toid arthritis and cancer, seem to support this contention.
Summary The matrix-degrading metalloproteinases are an intriguing family of enzymes that have evolved to digest specific extracellular matrix components. The expression of these enzymes is very highly regulated and can be controlled transcriptionally by a number of growth factors, tumor promoters, oncogenes, and hormones. It is suggested that the coordinated regulation of matrix metalloproteinases and their inhibitors by these agents modify the integrity of the extracellular matrix. These modifications may, at least in part, be responsible for mediating the effects of these factors on complex physiological processes.
Introduction The extracellular matrix (ECM) plays a central role in maintaining the structural integrity of primitive, multicellular organisms as well as highly complex mammals. In addition, the matrix plays an instructive role in directing the development and morphogenesis of vertebrate embryos and influences basic cellular processes such as proliferation, differentiation, migration, and adhesion. The early notion that the extracellular matrix is an inert and stable structure has been dispelled and it is clear that a dynamic equilibrium between synthesis and dcgradation of matrix components is required for matrix maintenance. This review focusses on the controlled degradation of cxtraccllular matrix components by a family of matrixdegrading metalloproteinases (MMPs). Although many proteases can cleave ECM molecules, the MMPs are believed to be the normal, physiologically relevant mediators of matrix degradation. The reasons for this arc several-fold: the MMPs are secreted proteins, placing them in the proper location for ECM degradation, and their enzymatic activities are most potent at pH values close to neutrality. There are multiple levels at which the expression and activity of MMPs arc regulated, suggesting tight control is required for the continuation of normal processes. The apparent consequences of abnormally high levels of expression of some metalloproteinases in pathological conditions, such as rheuma-
The Matrix Metalloproteinase Family Although the existence of bacterial collagenases has been long recognized, the first such mammalian enzyme was described in 1962 as the rotease that mediates dissolution of the tadpole tail?’). This enzyme, now referred to as interstitial collagenase, is the prototypic mcmber of what has turned out to be a family of matrixdegrading metalloproteinases. Elucidation of the primary sequence of these enzymes, by cloning of the cDNAs for at least 9 distinct members of the family over the past 6 years, suggests that they arose by duplication of a primordial gene. It has been useful to subdivide the MMPs into three subclasses based on substrate specificity, although distinctions are arbitrary and show a tendency to become less clear as more is known about the enzymatic activity of purified enzymes and substrates. Nomenclaturc for the MMPs i s also somcwhat confusing, since many of the enzymes have been given different names by different groups. The desigiiation used in this review was recently recommended by the International Union of Biochemistry and Molecular Biology(’). The various names and propertics of the MMPs are summarized in Table 1. There are two members of the collagenase subclass that cleave fibrillar collagens: interstitial collagenase and neutrophil collagenase. Both enzymes have the distinctive ability to cleave alpha chains of types T,TT, and I11 collagens at a single site, producing fragments approximately 3/4 and 1/4 the size of the original molecule. Although thece en7ymes hydrolyze other substrates in addition to fibrillar collagens, the cleavage of intact, fibrillar collagens is specifically limited to these enzymes. Once the higher order structure is destroyed, however, many proteases, including other metalloproteinases, can degrade the denatured collagen/gelatin substratcs. The two enzymes differ slightly in their specificity for the different fibrillar collagens: neutrophil collagenase has a preference for type I collagen, while type 111 collagen is preferen tially digested by interstitial collagenase(’). Tnterstitial collagenase is produced by fibroblasts and macrophages in particular, while the expression of neutrophil collagenase is restricted to cells of the neutrophil lineage. The gelatinases are generally thought of as having substrate specificity for denatured collagens (gelatins) and intact, type IV basement membrane collagen. Some reports suggest. however, that native t pe IV Type collagen is a poor substrate for these V collaged5) and elastin(‘) are also reported to bc good substrates for these cnzymes. Expression of the 72 kDa gclatinase A is widespread and is frequently elevated in malignant tumors. Gelatinase B, the 92 kDa cnzyme, was traditionally thought of as the macrophage gclatinase, but it has also been described a\ being
Table 1. Characteristics of the M M P family members Nanies Collagenasrs Interstitial collagenase (EC 3.3.24.7) MMP-1 Fibroblast collagenase Ncutrophil collagcnasc (EC 3.4.24.34) MMP-5 PMN collagenase Gelatinaxes Gelatinase A (EC 3.4.24.24) MMP-2 72 kDa gclatinase IV collagenase Gclatinase B (EC 3.4.23.35)
Deduced mass (kDa)
# of domains
54.1
4
collagens I, II,III* ,VII.X
53.4
3
collagens I".
TT,TTT
73.9
5
gelatins collagen IV?,V elastin
78.4
6
gclatins collagen IV?.V elastin
54.0
4
pro teogl ycans fibroncctin,laminin gelatins collagcns III,IV, V ,TX
MMP-9 Y2 kUa gclatinasc 92 kDa IV collagenase
Stromel ysins Stromelysin 1 (EC 3.4.24.17) MMP-3 transin proteoglycanase procollagenasc activator
ECM Substratcs
Stromelysin 2 (EC 3.3.24.22) MMP-10 transin-2
54.1
proteoglycans fibroncctin,laminin gelatins collagens lll,lV,V,IX
Matrilyin (EC 3.4.24.23) MMP-7 pump-I uterine mctalloproteinase
29.7
proteogi ycans* collagen IV? Iibronectin gclatins elastin
Others Stromelysin 3 Mctalloclastasc
54.6
4
?
53.9
4
elastin fibroncctin
Mcmhcrs of thc mctalloprotcinasc family can be divided into three categories depending on substrate specificity: the collagenases, gclatinases, and stromelysins. The name immediately preceding thc EC numbcr has bccn rccommcndcd by thc International Union of Riochcmistry and Molecular Biology('). The deduced mass of the entire protein starting with the initiating rncthionine is given in kilodallons (kDa)? although both proteolytic processing and glycosylation occurs on most of the cnzymcs so that scvcral forms of the proteins arc frequently observed. The number of protein domains is indicated and further described in Fig. 3 . The ECM substrates listed are rcpresentativc and not ncccssarily comprchcnsive. Components marked with an asterix (*) are exceptionally good substratcs. The question mark (?) indicates that there is some discrepancy in the literature regarding the cleavage o f the indicated substrate, or, in the case of stromclysin 3. that the substrate specificity for this enzyme has not yet been determined.
