Substrate Specificity of Human Collagenase 3 Assessed Using a Phage-displayed Peptide Library*

The substrate specificity of human collagenase 3 (MMP-13), a member of the matrix metalloproteinase family, is investigated using a phage-displayed random hexapeptide library containing 2 × 108 independent recombinants. A total of 35 phage clones that express a peptide sequence that can be hydrolyzed by the recombinant catalytic domain of human collagenase 3 are identified. The translated DNA sequence of these clones reveals highly conserved putative P1, P2, P3 and P1′, P2′, and P3′ subsites of the peptide substrates. Kinetic analysis of synthetic peptide substrates made from human collagenase 3 selected phage clones reveals that some of the substrates are highly active and selective. The most active substrate, 2,4-dinitrophenyl-GPLGMRGL-NH2 (CP), has a k cat/K m value of 4.22 × 106 m − 1s− 1 for hydrolysis by collagenase 3. CP was synthesized as a consensus sequence deduced from the preferred subsites of the aligned 35 phage clones. Peptide substrate CP is 1300-, 11-, and 820-fold selective for human collagenase 3 over the MMPs stromelysin-1, gelatinase B, and collagenase 1, respectively. In addition, cleavage of CP is 37-fold faster than peptide NF derived from the major MMP-processing site in aggrecan. Phage display screening also selected five substrate sequences that share sequence homology with a major MMP cleavage sequence in aggrecan and seven substrate sequences that share sequence homology with the primary collagenase cleavage site of human type II collagen. In addition, putative cleavage sites similar to the consensus sequence are found in human type IV collagen. These findings support previous observations that human collagenase 3 can degrade aggrecan, type II and type IV collagens.

Human collagenase 3 (MMP-13) 1 is a member of the collagenase subfamily of matrix metalloproteinases (MMPs) (1). Several potential roles for this enzyme have been proposed, including aggrecan degradation associated with osteoarthritis (2), cleavage of type II collagen in osteoarthritic cartilage explants (3), and in tumor progression and metastasis (1). Its expression is also detected in human T lymphocytes (4). Human collagenase 3 degrades type II collagen, the primary collagen in articular cartilage, more than 10 times as efficiently as collagenase 1, another MMP known to be up-regulated in arthritic diseases (5). In addition, collagenase 3 cleaves soluble type II collagen 5 to 6 times more efficiently than type I or type III collagen (6). Furthermore a preferential collagenase 3 inhibitor significantly reduces collagen degradation from unstimulated osteoarthritic cartilage (3). These cumulative data suggest that collagenase 3 may play a significant role in the cleavage of type II collagen in arthritic cartilage.
The specificity of this enzyme for aggrecan and other collagenous substrates has also been investigated. Collagenase 3 cleaves aggrecan at the same major and minor sites identified for other members of the MMP family as well as at a novel site (7). In addition, collagenase 3 can degrade fibronectin and type IV, IX, X, and XIV collagens (8).
MMPs are considered to be potential therapeutic targets. A highly active and selective substrate is a prerequisite for screening small molecule inhibitors in high throughput screening. Knowledge of the sequence specificity can facilitate the design and optimization of potent and selective inhibitors. Synthetic peptide substrates based on physiological cleavage site sequences are routinely used for enzymatic assays. These substrates can often be improved because physiologically based peptide substrates may not be the optimal substrate (9,10). To optimize protease substrates, a number of chemical methods have been developed. These include using defined substrate mixtures (11,12) and immobilized peptide libraries (13). Recently two biological systems using phage display have been developed to quickly search and decode a large peptide library for substrate optimization (14,15). The first system uses monovalent substrate phage display to study substrate specificity of proteases (14). The second system uses polyvalent substrate phage display to more quickly optimize protease substrates (15,16). These chemical and biological methods may also be useful in searching for substrates of proteases where the natural physiological substrate is unknown.
In this study, we have screened a phage-displayed random hexapeptide library of 2 ϫ 10 8 members using the catalytic domain of recombinant human collagenase 3. We have identified 35 phage clones that express a hexapeptide sequence cleaved by the enzyme. We searched sequence data bases for potential collagenase 3 cleavage sites by running FASTA and BLAST searches with the hexapeptide phage sequences as queries. These searches identified potential collagenase 3 cleavage sites in type IV collagen and the latency-associated peptide of TGF-␤Ϫ3. Using collagenase 3 and other closely related MMPs, we have also carried out kinetic studies for synthetic peptide substrates synthesized based on the hexapeptide sequence from a selected number of the 35 phage clones. These synthetic peptides are shown to be highly sensitive and selective to human collagenase 3.

