Vascular endothelial-cadherin (VE-cadherin) is an endothelial cell-specific adhesion molecule that is localized exclusively at cell-cell contacts referred to as adherens junctions. VE-cadherin-mediated adhesion is crucial for proper assembly of vascular structures during angiogenesis as well as for maintenance of a normal vascular integrity. We have shown previously that a monoclonal antibody (BV13) to VE-cadherin not only inhibits the formation of vascular tubes during tumor angiogenesis but also disrupts adherens junctions of normal vasculature with a concomitant increase in vascular permeability. The goal of the current studies was to block VE-cadherin function during angiogenesis without disrupting existing junctions on normal endothelium. Using in vitro screening assays to test for functional blocking of adherens junction formation and in vivo assays to detect antibody effects on vascular permeability in normal tissues, we have identified a novel blocking antibody (E4G10) that inhibits VE-cadherin function during angiogenesis but does not disrupt existing adherens junctions on normal vasculature. E4G10 inhibited formation of vascular tubes in vivo in the Matrigel plug and corneal micropocket assays. E4G10 also inhibited tumor growth in three models of mouse and human tumors via an antiangiogenic mechanism. Examination of normal mouse and tumor tissues showed that E4G10 bound to endothelial cells in a subset of tumor vasculature but not to normal vasculature. Bromodeoxyuridine labeling experiments showed that E4G10 specifically targeted a subset of tumor endothelium that is undergoing active cell proliferation, which likely reflects the activated, angiogenic endothelium. These findings indicate that VE-cadherin can be selectively targeted during states of pathological angiogenesis, despite its ubiquitous distribution throughout the entire vasculature. Our data also suggest that antibody E4G10 recognizes VE-cadherin epitopes that are only accessible on endothelial cells forming new adherens junctions, such as in angiogenic tumor vasculature.

VE-cad2 is an endothelial cell-specific adhesion molecule that is localized within specialized structures at cell-cell contacts, referred to as adherens junctions (1, 2, 3). Accumulating evidence implicates VE-cad in various aspects of vascular biology including endothelial cell migration (1), survival (4), contact-induced growth inhibition (5), vascular integrity (6, 7, 8), and most notably, endothelial cell assembly into tubular structures (9). The importance of VE-cad in angiogenesis was also demonstrated in studies of VE-cad null mice (4). The assembly of vascular structures is severely impaired in VE-cad null embryos, leading to embryonic lethality by day E9.5, thus establishing VE-cad as an important mediator of developmental angiogenesis.

The importance of VE-cad in postnatal angiogenesis and maintenance of vascular integrity has been demonstrated by targeting this molecule with antibodies. mAb BV13 was shown to inhibit tumor angiogenesis, tumor growth, and metastasis in various mouse models (8). BV13 also appears to target VE-cad molecules expressed on endothelial cells of normal quiescent vasculature. Systemic administration of BV13 at high levels induced vascular permeability in normal tissues, especially the lung, leading to pulmonary edema and death (7). These findings indicate that certain anti-VE-cad mAbs (such as BV13) may disrupt adherens junctions in all endothelia. Although this work established VE-cad as a potential target for antiangiogenic therapeutic intervention, these data also provide evidence for the limitations of this approach.

VE-cad is expressed constitutively throughout the entire vasculature; however, the cellular localization and structural organization of VE-cad molecules changes, depending on the functional state of the endothelium (3). Prior to assembly of tubular structures, VE-cad molecules are randomly distributed on the surface of endothelial cells. In newly forming vasculature, these adhesion molecules undergo redistribution and reorganization to form adherens junctions, a hallmark of established endothelium where endothelial cells remain in contact with each other (9). At the molecular level, the formation of adherens junctions is mediated by homophilic interactions of the extracellular regions of VE-cad molecules located on adjacent endothelial cells (1). The precise molecular nature of VE-cad-mediated adhesion has not yet been defined but is likely to be similar to that of other classic cadherins (10, 11, 12). The extracellular region of VE-cad consists of five ectodomains that appears to participate in the homophilic interaction in a complex manner, involving the dynamic interdigitation of several subdomains of VE-cad from the neighboring cells. Given the complexity of VE-cad-mediated adhesion, we hypothesized that certain regions of the VE-cad molecule may be exposed to the local environment only prior to the formation of endothelial cell-cell contacts. Such temporally exposed regions might be accessible to specific anti-VE-cad mAb. Furthermore, if mAb directed against these epitopes could be identified, they might be able to disrupt adherens junction formation without toxic liability attributable to interference with established adherens junctions and thus could be of potential therapeutic value as antiangiogenic agents.

In this report, we describe a novel mAb (E4G10) that inhibits adherens junction formation during angiogenesis but does not disrupt existing adherens junctions on normal vasculature.

Cells and Media.

Human A431 epidermis carcinoma cells (CRL-1555, American Type Culture Collection, Manassas, VA), C6-V10 glioma cells (from Dr. L. E. Benjamin, Harvard Medical School, Boston, MA; Ref. 13), and mouse endothelioma cells (H5V; from Dr. A. Vecchi, Instituto Mario Negri, Milan, Italy; Ref. 14) were grown in DMEM (Life Technologies, Inc. Gaithersburg, MD) containing 10% fetal bovine serum (HyClone Laboratories, Logan, UT). Mouse endothelial cells (1G11; Ref. 14) were grown in DMEM containing 20% fetal bovine serum supplemented with 100 μg/ml ECGs (Sigma Chemical Co., St. Louis, MO) and 100 μg/ml heparin (Sigma).

Antibody Generation and Production.

Antibody BV13 was generated as described (7). Other mAbs were generated using either affinity-purified soluble murine VE-cad-Ig fusion protein (15) or keyhole limpet hemocyanin-coupled synthetic peptides derived from the NH2-terminal subdomain of murine VE-cad as immunogens. Four peptides were chosen based on a structural model of the VE-cad NH2-terminal subdomain derived from the X-ray structure of the corresponding E- and N-cad domains (16, 17). These peptides correspond to the regions of AA46–60 (peptide 1), AA64–80 (peptide 2), AA78–95 (peptide 3), and AA116–131 (peptide 4). Peptides were synthesized, purified, and coupled with keyhole limpet hemocyanin at their COOH termini by Quality Controlled Biochemicals (Hopkington, MA). Supernatants from hybridoma clones were screened for binding to VE-cad-Ig immobilized on ELISA plates as described previously (7). Hybridoma supernatants derived from peptide immunization were also tested for their ability to bind to the respective peptides immobilized to ELISA plates (7).

