Endostatin Associates with Lipid Rafts and Induces Reorganization of the Actin Cytoskeleton via Down-regulation of RhoA Activity*

Endostatin, the C-terminal fragment of collagen XVIII, is a potent inhibitor of angiogenesis. Observations that endostatin inhibits endothelial cell migration and induces disassembly of the actin cytoskeleton provide putative cellular mechanisms for this effect. To understand the mechanisms of endostatin-induced intracellular signaling, we analyzed the association of recombinant endostatin with endothelial cell lipid rafts and the roles of its heparin- and integrin-binding properties in this interaction. We observed that a fraction of cell surface-bound endostatin partitioned in low density membrane raft fractions together with caveolin-1. Heparinase treatment of cells prevented the recruitment of endostatin to the lipid rafts but did not affect the association of endostatin with the non-raft fraction, whereas preincubation of endostatin with soluble α5β1 integrin prevented the association of endostatin with the endothelial cell membrane. Endostatin treatment induced recruitment of α5β1 integrin into the raft fraction via a heparan sulfate proteoglycan-dependent mechanism. Subsequently, through α5β1 integrin, heparan sulfate, and lipid raft-mediated interactions, endostatin induced Src-dependent activation of p190RhoGAP with concomitant decrease in RhoA activity and disassembly of actin stress fibers and focal adhesions. These observations provide a cell biological mechanism, which plausibly explains the anti-angiogenic mechanisms of endostatin in vivo.

The growth of solid tumors and their metastases is dependent on angiogenesis, the formation of new capillaries from pre-existing vessels. This process is stimulated by various growth factors and, on the other hand, inhibited by various endogenous angiogenesis inhibitors (1). Endostatin, a 183-amino acid C-terminal fragment of type XVIII collagen, is a naturally occurring inhibitor of angiogenesis capable of inhibiting tumor growth and metastasis in various animal models (2)(3)(4)(5)(6). In vitro endostatin interferes with various endothelial cell processes. It inhibits the migration of endothelial cells, induces cell cycle arrest, and promotes apoptosis (2,(7)(8)(9). In addition, endostatin can induce Src-dependent disassembly of the actin cytoskeleton and act as an inhibitor of Wnt signaling (10 -12, 13).
Both heparan sulfate proteoglycans and integrins serve as cell surface receptors for endostatin (14 -16, 11). However, the mechanisms of signal transduction via these receptors to the cytoplasm are incompletely defined. Glycosphingolipid/cholesterol-containing lipid rafts are specialized membrane microdomains, which serve as foci for the recruitment of transmembrane proteins and intracellular signaling molecules at the plasma membrane (17). Several membrane proteins such as glycosylphosphatidylinositol (GPI) 1 -anchored proteins or caveolin, as well as cytoplasmic proteins such as Src family kinases, are enriched in rafts (18,19). Recruitment of transmembrane receptors such as integrins into rafts promotes their interaction with local kinases and phosphatases in the new microenvironment, resulting in downstream signaling (20). The mechanisms by which the recruitment occurs have remained largely unknown, although various models have been proposed (21,22).
Lipid rafts can evidently associate with the actin-rich regions of the cell and harbor proteins involved in the regulation of these structures (23). One of the molecules involved is RhoA, a member of the Rho family of GTPases. Members of the Rho family of proteins regulate the coordinated alterations in cellmatrix interactions with repeated cycles of cell adhesion and detachment resulting in cell migration (24 -27). RhoA cycles between a GDP-bound inactive state and a GTP-bound active state and regulates the assembly of focal adhesions and actin stress fibers. These, in turn, are anchored to the cytoplasmic tails of ligand-bound integrins (28). Inactivation of RhoA is stimulated by GTPase-activating proteins (GAPs), which increase the intrinsic GTPase activity of small G-proteins (29 -31).
