S100P protein in human breast cancer cells is associated with reduced patient survival and, in a model system of metastasis, it confers a metastatic phenotype upon benign mammary tumour cells. S100P protein possesses a C-terminal lysine residue. Using a multiwell in vitro assay, S100P is now shown for the first time to exhibit a strong, C-terminal lysine-dependent activation of tissue plasminogen activator (tPA), but not of urokinase-catalysed plasminogen activation. The presence of 10 μM calcium ions stimulates tPA activation of plasminogen 2-fold in an S100P-dependent manner. S100P physically interacts with both plasminogen and tPA in vitro, but not with urokinase. Cells constitutively expressing S100P exhibit detectable S100P protein on the cell surface, and S100P-containing cells show enhanced activation of plasminogen compared with S100P-negative control cells. S100P shows C-terminal lysine-dependent enhancement of cell invasion. An S100P antibody, when added to the culture medium, reduced the rate of invasion of wild-type S100P-expressing cells, but not of cells expressing mutant S100P proteins lacking the C-terminal lysine, suggesting that S100P functions outside the cell. The protease inhibitors, aprotinin or α-2-antiplasmin, reduced the invasion of S100P-expressing cells, but not of S100P-negative control cells, nor cells expressing S100P protein lacking the C-terminal lysine. It is proposed that activation of tPA via the C-terminal lysine of S100P contributes to the enhancement of cell invasion by S100P and thus potentially to its metastasis-promoting activity.
Certain S100 proteins, S100A4 and S100P, when present individually in patients' cancers are associated with reduced patient survival [1–8], an association that is a likely consequence of the ability of these proteins to confer upon cells a metastatic phenotype. For S100A4, its elevated expression in two independent transgenic mouse models induces metastasis of oncogene-induced, non-metastatic tumours in the mammary gland [9,10]. In an in vivo model system of breast cancer metastasis, expression of either S100A4  or S100P  in a non-metastatic benign rat mammary tumour cell line results in the induction of a metastatic phenotype. It is now evident that intracellular S100A4 can enhance multiple properties contributing to the induction of a metastatic phenotype, including the activation of cell migration [12,13], arising from a specific interaction with non-muscle myosin isoform A , alteration in apoptotic capability, by interaction with the tumour suppressor protein, p53 , and enhanced invasion through the stimulation of expression of metalloproteinases . Furthermore, extracellular S100A4, as high-molecular mass associations, has been reported to act on endothelial cells to induce angiogenesis  and to activate plasminogen activators in vitro by a mechanism similar to that of S100A10 .
S100A10, a subunit of the annexin 2 tetramer, promotes plasminogen activation when its C-terminal lysines are displayed at the cell surface , particularly on endothelial cells . The displayed C-terminal lysine residues interact with lysine-binding Kringle domains in plasminogen and tissue plasminogen activator (tPA) proteins in the extracellular milieu . In macrophages, this interaction leads to enhanced invasion .
The molecular mechanisms whereby S100P induces a metastatic phenotype are much less well understood than for S100A4. In common with other S100 proteins, the S100P monomer contains a high- and a low-affinity calcium-binding site . The low-affinity calcium-binding site can bind Mg2+ and the high-affinity calcium-binding site also binds Zn2+ , although interaction is unlikely at recently determined free intracellular (cytoplasmic) and extracellular concentrations of zinc [25,26] and magnesium  ions. Using analytical ultracentrifugation experiments, it has been shown that the oligomeric state of the S100P molecule varies according to the calcium concentration .
S100P has been reported to interact with the cell membrane actin-cross-linking protein, Ezrin , and the cytoskeletal protein IQGAP1 ; however, any consequences of such interaction in the context of cancer metastasis or even cell invasion are so far unknown. S100P interacts specifically with, and causes dissociation of, non-muscle myosin II isoforms A and C, with a consequential reduction in vinculin-containing focal adhesions, an increase in cell motility and reduced cell adhesion . The likelihood is that, as with S100A4, S100P exerts its metastasis-inducing properties via many separate pathways, including cell invasion. However, apart from an association of S100P with the RAGE receptor and epithelial–mesenchymal transition , none of the present protein targets of S100P have been reported to have an S100P-associated role in the cellular invasion process. Since S100P contains a C-terminal lysine, it is possible that one such pathway is the activation of plasminogen. In the present paper, we show for the first time that S100P can activate plasminogen activators and that this activity contributes to the metastasis-associated biological property of cell invasion in S100P-expressing cells with metastatic potential .
