HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

Published online before print December 21, 2006

This Article Abstract Full Text (PDF) All Versions of this Article: jlb.1006629v1 81/4/870    most recent Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Piccard, H. Articles by Opdenakker, G. Search for Related Content PubMed PubMed Citation Articles by Piccard, H. Articles by Opdenakker, G.

(Journal of Leukocyte Biology. 2007;81:870-892.)
© 2007 by Society for Leukocyte Biology

Hemopexin domains as multifunctional liganding modules in matrix metalloproteinases and other proteins

Rega Institute for Medical Research, Laboratory of Immunobiology, University of Leuven, Leuven, Belgium

¹ Correspondence: Rega Institute for Medical Research, Laboratory of Immunobiology, University of Leuven, Minderbroedersstraat 10, 3000 Leuven, Belgium. E-mail: ghislain.opdenakker{at}rega.kuleuven.be

	ABSTRACT

TOP ABSTRACT INTRODUCTION HEMOPEXIN AS A MAJOR... THE HEMOPEXIN SUPERFAMILY FUNCTIONS OF THE HEMOPEXIN... CONCLUSIONS AND PERSPECTIVES REFERENCES

 
The heme-binding hemopexin consists of two, four-bladed propeller domains connected by a linker region. Hemopexin domains are found in different species on the phylogenetic tree and in the human species represented in hemopexin, matrix metalloproteinases (MMPs), vitronectin, and products of the proteoglycan 4 gene. Hemopexin and hemopexin domains of human proteins fulfill functions in activation of MMPs, inhibition of MMPs, dimerization, binding of substrates or ligands, cleavage of substrates, and endocytosis by low-density lipoprotein receptor-related protein-1 (LRP-1; CD91) and LRP-2 (megalin, GP330). Insights into the structures and functions of hemopexin (domains) form the basis for positive or negative interference with the formation of molecular complexes and hence, might be exploited therapeutically in inflammation, cancer, and wound healing.

Key Words: vitronectin • endocytosis • heme metabolism • chemokine • collagenolysis • pericellular proteolysis

	INTRODUCTION

TOP ABSTRACT INTRODUCTION HEMOPEXIN AS A MAJOR... THE HEMOPEXIN SUPERFAMILY FUNCTIONS OF THE HEMOPEXIN... CONCLUSIONS AND PERSPECTIVES REFERENCES

 
Free heme, a major molecule for the transport of oxygen throughout the body, is highly toxic. Several scavenger proteins of free heme are known, of which hemopexin is a major one. It is remarkable that numerous proteins contain one or several repeats that structurally (and functionally) share homology with hemopexin. These comprise viral, prokaryotic, and eukaryotic proteins and together, can be called "the hemopexin superfamily." Matrix metalloproteinases (MMPs), present in many species, form the major family subgroup. Despite the homology with hemopexin, hemopexin domains exert divergent, specific interactions with different proteins. Here, we compare the degree of sequence homology with human hemopexin and review the functions of hemopexin domains of different species, with the main focus on human MMPs. The knowledge of the divergent interactions and biological implications of hemopexin domains could help researchers to design inhibitors of the interaction and/or proteolysis of specific substrates and leave other functions unaffected, thus creating "disease-specific" inhibitors.

	HEMOPEXIN AS A MAJOR HEME SCAVENGER

TOP ABSTRACT INTRODUCTION HEMOPEXIN AS A MAJOR... THE HEMOPEXIN SUPERFAMILY FUNCTIONS OF THE HEMOPEXIN... CONCLUSIONS AND PERSPECTIVES REFERENCES

 
Oxygen, essential for survival of aerobic organisms, is transported from the lungs throughout the body by hemoglobin, the major protein of erythrocytes. Hemoglobin is also responsible for transporting most of the carbon dioxide and of protons from the peripheral tissues to the lungs [¹ ]. Human hemoglobin molecules consist of four globin subunits, each bearing a prosthetic "heme" group. Heme structures are also present in several other hemoproteins (listed in Tsiftsoglou et al. [² ]), such as myoglobin. Heme is also an essential component of enzymes or allosteric proteins such as certain catalases, peroxidases, and cytochromes. A heme molecule contains a porphyrin ring and a single atom of iron that can combine with small gas molecules, including oxygen, nitric oxide (NO), and carbon monoxide [² , ³ ]. In cases of destruction of erythrocytes (e.g., hemolysis) and enucleation of erythroblasts, which are accompanied by the release and metabolization of hemoglobin, heme is released into extracellular fluids. Plasma heme can also originate from other proteins with heme groups, for instance, from damaged tissues. This free heme can have detrimental consequences for health, as it can auto-oxidize spontaneously, which results in the formation of reactive oxygen species (ROS; radicals). Thus, free heme generates oxidative stress known to damage cells. Therefore, it also has proinflammatory effects and is potentially contributing to various pathologies, such as the development of intravascular hemolysis and atherosclerosis [⁴ , ⁵ ]. Free heme porphyrin is converted to biliverdin and bilirubin, possessing potent neurotoxicity, unless these molecules are detoxified by conjugation in the liver [⁶ ]. Moreover, heme is potentially toxic as a result of its ability to intercalate into lipid membranes [⁷ ]. Free heme is also an important source of essential iron for pathogenic microorganisms. Heme excess conditions predispose to infections. Fortunately, several mechanisms are present to eliminate free heme. For instance, free heme is bound by high-density lipoproteins and low-density lipoproteins (HDL and LDL, respectively), serum albumin, and hemopexin, and the complexes are removed from the circulation. Thus, the level of free heme is generally low. Nevertheless, the amount of free heme can increase under (pathological) conditions whereby massive hemolysis occurs [⁴ ]. In this case, ₁-microglobulin cooperates to heme scavenging [⁷ ]. In the further part of this introduction, we will highlight hemopexin as a heme scavenger.

Hemopexin is probably the primary specific carrier of plasma heme, having the highest binding affinity for heme [⁷ ]. Human hemopexin belongs to the acute-phase proteins whose expression can be induced by various cytokines in a context of inflammatory processes [⁸ ]. Hemopexin is expressed mainly in the liver but also in the CNS and peripheral nervous system and in the retina [⁷ ]. The protein transports heme to the LDL receptor-related protein-1 (LRP-1; also named CD91). LRP-1 is a receptor for heme:hemopexin complexes and is present on several cell types, such as hepatocytes, macrophages, neurons, and syncytiotrophoblasts [⁹ , ¹⁰ ]. Upon binding of heme:hemopexin to LRP-1, the complex becomes internalized via endocytosis into cells, mainly hepatocytes and macrophages in liver and spleen [⁹ , ¹¹ ]. Inside the cell, the heme:hemopexin complex is dissociated by lysosomal activity. Heme is catabolized by heme oxygenases into biliverdin, carbon monoxide, and iron [¹² ]. Studies suggest that hemopexin can be recycled as an intact molecule to the extracellular milieu [¹³ ]. However, Hvidberg et al. [⁹ ] indicate that most hemopexin is degraded in lysosomes.

In this manner, the binding of hemopexin to free heme limits the amounts of heme as a catalyst of radical formation in the circulation, makes the essential iron unavailable to invasive microorganisms, and contributes to recycling of iron components, as the heme iron atoms enter the intracellular iron pool [⁴ ]. Hemopexin-deficient mice recover more slowly after hemolysis and have more severe kidney damage than wild-type mice. These observations emphasize the protective role of hemopexin in hemolytic conditions [¹⁴ ].

The human hemopexin protein is a 60-kD plasma glycoprotein, which consists of two homologous, four-blade propeller domains, the aminoterminal and the carboxyterminal domain, linked by a 20-amino acid linker structure [⁷ ]. Each domain consists of four repeats of 50 residues (Fig. 1 ), which are suggested to originate from four duplications of an ancestral exon. Based on comparison analyses of the sequences of introns and exons, the two-domain structure of hemopexin most probably results from a gene duplication [¹⁵ ]. A disulfide bridge connects the first and fourth blades of each propeller. The two domains of hemopexin lock together with a 90° angle, and the face of the aminoterminal domain is packed against one edge of the carboxyterminal domain. Each propeller blade comprises a four-stranded, antiparallel ß-sheet. Connecting loops between the propeller blades provide the structural features that form the heme-binding site. Heme is bound reversibly at the center of hemopexin in a 1:1 ratio, with relatively high affinity between the two propeller domains and constrained by the linker between them. This linker plays a critical role in constituting the heme-binding site of hemopexin as it forms the outer boundary of the binding pocket. Moreover, the linker sequence provides, at Position 238, one of the His ligands for the heme iron ion, helping to stabilize the heme:hemopexin complex. The second histidine residue that coordinates the heme iron is located at position 291 in one of the loops of the carboxyterminal domain of hemopexin [¹⁶ ].

