Endostatin Associates with Integrin α₅β₁ and Caveolin-1, and Activates Src via a Tyrosyl Phosphatase-dependent Pathway in Human Endothelial Cells¹

The extracellular matrix is made of a complex fibrillar meshwork of proteins, proteoglycans, and glycosaminoglycans. In addition to providing mechanical strength to cells, it participates in the regulation of cell proliferation, migration, and attachment. Type XVIII collagen is a component of the matrix in endothelial and epithelial basement membranes. It possesses a unique COOH-terminal domain, a 182 amino acid polypeptide termed endostatin, which acts as an endogenous angiogenesis inhibitor when proteolytically cleaved from the native protein. Endostatin is a potent endothelial cell-specific inhibitor of angiogenesis and tumor growth in mice (1, 2). Growth factor-induced endothelial cell migration, an important step in angiogenic vessel sprouting, is impaired in response to endostatin (2, 3). Furthermore, endostatin treatment induces loss of focal adhesions and actin stress fibers, cellular structures crucial for cell motility, in cultured endothelial cells (4). Several cell surface-binding sites for endostatin have been reported (5, 6, 7), but intracellular mechanisms behind the antiangiogenic effect of endostatin are still poorly known.

Endostatin has been observed to associate with cell surface integrins, especially with α₅β₁, the major fibronectin receptor in endothelial cells (6). Binding of integrins to extracellular ligands initiates a series of events including integrin clustering, association with cytoplasmic linker proteins, cytoskeletal reorganization, and downstream signaling (8, 9, 10, 11). The interaction between integrins and extracellular matrix proteins is essential for endothelial cell proliferation and angiogenesis (12). Integrin α₅β₁ has been identified as one of the integrins involved in these processes (13).

Signaling through integrins requires physical interaction with other membrane proteins and subsequent association with cytoplasmic signal transduction proteins. Integrin α₅β₁ has been found to specifically associate with caveolin-1, a transmembrane anchor protein, which couples integrins to cytoplasmic signaling cascades such as the Src pathway (14). Phosphorylation signals generated as a result of integrin ligation regulate cellular events such as the assembly and disassembly of focal adhesions, and reorganization of the actin cytoskeleton (9, 15, 16). Members of the Src family of membrane-associated tyrosine kinases are important regulators of these events; they associate directly with caveolin-1 (17) and are regulated by tyrosine phosphorylation. A constitutively active variant of c-Src, v-Src, has been observed to induce disruption of the actin cytoskeleton, loss of focal adhesions, and impaired deposition of fibronectin matrix (18, 19, 20).

Fibronectin is deposited as a polymerized insoluble matrix by endothelial cells to provide support in cell migration and adhesion during angiogenesis (13, 21). The deposition is a cell-dependent event initiated with binding of soluble fibronectin to cell surface integrins in vinculin-containing focal adhesions (22, 23). Fibronectin deposition requires intact focal adhesions (24), which in turn serve as anchors for actin stress fibers involved in cell locomotion.

The current study was carried out to identify molecular interactions of recombinant endostatin at the endothelial cell surface that could explain the decreased migration of endostatin-treated endothelial cells. We found that exogenous recombinant endostatin associates with caveolin-1 and integrin α₅β₁ in microvascular endothelial cells. In addition, endostatin treatment initiates intracellular tyrosyl phosphatase-dependent signaling resulting in the activation of Src tyrosine kinase and the loss of actin stress fibers. Endostatin-induced disassembly of focal adhesions and actin stress fibers disturbs the deposition of the fibronectin matrix. These changes are likely to contribute to the antimigratory effect of endostatin.

