Active Immunization Against the Vascular Endothelial Growth Factor Receptor flk1 Inhibits Tumor Angiogenesis and Metastasis

Tumor metastasis is the main cause for failure of conventional cancer therapy, necessitating the reevaluation of current cancer therapy strategies. Conventional approaches for cancer immunotherapy usually target antigens expressed by tumor cells and are aimed at eradicating tumor cells by direct or indirect immunological attack (1–3). Despite the identification of tumor-associated antigens and the demonstrated effectiveness of immunotherapy in many experimental animal tumor models, immunotherapy currently has limited clinical utility for human cancers. The limitations of cancer immunotherapy are thought to be due in part to poor immunogenicity, immune tolerance, and escape of tumor cells from immune surveillance through antigen modulation, decreased MHC expression by tumor cells, lack of costimulatory molecules, or secretion of immunosuppressive molecules by tumor cells (4, 5). One possible alternative to overcome these obstacles of conventional immunotherapy is to target the blood vessels nourishing the growing tumor cells rather than the tumor cells themselves.

Angiogenesis, the growth of new blood vessels from preexisting vasculature, is a tightly regulated process important in fetal development and wound healing and in pathological conditions such as tumor growth and metastasis (6, 7). Vascular endothelial growth factor (VEGF)^* and its receptors fetal liver kinase 1 (flk1; murine) and KDR (human) play a critical role in regulating the process of normal and pathological angiogenesis. Targeted inactivation of the gene for VEGF or flk1 in mice results in embryonic lethality at day 7.5 due to severely impaired vascular development (8, 9). The importance of VEGF and flk1/KDR in tumor angiogenesis is exemplified in studies using a dominant-negative flk1 receptor (10), neutralization of VEGF by monoclonal antibodies (11), neutralizing flk1/KDR mAb (12, 13) or flk1/KDR kinase inhibitors (14), all of which were shown to inhibit angiogenesis and tumor growth. Moreover, overexpression of VEGF and KDR is strongly associated with invasion and metastasis in human malignant disease (15, 16). Therefore, overexpressed flk1/KDR could be a potential target for tumor immunotherapy.

We have hypothesized that a therapeutically effective immune response can be elicited that targets flk1 expressed on tumor blood vessels and thus, inhibit tumor angiogenesis and growth. To test this hypothesis, we used a dendritic cell (DC) immunization strategy, which has been successfully used in mice to overcome immune tolerance to other antigens (17). In this report, we demonstrate that immunization with DCs pulsed with soluble flk1 induces neutralizing antibody and CD8⁺ cytotoxic T cell responses, suppresses tumor angiogenesis, and strongly inhibits the development of metastasis in two mouse models.

Lewis lung carcinoma cell line D122–96 (H-2^b) was provided by Dr. L. Eisenbach (Weizman Institute of Science, Rehovot, Israel). H5V endothelial cell line (H-2^b), which expresses both flk1 and MHC class I, was provided by Dr. A. Vecchi (Istituto Mario Negri, Milan, Italy; reference 18). Mouse endothelial cell line bEND.3 (H-2^d), which expresses flk1 (19), was obtained from Dr. T. Sato (University of Texas Southwestern Medical Center, Dallas, TX). Melanoma cell line B16 (H-2^b), EL4 lymphoma cell line (H-2^b), and the NK-sensitive YAC-1 cell line were all purchased from American Type Culture Collection. The cell lines were maintained in DMEM media (Invitrogen) containing 10% FCS (HyClone Laboratories).

Female C57BL/6 mice, 6–8 wk of age, were purchased from Harlan Sprague Dawley, Inc. and housed under pathogen-free conditions.

Construction of the expression plasmid vector Aptag-flk1, which contains the cDNA corresponding to the extracellular domain of flk1 fused to alkaline phosphatase (AP), was described previously (20). Soluble flk1-AP protein was purified using an anti-AP affinity chromatography, followed by a size exclusion chromatography. It was then subjected to SDS-PAGE, and shown as a single band with molecular weight of ∼180 KD (>95% purity). The purified protein could be recognized by a flk1-specific monoclonal antibody, DC101 in Western blot. The binding of flk1-AP to VEGF was examined by an ELISA, and the protein was found to be active. The purified protein was tested for endotoxin using the Pyrogent^® plus Limulus Amebocyte Lysate assay kit (BioWhittaker). All protein preparations used in animal studies contained <1.25 EU/ml of endotoxin.

