Endostatin Binds Tropomyosin

The mechanism of action of Endostatin, an endogenous inhibitor of angiogenesis and tumor growth, remains unknown. We utilized phage-display technology to identify polypeptides that mimic the binding domains of proteins with which Endostatin interacts. A conformed peptide (E37) was identified that shares an epitope with human tropomyosin implicating tropomyosin as an Endostatin-binding protein. We show that recombinant human Endostatin binds tropomyosin in vitro and to tropomyosin-associated microfilaments in a variety of endothelial cell types. The most compelling evidence that tropomyosin modulates the activity of Endostatin was demonstrated when E37 blocked greater than 84% of the tumor-growth inhibitory activity of Endostatin in the B16-BL6 metastatic melanoma model. We conclude that the E37 peptide mimics the Endostatin-binding epitope of tropomyosin and blocks the antitumor activity of Endostatin by competing for Endostatin binding. We postulate that the Endostatin interaction with tropomyosin results in disruption of microfilament integrity leading to inhibition of cell motility, induction of apoptosis, and ultimately inhibition of tumor growth.

The mechanism of action of Endostatin, an endogenous inhibitor of angiogenesis and tumor growth, remains unknown. We utilized phage-display technology to identify polypeptides that mimic the binding domains of proteins with which Endostatin interacts. A conformed peptide (E37) was identified that shares an epitope with human tropomyosin implicating tropomyosin as an Endostatin-binding protein. We show that recombinant human Endostatin binds tropomyosin in vitro and to tropomyosin-associated microfilaments in a variety of endothelial cell types. The most compelling evidence that tropomyosin modulates the activity of Endostatin was demonstrated when E37 blocked greater than 84% of the tumor-growth inhibitory activity of Endostatin in the B16-BL6 metastatic melanoma model. We conclude that the E37 peptide mimics the Endostatinbinding epitope of tropomyosin and blocks the antitumor activity of Endostatin by competing for Endostatin binding. We postulate that the Endostatin interaction with tropomyosin results in disruption of microfilament integrity leading to inhibition of cell motility, induction of apoptosis, and ultimately inhibition of tumor growth.
Angiogenesis, the formation of new blood vessels from existing capillaries, is required for tumors to expand beyond 1-2 mm 3 in size and is the principal determinant for tumor growth (1,2). Consequently, antiangiogenic molecules offer promise as novel therapeutic modalities for the treatment of cancer. Endostatin, a 20-kDa cleavage fragment of collagen XVIII first identified in the conditioned medium of hemangioendothelioma cells (3), is a potent inhibitor of tumor angiogenesis and growth in murine models, and recombinant human (rh) 1 Endostatin is currently being evaluated in the clinic (4,6). Both recombinant murine (rm) and rhEndostatin have been successfully produced and shown to inhibit endothelial cell proliferation and migration in vitro, as well as tumor growth in in vivo tumor models (3,5,6). In fact, the first demonstration of tumor dormancy was achieved with systemic treatment of rmEndostatin (7).
It is increasingly recognized that alterations in the actin cytoskeleton play a crucial role regulating the proliferation and migration of endothelial cells (8,9). Tropomyosins are a large family of proteins; at least 20 different isoforms exist that are generated by alternative splicing of a multigene family (Refs. 10, 11 and references therein). They bind the ␣-helical groove of actin filaments to stabilize actin in the polymerized state directly influencing the integrity of microfilaments and thus play a role regulating reorganization of the actin cytoskeleton. Tropomyosins have been identified in organisms as diverse as yeast and man and are core components of the cell cytoskeleton. Many vertebrate non-muscle cells express between five and eight isoforms of tropomyosin in a tissue-specific manner, leading to speculation that tropomyosin isoforms may have evolved to perform specific functions in the microfilaments of non-muscle cells. Indeed, mutational analysis indicates that tropomyosin isoforms have distinct functions and that they play important roles in a variety of cellular functions, including contraction, cytokinesis, intracellular transport, secretion, motility, morphogenesis, and cell transformation (Ref. 11 and references therein).
