Membrane type 1 matrix metalloproteinase digests interstitial collagens and other extracellular matrix macromolecules.

Membrane type 1 matrix metalloproteinase (MT1-MMP) is expressed on cancer cell membranes and activates the zymogen of MMP-2 (gelatinase A). We have recently isolated MT1-MMP complexed with tissue inhibitor of metalloproteinases 2 (TIMP-2) and demonstrated that MT1-MMP exhibits gelatinolytic activity by gelatin zymography (Imai, K., Ohuchi, E., Aoki, T., Nomura, H., Fujii, Y., Sato, H., Seiki, M., and Okada, Y. (1996) Cancer Res. 56, 2707-2710). In the present study, we have further purified to homogeneity a deletion mutant of MT1-MMP lacking the transmembrane domain (ΔMT1) and native MT1-MMP secreted from a human breast carcinoma cell line (MDA-MB-231 cells) and examined their substrate specificities. Both proteinases are active, without any treatment for activation, and digest type I (guinea pig), II (bovine), and III (human) collagens into characteristic 3/4 and 1/4 fragments. The cleavage sites of type I collagen are the Gly775-Ile776 bond for α1(I) chains and the Gly775-Leu776 and Gly781-Ile782 bonds for α2(I) chains. ΔMT1 hydrolyzes type I collagen 6.5- or 4-fold more preferentially than type II or III collagen, whereas MMP-1 (tissue collagenase) digests type III collagen more efficiently than the other two collagens. Quantitative analyses of the activity of ΔMT1 and MMP-1 indicate that ΔMT1 is 5-7.1-fold less efficient at cleaving type I collagen. On the other hand, gelatinolytic activity of ΔMT1 is 8-fold higher than that of MMP-1. ΔMT1 also digests cartilage proteoglycan, fibronectin, vitronectin and laminin-1 as well as α1-proteinase inhibitor and α2-macroglobulin. The activity of ΔMT1 on type I collagen is synergistically increased with co-incubation with MMP-2. These results indicate that MT1-MMP is an extracellular matrix-degrading enzyme sharing the substrate specificity with interstitial collagenases, and suggest that MT1-MMP plays a dual role in pathophysiological digestion of extracellular matrix through direct cleavage of the substrates and activation of proMMP-2.

Matrix metalloproteinases (MMPs) 1 are zinc endopeptidases consisted of 14 different members and implicated in the extracellular matrix (ECM) degradation under both physiological and pathological conditions (1). Among the MMPs, MMP-2 (gelatinase A) is reported to be most related to invasion and metastasis in various human cancers (2). All these MMPs except for at least MMP-11 (stromelysin-3) are secreted as inactive zymogens (proMMPs), and thus their activation is one of the most important steps for the regulation of MMP activities. Research on the activation mechanisms of proMMP-2 has greatly progressed in recent years, since membrane type 1 MMP (MT1-MMP) has been cloned as an activator of proMMP-2 (3). The expression of MT1-MMP in human lung and gastric carcinomas is well correlated with the activation of proMMP-2 (4,5), suggesting that the proMMP-2 activation by MT1-MMP is a key step for the cancer cell invasion and metastases. On the other hand, one can expect that MT1-MMP possesses enzymic activity to the ECM macromolecules, since the structure of the catalytic domain of MT1-MMP is similar to that of other MMPs. Actually, we have recently demonstrated that a deletion mutant of MT1-MMP lacking the transmembrane domain (⌬MT1) and native MT1-MMP, both of which were isolated in the complex forms with tissue inhibitor of metalloproteinases 2 (TIMP-2), exhibit gelatinolytic activity after separation from TIMP-2 on gelatin zymography (6). However, information about the substrate specificity of MT1-MMP is still limited, although Pei and Weiss (7) have very recently reported that deletion mutants of MT1-MMP have some ECMdegrading activity.