Recombinant human endostatin was expressed and purified as described previously (6). Endostatin effectively inhibited fibroblast growth factor-induced endothelial cell tube formation in collagen, confirming biological activity (data not shown). The effective concentrations used in this study (40–50 nm; 0.8–1 μg/ml) are in agreement with the concentrations used commonly in cell culture (3). bFGF³ was from R&D Systems Inc. (Minneapolis, MN). OV was from Sigma Chemical Co. (St. Louis, MO). PP1 was purchased from Biomol Research Laboratories Inc. (Plymouth Meeting, PA). Rabbit polyclonal antibodies against human endostatin, mouse monoclonal antibodies against α₅β₁ integrin (clone JBS5), and goat polyclonal antibodies against α₅β₁ integrin were from Chemicon Co. (Temecula, CA). Rabbit polyclonal and mouse monoclonal anticaveolin-1 antibodies, and mouse monoclonal antipaxillin antibodies were from Transduction Laboratories (Lexington, KY). Rabbit polyclonal antibodies against Tyr-416 phosphorylated Src was from New England Biolabs Inc. (Beverly, MA). Rabbit polyclonal antifibronectin and mouse monoclonal antivinculin antibodies were from Sigma.

HDMECs were purchased from Promocell (Heidelberg, Germany) and were cultured in Endothelial Cell Growth Medium (Promocell), at 37°C in a humidified 5% CO₂ atmosphere. The cells used for the experiments were from passages 3–6. Unless indicated otherwise, the cells were washed twice and incubated in serum-free medium for at least 12 h before treatment with the various proteins or chemicals. All of the experiments were carried out under serum-free conditions.

Cells cultured on glass coverslips were washed with PBS [170 mm NaCl and 10 mm sodium phosphate buffer (pH 7.4)] and fixed with 3% paraformaldehyde at 4°C for 10 min. When required, the cells were permeabilized with 0.2% Triton X-100 in PBS for 3 min. Nonspecific protein binding sites were then saturated with 5% BSA in PBS for 30 min. The cells were then washed with PBS, and incubated with polyclonal antibodies against fibronectin or monoclonal antibodies against vinculin and paxillin. Rhodamine-conjugated phalloidin (Sigma) was used for the staining of the actin cytoskeleton. Unbound proteins were removed by washing, followed by incubation with Texas Red- or fluorescein isothiocyanate-labeled secondary antibodies (Jackson ImmunoResearch Laboratories, West Grove, PA) for 1 h. Hoechst fluorochrome 33258 (Sigma) was used for fluorescent labeling of the nuclei. The coverslips were then washed and mounted on glass slides using Vectashield (Vector Laboratories, Burlingame, CA). The fluorescent images were obtained using an epifluorescent microscope.

Cells were lysed with RIPA-DOC lysis buffer [50 mm Tris-HCl buffer (pH 8.0), containing 150 mm NaCl, 1% Triton X-100, 0.1% SDS, 1% sodium deoxycholate, 10 mm EDTA, 10 μg/ml aprotinin, 1 μg/ml pepstatin A, 1 μg/ml aminoethylbenzene sulfonyl fluoride, 1 mm OV, and 10 mm sodium fluoride] and subjected to immunoprecipitation. Cell lysates were preabsorbed by incubation with preimmune serum and Gamma-bind Sepharose (Amersham Pharmacia, Uppsala, Sweden) at 4°C in an end-over mixer for 2 h. The beads were removed by centrifugation, and the supernatants were incubated with polyclonal antibodies against α₅β₁ integrin or caveolin-1 on ice for 1 h. Gamma-bind Sepharose was then added, and the samples were incubated in an end-over mixer at 4°C for 1 h. The beads were subsequently collected by centrifugation and washed three times with RIPA-DOC buffer. The bound proteins were eluted with Laemmli sample buffer at 95°C and separated in 4–15% gradient PAGE in the presence of SDS (Bio-Rad, Hercules, CA) under reducing conditions and transferred onto nitrocellulose (Schleicher & Schuell, Dassel, Germany). Immunodetection of the proteins was performed essentially as described (25).

Cells were seeded on plastic culture dishes in complete culture medium and allowed to attach for 1.5 h. After washing, the cells were incubated with endostatin for 3 h in serum-free medium containing [³⁵S]methionine (80 μCi/ml, 1175 Ci/mmol; Amersham Pharmacia Biotech). Subsequently, aliquots of conditioned medium were subjected to immunoprecipitation with antibodies against fibronectin as described above. The eluted polypeptides were separated by 4–15% gradient PAGE in the presence of SDS under reducing conditions and visualized by autoradiography.