Materials and methods
Cell lines and pools
The benign rat mammary tumour cell line, Rama 37 , was used in the present experiments because expression of S100P in these cells has been shown to confer a metastatic phenotype . Rama 37 cells were transfected, as described previously , with the empty PBK-CMV plasmid expression vector or PBK-CMV vector containing a cDNA-encoding wild-type human S100P protein, or a mutant S100P protein in which the C-terminal lysine has been replaced with alanine, designated K95A, or a mutant S100P protein lacking the C-terminal lysine, designated ΔK95. Cell clones and pools were isolated as described previously  and cultured in Dulbecco's modified Eagle's medium (DMEM) with 10% (v/v) foetal calf serum, as described previously , with 1 mg/ml G418 included for all transfected cells. These cloned and pooled cell lines exhibit similar levels of S100P as determined by Western blotting (Supplementary Material S1). In some experiments, an existing S100P-inducible system, Rama 37-T25, was utilised , with cells cultured as above and wild-type human S100P protein production being induced with 1 μg/ml doxycycline for 24 h.
Transwell cell invasion assays were carried out using 8 µm pore-size BioCoat Matrigel Invasion plates (BD Biosciences, MA). The chambers were rehydrated for at least 2 h by adding warm, serum-free DMEM to the interior of the upper chambers and to the lower wells. Following rehydration, the wells were filled with 750 µl of DMEM containing 5% (v/v) foetal calf serum. The upper chambers were filled with 500 µl of the cell suspension containing 25 000 cells in DMEM with 1% (v/v) serum and were inserted into the wells. The plates were incubated for 22 h in a humidified tissue culture incubator at 37°C with a 10% (v/v) CO2 atmosphere. The non-invading cells on the upper surface were removed by gentle, but firm, scrubbing using a cotton-tipped swab. The invading cells on the lower surface of the membrane were stained by Diff-Quik solution, the membrane was removed using a sharp scalpel blade and the cells were counted using a light microscope .
Detection of S100P on the cell surface by enzyme-linked immunosorbent assay
For analysis of extracellular- and cell surface-bound S100P, quantification of the protein was carried out using an adapted enzyme-linked immunosorbent assay (ELISA) directly on living cells, as previously described . Briefly, 10 000 cells were seeded in 96-well plates and grown for 48 h. To assess the presence of S100P or eukaryotic elongation factor 1A (eEF1A) as an intracellular marker in both intracellular and extracellular fractions, cells were either incubated in the presence of antibody for 1 h prior to fixing, or fixed prior to antibody addition, or fixed, permeabilised and incubated with antibody directed against either S100P (R&D Systems) or eEF1A (Merck Millipore). Fixation was carried out using 3.7% (w/v) paraformaldehyde in phosphate-buffered saline (PBS) for 20 min at room temperature, while permeabilisation was achieved by incubating cells in 0.5% (v/v) Triton-X100 in PBS for 20 min at room temperature. Following the incubation with the primary antibody, wells were washed with blocking buffer [5% (w/v) dried skimmed milk in PBS, pH 7.4] prior to incubation with an appropriate horseradish-peroxidase-labelled secondary antibody for 2 h at room temperature. After subsequent washing, the reaction was developed using the o-phenylenediamine dihydrochloride substrate (Sigma, U.K.) and absorbance was read at 450 nm using a multiwell plate reader.
Production of recombinant wild-type and mutant S100P proteins
His-tagged wild-type and mutant recombinant (r)S100P proteins were produced in BL21 (DE3) Escherichia coli cells (Thermo Fisher), as described previously . rS100P-positive fractions eluted from a HisTrap column (GE Healthcare) were buffer-exchanged into 50 mM Tris–HCl (pH 8.0), 250 mM NaCl and the tag cleaved by the addition of thrombin–agarose resin (Sigma) for 20 h at room temperature. Residual uncleaved protein and the cleaved His tag were removed by passing the supernatant from the cleavage reaction through a HisTrap column, and the flow-through, containing rS100P protein without the His tag, was then passed down a Superdex® 75 gel filtration column (GE Healthcare) into 50 mM HEPES (pH 7.4) and 250 mM NaCl (Supplementary Material S2). Size-exclusion chromatography with multi-angle light scattering (SEC-MALS) was carried out as described previously .