Cell Cultures and Stable Transfectants of ⌬MT1-The SV40 early promoter of the pSG5 plasmid (Stratagene, La Jolla, CA) was used to express ⌬MT1, which lacks the COOH-terminal transmembrane and cytoplasmic domain of MT1-MMP (⌬Ala 536 -Val 582 ) (18). A cell line constitutively expressing ⌬MT1 was established by two-step selection of CHO cells lacking dihydrofolate reductase gene co-transfected with ⌬MT1 cDNA/pSG5 plasmids and pKG5 plasmid containing neomycin resistance gene and dihydrofolate reductase/pSV2 vector as described previously (6). The established cells were cultured in ␣-minimum Eagle's medium containing 0.2% lactalbumin hydrolysate for 4 days. MDA-MB-231 cells were grown in monolayer cultures in Dulbecco's modified Eagle's medium containing 10% fetal calf serum and treated with 20 g/ml concanavalin A (Wako Chem., Japan) for 4 days in serum-free Dulbecco's modified Eagle's medium containing 0.2% lactalbumin hydrolysate. These culture media were harvested and stored at Ϫ20°C until used for purification.
Purification of ⌬MT1 and MT1-MMP from Their TIMP-2 Complex Forms-We have recently isolated ⌬MT1 complexed with TIMP-2 from stable transfectants in CHO cells and native MT1-MMP⅐TIMP-2 complex from concanavalin A-stimulated MDA-MB-231 cells by a four-step protocol (6). The complexes were further subjected to the anti-MMP-1-IgG-Sepharose column to eliminate the possibility of MMP-1 contamination in the preparations, which was monitored by the sandwich enzyme immunoassay and immunoblotting for MMP-1 (19). ⌬MT1 and native MT1-MMP were finally purified to homogeneity by application of the complexes to anti-TIMP-2-IgG-Sepharose (clone 67-4H11, the antibody against the COOH-terminal tail domain of TIMP-2) (20) equilibrated with 50 mM Tris-HCl, pH 7.5, 0.15 M NaCl, 5 mM CaCl 2 , 0.05% Brij 35, and 0.02% NaN 3 ; ⌬MT1 and MT1-MMP were separated from TIMP-2 with 5 mM and 10 mM EGTA in the CaCl 2 -free buffer, respectively, and TIMP-2 was eluted with 6 M urea in the column buffer. Each fraction for the eluate contained 0.5 mM ZnCl 2 and 10 mM CaCl 2 (at final concentrations) to restore the metal ions immediately after the elution. ⌬MT1 and MT1-MMP were monitored by immunoblotting and gelatin zymography as described below. The combined fractions containing ⌬MT1 or MT1-MMP were dialyzed against 50 mM Tris-HCl, pH 7.5, 0.15 M NaCl, 10 mM CaCl 2 , 0.05% Brij 35 and 0.02% NaN 3 . The purified ⌬MT1 and MT1-MMP were analyzed on SDS-polyacrylamide gel electrophoresis (PAGE).
Immunoblot Analyses and Gelatinolytic Activity-Samples resolved by SDS-PAGE under reduction were transferred onto nitrocellulose filters. The filters were reacted with a monoclonal antibody against MT1-MMP, and protein bands were visualized by avidin-biotin-peroxidase complex method as described previously (8). The antibody specific to the catalytic domain of MT1-MMP (clone 114-1F2) was prepared and characterized previously (3). For detection of gelatinolytic activity, gelatin zymography and gelatinase assay using heat-denatured 14 C-acetylated type I collagen (gelatin) were performed according to the methods described by us (9). Digestion products of the gelatin were also analyzed by SDS-PAGE under reduction (9).
Iodination and Cross-linking Experiments-⌬MT1 and proMMP-2 were iodinated according to the methods by Fraker and Speck (21) and used for the cross-linking experiments and proMMP-2 activation by ⌬MT1. For cross-linking experiments, labeled ⌬MT1 was incubated with TIMP-2 or proMMP-2 in 50 mM Tris-HCl, pH 7.5, 0.15 M NaCl, 10 mM CaCl 2 , 0.05% Brij 35, and 0.02% NaN 3 for 2 h at 23°C, and then the buffer was replaced with 50 mM HEPES-KOH buffer, pH 7.5, 0.15 M NaCl, 5 mM CaCl 2 , and 0.02% Brij 35 by spin columns (6). Freshly prepared cross-linker (bis(sulfosuccinimidyl) substrate) (Pierce) was added to the samples at a final concentration of 4 mM and incubated for 45 min at 23°C. After termination of the reactions by incubation with 50 mM Tris for 15 min on ice, they were subjected to SDS-PAGE (10% total acrylamide) under reduction and the gels were autoradiographed. M r changes of proMMP-2 during activation with ⌬MT1 were also examined by autoradiography of the iodinated proMMP-2 after SDS-PAGE (9% total acrylamide).
