Differential effects of transforming growth factor-beta on the expression of collagenase-1 and collagenase-3 in human fibroblasts.

Collagenase-3 (MMP-13) is a matrix metalloproteinase (MMP) originally identified in breast carcinomas which is also produced at significant levels during fetal ossification and in arthritic processes. In this work, we have found that transforming growth factor beta1 (TGF-beta1), a growth factor widely assumed to be inhibitory for MMPs, strongly induces collagenase-3 expression in human KMST fibroblasts. In contrast, this growth factor down-regulated the expression in these cells of collagenase-1 (MMP-1), an enzyme highly related to collagenase-3 in terms of structure and enzymatic properties. The positive effect of TGF-beta1 on collagenase-3 expression was dose- and time-dependent, but independent of the effects of this growth factor on cell proliferation rate. Analysis of the signal transduction mechanisms underlying the up-regulating effect of TGF-beta1 on collagenase-3 expression demonstrated that this growth factor acts through a signaling pathway involving protein kinase C and tyrosine kinase activities. Functional analysis of the collagenase-3 gene promoter region revealed that the inductive effect of TGF-beta1 is partially mediated by an AP-1 site. Comparative analysis with the promoter region of the collagenase-1 gene which contains an AP-1 site at equivalent position, confirmed that TGF-beta1 did not have any effect on CAT activity levels of this promoter. Finally, by using electrophoretic mobility shift assays and antibody supershift analysis, we propose that c-Fos, c-Jun, and JunD may play major roles in the collagenase-3 activation by TGF-beta1 in human fibroblasts.

The matrix metalloproteinases (MMPs) 1 or matrixins form a family of structurally related metalloendopeptidases that are collectively capable of degrading the different macromolecular components of the extracellular matrix. These enzymes play a major role in normal tissue remodeling processes such as embryonic development, bone growth, and resorption, ovulation, uterine involution, and wound healing (1)(2)(3)(4). In addition, abnormal expression of these proteinases may contribute to a variety of pathological processes such as rheumatoid arthritis (5), atherosclerosis (6), pulmonary emphysema (7), and tumor invasion and metastasis (8). At present, the family of human MMPs is composed of 16 members that, according to structural and functional considerations, can be classified into four different families: collagenases, gelatinases, stromelysins, and membrane-type MMPs (1)(2)(3)(4), although there are some enzymes like macrophage metalloelastase (9), stromelysin-3 (10), MMP-19 (11), and enamelysin (12) that do not belong to these groupings. Recently, we have cloned from a breast carcinoma a novel MMP that has been called collagenase-3 (MMP-13) (13,14), since it represents the third member of the collagenase subfamily of human MMPs, the others being fibroblast and neutrophil collagenases. Biochemical characterization of recombinant human collagenase-3 has revealed that it degrades very efficiently the native helix of fibrillar collagens with preferential activity on type II collagen (15,16). In contrast, fibroblast collagenase (MMP-1 or collagenase-1) is more active against type III collagen (17) and neutrophil collagenase (MMP-8 or collagenase-2) preferentially cleaves type I collagen (18). Consequently, the three human collagenases characterized to date show distinct substrate specificities strongly suggesting that they have evolved as specialized enzymes to degrade tissues with different collagen composition. In addition to its degrading activity on fibrillar collagens, further analysis of the substrate specificity of collagenase-3 has revealed that this enzyme may also act as a potent gelatinase thus contributing to further degrade the initial cleavage products of collagenolysis to small fragments suitable for subsequent metabolism (15). Furthermore, very recent studies have shown that collagenase-3 is also able to degrade the large cartilage proteoglycan aggrecan and other components of the extracellular matrix and basement membranes, including the large tenascin isoform, fibronectin, and type IV collagen (15,18,19).
Analysis of the expression of collagenase-3 in human tissues has revealed that in addition to its presence in breast carcinomas, this enzyme is produced during fetal ossification (20,21), and in degenerative joint diseases including osteoarthritis and rheumatoid arthritis (21)(22)(23)(24)(25)(26)(27). At present, very little information is available on the mechanisms controlling collagenase-3 expression in both normal and pathological conditions. Thus, although several groups have reported that human collagenase-3 gene expression can be induced by IL-1␤ and tumor necrosis factor-␣ in chondrocytes from normal and osteoarthritic cartilage (22)(23)(24), much less is known on the mechanisms modulating its expression in tumor processes. In this * This work was supported in part by Grants SAF94-0892 and SAF97-0258 from the Comisión Interministerial de Ciencia y Tecnología, EU-BIOMED II (BMH4-CT96-0017), and Glaxo-Wellcome, Spain. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ Recipient of a fellowship from the Ayuntamiento de Oviedo, Asturias, Spain.
