstrated by enzymatic deglycosylation (11.7% of its M r corresponds to N-linked sugars). Amino acid sequencing revealed a lack of signal for the Asn 98 residue, thus it can be deduced that the glycosylation site N 98 LT carries N-linked sugars. This glycosylation site is conserved between collagenase-3, neutrophil collagenase (Knä uper et al., 1990b), and gelatinase-B and is occupied in all three enzymes. The role of the high levels of glycosylation observed for these three enzymes is not quite clear to date. It has been speculated that glycosylation of neutrophil collagenase and gelatinase-B might be important for targeting these enzymes to the specific granules of neutrophils, where they are stored prior to exocytosis. However, in the case of collagenase-3 it is not clear where the enzyme might be produced in vivo and why it carries a relatively high amount of N-linked sugars. It is most unlikely that the glycosylation will cause any changes in the enzymatic properties, activation, or TIMP interaction of collagenase-3, since studies on the natural and recombinant catalytic domain of neutrophil collagenase have shown that the unglycosylated recombinant protein has indistinguishable enzymatic properties (Knä uper et al., 1993a;Schnierer et al., 1993).
Activation of matrix metalloproteinases is one of the control mechanisms regulating extracellular connective tissue turnover. We have therefore studied the mechanisms leading to procollagenase-3 activation. Stromelysin activated procollagenase-3 by a two-step mechanism, which is similar to that observed for gelatinase-B (Shapiro et al., 1995;Ogata et al., 1992). In addition, neutrophil procollagenase was activated by stromelysin by a single-step mechanism (Knä uper et al., 1993b), while the fibroblast procollagenase cannot be directly activated by stromelysin (Murphy et al., 1987;Suzuki et al., 1990). The peptide bonds cleaved within procollagenase-3, neutrophil procollagenase and progelatinase-B seem to be readily accessible to stromelysin, while fibroblast procollagenase is resistant until proteolysis of upstream regions of the propeptide have been affected by combined trypsin-stromelysin treatment leading to "superactivation" (Murphy et al., 1987;Suzuki et al., 1990). In contrast, procollagenase-3 was very susceptible to either trypsin alone or trypsin in combination with stromelysin, which lead to the rapid loss of the C-terminal domain, thereby destroying the collagenolytic activity of the enzyme. Although relatively high amounts of stromelysin were needed to activate procollagenase-3 efficiently over 6 h, this activation pathway may still be of relevance in vivo, since very high levels of stromelysin have been observed under certain pathological conditions (Walakovits et al., 1992;Matrisian and Bowden, 1990).
Collagenase-3 can be assigned to the collagenase subfamily of matrix metalloproteinases, according to substrate specificity analysis, hydrolyzing the interstitial collagens I-III into 3/4 and 1/4 fragments preferentially cleaving type II collagen over type I and III. In contrast, fibroblast collagenase preferentially cleaves type III and neutrophil collagenase type I collagen  4. Inhibition of active collagenase-3 by the three homologous TIMPs. Active collagenase-2 (2 nM) was incubated with increasing concentrations of either TIMP-1 (q), TIMP-2 (f) or TIMP-3 (å). (Welgus et al., 1981;Hasty et al., 1987). Thus the three collagenases show distinct collagen substrate specificities, which implies that they may have evolved as specialized enzymes in order to dissolve different connective tissues, which vary in their collagen composition. Collagenase-3 may especially be important in the turnover of articular cartilage, which is rich in type II collagen. The specific activities of the three collagenases against type I collagen were in the range of 100 -120 g/min/ nmol enzyme with exception of "superactive" neutrophil collagenase, which cleaved 338 g/min/nmol. By comparison of the ratios of collagenolytic/gelatinolytic activity (Table III) or collagenolytic/peptidolytic activity (not shown) of the three enzymes, it becomes clear that fibroblast collagenase is the most specific collagenase within this group, although the specific collagenolytic activity of "superactive" neutrophil collagenase is 3 times higher.