View larger version (37K): [in this window] [in a new window]   Figure 1. Crystal structures of hemopexin domains. The crystal structure of rabbit hemopexin [16 ] is compared at the same scale with that of human gelatinase A/MMP-2 [34 ]. In the hemopexin structure, the aminoterminal domain is oriented in such a way to illustrate the four-bladed propeller (red box), also visible in gelatinase A/MMP-2.

	THE HEMOPEXIN SUPERFAMILY

TOP ABSTRACT INTRODUCTION HEMOPEXIN AS A MAJOR... THE HEMOPEXIN SUPERFAMILY FUNCTIONS OF THE HEMOPEXIN... CONCLUSIONS AND PERSPECTIVES REFERENCES

Numerous eukaryotic proteins are described to contain hemopexin domains. The phosphocholine-binding protein "limunectin" from Limulus amebocytes contains 10 serial, hemopexin-like repeats, and each repeat has a pair of cysteine residues, possibly forming a disulfide loop. The protein is therefore suggested to contain two or three hemopexin-like propellers. Limunectin is shown to bind to bacteria, fixed amebocytes, and several ECM components [¹⁸ ].

Jenne [¹⁹ ] described the first plant protein containing hemopexin-like repeats, pea albumin 2 (PA2), which is a cytosolic protein and a major albumin of the pea (P. sativum) seed. PA2 is composed entirely of four hemopexin-like repeats. Each repeat is predicted to contain four ß-strands, and two adjacent repeats are likely to form together a ß-barrel. Two such ß-barrels would then make up one globular hemopexin domain. This single hemopexin domain could be regarded as an intermediate form in the evolutionary pathway to the two-domain structure of hemopexin. The biological function of PA2 is yet unknown. However, a transport and storage function is assumed. The presence of hemopexin domains in a plant protein shows that hemopexin repeats most likely originated before the divergence of the animal and plant kingdoms [¹⁹ ].

Nectinepsin is described as a putative protein for which cDNA and mRNA are found in the neuroretina, liver, brain, and intestine of quails (C. coturnix) and in the murine retina. The deduced amino acid sequence, which is 60% identical to human vitronectin (vide infra) and shares homology with MMPs, indicates the presence of a hemopexin-like repeat. Specifically, nectinepsin contains a domain similar to the aminoterminal part of the second repeat of hemopexin but lacks the carboxyterminal part of this repeat and lacks the first, third, and fourth repeats of hemopexin domains. The mRNA level in the neuroretina increases during embryonic development. Further studies are necessary to elicit which role nectinepsin could play in the early steps of embryonic development [²⁰ ].

Human vitronectin, also known as serum-spreading factor or complement S-protein, is a glycoprotein found in plasma and the ECM. Vitronectin is multifunctional, serving as a cell-to-substrate adhesion molecule, interacting with proteins of the integrin receptor and serpin families, and being an inhibitor of membrane damage by binding to the membrane attack complex of the cytolytic complement pathway [¹⁹ , ²⁷ ]. Vitronectin consists of four structurally different domains: a somatomedin domain, Hemopexin 1 (N-glycosylated) and Hemopexin 2 domains, and a connecting region that links the somatomedin and Hemopexin 1 domains. Jenne and Stanley [¹⁵ ] suggested that vitronectin contains seven hemopexin-like repeats; this is one less than the hemopexin molecule. However, the alignment analysis presented in this review indicates the presence of five hemopexin-like repeats (Fig. 2 ). In the Hemopexin 2 domain, binding sites for heparin and the complement complex are localized. The interaction site for plasminogen activator inhibitor-1 is situated in the somatomedin B domain and/or Hemopexin 2 domain. Furthermore, binding capacities for sulfatide, phosphatidylserine, cholesterol 3-sulfate, collagen type I, and ß-endorphin are attributed to the Hemopexin 1 and Hemopexin 2 domains [²¹ ]. Moreover, these hemopexin-like repeats are found to be recognized by the bacteria S. pyogenes. Thus, pathogens could adhere to host cells via the hemopexin domains of host cell vitronectin molecules or gain complement resistance by binding vitronectin as a complement inhibitor. Moreover, S. pyogenes binds to hemopexin molecules [²² ].

View larger version (136K): [in this window] [in a new window]   Figure 2. Alignment of hemopexin-like motifs and hemopexin domains. A multiple-analysis sequence alignment, including color-coding, was conducted using AlignX of Vector NTI software (Invitrogen, San Diego, CA, USA) with default parameters. The amino acid sequences for each of the proteins were obtained from the "Entrez search and retrieval system" (www.ncbi.nlm.nih.gov). The accession numbers of the sequences and the amino acid numbers, which were used for this analysis, are indicated in Table 2 . The first column contains the names of the aligned proteins. Sequences of hemopexin domains of MMPs (MMP_HPX) are subdivided, and the denomination is colored as follows: gelatinases, dark blue; collagenases, dark green; stromelysins, pale blue; MT-MMPs, pale green; other MMPs, light orange. The first aligned residue number of each protein is indicated in the second column. The resulting consensus sequence, which contains the amino acid residues that are most common at a particular alignment position, is indicated at the bottom of the alignment. The amino acids in the alignment are colored as follows: black on default, nonsimilar residues at a given alignment position; green on default, residue that is weakly similar to the consensus residue at a given position; blue on cyan, consensus residue derived from a block of similar residues at a given position; black on green, consensus residue derived from the occurrence of more than 50% of a particular residue at a given position. The first and last cysteine residues, which define the hemopexin domains, are indicated in dashed boxes. At the N terminus of the carboxyterminal domain of hemopexin, two gaps were inserted between histidine (H) and leucine (L) to align the cysteine residue at Position 257. In Repeat 4, for all proteins except MMP-21, six gaps were inserted aminoterminally of the conserved cysteine residue to maximize the alignment. Residues indicated in bold red represent the interaction between the hemopexin domain and prodomain of MMP-1 and between MMP-2 and tissue inhibitors of metalloproteinase 2 (TIMP-2), respectively. HPX_Adom, Aminoterminal domain of hemopexin; HPX_Cdom, carboxyterminal domain of hemopexin; vtnc_rpt1, aminoterminal region of hemopexin-like repeats of vitronectin; vtnc_rpt2, carboxyterminal region of hemopexin-like repeats of vitronectin; prg4_HPX, hemopexin repeats encoded by the PRG4 gene.

MMPs have been studied profoundly in vertebrates but are also found in lower animals and plants. In human, 23 MMPs are identified. Human MMPs are multidomain enzymes and consist of at least a pro-, catalytic-, and Zn²⁺-binding domain. The catalytic- and Zn²⁺-binding domains together form the active site for proteolysis of substrates. The Zn²⁺-binding domain of MMPs contains a conserved sequence with three histidine residues, which coordinate with the catalytic Zn²⁺ ion. In proMMPs, the fourth ligand of the Zn²⁺ ion is a cysteine residue of the prodomain. Thus, the prodomain regulates the latency of the proteolytic activity of MMPs [³¹ ]. It is speculated that MMPs originate from single domains, which by gene fusion, became multidomain enzymes. This is likely to be an early evolutionary event, followed by diversification to result in the many different MMPs [²⁶ , ³² ]. In most human MMPs, except in matrilysins (MMP-7 and -26) and cysteine array-MMP/MMP-23, a carboxyterminal, hemopexin-like domain with various functions is present. This domain, consisting of 210 amino acids, shows structural similarity with hemopexin. The hemopexin domain forms a propeller with four blades, arranged around a central axis, of which the first blade is linked to the fourth by means of a disulfide bridge. Each blade consists of four antiparallel ß-sheets and one -helix [³³ ]. For comparison of hemopexin and the hemopexin domain of a MMP, the crystal structure of gelatinase A/MMP-2 [³³ ] is oriented in such a way in Figure 1 that the four-bladed propeller structure is visible. Whereas the structure around the active site is quite conserved between MMPs, the hemopexin domains are more differentiated [³⁵ ]. The hemopexin domain is absent in plant and nematode MMPs. A detailed evolutionary dendrogram of 64 hemopexin domains derived from the MMPs of different species has been elaborated [²⁶ ] and illustrates conservation and MMP-specific clustering.

