Extracellular Matrix Binding Properties of ...

Vol. 270, Na. 19, Issue af May 12, pp. 11555—11566, 1995 Printed in U.S.A.

ThE JOURNAL OF BIOLOGICAL CHEMISTRY © 1995 by The American Society for Biechemistry and Molecular Bialogy, Inc.

Extracellular Matrix Binding Properties of Recombinant Fibronectin Type 11-like Modules of Human 72-kDa Gelatinase/Type IV Collagenase HIGH AFFINITY BINDING TO NATIVE

TYPE

I COLLAGEN BUT NOT NATIVE

TYPE

IV COLLAGEN*

(Received for publication, December 12, 1994, and in revised form, February 2, 1995)

Bjorn Steffensent, U. Margaretha Wallon, and Christopher M. Overall* From the Faculty of Dentistry, University of British Columbia, Vancouver, British Columbia V6T 1Z3, Canada

72-kI)a gelatinase/type lv collagenase is an important matrix metalloproteinase in the degradation of basement membranes and denatured collagens (gelatin). These proteolytic processes are required for pathologic tissue destruction and physiologic tissue remodeling. To investigate the molecular determinants of substrate specificity of this enzyme, a 21-kDa domain of 72-kDa gelatinase, consisting of three tandem fibronectin type Il-like modtiles, was expressed in Escherichia coil. Similar to fulllength 72-kDa gelatinase and the type II modules in fibronectin, the recombinant (r) fibronectin-like domain of this proteinase bound denatured type I collagen with an apparent Kd in the micromolar range. This domain, designated the collagen-binding domain (rCBD123), possesses at least two collagen-binding sites that can each be simultaneously occupied. rCBD123 also avidly bound elastin and denatured types LV and V collagens, but neither native types lv and V collagens nor fibronectin, all of which are substrates of the enzyme. Although 72-kDa gelatinase is involved in basement membrane degradation, rCBD123 also did not bind reconstituted basement membrane, laminin, or SPARC. Native type I collagen, which is not degraded by 72-kDa gelatinase, competed with gelatin for a shared binding site on rCBD123. rCBD123 also displaced full-length 72-kDa gelatinase bound to native type I collagen, further demonstrating that the collagen binding properties of the recombinant domain closely mimicked those of the full-length enzyme. Since rCBD123 showed reduced binding to pepsin-cleaved type I collagen, either or both of the collagen telopeptide ends contain recognition sites for the 72-k.Da gelatinase fibronectin-like domain. This was confirmed by the avid binding of rCBD 123 to the al(I) collagen cyanogen bromide fragment CB2 from the NH 2-terminal telopeptide. rCBD123 also bound al(I)-CB7, which encompasses the fibronectinbinding site, and to al(I)-CB8, a fragment not bound by fibronectin Thus, type I collagen contains multiple binding sites for rCBD123 which are partially masked by the triple helical conformation of native collagen and fully exposed upon unfolding of the triple helix. The potential of the fibronectin-like collagen binding domain of 72-kDa gelatinase to bind extracellular matrix proteins may fa* This work was supported by a grant from the Canadian Medical Research Council. The costs of publication ofthis 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. t Recipient of a Medical Research Council of Canada Studentship. § Recipient ofa Medical Research Council ofCanada Dental Clinician Scientist Award. To whom correspondence should be addressed: Faculty of Dentistry, 2199 Wesbrook Mall, Vancouver, British Columbia V6T 1Z3 Canada. Tel.: 604-822-2958; Fax: 604-822-3562; E-mail: [email protected].

cilitate enzyme localization in connective tissue matrices.

