Research ArticleCancer

Dual angiopoietin-2 and VEGFA inhibition elicits antitumor immunity that is enhanced by PD-1 checkpoint blockade

See allHide authors and affiliations

Science Translational Medicine  12 Apr 2017:
Vol. 9, Issue 385, eaak9670
DOI: 10.1126/scitranslmed.aak9670

Antitumor attack on two fronts

The use of immune checkpoint inhibitors and other immunotherapies for the treatment of cancer is continuing to expand as these drugs demonstrate effectiveness in progressively more cancer types and therapeutic contexts. At the same time, the drugs are not perfect, and not all patients respond to them, so a key subject of research in this field is determining optimal ways to combine immune checkpoint therapies with other cancer treatments. Schmittnaegel et al. and Allen et al. focused their studies on the combination of antiangiogenic treatments with checkpoint inhibitors. The authors demonstrated how inhibition of tumor angiogenesis can facilitate the access of cytotoxic T cells to tumors, while the checkpoint inhibitors protect these T cells from exhaustion, enhancing their antitumor effects.

Abstract

Pathological angiogenesis is a hallmark of cancer and a therapeutic target. Vascular endothelial growth factor A (VEGFA) and angiopoietin-2 (ANGPT2; also known as ANG2) are proangiogenic cytokines that sustain tumor angiogenesis and limit antitumor immunity. We show that combined ANGPT2 and VEGFA blockade by a bispecific antibody (A2V) provided superior therapeutic benefits, as compared to the single agents, in both genetically engineered and transplant tumor models, including metastatic breast cancer (MMTV-PyMT), pancreatic neuroendocrine tumor (RIP1-Tag2), and melanoma. Mechanistically, A2V promoted vascular regression, tumor necrosis, and antigen presentation by intratumoral phagocytes. A2V also normalized the remaining blood vessels and facilitated the extravasation and perivascular accumulation of activated, interferon-γ (IFNγ)–expressing CD8+ cytotoxic T lymphocytes (CTLs). Whereas the antitumoral activity of A2V was, at least partly, CTL-dependent, perivascular T cells concurrently up-regulated the expression of the immune checkpoint ligand programmed cell death ligand 1 (PD-L1) in tumor endothelial cells. IFNγ neutralization blunted this adaptive response, and PD-1 blockade improved tumor control by A2V in different cancer models. These findings position immune cells as key effectors of antiangiogenic therapy and support the rationale for cotargeting angiogenesis and immune checkpoints in cancer therapy.

INTRODUCTION

Tumor angiogenesis—the formation of a tumor-associated vascular network—is essential for cancer growth and progression (1, 2). Vascular endothelial growth factor A (VEGFA) is a key promoter of tumor angiogenesis (3). The anti-VEGFA monoclonal antibody (mAb) bevacizumab was the first antiangiogenic drug to be approved for clinical use and is currently indicated for several cancer types, typically in the inoperable/metastatic setting and in combination with chemotherapy. However, the benefits of adding bevacizumab to standard-of-care therapies are, at best, measured in several months of increased progression-free survival (2). The lack of durable antitumoral responses after VEGFA inhibition mirrors that of other targeted drugs and can be attributed, at least in part, to compensatory proangiogenic signaling, either induced or preexisting in tumors (4, 5).

Angiopoietin-2 (ANGPT2) is a proangiogenic factor and target of antiangiogenic therapy (614). In mouse tumor models, the inhibition of ANGPT2 (i) decreases angiogenesis and increases pericyte coverage and maturation of the remaining blood vessels; (ii) inhibits tumor growth, often more prominently so when combined with anti-VEGFA drugs; and (iii) delays the onset of tumor resistance to anti-VEGFA therapy in some cancer types, suggesting that VEGFA and ANGPT2 function nonredundantly during tumor angiogenesis (7, 10, 11). Furthermore, ANGPT2 blockade markedly inhibits cancer metastasis via effects on both primary and metastatic tumors (6, 8, 15). Therefore, ANGPT2 inhibition may improve the therapeutic benefits of VEGFA neutralization in patients with cancer.

Cancer immunotherapies are treatments that restore or enhance the ability of the host immune system to recognize and eradicate cancer (16). These therapies either stimulate the functions of specific components of the immune system or counteract the signals—emanating from cancer cells or tumor-associated stromal cells—that suppress antitumor immune responses. In particular, mAbs that intercept signaling of inhibitory “immune checkpoints” expressed on T cells, such as programmed cell death protein 1 (PD-1) and cytotoxic T lymphocyte-associated antigen-4 (CTLA-4), have clinical efficacy in patients with immunogenic tumors, most notably melanoma and mismatch repair–deficient colorectal cancer (17, 18). Anti–PD-1 mAbs can reactivate T cells by removing an important source of T cell–suppressive signals conveyed by PD-1 ligands (PD-L1 and PD-L2) expressed on cancer cells and a variety of tumor-associated cells, including endothelial cells (ECs) and macrophages (17, 18).

Increasing data indicate that some of the structural and molecular features of the tumor blood vessels can limit the extravasation and antitumoral functions of T cells, including CD8+ cytotoxic T lymphocytes (CTLs), which are key effectors of antitumor immunity. Therefore, inhibiting proangiogenic signaling may facilitate the deployment of antitumor immunity to tumors and enhance the efficacy of cancer immunotherapies (1922). However, proof-of-concept demonstration of this notion remains scant. Here, we show that a murinized mAb that efficiently blocks both ANGPT2 and VEGFA, termed A2V (12, 23), promotes antitumor immunity and sensitizes mouse tumors to PD-1 checkpoint blockade (24) better than single ANGPT2 or VEGFA neutralization by LC06 (13, 15) and B20 (25), respectively. Our preclinical findings position dual ANGPT2 and VEGFA blockade as a tumor-conditioning strategy that may unleash or increase the efficacy of anti–PD-1/PD-L1 mAbs for cancer immunotherapy.

RESULTS

Combined ANGPT2 and VEGFA blockade provides marked survival benefits in both transgenic and transplant tumor models

We studied tumor responses to combined or single ANGPT2 and VEGFA blockade in six mouse cancer models: (i) transgenic MMTV-PyMT mice (FVB/n background), which develop metastatic breast cancer (6, 26); (ii) MMTV-PyMT tumors inoculated orthotopically in the mammary fat pad of FVB/n mice; (iii) 4T1 mammary tumors inoculated orthotopically in the mammary fat pad of Balb/c mice; (iv) transgenic RIP1-Tag2 mice (C57Bl/6 background), which develop pancreatic neuroendocrine tumors (7); (v) B16 melanomas expressing the chicken ovalbumin (OVA) surrogate neoantigen (B16-OVA), inoculated subcutaneously in C57Bl/6 mice (27); and (vi) MC38-OVA colorectal adenocarcinomas, inoculated subcutaneously in C57Bl/6 mice (28).

We treated 9-week-old MMTV-PyMT (Fig. 1A) or 11-week-old RIP1-Tag2 (Fig. 1B) transgenic mice carrying early established tumors (6, 7) with weekly, stoichiometrically matched doses of A2V, B20, or LC06. Control mice received isotype-matched IgGs, which were validated in the B16-OVA model (fig. S1, A to C). A2V, but not B20 or LC06 monotherapy, markedly extended the survival of both MMTV-PyMT and RIP1-Tag2 transgenic mice. Compared to B20 or LC06, A2V also improved tumor growth inhibition in mice with orthotopic MMTV-PyMT mammary tumor transplants (Fig. 1C). Notably, both spontaneous and transplanted MMTV-PyMT tumors responded rapidly to A2V treatment (Fig. 1C and fig. S1, D and E). A2V considerably extended the survival of C57Bl/6 mice carrying subcutaneous B16-OVA melanomas, whereas B20 provided more limited therapeutic benefits, and LC06 had no measurable antitumoral activity in this model (Fig. 1D and fig. S1F). Finally, A2V also showed antitumoral activity in the MC38-OVA tumor model (fig. S1G). Together, these results illustrate enhanced antitumoral activity of combined ANGPT2 and VEGFA blockade, as compared with single-agent therapies, in cancer models of diverse origin and genetic background.

