Membrane-type 1 matrix metalloproteinase-mediated progelatinase A activation in non-tumorigenic and tumorigenic humaneratinocytes

Elevated expression of type IV collagenases (MMP-2 and MMP-9) has been strongly correlated with tumour progression and metastasis in various tumours. Here, we analysed expression and activation of these MMPs in non-tumourigenic HaCaT cells and the malignant HaCaT variant II-4 rt. In monolayer cultures, both cell types secreted latent MMP-2 (proMMP-2) in comparable amounts, while MMP-9 production was clearly higher in II-4 rt cells. Upon contact with fibrillar collagen type I the malignant II-4 rt cells, but not the HaCaT cells, gained the capability to activate proMMP-2. This process is shown to be membrane-associated and mediated by MT1-MMP. Surprisingly, all membrane preparations from either HaCaT cells or II-4 rt cells grown as monolayers, as well as within collagen gels, contained considerable amounts of active MT1-MMP. However, within collagen gels HaCaT cells showed significantly higher TIMP-2 levels compared to II-4 rt cells. This indicates that TIMP-2 might play a central role for MT1-MMP-mediated gelatinolytic activity. Indeed, collagen type I-induced MT1-MMP-mediated proMMP-2 activation by II-4 rt membranes could be completely abolished by an excess of TIMP-2. In conclusion, our data suggest that MT1-MMP-mediated proMMP-2 activation might be associated with malignant progression of epidermal tumour cells. © 2000 Cancer Research Campaign

Local degradation of connective tissue in the vicinity of the cell surface is thought to play an essential role for tumour cell invasion and metastasis. Degradation of extracellular matrix components is accomplished through an array of proteolytic enzymes including members of both the matrix metalloproteinase (MMP) and serine protease families (Birkedal-Hansen et al, 1993;Stetler-Stevenson et al, 1993;Yu et al, 1997).
Among MMPs, the type IV collagenases MMP-2 and MMP-9 are considered to play a critical role in tumour progression. Elevated expression and activation of MMP-2 and MMP-9 have been strongly correlated with the invasive phenotype of tumours like squamous cell carcinomas (Pyke et al, 1992;Kusukawa et al, 1993), gastric , and breast (Okada et al, 1995) carcinomas.
Like most MMPs, MMP-2 (gelatinase A, 72-kDa type IV collagenase) is secreted as an inactive zymogen that has to be activated by removal of the N-terminal propeptide. In vitro, serine proteases have been shown to activate most MMPs (Birkedal-Hansen et al, 1993). However, MMP-2 is unique among the MMPs in that its activation is achieved in a membrane-associated manner by membrane-type matrix metalloproteinases (MT-MMPs), such as MT1-MMP (Sato et al, 1994).  is also synthesized as a proform which can be activated both intracellularly by the serine protease furin (Pei and Weiss, 1996), and extracellularly by plasmin (Okumura et al, 1997) and urokinase-type plasminogen activator (Kazes et al, 1998). Although five different membrane-bound MMPs have been described thus far (Sato et al, 1994;Takino et al, 1995;Will and Hinzmann, 1995;Puente et al, 1996;Butler et al, 1997;Pei, 1999), among them MT2-, MT3-, and MT5-MMP have been reported to activate proMMP-2, MT1-MMP is the one with the best documented correlation to the invasive phenotype of different types of cancer (Okada et al, 1995;Tsunezuka et al, 1996;Gilles et al, 1997;Nakamura et al, 1999). MT1-MMP production or activation has been shown in vitro to be up-regulated by fibrillar type I collagen (Azzam and Thompson, 1992;Seltzer et al, 1994;Haas et al, 1998;Kurschat et al, 1999), an interaction which is likely to occur in vivo when tumour cells invade the surrounding stroma. The activation of proMMP-2 by MT1-MMP has been reported to be dependent on the presence of low amounts of the tissue inhibitor of matrix metalloproteinases (TIMP)-2, which is required as a bridging molecule for the formation of a membrane-bound activation complex composed of MT1-MMP, TIMP-2 and latent MMP-2 (Strongin et al, 1995;Butler et al, 1998). At high concentrations, however, TIMP-2 has been shown to inhibit proMMP-2 activation, presumably by blocking the activity of MT1-MMP Itoh et al, 1998;Zucker et al, 1998).
The spontaneously immortalized non-tumorigenic human keratinocyte cell line HaCaT (Boukamp et al, 1988) and the c-ras Ha transfected HaCaT clones (Boukamp et al, 1990) provide an excellent model to study cellular changes associated with malignant progression of epithelial cells. While HaCaT cells reveal a nearly normal epidermal phenotype upon transplantation onto athymic nude mice (Breitkreutz et al, 1998), the malignant variants of the c-ras clones grow invasively and form squamous Membrane-type 1 matrix metalloproteinase-mediated progelatinase A activation in non-tumorigenic and tumorigenic human keratinocytes cell carcinoma-like tumours showing severe alterations at the cell-matrix interface (Tomakidi et al, 1999). In the present study, we analysed MMP-2 and MMP-9 expression in HaCaT cells and the c-ras Ha transfected HaCaT clone II-4 rt to determine whether alterations in the production of these MMPs are associated with acquisition of the malignant phenotype.

