Combining Antiangiogenics to Overcome Resistance: Rationale and Clinical Experience

The importance of tumor angiogenesis has been recognized for almost 80 years (1), and 4 decades ago, Folkman (2) proposed it as a therapeutic target, recognizing its critical role in the growth and survival of tumors larger than 1 mm³. However, it is only in the last 10 years that this knowledge has successfully translated to inhibitory strategies in clinical practice. Endothelial cells (EC) have been assumed to show genomic stability in comparison with the multiple somatic mutations acquired within tumors during their growth, leading to the belief that ECs would avoid resistance to therapy (3, 4). Targeting of angiogenesis has been validated by evidence of efficacy from several agents directed against components of the main proangiogenic VEGF signaling pathway at the level of the ligand (VEGF) and its receptors [VEGFR (5–10)]. Bevacizumab, a monoclonal antibody (mAb) targeting VEGF, was shown to result in improved overall survival (OS) in metastatic colorectal and lung cancer when combined with chemotherapy (6, 9), and progression-free survival (PFS) in metastatic breast and ovarian cancer (11, 12). In metastatic renal cell cancer (mRCC), bevacizumab was also shown to prolong the time to progression as monotherapy (10) and in combination with interferon (13). Tyrosine kinase inhibitors (TKI) targeting VEGFR, such as sunitinib (8) and sorafenib (5), as monotherapies have also produced an OS benefit for patients with mRCC. Similarly, sorafenib has now become the standard of care for advanced hepatocellular carcinoma because it was shown to increase OS compared with placebo (14). However, in other tumors, despite the initial benefit seen in some patients treated with VEGF pathway inhibitors, complete responses have not generally been documented, and most patients will experience tumor progression and succumb to their disease. Moreover, attempts to use these drugs in the adjuvant setting (e.g., in colorectal cancer) have thus far been disappointing, yielding the same results as chemotherapy alone (15).

Lessons could be learned from the paradigm of mRCC management, in which antiangiogenic strategies have been most successfully adopted in routine clinical practice. Acquired resistance to anti-VEGF therapy may mean that the order in which targeted agents are introduced to patients is important, such that a survival benefit may be derived from sequential monotherapy with multitargeted TKIs (16). An analysis of a prospective trial (17), numerous retrospective studies, and a subgroup analysis of expanded-access programs for sorafenib suggested that there is an increase in PFS for a TKI switch in mRCC (18). These data support a lack of cross-resistance between multitargeted TKIs; however, ongoing sorafenib/sunitinib crossover trials, such as SWITCH (www.ClinicalTrials.gov; NCT00732914), may address these issues more definitively.

An alternative to using VEGF-targeting agents sequentially is to focus on combining agents that affect different aspects of angiogenesis. Our knowledge about the complexity of angiogenesis is growing rapidly, and it is recognized that VEGF-independent drivers include multiple interactions among diverse growth factors and receptors involving ECs, NOTCH/δ-like ligand 4 (DLL4), angiopoietin (Ang)-Tie, placental growth factor (PlGF), tumoral cells (SDF1/CXCR4), pericytes [platelet-derived growth factor (PDGF) and transforming growth factor (TGF-β)], extracellular matrix (ECM) components (integrins and cadherins), inflammatory cells (tumor-associated macrophages and Tie-2–expressing monocytes), and bone-marrow–derived cells (Fig. 1; reviewed in refs. 19 and 20). A new paradigm for the development of malignant angiogenesis was recently proposed based on the finding in glioblastoma multiforme (GBM) that tumor ECs can arise directly through differentiation of CD133⁺/CD144⁺ tumor cells (21). In addition, preclinical work is providing insights into mechanisms of resistance to antiangiogenic therapy. These mechanisms are described as being intrinsic (preexisting) or adaptive (acquired), and they may explain why some tumors do not respond from the outset and why others progress after initial shrinkage (22). This information has driven the clinical drug development of a number of compounds that target various components of tumor angiogenesis (Table 1). In this review, we discuss the progress that has been made and the pitfalls involved in combining different antiangiogenic strategies with the aim of overcoming resistance.

Strategies that have been devised to overcome some of the known mechanisms of resistance to antiangiogenic therapies may conceptually be divided into 2 groups: (i) combining direct (targeting ECs) and indirect (targeting growth factors/receptors) inhibitors; and (ii) combining 2 indirect inhibitors, resulting in either horizontal blockade from 2 different inhibitors targeting proteins across separate signaling pathways or vertical blockade from targeting molecules within the same pathway (Fig. 1; ref. 23).