A central characteristic of metastatic tumor cells is their ability to degrade and penetrate basement membranes. Considerable evidence has linked elevated matrix metalloproteinase (MMP)’ expression by many tumor cells with these processes (1). Type IV collagen is the major structural component of basement membranes (2, 3). Therefore, the type IV collagenolytic activity of either of two MMPs, the 72- and 92-kiDa gelatinase/type IV collagenases (MMP 2/gelatinase A and MMP 9/gelatinase B, respectively), is viewed as a critical component of the metastatic process. The 72- and 92-kDa gelatinases are also important in other processes such as embryogenesis and tissue remodeling (reviewed in Refs. 4—6), osteoclastic activity (7), enamel formation (8), and lymphocyte cell migration (9). These enzymes also assist in completing the collagenolytic cascade by degrading the three denatured a-chains of cleaved collagen. Accordingly, discerning the molecular determinants of substrate specificity of the 72-kDa gelatinase is important in understanding the role of this enzyme in physiological and pathological processes. MMPs share a basic primary and tertiary structurehistidine (5, 10— 2~-binding 12). MMPs triad and a Ca2~-bindingmotif have a highly conserved in the catalytic Zn domain. A free cysteine in a highly conserved sequence in the prodomain coordinates with the catalytic zinc (II) ion and is responsible for enzyme latency (13, 14). A hemopexin/vitronectin-like carboxyl domain binds the specific tissue inhibitors of MMPs (TIMPs) (15—18) and, in the 72-kDa gelatinase, binds cell membranes (19) on ConA-activated fibroblasts (19—21). The hemopexinl vitronectin-like carboxyl domains of collagenase and stromelysin also bind native type I collagen (15, 22, 23). Removal of this domain from collagenase ablates collagenolysis but not catalytic competence, that is, the truncated collagenase lacking the carboxyl domain still degrades synthetic peptide substrates and casein, but not native type I collagen (22, 23). In contrast, the hemopexin-like domain of 72-kDa gelatinase does not bind collagen (24). In addition to the main structural elements characteristic of the MMPs, both the 72- and 92-kDa gelatinases contain three tandem copies of a 58-amino acid residue fibronectin type IIThe abbreviations used are: MMP, matrix metalloproteinase; BSA, bovine serum albumin; CB, cyanogen bromide; Dfl, dithioerytbreitol; gelatin, denatured collagen (denatured type I collagen consists of two nl(I) chains and one a2(I) chain); ME 2SO, dimethyl sulfoxide; PAGE, polyacrylamide gel electrophoresis; rCBD123, recombinant collagen-binding domain consisting of fibronectin type IT-like modules 1, 2 and 3; SDS, sodium dodecyl sulfate; SPARC, secreted protein which is acidic and rich in cysteine; TIMP-1, tissue inhibitor of matrix metalloproteinases-1.

11555

11556

The Fibronectin-ljke Domain of 72-kDa Gelatinase (MMP 2)

like module positioned immediately NH

2-terminal to the zincbinding site (25, 26). Fibronectin, a modular extracellular matrix glycoprotein, is composed of repeating homologous domains (Type I, IT, and TIll that bind a number of extracellular matrix proteins, fibrin, and cells (27). Although it is controversial whether fibronectin type IT modules alone bind denatured type I collagen (27—30), Banyai et a/. (31) reported that a recombinant type II module from fibronectin and a type IT-like module in bovine seminal fluid protein PDC-109 binds gelatin. Subsequent work demonstrated that the type TI-like modules in both the 72- (32, 33) and the 92-kDa (34) gelatinases also bind denatured type T collagen. Binding specialization of the different fibronectin type TI-like modules in the 72-kDa gelatinase may have occurred to generate exosites specific for the other collagens and extracellular matrix molecules degraded by the enzyme including native types TV, V, VIT, and X collagens, elastin, and fibronectin (25, 35). To further understand the function of the structural elements of the 72-kDa gelatinase, we have characterized the binding properties of the fibronectin-like domain of human 72-kDa gelatinase to a number of the enzymes substrates, reconstituted basement membrane, TTMP-l, and other extracellular matrix proteins. Reported here are experiments which establish that, in addition to binding denatured type I collagen, a recombinant fibronectin-like domain from human 72-kDa gelatinase, encompassing all three type IT-like modules, binds with high affinity to denatured types IV and V coil agens and elastin. Although human 72-kDa gelatinase cleaves native type IV collagen but not native type I collagen, surprisingly, the fibronectin-like domain avidly binds native type I collagen but neither native type IV collagen nor other basement membrane components. Thus, in addition to fulfilling the criteria as an exosite for a number of substrates, the fibronectin-like domain of 72-kDa gelatinase may have an ancillary role as an extracellular matrix localization domain by virtue, in particular, of its native type I collagen binding properties. EXPERIMENTAL PROCEDURES

