Complementary Actions of Inhibitors of Angiopoietin-2 and VEGF on Tumor Angiogenesis and Growth

Inhibitors of vascular endothelial growth factor (VEGF) that are widely used in cancer therapy reduce angiogenesis (1), cause regression of tumor vessels, slow tumor growth (2, 3), and may improve drug delivery (4, 5). VEGF inhibitors have documented antitumor effects (6–8), but drug resistance eventually occurs, and other strategies for affecting the tumor vasculature are being sought (9). Inhibitors of angiopoietins are among the promising antiangiogenic approaches under development (10–13).

Angiopoietins are Tie2 receptor ligands that play key roles in vascular development, angiogenesis, and remodeling. Angiopoietin-1 (Ang1) tends to stabilize blood vessels and promote vascular maturation (14). Ang2 is expressed in growing blood vessels (15, 16) and promotes angiogenesis and tumor growth (17, 18) by destabilizing blood vessels (16, 19, 20). Vascular destabilization may involve Ang2 acting as an inhibitor or partial agonist of Tie2, which opposes the stabilizing action of Ang1 on endothelial cells (21, 22).

Inhibition of Ang2 may promote vessel stability and reduce angiogenesis (10). Because Ang2 promotes the proangiogenic action of VEGF (16, 19, 20), and VEGF upregulates Ang2 expression in endothelial cells (23), inhibition of Ang2 and VEGF together could have complementary actions by reducing sprouting angiogenesis and tumor vascularity (19, 24, 25).

Multiple approaches have been used to inhibit Ang2 to explore effects on angiogenesis and tumor growth. Among these are genetic deletion of host-derived Ang2 (26), neutralization of angiopoietins by soluble Tie2 receptors that function as decoys (27, 28), selective Ang2-binding aptamers (29, 30), and antibodies and peptibodies (peptide-Fc fusion proteins) that selectively neutralize the action of Ang2 (13, 25).

The goals of the present study were to gain a better understanding of the mechanisms that underlie the slowing of tumor growth by Ang2 inhibitors and determine whether these actions can add to the effects of inhibiting VEGF. First, we sought to learn whether inhibition of Ang2 reduces the blood supply of tumors by decreasing endothelial cell sprouting, increasing tumor vessel regression, or both. Second, we determined whether the slowing of tumor growth results from reduced tumor cell proliferation, increased tumor cell death, or both. Finally, we asked whether the effects of Ang2 inhibitors complement the effects of concurrent inhibition of VEGF on tumor vessels and tumor cells.

Toward this end, we used an Ang2-binding peptibody called L1-7(N) to block the action of Ang2 in human Colo205 colon carcinoma cells implanted in nude mice (13, 31). The Ang2 inhibitor was administered alone or in combination with an anti-human VEGF antibody that neutralized the action of tumor cell–derived VEGF. The Colo205 tumor model was used because of its documented sensitivity to Ang2 inhibitors (13).

CD1 nu/nu mice with implanted human Colo205 colon xenografts were treated with the Ang2-selective peptibody L1-7(N) (13) and/or monoclonal anti-human VEGF antibody with a median neutralization dose of 0.04 to 0.08 μg/mL against rat and human VEGF (clone 26503; R&D Systems). L1-7(N) has an IC₅₀ of 71 pmol/L against murine Ang2 and comparable inhibitory activity against human Ang2 (13) but does not bind Ang1, Ang3, Ang4 (13), or angiopoietin-like proteins 3 and 4.³

The treatment groups were (a) hFc and normal IgG2, (b) L1-7(N), (c) anti-VEGF antibody, and (d) L1-7(N) and anti-VEGF antibody together. Control protein or isotype control antibody was added to the treatment groups to match the total amount of protein delivered in the combination group. Tumor length, width, and height were measured with calipers, and the volume was calculated using the formula for volume of an ellipsoid. All experimental procedures were conducted in accordance with institutional guidelines established by the Institutional Animal Care and Use Committees at the University of California, San Francisco and Amgen, Inc.

