Inhibition of Angiogenesis and Vascular Leakiness by Angiopoietin-Related Protein 4¹ Free

Blood vessel formation, which is required for embryogenesis, also contributes to several pathophysiological conditions such as neovascularization in tumorigenesis and inflammatory and ischemic disease in adulthood. Recent studies show that many critical growth factors regulate blood vessel formation and that the actions of these factors are carefully orchestrated in terms of time, space, and dose to form a vascular network (1, 2). Among them, endothelial-restricted receptors and their ligands are critical for such events. In particular, receptor tyrosine kinases, such as VEGF⁴ receptors and Tie receptors, play pivotal roles in blood vessel formation in both embryogenesis and pathophysiological conditions (3, 4). Tie2 ligands, angiopoietin-1, -2, -3, and -4, regulate both dynamic and quiescent states of endothelial cells through competitive interaction. Among the angiopoietins, angiopoietin-1 is critical for modulation of angiogenesis, playing complementary and coordinated roles with VEGF in vascular development (4).

To understand the roles of angiopoietin family proteins in the vascular system, we cloned four structurally homologous proteins containing characteristic coiled-coil domains within the NH₂ terminus and a fibrinogen-like domain in the COOH terminus. Among them, ARP4 (human ARP4, GenBank accession no. AB056477; mouse ARP4, GenBank accession no. AB054540) was identical to FIAF/PGAR, which has recently been found to be a PPAR target gene (5, 6). However, the role of ARP4 in angiogenesis has not been fully elucidated. Several lines of recent evidence indicate that ARPs also play important roles in regulating angiogenesis. For example, ANGPTL3, a member of the ARP family, exhibits biological activity as an angiogenic factor (7, 8). In contrast, ARP1/angioarrestin inhibits proliferation, migration, and tube formation of endothelial cells and significantly reduces HT1080 tumor nodule formation (9). Furthermore, recent reports indicate that ARP4/FIAF/PGAR expression is lost in two gastric cancer cell lines and reduced in primary gastric cancers (10). Therefore, we hypothesized that ARP4 could play important roles in modulating angiogenesis in tumorigenesis.

In this study, we show that ARP4 inhibits proliferation, migration, and tubule formation of endothelial cells and reduces vascular leakage. These conditions are found in multiple abnormalities of tumor vessels, and they influence neovascularization, which is required for both primary tumor growth and metastatic growth. Taken together, our findings indicate that ARP4 is a novel negative angiogenic regulator that may be useful in treating neovascular diseases such as cancer.

HUVECs (BioWhittaker, Inc., Walkersville, MD) and HDMVECs (BioWhittaker, Inc.) were cultured in EGM-2 medium (Clonetics, San Diego, CA). bEND3 (11), mouse brain capillary endothelial cells, were cultured in DMEM (Life Technologies, Inc., Gaithersburg, MD) supplemented with 10% FBS (JRH Biosciences, Lenexa, KS).

The coding region of ARP4 fused at the COOH terminus to the FLAG epitope was subcloned into pCEP4 (Invitrogen, Groningen, the Netherlands). HEK293 cells were cultured at 37°C in humidified 5% CO₂/air in DMEM supplemented with 10% FBS and transfected with pCEP4-ARP4-FLAG using FuGENE 6 (Roche Diagnostics, Mannheim, Germany). After transfection, HEK293 cells were selected in 300 μg/ml Hygromycin B (Life Technologies, Inc.) for 5 days, and the conditioned medium was collected and filtered with a 0.22-μm pore size filter (Millipore Co., Bedford, MA). To purify ARP4-FLAG fusion protein, it was transferred to an anti-FLAG antibody (M2) affinity gel (Sigma, St. Louis, MO). After washing the gel with PBS, protein was eluted by adding Gly-HCl (pH 3.0) and immediately neutralized with Tris-HCl (pH 8.0). The protein was visualized by SDS-PAGE with Coomassie Brilliant Blue Staining (Wako, Osaka, Japan). Western blot was performed using horseradish peroxidase-conjugated anti-FLAG antibody (M2; 1:500; Sigma). Development of peroxidase activity was detected using the enhanced chemiluminescence detection system (ECL; Amersham Biosciences Corp., Piscataway, NJ).

