Transmembrane-deletion mutants of the membrane-type matrix metalloproteinase-1 process progelatinase A and express intrinsic matrix-degrading activity.

Membrane-type matrix metalloproteinase-1 (MT-MMP-1) has been proposed to play a critical role in regulating the expression of tissue-invasive phenotypes in normal and neoplastic cells by directly or indirectly mediating the activation of progelatinase A. To begin characterizing MT-MMP-1 structure-function relationships, transmembrane-deletion mutants were constructed, and the processing of the zymogens as well as the enzymic activity of the mature proteinases was analyzed. We now demonstrate that pro-MT-MMP-1 mutants are efficiently processed to active proteinases following post-translational endoproteolysis immediately downstream of an Arg-Arg-Lys-Arg basic motif by a proprotein convertase-dependent pathway. The secreted form of active MT-MMP-1 not only displays an N terminus identical with that described for the processed wild-type enzyme at Tyr (Strongin, A. Y., Collier, I., Bannikov, G., Marmer, B. L., Grants, G. A., and Goldberg, G. I.(1995) J. Biol. Chem. 270, 5331-5338), but also directly mediated progelatinase A activation via a two-step proteolytic cascade indistinguishable from that observed with intact cells. Furthermore, although the only function previously ascribed to MT-MMP-1 is its ability to act as a progelatinase A activator, purified transmembrane deletion mutants also expressed proteolytic activities against a wide range of extracellular matrix molecules. Given recent reports that MT-MMP-1 ectodomains may undergo intercellular transfer in vivo (Okada, A., Bellocq, J.-P., Rouyer, N., Chenard, M.-P., Rio, M.-C., Chambon, P., and Basset, P.(1995) Proc. Natl. Acad. Sci. U. S. A. 92, 2730-2734), our data suggest that soluble forms of the proteinase confer recipient cells with the ability to not only process progelatinase A, but also directly degrade extracellular matrix components.

Members of the matrix metalloproteinase (MMP) 1 gene family have been implicated in the physiologic as well as pathologic remodeling of the extracellular matrix (ECM) in events ranging from organogenesis to tumor metastasis (1)(2)(3). Very recently, three new members of this family were discovered by screening cDNA libraries for homologies to conserved regions of the known MMP genes and named the membrane-type matrix metalloproteinases-1, -2, and -3 (MT-MMP-1, -2, and -3; Refs. 4 -6). Based on their predicted amino acid sequences, each of the MT-MMPs, like almost all previously characterized MMPs, contains (i) a candidate leader sequence, (ii) a propeptide region which includes a highly conserved PRCGXPD sequence that helps stabilize the MMP zymogen in a catalytically inactive state, (iii) a zinc-binding catalytic domain, and (iv) a hemopexin-like domain near their respective C termini (4 -7). In addition, in a pattern similar to that described for stromelysin-3, each of the MT-MMPs contains a short amino acid insert sandwiched between their proand catalytic domains that encodes a potential recognition motif for members of the proprotein convertase family (4 -8). Despite their considerable similarity to other MMP family members, 2 however, only the MT-MMPs contain ϳ75-100 amino acid extensions at their C termini, each of which includes a hydrophobic stretch consistent with the presence of a transmembrane (TM) domain (4 -6, 9). Thus, in contradistinction to all other MMPs, the MT-MMPs are expressed as membrane-associated ectoenzymes rather than soluble proteins.
