Tissue-type and urokinase plasminogen activators (t-PA and u-PA, respectively) are serine proteases that catalyse the activation of plasminogen in vivo.¹ The activity of plasmin is critical in a number of physiological processes, including extracellular proteolysis, cell migration, tissue remodelling, and angiogenesis.^2,3 Plasmin activity is also required for tumour progression, and the role of u-PA in activating plasminogen is well documented.^4,5 The activity of t-PA correlates with a poor prognosis in several cancers, including melanomas,^6–8 neuroblastomas,^9,10 acute non-lymphocytic leukaemia,¹¹ and cancer of the exocrine pancreas where its overexpression promotes growth and angiogenesis in vivo.^12–15

Plasminogen is readily activated by t-PA and u-PA when both are bound to cell surfaces.¹⁶ Plasmin generated at these sites, protected from the activity of inhibitors such as α₂-antiplasmin and α₂-macroglobulin,¹⁷ is responsible for most of the proteolytic activity generated at membrane sites. Several proteins may function as binding sites for plasminogen or as co-binding sites for plasminogen and t-PA.^18–21 Annexin II, a cytosolic phospholipid and Ca²⁺ binding protein, is a co-receptor for t-PA and plasminogen in endothelial cells²² that can form heterotetrameric complexes on the surface of HUVEC cells with the annexin II light chain (called S100A10 or p11) and this stimulates generation of t-PA dependent plasmin.^17,23 The annexin II/t-PA interaction is required for activation of microglia,²⁴ and for plasmin generation in nerve growth factor activated neuritogenesis.²⁵ Amphoterin, a heparin binding protein, binds t-PA and promotes neurite outgrowth in neuroblastoma cells.²⁶ Cytokeratin 18 and 8 bind t-PA in hepatocellular and breast carcinoma cells.²⁷ In addition, interactions of t-PA with surface membrane proteins dependent on its kringle domains or the catalytic site have been described.^28–31 Very recently, the transmembrane protein CKAP4 has been identified as a specific t-PA binding protein on the surface of vascular smooth muscle cells.³²

We have previously shown that overexpression of t-PA in pancreatic cancer cells potentiates their invasive capacity and induces proliferation and faster tumour growth in vivo.^13–15 The mechanisms by which t-PA promotes its different functions in pancreatic cancer cells have not been investigated. In the present work, we focused on the molecules that specifically bind active t-PA on the surface of cells and participate in invasion.

EXPERIMENTAL PROCEDURES

Cell culture and reagents

Cell lines were obtained from American Type Culture Collection (Rockville, Maryland, USA) and maintained in Dulbecco’s modified Eagle’s medium (DMEM) (Gibco-BRL, Gaithesburg, New York, USA) supplemented with 10% heat inactivated fetal bovine serum (FBS; Gibco-BRL) at 37°C and 5% CO₂.¹⁵ Goat antibodies recognising t-PA were from American Diagnostica (Greenwich, Connecticut, USA); antibodies to annexin I and annexin II were from Transduction Laboratories (Lexington, Kentucky, USA); anti-ERK 1,2, from Upstate Biotechnology (Lake Placid, New York, USA); antimouse FITC biotin labelled rabbit antigoat IgG and mouse IgGs from Dako (Glostrup, Denmark); antigoat TRITC from Pierce (Rockford, Illinois, USA); diisopropyl fluorophosphate (DFP), goat IgGs, streptavidin-agarose, and protein G agarose beads from Sigma (St Louis, Missouri, USA); and plasminogen and recombinant t-PA (rt-PA, Actilyse) from Boehringer Mannheim (Barcelona, Spain). Peptides LCKLSL and LGKLSL were synthesised at the Serveis Científico Tècnics, Universitat Pompeu Fabra (Barcelona, Spain).

Functional cell membrane binding assay for t-PA

Interaction of t-PA with the surface of pancreatic tumour cells was studied by analysing the plasminogen activating capacity of bound recombinant t-PA (rt-PA).^30,31 Briefly, cells were seeded (7×10⁴ cells/well) in 24 well plates and grew to confluence. Endogenous plasminogen activators were removed from cell surfaces with acid glycine buffer (0.05 M glycine (pH 3), 0.1 M NaCl) for three minutes and neutralised with 0.5 M HEPES (pH 7.5) and 0.1 M NaCl. Cells were incubated with different concentrations of rt-PA for 20 minutes at 37°C in phosphate buffered saline (PBS) containing 2% bovine serum albumin (BSA, binding buffer). After washing with binding buffer, bound t-PA was determined by incubating cells directly in 0.05 M Tris HCl, 0.1 M NaCl, pH 7.4, with plasminogen (0.2 μM) and H-D-Val-Leu-Lys-7-amido-4-methylcoumarin (AMC 0.2 mM) (Bachem, Budendorf, Switzerland), a plasmin specific fluorogenic peptide substrate.³⁰