Vascular disrupting agents (VDA) are direct inhibitors of angiogenesis that target established vasculature. The selectivity of VDAs for tumoral vessels results from structural differences, with tumor blood vessels being immature and therefore inadequately covered by pericytes. This makes them more susceptible to microtubule depolarization, leading to disruption of the actin cytoskeleton and subsequent cell reshaping that causes blood flow reduction and tumoral necrosis (24, 25). However, regrowth is characteristically observed from a peripheral rim of residual tumor cells via rapid recruitment of circulating endothelial progenitor cells (EPC). Pretreatment with VEGFR-2 inhibitors was associated with the disruption of EPC recruitment and enhancement of VDA-mediated blood flow reduction in tumor-bearing mice (26). Combinations of VDA with agents inhibiting the VEGF pathway have displayed synergy in xenograft models. For example, the VDA ZD6126 and the VEGFR-2 TKI vandetanib (ZD6474), when used in combination, produced 3.5-fold and 6-fold greater delays in tumor growth compared with that observed with either monotherapy (27). Similarly, the combination of bevacizumab and the VDA combretastatin A1 phosphate (OXi4503) resulted in significantly enhanced tumor growth delay in a renal cancer xenograft (27 days; P < 0.05) compared with bevacizumab (8 days) or OXi4503 (18 days) as monotherapy (28). The rationale and preclinical results for these combinations have encouraged clinical trials of different VDAs with VEGF inhibitors, especially bevacizumab.

Indirect inhibitors target receptors/ligands that stimulate EC growth/survival. Horizontal blockade encompasses the targeting of different classes of receptors. The intimate relationship between ECs and pericytes is associated with increased expression of PDGF receptor (PDGFR)-β, Tie-2, and Ang-1 (29). These factors may well contribute to the relative lack of efficacy seen with anti-VEGFR-2 (semaxinib) treatment in tumor xenografts compared with SU6668, an inhibitor of PDFGR, VEGFR, and fibroblast growth factor receptor [FGFR (30, 31)]. Of interest, although greater benefit was observed from the combination in a Rip1-Tag2 mouse model of pancreatic neuroendocrine tumors, when used as monotherapy, the VEGFR-2 inhibitor (semaxinib) was more efficacious at blocking initial angiogenic switching, whereas the PDGFR inhibitor (SU6668) was more effective against end-stage bulky disease (32). In this in vivo model, adaptive resistance to anti-VEGFR-2 therapy was also abrogated with the use of a dual FGF/VEGFR-2 TKI [brivanib (33)]. Similar strategies targeting both ECs and pericytes with TKIs that target VEGFR-2 (AEE788) and PDGFR (imatinib), respectively, showed that the reduced pericyte coverage as a result of PDGFR blockade with imatinib monotherapy was ineffective at controlling tumor growth; however, when imatinib was used in combination with AEE788, tumor control was superior to that seen with AEE788 alone (34).

The finding that tumors that show intrinsic resistance to anti-VEGF therapy display sensitivity to blockade of DLL4-NOTCH signaling led to interest in targeting the NOTCH-DLL4 and VEGF pathways together (35–37). NOTCH inhibition using γ secretase inhibitors may alter the self-renewing capacity of intestinal crypt cells; however, DLL4 seems to be restricted to the vascular compartment (38, 39). Combination therapy using antibodies against DLL4 and VEGF substantially inhibited tumor growth in non–small cell lung cancer xenografts [Calu6 and MV522 (37)]. However, recent studies showed that chronic DLL4 blockade can produce more vascular tumors in vivo, thus raising concerns over its clinical applicability (40). For this reason, regulators implicated in attenuation of NOTCH signaling, such as epidermal growth factor–like domain 7 (EGFL-7), are now being investigated as potential alternative targets (41).