The flk1 insert from the Aptag-flk1 vector was subcloned into the plasmid pET28a vector with a polyhistidine-encoding sequence at the amino-terminal (BD Biosciences/CLONTECH Laboratories, Inc.). The construct was verified by automatic sequence analysis. The recombinant flk1-His protein was expressed in Escherichia coli, purified from inclusion bodies by preparative electrophoresis, and assessed for purity by SDS-PAGE and binding to VEGF by ELISA.

DCs were generated from the bone marrow as described with modifications (21). Briefly, C57BL/6 mice were killed and bone marrow harvested from tibia and femurs. Bone marrow cells were depleted of existing T cells, B cells, macrophages, and granulocytes by incubation with a cocktail of antibodies including anti-CD4 (GK1.5), anti-CD8 (2.43), anti-Ia (B21–2), anti-B220 (RA3–3A1/6.1), and anti-Gr-1 (RB6–8C5/1; all from BD PharMingen), for 30 min at 4°C and then with rabbit complement (Accurate Chemical) for additional 30 min at 37°C. The remaining cells were cultured in 10% FCS supplemented RPMI 1640 in the presence of GM-CSF (20 ng/ml) and IL-4 (50 ng/ml; PeproTech) at 37°C, 5% CO₂, for 9 d. Nonadherent cells were then harvested and confirmed to be mature DCs by their morphology and phenotypic profile (CD40⁺, CD81⁺, CD86⁺, Ia⁺, and CD14⁻) on flow cytometric analysis.

DCs were pulsed with antigen, as described previously (22). Briefly, DCs were washed twice in the serum-free medium AIM V (Invitrogen) and incubated with soluble flk1-AP protein or human AP (50 μg/ml) in AIM V for 16 h. The cells were then washed twice in AIM V before use for vaccination. For immunization, mice were injected intravenously with 5 × 10⁴ flk1-AP-pulsed DCs, AP-pulsed DCs, or PBS (200 μl) per mouse at 8–10 d intervals.

Mice were immunized three times with DCs pulsed with flk1-AP (DC-flk1), DCs pulsed with AP (DC-AP), or PBS as described above. CTL response was assessed as described previously (23). Briefly, spleen cells were prepared from immunized mice (two mice per group) and restimulated with DCs pulsed with flk1-AP (at 100:1 effector/stimulator ratio) in a 24-well plate for 5 d in RPMI 1640 (Invitrogen) with 10% FCS. The CTL activity was tested in a 4-h ⁵¹Cr release assay against a panel of target cells including flk1⁺ H5V endothelial cells, flk1⁻ D122–96 tumor cells, flk1-AP-pulsed DCs, AP-pulsed DCs, and NK-sensitive YAC-1 cells. The percentage cytotoxicity was calculated using the formula: (experimental release − spontaneous release)/(maximum release − spontaneous release) × 100.

Mice were immunized three times with flk1-AP-pulsed DCs, AP-pulsed DCs, or PBS as described above. Blood samples were collected from mice before and 7 d after vaccinations. Anti-flk1 antibody in the sera was detected by ELISA. Briefly, a 96-well plate was incubated with 200 ng/well of flk1-His protein overnight at 4°C. After three washes with 0.1% Tween in PBS, 2% BSA was added to the plate and incubated at room temperature for 1 h. Diluted sera were then added to wells and incubated for 1 h. Wells were washed three times and then incubated with 100 μl goat anti–mouse peroxidase for 1 h. Wells were washed three times and then incubated with 50 μl of 3,3′, 5,5′-tetra-methylbenzidine (TMB) substrate (Kirkegaard and Perry Lab, Inc.) for 15 min. The reaction was stopped by adding 50 μl of 1 M phosphoric acid and wells read at 450 nm on a microtiter plate reader. For flk1-VEGF blocking assays, wells were coated with 100 ng of recombinant human VEGF₁₆₅ (provided by Dr. P. Kussie, ImClone) overnight at 4°C. Wells are blocked as described above and then incubated for 1 h at room temperature with 100 ng of flk-AP that had been preincubated for 2 h with various concentrations of sera from immunized mice. Wells were washed and incubated with p-nitrophenyl phosphate (PNPP; Sigma-Aldrich). Color was developed for 30 min at room temperature and was then read at 405 nm on a microtiter plate reader.