Fig. 1. Combined ANGPT2 and VEGFA blockade delays tumor progression.

(A) Kaplan-Meier survival curves of MMTV-PyMT transgenic mice treated as indicated (see Materials and Methods and Supplementary Materials and Methods for experimental design and drug regimens). Median (50%) survival (after treatment): immunoglobulin G (IgG), 48 days; B20, 47 days; LC06, 55 days; and A2V, 75 days. Number of experimental mice: IgG, n = 11; B20, n = 8; LC06, n = 7; and A2V, n = 14. Statistical analysis by log-rank test. Exact Pvalues are reported in table S2. (B) Kaplan-Meier survival curves of Rip1-Tag2 transgenic mice treated as indicated. Median (50%) survival (after treatment): IgG, 35 days; B20, 33 days; LC06, 39 days; and A2V, 49 days. Number of experimental mice: IgG, n = 15; B20, n = 12; LC06, n = 5; and A2V, n = 21. Statistical analysis as in (A). (C) Individual volumes of orthotopic MMTV-PyMT tumors treated as indicated. Mice were euthanized when the mean tumor size reached 1000 mm3 (IgG group) or 50 days after tumor inoculation (other groups). IgG, n = 8; B20, n = 8; LC06, n = 9; and A2V, n = 9. Arrow indicates start of treatment. (D) Kaplan-Meier survival curves of mice with subcutaneous B16-OVA tumors treated as indicated. Median (50%) survival (after treatment): IgG, 10 days; B20, 15 days; LC06, 13 days; and A2V, 45 days. Number of experimental mice: IgG, n = 10; B20, n = 9; LC06, n = 10; and A2V, n = 9. Statistical analysis by log-rank test. (E) Number of metastases (left) and total metastatic area (right) in the lungs of MMTV-PyMT transgenic mice from the experiment in (A). Each dot represents one mouse. IgG, n = 10; B20, n = 6 to 7; LC06, n = 6; and A2V, n = 6. Statistical analysis by Kruskal-Wallis test with Dunn’s correction for multiple comparisons. Red asterisks indicate analysis by Mann-Whitney test applied to the two indicated groups.

Unlike B20, LC06 suppressed spontaneous lung metastasis in transgenic MMTV-PyMT mice analyzed at the survival end point (Fig. 1E), in agreement with previous findings obtained with a different anti-ANGPT2 mAb (6). A2V further reduced the incidence and burden of lung metastases compared to LC06. Notably, most of the A2V-treated mice were virtually free of detectable metastases (in spite of the increased time for potential metastatic dissemination and colonization), whereas some of the LC06-treated mice had more and larger lung deposits. Together, these results indicate that A2V provides important survival benefits and blocks spontaneous metastasis in mouse models of cancer that are poorly responsive to anti-VEGFA or anti-ANGPT2 monotherapy.

Combined ANGPT2 and VEGFA blockade impairs tumor angiogenesis, increases tumor necrosis, and normalizes the remaining blood vessels

We analyzed tumor vascularization by immunostaining of tumor sections, flow cytometry of dispersed tumors, or intravital delivery of fluorescently labeled lectin. A2V decreased the relative CD31+ vascular area in the tumors of transgenic MMTV-PyMT (Fig. 2, A and B) and RIP1-Tag2 (Fig. 2C) mice analyzed at the termination end point. Notably, the relative tumor vascular area was only slightly reduced in B20- and LC06-treated mice, consistent with the onset of resistance to the antiangiogenic monotherapies during extended treatment regimens (4, 7). These findings indicate sustained antiangiogenic activity of A2V, but not B20 or LC06, in both MMTV-PyMT and RIP1-Tag2 mice.

Fig. 2. Combined ANGPT2 and VEGFA blockade inhibits tumor angiogenesis, increases necrosis, and normalizes the remaining blood vessels.

(A) Relative CD31+ area in transgenic MMTV-PyMT tumors. Each dot indicates one tumor per mouse and represents the average of 8 to 15 images. IgG, n = 6; B20, n = 7; LC06, n = 4; and A2V, n = 5. Statistical analysis by one-way analysis of variance (ANOVA) with Tukey’s correction for multiple comparisons. (B) Representative images of CD31 immunostaining (green) and 4′,6-diamidino-2-phenylindole (DAPI) nuclear staining (blue) of transgenic MMTV-PyMT tumors treated as indicated. Scale bars, 100 μm. (C) Relative CD31+ area in transgenic RIP1-Tag2 tumors. Each dot indicates one tumor and represents the average of 8 to 15 images; several tumors are present in each mouse. IgG, n = 15; B20, n = 32; LC06, n = 14; and A2V, n = 16. Statistical analysis by one-way ANOVA with Tukey’s correction for multiple comparisons. (D) Flow cytometry analysis of CD31+ ECs in orthotopic MMTV-PyMT tumors treated as indicated. Each dot indicates one tumor. IgG, n= 9; B20, n = 9; LC06, n = 8; and A2V, n = 8. Statistical analysis by one-way ANOVA with Tukey’s correction for multiple comparisons. (E) Relative CD31+ area in subcutaneous B16-OVA tumors. Each dot indicates one tumor per mouse and represents the average of 8 to 15 images. IgG, n = 6; B20, n = 6; LC06, n = 6; and A2V, n = 6. Statistical analysis by one-way ANOVA with Tukey’s correction for multiple comparisons. (F) Relative necrotic area in orthotopic MMTV-PyMT tumors. Each dot indicates one tumor. IgG, n = 5; A2V, n = 4. Statistical analysis by unpaired Student’s t test. Representative images of whole-tumor sections are shown on the right; necrotic areas appear brown. (G) Representative images of CD31 immunostaining (green) and DAPI nuclear staining (blue) of orthotopic MMTV-PyMT tumors treated as indicated. N, necrotic area. Scale bars, 200 μm. (H) Representative images of CD31 (green), α-smooth muscle actin (α-SMA) or NG2 (red) immunostaining, and DAPI nuclear staining (blue) of transgenic MMTV-PyMT tumors treated as indicated. Scale bars, 100 μm. (I and J) Relative proportion of α-SMA+ [(I) and (J)] or NG2+ (I) pericyte-covered blood vessels in transgenic (I) or orthotopic (J) MMTV-PyMT tumors treated as indicated. Each dot indicates one tumor and represents the average of 8 to 15 images. (I) IgG, n = 4; A2V, n = 6. (J) IgG, n = 5; A2V, n = 5. Statistical analysis by unpaired Student’s t test. (K) Representative images of CD31 (green), α-SMA or NG2 (red) immunostaining, and DAPI nuclear staining (blue) of orthotopic MMTV-PyMT tumors treated as indicated. Scale bars, 100 μm. (L) Relative proportion of NG2+ pericyte-covered blood vessels in orthotopic MMTV-PyMT tumors treated as indicated. Each dot indicates one tumor and represents the average of two to eight images. IgG, n = 8; B20, n = 8; LC06, n = 8; and A2V, n = 6. Statistical analysis by one-way ANOVA with Tukey’s correction for multiple comparisons.

