Differential Binding of Plasminogen, Plasmin, and Angiostatin4.5 to Cell Surface β-Actin: Implications for Cancer-Mediated Angiogenesis Free

Plasmin is a two-chain serine proteinase formed by cleavage of the zymogen plasminogen by urokinase-type plasminogen activator (uPA) or tissue plasminogen activator (tPA) through hydrolysis of the Arg⁵⁶¹-Val⁵⁶² peptide bond (1). In addition to being the primary fibrinolytic enzyme, plasmin cleaves matrix proteins, such as laminin, and activates matrix metalloproteinases (2, 3), which may promote angiogenesis and invasion by tumors. Angiostatin, first described by O'Reilly et al. in 1994, is an internal fragment of plasminogen (4, 5). Angiostatin is a potent inhibitor of vascular endothelial cell proliferation, migration, and tube formation and induces endothelial cell apoptosis (6, 7).

Several angiostatin isoforms have been reported, differing in kringle content. Angiostatins consisting of plasminogen kringles 1 to 3, kringles 1 to 4, or angiostatin4.5 (AS4.5), which includes kringles 1 to 4 plus 85% of kringle 5 (Lys⁷⁸-Arg⁵²⁹), have been generated in vitro by a variety of proteinases. However, it is still not well understood which angiostatin isoforms are present naturally in the human body (5, 8–13). We have shown previously that AS4.5 is a naturally occurring human form (14). AS4.5 is generated from plasminogen in a two-step process (12, 14). Plasminogen activators cleave plasminogen to form plasmin. Plasmin can then be converted to AS4.5 via autoproteolysis by cleavage within kringle 5 (12, 14). In vitro, plasmin autoproteolysis to AS4.5 may be mediated by an alkaline pH, phosphoglycerate kinase, or a free sulfhydryl donor (12, 15, 16).

We showed recently that PC-3 (human prostate cancer), HT1080 (human fibrosarcoma), and MDA-MB231 (human breast cancer) cells express globular β-actin on the surface of the cells (17, 18). We further showed that β-actin can mediate plasmin autoproteolysis to generate AS4.5 on the cancer cell surfaces in the absence of a free sulfhydryl donor (18). The mechanism of actin trafficking and binding to the cell surface is not well understood. In this study, we show colocalization of plasminogen to surface β-actin on PC-3 cells. Further, we characterized the binding of plasminogen, plasmin, and AS4.5 to actin.

Reagents. Plasminogen and AS4.5 were prepared as described previously (13). Pure plasmin was from Enzyme Research Laboratories (South Bend, IN). Recombinant kringle 5 was a gift from Donald Davidson of Abbott Laboratories (Abbott Park, IL). Synthetic peptides were made by NeoMPS (San Diego, CA). Purified nonmuscle actin was from Cytoskeleton (Denver, CO) and contains 80% β-actin. Anti-kringles 1 to 3 monoclonal antibodies (VAP) was a kind gift from V. Ploplis and F. Castellino of Notre Dame University (Notre Dame, IN). Goat anti-mouse IgG conjugated with alkaline phosphatase was from Kirkegaard & Perry Laboratories, Inc. (Gaithersburg, MD). ρ-Nitrophenyl phosphate and N-acetyl-l-cysteine (NAC) were from Sigma-Aldrich (St. Louis, MO). EZ-Link Sulfo-NHS-LC-Biotin was from Pierce Biotechnology, Inc. (Rockford, IL).

Immunofluorescent staining of PC-3 cells. PC-3 cells were plated on untreated glass coverslips for 2 days. Cells were washed twice with PBS and then treated with 40 μmol/L pefabloc (Roche, Indianapolis, IN) for 1 hour at 37°C followed by 2 mmol/L ε-aminocaproic acid (εACA; Sigma-Aldrich) for 1 hour at 37°C and a final wash of PBS. Cells were then treated with 1 μmol/L human Glu-plasminogen or AS4.5 for 1 hour at 37°C. Cells were washed twice with PBS followed by fixation in freshly prepared 2% formaldehyde. Cells were rinsed thrice with PBS. Coverslips were blocked with normal goat serum (NGS) for 30 minutes at 37°C followed by 30 minutes of incubation at 37°C with 1:100 dilution (in NGS/PBS) of primary antibody, a polyclonal rabbit anti-actin (Sigma-Aldrich). Coverslips were next rinsed with 0.05% Tween 20/PBS, twice with PBS, followed by 30 minutes of incubation at 37°C with a 1:200 dilution of secondary antibody, Alexa Fluor goat anti-rabbit 568 (Molecular Probes, Carlsbad, CA). The procedure was repeated with a primary mouse monoclonal antibody to plasminogen kringles (GMA-086, Green Mountain Antibodies, Burlington, VT) or VAP (a kind gift from V.A. Ploplis and F. Castellino) and a secondary antibody, Alexa Fluor goat anti-mouse 488 (Molecular Probes). Coverslips mounted on glass slides with gelvatol mounting medium containing 1.5 mg/mL DABCO (Sigma-Aldrich). Images of fixed, stained preparations were taken with a Zeiss LSM 510 microscope (Carl Zeiss, Thornwood, NY).