Although little is known with regard to the potential functions of the MT-MMPs, most attention has focused on the ability of MT-MMP-1 as well as MT-MMP-3 to induce the processing of the MMP zymogen, progelatinase A, to its activated form (i.e. [Tyr 81 ]gelatinase A) via a [Leu 38 ]gelatinase A intermediate (4,6,10). Given the ability of activated gelatinase A to cleave a wide range of ECM substrates (including native types I, IV, V, VII, and XI collagen, denatured collagens, elastin, proteoglycans, laminin, and fibronectin) as well as the association of gelatinase A activation with the expression of tissue-invasive phenotypes (1)(2)(3)(11)(12)(13)(14)(15), MT-MMPs have been dubbed as possible "master switches" that control ECM remodeling (16). Nonetheless, the processes that regulate the activation of the MT-MMP zymogens themselves to mature forms remain undefined as does the mechanism by which MT-MMPs mediate progelatinase A activation (4 -6, 9, 10). In large part, further progress in characterizing MT-MMP activities has been hindered by the technical problems associated with isolating and purifying membrane-associated molecules. Given that similar difficulties with other transmembrane enzymes have been negotiated by generating TM-deleted soluble mutants (17)(18)(19), we noted that, with the exception of the extended C-terminal domain, the modular organization of the MT-MMPs is identical with that of the secreted MMPs (7). Hence, two TM-deletion mutants of MT-MMP-1 were constructed by either truncating the molecule (i) immediately upstream of the start site of the TM domain (i.e. MT-MMP-1  ; herein referred to as MT-MMP-1 ⌬A ) or (ii) at the conserved cysteinyl residue found at, or near, the terminus of all hemopexin domain-containing MMPs (i.e. MT-MMP-1 1-508 or MT-MMP-1 ⌬B ). Utilizing these constructs, we now demonstrate that TM-deleted MT-MMP-1 (⌬MT-MMP-1) mutants undergo efficient post-translational endoproteolysis between Arg 111 -Tyr 112 by a proprotein convertasedependent pathway to generate fully active proteinases. Furthermore, the purified ⌬MT-MMP-1 mutants not only activate recombinant progelatinase A directly via a two-step activation cascade identical with that described for the membraneassociated enzyme, but they also express heretofore unsuspected ECM-degrading activities.
Plasmids-Following screening of an HT-1080 cDNA library constructed in the pBK-CMV vector with a 0.45-kilobase canine MT-MMP-1 cDNA isolated from an MDCK cDNA library, 3 a human clone was obtained and sequenced. The predicted amino acid sequence of full-length HT-1080 MT-MMP-1 was identical with that described by Okada et al. (13). To generate pro-MT-MMP-1 ⌬A and pro-MT-MMP-1 ⌬B , a 5Ј primer (ACCATGTCTCCCGCCCCAAGACCCTCCCGT) was paired with the 3Ј primers, TTCAGCTCACCGCCCCGCCGCC or TTCAGAA-GAAGAAGACTGCAAGGCC, respectively, in separate PCR reactions to generate cDNA fragments with the intended truncations at the C terminus (i.e. immediately prior to the TM domain as defined by Cao et al. (9) at Ser 538 or at the conserved Cys residue at position 508 (6, 7); see Fig. 1A). The PCR fragments were then cloned into pCR3-Uni vector (Invitrogen) and characterized by sequencing as described (8). Expression vectors for ␣ 1 -proteinase inhibitor (␣ 1 PI), the Pittsburgh mutant of ␣ 1 PI (␣ 1 PI Pitt ), and a TM-deletion mutant of furin were provided by A. Rehemtulla and R. Kaufman (University of Michigan).
Mutagenesis-Sequential PCR-based mutagenesis was performed as described previously, and the mutagenized fragments were cloned into pCR3-Uni vector and characterized by sequencing (8). The mutagenic primers used are as follows: ATCAAGGCCAATGTTGCAAGGAAGC-GCTACGCC for R108A, GCCAATGTTCGAAGGGCGCGCTACGCCA-TCCAG for K110A, AATGTTCGAAGGAAGGCCTACGCCATCCAG-GGT for R111A, and GTTCGAAGGAAGC GCTTTGCCATCCAGGG-TCTC for Y112F (bold nucleotides indicate the altered codons). MT-MMP-1 ⌬B expression constructs as well as those harboring the above mutations (1 g each) were transiently transfected into COS-7 cells (2 ϫ 10 5 /ml) by LipofectAMINE treatment, and, after a 24-h incubation, an aliquot of the cell-free supernatant (0.02 ml) was analyzed by Western blotting as described below.