EXPERIMENTAL PROCEDURES
Reagents-Escherichia coli K91 was obtained from Marc Navre of Affymax. E. coli TG1 was purchased from Amersham Pharmacia Biotech. All restriction enzymes were purchased from New England Biolabs (Beverly, MA). ECL system was purchased from Amersham Pharmacia Biotech. Pansorbin cells were purchased from Calbiochem. Monoclonal antibody 3-E7 was purchased from Gramsch Laboratories (Schwabhausen, Germany). Monoclonal antibody mAb 179 was prepared using standard methods. Dimethyl sulfoxide, p-aminophenylmercuric acetate, HEPES, and Brij 35 were purchased from Sigma.
Purification of Human Recombinant Collagenase 3-The catalytic domain of human collagenase 3 was prepared by PCR cloning from the original pNOT3␣ clone (1) using the 5-prime oligonucleotide 5Ј-CCCG-GGCATATGTACAATGTTTTCCCTCGTACGCTTAAATGGTCCAAAA-CGAATTTAACCTACAGA and the 3-prime oligonucleotide 5Ј-CCCGG-GATCCTTAACCATAGAGAGACTGAATCCCTTGTACATCGTCATCA-GGAAG. Base pairs 34 -36 of the 5-prime oligo change an ATG to CTT, which removes a problematic second start site, whereas base pair 29 of the 3-prime oligo changes a G to an A, which removes a naturally occurring BamHI site. The final product was cloned via NdeI and BamHI into pET11b. A 10-liter broth culture of E. coli BL21[DE3] bearing pETcoll3 was grown at 37°C to an A 600 of 1, shifted to 16°C, and induced with isopropyl-1-thio-␤-D-galactopyranoside, and growth was continued for 4 h until harvest. The E. coli cell pellet was suspended in 2 ml of lysis buffer (50 mM Tris, pH 7.5, 200 mM NaCl, 10 mM EDTA, 1 mM DTT, 50 g/ml APMSF (4-amidinophenyl)methanesulfonylfluoride), 20 g/ml aprotinin, 10 g/ml bestatin, 5 g/ml pepstatin A, and 10 g/ml leupeptin)/g of wet weight cells. Cells were lysed with two passes through a French press. Inclusion bodies were spun at 48,000 ϫ g and washed twice with 1.5 ml of Triton buffer (50 mM Tris, pH 7.5, 200 mM NaCl, 10 mM EDTA, 1 mM DTT, and 0.5% Triton X-100)/g of cells and once with 1.5 ml of low salt buffer (50 mM Tris, pH 7.5, 10 mM EDTA, 1 mM DTT, and 0.5% Triton X-100)/g of cells. The human recombinant collagenase 3 was solubilized with 1 ml of extraction buffer (50 mM Tris, pH 7.5, 1 mM DTT, and 6 M urea)/g of cells. Protein was applied to an anion exchange column (PerSeptive Biosystems Poros HQ20) equilibrated with column buffer (20 mM Tris, pH 7.5, 1 mM DTT, and 6 M urea) and eluted with a linear NaCl gradient (0 to 300 mM). Fractions containing collagenase 3 were identified by SDS-PAGE and pooled. Collagenase 3 was refolded at 4°C by rapid dilution to 100 g/ml (ϳ45-fold dilution) in folding buffer (50 mM Tris, pH 7.5, 200 mM NaCl, 50 M ZnCl 2 , and 1 mM CaCl 2 ). Unfolded or incorrectly folded protein was removed by filtration (0.2 m). The active enzyme was concentrated and stored at Ϫ80°C.