HUVECs and HDMVECs were cultured on collagen-coated 96-well plates (Becton Dickinson Labware, Bedford, MA) at a density of 5 × 10³ cells/well in 100 μl of EBM-2 (Clonetics) containing 0.5% FBS. After 24-h starvation at 37°C, the cells were washed twice with serum-free medium and incubated in fresh medium containing 0.5% FBS and VEGF (10 ng/ml; Pepro Tech EC Inc., London, United Kingdom) with or without various concentrations of ARP4. The BrdUrd incorporation assay was performed according to the manufacturer’s instructions (Roche Diagnostics). Cells were incubated with BrdUrd for 24 h. Each experiment was performed in triplicate and repeated three times.

A migration assay was performed by using Transwell polycarbonate membrane filters with 8.0-μm pore size (Costar, Cambridge, MA; Ref. 12). The lower surface of the membranes was precoated with 0.1% gelatin and dried overnight at room temperature. The coated membranes were washed in PBS and dried. EBM-2 containing 0.5% FBS with or without the indicated factors was loaded in the lower wells, and 3 × 10⁴ HUVECs, which were serum-starved for 24 h and suspended in 100 μl of EBM-2 medium, were inoculated onto each upper well. DMEM was used instead of EBM-2 medium for bEND3 cells (3 × 10⁴). After a 4-h incubation at 37°C, the membranes were fixed with 100% methanol and stained with Giemsa solution. The cells on the upper surface of the membranes were removed with cotton swabs. Cells migrating to the lower surface of the membrane were counted in three independent fields under a microscope. The number of cells counted in untreated samples was represented as 100%. Each experiment was performed in triplicate and repeated three times.

Ninety-six-well culture dishes were coated with 80 μl/well Matrigel (Becton Dickinson) on ice, and Matrigel was allowed to solidify for 30 min at 37°C. HUVECs (1 × 10⁴) in 100 μl of EGM-2 containing 2% FBS and 10 ng/ml VEGF and 10 ng/ml basic fibroblast growth factor (Pepro Tech EC Inc.) with or without various concentration of ARP4 per well were seeded onto each Matrigel-coated well. After a 24-h incubation at 37°C, endothelial tubule formation was photographed under phase-contrast microscopy. To quantify the length of tubular structures, tubule length was measured with NIH Image software. Tubules shorter than 100 μm were excluded from the measurement. Each experiment was performed in triplicate and repeated three times.

The corneal assay was performed as described previously (13, 14). In brief, under sterile conditions, slow-release pellets were prepared incorporating VEGF (150 ng) alone, ARP4 (1 μg) alone, and both VEGF (150 ng) and ARP4 (1 μg) in a casting solution of an ethynil-vinyl copolymer (Elvax-40; DuPont, Wilmington, DE) in 10% methylene chloride. After anesthesia with sodium pentobarbital (Dainippon Pharmaceutical Co., Ltd., Osaka, Japan), the pellet was implanted into the corneal micropocket of male 8-week-old C57BL/6 mice (SLC, Shizuoka, Japan). Erythromycin ophthalmic ointment was applied to each eye after implantation of the pellet. The eyes were examined by a slit-lamp biomicroscope (Kowa Co., Ltd., Nagoya, Japan) each day after pellet implantation. The neovascular response was measured as the maximal vessel length from limbal vasculature toward the pellet (R₁; in mm) and the contiguous circumferential zone of clock-hours of neovascularization (R₂; in mm). The neovascular area was calculated by the following formula: area (in mm²) = 0.2 × π × R₁ × R₂ (15). Each experiment was performed in triplicate and repeated three times.