The angiopoietins Ang-1, -2, -3, and -4 regulate EC homeostasis via interactions with their receptors Tie-1/Tie-2. This results in vascular quiescence/angiogenesis (Ang-1) or vessel regression and angiogenic sprouting (Ang-2), depending on the environmental cues present (42, 43). Ang-2–directed therapies have been assessed along with anti-VEGF strategies (44). Reduction of endothelial sprout formation and colon cancer xenograft growth following treatment with an Ang-2 binding peptibody, L1-7 (N), was even more significant when VEGF antibodies were added (45). Similarly, addition of an Ang-2 antibody, 3.19.3, to the VEGFR-1,2,3 TKI cediranib, the VEGFR-2 and epidermal growth factor receptor (EGFR) TKI vandetanib, AZ10167514 (VEGFR TKI), or DC101 (VEGFR-2 antibody) doubled the rate of tumor growth inhibition over that seen with each single agent (46). Other signaling pathways and molecules implicated in resistance to antiangiogenics, such as integrins, PI3K-AKT-mTOR, cMET/HGF, and TGF-β (endoglin, Alk1), are being targeted in combination with classic antiangiogenics to expand the options of novel combinations to overcome primary and secondary resistance.

Vertical blockade of the VEGF pathway at different levels has been extensively investigated in preclinical studies. Mouse models with the human head and neck cancer cell line CAL33 treated with cediranib (VEGFR-1,2,3 TKI) and bevacizumab showed a significant tumor growth reduction compared with either agent alone. Mice treated with single-agent therapy showed increased tumoral secretion of VEGF; however, this adaptive effect was abrogated with the combination therapy, suggesting that inhibition of both VEGF ligand and its receptor (VEGFR) can provide additional benefit over inhibition of either alone (47). The transmembrane glycoproteins neuropilins (NRP)-1,2 act as cofactors to enhance the binding of VEGF to VEGFR-2 (48). Recently, 2 high-affinity mAbs that bind discrete Sema3A and VEGF binding domains of NRP-1 were tested in combination with a VEGF antibody with good preclinical activity (49). Furthermore, blocking the same receptor at 2 different levels was recently observed to enhance receptor inactivation, and combining antibodies that separately inhibit ligand binding and VEGFR-3 dimerization more effectively inhibited tumor angiogenesis and lymphangiogenesis in vivo (50). The data gained from these preclinical studies combining inhibition at multiple points within the VEGF pathway are now informing a highly active focus for research in the clinical setting.

Current antiangiogenic strategies in clinical oncology practice either use single-agent, multitargeted TKIs with activity against different proangiogenic receptors (e.g., VEGFRs, PDGFR-β, FLT3, and FGFRs) or involve sequencing of drugs that work within the same vertical angiogenic axis. In addition, combination (horizontal) antiangiogenic strategies are now emerging in early clinical trials, and it will be important to establish their tolerability and determine whether additional gains in efficacy are offset by challenges in administration due to toxicity. Because bevacizumab is the best-established antiangiogenic agent, many combinations use this antibody as the backbone (Tables 2 and 3). Treatments targeting the VEGF pathway share well-recognized class-adverse effects related to VEGF blockade, namely, hypertension, proteinuria, hemorrhage, arterial thromboembolic events, and poor wound healing. These effects are typically downstream consequences of suppression of cellular signaling pathways important in the regulation and maintenance of the microvasculature, and they highlight the properties shared by tumor vessels and the vasculature of normal organs (51). In addition to the adverse effects related to VEGF-axis blockade, off-target effects result in frequently observed toxicities, such as nausea, diarrhea, hand–foot syndrome (HFS), and mucositis (52). A key issue in the development of antiangiogenic drugs is the lack of reliable predictive biomarkers for sensitivity to this strategy. A retrospective study on genetic polymorphisms in the VEGF gene showed correlations with OS and toxicity (53). Prospective confirmatory studies are needed, and many rationally selected candidates, such as serum levels of VEGF, FGF, PlGF, and other cytokines, have failed to show any ability to select patients who are more likely to benefit (54). The importance of establishing pharmacodynamic biomarkers early in the development of antiangiogenic combinations cannot be overstated. Hitherto, the most common approach was to use imaging studies to identify changes in tumor blood flow or vascular permeability (i.e., changes in blood flow by contrast-enhanced ultrasound or changes in K^trans by MRI). Serum levels of circulating markers (VEGF, sVEGFR-2, sVEGFR-3, and PlGF) are also usually modified by antiangiogenic therapies, but their value as early markers of activity remains to be determined. Current efforts are focused on measuring the soluble isoforms of circulating VEGFA-VEGFA121; however, at present, the lack of a reliable intermediate biomarker of efficacy in early-phase studies means that most combination antiangiogenic therapies will still require larger phase II/III trials to prove clinical benefit.

