A study of the collagen-binding domain of a 116-kDa Clostridium histolyticum collagenase.

The Clostridium histolyticum 116-kDa collagenase consists of four segments, S1, S2a, S2b, and S3. A 98-kDa gelatinase, which can degrade denatured but not native collagen, lacks the C-terminal fragment containing a part of S2b and S3. In this paper we have investigated the function of the C-terminal segments using recombinant proteins. Full-length collagenase degraded both native type I collagen and a synthetic substrate, Pz-peptide, while an 88-kDa protein containing only S1 and S2a (S1S2a) degraded only Pz-peptide. Unlike the full-length enzyme, S1S2a did not bind to insoluble type I collagen. To determine the molecular determinant of collagen binding activity, various C-terminal regions were fused to the C terminus of glutathione S-transferase. S3 as well as S2bS3 conferred collagen binding. However, a glutathione S-transferase fusion protein with a region shorter than S3 exhibited reduced collagen binding activity. S3 liberated from the fusion protein also showed collagen binding activity, but not S2aS2b or S2b. S1 had 100% of the Pz-peptidase activity but only 5% of the collagenolytic activity of the full-length collagenase. These results indicate that S1 and S3 are the catalytic and binding domains, respectively, and that S2a and S2b form an interdomain structure.

ture-function relationship of this unique enzyme is not known.
Multiple forms of collagenase are produced by C. histolyticum. Seven different forms have been identified, and they are divided into two classes based on possible similarities in their amino acid sequences and their specificities toward peptide substrates (9). In a previous study (10) we have cloned and sequenced a colH gene encoding the 116-kDa collagenase (ColH), which is abundant in many commercial enzymes (11). Comparison of the predicted amino acid sequence of ColH with those of the Vibrio alginolyticus (12) and Clostridium perfringens collagenases (13) revealed a segmental structure for these enzymes (10). ColH has been shown to consist of four segments, S1, S2a, S2b, and S3, and S2a and S2b are homologous. The molecular masses of S1, S2a, S2b, and S3 are 78.1, 10.0, 9.9, and 14.1 kDa, respectively (10). S1 contains the sequence HEXXH, a consensus motif located at the catalytic center of zinc-metalloproteases. A 98-kDa gelatinase that copurified with ColH hydrolyzed denatured but not native collagen. The two enzymes possessed identical N-terminal sequences and their peptide maps were almost identical. Therefore, the 98-kDa gelatinase is probably produced by cleaving off a C-terminal peptide from ColH (10). These observations led us to suspect that the C-terminal peptide forms a functional domain, which is involved in either providing accessibility to or binding of the enzyme to collagen.
To gain insights into the structure-function relationships of C. histolyticum collagenases, we have attempted a molecular dissection of ColH by constructing recombinant derivatives of the enzyme. In this paper, we examined the collagen binding activities of various C-terminal peptides fused to the C terminus of glutathione S-transferase (GST) 1 to localize the collagen-binding domain. We have also examined the enzymatic activities of C-terminal truncated species toward native type I collagen and a synthetic peptide substrate.

EXPERIMENTAL PROCEDURES
Bacterial Strains, Plasmids, and Enzymes-C. histolyticum JCM 1403 (ATCC 19401) was obtained from the Institute of Physical and Chemical Research (Saitama, Japan). Bacillus subtilis DB104 (14) was used to produce the recombinant collagenases. Escherichia coli DH5␣ (15) was used for the construction of all recombinant plasmids. Plasmid pCHC116⌬3 was generated by nested deletion of pCHC116 (10) from its 3Ј end. The plasmid contains a colH gene fragment from nucleotide 2,719 to 3,391 (numbered as in Ref. 10), which encodes the C-terminal segments of ColH, S2b, and S3. E. coli BL21 and the pGEX-4T series plasmids (Pharmacia Biotech, Uppsala, Sweden) were used as a hostvector system for the production of GST fusion proteins. A ColH fraction containing the gelatinase, prepared from C. histolyticum cultures as described previously (10), was used as partially purified ColH. ColH purified from cultures of recombinant B. subtilis DB104 (16) was used as recombinant ColH (rColH).