Phage-displayed Hexapeptide Library Screening-The construction of phage-displayed hexapeptide library fTC-LIB-N6 with 2 ϫ 10 8 individual sequence members was as described previously (15). A total of 2 ϫ 10 10 fTC-LIB-N6 phage were digested at 37°C for 1 h by 2 nM collagenase 3 in a 250-l reaction mixture containing 1ϫ MMP buffer (50 mM Tris-HCl, 5 mM CaCl 2 , 10 M ZnSO 4 , 200 mM NaCl, pH7.6) and 0.005% Brij 35. The reaction was terminated by the addition of 0.5 M EDTA to a final concentration of 5 mM. Then 100 g of mAb 179 and 10 g of mAb 3-E7 was added to the terminated reaction mixture followed by the addition of bovine serum albumin to 0.1%. Both of the monoclonal antibodies were used to capture uncleaved phage as described previously (15). After 30 min of incubation on ice, 100 l of Pansorbin cells were added, and the reaction mixture was rotated at 4°C for 1 h. The mixture was centrifuged for 2 min at 12,000 ϫ g, and the supernatant was recovered. The adsorption of uncleaved phage-mAb complex to Pansorbin cells was repeated once. Phage in the final supernatant were transfected into E. coli K91 and amplified overnight at 37°C. Phage in a small aliquot of the final supernatant was titered on E. coli K91. The above screening was performed for five rounds. Single phage clones from the titer plates were picked randomly and grown in 4 ml of LB for isolation of single-stranded DNA and for Western blotting analysis of phage.
Construction of Site-directed Mutant Phage, Consensus Phage, and Aggrecan-derived Phage-C2-12P3A, NF, DL, PV and consensus CP phage clones were constructed by inserting oligo cassettes into KpnI/ XhoI-digested fAFF-TC-LIB vector and confirmed by DNA sequencing. NF phage was made based on the major MMP site (PEN-FFG), and DL phage was made based on the minor MMP site (SED-LVV) in aggrecan. PV phage was constructed on the basis of a novel human collagenase 3 cleavage site (VKP-VFE), which was not observed for other MMPs in aggrecan (7). Oligonucleotides 5Ј-TCGAGCGGTGGTAGTGGTGGT-GCTTTGGGGTTGTCGCTGGTAC and 5Ј-CAGCGACAACCCCAAAG-CACCACCACTACCACCGC were used for the construction of C2-12P3A. Oligonucleotides 5Ј-TCGAGCGGTGGTAGTGGTCCGGAAAA-CTTCTTCGGTCTGGTAC and 5Ј-CAGACCGAAGAAGTTTTCCGGAC-CACTACCACCGC were used for the construction of NF phage clone. Oligonucleotides 5Ј-TCGAGCGGTGGTAGTGGTTCTGAAGACCTGGT-TGTTCTGGTAC and 5Ј-CAGAACAACCAGGTCTTCAGAACCACTAC-CACCGC were used for the construction of DL phage clone. PV phage was made by 5Ј-TCGAGCGGTGGTAGTGGTGTTAAACCGGTTTTCG-AACTGGTAC and 5Ј-CAGTTCGAAAACCGGTTTAACACCACTACCA-CCGC. Consensus CP phage clone was made with oligonucleotides 5Ј-TCGAGCGGTGGTAGTGGTCCTTTGGGTATGCGTGGTCTGGTAC and 5Ј-CAGACCACGCATACCCAAAGGACCACTACCACCGC.
Western Blotting Analysis and DNA Sequencing-Phage were precipitated by adding 200 l of 20% polyethylene glycol, 2.5 M NaCl to 1 ml of phage supernatant and centrifuging at 12,000 ϫ g for 15 min. Phage pellets were dissolved in 100 l of Tris-buffered saline. Three different enzymatic reactions were performed. For the first reaction, phage was incubated in 50 l of 1ϫ MMP buffer and 0.005% Brij 35 at 37°C for 1 h, but enzyme was not added. For the second reaction, phage was treated as described for the first reaction except 0.4 nM collagenase 3 was added. For the third reaction, phage were treated the same way as described for the first reaction except 0.4 nM collagenase 3 and 0.05 l of 10 M GW9471 inhibitor were combined (19). EDTA was added to 5 mM to terminate enzymatic reactions. Each reaction mixture of 10 l was analyzed on 12% SDS-polyacrylamide gels. Substrate peptidephage gene gIII fusions were detected by mAb 179. Phage cleaved by human collagenase 3 but whose cleavage was inhibited by the broad spectrum MMP inhibitor GW9471 was chosen for further study. Singlestranded DNA from individual phage clones was prepared using QIAprep M13 BioRobot kit in a Bio Robot 9600 system (Qiagen, CA). The phage substrate DNA sequences were determined by an ABI automated fluorescent DNA sequencer.
