37

Critical Reviews in Oral Biology and Medicine, 4(2): 197-250 (1993)

Matrix Metalloproteinases: A Review*

H. Birkedal-Hansen,1'2 W. G. I. Moore,1 M. K. Bodden,36 L J. Windsor/B. Birkedal-Hansen,4 A. DeCarlo,57 J. A. Engler27

Department of Oral Biology,1 Research Center In Oral Biology,2 Department of Restorative Dentistry,3Department of Diagnostic Sciences,4 Department of Periodontics, University of Alabama Schoolof Dentistry,5 Department of Pathology,6 and Department of Biochemistry,7 University of Alabamaat Birmingham, Birmingham, Alabama

* Address all correspondence to: Dr. Henning Birkedal-Hansen, Department of Oral Biology, SDB Box 54,University of Alabama School of Dentistry, University of Alabama at Birmingham, Birmingham,Alabama 35294, (205) 934-6154.

ABSTRACT: Matrix metalloproteinases (MMPs) are a family of nine or more highly homologous Zn++-endopeptidases that collectively cleave most if not all of the constituents of the extracellular matrix. The presentreview discusses in detail the primary structures and the overlapping yet distinct substrate specificities of MMPsas well as the mode of activation of the unique MMP precursors. The regulation of MMP activity at thetranscriptional level and at the extracellular level (precursor activation, inhibition of activated, mature enzymes)is also discussed. A final segment of the review details the current knowledge of the involvement of MMP inspecific developmental or pathological conditions, including human periodontal diseases.

KEY WORDS: matrix metalloproteinases, extracellular matrix, collagenase, regulation of tissue destruction,human periodontal diseases.

I. METABOLIC DEGRADATION OF THEEXTRACELLULAR MATRIX*

At present it is possible to identify elementsof four to five distinct pathways for degradationof extracellular matrices as summarized in Table1. Structural macromolecules of interstitial con-nective tissues and basement membranes may bedegraded by matrix metalloproteinase (MMP)-dependent, by plasmin (Pln)-dependent, and bypolymorphonuclear (PMN) leukocyte serine pro-teinase-dependent reactions, and in some casesby an apparently distinct phagocytic pathwaybased on intracellular digestion of internalizedmaterial by lysosomal cathepsins. Mineralizedmatrices are degraded by an entirely different

mechanism based on release of acidic thiol pro-teinases to a sealed microenvironment on thematrix surface (osteoclastic pathway). This re-view focuses specifically on matrix metallo-proteinases and their role in the metabolic degra-dation of the extracellular matrix in health anddisease and will only briefly summarize otherpathways.

At the outset it is important to recognize thatmolecular ^assembly of matrix macromoleculesfollows entirely different pathways than molecu-lar assembly. Consequently, the assembly pro-cess in itself is of little help in understanding howthe dissolution of tissue structures is brought about.On the other hand, a certain level of insight intothe structure of natural fibril systems is necessary

Abbreviations used throughout the text are listed after the Acknowledgments section and before the reference list (page231).

1045-4411/93/150 1993 by CRC Press, Inc.

197 by guest on October 12, 2015 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

TABLE 1Pathways for Metabolic Degradation of Extracellular Matrix

Pathway

Pig-dependentpathway

MMP pathway

PMN serlne proteinasepathway

Phagocytic pathway

Osteoclastic pathway

Tissue degraded

Interstitial connective tissues,basement membranes

Interstitial connective tissues,basement membranes

Insterstitial connective tissue,basement membranes

Insterstitial connective tissue

Bone, cementum, dentin

Effector enzymes

to begin to understand those factors that dictatewhether a particular macromolecule is degradedby one pathway or another. Proteolytic attack onsolid-phase matrix structures requires an initialextracellular step to initiate the fragmentationprocess. Because extracellular proteolysis proceedsin an environment with considerable molar ex-cess of either plasma or cellular proteinase inhibi-tors, it is necessary to compartmentalize the pro-cess, that is, to isolate the area to be destroyed andto selectively transport and retain reactants at thesite of action. The importance of compartmental-ization is perhaps most clearly seen in the case ofthe osteoclast that seals off the bone surface to bedegraded and creates a transiently closed microen-vironment that can be fully controlled (Vaes,1988). However, vectorial transport is not limitedto the osteoclast, and it is reasonable to assumethat most cells can selectively direct the move-ment of reactants (Unemori et aL, 1990). Thiseffect is reinforced by mechanisms to retain theenzymes at the site of action based on specificbinding affinities either for the substrate surface(binding of tissue-type plasminogen activator[t-Pa] and plasminogen to fibrin; binding ofgelatinases to collagen chain fragments) or for thecell surface through specific receptor-ligand in-teractions (binding of urokinase-type plasmino-gen activator (u-Pa) and plasmin to cell surfacereceptors) (Plow etai, 1986; Vasalli etaU 1985).

II. ALTERNATE PATHWAYS

A. Pig-Dependent Pathway

A body of evidence suggests that the Pin-dependent pathway plays an important role in the

Pin

MMPs

Cellular location

Peri/extracellular

Peri/extracellular

PMN elastase, cathepsin G Peri/extracellular

Cathepsins

CathepsinsIntraceliular

Peri/extracellular

remodeling of the extracellular matrix in cellmigration, trophoblast and tumor cell invasion,metastasis, embryonic development, and growth.A number of recent reviews have summarized thechemistry and biology of this pathway (Dan0 etal, 1985; Kruithof, 1988; Lijnen and Collen, 1988;Moscatelli and Rifkin, 1988; Saksela and Rifkin,1988; Takada and Takada 1988; Fears, 1989;Gerard and Meidell, 1989; Laiho and Keski-Oja,1989; Andreasen et al., 1990; Testa and Quigley,1990).

Dissolution of susceptible extracellular ma-trix proteins is mediated by cleavage by plas-min, a broad-spectrum serine proteinase that isconverted from its inactive circulating precur-sor form, plasminogen (Pig), to catalyticallyactive form by specific activating enzymes, Pigactivators. The concentration of Pig in plasma,interstitial fluids, and lymph is in the range of100 to 200 |ig/ml (1 to 2 \iM)9 which representsa phenomenal destructive potential. By com-parison, our best estimates suggest that tissueconcentrations of MMPs are at least 10- to 100-fold lower. The circulating Mr 92,000 single-chain Glu-Plg is activated by two cleavagesthat result in truncation of the molecule to yielda Mr 86,000 species composed of a Mr 63,000heavy-chain disulfide bonded to a Mr 25,000light chain. The activated enzyme (Pin) cleavesLys and Arg peptide bonds exposed on the sur-face of a wide range of native protein sub-strates, although at highly varying rates. Fibrinand fibronectin are cleaved rapidly at multiplesites, whereas type I and II collagens are notcleaved at all. In general terms, the Pin-depen-dent pathway appears to mediate the degrada-tion of provisional and rapidly remodelingmatrices but to leave intact the more slowly


remodeling scaffold structures, such as the typeI and II collagen fibrils.

Like other serine proteinases of the fibrin-olytic, coagulation, and complement cascades,Pig has evolved from a primordial prototypetrypsin-like proteinase by addition of regulatorydomains. The five so-called "kringle domains"that encode the affinity for Lys residues arethought to help retain the proteinase at the sub-strate surface by binding to COOH-terminal Lysresidues generated by cleavage of Lys-X bonds.The Pig activating enzymes, u-Pa and t-Pa, arealso serine proteinases that have evolved by theaddition of regulatory domains to a prototypetrypsin-like proteinase domain. U-Pa (Mr 55,000)contains an epidermal growth factor (EGF)-likedomain and a kringle domain, whereas t-Pa (Mr70,000) lacks the EGF-like domain but containstwo kringles.

Pin-dependent proteolysis is initiated by se-cretion of one or both activating enzymes atlocal tissue sites. Most cell types, including thosethat dominate the human periodontal tissues (fi-broblasts, keratinocytes, endothelial cells, PMNleukocytes), may be induced to express one orboth activating enzymes. The activity of Pin andof the activating enzymes is maximized by bind-ing to the substrate (Pig and t-Pa) or the cellsurface (u-Pa and Pig) (recently reviewed byVaheri et ai, 1990). This arrangement serves toconcentrate the reactants at the site of cleavageand under certain circumstances to protect againstcapture by plasma [a-2-macroglobulin (a2M),ocl-antiproteinase, oc2-antiplasmin] and cellular[plasminogen activator inhibitor-1 (PAI-1), and-2 (PAI-2), protease nexin-1 (PN-1)] inhibitors.U-Pa is secreted as a latent precursor that isactivated outside the cell, presumably by limitedproteolysis, by conversion of the single-chainprecursor to a two-chain form. Pin can mediatethis conversion in vitro, but it is uncertain howpro-u-PA and Pig interact to initiate the activa-tion reaction when both are in the precursorform.

B. Neutrophil Serine Proteinases

A body of evidence suggests that PMN leu-kocytes may mediate the degradation of extracel-

lular matrix macromolecules by release of twogranule serine proteinases, neutrophil elastase andcathepsin G. These proteinases are capable ofcleaving a variety of extracellular matrix proteinsincluding type IV collagen (but not type V col-lagen), laminin, fibronectin, and heparan sulfateand cartilage proteoglycans (Weiss, 1989; Jasinand Taurog, 1991; Heck et al, 1990; Janusz andDogerty, 1991).

C. Phagocytic Pathway

The preponderance of evidence for degrada-tion of extracellular matrix by a phagocytic path-way (reviewed by Melcher and Chan, 1981)comes from ultrastructural studies that have un-equivocally shown fragments of cross-striatedcollagen fibrils completely enclosed inside cellsand in some cases associated with lysosomalenzymes (Deporter and Ten Cate 1973, Garant1976, Schellens et al, 1982). Although it hasbeen possible to replicate elements of this pro-cess in vitro (Yajima and Rose, 1977; Svobodaet al\ 1979), it has not yet been possible toclearly define the component molecular reac-tions. Selective inhibition studies suggest thatlysosomal cathepsins but not MMPs, are involvedin the process (Everts et a/., 1985, 1989, 1990).At present it is difficult to reconcile the extracel-lular and intracellular pathways, but it is pos-sible that fragments of fibrils may be excisedperhaps by a collagenase-dependent reaction andthat these fragments are subsequently internal-ized by the cell for digestion in phagolysosomesmainly by thiol-proteinases of the cathepsin fam-ily. Phagocytic collagen degradation is particu-larly prevalent in areas of rapid collagen turn-over such as the involuting uterus (Parakkal,1969) and the periodontal ligament (Melcherand Chan, 1981; Schellens et al, 1982). Thephenomenon has been identified in human andanimal periodontal ligaments in vivo (Listgarten,1973; Melcher and Chan, 1981; Beertsen et al,1978; Schellens et al, 1979), in healing wounds(McGaw and Ten Cate, 1983), and in severalcultured cells including epithelial cells (Birek etal, 1980), fibroblasts (Svoboda et al, 1979),and macrophages (Parakkal, 1969; Deporter,1979).


