Matrix Metalloproteinases*

The timely breakdown of extracellular matrix (ECM) is essential for embryonic development, morphogenesis, reproduction, and tissue resorption and remodeling. The matrix metalloproteinases (MMPs), also called matrixins, are thought to play a central role in these processes. The expression of most matrixins is transcriptionally regulated by growth factors, hormones, cytokines, and cellular transformation (1, 2). The proteolytic activities of MMPs are precisely controlled during activation from their precursors and inhibition by endogenous inhibitors, a-macroglobulins, and tissue inhibitors of metalloproteinases (TIMPs). Table I lists currently known vertebrate matrixins. In addition, non-vertebrate members have been identified in sea urchins (3), Caenorhabditis elegans (4), soybean (5), and Arabidopsis thaliana (6). Most of these MMPs are the subject of individual chapters in the Handbook of Proteolytic Enzymes (7). This minireview focuses on recent progress in regulation of matrixin activities and their biological and pathological implications.

bronectin-type II domain inserted in the catalytic domain. These repeats interact with collagens and gelatins (15,16).
The C-terminal hemopexin-like domain (about 210 amino acids) has an ellipsoidal disk shape with a four bladed ␤-propeller structure; each blade consists of four antiparallel ␤-strands and an ␣-helix (17). The hemopexin domain is an absolute requirement for collagenases to cleave triple helical interstitial collagens (18), although the catalytic domains alone retain proteolytic activity toward other substrates (19). The hemopexin domain of MMP-2 is also required for the cell surface activation of pro-MMP-2 by MT1-MMP (20,21). The function of the proline-rich linker peptide that connects the catalytic and the hemopexin domains is not known, although its interaction with triple helical collagen is hypothesized based on molecular modeling (22). MMP-23 has cysteine-rich, proline-rich, and IL-1 receptor-like regions instead of the hemopexin domain (10). A transmembrane domain is found in the MT-MMPs, which anchors those enzymes to the cell surface.

Regulation of Matrixin Gene Expression
One of the striking features of the matrixins is that many of those genes are "inducible." The effectors include growth factors, cytokines, chemical agents (e.g. phorbol esters, actin stress fiberdisrupting drugs), physical stress, and oncogenic cellular transformation, etc., and the enhanced MMP gene expression may be down-regulated by suppressive factors (e.g. transforming growth factor ␤, retinoic acids, glucocorticoids) (1). Induction and suppression through the promoter regions of matrixin genes have recently been reviewed by Fini et al. (2).
Certain signaling pathways lead to expression of a particular MMP gene. Addition of soluble antibody to ␣ 5 ␤ 1 integrin causes disruption of the actin cytoskeleton and an increased expression of MMP-1 in rabbit synovial fibroblasts (32). This is because of activation of the GTP-binding protein Rac1, which generated reactive oxygen species and induced activation of NF-B (33). This leads to induction of IL-1␣, an autocrine inducer of MMP-1 expression. This pathway is distinct from integrin-dependent spreading, which leads to MMP-1 expression mediated through AP-1 and PEA3binding sites (34) but does not require the Rho-family GTPase pathway (33).
Inflammatory cytokines, TNF-␣ and IL-1, trigger the ceramide signaling pathways (35). The ceramide-dependent expression of MMP-1 in human skin fibroblast is mediated by three distinct MAP kinase pathways, i.e. ERK1/2, stress-activated protein kinase (SAPK)/JNK, and p38 (36). The maximal induction of MMP-1 gene expression in human astrocytes by oncostatin M is Raf1-dependent and requires cross-talk between the Janus kinase (JAK)-signal transducer and activator of transcription (STAT) and MAP kinase pathways (37).
Ultraviolet B irradiation up-regulates MMP-1, MMP-3, and MMP-9 expression in human dermal fibroblasts (38). Brenneisen et al. (39) reported that this was mediated by the activation of * This minireview will be reprinted in the 1999 Minireview Compendium, which will be available in December, 1999.This is the third article of four in the "Proteases in Cellular Regulation Minireview Series." This work was supported by National Institutes of Health Grants AR39189 and AR40994 (to H. N.) and AR16940 (to J. F. W.).