The mechanism of action of the antiangiogenic activity of Endostatin is unknown. To elucidate the mode of action of Endostatin, we screened a phage-display library to identify the binding domains of proteins with which Endostatin potentially interacts. Here we present evidence that human tropomyosin isoform 3 (hTM3) shares an epitope with an Endostatin-binding peptide, demonstrate an interaction between rhEndostatin and hTM3 in vitro, and show that rhEndostatin protein binds tropomyosin-containing mircofilaments of endothelial cells. We further show that the peptide mimotope of the hTM3 Endostatin-binding site blocks the antitumor activity of rhEndostatin in vivo and suggest that the antitumor activity of Endostatin results from an interaction with tropomyosin-containing microfilaments that leads to inhibition of microfilament function and an induction of apoptosis.

MATERIALS AND METHODS
Screening of the Phage-display and cDNA Libraries-The Ph.D.-C7C, a disulfide-constrained 7-mer phage-display library (New England BioLabs, Beverly, MA), was screened as recommended by the manufacturer. Following a third round of amplification, individual phage were isolated, and the peptide sequences were deduced by DNA sequencing.
Peptide Synthesis and Generation of Anti-peptide Sera-The E37 (CTHWWHKRCGGGS) and control (CSNSDKPKCGGGS) peptides were synthesized, cyclized at high dilution, and purified to at least 95% purity by high-performance liquid chromatography (Infinity Biotech Research and Resource, Aston, PA). The E37 peptide was coupled to hemocyanin from keyhole limpets (Sigma) in the presence of glutaraldehyde and used to immunize New Zealand White rabbits.
cDNA Library Screening-The Lambda ZAP ® II (Stratagene, La Jolla, CA) bFGF-stimulated HUVEC cDNA library was constructed by directional cloning of oligo (dT)-primed cDNA into EcoRI and XhoI cloning sites. 1.4 ϫ 10 6 recombinant phage were screened for immunoreactivity to E37 peptide antisera.
Native and Recombinant Proteins-Plasmid p21-1 was excised as a pBluescript phagemid (as recommended by Stratagene) from a recombinant phage isolated from the HUVEC cDNA library based on its recognition by E37 antisera. DNA sequencing of the cDNA insert of p21-1 showed it contained the complete hTM3 coding sequence fused inframe downstream of the Lac Z gene. Escherichia coli DH5␣ cells were transformed with p21-1 and pBluescript plasmids and crude bacterial lysates were prepared (12). Human cardiac tropomyosin and muscle tropomyosins were purchased from Trichem Resources (West Chester, PA) and Sigma, respectively.
ELISA-Plates were coated with 5 g/ml of bovine serum albumin or rhEndostatin and blocked. Dilutions of 21-1 or control lysate were added and incubated for 1 h at 37°C, followed by incubation with 1:500 diluted anti-tropomyosin TM311 ascites fluid (Sigma). Following incubation in 1:5000 diluted anti-mouse IgG (HϩL) conjugated to alkaline phosphatase (Promega, Madison, WI) wells were developed using Blue Phos Phosphatase Substrate (Kirkegaard and Perry, Gaithersburg, MD) and the A 635 measured.
Surface Plasmon Resonance (BIAcore)-Interactions between purified soluble rhEndostatin and rhTM3 were evaluated by surface plasmon resonance using the BIAcore 3000 (BiaCore, Piscataway, NJ). The purified rhTM3 was immobilized on the flow cell of a CM-5 BIAcore biosensor chip. The running buffer was 0.01 M HEPES, pH 7.4, 0.15 M NaCl, and 0.005% polysorbate 20 (v/v). The competition experiments involving soluble rhTM3 were performed using PBS, pH 7.4, as running buffer. All measurements were performed at 25°C. To correct for differences in bulk refraction all protein preparations were passed over an activated and blocked flow cell to which no protein had been coupled. Effects due to bulk shift were subtracted from experimental data.