Hybridoma supernatants from VE-cad-binding clones were further screened using two in vitro assays described below to examine the mAb ability to inhibit adherens junction formation and to disrupt existing adherens junctions. Antibodies were purified from conditioned medium by protein G-Sepharose chromatography and contained ≤1.25 EU/ml endotoxin, as assessed by the Limulus Amebocyte Lysate assay kit (BioWhittaker, Walkersville, MD).

[125I]-labeled mAb Binding Assay.

Antibodies were labeled with 125I (NEN Life Science Products, Inc., Boston, MA) following the previously described procedure (18). Various amounts of 125I-labeled mAb (0.0001–30 nm) at a specific activity of 5 × 107 cpm/μg were added to wells of 96-well plates, each containing 5 × 105 mouse endothelial cells (H5V) in 200 μl of PBS/0.1% BSA with or without 10 nm cold mAb. The reactions were allowed to proceed for 2 h at room temperature, followed by three washes with ice-cold PBS/0.1% BSA. The cell pellets were then resuspended in 200 μl of PBS/0.1% BSA, and the radioactivity of the mixtures was determined in a gamma counter. Specific 125I-labeled mAb binding was defined as the difference between the presence and absence of 10 nm nonradioactive corresponding mAb. Duplicate samples were analyzed, and binding affinities were calculated by Scatchard analysis (19, 20).

Calcium Switch Assay.

This assay was used to identify mAb that inhibit adherens junction formation. Briefly, VE-cad-mediated adherens junctions of endothelial monolayers established on glass cover slips were disrupted by chelating calcium from the culture medium. Reformation of adherens junctions was then allowed to take place in the presence of test mAb after resupplementing calcium. The assay was performed as described previously (15), with minor modifications. Monolayers of H5V or 1G11 cells were treated with 5 mm EGTA for 15 min at room temperature. After washing and restoration of a normal calcium concentration (2 mm), mAbs were added at various concentrations, and cells were incubated for 2 h at room temperature. Junction status was scored by microscopic evaluation of VE-cad distribution on cells. Control EGTA-treated cell monolayers were routinely evaluated by nuclear staining with 4,5-diamidino-2-phenylindole (Sigma) and by immunohistochemistry using anti-PECAM/CD31 mAb (BD PharMingen, San Diego, CA) to ensure that cells did not detach or retract, respectively, during EGTA treatment. The effect of mAb on junctional reformation was assessed by image analysis of fluorescence intensity in the areas of adherens junctions.

In Vitro Binding Assays.

These assays were used to test the ability of mAb to disrupt existing adherens junctions. Mouse endothelial cells (H5V) were seeded onto the 96-well tissue culture plates (5,000 cells/well in 200 μl) and cultured for 3 days to establish a complete monolayer. Control wells were stained with crystal violet to ensure proper formation of monolayers. Cell monolayers in half of the wells were treated with 5 mm EGTA for 15 min to artificially disrupt the adherens junctions. After three washes with PBS containing 0.02% Tween 20 (PBS-T), mAbs were added into EGTA-treated and control cell monolayers at various concentrations (1–50 μg/ml) and incubated with cells for 2 h at 37°C. After another three washes with PBS-T, 1 μg/ml of goat anti-rat-horseradish peroxidase (BioSource, International, Camarillo, CA) was added to reactions and incubated for 1 h at room temperature. Antibody binding to cells was assayed as described (15).

In Vitro Permeability Assay.

To quantify the disruptive effects of mAb on endothelial cell monolayers, H5V cells (104 cells/well) were seeded into the upper chamber of Costar transwells (24 wells, 0.4 μm pore; Corning Costar Corp., Cambridge, MA) and allowed to form a complete monolayer. The permeability of cell monolayers was assessed by adding test mAb and FITC-dextran (40S; Sigma) to the top chamber at a final concentration of 1 mg/ml and incubating for 2 h at room temperature, followed by determination of the FITC-dextran concentration in the bottom chambers of transwells by fluorimetry (model 450; Sequoia-Turner, Las Vegas, NV) using absorption/emission wavelengths set at 492/520 nm (15).

In Vivo Permeability Assay.

Various doses of test mAbs (ranging from 25 to 3000 μg/mouse) were administered to NIH nu/nu athymic mice and C57BL/6 mice (Harlan Sprague Dawley Inc., Indianapolis, IN) by single i.p. injection (four mice/group/dose). Vascular permeability in tissues was analyzed by a modified Miles assay as described previously (7). In brief, Evans blue dye (100 μl of 1 mg/ml; Sigma) was injected i.v. 21 h after mAb administration. Fifteen min later, mice were anesthetized by ketamine and perfused with >20 ml of saline via the left ventricle of the heart. Mouse organs were removed and homogenized in trichloroacetic acid:ethanol (1:1 v/v). Evans blue content in tissue homogenates was quantified by spectrophotometry (absorbance, 610 nm).

In Vivo Angiogenesis Assays.

The Matrigel plug and mouse corneal micropocket assays were performed as described previously (8). For Matrigel plug studies, 0.5 ml of Matrigel (BD, Biosciences, Bedford, MA) supplemented with 500 ng of bFGF (R&D Systems, Minneapolis, MN) and 10 μg of VEGF165 (made in-house) was injected s.c. into C57BL/6 mice. Antibodies were administered i.p. to mice twice weekly starting 24 h after plug implantation for a total of 21 days. Matrigel plug vascularization was quantified by injecting mice i.v. with FITC-dextran (200S; Sigma; at 20 mg/kg weight) on day 21 and measuring the uptake of FITC-dextran into plugs after their removal from mice by fluorimetry (8). For the mouse corneal micropocket assay, hydron-coated sucralfate pellets containing 50 ng of bFGF were implanted into mouse corneas 1 mm above the limbic vessel. Antibody was administered to mice i.p. at day 1, 3, and 5 after implantation of hydron pellets. Angiogenesis in mouse corneas was photographed 6 days after bFGF pellet implantation and quantified by measuring angiogenic area and vessel length using image analysis as described (8).

Tumor Models.

Studies using human A431 epidermoid tumor were performed in nude mice as described (8). For the C6-V10 glioblastoma (13) and H5V hemangioma (21) studies, 2 × 105 cells in 200 μl of saline were injected s.c. into the right flanks of athymic mouse. Mice were treated with either 50 μg/dose of BV13 or 1 mg/dose of E4G10 i.p. twice weekly, starting immediately after inoculating the rapidly growing C6-V10 or H5V cells or when tumors reached a size of 250 mm3 (A431). In one of the A431 studies, a range of doses (0.2–3 mg/every 3 days) of E4G10 was tested. Tumor size measurements were performed as described (8).