D. Osteoclastic Bone Resorption

The enzymatic mechanisms by which theorganic matrix of bones and teeth is removedare still incompletely understood, but a body ofevidence recently reviewed by Vaes and col-laborators (Vaes, 1988; Vaes et aL, 1992) sug-gests that both MMPs and lysosomal thiol-pro-teinases are involved. Osteoblasts respond toparathyroid hormone and other resorption-in-ducing stimuli by expression of a typical inter-stitial collagenase (Civitelli et aL, 1989; Heathet aL, 1985; Delaisse et aL, 1988), yet collage-nase has no proteolytic activity under the acidicconditions generated by the osteoclast. This andother findings have led to the hypothesis thatosteoblasts initiate the resorptive process bydissolution of the layer of osteoid using a col-lagenase-dependent (neutral) proteolytic pro-cess and thereby expose the underlying miner-alized bone surface. Osteoclast precursors at-tracted perhaps by signals emitted by the vacat-ing osteoblasts or released from the dissolvingbone surface populate the denuded mineral sur-face and differentiate. The microenvironmentbelow the osteoclast is sealed off, and the un-dulating membrane promotes the mixing of in-gredients released on the bone surfaces. Be-cause of the enormous capacity for acidifica-tion that permits a single adherent osteoclast tolower pH below its plasma membrane from 7.0to 3.0 in 6 min (Silver et aL, 1988), it is com-monly believed that the mineral is disposed ofby a simple acidic dissolution process. It isenvisioned that acidic thiol-proteinase(s) re-leased by the osteoclast are capable of dissolv-ing the collagenous matrix at the low pH. Apromising candidate, a cathepsin B-like enzymefrom chicken osteoclasts, has been isolated andsequenced (Blair et aL, 1986; Blair, personalcommunication). The detailed enzymatic ac-tion of this and other thiol proteinases againstthe irreversibly cross-linked collagenous ma-trix of bone and dentin is not yet well under-stood. Studies by Etherington (1977) andEtherington and Birkedal-Hansen (1987) haveshown that the collagenous network of bone issomewhat susceptible to cleavage by cathep-sins at acidic pH, particularly in high Ca2+-concentrations.

III. MATRIX METALLOPROTEINASEPATHWAY

For recent reviews, the reader is referred toBirkedal-Hansen (1987), Tryggvason et aL (1987),Sakamoto and Sakamoto (1988), Vaes (1988),Weiss (1989), Emonard and Grimaud (1990),Goldberg et aL (1990), Matrisian (1990), Jeffrey(1991), and Woessner (1991). A comprehensivereview of the field including a complete bibliog-raphy of the MMP literature through 1990 com-piled by Woessner (1992) was published recently(Birkedal-Hansen et aL, 1992a).

The MMP gene family encodes nine or moremetal-dependent endopeptidases with activityagainst most if not all extracellular matrix macro-molecules (Table 2). Eight human MMPs havebeen cloned and sequenced, and the proteolyticactivity of the natural or recombinant forms veri-fied for all but one, stromelysin-3 (SL-3) (MMP-11). Recently, a novel MMP, a metalloelastasefrom murine macrophages, has been added to thelist (Shapiro et aL, 1992). The enzymes share anumber of common structural and functional fea-tures but differ somewhat in terms of substratespecificity. For instance, the ability to cleave fibrilsof types I and II collagens, which is characteristicfor the PMN-type and fibroblast-type collagena-ses (PMN-CL, FIB-CL), is not shared by othermembers of the family. The physiologic basis forthis difference, however, is not well understood,nor is the apparent redundancy of evolution ofnine enzymes with overlapping substrate speci-ficities.

A. Modular Structure of MMPs

The nine members of the MMP family in-clude (1) interstitial collagenases (FIB-CL andPMN-CL), (2) stromelysins (SLs) (SL-1 andSL-2), (3) gelatinases (Mr 72,000 gelatinases/typeIII collagenase [Mr 72K GL] and Mr 92,000gelatinase/type IV collagenase [Mr 92K GL]),and (4) other MMPs (putative metalloproteinase[PUMP-1], SL-3, macrophage metalloelastase[MME]). The amino acid sequences of 16 humanand animal (rat, mouse, rabbit, pig) MMPs de-duced from cDNA data are shown in Appendix 1.The enzymes may be regarded as derivatives of


TABLE 2Matrix Metalloproteinase Family

Enzyme

Fibroblast-typecollagenase

PMN-typecoliagenase

Stromelysin-1

Stromelysin-2

Stromelysin-3

Macrophage metallo-elastase

Mr 72K gelatinasetype IV collagenase

Mr 92K gelatinase/type IV collagenase

Putative metallo-proteinase-1

Note: n.d.: not determined.

Abbreviation

FIB-CL

PMN-CL

SL-1

SL-2

SL-3

MME

Mr 72K GL

Mr 92K GL

PUMP-1

MMP# Mr Extracellular matrix substrates

MMP-1 57,000/ Collagen I, II, III, ( l l l l ) , VII, VIII, X;52,000 gelatin; PG core protein

MMP-8 75,000 Same as FIB-CL ( l l l l )

MMP-3 60,000 PG core protein; fibronectin;55,000 laminin; collagen IV, V, IX, X;

elastin; proCL

MMP-10 60,000/ Same as SL-155,000

MMP-11 n.d. n.d.

? 53,000 Elastin

MMP-2 72,000

MMP-9 92,000

MMP-7 28,000

Gelatin; collagen IV, V, VII, X, XI; elastin;fibronectin; PG core protein

Gelatin; collagen IV, V; elastin; PG coreprotein

Fibronectin, laminin, collagen IV, gelatin,proCL, PG core protein

MMP numbering according to Nagase, H.1:421-424 (1992).

A. J. Barrett and J. F. Woessner, Jr. et aL: Matrix. Spec. Supp. No.

the five-domain modular structure characteristicof collagenases and SLs formed either by additionor deletion of domains (Figure 1). The 17-29residue hydrophobic signal sequence is followedby a 77-87 residue propeptide that constitutes theNH7-terminal domain of the secreted MMP pre-cursor, a catalytic domain that contains the cata-lytic machinery including the Zn2+-binding site,and a 5-50 residue proline-rich hinge region thatmarks the transition to the -200 residuehemopexin- or vitronectin-like COOH-terminaldomain that appears to play a role in encodingsubstrate specificity. PUMP-1 does not possessthis last domain and therefore is considerablysmaller than the other members of the family. Thetwo gelatinases contain a single insert in the cata-lytic domain consisting of three tandem repeats offibronectin type II modules that endow the activeand latent enzymes with gelatin-binding proper-ties (Goldberg et a/., 1989; Collier et aL, 1992).Each of the enzymes contains a putative tridentateZn2+-binding site believed to constitute the activesite and to play a role in maintenance of catalyticlatency of the precursor form (Springman et aL,

1990, Van Wart and Birkedal-Hansen, 1990).Although no MMP crystal structures are avail-able, a body of evidence based on homology withother Zn2+-binding enzymes suggests that the Hisresidues in the HEXGH sequence constitute twoof the ligands of the Zn2+-binding site (Vallee andAuld, 1992). It is likely that the third ligand thatcompletes the HEXGHXXGXXH sequence is thehighly conserved His located six residues down-stream (Appendix 1). A fourth ligand site is pre-sumably occupied either by H2O (in the matureactive enzyme) or by a single unpaired Cys resi-due in the propeptide (in the latent precursor). AllMMPs contain a highly conserved Glu- and Asp-rich region between the Zn2+-binding site and thehinge region that is likely to constitute a Ca2+-binding site (Lepage and Gache, 1990). AnotherGlu/Asp-rich region, again a potential Ca2+-bind-ing site, is found in front of the fibronectin type IIdomain inserts (Table 3). The COOH-terminaldomain consists of four repeats (Appendix 1) thatshare some limited sequence homology withmodules also found in hemopexin and invitronectin (Matrisian et al. 1986b; Jenne and


PROTOTYPE

Fibronectin type IIDomain Inserts

MMP-9

MMP-2

MMP-1

MMP-8

MMP-3

MMP-10

MMP-11

MMP-?

MMP-7

a :: I II

Mr92KGL

I Mr 72K GL

I FIB-CL

] PMN-CL

SL-1

SL-2I I

I 1 1 SL-3

j MME

PUMP-1

FIGURE 1. Domain structure of MMPs.

Stanley, 1987). The four tandem modules are heldtogether by a single disulfide bond composed oftwo Cys residues that flank the hemopexin-likedomain (Figure 1; Appendix 1).

B. MMP Gene StructureMMP genes show a highly conserved modu-

lar structure (Figure 2). The CL, SL-1, and SL-2genes each contain ten exons and nine introns in8 to 12 kbp of DNA (Matrisian et al., 1985;Breathnach et al, 1987; Fini et al, 1987; Collieret al., 1988). Based on cDNA sequences, it may

be surmised that PUMP-1 lacks exons 7-10, whichencode the hemopexin-like domain, as well as allor most of exon 6, which encodes the hinge re-gion. Mr 72K GL and Mr 92K GL genes areconsiderably larger (26-27 kbp) and contain threeadditional exons, which encode the threefibronectin type II domains (Figure 2). The ex-tended hinge region of Mr 92K GL is encodedentirely in exon 6 (Huhtala et al, 1990a; 1991).The CL and SL-1 genes are located on the longarm of chromosome 11 (Spurr et al., 1988); theMr 72K GL gene is located on chromosome 16(Huhtala et al, 1990b).


TABLE 3Asp- and Glu-Rich, Putative Ca2+-Binding Sequences

Sequence Protein

DEDD

QDQDDS

E FDVDRD LD LE KDADKN 1DDDQD -D~DADD

HD

DDD

SNTDDDDDQDNDDDD

EE

11D

GGPQGGGGGNR-

-

D_

RQ

DND

DDDDNNNNDDDD-

D1

WW

GGG

DGGGGGGGCGGGG

T

TT

11D

TLRYYTTYE11-

-

D

NK

QQD

EDPNP111VP111

D

ND

AS1

DDQDDTDSNDDD-

FT

1LT

FLEVLTFAYD-

K-

TT

YYT

VLVA1KPAE

-

DD

EG

GG1

AVG1VEEEE

DDE

H-FIB-CL; site 1H-SL-1

H-FIB-CL; site 2H-SL-1Sea urchin hatching

enzyme

FN-R 1FN-R 2FN-R 3FN-R 4FN-R 5Calmodulin 1Calmodulin 2Calmodulin 3Calmodulin 4ThrombospondinMyosin light chainTroponin C consensusParvalbumin consensusUvomorulin Repeat B1Lactalbumin

Human CL and SL-1 sequences are from Appendix 1; sequences forfibronectin receptor (FN-R) sites 1-5, myosin light chain, troponin C,calmodulin, and parvalbumin compiled by Rouslahti, E.: Am. Rev.Biochem. 57:375-413 (1988); sequences for sea urchin hatchingenzyme, thrombospondin, human lactalbumin, and uvomorulin repeatB1 compiled by Lepage T. and C. Gache EMBO J. 9:3993-3012(1990); calmodulin sequences are from Cheung, W. Y.: Fed. Proc.41:2253-2257 (1982).

Propeptide

Catalytic Domain

FN-type II Inserts Hinge Region Pexin-like domain

HUMAN Mr 92K GL

L_13_

I 10

10

J HUMAN Mr 72K GL

HUMAN FIB-CL

RAT SL-1

FIGURE 2. Exon structure of human FIB-CL, Mr 72K GL, and Mr 92K GL and rat SL-1 (Redrawn from Huhtala,P., A. Tuuttila, L. T. Chou, J. Lohi, J. Keski-Oja and K. Tryggvason: J. Biol. Chem. 266:16485-16490 (1991). Withpermission.)