Growth Conditions-Recombinant B. subtilis strains were grown as described previously (16). For screening recombinant plasmids, E. coli DH5␣ transformants were grown in Luria-Bertani medium supplemented with 100 g of ampicillin/ml. For the preparation of GST fusion proteins, all recombinant E. coli BL21 cells were grown in 2YT-G medium consisting of the following ingredients: 16 g of tryptone, 10 g of yeast extract, 5 g of NaCl, 20 g of glucose, and 100 mg of ampicillin/liter. The expression of fusion proteins was induced by the addition of 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside (Wako Pure Chemical Industries, Ltd., Osaka, Japan).
Purification of the Recombinant ColH-An N-terminal peptide of ColH, consisting of S1 and S2a (S1S2a), was constructed as follows. A 2.9-kilobase HaeIII-PstI fragment containing the colHЈ gene (10) was ligated into the EcoRV and PstI sites of Bluescript II KS(ϩ). Into the downstream SacI site of the resulting plasmid (pCHC200) a synthetic oligonucleotide, 5Ј-GCTTAATTAATTAAGCAGCT-3Ј, was inserted so that translation from a colHЈ transcript terminates at the TAA codon. The resulting plasmid was designated as pCHC201. A 3.1-kilobase BssHII fragment, which encodes amino acids 1 to 766, (numbered as in Ref. 10) corresponding to S1S2a plus an extra 5-amino acid stretch (Ala-Arg-Gly-Ile-His), was isolated from pCHC201 and then ligated into the SmaI site of pHY300PLK. The resulting plasmid was introduced into B. subtilis DB104. The transformant was grown, and the culture supernatant was subjected to ammonium sulfate precipitation and gel filtration as described elsewhere (16). The fractions with Pz-peptidase activity after gel filtration were applied to a hydrophobic interaction column (Ether-Toyopearl, bed volume 14 ml; Toso, Tokyo, Japan), which was pre-equilibrated with 50 mM Tris-HCl (pH 7.5) containing ammonium sulfate (40% saturation). Proteins were eluted by a 280-ml linear gradient from 40 to 0% ammonium sulfate in 50 mM Tris-HCl (pH 7.5). The Pz-peptidase activity eluted at 29% ammonium sulfate. The active fraction was dialyzed against 50 mM Tris-HCl (pH 8.5), and then applied to a MonoQ column. Chromatography was performed as described elsewhere (16) except that the pH was changed to 8.5.
A recombinant N-terminal peptide containing only S1 was produced as follows. An oligonucleotide encoding the C-terminal peptide of S1 (Val 564 to Ser 683 ) followed by a stop codon was prepared by polymerase chain reaction using two primers (5Ј-GTGCCTTTTGTAGCTGATGA-3Ј and 5Ј-CCGCGGTTAGGAATCACCTTCGTTTGGTA-3Ј) and pCHC201 plasmid DNA. The amplified fragment was cloned into the pT7Blue T-vector (Novagen, Madison, WI). An 0.33-kilobase SphI-SacII fragment of this plasmid was substituted for a 0.61-kilobase SphI-SacII fragment of pCHC201. S1 was purified from cultures of recombinant B. subtilis as described above.
An inactive ColH mutant was produced as follows. Replacement of the GAA codon encoding Glu 416 , one of the residues forming a putative catalytic center, with a GAT codon (Asp) was performed using pCHC201 plasmid DNA and a Transformer site-directed mutagenesis kit (CLON-TECH Laboratories, Palo Alto, CA), according to the instruction of the manufacturer. Oligonucleotides, 5Ј-GCAAATAATGTGTATAATCATG-TCTAAATAATTC-3Ј and 5Ј-GTGACTGGTGAGGCCTCAACCAAGTC-3Ј, were used as a mutagenic primer and a selection primer, respectively. The mutant enzyme, designated as rColH(E416D), was produced using a shuttle vector, pAT19 (17) as described elsewhere (16).