HDMECs were treated with 50 nm endostatin for 50 min. The cells were subsequently lysed and subjected to immunoprecipitation with pan-Src antibodies (Src-2; Santa Cruz Biotechnology Inc., Santa Cruz, CA) as described above. The immunoprecipitates were washed three times with RIPA lysis buffer and twice with Src kinase buffer [20 mm HEPES (pH 7.35) and 3 mm MnCl₂]. Autophosphorylation was initiated by adding 5 μCi of [γ-³²P]ATP (5000 Ci/mmol; Amersham Pharmacia Biotech) and carried out in 40 μl of kinase buffer at 30°C for 15 min. The reactions were stopped by adding 10 mm EDTA on ice. The pellets were then washed twice with lysis buffer followed by SDS-PAGE and autoradiography. Incorporated radioactivity was detected and quantitated with Fujifilm BAS-2500 Image Analyzer and MacBAS 2.5 program.

Extracellular matrices were prepared according to Hedman et al. (26). Briefly, cells were seeded on plastic culture dishes in complete culture medium and allowed to attach for 1 h after which the medium was withdrawn and replaced with serum-free medium containing endostatin or various chemicals in concentrations indicated. After 4 h of incubation the cells were washed with PBS and treated three times with 0.5% sodium deoxycholate in 10 mm Tris-HCl buffer (pH 8.0) for 10 min. Subsequently, the plates were washed with PBS and allowed to dry. The resulting extracellular matrices were extracted in reducing Laemmli sample buffer, solubilized by boiling, and separated by SDS-PAGE. The proteins were then transferred onto nitrocellulose and detected by immunoblot analysis.

Confluent HDMECs on glass coverslips in complete culture medium were wounded with a sterile tip and washed with serum-free medium. The cells were treated with 50 ng/ml bFGF and with 50 nm endostatin or 50 nm endostatin in combination with OV immediately after wounding and again after 12 h. After 24 h of incubation the cells were fixed with paraformaldehyde and subjected to immunofluorescence analysis as described above. The migration distance was measured using an inverted phase contrast microscope and UltraView 4.0 software (Perkin-Elmer, Fremont, CA). The mean migration distances were calculated from three reference points on each coverslip by subtracting the distance between the migrating cell fronts measured after 24 h of incubation from that measured at the time of wounding. The results represent three independent experiments.

To ensure that the changes in microvascular endothelial cells observed to be induced by recombinant endostatin were specific for this protein, myelin basic protein was included in all of the experiments as an unrelated peptide control (data not shown).

Endostatin has been shown to associate with integrins α₅β₁ and α_vβ₃ (6). To characterize the interaction between endostatin and α₅β₁ integrin, we immunostained α₅β₁ on HDMECs after the administration of 50 nm recombinant endostatin into the culture medium of serum-starved cells. We observed that endostatin treatment induced clustering of α₅β₁ integrin at the HDMEC periphery (Fig. 1). The integrin clusters were associated laterally with bundles of actin stress fibers as visualized by coimmunostaining with rhodamine-phalloidin and monoclonal antibodies against α₅β₁ integrin. Concomitant loss of actin stress fibers was seen in the endostatin-treated cells compared with untreated control cells with no apparent integrin clusters (Fig. 1). These effects were rapid, becoming visible after 30 min of incubation.

To determine whether the clustering of α₅β₁ integrin involved increased association with membrane adaptor proteins to enable intracellular signaling, we coimmunostained the endostatin-treated cells with antibodies against caveolin-1, a known integrin-associated transmembrane anchor protein (14, 27). We observed that part of the clustered α₅β₁ integrin colocalized with caveolin-1 at the cell periphery (Fig. 2). The clustering of α₅β₁ integrin and its colocalization with caveolin-1 occurred in the same time scale, appearing after 30 min of incubation with endostatin and persisting up to 1 h of incubation (Fig. 2). After 2 h of incubation most of the α₅β₁ integrin became redistributed, but some clustering and colocalization with caveolin-1 could still be seen (Fig. 2 B).