Activation of Pro-matrixins: Cell Surface Activation of Pro-MMP2
Apart from a few members activated by furin (see above), most are secreted from the cell as inactive zymogens. Secreted promatrixins are activated in vitro by proteinases and by non-proteolytic agents such as SH-reactive agents, mercurial compounds, reactive oxygen, and denaturants. In all cases activation requires the disruption of the Cys-Zn 2ϩ (cysteine switch) interaction, and the removal of the propeptide proceeds often in a stepwise manner (41). In vivo, most pro-matrixins are likely to be activated by tissue or plasma proteinases or opportunistic bacterial proteinases. Using transgenic mice deficient in uPA, Carmeliet et al. (42) have suggested that the uPA/plasmin system is a pathophysiologically significant activator of pro-matrixins. The activation of pro-MMP-2, on the other hand, is thought to take place primarily on the cell surface.
In 1994, Sato et al. (43) cloned the first membrane-type MMP (MT1-MMP) and demonstrated it to be an activator of pro-MMP-2. Recent studies propose that this activation process requires both active MT1-MMP and the TIMP-2-bound MT1-MMP (21,44,45).The TIMP-2 in the latter complex binds, through its C-terminal domain, to the hemopexin domain of pro-MMP-2, which is assumed to localize the zymogen close to the active MT1-MMP (44). Although the major pro-MMP-2 activation pathway appears to be through MT-MMPs, Mazzieri et al. (46) reported that surface-bound pro-MMP-2 also can be activated by the cell surface-associated uPA/plasmin system, whereas soluble plasmin degrades it.
The expression of MT1-MMP and activation of pro-MMP-2 in tumor cells occur in specialized membrane extensions, invadopodia (47). This focal location of MT1-MMP is controlled by the transmembrane and cytoplasmic domains of MT1-MMP. Hiraoka et al. (48) reported that specific fibrinolysis by the plasma membrane-anchored MT1-MMP in endothelial cells is critical for neovascularization. MMP-2 may also be localized on the cell surface of angiogenic endothelial cells and melanoma cells by binding to ␣ v ␤ 3 integrin (49). This interaction is inhibited by a RGD peptide and an excess C-terminal hemopexin domain of MMP-2, which prevents invasive behavior of new blood vessels (50). In contrast, binding of pro-MMP-2 to mouse melanoma cells was not inhibited by the RGDS peptide (51), suggesting that the modes of MMP-2 binding to the cell surface may vary. Those studies, nonetheless, emphasize that membrane-bound proteinases such as MT1-MMP and MMP-2 play an important role in angiogenesis and tumor cell invasion.

Inhibition Mechanism of TIMPs
TIMPs (21-30 kDa) are the major endogenous regulators of MMP activities in the tissue, and four homologous TIMPs (TIMPs-1 to -4) have been identified to date (52). The crystal structure of the complex formed between TIMP-1 and the catalytic domain of MMP-3 was determined by Gomis-Rü th et al. (53). The critical residues involved in MMP inhibition are located around the disulfide bond between Cys 1 and Cys 70 (Fig. 2). The N-terminal ␣-amino and carbonyl groups of Cys 1 bidentately coordinate the catalytic Zn 2ϩ . The N-terminal segment Cys 1 -Thr-Cys-Val-Pro 5 binds to the active site cleft subsite S1 to S4Ј of the enzyme like the substrate, whereas Ser 68 and Val 69 in the connecting loop between  strand C and strand D are arranged in an orientation nearly opposite to that of the P3-P2 segment of a substrate. The side chain of Thr 2 extends into the large S1Ј specificity pocket of MMP-3. TIMP-2 binds to matrixins in a similar manner but has a longer strand A-strand B hairpin loop (54). The high resolution NMR structure of N-TIMP-2 (55) and the crystal structure of free TIMP-2 (56) were also reported. Proteinase binding causes an internal rotation between the N-terminal and the C-terminal domains of ϳ13°and other local conformational changes, especially in the AB ␤-hairpin loop in TIMP-2 (56). MMP-3 contact sites in N-TIMP-2 (55) and N-TIMP-1 contact sites in MMP-3 (57) were also mapped by NMR.

TIMPs Are Multifunctional Proteins
It is widely appreciated that TIMPs inhibit cell invasion in vitro, tumorigenesis, metastasis in vivo, and angiogenesis (52). TIMPs exhibit additional biological functions. TIMP-1 and TIMP-2 have mitogenic activities on a number of cell types, whereas overexpression of these inhibitors reduces tumor cell growth (52), and TIMP-2, but not TIMP-1, inhibits basic fibroblast growth factor-induced human endothelial cell growth (58). These biological activities of TIMPs are independent of MMP-inhibitory activities (59,60). The TIMP-1-procathepsin L complex was reported to stimulate steroidogenesis of rat testis and ovary in vitro (61), but TIMP-1-deficient mice gave little evidence for regulation of steroidogenesis in vivo (62). TIMP-1 also stimulates fibroblasts to produce MMP-1 (63). Recent studies of Zhao et al. (64) demonstrated TIMP-1 accumulates in the nuclei of human fibroblasts in a cell cycle-dependent manner (maximal at the S phase), suggestive of its participation in cell growth.
TIMP-3 binds to extracellular matrix (71) through the C-terminal domain of the inhibitor (72). A single mutation in the Cterminal domain (S156C, G166C, G167C, Y168C, or S181C) is associated with Sorsby's fundus dystrophy, an autosomal dominant macular disorder that causes irreversible loss of vision with onset in the third and fourth decade (73). TIMPs are therefore important regulators not only in matrix turnover but also in cellular activities.

