Paper | Published:

Reduction of tumor cell migration and metastasis by adenoviral gene transfer of plasminogen activator inhibitors

Gene Therapy volume 6, pages 227236 (1999) | Download Citation



Recombinant adenoviral vectors expressing u-PA, t-PA, PAI-1 and PAI-2 were employed to correlate the expression of components of the fibrinolytic system with the invasiveness of HT 1080 tumor cells. Migration through Transwell inserts in vitro in the presence of plasminogen was increased up to 22% by overexpression of u-PA, whereas t-PA had no effect. Gene transfer of PAI-1 or PAI-2 both reduced migration in a dose-dependent manner by up to 43% with PAI-1 and 29% with PAI-2. Two routes of gene transfer were used to alter metastasis of subcutaneously implanted HT 1080 cells expressing firefly luciferase in nude mice. Infection of cultured tumor cells with adeno- virus expressing either PAI-1 or PAI-2 before implantation significantly reduced the incidence of lung metastasis by 60% compared with control virus. However, only PAI-2 reduced the incidence of lung and brain metastasis following liver gene transfer. Although PAI gene transfer by either route reduced primary tumor size, it had little effect on tumor vascularization or host survival. The migratory and metastatic phenotype of HT 1080 tumor cells is thus directly dependent on u-PA expression levels and can be altered by gene transfer of u-PA or plasminogen activator inhibitors.


In many patients with malignant solid tumors, the final outcome of the disease is often determined by the spread of the primary tumor to distant organs. Metastasis is a multistep process that requires tumor cells to gain access to the lymphatic or blood vasculature, become blood-born and embolize into the capillary bed of another organ where they finally invade the surrounding tissue. Physiological barriers such as extracellular matrix and basement membranes must be traversed during several steps in this process. To accomplish this task, tumor cells employ a number of proteolytic enzymes.1,2

Plasmin, a broad spectrum serine proteinase, degrades the matrix proteins collagen, laminin and fibronectin and regulates the activation of other proteolytic enzymes, such as metalloproteinases.1 Plasminogen activation is controlled by both plasminogen activators and their inhibitors. Two physiological plasminogen activators appear to have different functions. While t-PA is primarily associated with the clearance of fibrin deposits from the vasculature,3 u-PA and its cellular receptor, u-PAR, mediate cell-associated plasmin generation during cell migration, inflammation and wound healing.4

u-PA activity is regulated by the irreversible plasminogen activator inhibitors, PAI-1 and PAI-2. PAI-1 is found in many different tissues, predominantly in liver and adipose tissue.5,6 It can be readily detected in plasma and exists in at least two distinct conformations, as an active inhibitor and a latent, inactive form. PAI-1 binding to vitronectin in plasma and in the extracellular matrix stabilizes it in the active conformation.7 Recent data suggest that PAI-1 reduces cellular migration, not only by direct inhibition of u-PA activity, but also by blocking u-PAR binding and the integrin-binding site on vitronectin,8,9 although under some circumstances PAI-1 can promote cellular migration by decreasing cell adhesion to vitronectin.10 Thus, there may be counteracting mechanisms involving the same molecules that differentially regulate cell adhesion, migration and metastasis.

The expression pattern of PAI-2 is different from that of PAI-1. It is mainly expressed in trophoblastic epithelium in the placenta and usually only detected in the plasma of pregnant women.11 PAI-2 has an inefficient, internal signal sequence that causes the majority of the protein to be retained in the cell12 where it seems to serve additional biological functions such as preventing TNFα-induced apoptosis in HT1080 cells.13 PAI-2 is a thermostable inhibitor and does not bind to vitronectin or other extracellular matrix proteins.14

The effects of u-PA and its inhibitors on cell migration and metastasis have been documented. Increased expression of u-PA augments tumor cell migration in vitro15 and accelerates metastasis of rat prostate carcinomas in vivo.16 Conversely, PAI-1 has been shown to reduce tumor cell migration in vitro and metastasis in vivo.17,18 The effects of PAI-2 on migration and metastasis have also been examined using an expression vector-transfected melanoma cell line19 or recombinant protein.20 Human HT1080 cells overexpressing PAI-2 have a reduced ability to degrade extracellular matrix and form a fibrous cap around the tumor in vivo. However, neither metastasis nor its inhibition could be demonstrated histologically.21

Recombinant adenoviral vectors can be produced in high titers and allow efficient gene transfer into a wide variety of mammalian cell and tissue types. Recently, the effect of adenovirus-mediated gene transfer of human PAI-1 on metastasis of human ocular melanoma cells in an orthotopic nude mouse model has been examined.22 Repeated intraocular injections of recombinant adenovirus, as well as liver gene transfer by intravenous injection reduced both the incidence and burden of liver metastasis and significantly prolonged survival.

The purpose of this study was to explore the effects of adenovirus-mediated gene transfer of plasminogen activator inhibitors on the metastasis of the more commonly used HT1080 tumor cell line in an animal model. This fibrosarcoma cell line expresses high levels of both u-PA and PAI-1, and is both highly migratory in vitro and metastatic in vivo. The relative efficacy of PAI-1 and PAI-2 was therefore compared with respect to their ability to reduce migration in vitro and metastasis in vivo. The results show that HT1080 cells readily metastasize in nude mice, that expression of plasminogen activator inhibitors by the tumor cells can reduce the incidence of metastasis significantly and that PAI-2 is more effective than PAI-1 following liver gene transfer.