¶ To whom correspondence should be addressed: regard, and based on in situ hybridization experiments and on co-cultures of fibroblast and epithelial breast cancer cells, we have recently proposed that collagenase-3 is predominantly expressed within fibroblasts adjacent to the invasive tumor cells in response to diffusible factors released from the breast cancer cells (28). A preliminary search of putative factors with ability to induce collagenase-3 expression in human fibroblasts revealed that only IL-1 and TPA were able to up-regulate the expression of this gene in KMST fibroblastic cells. By contrast, a series of cytokines and growth factors factors including aFGF, bFGF, platelet-derived growth factor, epidermal growth factor, tumor necrosis factor-␣, and TGF-␣, previously found to play important roles in up-regulating expression of other MMPs, did not show any effect on collagenase-3 expression by human fibroblasts. Since these findings strongly suggested that the mechanisms regulating collagenase-3 expression could be distinct to those operating in the control of other MMPs, we have extended our preliminary search for factors that could act as mediators of collagenase-3 expression in breast carcinomas. In this work, we show that TGF-␤, a growth factor widely assumed to be inhibitory for MMPs like collagenase-1, strongly up-regulates collagenase-3 expression in KMST fibroblasts. We also analyze the mechanisms mediating this induction with the finding that this growth factor acts through a signaling pathway involving tyrosine kinase and protein kinase C activities. Finally, we perform a functional characterization of the promoter region of the collagenase-3 gene looking for the putative elements with ability to mediate the induction of this gene by TGF-␤.

EXPERIMENTAL PROCEDURES
Materials-All media and supplements for cell culture were obtained from Sigma except for fetal calf serum, which was from Boehringer Mannheim (Mannheim, Germany). TGF-␤1, IL-1␣, IL-1␤, aFGF, bFGF, platelet-derived growth factor-BB, 12-O-tetradecanoylphorbol-13-acetate (TPA), cycloheximide, staurosporine, genistein, herbimycin A, calphostin C, and indomethacin were from Sigma. H-89 was from Calbiochem. Restriction endonucleases and other reagents used for molecular cloning were from Boehringer Mannheim. Double-stranded DNA probes were radiolabeled with [␣-32 P]dCTP (3000 Ci/mmol) from Amersham International (Buckinghamshire, United Kingdom) using a commercial random-priming kit from Amersham. Antibodies against members of the Fos and Jun family of transcription factors were from Santa Cruz Biotechnology (Santa Cruz, CA).
Cell Culture and Cell Growth Estimation-Human KMST-6 cells immortalized by ␥-ray irradiation of KMS-6 embryonic fibroblasts were kindly provided by Dr. M. Namba (Hyougo Medical College, Japan). NRK and Mv1Lu cells were kindly provided by Dr. F. Ventura (Universidad Autónoma, Barcelona, Spain). Cells were routinely maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 100 IU/ml penicillin, and 100 g/ml streptomycin in a humidified atmosphere of 5% CO 2 . Cells were subcultured weekly by incubation at 37°C for 2 min with 0.0125% trypsin in 0.02% EDTA, followed by addition of complete medium, washing and resuspension in fresh medium. For most experiments, approximately 5 ϫ 10 5 cells/well were plated out in 100-mm dishes and transferred to serum-free Dulbecco's modified Eagle's medium for 24 h and then exposed to the different growth factors, cytokines, and tumor promoters at the concentrations and for the times indicated. All inhibitors or antagonist compounds were added 1 h before treatment with the different stimulating factors. To test the effect of TGF-␤1 on cell number, KMST, NRK, and Mv1Lu cells were plated in 24-well plates and allowed to adhere to substrate for 24 h in Dulbecco's modified Eagle's medium containing 10% fetal calf serum. Afterward, the serum concentration was reduced to 1%, and TGF-␤1 was added at different concentrations. Cells were incubated for 7 days in the presence of TGF-␤1 with medium changes every 2 days. At the end of the incubation period, the total number of cells was estimated by the fluorometric protein assay essentially as described by Skehan et al. (29).
Isolation of RNA and Northern Blot Analysis-Total RNA from the cells was isolated by the guanidinium isothiocyanate procedure according to Chomczynski and Sacchi (30), separated by electrophoresis in 1% agarose-formaldehyde gels and blotted onto Hybond N nylon filters (Amersham Int.). Filters containing 20 g of total RNA were prehybridized at 42°C for 3 h in 50% formamide, 5 ϫ SSPE (1 ϫ ϭ 150 mM NaCl, 10 mM NaH 2 PO 4 , 1 mM EDTA, pH 7.4), 10 ϫ Denhardt's, 2% SDS, and 100 g/ml denatured herring sperm DNA and then hybridized with radiolabeled collagenase-3 full-length cDNA for 20 h under the same conditions. Filters were washed with 0.1 ϫ SSC, 0.1% SDS for 2 h at 50°C and exposed to autoradiography. RNA integrity and equal loading was assessed by hybridization with an actin probe or with a probe specific for human glyceraldehyde-3-phosphate dehydrogenase. Densitometry of the x-ray films was carried out with the BioImage software (Millipore Corp., Bedford, MA).
Western Immunoblot Analysis-Conditioned media were obtained after incubation of KMST cells in serum-free Dulbecco's modified Eagle's medium for 48 h or supplemented with TGF-␤1, filtered, and concentrated 25-fold in a Centricon filter with a molecular mass cut-off of 10 kDa. Proteins from conditioned media were separated by polyacrylamide gel electrophoresis under denaturing and reducing conditions and transferred to nitrocellulose membranes (Amersham Int.). After blocking with a 5% nonfat milk solution, the membranes were incubated with a 1:5000 dilution of rabbit polyclonal antisera raised against collagenase-3 (13) and then with a goat anti-rabbit IgG antisera conjugated to horseradish peroxidase. Finally, the membranes were washed and developed with a horseradish peroxidase chemiluminescence detection reagent (ECL system, Amersham Int.).