To characterize the interactions between recombinant endostatin, α₅β₁, and caveolin-1, coimmunoprecipitation analysis was carried out. Subconfluent cultures of HDMECs were treated with increasing concentrations of endostatin for 30 min and immunoprecipitated with polyclonal antibodies against integrin α₅β₁ followed by immunoblotting with antibodies against endostatin and caveolin-1. This analysis revealed that endostatin coimmunoprecipitated with integrin α₅β₁ from endostatin-treated cells (Fig. 3,A). The interaction between caveolin and integrin α₅β₁ was dependent on the concentration of endostatin administered to cells while the levels of integrin α₅β₁ protein were equal (Fig. 3,A, bottom panel). Similar coimmunoprecipitation results were obtained with monoclonal antibodies against integrin α₅β₁ (data not shown). No endostatin was coimmunoprecipitated with integrins α_vβ₃ or α₄ (data not shown). Also, endostatin and integrin α₅β₁ coimmunoprecipitated with caveolin-1 from extracts of endostatin-treated cells (Fig. 3 B). Antibodies against caveolin-1 or integrin α₅β₁ did not recognize recombinant human endostatin as confirmed by immunoblot analysis (data not shown).

Src tyrosine kinase activity is regulated via dephosphorylation of the inhibitory phosphorylation site Tyr-527 and phosphorylation of the activating Tyr-416 (28, 29). As Src is involved in integrin-mediated signaling and is thought to control actin stress fibers and focal adhesion turnover, we sought to determine the effect of endostatin on the kinase activity and phosphorylation status of Src. Subconfluent HDMECs were treated with increasing concentrations of endostatin for 50 min and then subjected to immunoprecipitation with antibodies against the COOH terminus of the Src family tyrosine kinases. Immunoprecipitates were washed and subjected to the in vitro autophosphorylation assay. Quantitation of the autoradiographs revealed that more than a 2-fold increase in kinase activity associated with a M_r ∼60,000 Src family kinase in endostatin-treated cells when compared with control cells (Fig. 4 A, bottom panel).

To determine whether the increased kinase activity was because of phosphorylation of caveolin-associated Src at the activating Tyr-416, HDMECs were treated with increasing concentrations of endostatin for 50 min. Immunoblotting analysis of caveolin-1 immunoprecipitates with antibodies against Tyr-416-phosphorylated Src revealed a dose-dependent increase in caveolin-1-associated M_r ∼60,000 phosphoprotein Src in the endostatin-treated cells (Fig. 4,B). This increase was dependent on tyrosyl phosphatase activity as orthovanadate treatment inhibited the endostatin-induced increase in Tyr-416-phosphorylated Src (Fig. 4 B). The levels of total Src protein associated with caveolin-1 were not affected by endostatin treatment (data not shown).

The endostatin-induced increase in Src phosphorylation became visible after 40 min and persisted up to 120 min of incubation (Fig. 4,B). In vitro kinase analysis of caveolin-1 immunoprecipitates additionally confirmed that increased autophosphorylation of a M_r ∼60,000 protein was associated with caveolin-1 in endostatin-treated cells (Fig. 4 C).

To additionally analyze the effect of endostatin on Src phosphorylation, immunofluorescence analysis with antibodies against Tyr-416-phosphorylated Src was carried out. Before the addition of recombinant endostatin, very low levels of Tyr-416-phosphorylated Src staining was visible in cytoplasmic plaque-like structures. After endostatin treatment, intense clusters of staining became apparent in the periphery of the cells, reaching maximal intensity after 60 min of incubation (Fig. 4 D).

We have observed previously that focal adhesions and actin stress fibers are disassembled in endostatin-treated HDMECs (4). Similar changes occur in cells treated with agents decreasing tyrosyl phosphorylation (15). To investigate whether the effects of endostatin on actin stress fibers and focal adhesions involve tyrosyl dephosphorylation, we administered the phosphatase inhibitor orthovanadate to HDMECs in combination with endostatin for 1 h. Whereas in untreated control cells distinct focal adhesions and a network of actin stress fibers were observed, in cells treated with endostatin the vinculin staining was restricted to cell-cell junctions and actin was present as cortical bundles with no evident stress fibers (Fig. 5,A). In cells treated with endostatin in combination with orthovanadate, as well as in cells treated with orthovanadate alone, focal adhesions and stress fibers were present, and cell morphology resembled that of the untreated control cells (Fig. 5 A). The endostatin-induced clustering of α₅β₁ integrin and its increased colocalization with caveolin-1 were not affected with orthovanadate treatment (not shown).