Investigators in clinical trials have pursued the rationale for combining VDAs with VEGF(-R) inhibitors to abrogate malignant progression at the viable tumor rim. To date, however, the schedules used have not followed the preclinical studies, in which abrogation of the circulating EPC spike was maximal when a VEGFR-2 inhibitor was administered prior to the VDA. Instead, in clinical trials conducted to date, the VDA was administered before the VEGF inhibitor for the purpose of monitoring safety. Combretastatin A4 phosphate (CA4P) added to 10 mg/kg of bevacizumab was well tolerated at doses up to 63 mg/m² CA4P every 2 weeks in a phase I study of 14 patients, with 2 grade 3 toxicities (asymptomatic self-limiting atrial fibrillation and hemorrhage) observed (55). One third of the patients had disease stabilization for more than 4 months (I. Judson, personal communication), and dynamic contrast-enhanced (DCE)-MRI showed statistically significant reductions in tumor perfusion and vascular permeability (56). The benefit of adding CA4P and bevacizumab to carboplatin and paclitaxel is currently being tested in a randomized phase II trial in 60 patients with non–small cell lung cancer, with preliminary results showing an increase in response rate (RR) from 38% to 55%, favoring the CA4P arm, and a similar PFS [median, 8.6 months vs. 8.9 in the control arm (57)]. A newer VDA, ombrabulin (AVE8062), is currently being tested in combination with bevacizumab in a phase I dose escalation study (ClinicalTrials.gov; NCT01193595).

Bevacizumab was assessed in combination with vandetanib in a phase I study in which hypertension was frequent and the dose-limiting toxicity (DLT) was allergic reaction (58). Despite 2 partial responses observed in patients with carcinomas of the small bowel and pancreas, chronic dosing was associated with a number of grade 2–3 toxicities, including myelosuppression, rash, prolonged QTc interval, and diarrhea, such that dose escalation was limited to 2 dose levels. The selected dose for further studies was vandetanib, 200 mg orally once daily with bevacizumab, 5 mg/kg every 21 days. Pharmacodynamic studies supported further exploration of this dual antiangiogenic approach. Although vandetanib alone would be expected to produce a 20% reduction in blood flow, DCE-MRI studies in the combination phase I trial revealed a 31% reduction in mean K_trans in all 7 patients following 2 cycles of therapy, with decreased levels of EPC in 5 patients and increased numbers of apoptotic mature ECs in 3 patients (59).

The combination of bevacizumab with sunitinib has shown clinical activity, albeit with significant toxicity, with high RRs in early-phase clinical trials, including all histologic subtypes of mRCC and other typically chemoresistant tumors, such as urothelial cancers and melanoma. The maximum tolerated dose (MTD) was judged to be 50 mg daily of sunitinib in a 4-weeks-on/2-weeks-off schedule, with 10 mg/kg bevacizumab every 2 weeks. However, this dose was considered too toxic for chronic therapy, with 87% of patients developing grade ≥3 adverse events [mainly hypertension, fatigue, thrombocytopenia, proteinuria, and HFS (60)]. In the phase I study in mRCC, 48% of the patients were withdrawn because of adverse events (61). The development of microangiopathic hemolytic anemia in 5 of 12 patients treated at the highest dose level was of major concern. Two further patients developed severe symptoms with hypertension, thrombocytopenia, renal insufficiency, proteinuria, and reversible posterior leukoencephalopathy syndrome, leading to a U.S. Food and Drug Administration warning (62) and discontinuation of this combination in other clinical trials (63).

