Physiology and pathophysiology of matrix metalloproteases.

Matrix metalloproteases or matrixins (see Table  1 ) are a subfamily of metalloproteases, which consist of 23 distinct proteases in humans (24 in mouse). The first MMP was identified in 1962 as the protease responsible for the degradation of fibrillar collagen in tadpole tails during metamorphosis and was thus dubbed interstitial collagenase (Gross and Lapiere 1962 ). After identification of a similar collagenase in human skin, this protease was renamed MMP-1. MMPs have since been identified as the major enzymes responsible for turnover of extracellular matrix by proteolytic degradation of virtually all proteinaceous components of the ECM (Woessner 1991 ). Table 1 Overview of the 23 identified human matrix metalloproteases and their common names MMP Alternative name MMP Alternative name 1 Collagenase-1 16 Membrane type-3 MMP Interstitial collagenase 2 Gelatinase A 17 Membrane type-4 MMP 72-kDa type IV collagenase 3 Stromelysin-1 19 None Transin-1 Human ortholog of Xenopus MMP-18 7 Matrilysin 20 Enamelysin Pump-1 8 Collagenase-2 21 None Neutrophil collagenase Human ortholog of Xenopus xMMP 9 Gelatinase B 23 Cysteine array MMP 92-kDa type IV collagenase Femalysin MMP-22 10 Stromelysin-2 24 Membrane type-5 MMP 11 Stromelysin-3 25 Membrane type-6 MMP Leukolysin 12 Macrophage metallo-elastase 26 Matrilysin-2/endometase 13 Collagenase-3 27 None 14 Membrane type-1 MMP 28 Epilysin 15 Membrane type-2 MMP Overview of the 23 identified human matrix metalloproteases and their common names MMPs are largely excreted proteins with several conserved domains (see Fig.  1 ). All MMPs contain the catalytic domain, which is shielded off in the inactive form of the enzyme by the prodomain. This propeptide interacts with the catalytic region through a conserved cysteine residue and the Zn 2+ ion in the catalytic pocket (the so-called cysteine switch) (Nagase 1997 ; Van Wart and Birkedal-Hansen 1990 ). Except for MMP-7, MMP-23 and MMP-26, all MMPs contain a C-terminal hemopexin-like domain that functions primarily as a recognition sequence for the substrate (Murphy and Knauper 1997 ). Although MMPs retain catalytic activity toward a wide range of substrates when missing this domain, the hemopexin domain, which is structured like a four-bladed propeller structure with each blade consisting of four antiparallel β-sheets and one α-helix, is an absolute necessity for the degradation of triple-helical collagens (Bode 1995 ). The gelatinases (MMP-2 and -9) further contain a series of three fibronectin type II inserts in the catalytic domain, which facilitate binding of gelatine and collagen (Bode et al. 1999 ). Fig. 1 Schematic representation of the domain structure of metzincin proteases. A signal peptide, B prodomain, C catalytic domain, D hemopexin-like domain, E fibronectin type II insert, F transmembrane domain, G cytoplasmic tail, H disintegrin domain, I cysteine-rich domain, J EGF-like domain, K thrombospondin type I-like repeat, L spacer region Schematic representation of the domain structure of metzincin proteases. A signal peptide, B prodomain, C catalytic domain, D hemopexin-like domain, E fibronectin type II insert, F transmembrane domain, G cytoplasmic tail, H disintegrin domain, I cysteine-rich domain, J EGF-like domain, K thrombospondin type I-like repeat, L spacer region MMP function is regulated at several levels. Firstly, induction of gene expression is controlled by number growth factors and cytokines, and may be suppressed by transforming growth factor β and glucocorticoids (Nagase and Woessner 1999 ; Vincenti 2001 ). Recent insights indicate an important modulatory role of epigenetic processes in the expression of MMPs (Chernov et al. 2009 ). Besides soluble factors, MMP expression may also be regulated by cell–cell contact or interaction of cells with ECM components such as EMMPRIN (extracellular matrix metalloprotease inducer or CD147) (Biswas et al. 1995 ). The expressed MMPs are largely excreted as inactive proenzymes with the propeptide effectively limiting entrance into and catalysis of a substrate in the catalytic pocket by blocking the catalytic zinc (II) ion via the cysteine switch mechanism. Activation of proMMPs can occur through several mechanisms [reviewed in Ra and Parks ( 2007 )], all of which lead to disruption of the cysteine switch. Perhaps, the most important mechanism is proteolytic removal of the prodomain by other endopeptidases such as furin (Pei and Weiss 1995 ). Removal of the prodomain of MMPs, which contains a furin-like proprotein convertase recognition site (RRKR or RxKR), has been described for nine MMPs including all membrane-type MMPs. Alternatively, the prodomain can be proteolytically removed by plasmin and other serine proteases, or even other MMPs. This mechanism is well described for MMP-2 where proMMP-2 binds the endogenous MMP inhibitor TIMP-2 (tissue inhibitor of metalloproteases 2). This complex in turn functions as a ligand for