Determination of Enzyme Concentrations-Concentrations of MMP-1, MMP-2, and ⌬MT1 were determined by titration of their activities against rTIMP-2 (concentration determined by amino acid analysis) in an assay using Mca-Pro-Leu-Gly-Leu-Dpa-Ala-Arg-NH 2 at 37°C for 1 h (22). Residual enzymic activities were measured and plotted versus TIMP-2 concentrations. A linear plot of activity against the inhibitor molarity was extrapolated to be zero activity at molarity of the enzyme solution.
Digestion of Extracellular Matrix Macromolecules-⌬MT1 was incubated with various ECM components and other protein substrates including carboxymethylated transferrin (23), ␣ 1 -proteinase inhibitor and ␣ 2 -macroglobulin in 50 mM Tris-HCl, pH 7.5, 0.15 M NaCl, 10 mM CaCl 2 , 0.05% Brij 35, and 0.02% NaN 3 at indicated temperatures. The reactions were stopped with 20 mM EDTA and digestion products were analyzed by SDS-PAGE. Specific activity of ⌬MT1 and MMP-1 against type I collagen and gelatin was determined using 14 C-acetylated type I collagen and its heat-denatured collagen (gelatin), respectively. The degradation of acid-soluble type I (guinea pig), II (bovine), and III (human) collagens was quantitated using the gel scanning protocol described by Welgus et al. (24). The activity of ⌬MT1 to cartilage proteoglycan and insoluble elastin were assayed as described previously (17,25). Synergistic effects of ⌬MT1 (or MMP-1) and MMP-2 on type I collagen digestion were assayed using 14 C-acetylated collagen.
Sequence Analyses of ⌬MT1-digested Type I Collagen-Human type I collagen was digested with ⌬MT1 at 27°C, and the fragments generated were separated by SDS-PAGE, and then transferred on polyvinylidine difluoride membranes. The band of 1/4-length fragments was sequenced by 492 sequencer (Applied Biosystem, Foster, CA).

Purification of ⌬MT1 and Native MT1-MMP
Dissociation of ⌬MT1 from ⌬MT1⅐TIMP-2 complex was efficiently performed by a stepwise elution method with EGTA and urea in the anti-TIMP-2-IgG-Sepharose column. ⌬MT1 was recovered in the EGTA eluate, and the final material (80.3 g) purified from the culture medium (500 ml) migrated as a single protein band of M r 56,000 under reducing conditions (M r 52,000 under nonreducing conditions) (Fig. 1A). Absence of MMP-1 contamination was verified by the sandwich enzyme immunoassay and immunoblotting for MMP-1 (data not shown). Native MT1-MMP was also purified from the MDA-MB-231 cell-derived MT1-MMP⅐TIMP-2 complex; the final material (5 g) was purified from 500 ml of the culture medium. Purified MT1-MMP showed a protein band of M r 56,000 under reducing conditions and that of M r 52,000 under nonreduction (Fig. 1A). ⌬MT1 and native MT1-MMP digested type I gelatin and carboxymethylated transferrin into the indistinguishable fragments (Fig. 1B), indicating that the activities of both enzymes are identical. Both proteinases were already active without any treatment and there was no decrease in the activity even after storage at 4°C for 4 months. Since a small amount of native MT1-MMP was purified, ⌬MT1 was used mainly for the following studies.

Reconstitution of ⌬MT1⅐TIMP-2 Complex and Activation of ProMMP-2 by ⌬MT1
To study the interaction of ⌬MT1 with TIMP-2, cross-linking experiments were carried out by incubating iodinated ⌬MT1 with TIMP-2. Radioiodination did not cause any significant changes in the intrinsic properties of the proteinase. When the reaction mixture was analyzed by autoradiography after SDS-PAGE, ⌬MT1 made a complex with TIMP-2 of ϳM r 73,000, while ⌬MT1 alone resulted in a broader band of M r ϳ56,000 (Fig. 2). In agreement with our previous data showing that ⌬MT1⅐TIMP-2 complex forms a trimolecular complex with proMMP-2 through the COOH termini of TIMP-2 and proMMP-2 (6), ⌬MT1 per se did not make a bimolecular complex with proMMP-2 (Fig. 2). Reconstituted ⌬MT1⅐TIMP-2 complex showed no enzymic activity in the assays using either Mca-peptide or [ 14 C]gelatin, as we have previously reported with the original ⌬MT1⅐TIMP-2 complex (6).