Purification of GST Fusion Proteins-The GST gene fusion system was employed to construct fusion proteins between GST and ColH C-terminal segments. A DNA fragment encoding segments 2a and 2b (S2aS2b, Pro 678 -Asp 860 ) was obtained by polymerase chain reaction using pCHC11 plasmid DNA (10), a 5Ј-primer, 5Ј-CCCGGGCCAAAC-GAAGGTGATTCCAA-3Ј, and a 3Ј-primer, 5Ј-CTCGAGTTAATCTGTA-ATCTTAATCTTCA-3Ј. A DNA fragment encoding segment 2b (S2b, Glu 767 -Asp 860 ) was also obtained in the same way using pCHC302 plasmid DNA, 5Ј-pGEX primer (Pharmacia), and the same 3Ј-primer as described above. These fragments were cloned into pT7Blue T-vector. They were inserted into pGEX-4T-2 vector DNA to express the enzyme fragments as GST fusion proteins. E. coli DH5␣ competent cells were transformed with each ligation mixture by electroporation. After the nucleotide sequence of each fusion gene was confirmed, E. coli BL21 was transformed with the recombinant plasmid. Each clone was grown in 100 ml of 2YT-G medium to an optical density at 600 nm of 2.5, and isopropyl-1-thio-␤-D-galactopyranoside was added. Purification of the fusion protein was performed with a glutathione-Sepharose column (bed volume, 0.5-ml; Pharmacia) as described by the manufacturer. The fusion proteins for S2aS2b and S2b were designated as FP342 and FP344, respectively. GST fusion proteins for C-terminal fragments of various lengths were also constructed. The insert DNA fragment in pCHC116⌬3 (see above) encoding S2b plus S3 (S2bS3) was deleted from its 5Ј end using appropriate restriction enzymes. Each fragment was inserted into the SmaI site of a suitable GST fusion vector of the pGEX-4T series (Pharmacia) so that the reading frame is intact. After the nucleotide sequence of each fusion gene was confirmed, fusion proteins carrying varying length C-terminal fragments were produced as described above, and designated as FP302-FP306.
DNA Manipulations and Sequencing-Restriction endonucleases were purchased from Takara Shuzo Co. (Kyoto, Japan), Toyobo (Osaka, Japan), and New England Biolabs (Beverly, MA). The DNA ligation kit was a product of Takara Shuzo. All recombinant DNA procedures were carried out as described by Sambrook et al. (18). All constructs were sequenced to confirm the reading frame on an automated nucleotide sequencer (model ABI PRISM 377, Perkin-Elmer, Foster City, CA). An ABI PRISM dye terminator cycle sequencing ready kit with AmpliTaq DNA polymerase, FS (Perkin-Elmer), and pGEX primers (Pharmacia) were used for sequencing the GST fusion constructs. A Thermo Sequenase fluorescent labeled primer cycle sequencing kit with 7-deaza-dGTP (Amersham Japan Inc., Tokyo, Japan) and M13 dye primers (Perkin-Elmer) were used for sequencing all other constructs.
Enzyme Assay, Protein Determination, and SDS-PAGE-The activities of the recombinant collagenases were determined using Pz-peptide (4-phenylazobenzyloxycarbonyl-Pro-Leu-Gly-Pro-D-Arg, Sigma) or insoluble collagen from bovine achilles tendon (Worthington Biochemical Co., Freehold, NJ), as described elsewhere (16). The GST activity of GST fusion proteins was assayed using 1-chloro-2,4-dinitrobenzene as described in the supplier's protocol. Protein concentrations were determined by the Bradford method (19) using the Bio-Rad protein assay reagent (Bio-Rad) with bovine serum albumin (BSA) as a standard. All the assays were carried out in triplicate. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) was performed using 7.5, 12.5, or 15% polyacryamide gels, and the gels were stained with Coomassie Brilliant Blue R as described previously (20). Band intensity was determined using a flat bed scanner and the public domain computer program, NIH Image (developed at the National Institute of Health, Bethesda, MD, and available from the Internet by anonymous FTP from zippy.nimh.nih.gov).