Fluorescent Staining of Endothelial Cells-To detect the binding of Alexa-labeled rhEndostatin 1 ϫ 10 4 HUVEC, HMVEC, and HAECs were plated on 1.5% gelatin-coated Nalge Nunc Lab Tek II chamber slides (Naperville, IL) and fixed in 10% neutral buffered formalin followed by a methanol wash. The chamber slides were incubated in PBS/1% calf serum containing 40 g/ml (2 M) Alexa 488-labeled rhEndostatin 3 for 1 h at room temperature. The slides were washed in PBS, coverslips mounted using fluorescent mounting media (Kirkegaard and Perry Laboratories), and cells were photographed under Alexa 488 excitation wavelength. Co-localization studies were performed as above except cells were incubated in PBS/1% calf serum containing 1:200 mouse anti-tropomyosin TM311 (Sigma) and 40 g/ml of Alexa 488labeled rhEndostatin followed by incubation with goat anti-mouse IgG (HϩL) conjugated to Alexa 594 (Molecular Probes, Eugene, OR). The same fields were photographed under Alexa 488 and Alexa 594 wavelengths. Competition experiments were done by including 100 M E37 or control peptide in the 40 g/ml (2 M) Alexa-rhEndostatin solution used to stain the cell.
In Vivo Studies-C57BL/6J mice (n ϭ 5) were injected intravenously (i.v.) with 5 ϫ 10 4 B16-BL6 cells/mouse. Three days later treatment commenced. 100 l of citrate-phosphate buffer (66 mM sodium phosphate, 17 mM citric acid, 59 mM NaCl, pH 6.2) containing 1.5 nmol of rhEndostatin and 100 l of PBS containing either 0, 15, 75, or 375 nmol of control or E37 peptide were mixed and injected subcutaneously (s.c.) daily for 11 days. All mice were then sacrificed, their lungs removed, and the number of surface metastases counted. Results were analyzed for statistical significance using the 2-tailed Student's t test.
To determine the pharmacokinetics, C57BL/6J mice (3 per time point) were dosed s.c. with 1.5 nmol of rhEndostatin alone or in combi-nation with 375 nmol of either E37 or control peptide. Sera was collected 0.17, 0.5, 1, and 2 h following s.c. administration. Serum drug levels were determined using an Endostatin EIA kit (CytImmune Sciences, College Park, MD) according to the manufacturer's instructions. Pharmacokinetic analyses of the serum concentration versus time data were performed by nonlinear least square minimization using RSTRIP (Version 4.03, MicroMath Scientific Software, Salt Lake City, UT).

RESULTS
An Endostatin-binding Peptide Shares an Epitope with hTM3-A cyclic (disulfide bond-constrained) random 7-mer peptide phage-display library was screened to identify peptides that interacted with rhEndostatin (6). The cyclic, disulfidebonded peptide, E37 (CTHWWHKRCGGGS) was found to bind rhEndostatin. A BLAST search of the National Center for Biotechnology Information database failed to identify proteins with any significant homology to the primary sequence of the E37 peptide. To identify proteins that share common epitope(s) with the E37 peptide, rabbit antisera was raised against KLHcoupled E37 peptide and used to probe an FGF-2 stimulated human umbilical vein endothelial cell (HUVEC) cDNA expression library. DNA sequencing of 9 clones recognized by E37 antisera identified seven as being unique cDNAs of different lengths encoding non-muscle human tropomyosin isoform 3 (hTM3), suggesting that E37 and hTM3 share a common epitope.