Selection and characterization of HT1080 cell clones

HT1080 cells were stably transfected with a luciferase expression plasmid to permit the sensitive detection of metastasis by biochemical assay of organ homogenates. Of the 72 primary G418-resistant clones analyzed, 44 expressed widely varying degrees of luciferase activity in cell lysates. Three rapidly growing subclones with the highest luciferase expression (20–37 light units per cell) were subcloned twice more from single cells using luciferase expression as the selection criterion to ensure both homogeneity and stability of the population. Each primary clone was characterized with respect to expression of u-PA, t-PA and PAI-1 in conditioned cell culture medium and in vitro migration (Table 1). While all three clones express high levels of both u-PA and PAI-1 and lower amounts of t-PA, the expression of these molecules varies two- to three-fold. It has been previously reported that HT1080 express no mRNA for PAI-2.21 In these experiments, the rates of cell migration in a Transwell assay were comparable. Clone 28.9.21 and clone 33.47.3, which have equal levels of luciferase expression, were further tested for metastasis in vivo (Table 1). Clone 28.9.21 demonstrated both the highest expression of u-PA in vitro and the most frequent metastasis in vivo, and was therefore used for all subsequent experiments.

Table 1: Table 1

Modulation of tumor cell migration by adenoviral gene transfer in vitro

Adenoviral gene transfer was used to alter cellular synthesis and secretion of plasminogen activators and their inhibitors. In the presence of serum, augmenting u-PA secretion by gene transfer stimulated cell migration across a Matrigel-coated filter in a viral dose-dependent manner (Figure 1a). At 30 infectious viral particles per cell, the number of migrated cells was 22% higher (P < 0.05) than in the control group. Cells that received t-PA gene transfer, however, migrated at the same rate as control virus-infected cells at all multiplicities of infection. At a MOI of 30, essentially 100% of the cells demonstrate expression of the transgene as determined by AdCMVβ-gal marker virus experiments (data not shown).

Figure 1
Figure 1

In vitro tumor cell migration. HT1080 cells were infected with recombinant adenovirus on the day before the migration assay. HT1080 cells (5 × 104) were seeded on top of Transwell inserts coated with Matrigel. After 18 h, cells on the upper side of the filter were mechanically removed, filters were fixed with 3.7% formaldehyde and stained with hematoxylin. The number of migrated cells was determined by counting 10 × 0.07 mm2 fields at × 400 magnification. Values are the mean ± standard deviation. The asterisk indicates values significantly different from controls (P < 0.05). (a) Effects of adenoviral gene transfer of human t-PA or u-PA on cell migration. AdRR5 control virus (n = 8, black bars), AdCMVt-PA (n = 8, hatched bars) and AdCMVu-PA (n = 8, striped bars). (b) Inhibition of cell migration by adenoviral gene transfer of human PAI-1 or PAI-2. AdRR5 control virus (n = 4, black bars), AdCMVPAI-2 (n = 4, gray bars) and AdCMVPAI-1 (n = 4, white bars). (c) Correlation of cell migration with u-PA levels in conditioned medium. The line was fitted by linear regression (r = 0.904, P  0.001). Values were obtained by ELISA of conditioned medium from parallel cultures of infected cells. (d) Inverse correlation of cell migration with PAI-1 levels in conditioned medium. The line was fitted by linear regression (r = −0.873, P  6 0.001).

In order to confirm that cell migration was dependent on the proteolytic activity of u-PA in the in vitro model system, the expression of its specific inhibitors PAI-1 and PAI-2 was also increased by viral gene transfer. PAI-1 expression reduced migration across the reconstituted matrix components in a dose-dependent way at all viral doses (P < 0.05), with a maximum reduction of 43% at an MOI of 30 (Figure 1b). At the highest viral dose, a 29% reduction for PAI-2 (P < 0.05) was observed in the number of migrated cells as compared with controls.

Parallel cultures of HT1080 cells infected for the migration studies described above were used to condition medium. ELISA was used to quantify secreted u-PA and PAI-1 antigen and correlate the levels of protein expression with cell migration. As shown in Figure 1c, an increase in the levels of u-PA secreted into the culture medium by u-PA gene transfer is positively correlated with an increase in cell migration, whereas a decrease in u-PA secretion by PAI-1 gene transfer reduces cell migration. Conversely, an increase in PAI-1 secretion by adenoviral PAI-1 gene transfer reduces cell migration and a decrease in PAI-1 secretion by u-PA gene transfer increases cell migration (Figure 1d). These experiments directly link the levels of synthesis and secretion of u-PA and PAI-1 by the tumor cells with their ability to migrate and presumably to metastasize.

Unlike PAI-1, PAI-2 gene transfer into HT1080, 28.9.21 cells did not lead to significant increases in secretion into the medium. Uninfected cells secreted background levels of PAI-2 (approximately 0.4 ng/106 cells per hour), which was only slightly increased to 1.3 ng/106 cells per hour by infection with AdCMVPAI-2 at an MOI of 30. This apparently low level of expression was solely due to the inefficient secretion of PAI-2. Cell extracts of uninfected HT1080 28.9.21 cells showed 3.4 ng/106 cells, and cells analyzed 40 h after infection demonstrated 180, 490 and 2400 ng/106 cells for MOIs of 3, 10 and 30, respectively. By contrast to PAI-1, PAI-2 gene transfer did not result in a decrease in u-PA antigen in conditioned medium at any MOI tested.

Metastasis after in vitro gene transfer

To examine whether the gene transfer of PAI-1 or PAI-2 also influenced metastasis in vivo, human HT1080 cells infected with recombinant adenovirus were implanted in nude mice. Luciferase activity in organ homogenates was used to determine both the incidence of metastasis and to quantify the burden of tumor cells in individual organs. Adenoviral gene transfer of plasminogen activator inhibitors before cell implantation reduced the incidence of metastasis to the lungs at 26 days by 60% for PAI-1 (P = 0.015) and by 64% for PAI-2 (P = 0.009) (Figure 2a). This was the earliest time-point at which lung metastasis could be reproducibly detected in 100% of animals. The low frequency of metastasis by HT1080 cells to the liver and brain allowed no meaningful analysis of the effect of in vitro gene transfer on the incidence of metastasis. Metastatic burden (Figure 2b) was not significantly different in any organ following treatment of tumor cells with adenoviral vector expressing either plasminogen activator inhibitor.