Stable Transfection of Human Collagenase-3 cDNA in KMST Fibroblasts-The full-length cDNA insert of human collagenase-3 was excised from plasmid pNot3a (13), filled-in with Klenow fragment, and ligated into the BamHI site of the eukaryotic expression vector pCMV, also blunt-ended by treatment with Klenow. The resulting plasmid, pCMV-col3ϩ, containing the collagenase-3 cDNA downstream of the constitutive CMV promoter, was transfected in KMST cells using the LipofectAMINE TM reagent (Life Technologies, Inc.) according to the manufacturer's instructions. A parallel transfection was performed with the unmodified pCMV vector as a control. Transfectants were selected with 500 mg/ml G418 (Life Technologies, Inc.). Individual antibiotic-resistant clones were isolated and expanded. A high-expression clonal cell line, called KMST-col3.3, was identified by Western blot analysis of the conditioned culture media and selected for further experiments. Control clones transfected with the empty vector were randomly chosen.
Construction of Chloramphenicol Acetyltransferase Fusion Plasmids-All plasmid constructs were prepared using standard methods (31). The promoterless basic plasmid pCAT-Basic (Promega Corp., Madison, WI) was used for cloning the different promoter fragments obtained from the human collagenase-3 gene upstream of the bacterial CAT gene. The different collagenase-3 promoter constructs (called, Ϫ1004 CAT, Ϫ402 CAT, and Ϫ56 CAT) were generated by polymerase chain reaction amplification with specific oligonucleotides or by endonuclease restriction. All constructs were verified by extensive restriction mapping and DNA sequencing. The Ϫ1004 CAT, Ϫ402 CAT, ϪCAT, and Ϫ56 mut CAT constructs were polymerase chain reactiongenerated by using the following 5Ј primers corresponding to nucleotides Ϫ1004 to Ϫ984 (5Ј-CTGCAGCCCTAGTTTTCTTGG-3Ј), nucleotides Ϫ402 to Ϫ383 (5Ј-TCTAGAATCAGTACTAAGTT-3Ј), nucleotides Ϫ56 to Ϫ38 (5Ј-AAGTGATGACTCACCATTG-3Ј), and nucleotides Ϫ56 to Ϫ38 with two mutations in the AP-1 site (5Ј-AAGTGATTTCTCAC-CATTG-3Ј). In all cases the same 3Ј primer was used, 5Ј-GGTCTAGAT-TGAATGGTGATGCCTGG-3Ј (nucleotides ϩ10 to ϩ27). The construct containing the AP-1 site of the collagenase-1 gene was kindly provided by Dr. P. Angel (University of Heidelberg, Germany). All recombinant plasmids used for transfection assays were purified by the QIAGEN plasmid kit (QIAGEN Inc., Chatsworth, CA).
DNA Transfections and Chloramphenicol Acetyltransferase Assays-For each transfection experiment, cells were seeded at 2 ϫ 10 5 cells/ 30-mm dish and transfected 18 h later with 1.5 g of the indicated reporter plasmid DNA and 0.5 g of pRSV␤gal (Promega Corp.) using the LipofectAMINE TM Reagent and following the manufacturer's indications. 16 h following the start of transfection, serum-free DNA-containing medium was replaced by fresh growth medium without serum and the indicated concentration of hormones to be tested. Transfected cells were harvested in phosphate-buffered saline after 24 h of hormone exposure for chloramphenicol acetyltransferase (CAT) assays. Cell extracts of transfected cells were prepared by three cycles of freeze-thaw and finally resuspended in 100 l of buffer A (15 mM Tris-HCl, pH 8.0, 60 mM KCl, 15 mM NaCl, 2 mM EDTA, 1 mM dithiothreitol). ␤-Galactosidase activity was assayed according to Sambrook et al. (31) in 30 l of the cell extracts. CAT activity was assayed essentially according to the method of Gorman et al. (32). 50 l of extract were incubated in 60 l of buffer A containing 0.2 Ci of [ 14 C]chloramphenicol and 2 mM acetylcoenzyme A for 4 h at 37°C. One ml of ethyl acetate was added, mixed by vigorous vortexing, and centrifuged for 5 min. The supernatant was evaporated under vacuum and reaction products were redissolved in 20 l of ethyl acetate, which were applied to the origin of a silica gel TLC plate. Plates were run in chloroform/methanol (95/5, v/v), dried, and exposed for autoradiography on Hyperfilm-MP (Amersham) for 16 h. For quantitation, radioactivity of the TLC plates was measured by using the InstantImager electronic autoradiography system (Packard Instrument Co., Meriden, CT). Stimulation of CAT activity was expressed as fold increase over activity of non-induced transfected cells and was based on at least three independent experiments.
