Anionic phospholipids are largely absent from the external leaflet of the plasma membrane of mammalian cells under normal conditions. Exposure of phosphatidylserine on the cell surface occurs during apoptosis, necrosis, cell injury, cell activation, and malignant transformation. In the present study, we determined whether anionic phospholipids become exposed on tumor vasculature. A monoclonal antibody, 9D2, which specifically recognizes anionic phospholipids, was injected into mice bearing a variety of orthotopic or ectopic tumors. Other mice received annexin V, a natural ligand that binds to anionic phospholipids. Both 9D2 and annexin V specifically localized to vascular endothelium in all of the tumors, and also to tumor cells in and around regions of necrosis. Between 15 and 40% of endothelial cells in tumor vessels were stained. No localization was detected on normal endothelium. Various factors and tumor-associated conditions known to be present in the tumor microenvironment were examined for their ability to cause exposure of anionic phospholipids in cultured endothelial cells, as judged by 9D2 and annexin V binding. Hypoxia/reoxygenation, acidity, thrombin, and inflammatory cytokines all induced exposure of anionic phospholipids. Hydrogen peroxide was also a strong inducer. Combined treatment with inflammatory cytokines and hypoxia/reoxygenation had greater than additive effects. Possibly, injury and activation of tumor endothelium by cytokines and reactive oxygen species induce exposure of anionic phospholipids, most likely phosphatidylserine. Anionic phospholipids on tumor vessels could potentially provide markers for tumor vessel targeting and imaging.
Anionic phospholipids are largely absent from the surface of resting mammalian cells under normal conditions. PS, 3 which is the most abundant anionic phospholipid of the plasma membrane, is tightly segregated to the internal leaflet of the plasma membrane in most cell types (1 , 2) . PI, another major anionic phospholipid, is also situated predominantly in the internal leaflet of the plasma membrane (3) . The minor anionic phospholipids, PA and PG, have only been examined in a few cells types, but they also appear to be mainly situated in the internal leaflet of the plasma membrane (4) . CL, another anionic phospholipid, is present in the mitochondrial membrane and is absent from the plasma membrane (5) . The neutral phospholipids are also asymmetrically distributed in the plasma membrane; PE is predominately on the internal leaflet, whereas the choline-containing phospholipids, PC and SM, are predominantly on the external leaflet.
PS asymmetry, along with that of PE, is maintained by an ATP-dependent aminophospholipid translocase that catalyzes the transport of aminophospholipids from the external leaflet to the internal leaflet of the plasma membrane (6) . Loss of PS and PE asymmetry results from the outward movement of these phospholipids in the plasma membrane and is caused either by inhibition of the translocase (7) or activation of scramblase, a Ca2+-dependent enzyme that transports all of the lipids bidirectionally (8) . Loss of asymmetry is observed under different pathological and physiological conditions, including apoptosis (9) , cell activation (10 , 11) , injury (12) , and malignant transformation (13) . Exposure of PS also plays a role in intercellular fusion (14 , 15) and cell migration (16) . Endothelial cells externalize PS in response to increased Ca2+ fluxes induced by thrombin (17) , calcium ionophore or phorbol esters (18) , hyperlipidemia (19) and nonlytic concentrations of complement proteins C5b-9 (20) .
Several major consequences follow membrane PS exposure. Phagocytic macrophages recognize, attach, and eliminate PS-positive senescent and apoptotic cells (21 , 22) . PS also mediates attachment of T lymphocytes to thrombin-activated endothelial cells (17) . The complement system is activated by PS and contributes to the lysis of PS-positive cells (23) . Finally, PS exposure contributes to a procoagulant shift on the endothelium (1 , 9) by providing a negatively charged lipid surface for assembly and activation of coagulation complexes (24 , 25) . The prothrombotic character of the tumor endothelium has long been recognized (26) .
In the present study, we hypothesized that anionic phospholipids become exposed on tumor vasculature because of increased stress conditions of the tumor microenvironment. Injury and activation of tumor endothelium have been shown to be caused by: (a) tumor-derived interleukin-1 and tumor necrosis factor, which activate the endothelium and induce expression of cell adhesion molecules (27 , 28) ; (b) ROS generated by leukocytes that adhere to the endothelium (28) ; and (c) ROS generated by tumor cells themselves as a byproduct of metabolism (27 , 29) or as a result of exposure to hypoxia followed by reoxygenation (30) . These observations suggested that Ca2+ fluxes might be generated by these stresses within the tumor endothelium that, in turn, cause exposure of PS, and probably also of PE, through activation of scramblase or inhibition of aminophospholipid translocase.
To detect cell surface anionic phospholipids, we generated a new monoclonal antibody, 9D2, which reacts with anionic but not neutral phospholipids. 9D2 antibody is more specific for anionic phospholipids than is the natural ligand, annexin V, which strongly binds to PE, in addition to anionic phospholipids. Annexin V has been used successfully to image activated platelets in thrombi, apoptotic cells in cardiac allografts undergoing rejection, cyclophosphamide-treated lymphomas, and anti-Fas antibody-treated livers in rodents (31) . We found that 9D2 and annexin V localize specifically to tumor endothelium after i.v. injection to mice bearing various types of solid tumors. This finding indicates that anionic phospholipids, most likely PS, routinely become exposed on the surface of tumor vascular endothelium.