Figure 2 represents a multiple-alignment analysis of all human proteins with hemopexin repeats or domains. A total of 20 hemopexin domains of human MMPs are aligned together with each of the two domains of human hemopexin. Furthermore, the sequence of the hemopexin domain encoded by the PRG4 gene and the hemopexin repeats of vitronectin are included in the analysis. The latter protein contains two distinct regions of hemopexin-like repeats, which are included separately in this alignment analysis. The hemopexin domains are defined based on the two cysteine residues that form the disulfide bound between the first and fourth blades of the propeller structure. In the vitronectin repeats, only one of these cysteines is conserved. Four sequence repeats are distinguished, which each form a blade of the hemopexin domain propeller. Each repeat has in the aminoterminal region the motif aspartic acid-alanine (DA). In collagenase-1/MMP-1, these aspartic acid residues of Repeats 1–3 are known to coordinate a central calcium ion of the hemopexin domain. Glutamic acid at Position 329 provides the fourth coordination to the ion [³⁶ ]. Pairwise alignments were conducted and show that hemopexin domains of human MMPs are, in general, slightly more homologous to the aminoterminal domain of the hemopexin protein than to the carboxyterminal one (Table 2 ). Only the hemopexin domains of collagenase-3/MMP-13 and MMP-28 are more related to the carboxyterminal domain of hemopexin.

Divergent functions have been ascribed to the hemopexin domain of MMPs. Although MMPs are detected in different species [³⁹ , ⁴⁰ ], in this review, the functions of the hemopexin domain of human MMPs are described in detail (summarized in Table 3 ). Therefore, the indicated amino acid residue numbers all refer to the human species, conform with the sequences and database accession numbers as listed in Table 2 . Specifically, we describe the contribution of the hemopexin domain of human MMPs to efficient activation and inhibition of several MMPs and how it influences the substrate specificity of the proteinases by the presence of exosites for the binding of substrates outside of the active site [⁴¹ ]. Moreover, the hemopexin domain is crucial for binding of MMPs to specific membrane proteins ("docking" molecules and endocytosis receptors) and thus, in the subcellular localization and/or catabolism of several MMPs.

	FUNCTIONS OF THE HEMOPEXIN DOMAINS OF HUMAN MMPs

TOP ABSTRACT INTRODUCTION HEMOPEXIN AS A MAJOR... THE HEMOPEXIN SUPERFAMILY FUNCTIONS OF THE HEMOPEXIN... CONCLUSIONS AND PERSPECTIVES REFERENCES

In the activation process of gelatinase A/MMP-2, MT1-MMP/MMP-14 plays a double role. First, cell surface MT1-MMP/MMP-14 binds a TIMP-2 molecule, together, forming a receptor complex for progelatinase A/MMP-2 [⁴³ ]. In this complex, TIMP-2 functions as an adaptor molecule, mediating the binding of progelatinase A/MMP-2 to MT1-MMP/MMP-14 by means of hydrophobic interactions between the carboxyterminal domain of TIMP-2 and the hemopexin-like domain of progelatinase A/MMP-2 (vide infra) [³³ , ⁴⁴ ]. In this manner, progelatinase A/MMP-2 is concentrated at the cell surface and brought in proximity of MT1-MMP/MMP-14 for catalysis.

As MT1-MMP/MMP-14 and TIMP-2 interact via the active site of MT1-MMP/MMP-14 and the inhibitory site of TIMP-2, the catalytic activity of MT1-MMP/MMP-14 is inhibited by TIMP-2 [⁴⁵ , ⁴⁶ ]. A second, nearby MT1-MMP/MMP-14 molecule, free of TIMP-2, removes part of the prodomain of the progelatinase-molecule [⁴⁷ ]. Thus, an intermediate, 68-kDa form of gelatinase A/MMP-2 is generated, which will be processed autocatalytically to the active, 66-kDa form of the MMP [⁴⁸ , ⁴⁹ ]. Studies to refine the activation mechanism reveal that TIMP-2 is essential for the autocatalytical activation cleavage of progelatinase A/MMP-2 [⁵⁰ ].

For the cell surface activation of gelatinase A/MMP-2 by MT1-MMP/MMP-14 molecules to occur efficiently, it is indispensable that receptor and activator MT1-MMP/MMP-14 molecules are in close proximity. Itoh et al. [⁵¹ ] demonstrated that MT1-MMP/MMP-14 molecules are expressed at the cell surface as homodimer structures. By the formation of such oligomeric structures, the activator and receptor MT1-MMP/MMP-14 molecules are positioned closely to each other, thus facilitating activation of progelatinase A/MMP-2. The oligomers are formed via the hemopexin, transmembrane, and cytoplasmic domains [⁵² ]. Within the cytoplasmic domain, dimerization occurs by the cysteine residue at Position 574, resulting in dimeric, covalently bound MT1-MMP/MMP-14 molecules [⁵³ ]. Unfortunately, until now, the exact participation of the hemopexin and transmembrane domains in oligomerization of MT1-MMP/MMP-14 remains obscure [⁵² ]. Several contradictory studies have been published about the role of the hemopexin domain of MT1-MMP in the activation process of gelatinase A/MMP-2. Thus, Wang et al. [⁵⁴ ] detected that MT1-MMP, without its hemopexin domain, can still activate progelatinase A/MMP-2. In contrast, Itoh et al. [⁵¹ ] described that by deletion or replacement of the hemopexin domain, the activation process of progelatinase A/MMP-2 is abolished. Furthermore, Atkinson et al. [⁵⁵ ] showed that MT1-MMP, bearing the hemopexin domain of MT4-MMP, is incapable of activating progelatinase A/MMP-2. However, in the latter study, the lack of activation of progelatinase A/MMP-2 was shown to be the result of a failure of the chimaeric proteins to reach the cell surface [⁵⁵ ]. The expression of the hemopexin domain of MT1-MMP/MMP-14 in cells interferes with gelatinase A/MMP-2 processing [⁵⁶ ].

Also, other mechanisms to concentrate gelatinase A/MMP-2 at the cell surface have been described (vide infra). For instance, it can bind noncovalently to heparin. This binding occurs via the hemopexin domain of gelatinase A/MMP-2 and leads to an increased activation status [⁵⁷ ].

Furthermore, other MT-MMPs can also activate progelatinase A/MMP-2 efficiently. MT2-MMP/MMP-15 processes progelatinase A/MMP-2 to the fully active form. In this activation process, TIMP-2 molecules are dispensable, but the hemopexin domain of gelatinase A/MMP-2 would be required for efficient activation at the cell surface by binding via the hemopexin domain, directly or indirectly, to MT2-MMP/MMP-15 [⁵⁸ ]. Furthermore, the second activation cleavage, from an intermediate to fully active form of gelatinase A/MMP-2, is dependent on the hemopexin domain of MT2-MMP/MMP-15 and is in the absence of TIMP-2, enhanced by a yet-unknown, soluble factor [⁵⁹ ]. Deletion of the hemopexin domains of MT2-, MT3-, MT5-, and MT6-MMP does not abolish the ability of these MMPs to activate progelatinase A/MMP-2 [⁵⁴ ].

Activation of progelatinase A/MMP-2 by MT-MMPs seems to be an important step in the invasion process of cancer cells, as active gelatinase A/MMP-2 can degrade components of the basement membrane, for instance, denatured collagen type IV and laminins, thus allowing cell migration [⁶⁰ ]. It is interesting that an intracerebral supply of an exogenous recombinant human hemopexin domain of gelatinase A/MMP-2 inhibits the growth of gliomas in mice, accompanied with a decrease in tumor vasculature and a change in tumor vessel morphology [⁶¹ ].

Activation of procollagenase-3/MMP-13 at the cell surface
Procollagenase-3/MMP-13 may be activated by means of soluble or cell-associated MT1-MMP/MMP-14 [⁶² ]. Initially, MT1-MMP/MMP-14 hydrolyzes the peptide bound between Gly₅₄ and Ile₅₅. Next, the formed, 56-kDa intermediate is converted autoproteolytically to the active form of 48 kDa. This activation process does not require the presence of TIMP-2. The hemopexin domain of procollagenase-3/MMP-13 is crucial for cell-associated activation, as a mutant form that lacks this domain and a chimeric form with the hemopexin domain of stromelysin-4/MMP-19 are not cleaved by MT1-MMP/MMP-14 [⁶³ ]. It is not known which receptor participates in the binding to the cell surface.