Electrophoretic Mobility Shift DNA Binding Assay-Nuclear extracts from KMST cells untreated or treated with 10 ng/ml TGF-␤1 for 2 h were prepared as described by Schreiber et al. (33). DNA probes corresponding to the AP-1 element identified at position Ϫ50 to Ϫ44 of the collagenase-3 promoter region were synthesized as two complementary oligonucleotides, 5Ј-AAGTGATGACTCACCATTG-3Ј and 5Ј-CAATGGT-GAGTCATCACTT-3Ј. Oligonucleotides were annealed, labeled with [␥-32 P]ATP by T4 polynucleotide kinase, and further purified by Sephadex G-25 column chromatography (Pharmacia Biotech Inc.). Nuclear extracts (2 g) were preincubated at room temperature for 15 min with 2 g of poly(dI-dC) in 25 mM Tris-HCl, pH 8, 60 mM KCl, 5 mM MgCl 2 , 1 mM EDTA, 10% glycerol, and 0.1 mM dithiothreitol in 20 l. The 30-min binding reaction was initiated by the addition of 2 l (0.1 pmol) of 32 P-labeled probe (5 ϫ 10 6 cpm/pmol). The amount of unlabeled competitor DNA added is indicated in the figure legends. Supershift reactions were carried out by adding 2-4 g of the corresponding antibodies to the binding reactions, 30 min after addition of the labeled probes. The whole mixture was incubated for 2 h at 4°C prior to electrophoresis. Samples were electrophoresed on prerun 4% polyacrylamide gels containing 2.5% glycerol in 50 mM Tris, 380 mM glycine, 2 mM EDTA buffer at 200 V for 4 h. Gels were dried and subjected to autoradiography.

Collagenase-3 Expression Is Induced by TGF-␤ in Human
Fibroblasts-To study the putative effect of TGF-␤ on the expression of collagenase-3 in human fibroblasts, KMST cells were treated with 5 ng/ml TGF-␤1 for 24 h and total cellular RNAs were purified and analyzed by Northern blot using a specific collagenase-3 cDNA probe (13). As shown in Fig. 1, TGF-␤1 induced the accumulation of two mRNA transcripts of 3.0 and 2.5 kilobases, corresponding to the two major collagenase-3 transcripts identified in breast carcinomas (13) and articular cartilage (22). These two transcripts are the result of utilization of different polyadenylation sites present in the 3Ј-noncoding sequence of the human collagenase-3 gene, as demonstrated by using a probe specific for the 3Ј-end of the collagenase-3 cDNA (13). Densitometric analysis of the x-ray films revealed that the observed stimulatory action of TGF-␤1 on collagenase-3 expression was about 2-and 4-fold lower than the respective effects of IL-1 and TPA, which had been previously characterized as collagenase-3 inducers in KMST cells (Fig. 1).
As a previous step to examine the possibility of differential regulation of collagenase-1 and collagenase-3 expression in human cells, the same filter was subsequently hybridized with a collagenase-1 probe and the results obtained are shown in Fig. 1. As can be seen, IL-1 and TPA increased the steady-state mRNA levels of collagenase-1 in a similar fashion as in the case of collagenase-3. However, and in marked contrast to the stimulatory effect of TGF-␤1 on collagenase-3, this factor strongly down-regulated the expression of collagenase-1 in KMST cells. Since these results indicated that TGF-␤1 displayed an opposite effect on these two highly related members of the MMP family, we undertook a detailed analysis of the TGF-␤1 induced up-regulation of collagenase-3 mRNA in human fibroblasts. A dose-response analysis showed that as little as 0.1 ng/ml TGF-␤1 induced a detectable expression of collagenase-3 mRNA, while incubation of the cells with 10 ng/ml induced a maximal accumulation of both collagenase-3 transcripts (10fold over the cells treated with 0.1 ng/ml TGF-␤1) ( Fig. 2A). In addition, a time-course analysis showed a consistent increase with time in the collagenase-3 mRNA levels of KMST cells treated with 10 ng/ml TGF-␤1, the maximal effect being reached after 24 h and declining at longer times of incubation (Fig. 2B). Interestingly, Northern blot analysis also revealed Northern blot analysis was performed using 10 g of total RNA from KMST cells incubated for 24 h in the presence of 5 ng/ml TGF-␤1, transforming growth factor-␣ (TGF-␣, 50 ng/ml), interleukin-1␣ and -␤ (5 ng/ml), epidermal growth factor (EGF, 10 ng/ml), platelet-derived growth factor-BB (PDGF-BB, 10 ng/ml), acidic fibroblast growth factor (aFGF, 10 ng/ml), basic fibroblast growth factor (bFGF, 10 ng/ml), or TPA (10 Ϫ7 M). Filters were hybridized with a collagenase-3 cDNA probe, stripped, and subsequently hybridized with a collagenase-1 cDNA probe, and with a ␤-actin probe to ascertain equal RNA loading for the different samples.

FIG. 2. Dose-response and time-course analysis of the effect of TGF-␤1 on collagenase-3 expression in KMST fibroblasts. A,
KMST cells were cultured for 24 h in the presence of different concentrations of TGF-␤1 and total RNA was analyzed by Northern blot, as described in the legend to Fig. 1. B, KMST cells were cultured in the presence of 10 ng/ml TGF-␤1 for the indicated times and total RNA from each culture was isolated and analyzed by Northern blot. In both cases, filters were hybridized consecutively with labeled probes for collagenase-3 and ␤-actin.
that the effects of TGF-␤1 and TPA on collagenase-3 expression were cooperative (Fig. 3A). In fact, densitometric analysis of the filters revealed that the magnitude of the induction of both collagenase-3 transcripts in cells simultaneously treated with TPA and TGF-␤1 was about 3-and 5-fold higher than those observed after incubation with TPA or TGF-␤1 alone, suggesting a synergistic effect of both collagenase-3 inducers.
