Report

Structure of Human Pro-Matrix Metalloproteinase-2: Activation Mechanism Revealed

See allHide authors and affiliations

Science  04 Jun 1999:
Vol. 284, Issue 5420, pp. 1667-1670
DOI: 10.1126/science.284.5420.1667

Abstract

Matrix metalloproteinases (MMPs) catalyze extracellular matrix degradation. Control of their activity is a promising target for therapy of diseases characterized by abnormal connective tissue turnover. MMPs are expressed as latent proenzymes that are activated by proteolytic cleavage that triggers a conformational change in the propeptide (cysteine switch). The structure of proMMP-2 reveals how the propeptide shields the catalytic cleft and that the cysteine switch may operate through cleavage of loops essential for propeptide stability.

Matrix metalloproteinases (MMPs) are key enzymes involved in connective tissue turnover in normal and pathological conditions (1). MMPs exist in both invertebrate and vertebrate species. In vertebrates, they are expressed mainly in connective tissue cells and in cells of bone marrow origin. MMPs are extracellular enzymes [except for the membrane-type MMPs (MT-MMPs)] that are secreted as proenzymes. Their activity is controlled by transcriptional regulation, zymogen activation, and specific tissue inhibitors of metalloproteinases (TIMPs) (1, 2).

MMP-2 and MMP-9 degrade type IV collagen, the major component of basement membranes and denatured collagen (gelatin) (1, 2). MMP-2 is primarily expressed in mesenchymal cells (mainly fibroblasts) during development and tissue regeneration. It was originally isolated from a malignant mouse tumor and was found to be highly expressed in stromal cells surrounding the invading front of metastasizing tumors (3). This indicated that type IV collagenolytic activity is required by metastasizing tumor cells to traverse basement membranes at tissue boundaries and in blood vessels. Therefore, these enzymes are a promising target for the development of antitumor drugs.

All the MMPs are multidomain enzymes containing propeptide, catalytic, and hemopexin (except matrilysin, MMP-7) domains. Additionally, MMP-2 and MMP-9 contain three contiguous fibronectin type II-like domains that are inserted within their catalytic domain. A cysteine residue, strictly conserved in the propeptide domain of all MMPs, has been shown to be essential for maintaining the MMPs in an inactive state (2, 4). It has been suggested that the sulfhydryl group of this cysteine residue is coordinated to the catalytic Zn2+ ion and that interruption of this interaction causes activation [cysteine-switch mechanism (4)]. Physiologic activation of MMPs is probably initiated by proteases that cleave specific sites within the propeptide, but final processing to the mature form of the active MMP that lacks the entire propeptide often requires intermolecular, autoproteolytic cleavage by the target MMP. When triggered with sulfhydryl-reactive compounds, such as organomercurials that interrupt the cysteine to Zn2+ coordination (5–7), processing of proMMPs to the active form can be entirely autoproteolytic. The physiologic activation of MMP-2 has been poorly understood, but recent evidence has shown that the MT-MMPs can initiate activation of proMMP-2 by cleaving at a specific site within the propeptide (Figs. 1 and2) (8, 9). Thus far, little is known about the structural background of the cysteine-switch activation because structural work on MMPs has concentrated on isolated domains (10). The structure of COOH-terminally truncated proMMP-3 (stromelysin) revealed that, as predicted, the catalytic cleft is occupied by the cysteine-switch peptide (11). It remains unclear, however, how the rest of the propeptide contributes to the stability of the proenzyme and how limited cleavage within the propeptide can initiate the activation process.

Figure 1

Structure of proMMP-2. The prodomain, catalytic domain, fibronectin domains, and hemopexin domain are shown in red, blue, green, and yellow, respectively. Zn2+ ions are indicated in red, and Ca2+ ions are magenta (24). Asterisk indicates the cleavage site for MT1-MMP.

Figure 2

Structure of the prodomain of proMMP-2 (24). (A) Comparison of the prodomains of proMMP-2 (red) and proMMP-3 (11) (yellow) with the noncatalytic domain of bacteriald-alanyl-d-alanyl–cleaving carboxypeptidase(blue) (16). (B) Sequence (numbering of amino acid residues is according to the SwissProt database) comparison of MMP prodomains and the NH2-terminus of d-alanyl-d-alanyl–cleaving carboxypeptidase, with helices indicated in red. Solid arrowheads point to cleavage sites for proMMP activation (5, 6). Open arrowhead indicates processing site for MT-MMPs (9). A proteolytically sensitive region for activation of proMMP-3 (5) is underlined.