Plasminogen activation assays in vitro
Assays were performed as described by Semov et al. . rS100P protein, 100 nM Glu-plasminogen (Millipore) and 125 μM chromozyme PL (Roche) were pre-incubated for 60 min at room temperature in a reaction buffer consisting of 50 mM Tris–HCl (pH 7.4) and 100 mM NaCl, with or without CaCl2. The reaction was initiated by the addition of tPA (Millipore) or urokinase-type plasminogen activator (uPA; Millipore) as indicated in the figure legends, and activation of plasminogen was measured by an increase in optical density at 405 nm for 2 h, at 2-min intervals.
For measurement of the activation of plasminogen activators by cultured cells, 2000 cells were seeded in each well of a Corning 24-flat-bottomed-well plate and allowed to attach for 24 h. Attached cells were washed twice in PBS and incubated in 100 nM Glu-plasminogen and 125 μM chromozyme PL (Roche) in PBS. Plasminogen activation was initiated by the addition of tPA in a final volume of 500 μl for 1 h. The activation of plasminogen by tPA was measured by the increase in optical density at 405 nm of samples withdrawn at 10-min intervals.
Interaction between S100P, plasminogen and plasminogen activators
Surface plasmon resonance (SPR) experiments were conducted in a Biacore X-100 instrument (GE Healthcare). Proteins were immobilised on CM5 chips, using amine coupling chemistry, as described previously . Experiments were carried out using a running buffer consisting of 10 mM HEPES (pH 7.4), 150 mM NaCl, added CaCl2 as specified in the Results section and 1 mg/ml carboxymethyl dextran (Sigma), at a flow rate of 30 µl/min at 25°C. Binding kinetics were studied using multi-cycle kinetic analysis, with surfaces regenerated using 1 mM EGTA between cycles. When used, 6-aminocaproic acid (6-ACA) and EGTA were added to the analyte samples only, not to the running buffer. Binding constants were calculated using the BiaEvaluation software 2.0 (GE Healthcare). Association and dissociation of S100P with tPA occurred very rapidly, and thus Kd values for tPA interactions were determined at equilibrium. Glu-plasminogen-binding constants were assessed using standard kinetic analysis, as described previously .
Stats Direct software version 3.0.181 was used for statistical analyses. For single comparisons, Student's t-test was used; for multiple comparisons, analysis of variance (ANOVA) with the Bonferroni post hoc test was used.
Plasminogen activator activation by S100P in vitro
S100A10 and S100A4 have been reported to activate plasminogen activators by virtue of their C-terminal lysine residues [18,38]. Since S100P also possesses a C-terminal lysine, we tested whether rS100P could regulate tPA-catalysed plasminogen activation using a multiwell assay. The production of active plasmin from Glu-plasminogen was monitored over time, by the appearance of a coloured product from a plasmin-specific substrate, as described in Materials and methods, using a buffer containing 5 mM calcium ions. In the absence of tPA, no plasmin activity was detected either without (Figure 1A–C) or with (not shown) 1 μM rS100P. The addition of excess tPA led to the generation of plasmin activity from Glu-plasminogen after a short lag period (Figure 1). Increasing concentration of rS100P in the presence of tPA resulted in reduction in the lag period and increasing rates of plasminogen activation for up to at least 3.5 μM rS100P (Figure 1A). Mutant rS100P proteins, ΔK95 (Figure 1B) or K95A (Figure 1C), were also tested in the multiwell plasminogen activation assay. To aid comparison between the various recombinant proteins, the optical density readings were replotted against time2 and the resulting linear gradients over the first 30 min of incubation were determined. Removal, or replacement, of the C-terminal lysine markedly reduced the ability of rS100P to activate tPA (Figure 1D), demonstrating that S100P and its C-terminal lysine are required for activation of tPA.
S100P enhances the rate of plasminogen activation by tPA in a cell-free multiwell assay.