MATERIALS AND METHODS
Na125I was obtained from Amersham (Arlington Heights, IL). Dulbecco’s modified Eagle’s tissue culture medium and Dulbecco PBS containing Ca2+ and Mg2+ were obtained from Life Technologies, Inc. (Grand Island, NY). FCS was obtained from Hyclone (Logan, UT). l-α-PS, l-α-PC, CL, l-α-PE, l-α-PI, SM, PA, PG, O-phenylenediamine, hydrogen peroxide, and thrombin were from Sigma (St. Louis, MO). Flat-bottomed plates with 24 wells were obtained from Falcon (Becton Dickinson and Co., Lincoln Park, NJ). Recombinant hepatocyte growth factor (or scatter factor) and actinomycin D was from Calbiochem (San Diego, CA). Recombinant murine interleukin-1 α, β, and TNF-α were purchased from R&D Systems (Minneapolis, MN). IFN of universal type I (hybrid protein that substitutes for all types of IFNs) was purchased from PBL Biomedical Laboratories (New Brunswick, NJ). Recombinant human VEGF 121, human platelet-derived growth factor-BB, IL-6, IL-8, IL-10, and human fibroblast growth factor-2 were purchased from PeproTech (Rocky Hill, NJ).
MECA 32, a pan mouse endothelial cell antibody, was kindly provided by Dr. Eugene Butcher (Stanford University, Stanford, CA) and served as a positive control for immunohistochemical studies. Details of this antibody have been published (32) . Rabbit antirat immunoglobulin, rat-antimouse immunoglobulin, and goat-antimouse and antirat secondary antibodies conjugated to HRP were purchased either from DAKO (Carpinteria, CA) or from Jackson Immunoresearch Labs (West Grove, PA).
L540Cy Hodgkin lymphoma cells, derived from a patient with end-stage disease, were provided by Prof. Volker Diehl (Medizinische Universitätsklinik 1, Köln, Germany). NCI-H358 human non-small cell lung carcinoma was provided by Dr. Adi Gazdar (Southwestern Medical Center, Dallas, TX). Meth A mouse fibrosarcoma and MDA-MB-231 human breast carcinoma were obtained from American Type Cell Collection (Rockville, MD). The mouse brain endothelioma line, bEnd.3, was provided by Prof. Werner Risau (Max Plank Institution, Munich, Germany). ABAE cells were purchased from Clonetics (San Diego, CA).
bEnd.3, ABAE cells, and all of the tumor cells except L540Cy lymphoma were maintained in DMEM supplemented with 10% FCS, 2 mm l-glutamine, 2 units/ml penicillin G, and 2 μg/ml streptomycin. L540Cy cells were maintained in RPMI 1640 containing the same additives. Cells were subcultured once a week. Trypsinization of bEnd.3 cells was performed using 0.125% trypsin in PBS containing 0.2% EDTA. For in vitro studies, endothelial cells were seeded at a density of 1 × 104 cells/ml in 1 ml of culture medium in 24-well plates and incubated 48–96 h before being used in the assays. Medium was refreshed 24 h before each experiment.
Growth of s.c. Implanted Tumors.
For localization studies, 2 × 107 L540 human Hodgkin’s lymphoma cells or 1 × 107 cells of other tumor types were injected s.c. into the right flank of SCID mice (Charles River, Wilmington, MA). Tumors were allowed to reach a volume of 0.4–0.7 cm3. A minimum of three animals per group was used. Experiments were replicated at least three times.
Orthotopic Model of Human MDA-MB-231 Breast Carcinoma.
Female nu/nu or SCID mice were purchased from Charles River. MDA-MB-231 human mammary carcinoma cells were implanted into the mammary fat pad according to a published protocol (33) . Briefly, mice were anesthetized, and a 5-mm incision was made in the skin over the lateral thorax. The mammary pad was exposed to ensure the correct site for injection of 1 × 107 MDA-MB-231 cells resuspended in 0.1 ml of saline.
Generation of 9D2 Rat Monoclonal Antibody Reactive with Anionic Phospholipids.
To generate monoclonal antibodies reactive with anionic phospholipids, female Lewis rats were immunized with bEnd.3 endothelial cells that had been treated with 200 μm of hydrogen peroxide for 2 h. The treatment caused translocation of anionic phospholipids to the external surface in 70–90% of cells as detected by 125I-labeled annexin V. Treated cells were washed, detached, and counted. Two-million cells were suspended in sterile PBS and injected five times i.p. with the interval of 3 weeks between injections. The titer of polyclonal antibodies to anionic phospholipids was determined 2 days after each immunization. Hybridomas were obtained by fusing splenocytes from immunized rats with myeloma partner P3 × 63AG8.653 cells (American Type Culture Collection). The reactivity of the selected antibody, rat IgM 9D2, with PS and CL was established by screening hybridoma supernatants on PS, CL, PE, and PC immobilized on plastic. Additional characterization of phospholipid specificity of 9D2 is given in “Results.”
Reactivity of 9D2 Antibody and Annexin V with Plastic-immobilized Phospholipids.