Expression of murine and human Ang2, Tie2, VEGF, and VEGFR-2/KDR was measured by Taqman real-time PCR (RT-PCR) using species-specific probe sets (Supplementary Table). DNase-treated total RNA (100 ng per reaction) was extracted from Colo205 tumors and 22 other human tumor xenografts implanted s.c. in immunocompromised mice (four to five mice per tumor type). In vitro transcribed mRNA for each gene of interest was used as the standard for calculating absolute copy numbers. Taqman RT-PCR was performed with the Taqman EZ RT-PCR kit (Applied Biosystems).

Mice were perfused with fixative and tumors were processed and stained as previously described (3). Viable regions of tumor were marked by YO-PRO-1 (1:1,000; Invitrogen/Molecular Probes), which is a fluorescent dye that stains intact nuclei (32). Necrotic regions of tumors did not stain with YO-PRO-1. Endothelial cells were identified by hamster anti-CD31 antibody (clone 2H8, 1:500; Thermo Scientific). Basement membrane was marked with rabbit anti–type IV collagen antibody (1:20,000; CosmoBio). Proliferating cells were identified with rabbit anti–phosphohistone-H3 (1:1,000; Millipore/Upstate Biotechnology), and apoptotic cells were marked with rabbit anti-activated caspase-3 (1:1,000; R&D Systems). Secondary antibodies were goat anti-hamster or anti-rabbit IgG labeled with FITC or Cy3 (1:400; Jackson ImmunoResearch). Tumors were stained with H&E for conventional light microscopy. Specimens were examined with a Zeiss Axiophot fluorescence microscope or a Zeiss LSM510 confocal microscope as previously described (31).

Sections were cut roughly through the central of each tumor and stained with YO-PRO-1 to distinguish viable regions (YO-PRO-1 positive) from necrotic regions (YO-PRO-1 negative). Multiple low-magnification (2.5× objective) digital images, each measuring 5.1 × 3.8 mm, were used to create a montage of composite images in Adobe Photoshop. After adjusting the green fluorescence threshold, typically 25 to 30 (range, 0–255), the total number of YO-PRO-1–positive pixels was determined with ImageJ software (33). The area of viable tumor was calculated as the number of YO-PRO-1–positive pixels, each having an area of 64 μm².

Area density and total area of blood vessels (CD31), proliferating cells (phosphohistone-H3), and apoptotic cells (activated caspase-3) were measured in five digital images (1280 × 960 μm in size; 10× objective lens). Four of the images were taken in each quadrant of the tumor perimeter and one in the center of an 80-μm-thick section. Fractional area (area density) of CD31, phosphohistone-H3, or activated caspase-3 immunoreactivity was measured as the number of pixels above the fluorescence threshold (typically 15–30) within YO-PRO-1–positive regions (3, 34). The total area of CD31, phosphohistone-H3, or activated caspase-3 immunoreactivity was calculated as the product of the fractional area and the total area of YO-PRO-1–positive pixels. Values for total area of immunoreactivity were thus influenced by tumor size, but those for area density were not.

Endothelial cell sprouts were counted in real-time fluorescence microscopic images of tumor vessels of sections 60 μm in thickness stained for CD31 (10× objective; 2× Optovar) as previously described (31, 35). The area density of platelet-derived growth factor receptor-β (PDGFR-β)–positive pericytes within 10 μm of tumor vessels was measured on confocal microscopic images as previously described (31).

Values are expressed as the mean ± SE for four to five mice per group. Tumor growth curves reflect seven mice per group. Effect of treatment on tumor growth was calculated as the difference between the volume increase of experimental tumors and the volume increase of control tumors expressed as a percentage of the control tumor volume increase. The significance of differences was determined by ANOVA followed by the Fisher's pairwise least significant difference or Dunn-Bonferroni post hoc test. Differences in tumor growth curves were tested by repeated-measures ANOVA and Scheffé's post hoc test. P values of <0.05 were considered statistically significant.

Taqman RT-PCR measurements revealed that untreated human Colo205 xenografts in nude mice expressed abundant human and mouse Ang2 and VEGF (Table 1). Expression of human Ang2 and human and mouse VEGF in Colo205 tumors was near the median value of 23 different xenograft models, and expression of mouse Ang2 was the highest (Table 1). Human VEGF in Colo205 tumors had 5-fold the expression of mouse VEGF and 17-fold the expression of human Ang2. Expression of human and mouse Tie2 in all tumor types and of human VEGFR-2/KDR in some tumor types was below the threshold of detection (<1,000 copies).