Collagenase 1 (MMP-1), Stromelysin 1 (MMP-3), and Gelatinase B (MMP-9) for Selectivity Studies-A 19-kDa collagenase 1 catalytic domain was expressed in E. coli. Purification was described previously (20). Full-length stromelysin 1 was also expressed in E. coli and purified from inclusion bodies using a protocol similar to that used for collagenase 1. Truncated progelatinase B (lacking the C-terminal domain) was expressed from baculovirus infected Trichoplusia ni cells and purified from the media using gelatin-agarose affinity chromatography. Purity was greater than 90%, as assessed by SDS gel electrophoresis.
Estimation of Specificity Constants k cat /K m -All synthetic 2,4-dinitrophenyl peptide substrates were ordered from Synpep (Dublin, CA) and prepared as 10 mM stock solutions in dimethyl sulfoxide. A substrate concentration of 10 M was used for all enzymatic reactions. Before the assay, collagenase 3 was diluted 100-fold to a 250 nM stock in 200 mM NaCl, 50 mM Tris, 5 mM CaCl 2 , 10 M ZnSO 4 , 0.01% Brij 35, pH 7.5 (assay buffer). Hydrolysis reaction concentrations were adjusted from 0.1 to 10 nM by further dilution. The samples were incubated at 37°C for 10 to 180 min. The reactions were terminated with the addition of an equal volume of 0.1% heptafluorobutyric acid. Peptide substrate and product were separated by reverse phase HPLC, and the peaks were evaluated by monitoring absorbance at 350 nm. Cleavage sites in synthetic peptides were determined by liquid chromatographymass spectrometry analysis of each reaction mixture.
The specificity constants k cat /K m were also determined for collagenase 1, stromelysin 1, and gelatinase B. Collagenase 1 was refolded by dilution to 5 M with extensive dialysis against 50 mM Tris/HCl, pH 7.5, containing 200 mM NaCl, 1 mM CaCl 2 , and 50 M ZnCl 2 (refolding buffer). Hydrolysis reactions were performed using 1-10 nM enzyme in assay buffer. Stromelysin 1 was refolded by diluting to 15 nM in assay buffer. After 1 h at room temperature, the enzyme was concentrated to a 150 nM stock solution using an Amicon CentriPrep 30 concentrator. Stromelysin 1 assay concentrations ranged from 5 to 50 nM. Gelatinase B was diluted to a 2.5 M stock in assay buffer and activated for 12 h with 1 mM p-aminophenylmercuric acetate at 37°C. The enzyme was desalted with Bio-Rad Bio-Spin 6 columns into assay buffer. The enzyme was diluted to a final assay concentration of 1-10 nM. All enzyme concentrations were verified by active site titration with the broad spectrum MMP inhibitor, GW9471 (19). The IC 50 for collagenase 3, collagenase 1, gelatinase B, and stromelysin 1 is 0.5, 1.2, 1.0, and 2.4 nM, respectively, for this inhibitor. A substrate concentration of 10 M was used for each enzymatic reaction. Hydrolysis reactions and HPLC analysis were performed under the same conditions as for collagenase 3. The standard error for specificity constant k cat /K m is less than or equal to 30% based on 3 independent experiments. The specificity constant k cat /K m for 2,4-dinitrophenyl-GPLGMRGL-NH 2 (CP) was also estimated by performing enzymatic reactions at a substrate concentration of 2.5 M, and the k cat /K m value was similar to that estimated at a substrate concentration of 10 M.

RESULTS
A total of 2 ϫ 10 10 fTC-LIB-N6 phage from the phage-displayed random hexapeptide library was treated with the catalytic domain of human collagenase 3. Undigested phage was removed by mAb 3E-7 and mAb 179 attached to Pansorbin cells. Cleaved phage was amplified in E. coli K91 cells and then subjected to four more rounds of enzyme digestion and screening. This screening yielded 35 phage clones, which were shown to be more sensitive by Western blot analysis to human collagenase 3 treatment than the positive control phage fTCol-3S (Table I). DNA sequencing of the 35 phage clones and sequence alignment analysis reveal that a proline appears in all 35 clones in a putative P3 subsite, a leucine appears in 13 clones at the P2 subsite, a glycine occurs in 12 clones at the P1 subsite, a methionine occurs in 8 clones at the P1Ј subsite, a basic residue arginine, lysine, or histidine occurs in 12 clones at the P2Ј subsite, and a glycine occurs in 7 clones at the P3Ј subsite (Table I). A consensus sequence, PLGMRG, can be deduced based on the preferred residue in each subsite position. A site-directed mutant phage C2-12P3A was constructed in which proline at the putative P3 subsite in phage clone C2-12 was replaced by alanine (Table II). Western blotting did not detect cleavage of the site-directed mutant by collagenase 3, whereas the parent phage was cleaved (Table II and Fig. 1). The k cat /K m for the synthetic peptide, however, was reduced 7-fold by the alanine substitution (Table V). Phage display screening resulted in five phage clones that shared sequence homology with a major MMP cleavage site NF for aggrecan (7). All 5 clones (C3-15, C3-20, C4 -14/5-25, C2-22, and C5-27) have a proline at the P3 subsite and a phenylalanine in the P1Ј subsite. Three clones (C3-15, C4 -14/5-25, and C3-20) have a glycine at P3Ј subsite (Table III). Three phage clones, DL, NF, PV, corresponding to minor, major MMP cleavage sites, and a novel collagenase 3 site, reported for aggrecan, were constructed (Table II). Of these, only the NF phage that was based on the major MMP cleavage site for aggrecan was obviously sensitive to human collagenase 3 treatment as shown by Western blot analysis (Table II and Fig. 1).