Phospholipids were dissolved in n-hexane to a concentration of 50 μg/ml. One-hundred μl of this solution was added to wells of 96-well microtiter plates. After evaporation of the solvent in air, the plates were blocked for 2 h with 10% fetal bovine serum diluted in DPBS containing 2 mm Ca2+ (binding buffer). 9D2 antibody or annexin V were diluted in the binding buffer in the presence of 10% serum at an initial concentration of 6.7 nm. Serial 2-fold dilutions were prepared in the plates (100 μl per well). The plates were then incubated for 2 h at room temperature. The plates were washed, and the 9D2 and annexin V were detected by goat antirat IgM conjugated to HRP and rabbit antihuman annexin V followed by goat antirabbit IgG conjugated to HRP (all diluted 1:1000), respectively. Secondary reagents were detected by using chromogenic substrate O-phenylenediamine followed by reading plates at 490 nm using a microplate reader (Molecular Devices, Palo Alto, CA). The specificity of the 9D2 antibody binding was validated by using control rat IgM of irrelevant specificity (PharMingen, San Diego, CA). The specificity of annexin V binding to phospholipids, which is Ca2+-dependent, was determined by diluting the reagent in the DPBS containing 5 mm EDTA. Additional negative controls consisted of washing the plates with the binding buffer containing 0.2% of a detergent Tween 20. This treatment extracts lipids, thus removing the phospholipid that was absorbed to plastic. Neither 9D2 antibody nor annexin V bound to detergent-washed plates.
Detection of Externally Positioned Anionic Phospholipids by 9D2 Antibody and Annexin V on the Surface of Cultured Endothelial Cells.
Endothelial cells were grown until they reached ∼70% confluence. To induce PS exposure, cells were treated with H2O2 (200 μm) for 1 h at 37°C. Control and treated slides were washed with DPBS containing Ca2+ and Mg2+, and fixed with 0.25% of glutaraldehyde diluted in the same buffer. Excess aldehyde groups were quenched by incubation with 50 mm of NH4Cl for 5 min. To examine the effect of detergents and organic solvents on detection of phospholipids, some slides were preincubated with acetone (5 min) or with PBS containing 1% (v/v) Triton X-100. Cells were washed with DPBS containing Ca2+, Mg2+, and 0.2% (w/v) gelatin, and incubated with 1 μg/ml of biotinylated annexin V (PharMingen) or with 1 μg/ml of 9D2 antibody. After 2 h of incubation, cells were washed with 0.2% gelatin buffer and were incubated with streptavidin-HRP (1:500 dilution) or with antirat-HRP. Rat IgM of irrelevant specificity and streptavidin alone were used as negative controls in these experiments. All of the steps were performed at room temperature. HRP activity was measured by adding O-phenylenediamine (0.5 mg/ml) and hydrogen peroxide (0.03% w/v) in citrate-phosphate buffer (pH 5.5). After 15 min, 100 μl of supernatant were transferred to 96-well plates, 100 μl of 0.18 m H2SO4 were added, and the absorbance was measured at 490 nm. Alternatively, PS-positive cells were detected by addition of carbazole substrate, resulting in insoluble red-brownish precipitate. Each experiment was performed in duplicate and repeated at least twice.
Inhibition of 9D2 and Annexin V Binding to Phospholipids by Liposomes.
Specificity of phospholipid recognition was additionally confirmed by competition assays with various liposomes. Liposomes were prepared from solutions of 5 mg of a single phospholipid in chloroform. The solutions were dried under nitrogen to form a thin layer in a round-bottomed glass flask. Ten ml of Tris buffer (0.1 m; pH 7.4) were then added, and the flask was sonicated five times for 2 min. 9D2 or annexin V (6.66 nm) were preincubated with 200 μg/ml of liposomal solution for 1 h at room temperature. The mixture was added to phospholipid-coated plates or endothelial cell monolayers. The ability of 9D2 to bind to an immobilized phospholipid or cell surface in the presence or absence of the different liposomes was determined as described above.
Competition of 9D2 and Annexin V for Binding to Immobilized PS.
Biotinylated 9D2 antibody and annexin V were prepared by incubating purified proteins with a 10-fold molar excess of N-hydroxysuccinimide biotin (Sigma) for 1 h at room temperature. Free biotin was removed by dialysis against PBS. The biotinylation procedure did not impair the PS-binding capacity of either protein. For competition experiments, unmodified and biotinylated proteins were premixed with a 10-fold molar excess of unmodified proteins. The mixtures were then added to PS-coated plates. Bound reagents were detected by streptavidin-HRP conjugate diluted 1:1000. The binding to PS of each reagent in the absence of a competitor was taken as the 100% value.
Detection of Externally Positioned Anionic Phospholipids by 9D2 Antibody and Annexin V in Tumor-bearing Mice in Vivo.
Immunohistochemical techniques, in which 9D2 or annexin V are applied directly to sections of frozen tissues, do not discriminate between anionic phospholipids on the inner leaflet and the outer leaflet of the plasma membrane. To detect externally positioned phospholipids, 50 μg of 9D2 or biotinylated 9D2 antibody, or 100 μg of biotinylated annexin V were injected i.v. into tumor-bearing SCID mice. Sixty min later mice were sacrificed, and their blood circulation was exsanguinated and perfused with heparinized saline as described previously (34) . All of the major organs and tumor were removed and snap-frozen for preparation of cryosections. Sections were blocked with PBS containing 10% serum. To prevent loss of phospholipids during slide processing, detergents and organic solvents were omitted from blocking and washing buffers. Rat IgM was detected using goat antirat IgM (μ-specific) HRP conjugate followed by development with carbazole or DAB (35) . This procedure was successfully used to detect PS-positive vessels in L540Cy tumors by using the mouse anti-PS antibody 3SB (36) . Biotinylated reagents were detected by streptavidin conjugated to HRP. Tumor sections derived from mice injected with saline or rat IgM of irrelevant specificity served as negative controls. Additional controls consisted of incubating the slides in 1% Triton solution or in acetone for 10 min. These treatments extract phospholipids. No signal was detected under these conditions. Staining of the sections by this method for the presence of 9D2 or annexin V detects cells having externalized anionic phospholipids that were accessible for binding by the reagents in vivo.