To determine if the up-regulating effect of TGF-␤1 on collagenase-3 mRNA levels in KMST cells was also reflected at the protein level, we performed Western blot analysis with conditioned medium from cells treated with 10 ng/ml TGF-␤1 for 24 h. As shown in Fig. 3B, there was an immunoreactive band of the expected molecular size (approximately 60 kDa) in the 25-fold concentrated conditioned medium from TGF-␤1 treated cells. This band was absent in medium obtained from control untreated cells. The same immunoreactive band against collagenase-3 polyclonal antibodies was detected in the conditioned medium from KMST cells treated with TPA. Western blot analysis also confirmed that the effect of TPA and TGF-␤1 was synergistic, since the intensity of the band detected in the medium from cells simultaneously incubated with both factors was much higher than those observed after incubation with TPA or TGF-␤1 alone (Fig. 3B).

TGF␤-mediated Induction of Collagenase-3 Expression in Human Fibroblasts Is Dependent upon Protein Kinase C and
Tyrosine Kinase Activities-To study the molecular mechanisms and signal transduction pathways underlying the upregulating effect of TGF-␤1 on collagenase-3 expression, we first performed cell culture experiments in the presence of the protein synthesis inhibitor cycloheximide. As shown in Fig. 4A, incubation of KMST cells with cycloheximide (10 g/ml, added 1 h before TGF-␤1 treatment) blocked the effect of this growth factor on collagenase-3 mRNA levels. We therefore conclude that de novo protein synthesis is required for the induction of collagenase-3 mRNA by TGF-␤1. We next evaluated the possibility that protein kinase C-mediated signaling pathways could be involved in the TGF-␤1 induced production of collagenase-3 in human fibroblasts. To this purpose, KMST cells were incubated with TGF-␤1 in the presence or absence of PKC inhibitors and the levels of collagenase-3 were examined by Northern blot. As shown in Fig. 4B, staurosporine and calphostin C are able to completely inhibit the up-regulating effect of TGF-␤1 on collagenase-3 expression, indicating the involvement of a protein kinase C in this process. To determine whether a tyrosine kinase is also involved in the collagenase-3 induction by TGF-␤1, we treated KMST cells with 10 ng/ml TGF-␤1 for 24 h in the presence or absence of genistein or herbimycin A, which inhibit tyrosine kinases by different mechanisms, and the levels of collagenase-3 mRNA were analyzed by Northern blot. As illustrated in Fig. 4B, both tyrosine kinase inhibitors were able to completely block the induction of collagenase-3 expression elic-ited by TGF-␤1. By contrast, incubation of fibroblasts with indomethacin, which inhibits prostaglandin synthesis, did not affect TGF-␤1 mediated induction of collagenase-3 (Fig. 4A). To rule out a possible nonspecific effect of these inhibitors on general transcriptional activity, we developed a KMST-derived cell line, KMST-col3.3, which stably expresses the collagenase-3 gene under the control of the constitutive CMV promoter (see "Experimental Procedures"). As can be seen in Fig. 4C, treatment of these cells with cycloheximide, staurosporine, calphostin C, genistein, or herbimicyn A at concentrations that completely abrogate the TGF-␤1-induced collagenase-3 expression in the parental KMST cells, failed to produce any significant reduction on the constitutive expression of the collagenase-3 mRNA by these transfected cells. Interestingly, treatment of KMST-col3.3 cells with cycloheximide resulted in a significant increase of the collagenase-3 mRNA levels, which reinforces the proposed specificity of these agents on the regulation of the endogenous collagenase-3 gene. Taken together, these results indicate that the positive effect of TGF-␤1 on collagenase-3 expression in human fibroblasts is exerted through a signal transduction pathway involving PKC and tyrosine kinase activities.

TGF-␤ Mediated Induction of Collagenase-3 Expression in Human Fibroblasts Is Independent of Cell Proliferation
Rate-It is well established that TGF-␤ exerts both positive and negative effects on cell growth, the final effect depending on a number of factors including cell type (34). Thus, it is widely assumed that TGF-␤ induces growth of mesenchymal cells and inhibits proliferation of epithelial cells. From these considerations, we next investigated if the effect of TGF-␤1 on collagenase-3 expression was correlated with a stimulation of cell proliferation in human fibroblasts. For this purpose, KMST cells were incubated for 7 days in the presence of different concentrations (0.1, 1, and 10 ng/ml) of TGF-␤1 and the cell growth was examined by the fluorimetric protein assay described by Skehan et al. (29). As shown in Fig. 5A, treatment of KMST cells in exponential growth phase with TGF-␤1 only resulted in a very slight increase in cell number. As a positive control of TGF-␤ induced cell proliferation we used NRK rat kidney fibroblasts which have been described to be sensitive to cell growth stimulation mediated by this factor (35), whereas Mv1Lu epithelial cells were used as control of cells inhibited in their proliferation by TGF-␤ (36). These results indicate that collagenase-3 induction by TGF-␤1 is independent of the effects of this factor on cell proliferation rate. To provide additional support to this observation, we also examined the possible occurrence of variations in the expression levels of different cyclin-dependent kinase inhibitors in KMST cells treated with TGF-␤. These inhibitors play essential roles in the regulation of the cell cycle and have been demonstrated to mediate the growth inhibitory signals of TGF-␤ in different cell types (36 -

FIG. 3. Northern blot and Western blot analysis of TGF-␤1and TPA-mediated induction of procollagenase-3 production by KMST cells.