Structural rearrangement of MMP-1 during its activation process
Recently, the crystal structure of procollagenase-1/MMP-1 showed not only interactions between the prodomain and catalytic domain but also revealed intramolecular interactions between the hemopexin domain and the prodomain of this MMP (Table 3) . The main contact between both domains is hydrophobic: the amino acids Phe, Tyr, and Pro (Positions 308–310) in the loop between the ß₃ and ß₄ sheets of the first blade of the hemopexin domain interacting with Gly₇₂, Leu₇₃-Lys₇₄, and Asp₈₀-Glu₈₂ in the prodomain of proMMP-1. In addition, the side-chain of Glu₈₂ in the prodomain forms hydrogen bonds to the side-chain of Tyr₃₀₉ and Asn₃₀₆ of the hemopexin domain. By these interactions, a closed, compact configuration for procollagenase-1/MMP-1 is generated. During the activation process, upon cleavage of the bait region of the proMMP, these interactions change, resulting in an altered, rotated position of the hemopexin domain in relation to the catalytic domain. Particularly, the Pro₃₀₈-Tyr₃₀₉-Pro₃₁₀ region of the hemopexin domain undergoes the most significant conformational change. Together, this results in an open, destabilized arrangement of active, interstitial collagenase-1/MMP-1 with a widened cleft between the catalytic and hemopexin domains, and the active site residues and a sequence in the catalytic domain, which is essential for collagenolysis becoming more accessible [³⁶ , ⁶⁴ ]. Thus, a partially active proteinase is generated, which allows further, complete activation by stromelysin-1/MMP-3, gelatinase A/MMP-2, or matrilysin-1/MMP-7 [^{99 100 101} ].

Inhibition of MMPs
Inhibition of MMPs by binding of TIMPs
TIMPs are the major, natural inhibitors of MMPs. In addition, other proteins with MMP inhibitory activity have been discovered, such as the procollagen C-terminal proteinase enhancer, the reversion-inducing cysteine-rich protein with Kazal motifs, and tissue factor pathway inhibitor 2. However, these proteins have modest inhibitory capacity against MMPs compared with TIMPs [¹⁰² ].

Until now, four different human TIMPs have been discovered: TIMP-1, -2, -3, and -4. TIMPs are (glyco-) proteins with a molecular weight between 20 and 30 kDa, which function as important physiological regulators of the activity of MMPs. TIMP molecules structurally consist of two domains: an inhibitory, aminoterminal domain (125 amino acids) and a carboxyterminal domain (65 amino acids), each containing three disulfide-bound loops [¹⁰³ ].

TIMP molecules form high-affinity, noncovalent, inhibitory complexes with MMPs in a 1:1 enzyme:inhibitor ratio [⁶⁵ ]. TIMPs have different specificities for different MMPs. For instance, TIMP-1 binds with high affinity to gelatinase B/MMP-9, in contrast with TIMP-2 and -3, which have low affinity for gelatinase B/MMP-9. TIMP-2, -3, and -4 interact with high affinity with gelatinase A/MMP-2. TIMP-1 is an inducible protein, as opposed to TIMP-2, which has a constitutive expression pattern [⁶⁶ ].

Binding of the aminoterminal domain of a TIMP molecule to the active site of a MMP inhibits the proteolytical activity [¹⁰³ ]. Moreover, a number of MMPs, including gelatinases, contain a site in the hemopexin domain for binding the carboxyterminus of TIMPs [⁶⁷ ]. Such carboxyterminal interactions have been described for TIMP-2, -3, and -4 with gelatinase A/MMP-2, for TIMP-1 and -3 with gelatinase B/MMP-9 and collagenase-3/MMP-13 and also for TIMP-1 and collagenase-1/MMP-1 [^{66 67 68 69 70} , ⁷³ , ⁷⁴ ].

Carboxyterminal interactions result in a higher affinity between TIMPs and MMPs. For instance, TIMP-1 inhibits deletion mutants of gelatinase B/MMP-9 without the intact, carboxyterminal, hemopexin domain, 17.5-fold less efficiently than the intact proteinase. In contrast, deletion of the hemopexin domain of gelatinase B/MMP-9 does not influence the efficiency of gelatinase B/MMP-9 inhibition by the less-specific TIMP-2. TIMPs not only bind activated MMPs, but TIMP molecules also bind inactive proMMPs, namely by interaction with the hemopexin domain. To bind TIMP-1, the hemopexin domain is indispensable for progelatinase B/MMP-9. An intact, disulfide bridge at Position 704 in the carboxyterminal part of the hemopexin domain is necessary for complex formation [⁷¹ ]. Moreover, recent experiments indicate that the O-glycosylated and hemopexin domains of progelatinase B/MMP-9 are critical for optimal binding of TIMP-1 [⁷² ].

For progelatinase A/MMP-2 and TIMP-2, two important interaction areas have been discovered by X-ray crystallography [³³ ]. This study provides an excellent view of the interaction of a hemopexin domain with a TIMP molecule at the atomic level. The most important cluster of hydrophobic interactions is localized around Met₁₄₉ of the GH loop of TIMP-2. Progelatinase A/MMP-2 recognizes this amino acid by means of Ala₆₁₂, Tyr₆₃₆, Leu₆₃₈, Val₆₄₈, and most significantly, Phe₆₅₀ of the fourth blade of the hemopexin domain. A second hydrophobic interaction is mediated mainly by Phe₁₈₈ of TIMP-2. The negatively charged carboxyterminal end of TIMP-2 is inserted into a cavity at the junction of the third and fourth blade of the hemopexin domain of progelatinase A/MMP-2, consisting of Tyr₅₈₁, Phe₅₈₈, Phe₆₀₂, Ala₆₀₉, and Trp₆₁₀ [³³ ]. According to Overall et al. [⁶⁵ ], the cationic cluster of lysine residues at Positions 576, 578, 579, and 595–597 in the third and fourth blade of the hemopexin domain of gelatinase A/MMP-2 is crucial for interaction with TIMP-2.

In contrast to TIMP-2, TIMP-1 does not contain a negatively charged carboxyterminus. Probably this difference between the two molecules explains the differences in specificity and kinetic properties of both TIMPs [³³ ].

The high affinity of TIMPs for MMPs has important physiological consequences. Normally, tissues contain an excess of TIMPs, originating from plasma or secreted by tissue cells. As TIMPs have only relatively low specificity for different MMPs, these inhibitors can protect several components of the ECM from being degraded [¹⁰³ ]. TIMP-3 binds strongly to components of the ECM and therefore, functions mainly in the area where it has been produced [¹⁰⁴ ]. The relatively high expression of TIMP-4 in the heart protects against cardiomyopathy [¹⁰⁵ ]. In conditions of excess of active MMP molecules relative to TIMP molecules, tissue degradation and several kinds of pathologies might result [⁶⁶ ].

Inhibition of MT6-MMP/MMP-25 by binding of clusterin
Membrane-type MMPs (MT-MMPs) form a subgroup of cell-associated MMPs, which are anchored to the cell membrane by a transmembrane domain and an additional cytoplasmic domain (MT1-, MT2-, MT3-, and MT5-MMP), or by a glycosyl-phosphatidyl inositol (GPI) anchor (MT4- and MT6-MMP) [³⁸ ]. MT6-MMP/MMP-25, also named leukolysin, is expressed almost exclusively by neutrophils, although it is also present on some brain tumor cells [¹⁰⁶ , ¹⁰⁷ ]. Therefore, MT6-MMP/MMP-25 is thought to play an important role in neutrophil function. A recombinant, soluble form of MT6-MMP/MMP-25 is shown to associate with the protein clusterin [⁷⁵ ]. Clusterin is also named Apolipoprotein J, testosterone-repressed prostate message-2, sulfated glycoprotein-2, and SP-40. It is a heterodimer abundantly present in body fluids and binds to many components, but the specific function of clusterin remains unknown [¹⁰⁸ ]. Two different forms of clusterin are described: a glycosylated, secreted form and nonglycosylated, nuclear clusterin, both intracellularly produced by alternative splicing and having distinct biological activities (reviewed in Shannan et al. [¹⁰⁹ ]). The expression of clusterin is induced in various physiological and pathological conditions and by stress [¹⁰⁸ ]. Clusterin binding inhibits the proteolytic activity of MT6-MMP/MMP-25 in a dose-dependent manner. However, clusterin has a lower affinity for MT6-MMP/MMP-25 in comparison with TIMPs. Experiments with deletion mutants of MT6-MMP/MMP-25 show that the hemopexin domain of MT6-MMP/MMP-25 is responsible for binding to clusterin. The exact mechanism by which clusterin binding to the hemopexin domain inhibits proteolytic activity is unknown. Possibly, the binding inhibits substrates to access the active site, clusterin itself also binds to the active site, or a combined effect takes place. A significant amount of MT6-MMP/MMP-25 of human neutrophils seems to form a complex with clusterin. Thus, it is speculated that clusterin binding regulates neutrophil functions in a negative way, preventing excessive tissue destruction [⁷⁵ ]. Clusterin-deficient mice indeed show more tissue damage at inflammatory sites than wild-type animals [¹¹⁰ ]. Clusterin also binds MT4-MMP/MMP-17. However, the binding of clusterin is not a nonspecific interaction with MMPs in general. It is unknown whether the binding of clusterin to MT4-MMP/MMP-17 also occurs through the hemopexin domain of this MMP [⁷⁵ ].