Synthetic peptide substrate CP, corresponding to the consensus sequence, showed the highest k cat /K m value of 4.22 ϫ 10 6 M Ϫ1 s Ϫ1 for hydrolysis by human collagenase 3. The k cat /K m value of peptide CP for human collagenase 3 is 1344-, 11-, and 820-fold higher than that for the MMPs stromelysin 1, gelatinase B, and collagenase 1, respectively. Peptide substrate CP has a 3-fold-improved k cat /K m value for hydrolysis by collagenase 3 when compared with the synthetic peptide substrate corresponding to positive control phage, fTCol-3S. Furthermore, the peptide substrate CP has a higher k cat /K m than described for previous substrates such as McaPChaGNva-HADpa-NH 2 , McaPLGLDpaAR-NH 2 (6), and the triple helical collagen substrate, ␣1(I)-772-786 triple helical peptide (22). With respect to selectivity for collagenase 3 over collagenase 1 and stromelysin 1, peptide CP also has a greater relative selectivity than the peptide corresponding to control phage, fTCol-3S (Tables V and VI). The NF synthetic peptide corresponding to the major MMP cleavage site in aggrecan has a 37-fold lower k cat /K m value for hydrolysis by collagenase 3 compared with peptide CP. It has a 49-, 1-, and 26-fold human collagenase 3 selectivity over related MMPs stromelysin 1, Cleavage site of type II␣1 collagen  gelatinase B, and collagenase 1. Substrates GPKGVYSL and GPFGFKSL, corresponding to phage clones C3-16 and C5-27, are good substrates for collagenase 3 and gelatinase B, but relatively poorer substrates for collagenase 1 and stromelysin 1. Highly selective substrates have been described (23,24) and shown to have utility for identifying specific MMPs in biological samples (25). No hydrolysis was observed by collagenase 3 for synthetic peptides PV and DL that were synthesized based on a reported minor MMP site and a novel collagenase 3 site, respectively, in aggrecan. We searched the SwissProt and translated EMBL protein sequence data bases for potential collagenase 3 cleavage sites by running FASTA and/or BLAST searches using the consensus sequence and each of the 35 hexapeptide phage sequences as queries. Considering the 17,000 human sequences available in March 2000, the 36 searches identified 12 different 6-residue sites that were identical to one or more of the phage sequences and numerous 6-residue sites that were similar to one or more of the phage sequences.
Eight of the identical matches occur in sequences that are poorly characterized or in hypothetical translations of DNA sequences; these are difficult to interpret. Three of the identical matches occur in proteins that are expected to be intracellular and, thus, unavailable for cleavage by collagenase 3. The one remaining identical match occurs in the sequence of TGF-␤3, a well characterized extracellular protein. This match involves the sequence PKGITS in the prodomain of TGF-␤3, which is identical to the sequence of phage clone C4 -19 (Table VII). This match is particularly interesting since activation of TGF-␤ involves proteolysis by metalloproteases (26).