Cells were cultured in the absence (control) or the presence of TGF-␤1 (10 ng/ml), TPA (10 Ϫ7 M), or both, for 72 h under serum-free conditions. After incubation, RNAs were extracted and analyzed by Northern blot (A), whereas conditioned media were concentrated 25-fold and analyzed by Western blot (B), along with recombinant purified procollagenase-3 (400 ng) (Cϩ). 39). In fact, TGF-␤ regulates the expression of at least three members of the cyclin-dependent kinase inhibitor family: p27 kip1 , p21 WAF1 , and p15 INK4B . However, Northern blot analysis of total RNA from TGF-␤1-treated KMST cells did not reveal any variations on p27 kip1 , p21 WAF1 , and p15 INK4B mRNA levels when compared with untreated cells (data not shown). These results provide additional evidence that the TGF-␤1 mediated induction of collagenase-3 expression in human fibroblasts is independent of its effects on cell cycle progression.

Induction of Collagenase-3 Expression by TGF-␤ in Human Fibroblasts Is Partially Mediated by an AP-1 Site-To further
analyze the molecular mechanisms involved in the TGF-␤1 mediated regulation of collagenase-3 expression by human fibroblasts, we undertook a functional analysis of the promoter region of this gene, looking for the putative response elements that could mediate the above observed up-regulatory effect. To this purpose, we first prepared a series of DNA constructs containing various lengths of a 1-kilobase fragment of the 5Јflanking region of the collagenase-3 gene (40), cloned in front of the CAT reporter gene. Three different collagenase-3 promoter constructs linked to the CAT reporter gene were used: Ϫ56 CAT containing an AP-1 consensus sequence (TGACTCA); Ϫ402 CAT including the AP-site and an additional PEA-3 site assumed to be important for MMP responsiveness to growth factors and tumor promoters; and Ϫ1004 CAT containing both sites and 596 base pairs of additional upstream sequence (Fig.  6A). These three constructs were transiently transfected into KMST cells and tested for inducibility by TGF-␤1. As can be seen in Fig. 6A, those cells transfected with the Ϫ56 CAT construct showed significant levels of CAT activity (2.2-fold over basal levels) after TGF-␤1 treatment. This increase in CAT activity was similar to that obtained after treatment of Ϫ56 CAT transfected cells with TPA. Furthermore, simultaneous treatment of these cells with TGF-␤1 and TPA led to a maximal inducibility of 4.3-fold, confirming the above observation of synergistic effect between these two factors in the induction of collagenase-3 expression. Addition of sequences located further upstream of the AP-1 site in the collagenase-3 promoter (constructs Ϫ1004 CAT and Ϫ402 CAT), abolished the stimulatory effect of both TGF-␤1 and TPA on CAT activity, suggesting the presence of putative inhibitory elements in this 5Ј-flanking region of the collagenase-3 gene (Fig. 6A). To perform a comparative analysis between collagenase-3 and collagenase-1 promoters, similar experiments were then carried out with a construct of the collagenase-1 promoter containing its corresponding AP-1 site located at equivalent position to that of collagenase-3 (Fig. 7). In agreement with previous studies (41), CAT levels in cells transfected with this construct were strongly induced by TPA. However, treatment with TGF-␤1 did not have any effect on CAT activity levels of the collagenase-1 promoter construct (Fig. 6A), confirming that despite its close similarity in many structural and functional aspects, collagenase-1 and collagenase-3 may respond in a completely opposite way to identical treatments with factors like TGF-␤1.
Since these results suggested that the AP-1 sequence found in the promoter of the collagenase-3 gene could be important for the observed TGF-␤1 inducibility of its expression, additional studies were performed to further clarify this question. Thus, we first introduced a double mutation (TGACTCA to TTTCTCA) in the AP-1 consensus sequence of the Ϫ56 CAT construct generating a novel construct (COL-3 AP1-MUT). This construct was transfected into KMST cells and its inducibility by TGF-␤1 was tested as above. As shown in Fig. 6B, the increase in CAT activity levels observed after treatment with this growth factor was completely abolished, indicating that the AP-1 element is involved in the TGF-␤ inducibility of col- KMST, NRK, and Mv1Lu cells were cultured in the absence or the presence of TGF-␤1 (0.1, 1, and 10 ng/ml), for 7 days under serum-free conditions. After incubation, the total number of cells was estimated by a fluorometric protein assay. The data are expressed as the means of triplicate wells in three independent experiments. Error bars indicate standard deviations. AU, arbitrary units. lagenase-3 expression in KMST cells. Similar results were obtained in the case of TPA treatment, thus extending previous observations in HeLa and COS cells indicating that the AP-1 element is involved in the TPA stimulatory effect on collagenase-3 transcription (40).