expressed in transformed and tumor-derived cells, neutrophils, corneal cpithelial cells, cytotrophoblasts, and keratinocytes. Thc third subclass of MMPs is the stromelysin class, so named because of the relatively broad substrate specificity of these enzymes. This family includes two highly homologous enzymes, stromclysin 1 and stromelysin 2, and a third, smaller enzyme now named matrilysin. Their natural substrates appear to be proteoglycans and glycoprotcins such as fibroncctin and laminin. Type IV collagen is cleaved by the stromelysins in the globular, but not the helical, domain. and there is conflicting data rcgarding the ability of matrilysin to
cleave native IV collagen. The possible role of the stromelysins and matrilysin in elastin degradation is also controversial. In general, it appears that matrilysin, but not the stromelysins, can degrade elastin. Stromelysin-1 is not widely expressed normally, but can be readily induced by growth factors, cytokines, tumor promoters, and oncogenes in cultured mesench ma1 cells, such as fibroblasts and chondrocytes( ). Stromelysin-2 and matrilysin were originally cloned from human tumor sampled8). Interestingly, the expression pattern of these enzymes is frequently distinct. For example, stromelysin is expressed in maturc macrophages("), while matrilysin is found at the
7-
earlier promonocyte stage(”). The differential expression of the various members of the stromelysin subclass of MMPs may help to explain why there are several family members with similar substrate specificities. An exciting recent finding is the cloning of a murine macrophage metalloelastase and its identification as a new member of the MMP family(”). This enzyme may rcpresent a fourth, elastin-degrading subclass of MMPs, although a rigorous comparison of the elastolytic activity of all MMPs has not yet been completed. Tn addition to degrading elastin. this cnzvmc also cleaves fibroncctin, casein, and al-antitrypsinTL1).The enzyme is distinctive in that the predicted molecular mass is 53 kDa, the same size as the stromelysins and collagenases, but the protein is rapidly processed at both the Nand C-terminus and is isolated from macrophages in a smaller, 21 kDa form. An additional MMP has been identified as being expressed in the stromal tissue surrounding breast adenocarcinomas(”). Although the cDNA clone was named stromelysin-3, the protein has not yet been expressed and its substrate specificity is unknown at this time. Stromelysin-3 transcripts have been identified in the interdigital regions of a developing human embryo and in normal uterus and placenta(’*). Domain Structure of the MMPs A comparison of the primary amino acid sequences of various members of the MMP family demonstrates that these proteins are divided into several distinct domains which are conserved among family members. These domains have been reviewed previously(13314), and are depicted in Figure 1.Briefly, all the MMPs are secreted proteins and the deduced amino acid sequences of their cDNAs reveal a leader sequence, or ‘pre’ domain. The ‘pro’ domain is removed when the enzymes are activated, and a highly-conserved region of this domain (PRCGVPD) is rcsponsible for the maintenance of enzyme latency. The catalytic domain contains the conserved sequence HEXGH, believed to be the zinc binding site by analogy to the zinc-binding site of bacterial metalloproteinases and by site-directed mutagenesis studies. With the exception of matrilysin, the enzymes also contain a carboxyl-terminal domain with sequence similarity to hemopexin, a heme-binding protein, and to the ECM component vitronectin. The gelatinases contain additional domains with sequence similarity to matrix proteins. A domain with sequence similarity to the fibronectin binding domain of collagen is found in both gelatinases, while a domain with similarity to type V collagen is found only in the larger gelatinase. A recent analysis of the exon and intron structure oC the genes for several MMPs suggests that the various family members evolved by gene duplication and exon shuffling(”) (Fig. 1). The boundaries of the ‘pre’, ‘pro’, and ‘catalytic’domains are found in all the enzymes and
do not always correspond to exon/intron Junctions. These domains are essential for enzyme function and appear to be components of a primordial gene. l h c ‘fibronectin’ and ‘hemopexin’ domains, however, are each contained within a discrete set of exons. This ic consistent with a model involving duplication of the primordial gene and addition of hemopexin-like exons. This 10 exon/4 domain metalloprotcinasc may have further duplicated and the 3-exon fibronectin-binding domain introduced to create the 72 kDa gelatinase A. Further duplication and addition of the type V collagen domain to the ninth exon could have created the 92 kDa gelatinase B. Support for this possibility is the observation that the genes for interstitial collagenase and stromelysin (containing the four domain structure) arc located on chromosome 11, while the genes for gelatinase A and B (containing these domains plus the additional fibronectin-like domain) are on chromosome 16(15). Interestingly, the mouse metalloelastase is located on chromosome 9(”), which is syntenic with the short arm of human chromosomc 19, suggesting that there is perhaps another cluster of MMP genes found on this chromosome. The addition of clusters of exons to the basic proteolytic structure seems to have endowed the resulting enzymes with their individual substrate specificities. The catalytic domain of the MMPs is suificient for proteolytic activity as measured by the degradation of casein. This is true for matrilysin(“), whose active form consists of only this domain, and for artificially truncated versions of stromel sin(”), neutrophilclS)’,and interstitial collagenase(”) created by recombinant DNA techniques. None of the truncated collagenases, however, retain the ability to cleave native collagen(18,‘9).The replacement of the carboxyterminal domain of either of the collagenases with the same domain from stromelysin docs not rcstorc collagen suggesting that the specificity for collagen degradation resides in this domain. The hemopexin domain of stromelysin binds coll a g e ~ ~ ( ’and ~ ) may be responsible for differences observed in substrate specificitiec of the ctromelysinc and matrilysin (Table 1). Along a similar line, the addition of a domain with sequence similarity to the collagen binding domain of fibroncctin to gelatinase A sccms to bc involved in the affinity of this enzyme for denatured collagen(20).It is not clear if the additional type V collagen-like domain that i s contained in the 92 kDa gelatinase B plays any role in substrate specificity, since only subtle ditferences between the activities of gclatinase A and B have been detected against the substrates that have been tested thus far. The smallest member of thc MMP family is matrilysin, which contains only the pre. pro, and catalytic domains. Preliminary data from this laboratory demonstrates that at least thc first 3 intron/exon junctions are identical to that found in the other MMPs for which this information is known (Fig. 2). Tt is tempting to speculate that matrilysin is the present-day
A
Protein Structure Pre
Pro
Catalytic
(Zn)
b ) (
Hemopexin
.......... ..........
Fibronectin
Collagen
Gene Structure
B Gelatinase B
m Emmmaa 157
rnEiEarnrnrnI 156
233 149 129
280 140151104
310
Gelatinase A
IlzZzmrnB 390
227 149 129
im ammm174 174 174
155
135 137160109
901
Interstitial collagenase
Cnmaarm 174
5 l 156
245 149 129
mmmm-
118 134 163 104
Matrilysin
mmmm 156 227
ii ?
149 129
representative of the primordial gene. The addition of the ‘hemopexin’ and ‘fibronectin’ domains may have been responsible for dictating the particular matrix substrate rccogniad by subsequent family members. We may, therefore, obtain insights into the evolution of metalloproteinase family members by making the assumption that the enzymes evolved in the same order as their matrix substrates. Matrilysin is described as having the most otent proteoglycanase activity of the P enzymes tested(-’). Both matrilysin and the strome-
602
Fig. 1. Protein and Gene structure of the MMP family members. Panel A: All MMPs contain at lcast threc protein domains: a ‘pre’ domain encoding thc leader sequence that targcts the enzymes for secretion, a ‘pro’ domain that is removed when the cnzymc becomes permanently activated, and thc ‘catalytic’ domain which contains the zinc binding rcgion. In gelatinase A and B, the catalytic domain i b separated by the ‘fibronechn’ domain, which has 5eyuence similarity to the collagen binding domain of fibronectin. The ‘hemopexin’ domain is found in all the enzymes with the exception of matrilysin. The ‘collagen’ domain has homology to typc V collagen and is found only in gelalinase B Panel B: The exons for four human MMPs, representing one example of each of the four different protein domain structures found in the MMPs, is depicted by a box with the number of nucleotides in the cxon indicated below. The filled regions of the boxes correspond to the protein domains depicted in Panel A . The open boxcs correspond to thc 5‘ and 3‘ untranslatcd regions. The gene structure data is derived from rcfcrenccs 15, 30, and 50. The intron/exon junctions of matrilysin are based on preliminary data from the authors laboratory and are currently incomplete, as indicated by the question mark.
lysins degrade glycoproteins such as fibronectin(“). Complex carbohydrate-containing polymers are a feature of the cell surface of procaryotic and eukaryotic cells. A glycoprotcin that has features that resemble both the cell surface glycoproteins and sulfated proteoglycans of eukaryotic cells has been identified in the halophilic bacteria of the genus Halobacterium(*’). The earliest precursors to the fibrillar collagens and the nonfibrillar collagens to be identified thus far have been isolated from the sponge, a primitive metazoantz3).
plasminogen
L
Fig. 2. Cascade of MMP activation by plasmin. The genes for urokinase, prostromelysin, and procollaTranscriptional genase contain AP-1 and PEA-3 activation +prostromelysin sites, suggesting that they can be (AP-LIPEA-3) coordinatelv induced bv activation collagenase\+ highly active of a pathw’ay involving transcrip+procollagenase tion factors recognizing these sites. collacenase Urokinase converts plasminogcn to plasmin. Plasmin activates both prostromelysin and procollagenase. The cleavage of collagenase by stromelysin results in a more highly activated collagenase. The degradation of collagens, proteoglycans, and glycoproteins can therefore be effected by a proteolytic cascade resulting in activation of several members of the MMP family.
