Evaluation of Novel Antimouse VEGFR2 Antibodies as Potential Antiangiogenic or Vascular Targeting Agents for Tumor Therapy

Abstract

We generated a panel of eight rat IgG_2a monoclonal antibodies with high affinity for mouse VEGFR2 (KDR/Flk-1), the main receptor that mediates the angiogenic effect of VEGF-A. The antibodies (termed RAFL, Rat Anti Flk) bound to dividing endothelial cells more strongly than they did to nondividing cells. Most of the RAFL antibodies blocked [¹²⁵I]VEGF₁₆₅ binding to VEGFR2. Three of eight antibodies localized to VEGFR2-positive tumor endothelium after intravenous injection into mice bearing orthotopic MDA-MB-231 breast carcinomas, as judged by indirect immunohistochemistry. An average of 60% of vessels in the tumors was stained. The majority (50–80%) of vessels were also stained in a variety of other human and murine tumors growing in mice. The antibodies did not bind detectably to the vascular endothelium in normal heart, lung, liver, and brain cortex, whereas the vascular endothelium in kidney glomerulus and pancreatic islets was stained. Treatment of mice bearing orthotopic MDA-MB-231 tumors with RAFL-1 antibody inhibited tumor growth by an average of 48% and reduced vascular density by 65%, compared to tumors in mice treated with control IgG. Vascular damage was not observed in normal organs, including kidneys and pancreas. These studies demonstrate that anti-VEGFR2 antibodies have potential for vascular targeting and imaging of tumors in vivo.

Introduction

Angiogenesis, the growth of new blood vessels from existing vessels [1], is of crucial importance for the growth, maintenance, and metastasis of solid tumors [2,3]. Vascular endothelial growth factor (VEGF-A) is one of the key angiogenic factors that stimulates vascularization of normal and neoplastic tissues. Evidence for the dominant role of VEGF in the development of tumor angiogenesis includes its high expression in a broad spectrum of malignancies [4,5]; induction of mitogenic, chemotactic, and antiapoptotic effects on cultured endothelial cells [6–8]; induction of vascular leakage [5,9]; and the ability to stimulate the full cascade of events required for angiogenesis in vivo [10,11].

VEGF-A exerts its effects on the vascular endothelium through binding to two high-affinity receptors, R1 (FLT-1/Flt-1) and R2 (KDR and Flk-1 for the human and mouse receptors, respectively). Several lines of evidence indicate that VEGFR2 plays the major role in transducing the angiogenic effect of VEGF on tumor vasculature. The expression of VEGFR2 is mainly restricted to vascular endothelial cells and hematopoietic cells [12,13]. Upon addition of VEGF to mouse or human endothelial cells that have both receptors, tyrosine phosphorylation is primarily detected in VEGFR2 [14]. Neither phosphorylation nor mitotic, migratory, and morphological changes associated with VEGF are apparent in endothelial cells that have been engineered to express VEGFR1 alone [14,15]. Hypoxia and genetic alterations in tumor cells continuously drive overexpression of both receptors on tumor vessels in vivo, with VEGFR2 being consistently found at higher levels than VEGFR1 [16]. Studies in our laboratory showed that the anti-VEGF antibody, 2C3, which blocks VEGF binding to VEGFR2 but not to VEGFR1, inhibits tumor growth and angiogenesis as efficiently as does 4.6.1 antibody [17], which neutralizes the binding of VEGF to both receptors [18]. Taken together, these observations indicate that VEGFR2 is able to mediate all known functions of VEGF on vascular endothelial cells whereas VEGFR1 might play a minor or indirect role [19,20].

The specificity of VEGFR2 expression, its location on the surface of the tumor vessels, and its predominant role in tumor angiogenesis make it a highly desirable target for the development of both antiangiogenic and vascular targeting drugs. Mouse tumor models represent a major testing system to evaluate such drugs. Surprisingly, only one monoclonal antibody (DC101) with definite specificity for mouse VEGFR2 (Flk-1) has been described [21]. DC101 has been shown to inhibit the phosphorylation of Flk-1, VEGF-dependent DNA synthesis in cultured cells, neoangiogenesis in Matrigel plugs and tumors in vivo, and tumor growth in animal models [21]. To our knowledge, immunohistochemical detection of native VEGFR2 by monoclonal antibodies has not been documented in the literature, although several polyclonal anti-VEGFR2 antibodies, including the TO14 antibody produced in our laboratory [22,23], have been successfully used for this purpose.

