Resistance to VEGF inhibitors is emerging as a major clinical problem. Notch signaling has been implicated in tumor angiogenesis. Therefore, to investigate mechanisms of resistance to angiogenesis inhibitors, we transduced human glioblastoma cells with retroviruses encoding Notch delta-like ligand 4 (DLL4), grew them as tumor xenografts and then treated the murine hosts with the VEGF-A inhibitor bevacizumab. We found that DLL4-mediated tumor resistance to bevacizumab in vivo. The large vessels induced by DLL4-Notch signaling increased tumor blood supply and were insensitive to bevacizumab. However, blockade of Notch signaling by dibenzazepine, a γ-secretase inhibitor, disrupted the large vessels and abolished the tumor resistance. Multiple molecular mechanisms of resistance were shown, including decreased levels of hypoxia-induced VEGF and increased levels of the VEGF receptor VEGFR1 in the tumor stroma, decreased levels of VEGFR2 in large blood vessels, and reduced levels of VEGFR3 overall. DLL4-expressing tumors were also resistant to a VEGFR targeting multikinase inhibitor. We also observed activation of other pathways of tumor resistance driven by DLL4-Notch signaling, including the FGF2-FGFR and EphB4-EprinB2 pathways, the inhibition of which reversed tumor resistance partially. Taken together, our findings show the importance of classifying mechanisms involved in angiogenesis in tumors, and how combination therapy to block DLL4-Notch signaling may enhance the efficacy of VEGF inhibitors, particularly in DLL4-upregulated tumors, and thus provide a rational base for the development of novel strategies to overcome antiangiogenic resistance in the clinic. Cancer Res; 71(18); 6073–83. ©2011 AACR.
Tumor angiogenesis is essential for tumor growth and metastasis and is triggered by a variety of proangiogenic factors, regulated by many angiogenic pathways. The most important and best characterized is the VEGF pathway (1). Blockade of VEGF signaling inhibits tumor angiogenesis and growth in various preclinical models (2, 3) and having systemic effects (4). The effect of anti-VEGF strategy has been validated in randomized phase III clinical trials. Disruption of VEGF signaling by using bevacizumab, a humanized anti-human VEGF antibody or sorafenib and sunitinib, 2 small chemical inhibitors of receptor-tyrosine kinases (RTK) including VEGF receptors (5) have yielded a significant improvement in progression-free survival but rarely in overall survival for patients with several types of cancers including metastatic colorectal cancer (6), advanced non–small cell lung cancer (7), metastatic breast cancer (8), advanced clear-cell renal-cell carcinoma (9–11) and advanced hepatocellular carcinoma (12). However, the duration of effect is often short in clinical trials (13). In preclinical models a number of tumors do not respond at all, or respond initially but become resistant at a late stage of VEGF inhibition (14, 15). In some experimental cases, such therapy may even increase tumor invasiveness and metastasis (16, 17). A further problem is that although this therapy is highly specific, patients are not selected on the basis of pathways of sensitivity or resistance in their tumors. Thus, tumor resistance to VEGF inhibition has become a major obstacle to the effectiveness of antiangiogenic agents in the clinic (18, 19). Clearly, understanding the resistance mechanisms will enhance anti-VEGF efficacy by developing new therapeutic approaches or rational drug combinations.
Notch signaling is a conserved pathway affecting many biological processes including angiogenesis (20). Ligand binding to the Notch receptor results in 2 successive proteolytic cleavages of the receptor. The second cleavage, catalyzed by γ-secretase, releases the Notch intracellular domain (NICD), which translocates to the nucleus where it activates the transcription of downstream targets such as EphrinB2, Hes1, and Hey1 (20). DLL4, the latest identified Notch ligand, is predominantly expressed in arterial endothelial cells (EC) during embryonic development, and haploinsufficiency of DLL4 leads to embryonic lethality from severe vascular defects (21). DLL4-Notch signaling inhibits physiologic angiogenesis by reducing the number of tip cells and thus decreasing angiogenic sprouting (22, 23). In adults, however, we have shown that DLL4 is strongly upregulated in tumor vasculature (24–28). Alterations of DLL4 expression inhibit multiple EC functions important in angiogenesis (28–30). Recently, we, and other groups, showed that DLL4-Notch signaling regulates tumor growth by decreasing angiogenesis but improving vascular function; conversely, blockade of DLL4-Notch signaling strikingly increases nonproductive angiogenesis but reduces tumor growth (26, 31–33).
