A novel antiangiogenic and vascular normalization therapy targeted against human CD160 receptor

We evaluated the antiangiogenic properties of CL1-R2 mAb in vivo by using two different ocular neoangiogenesis animal models. The vertebrate eye has an advantage for these studies in that it is considered to be an immunoprivileged site (Cursiefen, 2007), thus possibly devoid of immune cells that could bind CL1-R2. We first used a rabbit corneal pocket assay (Fons et al., 2006) to determine whether CL1-R2 inhibits fibroblast growth factor 2 (FGF2)–induced corneal neovascularization. The cornea is normally devoid of both blood and lymphatic vessels and actively maintains this avascularity (Cursiefen, 2007). In this model, neovessels are attracted from the limbus. Neovascularization was assessed 8 d after transplantation of corneal implants containing FGF2 and two subconjunctival injections of 100 µg CL1-R2 or control IgG1. Treatment with CL1-R2 significantly decreased corneal neovascularization when compared with control IgG1-treated rabbits (Fig. 1, A and B). These findings indicate that CL1-R2 treatment inhibits growth factor–induced corneal neovascularization.

Next, we investigated the effect of CL1-R2 in a mouse model of human retinopathy of prematurity by exposing premature newborn mice (postnatal day (P) 7–12) to high oxygen levels (Smith et al., 1994; Economopoulou et al., 2005). In 100% of these animals the return to normoxic conditions induced retinal ischemia and VEGF-dependent preretinal vascularization. We first performed a qualitative assessment of the retinal vasculature on flat-mounted FITC-dextran–perfused whole retinas after intravitreal injections of CL1-R2 or control IgG1 (Fig. 2 A). Retinas from normal untreated animals exhibited normal vascularization, i.e., both superficial and deep vascular layers which extended from the optic nerve to the periphery. The vessels formed a radial branching pattern in the superficial retinal layer and a polygonal reticular pattern in the deep retinal layer. Retinas from oxygen-treated animals with no intraocular injection (Fig. 2 A, Mock) or intraocular injection of control IgG1 (Fig. 2 A, Ctrl) displayed neovascular tufts that released fluorescein and had tortuous radial vessels and a central avascular zone, which is consistent with previous descriptions of this model. After intraocular injection of CL1-R2, avascular areas decreased in size and the retinas contained fewer neovascular tufts and fewer tortuous and dilated radial vessels (Fig. 2 A), suggesting better perfusion efficiency in the central vessels.

Eyes from the various untreated or oxygen-treated animals were analyzed further by histology. Serial ocular tissue sections were stained with periodic acid–Schiff reagent to visualize the nuclei of endothelial cells (Fig. 2 B). Unlike the retinas from mice in a normoxic environment (normal retina), the retinas of mock-treated mice typically contained abundant longitudinal and transverse aberrant microvessels of various sizes in the vitreous space and inner retina, as well as endothelial cell nuclei identified by periodic acid–Schiff staining (Fig. 2 B). Retinas from control IgG-treated animals displayed similar neovascularization with abundant aberrant microvessels and endothelial cell nuclei (Fig. 2 B). In contrast, retinas from mice injected with CL1-R2 had significantly fewer aberrant vessels, which were greatly reduced in size, especially in the vitreous space and within the retina, and fewer endothelial cell nuclei. Von Willebrand factor staining further confirmed that periodic acid–Schiff-positive cells were endothelial cells (Fig. 2 C). To quantify retinal neovascularization, endothelial cell nuclei and lumens of neovessels were counted in a large number of samples before and after the administration of CL1-R2 or control IgG1 (Fig. 2, D and E). These are crucial parameters for accurately evaluating any retinal antiangiogenic effect. Intravitreal injection of CL1-R2 significantly decreased (∼35%) the mean number of endothelial cell nuclei per section in both ganglion cell and inner nuclear layers as compared with animals injected with control IgG1 (P < 0.001; Fig. 2 D). Furthermore, CL1-R2 reduced the mean number of vessel lumens per section by ∼50% when compared with mock-treated (P < 0.001) or IgG1-treated control mice (P < 0.001), respectively (Fig. 2 E). We then compared the effect of CL1-R2 treatment to that of the widely used mAb bevacizumab in the same mouse model regardless of the controversy dealing with the specificity of bevacizumab to neutralize murine VEGF-A (Yu et al., 2008). Several studies have indeed conclusively demonstrated that bevacizumab, despite its weak affinity for the VEGF-A produced by mice, rats, guinea pigs, and rabbits, is efficient in treating experimentally induced corneal neovascularization in these animals (Bock et al., 2007; Manzano et al., 2007; Hurmeric et al., 2008; Hashemian et al., 2009; Habot-Wilner et al., 2010). Furthermore, a recent study showed unambiguously that bevacizumab had a very significant inhibitory effect on retinal angiogenesis in the oxygen-induced retinopathy mouse model (Zhang et al., 2009). These latter results are in full agreement with ours. After intraocular injection of bevacizumab, we observed normalized retinal vascularization on flat-mounted retinas and retinal tissue sections comparable to that obtained after CL1-R2 treatment (Fig. S2, A and B). Quantitative analysis indicated that the mean number of endothelial cell nuclei per section was not significantly different between CL1-R2–treated and bevacizumab-treated mice (Fig. S2 C), whereas the number of vessel lumens per section was significantly lower in CL1-R2–injected mice than in bevacizumab-treated control mice (P < 0.05; Fig. S2 D). Overall, these data show that CL1-R2 mAb monotherapy efficiently suppresses pathological angiogenesis in rabbit cornea and mice with ischemic retinopathy.