To determine whether the disruption of actin stress fibers and focal adhesions was because of increased activity of Src, we administered PP1, a specific inhibitor of Src family kinases (30), to the cells together with endostatin. The control cells displayed a prominent network of actin stress fibers and multiple cytoplasmic focal adhesion complexes (Fig. 5,B). The endostatin-treated cells as well as the cells treated with 100 μm PP1 alone displayed vinculin staining restricted to cell-cell junctions and cortical actin bundles with no stress fibers present (Fig. 5,B) When treated with endostatin in combination with PP1 and endostatin, the cells displayed cytoskeletal morphology comparable with untreated control cells with distinct stress fibers and cytoplasmic focal adhesion complexes (Fig. 5 B).

The assembly of insoluble fibronectin matrix fibrils occurs at specific assembly sites located in focal adhesions (22, 24). This process is dependent on the binding of soluble secreted fibronectin to integrin α₅β₁ (23, 31). As endostatin both associates with integrin α₅β₁ and disassembles focal adhesions, we sought to investigate the effect of endostatin on the deposition of the fibronectin matrix. Cells suspended in complete culture medium were allowed to attach onto glass coverslips for 1.5 h. The medium was then removed, and after extensive washing, endostatin was added in serum-free medium. After 3 h, the cells were fixed and stained with antibodies against vinculin and fibronectin. Immunofluorescence analysis revealed that the untreated control cells had formed distinct focal adhesions and produced a fibrillar network of fibronectin matrix underneath the cells. In endostatin-treated cells the vinculin staining was diffusely distributed with no visible cytoplasmic focal adhesions. The levels of fibronectin deposited underneath the endostatin-treated cells were decreased and it formed no apparent fibrillar structures (Fig. 6 A).

To investigate whether the modulation of fibronectin matrix deposition by endostatin was dependent on phosphatase activity, orthovanadate was administered to cells in combination with endostatin. HDMECs treated with the combination of orthovanadate and endostatin or with orthovanadate alone displayed prominent focal adhesions and a defined fibronectin matrix with fibrillar structures resembling those of untreated control cells (Fig. 6,A), indicating that tyrosyl phosphatase activity is required for endostatin-induced modulation of the fibronectin matrix assembly. The role of Src kinase activity in the altered deposition of fibronectin was also assayed. When endostatin was added in combination with the Src inhibitor PP1 (100 μm), the cells formed numerous focal adhesion structures and a fibrillar fibronectin matrix, indicating that Src kinase activity is involved in the endostatin-induced altered deposition of the fibronectin matrix (Fig. 6,A). To more quantitatively determine the effect of endostatin on fibronectin matrix deposition, preparations of deoxycholate-insoluble matrices were prepared as described under “Materials and Methods.” Immunoblot analysis of the matrices with antibodies against fibronectin indicated that the levels of deposited fibronectin were decreased in cells treated with 50 nm endostatin or 100 μm PP1 alone. In contrast, in cells treated with endostatin in combination with orthovanadate or orthovanadate alone the levels of fibronectin were comparable with that of untreated control cells. Also, in cells treated with endostatin together with PP1, the level of deposited fibronectin was higher than in cells treated with either of the chemicals alone (Fig. 6 B).

To determine possible alterations in the levels of soluble secreted fibronectin in endostatin-treated HDMEC cultures, cells in complete culture medium were plated on plastic culture dishes and allowed to attach for 1.5 h. The cells were then metabolically labeled, and increasing concentrations of endostatin were added. Aliquots of conditioned medium were collected after 3 h and subjected to immunoprecipitation. In addition, the cell layers were trypsinized, and the cells were counted. Analysis of the amount of fibronectin precipitated per cell revealed that endostatin treatment enhanced the accumulation of soluble fibronectin to the medium in a dose-dependent manner (Fig. 6 C).

