In normal epithelial cells, TGF-β acts as a strong inhibitor of cell growth.1 Binding of TGF-β1 to its type II cell-surface receptor (TGF-βRII) initiates an intracellular signaling cascade, leading to formation and nuclear translocation of a Smad2–Smad4 complex, which in concert with co-proteins regulates the expression of genes involved in cell differentiation and growth control.2 In addition, several other downstream signaling pathways, including the ERK-MAPK signaling pathway, p38, c-Jun N-terminal kinase and the phosphatidylinositol 3-OH kinase, can be activated by TGF-β and therefore potentially contribute to the response of the cell to TGF-β stimulation.3–5
In early tumor stages, transformed epithelial cells are usually still sensitive to TGF-β–mediated growth inhibition.6 However, in later stages of tumorigenesis, epithelial tumor cells frequently escape from TGF-β–induced growth control; and once this has occurred, TGF-β can act as a promoter of tumor progression.7, 8 TGF-β may then affect the plasticity and adhesion of tumor cells and induce tumor cell migration.9–11 In pancreatic cancer, high TGF-β expression levels are correlated with a more aggressive phenotype and increased local infiltration, suggesting that TGF-β may also stimulate the invasion of tumor cells to promote pancreatic tumor progression.12 However, it is not clear whether TGF-β directly affects the invasive potential of pancreatic cancer cells and, if so, whether this is dependent on their possession of an intact intracellular signaling pathway. This is of particular interest since genetic alterations of TGF-βRII and Smad4 are commonly found in pancreatic cancer and have been associated with more limited responses to TGF-β.13, 14 Furthermore, the molecular mechanisms through which TGF-β may promote pancreatic tumor cell invasion remain to be characterized. As the enzymatic degradation of extracellular matrix (ECM) barriers is a key step in tumor cell invasion, TGF-β may mediate tumor cell invasion by regulation of ECM-degrading proteinases.15 Among the increasing number of ECM-degrading proteinases implicated in tumor cell invasion, most attention has been focused on the family of matrix metalloproteinases (MMPs) and the plasminogen activator system.16–18 High expression and activation levels of MMP-2 have been found in various human cancer tissues, including breast and pancreatic cancer, and correlated with tumor stage and grade in several cases.19–21 We have demonstrated that increased expression and activation levels of MMP-2 are strongly associated with elevated expression levels of its activators MT1-MMP and MT2-MMP and that it plays a significant role in pancreatic tumor cell invasion.22 An increasing number of members of the ADAM (a disintegrin and metalloproteinase) family of proteins have been identified. These proteins possess integrin and metalloproteinase domains and are involved in proteolytic modification of the ECM; therefore, they might also be relevant in tumor cell invasion and metastasis.23, 24 However, only little is known about the role of ADAM proteins in physiological and pathological processes, and no data are available concerning expression and activation of ADAM proteins in the pancreas.
In contrast to the ADAM family of proteinases, several in vitro and in vivo studies have clearly shown an important role of the plasminogen activator systems [tissue plasminogen activator (tPA) and urokinase plasminogen activator (uPA) and its receptor (uPAR)] in epithelial tumor progression.25, 26 For instance, over-expression of uPA and uPAR have been reported in human cancer tissues, and high expression of uPA has been associated with poor relapse-free and overall survival in patients with primary invasive breast cancer.27, 28 Since both the family of MMPs and the plasminogen activator system may be responsible for TGF-β–induced tumor cell invasion, we studied not only the significance of TGF-β and the integrity of its downstream signaling pathway in the regulation of pancreatic cancer cell invasion but also the role of the MMP family and uPA in mediating these effects.
MATERIAL AND METHODS
Cell culture and reagents
Human pancreatic cancer cell lines were obtained from the following sources: IMIM-PC1 and IMIM-PC2 from Dr. F.X. Real29 (Departament d'Immunologia, Institut Municipal d'Investigacio Medica, Barcelona, Spain), MiaPaca2 and PANC-1 from the European Collection of Animal Cell Cultures (Salisbury, UK) and CAPAN-1 from the ATCC (Rockville, MD). Pancreatic cancer cell lines were cultured in DMEM (IMIM-PC1, IMIM-PC2, PANC-1 and MiaPaca2) or RPMI (CAPAN-1) (GIBCO BRL, Gaithersburg, MD) supplemented with 10% FCS (GIBCO BRL), penicillin (100 U/ml) and streptomycin (100 U/ml) at 37°C in 5% CO2 atmosphere. TGF-β (10 ng/ml) was obtained from R&D Systems (Wiesbaden, Germany), Matrigel from Becton Dickinson (Bedford, MA), PAI-1 from Calbiochem (Bad Soden, Germany), minocyclin from Sigma (Deisendorf, Germany) and batimastat (BB94) from British Biotech (Oxford, UK).