Extrocellular Matrix Proteins, Antibodies, and Chromotogrophy Media—Acid-soluble native type I collagen wns prepared from rat tail tendons as described by Piez (36) by extraction with 0.5 is acetic acid and differential precipitation with 1.7 11 NaCl. Pepsin-treated type I collagen was prepared by digestion of the acid-soluble type I collagen with pepsin (Sigma) at pH 2.0, 4 CC for 20 h, then precipitated with 1.7 M NaC1, redissolved in 0.15 ss acetic acid, and lyophilized. Gelatin was 4CiGlycine-labeled type I collagen, prepared from acid-soluble type I collagen (non-pepsin treated) by heat denaturation with a specificat activity 56 °Cfor of30 3.5mm. )< io~ I ~ disintegratioas/minlmg, was prepared by metabolic labeling and purified from conditioned cell medium by pepsin digestion and NaCI precipitation as described previously (37). To confirm the native collagen content of the metabolically labeled preparation, the labeled type I collagen was incubated with 0.1 or 0.01 p.g/ml trypsin (type XII bovine pancreas, Sigsna) (enzyme to substrate ratio —1:2 and 1:20) for 19 h at 20 CC. Intact protein was then precipitated in 10% lw/v) trichloroacetic acid, 1% (w/v) tannic acid for 2 h at 0 CC and the pellets collected by centrifugation at 10,000 x g for 20 mm at 0 T (37). The trypsin-digested denatured type I collagen content ‘was determined by scintillation counting of the trichloroacetic acid/tannic acid-soluble protein fraction and calculated to constitute < 14) with the appropriate chromatography buffer were performed after sample loading and between different eluants. Fractions were collected and analyzed by SDSPAGE at a constant dilution relative to the loaded sample volume to facilitate assessment of binding as described below. Columns eluted with ME2SO were not reused. The specificity ofrCBD123 to collagen interaction was also assessed by competition assays. Typically, rCBD123 was incubated for 90 mm at 20 ‘C with the competing ligand at various mole ratios between 1:0 to 1:9 before loading the reaction mixture onto either gelatin-Sepharose or native type I collagen affinity columns and eluting as described. The denatured and native type I collagen binding properties of 72kDa gelatinase were assessed by affinity chromatography of the enzyme in 1.5-mi aliquots of conditioned culture medium from rat osteoblastic cells (46). Chromatography and elution conditions were as described above for rCBD123. To ascribe the type I collagen binding properties of the parental 72-kDa gelatinase to a specific domain, the enzyme was bound to native type I collagen columns and, after 1 is NaC1 washes, the binding then competed with rCBD123 (40 pg 2 nmol) in chromatography buffer. Eluates were assayed for 72-kDa gelatinase by enzymography as described below. Reduction and Carboxymethylation of rCBDI23—To reduce intramolecular disulfide bonds in rCBD123, dithioerythreitol (DTT) was added to rCBD123 to a final concentration of 65 mis (>100-fold mis excess/ disulfide bond) for 30 mm

at 20 ‘C. This concentration of DTT was

maintained in all buffers in column assays. rCBDT23 was also reduced and carboxymethylated according to Creighton (1990) (47) and Hollecker (1990) (48), modified as follows. rCBD123 was first equilibrated

in denaturation buffer (8.0 is urea, 0.5 is Tris-HC1, 2 mis EDTA, pH 8.1) by gel filtration on a 10 DG column (1’, 10 ml) (Bio-Rad) and then reduced by addition of 100-fold molar excess of DTT (65 mis) over the estimated disulfide bond content and incubated at 50 ‘C for 1 h. After cooling to 20 ‘C, the alkylating agent, iodoacetic acid, was added to a 2-fold molar excess over DTT (130 mis) and reacted at20 ‘C for 30 mm. The reduced and carboxymethylated protein was then equilibrated in 50 mis Tris, pH 7.4, by chromatography over a 10 DG column and assessed by SDS-PAGE. Column assays with reduced and carboxymethylated rCBD 123 were performed in the absence of reductant under buffer conditions identical to those described for non-reduced rCBD123. SDS-Polyacrylnmide Gel Electrophoresis and Eaxymography—Proteins were separated by SDS-polyacrylamide gel electrophoresis according to Laemmli (1970) (49). Protein samples were analyzed without reduction or with the addition of65 mis DTT and heating at 95 ‘C for 5 mm. Gels were stained with Coomassie Brilliant Blue R-250 at 42 ‘C, and protein bands were quantitated by laser densitometry at 633 nm (LKB Ultrascan XL). For enzymography, non-reduced protein samples were electrophoresed on 10% (w/v) polyacrylamide gels containing 100 gg/ml heat-