We then evaluated the antiangiogenic effect of CL1-R2 in vivo on the mouse Matrigel plug model of angiogenesis (Passaniti et al., 1992). This assay, in which Matrigel plugs containing FGF2 are injected subcutaneously into mouse flanks, has proven to be a relatively easy method to evaluate antiangiogenic compounds in vivo. Initial indications of the levels of activity of antiangiogenic compounds can be assessed visually because of the color difference in the vascularized plugs when compared with the controls. The pale color of the plugs removed from CL1-R2–treated mice indicated little FGF2-induced vascularization over a 7-d period (Fig. 3 A, top). In contrast, the reddish color of the plugs in the control IgG1-treated mice reflected the development of dense neovascularization. We next compared the vessel density in CL1-R2–treated and control Matrigel plugs sections stained with the endothelial cell-specific marker isolectin B4. Representative sections are shown (Fig. 3 A, bottom). The vessels in plugs from CL1-R2–treated animals were smaller and fewer in number than those in plugs from the control mice. We quantified the level of angiogenesis by determining the hemoglobin content of the plugs and found a marked reduction (∼45%, P = 0.0153) in plugs from CL1-R2–treated mice when compared with those from IgG1-treated control mice (Fig. 3 B). In 14 out of 22 CL1-R2–treated Matrigel plugs, the hemoglobin content was extremely low. These data indicate that the CL1-R2 mAb significantly impaired angiogenesis in this in vivo assay.

To further evaluate the antiangiogenic activity of CL1-R2 in vivo, we used the transplanted mouse B16 melanoma tumor model, which is poorly immunogenic and highly vascularized. Histological analysis of these tumors showed that the vascularization was significantly lower in CL1-R2–treated mice than in control animals (Fig. 4, A and B; P = 0.0001). Histology gave an instantaneous picture of vascularization but did not reflect the functionality of blood vessels. So we used another approach based on the use of color Doppler ultrasonography (DUS), a noninvasive method of visualizing tumor vessels in situ at different times (Asselin-Paturel et al., 1999). DUS demonstrated that the vascular density of B16 tumors at 12, 15, and 18 d after CL1-R2 injections was significantly lower than that of tumors in control mice that received IgG1 injections (Fig. 4, C and D; and Videos 1 and 2). As B16 tumor cells are sensitive to cyclophosphamide (Hamano et al., 2004), we also investigated whether the observed vascular changes after CL1-R2 treatment increased susceptibility to this cytotoxic agent. The combination of both mAb and chemotherapeutic agent, administered according to a metronomic schedule, significantly reduced the growth of B16 tumor (Fig. 4 E; P = 0.0001) compared with control IgG1 with cyclophosphamide. Thus, CL1-R2 and cyclophosphamide act in concert to reduce neovascularization and slow the growth of B16 tumor.

The normalization of tumor vasculature with antiangiogenic therapy is an emerging concept (Jain, 2005). This vascular normalization implies that any antiangiogenic strategy should target immature dysfunctional vessels and fortify those remaining. To investigate the effect of CL1-R2 on tumor normalization, we used intravital microscopy. This method allows the continuous noninvasive analysis of the same tumor area in its orthotopic environment (in a dorsal skin-fold chamber) over a 2-wk period, including measurement of blood flow–related variables such as vessel diameter, microvascular density, and branching (Jain et al., 2002; Koehl et al., 2009). Wide-field intravital fluorescence microscopy showed that CL1-R2 injections together with cyclophosphamide significantly reduced B16 tumor growth over a 14-d period after tumor implantation when compared with IgG1-treated control mice with cyclophosphamide (not depicted), confirming results obtained by caliper measurements (Fig. 4 E). The first signs of angiogenesis, including dilation and tortuous elongation of preexisting blood vessels, were observed in the host tissues of control mice soon after melanoma implantation (Fig. 5, A and B). This phenomenon was a result of the proliferating tumor remodeling the host vessels, as no modification of the preexisting vessels was observed in the absence of a grafted tumor (Fig. S3). Treatment of mice with CL1-R2 limited host vessel dilation and tortuosity (Fig. 5 A and Fig. S4). The B16 tumors were vascularized over a substantial area in control mice, whereas the area of tumor vascularization was much smaller in CL1-R2–treated animals 12 d after implantation (Fig. 5 B). At day 14 after tumor implantation (the last day of the experiment), the vascular area of B16 tumors in CL1-R2–treated mice was 2.56-fold smaller than in control mice (Fig. 5 B). We then performed a fractal analysis to quantify the geometric complexity of the vascular network. This analysis offers an additional method for measuring tumor vascular damage given that tumor vasculature is characterized by loss of hierarchical architecture (Heymans et al., 2000). We found that the B16 tumor vasculature in CL1-R2–treated mice had a significantly smaller fractal dimension than in control mice (Fig. 5 C). These results indicate that treatment with CL1-R2 favored a better hierarchical architecture of the B16 tumor vessel network, which may influence the circulation. Consistent with this finding, B16 tumor vessel diameters in IgG1-treated mice varied from 2.35 to 48.33 µm, in contrast to the much more uniform tumor vessel diameters observed in CL1-R2–treated animals (6.46 to 25.94 µm; Fig. 5 D). Collectively, these observations, based on a wide range of vascular parameters, indicate that the CL1-R2 mAb, combined with cyclophosphamide chemotherapy, results in a reduced vascular density in transplanted B16 tumors and normalization of the remaining tumor vasculature when compared with IgG1-treated mice. Moreover, our data indicate that CL1-R2 treatment was effective not only if the mAb was injected on the same day as tumor cell implantation but also if injected 3 d after tumor inoculation (Fig. 4, C and D) or if a B16 tumor fragment was implanted under the skin (Fig. 5).

