The Propeptide Domain of Membrane Type 1 Matrix Metalloproteinase Is Required for Binding of Tissue Inhibitor of Metalloproteinases and for Activation of Pro-gelatinase A*

Activation of secreted latent matrix metalloproteinases (MMPs) is accompanied by cleavage of the N-termi-nal propeptide, thereby liberating the active zinc from binding to the conserved cysteine in the pro-domain. It has been assumed that an analogous mechanism is responsible for the activation of membrane type 1 MMP (MT1-MMP). Using recombinant wild-type MT1-MMP cDNA and mutant cDNAs transfected into COS-1 cells lacking endogenous MT1-MMP, we have examined the function of the propeptide domain of MT1-MMP. MT1-MMP was characterized by immunoblotting, surface biotinylation, gelatin substrate zymography, and 125 I-tis-sue inhibitor of metalloproteinases 2 (TIMP-2) binding. In contrast to wild-type MT1-MMP-transfected COS-1 cells, transfected COS-1 cells containing a deletion of the N-terminal propeptide domain of MT1-MMP or a chimeric construction (substitution of the pro-domain of MT1-MMP with that of collagenase 3) were functionally inactive in terms of binding of 125 I-labeled TIMP-2 to the cell surface and initiating the activation of pro-gelatinase A. These results support the concept that in its native plasma membrane-inserted form, the pro-do-main of MT1-MMP plays an essential sitometer. The gelatinolytic activity was calculated as percentage of degraded fragments versus total gelatin in the incubation Binding of 125 I-Labeled TIMP-2 to Membrane-extracted MT1-MMP— Plasma membrane-enriched fractions (100,000 3 g ) from COS-1 transfected cells were prepared following nitrogen cavitation and differential centrifugation as described previously (10). Membrane proteins were extracted using 0.25% polyoxyethylene ether (W1) in HEPES buffer. 125 I-TIMP-2 (40 n M final concentration) was incubated with membrane-extracted proteins for 1 h at 4 °C to permit binding. To assess the specificity of binding, a similar 125 I-TIMP-2 incubation with membrane-extracted proteins was performed in the presence of 10-fold excess unlabeled TIMP-2. Affinity-purified rabbit polyclonal antibodies to MT1-MMP (catalytic domain) were added to each reaction mixture to immunoprecipitate MT1-MMP and complexes. The mixtures were then incubated with protein A-coated Sepharose beads and then thoroughly washed in buffer. Beads were then added to SDS-PAGE sample buffer containing b -mercaptoethanol. After boiling, samples were subjected to SDS-PAGE followed by autoradiography to identify 125 I-TIMP-2 bound to MT1-MMP as described previously (10). Gelatin Substrate Zymography and Western Blotting— Basic proto-cols for these techniques have been described in our recent papers (14, 22).

Matrix metalloproteinases (MMPs, 1 matrixins) are a large family of neutral zinc endopeptidases, which display homologous structural features consisting of a N-terminal propeptide domain, a zinc-coordinated catalytic domain, and a C-terminal hemopexin-and vitronectin-like domain (1)(2)(3). During the process of activation of secreted latent MMPs in the pericellular and extracellular environment, conformational perturbation or limited proteolysis within the N-terminal propeptide domain causes a change in the molecule that disrupts the unpaired Cys-Zn 2ϩ interaction and frees the Zn 2ϩ to participate in proteolytic cleavage. The modified MMP then attacks the peptide sequence downstream of the PRCGVPD sequence in an autolytic manner and cleaves the propeptide, thus producing a lower molecular weight activated enzyme (popularly described as a cysteine switch or velcro mechanism (4)). Activation of MMPs is inhibited by a family of proteins collectively known as tissue inhibitors of metalloproteinases (TIMPs).