The COOH-terminal hemopexin-like domain of MMPs is linked to the catalytic domain by a hinge peptide, and it may determine the substrate specificity of MMPs (12). Its structure, a four-bladed propeller around a central cavity occupied by a Ca2+ ion, has been determined for MMP-1, -2, and -13 (13,14).

Here we report the crystal structure of the full-length proform of human MMP-2 (proMMP-2). Recombinant human proMMP-2 was produced in a mutant form in which Glu404, which is essential for catalytic activity of metalloproteases (15), was replaced by alanine. This mutant was stable against autoproteolysis and allowed the crystallization of full-length proMMP-2. We used molecular replacement (MR) techniques to determine the structure to 2.8 Å resolution (Table 1).

Table 1

Crystallographic data, phasing, and refinement. ProMMP-2 was produced as a proteolytically inactive mutant (Glu404 changed to Ala) with a baculovirus expression system and was purified by gelatin affinity chromatography and ion-exchange high-pressure liquid chromatography (6). Crystals were grown at 4°C by hanging drop crystallization. The drops contained a 1:1 mixture of protein solution (15 mg/ml) and reservoir buffer [0.2 M sodium glycine (pH 8.5) with 24 to 25% polyethylene glycol 550 monomethyl ether, 0.28 M Li2SO4, and 0.01 M dithiothreitol]. A heavy atom derivative was prepared by soaking the crystals in 5 mM Na2IrCl6. X-ray data from native and soaked crystals were collected at 100 K at beam line D2AM of the European Synchrotron Radiation Facility (ESRF, Grenoble, France) with a charge-coupled device detector and processed with the XDS software (23) followed by scaling and reduction with ROTAPREP and SCALA from the CCP4 package (23). The data showed a very high B factor and relatively strong anisotropy. An anisotropic correction was performed by SFCHECK (23). The crystals belonged to the tetragonal space group I4122 with cell dimensions a = b = 121.3 Å and c = 345.1 Å containing one 72-kD monomer per asymmetric unit. The structure was solved by MR in combination with a single heavy atom derivative and phases were improved by multicrystal averaging. MR was conducted with AMORE (23) using the model of the COOH-terminal domain of gelatinase A (MMP-2) and a model containing residues 100 to 204 of porcine fibroblast collagenase (MMP-1) (13). The iridium site determined from a difference Fourier map was refined by MLPHARE (23) using the model phases. Then SIR phases were combined with the model phases by using SIGMAA (23). The multicrystal averaging was performed with x-ray data for the COOH-terminal domain obtained from the PDB (1GEN) using DMMULTI (23). The protein model was built by using O (23) and refined to 2.8 Å resolution, including bulk solvent correction and grouped B-factor refinement with X-PLOR (23) and finally with highly restrained geometries with REFMAC (23). The model comprises 619 amino acid residues; 2 Zn2+, 3 Ca2+, 1 Na+, and 1 Cl ion; and 104 water molecules. No electron density was observed for the NH2-terminal residue Ala30 and for residues Ser448 to Leu461 of the hinge region. rmsd, root mean square deviation. The PDB accession number for proMMP-2 is 1CK7.

View this table:

An overall view of proMMP-2 is shown in Fig. 1. The propeptide of proMMP-2 forms a globular domain characterized by a three-helix fold that is stabilized by hydrophobic interactions and hydrogen bonds. This structural motif is similar to the NH2-terminal domain of d-alanyl-d-alanyl–cleaving carboxypeptidase from Streptomyces albus (Fig. 2A) (16). Comparison of the propeptide structures of proMMP-2 and proMMP-3 (11) shows that the cysteine-switch strand is bound into the catalytic cleft by several hydrogen bonds that are well conserved in both proMMPs. The helical part of the propeptides is similar, but the loops connecting the helices are different between proMMP-2 and proMMP-3 (11). The sites that get cleaved upon proteolytic activation are accommodated within these loops (Fig. 2B) (5, 6). MT-MMPs cleave proMMP-2 within the loop (Tyr58 to Asn66) that connects helices H1 and H2 (Fig. 2) (9). This loop contains a disulfide bridge (Cys60 to Cys65) that is unique to proMMP-2. The second loop (Phe81 to Ile94), stabilized by two internal hydrogen bonds, is also cleaved upon autoproteolytic activation (Fig. 2). This loop is probably important for the stability of the proenzyme because it contains hydrophobic side chains that isolate the catalytic cleft and the cysteine-switch strand from solvent molecules.