denatured acid-soluble type I collagen. Gels were processed as described previously (8). Briefly, after electrophoresis gels were equilibrated in 5% (v/v) Triton X-100, incubated in assay buffer (50 mis Tris, 200 mis NaCl, 5 mis CaC12) for 2—4 h at 37 ‘C, and the cleared bands, identifying the position of72-kDa gelatinase, revealed by counterstaining ofthe gelatin in the gels by Coomassie Brilliant Blue R-250. Reduced molecular mass markers used were rabbit muscle phosphorylase b (97 kDa), bovine serum albumin (BSA) (67 kDa), chicken egg ovalbumin (43 kDa), bovine carbonic anhydrase (29 kDa), horse heart myoglobin (18.8 kDa), chicken egg-white lysozyme (14.4 kDa), and

bovine insulin (6.2 kDa) (Sigma). Western Blot Analysis—Proteins were transferred to Immobilon-P polyvinylidene difluoride membranes (Millipore) after separation by SDS-PAGE. Transferred rCBD123 was then reacted with a polyclonal anti-72-kDa gelatinase antibody diluted 1:1,000 in TBS/Tween with 1% (w/v) BSA for 1 h, washed, and conjugates detected using enhanced chemiluminescence (ECL) reagents and Hyperfilm (Amersham Corp.). Microwell Substrate Binding Assay—To screen for potential interaction with rCBD123, a number ofknown substrates of72-kDa gelatinase as well as other extracellular matrix proteins were coated as films in

11557

96-microwell plates. Proteins included native type I collagen, pepsintreated native type I collagen (telopeptide-free), and heat-denatured type I collagen (gelatin); collagen alfi) cyanogen bromide fragments 2, 7, and 8; native and heat-denatured types IV and V collagens; elastin, Matrigel®, laminin, fibronectin, SPARC, and TIMP-l. Myoglobin and BSA served as negative control proteins for the assays. Microtiter plates were coated overnight at 4 ‘C with 10 pmol protein/well (typically 1—5 gg) in coating buffer (15 mis Na2CO5, 35 mis NaHCO3, 0.02% 4C]glycine-labeled (w/v) NaN3, pH 9.6). Consistent and equal binding ofprotein coatednato tive and denatured type I by collagens with the solutions as apmicrowells was confirmed incorporation of coating [‘ propriate and scintillation counting of the unbound supernatants. Matrigel® and native type I coBagen were also prepared as threedimensional gels as follows. Gels of reconstituted basement membrane were produced with Matrigel® diluted 1:3 in phosphate-buffered saline (100 p1/well) at 4 ‘C and incubated for 30 mm at 37 ‘C to assure solidification. Type I coBagen fibrillar gels were prepared by dissolving 10 pmol ofacid-soluble native collagen in 100 p1 of 50 mis Tris, 200 mis NaC1, pH 7.0, followed by incubation for 30 mm at 37 ‘C. Plates coated with protein films or gels were rinsed and thenblocked with 2.5% (w/v) BSA in phosphate-buffered saline for 1 h at 20 ‘C. After further extensive rinses with phosphate-buffered saline, serially diluted rCBD123 in 50 mis Tris, pH 7.4 (1 nmol to 0.125 pmollwell = 10 gis - 1.25 nis) was added for 1 h at 20 ‘C. The plates were rinsed thoroughly and bound rCBD 123 detected with a 1:1,000 dilution of polyclonal antibody raised against human 72-kDa gelatinase followed by reaction with a 1:5,000 dilution of alkaline phosphatase-conjugated goat anti-rabbit antibody (Bio-Rad). All antibody reactions and washes were performed in buffers containing 0.05% (v/v) Tween 20. For quantitation, p-nitrophenyl phosphate disodium (Sigma) was added as substrate and, to ensure linearity of the assay, the color intensity was determined at various times by measuring the absorbance at 405 nm in an automated enzyme-linked immunosorbent assay plate reader (Bio-Rad). Negative controls consisted of reaction mixture minus rCBD123, primary antibody, or secondary antibody.