+urokinase
plasmin
Although the earlier existence of collagen gene family members cannot be ruled out at this time, these observations suggest that the proteoglycans and glycoproteins arose before the extracellular structural proteins. It would follow, therefore, that the earliest MMP would recognize these substrates, while the addition of the collagen-binding domains would evolve with the appearance of the collagens. Since the carboxyl terminal domain of stromelysin binds collagen and is required for collagen degradation by the collagcnases(”), this argument would support matrilysin as the most primitive member of the family. An alternative view of the evolution of the matrilysin gene is that it may have arisen from the duplication of one of the genes containing the hemopexin domain but lost the exons encoding the C-terminal domain in the duplication process or by subsequent mutations. Support for this view is derived by comparison of the primary sequence of the catalytic domains of the MMPs, which places the matrilysin gene in a cluster with stromelysin-1, -2, and interstitial collagenase, and places stromelysin-3 closest to the bacterial metalloproteinasesC2‘). Although this analysis does not necescarily suggest a primordial relationship, stromelysin-3 is the nearest to the bacterial enzymes in an evolutionary sence and provides the firct link between the mammalian and bacterial metalloproteinases(2“).Proteolytic activity, substrate specificity, and role of the carboxyterminal domain has not yet been determined for stromelysin-3, so it is not clear if this possibility entirely contradicts the assumptions made regarding the specificity of the earliest MMP. The issue of the evolution of the metalloproteinase family can eventually be resolved by characterizing the MMPs in more primitive species.
Regulation of the MMPs The regulation of the MMPs is complex and occurs at different levels. The selective expression of MMP genes in specific cell types and their inducibility by a variety of biologically active agents such as growth factors, cytokines, oncogenes, and tumor promoters has sparked a great deal of interest. Recent analysis of the promoter sequences of several MMPs has shed light on the possible molecular basis for the inducibility, and may provide a partial explanation for the early observations that at least some of the MMPs are frequcntly co-ordinately regulated. The promoter regions of the genes for human and rat stromelysin-1 and -2 and human and rabbit interstitial collagenase have been shown to have certain common features important to their transcriptional regulation. These gcncs all contain TATA elements approximately 30 nucleotides upstream from the transcriptional start site, and have AP-1 sites and PEA-3 sites within the most proximal 210 bp. Mutational analysis has shown that AP-1 alone, or the combination of AP-1 plus PEA3 (and possibly other sites in addition), regulate the basal levels and inducibility of these genes by a variety
of agentdZ527). Transcription factors that recognize and transactivate through these elernen ts are proto-oncogenes, including c-fos and c-jun, which transactivate through the AP-1 element, and c-ets, which recognizes the PEA-3 element(28).It is. thereforc, relatively easy to visualize that stimulation of signal transduction pathways that activate these factors results in the coordinate induction of genes containing these elements in their promoters. This is certainly an oversimplification, since thcre are several examples in which one of these genes is selectively stimulated without an effect on the other. In general, however, the AP-1 elcmcnt seems to be an important component of a basic unit that can drive the transcription of these genes, while other elements, including the PEA-3 site, may contribute to the fine tuning mechanisms that modulate the inducibility of the MMP genes by agents such as growth factors, oncogenes, cytokines, and tumor promoters. The promoter regions of the gelatinax A(29)and neutrophil collagenase (K. Hasty. unpublished) gene are very different from that of stromelysin. stromclysin2, and collagenase, most notably by the absence of an obvious TATA box and AP-1 site. The gelatinax B gene contains TATA and AP-1-like motifs, but it is not clear that they are acting as true transcriptional elements("). These differences provide a molecular explanation for the differential regulation of thesc genes compared to stromelysin and collagenase. Anothcr gene in which the AP-l/PEA-3 combination of transcriptional elements plays a major role in regulating transcriptional activity is the urokinase plasminogen activator gene(”). Co-ordinate expression of urokinase with MMPs such as stromelysin and collagenasc may, therefore, initiate a cascade of proteolytic events that result in matrix degradation (Fig. 2). Urokinase converts plasminogen, a component of plasma and interstitial fluids, into the active enzyme plasmin. Plasmin can recognize and cleave basic resides in the ‘pro’domain of both collagenase and stromclysin. Removal of a portion of the pro domain destabilizes the association of the cysteine residue in the conserved PRCGVPD region with zinc and results in a conformational change that activates the enzyme. Activated stromelysin is autoproteolytic and cleaves itself at the phenylalanine residue at position 98, removing the entire ‘pro’ domain and resulting in a permanently active, low molecular weight form of the enzyme. Plasmin activation of collagenase makcs this enzyme more susceptible to proteolytic cleavage by stromelysin and cleavage and removal of the ‘pro’ domain of collagenase by stromelysin results in an approximate seven-fold increase in proteolytic activity of collagenas@). This cascade of protcolytic events, reminiscent of the clotting cascade, may serve as a controlled but powerful mechanism to coordinate complete degradation of multiple components of the ECM. Activation of MMPs by a conformation change which results in the release of the cysteine-zinc interaction has been referred to as the ‘cysteine switch’ mechanism of
MMP activation by Van Wart and his colleagues(”). Although there is evidence for activation of MMPs by plasmin in a co-culture system of keratinocytes and fibroblasts(33),it is likely that this mechanism may be only one of several that are operative in vivo. Other enzymes, such as cathepsin G and neutrophil elashave been shown to be activators in v i m and may participate in activation in vivo under certain circumstances. Interestingly, the 72 kDa gelatinase A is not cleaved by pla5min or several other enzymes, and would therefore require an alternative pathway of a c t i ~ a t i o n ( ~Neutrophil ). collagenase may be activated by an oxidative burst(3s). It is hoped that the next few years may bring insights into the normal mechanisms of metalloproteinase activation.