The action of ⌬MT1 on the processing of proMMP-2 was examined. Radiolabeled proMMP-2 was incubated with ⌬MT1 in different molar ratios ranging from 1:1 to 1:10, and M r changes of proMMP-2 molecule were analyzed by autoradiography and gelatin zymography after SDS-PAGE. As shown in Fig. 3A, ⌬MT1 processed proMMP-2 of M r 72,000 into the M r 69,000 intermediate species in a dose-dependent manner, while APMA treatment generated fully active form of M r 67,000. On gelatin zymography under nonreduction proMMP-2 of M r 68,000 was processed by ⌬MT1 to the intermediate species of M r 64,000 and active form of M r 62,000, the latter of which showed a very faint proteolytic band (Fig. 3B). This processing of proMMP-2 by ⌬MT1 confirmed the previous data showing that MT1-MMP initially cleaves proMMP-2 to the intermediate form, which is then autocatalytically processed to the fully active form only when higher concentrations of proMMP-2 are present (26).

Degradation of Extracellular Matrix Macromolecules
Digestion of Collagens by ⌬MT1 and MT1-MMP-⌬MT1 cleaved type I, II, and III collagens under the nondenaturing conditions, i.e. at 27°C, generating 3/4-and 1/4-length fragments of these collagens (Fig. 4A). On the other hand, the collagens were degraded into multiple fragments when incubated at 35-37°C, probably because of thermal denaturation of the substrates to gelatins (data not shown). Since the digestion products were similar to those obtained by the action of MMP-1, NH 2 -terminal sequence analyses on the 1/4 fragments of type I collagen were performed. The NH 2 terminus of ␣1(I) chains was Ile 776 -Ala-Gly-Gln-X-Gly-Val-Val-Gly-Leu and ␣2(I) chains had NH 2 -terminal Leu 776 -Leu-Gly-Ala-Hyp-Gly-Ile and Ile 782 -Leu-Gly-Leu-Hyp-Gly-Ser in approximately 1:2 molar ratio. Native MT1-MMP also digested type I, II, and III collagens into typical 3/4 and 1/4 fragments (Fig. 4B). However, no degradation of type IV, V, and VI collagens by ⌬MT1 and MT1-MMP were observed at the nondenaturing temperatures under 33°C, whereas type IV and V collagens, but not type VI collagen, were digested into fragments at 35 and 37°C (data not shown).
The catalytic efficiency of type I, II, and III collagens by ⌬MT1 and MMP-1 was estimated by incubation of the collagens with increasing concentrations of the proteinases. ⌬MT1 most preferentially digested type I collagen; the susceptibility of type I collagen was 6.5-and 4-fold higher than that of type II and III collagens, respectively. On the other hand, the activity of MMP-1 to type III collagen was approximately 4.4-and 25.6fold greater than that to type I and II collagens, respectively. Kinetic analyses of the type I collagen degradation by MMP-1 and ⌬MT1 were performed in the samples containing increasing amounts of type I collagen and constant amounts of the proteinases. Lineweaver-Burk plots were constructed from the velocity data, and values of K m and k cat were extracted (Table  I) (2.4 M Ϫ1 h Ϫ1 ) for type I collagen was ϳ7.1-fold less than that of MMP-1 (17.1 M Ϫ1 h Ϫ1 ). Consistent with the data, specific activity of ⌬MT1 (6.4 g/min/nmol) determined by a solution assay using 14 C-acetylated type I collagen was approximately 5-fold less than that of MMP-1.