The VEGF pathway interacts with DLL4-Notch signaling in the vasculature (20). VEGF induces the expression of DLL4 and Notch1 (28, 32, 34), while DLL4-Notch signaling reduces signaling in the VEGF pathway by decreasing VEGFR2 expression and increasing VEGFR1 expression (28–30, 32, 34). These characteristics have attracted many groups to further address the roles of DLL4 in tumor response to antiangiogenic therapy (26, 31, 32). Here, we report that DLL4-Notch signaling-mediated tumor resistance to anti-VEGF therapy by inducing the formation of larger vessels, which no longer depend on VEGF signaling, and by activating multiple other signaling pathways.
Materials and Methods
U87GM (U87, human glioblastoma) and HT1080 (human fibrosarcoma) cell lines were obtained from Cancer Research UK Cell Services. Generation of U87-EV and U87-DLL4 cells and growth conditions were described previously (26).
Xenograft mouse models
All protocols were carried out under Home Office regulations (26). The tumor cells (107 for U87, 5 × 106 for HT1080) were implanted into 6- to 8-week-old female BALB/c SCID mice (Harlan Sprague Dawley, Inc.) s.c. with 100 μL of cell suspension with an equal volume of Matrigel (BD Bioscience). Each group consisted of 5 mice. Tumor growth was monitored and measured using calipers. Tumor volume was calculated from a formula (V = L × W × H × π/6). When used, dibenzazepine (DBZ) and bevacizumab were injected intraperitoneally (i.p.) every 3 days at a dose of 8.1 mmol/kg and 10 mg/kg, respectively. DBZ was resuspended in 0.5% (w/v) Methocel E4M and 0.1% (w/v) Tween 80. Anti-EphrinB2 and sEphB4 were given i.p. 3 times a week at a dose of 20 mg/kg, respectively. Sorafenib was dissolved in a 50% Cremophor EL/50% ethanol as a stock solution (4×); once the compound was in solution, the aqueous component was added gradually to generate the dosing solution (12.5% Cremophor EL/12.5% ethanol/75% water; ref. 5). Sorafenib was administered daily at 60 mg/kg via oral gavage. PD173074 was given daily at 60 mg/kg in 50 mmol/L sodium lactate buffer, pH4.0 by oral gavage. Each animal experiment was repeated at least for 2 to 3 times.
The mouse CD31 and NG2 (desmin or α-SMA) stainings on cryostat sections were done as previously described (26). Immunohistochemistry on paraffin sections at 4 μm was done after antigen retrieval in Tris-EDTA buffer, pH 9.0 or in citrate buffer, pH 6.0 (for FGF2) in a Decloaker (Biocare Medical). Primary antibodies used were against CA9 (M75, 1/75, 1 hour), VEGF (VG1, 5 μg/mL, at 4°C for 16 hours), VEGFR3 (AF743, 1/125, 35 minutes), and FGF2 (sc79, 1/500, 1 hour). Tumor vessels were detected on paraffin sections by rat anti-CD34 monoclonal antibody (MEC14.7, 1/250, 1 hour) using a Novolink kit (Leica Microsystems) after the antigen was retrieved in retrieval buffer (Dako R1669). Primary antibodies were washed off in PBS and detected using a polymer-based system (Dako). The peroxidase reaction was developed using diaminobenzidine provided in the kit. Slides were washed and mounted in aqueous mountant.