Several of the near matches occur in collagens and collagenlike sequences, some of which may be substrates for collagenase 3. For example, Knauper et al. report that collagenase 3 degrades type IV collagen, although the actual cleavage site(s) is unknown (8). The FASTA searches identified several potential cleavage sites in the ␣2(IV) chain of type-IV collagen, including GPMGLK (residues 997-1002, similar to phage clone C3-9), GPLGLP (residues 851-856, similar to phage clone C2-12), and GPVGMK (residues 879 -884, similar to the consensus sequence). Cleavage of collagen by collagenase 1 and 2 is thought to involve local unwinding of the triple helix, with a tightly triple-helical region upstream from the cleavage site and a loosely triple-helical region downstream from the cleavage site (27). Thus, the local conformation and more distant flanking residues would presumably determine whether the ␣2(IV) chain is actually cleaved by collagenase 3 at any of these sites.
Additional near matches occur in proteins of the extracellular matrix. For example, biglycan, a small proteoglycan similar to decorin and found in connective tissue (28), contains a PKGVFS sequence (residues 175-180) that is almost identical to phage clone C3-16 (PKGVYS). The corresponding synthetic peptide was hydrolyzed by both collagenase 3 and gelatinase B (Table VI), suggesting that biglycan might be a substrate for either or both of these metalloproteases. DISCUSSION It has been proposed that collagenase 3 is associated with osteoarthritis, rheumatoid arthritis, and tumor progression and metastasis (2,3,5). To gain further insights into the substrate specificity of this enzyme, we screened a phage-displayed random hexapeptide library. We have identified 35 phage clones that are cleaved by the recombinant catalytic domain of collagenase 3. The presumed cleavage site and corresponding P3, P2, P1, P1Ј, P2Ј, and P3Ј positions were identified from a sequence alignment. All of the selected peptide substrates have proline at the P3 position and lipophilic amino acids at P1Ј. Most of the selected substrates have lipophilic amino acids at P2 and small amino acids at P1 and P3Ј. Many of the substrates have a basic amino acid at P2Ј. Our observation that the site-directed mutant C2-12P3A is 7-fold less sensitive to collagenase 3 treatment indicates the important role of the P3 proline in substrate specificity (Fig. 1, Tables II  and V). Earlier studies (24) showed that interstitial collagenase (MMP-1) and neutrophil collagenase (MMP-8) also have a preference for proline at P3 and for lipophilic residues at P1Ј. However, the preference for proline at P3 appears to be stronger in collagenase 3 than in the other two collagenases. Collagenase 3 and neutrophil collagenase can accommodate both aromatic and aliphatic amino acids at P1Ј, whereas interstitial collagenase has a stronger preference for aliphatic amino acids (24). This is consistent with x-ray structures, which show that neutrophil collagenase and collagenase 3 have deep lipophilic S1Ј pockets, whereas the S1Ј pocket in interstitial collagenase is partially blocked by an arginine side chain (29).
Screening selected seven substrate sequences with sequence homology to the primary collagenase cleavage site in type II collagen (PQG-LAG). This is in agreement with previous reports that human collagenase 3 cleaves type II collagen (3). Previous work (7) showed that collagenase 3 cleaves the interglobular domain of aggrecan at a major MMP site (PEN-FFG), a minor MMP site (SED-LVV), and a novel collagenase 3 site (VKP-VFE). Our phage display screening selected five substrates similar to the PEN-FFG site (Table III) but none similar to the SED-LVV or VKP-VFE sites. Western blot analysis showed that substrate phage clones DL and PV, corresponding to the minor MMP cleavage site and the novel collagenase 3 cleavage site, were not sensitive to collagenase 3 treatment (Fig. 1). No turnover is observed for the corresponding synthetic peptides DL and PV (Table V). Cleavage of these sites by collagenase 3 evidently requires additional flanking residues and/or the specific conformation adopted in the interglobular domain of aggrecan.