C. MMP Substrate Specificity

Substrate specificity studies have providedinformation about the catalytic properties of theenzymes and in a few instances yielded clues tobiologic function that may help us answer twoconnected questions: what are the natural sub-strates for individual MMP and how are specificMMPs degraded in the tissues? On both proteinand peptide substrates, the MMPs have somewhatoverlapping substrate specificities (Table 2; Tables7 to 11). For instance, virtually all of the enzymescleave gelatin and fibronectin at some rate, andmost cleave type IV and V collagens at suffi-ciently high temperatures. On the other hand, theunique ability of FTB-CLs and PMN-CLs to cleaveinterstitial collagens is not shared by other mem-bers of the family and suggests that this propertymirrors biologic function. MMPs cleave a widerange of largely hydrophobic bonds in native pro-teins as detailed under the description of the indi-vidual enzymes (Tables 7 to 11). Synthetic pep-tides carrying these sequences are also cleaved bythe enzymes and often at considerable rates (Fieldset al, 1987; Fields et al, 1990; Mallya et al,1990). Detailed analysis of the peptide cleavagerates have shown that although the MMPs haveoverlapping substrate specificities, sufficient dif-ferences exist to permit construction of optimizedpeptide substrates that can discriminate betweenthe enzymes (Netzel-Arnett etal, 1991b). Singleamino acid substitutions in the P4-P'4 sites mayyield greater than 20-fold differences in thekcat/KM values for these substrates, and multiplesubstitutions appear to be additive (Fields et al,1987; Netzel-Arnett et al, 1991a).

The molecular disassembly of intact solid-phase substrates such as natural collagen fibrils,proteoglycan aggregates, and basement mem-branes is considerably more complex than thecleavage of isolated soluble or reaggregated mol-ecules exposed to a single enzyme, and theseprocesses are still incompletely understood. Twoexamples illustrate this problem. The 3U -lU frag-ments of interstitial collagens that are producedby collagenases at temperatures considerably be-low the midpoint melting temperature (Tm) havenot been unequivocally identified in any biologicsystem. It is possible that they do exist as short-lived intermediates, but it is also possible that

further degradation takes place already at the fibrillevel. A recent study on natural fragments gener-ated during the degradation of the major cartilageproteoglycan, aggrecan, has underscored this com-plexity. When isolated proteoglycan moleculesare exposed to SL-1, fragments are generated thatidentify the cleavage site as a Asn-Phe bond in theIPEN*FFGV sequence in the G1/G2 interglobulardomain, and evidence based on analysis of humancartilage extracts suggests that this cleavage alsooccurs in vivo (Fosang et al, 1991; Flanneryet al, 1992) (Table 10). Bovine cartilageslices induced to degrade in culture, however,do not carry the Phe-Phe terminal sequence ex-pected from SL-1-mediated cleavage but carryinstead an Ala-Arg-Gly sequence locatedapproximately 30 residues downstream(PLPRNXTEXE*ARGXVILTXK) (Sandy et al,1991). The enzyme responsible for this cleavagehas not been identified.

D. Transcriptional Regulation of MMPGenes

In the intact organism, degradative tasks areaccomplished both by growth factor/cytokine-dependent and independent mechanisms. Amongthe nine members of the MMP gene family, it ispossible to identify two pairs of enzymes (PMN-and FIB-CL, Mr 72K GL and Mr 92K GL) withalmost identical substrate specificity but with dif-ferent transcriptional regulation. One member ofeach pair responds to growth factors and cytokineswhereas the other one does not. Growth factor-responsive MMPs (FIB-CL, SL-1, SL-3, and Mr92K GL) are regulated by closely related mecha-nisms. In contrast, the Mr 72 GL seems to bewidely expressed by most cell types, at least invitro, and appears to be only moderately inducedor repressed (usually two to four-fold) (Salo etal, 1991; Overall etal, 1991a). The regulation ofexpression of MMP in the PMN is uniquely dif-ferent from that of other cell types. Synthesis ofPMN-CL and Mr 92K GL is already completedby the time the PMN enters the vasculature, andany further regulation is mediated by granule re-lease rather than by transcriptional events. In spiteof major differences in the transcriptional regula-tion, storage, and utilization of the enzyme, PMNsappear to utilize the same gene for Mr 92K GL as


mesenchymal cells and keratinocytes. Transcrip-tion of MMP genes gives rise to mRNAs that arerelatively stable in the intracellular milieu, withhalf lives ranging from 12 to 150 h (Brinckerhoffetal, 1986; Overall etal, 1991a). Although mostof the regulatory mechanisms described in thefollowing are transcripional in nature, modulationof mRNA half-life may also play a role.Brinckerhoff et al. (1986) noted that FIB-CLmRNA half-life in rabbit synovial fibroblasts(t1/9 = 12 to 36 h) was significantly increased inresponse to TPA, and Overall et al (1991a) foundthat Mr 72K GL mRNA t1/2 in human fibroblastsincreased from 46 to 150 h after TGF-P stimula-tion, whereas those of CL (53 h) and tissue inhibi-tor of metalloproteinase-1 (TIMP-1) (60 h) re-mained unchanged.

1. Growth Factors and Cytokines

Stimulation or repression of growth factor-and cytokine-responsive MMP genes in manycases results in 20- to 50-fold changes in mRNAand protein levels. It is possible to broadly distin-guish between largely catabolic, interleukin-1(IL-1), tumor necrosis factor-a (TNF-a); largelyanabolic transforming growth factor-(3 (TGF-p);and hybrid growth factors and cytokines withvariable anabolic/catabolic effects, epidermalgrowth factor (EGF), TGF-a, platelet-derivedgrowth factor (PDGF), basic fibroblast growthfactor (bFGF). For example, transcription of theFIB-CL and SL-1 genes (and in some cells theSL-3 and Mr 92K GL genes as well) is induced byIL-lp, TNF-a, PDGF, TGF-a, EGF, bFGF, andnerve growth factor (NGF) and with few excep-tions (Salo et al., 1991) abrogated by TGF-P assummarized in Table 4, In spite of the redun-dancy, some level of specificity appears to beencoded in the process in that (1) different growthfactors and cytokines induce overlapping yet dis-tinct repertoires of MMP and inhibitors and (2)different cell types respond to the same growthfactors and cytokines by expression of unique anddistinct combinations of MMP and inhibitor genes.For example, IL-lp induces expression of FIB-CL and SL-1 in human fibroblasts but not inforeskin keratinocytes (MacNaul et al, 1990;Petersen et al, 1987); Mr 92K GL is repressed by

TGF-P in fibroblasts (Kerr et al, 1988b) but stimu-lated in keratinocytes (Salo et al, 1991); humanrheumatoid synovial fibroblasts respond to IL-lpby induction of SL-1, whereas TNF-a inducesprimarily FIB-CL (MacNaul et al, 1990); TNF-a, TGF-a, and EGF as well as 12-0-tetradecanoyl-phorbol-13-acetate (TPA) induce high-level ex-pression of SL-1 in fibroblasts but SL-2 inkeratinocytes (Birkedal-Hansen et al., unpub-lished).

Several growth factor and cytokine regula-tory pathways converge at the AP-1 binding site,which also constitutes the phorbolester-respon-sive element (TRE) (Angel et al, 1987a,b).AP-1 complexes are heterodimers of proteins ofthe two proto-oncogene families (Jun and Fos)that bind to a ATGAGTCA concensus sequencein the 5' upstream flanking region (-70 bp of thetranslation start site of the human FIB-CL gene).Oncogene and phorbol ester induction of FIB-CLproceeds along a c-fos dependent pathway(Schonthal et al, 1988) as does the induction ofSL-1 by PDGF but not by EGF (Kerr et al,1988a). AP-1 binding sequences have been iden-tified in the 5' flanking region of the FIB-CL,SL-1, and Mr 92K GL genes but are missing in theMr 72K GL gene (Angel et al, 1987b; Schonthalet al, 1988; Huhtala et al, 1991). Based on theobservation that expression of SL-3 (Basset et al,1990) and SL-2 genes (in human keratinocytes,Birkedal-Hansen et al, unpublished) can be in-duced by TPA, we predict that these genes alsocontain AP-1 binding sequences. Sirum andBrinckerhoff (1989) identified a putative AP-1binding site in the human fibroblast SL-2 pro-moter (ATGAATCA) but speculated that the singlebase change at position 5 (A for G) and perhapsa substitution of T for A in position 9 immediatelyfollowing the consensus sequence might accountfor the apparent lack of response of this promoterto TPA in fibroblasts. The promoter region ofthe human keratinocyte SL-2 gene, which ishighly responsive to TPA, has not yet been ana-lyzed. The AP-1 site is a necessary but not suffi-cient element for transcriptional activation ofFIB-CL/SL-1 genes (Auble and Brinckerhoff,1991; Gutman and Wasylyk, 1990; Buttice et al,1991). The FIB-CL promoter contains a TPA-and oncogene-responsive unit (TORU) composedof at least two elements, the AP-1 site and the


TABLE 4Stimulation and Repression of MMP Expression

Induction:growth factorsand cytokines

IL-1 oc,pTNF-ocTGF-ccEGFPDGFbFGFNGFTGF-p

Ref.

1, 22 ,3 ,4566, 78, 91011, 12

Induction:other

TPAOkadaic acidBacterial LPSPGE2Con AcAMPPTH

Ref.

8, 13, 14, 1516,17, 18192021, 2223

Repression

GlucocorticoidsProgesteroneTGF-pRetinoidscAMPIFN-y

Ref.

1,24259,2627,2829, 3031,32

Data from: (1) Frisch and Ruley, 1987; (2) MacNaul et al., 1990; (3) Dayer et al., 1985;(4) Brenner et al., 1989b; (5) Lin and Birkedal-Hansen, unpublished; (6) Kerr et al., 1988a;(7) Bauer et al., 1985; (8) Basset et al., 1990; (9) Edwards et al., 1987; (10) Machida et al.,1991; (11) Salo et al., 1991; (12) Overall et al., 1991 a; (13) Aggeler et al., 1984; (14) Angeletai, 1987a; (15) Wilhelm et al., 1989; (16) Kim et al., 1990; (17) Wahl et al., 1974; (18)Cury etai, 1988; (19) Wahl et al., 1977; (20) Overall and Sodek, 1990; (21) McCarthy etal., 1980; (22) Matrisian et al., 1986b; (23) Civitelli et al., 1989; (24) Jonat et al., 1990;(25) Newsome and Gross, 1977; (26) Kerr et al., 1990; (27) Brinckerhoff, 1990; (28)Nicholson etai., 1990; (29) Takahashi etai., 1991; (30) Kerr etai., 1988b; (31) Andrewsetai., 1990; (32) Wahl etai., 1990.

PEA3 binding site, which act synergistically toinduce maximal levels of transcriptional activa-tion, by TPA and nonnuclear oncogenes (v-src,Py middle T, Ha-ras, v-ra/, w-mos, and c-fos +c-jun) (Gutman and Wasylyk, 1990). PEA3, atranscription factor that also binds to the polyomavirus enhancer, binds to the GAGGATGT se-quence only a few bases from the AP-1 site:

5'93TCGAGAGGATGTTATAAAGCATGAGTCAG 3'

Although /

the other hand, the PEA 3 element, which playsan important role in transcriptional activation ofthe FIB-CL gene by phorbolesters and oncogenes(Gutman and Wasylyk, 1990; Auble andBrinckerhoff, 1991), does not appear to play amajor role in SL-1 induction.