The current work was carried out to investigate the association of endostatin with specific membrane microdomains on the surface of endothelial cells. We found that exogenous recombinant endostatin localized to lipid rafts through heparan sulfate proteoglycan and integrin-mediated interactions, and that the raft localization was indispensable for the downstream signaling by endostatin. In addition, endostatin treatment induced loss of actin stress fibers through Src-dependent activation of p190RhoGAP and subsequent down-regulation of RhoA activity. These events play potentially important roles in the action of endostatin.

MATERIALS AND METHODS
Reagents and Antibodies-Recombinant human endostatin was purchased from Calbiochem (San Diego, CA). Rabbit polyclonal antibodies against human endostatin were from Chemicon Inc. (Temecula, CA). Mouse monoclonal anti-vinculin antibodies were from Sigma Chemical Co. Rabbit polyclonal antibodies against caveolin-1 and mouse monoclonal antibodies against p190 and transferrin receptor were from Transduction Laboratories (Lexington, KY). Mouse monoclonal antibodies against glypican-1 were from Santa Cruz Biotechnology Inc.
Cell Culture and Treatment with Chemicals-Human dermal microvascular endothelial cells (HDMEC) were purchased from Promocell (Heidelberg, Germany) and were cultured in Endothelial Cell Growth Medium (Promocell), at 37°C in a humidified 5% CO 2 atmosphere. The cells used for the experiments were from passages 3-6. All experiments were carried out under serum-free conditions, in which the cells were washed twice and incubated in serum-free medium-199 for at least 8 h prior to treatment with the various proteins or chemicals. Where indicated, the cells were pretreated with heparinase III from Flavobacterium heparinum (Sigma) for 30 min prior to the addition of recombinant endostatin. The amount of heparinase used is expressed as conventional units.
To saturate the putative integrin binding sites of recombinant endostatin, 50 nM endostatin was preincubated with purified soluble ␣ 5 ␤ 1 integrin (5 g/ml, 33 nM; Chemicon) in phosphate-buffered saline⅐Ca/Mg (PBS⅐Ca/Mg; 170 mM NaCl, 10 mM sodium phosphate buffer, pH 7.4, 2 mM CaCl 2 , 1 mM MgCl 2 ) at ϩ4°C for 12 h. Subsequently the buffer containing the two proteins was applied to the cells and incubated for time points indicated. The interaction of the two proteins in vitro has been described previously (15), and was confirmed by us using a solid phase ligand-binding assay (see below).
Preparation of Low Density Membrane Fractions-Cells were treated with 50 nM endostatin for 20 min and lysed in 1% Triton X-100 in MES-buffered saline (25 mM MES, 150 mM NaCl, pH 6.5). Cell extracts were homogenized using a 22-gauge needle and adjusted to a final density of 40% by the addition of OptiPrep Density Gradient Medium (Sigma). Samples were placed in a centrifuge tube and overlaid with a discontinuous 30%/5% OptiPrep density gradient. Samples were then centrifuged at 39,000 rpm for 18 h, after which 1-ml fractions were collected and analyzed by immunoblotting as described previously (32).
Solid Phase Ligand Binding Assay-96-well plates were coated with 33 nM purified ␣ 5 ␤ 1 integrin at 4°C for 12 h. Nonspecific binding sites were then saturated with 5% bovine serum albumin at ϩ22°C for 1 h. Endostatin was then added in PBS⅐Ca/Mg at concentrations indicated in the figure and incubated at ϩ37°C for 3 h. Unbound protein was removed by washing the wells with 0.01% Tween-20 in Tris-buffered saline (TBS; 50 mM Tris-HCl, 150 mM NaCl, pH 7.6) containing 1% bovine serum albumin. Bound protein was detected with antibodies against human endostatin, peroxidase-conjugated streptavidin, and its substrate, diluted in 0.05% Tween-20 in TBS, containing 1% bovine serum albumin. Bound antibody was quantitated by measuring absorbance at a 450-nM wavelength.
