Antiangiogenic Effects of Latent Antithrombin through Perturbed Cell-Matrix Interactions and Apoptosis of Endothelial Cells¹

Angiogenesis, the formation of new capillaries, is a crucial process during embryogenesis, in wound healing, and in the female reproductive organs (1). A growing number of diseases,including cancer and inflammatory disorders, are characterized by excessive, deregulated angiogenesis (2), attributable to increased production of growth factors or to decreased production of angiogenesis inhibitors (3).

Endogenous angiogenesis inhibitors are fragments of abundant proteins,which by proteolytic cleavage gain new characteristics and specifically inhibit endothelial cell function. Examples of endogenous inhibitors are angiostatin, a fragment of plasminogen (4), and endostatin, a fragment derived from the COOH-terminal noncollagenous part of collagen XVIII (5, 6, 7). Angiostatin and endostatin have been shown to arrest tumor expansion in a synergistic manner(8). Furthermore, a M_r29,000 fragment of fibronectin (9) and a M_r 16,000 fragment of prolactin have been shown to possess antiangiogenic characteristics (10). Other strategies for inhibition of endothelial cell function include the use of agents that inhibit VEGF³or VEGF receptor signal transduction (11). Furthermore,synthetic peptides that disrupt the binding of integrins to specific extracellular matrix proteins interfere with endothelial cell survival(12).

The cleaved and latent forms of the serpin antithrombin have been shown recently to have antiangiogenic properties (13). Antithrombin is a heparin-binding protein and the major plasma inhibitor of coagulation proteases, primarily thrombin and factor Xa(14, 15). Antithrombin inhibits its target proteases by exposing its reactive-site loop as a pseudosubstrate for the protease (14). Cleavage at the reactive site triggers the insertion of the reactive-site loop into the main β-sheet of antithrombin. As a consequence of this insertion, the protease is translocated to the opposite pole of the inhibitor and concurrently inactivated. The reactive-site loop of antithrombin can be cleaved, and thereby inactivated, by a number of proteases (16). Moreover, an inactive form of antithrombin, denoted as latent, is produced by heat treatment of the inhibitor (15). In cleaved and latent forms of antithrombin, the reactive-site loop is no longer present on the surface of the antithrombin but is inserted into the main β-sheet of the molecule (17).

We show that latent antithrombin efficiently inhibited tumor angiogenesis in a mouse fibrosarcoma model, when administered s.c. at only 1 mg/kg/day. Latent antithrombin did not affect the activation of FGF receptor 1 but inhibited cell migration toward FGF-2 and activation of FAK, and induced an increase in apoptosis of endothelial cells.

The PAE cell line overexpressing FGFR-1 (18) was cultured in Ham’s F-12 medium supplemented with 10% FCS. Primary BCE cells,kindly provided by Dr. Rolf Christofferson (Uppsala University,Uppsala, Sweden) were cultured in DMEM containing 10% NCS and 3 ng/ml FGF-2 (Boehringer Mannheim). The T241 fibrosarcoma cell line, kindly provided by Dr. Lars Holmgren (Karolinska Institute, Stockholm,Sweden), was cultured in DMEM supplemented with 10% FCS. Media and sera were from Life Technologies.

Reactive loop-cleaved antithrombin was obtained by digestion of purified human plasma antithrombin (19) for 5 min at 37°C with thermolysin (20), followed by chromatography on a 5 ml of HiTrap Heparin-Sepharose column (Amersham Pharmacia Biotech, Uppsala, Sweden). Cleaved antithrombin appeared at about 0.3 m NaCl, appreciably before uncleaved antithrombin, which appeared at about 1 m NaCl. Its behavior in SDS-PAGE was characteristic of that of reactive-loop-cleaved antithrombin(21). SDS-PAGE under nondenaturing conditions showed no aggregates. The thrombin-inhibitory activity of the cleaved antithrombin was <2% of that of uncleaved antithrombin(16). Latent antithrombin was prepared by incubation of plasma antithrombin (4 g/l) for 24 h at 60°C in 10 mm Tris/HCl, 0.5 m sodium citrate, pH 7.4(15). The latent inhibitor was purified by heparin affinity chromatography and eluted at about 0.3 m NaCl. It migrated indistinguishably from native antithrombin in SDS-PAGE under nonreducing and reducing conditions. SDS-PAGE under nondenaturing conditions showed evidence of <5% of aggregates. The ability of the latent form to inhibit thrombin was <2% of that of native antithrombin (15).