Determination of the N-terminal Amino Acid Sequence of C-terminal Peptides-Two hundred picomoles of the isolated C-terminal fragments obtained from fusion proteins (FP302, FP305, FP342, and FP344) were blotted on a polyvinylidene difluoride membrane using ProSorb devices (Perkin-Elmer) as described by the supplier. Twenty amino acid residues from the N terminus were determined for each fragment on an automatic protein sequencer (Model 492, Perkin-Elmer). All the isolated fragments possessed the expected N-terminal amino acid sequences.
Determination of the Molecular Mass of a C-terminal Peptide-A GST fusion protein containing S2bS3 (FP302, 1.1 mg) was cleaved by incubation with 10 units of thrombin for 5 h at room temperature. The reaction mixture was dialyzed twice against 2 liters of phosphatebuffered saline (PBS) at 4°C to remove glutathione, and the cleaved N-terminal GST fragment was removed by adding glutathione-Sepharose beads (100 l). After incubation at room temperature for 30 min, the suspension was centrifuged at 500 ϫ g for 5 min and the bead treatment was repeated three times. The supernatant was dialyzed against 2 liters of distilled water at 4°C for 12 h with two changes, filtrated through an 0.45-m filter, and lyophilized. The sample was dissolved in 100 l of distilled water, mixed with 3,5-dimethoxy-4hydroxycinnamic acid solution (10 mg/ml of 50% ethanol), and analyzed by matrix-assisted laser desorption time-of-flight mass spectrometry (model Kompact MALDI III, Kratos, Manchester, United Kingdom) with BSA as an internal standard.
Collagen Binding Assay-Collagen binding was assayed as follows, unless otherwise stated. Five milligrams of insoluble collagen (type I, C-9879; Sigma) were added to an Ultrafree microcentrifugal device with an 0.22-m low-binding Durapore membrane (Millipore, Bedford, MA), which was placed in a microcentrifuge tube. All steps were carried out at room temperature. Two hundred microliters of collagen binding buffer (CB buffer: 50 mM Tris-HCl, 5 mM CaCl 2 , pH 7.5) were added to swell the collagen fibers. After incubation for 30 min, the tube was centrifuged at 15,000 ϫ g for 15 min. Centrifugation was repeated after changing the direction of the tube in the rotor. The collagen precipitate was resuspended in 60 l of CB buffer containing 100 pmol of enzyme and incubated for 30 min. The filtrate was collected by centrifugation at 15,000 ϫ g for 15 min, and used for GST assay or analysis by SDS-PAGE. To determine the binding affinities of rColH(E416D) and various C-terminal segments by Scatchard plot analysis, the collagen binding assay was carried out with the following modification. Collagen was washed with 200 l of CB buffer supplemented with 150 mM NaCl, and resuspended in 100 l of the same buffer containing various concentrations (25-400 g/ml) of the mutant enzyme or various fragments. A calibration curve was constructed for each sample by densitometory after SDS-PAGE, and was used to quantitate their amounts in the filtrates. The results obtained by triplicate assay were analyzed on a Scatchard plot, and the dissociation constant (K d ) and the number of binding sites on insoluble collagen (B max ) for each protein were determined by the least-square method.

RESULTS
Binding of ColH and Gelatinase to Collagen-To examine collagen binding, 6 g of ColH in 60 l of CB buffer was added to 5 mg of swollen collagen. After the mixture was incubated at room temperature for 30 min and centrifuged, a 20-l sample of the supernatant was analyzed by SDS-PAGE. As shown in Fig.  1, the gelatinase but not ColH was detectable in the filtrate, suggesting that ColH but not the gelatinase binds to insoluble type I collagen. Various smaller polypeptides were also detected in the filtrate. These might have resulted from hydrolysis of insoluble collagen by ColH. To inhibit the collagenolytic activity of ColH, the incubation was carried out at 4°C for 30 min, and then the filtrate was analyzed by SDS-PAGE (Fig. 1). The gelatinase was again detectable but neither ColH nor the smaller polypeptides were detected in the filtrate, indicating that ColH binds collagen without degrading it at 4°C. Therefore, subsequent collagen binding assays of active collogenases were carried out at 4°C.