The AP-1 site is also a target for repressionof MMP expression by glucocorticoids. Thesecompounds ablate transcription of responsiveMMP genes through formation of complexesbetween the glucocorticoid receptor (GCR) andthe JunlFos complex in a manner that preventsthe DNA-binding activities of both (Jonat et al,1990; Yang-Yen et al, 1990). It is of interest inthis context that the AP- 1-like binding site in thea-fetoprotein promoter (_160TGAACATAA)is located within the half-site of the GCRconsensus binding site centered at position-160(_166TGTCCTTGAACATAAG_151) (Zhanget al, 1991). In spite of the close functional andspatial relationship between the elements thatmediate glucocorticoid repression and phorbolester induction, it is possible to uncouple thesetranscriptional effects. Mutation of the TTAAsequence at position -102 to -99 of the rabbitCL promoter results in loss of dexamethasonerepressibility but does not affect phorbol induc-ibility (Auble and Brinckerhoff, 1991).

TGF-(3 and interferon-y (IFN-y) predomi-nantly down-regulate MMP gene expression.TGF-(3 represses both the constitutive andcytokine-induced expression of CL and SL-1 infibroblasts through a c-/

3. Cell Shape, Cell-Substrate Adhesion

Induction or stimulation of MMP expressionmay also occur in response to signals or eventsthat are physical rather than chemical in nature,including phagocytosis of particulate matter (la-tex beads, urate and mycostatin crystals [Werband Reynolds, 1974; Birkedal-Hansen et al,1976; Brinckerhoff et al, 1982]), heat-shock(Vance et al, 1989), or treatment with cytocha-lasin B, an actin cytoskeleton-disrupting agent(Harris et al, 1975; Werb et al, 1986). Theobservation that induction of MMP expressionby phorbol esters and by proteinases (Werb andAggeler, 1978) is also associated with profoundchanges in cell shape (rounding) prompted Werband co-workers to examine the relationship be-tween cytoskeletal rearrangement and regula-tion of MMP expression (Aggeler et al, 1984;Werb et al, 1986; Unemori and Werb, 1986).Their studies indicated that cell-shape changesoften, but not invariably, induce MMP expres-sion (Aggeler et al, 1984; Werb et al, 1986)and that it is the reorganization of polymerizedactin rather than cell rounding per se that islinked to induction of MMP expression(Unemori and Werb, 1986). Somewhat at vari-ance with these findings, Kuter et al (1989)concluded that, at least in homogenous primaryrabbit corneal stromal cells, expression of FIB-CL correlates with changes in cell shape onlyin the presence of TPA or cytokines and thatalteration of cell shape induced by a variety offactors is neither sufficient nor necessary forinduction of FIB-CL expression.

Because cell-shape changes are often dic-tated by cell-substrate adhesion, several studieshave suggested that the substrate alone canmodulate MMP expression, and the answerappears to be affirmative. Engagement or cross-linking of integrin receptors by monoclonalantibodies or fibronectin fragments (but notintact fibronectin) results in transcription ofMMP genes (Werb et al, 1989). Other studieshave shown that seeding on collagenous matri-ces of either type I collagen, type IV collagen,or matrigel stimulates expression of FIB-CL inkeratinocytes and fibroblasts (Emonard et al,1990; Petersen et al, 1992). Contact withlaminin stimulates expression of gelatinases by

sarcoma cells, and this effect has been attrib-uted to a 19 amino acid sequence of the lamininA chain (Turpeenniemi-Hujanen et al, 1986;Kanemoto et al, 1990).

4. Second Messenger Signaling

The signaling pathways that lead to induc-tion of MMP expression are still incompletelyunderstood, but certain patterns are beginning toemerge. Recent studies have implicated proteinkinase C (PKC), the major cellular receptor forphorbol esters such as TPA, as an importantmessenger in the transcriptional regulation ofgrowth factor-responsive MMP genes(McDonnell et al, 1990). However, PKC doesnot appear to be involved in all cases. SL-1induction by NGF in rat PC 12 pheochro-mocytoma cells requires multiple protein kinasesacting on a number of postreceptor steps butprobably not PKC (Machida et al, 1991). More-over, okadaic acid, a non-TPA-type tumor pro-moter that does not activate PKC but induces"apparent" activation of protein kinases by inhi-bition of protein phosphatases, also induces FIB-CL expression through an AP-1-dependent path-way (Kim et al, 1990). The role of 3'-5' cyclicadenosine monophosphate (cAMP) remains enig-matic in that a rise in cytoplasmic cAMP insome systems leads to stimulation of MMP ex-pression and in others to repression. cAMP stimu-lates expression of FIB-CL and SL-1 by guineapig peritoneal macrophages, rat UMR-106osteosarcoma cells, and rat fibroblasts (McCarthyet al, 1980; Civitelli et al, 1989; Matrisianet al, 1986b) but represses the constitutive ex-pression of FIB-CL by human GM637 fibro-blasts (Angel et al, 1992) and the IL-1-, EGF-,and oncogene-induced transcription of FIB-CLand SL-1 genes by rat and human fibroblasts(Kerr et al, 1988b; Takahashi et al, 1991).Recent studies suggest that cAMP activates adifferent transcriptional machinery than eitherTPA or growth factors/cytokines. A rise in intra-cellular cAMP induces transient expression ofJun-B, whereas both c-Jun and FIB-CL are re-pressed. Jun B, like c-Jun, is capable of bindingto the AP-1 site but fails to activate the FIB-CLpromoter (Angel et al, 1992).


E. Activation of MetalloproteinasePrecursors

The biologic activation of MMP is still in-completely understood. This gap in knowledgerepresents perhaps the most important obstacle toour understanding of how cells utilizemetalloproteinases to degrade the extracellularmatrix. A body of evidence suggests that the la-

tency of the virgin enzyme is maintained, at leastin part, by virtue of a putative Cys-Zn2+ bond thatlinks the unpaired propeptide Cys residue (Cys73in human FIB-CL) to the active site Zn i+(Springman et al., 1990; Van Wart and Birkedal-Hansen, 1990) (Figure 3) and displaces the H2Omolecule, which is necesssary for catalysis. TheCys residue remains cryptic in the latent form ofthe enzyme but is exposed during activation as a

CL(APMA)

LATENT (Mr 52K)

ACTIVE (Mr 52K, 46K, 43K)

ACTIVE (Mr 41K)

FIGURE 3. Latency/activation mechanism of human FIB-CL precursor. TRY:trypsin cleavage site; PKK: plasma kallikrein cleavage site; CL (PKK, PL): autolyticcleavage site following initial cleavage by plasma kallikrein or plasmin (PL); CL(APMA): autolytic cleavage site following reaction with 4-aminophenylmercuricacetate; SL: stromelysin cleavage site. Location of cleavage sites is from Suzuki,K., J. J. Enghild, T. Morodomi, G. Salvesen, and H. Nagase: Biochemistry.29:10261-10270 (1990).


result of conformational change (Lyons et al.,1991a). Disruption of this putative Cys-Zn2+ bondthat stabilizes the propeptide in its "docked" po-sition may be achieved by seemingly disparatechemical and physical means (Figure 3).Organomercurials, metal ions, thiol reagents, andoxidants presumably interact directly with thepropeptide Cys-residue to shift the equilibriumbetween the "closed" and the "open" form towardthe open form; chaotropic agents (3M KI, 3M

NaSCN) and detergents (1 to 2% SDS) inducepolypeptide chain conformational changes thatalso shift the equilibrium entirely toward the openform; proteolytic enzymes excise a portion of thepropepeptide so that the switch opens, most likelybecause of an entropy effect (Figure 4). Oncestabilized in the open form, the enzyme catalyzesseveral autolytic cleavages to generate the fullyprocessed form (Grant et al., 1987; Nagase et al.,1990; Suzuki et al., 1990). Detailed analyses of

PROTEOLYTICENZYMES

CONFORMATIONALPERTURBANTS

THIOL-REACTIVE AGENTS

LMW - ACTIVE

FIGURE 4. Activation of MMP. The reversible opening of the cysteine switch plays a centralrole in different activation pathways. The equilibrium may be shifted towards the open form byproteolytic cleavage of the propeptide (left column) or by reaction of the propeptide Cys residuewith metal ions, organomercurials, oxidizing agents, and thiol reagents (right column).


the successive activation cleavages of humanproCL and proSL-1 (Figures 3 and 5) have shownthat exogenous proteinases (trypsin, Pin, chymo-trypsin, neutrophil elastase, and plasma kallikrein)attack a short basic sequence exposed on the sur-face of the molecule (Nagase et al, 1990; Suzukiet al, 1990). This initial cleavage is sufficient topermit a second, autolytic cleavage 5 to 8 residuesupstream of the unpaired Cys residue (SL-1C s75,FIB-CLC 73). Exposure to organomercurials alsoleads to intramolecular cleavage of the region, 5to 8 residues upstream of the Cys residue (Nagaseet al, 1990; Stricklin et al, 1983; Grant et al,1987). The final mature form of the enzyme isproduced autolytically by mtermolecular trimmingof 14 to 18 residues from this intermediate, in-cluding the unpaired Cys residue.

autolytic cleavage of its own hinge region alreadyinside E. coli but more slowly, if at all, the au-tolytic activation cleavages of the propeptide. Evenafter complete selfcleavage at the hinge regionsite, the molecule still contains an intact propeptideand escapes capture by oc2M for a period of sev-eral hours and, by that criterion, is essentiallylatent (Windsor et al, 1991). These findings raisethe interesting question of whether it is possiblefor the enzyme to be "latent" on some substrates(oc2M bait region, FIB-CL propeptide) and "ac-tive" on others (FIB-CL hinge region).

A puzzling observation is the so-called"superactivation" of proFIB-CL by several otherMMPs including SL-1, SL-2, and PUMP-1(Murphy et al, 1987; He et al, 1989; Nicholsonetal, 1989; Quantin^a/., 1989; Birkedal-Hansen

NH 2 -terminus Catalytic Domain

PKK TRY, PKK CL(PKK,PL) CL(A?MA) CL(APMA)

FPATLETQEQDVDLVQKYLEKYYNLKNDGRQVEKRRNSGPVTO^ FIB-CL

SL-1

HNECT I PKK SL-l(APMA)

SL-1

i\\r \ \YPLDGAARGEDTSM^VQKYLENYYDLKKDVKQFVRRKDSGPWKK SL-1

ExogenousProteinaseClevage Site

AutolyticCleavage

Sitel

AutolyticCleavage

Site 2

FIGURE 5. Propeptide cleavages associated with activation of human FIB-CL and SL-1. Arrows identifycleavage sites by exogenous proteinases (plasma kallikrein [PKK], trypsin [TRY], chymotrypsin [CT], humanneutrophil elastase [HNE], and human stromelysin-1 [SL-1]) and autolytic cleavage sites after activation isinitiated either by plasma kallikrein (CL [PKK]) or APMA (CL [APMA]; SL-1 [APMA]). Data from Suzuki, K., J. J.Enghild, T. Morodomi, G. Salvensen and H. Nagase: Biochemistry. 29:10261-10270 (1990) and Nagase, H.,J. J. Enghild, K. Suzuki and G. Salvesen: Biochemistry. 29:5783-5789 (1990).