Statistical Analysis.

Tumor volumes and histological quantifications were analyzed using a Student’s t test. Analysis was computed using the Sigma Stat statistical package (v.205; Jandel Scientific, San Rafael, CA).

Histology.

Paraffin-embedded tissue sections (4 μm) were examined for antibody-induced lung pathology after staining with Mayer’s H&E (Sigma). Anti-CD31 staining and terminal deoxynucleotidyl transferase-mediated nick end labeling on the frozen sections of the A431 tumors were performed as described previously (8). Ki67 staining to analyze cell proliferation was performed using a polyclonal anti-human Ki67 antibody (DAKO Corp., Carpinteria, CA). In brief, deparaffinized tumor sections were rehydrated, incubated in the Target Retrieval Solution (DAKO) at 95–99°C for 40 min, and cooled to room temperature for 20 min. After H2O2 treatment (0.3% for 20 min) and protein blocked for 1 h, anti-Ki67 was added and incubated overnight at 4°C. Anti-rabbit-FITC (Jackson ImmunoResearch Laboratories, West Grove, PA) and anti-FITC-horseradish peroxidase (Roche Molecular Biochemicals, Mannheim, Germany) were added sequentially, and color was developed by 3,3′-diaminobenzidine (Zymed, San Francisco, CA) followed by hematoxylin (Sigma) counterstaining.

Binding of Fluorescently Labeled mAb in Vivo.

Lung, kidney, heart, liver, spleen, brain, and tumor tissue was examined for the extent of mAb binding to vasculature. For this purpose, antibodies BV13, E4G10, and control rat Ig (Jackson ImmunoResearch Laboratories) were labeled using Alexa-546 protein labeling kit (Molecular Probes, Eugene, OR) following the manufacturer’s instructions. Tumor-bearing animals were injected i.v. with a single dose of fluorescently labeled mAb (50–300 μg for BV13 and 0.3–1 mg of E4G10). Two h later, animals were perfused with 50 ml of saline via the left ventricle of the heart, and tissues were excised and fixed using Zinc Formal-Fixx (Shandon-Lipshaw, Pittsburgh, PA) and frozen. Frozen sections (6 μm) were stained for CD31 using FITC-anti CD31 (BD PharMingen), followed by Alexa-488-conjugated anti-FITC antibody (Molecular Probes).

BrdUrd Labeling.

The measurement of cell proliferation was performed using a BrdUrd labeling and Detection Kit I (Roche Molecular Biochemicals). Briefly, BrdUrd was injected i.v. (100 mg in 100 μl of PBS/mouse) into tumor-bearing mice (22). One h later, fluorescently labeled mAb (1 mg of E4G10/mouse) was injected i.v. into those mice and allowed to circulate for 2 h. After vascular perfusion, tumors were removed, snap-frozen, and sectioned (6 μm) for subsequent detection of BrdUrd labeling, following the instructions provided by the manufacturers.

Antibody E4G10 Inhibits the Formation of Adherens Junctions But Does Not Disrupt Existing Junctions.

Antibodies against murine VE-cad were generated by immunization of rats with a soluble recombinant protein (VE-cad-Ig) or peptides comprising regions of the NH2-terminal domain of murine VE-cad. Several hundred hybridoma clones were identified that secreted mAb and bound specifically to the extracellular region of mouse VE-cad in ELISA. Antibodies were then screened in the calcium switch assay for their ability to inhibit junction formation, and positive clones were further screened in an in vitro permeability assay for absence of permeability-inducing properties. One mAb (E4G10) generated against peptide 1 was selected as having the best respective in vitro properties.

E4G10 activity in the calcium-switch assay is shown in Fig. 1. The adherens junctions were assessed by immunofluorescent staining of clustered VE-cad molecules expressed on the mouse endothelial cells (H5V; Fig. 1,a). The adherens junctions were disrupted by EGTA treatment, leading to a loss of VE-cad molecules from the junctions (Fig. 1,b), which were restored by addition of calcium in the culture medium (Fig. 1,c). In the presence of E4G10, the reformation/reorganization of junctions after calcium switch was significantly inhibited, as evidenced by a lack of clearly identifiable VE-cad staining at cell-cell junctions. Rather, VE-cad molecules were diffusely distributed on the cell surface (Fig. 1,d). At a concentration of 10 μg/ml, E4G10 reduced the junctional VE-cad distribution to a similar extent as BV13 (80% inhibition), as quantified by image analysis (Fig. 1 e). When a wide range of antibody concentrations (0.1–100 μg/ml) was tested in the calcium switch assays, E4G10 exhibited a 2-fold lower potency than BV13 on inhibiting adherens junction formation (IC50, 2 and 1 μg/ml, respectively).

To compare the activity of E4G10 and BV13 on existing adherens junctions, antibodies were added directly to endothelial cell monolayers before cells were fixed. E4G10 did not bind to adherens junctions on intact cell monolayers even at concentrations up to 50 μg/ml (Fig. 1,f), whereas BV13 bound to intact junctions of cell monolayers in a dose-dependent manner (left panel). However, E4G10 and BV13 bound to VE-cad on cell monolayers to a similar extent when adherens junctions were disrupted by addition of EGTA to the culture (right panel). The differential binding activities of E4G10 and BV13 on cultured endothelial cell monolayers was further analyzed in a transwell system by assessing the changes in permeability of the treated monolayers. E4G10 had no detectable effect on monolayer permeability at concentrations up to 50 μg/ml (Fig. 1 g). In contrast, BV13 disrupted adherens junctions in a dose-dependent manner and increased monolayer permeability 3-fold at a concentration of 10 μg/ml. These results reveal distinct binding and disruptive activity of E4G10 and BV13 on established adherens junctions in culture, although these mAbs display comparable binding affinities to VE-cad molecules with Kds of 8.9 + 1.9 nm and 15.5 + 2.5 nm, respectively.

Antibody E4G10 Does Not Increase Vascular Permeability in Vivo.