Membrane-type matrix metalloproteinases (MT-MMPs) represent a newly described group of matrixins (5) that are widely but selectively distributed in tissues. As the name implies, MT-MMPs are localized to the plasma membranes of cells by a stretch of hydrophobic amino acids (transmembrane domain; Ref. 6) followed by a short cytoplasmic sequence. MT-MMPs have been the subject of great interest because of their role in activating a secreted MMP, pro-gelatinase A, at the cell surface (5). Expression of MT1-MMP in embryonic tissue (7) and in malignant tumors has been correlated with the degree of activated gelatinase A in these tissues.
The mechanism of pericellular activation of pro-gelatinase A appears to involve the formation of a unique bimolecular complex between TIMP-2 and MT1-MMP (5,8,9). Our recent study (10) documented that the N-terminal domain of TIMP-2 binds avidly to the catalytic domain of MT1-MMP, thereby producing a complex on the cell surface. The C-terminal domain of progelatinase A then binds to the available C-terminal domain of TIMP-2 (stabilization site), forming a trimolecular complex (9). According to this theory, a second MT1-MMP molecule (not complexed to TIMP-2) then attacks a single bond in complexed pro-gelatinase A, which is followed by an autolytic cleavage (11,12), thus resulting in pro-gelatinase A activation. Excess TIMP-2 interferes with this activation mechanism by binding and inhibiting all available MT1-MMP molecules.
A special feature of MT-MMPs, as well as stromelysin-3 (13), is a 10-residue insert immediately preceding the final processing site between the propeptide and catalytic domain, which harbors a paired basic amino acid cleaving enzyme recognition motif (RXKR) (5,6). Unlike stromelysin-3, which is a secreted MMP, latent membrane-bound MT1-MMP does not appear to be attacked at the RRKR sequence by the propeptide convertasedependent pathway (furin) in MT1-MMP transformed COS cells (14), but is attacked and converted to the activated form when secreted as a C-terminal truncated proenzyme (13). Secreted MT-MMPs are of unknown biological significance.
In the current study, we have examined structural-functional relationships within the propeptide domain of MT1-MMP. Whereas the peptide sequence 2 of the N-terminal do-main of MT-MMPs shares more than 70% homology with naturally secreted matrixins between amino acids Met 68 and Asp 97 , the peptide sequence (Ser 34 -Ala 67 ) following the signal peptide of MT-MMPs shares Ͻ10% homology with secreted matrixins. Considerable sequence homology (Ͼ60%) between Ser 34 -Ala 67 of MT1-MMP, MT2-MMP, MT3-MMP, and chicken MT-MMP (5,15,16) suggests that this peptide domain may be required for function of MT-MMPs. To examine this possibility, we have transfected COS-1 cells with human MT1-MMP cDNAs containing mutations within the propeptide domain. Whereas non-membrane-bound matrixins with critical mutations in the conserved PRCGVPD sequence of the N-terminal propeptide have been reported to be secreted as activated enzymes (stromelysin-1, collagenase, and matrilysin) (17)(18)(19)(20), we demonstrated that transfected COS-1 cells expressing a deletion or substitution of the N-terminal propeptide domain of MT1-MMP are functionally inactive in terms of activating progelatinase A and binding TIMP 2. Thus, contrary to observations with secreted matrixins, retention of the propeptide domain of MT1-MMP on the cell surface is required to maintain certain biological functions of the enzyme. Other studies have likewise identified only the 63-kDa latent form of MT1-MMP in cells demonstrating MT1-MMP-induced pro-gelatinase A activation (21).