The difference in the collagen-binding capability of the two enzymes suggests that the C-terminal peptide which is absent in the gelatinase is involved in collagen binding. However, it has not yet been proved that the 98-kDa gelatinase is produced from ColH by cleaving off the C-terminal region. To test this possibility, two recombinant enzymes, 116-kDa rColH and its truncated form consisting of S1S2a were purified from cultures of recombinant B. subtilis cells. Twenty-five picomoles of each enzyme were mixed with 5 mg of swollen insoluble collagen, and the mixture was incubated at 4°C for 30 min. A sample containing 1 g of protein was analyzed by SDS-PAGE (Fig. 2). While the full-length rColH bound to insoluble collagen, the truncated form did not. Activities against insoluble collagen and Pz-peptide, a synthetic water-soluble substrate, were determined for the two recombinant polypeptides (Table I). The Pz-peptide hydrolyzing activity of the truncated form was 73% of that of rColH, while its collagenolytic activity was only 7%. This result suggests that the C-terminal peptide is required for binding and hydrolysis of insoluble collagen.
Collagen Binding of GST Fused to the C-terminal Peptide-The C-terminal peptide, S2bS3, was fused to the C terminus of GST to see if it conferred the ability to bind to collagen on GST. The apparent molecular mass of the fusion protein estimated by SDS-PAGE (Fig. 3), 51 kDa, agreed with the 50,526 Da value calculated from the nucleotide sequence of the corresponding gene. The fusion protein was cleaved by thrombin and SDS-PAGE showed two peptides. One migrated to the same position as the product generated by cleavage of GST alone (26 kDa), and the other (28 kDa) reacted with anti-collagenase antiserum (data not shown). The peptide was purified, its Nterminal amino acid sequence was determined, and it coincided with the predicted sequence. Its apparent molecular mass differed from the value (24,378 Da) calculated for the S2bS3 containing peptide. To determine the molecular mass more accurately, the peptide was analyzed by mass spectrometry. Its molecular mass was determined to be 24,370 Da and the reason for its abnormal mobility in SDS-PAGE is unknown.  Samples of the GST fusion protein before and after cleavage with thrombin were combined, added to CB buffer containing BSA, chicken ovalbumin, and horse myoglobin (6 g each) and then the final volume was adjusted with CB buffer to 60 l. After incubation with 5 mg of insoluble collagen, and centrifugation, the supernatant was analyzed by SDS-PAGE (Fig. 3). All proteins except the fusion protein and S2bS3 were present in the filtrate, indicating that S2bS3 and the GST fusion protein specifically bound to collagen.
To determine the collagen binding condition, the S2bS3 peptide was incubated with insoluble collagen under various conditions. When 20 g of the peptide was incubated with varying amounts of insoluble collagen for 30 min at room temperature, it was almost completely bound to 10 mg or more collagen. When 20 g of the peptide was incubated with 15 mg of collagen for varying times, approximately 70% of the peptide bound to collagen after 2 min, and the binding reached its maximal level after 30 min. It was shown that the binding capacity of collagen for the S2bS3 peptide is more than 2 g/mg collagen, and that incubation at room temperature for 30 min is sufficient for completion of the binding reaction.