Degradation of Other Substrates-Type I gelatin was readily digested into multiple smaller fragments with ⌬MT1, and the ␤1, 2(I) chains and ␣2(I) chains were preferentially degraded compared to ␣1(I) chains (Fig. 5A). In an assay using 14 Clabeled type I gelatin, specific activity of ⌬MT1 was 6.3 g of gelatin degraded/min/nmol of enzyme at 37°C, which was 8-fold higher than that of MMP-1, while that of MMP-2 was approximately 80-fold higher. Cartilage proteoglycan was also digested by ⌬MT1 with an activity of 24.8 g of proteoglycan degraded/h/nmol enzyme at 37°C. ⌬MT1 degraded fibronectin into five major fragments of M r 178,000, 144,000, 123,000, 115,000, and 89,000 (under reduction) (Fig. 5B). Vitronectin was also degraded into two fragments with M r 41,000 and 40,000 (Fig. 5C). The ␣ chain of laminin-1 was slightly hydrolyzed by ⌬MT1 (Fig. 5D), but decorin and insoluble elastin were not substrates of ⌬MT1. ⌬MT1 also processed ␣ 1 -proteinase inhibitor to a major fragment of M r 52,500 and ␣ 2 -macroglobulin to those of M r 94,000 and 90,000, the digestion products similar to those obtained with MMP-1 (data not shown).

Synergistic Effects of ⌬MT1 and MMP-2 on Fibrillar Collagen Digestion
To assess the synergistic effect of ⌬MT1 and MMP-2 on type I collagen digestion, 14 C-labeled type I collagen was incubated at 35°C with ⌬MT1 (16 nM) in the presence of various amounts of MMP-2 ranging from 0 to 16 nM. As shown in Table II, the collagenolytic activity of ⌬MT1 was augmented up to 7.3-fold compared with ⌬MT1 alone, although MMP-2 itself showed no collagenolytic activity. Similar experiments were performed with MMP-1 (3 nM) and MMP-2 ranging from 0 to 16 nM. The activity of MMP-1 was also 6.1-fold enhanced in the presence of MMP-2. This effect was ascribed to the accelerated degradation of the collagenolytic fragments (gelatin) by MMP-2, since the collagen digestion products generated by ⌬MT1 treatment were completely hydrolyzed into peptides in the presence of MMP-2 (Fig. 6A). Similar data was obtained with type II and III collagens incubated with ⌬MT1 and MMP-2 (Fig. 6, B and C). On the other hand, no such effect was found with ⌬MT1 and MMP-1; the collagenolytic activity in the presence of ⌬MT1 and MMP-1 was equal to the sum of both proteinase activities.  (A and B, lanes 1, 3, and 5), with 500 ng of ⌬MT1 (A, lanes 2, 4, and 6) or with 100 ng of MT1-MMP (B, lanes 2, 4, and 6) at 27°C for 24 h. The digestion products were analyzed by SDS-PAGE under reduction after termination of the reaction with 20 mM EDTA. The gels contain 10% total acrylamide except for that for type II collagen digestion with ⌬MT1 (A, lanes 3 and 4), which contains 12.5% total acrylamide. Note appearance of characteristic 3/4 fragments of each collagen and 1/4 fragments of type I collagen in the samples incubated with the proteinases. ␣ and ␤ chains of each collagen are indicated.  5. Degradation of type I gelatin (A), fibronectin (B), vitronectin (C), and laminin-1 (D) with ⌬MT1. A, type I gelatin (15 g) was incubated with ⌬MT1 (500 ng) at 37°C and the reaction products were analyzed on SDS-PAGE (7% total acrylamide) under the reducing conditions. Lane 1, the substrate incubated with buffer alone for 24 h; lanes 2-5, the substrate incubated with ⌬MT1 for 2, 4, 8, and 24 h, respectively. B, fibronectin (10 g) was digested with ⌬MT1 (330 ng) at 37°C for 24 h, and the products were analyzed on SDS-PAGE (6% total acrylamide) under the reducing conditions. Lanes 1 and 2, the substrate incubated with buffer alone and ⌬MT1, respectively. C, vitronectin (5.3 g) was incubated with ⌬MT1 (180 ng) at 37°C for 24 h, and the products were analyzed on SDS-PAGE (12.5% total acrylamide) under the reducing conditions. Lanes 1 and 2 are as in B. D, laminin-1 (15 g) was incubated with ⌬MT1 (180 ng) at 37°C for 24 h, and the products were subjected to SDS-PAGE (6% total acrylamide) under the reducing conditions. Lanes 1 and 2 are as in B.