Homodimerization/-multimerization
Homodimerization/-multimerization of gelatinase B/MMP-9
Progelatinase B/MMP-9 exists as monomeric (92 kDa) and homodimeric molecule (200 kDa). This also counts for active gelatinase B/MMP-9 [⁷¹ ]. Monomeric and dimeric gelatinase B/MMP-9 molecules are present in culture media of different gelatinase B/MMP-9-producing cells, in biological fluids and tissues. This observation suggests that both forms are physiologically relevant. Dimerization/multimerization occurs intracellularly [⁶⁷ ]. As only the monomeric form of gelatinase B/MMP-9 is detected under reducing conditions, dimerization/multimerization presumably occurs via the formation of one or several disulfide bridges [⁶⁷ , ⁷¹ ]. Dimerization/multimerization is effected at the carboxyterminal domains of gelatinase B/MMP-9. The cysteine residues at Positions 516 and 704 in the hemopexin domain form a disulfide bridge. Thus, a connection is generated between the first and fourth blade of the propeller structure of the hemopexin domain, and a direct role of these residues in dimerization/multimerization is excluded [⁷¹ , ⁷⁶ ]. Based on studies of site-specific mutagenesis, participation of the cysteine residue at Position 674 also seems excluded [⁷¹ ].

However, gelatinase B/MMP-9 possesses, compared with other MMPs, a unique O-glycosylated domain with a cysteine residue at Position 468. It is interesting that deletion mutants of gelatinase B/MMP-9, which lack the O-glycosylated domain, are only observed as monomers. Thus, the O-glycosylated domain is essential for dimerization/multimerization. The cysteine at Position 468 was presumed to be involved. It is unexpected that dimerization/multimerization still occurs after mutation of this cysteine into an alanine [⁷² ]. Therefore, the exact mechanisms of how dimers/multimers are formed remain obscure. It can be concluded that cysteine residues of gelatinase B/MMP-9 (no dimers/multimers under reducing conditions) as well as the O-glycosylated domain (no dimers/multimers if the O-glycosylated domain is deleted) play important roles in dimerization/multimerization.

Different findings illustrate the potential of (isolated) hemopexin domains of gelatinase B/MMP-9 to form dimerization/multimerization. Cha et al. [⁷⁶ ] showed that hemopexin domains of gelatinase B/MMP-9 can dimerize by noncovalent and mainly hydrophobic interactions at the fourth blade of this domain. Furthermore, an 18 amino acid peptide, which binds selectively to the hemopexin domain of gelatinase B/MMP-9, inhibits dimerization of recombinant hemopexin domains of gelatinase B/MMP-9. TIMP-1 is unable to compete with this peptide for the binding of the hemopexin domain [⁷⁷ ]. In contrast, homodimerization/multimerization of progelatinase B/MMP-9 would prevent the formation of a complex with TIMP-1 and vice versa [⁷¹ ]. This phenomenon could be explained by the presence of overlapping contact areas for TIMP-1 and homodimerization/multimerization in the hemopexin domain [⁷⁶ ].

Homodimerization of other MMPs
MT1-MMP/MMP-14 molecules are present at the cell surface as homodimer structures [⁵¹ ]. To this end, the hemopexin, transmembrane, and cytoplasmic domains are important. By the formation of such oligomeric structures, the activation process of progelatinase A/MMP-2 is facilitated, which is described above [⁵² ]. Moreover, homodimerization of MT1-MMP molecules via their hemopexin domains is essential for cleavage of collagen type I fibers at the cell surface. Blockage of dimerization via hemopexin domains inhibits cell surface collagenolytic activity in a dose-dependent manner [⁷⁸ ]. Furthermore, the crystal structure of procollagenase-1/MMP-1 shows a dimer with contacts via the hemopexin domains. However, as proMMP-1 is present as a monomer in solution, the observed homodimer might be a result of a high, experimental MMP concentration and thus, not physiologically relevant [⁶⁴ ].

Binding and cleavage of substrates
Binding and cleavage of chemokines
Chemokines are chemotactic cytokines produced locally in tissues in response to a stimulus. These small-sized proteins (<10 kDa) are generally classified into four families: the C, CC, CXC, and CX₃C chemokine families. Depending on the presence of an amino acid between the first two aminoterminal cysteine residues, the CXC chemokine family, to which interleukin-8 (IL-8) and stromal-derived factor (SDF)-1 and -1ß belong, is distinguished from the CC-chemokine family with, for instance, monocyte chemotactic protein (MCP)-1, -2, -3, and -4 as important members. The CX₃C and C families contain only one and two chemokines, respectively, and will not be discussed further in this section [¹¹¹ ].

Chemokines form a chemotactic gradient within tissues and thus attract leukocytes out of the bloodstream to places of infection [¹¹² ]. After tissue injury, neutrophil granulocytes are attracted usually, first, under the influence of CXC chemokines. These chemokines will also stimulate the neutrophil granulocytes to degranulate. By exocytosis of secondary and tertiary granules, gelatinase B/MMP-9 and neutrophil collagenase-2/MMP-8, synthesized in advance and stored in granules, are released. These MMPs are important mediators in the early phase of matrix degradation, a main feature of inflammation. As a result of the proteolytic activity of these MMPs, leukocytes can migrate throughout tissues [¹¹³ , ¹¹⁴ ]. Second, under the influence of CC chemokines, chemotaxis, activation, and degranulation of monocytes and macrophages occur, and de novo-synthesized MMPs are released. Stimulated by proinflammatory cytokines produced by the infiltrated cells, stromal cells also secrete various MMPs, which enhance the acute, proteolytic phase of inflammation [¹¹⁵ ]. Gelatinase A/MMP-2 is produced constitutively by stromal cells [¹¹⁶ ]. After a while, the inflammatory infiltrate needs to be cleared. If not, this would result in a continuous process of tissue destruction and repair, such as is observed in chronic inflammation and granulomas.

In these processes, the substrate specificity of MMPs is not restricted to components of the ECM. MMPs also process chemokines. This property influences chemotaxis and promotes healing indirectly. This aspect is illustrated by means of the following examples.

An acute inflammation is mastered by the CXC chemokine IL-8, which attracts neutrophils, the predominant leukocyte type in the peripheral circulation. Some of the CXC chemokines contain in front of the CXC a Glu-Leu-Arg (ELR) motif and are angiogenic, whereas those without ELR are angiostatic [¹¹⁷ , ¹¹⁸ ]. Examples of ELR-positive CXC chemokines are IL-8, granulocyte chemotactic protein-2, and epithelial cell-derived neutrophil-activating peptide 78. These are aminoterminally processed by gelatinase B/MMP-9 [¹¹⁹ , ¹²⁰ ]. Owing to this, the biological activity of IL-8 increases tenfold. The truncated form of IL-8 attracts neutrophil granulocytes more efficiently and stimulates the cells to release, for instance, gelatinase B/MMP-9 [¹¹⁹ ]. This mechanism occurs during early phases of inflammation and ensures the acuteness of neutrophil influx, forming pus (Fig. 3A ). In a recent paper, it was shown that the cleavage of IL-8 by gelatinase B/MMP-9 is equally efficient for a gelatinase B/MMP-9 mutant without hemopexin domain [⁷² ], illustrating that the exosite principle is not applicable for this MMP-chemokine interaction.