Cells and and culture conditions
HaCaT cells, a non-tumorigenic keratinocyte cell line and the cras Ha transfected tumourigenic HaCaT cell clone II-4 rt were kindly provided by Dr N Fusenig (German Cancer Research Center, Heidelberg, Germany). Both cell lines were routinely cultured in Dulbecco's modified Eagle medium (DMEM) supplemented with 10% fetal calf serum (FCS), 2 mM glutamine and 100 U ml -1 each of penicillin and streptomycin. Three-dimensional collagen type I lattices were prepared as described previously (Mauch et al, 1988). Briefly, 4 × 10 5 cells ml -1 were seeded into collagen gels containing 2.4 mg ml -1 bovine collagen type I (Cellon, Cell Systems, Germany).

Preparation of conditioned media
Cells were cultured as monolayers on collagen type I-coated dishes (10 µg cm -2 ) or within collagen gels with daily medium changes. At indicated time-points the cultures were washed with phosphate buffered saline (PBS) and the medium replaced by serum-free DMEM. 24 h later the media were collected, the cells trypsinized, or released from the collagen gels by treatment with bacterial collagenase D (1 mg ml -1 ) and counted. Alternatively, conditioned media and cells grown as monolayers or in collagen gels were combined and homogenized by sonication. For inhibition experiments the cell suspensions were preincubated for 30 min with the synthetic furin inhibitor Decanoyl-Arg-Val-Lys-Argchloromethylketone (CMK; Bachem Biochemicals, Germany), with the CMK solvent methanol, with aprotinin (ROTH, Germany), or with 1,10 phenanthroline (Sigma, Germany) and then seeded into collagen gels. The inhibitors were added at the following concentrations: CMK = 25 and 50 µM; aprotinin = 10 µM; 1,10 phenanthroline = 10 and 20 µM. The peptides UKI and ni68 (20 µg ml -1 ) inhibiting uPA and interaction of uPA with its receptor, respectively, were also included. Both inhibitors were kindly provided by Dr V Magdolen (Department of Gynaecology, Technical University, Klinikum rechts der Isar, Munich, Germany). After 24 h the cultures were washed with PBS and further cultivated for 24 h in serum-free DMEM containing the inhibitor. Then the culture samples were processed as described before.

Gelatin zymography
For gelatin zymography supernatants corresponding to 1.5 × 10 4 cells or 10 µl homogenates of combined media and cells were separated on non-reducing 10% SDS polyacrylamide gels containing 1 mg ml -1 bovine gelatin (Sigma, Germany) (Herron et al, 1986). After electrophoresis the gels were washed in 2.5% Triton X-100 for 30 min and then incubated in enzyme substrate buffer (50 mM Tris-HCl, pH 8.5, 5 mM CaCl 2 ) overnight at 37°C. Gels were stained with Coomassie Blue R250 and destained in water.