The vascular permeability assay was performed as described previously (16). Male 8-week-old BALB/c mice (SLC) were anesthetized, and 100 μl of 1% Evans Blue dye (Wako) were injected into tail vein. Five min after injection, VEGF (50 ng) alone, a combination of ARP4 (1 μg) and VEGF (50 ng) in 20 μl of PBS, or PBS alone was injected intradermally at adjacent locations on the back. Extravasation of Evans Blue dye was recorded at 15 and 30 min from the underside of the back skin surrounding the injected site. Thirty min after intradermal injection, the mouse was perfused from the left ventricle with 1% paraformaldehyde in citrate buffer at pH 3.5. Skin surrounding the injected site was removed in the same square, gently blotted, and weighed, and Evans Blue dye was extracted from skin with 1 ml of formamide. The amount of extravasated dye was measured with a spectrophotometer (610 nm) and expressed as the content of dye per 1 mg wet weight of tissue. Each experiment was performed in triplicate and repeated three times.

To generate Tg mice that express ARP4 under control of the K14 promoter (K14-ARP4 Tg mice), a transgene was generated by inserting an XbaI fragment containing mouse ARP4 cDNA at the XbaI site of the expression vector under the regulation of the human K14 promoter (17). DNA containing the transgene was prepared for pronuclear injections as described previously (18). A total of 250 pronuclear injections in C57BL/6 mice were performed, resulting in 27 pups. PCR analysis of genomic DNA was performed with forward (5′-GCTCCTGGGCAACGTGCTGG-3′) and reverse (5′-TTGGGAGTCAAGCCAATGAG-3′) primers in a reaction consisting of 5 min at 95°C, followed by 30 cycles of denaturing at 95°C for 60 s, annealing at 60°C for 60 s, and extension at 72°C for 60 s. A final 7-min extension was done at 72°C. Northern blot analysis was performed using [α-³²P]dCTP-labeled mouse ARP4 cDNA probes generated by random priming using a BcaBEST Labeling Kit (Takara Shuzo Co., Ltd., Shiga, Japan). Membranes blotted with 20 μg of total RNA extracted from mouse skin using TRIzol (Life Technologies, Inc.) were used. The mice were housed in environmentally controlled rooms of the Laboratory Animal Research Center of Keio University under the guidelines of Keio University for animal and recombinant DNA experiments. To detect ARP4 protein in sections, we prepared antimouse ARP4 polyclonal antibodies that were produced by immunizing rabbits with a synthetic peptide corresponding to amino acids 84–97 of mouse ARP4 (KGKDAPFKDSEDRV). Fixed sections from back skin in K14-ARP4 Tg mice were stained with a 1:200 diluted antimouse ARP4 antibody.

Lectin staining of whole mount of ear skin with fluorescein-labeled Lycopersicon esculentum lectin (Vector Laboratories, Burlingame, CA) was performed as described elsewhere (16). Stained samples were analyzed using a fluorescence microscope (Olympus, Tokyo, Japan), and vascular density of the samples was assessed by the Chalkley Grid Method (19).

CMT93 murine colorectal tumor cells were cultured in DMEM supplemented with 10% FBS. Cells were trypsinized, washed with PBS, and resuspended in 300 μl of Matrigel at 1 × 10⁵ cells. Mice were inoculated intradermally in the flank. Five weeks after implantation, mice were sacrificed, and tumors were removed. Tumor volumes were estimated by the formula V = π/6 × a² x b, where the short diameter of tumor is represented as a, and the long diameter of the tumor is represented as b. Tumors were then prepared for immunostaining with a rat antimouse PECAM-1 monoclonal antibody to evaluate capillary formation. The secondary antibody was a horseradish peroxidase-conjugated antirat immunoglobulin antibody (Biosource, Camarillo, CA). Five random photographs per tumor were taken to quantify the tumor capillary vessel areas using NIH Image software.

Statistical evaluation of data was conducted using Student’s t test for per-comparison analysis.