Quantitative real-time PCR
RNA extraction from xenograft tumors, quantitative real-time PCR (qPCR) protocol, mHey1 and β-actin primers were previously described (26).
Protein extraction and Western blotting was done using standard techniques (28).
In situ hybridization
Mouse VEGFR2 mRNA was specifically localized in formalin-fixed tissue using an antisense riboprobe synthesized with SP6 RNA polymerase using 35S-UTP (∼800 Ci/mmol) and a plasmid encoding the 5′-end 2619bp of mouse VEGFR2. The riboprobe was used without hydrolysis (28).
All data are presented as mean ± SEM. Statistical analysis including t test for 2 groups or 1-way ANOVA for multiple groups, and parametric generalized linear model with random effects for tumor growth were done using GraphPad Prism 4.0b software. Statistical significance is indicated by *, where P < 0.05, and **, where P < 0.01.
DLL4-Notch signaling mediates tumor resistance to VEGF inhibition
SCID mice were implanted with U87GM cells that were transduced with empty vector retroviruses (U87-EV) or DLL4-encoded retroviruses (U87-DLL4) and treated with bevacizumab. DLL4 promoted tumor growth and reduced mouse survival, whereas bevacizumab inhibited tumor growth and increased animal survival of both U87-EV and U87-DLL4 tumors when the treatment was started on the day of tumor implantation (Fig. 1A) (26). To mimic the clinical situation where antitumor therapies are initiated only after patient has an established or advanced tumor, we initiated anti-VEGF treatment when the tumor reached 150 mm3 in size. Administration of bevacizumab was still effective in inhibiting the growth and improving animal survival of U87-EV tumors. In sharp contrast, U87-DLL4 tumors exhibited an intrinsic resistance to bevacizumab (Fig. 1B). Thus, DLL4 expression mediates tumor resistance to anti-VEGF therapy.
We utilized DBZ, a small molecule inhibitor of γ-secretase to block Notch signaling. DBZ strongly inhibited Notch signaling in vitro, as shown by reduced protein levels of the activated NICD and the Notch target Hes1 in ECs (Supplementary Fig. S1). Five injections of DBZ inhibited U87-DLL4 tumor growth and improved animal survival (Fig. 1C); continuous administration of DBZ completely abolished the tumor growth stimulated by DLL4. The effect of DBZ on Notch signaling in vivo was confirmed by assessing mHey1 expression by qPCR (Supplementary Fig. S2).
Next, we assessed the ability of DBZ to resensitize U87-DLL4 tumors to VEGF inhibition, initiating the therapy when the tumor reached 150 mm3 in size (Fig. 1D). For U87-EV tumors, DBZ only slightly inhibited the growth whereas bevacizumab delayed the growth; however, combining DBZ with bevacizumab resulted in a more striking inhibition of the growth. For U87-DLL4 tumors, DBZ significantly delayed the growth, particularly so at later points, whereas treatment with bevacizumab did not affect the growth at all. However, when treatments were combined, DBZ not only abolished the tumor resistance to bevacizumab but also synergistically enhanced the inhibition of tumor growth and improved the animal survival with bevacizumab.
Similar experiments with DBZ and bevacizumab were also done in the human fibrosarcoma HT1080 tumor model where the tumor exhibited an intrinsic resistance to VEGF inhibition. Similar therapeutic effects were observed (Supplementary Fig. S3), suggesting that blockade of endogenous Notch signaling by DBZ can also sensitize a tumor with intrinsic de novo resistance to anti-VEGF therapy.
DLL4-induced large vessels but not pericyte coverage contribute to anti-VEGF resistance
We investigated tumor neovasculature by immunostaining for the EC marker CD31 (Fig. 2ABC). DLL4 significantly reduced the vessel number but increased the vascular lumen size (26). DBZ increased the number of vessels by 30% and 120% in U87-EV and U87-DLL4 tumors, respectively and reduced the vascular lumen size in U87-DLL4 tumors.