Tissue Inhibitors of Metalloproteinases The MMPs can be inhibited by the broad-qpectrum serum inhibitor a2-macroglobulin and by a special class of tissue inhibitors of metalloproteinases (TIMPs). There are two members of the class of TIMPs, TTMP and TIMP-2. TTMP is a 28 kDa glycoprotein and TIMP2 is a smaller, 20 kDa protein, but this difference is mainly due to the lack of glycosylation of TTMP-2. Thc two proteins share a 40% sequence identity and complete conservation of 6 disulfide bonds. Both form a 1:l complex with the activated form of MMPs, such as collagenase and stromelysin, and inhibit their proteo14 tic activity by a mechanism that is not entirely clear. Interestingly, TTMP-2 can also inhibit thc autoactivation of interstitial procollagenase, whereas TIMP cannot(’”). TIMP-2 can be found in a complex with progelatinase A and prevents the autocatalysis of this enzyme also. It does not prevent gelatinolysis, however, and a second molecule of either TIMP or TIMP-2 is required to completely inhibit proteolysis of ECM substrates(37).It is believed, therefore, that TIMP-2 can bind to the 72 kDa gelatinase A at a site distinct from the active site, a situation which stabilizes the enzyme, while TTMP binds in the active site. The possibility of selective inhibition of either autoactivation or substrate catalysis by different TIMP family members, as well as differences in specificity for different MMPs, raises many interesting possibilities for the regulation of ECM degradation. Interestingly , the mouse TTMP promoter also contains an AP-1 site and can be transcriptionally activated by many of the same agents that activate collagenase or str~melysin(~*). Why would both a protease and its inhibitor be co-ordinately induced? A clue may lie in the realization that TIMP can inhibit active enzyme, but doe5 not appear to inhibit autoactivation. Thus, the enzyme could digest substrate to a very limited extent before being inactivated by the inhibitor, tightly controlling the extent of ECM degradation. Alternatively, there may be a slight difference in the kinetics of induction of the enzymes
and inhibitor that may also allow limited ECM degradation. Normal Roles for Metalloproteinases Although the complex regulation of MMPs and the importance of ECM in maintaining homeostasis suggests that MMPs play critical roles in normal processes, direct evidence to support this contention is sparse. The localization of MMPs in adult tissues or during embryonic devclopmcnt should provide clues to possible functions. The expression of TIMP in many adult and embryonic tissues has been well-established. including bone, cartilage, and endothelial cells(”). Similar studies to examine the expression of MMPs suggests that, in general, the specific enzymes are not as widely expressed. Bone growth and remodeling are likely candidates for the normal role for interstitial collagcnasc and stromelysin. Both enzymes have been localized to the distal femoral growth plate of newborn rabbits(‘”). By culturing the tissue in monensin, it was revealed that the growth-plate chondrocytes were the site of synthesis of the enzymes. Cultured chondrocytes are known to produce collagenase and stromelysin in response to many different biologically-activc agents. Stromelysin has also been identified by immunohistochemistry in the developing enamel matrix(41),another bony structure. Although fibroblasts in culture can be readily induced to express stromelysin and collagenare, we have not yet been able to detect stromelysin mRNA in normal stroma from either adult or embryonic mice (LMM, unpublished). Tt is possible, however, that normal matrix turn-over is mediated by MMPs. It is particularly intriguing to speculate that normal growth, division, and migration of fibroblasts in vivo requires the selective secretion of matrix-degrading enzymes. Such enzymes might be involved in the release of the cell from its normal matrix interactions to allow growth and division of the cell, directional degradation of matrix to allow migration of the daughter cells to their respective positions, and then re-establishment of normal matrix contacts. Such a scenario is supported by the observation that many of the agents that induce MMP expression in fibroblasts also induce growth; these include epidermal growth factor, phorbol ester tumor promotcrs, and oncogeneq. This correlation is not absolute, however, since serum, a potent inducer of cell growth. is not always stimulatory and, in fact, is sometimes inhibitory to MMP mRNA expression. The fact that growth and MMP expression can be dissociated in culture, however, does not necessarily imply that MMP production is not required tor cell proliferation in thc three dimensional environment found in normal stroma. T h e identification of MMP expression in this tissue may be very difficult: fibroblast growth, division, and migration in vivo is a sporadic and rare event and the chances of catching a cell ‘in the act’
FACTORS
.L Prometalloproteinasea 4 E C M components
Fig. 3. Alteration of cellular phenotype by soluble factors may be mediated by ECM changes involving MMPs. Soluble factors such as growth factors, cytokines, and hormones (FACTORS) may alter the pattern of gene expresyion in the nucleus of a cell to express MMPs, their activators, their
inhibitors, or ECM components This, in turn, can alter the integrity of thc ECM sur-
rounding the cell (shaded areas). The ECM, either highly organized as in the cell on the left, or partially degraded as in the cell on the right, in turn can affect the phenotype of the cell (indicated by the circular vs. thc square cell). These el'fects can be differences in properties such as cell proliferation, rnigrdtion, or differentiation. may be remote. MMP expression in bone plates may be more easily detected, since a number of cells in a specific region of the plate may be coordinated in their expression of degradative enzymes. MMPs have also been implicated in several processes specific to reproduction. Ovulation and the release of the mature ovum may involve control of metalloproteinases and their inhibitors(4'). Implantation of the embryo appears to involve specific elaboration of the 92 kDa gelatinase B by the embryo(43), as well as modulation of invasion by TIMP production from the decidua(u). MMPs have been implicated in mediating mammary gland involution following lactation(4s), and the MMP matrilysin has been isolated from thc involuting uterus following parturition(46). One commonality among these diverse processes is that they are all orchestrated events which are tightly regulated by hormones. A possible role for MMPs in regulating branching morphogenesis has been proposed. In particular, bacterial collagenase has been shown to inhibit branching morphogenesis o f cultured salivary gland and inhibitors of metallo roteinases have been shown to stimulate this processPJ7).In addition, we have recently shown that the growth factors Transforming growth factor-a (TGFa) and EGF have a similar effect to mammalian collagenase in inhibiting the branching of cultured lung rudimentd4'). Addition of TTMP to the medium of these lungs reversed the growth factor affect, suggesting that the activity of the factors on branching was mediated by their affect on the expression of MMPs. Are MMPs Mediators of the Biological Effects of Growth Factors, Cytokines, and Hormones? Growth factors, cytokines, and hormones can dramatically alter the physiological state of a cell. They can induce cellular proliferation, differentiation, and migration. Interestingly, the extracellular matrix can have similar affects on cellular function. There are several lines of evidence suggesting that the effects of these
biological modifiers may, at least in part. be mediated by specific alterations in the integrity or composition of the ECM by the MMPs. MMPs and their inhibitors can be induced by a wide variety of growth factors and cytokines and can bc regulated by hormones('). MMP expression in vivo is frequently associated with processes that are controlled by growth factors, cytokines, or hormones, as described in the previous section. And finally, in at least one organ culture system, growth factor effects on a developmental process such a5 branchin morphogcnesis can be altercd by inhibitors O ~ M M P SIt, ( ~therefore, . appears that at least some of the complex effects of these agents may be mediated through a mechanism that involvcs MMPs. The alteration of cellular morphology and function as a consequence of modulation of the ECM and the possible role of MMPs in this process is depicted in Fig. 3 . The activation of specific signal transduction pathways may co-ordinate the transcriptional regulation of MMPs, potential activators, and inhibitors to shift the balancc towards ECM degradation or towards synthesis. EGF could be an example of a factor which induces MMPs and urokinase. a MMP activator, resulting in a cascade of proteolytic events and ECM dissolution as described previously (Fig. 2). Transforming growth factor$ (TGFP), in contrast, is a potent inducer of ECM components and has been shown to elevate TIMP levels in some systems(49).TGFP also inhibits the transcription of some MMPs such as str~melysin(~). It is an example of a factor that could shift the balance towards ECM synthesis through its actions on MMPs and relatcd molecules. The remodeling of the matrix may then contribute to changes in the proliferation, differentiation, or migratory state of the cell. Such changes are the basis for more complex physiological responses such as wound healing or morphogcncsis. Although the effects of growth factors on these processes can occur at many different levels, aii understanding of how these agents regulate matrix metalloproteinases and their effect on matrix remodeling should provide interesting and important insights
in to the molecular basis of complex physiological
responses. Acknowledgements The author apologizcs to all contributors to the MMP field for thc restricted use of references and the inability to acknowledge all pertinent works. The kind cooperation and advanced information obtained from Drs. Howard Welgus, Gillian Murphy, and Karen Hasty, and the helpful discussions with Drs. Alexander Strongin, Tom Wight, Diane Blake, and Helene Sage are gratefully acknowledged. The author’s laboratory is supported by NIH grants CA-46843, CA-48799 and HD-
25580. References 1 Gross, J. and Lapiere. C. M. (1962). Collagenolytic activity in amphibian tissues: a tissue culture assay. Proc. Natl Acad. Sci. USA 54, 1197-1204. 2 (1992). Enzytne Nomenclati~re1991. Recommendations of [hi>Nomenclafur
Summary The matrix-degrading metalloproteinases are an intriguing family of enzymes that have evolved to digest specific extracellular matrix components. The expression of these enzymes is very highly regulated and can be controlled transcriptionally by a number of growth factors, tumor promoters, oncogenes, and hormones. It is suggested that the coordinated regulation of matrix metalloproteinases and their inhibitors by these agents modify the integrity of the extracellular matrix. These modifications may, at least in part, be responsible for mediating the effects of these factors on complex physiological processes.