Tumor blood vessels are immature and display functional defects, including vessel lumen collapse leading to restricted blood flow and poor access to the chemotherapeutic agents (Carmeliet, 2005; Jain, 2005; Hamzah et al., 2008). Open lumens are thus a good indicator of vessel maturation and functionality in solid tumors. By counting the vessels in different fields of each tumor section, we found that a greater proportion of B16 tumor vessels had an open lumen in CL1-R2–treated mice than in control animals (Fig. 6 A). Pericytes recruited along endothelial cells are another indicator of maturation and normal functioning of blood vessels (Carmeliet, 2005). We therefore evaluated the pericyte coverage of B16 tumor vessels by confocal microscopy. The vasculature of tumors from CL1-R2–treated mice exhibited strong staining for smooth muscle actin (SMA; a marker of pericytes), unlike tumor vessels from control mice which displayed little or no staining for SMA (Fig. 6 B). These data suggested that CL1-R2–treated tumor vessels might have a lower resistance to blood flow, which may improve the blood supply within the tumor. To evaluate blood perfusion, Hoechst 33342 fluorescent dye was administered intravenously and its uptake by tumor vessels was examined by fluorescence microscopy. This dye labels the nuclei of endothelial cells and cells adjacent to the walls of perfused vessels. Little perivascular labeling was observed in B16 tumor vessels from control mice despite a high level of vascularization (Fig. 6 C). In contrast, most of the tumor vessels in CL1-R2–treated mice displayed intense perivascular staining. Thus, treatment with CL1-R2 improved perfusion efficiency in B16 tumor vessels. We next evaluated the extent to which CL1-R2 influenced tumor necrosis as a result of cyclophosphamide. Histological studies showed more extensive necrosis in CL1-R2–treated tumors than in controls (Fig. 6 D), suggesting that CL1-R2 treatment improved the delivery of cyclophosphamide by normalizing the remaining tumor blood vessels. Together, these findings demonstrated that CL1-R2 mAb treatment not only reduced abnormal tumor vasculature but also favored the maturation and function of the remaining tumor vessels.

To evaluate the therapeutic benefit of CL1-R2 mAb injection and to make a comparison with anti-VEGF mAb (bevacizumab), we compared the B16 tumor burden after these different treatments combined with cyclophosphamide (Fig. S5). CL1-R2 mAb injection improved the mouse survival (35 d), compared with the earlier death (30 d) of mice injected with bevacizumab (P = 0.032) or control IgG1 (P = 0.016).

We investigated further whether CL1-R2 specifically targeted CD160 expressed by B16 tumor vessels. A representative section of B16 tumor showed intense staining of CD160 similar to that of the isolectin B4 endothelial marker (Fig. 7 A). In contrast, CD160 was barely detectable in the vessels of normal mouse heart in comparison with the strong isolectin B4 staining. We also evaluated the distribution of CD160 in human healthy colon and colon tumors. The vessels present in colon tumors were strongly stained by CL1-R2 with patterns resembling those stained for CD31, a human endothelial cell marker (Fig. 7 B). In contrast, CL1-R2 did not stain the blood vessels present in healthy colon. As CL1-R2 recognized newly formed vessels in solid tumors but not vessels from healthy tissues, its therapeutic antiangiogenic effect is likely to be restricted to these neovessels.