View larger version (30K): [in this window] [in a new window]   Figure 3. Interactions between MMPs and chemokines influence leukocyte trafficking. In response to various stimuli, stromal cells produce chemokines. These form a chemotactic gradient, by which leukocytes are recruited from the vasculature toward the stimulated site. The initial chemokine gradients are indicated as pink triangles. (A) A gradient of IL-8 chemoattracts neutrophils out of the blood vessels. These leukocytes are stimulated to release gelatinase B/MMP-9 (represented as green dots), thus allowing migration throughout the ECM. As a result of processing of IL-8 by gelatinase B/MMP-9 (green arrow), the chemotactic potency of IL-8 increases tenfold (white arrow; green triangle). Thus, MMP activity from the attracted neutrophils induces a positive feedback mechanism, enhancing the recruitment of neutrophils out of the blood vessels and the formation of pus at a stimulated place. (B) An initial gradient of MCP-3 directs mononuclear cells to the inflammatory site. The stromal and the attracted cells release MMPs (e.g., gelatinase A/MMP-2, represented as red dots), which allow leukocyte migration throughout the ECM. These MMPs bind, via their hemopexin domains, to MCP-3 molecules, which are processed subsequently (represented as red arrows) into antagonists. The white arrow indicates the resulting, diminished chemotaxis, mediated by hemopexin domain interactions. This panel was modified from McQuibban et al. [115 ]. (C) The function of the hemopexin domain of gelatinase A/MMP-2 in the binding (represented by black arrows) of MCP-3 is illustrated with the crystal structures of both molecules. The gelatinase A/MMP-2 structure [34 ] has been modified by removal of the propeptide and thus, represents a model of the activated enzyme approaching the MCP-3 substrate [121 ]. The N terminus of MCP-3 interacts with the active site of gelatinase A/MMP-2 (colored yellow), and four amino acids (colored yellow) are removed from MCP-3 (represented by the red arrow), thus converting MCP-3 into an antagonist.

Besides MCP-3, a number of other CC chemokines are also cleaved efficiently by MMPs. MCP-1, -2, and -4 are not cleaved by gelatinase A/MMP-2 but instead, by interstitial collagenase-1/MMP-1 and stromelysin-1/MMP-3. It has not yet been determined whether these chemokines also bind to the MMP via an exosite at the hemopexin domain. For MCP-2 and -4, it is known that the shortened forms are also antagonists of chemotaxis, leading to a strong decrease of inflammatory edema in vivo [¹¹⁵ ].

Gelatinase A/MMP-2 and other MMPs likewise remove a tetrapeptide from the aminoterminal end of the CXC chemokine receptor (CXCR)-4 ligands SDF-1 and -1ß. Efficient proteolysis of these chemokines results from the interaction with an exosite on the hemopexin domain, at least partially overlapping with the binding site for MCP-3. SDF-1 shows a higher relative affinity for gelatinase A/MMP-2 than MCP-3. However, upon cleavage of SDF-1 and -1ß, the affinity for the chemokine receptor CXCR-4 diminishes 100-fold. These chemokines lose their chemotactic capacity by proteolysis [⁸² ]. In the brain, cleavage of SDF-1 by HIV-induced gelatinase A/MMP-2 results in a highly neurotoxic protein, which induces neuronal apoptosis and neurodegeneration [¹²² ].

Triple helicase activity of MMPs
The hemopexin domain and linker peptide determine the substrate specificity of collagenolytic MMPs [⁴⁹ ]. The hemopexin domain is indispensable for classical collagenases (collagenase-1, -2, -3; MMP-1, -8, and –13, respectively) to cleave native, triple helical collagens. Without the hemopexin domain, these MMPs do not cleave collagen but retain their proteolytic activity for other substrates [³¹ ].

At present, more than 20 different types of human collagen have been identified. These proteins account for 30% of the total amount of protein in the mammalian body and are the main structural elements of the skin dermis and subcutis and of other connective tissues, including bone and teeth. Depending on structural features and supramolecular organization, collagens can be divided in several subgroups, such as the fibrillar collagens (types I–III, V, and XI). This subgroup represents 90% of the total collagen. Other subgroups will not be discussed in this review.

Fibrillar collagens consist of three polypeptide chains, together forming a triple helix structure. Several of such triple helices associate to form macromolecular fibrils [¹²³ ]. This supramolecular structure not only gives enormous tensile strength but also renders native collagens relatively resistant to proteolysis. However, at least five human MMPs can cleave native fibrillar collagen at the triple helical domain. Thus, collagen types I–III are cleaved by the classical collagenases-1, -2, and -3 (MMP-1, -8, and -13, respectively) and by MT1-MMP/MMP-14, and collagen type I is cleaved by gelatinase A/MMP-2 [⁸³ ]. The MMP hydrolyzes one specific peptide bond at approximately three-fourths of the molecule from the aminoterminal end. This requires an initial binding of a collagen triple helix to the MMP, followed by local unwinding of the structure. The latter step is necessary to allow cleavage of each chain of the triple helix individually, as the entire structure is too large to be accommodated by the active site of MMPs [⁸³ ]. The mechanism of collagenolysis has not been totally unraveled yet. However, several experiments have shown that the flexibility of the substrate as well as features of the catalytic and hemopexin domains and of the linker peptide of the above-mentioned MMPs are important in the process [⁸³ ].

Single catalytic domains of MMPs are unable to cleave triple helical collagen. MMPs require exosites for native collagen cleavage [⁴¹ ]. Collagenolytical MMPs contain at least two exosites for collagen: one in the catalytic domain and one in the hemopexin domain. Overall et al. [¹²⁴ ] suggested that by simultaneous binding of the catalytic and hemopexin domain to collagen, the triple helical structure is oriented and destabilized. After local unwinding of the cleavage site, proteolysis occurs in the triple helix [⁸³ ].

In silico experiments show that the linker peptide of interstitial collagenase-1/MMP-1 can position the hemopexin domain in such a way that it bends over the active site of the catalytic domain. Thus, the collagen would be captured between these two domains. However, the active site cannot accommodate the entire triple helix in a native state. The linker peptide would, by means of its collagen-like conformation, change the quaternary structure of the captured collagen. Interactions between proline residues of the collagenase and a specific region of the collagen would generate a "proline zipper," resulting in destabilization of the cleavage site area of the collagen. After destabilization, one chain of the triple helix fits in the active site of the MMP, and cleavage can occur. The efficiency of the collagenolytic process would then depend on the length of the linker peptide [¹²⁵ ].

Deletion of the hemopexin domain of classical collagenases (Residues 263–470 of interstitial collagenase-1/MMP-1, 263–487 of neutrophil collagenase-2/MMP-8, and 259–471 of collagenase-3/MMP-13, respectively) results in a loss of collagenolytic activity. A mutant form of interstitial collagenase-1/MMP-1, which only contains the first 242 amino acids of the proproteinase, can only cleave collagen type I if incubated with a mutant form of procollagenase-1/MMP-1, of which the first 199 aminoterminal residues are deleted. As each mentioned mutant in se cannot cleave native, triple-helical collagen, it has been suggested that the mutant form of interstitial collagenase-1/MMP-1 with the intact hemopexin domain orients the triple helical structure and helps to unwind it [⁸³ ]. The hemopexin domain of collagenase-1/MMP-1 is shown to bind to collagen type I [⁷⁴ ]. Furthermore, collagen-binding experiments indicated that procollagenase-3/MMP-13 and active collagenase-3/MMP-13, both without hemopexin domain, cannot bind to native collagen type I. Human interstitial collagenase-1/MMP-1 and collagenase-2/MMP-8 bind collagen much more strongly as active proteinases than as proforms [⁷³ , ⁷⁴ ].

Replacement of the second, third, and fourth blade of the hemopexin domain of neutrophil collagenase-2/MMP-8 by the corresponding blades of stromelysin-1/MMP-3 results in a chimeric molecule with only 16% of the collagenolytic activity of the intact collagenase. Therefore, Hirose et al. [⁸⁴ ] concluded that each of the four blades of the hemopexin domain of neutrophil collagenase-2/MMP-8 has characteristics that are important for efficient collagenolysis. When the linker peptide of collagenase-2 (16 amino acids) is replaced by the linker of MMP-3 (25 residues), the MMP also loses its collagenolytic capacity.

Active collagenases are unstable. As a result of autoproteolysis, the hemopexin domain and linker peptide of the intact MMP are released. After this, the specific collagenolytic activity toward collagen is lost. This finding also demonstrates that collagenolysis is comediated by carboxyterminal domains of the MMP and illustrates that enzymatic, negative feedback is an important control mechanism [⁷³ ].

