The hemopexin-like domain (C domain) of human gelatinase A (matrix metalloproteinase-2) requires Ca2+ for fibronectin and heparin binding. Binding properties of recombinant gelatinase A C domain to extracellular matrix and basement membrane components.

The binding properties of the COOH-terminal hemopexin-like domain (C domain) of human gelatinase A (matrix metalloproteinase-2, 72-kDa gelatinase) were investigated to determine whether the C domain has binding affinity for extracellular matrix and basement membrane components. Recombinant C domain (rC domain) (Gly417-Cys631) was expressed in Escherichia coli, and the purified protein, identified using two antipeptide antibodies, was determined by electrospray mass spectrometry to have a mass of 25,925 Da, within 0.1 Da of that predicted. As assessed by microwell substrate binding assays and by column affinity chromatography, the matrix proteins laminin, denatured type I collagen, elastin, SPARC (secreted protein that is acidic and rich in cysteine), tenascin, and Matrigel™ were not bound by the rC domain. Unlike the hemopexin-like domains of collagenase and stromelysin, the rC domain also did not bind native type I collagen. Nor were native or denatured types II, IV, V, and X collagen, or the NC1 domain of type VII collagen bound. However, binding to heparin and fibronectin (Kd, 1.1 × 10−6 M) could be disrupted by 0.58-0.76 and 0.3 M NaCl, respectively. Using nonoverlapping chymotrypsin-generated fragments of fibronectin, binding sites for the rC domain were found on both the 40-kDa heparin binding and the 120-kDa cell binding fibronectin domains (Kd values, ∼4-6 × 10−7 M). The Ca2+ ion, but not the potential structural Zn2+ ion, were found to be essential for maintaining the binding properties of the protein. The apo-form of the rC domain did not bind heparin, and both ethylenediaminetetraacetic acid and the specific Ca2+ ion chelator 1,2-bis(2-aminophenoxy) ethane-N,N,N′,N′-tetraacetic acid, but not the Zn2+ ion chelator 1,10-phenanthroline, eluted the holo form of the rC domain from both heparin-Sepharose and fibronectin. Inductive coupled plasma mass spectrometry also did not detect a Zn2+ ion in the rC domain. In contrast, reduction with 65 mM dithiothreitol did not interfere with heparin binding, further emphasizing the crucial structural role played by the Ca2+ ion. Together, these data demonstrate for the first time that the hemopexin-like domain of gelatinase A has a binding site for fibronectin and heparin, and that Ca2+ ions are important in maintaining the structure and function of the domain.

The matrix metalloproteinases (MMPs) 1 constitute a family of proteolytic enzymes that together can degrade all components of the extracellular matrix and basement membranes, with each MMP having a distinct substrate preference (1,2). MMP activity plays a major role during physiological and pathological processes, including embryogenesis, metastasis (3,4), and inflammatory diseases (5,6). Most soluble MMPs are secreted as proenzymes and share homologous primary and tertiary structures organized into distinct structural domains, with some differences in domain composition and number (6). These functionally and structurally defined domains include the NH 2 -terminal zymogen domain containing the conserved PRCGXPD motif involved in enzyme latency (7) and a Zn 2ϩ and Ca 2ϩ ion binding catalytic domain. As with other proteinases, the specificity of peptide bond cleavage is determined by the S and SЈ subsite defining amino acid residues (8). Equally important are discrete substrate binding domains, or smaller functional modules, located outside of the active site, which form specialized exosites (9) to target proteolytic activity in tissues and are essential for cleavage of some substrates. Immediately adjacent to the catalytic site in gelatinase A (MMP-2, 72-kDa gelatinase) and gelatinase B (MMP-9, 92-kDa gelatinase) is a fibronectin type II-like module triple repeat (10,11), which forms a collagen binding domain (CBD) with strong affinity for elastin and denatured types I (12,13), IV, and V collagens, and native types I, V, and X collagens (9, 13), 2 proteins degraded by the gelatinases. Following the catalytic domain is a variably long linker, which in collagenase-1 (MMP-1) may be important for triple helicase activity (14). The linker connects to the COOH-terminal domain (C domain), comprising four he-* This work was supported in part by Grant 006388 from the National Cancer Institute of Canada and initially by a grant from the Medical Research Council of Canada. 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. ‡ Present address: Dept. of Biochemistry, University of Washington, Seattle, WA.