The general sequence of events outlined aboveis supported by mutational analyses. Replacementof the Cys residue and of several other residues(Arg, Gly, Val, and Pro) in the highly conservedPRCGVPDV propeptide sequence (Appendix 1)results in mutant enzymes that appear to be par-tially or fully active (Windsor et al, 1991; Park etal, 1991; Sanchez-Lopez et al, 1988). The stateof activity of these mutants, however, remainssomewhat enigmatic. For example, the Cys-Sermutant of human FIB-CL rapidly catalyzes an

et al, unpublished). ProFIB-CL is capable ofcompleting all of the necessary autolytic activa-tion cleavages, either after excision of the first 35residues of the propeptide or after exposure toorganomercurials. The final cleavage of the se-quence -Asp-Val-Ala-Gln-Phe-Val-Leu-, whichmarks the transition from the propeptide to themature activated enzyme, yields either a Val orLeu amino terminus (Val-FIB-CL; Leu-FIB-CL)(Figure 3; Table 5). In the presence of proSL/SL,the predominant mature form, however, is Phe-


CL, which has a 5 to 8-fold higher catalytic activ-ity than either of the other two forms (Susukiet al9 1990). These findings have prompted specu-lation about whether SL plays a role in the bio-logic activation of proFIB-CL, but the questionhas not yet been resolved fully.

In the continuing search for a general, bio-logically relevant MMP activation mechanism,the short basic propeptide sequences that are thesites for the initial cleavage in trypsin/Pln-inducedactivation of FIB-CL and SL-1 have attractedconsiderable attention (Grant et al, 1987; Stricklin

TABLE 5Human MMP Sequences Surrounding the Autolytic Activation Sites

K V M K Q P R C G V P D V A Q - - - F *V *L T E G N P R W FIB-CLD M M K K P R C G V P D S G G - - - - - - - - - - F M * L T P G N P K W PMN-CLE V M R K P R C G V P D V G H _ _ _ _ _ _ _ _ _ *F R T F P G I P K W S L - 1E V M R K P R C G V P D V G H - - - - - - - - - - F S S F P G M P K W S L - 2S S L R P P R C G V P D P S D G L S A R N R Q K R F V L S G G - - R W S L - 3E I M Q K P R C G V P D V A E - - - - - - - - - - Y S L F P N S P K W P U M P - 1E T M R K P R C G N P D V A N - - - - - - - - - - * Y N F F P R K P K W M f 7 2 K G LK * A * M R T P R C G V P D L G R - - - - - - - - - - F Q T F E G D L K W M r 9 2 K G L

Note: Asterisks indicate autolytic cleavage sites verified by sequence analysis. The NH2-terminal sequences of theautolytic activation products of SL-2, SL-3, and PUMP-1 are not known.

Data from: FIB-CL: Grant, G. A., A. Z. Eisen, B. L Marmer, W. T. Roswit and G. I. Goldberg: J. Biol. Chem.262:5886-5889 (1987); Suzuki, K., J. J. Enghild, T. Morodomi, G. Salveson, and H. Nagase: Biochemistry.29:10261-10270 (1990); PMN-CL: Mallya, S. K., K. A. Mookhtiar, Y. Gao, K. Brew, M. Dioszegi, H. Birkedal-Hansenand H. E. Van Wart: Biochemistry. 29:10628-10634 (1990); SL-1: Nagase, H., J. J. Enghild, K. Suzuki, and G.Salveson: Biochemistry. 29:5783-5789 (1990); Mr72K GL: Stetler-Stevenson, W. G., H. C. Krutzsch, M. P. Wacher,I. M. K. Margulies, and L. A. Liotta: J. Biol. Chem. 264:1353-1356 (1989a); Mr 92K GL: Wilhelm, S. M., I. E. Collier,B. L. Marmer, A. Z. Eisen, G. A. Grant and G. I. Goldberg: J. Biol. Chem. 264:17213-17221 (1989); Tschesche,H., V. Knauper, S. Kramer, J. Michaelis, R. Oberhoff and H. Reinke: Matrix. Spec. Suppl. No. 1: 245-255 (1992).

Organomercurial activation of the Mr 72KGL results in autolytic cleavage, which removesa Mr 8K peptide by hydrolysis at the Tyrg5 -Asn86 bond, at a site homologous to that cleavedin other metalloproteinases, eight residues down-stream from the conserved propeptide Cys resi-due (Table 5) (Stetler-Stevenson et ai, 1989a).The activation of the Mr 92K GL appears to pro-ceed somewhat differently and with a differentendpoint. The enzyme is activated rather slowlyby either organomercurials or trypsin (Lyons etal., 1991a). Trypsin-activation of human PMNMr 92 GL results in cleavage of a Lys-Ala bondand yields an NH2-terminal sequence ofAMRTPRCGVD (Tschesche et al, 1992). Orga-nomercurial-induced autoactivation of this enzymeresults in cleavage only one residue downstream,namely at the Ala-Met bond, which yields aMRTPRCGVD NH2-terminal sequence (Wilhelmet al, 1989). In either case, the thiol-bearing se-quence appears to be retained in the mature en-zyme.

et al, 1983; Suzuki et al, 1990). Sequences sur-rounding the trypsin-sensitive site are only par-tially conserved, and it is not immediately appar-ent why some but not all MMPs are amenable toactivation by trypsin and Pin (Appendix 1). Forexample, PUMP-1 (KNAN) is activated quite wellby trypsin, PMN-CL (RKNG) and M 92K GL(KSLG) are activated only slowly, and Mr 72KGL (KDTL) not at all under the same conditions.The observation that activation of SL-1 also maybe initiated by cleavage by chymotrypsin (Phe-Val) and human neutrophil elastase (Val-Arg)suggests that the conformation around this site in addition to its sequence plays an importantrole in encoding the susceptibility to proteolysis.The question of whether cultured cells that se-crete the proenzymes possess all of the necessarycomponents to complete their activation has notbeen fully resolved. In fact, it has even provenquite difficult to achieve MMP activation in cellculture systems without addition of exogenousproteinases such as Pig or trypsin. MMP precur-


sors secreted in cell culture "spontaneously" acti-vate very slowly, if at all, even at elevated tem-peratures, perhaps because of coexpression ofMMP inhibitors (TIMP-1 and TIMP-2). In fibro-blast cultures, partial activation of precursors ofFIB-CL and of Mr 72K GL may be achieved byaddition of the ionophore A23187 (Unemori andWerb, 1988) and by Concanavalin A (Con A)(Overall and Sodek, 1990; Ward et ai, 1991).The latter investigators also provided evidencethat activation of the Mr 72K GL is mediated bya cell surface bound "activator". Our own studieshave shown that activation of proFIB-CL in se-rum-free cultures of fibroblasts (which expressSL-1) and of keratinocytes (which express SL-2)is very slow and incomplete for the first 3 to 5 dof incubation and that collagen breakdown isblocked during that period (Birkedal-Hansen etal., 1992b; Birkedal-Hansen et a/., 1989; Lin etaL, 1987; Lyons et a/., 1991a).

The degree of overlap between the Pig- andmetalloproteinase-dependent pathways, in general,and the role of Plg/Pln in the metabolic activationof MMP precursors, in particular, remain uncer-tain. Several studies using in vitro model systemssuggest that MMP-dependent proteolysis is greatlyaccelerated in the presence of Pig (Mignatti et aL,1986, He et al, 1989; Birkedal-Hansen et al.y1992b). The stimulating effect of Pig, while notanalyzed fully in detail, is believed to reside in theability of Pin to initiate the autolytic activation ofseveral MMP zymogens, including FIB-CL andSL-1. The reluctance to accept a generalized modelfor pericellular proteolysis of extracellular matrixsubstrates in which Plg/Pln play a role is focusedon two major points: (1) the lack of specificity ofPin cleavage reactions is not easily reconciledwith the fact that most regulatory proteinase cas-cades are composed of highly specific enzymesand (2) several MMPs are not activated by Pin.Moreover, examples do exist of cells that degrademodel matrices by seemingly Pig-independentprocesses. Such is the case for the degradation ofgelatin and collagen by PMN leukocytes (Peppinand Weiss, 1986; Weiss et a/., 1985). These in-vestigators provided strong evidence that, at leastin the PMN leukocyte, activation of PMN-CL andMr 92K GL may be achieved by oxidative path-ways, presumably by oxidation of the unpairedpropeptide Cys residue by HOC1.

IV. INDIVIDUAL MATRIXMETALLOPROTEINASES

A. Interstitial Collagenases

The PMN-CL gene is only expressed by PMNleukocytes, whereas that of FIB-CL is expressed(after appropriate stimulation) by fibroblasts frommany different sources (skin, synovium, mucosa,cornea, uterus), keratinocytes (Lin et ai, 1987;Petersen et ai, 1987), endothelial cells (Moscatelliet ai, 1980; Herron et aL, 1986), monocytes andmacrophages (Welgus et ai, 1985a; Campbell etai, 1987), chondrocytes (Lefebvre et ai, 1990),and osteoblasts (Otsuka et ai, 1984; Quinn et ai,1990). The protein cores of PMN-CL and FIB-CLare of virtually identical size (Appendix 1), butthe PMN enzyme is more highly glycosylated andhas a considerably larger molecular mass thanFIB-CL (Mr 75,000 vs. 57,000/52,000). It is specu-lated that the carbohydrate moiety of PMN-CLencodes targeting signals that direct the enzymeto granule storage sites. FIB-CL is partiallyglycosylated (Mr 57,000 form) and two potentialAsn-X-Thr N-glycosylation sequences have beenidentified, but the preponderance of the molecules(70 to 80%) remain unglycosylated (Nagase eta/., 1981,1986; Wilhelm etal, 1986; Goldberg etal, 1986). The function of the FIB-CL carbohy-drate moiety is unknown, but it does not appear tobe important for catalytic activity. The two en-zymes also differ in terms of activation mecha-nisms in that FIB-CL is readily activated by trypsinand Pin whereas the PMN-CL is not. Examina-tion of the primary structures (Appendix 1) re-veals that the tribasic KRR sequence in fibro-blast-type collagenase is altered to an RKN se-quence, which apparently fails to confer the samelevel of trypsin/Pln activatability.

The two collagenases differ markedly in termsof transcriptional regulation. PMN-type collage-nase is released instantly from granule storagesites of triggered PMNs, and it is uncertain whetherPMN-CL expression is subject to transcriptionalregulation in the bone marrow. In contrast, FIB-CL is not stored in cells but produced on demandby initiating transcription of the gene (for in-stance by the action of growth factors andcytokines). Because this process is dependent onactivation of a complex transcriptional apparatus,


the accumulation of the enzyme at the local site isdelayed by 6 to 12 h. The significance of thisdifference is evident by comparison of the effectof TPA on unleashing the destructive potential ofthese two cell types. In the PMN leukocyte, TPAgives rise to immediate release of most or all ofthe enzyme (ever) produced by the cell whereasin fibroblasts, 6 to 12 h elapse before any enzymeis released, but once that happens, production andsecretion can be sustained for several days. ThePMN leukocyte is capable of responding in fullforce instantly but not of sustaining any destruc-tive activity beyond minutes.