Tumors in mice treated with hFc increased >3-fold in volume from an average of 277 mm³ on day 1 to 871 mm³ on day 23 of treatment (Fig. 1A, i). Tumors grew more slowly in the other groups. Compared with the hFc group, the increase in tumor volume was reduced 62% in the L1-7(N) group, 77% in the anti-VEGF antibody group, and 91% in the combination group (Fig. 1A, i). Tumor growth in the combination group was 76% less than in the L1-7(N) group and 61% less than in the anti-VEGF antibody group. At the end of 26 days of treatment, in comparison with the hFc controls, tumors weighed 36% less after L1-7(N), 44% less after anti-VEGF antibody, and 59% less after the inhibitor combination (Fig. 1A, ii).

Viable regions of tumors were identified by YO-PRO-1 staining of nuclei (Fig. 1B, i), which served as a fluorescent equivalent to conventional staining by H&E (Fig. 1B, ii). Control tumors had large viable regions, but most also had large necrotic regions (Fig. 1C; Supplementary Fig. S1). After L1-7(N) or anti-VEGF antibody, the tumors were smaller in overall size and had smaller viable regions (Fig. 1C and D, i); however, the amount of necrosis was about the same as in control tumors (Fig. 1C and D, ii). Tumor cross-sectional area was even smaller after treatment with both inhibitors due largely to reduced viable tumor. Measurements revealed that the total area of viable tumor at day 26, relative to hFc-treated tumors, was 40% less after L1-7(N), 45% less after anti-VEGF antibody, and 55% less after the inhibitor combination (Fig. 1D, i).

After the dual inhibitors, the amount of necrotic tumor was slightly less in absolute amount but represented a larger proportion of the total tumor mass (Fig. 1D, i and ii). Necrosis tended to be greater near the center and less at the periphery in all groups of tumors (Supplementary Fig. S1). Central necrosis was particularly common after anti-VEGF antibody and the inhibitor combination but was also found in the other treatment groups (Supplementary Fig. S1).

Cell proliferation and apoptosis were assessed in viable regions of tumors stained for phosphohistone-H3 and activated caspase-3, respectively. Tumor cells were the predominant cell type in Colo205 tumors and were the predominant cell stained by markers of proliferation and apoptosis. At day 26, phosphohistone-H3–immunoreactive cells were abundant in tumors treated with hFc (Fig. 2A, i) but were less numerous after L1-7(N) (Fig. 2A, ii), anti-VEGF antibody (Fig. 2A, iii), or the inhibitor combination (Fig. 2A, iv). The area density of phosphohistone-H3 staining was significantly less after treatment with any of the active agents than after hFc, with reductions of 68% after L1-7(N), 63% after anti-VEGF antibody, and 67% after the inhibitor combination (Fig. 2B). Similarly, the total area of phosphohistone-H3 staining was 83% less after L1-7(N), 81% less after anti-VEGF antibody, and 82% less after the inhibitor combination (Fig. 2C).

Apoptotic cells identified by activated caspase-3 immunoreactivity were located throughout the viable regions of tumors in all groups treated for 26 days (Fig. 3A). Apoptotic cells tended to be more abundant after L1-7(N) (Fig. 3A, ii), anti-VEGF antibody (Fig. 3A, iii), or the inhibitor combination (Fig. 3A, iv). The amount of activated caspase-3 staining, expressed as a percentage of tumor area in comparison with the hFc control, doubled after L1-7(N) (Fig. 3B). The increase in apoptosis was not significant after the anti-VEGF antibody (48%) but was significant after the inhibitor combination (63%; Fig. 3B). The total area of staining expressed in absolute units (mm²) was not significantly different among the four treatment groups (Fig. 3C), where increases in percentage of tumor stained were offset by decreases in tumor size.