Effect of calcium ions on plasminogen activation
Although S100P binds calcium  and has been reported to bind magnesium and zinc ions , S100P is unlikely to bind zinc or magnesium ions at their free intracellular or extracellular concentrations [25–27]. Thus, the dependence on only calcium concentration of tPA activation by S100P was determined (Figure 2A). There were statistically significant differences between calcium concentrations, as determined by one-way ANOVA (F8,9 = 15.0, P = 0.0002). While plasminogen was activated in the presence of 1 μM rS100P and excess tPA, the rate of activation was significantly increased by the presence of 10 μM Ca2+ [95% CI (confidence interval) −0.00005 to −0.000018, P = 0.001, Bonferroni correction], but there was no further significant increase at higher calcium ion concentrations up to 10 mM (Figure 2A, 95% CI −0.000025 to 0.000007, P > 0.23, Bonferroni correction). In the absence of rS100P and in the presence of excess (5 mM) calcium ions, the rate of plasminogen activation was not significantly different from that with no calcium (95% CI −0.000007 to 0.000025, P = 0.25, Bonferroni correction), suggesting that the increase in plasminogen activator activity in the presence of calcium ions was due to an effect on the rS100P protein by calcium ions. In contrast, 1 μM bovine serum albumin (BSA), in the presence of calcium ions, did not significantly change the initial rate of plasminogen activation by tPA (Figure 2B; 95% CI −0.000011 to 0.000024, P = 0.32, Bonferroni correction), under conditions when there was a significant activation by 1 μM rS100P (95% CI −0.000076 to −0.000041, P = 0.002, Bonferroni correction), showing that the activation observed with rS100P is not a general property of lysine-containing proteins (Figure 2B). SEC-MALS experiments indicated that >96% of wild-type and mutant S100P proteins were in the dimeric form at calcium concentrations of both 50 and 500 μM (Supplementary Material S3).
S100P-dependent activation of plasminogen activator in a multiwell assay.
Effect of concentration of rS100P and its mutants on tPA-dependent activation of plasminogen
Increasing concentrations of rS100P up to 100 nM in the presence of 5 mM calcium ions resulted in a significantly increased initial rate of tPA-dependent plasminogen activation in vitro [one-way ANOVA (F6,7 = 24.4, P = 0.0002)]. Based on the initial gradients of plots of optical density vs. time2, this effect started to plateau at concentrations of rS100P above 100 nM (Figure 2C). Significant detectable enhancement of the initial rate of plasminogen activation by excess tPA occurred between 50 and 100 nM (600–1200 ng/ml) rS100P (50 nM, 95% CI −0.000031 to 0.000003, P = 0.09; 100 nM, 95% CI −0.000055 to −0.000021, P = 0.0012, Bonferroni correction; Figure 2C).
There were significant differences between the ability of wild-type and mutant S100P proteins to activate plasminogen activators [one-way ANOVA (F4,10 = 194.6, P < 0.0001)]. Initial rates of plasminogen activation with 1 μM K95A-mutant or ΔK95-mutant rS100P were significantly lower than those with wild-type rS100P (95% CI, K95A-mutant, 0.000031 to 0.000039 and ΔK95-mutant, 0.000028 to 0.000036, both P < 0.0001, Bonferroni correction; Figure 2D), but both still exhibited residual activation of tPA that was significantly higher than the no-S100P control (95% CI −0.00001 to −0.000003, P = 0.004 and 95% CI −0.000013 to −0.000005, P = 0.0003, respectively, Bonferroni correction), whereas, in the same experiment, 1 μM BSA was not significantly different from the no-protein control (95% CI −0.000004 to 0.000004, P = 0.98, Bonferroni correction; Figure 2D). The results show that the C-terminal lysine is largely required for the ability of S100P to enhance the activity of tPA. This enhancement might be due to interaction of the C-terminal lysines with lysine-binding Kringle domains of tPA and/or plasminogen.
Interaction between S100P, plasminogen and plasminogen activators in vitro
The binding of S100P to human plasminogen and tPA was measured using SPR as described in Materials and methods. Wild-type rS100P bound to immobilised human Glu-plasminogen (Figure 3A) with a Kd of 0.12 µM (Supplementary Material S4), a value that shows a greater affinity between Glu-plasminogen and S100P than between Glu-plasminogen and free lysine residues (43 μM; ). Wild-type rS100P bound to tPA with a Kd of 1.9 µM (Figure 3B and Supplementary Material S4). In the presence of the lysine analogue, 6-ACA, binding of wild-type rS100P to either Glu-plasminogen (Figure 3A) or to tPA (Figure 3B) was markedly reduced, displaying Kd values of 84.0 and 74.6 μM, respectively, confirming the involvement of lysine residues in the interaction of S100P with both plasminogen and tPA.
S100P binding to plasminogen, tPA and uPA.
Calcium chelation with EGTA resulted in only an ∼50% decrease in binding of wild-type rS100P to Glu-plasminogen (Figure 3A), but completely abolished binding to tPA (Figure 3B), suggesting that activation of S100P by calcium is essential for interaction with tPA, but not for interaction with Glu-plasminogen. In contrast, the loss of the C-terminal lysine virtually abolished the interaction of S100P with Glu-plasminogen (Figure 3A), but only partially reduced the extent of interaction of S100P with tPA (Figure 3B).