Endostatin has been shown to inhibit endothelial cell migration (2, 3). To correlate our observations with this effect, we introduced an artificial wound to confluent cultures of HDMECs using a sterile tip. The cells were stimulated to migrate using the angiogenic growth factor bFGF (25 ng/ml) and incubated with endostatin under serum-free conditions, as indicated. The treatments were repeated after 12 h. After 24 h of incubation, the cells were fixed and stained with antibodies against paxillin to visualize focal adhesions, and antibodies against fibronectin to visualize the fibrillar matrix. Staining of nuclei was used to mark the cells migrating on the coverslips. Measurement of the migration distance together with immunofluorescence analysis revealed that the untreated control cells migrated readily across the wound, were rich in focal adhesions at the leading cell edge, and produced a fibrillar fibronectin matrix (Fig. 7). The endostatin-treated cells remained immobile at the edges of the wound, lacked focal adhesions, and were unable to assemble a provisional fibronectin matrix (Fig. 7).

Orthovanadate treatment in combination with endostatin moderately increased cell migration, and inhibited the observed effects of endostatin on focal adhesions and fibronectin matrix in a dose-dependent manner. These cells displayed focal adhesions and deposited fibronectin fibrils that were comparable with those observed in untreated control cells or cells treated with orthovanadate alone (Fig. 7). Increasing the concentration of orthovanadate up to 500 μm did not notably increase the migration of the endostatin-treated cells (not shown).

In the present study we have analyzed the plasma membrane interactions of recombinant human endostatin in microvascular endothelial cell cultures. Endostatin-treated cells displayed clustering of α₅β₁ integrin with subsequent increased colocalization with caveolin-1. In addition, caveolin-1 and exogenous endostatin were recovered from cell extracts and subjected to immunoprecipitation with antibodies against α₅β₁ integrin and vice versa, additionally indicating colocalization. Studies on the intracellular effects of endostatin treatment revealed that endostatin enhanced Src kinase activity and increased the level of phosphorylated Src associated with caveolin-1 in a phosphatase-dependent manner. Endostatin treatment induced disassembly of focal adhesions and actin stress fibers, and disturbed the deposition of the fibronectin matrix through a mechanism involving Src and tyrosyl phosphatase activity. In addition, wound-induced endothelial cell migration studies indicated that the lack of focal adhesions and fibronectin matrix is associated with the antimigratory effect of endostatin.

Integrin α₅β₁ is an important regulator of tumor angiogenesis in vivo, and blocking this integrin by specific antibodies or peptides inhibits bFGF-induced angiogenesis in vitro (13). Furthermore, inhibition of integrin α₅β₁ function with specific blocking antibodies results in the loss of endothelial cell focal adhesions and actin stress fibers (32), resembling the effect of endostatin treatment. Endostatin has been shown to associate with integrin α₅β₁ in an RGD-independent manner (6). Our observations additionally elucidate the interaction between endostatin and molecules involved in cell-matrix interactions at the plasma membrane. Endostatin treatment induced the clustering of stress fiber-associated integrin α₅β₁ on the cell surface in a manner not resembling the pattern of integrin activation when occupied by an immobilized ligand. The endostatin-induced integrin clusters did not appear to colocalize with vinculin or paxillin in focal adhesions, but directly with lateral actin stress fibers.

Additional investigation of the interaction revealed colocalization of integrin α₅β₁ and the transmembrane anchor protein caveolin-1 in response to endostatin treatment. In addition, endostatin, through an undefined mechanism, associated with these two membrane-bound proteins. The presence of caveolin-1 in a complex with endostatin and integrin α₅β₁ suggest a role for caveolae, liquid-ordered domains involved in membrane trafficking, in the effects of endostatin. However, the interaction between integrin α₅β₁ and caveolin-1 has been suggested to occur independently of lipid rafts (14) as integrins are generally considered not to localize in caveolae (33). Other authors suggest that caveolae serve as temporary scaffolding domains for signaling units consisting of integrins, caveolin-1, and Src kinases, and these structures are disassembled in response to integrin activation and recruitment to focal adhesions (16). In the light of these findings, our observations imply that endostatin treatment inhibits the formation of focal adhesions and instead, leads to the capping of α₅β₁ integrins on the cell surface together with increased or stabilized association with caveolin-1. This interaction then results in the observed activation of the associated signaling molecules. The recruitment of integrin-endostatin complexes to caveolin-containing structures could also occur via a coreceptor for endostatin, such as a heparan sulfate proteoglycan (5, 7).