The effect of pH on collagen binding was examined by determining the GST activity of the GST fusion protein in the filtrate after incubating using the standard assay except the buffer was changed to: 50 mM bis-Tris-HCl containing 5 mM CaCl 2 (pH 5.0) or CB buffer (pH 7.0) or CB buffer (pH 9.0). Mean Ϯ S.D. of collagen binding at pH 5, 7, and 9 were 97.9 Ϯ 0.0, 97.5 Ϯ 0.3, and 100 Ϯ 0.0%, respectively. The Pz-peptidase activity of rColH determined in these buffers was 0.98 Ϯ 0.98, 76.5 Ϯ 2.4, and 60.1 Ϯ 1.2 units/nmol, respectively. When 5 mM EDTA was added to the buffer in place of 5 mM CaCl 2 , the Pz-peptidase activity of rColH was completely inhibited, while the collagen binding activity of the GST fusion was 88.6 Ϯ 1.2%. The addition of gelatin at a final concentration of 10 mg/ml to CB buffer did not affect binding of the GST fusion protein (96.0 Ϯ 0.2%). The addition of sodium chloride at a final concentration of 1.5 M slightly affected binding (92.3 Ϯ 0.3%).
Deletion Analysis of the Collagen-binding Domain-Two truncated peptides, S1S2a and S2bS3, exhibited Pz-peptidase and collagen binding activities, respectively. Given the homology between S2a and S2b, it appears that S3 is a collagenbinding domain, and that S2b and possibly S2a form an interdomain structure. To examine this possibility, S2bS3 was deleted to various lengths from its N terminus and fused to the C terminus of GST (Fig. 4A). The fusion proteins were purified to homogeneity as shown in Fig. 4B. These fusion proteins were incubated with insoluble collagen, and the GST activities in the filtrates were determined (Table II). Since GST activity differed slightly from one fusion protein to another, the collagen binding activity of each fusion protein was determined by calculating the difference of the GST activity in supernatants incubated with and without collagen. As the C-terminal peptide was shortened, collagen binding activity decreased. However, the binding activity of the fusion protein containing S3 (FP305) was still high, being more than 90% of that containing S2bS3 (FP302). When the C-terminal peptide was shorter than S3, binding decreased significantly. The fusion proteins were cleaved by thrombin, and the ability of the C-terminal peptides to bind insoluble collagen were tested by SDS-PAGE (Fig. 4C). S3 alone bound collagen and the shorter peptide bound weakly, as was expected. The difference in band intensity between peptides incubated with and without collagen matched the binding activity of the corresponding fusion protein determined by its GST activity.
Scatchard Analysis of Collagen Binding-Since GST is known to exist as a dimer (21), GST fusion proteins may also exist as a dimer. Therefore, the GST fusion proteins are not suitable for the quantitative evaluation of the collagen binding affinity of various C-terminal segments. Thus, the fusion proteins were cleaved by thrombin, and the collagen binding assay was performed with varying concentrations of the purified Cterminal peptides. Scatchard analysis of the data (Fig. 5) showed that S3 alone bound to insoluble collagen with low but  bound. S2bS3 showed biphasic plots, one with higher affinity (K d ϭ 3.39 ϫ 10 Ϫ7 M, B max ϭ 0.201 nmol/mg collagen) and the other with low affinity (K d ϭ 2.11 ϫ 10 Ϫ6 M, B max ϭ 0.628 nmol/mg collagen). The affinity of the full-length enzyme was determined by using rColH(E416D), in which the Glu 416 residue forming the putative catalytic center is replaced by an Asp residue. This replacement abolished the collagen hydrolytic but not the collagen binding activity. It had the highest affinity (K d ϭ 9.95 ϫ 10 Ϫ8 M) and lowest capacity (B max ϭ 0.108 nmol/mg collagen). Biphasic plots were not observed for this full-length enzyme. Since its achievable molar concentration was lower than that of S2bS3, it is uncertain if it could show biphasic binding. Catalytic Activity of the N-terminal Segment S1-By introduction of nested deletions into plasmid, pCHC201, plasmids encoding shorter peptides, which terminated at different positions such as Val 487 , Glu 496 , Arg 574 , Glu 586 , Ser 603 , Met 630 , Gly 675 , and Asp 682 were created. When we introduced these plasmids into E. coli DH5␣ and determined the Pz-peptidase activity of their cell lysates, only the sample from cells expressing a peptide ending at Asp 682 showed Pz-peptidase activity. 2 Since this residue is located near the end of S1, this result suggested that S1 is the catalytic domain. S1, ending at Ser 683 , was purified from the culture of recombinant B. subtilis. Its Pz-peptidase activity was even higher than that of full-length rColH, while its collagenolytic activity was only 5.2% of rColH (Table I). Use of an excess of S1 (75 g/assay) in the collagenase assay did not cause a proportional increase in activity (specific activity, 0.0022 Ϯ 0.0004 units/pmol).