Tumor blood vessels were examined to determine whether the decrease in proliferation and increase in apoptosis were attributable to vascular pruning. CD31-positive blood vessels in control tumors were abundant and tortuous, were uneven in size, and had numerous sprouts (Fig. 4A, i). Tumor vascular density was unchanged after treatment with L1-7(N) (Fig. 4A, ii), tended to be less after anti-VEGF antibody (Fig. 4A, iii), but was conspicuously reduced after treatment with the inhibitor combination (Fig. 4A, iv).

Measurements confirmed that the fractional area (area density) of CD31 immunoreactivity did not change after L1-7(N), had a small but not significant reduction after anti-VEGF antibody, and had a large (67%) and significant reduction after the inhibitor combination (Fig. 4B). When the fractional area of tumor vessels was scaled for tumor size (values expressed in square millimeters of tumor) to take into account treatment-related differences in overall tumor size, the values were 37% less after L1-7(N), 51% less after anti-VEGF antibody, and 82% less after the combination of inhibitors for 26 days (Fig. 4C). Values for the inhibitor combination were 71% less than those after L1-7(N) alone (Fig. 4C). In addition to the effects on total tumor vessel area, blood vessels in tumors treated with L1-7(N) and anti-VEGF antibody, alone or in combination, were smoother, had fewer sprouts, and were smaller and more uniform in size (Fig. 4A).

Under baseline conditions, tumor vessels had sparse coverage by PDGFR-β–immunoreactive pericytes, but after L1-7(N) or anti-VEGF antibody, the vessels had a thick envelopment of pericytes (Fig. 4D). Pericyte coverage after the inhibitor combination was intermediate in amount (Fig. 4D, iv). Measurements of PDGFR-β–positive pericytes within 10 μm of the endothelium reflected the differences evident by confocal microscopy (Fig. 4D, v).

The presence of empty basement membrane sleeves, indicative of blood vessel regression (3), was examined by costaining endothelial cells (CD31) and vascular basement membrane (type IV collagen) to determine whether the reductions in tumor vascularity were due to vessel regression or reduced angiogenesis. All treatment groups had scattered fragments of basement membrane (Fig. 5A, arrowheads), but empty basement membrane sleeves were more abundant after the anti-VEGF antibody or the inhibitor combination (Fig. 5A, iii and iv, arrows).

As an index of angiogenic activity, endothelial sprouts, identified as tapered endothelial processes that extended from the main axis of vessels, were common on blood vessels of control tumors (Fig. 5B, i). Sprouts were less numerous after L1-7(N) or anti-VEGF antibody (Fig. 5B, ii and iii) and rare after the combination of inhibitors (Fig. 5B, iv). Measurements of the number of sprouts revealed reduction of 46% after L1-7(N), 62% after anti-VEGF antibody, and 82% after the inhibitor combination (Fig. 5C).

This study sought to determine the effects on tumor cells and tumor blood vessels of the Ang2 inhibitor L1-7(N) and whether the changes complement the effects of inhibiting VEGF from tumor cells. Both the Ang2 and VEGF inhibitor decreased vessel sprouting and slowed tumor growth, but these changes were significantly greater when the two agents were combined than when either was given alone. The reduction in tumor growth was associated with decreased cell proliferation and increased apoptosis. Together, the results indicate that inhibition of Ang2 and VEGF together has additive effects on tumor growth and angiogenesis.

Inhibition of Ang2 or VEGF slowed tumor growth, and inhibition of both together slowed tumor growth even more. The amount of necrosis did not change much with treatment, but slowing of tumor growth was accompanied by reduction in the amount of viable tumor, reduced cell proliferation, and increased apoptosis. Tumor cell apoptosis also increases after dual inhibition of Ang1 and Ang2 (13). Increased tumor cell apoptosis without a measurable increase in amount of necrosis may reflect more efficient clearance of cell debris, but further studies are needed to understand the fate of tumor cells that die.

After inhibition of Ang2, the fractional area of proliferating cells decreased and the fractional area of tumor cell apoptosis increased, but the total area occupied by apoptotic cells did not change significantly because the higher frequency of apoptotic events was offset by the reduction in tumor size.