We also examined transplant tumor models. A2V decreased the frequency of CD31+ tumor ECs, as well as the relative CD31+ or lectin+ vascular area, in orthotopic MMTV-PyMT (Fig. 2D and fig. S2A), subcutaneous B16-OVA (Fig. 2E), and subcutaneous MC38-OVA (fig. S2B) tumors. The marked antiangiogenic effects of A2V were comparable to those achieved by a combination of B20 and LC06 antibodies in 4T1 mammary tumors (fig. S2C) and were superior to those of B20 or LC06 monotherapy in each tumor model tested. Likely as a consequence of acute vascular pruning and tissue ischemia, A2V-treated tumors displayed larger necrotic areas than IgG-treated tumors at early stages of tumor growth (Fig. 2, F and G).

We then analyzed the features of the tumor blood vessels after antiangiogenic therapy. To assess vascular maturation, we visualized pericytes by either NG2 or α-SMA immunostaining in transgenic and transplant MMTV-PyMT tumors. In both models, A2V increased pericyte coverage of the surviving tumor blood vessels (Fig. 2, H to L), which were mostly found at the tumor periphery or in perinecrotic and stromal tumor areas (see Fig. 2G above). Increased blood vessel coverage by pericytes in A2V-treated tumors was largely dependent on ANGPT2 blockade. Both A2V and LC06, but not B20, increased the proportion of pericyte-covered blood vessels in orthotopic MMTV-PyMT tumors and, to a lesser extent, in subcutaneous B16-OVA melanomas (fig. S2, D and E), consistent with previous studies that analyzed vascular phenotypes after the pharmacological inhibition or genetic deletion of ANGPT2 in tumors (68, 14, 29). These results indicate that combined ANGPT2 and VEGFA blockade effectively suppresses angiogenesis while increasing pericyte coverage of the remaining blood vessels in different mouse tumor models.

Combined ANGPT2 and VEGFA blockade activates tumor-infiltrating CD8+ T cells

Increased pericyte coverage of tumor blood vessels is typically associated with improved vascular function, a process referred to as “vascular normalization” (30). In addition to improving tumor oxygenation and drug penetration, vascular normalization may facilitate the trafficking and functions of effector T cells, including CD8+ CTLs, in the tumors (19, 20). We analyzed T cell infiltration and activation in the tumors. None of the antiangiogenic drugs significantly altered the proportions of CD8+ and CD4+ T cells within the CD45+ hematopoietic cell compartment of transgenic or transplant MMTV-PyMT tumors (Fig. 3, A and B, and fig. S3A). We noted, however, a relative expansion of CD11chiMHCIIhi M1-like tumor-associated macrophages (TAMs) and dendritic cells (DCs) (28) in the A2V-treated tumors (Fig. 3C), consistent with previous findings in a mouse glioma model (23). A2V increased the proportion of CD8+ CTLs that expressed an activated, interferon-γ (IFNγ)–positive or CD69+ phenotype in both transgenic and transplanted mammary tumors, whereas neither B20 nor LC06 appreciably altered such parameters of T cell activation (Fig. 3, D to F). On the other hand, the analysis of activation markers in the CD4+ T cell compartment revealed a more nuanced picture, with no change in CD69+ but increased IFNγ+ cells after A2V specifically in the transplanted mammary tumors.

Fig. 3. Combined ANGPT2 and VEGFA blockade activates intratumoral immune cells.

(A to K) Flow cytometry analysis of the indicated cell types (A to C and G), intracellular IFNγ (D, E, and H), CD69 expression (F and I), OVA-specific (dextramer+) Tcells (J), or SIINFEKL OVA peptide in complex with major histocompatibility complex class I (MHCI) (K), in transgenic MMTV-PyMT (A and D), orthotopic MMTV-PyMT (B, C, E, and F), and subcutaneous B16-OVA (G to K) tumors treated as indicated. Each dot indicates one tumor. MFI indicates the mean fluorescence intensity [in arbitrary units (a.u.)] of the indicated marker. The right panels in (H) show representative flow cytometry analyses of intracellular IFNγ in gated CD8+ T cells. The right panel in (K) shows representative flow cytometry analyses of H-2Kb/SIINFEKL staining in CD11bhiMHCIIhi cells. (A) Left: IgG, n = 12; A2V, n = 12. Right: IgG, n = 12; A2V, n = 9. (B) Left: IgG, n= 9; B20, n = 9; LC06, n = 8; and A2V, n = 8. Right: IgG, n = 7; B20, n = 9; LC06, n = 8; and A2V, n = 8. (C) IgG, n = 9; B20, n = 8; LC06, n = 8; and A2V, n = 8. (D) Left: IgG, n = 10; A2V, n = 13. Right: IgG, n = 12; A2V, n = 12. (E) Left: IgG, n= 9; B20, n = 8; LC06, n = 8; and A2V, n = 8. Right: IgG, n = 9; B20, n = 9; LC06, n = 8; and A2V, n = 8. (F) IgG, n = 9; B20, n= 9; LC06, n = 8; and A2V, n = 8. (G to I) IgG, n = 8; B20, n = 7; LC06, n = 6; and A2V, n = 9. (J) IgG, n = 7; B20, n = 7; LC06, n = 8; and A2V, n = 8. (K) IgG, n= 8; B20, n = 7; LC06, n = 8; and A2V, n = 9. Statistical analysis by unpaired Student’s t test (A and D) or one-way ANOVA with Tukey’s correction for multiple comparisons (B, C, and E to K). Red asterisks in (H) and (K) indicate analysis by unpaired Student’s t test applied to the two indicated groups. (L) Flow cytometry analysis of latex+ APCs after ex vivo incubation of total tumor-derived cells with latex beads. Each dot represents one tumor. The right panels show representative flow cytometry analysis of cells that ingested latex beads. IgG, n = 7; B20, n = 6; LC06, n = 8; and A2V, n = 8. Statistical analysis by one-way ANOVA with Tukey’s correction for multiple comparisons. FITC, fluorescein isothiocyanate.

We then analyzed CD8+ CTLs in immunogenic B16-OVA melanomas. Although the relative CTL abundance did not vary across the experimental groups (Fig. 3G), A2V-treated melanomas had greater proportions of CTLs expressing an IFNγ+ or CD69+ phenotype compared to IgG-treated tumors (Fig. 3, H and I), consistent with findings in the mammary tumors. Furthermore, A2V increased by about fivefold the proportion of intratumoral, OVA-specific CTLs (Fig. 3J and fig. S1C). Although A2V had more profound activating effects on T cells than B20 or LC06, we observed an expansion of OVA-specific CTLs also in tumors treated with B20 or LC06, suggesting that both VEGFA and ANGPT2 inhibition contributes to activate the CTLs, at least in this immunogenic tumor model. Together, our observations in mammary tumor and melanoma models suggest that VEGFA and ANGPT2 blockade, especially when combined, stimulates the activation of CD8+ CTLs, including tumor antigen–specific CTLs.