cDNA probes and antibodies
The cDNA probe for human MMP-2 was a kind gift from Dr. W.G. Stetler-Stevenson (Tumor Invasion and Metastasis Section, Laboratory of Pathology, National Cancer Institute, Bethesda, MD). MT1-MMP was provided by Dr. P. Basset (Strassburg, France). Image clone IMAGp998I061933, with 631 bp covering residues 313–944 of the MT2-MMP gene (accession D86331), and the clone IMAGp998O201778, containing the cDNA of uPAR, were obtained from the Resource Centre/Primary Database of the German Human Genome Project (Berlin, Germany). cDNA probes for MMP-9, tPA and uPA were cloned in our laboratory,30 and the cDNA probe for ubiquitin was previously cloned in our laboratory.31 Neutralizing anti-human TGF-β1 monoclonal antibodies (MAbs) were obtained from Sigma.
Northern blot analysis
Total RNA (30 μg) from pancreatic cancer cell lines was extracted as previously described,20 size-fractionated and transferred to Hybond N membranes (Amersham, Aylesbury, UK). Northern blots were hybridized with [32P]-labeled cDNA probes for MT1-MMP, MT2-MMP, MMP-2, MMP9, tPA, uPA and uPAR and washed to high stringency, as previously described.20 X-ray films (Kodak, Rochester, NY) were exposed for 1 to 7 days at –70°C using intensifying screens. To control the loading of individual lanes, all Northern blots were hybridized with a cDNA probe for the ubiquitously expressed ubiquitin.
The gelatinolytic activity of MMP-2 and uPA was determined in the supernatants of TGF-β–treated and untreated pancreatic cancer cell lines. For this purpose, cells were cultured for 24 hr in serum-free medium, washed twice and finally treated for 24 hr with TGF-β (10 ng/ml) or serum-free medium. After this procedure, supernatants were concentrated and the protein content was determined using the BCA Protein Assay Reagent (Pierce, Rockford, IL). Equal amounts of protein (50 μg/lane) were mixed with SDS sample buffer without reducing agents and incubated for 20 min at 37°C. After incubation, samples were applied to a 4.5% (w/v) stacking polyacrylamide gel and separated on a 7.5% (w/v) polyacrylamide gel containing 1 mg/ml gelatin for the detection of MMP-2 activity or 2 mg/ml casein plus 0.025 U/ml plasminogen for the determination of proteolytic activity of the plasminogen activator. After electrophoresis, gels were soaked in 2.5% Triton X-100 for 1 hr to remove SDS and incubated for 16 hr at 37°C in 50 mM Tris-HCl (pH 7.6) containing 150 mM NaCl, 10 mM CaCl2 and 0.02 % NaN3. Gels were stained for 1 hr in 45% methanol/10% acetic acid containing 0.5% Coomassie brilliant blue G250 and destained. Proteolytic activity was detected as clear bands on a blue background of the Coomassie blue–stained gel. Zymographic analyses were performed in at least 3 independent experiments.
In vitro invasion assays
The invasive potential of the pancreatic cancer cell lines PANC-1, IMIM-PC1, IMIM-PC2, MiaPaca2 and CAPAN-1 was determined using a modified 2-chamber invasion assay, as previously described.22 Briefly, 6-well Transwell plates (Costar, Cambridge, MA) with 8 μm pore size were coated with 1:2 diluted Matrigel (Becton Dickinson). The lower chamber was filled with DMEM containing 1% FCS plus TGF-β (10 ng/ml) or FCS alone. Tumor cells (2 to 3 × 105) in DMEM with 1% FCS were seeded in the upper chamber and incubated for 1, 6, 12, 24 or 48 hr. TGF-β–neutralizing antibodies and proteinase inhibitors were applied in the upper chamber as follows: TGF-β–neutralizing antibodies (1.5 μg/ml, Sigma), EACA (50 mM), PAI-1 (20 μg/ml), BB-94 (7.5 μM) and minocyclin (10 μg/ml). After incubation, cells on the upper side of the membrane were wiped off and the membrane was fixed in 4% paraformaldehyde and 0.25% glutaraldehyde. Cells on the lower side of the membrane were stained with 0.5% methylene blue in 50% methanol and counted. All invasion assays were done in triplicate.