A comparison of the primary sequence of hTM3 with that of the E37 peptide failed to reveal significant homology, suggesting that the E37 peptide sequence represents a conformational mimotope of the Endostatin-binding domain of tropomyosin. To confirm that the E37 peptide and human tropomyosin share an epitope, immunoblots of lysate from rhTM3 expressing E. coli  were probed with either TM311 anti-tropomyosin monoclonal antibody (Sigma) or anti-E37 polyclonal sera (Fig.  1). The anti-E37 sera and TM311 anti-tropomyosin antibody specifically recognized common bands of ϳ49 and 43 kDa in the rhTM3 expressing 21-1 lysate whereas neither band was detected in immunoblots of control lysate (Fig. 1). Furthermore, when a lysate of HUVECs was probed with anti-E37 sera or TM311 anti-tropomyosin antibody, a common band of ϳ38 kDa was identified that co-migrated with E. coli expressed rhTM3 (data not shown). These findings demonstrate that the E37 peptide and hTM3 share a common epitope. The anti-E37 sera however failed to recognize tropomyosins purified from human cardiac tissue, rabbit muscle, or chicken muscle (Fig. 1B). The amino acid sequences of rabbit muscle tropomyosin and human cardiac tropomyosin each share 86% identity with hTM3, diverging in two differentially spliced exons, exon 6 (amino acids 189 -212) and exon 9 (amino acids 258 -284). The inability of the anti-E37 sera to recognize either rabbit muscle or human cardiac tropomyosins suggests that the putative Endostatinbinding domain of hTM3 resides within either exon 6 or 9 of hTM3. Because the E37 peptide was identified based on its ability to bind rhEndostatin, the presence of the E37 mimotope may define the specific isoforms of tropomyosin that bind Endostatin.
rhEndostatin and rhTM3 Interact in Vitro-We characterized the interaction between rhEndostatin and tropomyosin by ELISA and showed that rhEndostatin binds rhTM3-expressing bacterial lysate (21-1) (Fig. 2). This result provides evidence of a binding interaction between rhEndostatin and rhTM3.
Surface plasmon resonance, with the BIAcore 3000, was used to assess the kinetics of rhEndostatin/rhTM3 binding. 3200 resonance units (RU) of purified rhTM3 were immobilized on the surface of a sensor chip via amine coupling and rhEndostatin injected over the chip as analyte. The rhEndostatin showed a dose-dependent increase in RU, demonstrating the real-time binding of rhEndostatin to immobilized rhTM3 (Fig. 3A). The dissociation constant (K D ) for the rhEndostatin/rhTM3 interaction was calculated to be ϳ100 M using a steady-state model. When, as a negative control, similar concentrations of rhAngiostatin K1-3 were passed over the rhTM3 chip, no evidence of specific binding to rhTM3 was observed (Fig. 3B), demonstrating the specificity of the rhEndostatin/rhTM3 interaction. To further evaluate the specificity of rhEndostatin/ rhTM3 binding, competition experiments using soluble rhTM3 were performed. Soluble rhTM3 competed for the binding of rhEndostatin to a flow cell containing 2000 RU of immobilized rhTM3 in a dose-dependent manner (Fig. 3C), further demonstrating the specificity of the rhEndostatin/rhTM3 interaction.
Endostatin Binding and Tropomyosin Co-localize to Actin Microfilaments-Incubation of formalin-fixed HUVECs, human microvascular endothelial cells (HMVECs), and human aortic endothelial cells (HAECs) with biologically active Alexa Fluor™ 488-labeled rhEndostatin showed that, given access to the cytosolic compartment, rhEndostatin displays a pattern of binding identical to that observed in cells immunostained for tropomyosin or filamentous actin (Fig. 4, A-C). Fluorescently labeled rhEndostatin and anti-tropomyosin antibodies co-localized to HUVEC microfilaments (Fig. 4, D-E). Similar results were obtained using anti-actin antibodies (data not shown). Thus, in fixed cells, Endostatin-binding, tropomyosin, and actin co-localize to microfilament bundles of the cell cytoskeleton. We postulate that Endostatin, which is specifically internalized by endothelial cells (13), binds tropomyosin to inhibit microfilament function, and as a consequence, tumor angiogenesis.
To determine whether the binding of Alexa-labeled rhEndostatin to the HUVEC cytoskeleton was the result of an interaction with tropomyosin competition experiments were performed. A 50-fold molar excess of the E37 peptide, but not the control peptide inhibited the binding of rhEndostatin to HUVEC microfilaments (Fig. 5, A-C), thus demonstrating the role tropomyosin plays in Endostatin binding to the HUVEC cytoskeleton. This observation further confirms the role of the E37 mimotope in the interaction between Endostatin and hTM3.