In the present study, we raised monoclonal anti-VEGFR2 antibodies that recognize native VEGFR2. The monoclonal antibodies have high affinity for Flk-1 in enzyme-linked immunosorbent assay (ELISA), recognize Flk-1 on frozen tissue sections and on cultured mouse endothelial cells, compete with [¹²⁵I]VEGF₁₆₅ for binding in vitro, and localize specifically to Flk-1-positive tumor vessels in vivo. These antibodies are useful reagents for monitoring the expression of the VEGFR2 protein in mouse tumor models and for exploring the vascular targeting and imaging potential of this tumor endothelial marker.

Materials and Methods

Materials

Dulbecco's modified Eagle's medium (DMEM), glutamine, sodium pyruvate, and nonessential amino acids were obtained from Life Technologies (Grand Island, NY). Human recombinant [¹²⁵I]VEGF₁₆₅ with a specific activity of 2000 µCi/ng was purchased from Perkin Elmer (Boston, MA). Purified mouse Flt-1 and Flt-4 proteins were purchased from R&D Systems (Minneapolis, MN).

Antibodies

Rat antimouse CD31 antibody was from PharMingen (San Diego, CA). Antimouse Flt-1 antibody (clone 10.2) was from Biodesign International (Saco, ME). Noncompeting mouse monoclonal (clone 1A8) and rabbit polyclonal antibodies (TO14) against murine Flk-1 were produced in our laboratory and previously described [22,23]. Goat antirat, antimouse, and antirabbit secondary antibodies conjugated to horseradish peroxidase (HRP) were purchased from Dako (Carpinteria, CA).

Cell Culture

Murine brain endothelioma bEnd.3 cells were provided by Prof. Werner Risau (Max Plank Institution, Munich, Germany). L540Cy human Hodgkin's lymphoma cells were provided by Prof. V. Diehl (Köln, Germany). NCI-H358 human nonsmall cell lung carcinoma was provided by Dr. Adi Gazdar (Southwestern Medical Center, Dallas, TX). Meth A mouse fibrosarcoma, 3LL mouse lung carcinoma, and MDA-MB-231 human breast carcinoma were obtained from American Type Cell Collection (Rockville, MD). Tumor and endothelial cells were maintained in DMEM supplemented with 10% fetal calf serum, 2 mM l-glutamine, 2 U/ml penicillin G, and 2 µg/ml streptomycin, and subcultured using 0.125% trypsin in phosphate-buffered saline (PBS) solution containing 0.2% EDTA. L540Cy cells were maintained in RPMI medium with the same supplements. Tumor cells intended for injection into mice were washed once and resuspended in serum-free medium. Cell number and viability were determined by staining with 0.2% trypan blue dye. Only single-cell suspensions of greater than 90% viability were used for in vivo studies.

Generation of Rat Monoclonal Antibodies Against Mouse VEGFR2 (Flk-1)

The soluble domain of mouse VEGFR2 was expressed in Sf9 insect cells and purified to homogeneity as previously described [22]. Normal, 6-week-old, female Lewis rats were purchased from Charles River (Wilmington, MA) and used for immunizations. Purified Flk-1 (100 µg per injection) was mixed with Titer-Max (Corixa, Seattle, WA) and injected at four subcutaneous sites. The injections were repeated four more times. Titer of polyclonal antibodies was determined 2 days after each immunization. When the titer reached 1,000,000, rats were sacrificed and their spleens were harvested for fusion with myeloma partner P3X63AG8.653 line (653 cells), obtained from ATCC. Alternatively, splenocytes from immunized rats were fused with 653 cells stably transfected with the apoptotic inhibitor, CrmA [24]. Prior studies determined that 653^CrmA fusants display improved survival and clonogenicity during the isolation and expansion of single hybridoma clones. Positive wells were identified by screening on immobilized Flk-1 and were subcloned three times using limiting dilution. The rat immunoglobulin isotype was determined using a kit from Zymed Laboratories (San Francisco, CA). This panel of monoclonal antibodies was termed RAFLs (Rat Anti Flk). Individual antibodies from this panel were identified by sequential numbers (RAFL-1, RAFL-2, etc.).

Antibody Purification

Antibodies were purified from supernatants of hybridoma lines grown in Integra flasks (Integra Biosciences, East Dundee, IL). Supernatants were centrifuged, filtered through 0.22-µm filters, and loaded onto Sepharose-Protein G columns. IgG was eluted with citric acid buffer (pH 3.5), dialyzed into PBS, and stored thereafter at 4°C in the same buffer. Purity was estimated by SDS-PAGE and was routinely >95%.