To analyze the biological function of ARP4, we prepared recombinant protein from the conditioned medium of HEK293 cells transfected with ARP4 cDNA. SDS-PAGE and Western blot analysis revealed that a major mouse and human ARP4 product greater than 50 kDa was detected in the culture medium and present at a low levels in the cell lysate. Under nonreducing conditions (without 2-mercaptoethanol), ARP4 oligomerized expectedly, as do other proteins containing coiled-coil motifs such as angiopoietins (Fig. 1; data not shown). These results indicate that ARP4 is a secreted and oligomerized protein, as has been reported previously for the angiopoietin family (20, 21, 22).

VEGF-induced angiogenesis and vascular leakiness occurs in the process of tumorigenesis. Therefore, we analyzed the role of ARP4 signaling in the vasculature by asking whether ARP4 affected VEGF-induced proliferation of HUVECs and HDMVECs. Addition of VEGF increased BrdUrd incorporation in HUVECs and HDMVECs, whereas VEGF-induced BrdUrd incorporation was inhibited in a dose-dependent manner by the addition of ARP4 (Fig. 2).

Next, we investigated whether ARP4 affected VEGF-induced chemotaxis of HUVECs and bEND3 cells. VEGF promoted migration of HUVECs and bEND3 cells into a microchemotaxis membrane. Such VEGF-induced chemotaxis was inhibited by the addition of ARP4 in a dose-dependent manner (Fig. 3). Treatment with ARP4 alone did not affect the basal level of migration of HUVECs and bEND3 cells.

We investigated the effect of ARP4 on tubule formation of HUVECs on Matrigel. HUVECs cultured on Matrigel undergo morphological changes with formation of a tubule, which possesses a lumen surrounded by endothelial cells adhering to one another (Fig. 4,A). We found that ARP4 inhibited tubule formation of HUVECs in a dose-dependent manner (Fig. 4, B–E). These tubular formations were quantified with NIH Image software by pixels (Fig. 4 F). Taken together, the results of the above in vitro analyses indicate that ARP4 directly inhibits the proliferation, chemotaxis, and tubule formation of endothelial cells.

Based on our findings in vitro, we investigated the effect of ARP4 on endothelial cells in in vivo angiogenesis using a corneal neovascularization assay in the mouse. Five days after implantation of a pellet containing VEGF alone, growth of new capillary vessels from the corneal limbus toward the pellet was detected (Fig. 5,A), whereas corneal neovascularization toward a pellet containing both VEGF and ARP4 was significantly inhibited (Fig. 5,B). The maximal vessel length, clock hours, and vessel area, which are parameters of corneal neovascularization, were markedly inhibited compared with VEGF-induced neovascularization (Fig. 5, D–F). No growth of new capillary vessels was observed in cornea-implanted pellets containing ARP4 alone (Fig. 5 C). These results demonstrate that corneas implanted with pellets containing ARP4 in addition to VEGF exhibit a significant reduction in angiogenesis, relative to those induced by VEGF alone.

Vessel leakiness is one of multiple abnormalities seen in tumor vessels, which influences angiogenesis, tumor growth, and metastasis. We therefore estimated the effect of ARP4 on the permeability of blood vessels using the Miles assay, in which extravasation of Evans Blue dye from vessels in the back skin of a mouse can be quantified. We found that ARP4 has a suppressive effect on VEGF-induced microvascular permeability (Fig. 6,A). Quantitative analysis revealed that about 4.1-fold less leakage occurred in skin injected with both ARP4 and VEGF compared with skin injected with VEGF only (Fig. 6 B). These results show that the presence of ARP4 led to a significant reduction in blood vessel permeability induced by VEGF.