Previous studies have shown that mAb BV13 administered systemically to mice at doses ≥100 μg increases vascular permeability in lung tissue and to a lesser extent the heart, resulting in pulmonary edema followed by death of the animals within 24–48 h (7, 8). We used the Evans blue assay to test whether mAb E4G10 had a similar effect on vascular permeability in mice. Single injection of E4G10 ranging from 25–3000 μg had a negligible effect on lung vascular permeability (Fig. 2,a, right panel). No mortality or signs of toxicity were observed in various strains of normal mice administered E4G10, even after repeated dosing (1 mg/dose, every 3 days for 6 weeks). In contrast, BV13 at doses of 50 μg and higher significantly increased vascular permeability in the lung (>2-fold; Fig. 2,a, left panel) and the heart (1.3-fold, data not shown). The BV13-induced increase in vascular permeability was dose and time dependent, with a maximal effect seen at 100 μg/dose 24 h after mAb injection. Histological examination of lung tissue from mice treated with mAb E4G10 and BV13 revealed distinct pathological findings. E4G10-treated lungs showed a healthy morphology, even after a dose of 1 mg (Fig. 2,b, top, right panel), in which vessels of all sizes were intact and lined by normally shaped endothelial cells. No abnormality was detected in any of the cell types comprising the alveolar-capillary complex. In contrast, a lethal dose of BV13 (100 μg) induced significant vascular leakage with an onset as early as 2 h after antibody administration (data not shown). Maximal pathology with massive hemorrhage was observed 24 h after antibody administration (Fig. 2,b, bottom, left panel). Morphological changes at this time point included bleb formation of endothelial and alveolar epithelial cells, aggregation of degranulated platelets and the formation of platelet microthrombi, and leukocyte activation and infiltration (Ref. 7 and data not shown). At a lower dose of 50 μg, BV13 induced pathological changes in the lung, as evidenced by small areas of hemorrhage from capillaries (Fig. 2 b, bottom, right panel). Altogether, these results show a correlation between antibody effects on vascular permeability and lung pathology, suggesting that vascular leakage is a triggering event for the pathological consequences described above.

E4G10 Inhibits Angiogenesis.

To examine the effects of E4G10 on angiogenesis in vivo, the mAb was tested in the Matrigel plug and mouse corneal micropocket assays. Mice with implanted Matrigel plugs were treated every 3 days with various doses of E4G10 (0.2, 0.5, and 1 mg/mouse/t.p.) for a total of 21 days. E4G10 inhibited neovascularization induced in Matrigel plugs in a dose-dependent manner (Fig. 3, a and b). The effect seen at a dose of 1 mg was comparable with that of BV13 at 50 μg. In the mouse corneal micropocket assay, systemically administered E4G10 (1 mg/dose) inhibited bFGF-induced neovascularization of the cornea by 55% and was comparable in efficacy to that of antibody BV13 (50 μg/dose; 65% inhibition; Fig. 3 c).

E4G10 Inhibits Tumor Growth.

The effect of E4G10 on tumor growth was examined in three xenograft models of the human A431 epidermoid (8), the mouse H5V hemangioma (21), and the C6-V10 rat glioblastoma (13). Preliminary dose-response studies performed in the A431 model indicated a maximal therapeutic effect at a dose of 1 mg of E4G10 administered twice weekly. E4G10 treatment inhibited tumor growth by 70, 65, and 55% in the A431, H5V, and C6-V10 models, respectively. Similar antitumor activity was observed in mice treated with the maximum tolerated dose of 50 μg of BV13 twice weekly (Fig. 4,a). Histological examination of A431 tumors treated with E4G10 or BV13 showed a similar decrease in microvessel density, increased tumor and endothelial cell apoptosis, decreased tumor cell proliferation, and extensive tumor necrosis, indicating a similar mode of action by these two mAbs (Fig. 4 b). All effects were observed as early as 2 weeks after treatment began and gradually increased as mAb therapy continued.

Antibodies E4G10 and BV13 Differentially Bind to Normal and Tumor Vasculature.

To investigate the mechanism underlying the distinct effects of E4G10 and BV13 on vascular permeability in mouse tissue, we examined the ability of these mAbs to bind to normal and tumor vasculature. Various amounts of fluorescently labeled E4G10 or BV13 were injected i.v. into tumor-bearing mice and allowed to circulate and bind to endothelium for 2 h. Normal tissue (lung, kidney, heart, liver, spleen, and brain) as well as tumor tissue was harvested after saline perfusion and processed for fluorescence microscopy. Analysis of tissues obtained from mice injected with up to 1 mg of E4G10 showed undetectable binding to normal endothelium in all tissues examined, consistent with the negligible effect of this mAb on vascular permeability in mice (Fig. 5,a, bottom panels). In contrast, a single injection of 50–300 μg of labeled BV13 showed a dose-dependent increase in binding of this mAb to a broad spectrum of endothelium in normal tissues, with a saturating concentration of 200 μg. Among the six normal tissues examined, lung and heart retained the largest amount of the disruptive antibody BV13 (Fig. 5 a, top panels). Lower level of BV13 binding was observed in the vasculature of the kidney and to a lesser extent in the liver, spleen, and brain (data not shown).

The binding of E4G10 and BV13 was also examined in tumor vasculature of several tumor types. E4G10 bound to a subset (20–40%) of the tumor vasculature, regardless of the tumor type (Fig. 5,b). E4G10 binding was preferentially located at the periphery of tumors. Examination of BrdUrd-labeled cells in tumors showed that a significant fraction of the E4G10-bound endothelium was undergoing active proliferation (Fig. 5 c). In contrast, BV13 bound to the majority of tumor vasculature in all tumors (>90% of CD31-positive endothelium), and there was no correlation with BrdUrd-labeled endothelium (data not shown).

Neovascularization requires the assembly of endothelial cells into vascular tubes. VE-cad plays a critical role in this process by mediating the formation of adherens junctions between endothelial cells. Targeting the process of vascular tube formation is an attractive approach to inhibit pathological angiogenesis because of the specificity of VE-cad expression in vasculature. Our previous work demonstrated that functional inhibition of adherens junctions using the VE-cad mAb BV13 could potently inhibit angiogenesis and tumor growth (8). However, treatment of mice with BV13 at elevated doses also resulted in disruption of adherens junctions in some normal tissue vasculature, most notably that of the lungs, leading to an increase in vascular permeability and pulmonary edema (7, 8). Hence, a more selective targeting/blockade of VE-cad function on tumor vasculature is an essential requirement for consideration of VE-cad as a therapeutic target.