Figure 2
Figure 2

Effects of in vitro gene transfer on tumor cell metastasis. HT1080 tumor cells expressing luciferase were infected with recombinant adenovirus at an MOI of 30 in vitro. Subsequently, 2 × 106 cells were implanted subcutaneously into nude mice. Animals were killed 26 days after tumor implantation and lungs, livers and brains assayed for luciferase activity. The asterisk indicates values significantly different from controls (P < 0.05) by ANOVA. (a) Incidence of metastasis. Samples with a signal exceeding background by 20% were considered to be positive for metastasis. Animals received cells infected with either control AdRR5 virus (n = 10, black bars), AdCMVPAI-1 (n = 10, white bars) or AdCMVPAI-2 (n = 11, gray bars). (b) Burden of metastasis in individual organs. Luminescence signals from positive organs were used to calculate the mean values for total organ burden in light units (lu), which are shown as solid squares. In the box plot, median and 25–75% percentiles are indicated by the open boxes, and 10–90% percentiles are given by the error bars. (c) Box plot comparing the weight of the primary tumors. The number of animals in each group is given in parentheses.

The effect of gene transfer of plasminogen activator inhibitors was also measured on the growth of the primary tumors (Figure 2c). Primary tumor weights were 1.4 g (median, range 0.15–2.90 g; n = 21) for the controls, 0.83 g (median, range 0.22–3.17 g; n = 22) for the PAI-1 group and 0.78 g (median, range 0.12–1.93 g; n = 11) for the PAI-2 group. Tumors in both the PAI-1 and PAI-2 groups were significantly smaller than controls (P < 0.05). Confidence intervals (95%) were 1.2–1.8 g for controls, 0.75–1.4 g for the PAI-1 group and 0.45–1.2 g for the PAI-2 group.

Metastasis after in vivo gene transfer to the liver

Injection of recombinant adenovirus into the venous circulation targets most of the viral particles to the liver, where infected hepatocytes will synthesize and secrete the transgene product into the blood. Previous results have shown a dose-dependent increase in circulating human PAI-1 following AdCMVPAI-1 injection in mice.23 Following gene transfer with 5 × 108 p.f.u. recombinant adenovirus, antigen levels of human PAI-1 in mouse plasma were increased five-fold from 25 ng/ml (median, range 3–160 ng/ml; n = 20) in control animals to 120 ng/ml (median, range 25–630 ng/ml; n = 23) in PAI-1 virus-treated animals (Figure 3a). The high levels of human PAI-1 in control mice were due to PAI-1 secretion from the primary human tumor cells, as non-tumor-bearing animals showed no human PAI-1 in plasma. A linear correlation (r = 0.79) between PAI-1 levels and tumor weight was observed (Figure 3b). Plasma PAI-1 levels were approximately 21 ng/ml for every gram of primary tumor burden. As PAI-2 is only inefficiently secreted, PAI-2 levels in plasma were undetectable in both the treated and control groups, except for two of 31 treated animals that had 1.0 and 4.5 ng/ml PAI-2 in plasma.

Figure 3
Figure 3

Expression of human PAI-1 in HT1080 tumor-bearing mice. (a) PA1-1 antigen levels in AdRR5 control (n = 20) and AdCMVPAI-1 injected (n = 23) mice. Antigen levels were determined at the time of death on day 26 after tumor cell implantation. Mean values are given by the solid squares. Median and 25–75% percentiles are given by the open boxes, and 10–90% percentiles are given by the error bars. The asterisk indicates values significantly different from controls (P < 0.05). (b) Individual PAI-1 levels in control virus-injected mice were correlated with primary tumor weight. Linear regression was used to fit the data to the line (r = 0.79).

In contrast to the results obtained with in vitro gene transfer, only PAI-2 reduced the incidence of metastasis to the lung by 22% (P = 0.049) and to the brain by 50% (P = 0.05) following liver gene transfer (Figure 4a). The apparent 33% reduction in incidence of metastasis to the brain in PAI-1 treated animals was not statistically significant. The increase in circulating levels of PAI-1 and PAI-2 antigen could not be directly correlated with a decrease in either the incidence or the burden of metastasis. Quantification of the tumor burden in different groups showed only minor effects of gene transfer. Although the mean tumor burdens were lower in treated groups than controls, only the 2.5-fold reduction in liver burden was significant following PAI-2 gene transfer (P < 0.05).

Figure 4
Figure 4

Effects of in vivo gene transfer to the liver on tumor cell metastasis. Gene transfer was performed by injection of 5 × 108 p.f.u. of recombinant adenovirus via the tail vein. Animals were treated with either control AdPacI virus (black bars), AdCMVPAI-1 (white bars) or AdCMVPAI-2 (gray bars). Uninfected HT1080 cells (2 × 106) were implanted subcutaneously in the flanks of nude mice. The asterisk indicates values significantly different from controls. (a) The incidence of metastasis was measured at day 26 after implantation. Samples with a signal exceeding background by 20% were considered to be positive for metastasis. The number of animals in each group is shown in parentheses. (b) Burden of metastasis in individual organs. Luminescence signals from positive organs were used to calculate the mean values for total organ burden in light units (lu), which are shown as solid squares. Median and 25–75% percentiles are indicated by the open boxes and 10–90% percentiles are given by the error bars. The number of animals in each group is shown in parentheses. (c) Size of primary tumors: AdRR5 (n = 35), AdCMVPAI-1 (n = 29), AdCMVPAI-2 (n = 35).