Biological and Pathological Roles of Matrixins
Matrixins participate in many normal biological processes (e.g. embryonic development, blastocyst implantation, organ morphogenesis, nerve growth, ovulation, cervical dilatation, postpartum uterine involution, endometrial cycling, hair follicle cycling, bone remodeling, wound healing, angiogenesis, apoptosis, etc.) and pathological processes (e.g. arthritis, cancer, cardiovascular disease, nephritis, neurological disease, breakdown of blood brain barrier, periodontal disease, skin ulceration, gastric ulcer, corneal ulceration, liver fibrosis, emphysema, fibrotic lung disease, etc.) (74). Although the main function of matrixins is removal of ECM during tissue resorption and progression of many diseases, it is notable that MMPs also alter biological functions of ECM macromolecules by specific proteolysis. For example, MMP-2 released by growth cones promotes neurite outgrowth by inactivating neurite-inhibitory chondroitin sulfate proteoglycans, thereby unmasking the neurite-promoting activity of laminin (75). Specific cleavage of the Ala 586 -Leu 587 bond in the ␣ 2 chain of laminin 5 by MMP-2 induces migration of normal breast epithelial cells by exposing a cryptic promigratory site on laminin 5 (76). The cleaved form of laminin 5 was found in tumors and in tissues undergoing remodeling but not in quiescent tissues (76). Amnion epithelial cells undergo apoptotic cell death before the onset of labor, which is closely associated with degradation of type I collagen by collagenase (most likely MMP-13) (77). Cleavage of type I collagen by MMP-1 and by MMP-13 ini-tiates keratinocyte migration during reepithelialization (78) and osteoclast activation (79), respectively.
Further insights into biological and pathological functions of matrixins have been provided by the use of transgenic animals and gene transfer techniques (80). The expression of MMP-9 in trophoblast giant cells appears to be critical during mouse blastocyst outgrowth and in early implantation stages in vivo (81,82) and in later embryonic skeletal tissue development (83). Vu et al. (84) reported that MMP-9-deficient (MMP-9 Ϫ/Ϫ ) mice exhibited no obvious phenotypic defects. However, these mice exhibited a delayed long bone growth associated with an abnormally thickened growth plate, which was accompanied by delayed apoptosis of hypertrophic chondrocytes, vascularization, and ossification. This abnormality was detected only during development, and normal appearance of bone was seen by 8 weeks of age; probably the deficiency was compensated by other MMPs. MMP-9 Ϫ/Ϫ mice are resistant to subepidermal blistering in experimental bullous pemphigoid, an autoimmune blistering disease, due to the lack of MMP-9 in neutrophils (85).
Mice deficient in MMP-12 (metalloelastase), in contrast to the wild-type, did not develop emphysema in response to long term exposure to cigarette smoke (86). MMP-12 Ϫ/Ϫ macrophages are unable to penetrate basement membrane in vitro and in vivo (87).
The matrixins are considered to participate in dissemination of cancer cells by breaking down ECM, but they are also important in creating an environment that supports the initiation and maintenance of growth of primary and metastatic tumors (88). Mice with adenomatous polyposis coli (Apc) gene mutation develop spontaneous intestinal tumors, but such mice deficient in MMP-7 showed significantly reduced tumor multiplicity and growth (89). MMP-11deficient mice also showed reduced chemically induced tumorigenesis (90). Those effects may be related to proteolytic release of growth factors and/or alteration of cellular environments.

Future Prospects
With a wealth of information provided in this field over the last few years we have begun to understand the significance of matrixins in biology and pathology. The actions of matrixins in vivo are complex and diverse; they are not restricted to simple breakdown of ECM, but they may reveal cryptic biological functions of ECM macromolecules. Both concepts need to be taken into consideration for our understanding of the timely alteration of cellular environments required in normal development and morphogenesis. We also anticipate the discovery of many more new MMPs, which will introduce further complexity in tissue matrix catabolism. Such discovery will, however, help us reveal more precise mechanisms of tissue matrix turnover. Detailed structural and functional analyses of MMPs led to the development of numerous potent synthetic inhibitors of matrixins, and some are in clinical trials to treat patients with cancer, arthritis, periodontal disease, and corneal ulceration. Such agents may be of great therapeutic value, but concerns remain about the consequences of inhibiting biologically functioning matrixins and related ADAMs. Alternative approaches may be tissue-targeted gene therapy with TIMPs or TIMP variants that selectively inhibit specific metalloproteinases. The further gain of knowledge about the mechanisms of cell/tissue-specific regulation of MMP gene expression and their signal transduction pathways may also lead to the rational design of inhibitors that perturb the production of MMPs in a particular cell type without affecting other cells. Such agents will be of great value not only for understanding of the basic biology of matrix and its turnover but also for intervention of diseases resulting from aberrant ECM degradation.