Construction of Plasmids-MT1-MMP cDNA encoding an open reading frame from amino acid residues Met 1 -Val 582 and mutant MT1-MMP were cloned in a pcDNA3 expression vector using a cytomegalovirus promoter as we have previously described (14). A mutant of MT1-MMP with Arg 108 , Lys 110 , and Arg 111 substituted with alanine was produced as described previously (14). Another mutagenesis strategy used overlap extension mutagenesis using two-step PCR (24). To generate a deletion mutant lacking the entire N-terminal propeptide of MT1-MMP (MT⌬pro), a PCR fragment coding for the signal domain of MT1-MMP (Met 1 -Phe 33 ) (25) was amplified using the MT1-MMP cDNA template with the MT1-MMP forward primer (5Ј to 3Ј, CACGAATTCCGGAC-CATGTCTCCCGCCCCAAGA) and the reverse primer (5Ј to 3Ј, AC-CCTGGATGGCGTAGAAGCTGCTGCTTTG) (bold nucleotides indicate the complementary region), which complemented with the C-terminal MT1-MMP fragment starting from the potential catalytic domain as described below. Another fragment extending from the beginning of the catalytic domain to the intracellular end (Tyr 112 -Val 582 ) was generated by amplifying the MT1-MMP cDNA template with the forward primer (5Ј to 3Ј, TACGCCATCCAGGGTCTCAAATGG), and MT1-MMP reverse primer (5Ј to 3Ј, CACGAATTCTCAGACCTTGTC-CAGCAGGAAAC). Both products were used as templates to generate full-length mutant MT1-MMP lacking the entire propeptide domain by PCR amplification with MT1-MMP forward and reverse primers. The resulting PCR fragment was cloned into the pcDNA3 expression vector. Using the same strategy, a deletion mutation of MT1-MMP lacking part of the N-terminal pro-domain (Ser 34 -Arg 51 ) was generated (MT⌬34 -51) using the reverse primer (5Ј to 3Ј, GAAGCTGCTGCTTTGGGC) and forward primer (5Ј to 3Ј, CAAAGCAGCAGCTTCACCCACACACAGC-GCTCA).
A chimera between collagenase 3 and MT1-MMP (Col-3/MT) was constructed by a two-step PCR using the introduced collagenase-3 fragment encoding the N-terminal signal and propeptide domain (from Met 1 to Arg 95 ) in an MT1-MMP plasmid that lacked the N-terminal signal and propeptide domains (from Met 1 to Arg 92 ). The mutagenic primers were as follows: a T7 primer that recognizes the T7 promoter in pcDNA 3 expression vector (5Ј to 3Ј, AATACGACTCACTATAG) was paired with the reverse primer of collagenase 3 (5Ј to 3Ј, GTCTGGAACACCA-CATCTTGGCTTTTTCAT) containing a complementary region with MT1-MMP to generate the N-terminal portion of the chimera by PCR with collagenase-3 cDNA as a template. A forward primer (5Ј to 3Ј, TGTGGTGTTCCAGACAAG) was paired with a reverse primer of MT1-MMP as described above to generate an MT1-MMP fragment lacking the signal and most of the propeptide domains. The full-length chimeric fragment was amplified by PCR using T7 and reverse primers using collagenase-3 and MT1-MMP fragments as templates. The PCR fragment was then cloned into pcDNA 3 vector. All of the mutants were confirmed by sequencing as described previously (14).
RNA Isolation and Northern blot Hybridization-Total RNA was extracted from COS-1 cells transfected with desired plasmids by guanidine solubilization, phenol-chloroform extraction, and serial precipitation as described previously (14). Approximately 15 g of total RNA was resolved by denaturing gel electrophoresis followed by Northern transfer to nylon membranes (Schleicher & Schuell, Keene, NH). Blots were hybridized to a 32 P-radiolabeled MT1-MMP insert at 42°C as described (22) and analyzed after 6-h exposure with an intensity screen at Ϫ80°C. The amount of the samples applied to the lanes was normalized by ␤-actin RNA.
Cell Surface Binding of TIMP-2-Recombinant TIMP-2 was iodinated to a specific activity of 5 ϫ 10 10 dpm/mg as recently described (26). Binding of 125 I-labeled TIMP-2 to COS-1 cells propagated in 24well dishes was performed in duplicate (10% variation between duplicates). For equilibrium binding experiments, dilutions of 125 I-labeled TIMP-2 (0.25-8.0 nM) in bovine serum albumin/PBS buffer were added to cells in 200 l of serum-free media in the presence or absence of excess unlabeled recombinant TIMP-2. After 3 h of incubation at 4°C, supernatant fluid and washes were collected as the unbound 125 I-TIMP-2 fraction. Cell monolayers were then lysed in 0.1% SDS in 0.5 M NaOH and collected as the bound fraction. Bound and unbound 125 I were measured by gamma counting. The residual radioactivity associated with cells in the nonspecific binding experiment (50-fold excess TIMP-2) was subtracted from the total bound fraction (no unlabeled TIMP-2) to give specific binding. Scatchard plot analysis of binding data used best fit curves (26).