Introduction The extracellular matrix (ECM) plays a central role in maintaining the structural integrity of primitive, multicellular organisms as well as highly complex mammals. In addition, the matrix plays an instructive role in directing the development and morphogenesis of vertebrate embryos and influences basic cellular processes such as proliferation, differentiation, migration, and adhesion. The early notion that the extracellular matrix is an inert and stable structure has been dispelled and it is clear that a dynamic equilibrium between synthesis and dcgradation of matrix components is required for matrix maintenance. This review focusses on the controlled degradation of cxtraccllular matrix components by a family of matrixdegrading metalloproteinases (MMPs). Although many proteases can cleave ECM molecules, the MMPs are believed to be the normal, physiologically relevant mediators of matrix degradation. The reasons for this arc several-fold: the MMPs are secreted proteins, placing them in the proper location for ECM degradation, and their enzymatic activities are most potent at pH values close to neutrality. There are multiple levels at which the expression and activity of MMPs arc regulated, suggesting tight control is required for the continuation of normal processes. The apparent consequences of abnormally high levels of expression of some metalloproteinases in pathological conditions, such as rheuma-
The Matrix Metalloproteinase Family Although the existence of bacterial collagenases has been long recognized, the first such mammalian enzyme was described in 1962 as the rotease that mediates dissolution of the tadpole tail?’). This enzyme, now referred to as interstitial collagenase, is the prototypic mcmber of what has turned out to be a family of matrixdegrading metalloproteinases. Elucidation of the primary sequence of these enzymes, by cloning of the cDNAs for at least 9 distinct members of the family over the past 6 years, suggests that they arose by duplication of a primordial gene. It has been useful to subdivide the MMPs into three subclasses based on substrate specificity, although distinctions are arbitrary and show a tendency to become less clear as more is known about the enzymatic activity of purified enzymes and substrates. Nomenclaturc for the MMPs i s also somcwhat confusing, since many of the enzymes have been given different names by different groups. The desigiiation used in this review was recently recommended by the International Union of Biochemistry and Molecular Biology(’). The various names and propertics of the MMPs are summarized in Table 1. There are two members of the collagenase subclass that cleave fibrillar collagens: interstitial collagenase and neutrophil collagenase. Both enzymes have the distinctive ability to cleave alpha chains of types T,TT, and I11 collagens at a single site, producing fragments approximately 3/4 and 1/4 the size of the original molecule. Although thece en7ymes hydrolyze other substrates in addition to fibrillar collagens, the cleavage of intact, fibrillar collagens is specifically limited to these enzymes. Once the higher order structure is destroyed, however, many proteases, including other metalloproteinases, can degrade the denatured collagen/gelatin substratcs. The two enzymes differ slightly in their specificity for the different fibrillar collagens: neutrophil collagenase has a preference for type I collagen, while type 111 collagen is preferen tially digested by interstitial collagenase(’). Tnterstitial collagenase is produced by fibroblasts and macrophages in particular, while the expression of neutrophil collagenase is restricted to cells of the neutrophil lineage. The gelatinases are generally thought of as having substrate specificity for denatured collagens (gelatins) and intact, type IV basement membrane collagen. Some reports suggest. however, that native t pe IV Type collagen is a poor substrate for these V collaged5) and elastin(‘) are also reported to bc good substrates for these cnzymes. Expression of the 72 kDa gclatinase A is widespread and is frequently elevated in malignant tumors. Gelatinase B, the 92 kDa cnzyme, was traditionally thought of as the macrophage gclatinase, but it has also been described a\ being
Table 1. Characteristics of the M M P family members Nanies Collagenasrs Interstitial collagenase (EC 3.3.24.7) MMP-1 Fibroblast collagenase Ncutrophil collagcnasc (EC 3.4.24.34) MMP-5 PMN collagenase Gelatinaxes Gelatinase A (EC 3.4.24.24) MMP-2 72 kDa gclatinase IV collagenase Gclatinase B (EC 3.4.23.35)
Deduced mass (kDa)
# of domains
54.1
4
collagens I, II,III* ,VII.X
53.4
3
collagens I".
TT,TTT
73.9
5
gelatins collagen IV?,V elastin
78.4
6
gclatins collagen IV?.V elastin
54.0
4
pro teogl ycans fibroncctin,laminin gelatins collagcns III,IV, V ,TX
MMP-9 Y2 kUa gclatinasc 92 kDa IV collagenase
Stromel ysins Stromelysin 1 (EC 3.4.24.17) MMP-3 transin proteoglycanase procollagenasc activator
ECM Substratcs
Stromelysin 2 (EC 3.3.24.22) MMP-10 transin-2
54.1
proteoglycans fibroncctin,laminin gelatins collagens lll,lV,V,IX
Matrilyin (EC 3.4.24.23) MMP-7 pump-I uterine mctalloproteinase
29.7
proteogi ycans* collagen IV? Iibronectin gclatins elastin
Others Stromelysin 3 Mctalloclastasc
54.6
4
?
53.9
4
elastin fibroncctin
Mcmhcrs of thc mctalloprotcinasc family can be divided into three categories depending on substrate specificity: the collagenases, gclatinases, and stromelysins. The name immediately preceding thc EC numbcr has bccn rccommcndcd by thc International Union of Riochcmistry and Molecular Biology('). The deduced mass of the entire protein starting with the initiating rncthionine is given in kilodallons (kDa)? although both proteolytic processing and glycosylation occurs on most of the cnzymcs so that scvcral forms of the proteins arc frequently observed. The number of protein domains is indicated and further described in Fig. 3 . The ECM substrates listed are rcpresentativc and not ncccssarily comprchcnsive. Components marked with an asterix (*) are exceptionally good substratcs. The question mark (?) indicates that there is some discrepancy in the literature regarding the cleavage o f the indicated substrate, or, in the case of stromclysin 3. that the substrate specificity for this enzyme has not yet been determined.