Although 1 μM wild-type rS100P enhanced tPA-dependent plasminogen activation in the multiwell assay, the same concentration of S100P had no effect on urokinase (uPA)-dependent plasminogen activation in the multiwell plasminogen activation assays, in the range 0–50 nM uPA. While the initial rate of plasminogen activation increased with increasing uPA, there was no significant change in activity by 1 μM wild-type rS100P at uPA concentrations of 1, 5 and 50 nM (P = 0.216, 0.089 and 0.273, respectively), with S100P just significantly reducing activity at 10 and 25 nM (both, P = 0.05, Student's t-test; Figure 3C). In SPR experiments, wild-type rS100P bound very poorly to uPA (Figure 3C) and the bound protein dissociated only slowly from the uPA surface, in the time-frame of the experiment. This poor uPA–rS100P interaction likely accounts for the lack of rS100P-enhanced activation of plasminogen by uPA, while the very slow dissociation of rS100P from uPA might account for the reduction in uPA-induced plasminogen activation in the presence of rS100P, as the build-up of ‘inactive’ rS100P–uPA complexes could inhibit further interaction with plasminogen. Thus, rS100P does not activate uPA, but acts specifically on tPA.
Cells expressing non-mutant S100P exhibit enhanced tPA-dependent plasminogen activation
Following the demonstration of plasminogen activation by S100P protein in vitro, experiments were conducted to find out whether cells expressing S100P exhibited enhanced plasminogen activation, compared with non-expressing cells. Plasminogen activation assays were carried out using cells growing in culture wells, as described in Materials and methods. PBS was found to be a satisfactory replacement for the Tris buffer system used in the plasminogen activation assays in vitro (Supplementary Material S5).
To compare the results of multiple experiments, the initial rates of colour development for the various cell lines in each experiment were expressed as a fraction of the result with cells expressing wild-type S100P. Figure 4 shows the results of six separate experiments. In each case, cells not expressing S100P, or cells expressing the ΔK95- or K95A-mutant S100P proteins, exhibited reduced ability to activate Glu-plasminogen in the presence of added tPA, compared with cells expressing wild-type S100P (Figure 4A, results summarised in Figure 4B). As further confirmation of the link between the production of S100P and the enhancement of tPA-catalysed plasminogen activation in cultured cells, an existing S100P-inducible system was employed . Induction of S100P with doxycycline for 24 h  increased the ability of the cells to activate plasminogen in the presence of tPA by 21.3% (P = 0.029, Student's t-test; Figure 4C).
Activation of plasminogen by S100P.
Detection of S100P on the outside of S100P-expressing cells
To understand the ability of S100P-expressing cells to activate extracellularly-added tPA, a quantitative ELISA was used to detect S100P directly on the surface of living cells by adding an S100P polyclonal antibody directly to the cell culture medium or after fixation or permeabilisation [one-way ANOVA (F5,12 = 10.9, P = 0.0004)]. S100P antibody binding was detected at a significant (95% CI −0.15 to −0.04, P = 0.0027, Bonferroni correction), 5.4-fold greater level in living unfixed cells expressing S100P than from the cell line transfected with the empty expression vector (Figure 5A, antibody in medium). A significantly higher level of S100P was detected also when the cells were fixed with formaldehyde (95% CI −0.11 to −0.00062, P = 0.05, Bonferroni correction), which was not significantly different from the level found when antibody was added directly to the medium (95% CI −0.088 to 0.022, P = 0.22, Bonferroni correction). Upon permeabilisation, a significantly higher optical density was evident with S100P-expressing cells compared with antibody added to the medium (95% CI 0.0017 to 0.112, P = 0.04, Bonferroni correction), possibly due to some of the intracellular S100P becoming available to the antibody. To check that the antibody added to the medium was not reacting with intracellular S100P as a result of cell leakiness, an antibody to the soluble intracellular protein, eEF1A , was tested [eEF1A, (F5,12 = 211.6), P = <0.0001]. In the ELISA assay, there was virtually no detectable binding of the eEF1A antibody when it was added to the medium in either S100P-expressing or control cells (Figure 5B). However, after permeabilisation, there was a significant 134-fold increase in detectable eEF1A antibody binding in both S100P-negative control cells (95% CI −0.94 to −0.74, P < 0.0001) and S100P-expressing cells (95% CI −1.21 to −1.02, P < 0.0001; Figure 5B), suggesting that under the conditions of the ELISA assay, the unfixed cells are intact and not leaking intracellular proteins, thus part of the detected S100P is displayed to the outside of the cells.