Because YO-PRO-1 stains all nuclei, the values for tumor area reflect all cellular constituents. The identity of proliferating and apoptotic cells was not determined, but the predominance of tumor cells in Colo205 tumors makes it reasonable to assume that most phosphohistone-H3–positive or activated caspase-3–positive cells were tumor cells. Taken together, our findings suggest that most of the reduction in tumor growth resulted from a combination of decreased tumor cell proliferation and increased tumor cell apoptosis.

Inhibition of Ang2 and VEGF together had additive effects on decreasing tumor volume but not on proliferation, apoptosis, or necrosis. However, readouts for cell proliferation and death were only a snapshot of their effect at the end of the experiment. Changes in rate of cell proliferation or death occurring earlier in the 26-day experiment, well in advance of when the measurements were made, could have had disproportionate effects in slowing tumor growth.

Necrosis was present in tumors of all treatment groups and tended to be greatest centrally. Central necrosis, which was especially prominent after anti-VEGF antibody and the inhibitor combination, is a common feature of implanted xenografts that is usually attributed to higher interstitial pressure and poorer blood flow at the center of tumors (36, 37). Less necrosis at the perimeter may also reflect well-perfused vasculature of the surrounding normal tissue, although other properties that favor resistance of this region to inhibitors cannot be excluded.

The present findings are consistent with the familiar view that tumor growth is restricted by expansion of its blood supply and that reduced blood vessel growth will slow tumor growth proportionately. We found that the tumor vascularity was reduced after inhibition of Ang2 and VEGF together but was not significantly different among the other treatment groups. However, the total area of tumor blood vessels, scaled for tumor size, was reduced after Ang2 inhibition alone or after VEGF inhibition alone and was reduced even more when treatments were combined. These findings suggest that tumors grew synchronously with their blood vessels and maintained a constant ratio of blood vessels to tumor area, unless perturbed by inhibitors of both Ang2 and VEGF.

Angiogenesis inhibitors can reduce the number of tumor vessels by inhibition of endothelial sprouting and new vessel growth (35, 38) or by causing regression of existing vessels (3, 39). In this study, we sought to determine whether inhibition of Ang2 and VEGF together had additive effects on the reduction of tumor vascularity and growth. The present findings are consistent with previous results (31), showing that inhibition of Ang2 was followed by reduced endothelial sprouting without much regression of tumor vessels. Ang2 inhibition also increased the pericyte coverage of tumor vessels (31). These actions are consistent with the concept that Ang2 acts as a partial antagonist of Tie2 signaling, whereby Ang2 promotes vessel destabilization, unlike Ang1, which promotes vessel stabilization (14, 15). Together, these results indicate that Ang2 inhibition suppressed the growth of tumor vessels by blocking the destabilizing action of Ang2 on tumor vessels and unmasking the stabilizing action of Ang1.

In contrast, VEGF inhibition led to regression of tumor blood vessels and reduced endothelial sprouting, as found in previous studies (3, 40). Inhibition of VEGF and Ang2 together reduced the number of endothelial sprouts to an extent greater than either inhibitor alone. However, the appearance of empty basement membrane sleeves was evident only after the anti-VEGF antibody or the inhibitor combination. Thus, the greater reduction in tumor vessels after inhibition of VEGF and Ang2 together is likely to result from additive effects on endothelial sprout suppression and tumor vessel regression.

Recent studies suggest that targeting VEGF alone, although effective in eliminating some tumor blood vessels, only temporarily halts tumor growth and may even promote tumor aggressiveness and metastasis (40, 41). Anti-VEGF therapy is most effective when combined with chemotherapy or radiation (42). Another potentially beneficial strategy is to target multiple receptor tyrosine kinases in the hope of killing more tumor blood vessels than with anti-VEGF treatment alone. Multikinase inhibitors that target VEGF, PDGF, and other receptors were developed for this purpose and can eliminate more tumor blood vessels than anti-VEGF treatment alone (43–45). Inhibition of VEGF and PDGF together is more efficacious against all stages of pancreatic islet tumors in RIP-Tag2 mice than either inhibitor alone (46). The present findings indicate that a reasonable multitargeted approach would be to combine inhibitors of Ang2 with inhibitors of VEGF signaling.