expressed in transformed and tumor-derived cells, neutrophils, corneal cpithelial cells, cytotrophoblasts, and keratinocytes. Thc third subclass of MMPs is the stromelysin class, so named because of the relatively broad substrate specificity of these enzymes. This family includes two highly homologous enzymes, stromclysin 1 and stromelysin 2, and a third, smaller enzyme now named matrilysin. Their natural substrates appear to be proteoglycans and glycoprotcins such as fibroncctin and laminin. Type IV collagen is cleaved by the stromelysins in the globular, but not the helical, domain. and there is conflicting data rcgarding the ability of matrilysin to
cleave native IV collagen. The possible role of the stromelysins and matrilysin in elastin degradation is also controversial. In general, it appears that matrilysin, but not the stromelysins, can degrade elastin. Stromelysin-1 is not widely expressed normally, but can be readily induced by growth factors, cytokines, tumor promoters, and oncogenes in cultured mesench ma1 cells, such as fibroblasts and chondrocytes( ). Stromelysin-2 and matrilysin were originally cloned from human tumor sampled8). Interestingly, the expression pattern of these enzymes is frequently distinct. For example, stromelysin is expressed in maturc macrophages("), while matrilysin is found at the
7-
earlier promonocyte stage(”). The differential expression of the various members of the stromelysin subclass of MMPs may help to explain why there are several family members with similar substrate specificities. An exciting recent finding is the cloning of a murine macrophage metalloelastase and its identification as a new member of the MMP family(”). This enzyme may rcpresent a fourth, elastin-degrading subclass of MMPs, although a rigorous comparison of the elastolytic activity of all MMPs has not yet been completed. Tn addition to degrading elastin. this cnzvmc also cleaves fibroncctin, casein, and al-antitrypsinTL1).The enzyme is distinctive in that the predicted molecular mass is 53 kDa, the same size as the stromelysins and collagenases, but the protein is rapidly processed at both the Nand C-terminus and is isolated from macrophages in a smaller, 21 kDa form. An additional MMP has been identified as being expressed in the stromal tissue surrounding breast adenocarcinomas(”). Although the cDNA clone was named stromelysin-3, the protein has not yet been expressed and its substrate specificity is unknown at this time. Stromelysin-3 transcripts have been identified in the interdigital regions of a developing human embryo and in normal uterus and placenta(’*). Domain Structure of the MMPs A comparison of the primary amino acid sequences of various members of the MMP family demonstrates that these proteins are divided into several distinct domains which are conserved among family members. These domains have been reviewed previously(13314), and are depicted in Figure 1.Briefly, all the MMPs are secreted proteins and the deduced amino acid sequences of their cDNAs reveal a leader sequence, or ‘pre’ domain. The ‘pro’ domain is removed when the enzymes are activated, and a highly-conserved region of this domain (PRCGVPD) is rcsponsible for the maintenance of enzyme latency. The catalytic domain contains the conserved sequence HEXGH, believed to be the zinc binding site by analogy to the zinc-binding site of bacterial metalloproteinases and by site-directed mutagenesis studies. With the exception of matrilysin, the enzymes also contain a carboxyl-terminal domain with sequence similarity to hemopexin, a heme-binding protein, and to the ECM component vitronectin. The gelatinases contain additional domains with sequence similarity to matrix proteins. A domain with sequence similarity to the fibronectin binding domain of collagen is found in both gelatinases, while a domain with similarity to type V collagen is found only in the larger gelatinase. A recent analysis of the exon and intron structure oC the genes for several MMPs suggests that the various family members evolved by gene duplication and exon shuffling(”) (Fig. 1). The boundaries of the ‘pre’, ‘pro’, and ‘catalytic’domains are found in all the enzymes and
do not always correspond to exon/intron Junctions. These domains are essential for enzyme function and appear to be components of a primordial gene. l h c ‘fibronectin’ and ‘hemopexin’ domains, however, are each contained within a discrete set of exons. This ic consistent with a model involving duplication of the primordial gene and addition of hemopexin-like exons. This 10 exon/4 domain metalloprotcinasc may have further duplicated and the 3-exon fibronectin-binding domain introduced to create the 72 kDa gelatinase A. Further duplication and addition of the type V collagen domain to the ninth exon could have created the 92 kDa gelatinase B. Support for this possibility is the observation that the genes for interstitial collagenase and stromelysin (containing the four domain structure) arc located on chromosome 11, while the genes for gelatinase A and B (containing these domains plus the additional fibronectin-like domain) are on chromosome 16(15). Interestingly, the mouse metalloelastase is located on chromosome 9(”), which is syntenic with the short arm of human chromosomc 19, suggesting that there is perhaps another cluster of MMP genes found on this chromosome. The addition of clusters of exons to the basic proteolytic structure seems to have endowed the resulting enzymes with their individual substrate specificities. The catalytic domain of the MMPs is suificient for proteolytic activity as measured by the degradation of casein. This is true for matrilysin(“), whose active form consists of only this domain, and for artificially truncated versions of stromel sin(”), neutrophilclS)’,and interstitial collagenase(”) created by recombinant DNA techniques. None of the truncated collagenases, however, retain the ability to cleave native collagen(18,‘9).The replacement of the carboxyterminal domain of either of the collagenases with the same domain from stromelysin docs not rcstorc collagen suggesting that the specificity for collagen degradation resides in this domain. The hemopexin domain of stromelysin binds coll a g e ~ ~ ( ’and ~ ) may be responsible for differences observed in substrate specificitiec of the ctromelysinc and matrilysin (Table 1). Along a similar line, the addition of a domain with sequence similarity to the collagen binding domain of fibroncctin to gelatinase A sccms to bc involved in the affinity of this enzyme for denatured collagen(20).It is not clear if the additional type V collagen-like domain that i s contained in the 92 kDa gelatinase B plays any role in substrate specificity, since only subtle ditferences between the activities of gclatinase A and B have been detected against the substrates that have been tested thus far. The smallest member of thc MMP family is matrilysin, which contains only the pre. pro, and catalytic domains. Preliminary data from this laboratory demonstrates that at least thc first 3 intron/exon junctions are identical to that found in the other MMPs for which this information is known (Fig. 2). Tt is tempting to speculate that matrilysin is the present-day
A
Protein Structure Pre
Pro
Catalytic
(Zn)
b ) (
Hemopexin
.......... ..........
Fibronectin
Collagen
Gene Structure
B Gelatinase B
m Emmmaa 157
rnEiEarnrnrnI 156
233 149 129
280 140151104
310
Gelatinase A
IlzZzmrnB 390
227 149 129
im ammm174 174 174
155
135 137160109
901
Interstitial collagenase
Cnmaarm 174
5 l 156
245 149 129
mmmm-
118 134 163 104
Matrilysin
mmmm 156 227
ii ?
149 129
representative of the primordial gene. The addition of the ‘hemopexin’ and ‘fibronectin’ domains may have been responsible for dictating the particular matrix substrate rccogniad by subsequent family members. We may, therefore, obtain insights into the evolution of metalloproteinase family members by making the assumption that the enzymes evolved in the same order as their matrix substrates. Matrilysin is described as having the most otent proteoglycanase activity of the P enzymes tested(-’). Both matrilysin and the strome-
602
Fig. 1. Protein and Gene structure of the MMP family members. Panel A: All MMPs contain at lcast threc protein domains: a ‘pre’ domain encoding thc leader sequence that targcts the enzymes for secretion, a ‘pro’ domain that is removed when the cnzymc becomes permanently activated, and thc ‘catalytic’ domain which contains the zinc binding rcgion. In gelatinase A and B, the catalytic domain i b separated by the ‘fibronechn’ domain, which has 5eyuence similarity to the collagen binding domain of fibronectin. The ‘hemopexin’ domain is found in all the enzymes with the exception of matrilysin. The ‘collagen’ domain has homology to typc V collagen and is found only in gelalinase B Panel B: The exons for four human MMPs, representing one example of each of the four different protein domain structures found in the MMPs, is depicted by a box with the number of nucleotides in the cxon indicated below. The filled regions of the boxes correspond to the protein domains depicted in Panel A . The open boxcs correspond to thc 5‘ and 3‘ untranslatcd regions. The gene structure data is derived from rcfcrenccs 15, 30, and 50. The intron/exon junctions of matrilysin are based on preliminary data from the authors laboratory and are currently incomplete, as indicated by the question mark.
lysins degrade glycoproteins such as fibronectin(“). Complex carbohydrate-containing polymers are a feature of the cell surface of procaryotic and eukaryotic cells. A glycoprotcin that has features that resemble both the cell surface glycoproteins and sulfated proteoglycans of eukaryotic cells has been identified in the halophilic bacteria of the genus Halobacterium(*’). The earliest precursors to the fibrillar collagens and the nonfibrillar collagens to be identified thus far have been isolated from the sponge, a primitive metazoantz3).
plasminogen
L
Fig. 2. Cascade of MMP activation by plasmin. The genes for urokinase, prostromelysin, and procollaTranscriptional genase contain AP-1 and PEA-3 activation +prostromelysin sites, suggesting that they can be (AP-LIPEA-3) coordinatelv induced bv activation collagenase\+ highly active of a pathw’ay involving transcrip+procollagenase tion factors recognizing these sites. collacenase Urokinase converts plasminogcn to plasmin. Plasmin activates both prostromelysin and procollagenase. The cleavage of collagenase by stromelysin results in a more highly activated collagenase. The degradation of collagens, proteoglycans, and glycoproteins can therefore be effected by a proteolytic cascade resulting in activation of several members of the MMP family.