Solid-phase ELISA for plasminogen binding to immobilized actin. All steps were conducted at room temperature. Microtiter plates (Nalgene Nunc International, Rochester, NY) were coated with 100 μL/well of 80 μg/mL nonmuscle actin or bovine serum albumin (BSA, Sigma-Aldrich) in PBS overnight and washed thrice with PBS with 1% BSA (PBS/BSA). Wells were blocked with PBS/BSA for 1 hour. Plasminogen (100 μL) at 10 to 500 nmol/L in PBS/BSA was added to duplicate wells coated with actin or BSA as control. After 1 hour of incubation, the wells were washed thrice with PBS/BSA, and VAP antibody (0.01 mg/mL in PBS/BSA) was added to the wells. After 1 hour of incubation, the wells were washed as above. Goat anti-mouse IgG conjugated with alkaline phosphatase (0.01 mg/mL diluted in PBS/BSA) was added to the wells, incubated as above, and wells were washed as above. Phosphatase substrate solution (1 mg/mL ρ-nitrophenyl phosphate in 0.05 mol/L NaHCO₃ with 1 mmol/L MgCl₂) was added to the wells. A₄₀₅ was read after 20 minutes of incubation. To test the influence of other agents on binding of plasminogen, putative binding inhibitors (εACA, synthetic actin-derived peptides, purified recombinant kringle 5, or AS4.5) were coincubated with 50 nmol/L plasminogen.

Measurement of AS4.5 generation. AS4.5 generation was measured by ELISA as described previously (18). For cell studies, PC-3 cells were cultured and fixed with glutaraldehyde as described previously (12, 18). Fixed PC-3 cells were incubated with 440 nmol/L (40 μg/mL) plasminogen alone or with synthetic peptide C, C1, C2, or εACA. In a cell-free system, plasminogen (440 nmol/L) was incubated alone or with urokinase (uPA), NAC, nonmuscle actin, and/or synthetic peptide C at 37°C as described previously (12, 18).

Binding of biotinylated plasminogen, plasmin, AS4.5, or kringle 5 to actin. Plasmin was inactivated with 10 mmol/L pefabloc for 5 minutes. Plasminogen, plasmin, AS4.5, and recombinant kringle 5 were buffer exchanged thrice with PBS. EZ-Link Sulfo-NHS-LC-Biotin was added to plasminogen, plasmin, AS4.5, or kringle 5 at a 10:1 molar ratio. After 30 minutes of incubation at room temperature, plasminogen, plasmin, AS4.5, and kringle 5 were buffer exchanged as above to remove unreacted biotin reagent.

Microtiter plate wells were coated with 100 μL of 80 μg/mL nonmuscle actin in PBS overnight at room temperature and then washed thrice with PBS/BSA. Wells were blocked with PBS/BSA for 1 hour at room temperature. Biotinylated plasminogen, plasmin, AS4.5, or kringle 5 was added to duplicate wells. After 1 hour of incubation at room temperature, wells were washed as above. Phosphatase-labeled streptavidin (100 μL of 0.001 mg/mL in PBS/BSA) was added to the wells. After 1 hour of incubation at room temperature, the wells were washed as above, and phosphatase substrate solution (200 μL/well) was added. A₄₀₅ was read after 20 minutes of incubation.

Binding of plasminogen and AS4.5 to PC-3 cells in culture. Using immunofluorescent staining, we compared plasminogen and AS4.5 binding to fixed PC-3 cells (Fig. 1). The cells had been pretreated with pefabloc to neutralize any serine proteinase activity on the cells and with εACA to remove any plasminogen-derived protein that may be present in the serum. Surface β-actin was observed on nonpermeabilized PC-3 cells as reported previously (18). Following treatment with plasminogen, there was an avid binding of GMA-086 to the nonpermeabilized cells, which colocalized with β-actin. In contrast, control cells and cells treated with AS4.5 did not bind two different antibodies to plasminogen kringles, GMA-086 (or VAP; data not shown), indicating absence of binding of AS4.5. These data confirmed that plasminogen does bind to the β-actin on the PC-3 cell surface and that AS4.5 does not bind.