We next performed a series of binding and competition studies with specific nucleotides and nuclear extracts from KMST cells, directed to confirm that this AP-1 sequence was recognized by nuclear factors induced by TGF-␤1 treatment. For this purpose, electrophoretic mobility DNA binding assays were performed with nuclear extracts obtained from KMST cells stimulated with TGF-␤1, and a 19-base pair synthetic oligonucleotide containing the collagenase-3 promoter AP-1 motif. As can be seen in Fig. 8, a retarded band that was competed by an excess of nonlabeled oligonucleotide was detected (lanes 1 and  2). The addition of antibodies against different members of the Fos/Jun protein family to the binding mixture further resulted in strong supershifted bands in the lanes corresponding to the incubation with the anti-c-Fos, anti-c-Jun, and anti-JunD antibodies, when compared with control untreated cells (Fig. 8,  lanes 3, 5, and 7, and data not shown). By contrast, no visible or very weak supershifted bands were detected by using antibodies against FosB, Fra-1, Fra-2, and JunB (Fig. 8, lanes 4  and 6, and data not shown). These data suggest that c-Fos, c-Jun, and JunD proteins could be the major transcription factors involved in the binding to the collagenase-3 AP-1 site after stimulation with TGF-␤1. Finally, and since several studies have shown that the TGF-␤1-elicited activation of the expression of diverse genes is mediated by the cell-specific induc-tion of distinct members of the Fos/Jun protein family (42)(43)(44)(45), we tried to correlate the TGF-␤1-mediated induction of collagenase-3 expression in KMST cells with variations in levels of these transcription factors. To do that, these cells were treated with 10 ng/ml TGF-␤1 at different times (1, 2, and 6 h) and the expression of c-fos, fosB, fra-1, fra-2, c-jun, junB, and junD was analyzed by Northern blot. As shown in Fig. 9, TGF-␤1 strongly enhanced the expression of junD and to a lesser extent that of c-fos. The mRNA levels of both transcription factors reached a maximum at 1 and 2 h, respectively, which is consistent with the fact that the different fos and jun family members are primary response genes induced by growth factors and tumor promoters (46). In marked contrast to this up-regulatory effect of TGF-␤1 on c-fos and junD, this growth factor did not affect the steady-state mRNA levels of c-jun, junB, fosB, fra-1, and fra-2 in KMST cells. These results, together with the above data derived from binding and competition studies, suggest that TGF-␤1-induced expression of collagenase-3 could be mediated by c-fos and junD, which are transcriptionally induced by this growth factor. Nevertheless, we cannot rule out the possibility that c-jun could be involved in binding the AP-1 site of the collagenase-3 promoter after post-transcriptional mechanisms of regulation induced by TGF␤, such as phosphorylation mediated by c-Jun N-terminal kinase (47). DISCUSSION In this work we have shown that expression of human collagenase-3, a matrix metalloproteinase produced by breast carcinomas and arthritic cartilage, is induced by TGF-␤1 in cultured human fibroblasts. This up-regulatory effect of TGF-␤1 on the production of a potent proteolytic enzyme like collagenase-3 is in marked contrast with the widely assumed role of this growth factor as an inducer of anabolic responses in mesenchymal cells. In fact, TGF-␤ has been implicated in the induction of connective tissue formation by stimulating the synthesis of several extracellular matrix components such as fibronectin, thrombospondin, and types I and III collagen (35, 48 -52). In addition, TGF-␤ suppresses the overall degrading activity on these matrix components through a concerted dual action involving the reduction in the production of a wide diversity of proteolytic enzymes including MMPs (53,54) and plasminogen activators (55), as well as the concomitant increase in the synthesis of their respective inhibitors (52, 56 -58). The ability of TGF-␤ family members to induce collagenase-3 expression is not exclusive of fibroblasts, since a recent report has shown that TGF-␤2 up-regulates its expression in transformed keratinocytes (59). By contrast, another recent report has shown that TGF-␤1 inhibits collagenase-3 expression in osteoblast cultures (60), suggesting that the effects of these growth factors are markedly dependent of the cell type. Nevertheless, the ability of TGF-␤ to stimulate the production of a proteolytic enzyme by human fibroblastic cells is not unprecedented, since Overall et al. (58,61) have reported that levels of gelatinase A (MMP-2) are increased in gingival fibroblasts, although the magnitude of this up-regulating effect is lesser than that observed in the present work for collagenase-3. The possibility that collagenase-3 and gelatinase A are coordinately regulated in fibroblasts is of interest in light of recent findings demonstrating that both enzymes form part of an activation cascade which can generate the extracellular collagenolytic activity requested for the connective tissue degradation occurring in both normal and pathological conditions (62). In this context, it is specially noteworthy that collagenase-1 (MMP-1), despite sharing with collagenase-3 its unique ability to initiate degradation of the native helix of fibrillar collagens, shows a completely opposite response to TGF-␤1 treatment. In fact, and as a consequence of the structural complexity of the extracellular matrix and basement membranes which must be proteolytically degraded by MMPs, these enzymes are often coregulated by the same cells in response to the same stimuli. In the case of collagenase-1 and collagenase-3 production by human fibroblasts, this statement seems to be true for cytokines like IL-1␣ and IL-1␤ and for tumor promoters like TPA, all of them displaying a marked up-regulatory effect on both MMP genes (Fig. 1). However, their divergent responses to TGF-␤1 clearly indicate that in addition to common regulatory elements, these genes have distinct transcriptional elements that determine their specific expression patterns. Functional studies of the promoter region of the collagenase-3 gene performed in this work have shown that the AP-1 site present in its 5Ј-flanking region is responsible, at least in part, for the TGF-␤1 mediated induction of this gene. AP-1 sites have also been reported to mediate the response to this growth factor of other genes such as type-1 plasminogen activator inhibitor (63), osteocalcin (64), retinoic acid and retinoid X receptors (65), as well as the autoinduction of the TGF-␤1 gene itself (66). Nevertheless, the extent of stimulation of collagenase-3 gene expression by TGF-␤1, as detected by Northern blot, was higher than the induction of promoter activity observed in transient cell transfection experiments with AP-1 containing constructs. Similar observations have been previously reported during the functional analysis of AP-1 sites present in other MMP genes (67, 68). Therefore, it seems likely that this AP-1 site present in the collagenase-3 gene is necessary but not sufficient for mediating its response to TGF-␤1 in human fibroblasts. The participation of additional elements which could be located further upstream in the 5Ј-flanking region of the collagenase-3 gene could contribute to explain the differences observed between data derived from Northern blot analysis and those from determination of relative CAT activity of reporter gene constructs. In addition, it is remarkable that other MMP genes, including collagenase-1, which are down-regulated by TGF-␤, also contain AP-1 consensus sequences at approximately the same position as that present in the collagenase-3 gene (Fig. 7), thus supporting the idea that sequences other than AP-1 influence the responsiveness of the different MMP genes to this growth factor. Finally, it is also likely that TGF-␤1, in addition to activating the collagenase-3 promoter, may increase the expression of this gene by post-transcriptional mechanisms such as stabilization of the corresponding mRNAs, which have been previously described to operate in the case of the collagenase-1 gene (69).