The procedure of the CAM assay followed essentially that described previously (22, 23). The CAM was exposed by making a 1 × 1-cm window in the shell of 10-day fertilized hen eggs. Filter discs (Whatman, Inc.) were saturated with 3 mg/ml cortisone acetate (dissolved in 70% ethanol, which was allowed to evaporate before addition of growth factors and inhibitors; Sigma) and soaked in buffer (30 μl for each filter) with or without FGF-2(Boehringer Mannheim; 0.2 μg/filter) and different forms of antithrombin (3, 0.3 or 0.03 μg/filter), and the disc was added to an avascular part of the CAM. After 3 days of incubation, the membrane was inspected in a light microscope (Nikon Eclipse TE 300; ×2.5 or ×4).

The animal work was approved by the local board of animal experimentation and performed according to the United Kingdom Coordinating Committee on Cancer Research guidelines (24). The animals were anesthetized with Isoflurane (Forene; Abbott) during all manipulations. Female C57BL6/J mice (M&B, Ejby, Denmark), 6–8 weeks of age, were acclimated and caged in groups of five. T241 fibrosarcoma cells, 0.5 × 10⁶ in 50 μl of DMEM, were injected s.c. into the left flank of the mouse. Animals carrying palpable tumors within 4 days after injection were randomized to 10-day treatment with 1 mg/kg/day of latent antithrombin,thermolysin-cleaved antithrombin, native antithrombin, or vehicle(PBS), given as a daily s.c. injection in the right flank. The tumors were measured with a caliper once a day, in a double-blind procedure,and their volumes were calculated by the formula π/6 × width² × length. ANOVA was used for statistical analysis. After 10 days of treatment, the mice were sacrificed with a lethal dose of pentobarbitone and perfused with 4% paraformaldehyde in Millonig’s phosphate buffer (pH 7.4). The tumors were then embedded in paraffin according to standard histological procedures and cut into sections 4 μm thick.

PCNA was detected with use of the monoclonal antibody PC10 (0.5μg/ml; Santa Cruz). Sections were pretreated for 2 × 7 min in a microwave oven at 750 W in 10 mm citrate buffer(pH 6.0), followed by immunohistochemical staining using PC10 on a Ventana NexES immunostainer with a diaminobenzidine peroxidase kit(Ventana Medical Systems, Tucson, AZ). The percentage of PCNA-positive cells in 2000 counted cells was estimated. Replacement of the primary antibody with an irrelevant mouse IgG served as a negative control.

TUNEL for detection of apoptotic cells was performed as described(25). Peroxidase-coupled Fab fragments raised against dUTP-digoxigenin (Roche) and diaminobenzidine peroxidase substrate were used for detecting positive reactions. Omission of terminal deoxynucleotidyl transferase enzyme served as a negative control. Sections were counterstained in Mayer’s hematoxylin, dehydrated, and mounted in Mountex resin (Histolab, Göteborg, Sweden). At least 1000 cells per tissue section were counted.

For analysis of tumor angiogenesis, hematoxylin-stained tumor sections were coded, and perfused vascular structures were counted at ×400,using an eyepiece grid of 10 × 10 squares. To adjust for the presence of apoptotic, necrotic, or hemorrhagic areas, the presence or absence of viable tumor tissue in the uppermost square to the far right of the grid was noted and used in the calculations of vascular parameters. Sampling of vision fields and stereological quantification were performed, using the vascular parameters described by Wassberg et al. (26).