To determine the linearity of plasminogen activation, fluorescence levels (relative fluorescence units (rfu)) were measured at different times after incubation with substrates at excitation and emission wavelengths of 380 nm and 480 nm, respectively, using a Fluoroskan Reader (Labsystems, Helsinki, Finland).

Equilibrium, saturation, and competition binding assays

To determine the time needed to reach equilibrium, PANC-1 cells were incubated with rt-PA (15 nM) at 37°C for different time periods, cells were washed, and plasminogen activation was determined. For equilibrium saturation binding assays, PANC-1 cells were incubated with rt-PA at different concentrations (0–500 nM) for 20 minutes, and plasmin generation was determined. To convert rfu data to t-PA protein concentration, a standard t-PA in a fibrin coupled enzymatic assay was used (Desafib-X, Biopool, Sweden).³³ Competitive binding assays were performed with peptides blocking the interaction of t-PA with annexin II (LCKLSL) or control peptides (LGKLSL) used at 0.1–10 mM.³⁴ To remove annexin II from cell surfaces, cells were rinsed with PBS-EDTA (0.5 mM) and incubated with PBS-EDTA plus EGTA (10–50 mM) for 20 minutes.³⁵ Supernatants from treated cells were concentrated by centrifugation with Ultrafree-4 filters Biomax-10K (Millipore, Bedford, Massachusetts, USA). To assay cells in suspension for t-PA binding, cells were trypsinised, washed, and resuspended in fresh medium for one hour at 37°C in polypropylene tubes before proceeding to the binding reaction.

For SK-PC-1 cells (overexpressing t-PA), to avoid contamination with the endogenous protein, bound rt-PA was rapidly eluted from cells in acid buffer, neutralised, and its activity measured using a fibrin coupled enzymatic assay.

Analyses of binding data

Equilibrium and saturation binding data were analysed by non-linear regression algorithms using Sigma Plot software (SPSS, Chicago, Illinois, USA). For saturation binding analyses, data were fit to a one or two binding site model.

Western blotting and immunoprecipitation

Western blotting was performed as described previously.¹⁵ For immunoprecipitation, subconfluent SK-PC-1 cells were lysed in buffer I (20 mM Tris HCl, pH 6.8, 3 mM MgCl₂, 50 mM NaCl, 300 mM sucrose, 1% Triton X-100 (v/v)) containing 1 mM phenylmethylsulphonyl fluoride and 2 μg/ml each of aprotinin, leupeptin, and pepstatin.³⁶ After centrifugation (13 000 g) the pellet containing annexin II was dissolved in buffer II (50 mM Tris HCl, 150 mM NaCl, 1% NP-40, 10 mM EGTA). Proteins (400 μg) from the supernatant and pelletted fractions were used for immunoprecipitation with antiannexin II, anti-t-PA, antiannexin-I antibodies, or goat IgGs, and recovered by protein G sepharose beads or antigoat biotinylated IgG and streptavidin agarose beads. After washing with buffer II without EGTA, beads were resuspended in Laemmli buffer, and immunoprecipitated proteins were identified by western blotting.

Immunohistochemical staining and confocal microscopy

Sections (5 μm) from paraffin embedded tissues were treated with heat in citrate buffer for antigen retrieval. Sections were incubated with antiannexin II and anti-t-PA antibodies, rinsed, and incubated with purified FITC conjugated antimouse IgG and TRITC conjugated antigoat IgG. After rinsing, sections were mounted with Immuno-Fluore Mounting Medium (ICN, Costa Mesa, California, USA). Fluorescence was visualised on an inverted fluorescence microscope DM IRBE (Leica, Wetzlar, Germany) and captured in a TCS-NT argon/kripton confocal laser microscope (Leica). Control sections were incubated with matched mouse or goat IgG and were found to be negative. Incubations with rabbit secondary antibodies were also unreactive.