Plasminogen-β-actin binding in a solid-phase ELISA. To characterize the plasminogen-actin interaction, we used a solid-phase ELISA to measure the binding efficiency. When microtiter plates were coated with actin, plasminogen binding was dose dependent up to ∼500 nmol/L where it reached a maximum (Fig. 2A). Using Scatchard analysis, the K_d of plasminogen binding to actin was 140 nmol/L. When plates were coated with BSA, plasminogen did not bind, indicating specific binding to actin. Because plasminogen contains several kringle domains, which are known to interact with lysines, we assessed the ability of εACA (1 mmol/L), a lysine analogue known to inhibit kringle binding, to inhibit plasminogen binding to actin. εACA inhibited plasminogen binding to actin by 84% to 91% (Fig. 2A). These data indicate that plasminogen binding to actin is kringle dependent.

Influences of actin-derived peptides on plasminogen-β-actin binding. Previous studies by Lu et al. suggested that Lys⁶¹ and Lys¹¹³, which are conserved in all actin isoforms, are on the surface of globular actin (19). Therefore, we tested three synthetic, actin-derived peptides to determine if these or other surface lysine residues contributed to plasminogen binding to actin (Fig. 3). Actin peptide A (amino acids 241-255) was used as a presumed negative control because, although a surface epitope, it did not contain a lysine residue. Peptide B consisted of amino acids 106 to 120, including Lys¹¹³ and Lys¹¹⁸. Peptide C consisted of amino acids 55 to 69, including Lys⁶¹ and Lys⁶⁸. These peptides, as well as εACA, were tested for their ability to competitively inhibit plasminogen binding to β-actin. εACA, at 0.2 and 1 mmol/L, inhibited plasminogen binding to β-actin by ∼74% (P < 0.02) and 91% (P < 0.01), respectively (Table 1). Peptide C also inhibited plasminogen binding to β-actin; 37% inhibition (P < 0.05) at 0.2 mmol/L and 79% inhibition at 1 mmol/L (P < 0.01), respectively (Table 1). Peptide A and B did not show significant inhibition (Table 1). These data indicate that Lys⁶¹ and Lys⁶⁸ are important in plasminogen binding.

To further confirm the role of Lys⁶¹ and/or Lys⁶⁸, two additional synthetic peptides were made, substituting each lysine residue in peptide C to an arginine residue. Peptide C1 (GDEAQSKRGILTLRY) has Lys⁶⁸ replaced with an arginine, and peptide C2 (GDEAQSRRGILTLKY) has Lys⁶¹ replaced with arginine. At 1 mmol/L, peptides C1 and C2 did not show significant inhibition of plasminogen binding to actin (Table 1), indicating that both Lys⁶¹ and Lys⁶⁸ may be necessary for the interaction between plasminogen and actin.

Influences of actin-derived peptides on cell surface AS4.5 generation. Because peptide C and εACA inhibited plasminogen binding to β-actin in a cell-free system, we investigated their ability to inhibit plasminogen conversion to AS4.5 on the surface of PC-3 cells. εACA inhibited AS4.5 generation by 88% (P < 0.001), and peptide C inhibited AS4.5 generation by 49% (P < 0.01), whereas peptides C1 and C2 had a negligible effect on AS4.5 generation (Fig. 4). This further supports that both Lys⁶¹ and Lys⁶⁸ have an important role in binding plasminogen to actin and also suggests that they function in the actin-mediated plasmin autoproteolysis that results in AS4.5 generation.

Effects of actin-derived peptides on cell-free AS4.5 generation. In a cell-free system, β-actin or a small-molecule free sulfhydryl donor can mediate plasmin autoproteolysis to generate AS4.5 (12, 18). To determine if peptide C could substitute for β-actin in this reaction, AS4.5 generation was measured in a cell-free system with or without this peptide. Peptide C did not significantly increase AS4.5 generation over the level of uPA alone (data not shown). These data indicate that, although peptide C was able to compete with β-actin for plasminogen binding, it was not sufficient to replace the full β-actin protein in AS4.5 generation.

Differential binding of plasminogen, plasmin, and AS4.5 to β-actin. Plasminogen and plasmin contain five kringle domains, which bind to lysines. AS4.5 shares the first four of the kringles, but kringle five is cleaved and truncated by the plasmin autoproteolysis. Therefore, we compared the different β-actin binding affinities of plasminogen, plasmin, and AS4.5 in a solid-phase ELISA. To compare the binding affinity, β-actin was coated onto plates, and biotinylated plasminogen, plasmin, or AS4.5 was added. Preliminary experiments indicated that biotinylation did not affect plasminogen, plasmin, or AS4.5 binding to β-actin (data not shown). When bound protein was measured, plasminogen and plasmin both bound to β-actin comparably and in a dose-dependent manner (Fig. 2B). AS4.5, however, did not show any significant binding to β-actin (Fig. 2B). Therefore, intact kringle 5 in plasminogen and plasmin may be necessary for actin binding because AS4.5 contains kringles 1 to 4 but has a truncated kringle 5.