Also, MT1-MMP/MMP-14 has collagenolytic capacity. The hemopexin domain and linker peptide are likewise important. An autoproteolytically formed, cell surface-bound form of MT1-MMP/MMP-14, comprising the linker and hemopexin domains, retains the ability to bind collagen and can suppress collagenolysis. The presence of an exosite for native collagen type I at the hemopexin domain is supposed, whereas the linker peptide cannot bind collagen [⁸⁵ ]. Recently, a similar recombinant fragment is shown to perturb the secondary structure of native collagen type I upon binding [⁸⁶ ]. The hemopexin domain of MT1-MMP/MMP-14 has been shown to be essential for invasion and growth of cells in three-dimensional collagen type I [⁵⁴ ]. In MT1-MMP/MMP-14 knockout mice, the formation of skeletal tissues is disturbed, and after birth, excessive fibrosis is apparent. Moreover, fibroblasts derived from these mice cannot degrade collagen type I [¹²⁶ ].

Gelatinase A/MMP-2 also cleaves collagen type I to three-fourths and one-fourth fragments. The presence of the catalytic and hemopexin domains of gelatinase A/MMP-2 is absolutely required for collagen cleavage, and the gelatin-binding fibronectin domain probably also participates [⁸⁶ , ⁸⁷ ]. However, the isolated hemopexin domain is shown not to bind native collagens, which suggests that the interactions of gelatinase A/MMP-2 with collagens occur in a context of an entire proteinase structure [¹²⁷ ]. Furthermore, prostromelysin-1/MMP-3 and the active MMP bind to collagen type I (cII) through the hemopexin domain. However, stromelysin-1/MMP-3 is unable to perform collagenolysis [⁷⁹ ].

Collagen type II is an important constituent of cartilage. In RA, it is degraded by the consecutive actions of collagenases and gelatinase B/MMP-9, leading to the release of autoimmunogenic epitopes [¹²⁸ ]. Whereas the hemopexin domain of collagenases is essential for the first cleavage into the three-fourth and one-fourth fragments (3/4 cIIn), recent experiments indicate that the proteolysis of 3/4 cIIn occurs equally efficiently by mutants of gelatinase B/MMP-9, which lack the hemopexin domain. Similarly, the digestion of fully denatured collagen (gelatin) is not altered significantly by the deletion of the hemopexin domain [⁷² ].

Binding and cleavage of a receptor of complement factor C1q by MT1-MMP/MMP-14
The 33-kDa receptor gC1qR, also known as p32, p32/TAP, and p33, has affinity for the globular heads of the subcomponent C1q of the human classical complement pathway. gC1qR is present in mature form in the cytosol and nucleus and on the membrane of many cells, including leukocytes [¹²⁹ ]. The precursor localizes in the mitochondria. The gC1qR protein is a chaperone-like regulator, thought to be involved in mitochondrial oxidative phosphorylation and in nucleus-mitochondrion interactions [¹³⁰ ]. Moreover, proliferating cells secrete soluble gC1qR into the surrounding milieu. The mature protein interacts with a variety of ligands, including cell surface receptors and several viral and bacterial proteins. Soluble gC1qR molecules inhibit, for example, the hemolytic activity of the complement system and regulate complement activation [¹³¹ ]. By releasing gC1qR, proliferating cells may evade destruction by the complement system. MT1-MMP/MMP-14 forms an enzyme-substrate complex with the released gC1qR molecules. In in vitro assays with pure proteins and in cell culture conditions, a gC1qR carboxyterminal fragment of 17 kDa and two minor fragments of 12 and 11 kDa are generated by active MT1-MMP/MMP-14. Biological effects of these cleavages remain unknown, but these might represent feedback regulation. At least two binding sites for gC1qR have been found on the MT1-MMP/MMP-14 protein: one on the cytoplasmic tail and one comprising the catalytic domain of MT1-MMP/MMP-14 [⁸⁸ , ¹³² ]. Moreover, a mutant form of MT1-MMP/MMP-14, which lacks the hemopexin domain, binds less efficiently to gC1qR and is incapable of cleaving the protein. Probably, the hemopexin domain also binds to gC1qR, via the loop connecting the ß₃ strand to the ß₄ strand of gC1qR. In this manner, the loop sequence could achieve the proper orientation relative to the active site of the proteinase to support cleavage of this loop sequence of gC1qR, generating the 17-kDa fragment [⁸⁸ ].

Binding of the IGFBP-3 by stromelysin-4/MMP-19
IGF-I and -II promote growth and survival of many cell types. The bioavailability of IGFs is modulated by binding to IGFBPs, of which six high-affinity human variants (IGFBP-1 to -6) are described. Various proteinases, such as different MMPs, proteolyze IGFBPs, thus mediating the bioavailability and bioactivity of IGFs [¹³³ ].

Recently, it is shown that the hemopexin domain of stromelysin-4/MMP-19, also named RASI, binds IGFBP-3 [⁸⁰ ], which is cleaved by stromelysin-4/MMP-19 and results in the release of IGF-1, a growth factor for keratinocytes [¹³³ ]. Thus, stromelysin-4/MMP-19 promotes proliferation of keratinocytes, as well as migration and adhesion of these cells on collagen type I [¹³⁴ ]. Generally, the expression of MMPs in the skin only becomes significant under pathological conditions, such as inflammation. However, stromelysin-4/MMP-19 is also expressed in healthy epidermal skin. It is remarkable that its expression is altered under pathological conditions, which are accompanied with increased epidermal proliferation, such as psoriasis [¹³⁵ ]. It is reasonable to assume that the proteolysis of IGFBP-3 by stromelysin-4/MMP-19 contributes to this disease with an inflammatory component.

Stromelysin-4/MMP-19 is not the only MMP that catalyzes IGFBP-3. This substrate is also cleaved by interstitial collagenase/MMP-1, stromelysin-1/MMP-3, gelatinase A/MMP-2, and possibly also gelatinase B/MMP-9 [^{136 137 138} ]. Also, other IGFBPs are substrates for MMPs. IGFBP-1 is cleaved by stromelysin-1/MMP-3, stromelysin-3/MMP-11, gelatinase A/MMP-2, and gelatinase B/MMP-9 [¹³⁹ , ¹⁴⁰ ]. In addition, metalloelastase/MMP-12, also named macrophage elastase, and gelatinase B/MMP-9 process IGFBP-6 [¹⁴¹ ]. At present, it is unknown whether the IGFBPs are also bound by the hemopexin domains of these different MMPs. It is remarkable that MT-1/MMP-7, which does not contain a hemopexin domain, cleaves all six IGFBPs [¹³⁷ , ¹³⁹ , ¹⁴² , ¹⁴³ ]. Thus, more research is necessary to elucidate the role of binding to hemopexin domains in cleavage of IGFBPs.

Binding and cleavage of progelatinase B/MMP-9 and chondroitin sulfate PRGs
PRGs are found in the ECM of all tissues and consist of glycoproteins bearing covalently bound, sulfated, negatively charged glycosaminoglycan (GAG) side-chains. PRGs differ from each other by their core proteins, length, and type of GAG side-chain, which consist mostly of chondroitin sulfate or heparin/heparan sulfate [¹⁴⁴ ]. Chondroitin sulphate proteoglycans (CSPGs) are produced by human macrophages. Upon stimulation, the expression and secretion of CSPGs increase [¹⁴⁵ ]. Until now, the biological functions of this phenomenon are unknown. However, it is shown that other secretion products can bind to CSPGs [¹⁴⁶ ]. Monomeric and dimeric progelatinase B/MMP-9 can, by forming one or several disulfide bridges, covalently bind to the core protein of CSPGs. Winberg et al. [¹⁴⁷ ] verified that the complex formation was not an experimental artifact, but the result of a cellular process. Addition of CaCl₂ to the progelatinase B/MMP-9:CSPG complex induces autocatalytic activation of progelatinase B/MMP-9. Moreover, the active MMP is released following a carboxyterminal cleavage of gelatinase B/MMP-9 and cleavage of the core protein of the CSPG. The truncation process is likely to be an intramolecular reaction, as the complex cannot process added progelatinase B/MMP-9. The released gelatinase B/MMP-9 molecules lack at least a part of the hemopexin domain, indicating that this domain participates in the covalent binding to CSPGs [⁸⁹ ]. Of the total amount of progelatinase B/MMP-9 secreted by macrophage THP-1 cells, 10–15% is covalently linked to CSPGs [¹⁴⁷ ]. As CSPGs bind various molecules, CSPGs are suggested to serve as carriers, directing gelatinase B/MMP-9 more efficiently to particular targets or to other sites than usually targeted. Thus, upon cleavage of the hemopexin domain, gelatinase B/MMP-9 is released from the CSPG into the surrounding tissue, where it could cleave molecules at a distance from its original attachment site [⁸⁹ ].