[24CJGlycine-labeled Collagen Binding Assay—To determine the potential for simultaneous binding of two or more molecules of type I

collagen by rCBD123, a sandwich type assay was employed. rCBDT23: [14Cjglycine-labeled type I collagen was quantitated in those rCBD123 complexes that could also bind unlabeled collagen films on microwell plates. Specifically, rCBD123 was serially diluted in 50mM Tris, pH 7.4, from 10,000 to 0.6 nis and incubated with 4,000 disintegrations/min native or heat-denatured [54C]glycine-labeled type I collagen (0.03 and 0.1 pmol, respectively) for 1 hat 20 ‘C. The reaction products were then transferred to microwell plates coated with 10 pmol/well of unlabeled native or denatured type I collagen as appropriate. That is, to avoid ligand displacement by competing substrates rCBD123 that had been incubated with [54C]glycine-native type I collagen was transferred to plates coated with native type I collagen, and rCBD 123 that had been incubated with [‘4C]glycine-labeled denatured type I coBagen was transferred to denatured type I collagen-coated plates. After 1 h of incubation at 20 ‘C, unbound material was removed by three rinses with 50 mis Tris, pH 7.4. Bound rCBD123:[’4C]glycine-labeled type I collagen complexes were dissociated by addition of 10% (v/v) Me 2SO in 50 mis Tris, pH 7.4, for 30 mm and then transferred to Scintillant and quantitated by scintillation counting. Binding reactions were performed in duplicate for eight serial dilutions with appropriate controls for both native and denatured type I collagens. RESULTS Characterization of Recombinant 72-kDa Gelatinase CBDI23 Protein—To determine the contribution of the fibronectin-like domain in 72-kDa gelatinase to the substrate binding and functional properties of the enzyme, the three fibronectin type TI-like modules of human 72-kDa gelatinase were expressed in F. coli as a single recombinant fusion protein, denoted as rCBD 123. Approximately 18% of total F. coli protein was rCBD 123 which was localized predominantly to inclusion bodies after 18 h of culture. Typically, the yield of purified gelatin binding rCBD123 from 1,000 ml of F. coli culture was approximately lb mg. Non-reduced rCBD 123 electrophoresed with an apparent molecular mass of 21.1 kDa whereas the reduced protein migrated with an apparent molecular mass of 22.1 kDa (Fig. IA), identical to that predicted from the sequence. Reduced and

11558

The Fibroneetin-like Domain of 72-kDa Gelatinase (MMP 2) \::e

A

::

M 97. 67 -‘ 43 29 --

/

/ /

+ +

+ + + +

~

/

/ /

+ + + + + +

CM DTT

-

—

ee 18.4

- —

14.4

-

fr

- -

—

rC BD 123

—~

rCBD123

B 97_ 67—s 43—s 29 —*

.

a

-

18.4.-. 14.4-.