DISCUSSION
The fact that no significant collagen binding activity was observed for the gelatinase suggests that the C-terminal pep-tide of ColH is essential for collagen binding. Binding of ColH was not inhibited by low temperature unlike activity, suggesting that binding is independent of hydrolysis. In a commercial preparation of C. histolyticum collagenase, Bond and Van Wart (22)(23)(24) reported the presence of multiple forms of collagenases which showed similar peptide maps and amino acid analyses. They suggested that this organism harbors multiple collagenase genes (4). Although the two C. histolyticum enzymes used in the present study showed identical N-terminal amino acid sequences and similar peptide maps (10), it could not be excluded that they are the products of different genes. If this is the case, collagen binding might depend on the specific amino acid sequence throughout the collagenase. To exclude this possibility, two recombinant enzymes were produced. One is a full-length form, rColH, and the other is a truncated form, rColHЈ(S1S2a), lacking segments 2b and 3. Their molecular masses estimated by SDS-PAGE were 116 and 87 kDa, respectively, which coincided with the calculated value, 112,977 and 89,661 Da. As expected, rColH did bind to insoluble collagen, but rColHЈ(S1S2a) did not. This result confirmed the necessity of the C-terminal fragment for collagen binding. However, hydrolytic activity on Pz-peptide was retained after truncation (70% of rColH activity), because binding is not necessary for hydrolysis of this water-soluble substrate.
The properties observed for the fusion proteins also support a multidomain structure. First, a GST fusion protein containing either peptide S2bS3 or S3 bound to collagen and exhibited GST activity. Second, the catalytic activity of ColH differed from the collagen binding activity in its pH dependence and requirement for divalent cations.
The site of the collagen-binding domain was examined using GST fusion proteins with varying sized C-terminal fragments. All of the fusion proteins (FP302 to FP306) bound efficiently to glutathione affinity columns. Since binding requires normal function of the GST moiety, it seems that these C-terminal fragments did not affect significantly the conformation of GST. Qualitative analysis by SDS-PAGE showed that S3 isolated from FP305 bound well to insoluble collagen, and that further deletion caused a marked decrease in its binding activity (the peptide from FP306). Scatchard analysis confirmed that S3 had collagen binding activity, although its K d value was 2 orders of magnitude higher than that of the mutant full-length enzyme. Therefore, S3 can be considered as a minimal region required to form a collagen-binding domain. The S3 in the construct, FP305, starts at Pro 861 . Since proline residues are often at the termini of ␣-helices or ␤-strands, the residue may be located at the beginning of the domain structure. Purification of GST fusion proteins containing shorter fragments such as one from Thr 924 to Arg 981 and one from Pro 954 to Arg 981 by affinity chromatography was unsuccessful probably due to interference with binding to glutathione by these shorter fragments. Therefore, it seems likely that the first half of S3 is essential for the formation of the domain structure. In C. perfringens collagenase, ColA, the S3 homologue is tandemly repeated at the C terminus. Whether or not this repeat enhances collagen binding activity remains to be determined.