In an attempt to determine the molecular basis of the somewhat paradoxical effect of TGF-␤1 on collagenase-3 expression in human fibroblasts, we have further investigated the mechanistic aspects underlying this up-regulatory effect. In this work, and by using a series of specific inhibitors for different signaling pathways, we have found that the TGF-␤1 action on collagenase-3 is mediated by PKC and tyrosine kinase signal transduction pathways. Since TGF-␤ receptors are Ser/Thr ki- FIG. 9. Effect of TGF-␤1 on expression of Fos/Jun family members in KMST cells. KMST cells were cultured for the indicated times in the presence of TGF-␤1 (10 ng/ml), and total RNA was analyzed by Northern blot with labeled probes for c-fos, fosB, fra-1, and fra-2 (A), as well as with probes for c-jun, junB, and junD (B). In both cases, filters were hybridized consecutively with a labeled probe for ␤-actin. nases, it is tempting to speculate that the activity affected by PKC inhibitors could be that intrinsic to the TGF-␤ receptors themselves. However, the observation that TGF␤-RI and -RII autophosphorylation is not affected by PKC inhibitors (70) suggests that additional kinase activities acting downstream from the TGF-␤ receptor are involved in mediating the induction of collagenase-3 by TGF-␤1. In addition, we have tried to correlate the TGF-␤1 positive effect on collagenase-3 expression with some of the pleiotropic effects elicited by this multifunctional growth factor. It is well known that TGF-␤ displays a wide variety of actions even in the same cell type, depending on a number of factors including the specific target, conditions of cell culture, and presence of other growth regulators. Consistent with this, the observation that collagenase-3 induction by TGF-␤1 in fibroblasts is accompanied by a weak stimulation of cell growth, together with data showing that in HaCaT keratinocytes TGF-␤ mediated up-regulation of this enzyme is accompanied by a potent inhibition of cell growth, 2 strongly suggest that induction of collagenase-3 by this growth factor is independent of its effects on cell proliferation.
Finally, in this work we have examined the possibility that cell specific induction of distinct members of the Fos/Jun family of transcription factors is responsible for the divergent regulation of collagenase-1 and collagenase-3 genes by TGF-␤1 in human fibroblasts. Electrophoretic mobility shift assays and antibody supershift analysis revealed that c-Fos, c-Jun, and JunD are found in complexes formed with nuclear extracts prepared from KMST cells treated with TGF-␤1. Analysis of the pattern of expression of these Fos/Jun proto-oncogenes in KMST cells treated with TGF-␤1 confirmed that collagenase-3 induction is preceded by an increase in levels of expression of c-Fos and JunD. According to these results, it is conceivable that the induction of high levels of c-Fos and JunD favors the formation of specific complexes which bind and transactivate the collagenase-3 promoter, thus resulting in the observed upregulation of this gene. Nevertheless, the participation of c-Jun in the process, after post-transcriptional mechanisms of regulation induced by TGF-␤ such as phosphorylation mediated by c-Jun N-terminal kinase (47), cannot be ruled out. In this regard, it is well known that the preferential binding to AP-1 sites exhibited by different Fos/Jun proteins is dependent upon specific flanking and core nucleotide sequences, thus allowing fine regulation of expression of the diverse AP-1 driven genes. c-Jun and c-Fos have been proposed to be fundamental intermediates for collagenase-1 and stromelysin-1 gene activation by tumor necrosis factor-␣ in fibroblasts, and by TGF-␤ in keratinocytes, whereas inhibitory effects have been reported to be mediated by transient elevation of JunB (42,45). However, in KMST cells, TGF-␤ had little if any effect on the expression of c-Jun or JunB, thus indicating that at least in these cells, other mechanisms should be involved in TGF-␤1 elicited downregulation of collagenase-1.
In summary, we have provided evidence that TGF-␤ dissociates production of collagenase-1 and collagenase-3 by human fibroblasts. The opposite response of both enzymes to TGF-␤1 in KMST cells makes them an appropriate model system for studying the molecular mechanisms controlling the expression of these two highly related enzymes in terms of structure and enzymatic properties, but displaying marked differences in their pattern of tissue expression. Further studies will be also required to evaluate the putative role of TGF-␤1 as an in vivo inducer of human collagenase-3 in those conditions in which this enzyme has been found at high levels, including breast carcinomas and arthritic processes.