Statistical analysis of the invasion data was performed by either ANOVA for studies involving time courses or Student's t-test. All invasion assays were done in triplicate in at least 2 different experiments. Summarized data are expressed as means ± SEM.
Increased invasiveness of pancreatic cancer cell lines by TGF-β1
The effect of TGF-β1 on the potential of cells to invade through reconstituted basement membranes was analyzed using Matrigel-coated invasion chambers. PANC-1, IMIM-PC1, IMIM-PC2, MiaPaca2 and CAPAN-1 cell lines were incubated either with or without TGF-β1 for 1, 6, 12, 24 or 48 hr. Figure 1a shows the invasive potential of tumor cells after 48 hr. A single dose of 10 ng/ml TGF-β1 induced a strong increase in the invasiveness of the cell lines PANC-1 and IMIM-PC1 but did not alter the invasive potential of IMIM-PC2, MiaPaca2 and CAPAN-1. The time course analysis for TGF-β–induced invasion of PANC-1 and IMIM-PC1 cell lines is demonstrated in Figure 1b. Tumor cell invasion started after 12 hr, reached statistical significance in both cell lines after 24 hr (p < 0.01) and most interestingly, continued even 48 hr after a single application of TGF-β1. This suggests that TGF-β1 might act in an autocrine manner to maintain tumor cell invasion.
Indeed, when neutralizing antibodies against TGF-β1 (1.5 μg/ml, Sigma) were applied to the upper chamber of the invasion assay after 24 hr of TGF-β treatment, a further increase of tumor cell invasion was completely prevented in both cell lines (Fig. 1b).
Together, these data show that TGF-β promotes tumor cell invasion in PANC-1 and IMIM-PC1 cells and strongly indicate that the long-term effect of TGF-β on tumor cell invasion is caused by an autoregulatory loop leading to sustained expression of TGF-β. Indeed, as previously reported,32 treatment of pancreatic cancer cells with 10 ng/ml TGF-β led to an increase of TGF-β gene expression in both responsive cell lines but did not affect TGF-β transcript levels in the non-responsive cell lines CAPAN-1, MiaPaca2 and IMIM-PC2.
Effect of TGF-β on expression and activation of MMPs
To study the role of MMPs in TGF-β–induced tumor cell invasion, we measured the expression of MMP-9 and MMP-2 and its activators MT1-MMP and MT2-MMP in pancreatic cancer cell lines before and after treatment with TGF-β (Fig. 2). MMP-2 mRNA was expressed in the cell lines IMIM-PC1, IMIM-PC2 and PANC-1 but was undetectable in unstimulated CAPAN-1 and MiaPaca2 cells. Treatment with TGF-β did not affect the transcript level of MMP-2 in IMIM-PC2 but led to a strong increase of MMP-2 mRNA in PANC-1 and IMIM-PC1, the same cell lines in which TGF-β up-regulated the invasive potential. TGF-β–induced up-regulation of MMP-2 occurred in a time-dependent manner, which reached peak values after 24 hr of treatment but remained elevated for a prolonged time, at least up to 48 hr after a single TGF-β application in both cell lines (Fig. 3a). Interestingly, treatment with TGF-β had no influence on transcript levels of MMP-9, MT1-MMP or MT2-MMP in any of the pancreatic cancer cell lines (Fig. 2).
In addition to our expression studies, we measured the ability of TGF-β to induce MMP-2 proteolytic activity. We therefore treated the cell lines PANC-1, IMIM-PC1 and IMIM-PC2 with 10 ng/ml TGF-β for 24 hr and measured the proteolytic activity of MMP-2 before and after treatment using gelatin zymography. As shown in figure 3b, the 72 kDa band of MMP-2 was found in all 3 cell lines, indicating protein expression of latent MMP-2. Treatment with TGF-β, however, significantly increased the intensity of the latent MMP-2 band in IMIM-PC1 and PANC-1 cells without affecting gelatinolytic activity in IMIM-PC2 cells. In addition and even more interesting, TGF-β increased the 62 kDa band of activated MMP-2 in both responsive cell lines, suggesting that TGF-β treatment not only stimulates the expression but also induces the activation of MMP-2 in those cell lines in which TGF-β induces tumor cell invasion.