Cell-based ligand-binding assays were performed as described with modifications (24). Flk1-expressing bEND.3 cells (10⁵ per well) were grown in 24-well plates in DMEM 10% FCS for 48 h. Cells were washed three times with binding buffer (DMEM with 0.1% BSA). Pooled immune sera from DC-flk1–immunized mice or from DC-AP–immunized mice and unlabeled recombinant human VEGF₁₆₅ were serially diluted in binding buffer as indicated. In a total volume of 400 μl/well, cells were incubated with the samples for 4 h at 4°C. Cells were then washed with binding buffer and incubated with 10 nCi [¹²⁵I]VEGF (Amersham Pharmacia Biotech) in 400 μl binding buffer per well for 1 h. After incubation, cells were washed using cold PBS with 0.1% BSA. Cells were harvested by adding 200 μl of 0.5 M NaOH. The radioactivity bound to the cells was determined in a Wizard 1470 gamma counter (PerkinElmer).

An alginate bead assay was designed to measure in vivo angiogenesis induced by tumor cells (25). Lewis lung tumor cells were suspended in a 1.5% solution of sodium alginate and added drop by drop into a swirling 37°C solution of 250 mM calcium chloride. Alginate beads were formed containing ∼5 × 10⁴ tumor cells per bead. Mice were anesthetized and four beads implanted subcutaneously through an incision made on the dorsal side. Incisions were closed with surgical clips. After 12 d, mice were injected intravenously with 100 μl of FITC-dextran solution (20 mg/ml). Animals were killed after 20 min, beads removed, and incubated overnight at room temperature in 1 ml of buffer (1 mM Tris-HCl, pH 8). The beads were ground briefly with a hand-held mixer and an additional 1 ml of buffer was added. Samples were then vortexed and centrifuged at 1,500 rpm for 5 min. Fluorescence of the sample supernatants was quantitated against a standard curve of FITC-dextran.

Mice were immunized three times with flk1-AP–pulsed DCs, AP-pulsed DCs, or PBS as described above. 10 d after the last immunization, mice were injected intravenously with 10⁶ B16 cells. Mice were killed based on the metastatic death in the control groups. Tumor load was assessed by counting the tumor nodules on the lung surface.

Mice were immunized three times with flk1-AP–pulsed DCs, AP-pulsed DCs, or PBS as described above. 10 d after the last immunization, mice were challenged with an intrafootpad injection with 2 × 10⁵ of D122–96 tumor cells. When tumors reached ∼5 mm in diameter, the tumor-bearing leg was surgically removed. Mice were killed based on the metastatic death in the control groups. Tumor load was assessed by counting the tumor nodules on the lung surface. In separate experiments, mice were monitored daily for survival.

1 d before DC-flk1 immunization, mice received intraperitoneal injection of 0.5 mg of either anti-CD4 (GK1.5), or anti-CD8 (clone 116), or control rat IgG (Jackson ImmunoResearch Laboratories). To ensure complete depletion of respective T cell population, one mouse from each group was killed the next day and splenocytes were analyzed by FACS^® after staining with FITC-conjugated anti-CD4 (L3T4) and PE-conjugated anti-CD8 (Ly2; BD PharMingen). The same depletion procedure was repeated every 2 wk to prevent recovery of depleted T cell populations. The mice were immunized three times with DCs pulsed with flk1-AP and then challenged with Lewis lung tumor as described in the Lewis lung metastasis model. Tumor load in the lungs were compared among the groups.

To test whether flk1 immunization could affect mouse pregnancy by interfering with prenatal angiogenic process, female mice were immunized with DCs pulsed with flk1-AP or PBS as described above. In an additional control, another group of mice were treated with intraperitoneal injections of the anti-flk1 antibody DC101 (800 μg per injection, twice weekly for 30 d; reference 12). Mice were then mated with males and monitored daily for signs of pregnancy. Number of pups of each delivery was recorded. The pups were also carefully examined for signs of sickness and abnormality.