Comparison of primary tumor size also showed a significant difference between treated groups and controls (Figure 4c). The median tumor weight in the control animals was 3.1 g (range 0.19–12.0, n = 35) compared to 1.3 g (range 0.14–9.8, n = 29) in the PAI-1 group and 1.7 g (range, 0.28–6.4, n = 35) in the PAI-2 group. Confidence intervals (95%) were 12.7–4.4 g for controls, 1.4–3.2 g for the PAI-1 group and 1.5–2.4 g for the PAI-2 group.

Effects of liver gene transfer on host survival

A survival study was performed to explore if the reduction of metastasis and smaller primary tumors had a beneficial effect on the disease outcome. Although the exact cause of death was not determined, the animals demonstrated cachexia and respiratory distress and therefore died of consequences of metastatic disease. At 42 days after tumor cell implantation, only three out of 21 animals in the control group, two out of 20 animals in the PAI-1 group and two out of 21 animals in the PAI-2 group remained alive (Figure 5). The cumulative probability of survival plotted as a function of time did not show any difference between treated and control groups. The mean survival time in the treated groups was 37.2 days for the control group, 39.5 days for the PAI-1 group and 39.4 days for PAI-2 group. This shows that the initial differences between treatment groups in the incidence of metastasis and smaller primary tumors found at day 26 did not result in significantly prolonged survival.

Figure 5
Figure 5

Survival of tumor-bearing animals following liver gene transfer. NMRI nude mice implanted with 2 × 106 HT1080 cells were intravenously injected with control AdRR5 virus (squares, n = 21), AdCMVPAI-1 (circles, n = 20) or AdCMVPAI-2 (triangles, n = 21) and checked daily for survival. The cumulative probability of survival was not significantly different.

Overexpression of plasminogen activator inhibitors and tumor perfusion

In order to address one possible mechanism responsible for the observed reduction in incidence of metastasis following ex vivo and in vivo gene transfer, primary tumor vascularization was determined. To quantify the extent of tumor microvascular density, blood perfusion of primary tumors was compared by embolization of 113Sn-labelled microspheres at day 16 after tumor implantation. This time-point was chosen because tumors were both small (approximately 0.3–0.4 g) and rapidly growing, criteria which are suggestive of active angiogenesis. Normalization of tumor perfusion to reference organs, the liver and kidney, were used to control for the variation in blood flow between animals. After in vitro gene transfer, the median ratios of tumor/liver perfusion in the PAI-1, PAI-2 and control groups were 0.49 (range 0.09–2.2; n = 8; P = NS versus control), 0.35 (range 0.13–1.6; n = 8; P = NS versus control) and 0.28 (range 0.12–1.8; n = 7), respectively (Figure 6). Similarly, no difference in tumor perfusion was observed following PAI-2 gene transfer by intravenous injection, as tumor/liver ratios were 0.72 (median, range 0.25–4.5; n = 19) and 0.93 (median, range 0.06–4.0; n = 16) for control and PAI-2 groups, respectively. PAI-1 expressing adenovirus significantly reduced blood flow to the tumor to 0.49 (median, range 0.13–1.6; n = 16; P < 0.05 versus control) if the tumor/liver ratio was used, but was not statistically significant if tumor/kidney ratios were compared. Thus, overexpression of plasminogen activator inhibitors had little if any effect on perfusion of tumor tissue following either gene transfer protocol.

Figure 6
Figure 6

Relative blood perfusion rates of tumors following gene transfer. Cells were infected before implantation (in vitro) or animals were treated by intravenous injection (in vivo). Tumor perfusion determined by microsphere embolization was normalized to that found in the liver for each animal. The number of animals in each group is shown in parentheses. The asterisk indicates a value significantly different from controls (P < 0.05) by ANOVA.


The aim of this study was to evaluate the role of plasminogen activators and their inhibitors in cell migration and tumor metastasis using adenoviral gene transfer. An HT1080 subclone (28.9.21) was isolated which was both highly migratory in vitro and metastatic in vivo and secreted large amounts of u-PA, t-PA and PAI-1. Although HT1080 human fibrosarcoma cells have not been previously shown to metastasize after subcutaneous implantation,21 stable transfection of these cells with the sensitive luciferase reporter gene permitted the detection of metastasis primarily to the lung, but also to a variety of other organs. This is in agreement with the report that this fibrosarcoma cell line was derived from a patient who died with metastatic malignancy.24

Clonal variation with respect to components of the fibrinolytic system has previously been reported in HT1080 subclones,25 as well as for other tumor cell lines.18 In the process of generating stably transfected cell clones expressing luciferase, we have isolated clones that both differed in their levels of u-PA expression and their ability to metastasize in vivo. Indeed, our HT1080 clone expressed low, but detectable, amounts of PAI-2. When stably transfected cell clones expressing plasminogen activator inhibitors are used to study cell migration, the possibility cannot be excluded that clonal variation in the expression of other genes involved in this cellular process, such as u-PA or u-PAR, distort the biological effects of the transgene. Thus, the use of adenoviral vectors to increase the expression of PAIs in mammalian cells permits their effect on cell migration to be studied in a clonal cell population that has been previously characterized with respect to its migratory and metastatic phenotype. This facilitates a more accurate comparison between different gene transfer treatments and controls.

If the balance of plasminogen activators and inhibitors was perturbed by adenovirus-mediated overexpression of either u-PA or inhibitors, the migratory behavior of these cells could be manipulated. While overexpression of u-PA promoted cell migration in vitro, both PAI-1 and PAI-2 reduced it. t-PA appeared not to be involved in the regulation of cell movement. The absence of an effect on cell migration may be due to the lack of fibrin in the in vitro assay, since t-PA is a fibrin-specific plasminogen activator.