In vitro invasion assays

The invasive potential of cultured tumour cells was tested using Matrigel coated Transwell filters (Costar, Cambridge, Massachusetts, USA), as previously described.¹⁵ Briefly, 1×10⁵ cells were resuspended in DMEM supplemented with 0.1% BSA and plated on the Matrigel coated filters in the upper chamber, and medium containing 10% FBS was placed in the bottom chamber. Plates were incubated at 37°C for 24, 48, or 72 hours, after which cells that reached the bottom surface of the filters were visualised by staining with crystal violet. Peptides LCKLSL and LGKLSL were tested on cell invasion by mixing them with cells before seeding. Quantitative analyses were performed by counting stained cells on the filters in 10 independent fields/Transwell. Invasion abilities of cells were assigned as follows: highly invasive (+++), ⩾100 cells/field at 24 hours; moderately invasive (++), ⩾100 cells/field at 48 hours; and poorly invasive (+), 50–100 cells/field at 72 hours. Quantitative analyses of stained filters were also obtained by elution in 10% acetic acid and measuring absorbance at 595 nm.

Statistics

Results are expressed as mean (SEM) and the Student’s t test was used for statistical analysis. A p value of <0.05 was considered statistically significant in all cases.

RESULTS

t-PA binds specifically and saturably to the surface of pancreatic cancer cells

Because pancreatic cancer cells express plasminogen activator inhibitor 1 (PAI-1) (Díaz and colleagues¹⁵ and data not shown), the physiological inhibitor of t-PA that binds with high affinity and inactivates t-PA, and the t-PA/PAI-1 complex can bind to cell surface receptors,¹⁶ t-PA binding sites other than PAI-1 can be analysed by evaluating the fraction of active enzyme bound to the cell surfaces.³⁰ To analyse putative receptors for t-PA that maintain the protease proteolytically active in pancreatic cancer cells, we used rt-PA and measured the activity of the bound protease by its ability to activate plasmin and a fluorogenic plasmin substrate. Firstly, we determined the linearity of plasmin generation by incubating PANC-1 cells with rt-PA (15 and 250 nM) over 20 minutes and, after washing out the unbound ligand, measuring hydrolysis of the AMC substrate at different times. As shown in fig 1A, generation of proteolysed AMC substrate is linear up to 30 minutes for t-PA at 15 nM (r = 0,990). When t-PA is used at 250 nM, plasmin generation is linear up to 20 minutes (r = 0.983). Therefore, all subsequent determinations of bound t-PA activity were performed after 15 minutes of incubation with substrate.

Next, we determined the time for t-PA binding to reach equilibrium. As shown in fig 1B, binding of t-PA to PANC-1 cells at 37°C was time dependent and reached equilibrium after 20 minutes. Using non-linear regression analyses, the data depicted in fig 1B fit to a monoexponential curve, suggesting a reversible bimolecular interaction.

We then determined the saturability of t-PA binding by incubating PANC-1 cells with increasing concentrations of rt-PA. As shown in fig 2A, cells demonstrated dose related and saturable binding of t-PA that reached a plateau at input doses of 100 nM. Non-linear regression analyses of these data showed that the best fit was to a one site binding model, with a K_D of 42 nM. In parallel experiments, binding of t-PA to SK-PC-1 cells gave very similar results (K_D 23.5 nM; fig 2B). These values are very similar to previously reported binding affinity values to purified annexin II (K_D 25 nM²²), to 293 cells transfected with annexin II cDNA (K_D 48 nM³⁷), or to an unidentified receptor molecule on vascular smooth muscle cells (K_D 25 nM³⁰). The specificity of this binding was demonstrated by gradual inhibition of bound t-PA in cells treated with increasing concentrations of DFP inactivated rt-PA; at 75 M excess of competitor t-PA, binding was reduced by 73.5% (fig 3A).

To discard the fact that the observed binding of t-PA involves interactions with acellular components, such as proteins of the extracellular matrix, cells in suspension were also assayed. Results from these experiments indicated saturable binding of t-PA to PANC-1 cells in suspension, with a K_D of 34 nM, very similar to the results obtained previously (not shown). Thus pancreatic cancer cells express a surface protein capable of binding t-PA specifically and with high affinity while maintaining its activity, and thus may be the mediator of t-PA activities in these tumours.¹⁵