10 d after the last immunization with flk1-AP–pulsed DCs, a full thickness wound, including the panniculus carnous, was excised from the dorsum of each mouse. A 1.60-cm² circular defect was outlined 2.0 cm from the nape of the animal's neck using a fine-tipped marking pen. The defect was created by elevating the skin and panniculus carnosus in the center of the outlined defect using forceps, followed by excision of the outlined area using microdissecting scissors. Wound area was measured twice weekly. 15 d after this excision, mice were killed, and scar tissues were removed for histological examination.

Lung samples and scar tissues were fixed overnight in 10% zinc formalin at 4°C, embedded in paraffin, and sectioned at 5 μm onto saline-coated slides. Hematoxylin and eosin staining was performed. Scar tissues from the wound healing study were additionally stained with a Masson's trichrome kit (Richard Allen Scientific) according to manufacturer's instructions.

All data including tumor counts and vessel measurement were analyzed using Student's t test with SigmaStat v.2.03.

C57BL/6 mice were immunized three times at 8–10 d intervals with DCs pulsed with soluble flk1-AP protein (DC-flk1), DCs pulsed with a control protein alkaline phosphatase (DC-AP), or with vehicle alone (PBS).

Sera were collected from immunized mice and tested for anti-flk1 antibody by ELISA in plates coated by flk1-His protein. A strong antibody response was generated in DC-flk1 vaccinated mice (Fig. 1 A). In contrast, anti-flk1 antibody was detected in only 1 of 10 mice from DC-AP group and in none of the PBS group. Antibody titer was significantly higher in all mice vaccinated with DC-flk1 compared with prevaccination sera (data not shown). Notably, the immune sera from DC-flk1 group inhibited the binding of soluble flk1 receptor to VEGF by ELISA, whereas the immune sera from DC-AP or PBS groups did not show any signification inhibition (Fig. 1 B), demonstrating the presence of a neutralizing anti-flk1 antibody. The specificity and neutralizing activity of the anti-flk1 antibody was further evaluated in a [¹²⁵I]VEGF binding assay using flk1⁺ bEND.3 cells (Fig. 1 C). Immune sera from DC-flk1 group strongly inhibited the binding of [¹²⁵I]VEGF to the native flk1 receptor on bEND.3 cells, compared with control sera from DC-AP group.

To analyze the cellular immune response in mice immunized with DC-flk1, splenocytes from the mice were harvested 10 d after the second boost immunization and further stimulated in vitro by cocultivation with flk1-AP–pulsed DCs for 14 d, followed by testing in a 4-h ⁵¹Cr release assay for CTL activity against flk1-positive and flk1-negative targets. Splenocyte cultures from mice immunized with DC-flk1 showed a significantly higher level of CTL activity against the flk1-positive endothelial cell line H5V, compared with DC-AP or PBS groups (Fig. 2 A). To determine whether this CTL response was specific for flk1, T cells from the DC-flk1 group were also tested against flk1-AP-pulsed DCs, and AP-pulsed DCs as targets. Only the DC-flk1-AP but not the DC-AP target cells were lysed by these T cells, indicating the specificity for flk1 antigen (Fig. 2 B). Moreover, these T cells did not lyse Lewis lung carcinoma or EL-4 lymphoma tumor cells, or the NK-sensitive YAC-1 cell line. These results demonstrate that both flk1-specific humoral and cellular immune responses can be induced in mice by immunization with DC-flk1.

DC-flk1–immunized mice appeared generally healthy, gained normal weight, and no gross toxicity was observed. However, breeding experiments revealed a striking effect of DC-flk1 immunization on pregnant female mice. Female mice were immunized on days 0, 10, and 20 and mated with nonimmunized male mice 10 d after the final immunization. Only 20% of DC-flk1–immunized mice became pregnant, compared with a 70% pregnancy rate in the control group. Moreover, the mean litter size was significantly smaller in the DC-flk1 group (2 vs. 6) compared with the PBS control group (Table I). The small number of pups from the DC-flk1–immunized mice were either stillborn or died within 10 d after birth and were found to have massive hemorrhage in all organs. Similar results were observed in a group of female mice treated with the neutralizing anti-flk1 monoclonal antibody DC101 (12). These data indicate an inhibitory effect of DC-flk1 immunization on fetal development.