Preparation of crude membranes
For preparation of crude plasma membranes the collagen gels were disrupted by mechanical shearing through a 10 ml syringe and subsequent incubation with bacterial collagenase D (1 mg ml -1 ) for approximately 10 min until no visible collagen fibres were left. After addition of half a volume of FCS the cells were pelleted (2000 g for 5 min at 4°C) and resuspended in PBS containing 1 mM Pefabloc (Boehringer Mannheim, Germany), 10 µg ml -1 aprotinin and 1 µg ml -1 leupeptin (Sigma, Germany). After cell lysis by three cycles of freezing in liquid nitrogen and thawing, the lysate was homogenized by sonification. The plasma membranes were pelleted by centrifugation (150 000 g for 30 min at 4°C) and resuspended in PBS. The protein concentration was determined using a commercial assay (Bio-Rad, Germany). For activity assays 10 µg of the membrane preparations were incubated overnight at 37°C with 20 µl fibroblast-conditioned medium containing proMMP-2. Incubations were performed in the presence of 1, 5 or 10 µg of anti-MT1-MMP peptide antibodies blocking specifically MT1-MMP-mediated proMMP-2 activation (raised against a peptide corresponding to the residues 160-173 of human MT1-MMP, clone 114-1F2, Fuji Chemicals, Japan) or in the presence of 3.8 or 7.6 nM recombinant TIMP-2 (Calbiochem-Novabiochem, Germany). 5 µl of the samples were analysed by gelatin zymography.

Northern blot analysis
Total RNA was isolated from cells grown as monolayers or within collagen gels using RNAzol™B according to the manufacturer's instruction (WAK-Chemie, Germany). 10 µg of total RNA were separated in 1% formaldehyde/agarose gels and blotted onto nylon membranes (Amersham-Pharmacia-Biotech, Germany). Filters were hybridized with 32 P-labelled cDNA probes for MT1-MMP (Sato et al, 1994), MMP-2 (Collier et al, 1988), and TIMP-2 (Stetler-Stevenson et al, 1990). As a control for equal loading and the integrity of the RNA the filters were rehybridized with a 32 P-labelled 18S rRNA oligonucleotide (Carlson et al, 1993).

Immunochemical analysis
For immunoblots crude plasma membrane preparations (20 µg) were separated on 10% SDS polyacrylamide gels under reducing conditions and transferred onto nitrocellulose membranes (Amersham-Pharmacia-Biotech). After blockage with 5% nonfat milk powder in PBS/0.5% (v/v) Tween the membranes were incubated overnight with 1 µg ml -1 of the anti-MT1-MMP peptide antibody 114-1F2 at 4°C, followed by an incubation with a horseradish peroxidase conjugated anti-mouse IgG antibody (Dako, Germany) for 1 h. Bound antibodies were detected with ECL according to the supplier's protocol (Amersham-Pharmacia-Biotech). For quantification of TIMP-2 protein levels conditioned media combined with sonicated cells or collagen gels (100 µl) were subjected to ELISA multiwell plates, precoated with anti-TIMP-2 antibodies, and measured following the supplier's instructions (Amersham-Pharmacia-Biotech). Serial dilutions of human recombinant TIMP-2 were used as internal standards.

Effect of cell-matrix interactions on proMMP-2 activation in non-tumorigenic HaCaT cells and in the tumorigenic HaCaT clone II-4 rt
HaCaT cells and the tumorigenic HaCaT cell clone II-4 rt were grown as monolayers on collagen type I coated dishes or within three-dimensional fibrillar collagen lattices. On days 1, 3, 5 and 7 conditioned media were prepared and analysed by gelatin zymography ( Figure 1). In monolayer cultures both cells types produced comparable amounts of MMP-2, but only in its latent form, while MMP-9 production did clearly differ, being higher in II-4 rt cells. When cultured within a three-dimensional matrix composed of native type I collagen fibrils, HaCaT cells continued to produce the latent form of MMP-2, whereas the tumorigenic II-4 rt cells gained the capability to convert proMMP-2 into its 62/59 kDa active forms. Surprisingly, the strong induction of proMMP-9 observed in II-4 rt monolayer cultures was greatly reduced by growing these cells in contact with fibrillar collagen. Preincubation of II-4 rt cells with the MMP-inhibitor 1,10 phenanthroline and the synthetic furin inhibitor CMK (Stieneke-Grober et al, 1992) resulted in complete suppression of collagen type I-induced activation of proMMP-2. The serine protease inhibitor aprotinin and the specific urokinase inhibitors did not affect this activation process (Figure 2). These observations indicate that MT-MMPs which have been shown to be activated intracellularly by furin-like protease are likely to be involved in proMMP-2 activation.
To test this, we prepared crude membrane fractions from HaCaT and II-4 rt cells cultivated as monolayers or within collagen gels and incubated them with fibroblast conditioned medium containing proMMP-2. Only the membrane fraction isolated from II-4 rt cells grown within the collagenous matrix showed activation of fibroblast derived proMMP-2 ( Figure 3).