Identification and Quantification of PS-positive Tumor Vessels.
Structures with localized 9D2 antibody or annexin V were identified as blood vessels by morphological appearance on DAB-stained sections and by coincident staining with the pan-endothelial cell marker, MECA 32, on serial sections of frozen tissues. Quantification on DAB-stained sections was done by counting vessels stained by MECA 32, 9D2, or annexin V in serial sections of a tumor. Six slides of each tumor type derived from 6 mice injected with 9D2 antibody, control rat IgM, or annexin V were examined. At least 10 random fields/section (0.317 mm2/field) were scored in blinded fashion by two independent observers. The mean numbers and SEs of vessels stained by 9D2, annexin V, or MECA 32 were calculated. The mean number of 9D2 or annexin V-positive vessels determined in each tumor type group was compared with the mean number of MECA 32-positive vessels in the same tumor group. The percentage of 9D2 or annexin V-positive vessels was calculated.
In additional experiments, mice bearing MDA-MB-231 tumors (0.3–0.7 cm3 in volume) were i.v. injected with 50 μg of biotinylated 9D2, control IgM, or annexin V (6 mice per group). Biotinylated reagents were first incubated with streptavidin-Cy3 conjugate, washed in PBS, then incubated with MECA 32 antibody followed by FITC-tagged antirat IgG secondary antibody. Single images, taken with appropriate filters for Cy3 (red) and FITC (green) fluorescence, respectively, were captured by digital camera and transferred to a computer. Images of 10 random fields (0.317 mm2/field) demonstrating yellow color (a product of merged green and red fluorescence) were superimposed with the aid of Metaview software. The same method was used to analyze tumors from mice injected with control rat IgM or saline. The percentage of vessels with localized 9D2 or annexin V was calculated as follows: mean number of yellow vessels per field divided by mean number of green (total) vessels multiplied by 100.
Iodination of Annexin V.
Recombinant human annexin V was purified from Escherichia coli transformed with ET12a-panionic phospholipid1 plasmid (a gift from Dr. Jonathan Tait, University of Washington, Seattle, WA). The purity of the protein and the binding to PS were confirmed on SDS-PAGE and on PS-coated plastic, respectively. Rabbit polyclonal, affinity-purified antiannexin V antibodies were used to detect annexin V bound to PS. Annexin V was radiolabeled with 125I using chloramine T as described by Bocci (37) . The specific activity was ∼1 × 106 cpm/μg of protein, as measured by a Bradford assay (38) .
Detection of Exposed PS on Cultured Endothelial Cells by 125I-labeled Annexin V.
After treatment with the reagents described above, treated and control cells were incubated with 7.1 pmols of 125I-labeled annexin V (200 μl/well) in the binding buffer. After 2 h incubation at room temperature, cells were washed extensively and dissolved in 0.5 m of NaOH. The entire volume of 0.5 ml was transferred to plastic tubes and counted in a gamma counter. Nonspecific binding was determined in the presence of 5 mm EDTA and was subtracted from experimental values. The results were expressed as net pmols of cell-bound annexin V, normalized per 1 × 106 cells. Maximal binding was determined on cells simultaneously treated with actinomycin D and TNF-α (50 ng/ml of each component). As has been reported previously, the above agents cause apoptosis and PS exposure in 90–100% of endothelial cells (39) . Basal binding of 125I-annexin V to untreated cells was determined in the presence of medium with 10% serum. The amount of 125I-annexin V that bound to the untreated cultures was subtracted from that in the treated cultures. The specific increase in the amount of externalized PS was calculated according to the following formula: (net experimental binding/net maximal binding) × 100. Each experiment was performed in duplicate and was performed at least three times.
Effect of Growth Factors, Cytokines, Inflammatory Mediators, Hydrogen Peroxide, Hypoxia, and Exposure to Low pH on Exposure of PS in Cultured Endothelial Cells.
Endothelial cells were treated with cytokines or growth factors at concentrations listed in Table 3 ⇓ . All of the reagents were diluted in medium containing 10% serum and incubated with the cells at 37°C for 24 h. To study the effect of hypoxia, cells were seeded on 24-well plates and were incubated in a humidified normoxic atmosphere (21% O2, 5% CO2) for 48 h before being transferred to a humidified hypoxic atmosphere (1% O2, 5% CO2, 94% N2) in a sealed chamber (Billups Rothenberg Inc., Del Mar, CA). Cells were incubated in a hypoxic chamber for 24 h at 37°C and were then returned to a normoxic environment for 4 h at 37°C. The cells were compared with a parallel culture from an identical passage, seeded on the same day, and maintained entirely under normoxic conditions. To examine the effect of an acidic microenvironment, cells were exposed to the growth medium lacking bicarbonate, which was adjusted to different pHs (ranging between 7.3 and 6.2) with HCl. Cells were incubated at 37°C in the absence of CO2. It was confirmed that culture medium held the assigned pH during the 24-h period of culture. These experimental conditions were not toxic to either bovine or mouse endothelial cells, and had no effect on cell morphology or viability of the attached monolayer.