E37 Peptide Blocks the Antimetastatic Activity of rhEndostatin in Vivo-To further evaluate the biological significance of the interaction between E37 and Endostatin, we assessed the effect of the E37 peptide on the antitumor activity of rhEndostatin. rhEndostatin has been shown previously to inhibit the growth of pulmonary metastases in the murine B16-BL6 experimental metastasis model in a dose-dependent manner (6). Daily administration of 1.5 nmol of rhEndostatin (1.5 mg/ kg/day) inhibited the growth of experimental B16-BL6 lung metastases by greater than 70% when compared with buffertreated control mice (Fig. 6). The number of metastatic lesions on the lungs of the mice treated with up to 375 nmol/day of E37 peptide in the absence of rhEndostatin was not statistically different from that of the control group, demonstrating that the E37 peptide alone has neither tumor promoting or inhibitory activity in this assay. Co-administration of rhEndostatin with E37 peptide dramatically blocked the tumor growth inhibitory activity of the rhEndostatin (Fig. 6). Co-injection of 1.5 nmol of rhEndostatin with 15, 75, and 375 nmol of the E37 peptide resulted in a dose-dependent decrease in the antitumor activity of the rhEndostatin of 11, 30.5, and 84.5%, respectively. The number of pulmonary metastases in mice co-treated with rhEndostatin and either 75 or 375 nmol of E37 peptide (50-and 250-fold molar excess, respectively) were significantly higher than those on the lungs of mice treated with either rhEndostatin alone or in combination with 75 or 375 nmol of control peptide, respectively (p Ͻ 0.02). The incidence of metastatic lesions on the lungs of mice co-injected with 1.5 nmol of rhEndostatin and up to 375 nmol of a cyclic disulfide-bonded control peptide, or in mice treated with 1.5 nmol of rhEndostatin alone, was statically indistinguishable (p Ͼ 0.11). Thus, in the experimental B16-BL6 lung metastases model, the antitumor activity of rhEndostatin is blocked in a dose-dependent manner by co-injection of E37 peptide. We conclude that because the E37 peptide mimics the binding pocket of TM3, E37 peptide competes with tropomyosin for rhEndostatin binding, resulting in disruption of the tropomyosin/rhEndostatin interaction and the rescue of angiogenesis and tumor growth. An alternative explanation for the inhibitory activity of E37 is that co-administration of peptide with rhEndostatin results in the formation of an insoluble precipitate at the site of injection, leading to reduced bioavailability of rhEndostatin. To test this hypothesis, we measured sera concentrations of rhEndostatin over time following subcutaneous (s.c.) administration of 1.5 nmol of rhEndostatin alone or in combination with either 375 nmol (250-fold molar excess) of the E37 or control peptide. rhEndostatin co-administered with a 250-fold molar excess of E37 peptide had an area under the curve last (AUC last ) equal to 85% of that observed when rhEndostatin was administered alone or when rhEndostatin was dosed with 375 nmol of control peptide (Table I). Assuming 100% bioavailability of rhEndostatin following its administration alone or with control peptide, then the bioavailable dose of rhEndostatin resulting from the co-administration of rhEndostatin and E37 peptide would correspond to 1.3 nmol (1.25 mg/kg). Inference from the B16-BL6 experimental metastasis model where a dose titration of rhEndostatin was used to inhibit metastatic growth (6) suggested that a dose of 1.3 nmol would inhibit lung metastases by 77%. In this study, 1.5 nmol of rhEndostatin co-administered with a 250-fold molar excess of the E37 peptide resulted in an 11% inhibition of lung metastasis (Fig. 6). Thus, despite having a lower AUC last , reduction of the activity of rhEndostatin in the B16-BL6 metastasis model by the E37 peptide cannot be attributed to decreased levels of circulating rhEndostatin. We thus conclude that the loss of tumor growth inhibition was caused by inactivation of rhEndostatin that results from the specific interaction of E37 with rhEndostatin. DISCUSSION We identified hTM3 as an Endostatin-binding protein. Using an epitope-specific antibody (Fig. 1B) we further demonstrated that the Endostatin-binding epitope of hTM3 is not present in all tropomyosins and consequently that Endostatin-binding is presumably not a characteristic of all tropomyosin isoforms. This finding has broad significance. Endostatin inhibits the growth of tumors by inhibiting endothelial cell function (3, 5, 6). Systemic rhEndostatin therapy was shown to induce a 50% reduction of intratumoral blood flow whereas other non-tumor affected organs in the same animal were unaffected. 4 Furthermore, other events that require angiogenesis such as wound healing remained unaffected by rhEndostatin (15,16). These observations suggest that the antiangiogenic effects of Endostatin are tightly controlled. Our finding that Endostatin may only affect specific isoforms of tropomyosin provides an explanation for how this tight control could be achieved and maintained. It is possible that specific isoforms of tropomyosin are expressed and regulated in endothelial cells during tumor angiogenesis. We speculate that these specific isoforms bind Endostatin preferentially leading to potent inhibition of tumor growth. We also postulate that expression of such isoforms of tropomyosin that interact with Endostatin with different affinities and dissociation constants provides the natural homeostatic balance of endothelial cell growth under normal physiological conditions.