Protein Purification-MDCK cells stably expressing MT-MMP-1 ⌬B were incubated in Opti-MEM (Life Technologies Inc.) supplemented with the synthetic MMP inhibitor, BB-94 (0.5 M, British Biotechnology; Ref. 21), to trap the active form of the proteinase as a reversible enzyme-inhibitor complex (20). After 48 h, ϳ2 liters of conditioned media were collected, dialyzed against buffer A (50 mM Tris, pH 7.5, 5 mM CaCl 2 ), and then loaded onto a Q-Sepharose column (1.5 ϫ 10 cm). Bound material was eluted with a NaCl gradient (0 to 1 M), and fractions containing MT-MMP-1 ⌬B (identified by Western blotting) were combined and dialyzed against buffer A. A heparin-Sepharose column (1 ϫ 10 cm) was then loaded with the dialyzed materials and developed with a NaCl gradient (0 to 1 M). Positive fractions were pooled and passed through a gelatin-Sepharose column (1 ϫ 5 cm) followed by gel filtration chromatography (Ultrogel ACA44, 1 ϫ 150 cm) in the absence of BB-94 to regenerate the active proteinase. In selected experiments, MT-MMP-1 ⌬A was purified from a batch culture of transiently transfected MDCK cells as described above.
SDS-PAGE, Western Blots, Zymography, and N-terminal Sequencing-Basic protocols for these techniques have been described (8,20). The MT-MMP antisera were raised in New Zealand rabbit using a bacterially generated recombinant fusion protein between glutathione transferase and MT-MMP 91-243 (20).
Enzymic Reactions-Enzyme assays of MT-MMP-1 ⌬A and MT-MMP-1 ⌬B were performed in buffer A supplemented with 150 mM NaCl at 37°C unless noted otherwise. Matrix substrates (devoid of contaminating progelatinase A activity as determined by gelatin zymography; data not shown) were obtained from Collaborative Research (type I, IV, and V collagens, fibronectin, laminin, vitronectin, and dermatan sulfate proteoglycan). Processing of pro-MT-MMP-1 ⌬B by purified soluble furin (specific activity 500 units/g; Ref. 18) was performed as described (8). Recombinant progelatinase A (purified as described in Ref. 22) was a gift from R. Fridman (Wayne State University, Detroit, MI), recombinant human TIMP-2 was supplied by Amgen, and soluble human furin was provided by R. Fuller, University of Michigan.

RESULTS
Expression of ⌬MT-MMP-1 Mutants and Detection of Enzymic Activity-COS-7 cells transiently transfected with either MT-MMP-1 ⌬A or MT-MMP-1 ⌬B cDNA each secreted a pair of major and minor products that were specifically recognized by polyclonal antibodies to a truncated form of the bacterially expressed protein (Fig. 1B). While the molecular mass of the minor secreted proteins (ϳ64 kDa for MT-MMP-1 ⌬A and ϳ60 kDa for MT-MMP-1 ⌬B ; lanes 2 and 3, respectively) were consistent with those of the predicted proforms of the metalloproteinases, the major soluble species detected with either TMdeletion mutant was a fragment ϳ10 kDa smaller in size. The generation of the major and minor forms was not specific to COS-7 cells since a similar profile was generated with trans-3 D. Pei and S. J. Weiss, unpublished observation. fected MDCK cells. 3 As expected, when cells were transiently transfected with wild-type MT-MMP-1, soluble forms of the enzyme were not detected in the conditioned media (Fig. 1B).
In intact cell systems, MMPs can be recovered in conditioned medium as a mixture of zymogens, processed active enzymes, zymogen-inhibitor complexes, or enzyme-inhibitor complexes (1)(2)(3). In the case of almost all MMP family members, many of these forms can be detected following SDS-PAGE in substrateimpregnated gels (1-3). Thus, serum-free conditioned media from control, MT-MMP-1 ⌬A , MT-MMP-1 ⌬B , or MT-MMP-1 transfected cells were depleted of endogenous gelatinases by gelatin affinity chromatography (⌬MT-MMP-1 does not bind to gelatin; see below) and electrophoresed in gelatin-containing gels. Proteinases were then allowed to renature following the removal of SDS and then incubated overnight at 37°C. As shown in Fig. 1C, supernatants recovered from MT-MMP-1 ⌬Aor MT-MMP-1 ⌬B -transfected cells each revealed the presence of a single band of gelatinolytic activity whose relative mobility matched that of the major form of ⌬MT-MMP-1 detected by Western blotting. Significantly, identical results were obtained when zymograms were performed with either ␤-caseinor -elastin-impregnated gels as well (data not shown). Regardless of substrate used, the band of proteolytic activity attributed to either ⌬MT-MMP-1 mutant was completely inhibited when zymograms were performed in the presence of the MMPspecific inhibitor, BB-94 (Fig. 1C).