Phospholipid Specificity of 9D2 Antibody and Annexin V.
9D2 antibody specifically recognized anionic phospholipids (PS, PA, CL, PI, and PG) and had no significant reactivity with neutral phospholipids (PE, PC, and SM) in ELISA (Fig. 1 ⇓ ; Table 1 ⇓ ). The order of strength of binding of 9D2 to phospholipids in ELISA was PA>PS = CL>PG = PI. The binding was antigen-specific because no binding was observed with several control rat IgM of irrelevant specificity. Also, binding of 9D2 to any of the anionic phospholipids adsorbed to ELISA plates was blocked by liposomes prepared from any of the anionic phospholipids but not by liposomes prepared from any of the neutral phospholipids. Annexin V also bound to anionic phospholipids but its binding was less specific than that of 9D2, in that it also bound strongly to the neutral phospholipid, PE. The order of strength of binding of annexin V to phospholipids in ELISA was PI>PS = PE = PA = CL>PG (Table 1) ⇓ . These findings for annexin V are consistent with earlier data (40) . Neither 9D2 nor annexin V bound detectably to heparin, heparan sulfate, or to double- or single-stranded DNA (data not shown). The binding of 9D2 was unaffected by the presence of 5 mm EDTA, showing it did not require Ca2+ for binding to anionic phospholipids. In contrast, the binding of annexin V to anionic phospholipids was abolished in the presence of 5 mm EDTA, as expected from its known dependence on Ca2+ for binding to anionic phospholipids or PE (41 , 42) . Neither 9D2 nor annexin V bound to ELISA plates that had been coated with phospholipids but then washed with 0.2% Tween in saline, confirming that their binding was to the absorbed phospholipids.
9D2 Antibody and Annexin V Do Not Cross-Block Each Other’s Binding to PS.
To examine whether 9D2 antibody and annexin V compete for binding to PS, cross-blocking experiments were performed using biotinylated proteins on PS-coated plates. Binding of biotinylated 9D2 antibody and annexin V was blocked by a 10-fold molar excess of unmodified 9D2 and annexin V, respectively (Table 2 ⇓ ). However, unmodified annexin V did not affect the ability of biotinylated 9D2 to bind to the PS plate. Likewise, addition of unmodified 9D2 antibody did not alter the ability of biotinylated annexin V to bind to the PS plate (Table 2) ⇓ . These results indicate that 9D2 antibody and annexin V do not cross-block each other binding to PS-coated plates, either because they recognize different epitopes on the PS molecule or different conformations of PS adsorbed on plastic.
9D2 Antibody and Annexin V Recognize Externalized Anionic Phospholipids on Cell Surfaces.
The binding of 9D2 antibody and annexin V to cell surfaces was examined using mouse bEnd.3 endothelioma cells or bovine ABAE cells. Neither 9D2 nor annexin V bound to nonpermeabilized monolayers of either cell type under quiescent conditions. This indicates that the majority of anionic phospholipids of the plasma membrane are normally sequestered to the cytosolic domain. In contrast, strong staining was observed when cells were preincubated with TNF-α and actinomycin D under conditions that caused apoptosis in 90–100% of the endothelial cells.
To confirm that 9D2 and annexin V were binding to phospholipids on cell surfaces, H2O2-treated bEnd.3 cells were incubated with 9D2 antibody or annexin V in the presence or absence of various competing liposomes. We had determined previously that anionic phospholipids become exposed on nonapoptotic, viable bEnd.3 cells when they are pretreated with a subtoxic concentration (100–200 μm) of H2O2 (43) . The binding of 9D2 antibody to H2O2-treated bEnd.3 cells was inhibited by liposomes containing anionic phospholipids but not by liposomes containing neutral phospholipids (Fig. 2) ⇓ . The magnitude of inhibition of 9D2 binding to cells varied in the order PA>PS>CL>PG>PI, in close agreement with the results obtained using plastic-immobilized phospholipids (Fig. 1) ⇓ . Similarly, the binding of annexin V to H2O2-treated cells was blocked by liposomes containing PS, PA, PE, CL, and, to a lesser extent, PI and PG. Liposomes containing SM or PC did not block annexin V binding to cells, all in agreement with the results obtained using plastic-immobilized phospholipids. These results confirm that 9D2 binds to anionic phospholipids in the H2O2-treated endothelial cells, whereas annexin V binds to PE in addition to anionic phospholipids.
Localization of 9D2 Antibody and Annexin V to Tumor Vessels in Mice.
The ability of 9D2 antibody and annexin V to localize to tumor vessels in mice was first determined by indirect immunohistochemistry. Mice bearing various types of solid tumors were injected i.v. with 9D2 antibody or biotinylated annexin V, and 1 h later, were exsanguinated, and the tumors and normal tissues were removed. The tumors were: human MDA-MB-231 breast tumor growing orthotopically in the mammary fat pads of SCID mice; human L540 Hodgkin’s tumor growing s.c.; human NCI-H358 NSCLC growing s.c.; mouse B16 melanoma growing s.c., and mouse Meth A fibrosarcoma growing s.c. Frozen sections of tissues were cut and stained with HRP-labeled antirat IgM or with HRP-labeled streptavidin to determine to which cells the 9D2 and annexin V had bound after injection. Blood vessels were identified morphologically and from their positive staining by the pan-endothelial cell antibody, MECA 32, on serial sections.