t-PA interacts with annexin II on the surface of pancreatic cancer cells

Because annexin II, the receptor for t-PA in endothelial cells, was previously determined to be overexpressed in pancreatic cancer,^13,38,39 we studied its contribution to the binding of t-PA to pancreatic cancer cells. Earlier work identified the sequence LCKLSL (residues 8–13) from the annexin II protein as the critical region involved in the interaction with t-PA.³⁴ This peptide inhibited 95% of binding of t-PA to annexin II.³⁴ As shown in fig 3B, binding of rt-PA to PANC-1 cells was significantly reduced to mean 60 (SEM 3)% by peptide LCKLSL but no inhibition was observed with the control peptide LGKLSL (binding was 99 (4)%), suggesting that annexin II may be involved in t-PA binding to pancreatic cells, and confirming that the Cys⁹ in annexin II is required for the interaction.³⁴ To prove these results, cells were treated with EGTA to reduce the amount of annexin II from the cell surface.³⁵ In the presence of EGTA, annexin II was recovered in the supernatant (fig 3C, inset) and binding of rt-PA to cells was reduced to 65.9 (4.4)% (fig 3 C). However, when EGTA was added in combination with peptide LCKLSL, t-PA binding was further reduced to 36.8 (2.4)%. These results support the notion that the interaction with annexin II accounts for most of the t-PA binding to pancreatic cell surfaces.

To confirm the interaction of t-PA with annexin II, we performed co-immunoprecipitation experiments using specific antibodies on SK-PC-1 cells that express high levels of both proteins. Isotype matched antibodies to annexin I were used as a control. Immunoprecipitations with antiannexin II antibody followed by western blotting analyses with anti-t-PA antibody revealed the presence of t-PA in immunoprecipitated proteins (fig 4). In contrast, no detectable t-PA was present in immunocomplexes formed by annexin I antibodies. Conversely, when cell extracts were immunoprecipitated with anti-t-PA antibody, annexin II was co-immunoprecipitated. However, not all t-PA secreted by SK-PC-1 cells co-immunoprecipitated with annexin II, and vice versa, suggesting that they may correspond to molecular forms that do not interact with each other.

t-PA and annexin II colocalise at the surface of human pancreatic tumours

To determine whether the interaction of t-PA and annexin II is present in vivo, we analysed the distribution pattern of annexin II and t-PA in human tumour tissue sections by double immunofluorescent staining and confocal microscopy. As shown in fig 5, both proteins were expressed at high levels in epithelial cancer cells and colocalisation was detected at membrane sites. t-PA and annexin II were also found at independent locations, in agreement with our previous results.

Annexin II and t-PA overexpression correlate with invasion in pancreatic cancer cells

Annexin II overexpression in pancreatic, colon, and gastric cancer cells correlates with a poor prognosis.^13,38–42 We studied expression of annexin II in several pancreatic cancer cell lines by western blotting and the invasion abilities of these cells in vitro, on Matrigel coated filters. As shown in fig 6, levels of annexin II were similar in all cell lines, despite differences in their ability to invade¹⁵ (table 1). Data summarised in table 1 show that levels of expression of t-PA, but not those of annexin II, directly relate to a higher capacity for invasion.

t-PA/annexin II interaction is critical for the invasive capacity of cancer cells in vitro

We analysed whether the interaction of t-PA with annexin II is required for cells to invade, through Matrigel in vitro. Figure 7 shows that SK-PC-1 cells efficiently invade Matrigel. However, in the presence of the blocking hexapeptide LCKLSL, cell invasion was significantly reduced to mean 62 (SEM 11)% of control cells at 72 hours. Invasion was also markedly inhibited by the use of this peptide at 24 and 48 hours. Invasion in the presence of a control peptide was no different from untreated cells, suggesting that the interaction of t-PA with annexin II is important for the pancreatic cell invasion process.

DISCUSSION

We have previously shown that t-PA, highly expressed in the majority of pancreatic tumours and cell lines but not in normal pancreatic tissue, is required for the invasion, proliferation, and angiogenesis of these tumours.^12,13,15 We have also shown that annexin II is overexpressed in pancreatic adenocarcinomas and localises at the basolateral compartment of SK-PC-1 cells.¹³ Localisation of active t-PA at the cell surface permits local activation of plasmin, imparts proteolytic and invasive advantages to tumour cells, and is also essential for t-PA dependent activation of signal transduction.¹⁵ However, so far the identities of the specific cell surface binding molecules for t-PA in pancreatic cancer cells have not been elucidated.

