INTRODUCTION

The matrix metalloproteinases mediate remodeling of extracellular matrix in healthy and diseased tissue (1, 2). Sequence analysis of the MMPs,¹ also referred to as the metzincin proteinase family (3), reveals a characteristic domain structure. The smallest MMP, matrilysin, consists of 19-kDa catalytic domain with an active site zinc ion coordinated within a conserved HEXXHXXGXXH motif. Proteolytic latency is maintained by an N-terminal propeptide of approximately 10 kDa in which a conserved cysteine is coordinated to the catalytic zinc (4). The larger members of the family, including interstitial collagenases, stromelysins, and gelatinases, possess additional functional and structural domains. In particular, collagenases have a C-terminal domain of about 30 kDa that exhibits sequence homology with hemopexin and vitronectin (5). This so-called pexin domain is essential for the expression of collagenolytic activity; without it, the enzyme retains general peptidase activity but does not hydrolyze native collagen (6). Matrilysin and mutant forms of collagenase and stromelysin possessing only a catalytic domain contain a second zinc ion (7-9). This structural zinc is absent from the full-length forms of stromelysin-1 and gelatinase A (10).

Calcium ions are also an essential components of the MMPs (11, 12). Crystallographic analysis of the catalytic domain reveals three bound Ca²⁺ in fibroblast collagenase (13, 14) and two in the neutrophil enzyme (15). Crystallography of full-length porcine MMP-1 also shows three Ca²⁺ in the catalytic portion and two Ca²⁺ in the pexin domain (16). Ca²⁺ undoubtedly binds to similar sites in the pexin domain of human collagenase because the binding sites are conserved. Previous studies suggest that Ca²⁺ stabilizes an active conformation of the MMP that is more stable to denaturants (17) and less susceptible to proteolysis (18). However, the precise stereochemical basis for Ca²⁺ stabilization is unknown because a three-dimensional model of a Ca²⁺-free MMP is not available for comparison.

To provide additional insight into the relationship between MMP structure and calcium binding, we correlated the effect of Ca²⁺ on catalytic activity and protein structure using intrinsic protein fluorescence and circular dichroism as structural indicators. Ca²⁺ concentration was manipulated either by the chelating agent EGTA or by dilution. EGTA was chosen because its affinity for Ca²⁺ is comparable with that of EDTA,² although its affinity for Zn²⁺ is much lower than that of EDTA (19). The following data document that Ca²⁺ binding to collagenase is cooperative and associated with changes in tertiary but not secondary structure. Furthermore, these changes were independent of the method used to alter calcium concentration.

EXPERIMENTAL PROCEDURES

A recombinant human fibroblast collagenase expression vector was generously supplied by Dr. G. McGeehan, Glaxo-Wellcome (Research Triangle Park, NC). The recombinant protein, which lacks the propeptide, was expressed in Escherichia coli and purified as described (20). Native procollagenase was purified from culture medium of human umbilical vein endothelial cells (20). The proenzyme was activated with trypsin (20 ng of trypsin/1 µg of proMMP-1) for 1 h at 37 °C (21); trypsin was subsequently inactivated with soybean trypsin inhibitor. The substrate peptide Dnp-Pro-Leu-Gly-Leu-Trp-Ala-D-Arg-NH₂ (DnpS) was synthesized by Stack and Gray (22); Ac-Pro-Leu-Gly-(2-mercapto-4-methyl-pentanoyl)-Leu-Gly-OEt (TPS) was purchased from Bachem (King of Prussia, PA). Type I collagen from calf skin, soybean trypsin inhibitor, 4,4-dithiopyridine, MOPS, and EGTA were from Sigma. CaCl₂ 2H₂O was from EM Science (Cherry Hill, NJ); it contained 6 ng zinc/g. Brij 35 (prepared and packaged under nitrogen) was from Pierce, and trypsin (sequencing grade) was from Promega (Madison, WI). Zinc reference standard solution was from Fisher.

MMP activity was measured with DnpS, TPS, or collagen as substrate. Hydrolysis of DnpS by rMMP-1 was assessed by high pressure liquid chromatography as described (22). Hydrolysis of TPS by HUVEC MMP-1 was measured in a continuous spectrophotometric assay using 4,4-dithiopyridine to trap the thiol peptide product, HSCH(iBu)-Leu-Gly-OEt (23). Assays were conducted in a 96-well UV-transparent microtiter plate with a Spectramax 250 UV-visible plate reader (Molecular Devices Corp., Sunnyvale, CA) set at 324 nm to record the kinetics of substrate hydrolysis. Collagenolytic activity was estimated under the conditions of Terato et al. (24). Collagen degradation products were separated by polyacrylamide gel electrophoresis (25); the fraction of degraded -chains was estimated by densitometric analysis of the Coomassie Blue-stained gels (26, 27). Fluorescence spectra were recorded with an SLM-Aminco SPF-500C spectrofluorometer interfaced to an IBM personal computer. All spectra have buffer blanks subtracted, and fluorescence intensities are presented in arbitrary units. Difference spectra were generated by subtracting the relevant absolute spectra using software supplied by SLM-Aminco. CD spectra were recorded with a Jasco J-710 spectropolarimeter as described previously (20). All assays and spectra were obtained at 25 °C. Zinc analyses were conducted by furnace atomic absorption analysis.