ELISA quantitation of S100P and eEF1A antibody binding to the surface of S100P-negative and S100P-positive living cells.
Cell invasion by cells expressing S100P or its mutants
Cells expressing wild-type S100P exhibited a significant 9.4-fold (Figure 6A; 95% CI 330.5 to 357.5, P < 0.0001, Bonferroni correction) or 8.7-fold (Figure 6B; 95% CI −162.9 to −99.7, P < 0.0001, Bonferroni correction) increase in invasion through Matrigel in independent transwell invasion assays, when compared with S100P-negative, vector-transfected control cells. Such an increase was evident in both a clone and a pool of S100P-expressing transfected cells when compared with vector-transfected or parental, untransfected, Rama 37 cells (Supplementary Material S6).
The effect of extracellular addition of ACA, an S100P antibody, aprotinin or α-2-antiplasmin on the rate of invasion through Matrigel by cells expressing wild-type and mutant S100P proteins.
In contrast, cells expressing the K95A or Δ95K C-terminal mutants of S100P protein exhibited levels of invasion that were highly significantly lower than those with wild-type S100P (K95A mutant: Figure 6A, 95% CI 316.9 to 343.8, P < 0.0001; Figure 6B, 95% CI 83.1 to 146.3, P < 0.0001; Δ95K mutant: Figure 6A, 95% CI 347.9 to 374.8, P < 0.0001; Figure 6B, 95% CI 98.742469–161.924197, P < 0.0001) and not significantly different or significantly less than S100P-negative, vector-transfected control cells in two independent experiments (K95A mutant: Figure 6A, 95% CI 0.18 to 27.15, P = 0.05; Figure 6B, 95% CI −48.3 to 14.9, P = 0.29 and Δ95K mutant: Figure 6A, 95% CI −30.8 to −3.9, P = 0.014; Figure 6B, 95% CI −32.6 to 30.6, P = 0.95). Thus, altering or deleting the C-terminal lysine of S100P, which we have shown to be important for interacting with tPA and plasminogen, reduces its ability to enhance invasion of the Rama 37 cells.
The effect of 6-ACA on S100P-dependent cell invasion
To confirm the possible role of S100P lysine residues in the observed changes in cell invasion, 10 mM 6-ACA had no significant effect on S100P-negative vector control cells (95% CI −20.1 to 6.8, P = 0.32), but significantly reduced by 32% the number of S100P-positive cells invading through Matrigel (Figure 6A, 95% CI 110.5 to 137.5, P < 0.0001). However, the invasion was still significantly higher than for S100P-negative control cells (95% CI 199.9 to 226.8, P < 0.0001). Invasion was also still significantly higher than with cells expressing the C-terminal mutants of S100P (K95A, 95% CI 192.9 to 219.8, P < 0.0001; ΔK95, 95% CI 223.9 to 250.8, P < 0.0001). This result indicates that lysine residues contribute significantly to S100P-induced invasion, but that the dramatic reductions in invasion seen with the C-terminal lysine mutant proteins are not entirely due to the absence of the C-terminal lysine.
The effect of S100P antibody added to the culture medium on S100P-dependent invasion
In invasion assays, the addition of the S100P antibody to the medium surrounding wild-type S100P-expressing cells (Figure 6B) reduced their invasion by a highly significant 57% from that of untreated cells (95% CI 52.4 to 115.6, P = <0.0001), whereas non-immune goat serum added to the S100P-expressing cells had no significant effect (Figure 6C; 95% CI −46.8 to 99.5, P = 0.43). Cells expressing K95A- or ΔK95-mutant S100P proteins, or cells expressing no S100P, showed no significant reduction in invasion upon treatment with the antibody (Figure 6B; K95A, 95% CI −47.3 to 15.9, P = 0.32; ΔK95, 95% CI −36.9 to 26.2, P = 0.73; vector control, 95% CI −47.3–15.9, P = 0.32). As it is unlikely that the S100P antibody can pass into the cells, these results suggest that the stimulatory effect of S100P on invasion depends on wild-type S100P being accessible from outside the cell and this is not evident with cells expressing S100P protein lacking the C-terminal lysine, S100P-negative cells or with non-immune serum.