Binding to Plastic Immobilized Soluble Domain of Mouse VEGFR2 on ELISA

All steps of the ELISA procedure were performed at room temperature. Microtiter plates were coated with the soluble domain of Flk-1 (1 µg/ml). Tissue culture supernatants from hybridoma lines or purified IgG were added to the wells in the presence of 2% BSA diluted in PBS. After 2 hours, plates were washed and incubated for 1 hour with goat antirat IgG (Fc gamma-specific) conjugated to HRP. Peroxidase activity was measured by adding O-phenylenediamine (0.5 mg/ml) and hydrogen peroxidase (0.03% wt/vol) in citrate-phosphate buffer (pH 5.5). The reaction was stopped by 100 µl of 0.18 M H₂SO₄. The absorbance was read at 490 nm. Experiments were performed in triplicate and were repeated three times.

Recognition of Mouse VEGFR2 on the Surface of Intact Mouse Endothelial Cells bEnd.3

bEnd.3 cells were seeded at a density of 50,000 cells/ml (0.5 ml per well) on sterile Lab-Tek II chamber glass slides (Nalge Nunc International, Naperville, IL) 2 days before the assay. Cells were washed twice with DPBS containing 2 mM Ca²⁺ and Mg²⁺ and 0.2% (wt/vol) gelatin as a carrier protein. In some experiments, cells were fixed with 0.25% glutaraldehyde prior to incubation with antibodies. Antibodies were added at the concentration of 2 µg/ml and incubated with cells for 2 hours at RT. The monolayer was subsequently washed and incubated with goat antirat IgG-HRP conjugate (1:500 dilution). VEGFR2-positive cells were detected by the addition of carbazole substrate, generating insoluble red brownish precipitate. The substrate was removed after 15 minutes of incubation and cells were counterstained using Mayer's hematoxylin solution (Sigma, St. Louis, MO). The control rat IgG and secondary antibody alone did not stain. Each experiment was repeated at least twice.

Competition of Anti-VEGFR2 Antibodies with [¹²⁵I]VEGF₁₆₅ for the Binding to VEGFR2 on Cell Surfaces

Binding assays were performed using mouse bEnd.3 endothelial cells. Cells were seeded in 48-well plates at a density of 50,000 cells/ml 48 hours before the assay. Cells were washed with DPBS containing 2 mM Ca²⁺ and Mg²⁺ and 0.2% (wt/vol) gelatin as a carrier protein (binding buffer). Human recombinant [¹²⁵I]VEGF₁₆₅ was diluted in the binding buffer to a final concentration of 25 pM. This concentration gives 90% of saturation binding on bEnd.3 cells. RAFL antibodies were immediately added at various concentrations (0.5–10 µg/ml). Rat IgG_2a of irrelevant specificity was used as a negative control. Cells were incubated with the mixture of [¹²⁵I]VEGF₁₆₅ and anti-Flk-1 antibodies for 4 hours at 4°C. Monolayers were washed three times with the binding buffer and dissolved by the addition of 0.5 M NaOH. Radioactivity was quantified using a gamma counter. Nonspecific binding was determined in the presence of a 100-fold molar excess of cold VEGF and was found to be 8% to 10% of total. The value of specific VEGF binding in the presence of control rat IgG_2a (5 µg/ml) was taken as 100%. Percent of blocking was calculated according to the following formula:

$$\frac{\text{Total net cpm}-\text{Net cpm in the presence of competitor}}{\text{Total cpm}}\text{x 100}$$

Each experiment was performed in duplicate and repeated three times.

Immunohistochemical Detection of Mouse VEGFR2 on Frozen Sections

Frozen sections of mouse normal organs and human tumors grown in nude mice were fixed with acetone for 5 minutes and rehydrated with PBST for 10 minutes. Sections were overlaid with undiluted hybridoma supernatants or purified antibodies solutions (10 µg/ml). All dilutions were prepared in PBST containing 0.2% BSA. Affinity-purified rabbit anti-Flk-1 antibody TO14, which was previously raised in our laboratory [23], was used as a positive control for immunohistochemical studies. Rat anti-CD31 antibody was used to identify endothelial cells on sections. Secondary antibodies were either antirat or rabbit IgG, conjugated to peroxidase. HRP activity was detected by developing with either carbazole or DAB substrates.

Growth of Tumors for Localization Studies

For localization studies, 2x10⁷ L540 human Hodgkin's lymphoma cells or 1x10⁷ cells of other tumor types lines were injected subcutaneously into the right flank of SCID mice (Charles River). MDA-MB-231 cells were injected into the mammary fat pad (MFP) of female nude mice (orthotopic model). Tumors were allowed to reach a volume of 0.4 to 0.7 cm³. A minimum of three animals per tumor group was used.