Immunofluorescence-Cells cultured on glass coverslips were washed with PBS and fixed with 3% paraformaldehyde at 4°C for 10 min. Nonspecific protein binding sites were saturated with 5% bovine serum albumin in PBS for 30 min. The cells were then washed with PBS, and incubated with monoclonal antibodies against vinculin for 1 h. Rhodamine-conjugated phalloidin (Sigma) was used to visualize the actin cytoskeleton. Unbound proteins were removed by washing, followed by incubation with fluorescein isothiocyanate-labeled secondary antibodies (Jackson ImmunoResearch Laboratories, West Grove, PA) for 1 h. The coverslips were finally washed and mounted on glass slides using Vectashield (Vector Laboratories, Burlingame, CA). The fluorescent images were obtained using an epifluorescent microscope.
Immunoprecipitation-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 g/ml aprotinin, 1 g/ml pepstatin A, 1 g/ml aminoethylbenzene sulfonyl fluoride, 1 mM sodium orthovanadate, and 10 mM sodium fluoride) and subjected to immunoprecipitation. Cell lysates were preabsorbed by incubation with preimmune serum and gamma-bind Sepharose (Amersham Biosciences) at 4°C in an end-over-end mixer for 2 h. The beads were removed by centrifugation, and the supernatants were incubated with polyclonal antibodies against p190 on ice for 1 h. Gamma-bind Sepharose was then added and the samples were incubated in an end-over-end 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 by 4 -15% gradient polyacrylamide gel electrophoresis in the presence of SDS (Bio-Rad) under reducing conditions, transferred onto nitrocellulose and subjected to immunoblot analysis. Intensities of the detected bands were quantified from scanned immunoblots using Adobe Photoshop software.

RESULTS
Endostatin Associates with Triton X-100-insoluble Membrane Fractions-Membrane rafts are characterized by their resistance to detergent solubilization at low temperatures and light buoyant density. This allows isolation of a light glycolipidenriched membrane fraction using Triton-X flotation gradients (34,35). Recombinant endostatin administered to endothelial cells associates with caveolin-1, a marker protein for a subclass of lipid rafts (11). To investigate whether this interaction involves the association of endostatin with the lipid rafts, we isolated Triton-X-100 insoluble membrane fractions of cells treated with 50 nM endostatin (1 g/ml) using a discontinuous OptiPrep density gradient. Proteins associated with low density lipid rafts migrate to the interface of 5 and 30% OptiPrep layers (35). We found that a fraction of endostatin localized in the caveolin-and glypican-1 -enriched top fractions of the flotation gradient, indicating raft localization (Fig. 1A, fraction 3). Transferrin receptor was used as a marker for the non-raft fractions (36).
To assess the role of heparan sulfate proteoglycans in the cell surface association of recombinant endostatin, cells were pretreated with heparinase III for 30 min prior to the addition of endostatin. Treatment of cells with heparinase III (5 milliunits/ ml) for 30 min prior to the addition of endostatin significantly reduced the amount of endostatin associated with the low density fractions without appreciably affecting the levels of endostatin associated in the non-raft fraction. No changes in the distribution of caveolin-1 or transferrin receptor in the raft and non-raft fractions were observed in the heparinase-treated cells. The antibody used did not detect glypican-1 from the raft fractions after heparinase III treatment, which was most likely due to efficient cleavage of the heparan sulfate side chains of the protein (Fig. 1B).
To investigate the role of ␣ 5 ␤ 1 integrin in the cell surface association of endostatin, recombinant endostatin (1 g/ml, 50 nM) was preincubated with purified soluble ␣ 5 ␤ 1 integrin (5 g/ml, 33 nM) for 12 h to saturate the putative integrin binding sites of endostatin. Endostatin preincubated with soluble integrin ␣ 5 ␤ 1 displayed decreased association with both high and low density fractions of the cell membrane (Fig. 1C). To confirm that soluble integrin ␣ 5 ␤ 1 interacts directly with endostatin, solid-phase ligand binding assay was performed. The analysis revealed that immobilized, purified ␣ 5 ␤ 1 integrin bound endostatin in a saturable and dose-dependent manner, and that saturation was reached beyond the 50 nM concentration of endostatin (Fig. 1D).