+urokinase
plasmin
Although the earlier existence of collagen gene family members cannot be ruled out at this time, these observations suggest that the proteoglycans and glycoproteins arose before the extracellular structural proteins. It would follow, therefore, that the earliest MMP would recognize these substrates, while the addition of the collagen-binding domains would evolve with the appearance of the collagens. Since the carboxyl terminal domain of stromelysin binds collagen and is required for collagen degradation by the collagcnases(”), this argument would support matrilysin as the most primitive member of the family. An alternative view of the evolution of the matrilysin gene is that it may have arisen from the duplication of one of the genes containing the hemopexin domain but lost the exons encoding the C-terminal domain in the duplication process or by subsequent mutations. Support for this view is derived by comparison of the primary sequence of the catalytic domains of the MMPs, which places the matrilysin gene in a cluster with stromelysin-1, -2, and interstitial collagenase, and places stromelysin-3 closest to the bacterial metalloproteinasesC2‘). Although this analysis does not necescarily suggest a primordial relationship, stromelysin-3 is the nearest to the bacterial enzymes in an evolutionary sence and provides the firct link between the mammalian and bacterial metalloproteinases(2“).Proteolytic activity, substrate specificity, and role of the carboxyterminal domain has not yet been determined for stromelysin-3, so it is not clear if this possibility entirely contradicts the assumptions made regarding the specificity of the earliest MMP. The issue of the evolution of the metalloproteinase family can eventually be resolved by characterizing the MMPs in more primitive species.
Regulation of the MMPs The regulation of the MMPs is complex and occurs at different levels. The selective expression of MMP genes in specific cell types and their inducibility by a variety of biologically active agents such as growth factors, cytokines, oncogenes, and tumor promoters has sparked a great deal of interest. Recent analysis of the promoter sequences of several MMPs has shed light on the possible molecular basis for the inducibility, and may provide a partial explanation for the early observations that at least some of the MMPs are frequcntly co-ordinately regulated. The promoter regions of the genes for human and rat stromelysin-1 and -2 and human and rabbit interstitial collagenase have been shown to have certain common features important to their transcriptional regulation. These gcncs all contain TATA elements approximately 30 nucleotides upstream from the transcriptional start site, and have AP-1 sites and PEA-3 sites within the most proximal 210 bp. Mutational analysis has shown that AP-1 alone, or the combination of AP-1 plus PEA3 (and possibly other sites in addition), regulate the basal levels and inducibility of these genes by a variety
of agentdZ527). Transcription factors that recognize and transactivate through these elernen ts are proto-oncogenes, including c-fos and c-jun, which transactivate through the AP-1 element, and c-ets, which recognizes the PEA-3 element(28).It is. thereforc, relatively easy to visualize that stimulation of signal transduction pathways that activate these factors results in the coordinate induction of genes containing these elements in their promoters. This is certainly an oversimplification, since thcre are several examples in which one of these genes is selectively stimulated without an effect on the other. In general, however, the AP-1 elcmcnt seems to be an important component of a basic unit that can drive the transcription of these genes, while other elements, including the PEA-3 site, may contribute to the fine tuning mechanisms that modulate the inducibility of the MMP genes by agents such as growth factors, oncogenes, cytokines, and tumor promoters. The promoter regions of the gelatinax A(29)and neutrophil collagenase (K. Hasty. unpublished) gene are very different from that of stromelysin. stromclysin2, and collagenase, most notably by the absence of an obvious TATA box and AP-1 site. The gelatinax B gene contains TATA and AP-1-like motifs, but it is not clear that they are acting as true transcriptional elements("). These differences provide a molecular explanation for the differential regulation of thesc genes compared to stromelysin and collagenase. Anothcr gene in which the AP-l/PEA-3 combination of transcriptional elements plays a major role in regulating transcriptional activity is the urokinase plasminogen activator gene(”). Co-ordinate expression of urokinase with MMPs such as stromelysin and collagenasc may, therefore, initiate a cascade of proteolytic events that result in matrix degradation (Fig. 2). Urokinase converts plasminogen, a component of plasma and interstitial fluids, into the active enzyme plasmin. Plasmin can recognize and cleave basic resides in the ‘pro’domain of both collagenase and stromclysin. Removal of a portion of the pro domain destabilizes the association of the cysteine residue in the conserved PRCGVPD region with zinc and results in a conformational change that activates the enzyme. Activated stromelysin is autoproteolytic and cleaves itself at the phenylalanine residue at position 98, removing the entire ‘pro’ domain and resulting in a permanently active, low molecular weight form of the enzyme. Plasmin activation of collagenase makcs this enzyme more susceptible to proteolytic cleavage by stromelysin and cleavage and removal of the ‘pro’ domain of collagenase by stromelysin results in an approximate seven-fold increase in proteolytic activity of collagenas@). This cascade of protcolytic events, reminiscent of the clotting cascade, may serve as a controlled but powerful mechanism to coordinate complete degradation of multiple components of the ECM. Activation of MMPs by a conformation change which results in the release of the cysteine-zinc interaction has been referred to as the ‘cysteine switch’ mechanism of
MMP activation by Van Wart and his colleagues(”). Although there is evidence for activation of MMPs by plasmin in a co-culture system of keratinocytes and fibroblasts(33),it is likely that this mechanism may be only one of several that are operative in vivo. Other enzymes, such as cathepsin G and neutrophil elashave been shown to be activators in v i m and may participate in activation in vivo under certain circumstances. Interestingly, the 72 kDa gelatinase A is not cleaved by pla5min or several other enzymes, and would therefore require an alternative pathway of a c t i ~ a t i o n ( ~Neutrophil ). collagenase may be activated by an oxidative burst(3s). It is hoped that the next few years may bring insights into the normal mechanisms of metalloproteinase activation.