Combined ANGPT2 and VEGFA blockade increases tumor antigen presentation

To identify potential mechanisms that improved CTL activation in A2V-treated tumors, we analyzed CD11chiMHCIIhi M1-like TAMs and DCs, which encompass professional antigen-presenting cells (APCs), in the B16-OVA melanomas. A2V, but not B20 or LC06, increased the proportion of APCs that cross-presented the MHCI-restricted, OVA-derived SIINFEKL peptide (Fig. 3K and fig. S1B). Furthermore, APCs isolated from A2V-treated and, to a lesser extent, B20- or LC06-treated tumors, displayed enhanced phagocytic activity, as shown by the ingestion of latex beads ex vivo (Fig. 3L). Therefore, combined ANGPT2 and VEGFA blockade increased both MHCI-mediated antigen cross-presentation, which may be conducive to the activation of tumor-specific CD8+ CTLs, and the phagocytic activity of tumor-associated APCs, which primarily stimulate CD4+ T helper functions through MHCII-mediated antigen presentation (31). These results support the notion that A2V can reprogram the intratumoral immune cell compartment toward increased tumor antigen presentation and T cell activation.

Combined ANGPT2 and VEGFA blockade promotes perivascular T cell accumulation

As mentioned above, A2V did not alter the relative abundance of either CD8+ or CD4+ T cells in the hematopoietic cell infiltrate of transplant MMTV-PyMT tumors. However, it increased their absolute numbers specifically in the perivascular space of both transplant MMTV-PyMT (Fig. 4, A to C, and fig. S3B) and MC38-OVA (Fig. 4D) tumors, as shown by immunofluorescence staining and morphometric analysis of tumor sections. Notably, neither B20 nor LC06 increased perivascular CD8+ or CD4+ T cells in the mammary tumors. Also, none of the antiangiogenic treatments significantly altered the overall density of CD8+ T cells in the viable areas (comprising both perivascular and nonperivascular spaces) of either mammary (fig. S3C) or MC38-OVA (fig. S3D) tumors. On the other hand, there were more abundant CD4+ T cell infiltrates in the viable regions of A2V-treated mammary tumors; these cells occasionally formed discrete clusters at the tumor periphery (fig. S3E). In summary, these results indicate that A2V promotes the extravasation and perivascular accumulation of T cells, likely as a consequence of its normalizing effects on the tumor blood vessels surviving antiangiogenic therapy.

Fig. 4. Combined ANGPT2 and VEGFA blockade promotes the extravasation and perivascular accumulation of T cell in tumors.

(A) Representative images of CD8+ or CD4+ (green) and CD31 (red) immunostaining and DAPI nuclear staining (blue) in orthotopic MMTV-PyMT tumors treated as indicated. Scale bars, 100 μm. (B to D) Quantification of perivascular CD8+ or CD4+ T cells in orthotopic MMTV-PyMT (B and C) or subcutaneous MC38-OVA (D) tumors treated as indicated. Each dot indicates one tumor and represents the mean of 8 to 15 images per tumor. (B) IgG, n = 5; A2V, n = 5. (C) IgG, n = 7; B20, n = 8; LC06, n = 7; and A2V, n = 6. (D) IgG, n = 7; A2V, n = 5. In (C), the IgG-treated mice were euthanized at day 21 after the initiation of therapy, whereas the other mice were euthanized at day 38. Statistical analysis by unpaired Student’s t test (B and D) and one-way ANOVA with Tukey’s correction for multiple comparisons (C). Red asterisk in (C) indicates analysis by unpaired Student’s t test applied to the two indicated groups.

Combined ANGPT2 and VEGFA blockade induces a proinflammatory gene signature in tumor-associated ECs

Although combined neutralization of ANGPT2 and VEGFA stimulated immune cell activation and slowed tumor growth to extend mouse survival, it did not regress tumors. These observations may suggest the occurrence of immunosuppressive mechanisms limiting the full expression of A2V-elicited antitumor immunity. Increasing experimental evidence has implicated myeloid cells—namely, macrophages, neutrophils, and subsets thereof—as a source of immunosuppressive cues in the tumor microenvironment (32, 33). However, immunosuppressive signals may also emanate from activated adaptive immune cells through negative feedback regulatory mechanisms, such as the induction of immune checkpoints (for example, PD-L1) in response to T cell–derived IFNγ (34). This notion led us to hypothesize that the vascular ECs of A2V-treated tumors had acquired immunosuppressive properties, possibly in response to IFNγ-expressing perivascular T cells. We therefore studied the effects of A2V on the transcriptomes of tumor ECs and other cell types of interest by RNA sequencing (RNA-seq) analysis. We used fluorescence-activated cells sorting (FACS) to isolate vascular ECs, TAMs, and CD4+ or CD8+ T cells (n = 4 samples per cell type per treatment) from transplanted MMTV-PyMT tumors treated with A2V or control IgGs. The analysis of selected, cell type–specific genes confirmed the purity of the sorted cells (Fig. 5A and fig. S4A).

Fig. 5. Combined ANGPT2 and VEGFA blockade induces a proinflammatory gene signature in tumor-associated ECs.

(A) RNA-seq analysis of gene transcripts of interest in ECs, CD4+ and CD8+ T cells, and TAMs sorted from orthotopic MMTV-PyMT tumors 5 days after treatment (n = 4 samples per cell type per treatment). Data are displayed as log2-transformed RPKM (reads per kilobase per million mapped reads) values. (B and C) RNA-seq analysis of ECs sorted as in (A). The schematics show the predicted upstream regulators of genes differentially expressed in A2V- versus IgG-treated tumor ECs, according to IPA, using red/blue tones for activatory/inhibitory effects. (D) RNA-seq analysis of gene transcripts of interest in ECs sorted as in (A). Statistical analysis by unpaired Student’s t test. (E) Flow cytometry analysis of PD-L1 expression in the ECs of orthotopic MMTV-PyMT tumors treated as indicated. Each dot represents one mouse. IgG, n = 5; A2V, n = 5. Statistical analysis by unpaired Student’s t test. (F) PD-L1 MFI in CD31+ blood vessels of orthotopic MMTV-PyMT tumors treated as indicated. Each dot indicates one tumor. IgG, n = 5; A2V, n = 5. Statistical analysis by unpaired Student’s t test. IF, immunofluorescence. (G and H) Flow cytometry analysis of PD-L1 expression in the ECs of transgenic MMTV-PyMT tumors (G) or subcutaneous B16-OVA melanomas (H) treated as indicated. Each dot represents one mouse. (G) IgG, n = 13; A2V, n = 11. (H) IgG, n = 7; A2V, n = 9. Statistical analysis by unpaired Student’s t test. (I) Correlation between the number of perivascular CD8+ T cells and the PD-L1 MFI in CD31+ blood vessels of orthotopic MMTV-PyMT tumors treated as indicated. IgG, n = 5; A2V, n = 5. Statistical analysis by linear regression analysis. (J) Representative images of CD8 (green), CD31 (red), and PD-L1 (white) immunostaining and DAPI nuclear staining (blue) of orthotopic MMTV-PyMT tumors treated as indicated. Scale bars, 100 μm. (K) Quantitative polymerase chain reaction (qPCR) analysis of Pdcg (PD-L1), Stat1, or Cxcl10 in lysates of orthotopic MMTV-PyMT tumors treated as indicated. Data indicate the mean fold change over the reference sample (IgG) after normalization to the average expression of Hprt and Gapdh. Each dot represents one tumor. IgG (including mouse and rat IgG), n = 6; A2V (including rat IgG), n = 6; IgG plus α-IFNγ, n = 5; and A2V plus α-IFNγ, n= 5 to 6. Statistical analysis by one-way ANOVA with Tukey’s correction for multiple comparisons.