MT1-MMP is constitutively expressed in HaCaT and II-4 rt cells under both culture conditions
In order to elucidate the mechanism underlying proMMP-2 activation in II-4 rt cells, we analysed MMP-2 and MT1-MMP expression on both transcript and protein levels. RNA analysis (Figure 4) revealed low MMP-2 mRNA levels in HaCaT and II-4 rt cells under monolayer conditions. After contact with fibrillar collagen the transcript levels increased significantly in both cell types. Surprisingly, HaCaT cells as well as II-4 rt cells displayed constitutive expression of MT1-MMP mRNA in monolayer cultures with higher levels found in II-4 rt cells. In collagen gels both cell types showed an induction of MT1-MMP mRNA to comparable levels.
As shown in Figure 5, membranes purified from HaCaT cells grown as monolayers or within collagen gels for 24 h displayed two immunoreactive bands of equal intensity, with the 63 kDa band corresponding to proMT1-MMP and the 60 kDa band to the active processed form of MT1-MMP (Lehti et al, 1998;Maquoi et al, 1998). In contrast, membranes from II-4 rt cells grown as monolayers and within the collagenous matrix contained mainly  Lehti et al (1998).

MT1-MMP and TIMP-2 play substantial roles for proMMP-2 activation induced by fibrillar collagen in II-4 rt cells
To test whether MT1-MMP is involved in proMMP-2 activation by II-4 rt cells, membranes were isolated from II-4 rt cells grown in collagen gels for 24 h and incubated overnight with fibroblast HaCaT (black bars) and II-4 rt (white bars) cells were grown as monolayers and in collagen gels for 24 and 72 h. At these time-points the cultures were washed with PBS and continued in serum-free medium for a further 24 h. Then the cells/collagen gels were combined with their conditioned media and homogenized. TIMP-2 levels were determined using an ELISA as described in Materials and Methods. Each bar represents the mean ± SEM of two independent experiments performed in duplicate (P < 0.05) conditioned medium containing proMMP-2. Analysis by gelatin zymography revealed activation of fibroblast derived proMMP-2 to the 62/59 kDa active forms ( Figure 6, lane 2, see also Figure 3). However, when the membranes were preincubated with increasing amounts of MT1-MMP-specific antibodies, binding to the catalytic center of MT1-MMP, proMMP-2 activation was completely abolished in a concentration-dependent manner ( Figure 6, lanes 3-5) indicating that MT1-MMP is the membraneassociated activator for proMMP-2 in II-4 rt cells. The higher molecular protein bands in this zymogram are presumably caused by membrane aggregates containing MT1-MMP, MMP-2 and TIMP-2. Incubation of the membrane fraction with the control antibody anti-HLA-ABC did not result in inhibition of proMMP-2 activation (data not shown).
As both cell types showed active MT1-MMP on their cell surface, although in different amounts, under both culture conditions we asked whether different expression levels of TIMP-2 by HaCaT and II-4 rt cells might be critical for proMMP-2 activation. Analysis of TIMP-2 mRNA expression did not reveal significant differences between HaCaT and II-4 rt cells grown in collagen gels. (Figure 4). However, the amounts of TIMP-2 protein were strikingly different (Figure 7). Upon contact with fibrillar collagen, TIMP-2 production in II-4 rt cells was found to be strongly reduced when compared to monolayer culture, whereas in HaCaT cultures the amount of TIMP-2 increased, resulting in about 6-fold higher levels when compared to II-4 rt cells. In contrast, under monolayer conditions both cells types displayed comparable high amounts of TIMP-2 protein. In addition, proMMP-2 activation by membranes isolated from II-4 rt cells could be completely inhibited by addition of recombinant TIMP-2 ( Figure 6, lanes 6 and 7).