the membrane-bound MMP-14 (or membrane-type 1 MMP) leading to activation of MMP-2 (Cao et al. 1996 ). The cysteine switch may also be broken by chemical reactions, either physiologically by oxidation of cysteine by reactive oxygen species, or artificially by mercury-containing compounds such as 4-aminophenylmercuric acetate (APMA) or denaturing surfactants such as sodium dodecyl sulfate (SDS). This disruption of the thiol–zinc interaction leads to allosteric relocation of the prodomain leading to active forms of the enzyme with the propeptide still attached or to autoproteolytic removal of the relocated prodomain. MMPs are inhibited by the general protease inhibitor α 2 -macroglobulin and a small family of natural inhibitors specifically geared toward inhibiting metalloprotease activity. These TIMPs are a group of four proteins (21–30 kDa in size) that as a group effectively inhibit all MMPs in vivo (Gomez et al. 1997 ). Excreted MMPs are generally classified according to their substrate specificity, leading to four classes: the collagenases (MMP-1, -8 and -13), the gelatinases (MMP-2 and -9), the stromelysins (MMP-3, -10 and -11) and a heterogeneous group containing matrilysin (MMP-7), metallo-elastase (MMP-12), enamelysin (MMP-20), endometase (MMP-26) and epilysin (MMP-28). In this nomenclature, the membrane-anchored MMPs (MMP-14, -15, -16, -17, -24 and -25) are considered as a separate class. An alternate classification arranges the MMPs according to their domain structure (Sternlicht and Werb 2001 ).

MMP-12 was first described in 1981 (Banda and Werb 1981 ) as murine metalloelastase and later identified as a member of the MMP family (Shapiro et al. 1992 ). In 1993, a human ortholog was found (Shapiro et al. 1993 ). MMP-12 was identified as an elastolytic metalloprotease produced by alveolar macrophages, which led to the trivial name (murine) macrophage metalloelastase (MME) or human macrophage elastase (HME). MMP-12 is expressed as a 54-kDa inactive proenzyme and is activated to a 45-kDa active enzyme by removal of the propeptide sequence. The mature enzyme can be further truncated to a 22-kDa active form on C-terminal processing mediated by serine proteases or through autocatalytic cleavage (Shapiro et al. 1993 ). Autocatalytic removal of the C-terminal domain is possible in many other MMPs, but occurs very slowly in contrast to MMP-12, which is readily processed to the smaller form. Expression of MMP-12 is limited to macrophages and is not observed in blood monocytes. As the name reveals, a major substrate for MMP-12 is elastin, but MMP-12 is capable of degrading other ECM constituents (but not gelatin) (Banda et al. 1983 ) and many non-matrix proteins in vitro (Chandler et al. 1996 ). MMP-12 null mice show normal development in the absence of inflammatory stress, but litter size is smaller, presumably due to abnormalities of the placenta during gestation. Macrophages obtained from knockout mice retain only a small fraction of their elastolytic activity, indicating that MMP-12 is indeed the most important elastin-degrading enzyme (in mice). MMP-12 is necessary for the penetration of macrophages through the basement membrane, as demonstrated by a complete inhibition of macrophage migration from MMP-12 null mice both in vitro and in vivo (Shipley et al. 1996 ). MMP-12 seems to play an intriguing role in cancer, which is different from other MMPs. MMP-12 is the primary protease responsible for proteolytic liberation of angiostatin from plasminogen. Angiostatin is a 38-kDa protein with anti-angiogenic properties due to the selective inhibition of endothelial cell proliferation. MMP-2 and -9 are also capable of degradation of plasminogen in vitro, but were found to make only a minor or no contribution to the production of angiostatin in an animal model (Dong et al. 1997 ; Houghton et al. 2006 ). This is another example of a protective effect of MMP activity in tumor genesis, an indication that knowledge of the actual biochemical mechanism of the involvement of a protease in disease is indispensable and cannot be replaced by mere association of expression levels in tumors. Since human MMP-12 was first cloned from alveolar macrophages, it is not surprising that the role of MMP-12 in lung disease has been extensively studied. The involvement of MMP-12 in fibrotic processes in the lung is well established, but studies concerning the role of MMP-12 in emphysema are diffuse due to species differences between murine models and humans. In humans, it was reported that MMP-12 is not upregulated in macrophages of patients with emphysema compared to controls, but rather that macrophage-derived MMP-9 and MMP-1 are upregulated (Finlay et al. 1997 ). MMP-12, on the other hand, was shown to be an absolute requirement for the development of emphysema after cigarette smoke exposure in a knockout animal model (Hautamaki et al. 1997 ; Matute-Bello et al. 2007 , 2008 ). The interspecies difference may have caused bias with respect to the