Work described in this report was undertaken to explore possible avenues to overcome antibody toxicity attributable to nonselective interference with adherens junctions in normal vasculature. We argued that mAbs might be identified that: (a) recognize epitopes in certain regions of the VE-cad extracellular domain that are exposed only in endothelium undergoing formation of new adherens junctions (i.e., during angiogenesis); and (b) these epitopes are not accessible to mAb in established adherens junctions of normal vasculature. We based this hypothesis on the distinct cellular localization and structural organization of VE-cad molecules in different functional states of endothelium as well as the complexity of VE-cad-mediated adhesion (10, 11, 12). In support of our hypothesis, we characterized a large panel of mAbs and identified one mAb, E4G10, which distinguishes itself in that it exhibits antiangiogenic and antitumor activities but has a negligible effect on normal vascular permeability.

In this context, we carefully studied the effects of E4G10 on normal vasculature and specifically looked for evidence of toxicity. The data show that E4G10 was well tolerated when frequently administered to mice at high doses (>1 mg/every 3 days) for periods up to 6 weeks. Mice treated in this manner showed no signs of antibody toxicity. In contrast, mice treated with the toxic mAb BV13 succumbed rapidly to vascular leak in the lungs. Furthermore, we investigated whether any abnormal, subtoxic permeability in susceptible organs, especially the lung, was associated with E4G10 therapy. We did not observe any degree of abnormal permeability in the lung or other tissues, i.e., the heart and the kidney, after administration of large amounts of E4G10 (10 mg/dose). Again, the toxic mAb BV13 acutely and strongly increased vascular permeability in these organs. Finally, we demonstrated by an in vivo labeling method that E4G10 only bound to endothelial cells in tumors but not to normal vasculature in a variety of organs, whereas BV13 bound to tumor and normal vasculature to a similar extent. We are therefore confident that E4G10 represents a prototypic tumor vasculature-selective mAb with antiangiogenic and antitumor activity but does not have vascular permeability-inducing properties.

A most interesting aspect of our work is the finding that mAb E4G10 binds only to a subset (20–40%) of tumor vasculature. To our knowledge, this finding provides the first example that a constitutively expressed molecule can be selectively targeted on tumors. We hypothesize that the endothelium labeled by E4G10 represents angiogenic endothelium, i.e., containing the actively proliferating endothelial cells that are in the process of forming new vascular tubes. In support of this hypothesis, we have found that E4G10 binding is restricted to the highly proliferative, peripheral areas of tumors. The periphery of growing tumors is well recognized as the major site of endothelial cell proliferation and ongoing angiogenesis, as well as tumor cell proliferation (see Ki67 staining in Fig. 4 b and Refs. 23, 24). Furthermore, our data suggest that there may be a correlation between sites of VE-cad binding and mitotic endothelial cells, as evidenced by BrdUrd staining.

We propose that E4G10 recognizes an epitope on VE-cad that is accessible in neovasculature and becomes inaccessible once adherens junctions form in resting vasculature, whereas BV13 recognizes a “constitutive” epitope that is accessible in all types of vasculature (Fig. 6). Supporting this hypothesis are our observations that mAbs BV13 and E4G10 bind differentially to VE-cad expressed on cultured endothelial cell monolayers (see Fig. 1,f) and on normal and tumor vasculature in vivo (see Fig. 5); yet, both mAbs bind to VE-cad in solution with similar affinities. Therefore, we believe that it is the accessibility of mAb epitopes rather than the mAb binding affinities that determines the functional distinctions of these two mAb. A number of integrins serve as precedents for such functional epitopes, i.e., specific mAbs that recognize epitopes corresponding to different functional states of integrins have been reported (25, 26, 27).

Antibody E4G10 was generated against a polypeptide sequence (peptide 1) corresponding to the NH2-terminal region of VE-cad. BV13 recognizes a distinct epitope also located within the NH2-terminal domain of VE-cad.3 These findings indicate that the NH2-terminal domain of VE-cad can elicit functionally distinct mAbs. Although we were unable to generate another E4G10-like mAb against other peptide sequences or the entire extracellular region of VE-cad, we cannot exclude the possibility that mAbs raised against different epitopes of VE-cad may have characteristics similar to E4G10. Indeed, we have shown previously that mAbs targeting the NH2-terminal domain and the third domain of human VE-cad can have similar blocking effects in vitro(15). Therefore, it is conceivable that mAbs against various regions of VE-cad may exhibit properties similar to E4G10. Additional studies are required to identify the precise binding sites of various mAb candidates with distinct antitumor and permeability properties and to correlate epitope location with functional activity.

It should be noted that a tumor-inhibiting dose of antibody E4G10 is 20-fold higher than that of antibody BV13 (1 mg versus 50 μg, respectively). The basis for the difference in required doses is unknown. As described above, the distinct activities between these two mAbs are not attributable to a difference of their binding affinities. One possible explanation is that the densities of the epitopes for these mAbs determine the necessary dosages (28). It is at least theoretically possible that the lower doses required for BV13 are attributable to its disruptive targeting of the entire tumor vasculature, in which virtually every VE-cad molecule moves to the cell surface in the accessible configuration. Thus, a low dose of BV13 can more extensively affect the entire tumor blood supply. In contrast, E4G10 targets a subset of tumor vasculature, i.e., presumably the dynamic, angiogenic subset of vessels. Treatment with E4G10 requires a sustained level of antibody over a longer period of time to achieve the same, cumulative therapeutic effect as BV13. However, it is important to note that E4G10 has a much broader therapeutic window than BV13 because of its negligible effect on vascular permeability in normal tissues, even at high doses.

In summary, our data provide proof-of-concept that specific targeting of VE-cad on tumor vasculature is possible, despite ubiquitous VE-cad expression in normal tissue vasculature. This finding may have therapeutic implications; however, additional studies will need to address additional questions regarding appropriate epitopes, safety, and efficacy. If these findings are confirmed and structure/function relationship defined, the notion that an anti-VE-cad antibody can exclusively target active, angiogenic endothelium is novel and has possible far-reaching consequences for the development of highly specific, antiangiogenic therapeutic agents. Therapeutic efforts to design potent and specific antiangiogenic agents for the therapy of tumors and other angiogenic diseases are increasingly focusing on the identification of specific targets that occur only on tumor vasculature. Examples of such targets are αvβ3 integrin (29), endoglin (30), and most prominently, the vascular endothelial growth factor receptor-2 (flk1/KDR; Refs. 31, 32). However, even these highly specific markers of angiogenic endothelium are not uniquely present on tumor endothelial cells but are also expressed to various degrees on normal vasculature, and thus potentially limit the therapeutic usefulness of such targets (33). Antagonists such as E4G10 may overcome the limitations of these less specific targets by inhibiting uniquely the angiogenic vasculature. The ability to specifically target tumor vasculature with antibodies such as E4G10 may also open up the possibility of using immunoconjugates to target angiogenic endothelium with cytotoxic payload, a feat that is difficult to achieve safely with currently available targets and targeting agents. Therefore, the advent of an epitope that is restricted to angiogenic endothelium is of considerable interest for the development of more specific antiangiogenic therapies.