Purification of ⌬MT-MMP-1 and Characterization of Zymogen Processing-Because both MT-MMP-1 ⌬A and MT-MMP-1 ⌬B appeared to undergo a similar, if not identical, processing event to generate active proteinases as assessed by zymography, one of the mutants (i.e. MT-MMP-1 ⌬B ) was stably expressed in the MDCK cell line for further analyses. As shown in Fig. 2, serumfree conditioned media from stable transfectants expressing MT-MMP-1 ⌬B were subjected to a combination of gelatin-Sepharose affinity, Q-Sepharose affinity, and heparin-Sepharose affinity chromatography followed by gel filtration chromatography. Utilizing this protocol, a single immunoreactive ϳ50-kDa species was isolated that co-migrated with the band of activity detected by gelatin-zymography (Fig. 2C). Following 10 cycles of N-terminal sequence analysis, the material was identified as MT-MMP-1 with a single start site at Tyr 112 (Fig. 2D). Interestingly, this N terminus not only aligns with that of the active forms of all other MMPs (6), but it is also identical with that reported for the mature form of wild-type MT-MMP-1 recovered from the HT-1080 fibrosarcoma cell line (10). Thus, while almost all MMPs are synthesized and secreted as inactive zymogens, MT-MMP-1 ⌬B (as well as MT-MMP-1 ⌬A ; see below) underwent further processing to its active form.
Like stromelysin-3, the only other membrane of the MMP family to be secreted as a fully processed active proteinase, MT-MMP-1 contains a motif of basic amino acids (i.e. 114 RRKR) immediately upstream of its catalytic domain (see Fig. 3A; Refs. 4 and 8). Recently, the RXKR array in stromelysin-3 (X ϭ nonbasic amino acid) was shown to act as an endoproteolytic processing signal for an intracellular serine proteinase belonging to the proprotein convertase family (8). Because specific proprotein convertases can display varying requirements for basic residues at positions Ϫ1, Ϫ2, and Ϫ4 relative to the scissile bond (i.e. P Ϫ1 , P Ϫ2 , and P Ϫ4 , respectively; Refs. 23 and 24), a potential role for this enzyme class in ⌬MT-MMP-1 processing was initially assessed by successively substituting each basic residue with an Ala moiety in transient transfection assays. As shown in Fig. 3B, each of these substitutions almost completely blocked MT-MMP-1 ⌬B processing (lanes 2-4). In contrast, a Tyr 3 Ala substitution at the less critical P ϩ1 site (23) did not affect MT-MMP-1 ⌬B processing (Fig. 3, lane 5). Given that COS are known to express only two members of the proprotein convertase family that recognize RXKR motifs (i.e. furin and PACE4; Ref. 24), cells were co-transfected with MT-MMP-1 ⌬B and the Pittsburgh mutant of ␣ 1 PI (␣ 1 PI Pitt ), a reactive site variant that inhibits furin (but not PACE4) activity in situ (25,26). Under these conditions, ␣ 1 PI Pitt completely blocked MT-MMP-1 ⌬B processing while wild-type ␣ 1 PI exerted no inhibitory effect (Fig. 3C). MT-MMP-1 ⌬B processing was similarly inhibited by ␣ 1 PI Pitt in MDCK cells (data not shown). These results (together with the demonstration that soluble furin processes pro-⌬MT-MMP-1 to its active form under cellfree conditions; see below) indicate that pro-⌬MT-MMP-1 maturation is regulated by a furin-dependent pathway in an intact cell system.