Cell Surface Biotinylation and Immunoprecipitation-COS-1 cells transfected with vector pcDNA3, MT1-MMP, or MT⌬pro cDNA were washed twice with PBS. Sulfo-NHS-LC-biotin (Pierce) at a concentration of 1 mg/ml in PBS was added, and cells were incubated at 4°C with gentle shaking. After 30 min, the cells were washed three times with PBS containing 100 mM glycine and further incubated with the same buffer for 20 min to remove unincorporated biotin. After washing three times with PBS, surface-biotinylated cells were solubilized at 4°C for 30 min in 50 mM Tris-HCl (pH 7.5), 10 mM EDTA, 150 mM NaCl, 1% IGEPAL CA-630 (a nonionic detergent), 0.25 mM dithiothreitol, and protease inhibitors (1 g/ml leupeptin, 1 g/ml aprotonin). Nuclei and intact cells were removed by centrifugation at 10,000 rpm for 5 min. Aliquots of biotin labeled cell extracts were incubated overnight at 4°C with polyclonal anti-MT1-MMP antibodies. Immune complexes were precipitated with protein A-agarose followed by brief centrifugation and washing. Surface-biotinylated immunoprecipitates were resolved by SDS-PAGE (10%) and transferred to nitrocellulose filters. The filters were blocked for 60 min using 3% (w/v) bovine serum albumin/PBS, and then incubated for 60 min in the same buffer containing streptavidin conjugated to horseradish peroxidase. After extensive washes in 20 mM Tris-HCl, pH 7.6, 137 mM NaCl, and 0.1% Tween 20, protein was detected by enhanced chemiluminescense (ECL, Amersham Pharmacia Biotech).

Digestion of 3 H-Labeled Type I Gelatin by Membrane
Fractions-Crude plasma membrane fractions of COS-1 cells transfected with vector (control), wt MT1-MMP, MT⌬pro, and Col-3/MT were prepared following nitrogen cavitation and differential centrifugation as recently described (22). Two g of each membrane fraction was incubated with 10 g of [ 3 H-methyl]type I gelatin at 37°C for 4 -18 h in the presence or absence of 0.5 nM TIMP-2. The reaction was terminated by adding 10ϫ SDS-sample buffer under reducing conditions. The degradation of type I gelatin was quantified by scanning the fluorogram with a laser den-sitometer. The gelatinolytic activity was calculated as the percentage of degraded fragments versus total gelatin in the incubation mixture.
Binding of 125 I-Labeled TIMP-2 to Membrane-extracted MT1-MMP-Plasma membrane-enriched fractions (100,000 ϫ g) from COS-1 transfected cells were prepared following nitrogen cavitation and differential centrifugation as described previously (10). Membrane proteins were extracted using 0.25% polyoxyethylene ether (W1) in HEPES buffer. 125 I-TIMP-2 (40 nM final concentration) was incubated with membraneextracted proteins for 1 h at 4°C to permit binding. To assess the specificity of binding, a similar 125 I-TIMP-2 incubation with membraneextracted proteins was performed in the presence of 10-fold excess unlabeled TIMP-2. Affinity-purified rabbit polyclonal antibodies to MT1-MMP (catalytic domain) were added to each reaction mixture to immunoprecipitate MT1-MMP and complexes. The mixtures were then incubated with protein A-coated Sepharose beads and then thoroughly washed in buffer. Beads were then added to SDS-PAGE sample buffer containing ␤-mercaptoethanol. After boiling, samples were subjected to SDS-PAGE followed by autoradiography to identify 125 I-TIMP-2 bound to MT1-MMP as described previously (10).