Localization of Anti-VEGFR2 Antibodies to Flk-1-Positive Vessels In Vivo

Tumor-bearing mice were intravenously injected with 50 µg of anti-VEGFR2 or control rat IgG_2a antibodies. One hour later, mice were sacrificed and exsanguinated. The tumor and organs were removed and snap-frozen for preparation of cryosections. The localized rat IgG was detected on frozen sections using goat antirat antibody conjugated to peroxidase (10 µg/ml). This secondary antibody does not recognize mouse immunoglobulins present in nude mice.

Inhibition of Tumor Microvasculature in Orthotopic Model of Breast Cancer

Nude mice (six animals per group) bearing orthotopic MDA-MB-231 tumors were treated with RAFL-1 antibody (100 µg per dose, i.p., every second day). A control group of mice bearing tumors of similar size (∼ 150 µl) was injected with saline. The tumor size and general health of the animals were recorded every 2 to 3 days. Perpendicular tumor diameters were measured using a Vernier scale caliper. Tumor volumes were calculated according to the following equation: volume = D x d² x π/6, where D is the larger tumor diameter and d is the smaller diameter. Animal care was in accordance with institutional guidelines.

After 5 weeks of treatment, mice were anesthesized, and their blood circulation was perfused with heparinized saline as described before [25]. The tumor and major organs were removed and snap-frozen in liquid nitrogen. Cryostat sections of the tissues were cut and stained for vessels using pan-endothelial rat antibody antimouse CD31 (PharMingen). Vessels were counted in 10 fields (two fields from each quadrant of a cross section and two in the center) at a final magnification of x100. The mean number of vessels per square millimeter was calculated.

Statistical Analysis

Results are expressed as mean±SEM, unless otherwise indicated. Statistical significance was determined by the Student's t-test with two-tail analysis. A P value of <.05 was taken as statistically significant.

Results

Generation of Monoclonal Antibodies Against Mouse VEGFR2

Monoclonal hybridomas were generated by fusing splenocytes from immunized rats with 653^CrmA cells or 653 cells. The initial screening of supernatants derived from 653^CrmA fusants on immobilized Flk-1 antigen in ELISA yielded 110 wells having supernatants that were highly positive (higher than 2 OD), compared with only two wells in a similarly performed fusion deriving from the 653 cells. The higher fusion efficiency of the CrmA-transfected myeloma cells has been observed in several fusions and is possibly due to stable expression of the antiapoptotic protein, CrmA [24], by myeloma partner cells. From this extensive primary pool, we selected eight stable clones (RAFL-1 to RAFL-8) secreting high-affinity antibodies with diverse functional properties. All the antibodies were rat IgG_2a. All but RAFL-8 had κ light chain.

Binding of Anti-VEGFR2 Antibodies to Immobilized Soluble Domain of Flk-1 in ELISA

RAFL-1 to RAFL-8 antibodies bound strongly and specifically to sFlk-1 in ELISA. Half-maximal binding was observed at concentrations that ranged between 10 and 67 pM (Figure 1). RAFL-4 was the antibody having the strongest binding from this panel with a half-maximal binding of 10 pM. All antibodies reached saturation at concentrations of 0.2 to 0.4 nM. None of the antibodies reacted with purified mouse Flt-1 or Flt-4 proteins, which have structural similarity to Flk-1 (data not shown).

Figure 1.

Binding of RAFL antibodies to mouse VEGFR2 in ELISA. The extracellular domain of mouse VEGFR2 was immobilized on plastic by incubating 96-well plates with 1 µg/ml purified protein. RAFL antibodies were added at concentrations ranging from 6.7 to 0.005 nM. The bound antibodies were detected with goat antirat IgG conjugated to peroxidase. Results show absorbance at 490 nm that reflects peroxidase activity. Each point is the average of triplicate measurements, the standard deviations of which did not exceed 5% of the mean value.

Recognition of Mouse VEGFR2 Expressed on Surface of Cultured EC

The ability of RAFL antibodies to bind to VEGFR2 on intact fixed or unfixed mouse bEnd.3 endothelial cells was examined. RAFL-1, RAFL-5, and RAFL-8 stained unfixed cells but not cells after glutaraldehyde fixation, indicating preferential recognition of native epitope(s). RAFL-6 stained fixed cells but not unfixed cells, suggesting that it recognizes an epitope in denatured VEGFR2. RAFL-2, RAFL-3, and RAFL-7 bound equally well to native and fixed cells (Table 1). RAFL-4 did not stain fixed or unfixed cells.