The binding of rhEndostatin to immobilized rhTM3 displays both rapid rates of association and dissociation and a K D of ϳ100 M (Fig. 3A). Similar binding kinetics have been observed between T cell receptors and their cognate ligands on the surfaces of antigen-presenting cells and between the glycosylationdependent cell-adhesion molecule-1 (GlyCAM-1) and L-selectin (17)(18)(19). Kinetic analysis of these interactions shows they have relatively low affinities that are also the result of rapid disso-ciation rates. Such kinetics are suggestive of highly dynamic interactions. It is possible that Endostatin may have higher affinity for an as yet undefined isoform of tropomyosin or that a third, as yet unidentified component is involved in stabilizing the Endostatin/hTM3 interaction in vivo, and that in the presence of this molecule(s) the K D of binding would be significantly lower.
Whereas pharmacokinetic analysis indicates that parenteral administration of rhEndostatin results in circulating levels in the nM range (6), local concentrations in the tumor microenvironment may be significantly higher. Planar imaging of mammary tumor-bearing rats following i.v. injection of technetium ( 99m Tc) labeled rhEndostatin (100 Ci/rat) showed that the tumor could be visualized from 0.5-4 h post-injection. 5 This observation suggests that Endostatin targets the tumor microenvironment and indicates significantly higher Endostatin concentrations at the tumor bed in comparison to circulating levels. These elevated local concentrations may favor the interaction of Endostatin with hTM3 in the tumor vasculature.
Whereas normally located intracellularly, tropomyosin has been localized to the surface of colon epithelial cells in ulcerative colitis (20) but surface expression has not been reported for endothelial cells. This does not eliminate the possibility that tumor endothelial cells more closely resemble damaged colonic epithelium in this regard. Dixelius et al. (13) recently showed that endothelial cells, but not fibroblasts internalized Endostatin by an endocytic pathway. This observation together with well established differences that exist between the endothelial cells of normal and tumor tissues, may explain the anti-endothelial cell activity of Endostatin in vitro and the tumor specificity in vivo. However, the precise mechanism(s) by which endocytosed Endostatin gains access to the cytosolic compartment remains to be elucidated.
For the first time we present evidence that links a mechanism of action to the in vivo antitumor activity of Endostatin. Co-injection of the E37 peptide with rhEndostatin resulted in a dose-dependent inhibition of the antitumor activity of rhEndostatin in the B16-BL6 experimental metastasis assay (Fig.  6). We postulate that the E37 peptide mimics the Endostatin protein-binding epitope of hTM3 to compete for Endostatin FIG. 6. The E37 peptide inhibits the antimetastatic activity of rhEndostatin. C57BL/6J mice were injected via the tail vein with 5 ϫ 10 4 B16-BL6 melanoma cells on day 0. On day 3, groups of five mice received daily subcutaneous doses of either buffer control, 1.5 nmol (1.5 mg/kg/ day) of rhEndostatin in combination with 0, 15, 75, or 375 nmol of control or E37 peptide, or control or E37 peptide alone for eleven days. All mice were sacrificed on day 14, and the number of pulmonary surface metastases in each animal counted. Error bars indicate mean Ϯ S.D.