The effect of protease inhibitors on the rate of S100P-dependent invasion
To find out whether the generation of plasmin was associated with the lysine-associated invasive behaviour of the cells, the effects of protease inhibitors on cell invasion were tested. Aprotinin, an inhibitor of the serine proteases, plasmin, trypsin, chymotrypsin, kallikrein and activated protein C , and α-2-antiplasmin, which targets predominantly neutrophil elastase and plasmin , were used.
There were significant reductions in invasive activity of 55% (Figure 6A; 95% CI 199.5 to 226.5, P < 0.0001) and 70% (Figure 6B; 95% CI 71.7 to 134.9, P < 0.0001) when S100P-expressing cells were treated with 25 μg/ml aprotinin or 10 μg/ml α-2-antiplasmin, respectively. There was no significant reduction in the invasion of control S100P-negative cells by either aprotinin (Figure 6A; 95% CI −5.5 to 21.5, P = 0.24) or α-2-antiplasmin (Figure 6B; 95% CI −31.9 to 31.3, P = 0.98). Whereas α-2-antiplasmin did not significantly reduce the invasive potential of cells expressing the mutant S100P proteins (Figure 6B; K95A, 95% CI −39.3 to 23.9, P = 0.62; ΔK95, 95% CI −37.3 to 25.9, P = 0.72), aprotinin reduced the invasive potential of cells expressing the K95A-mutant protein (Figure 6A; K95A, 95% CI 16.9 to 43.8, P < 0.0001), but not significantly for the ΔK95 mutant (95% CI −17.5 to 9.5, P = 0.55). These results demonstrate that the stimulatory effect of S100P on invasion involves the plasmin/plasminogen pathways, as these specific protease inhibitors significantly disrupted cell invasion.
It has been shown here for the first time that soluble rS100P protein can activate tPA in vitro. The 7.6-fold activation over that with tPA alone is comparable with that of ∼9-fold for S100A4 , but less than the 46-fold activation reported for S100A10 . Deleting the C-terminal lysine of S100P (Δ95K, exposing a C-terminal leucine) or changing it to alanine (K95A), with no change in the dimeric state of the rS100P, reduced the rates of activation of tPA to only 16 and 22%, respectively, of the non-mutant S100P protein, similar to the 15% of activity remaining upon deletion of the two C-terminal lysines of S100A10 , reflecting the involvement of non-C-terminal lysine side chains .
Direct interaction in vitro between non-mutant S100P and Glu-plasminogen or tPA, individually, has been demonstrated, with Kd values (0.12 and 1.9 μM, respectively) similar to those for S100A10 binding to Glu-plasminogen [0.66 µM, calculated in the present experiments (not shown) and 1.81 µM reported previously ] and to tPA of 0.45 µM . However, for S100P, there were differences between the interactions of S100P with Glu-plasminogen and tPA. Whereas the interaction of S100P with tPA required the presence of calcium ions, S100P still bound to Glu-plasminogen in the absence of calcium. In contrast, the C-terminal lysine mutants of S100P did not bind to Glu-plasminogen, but did bind to tPA, a result which contrasts with that obtained with S100A10 . Thus, the observed dependence of rS100P-mediated plasminogen activation on both the presence of calcium ions and the presence of the C-terminal lysine residue is explained by the former requirement for binding of S100P to tPA and the latter requirement for binding to Glu-plasminogen.
In the present multiwell experiments, significant activation of plasminogen activator was detectable with only 50–100 nM rS100P (<1 μg/ml). S100P has been reported to be present in the extracellular environment [44,45], but at presently unknown concentration. However, the observation that cells constitutively expressing either of the mutants exhibited reduced plasminogen activation rates compared with those expressing non-mutant rS100P protein and in a separate cell system, inducible up-regulation of S100P enhanced plasminogen activation shows for the first time that cell-produced S100P can modulate plasminogen activation.
The main findings in the present paper are that for the first time, the activation of tPA by an S100 protein in a tumour setting is linked to cell invasion. Elevated levels of tPA have been associated with tumour-associated cell activities in pancreatic cancer cells, but these were not linked specifically to an increase in plasmin activity .
In the present experiments, plasmin activity, and thus plasminogen activation, has been directly and specifically linked to S100P-directed cell invasion by the use of inhibitors of plasmin activity, aprotinin or α-2-antiplasmin. The specificity of α-2-antiplasmin for plasmin  has been used previously to show the involvement of plasmin and tPA in chemotactic migration of human T cells . Aprotinin and α-2-antiplasmin, individually, significantly inhibited S100P-promoted invasion, but the inhibitors had no effect on invasion by S100P-negative cells, and α-2-antiplasmin had no inhibitory effect on cells expressing the K95A- or ΔK95-mutant S100P proteins. These results show a direct involvement of plasmin activation in S100P-driven cell invasion. Enhancement of cell invasion by plasmin proteolytic activity is well established in normal tissue  and tumours [49,50].