The site of plasminogen activator inhibitor expression, as well as their stability and level of expression determines their efficacy in inhibiting metastasis in different animal models. While stably transfected cell lines that express PAI-117,18 or PAI-219 or local injection of recombinant PAI-2 protein20 show biological effects in various tumor cell lines, elevation of systemic PAI-1 levels in transgenic mice has no effect on lung colonization by B16 melanoma cells.26 Human ocular melanoma cells implanted orthotopically in the eyes of nude mice are reduced in their metastatic potential by both local gene transfer of PAI-1 into the tumor and systemic elevation of PAI-1 levels using liver gene transfer.22 The results presented here extend these previous observations. Both PAI-1 and PAI-2 were equally effective at inhibiting metastasis when expressed by the tumor cells, suggesting important autocrine and paracrine effects of these molecules. PAI-1, however, was not effective when overexpressed in the circulation. Although PAI-2 was only detectable in a few animals, it significantly reduced HT1080 metastasis to the lung and brain following liver gene transfer. One possible explanation is that PAI-1 quickly becomes latent after secretion, while this does not occur with PAI-2 even though it is inefficiently secreted. Perhaps a fully secreted form of PAI-212 would function even more effectively than the wild-type protein in this setting. Nevertheless, there may be important autocrine effects of PAI-2 on the tumor cells that are unrelated to cell migration. Indeed, expression of both PAI-1 and PAI-1 had significant effects on primary tumor mass following in vitro gene transfer which cannot be explained by reduced tumor perfusion.

The level of expression of u-PA by different tumor cell types was also a factor in determining the inhibition of metastasis by PAI gene transfer. While HT1080 cells secrete high levels of u-PA, the amounts secreted by ocular melanoma cells are almost undetectable. The 10-fold difference in u-PA expression by the two cell types might explain why PAI-1 liver gene transfer reduced metastasis and prolonged survival in the ocular melanoma model but not in the present study. The expression of other proteases, such as metalloproteinases, could also augment the metastasis of HT1080 cells by comparison to ocular melanoma. Plasminogen activation by u-PA has been shown to be directly involved in metalloproteinase activation in vivo that leads to degradation of elastin and collagen in the extracellular matrix.27

The expression of u-PA and its inhibitors could theoretically affect metastasis by alterations in angiogenesis and access of the tumor cells to the vascular system. Local overexpression of human PAI-1 in PC-3 human prostate carcinoma cells has previously been reported to reduce vascularity in primary tumors.18 Except for a two-fold effect on HT1080 tumor perfusion with PAI-1 liver gene transfer, the degree of tumor vascularization was largely unaffected by increasing the expression of either plasminogen activator inhibitor. The absence of an effect of in vitro PAI-1 gene transfer on perfusion in HT1080 primary tumors suggests that the reduction in metastasis observed can be attributed solely to the direct effects of plasminogen activator inhibitors on uPA-mediated tumor cell migration and not to differences in neoangiogenesis. Even though HT1080 cells synthesize VEGF,28 the angiogenic activity of VEGF is not associated with perturbations in the expression of uPA, tPA or PAI-1 in endothelial cells.29

The present data further emphasize the reduction of tumor cell migration and metastasis by u-PA inhibition. Nevertheless, the ability of tumor cells to access the vascular system and migrate to distant sites is an important control point in the metastatic process. Overexpression of PAI-1 shows no effect on metastasis when tumor cells are directly injected in the circulation,26 but it can effectively inhibit the process when tumor cells are forced to escape from a confined site.22 The antimetastatic effects of PAI gene transfer were only partially effective when tumor cells are injected subcutaneously (this report). Apparently, metastasis was only delayed by gene transfer, as survival was not markedly prolonged. Therefore, impaired intravasation of tumor cells from the tumor to the bloodstream is the most likely explanation for the antimetastatic effects of u-PA inhibition.

While gene therapy of malignant tumors using suicide genes and cytokines has already progressed to the stage of clinical trials, it remains to be seen whether overexpression of plasminogen activator inhibitors by viral vectors could be used as an effective adjunctive therapy to inhibit metastasis. Research and development of small molecules that are both potent and specific for u-PA might be a preferable alternative.30

Materials and methods

In vitro assays

Luciferase assay:

Samples were prepared by addition of lysis buffer (25 mM Tris-phosphate pH 7.8, 2 mM dithiothreitol (DTT), 2 mM 1, 2 diaminocyclohexane-N,N,N′,N′-tetraacetic acid, 10% glycerol, 1% Triton × 100). Assays were performed in 96-well plates in a MICRO-Lumat LB96P luminometer (EG&G Berthold, Germany). Samples or extracts (20 μl) were mixed with 30 μl L-buffer (25 mM Tris-HCl pH 8.0, 2 mM EDTA, 2 mM DTT, 10% glycerol). Luciferin reagent (50 μl of 20 mM tricine pH 7.8, 1.07 mM (MgCO3)4Mg(OH)2*5H2O, 2.67 mM MgSO4, 0.1 mM EDTA, 33.3 mM DTT, 270 μM coenzyme A, 470 μM luciferin, 530 μM ATP) was injected and luminescence was measured for 1 min.

ELISAs for human plasminogen activators and inhibitors:

Human u-PA antigen was assayed using a sandwich ELISA consisting of monoclonal antibody MA-4D1E8 as capture antibody and MA-2L3-horseradish peroxidase conjugate for detection.31 Human t-PA antigen was assayed using a sandwich ELISA consisting of two monoclonal antibodies (MA-62E8 and MA-3B6) for antigen capture and MA-29b9-horseradish peroxidase conjugate for detection.32 Human PAI-1 antigen was assayed using a sandwich ELISA consisting of monoclonal antibody MA-15H12 as capture antibody and MA-12A4-horseradish peroxidase conjugate for detection.33 An Imubind human PAI-2 ELISA kit was obtained from American Diagnostica (Greenwich, CT, USA). Assay of human PAI-2 was performed according to the manufacturer’s directions using the standards provided. Mouse plasma (20% vol/vol) was found partially to inhibit the binding of the human PAI-2 standards provided. All plasma values were therefore corrected by a factor of 1.5. The lower limit of detection in the various ELISA assays was approximately 0.5–1.0 ng/ml.