Even though the primary and tertiary structures of the MMP C domains share extensive homology, they possess a range of different properties that are MMP-specific. The C domain of gelatinase B binds the natural tissue inhibitor of matrix metalloproteinases-1 (TIMP-1) (11,19,20) whereas the C domain of gelatinase A is involved in binding to TIMP-2 (21) and to cell membranes (9,22) on concanavalin A (ConA) activated normal cells (23) and tumor cells (24,25). In addition, latent gelatinase A, in complex with TIMP-2, is activated on interaction with active membrane type MMPs (MT-MMPs) (26). Deletion mutants have demonstrated the requirement of the C domain of gelatinase A for cell binding and activation of the latent enzyme on cell surfaces (27), whereas the C domain of the collagenases mediates substrate specificity for native, triple helical type I collagen. In the absence of the C domain, the catalytic domain alone of collagenase-1 (28,29) and neutrophil collagenase (30,31) cleaves synthetic peptides and casein, but not native collagen. Indeed, the C domain alone of collagenase-1 binds native type I collagen (32)(33)(34), as can the C domain of stromelysin-1 (33,34), even though collagen is not cleaved by stromelysin-1. In contrast, neither the activity nor the substrate specificity of the stromelysins is dramatically modified by the removal of the C domain (35)(36)(37)(38). Likewise, the gelatinases degrade gelatin and native type IV collagen in a manner that appears independent of the presence of the C domain (39).
The gelatinases also efficiently degrade fibronectin, native types V, VII, and X collagen, and elastin (40 -42). However, the importance of the C domain for these activities is unknown. Indeed, other than for elastin, native types I and V collagen, and gelatins that are bound by the CBD (12,13,42), exosites conferring specificity for these substrates have yet to be identified. The gelatinases may depend on the C domain for binding fibronectin or native type VII and X collagens as a requisite for efficient catalysis and triple helicase activity, but this has not been directly tested. In the present study we investigated the potential binding properties of a recombinant C domain of human gelatinase A for a number of extracellular matrix and basement membrane components. We found that the C domain exhibits strong binding properties for fibronectin and heparin and that the binding is dependent on the structural Ca 2ϩ but not Zn 2ϩ ion in the C domain.

EXPERIMENTAL PROCEDURES
Materials and Reagents-Human types IV and V collagen (nonpepsin-treated) were from Becton Dickinson and Life Sciences; type X collagen and elastin were from Sigma; fibronectin fragments (40 and 120 kDa) and laminin were from Life Sciences; Matrigel TM was obtained from Beckton Dickinson; tenascin was from Chemicon and Life Sciences; Affi-Gel heparin, Affi-Gel 10, and 10-DG columns were from Bio-Rad; chelating Sepharose 6B, gelatin-Sepharose 4B, heparin-Sepharose CL-6B, and CM-Sepharose Fast Flow were from Pharmacia Biotech Inc.; G-10 Sephadex, 1,2-bis(2-amino phenoxy)-ethane-N,N,NЈ,NЈ-tetraacetic acid (BAPTA), and keyhole limpet hemocyanin were from Sigma; and Freund's complete and incomplete adjuvants were from Calbiochem. Human gelatinase A cDNA was a kind gift from Dr. K. Tryggvason (Karolinska Institute, Stockholm, Sweden) (43) secreted protein that is acidic and rich in cysteine (SPARC) was kindly provided by Dr. J. Sodek (University of Toronto, Toronto, Ontario Canada); the NC1 domain of type VII collagen was generously provided by Dr. P. Rousselle (Centre National de la Recherche Scientifique, Lyon, France); and type II collagen was a gift from Dr. J. Mort (Shriners Hospital, Montreal, Quebec, Canada).
Acid-soluble native type I collagen was prepared from rat tail tendons by extraction with 0.5 M acetic acid and purified by differential precipitation with 1.7 M NaCl (44). Fibronectin was affinity purified from bovine serum by binding to gelatin-Sepharose 4B and elution with a 0 -10% Me 2 SO gradient (45). The Me 2 SO was then removed from the fibronectin using a 10-DG column. Purity was confirmed by SDS-PAGE analysis under reducing and nonreducing conditions.
Polyclonal Antipeptide Antibodies-Rabbit polyclonal ␣His 6 and ␣72ex12 antipeptide antibodies were raised against a peptide present in the polyHis fusion tag and from part of a surface exposed ␤-strand of hemopexin module III of the gelatinase A C domain, respectively. Synthetic peptides were purified by high performance liquid chromatography before coupling to keyhole limpet hemocyanin (46). After emulsifying in Freund's complete adjuvant 1 mg of each peptide was injected subcutaneously into two rabbits each. Three to four biweekly boosts of peptide in incomplete Freund's adjuvant were administered intramuscularly. Whole serum was affinity-purified against peptide coupled to Affi-Gel 10 and quantitated by enzyme-linked immunosorbent assay.
Recombinant Gelatinase A Hemopexin-like C Domain Expression Construct-The 5Ј-and 3Ј-boundaries of the cDNA encoding the C domain and linker of gelatinase A (Gly 417 -Cys 631 ) were defined by the 5Ј-boundary of exon 9 and the natural stop codon in exon 13, respectively. The cDNA was polymerase chain reaction-amplified from the full-length cDNA of human gelatinase A using the primers 5Ј-GCTAGCTAGCGGGGCCTCTCCTGACATT-3Ј and 5Ј-CTTAAGCT-TCAGCAGCCTAGCCAGTC-3Ј, which added an NheI site and a BamHI site in the 5Ј-and 3Ј-extensions, respectively. The purified 688-base pair product was digested with NheI and BamHI and ligated into the expression vector pGYMX (13), which expresses recombinant protein with a short NH 2 -terminal fusion tag comprising an initiation methionine, a His 6 tag, and a factor X a cleavage site (13). Sequencing (47) of the plasmid, pGYMX9 -13, confirmed the fidelity and reading frame of the cDNA.