PAE/FGFR-1 cells were seeded into 24-well dishes (2 × 10⁴ cells/well). After 3 h, the medium was changed to Ham’s F-12 containing 0.1% FCS. After an additional 12 h, FGF-2 (20 ng/ml), antithrombin (10 μg/ml), or FGF-2 (20 ng/ml) in combination with antithrombin at 1 or 10 μg/ml was added,and added again at days 2 and 4. Numbers of cells in triplicates were scored after 5 days of incubation, using a Coulter counter.

PAE/FGFR-1 cells cultured in Ham’s F-12, 10% FCS were incubated with different forms of antithrombin (3 μg/ml). After 8 h, the medium was changed to Ham’s F-12, 0.1% FCS and fresh antithrombin with or without FGF-2 (10 ng/ml). After 18 h of incubation, the cells were prepared according to the Annexin-V-FLUOS kit (Boehringer Mannheim). Briefly, cells were trypsinized, stained with fluorescein-conjugated Annexin V and 2.5 μg/ml propidium iodide, and analyzed with a flow cytometer (FACSCalibur), with 488 nm excitation and a collecting light scatter, green and red fluorescence. Apoptotic cells were defined as cells with enhanced Annexin V fluorescence simultaneously exhibiting normal propidium iodide staining. The frequency of necrotic cells (with strongly increased propidium iodide staining) showed no differences between the different experimental conditions used.

BCE cells were cultured in DMEM containing 10% NCS with 1 ng/ml FGF-2. After 24 h, the medium was changed to starvation medium, DMEM with 0.25% NCS, containing latent antithrombin (3 μg/ml) or native antithrombin (3 μg/ml). Forty h later, the cells were harvested,fixed, and prepared according to the In situ Cell Death Detection kit-Fluorescein kit with fluorescein-conjugated dUTP (Roche Diagnostics). The percentage of TUNEL-positive cells in 1000 counted cells was determined.

The migration assay was performed in a modified Boyden chamber(27), using micropore nitrocellulose filters (8 μm thick, 8-μm pores) coated with type-1 collagen solution at 100μg/ml (Vitrogen 100; Collagen Corp.). Endothelial cells were preincubated with latent antithrombin (3 μg/ml) for 30 min,trypsinized, and resuspended at a concentration of 5.5 × 10⁵ cells/ml in Ham’s F-12 medium containing 0.1% FCS. The cell suspension was placed in the upper chamber, and serum-free medium containing 0.25% BSA and 5 ng/ml FGF-2 or 5 μg/ml latent antithrombin, individually or in combination, was placed below the filter. FCS at 10% was used as a positive control. After 4 h at 37°C, the medium was removed, and cells adhering to the filter were fixed in pure methanol and stained with Giemsa stain. Cells on the lower side of the filter were counted in three separate microscopic fields. Samples were analyzed in triplicate on three separate occasions.

PAE/FGFR-1 cells were serum starved overnight and stimulated or not with FGF-2 (100 ng/ml), latent antithrombin (3 μg/ml) individually or in combination, for 10 min at 37°C. The cells were lysed in NP40-containing buffer, and samples were separated by SDS-PAGE and transferred to Hybond-C Extra (Amersham Pharmacia Biotech). The membranes were immunoblotted with anti-phosphoFAK antibodies (Biosource International) and subsequently with anti-FAK antibodies (Biosource International). Immunoreactive proteins were visualized by a chemiluminescence detection system (28).

Serum-starved PAE/FGFR-1 on glass coverslips were incubated with or without FGF-2 (100 ng/ml) and latent antithrombin (3 μg/ml) for 20 min at 37°C. The cells were fixed in 3.7% paraformaldehyde,permeabilized with 0.2% Triton X-100, rinsed, and incubated either with FITC-labeled phalloidin in PBS (0.66 μg/ml; Ref.29) or with paxillin antibodies (Santa Cruz) and tetramethylrhodamine isothiocyanate-coupled secondary antibody.