column chromatography. Supplementation of CaCl 2 and ZnCl 2 to the column fractions restored the structural integrity of the proteinases, since they reconstitute the complex with TIMP-2 and retain stable activities during storage. Although ⌬MT1 appears to form a stable complex with TIMP-2 through the binding of the catalytic domain of ⌬MT1 with the inhibitor domain of TIMP-2 as demonstrated with the original ⌬MT1⅐TIMP-2 complex (6), the present data do not exclude the possibility that other domains are also involved during the complex formation. Unlike other most MMPs, such as MMP-1, both ⌬MT1 and MT1-MMP were secreted into the culture media in active forms. Our recent study (6) suggested that the furin-recognition site, a unique insertion of the RRKR amino acid sequence between the propeptide and catalytic domains of MT1-MMP, is essential to the intracellular processing of ⌬MT1. Indeed, Pei and Weiss (7) have very recently demonstrated that furin is responsible for the NH 2 -terminal processing of MT1-MMP. Since MMP-11, which possesses an insertion containing the RQKR sequence, is also processed to an active form by furin (27), it seems likely that all the members with the R(R/Q)(K/R)R sequences of the MMP gene family, i.e. MT-MMPs and MMP-11, are intracellularly processed to active forms.
Fibrillar collagen types I, II, and III are resistant to many animal proteinases because of their triple helical structures, and they are cleaved only by interstitial collagenases, MMP-1, -8, and -13. CHO cells and MDA-MB-231 cells produced neither MMP-8 nor MMP-13, at least at the protein level determined by immunoblotting. 2 Thus, contamination of MMP-1 in the purified preparations was carefully ruled out since it is secreted by both cell lines. During the purification steps, proMMP-1 was eliminated by using anti-MMP-1-IgG-Sepharose and anti-TIMP-2-IgG-Sepharose column chromatographies, and no contamination in the preparations was verified by sandwich enzyme immunoassay and immunoblotting for MMP-1. This was also supported by the data that the final products have a single protein band of M r 56,000 and were already active without any treatment for activation. In addition, the collagenolytic activity of ⌬MT1 was different from that of MMP-1. ⌬MT1 preferentially cleaved type I collagen over type II and III collagens, whereas MMP-1 cleaved preferentially type III collagen over type I and II collagens. Comparison of the type I collagenolytic activity of ⌬MT1 with that of MMP-1 revealed that ⌬MT1 is 5-ϳ7.1-fold less efficient than MMP-1. In contrast, gelatinolytic activity of ⌬MT1 was 8-fold higher than MMP-1. Knä uper et al. (28) recently reported that three collagenases have distinct substrate specificity to collagens and gelatins; MMP-1, -8, and -13 preferentially digest collagen types III, I, and II, respectively, and both MMP-8 and MMP-13 exhibit 4.9-and 41-fold higher activity against type I gelatin than does MMP-1. Thus, the present data suggest that ⌬MT1 shares the proteolytic characteristics with MMP-8. Higher gelatinolytic efficiency of MMP-8, MMP-13, and gelatinases (MMP-2 and -9) is explained by the presence of key residues specifically conserved in the active sites of these MMPs (28). ⌬MT1 also conserves the residues including Ile in the SЈ 1pocket, negatively charged Glu just preceding the third His residue and invariant Pro three amino acids after the His residue.