Gelatin Substrate Zymography and Western Blotting-Basic protocols for these techniques have been described in our recent papers (14,22).

N-Terminal Pro-domain-deleted MT1-MMP Loses Pro-gelatinase A Activation but Not Gelatinolytic Function-
The expression plasmid encoding cDNA for a deletion mutant of MT1-MMP (MT⌬pro) lacking the entire propeptide domain from Ser 34 to Arg 111 (5,25) was constructed by a PCR-based overlap extension method as described under "Materials and Methods" (Fig. 1). COS-1 cells, which do not produce endogenous MT1-MMP proteins, as demonstrated by Northern blotting (Fig. 2), were used as the recipient for the transfection assay. Substituted mutant MT1-MMP Ala 108 -Ala 110 -Ala 111 (MT ARAA ), which is not attacked by furin (14), was examined in selected experiments. wt MT1-MMP, MT⌬pro, and MT ARAA proteins were expressed in COS-1 cells by transient transfection of plasmids. As demonstrated by Western blotting using an antibody to the catalytic domain of MT1-MMP, expressed mutant proteins migrated predictably based on the truncated length of the propep-tide domain and stained with similar intensity (measured by densitometry); MT⌬pro, lacking the entire N-terminal propeptide, appeared as a protein band of 53 kDa, whereas MTAARA was detected as a 63-kDa protein, identical to that of MT1-MMP (Fig. 3A, left panel). The molecular mass of wt MT1-MMP and its mutants was not unique to COS-1 cells, because the same profile was generated with transfected MDA-MB-436 cells (human breast cancer cell line) and NIH-3T3 cells (data not shown). As we previously showed, the molecular mass of MT1-MMP was not altered by co-transfection of furin cDNA along with wild-type MT1-MMP cDNA into COS-1 cells (14).
To validate the presence of the N-terminal propeptide of MT1-MMP in transfected COS-1 cells, an antibody specific for the pro-domain was used in Western blotting. A specific protein band was not identified in plasma membranes isolated from MT⌬pro-transfected cells, whereas the anticipate 63-kDa protein band was identified in wt MT1-MMP membranes (Fig. 3A,  (Fig. 3B); pro-gelatinase A activation was not induced even when the dose of transfected MT⌬pro plasmid was increased 2-fold (data not shown). In a similar vein, nitrogen-cavitated plasma membranes (adjusted to 1 g protein/sample) isolated from wt MT1-MMP-transfected cells readily activated pro-gelatinase A; membranes isolated from MT⌬pro-transfected COS-1 cells failed to activate pro-gelatinase A (data not shown). In contrast to results achieved with intact plasma membranes, detergent-extracted membranes from MT⌬pro-transfected cells induced pro-gelatinase A activation (72 to Ͼ62 kDa) almost comparable with that achieved with wt MT1-MMP-extracted membranes (Fig. 3C). These data indicate that membrane-bound and membrane-solubilized MT⌬pro differ in their capacity for pro-gelatinase A activation.
The general proteolytic activity of MT⌬pro was compared with wt MT1-MMP by incubating crude plasma membranes isolated from transfected COS-1 cells with [ 3 H]gelatin substrate. As shown in Fig. 4, wt MT1-MMP had somewhat greater [ 3 H]gelatin degrading capacity after 18-h incubation compared with MT⌬pro; after 4 h of incubation, the degradation of [ 3 H]gelatin was 17 and 14%, respectively. Addition of TIMP-2 at the initiation of the incubation with substrate (no preincubation) resulted in 56% inhibition of wt MT1-MMP membrane degradation of gelatin; TIMP-2 had a minimal effect on MT⌬pro membranes. Membranes from vector-transfected cells lacked endogenous gelatinolytic activity. These results indicate that the deletion of the N-terminal propeptide domain of (membrane-bound) MT1-MMP does not impair the proteolytic activity of the enzyme against all substrates but is selective for impairing pro-gelatinase A activation.   (Fig. 6), binding of TIMP-2 to MT⌬pro transfected COS-1 cells was negligible (nonspecific binding exceeded specific binding). 125 I-labeled TIMP-2 binding to MT ARAA -transfected COS-1 cells was not impaired (K d , 3.7 nM; data not shown), further suggesting that furin cleavage of the propeptide domain of MT1-MMP is not required for functional activity of MT1-MMP. These data suggest that an intact N-terminal propeptide domain of membrane-bound MT1-MMP is required for binding of TIMP-2 at the cell surface and subsequent activation of pro-gelatinase A.