Bevacizumab greatly decreased the vessel number by 12-fold in U87-EV tumors, while only moderately by 2.5-fold in U87-DLL4 tumors. However, bevacizumab did not affect vascular lumen size at all in U87-DLL4 tumors. In contrast, the combination therapy not only reduced the vessel number in both tumors but also disrupted the large vessels in U87-DLL4 tumors. Thus, the results suggest that the large vessels triggered by DLL4-Notch signaling contribute to the tumor resistance.
We costained for CD31 and the pericyte marker NG2 (Supplementary Fig. S4). DLL4 decreased the pericyte coverage of tumor vessels (26); DBZ increased the pericyte coverage in both U87-EV and U87-DLL4 tumors. However, bevacizumab alone did not affect the level of pericyte coverage in either tumor compared with their relative controls. In the tumors cotreated with DBZ and bevacizumab, the levels of vessel coverage by pericytes were similar to those observed in the tumors treated with DBZ alone.
Reductions in hypoxia and VEGF contribute to anti-VEGF resistance
Hypoxia increases VEGF production in tumors and we hypothesized reduced hypoxia would decrease VEGF expression and hence reduce tumor dependency on VEGF. We assessed hypoxia levels before and following therapeutic treatments as indicted by response of the hypoxia inducible factor-1 target carbonic anhydrase IX (CA9; Fig. 3AB). Expression of CA9 was lower in U87-DLL4 tumors compared with the U87-EV control. DBZ treatment significantly increased tumor hypoxia, as shown by strong CA9 staining in both tumor types. Bevacizumab also induced hypoxia/CA9 expression in U87-EV tumors because of reduced vessel numbers; however, it did not significantly increase hypoxia in U87-DLL4 tumors.
Combined DBZ and bevacizumab treatment increased tumor hypoxia much more efficiently than DBZ alone in either tumor type. In addition, the combination therapy not only increased the number of cells expressing CA9 but also the intensity of CA9 staining (Fig. 3A). Consistent with hypoxia (Fig. 3B), similar results were obtained for tumor necrosis (Fig. 3C). Less hypoxia and less necrosis in DLL4 tumors could be attributed to the persistence of functional large vessels in bevacizumab-treated DLL4 tumors.
We costained tumor tissues with anti-CA9 and anti-VEGF antibodies. VEGF staining was observed on vessels along with weak cytoplasmic positivity in the tumor which seemed to increase in intensity around necrotic areas (Fig. 3A). Overall, the relative proportional expression of tumoral VEGF in each group (Fig. 3D) was similar to CA9 in each group (Fig. 3B). The level of VEGF in U87-DLL4 tumors was much lower than in U87-EV tumors. VEGF staining on vessels was clearly detected in U87-EV tumors as opposed to very faint staining in U87-DLL4 tumors. DBZ increased VEGF expression in both tumor types. In addition, DBZ increased VEGF staining on vessels in U87-DLL4 but not U87-EV tumors. Bevacizumab significantly increased VEGF expression only in U87-EV but not U87-DLL4 tumors. Combined DBZ with bevacizumab substantially increased VEGF expression in both tumors but decreased the VEGF staining on vessels although this effect was much clearer in U87-EV than in U87-DLL4 tumors (Fig. 3A–D). Bevacizumab treated U87-EV tumors showed vascular disruption, a neutrophilic vasculitis, and fibrinoid necrosis, an effect that was notably absent in U87-DLL4 tumors. Taken together, these results suggest that blockade of Notch signaling by DBZ induced tumor hypoxia and necrosis by disrupting the large vessels, and consequently increased the level of VEGF.