9D2 antibody and annexin V localized to tumor vessels in all of five tumors included in this study (Table 3 ⇓ ; Fig. 3 ⇓ ). 9D2 antibody and annexin V gave essentially the same patterns of staining. Vascular endothelium in the tumors showed distinct membrane staining (Fig. 4) ⇓ . The percentage of 9D2 and annexin V-positive vessels ranged from 40% in MDA-MB-231 tumors to 15% in NCI-H358 tumors relative to the number of MECA 32-positive vessels (Table 3) ⇓ . Anionic phospholipid-positive vessels were present on the luminal surface of capillaries and vessels in all regions of the tumors, but were particularly prevalent in and around regions of necrosis. Most anionic phospholipid-positive vessels did not show morphological abnormalities that were apparent by light microscopy. Occasional vessels, particularly those located in necrotic areas, showed morphological signs of deterioration. Localization of the 9D2 antibody to tumor vessels appeared to be specific because membrane staining of tumor endothelium was not observed in tumors from mice that had been injected with rat IgM of irrelevant specificity. Presumably, leakage of the control rat IgM out of tumor vessels occurred to some extent, but the staining of extravascular IgM was too diffuse or too weak to discern by indirect immunohistochemistry. 9D2 antibody and annexin V also localized to the membrane and cytosol of necrotic and apoptotic tumor cells, whereas localization of the control IgM was not detectable (Fig. 3) ⇓ .
No vascular localization of 9D2 antibody or annexin V was observed in 9 of the 10 normal organs that were examined (Table 3) ⇓ . In the kidney, staining of tubules was observed that appeared not to be antigen specific. Tubules were stained in both 9D2 and control rat IgM recipients, presumably because of secretion of IgM or its metabolites through this organ. The ovaries, a site of physiological angiogenesis, were not examined.
Double-staining experiments were also performed in which mice bearing orthotopic MDA-MB-231 breast tumors were injected i.v. with biotinylated 9D2 antibody, biotinylated control IgM, or biotinylated annexin V. One h later, the mice were exsanguinated, their tumors were removed, and frozen sections were cut. The tumor sections were then stained with Cy3-conjugated streptavidin to detect the biotinylated proteins and with FITC-conjugated MECA32 to detect vascular endothelium. This detection method labeled the biotinylated proteins and the vascular endothelium by red and green, respectively. Where the biotinylated proteins are bound to the endothelium, the converged image appears yellow. The biotinylated 9D2 and annexin V appeared mostly to be bound to the vascular endothelium, because their staining patterns converged with that of MECA 32 (Fig. 4, A and B) ⇓ . About 40% of MECA 32-positive vessels bound 9D2 and annexin V, in close agreement with the results obtained by indirect immunohistochemistry. However, leakage of the biotinylated proteins into the tumor interstitium was apparent by double staining, whereas it was not apparent by indirect immunohistochemistry. Biotinylated proteins were visible outside the vascular endothelium around a minority (∼5%) of vessels. In tumors from mice that had been injected with biotinylated rat IgM of irrelevant specificity, the biotinylated IgM had also leaked into the tumor interstitium around a similar percentage (∼5%) of vessels, but mostly appeared not to be bound by the vascular endothelium. (Fig. 4C) ⇓ . Presumably, the detection of extravasated 9D2 and annexin V by the double-staining technique but not by the indirect immunohistochemistry technique reflects the greater sensitivity of the former technique and the greater precision with which two staining patterns can be compared. Noninjected control tumors were completely unstained by streptavidin-Cy3 (Fig. 4D) ⇓ , indicating that red fluorescence corresponds to a localized protein.
Perturbation of Phospholipid Asymmetry on Tumor Vessels Might Be Induced by Oxidative Stress.
Mouse bEnd.3 or bovine ABAE cells in vitro were treated for 24 h with various concentrations of factors and conditions that are present in the microenvironment of many tumors (44 , 45) . Externalization of anionic phospholipids was quantified by measuring 125I-annexin V binding. The amount of annexin V binding was compared with that of cells in which apoptosis of 90–100% of cells had been induced by combined treatment with actinomycin D and TNF-α. Actinomycin D and TNF-α induced the binding of 6.2 pmols of annexin V per 106 cells (3.8 × 106 molecules of annexin V per cell) on both cell types, in good agreement with literature reports (46) . This value was taken as the maximal level of externalized anionic phospholipids.
Untreated cells were largely devoid of externalized PS, as judged by annexin V or 9D2 binding (Table 4) ⇓ . The basal binding in the presence of growth medium alone was 0.44 and 0.68 pmols of 125I-annexin V for ABAE and bEnd.3 cells, respectively. This corresponds to 7.06% and 10.9% of the maximal binding for ABAE and bEnd.3 cells, respectively, which correlated well with the finding that ∼10% of cells bound biotinylated annexin V under the same conditions.