Recombinant Protein Expression in Escherichia coli-E. coli strains were screened for recombinant protein expression after growth in 1% (w/v) tryptone (Becton Dickinson), 0.8% (w/v) yeast extract (Becton Dickinson), 0.5% (w/v) NaCl, and 100 g/ml ampicillin, pH 7.4. Log phase seed cultures were used to inoculate either 35-or 60-liter Chemap fermentor cultures, which were then grown under controlled constant temperature and aeration conditions for 24 h. Collected cells were washed and then lysed, inclusion bodies were dissolved in guanidinium-HCl, and recombinant protein was refolded prior to purification using procedures slightly modified (48) from those previously demonstrated (13) to produce correctly folded, disulfide cross-linked monomeric recombinant protein. In brief, disulfide bond exchange was performed in 0.1 M sodium borate, pH 10, under highly aerated conditions for 2 h, and the guanidinium was then slowly removed by step dialysis to refold the protein.
Purification of the Recombinant C Domain-Refolded protein was loaded on a Zn 2ϩ -charged chelating Sepharose 6B Column (V t , 20 ml). After extensive washes with chromatography buffer (100 mM sodium dibasic phosphate buffer, 0.5 M NaCl, pH 8.0), nonspecifically bound bacterial proteins were eluted with a step pH gradient (pH 8.0 -6.0) in 1.0 M NaCl. Recombinant protein was then eluted with a 0 -400 mM imidazole gradient developed over 100 ml, and the pooled fractions buffer was exchanged to 50 mM Tris-HCl, pH 7.4, by gel filtration prior to snap freezing in liquid N 2 .
SDS-PAGE and Enzymography-Heat-denatured protein samples were separated under reducing (65 mM DTT) or nonreducing conditions by SDS-PAGE according to the method of Laemmli (49) using 15% polyacrylamide gels. Protein bands were stained with Coomassie Brilliant Blue R-250. Samples analyzed by enzymography were electrophoresed nonreduced on 10% polyacrylamide gels copolymerized with 1 mg/ml gelatin according to the method of Overall and Limeback (50).
Immunoblot Analysis-To confirm the identity of the purified protein, reduced and nonreduced recombinant protein was blotted onto a polyvinylidene difluoride membrane (Millipore), reacted with affinitypurified ␣His 6 or ␣72ex12 antibodies, and then detected with peroxidase-conjugated goat anti-rabbit antibody and enhanced chemiluminescence reagents (Amersham Corp.) on Kodak SB-5 film (Eastman Kodak Co.).
Mass Spectrometry-The mass of the rC domain was measured by electrospray mass spectrometry using a PESCIEX API 300 spectrometer after sample injection on a C18 high performance liquid chromatography column at 50 l/min. The Zn 2ϩ ion content of the rC domain was measured three times by inductive coupled plasma mass spectrometry using 8.6 -10 mg protein per analysis on a Sola (Finnigan-MAT) spectrometer.
Microwell Substrate Binding Assay-Binding of the rC domain to extracellular matrix proteins and substrates of gelatinase A was deter-mined using an enzyme-linked immunosorbent type assay (13). Proteins tested were soluble fibronectin, laminin, SPARC, tenascin, native and denatured types I, II, IV, V, and X collagens, the NC1 domain of type VII collagen, and reconstituted basement membrane Matrigel TM . Denatured collagens were prepared by heat denaturation at 56°C for 30 min. Ovalbumin was used as a negative control. To map C domain binding sites on fibronectin, nonoverlapping 40-and 120-kDa chymotrypsin fragments of fibronectin were used. Proteins (0.5 g) in 15 mM Na 2 CO 3 , 35 mM NaHCO 3 , and 0.02% (w/v) NaN 3 , pH 9.6, were coated on 96-well microtiter plates for 18 h at 4°C. Consistent and equal coating of test substrates in this assay has been previously confirmed (13). Wells were then blocked with 2.5% (w/v) ovalbumin, and serially diluted rC domain was added (1024 to 4 pmol/100 l (10.24 M to 40 nM) in 20 mM Tris-HCl, pH 7.4, for 1 h. Extensive washes with phosphate-buffered saline, 0.02% Tween 20 (v/v) followed, and the bound rC domain was then quantitated using either the ␣His 6 or the ␣72ex12 antibodies followed by incubation with goat anti-rabbit alkaline phosphatase-conjugated secondary antibody and p-nitrophenyl phosphate disodium as substrate. Assays were performed at least in duplicate per plate, and data were only compared for experiments on the same plate. All experiments were repeated six times, except for type II and X collagens, the type VII collagen NC1 domain, SPARC, and tenascin, which were performed in quadruplicate. Specificity was confirmed by comparing the binding of the rC domain with another recombinant domain from gelatinase A, the fibronectin type II-like triple repeat previously designated the collagen binding domain (13). The rCBD contains the same fusion tag and was expressed in the same E. coli strain used for rC domain expression.