Effect of TGF-β on the plasminogen activator system
Since the uPA system has also been suggested to be involved in pancreatic cancer cell invasion, we analyzed the expression and activity of the plasminogen activator systems (tPA, uPA and uPAR) and studied their regulation by TGF-β in the 5 pancreatic cancer cell lines (Fig. 4a). Northern blot analysis revealed high expression levels of uPA and uPAR in cultured IMIM-PC2 cells, whereas only low amounts of uPA were detected in the cell lines PANC-1, IMIM-PC1, MiaPaca2 and CAPAN-1. However, treatment with TGF-β up-regulated expression of uPA in PANC-1 cells and increased expression of uPA and uPAR in IMIM-PC1 cells. In contrast, TGF-β treatment did not affect the expression level of either uPA or uPAR in IMIM-PC2, MiaPaca2 and CAPAN-1 cells. Time course analysis for uPA expression in PANC-1 and IMIM-PC1 cells showed a rise of uPA transcript levels 6 hr (PANC-1) to 12 hr (IMIM-PC1) after the start of TGF-β treatment, with transcript levels remaining high up to 48 hr (Fig. 4b). To examine the effect of TGF-β on the activity of uPA, we treated PANC-1, IMIM-PC1 and CAPAN-1 cells with TGF-β for 24 hr and measured the proteolytic activity in the supernatants of stimulated and unstimulated cells. As illustrated in Figure 4c, a single application of TGF-β increased the level of activity in both responsive cell lines but did not significantly change the proteolytic activity of CAPAN-1 cells. This suggests that both the expression and the activity of the uPA system are up-regulated by TGF-β. In contrast, tPA was not expressed at significant levels in either TGF-β–treated or untreated pancreatic cancer cell lines.
Influence of proteinase inhibitors on TGF-β–induced invasion
To confirm that TGF-β–induced invasiveness of PANC-1 and IMIM-PC1 cells is mediated by uPA and MMP-2, PANC-1 and IMIM-PC1 cells were pre-treated with a selection of metalloproteinase inhibitors and inhibitors of the plasminogen activator system. Inhibition of MMPs with 7.5 μM batimastat (BB-94) or 10 μg/ml minocyclin led to a significant reduction of TGF-β–induced invasion in both cell lines, suggesting that the increase in invasiveness in the PANC-1 and IMIM-PC1 cell lines is caused mainly by activation of the MMP family (Fig. 5). The applied MMP inhibitors are not specific for MMP-2 and might interact with other MMP members involved in tumor cell invasion, e.g., MMP-9, MT1-MMP and MT2-MMP. However, since MMP-9, MT1-MMP and MT2-MMP are not affected by TGF-β treatment, our data suggest that MMP-2 is a major mediator of TGF-β–induced invasion.
Nevertheless, inhibition of the MMP family did not completely prevent TGF-β–stimulated tumor cell invasion, suggesting that MMP-2 is not the only downstream mediator of TGF-β–induced invasion. Indeed, inhibition of the plasminogen activator system with 50 mM EACA or 20 μg/ml PAI-1 led to a similar, though less pronounced, reduction of TGF-β–mediated tumor cell invasion in both cell lines, indicating that both the MMP family (e.g., MMP-2) and the uPA system are involved in this process. Since tPA could not be detected in either stimulated or unstimulated cells, it appears highly likely that the inhibiting effects of EACA and PAI-1 on TGF-β–induced tumor cell invasion are caused by interaction of the inhibitors with uPA and uPAR.