Protein phosphorylation and dephosphorylation serve as molecular switches in the control of endothelial cell migration and adhesion. The migration-associated assembly and disassembly of focal adhesions and actin stress fibers is regulated through their phosphorylation status (15). Important regulators of these processes are the Src family kinases that are activated by integrin ligand binding and regulate the turnover of the focal adhesions. Our results indicate that endostatin acts as a soluble agonist, not an antagonist, for integrin α₅β₁, and endostatin treatment leads to increased activation of M_r ∼60,000 Src associated with caveolin-1. However, these events may occur independently of each other. The activation of Src was inhibited by OV, suggesting phosphatase-dependent activation of the M_r ∼60,000 Src. Interestingly, the increase in Src activity was relatively sustained, persisting for up to 120 min, which is in contrast with the transient induction of Src phosphorylation by integrins ligated by physiological, immobilized substrates (34). This observation, together with the finding that endostatin-induced disassembly of focal adhesions and actin stress fibers was antagonized by the Src inhibitor PP1, which alone is capable of inducing disassembly of the cytoskeleton, suggests that endostatin binding to the cell surface leads to aberrant and perturbed signaling, resulting in excessive turnover of focal adhesions and actin stress fibers. Analogously, a similar absence of focal adhesions and actin stress fibers can be observed in cells transformed with the constitutively active v-Src (35, 36). Here it should also be noted that various protein tyrosine phosphatases function in the disassembly of focal adhesions in multiple cell systems (37, 38, 39, 40, 41).

Previous studies have indicated that intact focal adhesions and actin stress fibers are required for fibronectin matrix deposition (24). We observed that endostatin-treated endothelial cells lacked focal adhesions and actin stress fibers, and were unable to assemble fibronectin matrix fibers. In addition, increased levels of soluble secreted fibronectin were found in the conditioned medium of endostatin-treated cells suggesting that the cells are unable to deposit the secreted soluble fibronectin into insoluble matrix fibers. Inhibition of the endostatin-induced focal adhesion disassembly and loss of fibronectin in cells treated with the phosphatase inhibitor orthovanadate or the Src kinase inhibitor PP1 established an association between these two events. This suggests that the Src-mediated disassembly of focal adhesions by endostatin results in altered fibronectin deposition and subsequent accumulation of soluble fibronectin in the culture medium. The disturbed deposition of fibronectin also occurs in v-Src transformed cells (20). However, the contribution of outside-in signaling from the deficient fibronectin matrix to the disassembly of the cytoskeletal structures cannot be excluded on the basis of our results. Interestingly, the fibronectin matrix has been described to play an important role in angiogenesis in vivo (13), and agents disrupting the fibronectin matrix have been observed to inhibit angiogenesis, tumor growth, and metastasis (42, 43), implying that these observations could correlate with the biological functions of endostatin.

To study the functional significance of the endostatin-induced disassembly of focal adhesions and decreased assembly of the fibronectin matrix, we carried out a bFGF-induced wound migration assay. In agreement with previous observations (2), we found that endostatin-treated cells stimulated with bFGF were unable to migrate across a wound made to the culture monolayer. In addition, the cells located directly on the edges of the wound lacked focal adhesions on the leading edge and were unable to deposit a provisional fibronectin matrix. This was in marked contrast to the untreated control cells, which had focal adhesions and fibronectin matrix fibers, and which migrated readily across the wound. The finding that impaired migration of endostatin-treated endothelial cells could partly be restored by adding the phosphatase inhibitor orthovanadate, supports the hypothesis that the disassembly of focal adhesions together with decreased deposition of fibronectin contribute to the antimigratory effect of endostatin. However, additional mechanisms are likely to be involved in the endostatin-induced inhibition of endothelial cell migration, because even high doses of orthovanadate were not able to restore the migration of endostatin-treated cells to the level of untreated control cells. One putative mechanism is increased apoptosis, which has been shown to occur in endothelial cells in response to treatment with antagonists of integrins and also endostatin (44, 45).

We thank Drs. Marko Rehn and Kristiina Vuori for providing recombinant human endostatin. We also thank Sami Starast for fine technical assistance.