Sequence Specificity of the Propeptide Domain of MT1-MMP Required for Function-To address whether the amino acid sequence of the propeptide domain of MT1-MMP is unique in function or whether the pro-domain of other MMPs could function in this capacity, we constructed a substituted mutation in the pro-domain of MT1-MMP. Because the pro-domain of collagenase 3 (which classically is responsible for maintaining latency of this MMP) was determined to share minimal identity (Ͻ30% homology) with that of MT1-MMP, the signal and prodomain of cDNA from Met 1 to Arg 92 (in the conserved sequence PRCGVPD) in MT1-MMP was replaced by the homologous cDNA of collagenase 3 (Met 1 to Arg 95 ) using overlap extension PCR as described under "Materials and Methods" (Fig. 1). To confirm that transfected COS-1 cells synthesize chimeric MT1-MMP protein, Western blotting using an anti-MT1-MMP polyclonal antibody was performed with cell lysates and analyzed by SDS-PAGE followed by the ECL detection system. Compared with wild-type MT1-MMP, Col-3/MT chimeric protein was detected as a 65-kDa band (Fig. 7A). The chimeric protein was then examined for function in terms of pro-gelatinase A activation. As demonstrated by gelatin substrate zymography, Col-3/MT-transfected COS-1 cells (Fig. 7B), intact plasma membranes (data not shown), or membrane-extracted proteins isolated from these cells (Fig. 3C) did not induce pro-gelatinase A activation. To investigate whether substitution of the Nterminal propeptide domain of MT1-MMP with that of collagenase 3 resulted in TIMP-2 binding activity, we performed 125 I-TIMP-2 binding studies on COS-1 cells transfected with Col-3/MT compared with wild-type MT1-MMP cDNA. TIMP-2 binding was not demonstrated in COS-1 cells transfected with Col-3/MT plasmid or pcDNA3 vector alone (data not shown). Thus, we concluded that the TIMP-2 binding effect provided by the pro-domain of MT1-MMP is not replaceable by the prodomain of secreted MMPs.
To further explore the unique property of MT1-MMP, a deletion mutant of MT1-MMP lacking the region from the beginning of pro-domain Ser 34 -Arg 51 (MT⌬34 -51) was constructed. This mutation was expressed in COS-1 cells transiently transfected with the plasmid (Fig. 7A). As shown in Fig. 7B, no pro-gelatinase A activation (gelatin zymography) was noted in MT⌬34 -51-transfected COS-1 cells. These data further strengthen the concept that the N-terminal portion of the propeptide domain of MT1-MMP is necessary for MT1-MMP induced pro-gelatinase A activation.