Table 1.

Functional Properties of Rat Monoclonal Antibodies Against VEGFR2.

Antibody	ELISA (pM)*	VEGF Blocking (%)†	Binding to Cell Surface‡		Binding to Frozen Sections§	Localization to Vessels In Vivo¶
			Unfixed	Fixed
RAFL-1	15	34	Yes	No	4+	Yes
RAFL-2	23	55	Yes	Yes	4+	Yes
RAFL-3	67	48	Yes	Yes	2+	Yes
RAFL-4	10	12	No	No	-	No
RAFL-5	50	51	Yes	No	2+	No
RAFL-6	27	16	No	Yes	2+	No
RAFL-7	30	49	Yes	Yes	2+	No
RAFL-8	70	42	Yes	No	1+	No
TO14	25	32	Yes	Yes	4+	Yes
Control rat IgG	N/A	10	No	No	-	No

^*Concentration of antibody required to achieve half-maximal binding to immobilized VEGFR2 in ELISA.

^†Maximal percent of [¹²⁵I]VEGF₁₆₅ displacement by 2 µg/ml antibody added to competitive assay.

^‡Determined by staining of either 0.25% glutaraldehyde fixed or native bEnd.3 cells using 2 µg/ml primary antibody.

^§Determined by staining frozen sections of tumors and kidney using 10 Ég/ml antibody. Intensity of staining was scored by comparing to that produced by a panendothelial marker MECA 32: (-) no staining; (1+) weak; (2+) moderate; (3+) moderate to strong; (4+) strong, MECA 32-like intensity.

^¶Determined by specific detection of rat IgG on vasculature of tumors and kidney derived from nude mice bearing orthotopic MDA-MB-231 human breast tumors.

RAFL antibodies bound preferentially to cells in mitosis (Figure 2, A and B, arrows). Strong binding to cells during metaphase (Figure 2A) and telophase (Figure 2B) was observed. Nondividing cells, having distinct nuclear membranes and nucleoli, were more weakly stained (Figure 2, arrowheads). In contrast, antimouse CD31 antibody uniformly stained all endothelial cells with strong intensity regardless of stage of division (data not shown). This observation suggests that VEGFR2 on endothelial cells is upregulated during mitotic division.

Figure 2.

Binding of RAFL-1 and RAFL-6 antibodies to mouse endothelial cells. Cultures of bEnd.3 endothelial cells, which contained both dividing and nondividing cells, were grown on sterile glass slides. Binding of RAFL antibodies to unfixed cells (A) or fixed cells (B) was examined. Arrows point to strongly stained endothelial cells that display morphological characteristics typical of cells during metaphase (A) or telophase (A,B). Arrowheads point to neighboring cells lacking morphological signs of cell division that were not stained by either RAFL-1 or RAFL-6.

Blocking of Binding of [¹²⁵I]VEGF₁₆₅ to VEGFR2 on Mouse Endothelial bEnd.3 Cells

The six RAFL antibodies that bound to intact unfixed cells were tested for their ability to block the binding of [¹²⁵I]VEGF₁₆₅ to VEGFR2 on bEnd.3 cells. We established in prior experiments that bEnd.3 cells express approximately equal numbers of Flk-1 and Flt-1 receptors. This is based on the observation that both antimouse Flk-1 antibody TO14 and antimouse Flt-1 antibody 10.2, when added at a saturating concentration of 10 µg/ml, were able to compete out only ∼50% of [¹²⁵I]VEGF₁₆₅ binding (Figure 3). Thus, antibodies reactive exclusively to the Flk-1 receptor are expected to block the binding of VEGF by a maximum of 50%. Noncompeting anti-Flk-1 antibody, 1A8 (10 µg/ml), reduced the binding of [¹²⁵I]VEGF₁₆₅ by 10% under the same conditions, indicating that 90% of [¹²⁵I]VEGF₁₆₅ was binding specifically to cell surface receptors.

Figure 3.