We then focused on tumor ECs, which are the first cellular barrier regulating T cell extravasation to tumors (19, 20). Ingenuity Pathway Analysis (IPA) revealed an inflammatory gene signature in the ECs of A2V-treated tumors (Fig. 5, B and C), characterized by the activation of inflammatory signaling pathways involving tumor necrosis factor–α (TNF-α), type I and II IFN, interleukin-1 (IL-1) and IL-6, signal transducer and activator of transcription 1 (STAT1), and nuclear factor κB (NFκB). Consistent with inflammatory EC activation, the IFNγ/STAT1-inducible genes Pdcg (PD-L1), Cxcl10, and Vcam1 were up-regulated by A2V (Fig. 5D). These results indicate that the vascular ECs of A2V-treated mammary carcinomas acquire a distinctly activated, IFNγ-regulated inflammatory phenotype, which may influence immune cell trafficking and function in the tumors.

Perivascular CD8+ T cells elicited by antiangiogenic therapy up-regulate endothelial PD-L1 through IFNγ

The finding that A2V had up-regulated Pdcg/PD-L1 expression in tumor ECs prompted further analyses. A2V increased endothelial PD-L1 expression also at the protein level, as shown by flow cytometry (Fig. 5E and fig. S4B) and immunofluorescence staining (Fig. 5F) of orthotopic MMTV-PyMT tumors. We observed a similar response in the ECs of transgenic MMTV-PyMT tumors (Fig. 5G) and subcutaneous B16-OVA melanomas (Fig. 5H and fig. S4C).

IFNγ is a key inducer of PD-L1 (34). Accordingly, recombinant IFNγ, but not hypoxia, increased the amount of PD-L1 protein in cultured mouse ECs (fig. S4D). We then hypothesized that A2V had up-regulated PD-L1 expression in tumor ECs in response to increased IFNγ secreted by the perivascular CTLs. Tumor imaging revealed a direct correlation between the number of perivascular CD8+ T cells and the expression of PD-L1 in the blood vessels of transplanted MMTV-PyMT tumors (Fig. 5, I and J). Furthermore, systemic IFNγ neutralization by anti-IFNγ mAbs (28) blunted the A2V-induced up-regulation of Pdcg/PD-L1, Stat1, and Cxcl10 transcripts in transplanted MMTV-PyMT tumors (Fig. 5K).

Notably, monotherapy with B20 or LC06 enhanced PD-L1 expression in the ECs of both mammary tumors and melanomas (fig. S4, B and C), suggesting that VEGFA and ANGPT2 inhibition exerts complementary immunomodulatory actions that converge on stimulating PD-L1 gene expression or stability. Moreover, angiogenesis inhibition increased the expression of PD-L1 also in EpCAM+ (epithelial cell adhesion molecule–positive) cancer cells and CD45+ leukocytes of orthotopic MMTV-PyMT tumors (fig. S4, E and F). Together, these data illustrate a potentially immunosuppressive mechanism involving both vascular and nonvascular induction of PD-L1 consequent to antiangiogenesis-mediated immune cell activation.

PD-1 blockade improves the antitumoral activity of combined ANGPT2 and VEGFA blockade

We then reasoned that blocking PD-L1/PD-1 signaling could help prevent the early exhaustion of CD8+ T cells promoted by A2V through a counter-regulatory mechanism involving enhanced IFNγ production in the tumor microenvironment. Although the addition of an anti–PD-1 mAb provided only minimal benefits on top of the substantial antitumoral activity of A2V in the B16-OVA melanoma model (Fig. 6A), the combination of the two improved tumor growth control in the MC38-OVA model (Fig. 6B). These results show that the magnitude of the therapeutic benefit achieved by combining PD-1 blockade with A2V may vary with the tumor model.

Fig. 6. PD-1 blockade enhances the antitumoral activity of A2V.

(A) Left: Volume (mean values) of subcutaneous B16-OVA tumors treated as indicated. Arrow indicates start of treatment. Right: Tumor weight at necropsy (fold change over IgG). Each dot represents one tumor. IgG, n = 11; A2V, n = 14; α–PD-1, n = 12; and A2V plus α–PD-1, n = 13. Statistical analysis by two-way ANOVA with Tukey’s correction (left) or one-way ANOVA with Tukey’s correction for multiple comparisons (right). Red asterisks indicate analysis by two-way ANOVA with Sidak correction (left) or unpaired Student’s t test (right) applied to the two indicated groups. (B) Top: Volume (mean values) of subcutaneous MC38-OVA tumors treated as indicated. Each dot represents one tumor. Bottom: Growth of individual tumors. Arrows indicate start of treatment. IgG, n = 8; A2V, n = 10; α–PD-1, n = 9; and A2V plus α–PD-1, n = 9. Statistical analysis by one-way ANOVA with Tukey’s correction. Red asterisks indicate analysis by unpaired Student’s t test applied to the two indicated groups. (C) Flow cytometry analysis of the indicated cell types in subcutaneous B16-OVA tumors treated as indicated. For each cell type, the right panel shows the proportion of cells positive for intracellular IFNγ. Each dot represents one tumor; data are combined from two independent experiments. IgG, n = 14 to 16; A2V, n = 15 to 18; α–PD-1, n = 14 to 15; and A2V plus α–PD-1, n = 16 to 20. Statistical analysis by one-way ANOVA with Tukey’s correction for multiple comparisons. Red asterisks indicate analysis by unpaired Student’s t test applied to the two indicated groups. (D) Flow cytometry analysis of dextramer+, OVA-specific CD8+ T cells in subcutaneous B16-OVA tumors treated as indicated. Each dot in the left panel represents one tumor; data are combined from two independent experiments. Right panels show representative flow cytometry contour plots; FMO, fluorescence minus one. IgG, n = 11; A2V, n = 10; α–PD-1, n = 11; and A2V plus α–PD-1, n = 10. Statistical analysis by Kruskal-Wallis test with Dunn’s correction for multiple comparisons. Red asterisk indicates analysis by unpaired Student’s t test applied to the two indicated groups. (E) Kaplan-Meier survival curves of MMTV-PyMT transgenic mice treated as indicated. Median (50%) survival: IgG, 48 days; A2V, 75 days; α–PD-1, 45 days; A2V plus α–PD-1, 77 days; B20 plus α–PD-1, 67 days; and LC06 plus α–PD-1, 54 days. Number of experimental mice: IgG, n = 11; A2V, n = 14; α–PD-1, n = 8; A2V plus α–PD-1, n = 17; B20 plus α–PD-1, n = 8; and LC06 plus α–PD-1, n = 8. Statistical analysis by log-rank test. Note that some of the experimental groups (IgG and A2V) are also shown in Fig. 1A.

We then studied immune cell infiltrates in B16-OVA tumors, which allowed the analysis of tumors of a similar size in the three treatment cohorts (A2V, anti–PD-1, and A2V plus anti–PD-1). Although tumor infiltration by CD8+ T cells, CD4+ T cells, and NK1.1+ natural killer (NK) cells was similar across the treatment groups, we observed increased proportions of IFNγ+ T and NK cells in the A2V plus anti–PD-1–treated tumors, compared to tumors treated with A2V alone (Fig. 6C). Furthermore, A2V plus anti–PD-1 increased, as compared to A2V, the proportion of OVA-specific CTLs in some of the tumors (Fig. 6D). Thus, the combination of A2V and PD-1 blockade promotes broader immune cell activation than either alone.