No effects on the association of endostatin with the membrane fractions were seen with non-interacting soluble ␣ 3 ␤ 1 integrin (5 g/ml) or chondroitin sulfate (10 g/ml) (Fig. 1E).
Endostatin Treatment Induces Association of Integrin ␣ 5 ␤ 1 with Lipid Rafts-Several integrins can interact with the lipid raft scaffold protein caveolin-1 (37,38). It has, however, been unclear whether this interaction occurs in lipid rafts. We have previously observed that endostatin treatment induces marked relocation of ␣ 5 ␤ 1 integrin into membrane clusters distinct from the focal adhesions, with concomitant association of these two proteins with caveolin-1 (11). To investigate whether endostatin induces the translocation of the ␣ 5 ␤ 1 integrin to the lipid rafts, we prepared Triton-X-100 flotation gradients from the membranes of HDMECs and collected 8 fractions. Based on the distribution of the raft marker protein caveolin-1 and the non-raft marker protein transferrin receptor in these fractions we pooled fractions 3-5 and 7-8 and marked these fractions as raft and non-raft, respectively. These pools were then analyzed by immunoblotting with polyclonal antibodies against ␣ 5 ␤ 1 integrin and caveolin-1 and with monoclonal antibodies against the transferrin receptor. Very low levels of ␣ 5 ␤ 1 integrin were visible in the raft fraction of the untreated control cells, while the non-raft fraction displayed intensive staining. In endostatin treated cells, the levels of ␣ 5 ␤ 1 integrin in the raft fraction were increased compared with control cells (Fig. 2A,  upper panel). When cells were pretreated with heparinase III (5 milliunits/ml) for 30 min prior to the addition of endostatin, the amount of ␣ 5 ␤ 1 integrin in the raft fraction was decreased to the level of control cells (Fig. 2, upper panel). The treatments had no effect on the distribution of caveolin-1, which was ob-served almost exclusively in the raft fraction ( Fig. 2A, lower  panel). Transferrin receptor staining was restricted to the nonraft fraction ( Fig. 2A, middle panel). Quantification of the scanned immunoblots revealed a Ͼ2-fold increase in the ratio of raft versus non-raft associated ␣ 5 ␤ 1 integrin in endostatintreated cells compared with control cells (Fig. 2B).
Endostatin Induces Phosphorylation of p190RhoGAP and Down-regulation of RhoA Activity-Cell adhesion to extracellular matrices and the concomitant formation of focal adhesions and actin stress fibers are associated with the up-regulation of RhoA activity (28,25). To investigate whether changes in RhoA activity were associated with the endostatin-induced disassembly of focal adhesions and actin stress fibers, Rho activity assays were carried out. Endothelial cells treated with increasing concentrations of endostatin for 45 min were washed and lysed, after which the active GTP-binding fraction was isolated using Rhotekin Rho-binding domain affinity precipitation. Immunoblotting analysis of the isolated polypeptides with antibodies against RhoA revealed a dose-dependent decrease in the levels of active RhoA in the endostatin-treated HDMECs (Fig. 3A).
We have previously observed that endostatin activates the fraction of Src associated with caveolin-1, and that treatment of endothelial cells with the Src inhibitor PP1 can prevent the disassembly of cytoskeletal structures induced by endostatin (11). To characterize the role of Src in the endostatin-induced decrease of RhoA activity, cells were treated with endostatin in combination with PP1. These cells displayed higher levels of active RhoA than the cells treated with endostatin alone (Fig. 3B).