DISCUSSION
The proteolytic activity of MMP-2 and MMP-9, both degrading type IV collagen in basement membranes, has been proposed to play a critical role for tumour cell invasion which depends on the destruction of tissue barriers.
In a recent study (Meade-Tollin et al, 1998) enhanced expression of these two MMPs was proposed to be associated with acquisition of the tumorigenic phenotype of c-ras Ha -transfected HaCaT clones. In contrast to this report, we did not detect differences in the amount of latent MMP-2 produced by the non-tumorigenic HaCaT cells and the c-ras Ha -transfected tumorigenic HaCaT variant II-4 rt under comparable culture conditions. However, for MMP-9 the situation was markedly different, showing high amounts of MMP-9 only in the malignant II-4 rt cells. This is in good agreement with reports demonstrating that MMP-9 induction is mediated through the ras proto-oncogene (Gum et al, 1996).
Upon contact with fibrillar collagen the malignant II-4 rt cells, but not the non-tumorigenic HaCaT cells, showed activation of proMMP-2 to its 62/59 kDa active forms. The membraneassociated activation of proMMP-2 was completely abolished in the presence of MT1-MMP-specific antibodies, clearly indicating that other MT-MMPs, although low transcript levels were detected by RT-PCR (data not shown), are not involved in the activation process.
As are all other MMPs, MT1-MMP is synthesized as an inactive zymogen. The mechanisms underlying proMT1-MMP activation and the necessity for cleavage of the N-terminal peptide for its activity are still controversially discussed. MT1-MMP can be activated intracellularly by the protease furin (Pei and Weiss, 1996) and extracellularly by the serine proteases plasmin and urokinase (Okumura et al, 1997;Kazes et al, 1998). However, MT1-MMP-overexpressing COS cells have been shown to activate proMMP-2 without cleavage of proMT1-MMP by furin-like proteases (Cao et al, 1996). In II-4 rt cells conversion of latent MMP-2 to the active 62/59 kDa form could be inhibited by the furin inhibitor CMK (Stieneke-Grober et al, 1992), which prevents intracellular activation of proMT1-MMP Kurschat et al, 1999). In contrast, the serine protease inhibitors had no effect on the activation process. Furthermore, as we obtained complete inhibition by the furin inhibitor our data convincingly demonstrate that activation of proMT1-MMP by furin is a prerequisite for proMMP-2 activation in II-4 rt cells.
Conversion of latent MMP-2 to its active 62/59 kDa forms was observed with II-4 rt cells upon growth within collagen gels but not on a collagenous substrate, indicating that activation of proMMP-2 is highly dependent on interactions of the cells with the surrounding matrix. Induction of proMMP-2 activation upon contact with fibrillar collagen type I has also been described for other cell types including fibroblasts (Azzam and Thompson, 1992;Seltzer et al, 1994;Gilles et al, 1997), microvascular endothelial (Haas et al, 1998), and melanoma (Kurschat et al, 1999) cells. In contrast to fibroblasts and endothelial cells, which showed a collagen-induced upregulation of MT1-MMP mRNA, in melanoma cells MT1-MMP transcript levels were barely affected. Instead, melanoma cells showed a collagen-dependent activation of MT1-MMP which, however, did not correlate with the invasive potential of the melanoma cell lines. In our cell system significant amounts of active MT1-MMP protein were detected in all membrane preparations, independent of the tumourigenic phenotype or culture conditions. However, proMMP-2 activation was only obtained with membranes from II-4 rt cells grown in collagen gels. This observation strongly suggests that the presence of active MT1-MMP is not sufficient to explain the activation differences observed with HaCaT and II-4 rt cells.
As pointed out by many reports, MT1-MMP-mediated activation of proMMP-2 is critically dependent on the amount of TIMP-2 present in the system (Strongin et al, 1995;Will et al, 1996;Butler et al, 1998;Itoh et al, 1998;Zucker et al, 1998). In monolayer cultures TIMP-2 levels were almost comparable in both cell types. However, cultivation in fibrillar collagen induced a dramatic reduction of TIMP-2 protein in II-4 rt cells when compared to HaCaT cells. Furthermore, proMMP-2 activation by II-4 rt membranes could be completely abolished by an excess of recombinant TIMP-2. These findings indicate that TIMP-2 levels might be critical for the activation process in epidermal cells. Thereby the relative low TIMP-2 levels in II-4 rt cells grown in contact with the collagenous matrix might enable MT1-MMPmediated proMMP-2 activation, whereas the high levels in HaCaT cells might be inhibitory for this process. Addition of recombinant TIMP-1, which has been shown to be a poor inhibitor of proMMP-2 activation , had no effect on proMMP-2 activation (data not shown).
Our study indicates that activation of proMMP-2 rather than enhanced expression might be associated with acquisition of the tumourigenic phenotype in epidermal cells. However, differences in the malignancy of cells might not become obvious unless such more complex in vitro culture systems as three-dimensional collagen lattices are used.