Tissue Inhibitors of Metalloproteinases The MMPs can be inhibited by the broad-qpectrum serum inhibitor a2-macroglobulin and by a special class of tissue inhibitors of metalloproteinases (TIMPs). There are two members of the class of TIMPs, TTMP and TIMP-2. TTMP is a 28 kDa glycoprotein and TIMP2 is a smaller, 20 kDa protein, but this difference is mainly due to the lack of glycosylation of TTMP-2. Thc two proteins share a 40% sequence identity and complete conservation of 6 disulfide bonds. Both form a 1:l complex with the activated form of MMPs, such as collagenase and stromelysin, and inhibit their proteo14 tic activity by a mechanism that is not entirely clear. Interestingly, TTMP-2 can also inhibit thc autoactivation of interstitial procollagenase, whereas TIMP cannot(’”). TIMP-2 can be found in a complex with progelatinase A and prevents the autocatalysis of this enzyme also. It does not prevent gelatinolysis, however, and a second molecule of either TIMP or TIMP-2 is required to completely inhibit proteolysis of ECM substrates(37).It is believed, therefore, that TIMP-2 can bind to the 72 kDa gelatinase A at a site distinct from the active site, a situation which stabilizes the enzyme, while TTMP binds in the active site. The possibility of selective inhibition of either autoactivation or substrate catalysis by different TIMP family members, as well as differences in specificity for different MMPs, raises many interesting possibilities for the regulation of ECM degradation. Interestingly , the mouse TTMP promoter also contains an AP-1 site and can be transcriptionally activated by many of the same agents that activate collagenase or str~melysin(~*). Why would both a protease and its inhibitor be co-ordinately induced? A clue may lie in the realization that TIMP can inhibit active enzyme, but doe5 not appear to inhibit autoactivation. Thus, the enzyme could digest substrate to a very limited extent before being inactivated by the inhibitor, tightly controlling the extent of ECM degradation. Alternatively, there may be a slight difference in the kinetics of induction of the enzymes
and inhibitor that may also allow limited ECM degradation. Normal Roles for Metalloproteinases Although the complex regulation of MMPs and the importance of ECM in maintaining homeostasis suggests that MMPs play critical roles in normal processes, direct evidence to support this contention is sparse. The localization of MMPs in adult tissues or during embryonic devclopmcnt should provide clues to possible functions. The expression of TIMP in many adult and embryonic tissues has been well-established. including bone, cartilage, and endothelial cells(”). Similar studies to examine the expression of MMPs suggests that, in general, the specific enzymes are not as widely expressed. Bone growth and remodeling are likely candidates for the normal role for interstitial collagcnasc and stromelysin. Both enzymes have been localized to the distal femoral growth plate of newborn rabbits(‘”). By culturing the tissue in monensin, it was revealed that the growth-plate chondrocytes were the site of synthesis of the enzymes. Cultured chondrocytes are known to produce collagenase and stromelysin in response to many different biologically-activc agents. Stromelysin has also been identified by immunohistochemistry in the developing enamel matrix(41),another bony structure. Although fibroblasts in culture can be readily induced to express stromelysin and collagenare, we have not yet been able to detect stromelysin mRNA in normal stroma from either adult or embryonic mice (LMM, unpublished). Tt is possible, however, that normal matrix turn-over is mediated by MMPs. It is particularly intriguing to speculate that normal growth, division, and migration of fibroblasts in vivo requires the selective secretion of matrix-degrading enzymes. Such enzymes might be involved in the release of the cell from its normal matrix interactions to allow growth and division of the cell, directional degradation of matrix to allow migration of the daughter cells to their respective positions, and then re-establishment of normal matrix contacts. Such a scenario is supported by the observation that many of the agents that induce MMP expression in fibroblasts also induce growth; these include epidermal growth factor, phorbol ester tumor promotcrs, and oncogeneq. This correlation is not absolute, however, since serum, a potent inducer of cell growth. is not always stimulatory and, in fact, is sometimes inhibitory to MMP mRNA expression. The fact that growth and MMP expression can be dissociated in culture, however, does not necessarily imply that MMP production is not required tor cell proliferation in thc three dimensional environment found in normal stroma. T h e identification of MMP expression in this tissue may be very difficult: fibroblast growth, division, and migration in vivo is a sporadic and rare event and the chances of catching a cell ‘in the act’
FACTORS
.L Prometalloproteinasea 4 E C M components
Fig. 3. Alteration of cellular phenotype by soluble factors may be mediated by ECM changes involving MMPs. Soluble factors such as growth factors, cytokines, and hormones (FACTORS) may alter the pattern of gene expresyion in the nucleus of a cell to express MMPs, their activators, their
inhibitors, or ECM components This, in turn, can alter the integrity of thc ECM sur-
rounding the cell (shaded areas). The ECM, either highly organized as in the cell on the left, or partially degraded as in the cell on the right, in turn can affect the phenotype of the cell (indicated by the circular vs. thc square cell). These el'fects can be differences in properties such as cell proliferation, rnigrdtion, or differentiation. may be remote. MMP expression in bone plates may be more easily detected, since a number of cells in a specific region of the plate may be coordinated in their expression of degradative enzymes. MMPs have also been implicated in several processes specific to reproduction. Ovulation and the release of the mature ovum may involve control of metalloproteinases and their inhibitors(4'). Implantation of the embryo appears to involve specific elaboration of the 92 kDa gelatinase B by the embryo(43), as well as modulation of invasion by TIMP production from the decidua(u). MMPs have been implicated in mediating mammary gland involution following lactation(4s), and the MMP matrilysin has been isolated from thc involuting uterus following parturition(46). One commonality among these diverse processes is that they are all orchestrated events which are tightly regulated by hormones. A possible role for MMPs in regulating branching morphogenesis has been proposed. In particular, bacterial collagenase has been shown to inhibit branching morphogenesis o f cultured salivary gland and inhibitors of metallo roteinases have been shown to stimulate this processPJ7).In addition, we have recently shown that the growth factors Transforming growth factor-a (TGFa) and EGF have a similar effect to mammalian collagenase in inhibiting the branching of cultured lung rudimentd4'). Addition of TTMP to the medium of these lungs reversed the growth factor affect, suggesting that the activity of the factors on branching was mediated by their affect on the expression of MMPs. Are MMPs Mediators of the Biological Effects of Growth Factors, Cytokines, and Hormones? Growth factors, cytokines, and hormones can dramatically alter the physiological state of a cell. They can induce cellular proliferation, differentiation, and migration. Interestingly, the extracellular matrix can have similar affects on cellular function. There are several lines of evidence suggesting that the effects of these
biological modifiers may, at least in part. be mediated by specific alterations in the integrity or composition of the ECM by the MMPs. MMPs and their inhibitors can be induced by a wide variety of growth factors and cytokines and can bc regulated by hormones('). MMP expression in vivo is frequently associated with processes that are controlled by growth factors, cytokines, or hormones, as described in the previous section. And finally, in at least one organ culture system, growth factor effects on a developmental process such a5 branchin morphogcnesis can be altercd by inhibitors O ~ M M P SIt, ( ~therefore, . appears that at least some of the complex effects of these agents may be mediated through a mechanism that involvcs MMPs. The alteration of cellular morphology and function as a consequence of modulation of the ECM and the possible role of MMPs in this process is depicted in Fig. 3 . The activation of specific signal transduction pathways may co-ordinate the transcriptional regulation of MMPs, potential activators, and inhibitors to shift the balancc towards ECM degradation or towards synthesis. EGF could be an example of a factor which induces MMPs and urokinase. a MMP activator, resulting in a cascade of proteolytic events and ECM dissolution as described previously (Fig. 2). Transforming growth factor$ (TGFP), in contrast, is a potent inducer of ECM components and has been shown to elevate TIMP levels in some systems(49).TGFP also inhibits the transcription of some MMPs such as str~melysin(~). It is an example of a factor that could shift the balance towards ECM synthesis through its actions on MMPs and relatcd molecules. The remodeling of the matrix may then contribute to changes in the proliferation, differentiation, or migratory state of the cell. Such changes are the basis for more complex physiological responses such as wound healing or morphogcncsis. Although the effects of growth factors on these processes can occur at many different levels, aii understanding of how these agents regulate matrix metalloproteinases and their effect on matrix remodeling should provide interesting and important insights
in to the molecular basis of complex physiological
responses. Acknowledgements The author apologizcs to all contributors to the MMP field for thc restricted use of references and the inability to acknowledge all pertinent works. The kind cooperation and advanced information obtained from Drs. Howard Welgus, Gillian Murphy, and Karen Hasty, and the helpful discussions with Drs. Alexander Strongin, Tom Wight, Diane Blake, and Helene Sage are gratefully acknowledged. The author’s laboratory is supported by NIH grants CA-46843, CA-48799 and HD-
25580. References 1 Gross, J. and Lapiere. C. M. (1962). Collagenolytic activity in amphibian tissues: a tissue culture assay. Proc. Natl Acad. Sci. USA 54, 1197-1204. 2 (1992). Enzytne Nomenclati~re1991. Recommendations of [hi>Nomenclafur