Affinity Chromatography-Binding properties of the rC domain were also determined by affinity chromatography. Minicolumns containing either gelatin-Sepharose 4B, elastin/Sephadex G-10, heparin-Sepharose CL-6B, fibronectin coupled to gelatin-Sepharose 4B, or fibronectin coupled to Affi-Gel 10 (V t , 100, 200, 100, 70, and 50 l, respectively) were used in 50 mM Tris-HCl with or without 0.15 M NaCl, pH 7.4, chromatography buffer as appropriate (51). To ensure saturation of binding sites on the affinity matrix, a standardized quantity (50 g, 2 nmol) of the rC domain was loaded onto the columns, which would result in excess protein being recovered in the unbound and wash fractions. To confirm binding to affinity matrices, any rC domain recovered in the unbound material was reapplied to another column of the affinity matrix, and binding was compared. Binding of the rC domain was typically assessed by step elution with NaCl to 1.0 M followed by a step gradient of Me 2 SO (1-10%) in chromatography buffer. Chromatography fractions were analyzed by SDS-PAGE using 15% gels. Affinity chromatography experiments were performed at least six times for each matrix. FPLC was also performed using both heparin-Sepharose and Affi-Gel heparin 1.0-ml columns. After sample loading, a 0 -1.0 M NaCl gradient was developed over 10 ml at 1.0 ml/min. Chelation Experiments-To assess the role of the structural divalent cations on the binding of the rC domain to heparin and fibronectin, chelators (EDTA, 1,10-phenanthroline, EGTA, and BAPTA) were used to attempt elution from these affinity matrices. To confirm the elution by chelators, columns were then eluted with 1.0 M NaCl to recover any remaining bound rC domain. The apo rC domain was prepared before Affi-Gel heparin chromatography by chelation of structural Ca 2ϩ ions for 1 h by 50 or 100 mM EDTA followed by desalting on a Sephadex G-10 spun column.

Characterization of the rC Domain of Human Gelatinase
A-To investigate the properties of the hemopexin-like C domain of human gelatinase A, recombinant protein encoded by exons 9 -13 of gelatinase A (Gly 417 -Cys 631 ) was expressed in E. coli. The purified rC domain electrophoresed as a single band on 15% SDS-PAGE gels with an apparent molecular mass of ϳ26.5 kDa under reducing conditions (Fig. 1A). The absence of intermolecular disulfide linked multimeric forms of the recombinant protein was evident following electrophoresis under nonreducing conditions (Fig. 1A). The downshift in apparent molecular mass (⌬0.8 kDa) for nonreduced samples indicated the presence of an intact disulfide bond between Cys 440 and Cys 631 (20) within the protein. The precise mass of the rC domain was measured to be 25,925.0 Da by electrospray mass spectrometry, within 0.1 Da of the predicted mass of a NH 2terminal methionine-processed form of the recombinant pro-tein (25, 924.9 Da), confirming the fidelity of correct expression. Immunoreactivity with two affinity purified antipeptide antibodies, ␣72ex12 (Fig. 1B) and ␣His 6 (data not shown), further verified the identity of the purified protein. Importantly, essentially no disulfide cross-linked multimeric forms were observed on the immunoblots even when the ECL reaction was allowed to proceed well beyond the linear range.
Characterization of rC Domain Binding to Fibronectin, Laminin, SPARC, and Tenascin-Binding of the rC domain to basement membrane components was quantitated by the microwell substrate binding assay. Although the rC domain did not bind Matrigel TM and laminin ( Fig. 2A) or SPARC and tenascin (not shown), saturable binding to fibronectin in either NaCl-free 50 mM Tris buffer or phosphate-buffered saline was found, with an apparent K d of 1.1 ϫ 10 Ϫ6 M. A recombinant collagen binding domain (13) from gelatinase A with the identical fusion tag did not bind fibronectin (Fig. 2B), further confirming the specificity of the rC domain interaction. Using nonoverlapping chymotrypsin-generated fragments of fibronectin, binding sites for the rC domain were found on both the 40-kDa heparin binding and the 120-kDa cell binding fibronectin fragments with a similar but slightly stronger (apparent K d values, 4 ϫ 10 Ϫ7 and 6 ϫ 10 Ϫ7 M, respectively) affinity than for intact fibronectin (Fig. 3).