1. Cleavage of Interstitial Coiiagens

The cleavage of interstitial coiiagens by col-lagenases has been studied in greater detail thanany other MMP-catalyzed reaction, but a numberof questions that pertain to the substrate specific-ity of these enzymes have been only partiallyresolved. These include (1) the structural basis forthe unique ability of collagenases, but not otherMMP, to cleave coiiagens type I, II, and III and(2) the structural basis for the susceptibility of thecollagenase-sensitive Gly-Ile bond but not otherpotentially cleavable bonds along the triple helix.

scission of the Gly775-Ile776 or Gly775-Leu776 bondsof the component a-chains (Table 6). The pri-mary structure of the collagenase-sensitive site isan important factor in determining the suscepti-bility to collagenases, and Wu etal. (1990) showedthat most substitutions in and around the scissilebond by site-directed mutagenesis are unfavor-able and yield either uncleavable or poorly sus-ceptible collagen molecules. However, studiesusing synthetic peptides modeled after the colla-genase-sensitive region have shown that the uniquespecificity cannot be accounted for entirely by theamino acid sequence and that there are severalGly-Leu-Ala and Gly-Ile-Ala sequences through-out the triple helix of interstitial coiiagens that arenot cleaved (Fields et al, 1987; Netzell-Arnett etal, 1991a). In search of an explanation for theunique susceptibility of the scissile Gly-Ile/Gly-Leu bonds, several investigators have suggestedthat the collagenase-sensitive region is a locus ofminor resistance that more readily unfolds andrelaxes its triple helical structure than other partsof the molecule (Wang et al, 1978; Fields et al,1987; Birkedal-Hansen, 1987). In support of thisevidence is the finding that several bonds in thecollagenase-sensitive region of type III collagenare cleaved by trypsin, thermolysin, pronase, andPMN elastase (Wang et al, 1978; Miller et al,

TABLE 6Collagenase Cleavage Sites in Interstitial Coiiagens

Substrate

Calf and chick cc1(l)Calf o2(l)Chick o2(l)Human oc1 (II)Human a1(lll) (skin)Human oc1 (III) (liver)Calf a1 (III)

Sequence

Gly-Pro-Gln-Gly-Ile-Ala-Gly-GlnGly-Pro-Gln-Gly-Leu-Leu-Gly-AlaGly-Pro-Gln-Gly-Ile-Leu-Gly-Ala

-Ile-Ala-Gly-Gln-Leu-Ala-Gly-Leu

Gly-Pro-Leu-Gly-Ile-Ala-Gly-IleGly-Pro-Leu-Gly-Ile-Ala-Gly-Leu

Ref.

1,2,3456786

Data from: (1) Gross etal., 1974; (2) Highberger etal., 1982; (3) Glanville etal., 1983; (4) Bomstein and Traub, 1979; (5) Dixit etal., 1979; (6) Miller etal., 1976; (7) Seyer and Kang, 1981; (8) Lang etal., 1979. (Modified fromNetzel-Arnett, S., G. Fields, H. Birkedal-Hansen and H. E. Van Wart: J. Biol.Chem. 266:6747-6755 (1991a).

At temperatures well below the denaturationtemperature, type I, II, and III coiiagens in solu-tion are cleaved exclusively at a single locus by

1976; Mainardi etal, 1980a; Birkedal-Hansen etal, 1985). It is more difficult to demonstrate asimilar proteinase-sensitive region in type I col-


lagens, but it can be done under narrowly definedconditions (Ryhanen et al., 1983). The higherrates of cleavage of type III collagen by nonspe-cific proteinases appears to mirror its faster cleav-age also by collagenases, at least in some species(Tables 7 and 8). These observations when heldtogether suggest that the collagenase-sensitiveregion in most species unfolds more readily intype III collagen than in type I. That the collage-nase-sensitive site represents a locus of minorresistance also in type I collagen is indirectlysupported by the observation that this site is thepreferred cleavage site by trypsin/chymotrypsin(Ryhanen et al., 1983) and by unrelatedcollagenolytic enzymes from the microorganismAchromobacter iophagus (Lecroisey and Keil,1979), the fiddler crab (Uca pugilator) hepato-pancreas (Welgus and Grant, 1983), and the der-mal larva Hypoderma lineatum (Lecroisey andKeiL 1979). The hypothesis that the collagenase-sensitive site has a greater degree of molecularmobility than other sites is indirectly supportedby microcalorimetric studies, which have shownthat a segment between 3 and 30 triplets in lengthstarts to unfold as much as 10C below the dena-turation temperature (Privalov et al., 1979). Al-though these data do not permit us to identify thelocation of the unfolding segment, it is highlylikely, when viewed in the context of the extraor-dinary proteinase susceptibility of type III col-lagen, that unfolding is initiated around the colla-genase-sensitive site.

The kinetic parameters of FIB-CL from sev-eral different laboratories are summarized in Table7. Although the data were obtained by use ofdifferent enzymatic assays, the consistency ofcertain key determinations such as (1) the rates ofcleavage of type I collagens in solution at 25 to30C, (2) the rates of cleavage of types I and IIIcollagen in solution at 37C, and (3) the KM val-ues on collagen substrates, suggest that meaning-ful comparisons may be made from this uniquedata set. The cleavage of type I collagen at 25 to30C is a rather slow reaction by any criteria inthat less than one molecule per minute is cleavedper molecule of enzyme (16 to 53 h"1). By com-parison, casein is cleaved 10-fold, and oc2M 50-fold, faster than type I collagen. The three inter-stitial collagens are cleaved at very different ratesby FIB-CL in the same temperature range (25 to

30C). For most species, there is in essence a ten-fold difference between collagen types III and I(=300 hr1 and =30 h"1) and again between col-lagen types I and II (-30 h"1 and -3 h"1). Thecleavage of interstitial collagens is highly tem-perature-dependent particularly in thesubdenaturation temperature range (Tm = 40 to42C). A rise in temperature between 25 to 30Cand 35 to 37C increases the rate of cleavage oftype I collagen by 50 to 80-fold (16 to 53 h"1 vs.1600 to 1700 tr1) and all but eliminates the ratedifference between types I and type III collagens.It is also apparent that native triple helical col-lagen (1600 to 1700 h"1) is a much better substratethan denatured collagen (230 to 750 h"1) at thesame temperature (37C). Not only is the rate ofcleavage 2 to 7-fold higher, but kM is 4 to 8-foldlower so that, based on kcat/KM comparisons, triplehelical type I collagen is a 10 to 60-fold bettersubstrate for FIB-CL than either of its componentrandom coil a-chains. The rates of cleavage ofreconstituted fibrillar substrates are much lowerthan those of soluble collagen molecules. At 35 to37C, both type I and III collagen fibrils are cleavedat rates in the range of 11 to 60 h"1 when com-pared with 1600 to 1700 hr1 for the same col-lagens in solution. It is likely that this differenceis due in part to the increased thermal stability offibrillar structures (Tm flbnls = 47 to 49C; Tm sol =40 to 41C) (Birkedal-Hansen et al, 1985). It isinteresting to note that the difference in catalyticrates between types I and III collagen, whichexists at 25 to 30C, all but disappears at 35 to37C whether in soluble or fibrillar form. Thedata shown in Table 7 also reveal several interest-ing species differences, particularly with respectto cleavage of type III collagen. Whereas human,dog, and cat type III collagens are cleaved veryrapidly (350 to 627 h"1), those of guinea pig andchick are not; in fact, chick type III collagen iscleaved about as slowly as collagen type II. PMN-CL is a considerably more efficient enzyme (10 to30-fold) than FIB-CL on virtually all substratesexcept for type III collagen (Hasty et al, 1987;Netzell-Arnett et al., 1991a). A comparison of thecatalytic properties of FIB-CL and PMN-CL isshown in Table 8. The proteolytic activity of col-lagenases is not limited to collagen and collagen-like synthetic peptides. Recent studies have shownthe collagenases cleave a variety of susceptible


TABLE 7Catalytic Properties of Human Fibroblast-Type Coliagenase

Type 1 collagen, rat, solnType 1 collagen, rat, solnType 1 collagen, human, solnType 1 collagen, human, solnType 1 collagen, calf, solnType 1 collagen, guinea pig, solnType 1 collagen, dog, solnType 1 collagen, cat, solnType 1 collagen, chick, solnType II collagen, calf, solnType II collagen, calf, solnType II collagen, human, solnType II collagen, rat, solnType III collagen, human, solnType III collagen, human, solnType III collagen, dog, solnType III collagen, cat, solnType III collagen, guinea pig, solnType III collagen, chick, solnType 1 collagen, calf, solnType III collagen, calf, solnType III collagen, calf, fibrilsType III collagen, human, fibrilsType III collagen, dog, fibrilsType III collagen, cat, fibrilsType III collagen, guinea pig, fibrilsType III collagen, chick, fibrilsType 1 collagen, calf, fibrilsType 1 collagen, dog, fibrilsType 1 collagen, guinea pig, fibrilsType 1 collagen, guinea pig, solnType 1 gelatin, oc1(l), guinea pigType I gelatin, a2(l), guinea pigType I gelatin, ratP-caseinP-caseinGly-Pro-Gln-Gly-lle-Ala-Gly-GInGly-Pro-Gln-Gly-Leu-Ala-Gly-GInG!y-Pro-Leu-Gly-lle-Ala-Gly-G!nGly-Pro-Gln-Ala-lle-Ala-Gly-GInoc2-M, humanOvostatin, chick

kcat(h"1)

1619.54453.434.222.532.222.835.1

3.22.71.04.5

350565472627

18.05.4

16531477

13.145.753.960.74.10.9

11.421.025.0

1700230750

24420160730970

120045001739

2.1

KM(HM) 1

0.80.90.80.80.80.91.51.11.02.41.62.11.11.71.41.11.80.71.2

n.d.n.d.n.d.n.d.n.d.n.d.n.d.n.d.n.d.n.d.n.d.

0.97.03.79.8

n.d.710

3300280036002800

0.170.32

( c A P'

1 h"

1 )

19.021.754.066.842.825.021.520.735.1

1.31.70.484.1

205.9403.6429348

25.74.5

n.d.n.d.n.d.n.d.n.d.n.d.n.d.n.d.n.d.n.d.n.d.

1888.932.9

202.72.5

n.d.0.230.220.350.331.6

102296.6

Temp. (C)

302530252525252525302525253025252525253535353737373737353737373737303030303030302525

Ref

121222333122212332344433333435666787999

101111

Note: n.d.: not determined.

Data from: (1) Mallya etal., 1990; (2) Welgus etal., 1981; (3) Welgus etal., 1985c; (4) Birkedal-Hansenetai, 1985; (5) Welgus etal., 1980; (6) Welgus etal., 1982; (7) Fields etal., 1990; (8) Windsor etal.,1991; (9) Fields eta/., 1987; (10) Netzel-Arnett etal., 1991a; (11) Enghild etal., 1989.


TABLE 8Kinetic Parameters for the Hydrolysis of Soluble Types I, II, and III Collagens andSynthetic Peptides by Human FIB-CL and PMN-CL at 30 C

FIB-CL PMN-CL

Substrate

Rat type 1Human type 1Calf type II

Human type IIIGPQGIAGQGPQAIAGQ

Kcath"1

1644

3.2

350730

4,500

KM,M

0.80.82.4

1.73,3002,800

K c a t /K M ,M-h- i

1954

1.3

2100.221.6

690460130

20039,00087,000

1.01.02.3

2.56,9004,800

69046057

805.7

18

From Netzel-Arnett, S., G. Fields, H. Birkefal-Hansen and H. E. Van Wart: J. Biol. Chem.266:6747-6755 (1991a).

peptide bonds in several different proteins, assummarized in Table 9. Collagenases also cleave(3-casein quite readily (Table 7), but the suscep-tible bond(s) have not been identified.

B. Stromelysins

The stromelysin group of MMPs includes atleast two members, SL-1 and -2. It is more uncer-

TABLE 9Collagenase Cleavage Sites in Noncoilagenous Proteins

Substrate Sequence Ref.