Overall, our findings show that in monotherapy CL1-R2 mAb efficiently blocks pathological angiogenesis in both rabbit cornea and mice with ischemic retinopathy, whereas in combined therapy it restricts the development of abnormal B16 tumor vasculature and normalizes the remaining vessels, allowing a better delivery of a cytotoxic agent. Although CL1-R2 therapy could be used with combined angiostatic drugs that inhibit other parts of the angiogenic process (Dorrell et al., 2007; Takeda et al., 2009), our data suggest that CL1-R2 might be effective as a monotherapy for several human ocular diseases associated with neovascularization, such as diabetic retinopathy and AMD with choroidal neovascularization. These diseases remain a major cause of blindness and poor vision worldwide with an expected doubling of the prevalence of AMD in the coming decades as a result of the projected increase in the aging population.

We have previously demonstrated that, in vitro, CL1-R2 mAb inhibited FGF2-mediated tubule vessel growth from human umbilical vein endothelial cells expressing CD160 and induced their apoptosis (Fons et al., 2006). It is likely that the CL1-R2–mediated inhibition of in vivo ocular and cancer neoangiogenesis described in this paper is the consequence of a similar caspase-dependent apoptosis mechanism. Such a mechanism differs from the existing antiangiogenic therapies targeted to VEGF/VEGFR and thus may be useful when resistance to these latter therapies occurs. We cannot exclude the possibility of a synergic antiangiogenic effect caused by both cyclophosphamide and the CL1-R2 mAb, as chemotherapeutic drugs administered with a metronomic schedule have been reported to display antiangiogenic properties (Man et al., 2002). However, this is unlikely in the B16 melanoma tumor model, as cyclophosphamide treatment alone had no effect on tumor vascular density, suggesting that CL1-R2 might be solely responsible for the antiangiogenic effect observed when combined with cyclophosphamide.

In addition to its antiangiogenic effect, we demonstrated that CL1-R2 mAb contributes to the normalization of remaining B16 vessels by producing unbranched vessels with homogeneous diameters and open lumens, resulting in a better organized network similar to the vasculature of normal tissues. Importantly, CL1-R2 treatment strongly increased the association of pericytes with the B16 tumor vasculature, demonstrating their mature status. Consistent with this, we observed in a transplant model of ALK-induced fibroblast tumors in athymic nude mice (Giuriato et al., 2007), treated with CL1-R2 mAb alone (without any combined chemotherapy), that the tumor vessels had a mature phenotype, i.e., open lumen and pericyte coverage, which is in contrast with cyclophosphamide monotherapy (Fig. S6). These latter observations made in a second tumor model confirm that CL1-R2 has a direct effect on the maturation of tumor vascularization. This property is important, as it has been shown that anti-VEGF therapy inhibited tumor vessel maturation and pericyte function (Greenberg et al., 2008). Mature vessels are more resistant to the pressure exerted by proliferating tumors, favoring drug delivery within the tumor (Padera et al., 2004). This should be the case for B16 tumors treated with CL1-R2 mAb, as we observed extensive intratumor necrosis. Moreover, it has been recently shown that the level of vessel maturation significantly contributed to the response toward antiangiogenic therapy in melanoma (Helfrich et al., 2010). Thus, CL1-R2 induced normalization of B16 tumor vessels, leading to the establishment of a stable and functional vessel network. It was shown that antiangiogenic drugs could generate a normalization window for improved delivery of chemotherapeutic drugs (Lin and Sessa, 2004). Because we gave cyclophosphamide at the same time as CL1-R2 mAb, i.e., from day 0 of B16 tumor implantation, it is very unlikely that a normalization window occurred in our tumor model. For the same reason, it is also unlikely that B16 tumors acquired an invasive and metastatic phenotype after CL1-R2 treatment, as has been recently described in other mouse tumor models (Pàez-Ribes et al., 2009). Normalization by CL1-R2 treatment of retinal vascularization in mice with ischemic retinopathy was also observed by ex vivo angiography (Fig. 2 A). These results suggest that the antiangiogenic and normalization effects are not mutually exclusive.

CL1-R2 is a bifunctional mAb, as it blocks neoangiogenesis, normalizing the remaining tumor vessels (this study), but may also activate the innate immune system (Le Bouteiller et al., 2002; Barakonyi et al., 2004). Although we did not investigate whether CL1-R2 mAb therapy activated NK cells, this possibility cannot be excluded as we previously demonstrated that specific binding of CL1-R2 mAb to CD160 on human peripheral blood NK cells triggered interferon-γ, TNF, and IL-6 production, which participated in the inflammatory response (Barakonyi et al., 2004). Activation of NK cells by CL1-R2 may thus contribute to slowing tumor growth in addition to the antiangiogenic and normalization effects of the mAb. In contrast, in cornea or retina such CL1-R2–mediated NK cell activation is unlikely, as these cells are generally not present in these tissues (Ambati et al., 2006).