importance of MMP-12 since, although the failure of MMP-12 null mice to develop emphysema upon smoke exposure makes a more compelling point than determination of expression levels, MMP-12 seems to be the major MMP in murine macrophages, while human macrophages produce several MMPs (Parks and Shapiro 2001 ). Evidence for a role of MMP-12 in lung function is based on a polymorphism in the MMP-12 gene that has been described to have predictive value for lung function decline in COPD (Joos et al. 2002 ), a finding that was recently corroborated in a large association study between a single nucleotide polymorphism (SNP) in the MMP-12 promotor, which decreases MMP-12 expression, and a protective effect against lung function decline in asthma and COPD patients (Hunninghake et al. 2009 ). Elevated MMP-12 levels have been measured in induced sputum of COPD patients (Demedts et al. 2006 ), while a recent study has found only a slight increase of MMP-12 in stage 0 of the disease (Ilumets et al. 2007 ). Proteolytic fragments of elastin have been implicated as chemotactic factors in macrophage recruitment (Senior et al. 1980 ), giving a possible explanation for an early role of MMP-12 in emphysema development, with severe tissue destruction at later stages being mainly caused by other macrophage-derived proteases.

Matrix metalloproteases (MMPs) form a large family of enzymes with broad substrate specificity. While their initial role was mainly ascribed to turnover of the extracellular matrix, this view must likely be revised based on recent, more comprehensive studies. MMPs are involved in a delicate balance between proteolytic and anti-proteolytic activity to confine them in time and space. When this balance is disturbed, many disease phenotypes are observed ranging from tissue destruction in chronic inflammatory conditions, such as rheumatoid arthritis and COPD, to cancer metastasis and neurological disorders. The fact that MMPs degrade proteins into fragments with often new, completely different biological activities, supports the view of protein species.

The obvious relation of gelatinases to tumor metastasis and angiogenesis has led to a plethora of research papers on the role of MMP-2 and -9 in diverse malignant processes. This hypothesis was first affirmed by the observation that MMP-2 knockout mice show decreased tumor angiogenesis and progression (Itoh et al. 1998 ). Since then, MMPs have been identified as important players in angiogenesis, growth and metastasis of tumors. The development of a new vascular system is necessary for tumor growth, since without new blood vessels the size of a tumor will be restricted. Gelatinases are primarily involved in this process by enabling proteolytic degradation of the vascular basal membrane, opening the way for migration of endothelial cells to form new blood vessels (Risau 1997 ). MMP-2 is further capable of cleaving laminin-5, which after degradation yields a cryptic site that increases endothelial cell migration (Risau 1997 ) and the release of VEGF that stimulates angiogenesis, not only under physiological conditions but also during tumor development. Tumor growth can be stimulated by gelatinase activity, since MMP-2 and -9 have been shown to release growth factors (Levi et al. 1996 ). Tumor metastasis is a process that involves the release of single tumor cells, migration of these cells to a blood vessel, penetration into the blood stream or lymph system and finally adhesion to vessel endothelium and extravasation into the tissue at the metastatic location. The ECM degrading properties of gelatinases are crucial in both exit of the metastatic cells from the bulk tumor as well as their entrance at the site of metastasis. Increased gelatinase expression and activity has been described in hundreds of publications related to malignant diseases ranging from breast cancer (Somiari et al. 2006 ), urogenital cancers (Maatta et al. 2007 ; Sier et al. 2000 ; Takemura et al. 1992 ), brain tumors (Forsyth et al. 1999 ), lung cancer (Nawrocki et al. 1997 ), skin cancer (Pyke et al. 1992 ), colorectal cancer (Murnane et al. 2009 ) and many more. Interestingly, many authors have found a positive correlation between gelatinase expression or activity and the invasive potential of the tumor, again stressing the crucial role that MMP-2 and -9 play in metastasis. A comprehensive review of the literature on the role of gelatinases in individual cancers is beyond the scope of this article, but excellent reviews are available (Van den Steen et al. 2002 ). Besides in pulmonology and oncology, gelatinase activity is under investigation in several other research fields. MMP-2 has, for example, been identified as a possible target in cardiovascular disease, since it was identified as the protease responsible for degradation of the vasodilator peptide andrenomedullin, with one of the resulting fragment peptides having vasoconstrictive properties, possibly leading to hypertension (Martinez et al. 2004 ).