Human oc2-MHuman PZPHuman PZPHuman PZPRat a1 MRat ot1 MRat a2MRat oc2MRat a1-l3, variant 2

Rat a1-l3, variant 1Human-FIB-CL, hingeHuman-FIB-CL, propeptideHuman-FIB-CL, propeptideHuman-FIB-CL, propeptideHuman-PMN-CL, propeptideHuman-oc1PIHuman-oc1PIHuman alACTChick-ovostatin

Gly-Pro-Glu-Gly-Leu-Arg-Val-GlyTyr-Gly-Ala-Gly-Leu-Gly-Val-ValAla-Gly-Leu-Gly-Val-Val-Glu-ArgAla-Gly-Leu-Gly-Ile-Ser-Ser-ThrGlu-Pro-Gln-Ala-Leu-Ala-Met-SerGln-Ala-Leu-Ala-Met-Ser-Ala-IleAla-Ala-Tyr-His-Leu-Val-Ser-GlnMet-Asp-Ala-Phe-Leu-Glu-Ser-SerGlu-Ser-Leu-Pro-Val-Val-Ala-Val

Ser-Ala-Pro-Ala-Val-Glu-Ser-GluPro-Val-Gln-Pro-Ile-Gly-Pro-GlnAsp-Val-Ala-Gln-Phe-Val-Leu-ThrVal-Ala-Gln-Phe-Val-Leu-Thr-GluAla-Gln-Phe-Val-Leu-Thr-Glu-GlyGly-X -Phe-Met-Leu-Thr-Pro-GlyGly-Ala-Met-Phe-Leu-Glu-Ala-IleGlu-Ala-Ile-Pro-Met-Ser-Ile-ProLeu-Leu-Ser-Ala-Leu-Val-Glu-Thr

Leu-Asn-Ala-Gly

111111111

1233345, 66, 75, 68

Abbreviations: PZP: pregnancy zone protein; a1M: alpha-1-macroglobulin; OC1I3:alpha-1 -inhibitor-3;

tain whether the novel SL-3 (Bassett et al, 1990)merits inclusion in this group or in a separate,new group. SLs share with collagenase the typicalfive-domain structure. The only major structuraldifference is a slightly longer hinge region in SLs(26 residues) than in collagenases (16 residues)(Appendix 1). Comparison of the enzymatic prop-erties of n-SL-1 and r-SL-2 (Nicholson et al,1989) suggests that the two enzymes have virtu-ally identical substrate specificities and cleave awide range of extracellular matrix protein sub-strates, including proteoglycan core protein, typeIV and V collagen, the nonhelical NH2 and COOH-terminal peptides of type II collagen, fibronectin,and laminin (see Table 2). Peptide bonds in natu-ral proteins cleaved by SL-1 are shown in Table10.

keratinocytes in the human, but an SL-1 (orSL-2) homologue is induced in murine skin epi-dermis by phorbolester treatment (Matrisian etal, 1986a; Wilhelm et al, 1987). SL-1 has beenextracted and purified from human patellar carti-lage (Gunja-Smith et al, 1989).

SL-2 transcripts appear to be expressed lessabundantly than those of SL-1. Sirum andBrinckerhoff (1989) found little or no SL-2 mRNAin cultured human foreskin and synovial fibro-blasts and were unable to significantly stimulatemRNA levels either by growth factors (EGF andIL-1) or by phorbolesters. On the other hand,Muller et al (1988) observed that SL-2 messagewas 4 to 5 times more abundant than that ofSL-1 in RNA extracts of human head and necktumors, and Breathnach et al (1987) identified rat

TABLE 10Stromelysin Cleavage Sites in Natural Proteins

Substrate

SL-1SL-1PG core proteinPG link protein,human

PG link protein,rat

ori-PIa1-ACTAT-IIISb Pcc2Mcc2MOvostatin

Sequence

Asp-Thr-Leu-Glu-Val-Met-Arg-LysAsp-Val-Gly-His-Phe-Arg-The-PheIle-Pro-Glu-Asn-Phe-Phe-Gly-ValArg-Ala-Ile-His-Ile-Gln-Ala-Glu

His-Ile-Gln-Ala-Glu-Asn-Gly-Pro

Glu-Ala-Leu-Leu-Ile-Ala-Lys-Pro-Gly-Pro-Arg-Val-Leu-Asn-

Ile-Pro-Met--Ser-Ala-Leu-Gly-Arg-Ser-Gln-Gln-Phe--Glu-Gly-Leu-Gly-Phe-Tyr-Ala-Gly-Phe-

Ser-Ile-ProVal-Glu-ThrLeu-Asn-ProPhe-Gly-LeuArg-Val-GlyGlu-Ser-AspThr-Ala-Ser

Ref.

1123

4445666

Abbreviations: PG: proteoglycan; a1-PI: alpha-1-proteinase inhibitor,alpha-1-antitrypsin; oc1-ACT: alpha-1-antichymotrypsin; AT-III:antithrombin-lll; Sb P: substance P.

Data from: (1) Nagase et al, 1990; (2) Fosang et al., 1991;(3) Nguyen et al, 1989; (4) Mast et al, 1991; (5) Harrison et al,1989; (6) Enghild et al, 1989.

SL-1 is expressed by stromal cells either con-stitutively or after induction by growth factors/cytokines (IL-1, EGF, TNF-oc, PDGF) or phorbolesters (Chin et al, 1985; Herron et al, 1986;Wilhelm et al, 1987). The enzyme does not ap-pear to be expressed by PMN leukocytes and

SL-2 transcripts in Rous sarcoma virus trans-formed rat embryo fibroblasts. Moreover, SL-2transcripts were inducible by TPA in the rat in thepresence of some, but not all, fetal bovine sera.We have recently observed that TPA, TGF-a,TNF-a, and EGF, which induce expression of the


SL-1 gene in human fibroblasts, induce expres-sion exclusively of the SL-2 gene in humankeratinocytes (Birkedal-Hansen et ai, unpub-lished).

C. Gelatinases/Type IV Collagenases

The Mr 72K GL is perhaps the most widelydistributed of all MMP and has been identified inskin fibroblasts (Seltzer etai, 1981), keratinocytes(Salo et ai, 1991), chondrocytes (Lefebvre et ai,1991), endothelial cells (Kalebic et ai, 1983),monocytes (Garbisa et ai, 1986), osteoblasts(Overall and Sodek, 1987), and in a number ofother normal and transformed cells (Sang et ai,1990; Arthur et ai, 1989; Liotta et ai, 1979; Saloet ai, 1983). Mr 72K GL does not appear, how-ever, to be expressed by PMN leukocytes but ispresent in a circulating form in plasma (Johanssonand Smedsrod, 1986). The Mr 92K GL is pro-duced by keratinocytes (Wilhelm et ai, 1989;Salo et ai, 1991), monocytes and alveolar mac-rophages (Mainardi et ai, 1984), PMN leuko-cytes (Hibbs et ai, 1985; Murphy et ai, 1989a)and in a number of malignant or transformed cells(Lyons et ai, 1991a; Wilhelm et ai, 1989; Mollet ai, 1990, Davis and Martin, 1990). It is inter-esting to note that PMN, which express a uniqueand distinct collagenase gene, utilize the samegene for the Mr 92K GL, although in a mannerthat yields a storable rather than a secreted geneproduct.

The two gelatinases are similar in many char-acteristics, including high affinity for gelatin bothin the latent and activated form. The main struc-tural difference is the extended 54 amino acidhinge region sequence in the Mr 92K GL thatshares some homology with the oc2 chain of typeV collagen (Wilhelm et ai, 1989) (Appendix 1).Another difference is that the Mr 72K GL precur-sor is tightly associated with TIMP-2, whereasthe Mr 92K GL proenzyme is associated withTIMP-1 (Goldberg et ai, 1989; Wilhelm et ai,1989). Howard et ai (1991a) showed that TIMP-2 remained bound to the gelatinase even aftermercurial activation, but was able to retardautoactivation.

Mr 72K GL is expressed constitutively bymost cells in culture but is only moderately re-sponsive (two- to four-fold) to TPA and growth

factors that induce the Mr 92K GL (Salo et ai,1985;Templeton^a/., 1990; Huhtala

TABLE 11Mr 72K GL Cleavage Sites

Substrate

Gelatin, oc1(l)-CB8Gelatin, oc1(l)-CB8Gelatin, a1(l)-CB8Gelatin, oc1(l)-CB8Gelatin, oc1(l)-CB8Gelatin, a1(l)-CB8Gelatin, cc1(l)-CB7Gelatin, cc1(l)-CB7Gelatin, oc1(l)-CB7Mr 72K GL

Gly-Pro-Gln-Gly-Pro-Ala-Gly-Pro-Ser-Gly-Pro-Ala-Gly-Pro-Ala-Gly-Ala-Lys-Gly-Pro-Ala-Gly-Pro-Ile-Gly-Pro-Hyl-Asp-Val-Ala-

Sequence

Gly-Val-Gly-Val-Gly-Leu-Gly-GluGly-GlxGly-Leu-Gly-PheGly-AsnGly-SerAsn-Tyr

Arg-Gly-GluGln-Gly-Pro-Hyp-Gly-Pro-Arg-Gly-SerAsp-Gly-Pro-Thr-Gly-Ser-Ala-Gly-Pro-Val-Gly-Ala-Arg-Gly-Ala-Asn-Phe-Phe

Ref.

1111111112,3

Data from: (1) Seltzer et al., 1990; (2) Stetler-Stevenson et al.,1989a; (3) Howard et al., 1991a.

D. Other MMPs

SL-3 has been studied only at the messagelevel. SL-3 mRNA is expressed in human mam-mary tumors by mesenchymal cells adjacent toinvading epithelial tumor cells. The observationthat mRNA transcripts could be induced by TPAand by growth factors in embryonic fibroblastssuggests that the transcriptional regulation of thisenzyme is similar to that of FIB-CL, SL-1, and Mr92K GL. The enzyme protein has not yet beenidentified or isolated, but the deduced amino acidsequence is consistent with the notion that it en-codes a functional metalloproteinase. SL-3 hasthe same principal domain structure as FIB-CL,SL-1, and SL-2 but differs by an insert of 10residues at the autolytic activation site (Basset etal., 1990) (Table 5).

PUMP-1, which lacks the entire pexin-likedomain as well as the hinge region, cleaves awide range of substrates including fibronectin,laminin, casein, gelatin, and proCL. The onlyexisting sequence information comes from cleav-age of the insulin P chain (Woessner and Taplin,1988):

FVNQHLCGSHLVEALYLVCGERGFFYTPKA

PUMP-1 cleaves Ala14-Leu15 and Tyr16-Leu17 inthe middle of the p chain. The enzyme is ex-pressed in gingival fibroblasts (Overall and Sodek,1991a) and has been isolated from the involutingrat uterus and from rectal carcinoma cells

(Woessner and Taplin, 1988; Miyazaki et al.,1990). PUMP-1 is induced by Con A and TPA inhuman fibroblasts (Overall and Sodek, 1990).

E. Expression of Recombinant MMP

Expression of recombinant wild-type or mu-tant forms of MMP has proven to be a powerfultool in analyzing structure-function relationshipsand catalytic properties of the enzymes. SeveralMMP may be expressed in E. coli with minimalrefolding efforts. These include FIB-CL (Windsoret al., 1991), a truncated version of this enzymelacking the pexin domain (mini-CL), PUMP-1(Windsor et al., unpublished), and a truncatedform of SL-1 (mini-SL-1) (Marcy et al, 1991).The full-length forms of SL-1 and of the two GLscan be obtained by expression in E. coli, but theenzymes are not catalytically active, suggestingthat refolding is more involved. Expression isdriven from the bacteriophage T7 promoter byintroduction of the plasmid carrying the cDNAconstruct into E. coli DE3 cells, which contain achromosomal copy of the T7 polymerase geneunder lac control. Induction with isopropyl P-D-thiogalactoside (IPTG) results in T7 RNA poly-merase production, which allows for transcriptionof mRNA and thus translation of the protein. It isinteresting to note that the recombinant enzymesform intact latent constructs. This suggests thatthe latent structure forms spontaneously duringand after synthesis.