We examined the effects of different forms of antithrombin on growth factor-induced angiogenesis in the CAM. FGF-2 induced a branching network of capillaries, which was efficiently inhibited by coincubation with latent antithrombin at a 10-fold molar excess (Fig. 1) or even at an equimolar concentration (Table 1). VEGF-induced angiogenesis in the CAM was similarly inhibited by latent antithrombin (data not shown). In contrast, native and thermolysin-cleaved antithrombin preparations had only a weak inhibitory effect on growth factor-induced angiogenesis (Fig. 1 and Table 1).

The effects of different forms of antithrombin on the growth of syngeneic T241 fibrosarcoma tumors in C57BL6/J mice were studied. Mice carrying palpable tumors on the left flank were treated with latent,native, or thermolysin-cleaved antithrombin (1 mg/kg) daily by s.c. injections in the right flank. After 10 days of treatment, when the control animal tumor size had reached 2 cm³, the animals were sacrificed. As shown in Fig. 2,A, the tumor volume in the PBS-treated animals was about three times that in mice treated with latent antithrombin. The tumors in mice treated with thermolysin-cleaved antithrombin were only slightly smaller than those in the control animals. Native antithrombin also inhibited tumor expansion, although less efficiently than latent antithrombin (Fig. 2,B). Thus, at 1 mg/kg/day, latent antithrombin was the most efficient of the different forms of antithrombin in halting tumor expansion in this model, which is in agreement with the results obtained in the CAM angiogenesis assay (Fig. 1).

Samples of tumors from the same untreated and latent antithrombin-treated animals, as referred to in Fig. 2,A,were examined with regard to proliferation, apoptosis, and angiogenesis. As shown in Fig. 2,C, there was no difference in the proliferative index between tumors from animals treated with latent antithrombin and with vehicle. The number of apoptotic tumor cells was determined by TUNEL labeling, which specifically labels DNA strand breaks, indicative of apoptosis (25). The number of TUNEL-positive cells increased significantly from 1.5 to 3.1% with latent antithrombin treatment (Fig. 2,C). As shown in Table 2, treatment with 1 mg/kg/day of latent antithrombin led to a decrease in the vessel length density of the tumors, as compared with control(P < 0.05) and to a tendency toward a decrease in both the volumetric and surface density of the vessels. As a control, the mean vessel section area, boundary length, and section diameter were estimated and were found not to vary among the different conditions (data not shown).

The molecular mechanisms underlying the effect of latent antithrombin on angiogenesis were further studied with use of endothelial cells in culture. PAE cells expressing FGFR-1 were used to analyze the effect of latent antithrombin on FGF-2-induced cell growth. Fig. 3 shows that FGF-2 treatment induced an increase in the number of cells to 340% of that in the controls in 5 days. In cultures treated with FGF-2 together with latent antithrombin, the number increased to only 230%. In cultures treated with latent antithrombin alone, the number of cells was slightly decreased compared with the basal condition. The VEGF-induced increase in the number of PAE cells expressing VEGF receptor-2 was similarly diminished by latent antithrombin (data not shown). In contrast, FGF-2-induced proliferation of COS (monkey kidney epithelial) cells was not affected by latent antithrombin (data not shown).

The decrease in the number of endothelial cells in cultures treated with latent antithrombin could be attributable to induction of apoptosis. To investigate this possibility, serum-starved PAE/FGFR-1 cultures were incubated overnight in the presence or absence of latent antithrombin and subsequently stained for Annexin V to detect early plasma membrane changes indicative of the apoptotic process(30). In cultures treated with FGF-2, the fraction of Annexin V-positive cells decreased in comparison with that in the controls (Fig. 4), presumably as a result of growth factor-induced survival. Treatment with latent antithrombin led to a nearly 2-fold increase in Annexin V-positive cells. Addition of FGF-2 to cells treated with latent antithrombin gave slight protection against apoptosis, but comparison of the cultures treated with FGF-2 with and without latent antithrombin showed that the survival signal by FGF-2 was strongly counteracted by latent antithrombin. Primary BCE cells also underwent apoptosis when exposed to latent antithrombin for 40 h, as assessed by TUNEL labeling (data not shown). The number of TUNEL-positive cells increased from 1.5 to 4.2% among cells treated with latent antithrombin. As a control, native antithrombin was added to the BCE cell medium. This led to an increase in TUNEL-positive cells to 2.9% of the cell population. Thus, native antithrombin was less efficient than latent antithrombin in inducing endothelial cell apoptosis but was still active in this respect. In contrast, T241 cell cultures did not show increased apoptosis when treated with latent antithrombin under similar conditions (data not shown).