Cell culture procedures

All cell lines were cultured in Dulbecco’s minimal essential medium (DMEM; GibcoBRL, Gaithersburg, MD, USA). Medium was supplemented with 10% (vol/vol) fetal bovine serum (FBS; GibcoBRL). Standard cell culture conditions were 37°C and 5% CO2. Human embryonic retinoblasts transformed with adenoviral left end sequences (911 cells)34 were obtained from IntroGene (Leiden, The Netherlands) and were used for propagation and titration of recombinant adenoviral vectors.

Recombinant adenovirus

Recombinant adenovirus expressing human PAI-123 and empty vectors AdPacI (containing a PacI restriction site) and AdRR5 (containing a pUC19 polylinker)35 have been previously described. All recombinant vectors contain a deletion of np 455–3333, inclusive, in the E1 region, and the dl309 E3 deletions/insertion of salmon DNA,36 except for AdPacI which contains the dl327 XbaI deletion in E3.

Construction of recombinant AdCMVu-PA and AdCMVPAI-2:

A plasmid containing the human pro-urokinase cDNA, pu-PA, was provided by William Bennett of Genentech (San Francisco, CA, USA). The uPA cDNA fragment from HindIII to SspI, containing 75 bp of 5′ untranslated and 600 bp of 3′ untranslated sequence, was inserted into the adenovirus construction plasmid, pACskCMV2. This plasmid contains the left end of adenovirus type 5 including the origin of replication and packaging sequences (np 1–454), the SV40 ori/hGH terminator from pCMV5,37 a polylinker to facilitate cDNA insertion, the CMV promoter and flanking adenovirus sequences (np 3334–5779) serving as the target for homologous recombination. It was kindly provided by Joseph Alcorn of UT Southwestern Medical Center (Dallas, TX, USA). The resulting plasmid, pACskCMVu-PA, was cotransfected with ClaI digested Ad5dl309 into 293 cells using standard methods.35

The cDNA of human PAI-238 provided by Toni Antalis (Queensland Institute of Medical Research, Brisbane, Australia) was used for the construction of AdCMVPAI-2. An EcoRI–HindIII fragment was subcloned into the shuttle plasmid pACCMVpLpA(−)35 to yield pACCMVPAI-2. For the generation of recombinant adenovirus, pACCMVPAI-2 (5 μg) and pJM17 (2 μg) were cotransfected into 293 (Ad5-transformed human embryonic kidney) cells using standard procedures.35

Viral clones were verified by restriction enzyme analysis and Southern blotting. Secretion of functional u-PA activity into conditioned medium by virus-infected 293 cells was verified by standard chromogenic substrate assay for plasminogen activation.39 Expression of functional PAI-2 by the vector was assessed in a similar manner by testing for u-PA inhibition. Viral stocks were prepared and frozen in aliquots at −80°C. Viral titers were determined on 6-cm dishes of 911 cells by plaque assay, using absorption for 1 h at 37°C in 0.5 ml medium, as described previously.35

Preparation of high titer purified adenovirus:

Large-scale production and purification of recombinant adenovirus on CsCl gradients was performed as described.35 Virus was supplemented with 10% glycerol (vol/vol) and frozen in aliquots at −80°C. Titers of infectious particles were determined as described above and were routinely greater than 1010 plaque forming units per ml. The p.f.u. per particle ratio was typically 10–12.

Generation of HT1080 clones stably expressing luciferase

Generation of expression plasmid of firefly luciferase:

The complete luciferase transcription unit on a NotI restriction fragment consisting of the cDNA, a CMV promoter and the SV40 splice/polyadenylation signal was subcloned from pACCMVluc40 into pbluescript (Stratagene, La Jolla, CA, USA) to yield pBSCMVluc.

Transfection of HT1080 fibrosarcoma cells:

The standard protocol for calcium phosphate mediated transfection of adherent eukaryotic cells was used for the cotransfection of pBSCMVluc (1 μg) and pkOneo41 (200 ng) into human HT1080 cells. Transformants were selected with 300 μg/ml G-418 and screened for luciferase expression. Positive clones were subcloned twice to ensure homogeneity of the population.

In vitro migration assay

HT1080 cells were infected with recombinant adenovirus on the day before the migration assay. 12-mm Transwells (Costar, Cambridge, MA, USA) with polycarbonate membranes (12 μm pore size) were coated with 400 μl Matrigel (550 μg/ml) for 1 h at 37°C. Cells (5 × 104) were seeded on top of the Matrigel coat in DMEM containing 2% FBS and allowed to migrate towards DMEM containing 10% FBS for 18 h. Cells were fixed in 3.7% formaldehyde in PBS and stained with hematoxylin for light microscopy. Non-migrated cells were removed from the upper chamber with a cotton swab to aid visualization of migrated cells. The number of migrated cells in an individual well was determined by counting 10 random microscopic fields of 0.07 mm2 at × 400 magnification.

Animals and experimental protocols

Female NMRI nude mice, 5–8 weeks old, were purchased from CERJ France (Le Geneste St Isle, France). Animal use and care was according to institutionally approved protocols.