In the present study, we demonstrated the presence of high affinity binding sites on cell membranes that maintain t-PA as an active enzyme and contribute to the invasive properties of SK-PC-1 cancer cells. By measuring only active t-PA bound to extracellular membranes of cells, we have detected specific and saturable binding for t-PA in SK-PC-1 and PANC-1 cells. This binding was significantly inhibited by a specific peptide corresponding to the N terminus of the annexin II protein (residues 8–13), and also by treatment with EGTA that removes most of annexin II from the cell surface, suggesting that annexin II is a binding site for t-PA. In addition, the interaction was demonstrated by co-immunoprecipitation in cell extracts in vitro and co-immunolocalisation in tumours. Finally, the significance of the interaction was functionally demonstrated by inhibition of invasion detected in the presence of the blocking hexapeptide. Based on these data, we conclude that annexin II is a t-PA binding site on the surface of pancreatic cancer cells and that this interaction is important for the invasive capacity of cells.

In HUVEC cells, treatments with EGTA, antisense oligonucleotides to annexin II, or antiannexin II antibodies results in a reduction in specific binding of t-PA to 40–50% of controls, values comparable with our results in pancreatic cells obtained with EGTA or the peptide LCKLSL.^34,37 Interestingly, the inhibitory effect on t-PA binding to pancreatic cancer cells after treatment with EGTA combined with the peptide LCKLSL was higher than inhibition from individual treatments. These results may be explained by the following: (i) annexin II may be present on surface membranes as monomers or heterotetramers with p11^17,23; (ii) in HUVEC cells, a portion of annexin II, likely the heterotetrameric annexin II/p11, binds to membranes in a Ca²⁺ independent way and is thus resistant to EGTA treatment^23,43; and (iii) the activity of t-PA is increased when it is bound to the heterotetrameric annexin II with respect to monomeric annexin II.²³ Thus in our cells the efficacy of EGTA treatment on t-PA binding to annexin II may be partially occulted by the increased activity of t-PA bound to the tetrameric annexin II. In summary, inhibition of the interaction obtained by EGTA and the inhibitory peptide is significant, consistent with annexin II being a major binding site for t-PA on the surface of pancreatic tumour cells.

None the less, in polarised SK-PC-1 cells, t-PA is also bound to external apical membranes not containing annexin II (VD and RP, unpublished observations), suggesting that an additional protein(s) may function as a binding site for t-PA in these cells. In vascular smooth muscle cells, a high affinity binding site different from annexin II that maintains the protease active and shows a K_D (25 nM) very similar to that described here³⁰ has recently been identified as the transmembrane protein CKAP4.³² Amphoterin, that binds t-PA in the central nervous system and is implicated in invasion,^26,44 is also expressed by SK-PC-1 cells at high levels (VD and RP, unpublished observations). However, the non-saturable binding kinetics of t-PA to this site²⁶ suggest that it is not the binding site detected in our experiments. Cytokeratin 8 also binds t-PA in a specific and saturable manner in breast cancer cells⁴⁵ and promotes cell surface plasmin generation, but the reported K_D (250 (48) nM) is significantly different from that determined in our experiments.²¹ These observations suggest that pancreatic tumour cells may have more than one membrane molecule, with similar binding affinities for t-PA, possibly to focus plasmin generation at different membrane sites, the relevance of which in the process of tumour progression must be investigated.

Amphoterin, cytokeratin 8, and annexin II are cytosolic proteins lacking a signal sequence, and yet extracellular functions for these proteins have been described. In this regard, annexin II has been found in exosomes, an unconventional secretion pathway for proteins lacking a signal sequence.⁴⁶

Our co-immunoprecipitation experiments clearly demonstrated the interaction of t-PA with annexin II. However, only a fraction of the two proteins interact with each other, likely corresponding to the fractions localised at the surface compartment of cells. The partial co-immunolocalisation of both proteins in vivo supports this notion, showing both t-PA and annexin II overexpression in tumours and colocalisation at particular membrane sites.

The LCKLSL peptide inhibits t-PA binding to cells and invasion through Matrigel at comparable levels, confirming that the active t-PA bound to surface annexin II is needed for invasion. SK-PC-1 and PANC-1 cells also express high levels of u-PA and u-PAR that contribute to the invasion process.^13,15,47 Thus inhibition of invasion obtained by the disruption of annexin II/t-PA interaction in the experiments shown here may account for most of the t-PA dependent invasive capacity.

Our results are the first to show the specific interaction of t-PA with annexin II on pancreatic cancer cell surfaces and its relevance for the invasive process. Understanding the precise mode of interaction of t-PA with pancreatic cancer cells and the molecular mechanisms used for invasion will help in the understanding of the contribution of this protease to cancer progression and in the development of more effective therapeutic strategies in the future.