We also studied tumor responses in transgenic MMTV-PyMT mice. Whereas most of the mice did not experience survival benefits, about 30% of the mice in the A2V plus anti–PD-1 treatment group survived longer than those treated with A2V (Fig. 6E). It is noteworthy that anti–PD-1 monotherapy did not provide measurable survival benefits in this transgenic tumor model, nor did its combination with B20 or LC06. Together, our findings in both melanoma and mammary tumor models illustrate the potential benefits of combining PD-1 blockade with dual ANGPT2 and VEGFA neutralization.

The antitumoral activity of combined ANGPT2 and VEGFA blockade is CTL-dependent in an immunogenic tumor model

Collectively, our data implicate adaptive immune cells, particularly CD8+ CTLs, in the antitumoral activity of optimized antiangiogenic drugs. To formally establish their involvement, we depleted CTLs from MC38-OVA tumor-bearing mice using a rat anti-CD8 mAb, as described previously (28). Mice carrying early established tumors were treated with anti-CD8 or control IgGs starting 3 days before the administration of either A2V or A2V plus anti–PD-1. Anti-CD8 treatment effectively depleted the CD8+ T cells in the peripheral blood of tumor-bearing mice (Fig. 7A) and negated the antitumoral benefits of both A2V and its combination with PD-1 blockade (Fig. 7B). Although these results are limited to one immunogenic cancer model—and we could not extend the observation window beyond 2.5 weeks from the treatment start because of the potential neutralization of rat IgGs—they suggest that CTLs may play important roles in the antitumoral responses stimulated by angiogenesis inhibitors.

Fig. 7. The antitumoral activity of A2V is CTL-dependent in an immunogenic cancer model.

(A) Flow cytometry analysis of CD8+ T cells in the blood of mice carrying subcutaneous MC38-OVA tumors treated as indicated. Each dot represents one mouse. Rat IgG refers to the isotype control for anti-CD8. Rat IgG plus A2V, n = 5; rat IgG plus A2V plus α–PD-1, n = 5; α-CD8 plus A2V, n = 6; and α-CD8 plus A2V plus α–PD-1, n = 6. Statistical analysis by one-way ANOVA with Tukey’s correction for multiple comparisons. (B) Volume (mean values ± SEM) of subcutaneous MC38-OVA tumors treated as indicated. Gray and red arrows indicate start of treatment with α-CD8/control rat IgG or therapeutic antibodies, respectively. IgG, n = 7; rat IgG plus A2V, n = 9; rat IgG plus A2V plus α–PD-1, n = 7; α-CD8 plus A2V, n = 10; and α-CD8 plus A2V plus α–PD-1, n = 10. Statistical analysis by two-way ANOVA with Tukey’s correction (only the relevant comparisons are shown).

DISCUSSION

We report that the concurrent neutralization of VEGFA and ANGPT2 by A2V promotes the development and deployment of antitumor immunity in mouse models of cancer. A2V increases the intratumoral representation of phagocytes that cross-present tumor antigens and activates CTLs and facilitates their extravasation from the tumor blood vessels that persist after antiangiogenic therapy. These immunostimulatory properties were associated with survival benefits in several mouse cancer models. Our findings, along with those reported in the article by Allen et al. (35), underscore the ability of optimized antiangiogenic regimens to stimulate antitumor immune responses but also highlight the potential occurrence of counter-regulatory mechanisms, such as the induction of PD-L1 in both vascular and nonvascular cells in response to A2V-elicited CTLs and IFNγ.

Tumors can rapidly adapt to the neutralization of individual proangiogenic factors, including VEGFA, through routes that involve metabolic reprogramming, the enforcement of compensatory proangiogenic signals, or the acquisition of angiogenesis-independent modes of invasive growth (4, 5, 3638). An important result of our study was that A2V markedly extended the survival of MMTV-PyMT transgenic mice, which develop multiple mammary tumors that progress to end-stage by 14 to 16 weeks of age and metastasize to the lung (26). LC06 decreased angiogenesis and delayed tumor growth in a transplanted MMTV-PyMT tumor model, consistent with data obtained with a different anti-ANGPT2 antibody (6, 7). However, neither LC06 nor B20 provided survival benefits in transgenic MMTV-PyMT mice. The improved antitumoral activity of dual ANGPT2 and VEGFA blockade in MMTV-PyMT breast cancer models thus suggests that VEGFA and ANGPT2 function nonredundantly and control processes that likely extend beyond tumor angiogenesis.

Besides rebound sprouting angiogenesis mediated by compensatory proangiogenic pathways, alternative modes of tumor vascularization, such as vascular co-option or mimicry, may play roles in resistance to anti-VEGFA therapy. For example, the occurrence of vessel co-option in colorectal cancer liver metastases is associated with poor response to bevacizumab (38). ANGPT2 neutralization might exacerbate vascular co-option by cancer cells (39), and there is also evidence that genetic Angpt2 deficiency may aggravate the progression of colorectal liver metastasis in mice (40). Therefore, it is unlikely that A2V would provide therapeutic benefits in metastatic colorectal cancer and, possibly, other metastatic tumors that exploit vessel co-option to enforce their blood supply. On the other hand, our data support the rationale for applying A2V in the context of unresectable primary cancers that have not yet metastasized, which may best benefit from both antiangiogenic and antimetastatic actions of ANGPT2 blockade. ANGPT2 neutralization abated the incidence and growth of spontaneous pulmonary metastasis from primary MMTV-PyMT tumors, in agreement with previous findings (6). A2V further improved metastasis control, suggesting that cotargeting VEGFA may enhance the antimetastatic potential of ANGPT2 blockade, at least in breast cancer.

The complementary actions of VEGFA and ANGPT2 on the tumor microenvironment may contribute to suppressing antitumor immunity, in part through their effects on tumor blood vessels. Under the influence of VEGFA, the tumor blood vessels acquire structural and molecular features that limit the trafficking of T cells (20, 21). T cells need to cross the EC layer of tumor blood vessels through a multistep process that includes binding to adhesion molecules [such as intercellular adhesion molecule–1 (ICAM1) and vascular cell adhesion molecule–1 (VCAM1)] expressed on ECs. VEGFA causes the ECs of tumor blood vessels to down-regulate the expression and abrogate the clustering of ICAM1 and VCAM1, hence limiting T cell adhesion and extravasation (41). Furthermore, angiogenic tumor ECs express surface receptors or soluble mediators, such as endothelin B receptor, which can blunt the effector functions of T cells newly recruited to the tumor (42, 43). There is also experimental evidence for VEGFA suppressing antitumor immunity through direct effects on innate immune cells. For example, VEGFA can directly impair DC maturation, and neutralization of VEGFA enhances the antigen-presenting capacity of DCs in mouse tumor models (44, 45).

Initial clinical data support the notion that inhibiting VEGFA signaling may improve T cell–mediated antitumor immunity, particularly in association with immune checkpoint blockade (22, 46, 47). The administration of bevacizumab to cancer patients increased the allostimulatory capacity and proliferation of DCs and T cells, respectively, in response to recall antigens ex vivo (48). Furthermore, the analysis of melanoma samples showed increased trafficking of CD8+ T cells across the vasculature of tumors from patients treated with bevacizumab in combination with CTLA-4 blockade (ipilimumab), as compared with ipilimumab alone (47). Bevacizumab plus ipilimumab also increased the proportion of circulating T cells expressing a memory phenotype, as well as the titers of serum antibodies against galectin family members (47).