In vitro gene transfer protocol:

Cells were infected in vitro at a multiplicity of infection (MOI) of 30 plaque forming units (p.f.u.) per cell with recombinant adenovirus in DMEM supplemented with 2% FBS. Two hours after addition of the infectious medium, cells were trypsinized and cell density was adjusted. Infected cells (2 × 106 in 300 μl DMEM) were implanted subcutaneously into the flank of nude mice and allowed to form tumors.

Liver gene transfer protocol:

Liver gene transfer was performed by intravenous injection of adenovirus into the tail vein of mice with 5 × 108 p.f.u. recombinant virus in 250 μl isotonic saline. On the same day, clone 28.9.21 HT1080 cells (2 × 106) that stably express luciferase were implanted subcutaneously as described above.

Assay of metastasis:

Animals were killed on day 26 after tumor cell implantation. Citrate-treated mouse plasma (1:10; vol/vol) was obtained from vena cava for PAI antigen determination. Lungs, liver, brain and the primary tumor were excised and the tissue specimens were weighed. The liver was divided into two samples that were assayed individually and the results later combined. Lung, brain, liver and tumor samples were homogenized in 2 ml luciferase lysis buffer. Cell debris was removed by centrifugation and luciferase activity was assayed as described above on 20 μl samples. Light units per organ were calculated by correcting for instrument background, sample weight and dilution factor. In order to assess incidence of metastasis, organs that had a luminescence signal 100 light units above instrument background (500 light units) were rated positive. This value was chosen because it represented the minimum signal which could be detected which was reproducibly above background. For reference purposes, minimum values for organs were 11000 light units for the lung, 11750 for the brain and 15000 for the liver when multiplied by the total homogenized tissue sample. Burden of metastasis was calculated by statistical analysis of positive luminescence signals.

Survival study after liver gene transfer:

Liver gene transfer and implantation of HT1080 tumor cell clone 28.9.21 was performed as described above. Animals were checked daily for survival. Animals that became moribund or comatose were killed and analyzed on that date.

Blood perfusion studies:

Blood perfusion of tumor tissue was measured following in vitro or liver gene transfer as described above. On day 16 after tumor cell implantation, animals were anesthetized by intraperitoneal injection of 3.3 mg pentobarbital. Animals were ventilated after a right parasternal thoracotomy and 2.5 μCi 113Sn-labelled microspheres 15.5 μM in diameter (DuPont-NEN, Boston, MA, USA), corresponding to approximately 120000 microspheres, were injected into the left ventricle of the heart. This method for measuring blood flow to organs is both sensitive and accurate and has been described previously.42 Animals were killed 3 min after injection by cervical dislocation. Liver, kidney and tumor tissue were excised, blotted on tissue paper, weighed and γ radiation was counted in a MINAX γ counter (Packard, Downers Grove, IL, USA). Perfusion was expressed as counts per miiligram wet tissue and the ratios of tumor/liver and tumor/kidney were calculated.

Statistical analysis

Values are expressed as means ± standard deviation unless otherwise noted. Differences in tumor incidence were calculated by χ2 contingency table analysis with continuity correction and differences in tumor weight was assessed by the Mann–Whitney rank sum test. One factor ANOVA with Scheffe’s post-hoc test was used to determine significance for all other statistics. P values of <0.05 were considered significant.


We thank Huberte Moreau, Veerle Laurysens and Benny Buelens for their expert technical assistance. MP was the recipient of a post-doctoral fellowship from the Ernst Schering Research Foundation.


  1. 1.

    , , , . Tumor invasion, proteolysis and angiogenesis J Neuro-Oncol 1994 18: 89–103

  2. 2.

    , , , . The urokinase-type plasminogen activator system in cancer metastasis: a review Int J Cancer 1997 72: 1–22

  3. 3.

    , . Basic and clinical aspects of fibrinolysis and thrombolysis Blood 1991 78: 3114–3124

  4. 4.

    . Urokinase and urokinase receptor: a paracrine/autocrine system regulating cell migration and invasiveness BioEssays 1993 15: 105–110

  5. 5.

    , , , . Distribution of plasminogen activator inhibitor (PAI-1) in tissue J Clin Pathol 1991 44: 139–143

  6. 6.

    , . Induction of plasminogen activator inhibitor 1 gene expression in murine liver by lipopolysaccharide. Cellular localization and role of endogenous tumor necrosis factor-alpha Am J Pathol 1997 150: 579–590

  7. 7.

    , , , . Plasminogen activator inhibitor 1 (PAI) is bound to vitronectin in plasma FEBS Lett 1988 19: 125–128

  8. 8.

    , . The serpin PAI-1 inhibits cell migration by blocking integrin avb3 binding to vitronectin Nature 1996 383: 441–443

  9. 9.

    et al. Plasminogen activator inhibitor-1 represses integrin- and vitronectin-mediated cell migration independently of its function as an inhibitor of plasminogen activation Exp Cell Res 1997 232: 420–429

  10. 10.

    et al. Plasmin and plasminogen activator inhibitor type 1 promote cellular motility by regulating the interaction between the urokinase receptor and vitronectin J Clin Invest 1997 100: 58–67

  11. 11.

    et al. Urokinase binds to a plasminogen activator inhibitor type-2-like molecule in placental microvillous membranes Biochim Biophys Acta 1989 986: 135–140

  12. 12.

    et al. The efficiency of the uncleaved secretion signal in the plasminogen activator inhibitor type 2 protein can be enhanced by point mutations that increase its hydrophobicity J Biol Chem 1991 266: 15240–15243

  13. 13.

    , . Protection from tumor necrosis mediated cytolysis by overexpression of plasminogen activator inhibitor type-2 J Biol Chem 1991 266: 20960–20964

  14. 14.

    , , , . Plasminogen-activator inhibitor type 2 (PAI-2) is a spontaneously polymerising SERPIN. Biochemical characterisation of the recombinant intracellular and extracellular forms Eur J Biochem 1993 218: 1071–1082

  15. 15.