Metzincins are a ubiquitously expressed family of multi-domain zinc (II)-dependent endopeptidases (Stocker and Bode 1995 ), the members of which include well-known metalloproteases such as the matrix metalloproteases (MMPs) (Nagase and Woessner 1999 ), a disintegrin and metalloproteases (ADAMs) (White 2003 ), the ADAMs with a thrombospondin motif (ADAMTS) (Tang 2001 ), the bacterial serralysins (Nakahama et al. 1986 ) and proteases such as the astacins (including the meprins) (Bode et al. 1992 ; Butler et al. 1987 ). This superfamily of proteases is defined by the presence of a Zn 2+ ion at the catalytic center, which is coordinated by three histidine residues in the zinc-binding consensus sequence HExxHxxGxxH that is present in all proteolytically active metzincins, and a characteristic, strictly conserved methionine-containing tight 1,4 beta turn forming a hydrophobic cleft for the catalytic zinc ion (Bode et al. 1993 ). Catalysis of protein substrates is (most probably) carried out via a general base mechanism involving activation of a zinc-bound water molecule by the carboxylate group of the conserved glutamate residue in the catalytic pocket followed by attack of water on the polarized carbonyl group in the substrate’s scissile bond (Browner et al. 1995 ). The main physiological function of these proteases was originally ascribed to the modulation and regulation of extracellular matrix (ECM) turnover by direct proteolytic degradation of the ECM proteins (e.g., collagen, proteoglycans and fibronectin) (Woessner 1991 ). Another important function is the liberation of biologically active proteins such as cytokines, growth factors and chemokines from their membrane-anchored proforms (so-called shedding). More comprehensive “degradomics” approaches indicate, however, that this view is likely too limited (Butler and Overall 2009 ). MMPs thus contribute to the generation of protein species with vastly differing activities from a single, original gene product.

In addition to the soluble matrix metalloproteases, a small group of membrane-anchored MMPs has been described. The first member of this subfamily, MMP-14 or membrane type-1 MMP (MT1-MMP), was discovered in 1994 (Sato et al. 1994 ) and cloning experiments have since revealed the existence of five additional MT-MMPs. The domain structure of MMP-14 is very similar to that of soluble MMPs (see Fig.  1 ) with the characteristic zinc-binding catalytic domain, the prodomain in the inactive proenzyme and a linked hemopexin-like domain. Most MT-MMPs are membrane anchored by a single-pass transmembrane domain followed by an intracellular cytoplasmic tail that contains three putative phosphorylation sites and is presumed to be important for localizing the enzyme on the cell surface (Lehti et al. 2000 ). All MT-MMPs described contain the furin-like recognition site in their prodomain, allowing activation of the proenzyme by proteolytic removal of this domain by furin and other proprotein convertases (Yana and Weiss 2000 ). MMP-14 is present at the cell surface as a 55–60 kDa active protease, but may be processed by autocatalysis into smaller species of 45 kDa by proteolytic removal of the catalytic domain. This truncated form, which still contains the hemopexin-like domain, is assumed to play a role in autoregulation of MMP-14 catalytic activity (Toth et al. 2002 ). The substrate specificity of MMP-14 is well described in literature [reviewed in Barbolina and Stack ( 2008 )]. The enzyme is capable of proteolytic degradation of type I, II and III collagens following the characteristic cleavage pathway used by collagenases (d’Ortho et al. 1997 ), a finding that was corroborated by knockout experiments that confirmed the role of MMP-14 as an important interstitial collagenase. MMP-14 null mice die within 3 weeks, showing severe developmental abnormalities related to deficiencies in ECM processing, such as dwarfism, skeletal dysplasia and defective vascularization (Holmbeck et al. 1999 ; Zhou et al. 2000 ). Expression of MMP-14 may be regulated by the interaction of migrating cells with a three-dimensional collagen matrix, which triggers clustering of cell surface integrins and lead to transcriptional activation of the MMP-14 