The effects of ANGPT2 on the immune system are less well understood. ANGPT2 promotes vascular leakiness and extravasation of myeloid cells to tumors, which can have direct suppressive effects on T cells (6, 4951). Blocking ANGPT2 prunes immature tumor blood vessels while promoting the maturation and stabilization of the remaining vessels (68, 14, 29); these vascular-normalizing effects may facilitate T cell trafficking in tumors (19, 20). Also, blocking ANGPT2 impedes the association of TIE2-expressing macrophages (TEMs) with nascent tumor blood vessels (6). TEMs were shown recently to promote vascular permeability in tumors through their secretion of VEGFA (52), so blocking ANGPT2 may normalize the tumor-associated vasculature also through effects on these perivascular TAMs (53).

The aforementioned results and considerations support the notion that cotargeting VEGFA and ANGPT2 signaling may not only render the tumor blood vessels more permissive to T cell extravasation but also relieve suppressive cues on, or emanating from, innate immune cells to facilitate the development and deployment of antitumor immunity. To analyze antigen-specific immune responses after A2V, B20, or LC06, we used a B16-OVA melanoma model. Several observations indicate that cotargeting VEGFA and ANGPT2 may provide a better proimmune conditioning strategy than targeting either factor alone. First, A2V increased the abundance and the phagocytic and cross-presentation capacities of intratumoral APCs to a greater extent than B20 or LC06. In line with these findings, A2V was shown to skew the polarization of TAMs toward an M1-like (potentially immunostimulatory) phenotype in transplant glioma models (23). Second, A2V was more effective than B20 or LC06 in increasing the proportion of activated (IFNγ+ or CD69+) CTLs and OVA-specific CTLs in B16-OVA or mammary tumor models. Mechanistically, extensive tumor necrosis induced by A2V may increase the availability of tumor material for engulfment by intratumoral phagocytes/APCs. This, combined with increased APC maturation in VEGFA-depleted microenvironments (44, 45), may enhance the uptake, processing, and presentation of tumor-associated antigens to both CD4+ and CD8+ T cells, resulting in their enhanced activation and expansion. Third, A2V inhibited tumor angiogenesis but also normalized the residual blood vessels, which was associated with increased T cell extravasation into the tumor microenvironment. Collectively, these findings suggest that dual VEGFA and ANGPT2 inhibition promotes antitumor immunity through multidimensional effects on both innate and adaptive immune cells and complementary actions on the tumor blood vessels.

We found that CTL depletion abated the antitumoral efficacy of A2V in an immunogenic transplant tumor model. This observation suggests that CTLs may mediate key antitumoral responses in the context of antiangiogenic therapy. Such responses may involve both conventional cytotoxicity toward cancer cells and, provocatively, direct antiangiogenic effects. Previous studies have shown that T cell–derived IFNγ can directly inhibit tumor angiogenesis by inducing EC apoptosis independent of direct contacts between T cells and ECs (54). Also, angiogenic ECs may express TRAIL (TNF-related apoptosis-inducing ligand) and FAS receptors and become targets of CTL-mediated killing (55). Thus, it is tempting to speculate that, in tumors, dual ANGPT2 and VEGFA blockade may evoke canonical antiangiogenic mechanisms that are reinforced through the vascular-targeted actions of the CTLs. From a complementary viewpoint, certain immunotherapy regimens may trigger antiangiogenic responses in tumors. For example, CTLA-4 blockade in combination with a tumor vaccine triggered a form of ischemic tumor vasculopathy associated with enhanced T cell infiltration and humoral responses against ANGPT2 and VEGFA in some long-term responding patients (56). Furthermore, tumor responses after PD-1 or CTLA-4 blockade are associated with increased titers of autoantibodies against ANGPT2, whereas higher pre- or on-treatment ANGPT2 serum levels are associated with tumor refractoriness or resistance to immune checkpoint blockade (57). Together, these findings highlight the intertwining roles played by antiangiogenic and immune-targeted therapies in the regulation of angiogenic blood vessels.

There is increasing clinical evidence for antiangiogenic drugs profoundly modulating immune cell composition in human cancers, with potential consequences for antitumor immunity (1921, 58, 59). For example, both bevacizumab and sunitinib—a broad-spectrum tyrosine kinase inhibitor that potently blocks tumor angiogenesis—increased infiltration by CD4+ and CD8+ T cells and up-regulated PD-L1 expression in primary renal cell carcinomas (60). Although T cell numbers and PD-L1 expression unexpectedly correlated inversely with progression-free and overall patient survival, the findings of Liu et al. (60) are consistent with our observation that effective antiangiogenic therapy increases activated CTLs but also induces PD-L1 expression in the tumors, likely through CTL-derived IFNγ. While A2V elicited CTL activation, it also triggered a potentially immunosuppressive counter-regulatory response in the tumor endothelium. RNA-seq analysis of MMTV-PyMT tumor–derived ECs revealed the up-regulation of PD-L1 in response to A2V, a finding that we confirmed in B16-OVA melanomas. A2V induced a proinflammatory signature in the tumor ECs that involved, according to IPA and qPCR validation data, IFNγ and TNF-α signaling and downstream STAT1 activation. Systemic IFNγ blockade negated the up-regulation of PD-L1 after A2V treatment, suggesting that IFNγ signaling indeed controls PD-L1 expression in the vascular ECs. Our finding that perivascular CD8+ T cell numbers correlated with the expression of PD-L1 in the corresponding vascular segments supports the notion that gradients of IFNγ produced by the perivascular CTLs directly regulate endothelial PD-L1 expression. Hence, concomitant inhibition of VEGFA and ANGPT2 proved to be immunostimulatory for innate and adaptive immune cells and partly immunosuppressive via a CTL-EC cross-talk. Because the antitumoral potential of antiangiogenic therapy may be both enhanced and limited by the ensuing host immune response against the tumor, blocking rebound immunosuppression induced by antiangiogenesis may tilt the balance toward enhanced antitumor immunity, greater tumor inhibition, and improved clinical benefit. In this regard, it was shown that blocking antiangiogenesis–induced immunosuppression with a PI3KG (phosphatidylinositol 3-kinase γ) inhibitor translated into a better tumor control in mouse models of cancer (61).

In the transgenic MMTV-PyMT model, the addition of anti–PD-1 mAb to A2V extended the survival of a fraction of the mice. This result may be reminiscent of findings in cancer patients receiving immune checkpoint blockade: Although the antitumoral responses can be marked and long-lasting in some cases, they typically involve only a minority of the patients (17). In the B16-OVA model, A2V plus anti–PD-1 increased both activated (IFNγ+) T and NK cells and the proportion of OVA-specific CTLs in some of the tumors, as compared to A2V-treated mice. Although these observations suggest that PD-1 blockade may reinforce antigen-specific immune responses initiated by A2V, our results also highlight the limitations of conventional mouse cancer models for interrogating tumor immunity and response to immunotherapy. Anti–PD-1 monotherapy provided no survival benefits in transgenic MMTV-PyMT mice and only had modest inhibitory effects on B16-OVA and MC38-OVA tumor growth. This is consistent with previous studies showing that experimental mouse tumors, including mammary tumors (62) and B16 melanomas (63), are essentially refractory to PD-1 or PD-L1 blockade. The low mutational burden of genetically engineered mouse models of cancer and some of the commonly used mouse cancer cell lines may underlie their poor immunogenicity (62, 64). Also, whereas PD-1 blockade improved tumor inhibition by A2V in each tumor model tested, the combined regimen failed to regress established tumors. It remains to be seen whether the lack of tumor regressions is indicative of either poor tumor immunogenicity or the rapid onset of adaptation and/or resistance to the treatment. Ongoing clinical trials of A2V in combination with the anti–PD-L1 mAb atezolizumab (NCT01688206; www.clinicaltrials.gov) will provide important information on the benefits of such combination. Finally, our study focused on primary tumors, so the results may not be generalized to metastatic disease, which is the most relevant condition in clinical oncology.