Src regulates Rho activity by tyrosine phosphorylation of p190RhoGAP, a GTPase activating protein with high specific-FIG. 1. Endostatin associates with Triton X-100-insoluble lipid fractions. A, HDMECs treated with 50 nM endostatin for 20 min were lysed in 1% Triton X-100 buffer after which membrane fractions were prepared using density gradient centrifugation. Aliquots of each fraction were analyzed by immunoblotting with antibodies against endostatin. Caveolin-1 and transferrin receptor (Transferrin R) were used as loading controls as well as raft and non-raft markers, respectively. Glypican-1 was used as an additional raft marker. Note the presence of endostatin in the raft fractions 3 and 5 with a peak in fraction 3 and also in the non-raft fractions 7-8. B, heparan sulfate proteoglycans mediate raft localization of endostatin. Cells pretreated with heparinase III were incubated with 50 nM endostatin after which Triton X-100 flotation gradients were prepared. Note that cleavage of heparan sulfate side chains of cell surface proteins using heparinase leads to drastically decreased levels of endostatin and glypican-1 in the raft fractions 3-5. C, interaction with integrin ␣ 5 ␤ 1 is required for the cell surface association of endostatin. Endostatin (50 nM) was preincubated with soluble integrin ␣ 5 ␤ 1 (sol-␣ 5 ␤ 1 ) for 12 h before addition into the cell culture medium. Triton X-100 flotation gradients were then prepared, and the fractions were analyzed by immunoblotting. Note the absence of endostatin in the raft fractions 3-5 and decreased levels of endostatin in the non-raft fractions 7-8 compared with the levels in panel A. D, endostatin binds purified integrin ␣ 5 ␤ 1 . Solid-phase ligand binding analysis revealed that immobilized purified ␣ 5 ␤ 1 integrin (33 nM) bound endostatin in a saturable and dose-dependent manner. Saturation was reached beyond the 50 nM concentration of endostatin. E, no effects on the association of endostatin with the raft membrane fractions (3)(4)(5) were observed with preincubation of endostatin with soluble unrelated integrin ␣ 3 ␤ 1 (sol-␣ 3 ␤ 1 ) or pretreatment of cells with chondroitin sulfate. ity for Rho (31,39,40). To determine whether p190RhoGAP is involved in the endostatin-induced down-regulation of Rho, we analyzed the effect of endostatin on tyrosine phosphorylation of p190RhoGAP. Endostatin treatment led to an increase in the tyrosine phosphorylation of p190RhoGAP as observed by phosphotyrosine immunoblotting of cell extracts immunoprecipitated with antibodies against p190. This effect was partly inhibited when PP1 was added to cells in combination with endostatin (Fig. 3C).
Lipid Raft Localization Is Required for the Cytoskeletal Effects of Endostatin-To determine whether the association of endostatin with lipid rafts is required for the cytoskeletal effects of endostatin, endostatin was depleted from lipid rafts by treating the cells with the cholesterol-sequestering agent filipin III (2.5 g/ml) or the cholesterol-chelating agent methyl-␤cyclodextrin (5 mM) (41). After 30 min of treatment endostatin (1 g/ml) was added and incubated for 15 min. The cells were then lysed in 1% Triton X-100, after which flotation gradients were prepared. Eight fractions were collected and based on the distribution of the raft marker protein caveolin-1 and the nonraft marker protein transferrin receptor we pooled the fractions 3-5 and 7-8 and marked these fractions as raft and non-raft, respectively. These pools were then analyzed by immunoblotting with polyclonal antibodies against endostatin. Analysis of the immunoblots revealed that both filipin III and cyclodextrin treatments inhibited the trans-localization of endostatin into the raft fraction. Caveolin-1 became only slightly redistributed, probably due to the low concentration of the cholesterol binding chemicals used (Fig. 4A). Immunofluorescence analysis of focal adhesions and actin stress fibers was then performed. As observed previously, endostatin treatment led to the disruption of actin stress fibers and focal adhesions. However, when the raft localization of endostatin was inhibited using filipin III treatment, the fibers and focal adhesions were comparable to those of control cells. Filipin III had no independent effects on the cytoskeletal structures (Fig. 4B).