To further investigate the binding of the rC domain to fibronectin, the recombinant protein was applied to minicolumns of fibronectin coupled to gelatin-Sepharose columns in chromatography buffer containing 0.15 M NaCl. After adding the rC domain to saturation, as determined by the eventual detection of the recombinant protein in the unbound fraction (Fig. 4A, U), the bound rC domain was step eluted with 0.3-0.5 M NaCl in 50 mM Tris-HCl, pH 7.4. No further rC domain was recovered with 1.0 M NaCl or when Me 2 SO was used to elute the gelatin-bound fibronectin (not shown). That the rC domain did not bind to the gelatin-Sepharose was confirmed in control experiments (Fig.  4B). Identical elution profiles were obtained when the rC domain was loaded in chromatography buffer with NaCl omitted (data not shown). To ensure that the small amount of unbound rC domain did not represent a misfolded form of the protein, this material was reapplied to another fibronectin affinity column. Fibronectin binding was again demonstrated (data not shown), confirming that the unbound rC domain was the result of overloading after saturation of fibronectin binding sites. rC Domain Interaction with Type I, II, IV, V, VII, and X Collagens and Elastin-Although the C domains of collagenase-1 and stromelysin-1 bind native type I collagen, there was no binding of the gelatinase A rC domain to either native or denatured type I collagen in the microwell substrate binding assay compared with the negative control ovalbumin (Fig. 5A). The rC domain also did not bind gelatin-Sepharose (Fig. 4B), being quantitatively recovered in the unbound and wash fractions with no protein recovered in any elutes. In contrast, the rCBD from gelatinase A, used as a positive control for collagen interaction, showed avid binding to both forms of type I collagen (Fig. 5A).
To determine whether the gelatinase A C domain contributes to the binding specificity for other substrates of the enzyme, the rC domain was tested for interaction with types IV and V collagen and elastin. As for type I collagen, the rC domain did not bind these collagens in either their native or denatured forms (Fig. 5B). Native and denatured types II and X collagen and the NC1 domain of type VII collagen were also not bound by the rC domain (not shown). In confirmation of these data using the ␣His 6 antibody, identical results were also obtained with the ␣72ex12 antibody (not shown). rC domain binding to elastin was assessed using 6 mg of insoluble elastin mixed with 100 l of Sephadex G-10. All the loaded rC domain protein was recovered in the unbound and first wash fractions, and none was eluted with NaCl or Me 2 SO (data not shown). This indicated that the rC domain does not contribute to the elastin binding properties of the enzyme that are localized to the fibronectin type II-like CBD of gelatinase A (13) and shown to be essential for elastinolysis by gelatinase B (42).
The rC Domain Binds to Heparin-Heparin-Sepharose minicolumns and FPLC columns were used to investigate the binding properties of the C domain of gelatinase A to the heparan sulfate component of basement membranes and cell surface proteoglycans. The bound rC domain was step eluted off heparin-Sepharose minicolumns with 0.5 M NaCl (Fig. 6A) and at 0.58 M NaCl on a NaCl gradient developed on a 1.0-ml heparin-Sepharose FPLC column (data not shown). Taking into consid-

FIG. 3. Interaction of rC domain with fibronectin domains.
Serially diluted rC domain was added to microwell plates coated with 0.5 g each of the 40-and 120-kDa fibronectin domains and intact soluble fibronectin as indicated. The bound rC domain was measured by enzyme-linked immunosorbent assay reaction with affinity-purified ␣His 6 -antipeptide antibody. Curve-fitting and data points from the means of duplicate measurements from one enzyme-linked immunosorbent assay plate performed in triplicate are presented. eration the calculated lag time for protein elution off the 1.0-ml column, the results of the minicolumn and FPLC were in accordance. Binding to Affi-Gel heparin was consistently stronger, requiring 0.76 M NaCl for peak elution (Fig. 7A). Heparin specificity was confirmed by the absence of binding to the negatively charged CM-Sepharose and that the unbound overloaded protein from the minicolumns could bind heparin-Sepharose on subsequent chromatography (data not shown).
To compare the heparin binding properties of the rC domain with gelatinase A, active enzyme was obtained from confluent cell cultures treated for 18 h with 20 g/ml ConA (23) and then chromatographed over heparin-Sepharose minicolumns. Similar to the rC domain and confirming our previous studies (23,51), the activated gelatinase A was also eluted off the column by 0.5 M NaCl (Fig. 6B). Previously we have shown that the gelatinase A CBD has a low affinity heparin binding site (13), and our recent mutagenesis studies 3 have identified Lys 357 as an essential residue in this site. Therefore, to compare the relative importance of this and other possible heparin binding sites on gelatinase A, active enzyme was loaded onto heparin-Sepharose, and competition was attempted with Ͼ20-fold molar excess rC domain added before elution was continued. Since only a small amount of the bound enzyme could be competed off the column by the rC domain, with the bulk of the enzyme being recovered in the 0.5 M NaCl elute (Fig. 6C), this indicated that other heparin binding sites on gelatinase A remained associated with the heparin. Since the previously described "mini-gelatinase," representing a ConA-induced processed form of gelatinase A lacking the C domain (23, 52) (Fig. 6, B and C, ⌬C-Gs), was predominantly but not quantitatively recovered in the unbound and wash fractions, this reveals the importance of the C domain heparin binding site relative to the other heparin  (n ϭ 4). After washes (W) with chromatography buffer, bound gelatinase A was eluted from the columns with a step gradient of 0.1-1.0 M NaCl, as before. Samples were separated on 10% SDS-polyacrylamide gels copolymerized with 1 mg/ml gelatin. After zymography, gelatinase activity was detected as cleared bands after staining with Coomassie Brilliant Blue R-250. Std, latent gelatinase A standard; ⌬C-Gs, mini-gelatinase lacking the C domain produced by autolytic processing of the gelatinase A on ConA stimulation of cultures. In C, the rC domain was applied to a column of heparin-bound active gelatinase A (n ϭ 4) in an attempt to compete for binding sites. After extensive washes (W, 3 and 4), bound proteins were eluted by a step gradient of NaCl as shown.
binding sites on human gelatinase A.