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

2

The abbreviations used are: VE-cad, vascular endothelial cell-cadherin; mAb, monoclonal antibody; bFGF, basic fibroblast growth factor; BrdUrd, 5-bromo-2′-deoxyuridine.

3

Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org).

Fig. 1.

Effects of BV13 and E4G10 on adherens junctions. The calcium-switch assay was performed on mouse endothelial cell (H5V) monolayers. Adherens junctions were assessed by immunofluorescent staining of clustered VE-cad molecules. a, in confluent H5V monolayers, VE-cad molecules are clustered in adherens junctions. b, VE-cad molecules disappear from the junctions upon EGTA treatment. Blue spots are nuclei stained by 4,5-diamidino-2-phenylindole. c, junctions were restored by adding calcium into the culture medium, as evidenced by VE-cad reappearance at the cell-cell contacts. d and e, addition of antibodies E4G10 (d) or BV13 (e) inhibits adherens junction reformation at a concentration of 10 μg/ml. f, binding of mAb E4G10 and BV13 to VE-cad molecules expressed on cultured H5V cell monolayers with and without EGTA treatment. Antibody E4G10 only binds to VE-cad molecules after junctions are artificially disrupted, whereas BV13 can bind to VE-cad molecules in junctions of intact monolayers. g, quantification of the disruptive effects of E4G10 and BV13 on existing monolayers in the transwell system.

Fig. 1.

Effects of BV13 and E4G10 on adherens junctions. The calcium-switch assay was performed on mouse endothelial cell (H5V) monolayers. Adherens junctions were assessed by immunofluorescent staining of clustered VE-cad molecules. a, in confluent H5V monolayers, VE-cad molecules are clustered in adherens junctions. b, VE-cad molecules disappear from the junctions upon EGTA treatment. Blue spots are nuclei stained by 4,5-diamidino-2-phenylindole. c, junctions were restored by adding calcium into the culture medium, as evidenced by VE-cad reappearance at the cell-cell contacts. d and e, addition of antibodies E4G10 (d) or BV13 (e) inhibits adherens junction reformation at a concentration of 10 μg/ml. f, binding of mAb E4G10 and BV13 to VE-cad molecules expressed on cultured H5V cell monolayers with and without EGTA treatment. Antibody E4G10 only binds to VE-cad molecules after junctions are artificially disrupted, whereas BV13 can bind to VE-cad molecules in junctions of intact monolayers. g, quantification of the disruptive effects of E4G10 and BV13 on existing monolayers in the transwell system.

Close modal
Fig. 2.

Effects of BV13 and E4G10 on vascular permeability in lung tissue. a, effects of various doses of antibodies BV13 and E4G10 on vascular permeability in lung tissue. Data are presented as means (n = 5); bars, SE. b, different pathological effects of BV13 and E4G10 on lungs treated with either antibody for 24 h, as examined by H&E staining. Boxed regions (BV13, 50 μg) indicate areas of hemorrhage. Representative pictures from each experimental group (n = 3/group) are presented.

Fig. 2.

Effects of BV13 and E4G10 on vascular permeability in lung tissue. a, effects of various doses of antibodies BV13 and E4G10 on vascular permeability in lung tissue. Data are presented as means (n = 5); bars, SE. b, different pathological effects of BV13 and E4G10 on lungs treated with either antibody for 24 h, as examined by H&E staining. Boxed regions (BV13, 50 μg) indicate areas of hemorrhage. Representative pictures from each experimental group (n = 3/group) are presented.

Close modal
Fig. 3.

Antiangiogenic effect of antibodies E4G10 and BV13. a, immunohistochemical staining of the Matrigel plug sections with anti-CD31 mAb (×100) shows a mark reduction of CD31-positive vessels in E4G10-treated plug sections. b, quantification of neovascularization in the Matrigel plugs indicates an inhibitory of E4G10 on microvessel density in a dose-dependent manner. Data are presented as means (n = 10); bars, SE. c, inhibition of angiogenesis in mouse corneas. Eyes were photographed 6 days after implantation of hydron pellets containing 50 ng of bFGF and three systemic treatments with antibodies at days 1, 3, and 5. Representative eyes from three independent experiments (n = 6/each experiment) are shown.

Fig. 3.

Antiangiogenic effect of antibodies E4G10 and BV13. a, immunohistochemical staining of the Matrigel plug sections with anti-CD31 mAb (×100) shows a mark reduction of CD31-positive vessels in E4G10-treated plug sections. b, quantification of neovascularization in the Matrigel plugs indicates an inhibitory of E4G10 on microvessel density in a dose-dependent manner. Data are presented as means (n = 10); bars, SE. c, inhibition of angiogenesis in mouse corneas. Eyes were photographed 6 days after implantation of hydron pellets containing 50 ng of bFGF and three systemic treatments with antibodies at days 1, 3, and 5. Representative eyes from three independent experiments (n = 6/each experiment) are shown.

Close modal
Fig. 4.

Antitumor effects of E4G10 and BV13 in vivo. a, inhibitory effects of mAb on the growth of human A431 epidermoid, H5V hemangioma, and C6-V10 glioblastoma tumors. Data are presented as means (n = 10); bars, SE. b, histology of A431 tumor sections from animals treated for 2 weeks. Top panel, H&E staining of paraffin-embedded sections (×40). Middle panels, cell proliferation by Ki67 staining and CD31-expressing vessels (×100). Bottom panel, terminal deoxynucleotidyl transferase-mediated nick end labeling/CD31 double staining on frozen tumor sections (×400).

Fig. 4.

Antitumor effects of E4G10 and BV13 in vivo. a, inhibitory effects of mAb on the growth of human A431 epidermoid, H5V hemangioma, and C6-V10 glioblastoma tumors. Data are presented as means (n = 10); bars, SE. b, histology of A431 tumor sections from animals treated for 2 weeks. Top panel, H&E staining of paraffin-embedded sections (×40). Middle panels, cell proliferation by Ki67 staining and CD31-expressing vessels (×100). Bottom panel, terminal deoxynucleotidyl transferase-mediated nick end labeling/CD31 double staining on frozen tumor sections (×400).