RESULTS

To assess the effect of Ca²⁺ binding on rMMP-1 secondary structure, we recorded the low UV CD spectrum of proMMP-1 in the presence and the absence of the chelating agent, EGTA. As shown in Fig. 1A, EGTA had no effect on the CD spectrum of the enzyme. The addition of Ca²⁺ to the EGTA-treated protein also elicited no observable change in CD spectrum. Similar results were obtained with the recombinant MMP-1 (data not shown). Thus, we conclude Ca²⁺ binding to MMP-1 resulted in no global changes in secondary structure.

We also measured the emission spectrum of rMMP-1 in the presence of 5 mM CaCl₂ before and after the addition of 6 mM EGTA to assess Ca²⁺-dependent changes in tertiary structure. Fig. 1B shows that the addition of EGTA resulted in decreased fluorescence intensity of 10% at 340 nm and a red shift in the emission maximum from 334 to 340 nm (compare spectra 1 and 2). Similar changes in emission intensity were observed when HUVEC procollagenase was diluted into Ca²⁺-free buffer (Fig. 1B, spectrum 4) and then supplemented with added Ca²⁺ (Fig. 1B, spectrum 3). This suggests that the fluorescence changes in spectra 1 and 2 result from dissociation of Ca²⁺ induced by the chelating agent. Lack of an obvious shift in emission maximum in comparing spectra 3 and 4 may result from the presence of the propeptide. This portion of the protein contains one tryptophan, which could have obscured the shift in emission maximum associated with Ca²⁺ dissociation from the enzyme lacking the propeptide.

The emission characteristics of the quenched residue (shown in Fig. 1B, spectrum 5) were assessed by subtracting the spectrum measured in the presence of EGTA (spectrum 2) from that obtained in its absence (spectrum 1). The maximum in the difference spectrum 5 at 320 nm indicates that the quenched fluorophores are predominantly tryptophyl residues that reside in a hydrophobic environment (28).

The contribution of tyrosyl residues to rMMP-1 fluorescence was ascertained by subtracting the emission spectrum obtained with excitation at 297 nm (where Trp alone absorbs) from the spectrum obtained with excitation at 280 nm (Trp and Tyr absorb). The resulting difference spectra (Fig. 2) allow comparison of the emission properties of tyrosyl residues in the Ca²⁺-free and Ca²⁺-bound forms of the enzyme. Each spectrum exhibits a maximum near 315 nm, which is strongly red shifted from that expected for tyrosine in aqueous solution (28). A small decrease in emission intensity was observed, suggesting that the Ca²⁺-dependent conformational differences are sensed by tyrosyl as well as tryptophyl residues.

Titration data illustrating the dependence of fluorescence and catalytic activity on [Ca²⁺] are shown in the Hill plots of Fig. 3. Ca²⁺ binding as assessed by changes in fluorescence was cooperative (Hill coefficient, 2.9; 50% saturation at 0.43 mM Ca²⁺). When the Ca²⁺ dependence of rMMP-1 activity was assayed with a peptide substrate, the resulting Hill plot also indicated cooperative Ca²⁺ binding (Hill coefficient, 1.7; 50% saturation at 0.23 mM [Ca²⁺]). When the [Ca²⁺] was manipulated by dilution rather than EGTA, cooperative Ca²⁺ binding to native HUVEC MMP-1 was also observed (Hill coefficient, 2.0; 50% saturation at 0.20 mM Ca²⁺). Because similar results were obtained whether [Ca²⁺] was altered by EGTA or dilution, we conclude that the effects of the chelating agent result from Ca²⁺ dissociation from the enzyme rather than removal of catalytic Zn²⁺.

To determine if the effects of Ca²⁺ on activity and fluorescence were reversible, we compared the catalytic activity of the recombinant enzyme with DnpS and collagen prior to Ca²⁺ removal, after Ca²⁺ removal with EGTA, and after reconstitution with Ca²⁺. Previous reports (11, 12) indicate that prolonged incubation of collagenase with EDTA resulted in removal of zinc and irreversible denaturation of the protein. The data in Table I show that activity against both DnpS and collagen was completely lost on treatment with EGTA. Loss of activity and fluorescence changes were reversed by adding Ca²⁺, even 24 h after removal of the metal ion.