Recombinant full-length proSL-1 and proCLhave also been expressed in eukaryotic systemssuch as Cos cells and mouse mammary tumorC127 cells using either the SV40 late promoter orthe mouse metallothionein I promoter with theSV40 early or late polyadenylation site (Murphyet al., 1987; Park et al, 1991). The enzymesbehave essentially as their natural counterparts,except that r-FIB-CL displays lower activity oncollagen than the natural enzyme (Murphy et al,1987; Docherty and Murphy, 1990), an observa-tion also made when FIB-CL is expressed in E.coli (Windsor et al, 1991). Quantin et al (1989)expressed PUMP-1 as a fusion protein with thestaphylococcal protein A IgG-binding domain inCos cells using the plasmid pPROTA containingthe SV40 early gene promoter. This constructpermits easy purification of enzymes by passageof the culture medium over IgG-Sepharose fol-lowed by elution with buffer containingorganomercurials. The recombinant PUMP-1 pro-tein was shown to degrade the same substrates asthe native enzyme and to enhance or activateproFIB-CL/CL. Full-length proSL-1 was recentlyexpressed in Hela cells by coinfection with avaccina virus vector containing the T7 promotorand a T7 RNA polymerase construct (Galazka etal, unpublished data). The recombinant enzymeis fully latent yet catalytically competent. RabbitproCL has been expressed in baby hamster kid-ney (BHK) cells using the mouse metallothioneinI promoter (Brinckerhoff et al., 1990). The re-combinant protein produced in this system, as itsnatural counterpart, required SL-1 for maximalactivation. Devarajan et al. (1991) expressed the52K unglycosylated form of human PMN-CL invitro in a reticulolysate system.

Several studies have been aimed at identify-ing the role of the various domains of MMP bysite-directed mutagenesis (Windsor et al., 1991;Sanchez-Lopez et al, 1988; Park et al, 1991).These studies have shown that the conservedpropeptide sequence PRCGVPDV is instrumen-tal in maintaining the latency of the enzyme andthat the hemopexin-like domain plays a majorrole in determining substrate specificity. "r-Mini-CL", formed by deletion of the pexin-like domainby mutagenesis (rather than autolysis), is inactiveagainst collagen yet has retained essentially fullcatalytic activity against unfolded substrates suchas casein. If disulfide bond formation of the pexin-

like domain is abolished by site-directed mutagen-esis, the activity against casein is only moderatelyaffected but collagen-cleaving activity is lost(Windsor et al, 1991). Collier et al (1992) re-cently made alanine scanning mutations in thefibronectin type II domains of the human 92KGL. Several of these mutations destroyed gelatin-binding properties but did not adversely affect therate of cleavage of gelatin. This observation sug-gests that the gelatin-binding domains are notrequired for catalytic activity against unfoldedcollagen chains.

V. INHIBITION OFMETALLOPROTEINASE ACTIVITY

A. Synthetic Inhibitors

Chelating agents that interact with (or remove)Zn2+ at the active site such as 1,10-phenanthrolineand EDTA are potent inhibitors of MMP but showlittle if any selectivity and are therefore of limitedanalytical or therapeutic potential. Several ap-proaches have been employed to utilize substratespecificity information to generate more selectiveinhibitors based, in general, on synthesis of shortsubstrate analogue peptide sequences linked tosuspected chelating moieties such as hydroxamate,thiol, phosphonamidate, phosphinate, andphosphoramidate groups. The most potent of thesehave reached ICs in the low nanomolar range. ALeu-Phe-Ala thiol derivative synthesized by Grayetal. (1986, 1987):

HS-CH2-(R,S)-CH-[CH2-CH(CH3)2]-CO-L-Phe-L-Ala-NH2

reached IC50 values

sulfone analogs gave less-potent inhibitors.Hydroxamate inhibitors have been produced byJohnson et al. (1987) the best with a K{ of 5 nM:

HONH-CO-CH2-CH(Bu)-CO-NH-CH(Bu)-CO-NH-CH(Me)-COOEt

It is speculated that the terminal hydroxyl groupand the vicinal carbonyl group together form abidentate Zn2+ ligand. Phosphonamidate andphosphoramidate peptide-analogue inhibitorsagainst FIB-CL and PMN-CL, which span thescissile Gly-Leu or Gly-Ile bond and in which theGly carbonyl group is replaced by a P(=O) (OH),have reached K{ values down to 20 to 30 nM(Kortylewicz and Galardy, 1990; Mookhtiar etal, 1987; Galardy etal, 1992). These include thecommercially available:

phthaloyl-Glyp-Ile-Trp-NHBzl

with a Kj of 25 nM with FIB-CL (Kortylewiczand Galardy, 1989). Substitution of the scissilepeptide bond with a phosphinate linkage (PO2-CH2) also gives potent inhibitors such as:

napthoyl-Glyp-cLeu-Trp-NHBzl

with K{ values near 10 nM (Galardy et al, 1992).Tetracyclines and certain synthetic analogueswithout antibiotic activity inhibit PMN-CL withKj in the micromolar range (Golub et al, 1983,1985, 1987). The mechanism of inhibition is notknown, but it is suspected that it depends on thechelating properties of the compounds. Tetracy-clines are considerably less effective against FIB-CL, but the reason for this selectivity is not known.

MMP inhibitor efficacy has mostly been de-termined with isolated enzyme preparations but anumber of studies suggest that these inhibitorsmay have utility also in cell- or organ-based sys-tems and may ultimately have therapeutic poten-tial in intact organisms. A synthetic collagenaseinhibitor, ^-[-^-(benzyloxycarbonyO-amino-1 -(R)-carboxypropyl]L-Leu-Tyr(OMe)-NHMe(CI-1; SC 40827), produced by Searle, inhibits ovula-tion in perfused rat ovaries, whereas its inactive(S) stereoisomer (CI-2; Searle SC40844) does not(Brannstrom et al, 1988). This inhibitor alsoblocks bone resorption in tissue culture (Delaisseet al, 1985). Librach et al (1991) showed that a

hydroxamate inhibitor of Mr 72K GL and Mr 92KGL, HONHCO-CH2CH(i-Bu)CO-Tyr(OMe)-NHMe (Searle SC39026; K. 2 nAf [Reich et al,1988]) blocks trophoblast invasion in vitro,whereas the noninhibitory analog (f-butoxy-Leu-Tyr (OMe)-NHMe) does not.

The studies summarized above have shownthat potent inhibitors of MMP can be producedbased on existing knowledge of substrate speci-ficity coupled with trial-and-error experimenta-tion. The best of these, with low nanomolar Kivalues, are clearly potent enough to hold thera-peutic potential. It has proven more difficult totarget specific enzymes for inhibition while leav-ing others unaffected, although recent studies haveshown promise for this line of investigation. Forexample, Netzel-Arnett et al (1991b) showed thatit is possible to optimize synthetic peptide sub-strates to yield as much as a 200-fold differencein Kcat/KM between two highly homologous en-zymes such as FIB-CL and PMN-CL. This find-ing inspires confidence that it will be possible toconstruct synthetic inhibitors that for all practicalpurposes target only a single MMP.

B. Inhibiting Antibodies

Production of blocking antibodies representsanother attractive approach to the design of effec-tive and specific inhibiting reagents. Affinity-purified polyclonal antibodies as well as mono-clonal antibodies are both highly effective andattain IC50 in the 1 to 10 |ug/ml range (20 to 200nM). The best of these are almost as potent asgood synthetic inhibitors and are frequently highlyspecific and readily discriminate between closelyrelated enzymes such as FIB-CL and PMN-CL(Birkedal-Hansen etal, 1988). The production ofinhibiting monoclonal antibodies has been suc-cessfully accomplished for FIB-CL, Mr 72K GL,and Mr 92K GL (Hoyhtya et al, 1988; Birkedal-Hansen et al, 1988; Hoyhtya et al, 1990; Lyonsetal, 1991a). The inhibition obtained with mono-clonal antibodies to collagenase on collagenoussubstrates is virtually complete (>90%) but simi-lar antibodies against GLs on gelatin substratesappear to be less effective (40 to 60% inhibition)(Hoyhtya et al, 1988; Birkedal-Hansen et al,1992c). Inhibiting antibodies have been shown to


block matrix degradation by live cells in severalmodel systems (Birkedal-Hansen et al, 1989;Hoyhtya et al, 1990; Librach et al, 1991;Gavrilovics et al, 1985; Meikle et al, 1989).

C. a-Macroglobulins

a-Macroglobulins inactivate susceptible pro-teinases by entrapment following cleavage of thebait region (Sottrup-Jensen, 1989). The protein-ase cleaves one or more bonds in the 40 residuebait region and thereby initiates a conformationalchange that leads to entrapment of the proteinase(Sottrup-Jensen et al, 1989). In all a-macroglo-bulins except chicken ovostatin, this conforma-tional change leads to hydrolysis of one internalthiolester bond [-C(=O)-S-] per subunit and togeneration of a highly reactive glutamyl residue(Sottrup-Jensen et al, 1989). The nascent glutamylresidue reacts with a lysyl side chain exposed onthe surface of the attacking proteinase to covalentlycross-link the proteinase to the inhibitor by an-lysyl-y-glutamyl bond.

Pro-Tyr-Gly-Cys-Gly-Glu-Glu-Asn-Met-Vali Is oo

Pro-Tyr-Gly-Cys-Gly-Glu-Glu-Asn-Met-ValI I

SH O OI

NHI

Proteinase

Sottrup-Jensen et al. (1983) identified five lysineresidues located on patches on opposite sides onthe surface of the molecule as targets for e-lysyl-y-glutamyl crosslinks in the ct2M-trypsin com-plex. The trapping mechanism does not block theactive site per se and still permits access of low-molecular-weight substrates. On high-molecular-weight substrates, however, inhibition is com-plete because of steric hindrance. Although cova-lent capture following the initial entrapment ap-pears to be an integral part of the inhibition mecha-nism by a2M, evidence suggests that the covalentcross-linking of the proteinase is not necessary

for inhibition. Chicken ovostatin lacks thethiolesters and therefore fails to covalently cross-link the proteinase yet has retained inhibitoryactivity (Nagase and Harris, 1983). Moreover, itis possible under certain conditions to retain theinhibitory activity after disruption of the thiolestersby treatment with nucleophilic agents such asmethylamine. The entire complement of cleavagesites identified in the oc-macroglobulin bait regionis summarized in Figure 6. No information is yetavailable on the conformation of the inhibitor orits bait region, but the extraordinary susceptibilityto proteolysis suggests that the bait region con-sists of a readily accessible and highly flexiblestretch of peptide that is able to adapt to mostproteinases and present one or more scissile bonds.Bait region cleavages catalyzed by FIB-CL andSL-1 have been analyzed recently (Sottrup-Jensenand Birkedal-Hansen, 1989; Enghild eftf/., 1989).Some of these bonds are quite predictable (suchas the Gly-Leu bond of human oc2M), whereasothers are more surprising (Table 9,10; Figure 6).The proteinase-inhibitor complexes formed withoc2M are rapidly cleared from circulation in theliver by uptake by receptors that recognize thecomplex but not the unreacted inhibitor. Recep-tor-mediated clearing may also ta

37

Documents