    , . Binding of urokinase to its receptor promotes migration and invasion of human melanoma cells in vitro Cancer Res 1994 54: 3066–3071

  16. 16.

    , , , . The effect of antisense inhibition of urokinase receptor in human squamous cell carcinoma on malignancy EMBO J 1994 13: 3983–3991

  17. 17.

    et al. Tissue-type plasminogen activator induced invasion and metastasis of murine melanomas Curr Eye Res 1995 14: 449–458

  18. 18.

    et al. Expression of plasminogen activator inhibitor type 1 in human prostate carcinoma cells inhibits primary tumor growth, tumor-associated angiogenesis, and metastasis to lung and liver in athymic mouse model J Clin Invest 1996 96: 2593–2600

  19. 19.

    , , . Overexpression of plasminogen activator inhibitor 2 in human melanoma cells inhibits spontaneous metastasis in scid/scid mice Proc Natl Acad Sci USA 1995 92: 205–209

  20. 20.

    , , . Urokinase-producing tumour growth in SCID mice inhibited by recombinant PAI-2 Fibrinolysis 1995 9: 175–177

  21. 21.

    et al. Inhibition of invasion of HT1080 sarcoma cells expressing recombinant plasminogen activator inhibitor 2 Cancer Res 1993 53: 6051–6057

  22. 22.

    et al. Inhibition of metastasis of intraocular melanomas by adenoviral gene transfer of plasminogen activator inhibitor type 1 (PAI-1) in an athymic mouse model Blood 1997 90: 2738–2746

  23. 23.

    et al. Inhibitory role of pharminogen activator inhibitor-1 in arterial wound healing and neointima formation. A gene targeting and gene transfer study in mice Circ 1997 96: 3180–3190

  24. 24.

    et al. Characterization of a newly derived human sarcoma cell line (HT-1080) Cancer 1974 33: 1027–1033

  25. 25.

    et al. Clonal variation of expression of the genes coding for plasminogen activators, their inhibitors and the urokinase receptor in HT-1080 sarcoma cells Int J Cancer 1992 52: 298–304

  26. 26.

    et al. Lack of plasminogen activator inhibitor-1 effect in a transgenic mouse model of metastatic melanoma Blood 1996 87: 4718–4722

  27. 27.

    et al. Urokinase-generated plasmin is a candidate activator of metalloproteinases during atherosclerotic aneurism formation Nature 1997 17: 439–444

  28. 28.

    , , . Significance of vascular endothelial growth factor/vascular permeability factor for solid tumor growth, and its inhibition by the antibody Biochem Biophys Res Commun 1993 194: 1234–1241

  29. 29.

    et al. Interaction of vasculotropin/vascular endothelial growth factor with human umbilical vein endothelial cells: binding, internalization, degradation and biochemical effects J Cell Physiol 1991 149: 50–59

  30. 30.

    , , , . Why drinking green tea could prevent cancer Nature 1997 387: 561

  31. 31.

    , , , . An enzyme-linked immunosorbent assay for urokinase-type plasminogen activator (u-PA) and mutants and chimeras containing the serine protease domain of u-PA Thromb Haemostas 1992 67: 95–100

  32. 32.

    , , . Assay of tissue-type plasminogen activator with enzyme-linked immunosorbent assay (ELISA) based on three murine monoclonal antibodies to t-PA Thromb Haemostas 1985 54: 684–687

  33. 33.

    et al. Measurement of PAI-1 in biological fluids with murine monoclonal antibody-based enzyme linked immunoassay Blood 1988 8: 220–225

  34. 34.

    et al. Characterization of 911: a new helper cell line for the titration and propagation of early region 1-deletedadenoviral vectors Hum Gene Ther 1996 7: 215–222

  35. 35.

    , . Adenovirus vectors. In: Hames BD, Glover D (eds) DNA Cloning – A Practical Approach: Mammalian Systems Oxford University Press: Oxford, 1995, pp 285–307

  36. 36.

    , , . DNA sequence of the deletion/insertion in early region 3 of Ad5dl309 Virus Res 1995 39: 75–82

  37. 37.

    , , , . Human 3β-hydroxysteroid dehydrogenase/Δ54isomerase from placenta: expression in nonsteroidogenic cells of a protein that catalyzes the dehydrogenation/isomerization of C21 and C19 steroids Endocrinology 1990 126: 2493–2498

  38. 38.

    et al. Cloning and expression of a cDNA coding for a human monocyte-derived plasminogen activator inhibitor Proc Natl Acad Sci USA 1988 85: 985–989

  39. 39.

    , , , . Mutational and immunochemical analysis of plasminogen activator inhibitor 1 J Biol Chem 1990 265: 18379–18385

  40. 40.

    , . Adenovirus-mediated transfer of low density lipoprotein receptor acutely accerates cholesterol clearance in normal mice Proc Natl Acad Sci USA 1993 90: 2812–2816

  41. 41.

    , , . Infection of eucaryotic cells by helper-independent recombinant adenoviruses: early region I is not obligatory for integration of viral DNA J Virol 1994 50: 606–614

  42. 42.

    , , . Reproductive organ blood flow measured using radioactive microspheres in diestrous and estrous mice Reg Integ Comp Physiol 1992;31: R666–R670

Download references

Author information


  1. Center for Transgene Technology and Gene Therapy, Flanders Interuniversity Institute for Biotechnology, KU Leuven, Belgium

    • M Praus
    • , K Wauterickx
    • , D Collen
    •  & R D Gerard


  1. Search for M Praus in:

  2. Search for K Wauterickx in:

  3. Search for D Collen in:

  4. Search for R D Gerard in:

Corresponding author

Correspondence to R D Gerard.

About this article

Publication history





Further reading