VEGF, hepatocyte growth factor, fibroblast growth factor, TGF-β1, platelet-derived growth factor, IL-6, IL-8, and IL-10 did not increase binding of 125I-annexin V above the basal level for untreated cells. Inflammatory mediators (IL-1α, IL-1β, TNF-α, and IFN) caused a small but reproducible increase in anionic phospholipid translocation that ranged from 5% to 8% of the maximal level for ABAE cells and from 3% to 14% for bEnd.3 cells (Table 4) ⇓ . Hypoxia/reoxygenation, thrombin, or acidic external conditions (pH 6.8–6.6) induced a moderately high externalization of anionic phospholipid that ranged from 8% to 20% of the maximal level for ABAE cells and from 17% to 22% of the maximal level for bend.3 cells. The largest increase in anionic phospholipid translocation was observed after treatment with 100–200 μm of hydrogen peroxide. This treatment caused nearly complete (95%) externalization of anionic phospholipid in both cell types as judged by 125I-annexin V binding (Table 4) ⇓ . More than 70% of ABAE and bEnd.3 cells bound biotinylated annexin V, as judged immunohistochemically (data not shown). Endothelial cells in which anionic phospholipid translocation was generated by treatment with hypoxia/reoxygenation, thrombin, acidity, TNF-α, IL-1, or H2O2 remained attached to the matrix during time period of the assay (24 h), retained cell-cell contact, and retained their ability to exclude trypan blue dye. Normal anionic phospholipid orientation was restored 24–48 h later in the majority of the cells after the inducing factor was removed or the culture conditions were returned to normal. These results indicate that hypoxia/reoxygenation, acidity, thrombin, and inflammatory cytokines all trigger a transient translocation of anionic phospholipid on viable endothelial cells.
Combined Effects of Inflammatory Cytokines and Hypoxia/Reoxygenation on Anionic Phospholipid Exposure by Endothelial Cells in Vitro.
Enhanced anionic phospholipid exposure was observed when ABAE cells were subjected to hypoxia/reoxygenation in the presence of IL-1α or TNF-α. In the absence of the cytokines, hypoxia/reoxygenation conditions increased PS exposure to 15% of the maximum level for cells treated with apoptotic concentrations of actinomycin D and TNF-α. In the presence of subtoxic concentrations of IL-1α or TNF-α, hypoxia/reoxygenation increased anionic phospholipid exposure to 26% and 33%, respectively, of the maximum (Fig. 5) ⇓ . Cytokines in the absence of hypoxia/reoxygenation increased annexin V-binding sites by <7% indicating that the combination of cytokines and hypoxia/reoxygenation had greater than additive effects on PS-exposure. Thus, in tumors, the exposure of anionic phospholipids induced by hypoxia/reoxygenation may be amplified by inflammatory cytokines and possibly by such other stimuli as acidity and thrombin.
The major finding to emerge from this study is that anionic phospholipids are exposed on the surface of tumor endothelium. This phenomenon was demonstrated using two independent reagents that bind selectively to anionic phospholipids: a monoclonal antibody, 9D2, and annexin V.
9D2 antibody and annexin V bound with high affinity and specificity to anionic phospholipids adsorbed to plastic, as liposomes, or presented on the membrane surface of activated or apoptotic endothelial cells in vitro. 9D2 bound strongly to PS, PA, and CL but more weakly to PI and PG. Annexin V bound to PE in addition to PS, CL, PA, PI, and PG, as found previously by others (40, 41, 42 , 47) . Recognition of anionic phospholipids by 9D2 antibody was identical in the presence and absence of serum, indicating that binding does not require serum cofactors. Binding of 9D2 to anionic phospholipids did not require Ca2+ ions, whereas the binding of annexin V required Ca2+. Cross-blocking experiments on PS-coated plates showed that 9D2 and annexin V do not block each other’s binding to PS. This indicates that the two reagents recognize different epitopes on the PS molecule, or, more likely, differently packed forms of PS. Annexin V is thought to bind to planar PS surfaces, whereas anti-PS antibodies are thought to bind to hexagonally packed PS (48) . Both forms are probably present on PS-coated plates.
9D2 antibody and annexin V specifically localized to tumor vessels, and to tumor cells in and around necrotic regions of tumors, after i.v. injection into tumor-bearing mice (Fig. 3) ⇓ . Between 15% and 40% of blood vessels (Table 3) ⇓ in all five types of tumors that we examined in vivo had anionic phospholipid-positive endothelium. In contrast, none of the blood vessels in normal tissues had detectable externalized anionic phospholipids. The specificity of staining of tumor endothelium by 9D2 was demonstrated by: (a) coincidence of staining by 9D2 and the pan-endothelial antibody, MECA 32, as detected by indirect immunohistochemistry and double-staining techniques (Figs. 3 ⇓ and 4 ⇓ ); (b) the weaker and more infrequent staining of vascular endothelium in tumors by control rat IgM; (c) the finding that extraction of phospholipids from tumor sections with detergents or organic solvents abolished staining; (d) the lack of localization of 9D2 or annexin V to the quiescent endothelium in normal organs; and (e) the blocking of 9D2 or annexin V binding to H2O2-treated endothelial cells in vitro by liposomes prepared from anionic phospholipids but not neutral phospholipids.