Unselected human breast cancers are poorly responsive to both anti-VEGFA therapy and immune checkpoint blockade (65). Our preclinical findings in a mouse mammary tumor model somewhat recapitulate the clinical findings but also highlight the potential therapeutic benefits of combining improved antiangiogenic drugs with checkpoint inhibitors in breast cancer. BRCA-mutant breast cancer may provide a suitable subtype for testing such drug combinations because higher CTL numbers are associated with complete pathological responses after platinum-based chemotherapy in this breast cancer subtype (66). Although we have not examined the nature of potential immunogenic antigens in MMTV-PyMT mice, it is tempting to speculate that, in the long-term responder MMTV-PyMT mice, A2V promoted the development of tumor antigen–specific T cell clones, whose antitumoral activity could be improved by anti–PD-1 therapy. These considerations encourage further preclinical testing of combined antiangiogenics and checkpoint inhibitors in breast cancer and other cancer types.

MATERIALS AND METHODS

Study design

This study was designed to identify potential immunological mechanisms involved in the antitumoral actions of angiogenesis inhibitors targeting the VEGFA and ANGPT2 pathways. To this end, we used immunocompetent mouse cancer models, including genetically engineered (transgenic) mice. In the first part of the study (Figs. 1 and 2), we analyzed the effects of angiogenesis inhibitors on tumor angiogenesis and progression. In the second part of the study (Figs. 3 to 5), we focused on the immunomodulatory actions of the angiogenesis inhibitors. In the third part of the study (Figs. 6 and 7), we tested the hypothesis that angiogenesis inhibitors function, at least in part, by eliciting antitumoral immune responses that can be enhanced through immune checkpoint blockade. All procedures were performed according to the protocols approved by the Veterinary Authorities of the Canton Vaud according to Swiss law (protocols 2574, 2574.a, 2577, 2577.a, and 3049).

The design of the experimental treatment trials (drug regimens, age of mice or tumor size at the time of enrollment, etc.) is presented in the Supplementary Materials and Methods. Detailed information on the sample size and statistical methods is presented in the figures and associated legends. Table S1 provides a checklist of the experiments and associated analyses shown in the study. Briefly, survival studies involving large cohorts of mice (transgenic MMTV-PyMT and RIP1-Tag2 mice) were performed only once by preplanned enrollment of mice when they reached a defined age (65 ± 3 and 78 ± 3 days, respectively). The transgenic mice were randomized to the various experimental cohorts by excluding obvious outliers (such as unusually large mammary tumors). Studies involving transplant tumor models were typically performed several times. Treatment trials were initiated when the transplanted tumors reached a mean volume of 20 to 200 mm3, depending on the tumor model. Analytical studies (characterization and quantification of immune cell infiltrates, vascular parameters, qPCR, etc.) were typically performed several times in independent experiments, implementing fixed time points of analysis for all experimental groups, unless indicated otherwise (for example, survival studies). Fixed time points are shown in the figures and indicate the time elapsed from the treatment start. The investigators were not blinded when assessing the results or analyzing the data. On rare occasions, outliers at the end point were excluded by using the ROUT method (provided in GraphPad Prism) to identify outliers. In some cases, selected samples were excluded from specific analyses because of technical flaws during sample processing or data acquisition.

Statistical analysis

Error bars indicate the standard error of the mean (SEM). The number of biological (nontechnical) replicates for each experiment is indicated in the figure legends. Independent experiments are presented individually or combined, as explained in the figure legends and table S1. Analysis of experiments with more than two groups was performed using one-way ANOVA with Tukey’s correction for multiple comparisons, unless indicated otherwise. In some experiments involving multiple groups, in addition to multiple comparisons, we directly compared two experimental groups of interest and applied Student’s t test with 95% confidence interval; in these cases, the statistical significance is indicated for the overall experiment (by one-way ANOVA with correction for multiple comparisons) as well as the two groups of interest (by Student’s t test or Mann-Whitney test), using a color code (black versus red symbols) explained in the figure legends. In experiments in which the determinations were considered interdependent (such as tumor growth over time), we applied two-way ANOVA with Sidak or Tukey’s correction, as indicated in the figure legends. For experiments with two groups, statistical analysis was performed using Student’s t test with 95% confidence interval. Statistical analyses in the survival experiments were performed by log-rank (Mantel-Cox) test. Statistical significance is indicated in the figures as follows: *0.01 ≤ P < 0.05; **0.001 ≤ P < 0.01; ***P < 0.001. Exact P values for all comparisons are shown in table S2; P values >0.05 are not indicated.

SUPPLEMENTARY MATERIALS

www.sciencetranslationalmedicine.org/cgi/content/full/9/385/eaak9670/DC1

Materials and Methods

Fig. S1. Combined ANGPT2 and VEGFA blockade delays tumor progression.

Fig. S2. Combined ANGPT2 and VEGFA blockade impairs tumor angiogenesis and normalizes the remaining tumor blood vessels.

Fig. S3. Combined ANGPT2 and VEGFA blockade promotes the perivascular accumulation of T cells.

Fig. S4. IFNγ, but not hypoxia, induces PD-L1 expression in ECs.

Table S1. Checklist of the experiments performed and associated analyses (provided as an Excel file).

Table S2. Exact P values associated with the main and supplementary figures (provided as an Excel file).

REFERENCES AND NOTES

  1. Acknowledgments: We thank C. H. Ries and S. Hoves (Roche, Penzberg) for fruitful discussions and advice on drug regimens and experimental design; D. Pais-Ferreira, J.-B. Pignier, and A. Bellotti for help with some experiments; A. Palini (Nestlé, Lausanne) and the Flow Cytometry Core Facility at École Polytechnique Fédérale de Lausanne (EPFL) for cell sorting; the BioImaging and Optics Platform at EPFL for advice on confocal microscopy; and K. Breckpot (Vrije Universiteit Brussel) and N. Haynes (University of Melbourne) for providing B16-OVA and MC38-OVA cells, respectively. Funding: D.L. was supported by an EMBO Short-Term Fellowships (ASTF no. 289-2015). This study was supported by grants from the Leenaards Foundation, the Swiss Cancer League (grants KFS-3007-08-2012 and KFS-3759-08-2015), the San Salvatore Foundation, and Roche (to M.D.P.). Author contributions: M.S. and N.R. designed and performed the experiments, collected and analyzed the data, interpreted the results, and drafted the paper. E.K., A.C., and D.L. designed and performed the experiments and analyzed the data. C.W.R. performed the treatment trials in MMTV-PyMT and RIP1-Tag2 mice and collected the data. A.K. performed RNA-seq. Y.K. and H.-J.M. provided the therapeutic mAbs and intellectual input. C.-H.O. analyzed the RNA-seq data. M.D.P. designed the experiments and interpreted the results, supervised and coordinated the study, and wrote the paper. Competing interests: A.K., Y.K., H.-J.M., and C.-H.O. are Roche employees. The other authors declare that they have no competing interests. Data and materials availability: Primary RNA-seq data sets have been deposited in the Gene Expression Omnibus under accession code GSE94920. The mAbs A2V, LC06, and B20 and isotype-matched control IgGs were provided by Roche (Penzberg, Germany) under conditions expressed in a sponsored research agreement between Roche and the corresponding author’s institution (EPFL). Interested researchers may contact Roche for requests concerning the aforementioned materials.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article