Close modal
Fig. 5.

Binding of E4G10 and BV13 on normal mouse tissue and established tumors. a, binding of fluorescently labeled antibodies (200 μg of BV13, 1 mg of E4G10 or control rat Ig) in vasculature of lung, heart, and kidney tissue. Micrographs derive from the overlaying images of the CD31-stained vessels (green) and the antibody-labeled tissues (red; ×400). b, binding of E4G10 and BV13 to various tumor tissues. c, proliferating cells (green) in tumor tissue, as examined by BrdUrd labeling. E4G10 (red) localizes to a subset of tumor endothelial cells that are also actively proliferating. The majority of the proliferating cells are the tumor cells adjacent to vessels.

Fig. 5.

Binding of E4G10 and BV13 on normal mouse tissue and established tumors. a, binding of fluorescently labeled antibodies (200 μg of BV13, 1 mg of E4G10 or control rat Ig) in vasculature of lung, heart, and kidney tissue. Micrographs derive from the overlaying images of the CD31-stained vessels (green) and the antibody-labeled tissues (red; ×400). b, binding of E4G10 and BV13 to various tumor tissues. c, proliferating cells (green) in tumor tissue, as examined by BrdUrd labeling. E4G10 (red) localizes to a subset of tumor endothelial cells that are also actively proliferating. The majority of the proliferating cells are the tumor cells adjacent to vessels.

Close modal
Fig. 6.

Model for selective binding of E4G10 on tumor vasculature. In the neovasculature, VE-cad-mediated adherens junctions are loosely organized, in which most regions of the VE-cad molecules are exposed, giving rise to accessible epitopes, such as the one for E4G10. In established, quiescent vasculature adherens junctions are well organized, in which VE-cad molecules are tightly engaged, and certain regions of the molecules may be embedded and become inaccessible, such as the epitope for E4G10. The epitope for BV13 does not distinguish between VE-cad molecules engaged in adherens junctions or loosely organized in neovasculature.

Fig. 6.

Model for selective binding of E4G10 on tumor vasculature. In the neovasculature, VE-cad-mediated adherens junctions are loosely organized, in which most regions of the VE-cad molecules are exposed, giving rise to accessible epitopes, such as the one for E4G10. In established, quiescent vasculature adherens junctions are well organized, in which VE-cad molecules are tightly engaged, and certain regions of the molecules may be embedded and become inaccessible, such as the epitope for E4G10. The epitope for BV13 does not distinguish between VE-cad molecules engaged in adherens junctions or loosely organized in neovasculature.

Close modal

We thank Drs. L. E. Benjamin and A. Vecchi for generous gifts of the C6-V10 cell line and the H5V and 1G11 cell lines, respectively. We are grateful to Dr. William Betzer for critical reading of the manuscript and providing valuable discussion.