The main anionic phospholipid that is localized by 9D2 or annexin V on tumor vasculature is likely to be PS. This is the most abundant anionic phospholipid, and its exposure on the cell surface is known to be regulated by environmental influences or injury. We cannot exclude the possibility that other anionic phospholipids (e.g., PI, PA, and PG) are also exposed, but they are less abundant and their membrane position is not as tightly regulated by environmental conditions. However, it is possible that the major neutral phospholipid, PE, is also exposed on tumor endothelium and contributes, together with PS, to the annexin localization that we observe on tumor vessels. The position of PE in the plasma membrane is regulated in a similar manner to PS. PE is segregated to the internal leaflet of the plasma membrane in part by aminophospholipid translocase, although at a slower rate than PS (49) , and is transported to the external surface by scramblase (50) . Recent work has shown that PE, like PS, is exposed during apoptosis and cell activation (51) .
To examine the mechanism of exposure of anionic phospholipids on tumor endothelial cells, a series of experiments was performed in which endothelial cells in vitro were treated with various factors and conditions known to be present in the tumor microenvironment. Hypoxia followed by reoxygenation, acidity, and thrombin increased PS exposure on viable endothelial cells to between 10% and 22% of the level seen when all of the cells are apoptotic (Table 4) ⇓ . Inflammatory cytokines (TNF-α and IL-1) also caused a weak but definite induction of PS exposure. Our findings (Fig. 5) ⇓ are consistent with the possibility that, in tumors, exposure of anionic phospholipids on the vascular endothelium is induced by hypoxia/reoxygenation in combination with inflammatory cytokines, thrombin, and acidity. ROS may be generated by tumor cells as a byproduct of metabolism or in response to hypoxia (30) . Cytokines released by tumor cells may induce leukocyte adhesion molecules on the endothelium that mediate adherence of activated macrophages, polymorphonuclear cells, and platelets to tumor endothelium and additional secretion of ROS. The ROS may then induce PS translocation through oxidation of thiol-containing transport molecules or peroxidation of lipids (52) , possibly by causing an influx of Ca2+ or release of Ca2+ from intracellular stores (53) . It is also possible that exposure of anionic phospholipids occurs during certain stages of cell division and contributes to the staining of the proliferating endothelium in tumors: the relationship between angiogenesis and exposure of anionic phospholipids was not examined in the present study.
Exposure of PS and other anionic phospholipids may in part explain the procoagulant status of tumor endothelium that has long been recognized (26) . The anionic phospholipids would provide the surface on which coagulation factors concentrate and assemble (24 , 25) . It also would provide an attachment site for circulating macrophages (21) , T lymphocytes (17) , and polymorphonuclear cells that assists in leukocyte infiltration into tumors.
Antibodies, annexins, and other ligands that bind to anionic phospholipids might be used for the targeting or imaging of tumor blood vessels. Anionic phospholipids are attractive as tumor vessel targets for several reasons: they are abundant (PS is present at >106 molecules per cell); they are on the luminal surface of tumor endothelium, which is directly accessible for binding by vascular targeting agents in the blood; they are present on a significant percentage of tumor endothelial cells in diverse solid tumors; and they appear to be absent from endothelium in all of the normal tissues. However, PS may be exposed on vascular endothelium in nonmalignant lesions (e.g., sites of inflammation), where cytokines, hypoxia, and ROS might induce PS translocation. It is possible this could lead to toxicity with a vascular targeting strategy, making it necessary to exclude patients with these conditions from treatment. PS is also present on cells undergoing physiological apoptosis, but, because these cells are destined to die, toxicity through targeting these cells is of lesser concern. PS is also found on the surface of apoptotic vesicles, such as are shed from tumors. These could bind to anti-PS antibodies in the blood circulation or diminish localization to tumor vasculature; nevertheless, access of the antibody to tumor endothelium does not appear to be a problem in the present studies performed in mice.
In conclusion, anionic phospholipids on tumor vessels may provide target molecules for tumor therapy. In addition, anionic phospholipids exposed on apoptotic and living tumor cells may contribute to imaging intensity. Annexin V and antibodies to anionic phospholipids might be used to deliver a cytotoxic drug, radionuclide, or coagulant to tumor vessels for the vascular targeting of tumor vessels in humans.
We thank Linda Watkins and Maria Sambade for technical assistance, and Samia Burns for help in preparing this manuscript.
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.
↵1 Supported in part by NIH Grants 1RO1CA74951-01, 5RO1CA54168-05, Lung Cancer Special Program of Research Excellence (SPORE) P50-CA70907, Susan G. Komen Breast Cancer Foundation Grant BCTR-2000-269, and a Sponsored Research Agreement with Peregrine Pharmaceuticals, Inc.
↵2 To whom requests for reprints should be addressed, at University of Texas Southwestern Medical Center, 2201 Inwood Road, NC7.304, Dallas, TX, 75390-8594. Phone: (214) 648-1499; Fax: (214) 648-1613; E-mail:
↵3 The abbreviations used are: PS, phosphatidylserine; PI, phosphatidylinositol; PA, phosphatidic acid; PG, phosphatidylglycerol; CL, cardiolipin; PE, phosphatidylethanolamine; PC, phosphatidylcholine; SM, sphingomyelin; ROS, reactive oxygen species; IL, interleukin; TNF, tumor necrosis factor; DPBS, Dulbecco’s phosphate buffered saline; HRP; horseradish peroxidase; SCID, severe combined immunodeficient; VEGF, vascular endothelial growth factor; ABAE, adult bovine aortic endothelial; DAB, 3,3′-diaminobenzidine.
- Received May 30, 2002.
- Accepted September 6, 2002.
- ©2002 American Association for Cancer Research.