Activation of Recombinant Progelatinase A by Purified ⌬MT-MMP-1-Current evidence indicates that MT-MMP-1-expressing cells initiate progelatinase A activation via a two-step process that involves an initial cleavage of the Asn 37 -Leu bond followed by an autocatalytic conversion of the Leu 38 intermediate into a 62-kDa active enzyme with an N-terminal Tyr 81 residue (10,27). Nonetheless, the ability of MT-MMP to directly cleave progelatinase A is unclear, and it has been postulated that additional intermediates may be involved (9, 10). Thus, purified active MT-MMP-1 ⌬B was incubated with recombinant progelatinase A and processing monitored by gelatin zymography and N-terminal sequence analysis. Following a 2-h incubation at 37°C, MT-MMP-1 ⌬B initially cleaved progelatinase A (which migrates as a ϳ68-kDa species) into a ϳ64-kDa fragment (Fig. 4A). N-terminal sequence analysis of the ϳ64-kDa gelatinase A fragment yielded the Leu 38 form of the enzyme (Fig. 4A). Subsequently, the 64-kDa form of the enzyme underwent further processing to a 62-kDa product whose N terminus confirmed the generation of [Tyr 81 ]gelatinase A (Fig.  4A). As expected, the ability of MT-MMP-1 ⌬B (as well as MT-MMP-1 ⌬A ; data not shown) to activate progelatinase A was completely blocked by the addition of the MMP inhibitor, TIMP-2 ( Fig. 4A) or BB-94 (data not shown). Interestingly, while earlier studies have demonstrated that membrane-associated forms of MT-MMP-1 can also process progelatinase A into a 42-kDa active species (10,27,28), this product was not detected under the standard conditions employed. However, when either MT-MMP-1 ⌬A or MT-MMP-1 ⌬B were incubated with progelatinase A in a low ionic strength buffer identical with that used previously (10, 27, 28), MT-MMP-1 ⌬A/B -dependent progelatinase A activation was significantly accelerated and the 42-kDa form of gelatinase generated (Fig. 4A, lanes  8 -10).
The ability of mature MT-MMP-1 ⌬A/B to mediate progelatinase A activation is consistent with a model wherein active MT-MMP-1 ⌬A/B directly cleaves the gelatinase zymogen, but the data do not rule out the possibility that the ⌬MT-MMP-1 zymogen only induces progelatinase A to undergo autocatalytic processing to its active form. Hence, purified pro-⌬MT-MMP-1 was isolated (i.e. from cells co-transfected with MT-MMP-1 ⌬B and ␣ 1 PI Pitt ), and its ability to mediate progelatinase A activation was examined. As shown in Fig. 4B, pro-MT-MMP-1 ⌬B was unable to stimulate progelatinase A activation. However, when pro-MT-MMP-1 ⌬B was processed to its active form ex situ with a TM-deleted soluble form of furin (Fig. 4B, lane 2) and then incubated with progelatinase A, the gelatinase zymogen was readily activated (lane 5). Thus, only the processed active form of ⌬MT-MMP-1 is able to mediate progelatinase A activation.
ECM-degrading Activity of ⌬MT-MMP-1-While previous attention has focused solely on the role of MT-MMP-1 in progelatinase A activation (4,9,10,13,27), the ability of ⌬MT-MMP-1 to degrade gelatin, ␤-casein, or -elastin following zymography suggested that the proteinase might express activity against a wider range of targets. Thus, purified MT-MMP-1 ⌬A/B was incubated with either basement membrane-or interstitium-associated ECM molecules, and proteolysis was assessed in the absence or presence of BB-94. As shown in Fig. 5, while MT-MMP-1 ⌬B did not degrade native type I, IV, or V collagens, the enzyme readily proteolyzed gelatin as well as fibronectin, the B chain of laminin, vitronectin, and dermatan sulfate proteoglycan via a BB-94-sensitive process. Similar, if not identical, results were obtained with purified MT-MMP-1 ⌬A (data not shown). Given that none of these substrates were contaminated with detectable quantities of progelatinase A (see "Experimental Procedures"), we conclude that ⌬MT-MMP-1 mutants can express intrinsic matrix-degrading activities.

DISCUSSION
Sequence alignments of the 13 human MMPs that have been characterized to date indicate that amino acids 1-508 of the 538-residue-long extracellular domain of MT-MMP-1 contain all of the major structural elements of the secreted members of this gene family (i.e. a propeptide and catalytic domain as well as a hemopexin-like region that is bounded by a pair of highly conserved cysteinyl residues; Refs. 4 -7). Given that the C termini of virtually all secreted MMPs end at, or extend no more than 8 amino acids beyond, the final cysteinyl residue in the hemopexin domain (7), we reasoned that TM-deletion mutants of MT-MMP-1 that retained this modular organization would encode functional proteinases. Indeed, as demonstrated, regardless of whether soluble mutants of MT-MMP-1 were truncated either at the edge of the TM domain or at the end of the hemopexin domain, the expressed proteins displayed similar, if not identical, activities as assessed by zymography, progelatinase A processing, or substrate specificity.