Several lines of in vitro experiments in this study characterized ARP4 as an antiangiogenic factor. Tg mice expressing a gene targeted to the epidermis under the control of keratinocyte-specific promoter K14 represent a powerful strategy to investigate whether a targeted gene affects angiogenesis (16, 23, 24). To further investigate the roles of ARP4 in vivo, we therefore generated K14-ARP4 Tg mice, which express ARP4 in basal epidermis (Fig. 7,A). From 27 pups, we identified 3 Tg mice by PCR analysis (data not shown). Subsequently, we selected two independent K14-ARP4 Tg mouse lines expressing ARP4 in the skin as determined by Northern blot analysis (Fig. 7,B). ARP4 transgene was expressed in the basal keratinocytes of the epidermis and the hair follicle outer sheath (Fig. 7,C). K14-ARP4 Tg mice were grossly normal, and their skin color was also normal (Fig. 7, D and E). Histological analysis of the skin tissues of 3-month-old K14-ARP4 Tg mice showed no alteration in the thickness and morphology of both the epidermis and dermis (Fig. 7,F). Fluorescein-labeled L. esculentum lectin staining revealed that the vessels in the ear skin of K14-ARP4 Tg mice resembled vessels in the ear skin of control mice in numbers, density, morphology, and size (Fig. 7, G and H). Based on these observations, these two lines were chosen for further study, and both showed essentially equivalent phenotypes.

We analyzed whether ARP4 from keratinocytes would suppress tumor growth using an intradermal CMT93 xenografted tumor model. Five weeks after intradermal CMT93 tumor cell inoculation, the size of tumors seen in K14-ARP4 Tg mice was significantly smaller than that seen in control littermates (Fig. 8,A). We found that the number of invaded capillary vessels in K14-ARP4 Tg mice was markedly decreased and that capillary lumens were significantly smaller compared with those of control littermates by PECAM-1 immunostaining (Fig. 8, B and C). Taken together, these findings suggest that ARP4 suppresses tumor growth by its antiangiogenic activity.

ARP4 expression was recently found to be reduced in two gastric cancer cell lines (MKN28 and MKN74) in association with methylation of CpG islands in the 5′ region of ARP4 (10). A reduced expression level of ARP4 was also observed in primary gastric cancer, but not in normal tissue (10). We have also detected a marked reduction of ARP4 expression in many tumor cells compared with normal tissues (data not shown). These findings indicate an inverse correlation between ARP4 expression and tumorigenesis. Recent findings also indicate that down-regulation of expression of antiangiogenic factors, such as maspin (25) and thrombospondin-1 (26), correlates with tumor progression and metastasis. Based on these observations, we hypothesized that ARP4 signaling might negatively regulate tumorigenesis. Indeed, we found that ARP4 inhibited the proliferation, chemotaxis, and tubule formation of endothelial cells, supporting such a role.

Furthermore, we found that ARP4 had a potential inhibitory effect on VEGF-induced vascular leakage in vivo. Because many tumor cells secrete VEGF, plasma leakage is a key feature of new vessels in tumorigenesis (27). This finding also indicates that ARP4 could negatively regulate angiogenesis in tumorigenesis. This antipermeability effect is similar to that of angiopoietin-1, which potentially reduces both VEGF- and inflammation-induced plasma leakage (16). Angiopoietins and ARPs have similar structures exhibiting coiled-coil domains and fibrinogen-like domains in which an essential binding site for Tie2 receptor exists. ANGPTL3 (7, 8), ARP1/angioarestin (9, 28), and ARP2 (28), which are ARP family members, act on the endothelial cells to regulate angiogenesis, although previous reports revealed that ARP family members in general bound to neither Tie1 nor Tie2 receptor. Therefore, it is necessary to identify the ARP4 receptor and determine not only where it is expressed but how it signals to address the antiangiogenic and antipermeability mechanism of ARP4. Such characterization should clarify how decreases in ARP4 expression could potentially lead to tumor formation and promote vascular leakage.