TGF-β is a strong inhibitor of cell growth in normal epithelial cells and acts as a tumor suppressor in early tumor stages.6 However, tumor cells frequently escape from growth regulation by TGF-β and once the negative growth response to TGF-β is attenuated, other responses to TGF-β can manifest, leading to a more malignant phenotype.8, 9 In the present study, we demonstrate the potential of TGF-β to promote tumor cell invasion in pancreatic cancer. A single application of TGF-β to the pancreatic cancer cell lines PANC-1 and IMIM-PC1 strongly increased their potential to migrate through Matrigel-coated invasion membranes. TGF-β–induced invasiveness started 12 hr after treatment and reached statistical significance after 24 hr. Interestingly, the invasive potential continued to increase in both cell lines even 48 hr after treatment, suggesting that TGF-β may act in an autocrine manner to maintain the invasive phenotype. Indeed, TGF-β up-regulates its own expression in both cell lines,32 and application of TGF-β–neutralizing antibodies 24 hr after TGF-β treatment completely prevented any further increase of tumor cell invasion. Our findings are strongly supported by studies indicating that TGF-β acting in an autocrine fashion mediates and maintains epithelial tumor cell invasion so as to develop metastasis.33–36 For instance, auto-induction of TGF-β has been shown in transformed mammary cells, rendering the cells highly invasive, whereas normal mammary cells remained unaffected by TGF-β treatment.9 Consequently, tumor cells derived from metastases often express significantly higher levels of TGF-β than primary cancer–derived cell lines.37 However, disruption of TGF-β signaling by transfection of a dominant negative TGF-βRII has been shown to abolish the potential of colon-cancer cells to metastasize,35 indicating that a functional signaling pathway is a prerequisite for tumor cells to respond to TGF-β with increased invasiveness. Since mutation-based disturbances of the TGF-β signaling pathway are particularly common in pancreatic cancer, we followed this idea and studied the effect of TGF-β on the invasiveness of pancreatic cell lines with defects of the signaling pathway. According to our predictions, TGF-β did not affect the invasive potential in cells with inactivating mutations of the type II receptor (MiaPaca2) or in cells with lack of Smad4 expression (CAPAN-1 and IMIM-PC2). Together with our previous observation that TGF-β stimulates its own transcription in the responsive cell lines IMIM-PC1 and PANC-1 but not in MiaPaca2, CAPAN-1 and IMIM-PC2 cells,32 our results indicate that mutational alterations of the TGF-β signaling pathway interrupt the autoregulatory loop of tumor cells and negatively affect their potential to respond to TGF-β with increased invasiveness.
Therefore, it is not surprising that an increasing body of evidence shows a strong relation between the expression level of TGF-β in tumor tissues and the malignant phenotype.38–40 High levels of TGF-β have been described in advanced epithelial tumors, e.g., breast, gastric and colon cancers, and correlated with decreased survival.41–43 In pancreatic cancer, high expression levels of TGF-β have been detected and correlated with a more aggressive phenotype and increased local infitration.12 However, little was known about the molecular mechanisms through which TGF-β promotes tumor cell invasion in pancreatic cancer. We previously showed that the MMP family, in particular MMP-2, is involved in pancreatic cancer cell invasion.22 In the current study, we show that MMP-2 and the uPA system are mediators of TGF-β–induced tumor cell invasion. Strong increases in MMP-2 gene expression and activation were observed in the TGF-β-responsive cell lines PANC-1 and IMIM-PC1 after TGF-β treatment, and application of the MMP inhibitors batimastat and minocyclin significantly reduced TGF-β–stimulated tumor cell invasion. Interestingly, TGF-β did not affect transcript levels of MMP-2 in cell lines with defects of the TGF-β pathway, nor did it change expression levels of MMP-9, MT1-MMP and MT2-MMP in any of the analyzed cell lines. Thus, MMP-2 appears to be the only member of the MMP family involved in TGF-β–induced pancreatic cancer cell invasion. However, since neither batimastat nor minocyclin completely abolished TGF-β–induced invasion, it appears unlikely that MMP-2 is the only effector of TGF-β. For this reason, we studied the involvement of the plasminogen activator system, which has also been shown to participate in tumor progression and whose constituents are highly expressed in pancreatic cancer.27, 44 TGF-β treatment up-regulated the expression of both uPA and uPAR in IMIM-PC1 cells and of uPA in PANC-1 cells but failed to cause a change of tPA expression. In addition, TGF-β increased the level of uPA activity in both cell lines, and application of the plasminogen activator inhibitor PAI-1 or the plasminogen/plasmin binding inhibitor EACA resulted in a relevant decrease of tumor cell invasiveness.
Taken together, our results provide further evidence that TGF-β functions as a promotor of tumor progression in pancreatic cancer. In summary, in cells with a functional TGF-β signaling pathway, TGF-β may act as a strong mediator of tumor cell invasion by up-regulating MMP-2 and the uPA system. Our data further suggest that TGF-β may act in an autocrine manner to maintain tumor cell invasion and, therefore, may contribute to the expansion of responsive cell clones, leading to distant metastases.
This work was supported by grants from the Deutsche Forschugsgemeinschaft (SFB 518, project B1) and the European Union (BMH4-CT98-3085) to TMG.