Both Integrins and Heparan Sulfates Are Involved in Mediating the Cytoskeletal Effects of Endostatin-To further explore the roles of integrin ␣ 5 ␤ 1 and heparan sulfates in the endostatin-induced cytoskeletal disassembly, cells were incubated with endostatin (50 nM) in the presence of soluble integrin ␣ 5 ␤ 1 (33 nM) or pretreated with heparinase III (5 mU/ml) for 30 min prior to the addition of endostatin. Subsequently, RhoA activity assays as well as immunofluorescence analysis of focal adhesions and actin stress fibers were carried out. In cells pretreated with heparinase III for 30 min before the addition of recombinant endostatin, no down-regulation of RhoA activity was seen (Fig. 5A). In addition, in contrast to the endostatintreated cells, which lacked actin stress fibers and focal adhesions, cells pretreated with heparinase III before the addition of endostatin displayed intact stress fibers and distinct focal adhesions comparable to untreated control cells (Fig. 5B). Saturation of integrin binding sites with preincubation of endostatin with soluble ␣ 5 ␤ 1 reverted the activity levels of RhoA to the level of control cells (Fig. 5C). In addition, intact focal adhesions and actin stress fibers were seen in cells that were treated FIG. 2. Endostatin induces association of integrin ␣ 5 ␤ 1 into lipid rafts. A, Triton X-100 flotation gradients were prepared from endostatin-treated cells using a 5/30/40% discontinuous OptiPrep gradient. Fractions collected from the 5%/30% interface were termed raft (Ra) and fractions from the 40% suspension non-raft (non-Ra). Aliquots of the fractions were analyzed by immunoblotting using antibodies against ␣ 5 ␤ 1 integrin, transferrin receptor (Transferrin R) or caveolin-1. Note increased level of ␣ 5 ␤ 1 integrin in the raft fraction of endostatintreated cells. Decreased levels of ␣ 5 ␤ 1 integrin in the raft fraction were detected in endostatin-treated cells preincubated with heparinase III (Hep III). Caveolin-1 and transferrin receptor stainings were used as raft and non-raft markers. B, quantification of the scanned immunoblots revealed Ͼ2-fold increase in the ratio of raft versus non-raft associated ␣ 5 ␤ 1 integrin in endostatin treated cells compared with control cells and cells treated with heparinase prior to the addition of endostatin.

FIG. 3. Endostatin induces down-regulation of RhoA with concomitant phosphorylation of p190.
A, subconfluent HDMECs were incubated with endostatin (50 nM) for 45 min, lysed, and subjected to the GTP-Rho affinity pull-down assay followed by immunoblotting with RhoA antibodies. Note the dose-dependent decrease in Rho activity in cells treated with endostatin. Quantification of RhoGTP/Rho signal ratios is indicated below. B, Rho activity assays were performed on cells treated with 50 nM endostatin and increasing concentrations of PP1. Higher levels of Rho activity were detected in cells treated with endostatin in combination with PP1. C, subconfluent HDMECs were incubated with endostatin for 45 min, lysed, and subjected to immunoprecipitation with antibodies against p190. The immunoprecipitates were then analyzed by immunoblotting with antibodies against phosphotyrosine (pTyr). Note the increase in tyrosine-phosphorylated p190 in endostatin-treated cells but not in cells treated with endostatin in combination with PP1 (10 M). Quantification of pTyr/p190 signal ratios is indicated below.
with endostatin preincubated with soluble ␣ 5 ␤ 1 (Fig. 5D). These data indicate that both heparan sulfate proteoglycans and integrin ␣ 5 ␤ 1 modulate the cytoskeletal effects of endostatin. DISCUSSION In the current study we have investigated the role of lipid rafts on certain central cell biological effects such as cytoskeletal disassembly induced by human endostatin. To understand the mechanism behind this, we explored changes in RhoA activation as a result of endostatin treatment. The results indicated that endostatin associated with lipid rafts, and that the association involved integrin ␣ 5 ␤ 1 -dependent binding of endostatin to the cell surface and heparan sulfate-mediated translocation of integrin ␣ 5 ␤ 1 together with endostatin into the raft fraction. Furthermore, the raft localization as well as interactions of endostatin with both ␣ 5 ␤ 1 integrin and heparan sulfate proteoglycans were found to be essential for the endostatin-induced disassembly of the actin cytoskeleton. In addition, we observed that endostatin induced Src-dependent phosphorylation of p190RhoGAP with concomitant down-regulation of RhoA activity.