Inhibition of [¹²⁵I]VEGF₁₆₅ binding to the surface of endothelial cells by RAFL antibodies. Log phase or confluent endothelial cell cultures were incubated with [¹²⁵I]VEGF₁₆₅ (25 pM) in the presence of various RAFL antibodies; nonblocking anti-VEGFR2 IgG, 1A8; blocking rabbit polyclonal anti-VEGFR2, TO14; blocking mouse monoclonal anti-VEGFR1, clone 10.2; or unlabeled VEGF₁₆₅ (2.5 nM). All antibodies were added at a final concentration of 5 µg/ml. After 4 hours of incubation at 4°C, cells were washed and specific VEGF₁₆₅ binding was determined as described under Materials and Methods section. The mean percent of VEGF blocking ±SEM is presented. Insert: Relationship between [¹²⁵I]VEGF₁₆₅ binding and concentration. A VEGF₁₆₅ concentration of 25 pM gives about 90% maximal binding.

RAFL-1, RAFL-2, RAFL-3, RAFL-5, RAFL-7, and RAFL-8 inhibited the binding of [¹²⁵I]VEGF₁₆₅ to dividing, sparsely seeded endothelial cells by 33% to 55%, indicating that RAFL antibodies are efficient blockers of VEGF interaction with R2. RAFL-2 was the most efficient antibody in this assay (Table 1). All competing antibodies had little or no effect on [¹²⁵I]VEGF₁₆₅ binding to confluent monolayers (Figure 3). This is in accord with the above observation that dividing endothelial cells have higher surface expression of VEGFR2 than nondividing cells (Figure 2).

Binding to Vascular Endothelial Cells on Frozen Tumor Sections

The ability of RAFL antibodies to recognize VEGFR2 on tissue sections was of particular interest because of the paucity of available reagents having this ability. Two of the RAFL antibodies (RAFL-1 and RAFL-2) showed strong recognition of vascular endothelial cells on frozen tumor sections, ranked as 4+ (Table 1, Figure 4). Four additional antibodies (RAFL-3, RAFL-5, RAFL-6, and RAFL-7) had moderate binding, ranked as 2+. RAFL-4 antibody, which had high affinity for immobilized Flk-1 in ELISA, had no recognition of the cell surface receptor and did not stain tumor vessels.

Figure 4.

Detection of VEGFR2 on vasculature of various tumors by RAFL-1 antibody. NCI-H358 human nonsmall lung carcinoma and L540 Hodgkin's disease tumors were grown in SCID mice. Meth A mouse fibrosarcoma and 3LL mouse lung carcinomas were grown in BALB/c and C57BL/6 mice, respectively. VEGFR2 was detected on frozen tumor sections by incubating them with RAFL-1 antibody (10 µg/ml) followed by goat antirat IgG HRP, as described under Materials and Methods section. Sections were developed with DAB and counterstained with hematoxylin. Representative images from each tumor type are shown.

Six different tumors growing in mice were tested for staining with RAFL-1, RAFL-2, RAFL-3, and RAFL-5. The tumors were NCI-H358 human NSCLC (Figure 4), L540 human Hodgkin's disease (Figure 4), Meth A mouse fibrosarcoma (Figure 4), 3LL mouse lung tumor (Figure 4), orthotopic MDA-MB-431 human breast carcinoma (Figure 5), and orthotopic L3.6pl human pancreatic carcinoma (Figure 6). The four antibodies gave similar patterns of staining in each individual tumor. They stained moderately to strongly 50% to 80% of vessels expressing CD31, a pan-mouse endothelial marker (Figure 4). This corresponds to 50 to 160 vessels/mm². Rat IgG_2a of irrelevant specificity did not stain tumor sections. TO14, a rabbit polyclonal antibody to mouse VEGFR2 [23], showed the same intensity and pattern distribution of vascular staining as did RAFL antibodies. However, TO14 cross-reacts with cytosolic components of the human tumor cells, giving a high background. In contrast, RAFL antibodies specifically recognized mouse tumor blood vessels.

Figure 5.

Biodistribution of RAFL-1 antibody in mice bearing orthotopic MDA-MB-231 human breast xenografts. Nu/nu mice bearing MDA-MB-231 tumors in their MFPs were injected intravenously with 50 µg of RAFL-1 antibody. One hour later, their blood circulation was perfused with saline, and tumors and normal organs were harvested and prepared for cryosectioning. VEGFR2 protein was detected by incubating frozen sections with RAFL-1 antibody and appropriate secondary antibody (left column). The RAFL-1 antibody that localized to the vasculature after intravenous injection was detected indirectly by incubating sections with secondary antibody alone (center column). To identify vessels, parallel sections were stained with a pan-endothelial marker, anti-CD31 antibody (right column). Note that the majority of tumor vessels both express VEGFR2 and display localized RAFL-1 antibody. Kidney glomeruli behave likewise. Neither VEGFR2 protein nor localized RAFL-1 antibody was detected on vessels of lung or heart.