Collagenase-3 cleaved gelatin and the two synthetic peptide substrates with highly improved efficiency when compared with fibroblast or neutrophil collagenase. Thus, it appears that collagenase-3 not only efficiently degrades type I collagen, but it might also act as a gelatinase to further degrade the initial cleavage products of collagenolysis to small peptides suitable for further metabolism. This is in agreement with results obtained earlier for rat collagenase, which shows relatively high levels of gelatinolytic activity (Welgus et al., 1985) and shares the highest degree of homology with human collagenase-3, as does mouse collagenase (Henriet et al., 1992;Quinn et al., 1990). According to the high degree of functional and sequence homology between human collagenase-3 and the rodent collagenases, these enzymes belong to the collagenase-3 subfamily (MMP-13) of matrix metalloproteinases and are distinct from human fibroblast collagenase (MMP-1). We therefore propose to introduce a revised nomenclature for the rodent collagenases to prevent further confusion in the literature assigning them as MMP-13. Indeed, it may be concluded that rat and mouse cells express only collagenase-3 (MMP-13), there being no evidence to date for a homologous MMP-1 in either rat or mouse. The relative distribution of fibroblast collagenase (MMP-1) and collagenase-3 (MMP-13) in human tissues awaits detailed studies, but initial observations suggest that MMP-1 is predominant.
Comparison of the ratios of gelatinolytic over peptidolytic activity of collagenase-3 with those values obtained for human gelatinase A revealed that collagenase-3 is 10 times less efficient than wild-type gelatinase A (Murphy et al., 1994). The high efficiency of wild-type gelatinase A against gelatin as a substrate can be attributed to the fibronectin-like type II repeats, since a gelatinase A deletion mutant (⌬ V191-Q364gelatinase A) lacking these sequence motifs has a similar ratio of gelatinolytic over peptidolytic activity to collagenase-3 (Murphy et al., 1994). Thus collagenase-3 shares some proteolytic characteristics with the gelatinase subfamily of matrix metalloproteinases, which is reflected in common structural elements shared by collagenase-3 and the gelatinases being localized within the active site cleft as discussed below.
Sequence alignments of the active site residues of the collagenases with the gelatinases revealed that the Arg (Fig. 5, number 1) in fibroblast collagenase is changed to Ile or Leu in collagenase-3, the rodent collagenases, neutrophil collagenase, and in the gelatinases. It has been noted by Stams et al. (1994) that the SЈ 1 -pocket in neutrophil collagenase is significantly larger than the equivalent pocket in fibroblast collagenase and that we can deduce that due to the presence of Leu within collagenase-3 and the gelatinases that these have a similar enlarged SЈ 1 -pocket and structure. Hence these enzymes should be able to hydrolyze a broader range of substrates. Second, collagenase-3, neutrophil collagenase, and the rodent homologues share a Pro residue (Fig. 5, number 3) with the gelatinases, while fibroblast collagenase has an Ile residue in this position. Furthermore, collagenase-3, the rodent enzymes, and the gelatinases contain negatively charged residues just preceding the third His residue of the catalytic zinc binding motif (either Asp or Glu; Fig. 5, number 2). In contrast, this residue corresponds to Ser or Ala in fibroblast or neutrophil collagenase. The presence of a negatively charged residue in collagenase-3 and the gelatinases might well have implications on the polarization of the zinc-bound water molecule within these enzymes, possibly increasing its nucleophilic nature (Fig. 6). This would certainly account for the increased proteolytic efficiency of collagenase-3 and the gelatinases, as indicated by our experimental results, but it remains to be confirmed by site-directed mutagenesis.
Analysis of the inhibition profile of collagenase-3 by the three homologous TIMPs revealed that all react in 1:1 stoichiometry by forming noncovalent tight-binding complexes, which is in agreement with earlier published data on other matrix metalloproteinases (for review see, Murphy and Willenbrock (1995)). Key residues specifically conserved between the gelatinases, collagenase-3 (and partially MMP-8), which may be of importance for gelatinolytic specificity are indicated in boldface italics.
Comparison of the efficacy of two synthetic hydroxamate inhibitors against collagenase-3 confirmed the structural similarity to the gelatinases. CT1399, which has a K i of less than 10 pM for gelatinase A and 16 pM for gelatinase B, had an approximate K i of ϳ4 pM for collagenase-3 and a K i of 385 nM for MMP-1. Similarly, CT1847, which has a K i of 1.55 nM against gelatinase A and 2.1 nM against gelatinase B had K i values of 0.54 nM against collagenase-3 and of 2.9 nM against MMP-1. 4 It may be concluded that inhibitors directed against gelatinases will also be efficient in the control of collagenase-3.
Our studies have indicated that human collagenase-3 is a potent proteinase with a broad spectrum of activity against extracellular matrix proteins (data not shown) as well as collagenolytic and high gelatinolytic activity. The regulation and location of its expression relative to the more specific fibroblast collagenase will be a matter of great importance for future study.