promotor (Ellerbroek et al. 2001 ). Besides collagen, MMP-14 is capable of degradation of many other ECM components, such as fibronectin, vitronectin, tenascin, nidogen, aggrecan, fibrin, fibrinogen and laminin-5, leading to a possible stimulation of cellular migration as described above (Barbolina and Stack 2008 ). Although the major proteolytic function of MMP-14 lies in the cleavage of extracellular substrates, some studies indicate a role in intracellular proteolysis after incorporation and accumulation of active MMP-14 in the centrosomal compartment, where it could contribute to development of mitotic spindle changes by degradation of pericentrin (Golubkov et al. 2005a , b ). This mechanism could give rise to a role for MMP-14 in malignant transformation of cells. Intracellular protein–protein interaction between MMP-14 and adenosine nucleotide translocator has been described and is independent of the catalytic activity of MMP-14. This observation may implicate MMP-14 as a player in the regulation of the changing energy metabolism in malignant cells during tumor formation and migration (Radichev et al. 2009 ). MMP-14 has further been identified as a cell surface sheddase and is capable of cleavage of many membrane-anchored proteins such as E- and N cadherin, integrins, hyaluronan receptor CD44, receptor activator of NF-κB ligand (RANKL) and several cell surface proteoglycans and their receptors (Barbolina and Stack 2008 ). Ongoing research by proteomics techniques has revealed an ever-growing array of possible endogenous substrates of MMP-14 (Niiya et al. 2009 ). MMP-14 was originally identified as the extracellular protease responsible for activation of proMMP-2, and this process remains the best-described proteolytic function of the enzyme. In this process, one of the units of an MMP-14 dimer forms a trimeric complex with proMMP-2 and TIMP-2 at the cell surface leading to proteolytic removal of the propeptide of the proMMP-2 by the ‘free’ MMP-14 unit. Besides proMMP-2, also proMMP-13 (Knauper et al. 2002 ) and proMMP-8 (Holopainen et al. 2003 ) have been identified as possible targets for activation through this mechanism. The role of MMP-14 in angiogenesis is well described and demonstrated by insufficient vascularization in knockout mice. MMP-14 is capable of degradation of the deposited fibrin matrix after vascular injury, effectively disrupting the repair mechanism and allowing endothelial cell invasion (Hiraoka et al. 1998 ). MMP-14 may also be involved in migration of endothelial cells into the ECM, by proteolytic degradation of extracellular matrix proteins and by processing of various adhesion molecules (Galvez et al. 2001 ). The formation and stabilization of the newly formed capillary tubes may also be dependent on MMP-14 activity, as demonstrated by impaired capillary formation in knockout models and RNA interference experiments (Robinet et al. 2005 ). Finally, MMP-14 is able to release vascular endothelial growth factor A (VEGF-A) by shedding, again promoting neovascularization (Sounni et al. 2004 ). MMP-14 together with MMP-15 may even be the determining factor in transforming cancer cells into the invasive phenotype, a process that is regulated by zinc-finger proteins such as Snail-1 (Ota et al. 2009 ). MT2-MMP (MMP-15) was first described in 1995 as the second member of the membrane-anchored MMP subfamily (Takino et al. 1995 ). MMP-15 is a ubiquitously expressed enzyme with largely overlapping substrate specificity with MMP-14 (d’Ortho et al. 1997 ). Although the physiological function of this protease is not as well described as for MMP-14, some studies have indicated a role in follicle rupture during ovulation (Ogiwara et al. 2005 ) and the generation of tubular structures during angiogenesis (Lafleur et al. 2002 ). MMP-15 has also anti-apoptotic properties (Abraham et al. 2005 ). MMP-15 is capable of activating the MMP-2 proenzyme, but contrary to MMP-14 the activation mechanism is not dependent on the presence of TIMP-2, but rather on interaction with the hemopexin-like domain of MMP-2 (Morrison et al. 2001 ). Involvement of MMP-15 in pathology is still unclear, but considering the similarity with MMP-14 a role in cancer is expected. Indeed, MMP-15 