Alterations of expression of VEGFRs contribute to anti-VEGF resistance
We investigated the expression of VEGFR1 and VEGFR2 in stromal cells by qPCR using mouse specific primers. In vitro, DLL4 increased VEGFR1 (and soluble VEGFR1) expression and significantly decreased VEGFR2 expression in ECs (29, 30). In vivo, DLL4 increased the mRNA levels of VEGFR1 but VEGFR2 transcription was not deceased in U87-DLL4 tumors (Fig. 4A). We then analyzed VEGFR2 transcripts in tumor blood vessels using in situ hybridization (Fig. 4B). DLL4 reduced VEGFR2 expression only in the large vessels, whereas DBZ restored VEGFR2 expression in tumor vessels.
However, bevacizumab did not affect the expression of VEGFR2 in either U87-EV or U87-DLL4 tumors. In contrast, combined treatment of DBZ with bevacizumab restored the expression of VEGFR2 in the few remaining vessels, similar to the effects of DBZ alone. Indeed, the protein levels of VEGFR2 expressed in the large vessels were also decreased in U87-DLL4 tumors compared with U87-EV tumors (26). We further investigated the expression of VEGFR3. DLL4 reduced the mRNA level of VEGFR3 in human microvascular EC (hMEC-1) in vitro and the protein level in tumor neovasculature in vivo (Supplementary Fig. S5). Taken together, these results suggest that the large vessels were insensitive to VEGF due to the low expression of VEGFR2, and increased expression of VEGFR1 further diminished the sensitivity of VEGF-VEGFR2 signaling in tumors.
To inhibit mouse and human VEGF effects and exclude the possibility that antibody transport was a mechanism of resistance, we treated both tumor types with sorafenib to inhibit VEGFR1-R3. Sorafenib delayed the growth and improved animal survival of U87-EV tumors (Fig. 4C) but the efficiency was less than bevacizumab treatment (see Figure 1). Sorafenib marginally inhibited U87-DLL4 tumor growth and significantly improved animal survival of U87-DLL4 tumors, although this effect was greater than bevacizumab treatment alone (see Figure 1), presumably due to the blockade other RTKs (5).
Activation of FGF2-FGFR signaling contributes to anti-VEGF resistance
DLL4-Notch signaling significantly increased the expression of FGF2 in ECs in vitro (29) and in U87-DLL4 tumors in vivo at the protein level (Fig. 5A). We then treated both U87-EV and U87-DLL4 tumors with PD173074 to inhibit FGF2-FGFR signaling (35). As little as 100 nmol/L of PD173074 could inhibit FGF2-FGFR signaling in ECs in vitro, reflected by elimination of tyrosine phosphorylation of ERK induced by FGF2 (Fig. 5B). In U87-EV tumors, PD173074 did not significantly inhibit the growth but slightly enhanced the efficacy of bevacizumab (Fig. 5C). In contrast, PD173074 slightly inhibited U87-DLL4 growth but significantly improved the tumor response to bevacizumab in the combination treatment (Fig. 5D).
Activation of EphB4-EphrinB2 signaling contributes to anti-VEGF resistance
The expression of EphB4 and EphrinB2 in tumors was assessed by qPCR. EphB4, but not EphrinB2, was significantly upregulated in U87-DLL4 tumors compared with the control (Fig. 6A), suggesting that EphB4-EphrinB2 signaling is enhanced as a result of increased EphB4 in the tumor. We therefore treated both tumor types with anti-EphrinB2 antibody or soluble EphB4 protein (sEphB4) that inhibit internalization and phosphorylation of EphrinB2, respectively. Single therapy with anti-EphrinB2 or sEphB4 did not significantly affect the growth of either tumor type (Fig. 6B), but combined anti-EphrinB2 or sEphB4 with bevacizumab significantly inhibited the growth and improved survival of U87-DLL4 tumors (Fig. 6C). Neither anti-EphrinB2 nor sEphB4 enhanced the efficacy of bevacizumab in combination treatments in U87-EV tumors (Supplementary Fig. S6). Thus, the results suggest that increased EphrinB2 reverse signaling in DLL4-tumors contributes to the tumor resistance to anti-VEGF therapy.