It has been established that MMP-1 and MMP-8 cleave type I, II, and III collagens at a specific single site after the Gly residue of the partial sequences Gly-(Ile or Leu)-(Ala or Leu) located approximately 3/4 from the NH 2 terminus in these collagens. Unlike MMP-1 and MMP-8, however, MMP-13 hydrolyzes ␣ chains of type II collagen at the Gly 906 -Leu 907 and Gly 909 -Gln 910 bonds. The present study also demonstrated that ⌬MT1 cleaves the Gly 775 -Leu 776 and Gly 781 -Ile 782 bonds of ␣2(I) chains. Since ␣1(I) chains were hydrolyzed only at the Gly 775 -Leu 776 bond, cleavage of the Gly 781 -Ile 782 bond may be a secondary cleavage. This two-site cleavage by ⌬MT1 may be related with higher gelatinolytic activity of this enzyme, but its biological function remains unclear at the present time. Pei and Weiss (7) have reported that a deletion mutant of MT1-MMP (⌬Pro 509 -Val 582 ) (MT1-MMP 1-508 ) digests several ECM macromolecules including gelatin, fibronectin, laminin, vitronectin, and dermatan sulfate proteoglycan. However, MT1-MMP 1-508 had no ability to cleave type I collagen. The structural difference between ⌬MT1 in the present study and MT1-MMP 1-508 is that ⌬MT1 is longer with 27 amino acid residues in its COOH terminus. Since the substrate specificity of ⌬MT1 is almost identical to that of MT1-MMP 1-508 except for the activity to fibrillar collagens, it seems likely that the COOH-terminal sequence of the 27 amino acid residues is essential to the collagenolytic activity probably because the sequence is necessary for the intact conformation of the hemopexin-like domain of MT1-MMP, which may interact with the collagen molecules. The present data that MT1-MMP derived from MDA-MB-231 cells also possesses collagenolytic activity indicate that collagenase activity of MT1-MMP is not artificial, but natural.
Our previous study (6) demonstrated that active ⌬MT1 has Ala 113 at the NH 2 terminus, indicating that the Tyr 112 is lost during the intracellular activation. In MMP-1 and MMP-8, the 2 E. Ohuchi K. Imai, and Y. Okada, unpublished data.  Phe at their NH 2 termini, which corresponds to Tyr 112 of ⌬MT1, is essential to keep the conformational integrity and express their full activity. Lack of the Phe in these species generated by APMA activation results in only partial collagenase activity of MMP-1 and MMP-8 (29,30). Thus, it may be possible that the Tyr 112 -⌬MT1 would have higher activity against the fibrillar collagens than the Ala 113 -⌬MT1. When ⌬MT1 purified from cDNA-transfected Escherichia coli was treated with furin, Tyr 112 -⌬MT1 was obtained by the cleavage at the Arg 111 -Tyr 112 bond. 3 Comparative study on the collagenolytic activities of Tyr 112 -⌬MT1 and Ala 113 -⌬MT1 is under way in our laboratory.
The degradation of the fibrillar collagens is considered to be sequentially performed by gelatinases after the initial cleavage at the collagen triple helix by collagenases (1). A synergistic effect of MMP-1 and MMP-2 on the fibrillar collagen digestion has been reported (31) and further confirmed in the present study. Similar accelerated digestion of the collagens was demonstrated with ⌬MT1 and MMP-2. The combination of MT1-MMP and MMP-2 may be crucial for the pericellular collagen degradation in cancer invasion and metastasis, because MT1-MMP can activate proMMP-2 on the carcinoma cell surfaces where both MMPs may act in concert. Indeed, we have demonstrated that MMP-2 is localized on the cell membranes of the MT1-MMP expressing carcinoma cells in human stomach cancers (5). It is also notable that ⌬MT1 digests aggrecan, even if the specific activity in vitro is approximately one-third of MMP-3 activity. The initial event in osteoarthritic cartilage is depletion of aggrecan from the articular cartilage, leading to loss of tensile strength of the tissue. So-called "aggrecanase" which clips the aggrecan molecules at the Glu 373 -Ala 374 bond is reported to be a key enzyme for the cartilage degradation (32). Although aggrecanase is thought to be a metalloproteinase (33), none of MMPs except for MMP-8 cleaves the Glu 373 -Ala 374 bond. MMP-8 can digest aggrecan molecules at the aggrecanase-site only when the enzyme is incubated with the substrate in a very high concentration (34). Previous studies (35) have demonstrated that the aggrecanase activity is a cellmediated event, suggesting the pericellular proteolysis of aggrecan. In fact, our preliminary studies showed that MT1-MMP is highly co-expressed with MMP-2 in human osteoarthritic chondrocytes. 4 It is, therefore, reasonable to speculate that aggrecanase may be MT1-MMP or combined action of MT1-MMP and MMP-2. This possibility should be elucidated by further studies.