fr-s Fen. 1. SUS-PAGE and Western blot analysis of non-reduced, reduced, and carboxymethylated rCBD123. Samples of rCBD123 (1 pg/lane) were analyzed by SDS-PAGE on 15% (w/v) cross-linked polyacrylamide minislab gels under non-reducing ) —DTT) or reducing ÷DTT)conditions and stained with Coomassie Brllmant Blue R-250 (panel A) or transferred by Western blotting to PVDF membranes and reacted with an anti-72-kDa gelatinase polyclonal antibody as described under “Experimental Procedures” (panel B). The antibody reaction to rCBD123 was consistently greater for non-reduced samples compared to reduced or reduced and carboxymethylated (CM+ ( samples. rCBD123 migrated with a relative molecular mass of 21.1 kDa non-reduced, 22.1 kDa reduced, and 24.1 kDa reduced and carboxymethylated. Panel A, reduced and carboxymethylated rCBD123 was analyzed for gelatin binding by affinity chromatography over minicolumns of gelatmn-Sepharose as described under “Experimental Procedures.” Reduced and carboxymethylated rCBD123 before chromatography (B) was fully recovered in the unbound (U) and wash 1 (Wi) and wash 2 (W2) fractions and did not bind to gelatin as shown by the absence of protein in the 1.0 a NaC1 (N) and 10% (v/v) ME,SO ID) eluates. M~,molecular weight markers x 10 ‘ as indicated; fr’ dye front. carboxymethylated rCBD 123 electrophoresed with an apparent molecular mass of 24.1 kDa. The 1.0 kDa decrease of apparent molecular mass in non-reduced compared with reduced samples indicated that rCBD123 was folded and contained intact disulfide bonds (50). To further substantiate that the purified protein was indeed the recombinant product, an antibody to human 72-kDa gelatinase was found to react with the recombinant protein (Fig. iF). Moreover, antibodies raised against the fibronectin T-like domain, to a synthetic peptide corresponding to the poly-His tract of the fusion protein, and to a peptide in the third fibronectin type Il-like module, all reacted strongly with the purified protein (not shown). Recombinant protein in the gelatin-Sepharose ME2SO elute was predominantly monomeric with low amounts of dimers (100-fold mole excess of DTT was maintained in all chromatography buffers. This treatment had no effect on the binding of rCBD 123 to denatured type I collagen. Further, the reduced rCBD123 showed an elution profile from gelatmn-Sepharose identical to that of non-reduced rCBlJl23 (not shown, n = 3). Therefore, it appears that rCBD123 can retain a biologically

11559

The Fibronectin-like Domain of 72-kDa Gelatinase (MMP 2) A

%DMSO MrB U N 0 1 2 3 4 5 6 7 8

1.5~ O

97—) — 67—s — 43—) P

E C

LO 29

Type i fibrila • Type Idenaiured O Type I native A Myoglabin

—.4 ~

—

0

rCBvl23

——----

1.0•

0

a)

lB.4 —s S 14.4-se 6.2_s

0 C

-o 0.5~ 0 ci)

-o

cc

B

1:0.3

1:)

1:3

1:9

BU12

BU12

0.0~ ~a—a---— A 01 ~ ~

WI

M, B U 1 2 97.-) 67_s 43—s

— — —

29 --s

—

BU12

A

456

Concentration rCBD123 (RM) ~—

a-chains

Fin. 3. rCBD123 affinity for type I collagen in different conformations. Type I collagen as a three-dimensional gel (Type 1 fibrels), an

—

—

~

,—rCBDI23

18.4-. — 14.4—s —

Fin. 2. Denatured type I collagen affinity chromatography of rCBD123. Recombinant CBD123 (40 pg, 1.9 nmol in 100 p1) was loaded (B) onto mini-columns of gelatin-Sepharose and chromatographed as described under “Experimental Procedures.” After extensive washes with chromatography buffer, 1.0 is NaC1 was applied to the column in chromatography buffer followed by a step gradient ofME 2SO from 1 to 8% (v/v) in chromatography buffer. Eluates were analyzed on 15% SDS-PAGE gels under reducing conditions (panel A). rCBD 123 bound avidly to gelatin-Sepharose as shown by the absence of rCBD123 in unbound fractions (U), washes (not shown), or the 1.0 is NaC1 (N) elute. Peak elation ofrCBD123 was at 3% (v/v( MR2SO. The specificity ofthe interaction between rCBD 123 and gelatin was confirmed by a competition assay in which rCBD 123 was incubated with increasing mole ratios of gelatin (1:0.3 to 1:9) for 90 mm at 20CC prior to loading onto gelatin-Sepharose columns. The eluates were analyzed under reducing conditions on 15% SDS-PAGE gels. Panel B shows that competition with increasing mole amounts of gelatin caused both a graduated increase in the amount of rCBD123 in the unbound (U) and the two wash fractions (W; 1, 2) and a corresponding decrease in the column-bound rCBD123 (not shown). The positions of collagen n-chains in the unbound and wash fractions from the competition reaction, rCBD 123, and the kDa of marker protein standards (34,) are indicated. functional folded structure at 22 ‘C in the absence of disulfide cross-links. However, when rCBD123 was reduced and carboxymethylated, gelatin binding was lost (Fig. 1) indicating that either structural perturbation or alterations in charge introduced by the carboxymethylation of cysteines were responsible for the loss of gelatin binding properties. rCBDJ23 Binds Native Type I Collagen—Human 72-kDa gelatinase does not cleave nor has it been reported to bind native type I collagen. However, since fibronectin binds native type I collagen near the tissue collagenase cleavage site (Si— 53), we assessed rCBD 123 for binding to native type I collagen. Using microwell substrate binding assays with native type I collagen coated as a two-dimensional film, rCBD123 was found to bind to native type I collagen in a saturable manner (Fig. 3). In comparison, binding was somewhat stronger to denatured type I collagen with an apparent Kd in the gis range. In these experiments, an equal number of moles of molecules of denatured collagen a-chains and native type I collagen were coated per well. Since there are three a-chains/triple helical molecule of native type I collagen, then each well of native type I collagen contained three times the number of moles of a-chains as the denatured collagen-coated wells. Thus, together with the slightly lower binding of rCBD123 to native type I collagen, this showed that there were at least three times fewer binding sites/mole of a-chains when in the native triple helical confor-