Table I. Effect of EGTA and Ca2+ on the activity of rMMP-1 rMMP-1 (0.2 µM) in 150 mM Tris-Cl, 200 mM NaCl, 0.05% Brij 35, 5 mM CaCl2 was treated with EGTA under the indicated conditions. Fluorescence emission spectra were recorded, after which aliquots of the enzyme solutions were assayed with either collagen or DnpS as substrate under the same conditions. The emission spectra were identical to those in Fig. 1 and are not shown. Addition Activity (pmoles degraded/min) Collagena DnpSb None 2.7 9.7 7 mM EGTA 0 0 7 mM EGTA followed after 10 min by 8 mM CaCl2 2.6 10.2 7 mM EGTA followed after 24 h by 8 mM CaCl2 2.7 11.2 a  Activity determined at 25 °C for 50 min with 0.7 mg/ml collagen and 0.3 µM rMMP-1. b  Activity determined at 25 °C for 60 min with 20 µM DnpS and 0.2 µM rMMP-1.

DISCUSSION

The present studies, conducted with full-length rMMP-1 that includes both the catalytic and pexin domains, demonstrate that acquisition of catalytic activity by collagenase is coupled to protein conformational changes induced by Ca²⁺ binding. Binding was cooperative as assessed by both structural and activity changes. The observation that a lower Ca²⁺ concentration is required for catalytic activity compared with development of the fluorescence changes suggests that fewer binding sites influence catalysis, whereas the fluorescence changes involve sites other than those necessary for catalytic efficiency. Our studies with the full-length protein extend previous studies conducted with the catalytic domain alone. Lowry et al. (17) showed that both Ca²⁺ and Zn²⁺ stabilized a recombinant collagenase catalytic domain against denaturation by GdnHCl, and Housley et al. (18) reported that Ca²⁺ stabilized the catalytic domain of recombinant stromelysin against heat denaturation. In addition, activation of prostromelysin-1 by an organomercurial in the presence of 0.1 mM Ca²⁺ resulted in autolysis that could be prevented with 5 mM Ca²⁺ (18). These data indicate that Ca²⁺ stabilizes a compact protein structure that is less susceptible to denaturation and proteolysis. Our results show that Ca²⁺ binding to full-length rMMP-1 results in local conformational change(s) that do not detectably alter the secondary structure of the protein but do change tertiary structure as reflected in the environment of tryptophyl and tyrosyl residues. The change in tertiary structure in the absence of secondary structural changes suggests that Ca²⁺ removal may produce a "molten globule-like" structure (20).

The emission maximum of denatured MMP-1 is at 356 nm (20), characteristic of tryptophan in water, whereas the emission maximum of the Ca²⁺-deficient enzyme shifts only to 340 nm. Thus the change in tryptophyl environment brought about by Ca²⁺ release is less drastic than complete unfolding. In our study of GdnHCl-induced denaturation of rMMP-1, we observed a folding intermediate with an emission maximum at 340 nm at 1 M denaturant (20). However, it is unlikely that the Ca²⁺-free form of MMP-1 in the present study is the same as this intermediate because in 1 M GdnHCl, there was an increase, rather than a decrease, in emission intensity.

MMP-1 contains one tryptophyl residue in the propeptide, three in the catalytic domain and four in the pexin domain. Assignment of the fluorescence changes to particular residues is of course impossible from the data at hand. Lovejoy et al. (13) suggested that in 19-kDa collagenase one of the calcium ions and the structural zinc may stabilize a surface loop by fastening it to two  strands. Neither the loop nor the two strands contain tyrosyl or tryptophyl side chains that might be directly affected by Ca²⁺. Thus, it is likely that any conformational change induced by Ca²⁺ binding at this site is propagated to other regions of the structure. The two Ca²⁺ in the pexin domain of pig collagenase also appear to link structural domains together: -sheet-1 to sheet 3 and -sheet-4 to sheet 2 (16). Two of the four pexin tryptophyl residues are located within these sheets; Trp³⁴⁹ is in sheet 2, and Trp³⁹⁸ is in sheet 3. Thus, it is reasonable to assign at least a portion of the Ca²⁺-dependent fluorescence change to them. In addition, both tryptophyls are close in the primary structure to tyrosyl residues [YW³⁴⁹A and YW³⁹⁸RY (29)]. Variation in the relationship between these groups should influence the efficiency of excitation energy transferred from tyrosyl to tryptophyl residues. However, the quenching of tyrosyl emission observed on Ca²⁺ dissociation was accompanied by decreased, rather than increased, tryptophyl emission, as would be expected if altered transfer efficiency were the sole basis for the altered fluorescence.