The ability of PAE/FGFR-1 cells to migrate in a mini-Boyden chamber in the presence or absence of latent antithrombin was examined(27). Cells preincubated with latent antithrombin for 30 min were seeded on one side of a collagen-coated nitrocellulose filter,and FGF-2 with or without latent antithrombin was added to the wells under the filter. The number of cells migrating through the filter,toward the potential stimulator or inhibitor on the opposite side of the filter, during 4 h of incubation at 37°C was counted. As shown in Fig. 5, the relative number of migrating cells increased from 100% (control,BSA-treated) to 150% when stimulated with FGF-2. When latent antithrombin was added together with FGF-2 during the Boyden chamber assay, the number of migrating cells decreased to 105%. Latent antithrombin in itself did not affect migration under these conditions.

With the aim of identifying the molecular mechanisms underlying the effect of latent antithrombin on endothelial cell migration, actin reorganization was studied. Staining with rhodamine-coupled phalloidin(Fig. 6 A) showed that treatment with FGF-2 for 20 min led to formation of dense, actin-containing membrane structures, denoted as edge ruffles. In cells treated with latent antithrombin together with FGF-2 for 20 min, actin stress fibers were present, but the ruffle formation was almost attenuated.

The cytoplasmic tyrosine kinase FAK is localized in focal adhesions and has a critical role in the regulation of cell migration(31). We examined the effect of latent antithrombin on FAK tyrosine phosphorylation by immunoblotting with an anti-phosphoFAK antibody. Fig. 6,B shows that latent antithrombin treatment abolished FGF-2-induced FAK tyrosine phosphorylation, without affecting the FAK protein levels. Treatment with latent antithrombin alone, but not with native antithrombin, led to increased levels of tyrosine phosphorylated FAK. Induction of FGFR-1 tyrosine kinase activity by FGF-2 in PAE/FGFR-1 cells was unaffected by treatment with latent antithrombin (Fig. 6 C). The basal level of FGFR-1 activation in endothelial cells treated with latent antithrombin alone was slightly increased.

PAE/FGFR-1 cells were further stained with an antiserum recognizing paxillin, which is localized in focal adhesion contacts (Fig. 6 D). FGF-2 stimulation of the cells for 20 min led to the appearance of a punctate radial pattern, typical for focal adhesion contacts. Coincubation of the cells with FGF-2 and latent antithrombin prevented formation of focal adhesion contacts. There was no effect on formation of focal adhesion contacts in cells treated with latent antithrombin alone. Together, these results indicate that latent antithrombin exerts its antiangiogenic effects by negative regulation of the actin cytoskeleton, FAK, and focal adhesion formation.

Latent antithrombin, which is inactive as a protease inhibitor, is created by heat denaturation, but a slow conversion of native antithrombin to the latent form may also take place under in vivo conditions (32). We have shown in the present study that latent antithrombin is effective in halting tumor growth at the low dose of 1 mg/kg/day. The observed effects on blood vessel density (Table 2), the unchanged proliferation index, and the increased apoptosis of tumor cells (Fig. 2,C), together with the results from the CAM assay, imply that latent antithrombin exerts its antitumoral effects through inhibition of angiogenesis. Native antithrombin also decreased tumor growth, although less efficiently,whereas tumors in animals treated with thermolysin-cleaved antithrombin were similar in size to those in the control animals. The latent but not the cleaved form of antithrombin was also found to be a potent angiogenesis inhibitors in the CAM. However, native antithrombin did not effectively inhibit neoangiogenesis in the CAM (Fig. 1).