Most of the synthetic peptides corresponding to phage clones have reasonable specificity constants. The synthetic peptide substrate corresponding to the consensus sequence CP has the highest collagenase 3 k cat /K m value, 4.22 ϫ 10 6 M Ϫ1 s Ϫ1 . This k cat /K m is 1300-, 11-, and 820-fold better than that of related MMPs stromelysin 1, gelatinase B, and collagenase 1 (Tables V  and VI). Interestingly, the fTCol-3S peptide substrate GPLGL-WAR has a k cat /K m value of 1.4 ϫ 10 6 M Ϫ1 s Ϫ1 for human  Chain of Type IV collagen  aa 1000  Triple-helical region  GPVGMK  ␣2 Chain of Type IV collagen  aa 881  Triple-helical region  C2-12  GPLGLS  GPIGLS  ␣4 Chain of Type IV collagen  aa 86  Triple-helical region  C3-16  PKGVYS  PKGVFS  Biglycan  aa 177  Mature protein  C4-19  PKGITS  PKGITS  TGF-␤3  aa 150 Latency-associated peptide collagenase 3. However, the corresponding control phage is only weakly hydrolyzed as assessed by Western blot. It is possible that steric interference of phage protein on the displayed control substrate sequence may have caused the reduction of turnover. Therefore it is possible that some of the substrate sequences in the phage-displayed hexapeptide library may not have been selected because of steric interference and/or the lesser sensitivity of Western blotting. The k cat /K m value for collagenase 1 is also significantly higher for GPLGLWAR than what has been reported previously in the literature (17,18). The explanation is unclear but may be due to differences in detergent concentration or buffer composition as well as potential differences related to the use of a truncated construct. This work demonstrates the utility of phage display in identifying highly active and selective substrates. The most active substrate, CP, and selective substrates, C3-16 and C5-27, may be useful in the screening and design of selective collagenase 3 inhibitors for cancer and arthritic diseases. In addition, the more selective substrates may be used to monitor activity of collagenase 3 in disease samples. An example of the application of a selective substrate was reported for stromelysin, where levels of the enzyme were determined in rheumatoid synovial fluid by a fluorogenic assay using a selective peptide substrate (25). Phage display can also help define the substrate specificity of a protease more completely. In general, this could be useful in recognizing physiologically relevant substrates.
In this particular case, data base searches identified sites in many proteins with similarity to one or more of the phage query sequences. However only a few proteins showed identical or one conservative mismatch sequences to the phage query sequences. Most of these sites occur in poorly characterized or hypothetical proteins or in proteins that are presumably intracellular and, thus, unavailable to collagenase 3. However, several sites seem plausible enough to suggest further experiments. Of these, the most interesting is TGF-␤3, which contains a PKGITS sequence identical to that of clone C4 -19. TGF-␤3 is a potent angiogenesis factor that also plays a role in wound healing and bone homeostasis (30). TGF-␤ is released from cells as an inactive complex consisting of the mature TGF-␤ protein bound to its cleaved prodomain, which is then referred to as the TGF-␤ latency-associated protein. (31). This inactive form of TGF-␤ can be activated by enzymatic or chemical treatment, allowing it to bind to its receptor (32). Recently, Yu and Stamenkovic (26) discovered that cell surface gelatinase B proteolytically activated TGF-␤1, 2, and 3, thereby acting as an angiogenesis factor and promoting tumor invasion and metastasis. The PKGITS sequence identified by our phage display work lies near the middle of the TGF-␤3 prodomain. Collagenase 3 might thus cleave the TGF-␤3 latency-associated protein, possibly breaking up the complex and thereby activating TGF-␤3. The most similar synthetic peptide contains the PKGVYS sequence, corresponding to phage clone C3-16. This synthetic peptide was cleaved efficiently by collagenase 3 and gelatinase B, suggesting that either of these enzymes might cleave TGF-␤3 at this position. Of course, this site might be inaccessible in the three-dimensional structure of the complex. Note also that activation by gelatinase B depends on its localization to the cell surface by the hyaluronan receptor, CD44 (26). Activation by collagenase 3 might require similar localization, possibly by a different mechanism.
Data base searching identified potential collagenase 3 cleavage sites in type IV collagen and biglycan. Type IV collagen is the major structural protein of basement membranes. Degradation of type IV collagen could lead to breakdown of the basement membrane, which occurs in cancer (33) and multiple sclerosis (34), thus suggesting a possible role for collagenase 3 in these diseases. Biglycan is a small proteoglycan that has homol-ogy to decorin and is found in connective tissue (28). Degradation of biglycan could be important in joint disease (21).
In summary, we have used phage display technology to identify a number of new collagenase 3 substrate sequences. Many of the new substrate sequences are similar to known cleavage sites in type II collagen or aggrecan. However, others deviate significantly from previously known cleavage sequences. Considered together with previous work, the phage display results help define the overall specificity of collagenase 3. We have also used this knowledge of collagenase 3 specificity to speculatively identify other possible substrates. Our data base search results support previous studies demonstrating that gelatinase B can degrade TGF-␤ (26) and provide some evidence for the first time that collagenase 3 may also be involved in processing of the latency-associated peptide to activate this growth factor. Future studies will help to determine if any of the proteins identified by data base searches are actual substrates for collagenase 3.