By itself, our work does not rule out the possibility that the TM or cytosolic domains of MT-MMP-1 convey additional structural information to the processed proteinase (i.e. beyond acting as a membrane anchor). However, while our work was in progress, Cao et al. (9) reported that an MT-MMP-1 chimera generated by exchanging the TM and cytosolic domains of the metalloproteinase with those of the IL-2 receptor functioned normally in terms of its ability to mediate progelatinase A activation (9). Although this result is consistent with our conclusion that the extracellular domain of MT-MMP-1 confers the proteinase with its distinct characteristics, these authors also reported that a TM-deletion mutant encoding residues 1-535 of MT-MMP-1 was unable to process progelatinase A (9). In comparing our experimental approaches, it is important to note that the soluble MT-MMP-1 generated in their study was not isolated nor were its interactions with progelatinase A examined directly (9). Instead, Cao et al. (9) judged their TM-deletion mutant to be inactive on the basis of its inability to process endogenously derived progelatinase A secreted by COS-1 cells in a transient transfection assay system (9). Under these conditions, however, attempts to assess the activity of secreted MT-MMP-1 would be complicated by the presence of cell-derived TIMPs which can interfere with progelatinase A processing by either inhibiting ⌬MT-MMP-1 activity directly, or, in the case of TIMP-2, by binding to the C-terminal domain of the gelatinase zymogen (4,27,28). Indeed, when endogenous levels of TIMP are overwhelmed by co-transfecting COS cells with MT-MMP-1 ⌬A/B and progelatinase A, gelatinase activation can be readily detected in the intact cell system as well as our purified system. 3 Thus, while anchoring a proteinase to the cell membrane might be predicted to more effectively shield an active proteinase from soluble inhibitors (and to perhaps provide a surface more conducive for accelerating processing events) (e.g. Refs. 29 -31), our data demonstrate that the TMdeletion mutants retain the key functional properties of the wild-type enzyme.
As a consequence of our attempts to characterize the activity of ⌬MT-MMP-1, we also discovered that the TM-deletion mutants are capable of undergoing rapid processing to their mature forms. This finding is noteworthy since, as a general rule, MMPs are synthesized and secreted as inactive zymogens (1)(2)(3)7). However, we recently reported that in a fashion similar to that observed for the ⌬MT-MMP-1 mutants, prostromelysin-3 is secreted as an active enzyme following its intracellular processing within the constituitive secretory pathway (8). In this case, activation was dependent upon a decapeptide insert that is sandwiched between the pro-and catalytic domains of stromelysin-3 and encrypted with an extended furin recognition motif (i.e. RXRXKR). At the time that these earlier studies were completed, stromelysin-3 was the only member of the MMPs family known to contain this recognition sequence. However, with the recent cloning of MT-MMP-1, -2, and -3, it is clear that all three of these enzymes contain homologous inserts which include an array of basic residues (i.e. RRKR) that match the recognition motif of the proprotein convertases (i.e. RX(K/R)-R; Refs. 4 -6). 4 Consistent with this prediction, (i) the N terminus of ⌬MT-MMP-1 was located at Tyr 112 on the Cterminal side of the Arg-Arg-Lys-Arg motif, (ii) ⌬MT-MMP-1 processing could be inhibited by either inserting point mutations in the RRKR motif or by co-transfecting cells with the furin-specific inhibitor, a 1 PI Pitt , and (iii) the ⌬MT-MMP-1 zymogen could be processed to its active form ex situ by soluble furin. While we have not yet identified the intracellular/extracellular compartments in which the ⌬MT-MMP-1 zymogen undergoes processing in the intact cell, furin is a membraneassociated endoprotease that not only cycles between the trans-4 Although stromelysin-3 and MT-MMP-1 both contain proprotein convertase recognition motifs, comparisons of their genomic organization and chromosomal localization indicate that the two metalloproteinases are not closely related and belong to separate branches of the phylogenetic tree (D. Pei and S. J. Weiss, unpublished observation). Golgi network and the cell surface, but also undergoes processing to a soluble form that accumulates extracellularly (32)(33)(34). Indeed, the possibility that ⌬MT-MMP-1 may undergo extracellular processing is further supported by our results with the TM-deleted form of soluble furin. Nonetheless, in spite of the fact that furin is the most credible MT-MMP-1 activator identified to date, caution should be exercised in terms of extrapolating processing pathways that are operative for ⌬MT-MMP-1 to the wild-type enzyme. Indeed, in contrast to the results obtained with ⌬MT-MMP-1, we and others have found that COS cells transfected with wild-type MT-MMP-1 route most of the enzyme to the cell surface as the unprocessed zymogen rather than the mature enzyme (4,9,31). 3,5 Utilizing chimeric constructs between stromelysin-3 and wild-type MT-MMP-1, it appears that while the furin recognition motif in either of the secreted metalloproteinases can be processed effectively, the TM domain of MT-MMP-1 appears to "shield" the recipient proteinase from undergoing rapid intracellular processing. 3 The mechanisms responsible for controlling the intracellular and extracellular processing of wild-type MT-MMP-1 require further analysis, but the fact that the active form of the full-length (10) and mutant enzyme display an identical N terminus directly downstream of the proprotein convertaserecognition motif strongly suggests a role for furin or, perhaps, a related proprotein convertase (e.g. PC6; Refs. 24 and 35) in zymogen activation.