In summary, we have demonstrated that the CL1-R2 mAb exerts an antiangiogenic effect on tumors in vivo; by normalizing the vasculature, it thus improves perfusion and enhances the accessibility of the tumor to chemotherapeutic drugs. These studies indicate that targeting the CD160 receptor with a specific mAb, CL1-R2, is a promising therapeutic option that could be used in monotherapy or combination therapy for the treatment of many diseases with associated pathological neovascularization. The spectrum of neovascular diseases that could be treated is potentially large, including several vision-threatening eye diseases and many solid tumors associated with high levels of angiogenesis. CL1-R2 monotherapy might be particularly apt for treatment of ocular neovascular diseases in humans, as this mAb has antiangiogenic activity alone in animal models of both corneal and retina neoangiogenesis. In combination with intravenous chemotherapy, CL1-R2 might be useful for highly vascularized cancers as this mAb also triggers vessel normalization, opening the way to a successful chemotherapy. Further investigations are ongoing to ascertain the appropriate CL1-R2 mAb dosing regimens, best route of delivery, and duration of therapy to manage the toxicities and risks of systemic exposure and to select the most effective humanized mAbs with the lowest level of immunogenicity that may be successful for treating ocular neovascular diseases and cancer (Cao and Langer, 2010).

The mouse anti-CD160 CL1-R2 mAb (IgG1) was developed in our laboratory and was evaluated as an anti-CD160 mAb during the seventh Human Leukocyte Differentiation Antigen Workshop (Bensussan, 2000). We produced the CL1-R2 mAb from a specific secreting hybridoma cell line by using the high cell density system CELLine (Valdea Biosciences). The endotoxin unit concentration was <0.6 EU/mg of IgG. The mouse IgG1 isotype control monoclonal antibody He6 was developed by immunizing mice with hepatitis B surface antigen. Both CL1-R2 and IgG1 isotype control were similarly purified by affinity chromatography on a HiTrap protein G column (GE Healthcare) in an ÄKTA purifier system (GE Healthcare), dialyzed against PBS, pH 7.0, concentrated, and filtered through 0.22-µm filters. The cross-reactivity of CL1-R2 mAb with CD160 mouse proteins has been demonstrated by different means (Fig. S1): (1) using flow cytometry, we found that CL1-R2 stained mouse SVR endothelial cells as well as the BCL-1 mouse B cell chronic lymphocytic leukemia cell line (Fig. S1 A); (2) using immunoprecipitation and Western blotting, we found that the CNX46-3 anti–mouse CD160 mAb (Cedarlane laboratories) immunoprecipitated protein bands from mouse spleen or BCL-1 cells that were blotted by the CL1-R2 mAb (Fig. S1 B); and (3) five CD160 amino acids were identified by mass spectrometry in the CNX43-6–immunoprecipitated 37-kD band blotted with CL1-R2 mAb with a minimum coverage of 37.5% (Fig. S1 C).

We used BALB/c, C57BL/6J, and NMRI-nu (nu/nu) Nude mice (Janvier Laboratories). Mice were 7–10 wk old, except for those used for the ischemic retinopathy experiments which were 7 d old. Animals were housed in a conventional temperature-controlled room (21°C), exposed to a daily 12-h period of light and dark, and fed ad libitum with a balanced diet as determined by The Jackson Laboratory for the C57BL6/J mouse strain. We used male New Zealand albino rabbits from Institut National de la Recherche Agronomique (Castanet-Tolosan). For the mouse retina experiments, animals were handled according to the guidelines of the institutional animal care committee, using protocols approved by the Ethics Committee and the ARVO Statement for use of Animals in Ophthalmic and Vision Research. All other animal experiments were performed in agreement with the European Union guidelines and the Midi-Pyrénées Ethic Committee for the care and use of laboratory animals (MP/02/37/06/08).

The corneal pocket assay used in this study has been previously described (Fons et al., 2006). We made an incision in the upper side of the cornea, 2 mm from the limbus in anesthetized rabbits. FGF2-treated implants (500 ng; R&D Systems) were inserted into this pocket. Subconjunctival injections of CL1-R2 mAb or control IgG1 (100 µg in 30 µl PBS) were administered to the upper side of the limbus 24 and 72 h after corneal implantation. Corneal neovascularization was measured 8 d after implantation and was scored on a four-grade scale based on the length of the newly formed vessels from the limbus to the FGF2-containing implant (Fons et al., 2006).

Retinal neovascularization was induced in mouse C57BL/6J pups using a well established and reproducible model of oxygen-induced retinopathy (Smith et al., 1994). In brief, mice (7 d old, P7) and their nursing mothers were placed in an airtight incubator and exposed to a 75 ± 2% oxygen atmosphere for 5 d. The oxygen level was continuously monitored with an oxygen analyzer (model 110; PROOX; BioSpherix). Mice were removed on P12 and maintained in normoxic conditions (room air) until P17. Mice were injected intravitreally under an operating microscope. With the exception of the noninjected group, each pup received an intravitreal injection in their left and right eyes on P12. In brief, mouse pups were anesthetized with an intramuscular injection of ketamine (100 mg/kg body weight) and xylazine (10 mg/kg body weight). The palpebral fissures were opened with microscissors and pupils were dilated with topical 10% phenylephrine and 0.5% tropicamide. The tip of a 10-mm 33-gauge steel needle, mounted on a 5-µl Hamilton syringe, was pushed through the sclera, 1 mm posterior to the corneoscleral limbus, into the vitreous body. Approximately 1 µl CL1-R2 mAb (5 µg/µl), bevacizumab (25 µg/µl; Roche), or IgG1 isotype control mAb (5 µg/µl) was injected into the vitreous cavity.