In a recent publication by O’Reilly et al.(13), pancreatic elastase-cleaved and latent antithrombin were shown to be equally potent in arresting expansion of SK-NAS neuroblastoma and Lewis lung carcinoma when administered at 25 mg/kg/day, whereas native antithrombin had no effect in the tumor models investigated. Our results indicate that cleaved antithrombin has a lower antiangiogenic and tumor-inhibitory ability than latent antithrombin. This difference may have been obscured in the study by O’Reilly et al. (13) by the 25-fold higher dose and the longer duration of the treatment. The different effects of cleaved and latent antithrombin are compatible with the marked difference in their three-dimensional structures. Most strikingly, in cleaved antithrombin, residues 402–407 form strand 1 ofβ-sheet C, and residues 394–401 are located in the region originally occupied by the uncleaved reactive-site loop (on “top” of the molecule in the standard orientation) but are highly mobile(17). In contrast, in latent antithrombin, sheet C is disrupted, and the segment between residues 394 and 407 forms an exposed loop on the “side” of the protein (33). Nevertheless, we cannot exclude the possibility that the use of different proteases, in preparing cleaved antithrombin namely elastase in the study by O’Reilly et al. (13) and thermolysin in the present work may have been partly responsible for the different effects in the two investigations. The native antithrombin used in the present study was prepared in-house from human plasma, and according to criteria such as heparin binding and thrombin-inhibitory ability (19), the preparation was homogeneous, native antithrombin. Conversion of a certain amount of the native inhibitor to the latent form in the mice, as suggested for antithrombin in humans (32), is a possibility that would be in agreement with the lack of effect of native antithrombin on CAM angiogenesis.

Latent antithrombin binds heparin with a low affinity(K_d of∼10⁻⁴ m; Ref.34), which implies that binding to heparin or heparan sulfate is not critical in the antiangiogenic effects of latent antithrombin. Furthermore, latent antithrombin does not bind thrombin,which would indicate that the effect is not mediated through interaction with thrombin bound to the endothelial cell receptor,thrombomodulin (35). However, it is possible that the receptor for plasma clearance of antithrombin-protease complexes, the low density lipoprotein receptor-related protein, may be involved. This receptor has been shown to bind such complexes but not cleaved antithrombin (36). It cannot be excluded that lipoprotein receptor-related protein binds latent antithrombin, which would explain the difference in antiangiogenic effect between latent and cleaved antithrombin observed in the present study.

Our data show that activation of FAK in the FGF-stimulated endothelial cells was attenuated in the presence of latent antithrombin. FAK activation has been shown to correlate with cell survival(37). Thus, the increased rate of apoptosis that we observed in the latent antithrombin-treated endothelial cells could be explained on the basis of the block in FAK signaling. Moreover, we observed an increase in FAK activation in endothelial cells treated with latent antithrombin alone. The mechanism of FAK perturbment might involve displacement of integrin-binding to the extracellular matrix, e.g., by presentation of a high affinity binding site in latent antithrombin. However, there is no apparent integrin-binding site, such as an R-G-D tripeptide, in the antithrombin sequence(38). Cells from mice lacking FAK expression as a result of gene inactivation fail to migrate, possibly on account of a defect in focal adhesion turnover (31). We observed a decrease in migration and decreased formation of focal adhesions in response to FGF-2 when cells were treated with latent antithrombin. The regulation of actin filament organization was also disturbed in the latent antithrombin-treated cells. Regulation of the actin cytoskeleton involves Rho family GTPases, which have also been implicated in the regulation of FAK and paxillin phosphorylation during cell adhesion(39). Thus, it is possible that Rho GTPases are primary targets of latent antithrombin and that FAK activation is subsequently perturbed.

We thank Charlotte Wikner, Mari-Anne Carlsson, and Helena Hermelin for excellent technical assistance.