We investigated tumor neovasculature further by immunostaining for CD31 (Fig. 6D). Anti-EphrinB2 or sEphB4 significantly reduced the vessel number but not the vessel size in U87-EV tumors. However, both anti-EphrinB2 and sEphB4 increased the vessel number and reduced the vessel size in U87-DLL4 tumors. Combined anti-EphrinB2 or sEphB4 with bevacizumab not only decreased the vessel number but also abolished most, if not all, of the large vessels in DLL4 tumors. Taken together, the results suggest that blockade of EphrinB2 reverse signaling partially resensitized the tumor response to anti-VEGF therapy by disrupting the majority of large vessels induced by DLL4-Notch signaling.
Multiple mechanisms are involved in tumor resistance to anti-VEGF therapy. These include preexistence/activation of alternative angiogenic pathways, recruiting vascular progenitor cells from the bone marrow, increasing pericyte coverage, and enhanced capability of tumor cells to invade (13, 15). Here, we show that DLL4-Notch signaling mediates tumor resistance to bevacizumab in vivo. DLL4 is predominantly upregulated in tumor vasculature in preclinical models (21, 26, 31, 33) and in clinical tumor samples (24–28). However, DLL4 is also substantially upregulated in a subset of tumor cells in several human cancers including glioblastoma, colon, and breast cancer (24–28). Thus, our model by upregulating DLL4 in glioblastoma cells growing in immunodeficient mice is clinically relevant. Our findings may explain why so many patients in clinical trials do not respond to bevacizumab (13). Importantly, disruption of DLL4-Notch signaling by DBZ not only abolished the tumor resistance but also synergistically enhanced the therapeutic efficacy of bevacizumab.
All current clinical trials investigate the therapeutic efficiency of VEGF inhibitors without examining the expression of markers such as DLL4 and their subsequent impact in tumor and patient response to therapy. We investigated DLL4 expression in advanced breast tumors from AVF2119g, a randomized phase III trial by adding bevacizumab to capecitabine chemotherapy. When the tumors did not express DLL4 or had very low level expression, patients had a significant prolongation of progression free-survival upon adding bevacizumab; however, when DLL4 was highly expressed there was no benefit to the patients on adding bevacizumab (36). Thus our findings are clinically corroborated by the investigation of this randomized trial.
Our results indicated that the large vessels triggered by DLL4-Notch signaling play a central role in tumor resistance to bevacizumab. Bevacizumab only decreased the numbers of small vessels but did not affect the large vessels. These large vessels effectively delivered oxygen and nutrients to tumor cells as revealed by an increase in blood perfusion (26), resulting in a reduction in hypoxia and necrosis in DLL4 tumors. Reduced hypoxia decreased VEGF production and thus decreased the tumor dependency on VEGF. DLL4 decreased VEGFR2 expression in the large vessels and increased VEGFR1 expression in the tumor microenvironment, further diminishing the sensitivity of VEGF-VEGFR2 signaling. Blockade of DLL4-Notch signaling by DBZ abolished the large vessels, leading to a substantial increase of hypoxia and VEGF expression in DLL4 tumors, and consequently enhanced the therapeutic efficacy of bevacizumab. We found increased VEGF expression and although this may appear surprising in the context of bevacizumab treatment, we would still expect to detect intracellular VEGF in cells synthesizing or internalizing it. The observation of a neutrophilic vasculitis in the EV-tumors treated with bevacizumab, was most likely a response to the necrosis induced in the control tumors in contrast to the DLL4 tumors.
Increased and tight pericyte coverage of newly formed vessels has been shown to contribute to tumor resistance to anti-VEGF therapy (13, 37). However, we found that this is not the case in our models since DLL4 reduced the pericyte coverage and the reduction of pericyte coverage did not increase the sensitivity of bevacizumab. The SDF-1α/CXCR4 axis was reported to be essential for the pericyte recruitment in the PDGF-BB–expressing tumors (38). We previously reported that DLL4 inhibits CXCR4 expression in vitro and in vivo (39), thus providing a mechanism for the decreased pericyte coverage in DLL4-tumors.