Mice were killed at P17 to analyze neovascularization by histology and quantitative measurements. Some mice underwent retinal angiography with fluorescein-dextran. For this qualitative assessment, we anesthetized mouse pups as previously described and perfused the heart through the left ventricle with 1 ml PBS containing 50 mg/ml fluorescein-labeled dextran (2 × 10⁶ mean molecular weight; Sigma-Aldrich) that had been cleared by centrifugation for 5 min at 10,000 rpm. The eyes were enucleated and fixed in 4% paraformaldehyde for 3 h. The cornea and lens were removed and the retina was dissected from the eyecup. The retina was cut into four quadrants and flat mounted in Vectashield under a coverslip for examination by fluorescence microscopy. At least 12 eyes from each treatment were examined. For the histological analysis, mouse pups were killed and their eyes enucleated, fixed in 4% paraformaldehyde for at least 16 h at 4°C, and embedded in paraffin. We prepared sagittal 5-µm sections with a microtome (HM355, MICROM MICROTEC), stained sections with periodic acid–Schiff reagent, and counterstained with hematoxylin and eosin. We counted five to eight sections on each side of the optic nerve. Two trained investigators blindly counted the number of neovascular endothelial cells and vessel lumens across the entire retinal sample in each intact section at 100× magnification. Mean numbers of endothelial cell nuclei and vessel numbers were determined using a Poisson regression model for clustered data. 95% confidence intervals of the mean number estimates are figured as error bars. P-values were corrected for post-hoc group comparisons by the Bonferroni method. Endothelial cell–specific staining was performed by immunohistofluorescence. Paraffin was removed and sections were rehydrated and incubated for 20 min in 1× citrate buffer in a microwave at 500 W. Sections were treated with 0.3% Triton X-100 for 5 min and then incubated overnight with polyclonal antibody raised against Von Willebrand Factor at a 1:500 dilution (Abcam). Negative control immunohistochemical experiments were systematically performed with the exclusive omission of primary antibody. Sections were then incubated for 2 h at room temperature with a secondary Alexa Fluor 488 anti–sheep antibody at a 1:300 dilution (Invitrogen). After counterstaining for 10 min with DAPI at a 1:1,000 dilution (Sigma-Aldrich), washed sections were mounted in Fluorescent Mounting Medium (Dako). Optically sectioned images were acquired by confocal laser-scanning microscopy (LSM 700; Carl Zeiss).

500-µl Matrigel plugs (BD), supplemented with 600 ng/ml FGF2 (R & D Systems) and 25 µg of either CL1-R2 mAb or an IgG1 isotype control, were injected subcutaneously into the two flanks of BALB/c mice. Every other day, 200 µg CL1-R2 mAb or IgG1 mAb was injected in vivo directly into the Matrigel plug through the skin. 7 d after injection, the animals were killed, the plugs removed, and their hemoglobin concentration evaluated sing Drabkin’s reagent kit (Sigma-Aldrich). For histological analyses, Matrigel plugs were fixed in 4% paraformaldehyde and embedded in paraffin. Sections were stained by incubation with biotinylated isolectin B4 (Vector Laboratories) followed by streptavidin-peroxidase and DAB substrate.

Mouse B16 tumor cells expressing GFP (a gift from J. Di Santo [Institut Pasteur, Paris] and B. Couderc [Institut National de la Santé et de la Recherche Médicale U563, Toulouse, France]) were injected subcutaneously into the flanks of C57BL/6 mice (10⁵ cells per mouse). These mice received i.p. injections of 500 µg CL1-R2 mAb or IgG1 control or an equivalent volume (100 µl) of PBS three times per week starting on the day of tumor inoculation. Mice were also treated with 10 mg/kg body weight of cyclophosphamide (Sigma-Aldrich) administered daily in the drinking water. No apparent signs of toxicity, such as weight loss or lethargy, were observed in response to treatment. Tumor volume was determined three times per week by measurement with calipers. The volume was calculated as width² × length × 0.52.

Mouse MEF fibroblasts expressing the active mutant form of the nonmuscular TPM3-ALK were injected subcutaneously into nude mice as previously described (Giuriato et al., 2007). Mice were randomized in four treatment groups of five mice each that received IgG1 control, cyclophosphamide, CL1-R2, or CL1-R2 combined to cyclophosphamide (same protocol and doses as the B16 tumor model described in the previous section). Treatments were stated the same day as tumor implantation. Growing tumors were harvested 21 d later, fixed, and paraffin embedded for subsequent immunohistochemical or confocal microscopy analysis.