Structural Requirements for Heparin Binding-The structural importance of the Cys 440 -Cys 631 disulfide bond for heparin binding was examined after treatment of the rC domain with 65 or 130 mM DTT for 60 min prior to heparin-Sepharose chromatography. All minicolumn solutions also included freshly prepared 65 mM DTT to ensure that reducing conditions were maintained throughout chromatography. As shown in Fig. 8A, there was no change in the heparin-Sepharose elution profile of the reduced rC domain compared with unreduced samples (see Fig. 6A). When more precisely analyzed by Affi-Gel heparin FPLC, the reduced rC domain (130 mM DTT; Fig.  7B) was found to possess essentially identical elution properties as the unreduced protein (Fig. 7A) at 22°C.
The recently published three-dimensional structures of the collagenase (18) and gelatinase A C domains (16,17) show a divalent ion, modeled as a Ca 2ϩ ion, in the central channel of the four-bladed ␤-propeller structure, together with either a Na ϩ -Cl Ϫ (16) or Ca 2ϩ -Cl Ϫ (17) ion pair. In addition, a potential Zn 2ϩ ion was also modeled in hemopexin module IV (16). In support of an important structural and/or functional role for the Ca 2ϩ ions and a possible Zn 2ϩ ion for binding of the rC domain to matrix components, the apo-rC domain was found to have lost binding potential for Affi-Gel heparin (Fig. 7C) and heparin-Sepharose (not shown) on FPLC. The loss of binding properties on removal of the structural Ca 2ϩ was also confirmed by the elution of the heparin-Sepharose-bound holo-rC domain by 50 mM EDTA on minicolumns (Fig. 8B). Elution was quantitative, with no further protein recovered by 100 mM EDTA or in the subsequent 0.5 M NaCl elution. Lower concentrations of EDTA (5, 10, and 20 mM) did not elute the bound rC domain (data not shown), suggesting a tight coordination of the divalent cations with the protein. Moreover, the rC domain treated with 15 mM EDTA prior to FPLC over Affi-Gel heparin retained heparin binding and elution properties identical to that of the holo protein (data not shown).
Identification of the Structural Divalent Cations-Since EDTA is a chelator of several divalent cations, including Ca 2ϩ and Zn 2ϩ , more specific chelators were used to identify the important structural ions in the rC domain. The well charac- terized Zn 2ϩ ion chelator 1,10-phenanthroline had no apparent effect on the binding of the rC domain to heparin, with all bound protein being recovered with the final 0.5 M NaCl elution (Fig. 9A). Thus, either no Zn 2ϩ ions are normally ligated to the C domain, or if present, then chelation by 1,10 phenanthroline does not alter the binding properties of the protein. Moreover, inductive coupled plasma mass spectrometry analysis of the rC domain failed to detect a biologically significant Zn 2ϩ ion content in the protein (8.5 ϫ 10 Ϫ6 mol of Zn 2ϩ /mol of rC domain).