To determine whether the anti-flk1 immune responses could inhibit tumor-induced angiogenesis, we conducted an in vivo angiogenesis assay using alginate beads. Lewis lung tumor cells encapsulated alginate beads were implanted subcutaneously in DC-flk1–vaccinated mice. Growth factors produced by the encapsulated tumor cells induced vascularization of the beads, which was then measured by uptake of FITC-dextran. Beads removed from control mice after 12 d were extensively vascularized whereas beads from DC-flk1–immunized mice were essentially avascular (Fig. 3). Quantitative analysis of tumor cell–induced vascularization decreased by 65% in mice vaccinated with DC-flk1 compared with DC-AP group (P < 0.01).

To determine whether active immunization against flk1 could inhibit tumor metastasis, we tested immunized mice in a metastasis model of Lewis lung carcinoma. Mice immunized with flk1-AP–pulsed DCs developed significantly reduced numbers of metastases in the lungs (P < 0.01) compared with mice immunized with either AP-pulsed DCs or PBS (Fig. 4, A and B). Most significantly, 6 of 10 mice were found to be tumor-free in the DC-flk1–immunized group, whereas all mice in the two control groups developed extensive pulmonary metastases. Macroscopic observations were further confirmed by microscopic examination of lung samples showing significant reduction or complete absence of tumor cells in the lungs of DC-flk1–immunized mice, whereas extensive tumor infiltration was found in lungs of control groups (Fig. 4 C). In addition to examination of tumor metastasis in immunized mice, a similar experiment was conducted to evaluate the effect of flk1 immunization on survival. As shown in Fig. 5, while all mice had died of tumor by day 63 in both control groups, 40% mice in the DC-flk1 group survived over 200 d. The other 60% of DC-flk1 group died almost as quickly as the control mice, suggesting that it was “all or none” protection.

DC-flk1 immunization was also tested in a B16 melanoma model. After three immunizations with flk1-AP–pulsed DCs at days 0, 10, and 20, mice were challenged with an intravenous injection of B16 tumor cells. As shown in Fig. 6, DC-flk1 immunization significantly inhibited the formation of lung metastases compared with PBS or DC-AP control groups (P < 0.01). Interestingly, DC-AP group also showed a certain level of metastatic inhibition compared with PBS group.

To determine which T cell population (i.e., CD4, CD8, or both) was responsible for the antitumor activities, we conducted an in vivo T cell depletion assay in mice vaccinated with flk1-AP-pulsed DCs in the Lewis lung metastatic model (23). Mice were treated with either anti-CD4, anti-CD8, or control Ab before vaccinations with DC-flk1 and again 2 wk later. Depletion of respective T cell population was confirmed by FACS^® analysis of splenocytes of one mouse from each group. The antitumor effect of vaccination against flk1 was abrogated in CD8-depleted but not in CD4-depleted mice (Fig. 7). These results suggest that the CD8 population is mainly responsible for the T cell–mediated anti-flk1 immune response.

As described above, breeding experiments indicated an effect of DC-flk1 immunization on prenatal development. To further examine whether active immunization against flk1 has an effect on normal physiological angiogenesis, we analyzed wound healing in immunized mice using an excisional cutaneous wound model. Full-thickness wounds were created on the dorsum of control or DC-flk1–immunized mice 10 d after three immunizations on days 0, 10, and 20. Wound areas were measured twice weekly until they were completely healed. No significant difference in wound healing was observed between DC-flk1–immunized and control mice (Fig. 8 A). Microscopic examination of wound tissues revealed no difference in wound epithelialization, vascularity or collagen deposition (Fig. 8 B).

By using DCs to break tolerance to self-antigen, we demonstrate that active immunization against the VEGF receptor flk1 can inhibit tumor angiogenesis and metastasis in mice. We chose flk1 as the molecular target for these studies because of its recognized importance in angiogenesis and tumor growth. These results provide strong proof of concept that an active immunization approach against an angiogenic target can be effectively used for tumor therapy. Our data are consistent with those of other studies that have used antagonists of the VEGF pathway to inhibit tumor-associated angiogenesis (10–14). Recently, Wei et al. reported that cross-immunization with xenogeneic endothelial cells was effective in inducing an endothelial cell–specific immunity and protecting mice from tumor challenge. In their study, peptides within αv integrin and EGFR-2 were identified as the possible antigen epitopes (26). Plum et al. showed that administration of a liposomal FGF-2 peptide vaccine led to abrogation of FGF-mediated angiogenesis and tumor development in mouse models (27). These studies suggest that angiogenesis-directed immunotherapy approach may potentially extend to other angiogenic targets such as VEGF, the Tie2 receptor (28), angiopoietin-1 (29, 30), αvβ3 (31), and VE-cadherin (32).