The architecture of the catalytic domain, known as the matrixin fold (17), consists of a five-stranded β sheet and three α helices. This structure is highly conserved in MMPs and is unaffected by insertion of the fibronectin domains. The catalytic domain of proMMP-2 is similar to that of proMMP-3. The same residues form the substrate binding pockets and coordination of the catalytic Zn2+ ion is similar. Replacement of the charged Glu404 with alanine in proMMP-2 has no influence on the architecture of the active site. Also, the binding site for the structural Zn2+ ion is identical to a well-conserved motif found in all known MMP structures. The catalytic domain of proMMP-2 revealed one Ca2+ ion within the S-shaped loop and a second Ca2+ ion bound by two peripheral loops (Fig. 1).

The fibronectin domains of proMMP-2 (Figs. 1 and3) are inserted between the fifth β strand and helix 2 in the catalytic domain. The basic fold of the fibronectin type II-like domain (18) comprises a pair of β sheets, each made from two antiparallel strands, that are connected with a short α helix. The two β sheets form a hydrophobic pocket that is accessible from the outside. A cis-proline following the first β sheet is part of a hairpin turn, which orients the surrounding aromatic side chains into the hydrophobic pocket. These pockets are the structural hallmark of the fibronectin domains and probably account for substrate binding. In MMP-9, gelatin binding residues have been mapped to the hydrophobic pocket (Fig. 3) (19). In proMMP-2, the side chain of Phe37in the propeptide inserts into the hydrophobic pocket of the third fibronectin domain (Fig. 4A). The propeptide of proMMP-2 is also bound to the third fibronectin domain by a hydrogen bond and a salt bridge (Fig. 4A). This interaction probably mimics the binding of gelatin to fibronectin type II-like domains that, based on biochemical evidence, is predicted to have all three types of interactions (18). The binding sites of the three fibronectin domains are not oriented toward each other to form a continuous binding motif as previously proposed (20). On the contrary, they turn outward as in a three-pronged fishhook (Fig. 3).

Figure 3

Structure of fibronectin domains (24). (A) Stereoview of the electron density map contoured around cis-Pro236 (P236) and the Cys233 (C233) to Cys259 (C259) disulfide bridge. Sulfur atoms are yellow. (B) Ribbon representation showing the secondary structure with the disulfide bridges. (C and D) Molecular surface of the fibronectin domains show polar (green) and hydrophobic (red) residues that bind to gelatin (19). The orientation in (B) and (C) is the same, and the molecule in (D) is rotated by 180°. F, Phe; L, Leu; W, Trp.

Figure 4

Interaction between propeptide and fibronectin domains (24). (A) Stereoview of the propeptide residues in contact with the hydrophobic pocket of the third fibronectin domain. Hydrophobic residues from the fibronectin domain are magenta and propeptide residues are yellow. Arg368(R368) and Gly367 (G367) (cyan) form hydrogen bonds with the propeptide residues Asp40 (D40) and Ile35(I35), respectively. (B) Molecular surface of full-length proMMP-2. Colored surfaces associated are with negatively charged (red) and positively charged (blue) residues. Propeptide residues are represented as a ball-and-stick model. Y, Tyr.

The hemopexin domain shows a four-blade propeller fold (13,14). The first and second propeller blades are oriented toward the catalytic domain and are linked to the first fibronectin domain by a hydrogen bond (Glu243 to Arg550). This orientation turns propeller blades 3 and 4 away from the catalytic domain. On blades 3 and 4 are the binding sites for TIMP-2, a protein inhibitor that specifically interacts with proMMP-2 (21). It is unclear whether binding of TIMP-2 is the only function of the hemopexin domain in MMP-2 or whether it also modulates substrate specificity. Our structure supports the latter as the surface structure of proMMP-2 (Fig. 4B) reveals that the hemopexin domain contributes to a groove that may be involved in substrate binding.

These results provide a structural basis for understanding the activation mechanism of proMMP-2. Loops within the propeptide domain function as bait for activating proteases. Upon cleavage, the prodomain structure breaks down and its shielding of the catalytic cleft is withdrawn, allowing water to enter and hydrolyze the coordination of the cysteine to the Zn2+ ion.

MMPs, particularly MMP-2, have been a target for the development of antitumor therapeutics that inhibit the motility of malignant cells, a prerequisite for tumor invasion and formation of metastases. MMP inhibitors against the active site have been designed (22), but structural homology of their catalytic domains has made specificity a problem. The full-length proMMP-2 structure may provide alternative concepts for development of specific MMP-2 antagonists.

  • * To whom correspondence should be addressed. E-mail: gunter{at}alfa.mbb.ki.se (G.S.); karl.tryggvason{at}mbb.ki.se (K.T.).

REFERENCES AND NOTES

View Abstract

Stay Connected to Science

Navigate This Article