TABLE I Bioavailability of rhEndostatin administered in the presence of E37
peptide in mice Groups of mice (3 mice/time point) were injected s.c. with 1.5 nmol of rhEndostatin alone or in combination with a 250-fold molar excess of either E37 or control peptide. Sera were collected at 0.17, 0.5, 1, and 2 hours following s.c. administration, and assessed for serum rhEndostatin levels as previously described (6) binding and consequently inhibits the antitumor activity of Endostatin. We believe that the antiangiogenic effect of Endostatin is mediated via an interaction with tropomyosin-containing microfilaments. Rearrangement of the actin cytoskeleton has been reported to be characteristic of, and sufficient to induce endothelial cell apoptosis (21). Taken together, the ability of rhEndostatin to induce endothelial cell apoptosis (3,(22)(23)(24) and our observations that rhEndostatin binds tropomyosincontaining mircofilaments, suggests that Endostatin disrupts microfilament function to initiate apoptosis. Further observations consistent with Endostatin exerting an effect upon the actin cytoskeleton come from an endothelial cell monolayerwound assay (6,25). In this assay, cell migration at the wound edge has been correlated with increased incorporation of monomeric actin into filaments and a reorganization of the actin cytoskeleton (26). Endostatin binding to microfilaments may inhibit the cytoskeletal reorganization required for cell migration, leading to an inhibition of migration.
Whereas the mechanism by which the Endostatin and hTM3 interaction may induce endothelial cell apoptosis is poorly defined, tropomyosin and the cytoskeleton have critical roles in cell survival. Ingber and co-workers (27)(28)(29)(30) have demonstrated that actin-based microfilaments play a crucial role in cell shape-regulated determination of cell fate and that whereas growth factors and integrin signaling are required for endothelial cell growth, they are not sufficient (31). Progression through the cell cycle is inhibited following disruption of the cytoskeleton or by release of cytoplasmic tension, implicating the actin-based cytoskeleton and its ability to generate tension against integrin-ECM contacts in the regulation of endothelial cell cycle progression and cell fate (29,32). Thus, we surmise that Endostatin, which has been reported to regulate cell cycle progression in endothelial cells (13,23), may bind cytoskeletal tropomyosin leading to a release of cytoskeletal tension and subsequent induction of cell apoptosis.
Our data indicate that Endostatin exerts its antitumor and antiangiogenic effects as a result of its interaction with tropomyosin isoforms containing the E37 mimotope. We suggest that this interaction leads to disruption of the actin cytoskeleton, that blocks cell migration and contributes to the induction of apoptosis. Interestingly troponin I, better known for its role in muscle contraction where it forms a regulatory complex with troponin T, troponin C, and tropomyosin, has been identified as cartilage-derived inhibitor of tumor angiogenesis and metastasis (33). Troponin I is the actin-binding component of the troponin complex, and thus may inhibit endothelial cells via an interaction with microfilaments that disrupts cytoskeletal function. Thus, troponin I and Endostatin may represent a family of naturally occurring antiangiogenic proteins that block microfilament function.
We recognize that the antiangiogenic activity of Endostatin is complex and potentially involves multiple mechanistic pathways, the convergence of which results in the inhibition of endothelial cell migration, proliferation, differentiation, and an induction of apoptosis. Our study has identified tropomyosin as an Endostatin-binding protein, thus providing a mechanism that may explain the antitumor activity of Endostatin. Endothelial cells rapidly internalize Endostatin (13), and we showed here that rhEndostatin and tropomyosin co-localize to the microfilaments of formalin-fixed human endothelial cells (Fig. 4). The biological relevance of the tropomyosin/rhEndostatin interaction to the antiangiogenic activity of Endostatin was dem-onstrated by blocking antitumor activity of rhEndostatin with a peptide mimotope of tropomyosin. We postulate that the antiangiogenic activity of Endostatin may result, at least in part, from internalization of Endostatin by endothelial cells and its subsequent interaction with tropomyosin; leading to a disruption of microfilament integrity, inhibition of cell motility and induction of apoptosis, and ultimately tumor growth inhibition.