VEGFR3 is also important in tumor angiogenesis (40, 41) and is regulated by Notch signaling (23, 41, 42). In U87-DLL4 tumors, the protein level of VEGFR3 was significantly downregulated, which may also produce resistance to VEGFR3 inhibitors. Sorafenib had a marginal effect on the growth of DLL4 tumors, suggesting that DLL4 may also contribute to tumor resistance to other antiangiogenic agents for example, blockers of VEGFR3 or VEGFC. FGF2 has been shown to confer tumor resistance to anti-VEGF signaling in the RIP-Tg2 pancreatic islet tumor model, some syngeneic mouse models, and in vitro cultures (14, 43, 44). We found that FGF2 was significantly increased and bevacizumab further increased FGF2 expression in DLL4 tumors. However, the FGF2-FGFR inhibitor, PD173074, marginally inhibited the growth of DLL4 tumors but significantly improved the tumor response to bevacizumab, suggesting that FGF2-FGFR minimally contributes to the tumor resistance.
Recent studies showed that interaction of EphB4 with EphrinB2 inhibited EC sprouting and promoted circumferential growth of vessels, and reverse signaling via EphrinB2 decreased vessel numbers, but increased vessel lumen diameter, resulting in tumor progression (45, 46). EphrinB2 is a well-known downstream target of DLL4-Notch signaling (29, 30) and VEGF can effectively induce EphrinB2 but repress EphB4 (47). However, VEGF levels are reduced in DLL4 tumors due to reduced hypoxia, which may reverse this phenotype. Indeed, we found that EphB4, but not EphrinB2, was significantly higher in DLL4 tumors. sEphB4 was shown to antagonize forward and reverse signaling of EphrinB2, reduce vessel density and inhibit xenograft tumor growth (48). Anti-EphB4 antibody inhibited tumor angiogenesis and growth alone and in combination with bevacizumab (49). Combined therapy of sEphB4 with soluble DLL4 displayed cumulative efficacy against mouse insulinoma growth (50). We found that both sEphB4 and anti-EphrinB2 reduced the vessel number in U87-EV tumors, but increased the vessel number and reduced the vessel size in U87-DLL4 tumors. However, neither sEphB4 nor anti-EphrinB2 alone inhibited the growth of either tumor type. Importantly, both sEphB4 and anti-EphrinB2 resensitized the tumor response to bevacizumab by disrupting the large vessels, further highlighting the role of the large vessels in tumor resistance to anti-VEGF therapy. This also shows the heterogeneity of response to such therapies in different tumor models.
In conclusion, we found that DLL4-induced Notch signaling mediates tumor resistance to anti-VEGF therapy by inducing the formation of large vessels and activating multiple pathways in tumors. Importantly, blocking these bypass pathways had little effect on the more sensitive parent tumors, but showed greater effects in combination with bevacizumab in DLL4-tumors. The clinical implications for clinical trial design are that patients need to be carefully categorized by analysis of pathways in tumors, and combination strategies designed. Our findings provide a rational base for the development of novel strategies to overcome antiangiogenic resistance, and suggest that combination therapy by blocking both DLL4 and VEGF pathways may enhance the antitumor efficacy in the clinic, and DLL4 levels in tumor and vascular tissue should be assessed as a predictive biomarker.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
We thank Sandra Peak, Del Watling, Rachel Hayes, Sarah Scott, Christine White, and Barbara Cross (Cancer Research UK Biological Resources Unit) for assistance with xenograft experiments, Chunning Yang and Kristen Lekstrom (Antibody Discovery and Protein Engineering, MedImmune) for the production of anti-EphrinB2 antibody and sEphB4 recombinant protein, and Rosemary Jeffery (Cancer Research UK London Research Institute) for help with in situ hybridization.
This work was supported by Cancer Research UK and European Union 6th Framework Grant Angiotargeting.
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.