We detected a strong CTL response after DC-flk1 immunization. It is likely that flk1-specific CTLs exert antitumor effect by destroying endothelial cells in the neovasculture surrounding tumor nodules. Notably, flk1-specific neutralizing antibody was also present in the sera of immunized mice. The antibody effectively blocked the binding of VEGF to native flk1 at the surface of endothelial cells. Blockade of the VEGF-flk1 pathway could play an important role in the suppression of angiogenesis observed in DC-flk1–immunized mice. In the in vivo T cell depletion experiment, depletion of CD8⁺ T cells abrogated the antitumor effect in the Lewis lung metastasis model whereas CD4⁺ T cell depletion had a marginal effect. This result suggested a dominant role of CD8⁺ CTLs in the antitumor activity while the role of flk-1 neutralizing antibody could not be excluded. This finding is consistent with the notion that antitumor immunity depends on CD8⁺ T cells in many experimental tumor models (33). Given that an anti-flk1 neutralizing antibody response occurred in DC-flk1–immunized mice, however, we were surprised that CD4⁺ T cell depletion did not affect the antitumor activity. One possible explanation for this observation is that the antibody induced by DC-flk1 immunization recognized a flk1 epitope that is CD4 independent. In the study by Wei et al., CD4⁺ T cells and antibody were reported to be important in the antitumor activity induced by xenogenic endothelial cell vaccine (26). Plum et al. found that an anti-FGF-2 antibody was induced after immunization with liposomal FGF-2 peptide vaccine (27). The importance of CD8⁺ versus CD4⁺ T cell immunity in these studies may have been determined by the use of different target antigens, adjuvants, and routes of immunization.

Flk1 is expressed in normal vascular endothelium albeit in lower levels compared with that in tumor vasculature and is important in fetal development (9, 34–37) and in physiological angiogenesis such as wound healing (34, 38, 39). One important observation in our studies was the adverse effect of DC-flk1 immunization on mouse pregnancy. This finding is consistent with the role of angiogenesis in multiple aspects of female reproductive function and fetal development. In nonpregnant adult mice, however, vaccination did not induce observable toxicity. Wound healing was not affected in the immunized mice. This result is compatible with the report that the angiogenesis inhibitor vasostatin does not impair wound healing at tumor-inhibiting doses (40). The lack of toxicity suggests a different sensitivity of malignant tumor growth than physiologic wound healing to inhibition of angiogenesis. Another possible explanation is that the level of expression of flk1 in quiescent normal vascular endothelium may be too low to be recognized by flk1-specific CTLs.

As an antiangiogenesis therapy, active immunization affords the opportunity to elicit a sustained effect in chronic disease settings. In contrast, antagonist drug therapies must be administered for long periods of time. In our experiments, the anti-flk1 immune response was monitored as long as 20 wk after the last booster immunization and was found to gradually diminish after the last immunization but could be restored to a therapeutic level with 1–2 booster immunizations (data not shown). Thus, the immune response against an angiogenesis-related target (and any effects associated with the immune response) may be short lived but could be recalled by repeated immunizations. In the context of immunotherapy, this approach benefits from the advantages of targeting an antigen expressed on nontransformed cells, i.e., endothelial cells. Endothelial cells do not bear the problems associated with targeting antigens on tumor cells, such as genetic instability, mutations, or loss of antigen expression. Another advantage is that one single antiangiogenesis vaccine may be used against multiple tumor types. This notion is supported by this study and others (26, 27). In conclusion, angiogenesis-targeted immunotherapy offers the potential for a new approach to treatment of cancer and other angiogenesis-related diseases.

We thank Drs. B. Pytowski, P. Kussie, and F. Liao for critical comments, and M.J. Kim and D. Burtrum for technical assistance.

This work was supported in part by Small Business Innovation Research (SBIR) grant CA86648-01 from the National Institutes of Health to Y. Li. An abstract of this paper was first presented on March 3, 2000, at the Keystone Symposium “Experimental and Clinical Regulation of Angiogenesis,” in Salt Lake City, UT.