The downstream effects of angiogenesis inhibitors on tumor cells via blood vessels are incompletely understood, but these inhibitors can have complex effects on tumor growth (4, 34). Angiogenesis inhibitors are thought to slow tumor growth by eliminating blood vessels. VEGF inhibition reduces growth of some human cancers by stopping angiogenesis, destroying existing tumor vessels, and reducing blood flow (47, 48). Inhibition of PDGF-B can increase tumor growth in a preclinical model despite the regression of 80% of the blood vessels (34). Some angiogenesis inhibitors that clearly reduce tumor vascularity can paradoxically accelerate tumor invasiveness and growth (40, 41). Both the quantity and the quality of tumor vessels determine whether tumors grow or regress. The results of the present study revealed that tumor vascularity and tumor growth are tightly coupled in Colo205 tumors. Inhibition of Ang2 or VEGF alone led to decreased tumor vascularity and reduced tumor size, and the combination had complementary effects.

By using species-specific probe sets, expression in tumors of human Ang2, Tie2, VEGF, and VEGFR-2/KDR in human Colo205 tumor cells could be measured separately from the corresponding mouse proteins in stromal cells. Measurements by Taqman RT-PCR revealed that expression of human tumor cell–derived VEGF was higher than mouse VEGF, but expression of mouse stromal cell–derived Ang2 and VEGFR-2 was higher than human. As there are far greater numbers of tumor cells than endothelial cells, these species differences in Ang2 and VEGFR-2 expression would be magnified if determined on a per-cell basis. These observations fit with published evidence for Ang2 expression in blood vessels of Colo205 tumors (13) and similar unpublished data for VEGFR-2 in these tumors. Tie2 expression was not detectable by quantitative RT-PCR or in situ hybridization, consistent with low expression or sparse distribution of Tie2 expression in endothelial cells, monocytes, or macrophages in Colo205 tumors. Together, these results are consistent with direct effects of Ang2 and VEGF inhibitors mainly on blood vessels in tumors. The reduction in tumor cell proliferation and increase in apoptosis are likely secondary to the vascular changes.

Inhibition of Ang2 may have additional beneficial effects in cancer therapy, which are mediated by mechanisms that are separate from the destabilizing effect of Ang2 on tumor vessels. Recent studies have reported that tumor-associated monocytes or macrophages express Tie2 receptors, can be recruited and/or activated by angiopoietins, and promote tumor angiogenesis by release of proteases that liberate sequestered VEGF from tumor matrix (49, 50). Further studies will be needed to determine whether inhibition of Ang2 signaling suppresses a macrophage-dependent mechanism that complements the direct effects of Ang2 inhibition on endothelial cells.

Possible mechanisms that explain the effects of inhibitors of Ang2 or VEGF, given alone or in combination, on tumor growth and angiogenesis are illustrated in Fig. 5D. Ang2 inhibition prevents the growth of new vessels by endothelial sprout formation. Reduced sprouting leads to less tumor cell proliferation, more tumor cell apoptosis, and smaller tumors. VEGF inhibition is followed by tumor vessel regression, the formation of empty basement membrane sleeves, and reduced tumor growth. Blockade of Ang2 and VEGF together has additive effects on sprouting and vessel regression and produces greater slowing of tumor growth. Together, our findings suggest that Ang2 inhibitors and VEGF inhibitors have complementary antiangiogenic actions and effects on reducing tumor growth.

A. Coxon, D. Yu, J.V. Bready, and J.D. Oliner: employees and stockholders of Amgen, Inc. D.M. McDonald: commercial research grant, Amgen. The other authors disclosed no potential conflicts of interest.

We thank Jeyling Chou, Jennifer Feng, Talia Romano, Anne Saiki, and Ji-Rong Sun for technical assistance and Tatsuma Okazaki, Barbara Sennino, and Weon-Kyoo You for valuable discussions.

Grant Support: NIH/National Cancer Institute grant CA82923; NIH/National Heart, Lung, and Blood Institute grants HL24136, HL59157, and HL96511; Amgen; and AngelWorks Foundation (D.M. McDonald).

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.