1
Breviario F., Caveda L., Corada M., Martin-Padura L., Navarro P., Golay J., Introna M., Gulino D., Lampugnani M-G., Dejana E. Functional properties of human vascular endothelial cadherin (7B4/cadherin-5), an endothelium-specific cadherin.
Artherioscler. Throm. Vasc. Biol.
,
15
:
1229
-1239,  
1992
.
2
Lampugnani M. G. The molecular organization of endothelial cell to cell junctions, differential association of plakoglobin, β-catenin, and α-catenin with vascular endothelial cadherin (VE-cadherin).
J. Cell Biol.
,
129
:
203
-217,  
1995
.
3
Bazzoni G., Dejana E., Lampugnani M. G. Endothelial adhesion molecules in the development of the vascular tree: the garden of forking paths.
Curr. Opin. Cell Biol.
,
11
:
573
-581,  
1999
.
4
Carmeliet P., Lampugnani M. G., Moons L., Breviario F., Compernolle V., Bono F., Balconi G., Spagnuolo R., Oosthuyse B., Dewerchin M., Zanetti A., Angellilo A., Mattot V., Nuyens D., Lutgens E., Clotman F., Ruiter M. A. D., Groot A. G., Poelmann R., Lupu F., Herbert J. M., Collen D., Dejana E. Targeted deficiency or cytosolic truncation of the VE-cadherin gene in mice impairs VEGF-mediated endothelial survival and angiogenesis.
Cell
,
98
:
147
-157,  
1999
.
5
Caveda L., Martin-Padura I., Navarro P., Breviario F., Corada M., Gulino D., Lampugnani M. G., Dejana E. Inhibition of cultured cell growth by vascular endothelial cadherin (cadherin-5/VE-cadherin).
J. Clin. Investig.
,
98
:
886
-893,  
1996
.
6
Dejana E. Endothelial adherens junctions: implications in the control of vascular permeability and angiogenesis.
J. Clin. Investig.
,
100
:
S7
-S10,  
1997
.
7
Corada M., Mariotti M., Thurston G., Smith K., Kunkel R., Brockhaus M., Lampugnani M. G., Martin-Padura I., Stoppacciaro A., Ruco L., Dejana E. Vascular endothelial-cadherin is an important determinant of microvascular integrity in vivo.
Proc. Natl. Acad. Sci. USA
,
96
:
9815
-9820,  
1999
.
8
Liao F., Li Y., O’Connor W., Zanetta L., Bassi R., Santiago A., Overholser J., Hooper A., Mignatti P., Dejana E., Hicklin D. J., Bohlen P. Monoclonal antibody to VE-cadherin is a potent inhibitor to angiogenesis, tumor growth, and metastasis.
Cancer Res.
,
60
:
6805
-6812,  
2000
.
9
Bach T. L., Barsigian C., Chalupowicz D. G., Busler D., Yaen C. H., Grant D. S., Martiney J. VE-Cadherin mediates endothelial cell capillary tube formation in fibrin and collagen gels.
Exp. Cell Res.
,
238
:
324
-334,  
1998
.
10
Sivasankar S., Brieher W., Lavrik N., Gumbiner B. M., Lecjband D. Direct molecular force measurements of multiple adhesive interactions between cadherin ectodomains.
Proc. Natl. Acad. Sci. USA
,
96
:
11820
-11824,  
1999
.
11
Leckband D., Sivasankar S. Mechanism of homophilic cadherin adhesion.
Curr. Opin. Cell Biol.
,
12
:
587
-592,  
2000
.
12
Chappuis-Flament S., Wong E., Hicks L. D., Kay C. M., Gumbiner B. M. Multiple cadherin extracellular repeats mediate homophilic binding and adhesion.
J. Cell. Biol.
,
154
:
231
-243,  
2001
.
13
Benjamin L. E., Keshet E. Conditional switching of vascular endothelial growth factor (VEGF) expression in tumors: induction of endothelial cell shedding and regression of hemangioblastoma-like vessels by VEGF withdrawal.
Proc. Natl. Acad. Sci. USA
,
94
:
8761
-8766,  
1997
.
14
Dong Q. G., Bernasconi S., Lostaglio S., Decalmanovici R. W., Martin-Padura I., Breviario F., Garlanda C., Ramponi S., Mantovari A., Vecchi A. A general strategy for isolation of endothelial cells from murine tissues. Characterization of two endothelial cell lines from the murine lung and subcutaneous sponge implants.
Arterioscler. Thromb. Vasc. Biol.
,
17
:
1599
-1604,  
1997
.
15
Corada M., Liao F., Lindgren M., Lampugnani M-G., Breviario F., Frank R., Muller W. A., Hicklin D. J., Bohlen P., Dejana E. Monoclonals directed to different regions of VE-cadherin extracellular domain may affect adhesion and clustering of the protein and modulate endothelial permeability.
Blood
,
97
:
1679
-1684,  
2001
.
16
Shapiro L., Fannon A. M., Kwong P. D., Thompson A., Lehmann M. S., Grubel G., Legrand J. F., Als-Nielson J., Colman D. R., Hendrickson W. A. Structural basis of cell-cell adhesion by cadherins.
Nature (Lond.)
,
374
:
327
-337,  
1995
.
17
Nagar B., Overduin M., Ikura M., Rini J. M. Structural basis of calcium-induced E-cadherin rigidification and dimerization.
Nature (Lond.)
,
380
:
360
-364,  
1996
.
18
Coligan J. E., Kruisbeek A. M., Margulies D. H., Shevach E. M., Strober W. Iodination of soluble and membrane-bound proteins .
Current Protein Immunology
,
Vol. 2
:
8.11.1
-8.11.4,  
1998
.
19
Scatchard G. The attractions of proteins for small molecules and ions.
Ann. NY Acad. Sci.
,
51
:
660
-672,  
1949
.
20
Coligan J. E., Kruisbeek A. M., Margulies D. H., Shevach E. M., Strober W. Calculation of IL-2 binding affinity and receptor number by Scatchard plot analysis .
Current Protein Immunology
,
Vol. 1 (Supplement 23)
:
6.1.5
-6.1.7,  
1998
.
21
Garlanda C., Prravicini C., Sironi M., De Rosi M., Wainstok D., Calmanovici R., Carrozzi F., Bussolino F., Colotta F., Mantovani A., Vecchi A. Progressive growth in immunodeficient mice and host cell polyoma middle-sized T antigen: implications for the pathogenesis of opportunistic vascular tumors.
Proc. Natl. Acad. Sci. USA
,
91
:
7291
-7295,  
1994
.
22
Ezaki T., Baluk P., Thurston G., Barbara A. L., Woo C., McDonald D. M. Time course of endothelial cell proliferation and microvascular remodeling in chronic inflammation.
Am. J. Pathol.
,
158
:
2043
-2055,  
2001
.
23
Thompson W. D., Shiach K. J., Fraser R. A., McIntosh L. C., Simpson J. G. Tumours acquire their vasculature by vessel incorporation, not vessel ingrowth.
J. Pathol.
,
151
:
323
-332,  
1987
.
24
Shweiki D., Itin A., Soffer D., Keshet E. Vascular endothelial growth factor induced by hypoxia may mediate hypoxia-induced angiogenesis.
Nature (Lond.)
,
359
:
843
-845,  
1992
.
25
Humphries M. J. Integrin activation: the link between ligand binding and signal transduction.
Curr. Opin. Cell Biol.
,
8
:
632
-640,  
1996
.
26
Koyama N., Seki J., Vergel S., Mattsson E. J., Yednock T., Kovach N. L., Harlan J. M., Clowes A. W. Regulation and function of an activation-dependent epitope of the β1 integrins in vascular cells after balloon injury in baboon arteries and in vitro.
Am. J. Pathol.
,
148
:
49
-761,  
1996
.
27
Felding-Habermann B., O’Toole T. E., Smith J. W., Fransvera E., Rugger Z. M., Ginsberg M. N., Hughes P. E., Pampori N., Shattil S. J., Saven A., Mueller B. M. Integrin activation controls metastasis in human breast cancer.
Proc. Natl. Acad. Sci. USA
,
98
:
1853
-1858,  
2001
.
28
Zuckier L. S., Berkowitz E. Z., Sattenberg R. J., Zhao Q. H., Deng H. F., Scharff M. D. Influence of affinity and antigen density on antibody localization in a modifiable tumor targeting model.
Cancer Res.
,
60
:
7008
-7013,  
2000
.
29
Brooks P., Clark R. A., Cheresh D. A. Requirement of vascular integrin αvβ3 for angiogenesis.
Science (Wash. DC)
,
264
:
569
-571,  
1994
.
30
Miller D. W., Graulich W., Karges B., Stahl S., Ernst M., Ramaswamy A., Sedlacek H. H., Muller R., Adamkiewicz J. Elevated expression of endoglin, a component of the TGF-β-receptor complex, correlates with proliferation of tumor endothelial cells.
Int. J. Cancer.
,
81
:
568
-572,  
1999
.
31
Prewett M., Huber J., Li Y., Santiago A., O’Connor W., King K., Overholser J., Hooper a., Pytowski B., Wite L., Bohlen P., Hicklin D. J. Antivascular endothelial growth factor receptor (fetal liver kinase 1) monoclonal antibody inhibits angiogenesis and growth of several mouse and human tumors.
Cancer Res.
,
59
:
5209
-5218,  
1999
.
32
Skobe M., Rockwell P., Goldstein N., Vosseler S., Fusenig N. E. Halting angiogenesis suppresses carcinoma cell invasion.
Nat. Med.
,
3
:
1222
-1227,  
1997
.
33
Hicklin D. J., Witte L., Zhu Z., Liao F., Wu Y., Li Y., Bohlen P. Monoclonal antibody strategies to block angiogenesis.
Drug Discovery Today
,
6
:
517
-528,  
2001
.

Supplementary data