Figure 6.

Expression of VEGFR2 on vessels of normal mouse pancreas and orthotopically implanted human pancreatic tumor, L3.6pl. Human pancreatic tumor, L3.6pl, was orthotopically grown in nude mice. Frozen sections of normal pancreas from nontumor bearing mice and L3.6pl tumors were stained with RAFL-1 antibody as described above. Arrows indicate stained blood vessels in both types of tissue. Control rat IgG_2a did not stain either of the sections.

Binding to Vascular Endothelial Cells on Sections of Normal Tissues

A limited panel of normal tissues including heart, lung, liver, kidney, pancreas, and brain was examined with RAFL antibodies by direct staining. Vessels in the heart and lungs (Figure 5), liver, and brain cortex (not shown) were not visibly stained. In the kidney, RAFL antibodies discretely stained glomerular endothelium (Figure 5). In the pancreas, most of the CD31-positive vessels in the islets were positive for VEGFR2 (Figure 6). Vascular expression of VEGFR2 in normal kidney glomeruli and pancreas was also confirmed by a rabbit anti-VEGFR2 antibody, TO14.

Localization of Anti-VEGFR2 Antibodies to Tumor Vessels in Orthotopic MDA-MB-231 Breast Tumors in Mice

RAFL-1, RAFL-2, RAFL-3, and a control rat IgG^2a antibody were intravenously injected into nude mice with human MDA-MB-231 breast tumors growing in their MFPs. One hour later, the mice were sacrificed and their blood circulation was perfused to remove free antibody. Frozen sections were prepared from the tumor and normal organs and examined immunohistochemically to determine the localization of the RAFL antibodies or control rat IgG (Figure 5). About 60% of tumor vessels that were positive for CD31 were also positive for localized RAFL-1, RAFL-2, and RAFL-3. Vessels with bound RAFL antibodies were homogeneously distributed within the tumor vasculature. All VEGFR2-positive vessels detected by direct staining of tumor sections also had localized RAFL-1, indicating that all VEGFR2-positive vessels were accessible to intravenously administered antibody. Vessels in normal organs were unstained, with the exception of the kidney glomeruli (Figure 5) and the pancreas (Figure 6). Control rat IgG did not localize to tumor or normal vessels in any of the mice, indicating specificity of detection of anti-VEGFR2 antibodies.

Inhibition of Tumor Vasculature by RAFL-1 Antibody in MDA-MB-231 Breast Cancer Model

Groups of six tumor-bearing mice were treated with either saline or RAFL-1. Mice were sacrificed when the mean tumor volume of the control and treated groups reached 1250 and 640 mm³, respectively. Treatment with RAFL-1 reduced the tumor volume by an average of 48%. Tumors from all treated and control mice were excised and microvessel density and morphology was assessed as described under Materials and Methods section.

The mean number of vessels per square millimeter was 164±7 and 57±5 in control and RAFL-1-treated groups, respectively. Vessels in control tumors mainly consisted of capillaries that were homogeneously distributed in all regions except necrotic areas. The vessel phenotype, number, and distribution of tumors from RAFL-1-treated mice were strikingly different from those of control tumors (Figure 7). Microvasculature was almost absent from treated tumors and residual vessels mainly consisted of mid-sized mature vessels coated by smooth muscle cells/pericytes. Necrotic areas occupied 80% to 90% of cross sections, leaving only a rim of viable cells of approximately 130 µm in width (Figure 7). No effects were observed on the vascularization of normal MFP epithelium adjacent to treated tumors. RAFL-1-treated and control mice had similar numbers and patterns of vessels in normal MFP epithelium as well as in other normal organs. These observations indicate that RAFL-1 treatment specifically inhibits VEGF-dependent angiogenesis in tumors.

Figure 7.

RAFL-1 antibody reduces microvascular density in MDA-MB-231 human breast tumors growing in the MFP of mice. Tumors from RAFL-1-treated (A) or rat IgG control antibody-treated (B) mice were removed, frozen sections were cut, and microvessels were detected by anti-CD31 antibody. Microvascular densities were determined on six individual slides per experimental group and the results were averaged. The mean number of vessels per square millimeter was 164±7 and 57±5 in control and RAFL-1-treated groups, respectively. Representative images of tumors derived from control and RAFL-1-treated mice are shown. Note the absence of microvasculature and the extent of necrosis in tumor from RAFL-1-treated mice compared to a fully viable and highly vascularized tumor from the control group.