absorbed film (Type I native), and as heat denatured al-chains (Type I denatured) was coated on the bases of weDs in 96-microwell plates as described under “Experimental Procedures.” None of these collagens had been pepsin-treated. Myoglobin was coated as a control protein for nonspecific binding. Serially diluted rCBD 123 was added to the wells in a volume of 100 p1 and the amounts of bound rCBD123 quantitated after reaction with anti-72-kDa gelatinase and alkaline phosphatase conjugated secondary antibody by reading the absorbance of the reaction mixture at 405 nm as described under “Experimental Procedures.” Data are plotted as the mean values oftwo experiments. mation than in the denatured conformation. This indicates that the triple helical structure masks binding sites on the constituent a-chains. Even though identical amounts of native type I collagen were used to form fibrils (10 pmol), a substantially higher amount of rCBD123 bound to the fibrillar gel form than to either native or denatured type I collagen coated as films (Fig. 3). These results may reflect improved access to binding sites on all faces of the type I collagen which are otherwise occluded when the collagen is adhered to plastic surfaces. To confirm that rCBD123 bound native type I collagen in solution, affinity chromatography on mini-columns coupled with native type I collagen was performed (Fig. 4A). No rCBD 123 was detected in the wash fractions, only a trace was found in the 1.0 is NaC1 elute, and essentially all the rCBD123 was recovered using ME2 SO in the chromatography buffer with a peak elution at —2% (v/v) ME2SO. Notably, this concentration of ME2SO was reproducibly lower than that required to elute rCBD123 from denatured type I collagen (—3%). In contrast to these studies and those shown in Fig. 3 where the native type I collagen contained intact telopeptides, rCBD123 did not bind pepsin-cleaved native type I collagen coupled to Affi-Gel as evidenced by the quantitative recovery of rCBD 123 in the unbound and wash fractions after chromatography (Fig. 4B). Since pepsin digestion removes the collagen amino- and carboxyl-terminal telopeptides, this shows that the telopeptide regions contribute significantly to the rCBD 123 binding of native type I collagen. Repeated column assays consistently failed to detect rCBD123 binding to pepsin-treated native type I collagen. However, the highly sensitive microwell substrate binding assay showed some, although weaker, binding of rCBD123 to pepsin-treated type I collagen (not shown). rCBD123 Binding to Native and Denatured Type I Collagen Mimics the Binding of Full-length 72-hDa Gelatinase—To verify that the binding interaction observed between the rCBD 123 domain and type I collagen corresponds to the properties of the full-length enzyme, we compared the binding of 72-kDa gelatinase to denatured and native type I collagen. 72-kDa gelatinase in conditioned cell culture medium was chromatographed over gelatin-Sepharose. Enzyme in the column fractions was revealed by enzymography. Specific binding of 72-kDa gelatin-

11560

A

The Fibronectin-like Domain of 72-kDa Gelatinase (MMP 2) W B U 123

NaG! 1 201

%DMSO 23456

A

%DMSO N 0 1 2 3 4 5 10

B

%DMSO N 0 1 2 3 4 5 10

43—s 29—s S.

—--—.

(—rCBD123

18~4—s 14.4—s

6.2 —s 66 62 59

B 43—s 29—s

‘5

0