Our cloned ARP4 is identical to the previously reported FIAF (5) and PGAR (6) proteins. FIAF/PGAR is a PPAR target gene. Investigators of FIAF/PGAR propose that those proteins may play a role in regulating systemic lipid metabolism or glucose homeostasis. PPARγ is expressed in endothelial cells, and PPARγ ligands, inducers of FIAF/PGAR/ARP4 expression, are potent inhibitors of endothelial tube-like structures and proliferation in vitro and also suppress VEGF-induced angiogenesis in vivo (29). Moreover, PPARγ ligands induce endothelial cell apoptosis in vitro (30) and reduce the expression of metalloproteinases, such as matrix metalloproteinase 9 (31), which are implicated in tumor angiogenesis and invasion (32). Furthermore, PPARγ ligands can inhibit primary tumor growth and metastasis (33, 34, 35) through inhibition of their proliferation induced by growth factors or induction of apoptosis and fibrosis of injected tumor cells (32). In this study, we found significantly suppressed growth of grafted tumor cells and decreased numbers of invading blood capillary vessels in grafted tumors in the dermal layer of K14-ARP4 Tg mice relative to control littermates. Given that angiogenesis is essential for tumor growth, these findings also suggest that ARP4 may function in tumorigenesis as a negative modulator by suppressing tumor angiogenesis. During preparation of this manuscript, ANGPTL4, which is identical to ARP4, was reported to have a proangiogenic effect and to be induced under hypoxic conditions (36). Although we also confirmed that ARP4 mRNA was induced in endothelial cells under hypoxic conditions (data not shown), we could not find a proangiogenic effect of ARP4 by any in vitro experiments. Moreover, we generated Tg mice expressing ARP4 in the epidermis (K14-ARP4 Tg) to examine whether ARP4 acts in angiogenesis. Because both VEGF and Ang1 are proangiogenic factors, we therefore compared skin color and vasculature in the dermis of K14-ARP4 Tg mice with those of K14-VEGF and K14-Ang1 Tg mice (both gifts from Dr. George D. Yancopoulos; Regeneron Pharmaceuticals, Tarrytown, NY), which had shown angiogenic activity in their dermal layer (16). This investigation revealed no alteration in skin color or in the number and size of microvessels in the dermis of K14-ARP4 Tg mice compared with their controls, whereas both K14-VEGF and K14-Ang1 Tg mice were grossly red, and K14-VEGF Tg mice showed an increased number of microvessels, and K14-Ang1 Tg mice showed an enlarged size of microvessels in their dermis (Ref. 16; data not shown) as reported elsewhere. These findings suggest that ARP4 is not a proangiogenic factor, but an antiangiogenic factor. Critical to understanding the function of a novel ligand is identification and characterization of its cognate receptor. The fibrinogen-like domain at the COOH terminus of angiopoietins is a binding site for the Tie2 receptor. ARP4 also contains this domain, suggesting that ARP4 could be a ligand for either Tie1 or Tie2. However, we observed that, like other ARPs, ARP4 did not bind to either immobilized Tie1-Fc or Tie2-Fc protein by a BIAcore binding assay (data not shown). We need further investigation to identify the ARP4 receptor to understand the function of ARP4 in endothelial cells.

In summary, we demonstrate here that ARP4 inhibits proliferation, migration, and tubule formation of endothelial cells in vitro and VEGF-induced angiogenesis and vascular leakage in vivo. Because several lines of evidence indicate that both angiogenesis and vascular leakiness play a crucial role in the process of growth of solid tumors and metastasis, we hypothesize that ARP4 might play an important role as a negative regulator in tumorigenesis. Finally, we show that tumors derived from inoculated CMT93 tumor in K14-ARP4 Tg mice are significantly smaller than those seen in control littermates due to inhibition of tumor angiogenesis. Neovascularization and plasma leakage are common key features of rheumatoid arthritis, diabetic retinopathy, and cancers. Our in vitro and in vivo functional analyses provide insight into the role of ARP4 in the vascular system as an antiangiogenic regulator. These findings suggest that ARP4 could act as a therapeutic agent not only in tumors but also in various pathological conditions related to neovascularization.

We thank T. Ota, T. Isogai, T. Keida, Y. Saito, and M. Kitamura for experimental assistance and Dr. Y. Suzuki for providing a full-length cDNA library.