Purified kringle 5-β-actin binding in a solid-phase ELISA. To test the ability of kringle 5 to bind actin, we obtained purified recombinant kringle 5 protein. Kringle 5 bound to β-actin in a dose-dependent manner (Fig. 2C). In contrast, AS4.5 did not bind to β-actin in concentrations as high as 40 μmol/L (Fig. 2B and C). Furthermore, when tested for its ability to competitively inhibit plasminogen binding to β-actin, intact kringle 5 inhibited plasminogen binding by ∼40%, whereas AS4.5 did not inhibit plasminogen-actin binding at the same concentrations. These data further support a critical role for kringle 5, but not kringles 1 to 4, in plasminogen-plasmin binding to β-actin.

We and others have shown that cell surface actin can bind plasminogen (18, 20–22), but the nature and functions of this binding were uncharacterized previously. Further, we have published recently that β-actin on the surface of cells plays a key role in the proteolysis of plasminogen to form AS4.5 by cancer cells (18). Plasminogen contains 5 kringle domains, which are known to interact with lysines, and kringle-containing proteins can be isolated using lysine-binding methods (13). Here, we have shown that εACA inhibited plasminogen binding to actin, indicating that this interaction is lysine-kringle dependent. We also showed that Lys⁶¹ and Lys⁶⁸ in actin are critical for binding plasminogen/plasmin because when either was substituted for an arginine in a synthetic peptide, the ability to inhibit plasminogen binding to actin was lost. Lu and Szilagyi have shown that Lys⁶¹ is a surface lysine residue in actin molecules (19), and because Lys⁶¹ and Lys⁶⁸ are conserved in all actin isoforms, this explains the ability of plasminogen to bind to different actin isoforms (18, 20–22).

Plasminogen and plasmin have intact kringles 1 to 5 and bind actin efficiently (K_d, ∼140 nmol/L), whereas AS4.5 with intact kringles 1 to 4 but a truncated kringle 5 does not bind actin at concentrations as high as 40 μmol/L. We thus conclude that kringle 5 is the key domain of plasminogen and plasmin, which binds to actin. Kringle 5 of plasminogen/plasmin contains 80 amino acids that are structurally similar to kringles 1 to 4 (23). However, our studies showed that kringle 5 alone could bind to actin and, in competitive binding studies, could inhibit plasminogen binding to actin. We propose a model of plasminogen/plasmin interaction with actin where plasminogen/plasmin is bound, but on autoproteolysis of plasmin, which cleaves plasmin within kringle 5, the generated AS4.5 is released from actin, thus, from the cell surface (Fig. 5).

Interestingly, Dudani et al. proposed that endothelial cell surface actin can bind both plasminogen and angiostatin (24). As they used a different angiostatin isoform (kringles 1-4 produced by limited enzymatic proteolysis of plasminogen, by Hematologic Technologies, Essex Junction, VT) and not the native AS4.5 human isoform, direct comparison of their results with ours is not possible. It is possible that endothelial cell actin may have a role in differential angiostatin binding to endothelial and PC-3 cells. However, we have not been able to show AS4.5 binding to actin in either a cell-dependent or a cell-free system.

Our model may explain some important observations related to angiogenesis and cancer. Although AS4.5 is known to inhibit angiogenesis, plasmin is believed to promote angiogenesis (5, 25, 26). For example, plasminogen activation to plasmin by tPA on endothelial cells results in a 2.5- to 3.0- fold increase in endothelial cell penetration of matrix (25). In another study, Tarui et al. observed that plasmin binds to integrin α_vβ₃ via the kringle domains and that this interaction induces the migration of endothelial cells (26). This migration required both plasmin binding to integrin α_vβ₃ and the plasmin catalytic activity (26). Angiostatin can inhibit both plasmin-induced endothelial cell invasion and migration and has been reported to bind to α_vβ₃ (25, 26). The proangiogenic activity of plasmin may also be mediated through induction of other angiogenic factors. Vascular endothelial growth factors [VEGF; (VEGF-C and VEGF-D)] are potent angiogenesis inducers (27, 28). Plasmin can activate VEGF-C and VEGF-D by propeptide cleavage, generating mature forms that have markedly higher binding affinity to their receptors (29).

Therefore, in the setting of cancer, plasminogen activation to plasmin may have complex effects. The locally generated plasmin may induce angiogenesis by either inducing migration and invasion and/or activating other angiogenic inducers (25, 26, 29). However, if the cancer cell surface also expresses β-actin, then the generation and release of AS4.5 may inhibit metastatic spread as has been postulated (25, 26). The plasminogen-β-actin system may thus help explain the paradigm that angiogenesis inducers act locally, whereas angiogenesis inhibitors act systemically.

Grant support: National Cancer Institute grants P50 CA90386 and P50 CA89018-02.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Donald Davidson for the gift of recombinant kringle 5 protein and valued discussions on this article.