The biological role of the endostatin fragment of collagen type XVIII is currently unclear, as mice lacking type XVIII collagen display only a mild vascular phenotype. Furthermore, no effects on tumor growth or tumor angiogenesis have so far been observed in the gene-targeted mice (42). However, the concentrations of endostatin used to achieve anti-tumor effects are 10-fold higher than the levels of endostatin in mouse circulation, arguing that the pharmacological effects of high concentrations of endostatin are distinct from its physiological effects. Furthermore, endostatin is sequestered in the basement membranes, where it is not accessible for interaction with cell surface receptors (42). Recent findings have elucidated the FIG. 4. Association of endostatin in lipid rafts is required for endostatin-induced disassembly of actin cytoskeleton. A, cells were treated with Filipin (2.5 g/ml) or ␤-methyl-cyclodextrin (5 mM) for 30 min prior to the addition of endostatin. Triton X-100 flotation gradients were then prepared using a 5/30/40% discontinuous OptiPrep gradient. Fractions collected from the 5/30% interface were termed raft (Ra) and fractions from the 40% suspension non-raft (non-Ra). Aliquots of the fractions were analyzed by immunoblotting using antibodies against endostatin. Note the absence of endostatin in the raft fraction of Filipin-and cyclodextrin-treated (␤-cyclodx) cells. B, The cells were grown to subconfluency on glass coverslips and treated with 50 nM endostatin. After 1 h of incubation the cells were fixed with 3% paraformaldehyde and immunostained with antibodies against vinculin or rhodamine-phalloidin. Note the disassembly of focal adhesions (vinculin staining) and actin stress fibers in cells treated with endostatin but not in untreated control cells or in cells pretreated with Filipin III for 30 min. dual role of integrins as negative as well as positive regulators of angiogenesis (43,44). It has been postulated that the inhibitors of angiogenesis that bind integrins would not serve as inhibitors of physiological matrix-integrin signals, but would, instead, act as signaling molecules themselves (45). They would thus induce signals regulating cell growth, survival, and migration specific for these inhibitors. Our observations on the intracellular signals induced by recombinant endostatin resulting in disassembly of cytoskeletal structures support these hypotheses.
The current findings implicate lipid rafts as specific sites for endostatin binding and downstream signaling. Both heparan sulfates and ␣ 5 ␤ 1 integrins were required for this association and subsequent intracellular signaling. More specifically, the integrin binding properties of endostatin seemed to regulate the binding of endostatin to the endothelial cell surface, and the heparin-binding capacity would, in turn, translocate ␣ 5 ␤ 1 integrin and endostatin into the lipid rafts. This could occur via a raft-localized heparan sulfate proteoglycan acting as a coreceptor for endostatin. Several heparan sulfate proteoglycans have been found to localize in these rafts (46,47). The heparan sulfate proteoglycan glypican, a putative cell surface receptor for endostatin, belongs to the GPI-anchored proteins (14), which are a major class of proteins in the lipid rafts (48,49). Interestingly, we observed that glypican and endostatin were localized in the same detergent-insoluble membrane fraction. A role for lipid rafts in endostatin-mediated signaling was additionally underlined by the observation that the raft localization is necessary for the effects of endostatin on the actin cytoskeleton, as removal of endostatin from the raft fraction with Filipin treatment inhibited the endostatin-induced disassembly of actin stress fibers and focal adhesions. This suggests that the lipid rafts serve as scaffolds for endostatin, integrin ␣ 5 ␤ 1 , and heparan sulfate proteoglycans to interact with their specific downstream signaling partners, caveolin-1-bound Src and p190RhoGAP.