Discussion

The goal of the present study was to generate a panel of antimouse VEGFR2 monoclonal antibodies to detect mouse VEGFR2 on frozen sections and to localize to tumor vessels in vivo. Hybridomas secreting high-affinity antibodies were generated using a myeloma partner that had been stably transfected with the antiapoptotic protein, CrmA [24]. The hybridomas resist apoptosis, enabling a large primary pool of positive clones to be selected. Eight hybridomas secreting high-affinity antibodies to Flk-1 were eventually selected from this pool (Figure 1). Most of the antibodies competed with [¹²⁵I]VEGF₁₆₅ for the binding to cell surface receptor (Figure 3). Most RAFL antibodies strongly stained dividing endothelial cells and had weak reactivity with nondividing cells present in the same cultures (Figure 2). Similar observations have been previously reported for mouse antihuman VEGFR2 antibodies [26].

The most intense detection of VEGFR2 on frozen sections of various tumors (Figures 4–6) and the most intense localization to tumor vessels in vivo were obtained with RAFL-1 and RAFL-2 antibody (Figure 6). VEGFR2 was observed with this antibody to be a selective tumor endothelial surface marker in ectopic and orthotopic mouse tumors and human xenografts. VEGFR2 was found in the majority of the tumor vessels and was homogeneously distributed within the vascular network. This pattern of expression is particularly desirable for targeting tumor vasculature as it enables the attack of both central and peripheral tumor regions.

In prior studies, mRNA for VEGFR2 was detected in the vascular endothelium in many normal organs [27]. In the present study, we did not detect VEGFR2 protein on vessels in normal lung, liver, heart, and brain cortex, although it was detectable on vessels in kidney glomerulus and pancreatic islets. The lack of staining of vessels in certain organs may be due to differences in the sensitivity of mRNA and protein detection techniques, or to rapid turnover of VEGFR2 protein on vessels in some organs. The expression of VEGFR2 in normal kidney glomeruli and pancreatic islets is thought to be required for maintenance of high vascular permeability [28,29].

The expression of VEGFR2 in kidney glomerulus and pancreatic islets raises the question of whether agents directed against VEGFR2 have sufficient specificity for tumor vessels to be useful for vascular targeting therapy. Several lines of evidence suggest that cytotoxic or prothrombotic effects of agents targeted to VEGFR2 might be tolerated by normal vessels. First, the level of VEGFR2 expression is higher on dividing endothelial cells, which are abundant in malignant tumors. Second, intracellular trafficking of an internalized anti-VEGFR2-toxin conjugate may differ between dividing and quiescent endothelia in that only in dividing cells can the conjugates reach a compartment from which the toxin moiety can escape into cytosol. This mechanism has been previously proposed to explain why dividing endothelial cells are 60-fold more sensitive to VEGF gelonin toxin than are confluent, quiescent endothelial cells [30]. Third, vascular targeting agents containing tissue factor as an effector moiety require coincident expression of a target molecule and phosphatidylserine for the induction of the coagulation cascade [31–33]. Endothelial cells in tumors expose phosphatidylserine on their luminal surface, whereas endothelial cells in resting normal organs do not [34]. We have previously demonstrated that the absence of externalized phosphatidylserine from the surface of VCAM-1-positive normal vessels prevented the action of anti-VCAM-1-tissue factor conjugate and averted toxicity [34]. Renal or pancreatic toxicities have not been reported for other VEGFR2 targeting or blocking reagents either in the mouse or clinical studies despite expression of VEGFR2 on vessels in these organs [35–37].

We demonstrate here that unconjugated RAFL antibodies directed against native Flk-1 receptor efficiently blocked VEGF binding to dividing endothelial cells (Figure 3) and inhibited the formation of tumor neovasculature in vivo (Figure 7). Treatment with anti-VEGFR2 antibody had no effect on morphology or number of vessels in normal tissues, indicating specific inhibition of VEGF-driven angiogenesis. These findings are in accord with prior observations [21,38]. It is possible that inhibition of binding of VEGF to VEGFR2 denies dividing remodeling endothelial cells VEGF-mediated survival signals, leading to their demise, whereas quiescent endothelium is not dependent on VEGF signals for survival.

In summary, our findings with RAFL antibodies confirm that VEGFR2 is strongly and consistently expressed on blood vessels in a variety of solid tumors, and suggest that anti-VEGFR2 antibodies could be used to deliver toxins, tissue factor, or radionuclides to tumor vasculature.

Abbreviations

VEGF

vascular endothelial growth factor

VEGFR2

VEGF receptor 2

HRP

horseradish peroxidase

HPF

high power field

MFP

mammary fat pad