Cleavage of fibrinogen by gelatinase A/MMP-2
During blood coagulation, thrombin converts fibrinogen, a glycoprotein composed of six chains, into fibrin monomers, which associate into clots. It has been shown that gelatinase A/MMP-2 cleaves fibrinogen, resulting in products with less clotting activity [¹⁴⁸ , ¹⁴⁹ ]. Gelatinase A/MMP-2 interacts through its hemopexin domain with fibrinogen, as the fragmentation of fibrinogen by gelatinase A/MMP-2 is inhibited significantly by addition of the isolated hemopexin domain. Fragmentation of fibrinogen occurs at a 20-fold lower rate by a gelatinase A/MMP-2 variant without hemopexin domain than by the intact proteinase, and an additional fibrinogen fragment, which is not observed upon cleavage by full-length gelatinase A/MMP-2, is formed. These findings implicate a role for the hemopexin domain of gelatinase A/MMP-2 in (correct) recognition of fibrinogen [¹⁴⁹ ]. Also, the catalytic domains of neutrophil collagenase-2/MMP-8, metalloelastase/MMP-12, collagenase-3/MMP-13, and MT1-MMP/MMP-14 are shown to cleave fibrinogen [¹⁵⁰ ]. However, more research is necessary to unravel the role of the hemopexin domains of these other MMPs in fibrinogen proteolysis.

Attachment of MMPs to the cell surface
Localization of MMPs at cell surfaces can facilitate their activation (vide supra) and allow efficient pericellular proteolysis, both essential functions in leukocyte biology. Besides MT-MMPs, other MMPs can also be attached to the cell surface, albeit by using other mechanisms. For a number of MMPs the hemopexin domain is crucial. However, the MMP domain responsible for the interaction with docking molecules has not always been characterized. A number of (cell surface) binding reactions for which it is known that the hemopexin domain of MMPs is involved, will be described in this section.

Binding of MMPs to integrins
In case of an intact epidermal layer of the human skin, keratinocytes with ₂ß₁-integrins at the cell surface are present above the basement membrane [¹⁵¹ ]. These proteins are also cell surface receptors mediating adhesion between endothelial cells and the ECM. Integrins are produced constitutively by keratinocytes, forming heterodimeric receptors for collagen type I [¹⁵² ], which, however, is absent in this area under normal conditons. Following epidermal damage, the mentioned keratinocytes will move from the edge of the wound to the underlying dermis. In this skin layer, collagen type I is abundant. As a result of interaction of keratinocytes and dermal collagen, a site-specific signal is generated. This signal, together with other processes will start wound healing. By binding of collagen type I and ₂ß₁-integrins, expression and secretion of interstitial collagenase-1/MMP-1 by keratinocytes are induced [¹⁵³ ]. Next, pro- and active interstitial collagenase-1/MMP-1 can bind to the I-domain of the ₂-subunit of ₂ß₁-integrin at the cell surface. This binding is partially dependent on divalent cations.

Experiments using chimeric MMP molecules reveal that the linker peptide as well as the hemopexin domain of interstitial collagenase-1/MMP-1 participate in binding to ₂ß₁-integrin [⁹⁰ ]. Each of both domains in se mediates binding, although is not sufficient for optimal binding. Probably, the physiological ligand is restricted to the proform of interstitial collagenase-1/MMP-1 [¹⁵⁴ ]. As a result of the proteolytic activity of interstitial collagenase-1/MMP-1 on collagen type I, which is also bound to ₂ß₁-integrin, keratinocytes migrate along dermal, collageneous matrices and participate in wound healing [¹⁵³ , ¹⁵⁴ ].

Brooks et al. [⁹¹ ] observed that the hemopexin domain of gelatinase A/MMP-2 participates in binding to _Vß₃-integrin. Active gelatinase A/MMP-2, via the hemopexin domain, binds directly to _Vß₃-integrin at the surface of invasive cells and angiogenic blood vessels, thereby enhancing tumor cell growth. As gelatinase A/MMP-2 was not expressed by these cells, this MMP is probably produced by surrounding cells. A mutant form of gelatinase A/MMP-2, lacking the hemopexin domain, did not bind to the surface of cells expressing _Vß₃-integrin. The interaction of a recombinant form of the hemopexin domain of gelatinase A/MMP-2 with _Vß₃-integrin prevents binding of gelatinase A/MMP-2 and this integrin. In addition, Nisato et al. [⁹² ] observed that the in vitro experimental conditions are determining whether the hemopexin domain of gelatinase A/MMP-2 binds _Vß₃-integrin and affects angiogenesis.

The hemopexin domain of gelatinase A/MMP-2 seems to be present in tumor tissue as a natural breakdown product of this MMP. The presence of this product depends on the expression of _Vß₃-integrin. The hemopexin domain of gelatinase A/MMP-2 inhibits in vivo angiogenesis and tumor cell growth. The concentration of the single hemopexin domain increases during the invasive stage of neovascularisation to reach maximal concentrations during the maturation stage of the formed blood vessels. Thus, the hemopexin domain could be an important negative regulator of cell-bound proteinase activity during angiogenesis and vasculogenesis [¹⁵⁵ ].

Most ligand molecules of integrins contain an arginine-glycine-aspartic acid recognition sequence. As gelatinase A/MMP-2 molecules do not have such a tripeptide sequence, the MMP must be recognized by _Vß₃-integrin in another way than most ligands of this integrin [¹⁵¹ ]. At the moment, it is uncertain whether gelatinase A/MMP-2 only binds to _Vß₃-integrin via the hemopexin domain.

Progelatinase A/MMP-2 (72 kDa) is converted into an intermediate form of 68 kDa and subsequently, by autocatalysis, to an active form of 66 kDa [⁴⁸ ]. The _Vß₃-integrin could, by binding gelatinase A/MMP-2, participate in membrane-bound activation of the MMP (vide supra). Puyraimond et al. [¹⁵⁶ ] detected colocalization of the gelatinase A/MMP-2:MT1-MMP/MMP-14:TIMP-2 complex with _Vß₃-integrin in caveolae. Possibly, binding of gelatinase A/MMP-2 to MT-MMPs results in partial activation, allowing binding to _Vß₃-integrin. However, this remains uncertain at this moment.

Furthermore, gelatinase B/MMP-9 binds specifically to a protein fragment comprising the integrin epidermal growth factor (I-EGF)-like Domains 2 and 3 of ß₅ integrins. Björklund et al. [⁷⁷ ] constructed an inhibitor peptide that bears similarity to the EGF-like Domain 2. This inhibitor binds selectively to the hemopexin domain of gelatinase B/MMP-9. This binding cannot be competed by TIMP-1. The peptide does not directly inhibit the proteolytic activity of gelatinase B/MMP-9, although it prevents the activation of the proMMP in cell culture medium. Moreover, the peptide prevents in vitro cell migration and in vivo tumor cell growth. The hemopexin domain of gelatinase B/MMP-9 inhibits invasion with a similar potency as the protein fragment comprising the I-EGF-like Domains 2 and 3. Progelatinase B/MMP-9 is shown to associate with ₅ and ß₅ integrins, indicating that ₅ß₁ and _Vß₅ are the major integrins involved in the binding of progelatinase B/MMP-9 by cells. The hemopexin domain of gelatinase B/MMP-9 binds to cells that are transfected to express _Vß₅ but not to untransfected, control cells. Gelatinase B/MMP-9 and ß₅-integrin colocalize in the leading edge of invasive cells.

The peptide mentioned interferes specifically with integrin-mediated interactions without affecting the proteolytic activity of gelatinase B/MMP-9. This finding indicates that specific drugs can be developed that selectively prevent specific functions of MMPs mediated by interactions of hemopexin domains [⁷⁷ ].

Binding of MT1-MMP/MMP-14 to CD44H
The membrane-bound glycoprotein CD44 participates in many cell-cell and cell-matrix interactions. For instance, CD44 forms an important receptor for hyaluronic acid, a GAG that is present abundantly in the ECM. CD44 contains a ligand-binding domain that is linked to the cell surface through a stem sequence and