The experimental protocol was similar to that used in previous studies (Makale, 2007). C57BL/6 mice were anesthetized, and then a double layer of skin was sandwiched between two titanium frames and a layer of skin was surgically removed to create an observation window that was then covered with a glass coverslip with a 12-mm diameter. At the end of the surgery and the day after, mice received an intramuscular injection of 10 mg/kg ketoprofen (50 µl in each thigh) to prevent inflammation. After implantation, C57/Bl6 mice were housed individually and allowed to recover for 1 d. We temporarily removed the coverslip and deposited a B16-GFP tumor fragment (∼1-mm diameter) in the middle of the chamber. Mice tolerated the skin-fold chamber well and showed no signs of discomfort or changes in sleeping and feeding behavior. Mice were given an i.p. injection of 500 µg CL1-R2 or IgG1 control three times per week together with 10 mg/kg cyclophosphamide administered in the drinking water. Animals anesthetized with 2% isoflurane were analyzed by intravital microscopy on days 0, 2, 5, 7, 9, 12, and 14 after the tumors were implanted. These observations were performed with a MacroFluo microscope (Leica) equipped with wide-field objectives (Plan apo 2× for Z6/Z16 with 0.57×, 2×, and 4× objectives) and a cooled charge-coupled device camera (CoolSNAP HQ; Roper Scientific). Fluorescence excitation was obtained with a Mercury arc lamp (HBO; Osram) and a GFP filter (Leica). We used MetaVue software (Molecular Devices) for image acquisition. 50 µl of 0.5% tetramethylrhodamine isothiocyanate–labeled dextran (TRITC-dextran; Sigma-Aldrich) in PBS was injected intravenously to obtain better contrast when viewing vessels. To calculate the tumor area, mean vessel diameter, vascular area, and microvessel fractal dimension, digital images were obtained from up to five locations within the tumor and analyzed with ImageJ software (1.36b; National Institutes of Health). By applying a Gaussian Blur filter and threshold, nonspecific pixels were removed, blood vessels were identified, and vascular areas were calculated. The binary images were reduced to a 1-pixel-wide skeleton. We calculated the microvessel fractal dimension with the FracLac v2.4e plug-in, available in the Image J environment, using the box-counting method. The image was divided into increasingly smaller boxes and the number of boxes of size ε required to cover the vasculature was counted and indicated as N(ε). Fractal dimension was calculated and indicated as N(ε). Fractal dimension (Df) is given by: Df = lim_ε→0[log N(ε)/log(ε)].

C57/BL6 mice were injected subcutaneously with 10⁵ B16 cells and treated 3 d later with i.p. injections (three times per week) of 500 µg of either CL1-R2 or control IgG. Cyclophosphamide was added to the drinking water. We analyzed changes in the number of intratumoral blood vessels by carrying out DUS analysis on days 12, 15, and 18 after the first CL1-R2 injection, as previously described (Elie et al., 2007). In brief, mice were anesthetized 15 min before each DUS analysis by an i.p. injection of 0.2 ml of a mixture of 10 mg/ml ketamine (Pfizer) and 0.1% xylazine (Bayer). One group of mice received repeated i.p. injections of 500 µg CL1-R2 mAb, whereas another group received IgG1 control antibody (five mice per group). We performed examinations with a diagnostic ultrasound system (Aplio SSA-770A; Toshiba) with two high-frequency large-bandwidth probes: 9 MHz (PZT-PLT 604AT; Toshiba) and 14 MHz (PZT-PLT1204AT; Toshiba). We optimized the sonograph settings for small and superficial tumors, and these settings remained unchanged for all examinations. We used a 50.0-mm Eko-Gel high polymer ultrasound standoff pad (Eurocamina SRL) between the transducer and the tumor surface. Color DUS scans were performed in the power Doppler mode for evaluation of the number of blood vessels within tumors. We determined the number of vessels for multiple cross sections of the tumor.

Human normal or adenocarcinoma colon biopsies were obtained from P. Brousset (Laboratory of Anatomy and Pathological Cytology, Purpan Hospital, Toulouse, France). Grafted Matrigel plugs, B16 tumors at day 16, or human colon biopsies specimens were fixed in 4% paraformaldehyde and embedded in paraffin. Mouse vessels were stained with the biotinylated isolectin B4 (Vector Laboratories) at 1:200 dilutions and human vessels with the mouse anti-CD31 antibody (Dako) at 1:200 dilutions, followed by biotinylated rat anti–mouse secondary antibody (BD). Immune complexes were visualized using the VECTASTAIN Elite ABC system (Vector Laboratories). For CD160 visualization, tumor sections were stained with 6 µg/ml of the CL1-R2 anti-CD160 mAb and incubated with the TSA tyramide amplification kit (Dako) according to the manufacturer’s instructions. For necrosis analysis, B16 tumors sections were stained with hematoxylin and eosin. In perfusion experiments, mice received an i.v. injection of 40 mg/kg of the fluorescent DNA-binding dye Hoechst 33342 (Sigma-Aldrich) 1 min before killing. Tumors were removed and immediately frozen in liquid nitrogen. Frozen tissue sections were stored in the dark at −80°C. Labeled vessels were viewed under a microscope using 350-nm excitation filters to observe the blue fluorescence. Vessels were stained with biotinylated isolectin B4 (Vector Laboratories) at a 1: 10 dilution, followed by streptavidin-phycoerythrin (1:100 dilution; Dako). For confocal microscopy, paraffin-embedded sections were stained with 2 µg/ml of rabbit anti-SMA (Spring Bioscience) or rat anti-CD34 (2 µg/ml; AbD Serotec), followed by anti–rabbit or anti–rat antibodies conjugated with Alexa Fluor 543 or Alexa Fluor 488 (Invitrogen). Nuclei were counterstained with DAPI. Slides were analyzed using a confocal microscope (LSM 710; Carl Zeiss).