is present in many investigated tumors, such as glioblastoma (Zhang et al. 2005 ), non-small cell lung carcinoma (Atkinson et al. 2007 ), and ovarian (Davidson et al. 2001 ) and breast carcinoma (Ueno et al. 1997 ), and seems to correlate with tumor invasiveness (Zhang et al. 2005 ). The two latest additions to the membrane-spanning MMP family are MT3-MMP (MMP-16), first described in 1997 (Matsumoto et al. 1997 ), and MT5-MMP (MMP-24), first described in 1999 (Llano et al. 1999 ). These enzymes are still poorly described in literature. The crystal structure of MMP-16 has been elucidated, and shows extensive similarity to MMP-14 (Lang et al. 2004 ). MMP-16 activity is regulated by an autoproteolytic shedding process where a soluble form of the enzyme is released from the cell surface. The active enzyme shows high affinity to TIMP-3, as opposed to TIMP-1. Although originally anticipated to be a brain-specific enzyme (Pei 1999a ), MMP-24 is possibly involved in remodeling events in endometrial lesions and endometriosis (Gaetje et al. 2007 ). Like the other MT-MMPs, both MMP-16 and MMP-24 are capable of activating proMMP-2 (Wang et al. 1999a ; Zhao et al. 2004 ). The final two MT-MMPs are structurally different from the other four with respect to their interaction with the cell membrane. MT4-MMP (MMP-17) (Pendas et al. 1997 ) and MT6-MMP (MMP-25) (Pei 1999b ) are linked to the cell membrane via a glycosyl-phosphatidyl-inositol (GPI) anchor, as opposed to containing a membrane-spanning domain. This anchor moiety is linked to the hemopexin-like domain by a 35–45 amino acid long hydrophilic linker, or stem. After production of the enzyme, this stem region is linked to a short hydrophobic tail, which is exchanged for a GPI anchor in the endoplasmic reticulum (Sohail et al. 2008 ). The stem region further contains two or three cysteine residues, which may have a function in formation of dimers or trimers, as demonstrated by the presence of ~120 and ~180-kDa protein species of MMP-25 that are dissociated into the mature 57-kDa form of the enzyme under reducing conditions (Sun et al. 2007 ). The GPI anchor gives these enzymes the possibility of interaction with lipid raft microstructures and may be involved in internalization and recycling of the enzymes (Sohail et al. 2008 ). The TIMP-inhibition profile of the GPI-anchored MT-MMPs is different from the membrane-anchored MT-MMPs. While the latter are relatively resistant to inhibition by TIMP-1 due to incompatibility of the Thr 98 residue with the S’1 selectivity pocket, MMP-17 and MMP-25 are effectively inhibited by TIMP-1, as well as by TIMP-2 and TIMP-3 (English et al. 2001 ; Wang et al. 1999b ). GPI-anchored MT-MMPs can be shed from various cells in exosomes, possibly leading to paracrine transfer to other cells (Itoh et al. 1999 ). Both MMP-17 and MMP-25 are capable of degrading ECM proteins, albeit MMP-17 in a limited fashion with cleavage demonstrated for gelatin, fibrin and fibrinogen. MMP-25 can cleave a wider range of ECM constituents including fibronectin, type IV collagen and proteoglycans (Kang et al. 2001 ). In vitro experiments have revealed a multitude of possible substrates (including TNF alpha, indicating sheddase activity), but the physiological relevance is unclear [reviewed in Sohail et al. ( 2008 )]. Interestingly, MMP-17 is not able to activate proMMP-2 even in vitro, making this the only MT-MMP that lacks this trait (English et al. 2000 ). MMP-25 does activate proMMP-2, but generates a different form of the active enzyme than the other MT-MMPs, indicating that the interaction between MMP-2 and MMP-25 is different (Nie and Pei 2003 ). This activation mechanism is possibly dependent on the tight junction protein claudin-5, since cells that do not produce this protein are incapable of proMMP-2 activation by MMP-25 (Miyamori et al. 2001 ). MMP-17 has been described as an activator of the aggrecanase ADAMTS-4 (Gao et al. 2004 ). Both GPI-anchored MT-MMPs are highly expressed in a wide variety of cancer cells ranging from breast carcinoma to glioma and colon cancers [reviewed in Sohail et al. ( 2008 )], but the clinical relevance of the presence of these proteins in malignant cells is not yet clear.