Surprisingly, 100 mM EGTA was ineffective in eluting the rC domain from heparin-Sepharose (Fig. 9B). This was confirmed by Affi-Gel heparin FPLC analysis in which elution was achieved only at 180 -200 mM EGTA (not shown). Since EGTA is a Ca 2ϩ and Mg 2ϩ ion chelator, this result was paradoxical. However, several Ca 2ϩ binding proteins are known to directly bind EGTA (53) due to the high electronegativity of the chemical. Therefore, to confirm the identity of the central Ca 2ϩ ion, a newer and one of the most highly specific chelators of Ca 2ϩ ions, BAPTA, was used. Low concentrations of BAPTA were ineffective in eluting rC domain (1 and 5 mM), but at 20 mM BAPTA (Fig. 9C) the bound rC domain was quantitatively eluted from the heparin-Sepharose. No further protein was recovered with 0.5 M NaCl (cf. 1,10-phenanthroline and EGTA; Fig. 9, A and B), confirming the importance and strength of association of the Ca 2ϩ ion. The rC Domain also Binds Fibronectin in a Ca 2ϩ Ion-dependent Manner-To determine whether the structural Ca 2ϩ ion has a similar influence on the binding properties of the C domain for fibronectin, minicolumns of fibronectin coupled to gelatin-Sepharose were loaded with the rC domain. Both 50 mM EDTA (Fig. 10A) and 20 mM BAPTA (Fig. 10B) quantitatively dissociated the rC domain from the fibronectin, with no further rC domain being recovered with 1.0 M NaCl. Overall, these experiments exclude the possibility of a structurally important Zn 2ϩ ion in the rC domain, chemically identify the essential divalent cation in the rC domain as a Ca 2ϩ ion, indicate the strength of the Ca 2ϩ ion coordination with the protein, and reveal its important role in defining the binding properties of the rC domain to heparin and fibronectin. DISCUSSION Reported here are the fibronectin and heparin binding properties of the COOH-terminal hemopexin-like domain of human gelatinase A, the first such report for any of the MMPs. Fibronectin is a complex mutidomain matrix glycoprotein with multiple binding sites for extracellular matrix components and cell membrane proteins. Indeed, the rC domain binds two nonoverlapping chymotrypsin-generated fragments of fibronectin with similar K d values (10 Ϫ7 M range). The display of binding sites on fibronectin is conformation-dependent, with some cryptic sites being exposed on denaturation, surface binding, or cleavage (54), possibly explaining the slight difference in binding strength of the rC domain to the intact protein (apparent K d , 1.1 ϫ 10 Ϫ6 M) compared with the 40-and 120-kDa fibronectin fragments. Fibronectin and its alternately spliced variants also occur as matrix-bound and soluble forms in serum. Accord- FIG. 9. Heparin binding by rC domain shows a Ca 2؉ but not a Zn 2؉ ion requirement. To investigate the role of a potential structural Zn 2ϩ ion in the rC domain, the recombinant domain was bound to heparin-Sepharose, and after extensive washes (W) to ensure no carryover of overloaded unbound (U) protein, 1,10-phenanthroline (prepared from a stock solution of 1,10 phenanthroline dissolved in 70% (v/v) ethanol) was applied to the columns with an intervening wash (A, W). Finally, 0.5 M NaCl in chromatography buffer was applied to step elute any remaining bound rC domain (see Fig. 6A). B and C show the results using EGTA and BAPTA, respectively, for elution from heparin-Sepharose. Aliquots from each fraction, together with a sample of the rC domain before chromatography (B), were separated on 15% SDS-PAGE gels. M r , molecular mass marker proteins in Da as indicated; front, dye front. ingly, the exact binding stoichiometry and dynamics of the gelatinase A C domain association with fibronectin in tissues is likely a complex interaction. Thus, although these experiments directly revealed the presence of multiple potential gelatinase A binding sites on the fibronectin chains, these sites may not always be available for concurrent occupancy due to structural differences between the intact tertiary fold of the protein and the cleavage fragments or to steric clashes, which may prevent multiple binding.
Fibronectin cleavage by gelatinase A results in prominent degradation fragments (40) that include the amino-and carboxyl-terminal fibrin binding domains and the larger central cell binding domain (54). Other fibronectin-degrading MMPs include the stromelysins, matrilysin, and metalloelastase (2), but as with gelatinase A, their fibronectin-degradative activities have received little study. With the exception of matrilysin, a fibronectin binding site might also be located on the C domain of these MMPs. Fibronectin binding may be a requisite for efficient catalysis, but previous C domain deletion mutant studies of gelatinase A (27,55) or other MMPs (39, 55) did not investigate fibronectin binding or degradation. Fibronectin binding might also represent another means by which gelatinase A can bind cell surfaces. This may produce a reservoir of latent gelatinase A in proximity to the cell surface for eventual interaction with TIMP-2 and its receptor (56), poised for cis activation by MT-MMP-expressing cells (26,57).
Gelatinase A also degrades native type IV basement membrane collagen and is considered pivotal for basement membrane remodeling during embryogenesis 4 and in tumor cell invasion and metastasis (58). However, the rC domain did not bind basement membrane components either individually (type IV collagen, tenascin, laminin, and SPARC) or in combination as reconstituted basement membrane (Matrigel TM ) or to the NC1 domain of anchoring fibril type VII collagen. This is consistent with the findings of others that C domain deletion mutants of gelatinase A did not bind (55) but still degraded (27) native type IV collagen and laminin. The rC domain also did not bind native type I, V, or X collagens, which are degraded by gelatinase A, or other collagens such as type II, which are not. Thus, the C domain of gelatinase A differs from those of the collagenases and stromelysins, which bind native type I collagen (32,(35)(36)(37)55) and which, for the collagenases, are absolutely essential for collagenolytic activity (28 -31). This indicates that the molecular mechanism involved in the triple helicase activity of gelatinase A for native type I, IV, V, and X collagens is fundamentally different from that of the collagenases. Indeed, a C domain deletion mutant of gelatinase A showed only slight alteration in type IV collagen cleavages (27). Alternatively, the collagen binding properties of the CBD of gelatinase A may compensate for the absence of collagen tethering by the C domain to potentiate triple helicase activity. Nonetheless, this does not explain the mechanism of type IV collagenolytic activity, since the isolated CBD does not bind native type IV collagen (13).