For the mouse and rabbit model experiments, quantitative data (presented as mean ± SEM) were analyzed with Prism 4 or Prism 5 (GraphPad Software). A mean value for each vascular variable (intravital microscopy and histological analysis) was determined for each animal, and these values were used to calculate the overall mean for all the animals in each experimental group. Before carrying out statistical tests, we determined whether the data were normally distributed and evaluated their variance. We then performed the appropriate tests as indicated. For in vivo time course experiments, we used two-way ANOVA analysis or Student’s t test. We report the actual p-value for each test. P < 0.05 was considered statistically significant. For the retina counts, because there are two nested levels of dependence between histological sections pertaining to the same eye and the same mouse, counts of cell nuclei and vessel lumens were analyzed by a Poisson generalized linear mixed model (Wolfinger and O’Connell, 1993) with proc GLIMMIX of the SAS statistical package v9.1.3 (Sas Institute). We considered experimental treatment groups as fixed effect factors and individual eyes as random effects. At the upper limit, robust empirical variance of the fixed effect estimates (Liang and Zeger, 1986) was computed by defining the mice as clusters. P < 0.05 was considered statistically significant.

Fig. S1 shows that CL1-R2 mAb cross reacts with mouse CD160, using immunoprecipitation, Western blotting, and mass spectrometry analysis. Fig. S2 compares the effects of intravitreal injection of CL1-R2 and bevacizumab on retinal neovascularization in a mouse model of oxygen-induced retinopathy. Fig. S3 shows that dilation of preexisting host vessels does not occur in the absence of grafted B16 tumor. Fig. S4 shows that CL1-R2 treatment inhibits tumor-induced host vessels dilation. Fig. S5 shows that CL1-R2 injection improves the B16 tumor-bearing mouse survival when compared with bevacizumab treatment. Fig. S6 shows that CL1-R2 treatment induces vessel maturation in transplanted TPM3-ALK MEF tumor.

We thank Bettina Couderc (Institut National de la Santé et de la Recherche Médicale U563, Toulouse) and James Di Santo (Institut Pasteur, Paris) for providing the B16-GFP cell lines, Pierre Brousset (Hôpital Purpan, Toulouse, France) for human colon biopsy specimens, and Alain Razafindratsita (Matbiopharma, Evry) for production of the mAb CL1-R2. We thank Talal al Saati and Florence Capilla (Histologie Expérimentale Platform of Institut Fédératif de Recherche Bio-Médicale de Toulouse, IFR-BMT/150) for their assistance with immunohistology, Alain Regans (Hôpital Purpan, Service d’Ophtalmologie, Toulouse), and the staff of the Génopôle Toulouse-Midi Pyrénées, France, and Plateforme d’imagerie de l’Institut de Pharmacologie et de Biologie Structurale, Institut des technologies avancées en sciences du vivant, Toulouse RIO cellular imaging for their valuable technical assistance. We also thank Jean-Pierre Kinet and Hervé Prats for discussions and comments on the manuscript, and Carol Featherstone for comments and editing.

S. Chabot was a recipient of an ANR postdoctoral research fellowship (AGIR-RIB-2006) and J. Tabiasco was supported by the Fondation pour la Recherche Médicale, and the Institut National de la Santé et de la Recherche Médicale (INSERM). N. Jabrane-Ferrat was supported by Centre National de la Recherche Scientifique. This research was supported by grants from the INSERM (P. Le Bouteiller and A. Bensussan), the Université de Toulouse, Université Paul Sabatier (P. Le Bouteiller), the Agence Nationale de Recherche (ANR-R06459BS AGIR-RIB-2006; J. Kadouche., P. Le Bouteiller, and A. Bensussan), MATBiopharma (A. Bensussan, P. Le Bouteiller, and M. Abitbol), the Association RETINA-FRANCE (M. Abitbol and P. Le Bouteiller), the Association Valentin Haüy (P. Le Bouteiller), and the Ligue Régionale Midi-Pyrénées Contre le Cancer (N. Jabrane-Ferrat).

The authors have no conflicting financial interests. A. Bensussan and P. Le Bouteiller are inventors on a filed patent regarding this work.