Membrane lipids have been observed to regulate integrin and integrin-associated protein activity (50,51). Furthermore, adhesion of cells to fibronectin localizes ␣ 5 ␤ 1 integrin and Src family tyrosine kinase activity in the rafts (52). These studies have, however, not indicated a role for caveolin-1 or heparan sulfate proteglycans in these rafts. Endostatin associates with both of these proteins, suggesting that the downstream signaling induced by endostatin is not completely analogous to that of fibronectin bound to ␣ 5 ␤ 1 integrin. This is further supported by the observation that, in contrast to soluble endostatin, soluble fibronectin did not induce prolonged down-regulation of RhoA activity. 2 Another possible scenario is that internalization of endostatin is required for its cell biological effects and that this internalization occurs via caveolae or a related raft pathway. There is some evidence for recombinant endostatin being internalized by endothelial cells, but the functional consequences of this process have not been assessed (8).
The disassembly of focal adhesions and actin stress fibers by endostatin has previously been observed to involve tyrosyl phosphatase-dependent activation of Src tyrosine kinase (11). The current results identify other proteins potentially involved in this effect, as down-regulation of RhoA activity was observed in endostatin-treated endothelial cells. Interestingly, Src is involved in the transient down-regulation of RhoA activity occurring as the initial event of integrin ligand binding and this is accompanied by disassembly of actin stress fibers (39). This process promotes plausibly efficient cell spreading and is followed by an increase in Rho activity resulting in reformation of actin stress fibers and focal adhesions. We found evidence for the involvement of Src in the down-regulation of Rho activity by endostatin. Treatment of cells with the Src inhibitor PP1 in combination with endostatin was found to reduce the ability of endostatin to downregulate RhoA activity. Interestingly, the kinetics of this down-regulation was significantly slower than that resulting from the interaction of deposited fibronectin with integrin ␣ 5 ␤ 1 (39), further underlining the exceptional nature of the signals by soluble endostatin. Rho is thought to possess a dual role in the regulation of cell migration. High levels of active Rho inhibit cell migration through promotion of excessive cell adhesion. On the other hand, decreased levels of Rho activity inhibit cell migration probably through disruption of cell-matrix interactions, reducing membrane ruffling and tail-retraction of the migrating cell, and delaying the turnover of actin stress fibers and focal adhesions (53)(54)(55). Since endostatin induces loss of stress fibers persisting for several hours (10), the endostatin-induced down-regulation of RhoA could significantly reduce the speed of endothelial cell migration by delaying the dynamics of actin stress fiber and focal adhesion turnover.
Based on these results and previous data we suggest the following mode of action for endostatin (Fig. 6). Endostatin binds ␣ 5 ␤ 1 integrin on the cell surface. Its simultaneous or subsequent interaction with a heparan sulfate proteoglycan leads to the association of endostatin, ␣ 5 ␤ 1 integrin, and caveolin-1 in the lipid rafts. These interactions induce activation of Src tyrosine kinase associated with caveolin-1. Src then phosphorylates and activates p190RhoGAP, which in turn induces inactivation of RhoA. Finally, down-regulation of RhoA activity leads to the disassembly of actin stress fibers and focal adhe-2 S. Wickstrom, unpublished observations. FIG. 6. Model for endostatin signaling. Endostatin binds to integrin ␣ 5 ␤ 1 on endothelial cell surface, where it simultaneously or subsequently interacts with a heparan sulfate proteoglycan (HSPG) and caveolin-1 (cav-1) and is recruited to lipid rafts. Caveolin-associated Src kinase is then activated, and Src subsequently phosphorylates p190RhoGAP. RhoA is inactivated by p190RhoGAP resulting in the disassembly of actin stress fibers and focal adhesions. These changes may contribute to the anti-migratory effects of endostatin. sions resulting in reduced migratory capacity of the endothelial cell. This cascade represents a novel mechanism by which a fragment of the extracellular matrix and an inhibitor of angiogenesis transmits signals to the cytoplasm and regulates cell adhesion.