The heparin binding properties of gelatinase A have previously been described (51,59) and shown to be important for enzyme activation (59). A gelatinase A heparin binding site has been previously located within the CBD (13). 3 The heparin binding site located in the C domain is specific, since the rC domain did not bind and was only slightly retarded by interaction with the negatively charged CM-Sepharose. Although no match to the canonical heparin binding consensus sequences (XBBXBX and XBBBXXBX; B, basic amino acids; X, undefined) (60) is found in the C domain, a site that matches the XBBXBX site in reverse (SKNKKT) and a sequence with high similarity (VKKKMDPG) are located in the lysine-and arginine-rich hemopexin-like module III. However, competition experiments using an antipeptide antibody raised against this sequence (␣72ex12) had no effect on the binding of the rC domain to heparin (data not shown). These and other positively charged clusters within hemopexin-like modules III and IV are now being studied by site directed mutagenesis to identify the binding sites for heparin and the negatively charged COOH-terminal peptide of TIMP-2. 5 Thus, heparan sulfate binding to the C domain may alter TIMP-2 interaction, which could in turn modulate cell surface activation of gelatinase A.
The Ca 2ϩ ion dependence of the C domain interaction with fibronectin and heparan sulfate proteoglycans points to a new potential therapeutic target for gelatinase A. Confirmation of the importance and identity of the structural Ca 2ϩ ions was shown in the apo form of the rC domain by disruption of the structural properties important for ligand binding. Chelation with EDTA and BAPTA, a new derivative of EGTA that is a highly specific Ca 2ϩ chelator (61) but which has greatly improved rate constants and pH insensitivity over EGTA (61,62), also eluted the bound holo-rC domain from heparin. The unexpected results obtained with EGTA may be due to direct association of EGTA with the protein (53) or the neutral pH used. At pH 7, EGTA occurs as a dianion, which results in a reduced ability to bind Ca 2ϩ ions by 2 to 3 orders of magnitude compared with pH 8.6 (61). Together, these effects minimize the effective chelating capacity of the free EGTA for the heptacoordinated Ca 2ϩ ion. Therefore, the chelation results obtained with BAPTA were important in clearly establishing the identity and importance of the Ca 2ϩ ions in the rC domain. The chelation experiments further revealed the tight coordination of the Ca 2ϩ ions in the protein, supporting the conclusion that the Ca 2ϩ ions play an essential stabilizing role in the domain, possibly acting like a hub to centrally pin the four ␤-blades together. Last, since inductive coupled plasma mass spectrometry analysis did not measure any Zn 2ϩ ions in the rC domain and 1,10-phenanthroline did not alter the binding properties of the rC domain, these data failed to confirm the presence of a structural Zn 2ϩ ion in the C domain of gelatinase A. Together, these data indicate that the Zn 2ϩ ion modeled in a potential binding site on module IV of the C domain was likely an artifact of crystallization in 150 mM Zn 2ϩ acetate (16).
The central Ca 2ϩ ion was also more effective than the disulfide bond in maintaining the structural integrity of the rC domain at 22°C. Of note, these chromatography experiments were performed entirely under reduced conditions and not with reduced and alkylated protein. The introduction of blocking groups or a charged moiety during alkylation after reduction and protein denaturation by 6 M guanadinium or by heat can alter the structural properties of the protein distinct from those due to disulfide bond reduction alone. Indeed, reduced and alkylated rCBD from gelatinase A does not bind gelatin (13,63), whereas rCBD protein chromatographed and analyzed under continuous reducing but nondenaturing conditions retains gelatin interaction equal to nonreduced rCBD (13). Thus, these experiments indicate that the integrity of the four-bladed ␤-propeller structure of the rC domain was not markedly perturbed at 22°C by reduction of the disulfide bond but was altered sufficiently by chelation and loss of the structural Ca 2ϩ ions to disrupt the heparin and fibronectin binding sites.
The widespread and near constitutive expression of gelatin-ase A indicates that the most important level of regulation of gelatinase A activity may be cellular activation and inhibition by TIMPs rather than by regulated transcription (1,23,65). In this regard, we have previously proposed that latent gelatinase A may bind native type I collagen through the CBD (13) and so may remain localized in tissues, poised for proteolysis of gelatin on collagenase cleavage of the native collagen and trans activation by other MT-MMP-expressing cells. A similar reservoir of enzyme may also be important for elastinolysis, since binding to elastin also occurs through the CBD of gelatinases (13,42). 2 The binding of rC domain to fibronectin and potentially to heparan sulfate proteoglycans reveals additional extracellular or pericellular matrix components that may serve as anchors to sequester gelatinase A in tissues and to the cell surface. This may render the enzyme readily accessible to MT-MMPs for activation and to substrates, such as fibronectin and proteoglycan core proteins, on activation for cleavage. Together with the potential for the CBD to bind pericellular collagen, 3 this also provides a novel additional mechanism for cell surface interaction that may complement those involving ␣ v ␤ 3 -integrin, MT-MMPs, and TIMP-2 receptors.