Binding of 125 I-Labeled TIMP-2 to Membrane-extracted MT1-MMP-Binding of 125 I-TIMP-2 to membrane extracted-MT1-MMP, as depicted by an intense 22-kDa band ( 125 I-TIMP-2⅐MT1-MMP complex dissociates after ␤-mercaptoethanol treatment and heating), was demonstrated with wt MT1-MMPand MT⌬pro-transfected cell proteins (Fig. 8, lanes 2 and 4,  respectively); minimal binding of 125 I-TIMP-2 to vector-trans-fected and Col-3/MT-transfected cell membranes was noted (Fig. 8, lanes 1 and 3). These data are consistent with binding of TIMP-2 to soluble membrane MT1-MMP and MT⌬pro followed by separation of the noncovalent complexes on SDS-PAGE. In the presence of 10-fold excess unlabeled TIMP-2 (competition with 125 I-TIMP-2), a 22-kDa band was no longer detected with wt MT1-MMP-and MT⌬pro-extracted proteins (data not shown). Omission of anti-MT1-MMP antibodies from the experiment resulted in the absence of 125 I-TIMP-2 binding to all transfected cell types (data not shown). DISCUSSION The structure and function of secreted matrixins have been the subject of intense scrutiny for the past decade. The unique feature of MT-MMPs is their insertion into the plasma membrane of cells by a stretch of hydrophobic amino acids followed by a cytoplasmic sequence at the C-terminal tail of the molecule. Based on considerable homology with other members of the matrixin family, it has generally been assumed that other aspects of MT-MMP function, such as cleavage of the N-terminal propeptide domain during activation of the proenzyme, are analogous to secreted MMPs (1,18,19,27,28). Because purified MT-MMPs cannot readily be evaluated in their membranebound state, supportive data for the pro-domain cleavage hypothesis has been derived from experiments using secreted mutant forms of MT1-MMP lacking the transmembrane domain (29 -32).
In the current study examining native membrane-bound MT1-MMP, we have demonstrated that the N-terminal prodomain of MT1-MMP expressed in transfected COS-1 cells is required for function of the intact membrane-bound enzyme (function defined as the activation of pro-gelatinase A and binding of TIMP-2). The presence of an intact N-terminal propeptide in MT1-MMP was documented using a specific antibody to the pro-domain. These data are consistent with our hypothesis that conformational effects induced by the plasma membrane provide functional activity to membrane-bound MT1-MMP without cleavage of the molecule (13). In contrast, using detergent extracts of crude plasma membranes, we have confirmed that the pro-domain of MT1-MMP is not required for function of the soluble form of MT1-MMP.
By transfection of mutant MT1-MMP cDNAs into COS-1 cells, we have demonstrated that deletion of the entire Nterminal propeptide sequence of MT1-MMP resulted in loss of both 125 I-TIMP-2 binding activity and pro-gelatinase A activation function. Loss of function of this mutein was not attributable to a detectable defect in protein synthesis or insertion into the plasma membrane, as studied by cell surface biotinylation (Fig. 5). As anticipated, MT⌬pro reacted with an antibody generated against a peptide sequence contained within the catalytic domain of MT1-MMP but did not react with an antibody against a pro-domain sequence of MT1-MMP. Based on the observation that cell membranes isolated from mutant MT1-MMP (MT⌬pro)-transfected COS-1 cells exhibited gelatinolytic activity almost equivalent to that of wild-type MT1-MMP-transfected cells, it would appear that the mutant protein is properly folded and functional in the cell membrane. The fact that a chimeric mutant containing the N-terminal domain of collagenase 3 and the remainder of intact MT1-MMP (Col-3/MT) inserted into the plasma membrane was unable to bind TIMP-2 and activate pro-gelatinase A in intact COS-1 cells further emphasizes that the propeptide of MT1-MMP plays an essential role in the function of the molecule, presumably related to facilitating binding of TIMP-2. Our observation that TIMP-2 inhibited the gelatin-degrading activity of plasma membranes (nonsolubilized) isolated from wt MT1-MMP, but not membranes isolated from MT⌬pro-transfected COS-1 cells, is consistent with TIMP-2 interacting with the propeptide domain of membrane-bound MT1-MMP. A deletion mutant of MT1-MMP lacking the region from Ser 34 to Arg 51 (MT⌬34 -51) likewise lacked pro-gelatinase A activation function in transfected COS-1 cells. These data are consistent with the concept that the conserved N-terminal amino acid sequence following the signal peptide of MT1-MMP (E 36 WLQ⅐YGYLPP 47 ) is required for proper conformation and function of the enzyme in the plasma membrane. The high degree of homology in this propeptide sequence among MT1-MMP, MT2-MMP, MT3-MMP, and chicken MT-MMP (Ͼ80% identify) and minimal identity with other MMPs is of special interest, because this sequence is unique only for membrane-bound MMPs. Of relevance to potential function of the pro-domain of other MMPs, the N-terminal propeptide of pro-gelatinase A has been reported to contribute to the binding of TIMP-2 (33). Likewise, a conserved stretch of amino acids located in the comparable region of prostromelysin-1 and procollagenase-1 (L 16 VQKYLE 22 ) appears to play a central role in maintaining the latency mechanism of secreted MMPs. The Tyr-Leu sequence of this region is conserved among most secreted matrixins but differs in MT-MMPs (W 38 L 49 ) (34).