In the presence of purified active ⌬MT-MMP-1 (but not its zymogen), progelatinase A was processed to its mature form (i.e. [Tyr 81 ]gelatinase) via the formation of the Leu 38 intermediate. This two-step, TIMP-2-sensitive activation cascade is identical with that previously established for crude preparations of plasma membrane-associated MT-MMP-1 (27) and allows us to conclude that ⌬MT-MMP-1 can initiate the processing event independently of additional co-factors or substrates. Interestingly, the ability of ⌬MT-MMP-1 to directly activate progelatinase A under cell-free conditions contrasts with a recent report by Strongin et al. (10) wherein an MT-MMP-1⅐TIMP-2 complex (rather than MT-MMP-1 alone) was proposed to function as the membrane-associated activator of the gelatinase zymogen. We were unable to reproduce this finding with MT-MMP-1 ⌬A/B , but cannot rule out the possibility that TIMP-2 plays a more complex role on the membrane surface. However, an interpretation of the data presented by Strongin et al. (10) is complicated by the fact that even in the apparent absence of TIMP-2, MT-MMP-1 continued to process progelatinase A to the [Leu 38 ]gelatinase intermediate, but not the final mature form. Thus, it remains possible that TIMP-2 exerts its stimulatory effect by accelerating the inter-or intramolecular autocatalytic conversion of [Leu 38 ]gelatinase to [Tyr 81 ]gelatinase on the cell surface (31) rather than by stimulating MT-MMP-1 activity directly.
To date, the only function ascribed to MT-MMP-1 has been its ability to activate progelatinase A (4,9,10,13,27,28). However, we have demonstrated that purified MT-MMP-1 ⌬A/B can also degrade a number of extracellular matrix components. These data indicate that the ability of MT-MMP-1-transfected cells to express a heightened invasive potential may not necessarily be linked to progelatinase A activation alone (4). Although our studies have employed a soluble form of MT-MMP-1, we believe that the modular organization of MT-MMP-1 strengthens the likelihood that the membrane-tethered form displays a similar substrate specificity. Furthermore, the potential physiologic relevance of ⌬MT-MMP-1 mutants have been heightened by recent findings which suggest that soluble forms of MT-MMP-1 may be generated in vivo (13). Thus, while the MT-MMP-1 antigen has been immunodetected on the surface of cancer cells in vivo (4), RNA in situ hybridization studies have more recently demonstrated that MT-MMP-1 transcripts are confined to the surrounding stromal cells (13). Should MT-MMP-1 undergo solubilization and intercellular transfer in situ (4,13), tumor cells could potentially use the stroma-derived enzyme to assemble a multicatalytic complex on their surface that would not only arm them with the ability to catalyze progelatinase A activation, but also to express an additional repertoire of proteolytic activities. Additional studies will be required to directly compare the soluble and membrane-anchored forms of MT-MMP-1, but the established catalytic activity of the TM-deletion mutants should provide a useful tool for characterizing the enzymic properties of this new family of membrane-anchored MMPs.