There is no significant similarity in amino acid sequence between bacterial and eukaryotic collagenases. The three-dimensional structure of a fibroblast collagenase determined by Li et al. (25), has two domains, a catalytic domain (161 amino acid) and a hemopexin-like one (189 amino acid), which are connected by a short linker sequence (17 amino acid). The latter domain, which is believed to participate in the recognition of the macromolecular substrate, has a four-bladed ␤-propeller structure consisting of 16 ␤-strands. The secondary structure of S3 (ϳ120 amino acid) predicted by the method of Chou and Fasman (26) and Garnier et al. (27) contains at most 9 ␤-strands. Considering the necessity of at least six blades for the formation of a stacked propeller structure suggested by Murzin (28) and the four-bladed perpendicular propeller structure of the eukaryotic collagen-binding domain, it seems unlikely that S3 forms a ␤-propeller structure. Determination of the three-dimensional structure of the collagen-binding domain in bacterial collagenases would facilitate understanding of the molecular mechanism for the interaction between these enzymes and collagen.
Although neither S2aS2b nor S2b showed any binding activities, addition of S2b (FP302) to S3 increased its binding affinity close to that of the full-length enzyme. This enhancement could be explained by the hypothesis that S2b assists proper folding of the minimal collagen-binding domain, S3. There is an alternative explanation that S3 bound to insoluble collagen could modify the local structure of the substrate, where S2b could bind subsequently. The International Polycystic Kidney Disease Consortium (29) showed that the amino acid sequence of S2 is homologous to that of a repetitive sequence (PKD domain) of the protein encoded by the PKD1 gene, the most common gene responsible for autosomal dominant polycystic kidney disease. They speculated that the S2-encoding fragment has been horizontally transferred from eukaryotic cells to prokaryotic cells. It may be possible that the bacterial collagenase genes have evolved from an ancestral bacterial zinc-metalloprotease gene by insertion and duplication of a short stretch between the catalytic domain and the binding domain. Further precise characterization of S2 is necessary to elucidate its functional role.
The collagen binding activity of the C-terminal segments was not inhibited by gelatin, a denatured collagen which lacks a triple helical structure, suggesting that the interaction of the collagen-binding domain with native collagen is conformationdependent. Type I insoluble collagen exhibits a maximal binding capacity for rColH(E416D) of 108 pmol of enzyme/mg of collagen, which gives an average of one binding site every 32.5 tropocollagen molecules. It is likely that only binding sites on the surface of the collagen fibrils are accessible to the collagenase. Binding of the enzyme may occur only at the sites where multiple tropocollagen molecules form a staggered array of collagen fibrils, where there is a hole between tropocollagen molecules. The C-terminal peptides are smaller and hence could access more binding sites than the full-length enzyme. All of the C. histolyticum collagenases initially cleave collagen molecules at hyperactive sites, and thereafter successively degrade the fragments into small, dialyzable peptides (7). Hyperactivity might result from the selective binding of the enzyme to specific sites.
Preliminary experiments showed that a fusion protein carrying the C-terminal fragment (S2bS3), when injected subcutaneously into nude mice, remained at the sites of injection for up to 10 days. Consequently, the collagen binding ability of the C-terminal fragment is not restricted to the substrate preparation used in the present study but it can bind to "living collagen fibers" in vivo, suggesting an important role of the segments in the tissue degrading activity of ColH.
All C. histolyticum collagenases contain one zinc molecule (4), and its depletion by treatment with 1,10-phenanthroline abolishes their hydrolytic activity. Within S1 there is a zinc binding consensus sequence, HEXXH, which is at the catalytic center in other zinc-metalloproteases. Since a truncated enzyme, rColHЈ(S1), showed higher activity against Pz-peptide than the full-length enzyme, the catalytic center seems to be folded into its normal conformation without S2a, S2b, or S3. Further deletions abolished its catalytic activity, and, thus, S1 seems to be the minimal region necessary to form a catalytically active structure. The residual collagenolytic activity of the truncated enzymes may reflect catalytic activity dependent on random collision which is greatly stimulated by a collagenbinding domain. Alternatively, swollen type I collagen may be partly denatured and the residual activity may be due to the gelatinolytic activity of the peptide. Since a collagenase assay using an excess amount of the truncated enzyme reached a saturation level, the latter case seems to be likely. More precise kinetic study is necessary to draw a final conclusion about this point.