Given the link between plasmin activation and cell invasion in S100P-expressing cells, the activation of tPA, demonstrated in vitro and in cultured cells, might contribute to the previously reported metastasis-inducing properties of S100P . Neither the S100P-expressing cells, nor the control cells, exhibited detectable plasminogen mRNA by PCR (Supplementary Material S7); however, the average blood concentration of plasminogen (1.33 μM; ) is much greater than both the Glu-plasminogen concentration used in the in vitro assays (100 nM) and the experimentally determined Kd values for the S100P : Glu-plasminogen interaction (0.12 μM); thus, at suitable S100P concentrations in vivo, S100P : plasminogen interaction could take place during the metastatic process. However, in contrast, the average blood concentration of tPA of only 0.143 nM  is much lower than both the concentration of tPA used in the in vitro multiwell assay, in order to detect suitably large optical density changes, and the measured Kd for the interaction between S100P and tPA (1.9 μM). However, the Rama 37 cells do produce PCR-detectable tPA mRNA (Supplementary Material S7), which, if translated into active tPA protein, could locally raise the tPA concentration in the surrounding medium.
S100P has been detected on the cell surface from outside the cell using an antibody (Figure 5), which also reduced the cell invasion (Figure 6). Thus, on the cell surface, S100P might be complexed with other cell components. S100A10, a major tPA activator, exhibits a 340-fold increase in tPA activation when it is associated with its natural cell surface receptor, annexin AII . The cells used in the present experiments exhibit abundant levels of annexin AII mRNA and protein by PCR (not shown) and Western blot (Supplementary Material S8), respectively. However, so far, it has not been possible to show, experimentally in vitro, a physical interaction between S100P and commercially available recombinant annexin AII protein under conditions in which S100A10 (Kd = 3.7 ± 1.1 × 10−8 M) and S100A4 (Kd = 3.45 ± 1.7 × 10−6 M) bind to immobilised annexin AII (Sourav, V. and Barraclough, R, unpublished observation). In other systems, S100P interacts/colocalises in surface protrusions with the cytoskeletal protein ezrin [28,53] and interacts with the adaptor protein, IQGAP1 . Both these proteins localise to the cortical cytoskeleton in contact with the cell membrane . Elsewhere, extracellular S100P has been reported to interact with the RAGE receptor . Based on the assumption that binding of S100P to these proteins would allow contact of the S100P C-terminal lysine with tPA, these various proteins might provide a means for displaying S100P on the outside of the cell and thereby enhance its ability to activate plasminogen activator.
analysis of variance
bovine serum albumin
Dulbecco's modified Eagle's medium
eukaryotic elongation factor 1A
enzyme-linked immunosorbent assay
receptor for advanced glycation end products
size-exclusion chromatography with multi-angle light scattering
surface plasmon resonance
tissue plasminogen activator
urokinase-type plasminogen activator
R.B. conceived and co-ordinated the study and wrote the manuscript. M.S. carried out pilot experiments and contributed to the manuscript. T.M.I. produced and characterised the S100P- and S100P-mutant expression systems and developed the cell lines expressing wild-type and mutant S100P proteins. S.R.G. carried out all the S100P ELISA assays, contributed text and commented scientifically on the manuscript. M.A.-M. planned and carried out the invasion assays. All other practical aspects of the project were carried out by C.J.C. who also contributed text for and commented on the manuscript. P.S.R. made available the Rama 37 cell line, contributed scientifically to, and commented on, the manuscript.
The authors thank The Cancer and Polio Research Fund and the James Tudor Foundation for supporting this work through research grants to R.B., the Aston University Biomedical Sciences Research Funds for consumables to S.R.G. and the Ministry of Higher Education and Scientific Research, Kufa University, Iraq for a studentship to M.A.-M.
We thank Dr Min Du (Institute of Integrative Biology) and Dr Guozheng Wang (Institute of Global Health) for providing the S100P-inducible Rama 37 cell line and Dr Caroline Dart (Institute of Integrative Biology) for providing control rat liver cDNA.
The Authors declare that there are no competing interests associated with the manuscript.
These authors contributed equally to this work.