Our attempts to further explore the pro-domain of MT1-MMP by expressing other mutations in COS-1 cells have been frustrated by the observation that some of the mutant proteins are not identified in COS-1 cells as intact proteins at the anticipated molecular weight probably because of aberrant cleavages, autolysis, or fragmentation attributable to incorrect protein folding (i.e. single-amino acid deletion of Ser 34 resulted in synthesis of a 40-kDa protein as demonstrated by immunoblotting, and deletion of G 43 YLPP 47 resulted in total lack of expression of the protein; data not shown). Defective protein expression in cells transfected with mutations of the propeptide domain have been described previously with other MMPs (19,34).
Other studies using secreted forms of MT1-MMP (29, 31, 32) and MT2-MMP (28) lacking the transmembrane domain have indicated that cleavage of the N-terminal domain of MT-MMPs (presumably by furin) is required for activation of pro-gelatinase A and binding to TIMP-2. Secreted MT1-MMP is capable of degrading interstitial collagens and extracellular matrix molecules. To resolve differences between these studies and our findings with intact cells, we performed a binding experiment using 125 I-labeled TIMP-2 and detergent-extracted proteins from transfected COS-1 cells. Using extracted membrane proteins rather than intact cells, 125 I-labeled TIMP-2 bound to N-terminal propeptide-deleted MT1-MMP (MT⌬pro) as well as wt MT1-MMP, but not to Col-3/MT or vector-transfected proteins. Likewise, solubilized MT⌬pro, but not membrane-bound MT⌬pro, was capable of activating pro-gelatinase A. These data indicate that soluble forms of MT1-MMP react differently than membrane-bound enzymes; Butler et al. (12) reached the same conclusion in kinetic studies of soluble and membranebound MT1-MMP. Stoichiometric concentrations of TIMP-2 (followed by formation of a triplex among plasma membrane MT1-MMP, TIMP-2, and pro-gelatinase A) do not appear to be required for activation of pro-gelatinase A by (soluble) membrane domain-deleted MT1-MMP (12). As with naturally secreted MMPs, TIMP-2 function is limited to binding and inhibition of soluble pro domain-deleted MT1-MMP (12,30,32).
To further examine the role of the pro-domain of membranebound MT1-MMP, we have recently inserted the cDNA encoding the propeptide sequence of MT1-MMP (MT 1-109 ) in an expression vector. Co-transfection of COS-1 cells with both MT⌬pro cDNA and MT 1-109 cDNA resulted in reconstitution of MT1-MMP function; co-transfected cells activated recombinant pro-gelatinase A. Co-transfection of COS-1 cells with Col-3/MT cDNA and MT 1-109 cDNA or MT⌬pro cDNA and interstitial collagenase 1-99 cDNA (containing signal and propeptide domains) did not reconstitute pro-gelatinase A activation function. 3 These experiments suggest that the isolated pro-domain of MT1-MMP is capable of binding to MT⌬pro in the plasma membrane and thereby reconstituting the function of MT1-MMP.
The issue of whether cells contain non-TIMP-2-related receptors for gelatinase A is disputed (9,(35)(36)(37). Recently Brooks et al. (38) suggested that the ␣v␤3-integrin functions as a cell surface receptor for pro-gelatinase A. Other investigators have been unable to reproduce this observation (12,39). The role of MT-MMP in pro-gelatinase A activation in this scenario remains to be determined.