The combination of bevacizumab with sorafenib seems to be safer, although toxicity remains a concern. A phase I combination trial established an MTD at dose level 1, using sorafenib, 200 mg twice daily and bevacizumab, 5 mg/kg every 2 weeks (64). Dose-limiting toxicities were proteinuria and thrombocytopenia; however, 74% of the patients required dose reductions, mainly due to HFS, fatigue, and hypertension. A high rate of antitumor activity (43% RR for heavily pretreated patients with ovarian cancer, with a median of 5 lines of prior chemotherapy) encouraged the development of better-tolerated regimens. An intermittent schedule of sorafenib, 200 mg twice daily 5 days per week, and bevacizumab, 5 mg/kg every 2 weeks, was associated with fewer dose reductions (41%) while maintaining a high RR of 47% in ovarian cancer (65). This schedule was also tested in 39 patients with GBM, resulting in a 37% RR, although poor tolerance led to one third of the patients discontinuing treatment. Dose reduction to sorafenib, 200 mg daily, significantly improved tolerability, with 15% of the patients experiencing grade ≥3 toxicities (66). The same schedule was tested in a phase II trial in 44 patients with advanced neuroendocrine tumors, with an RR of 9.8% reported; however, 20% of patients developed grade ≥3 HFS and 15% developed grade ≥3 asthenia (67). No objective responses were observed when this schedule was used in phase II trials in advanced melanoma and colorectal cancer (68, 69). Two phase I studies (70, 71) are investigating the safety and tolerability of this combination with paclitaxel in solid tumors, and preliminary reports suggest some activity, with 3 complete responses (ovarian, endometrial, and urachal cancers) and 8 partial responses (6 ovarian, 1 endometrial, and 1 prostate) among 19 evaluable cancer patients.

Following a similar rationale, investigators combined bevacizumab with axitinib (AG-013736) in a phase I study with FOLFOX chemotherapy. The DLT was hypertension, observed in 81% of patients on combination treatment compared with 26% on axitinib monotherapy. The MTD was bevacizumab 2 mg/kg q14d with axitinib 5 mg twice daily. A phase II study in advanced CRC patients (ClinicalTrials.gov; NCT00460603) leading on from this study is ongoing, and an RR of 33% has been reported (72).

Theoretically, angiogenic blockade at different receptors could overcome the drawback of overlapping toxicity, and horizontal combinations are now being explored in early clinical trials. Thrombospondin-1 (TSP-1) is an endogenous inhibitor of angiogenesis that mediates its effects through interactions with different receptors (α_v and β integrins, CD47, and CD36), and it can also inhibit nitric oxide signaling (73). A synthetic analogue of the N-terminal region of TSP-1, ABT-510, in combination with bevacizumab (10 mg/kg) has shown no DLT at doses up to 100 mg s.c. twice daily in a phase I study, suggesting a lack of overlapping toxicities clinically; however, efficacy data are yet to be published (74). Proangiogenic interactions between Ang-1,2 and the Tie-2 receptor can be suppressed therapeutically with F_c-peptide fusion proteins, such as AMG-386, that comprise the F_c domain of immunoglobulin G1 fused to a synthetic peptide with potent binding to Ang-1,2 and result in inhibition of Tie-2 receptor activation (75). The results of a randomized, double-blind, placebo-controlled phase II combination study of AMG-386 and sorafenib in 152 patients with mRCC were recently reported (76). Patients were randomized to sorafenib, 400 mg twice daily, with AMG-386, 10 mg/kg; AMG-386, 3 mg/kg; or placebo. The combination was well tolerated, and although an increase in the incidence of diarrhea was reported (70%, 67%, and 56%, respectively), other common toxicities, such as HFS and hypertension, were similar among the 3 arms. Despite a higher RR with the combination of sorafenib + AMG-386 compared with the sorafenib + placebo combination (37% vs. 24%), the primary endpoint was not met, as the median PFS of ∼9 months was similar in all 3 arms (P = 0.52).

It therefore remains to be determined whether, in addition to their greater tolerability, these combinations can provide a benefit beyond increased RRs.

Although some encouraging examples of activity from antiangiogenics in a number of tumors have been reported, resistance to these agents is a major hurdle in the further development of this class of cancer therapy. Mechanisms of resistance are still being defined, and it is anticipated that the knowledge gained will give rise to increasingly rational attempts to overcome them. Active efforts are under way to address the optimal scheduling and combinations of antiangiogenics. It remains to be determined whether vertical inhibition at numerous levels within the VEGF pathway will improve clinical outcomes. Likewise, despite positive data from preclinical studies, the anticancer efficacy of horizontally combining agents that act across different antiangiogenic pathways remains to be established in clinical practice. Although these preclinical studies were frequently performed using different agents compared with those tested in clinical trials, they generally involved the same mechanisms of action. Other explanations, such as a lack of specificity or different potency for the target, might therefore explain some of the discrepancies seen so far between preclinical and clinical studies.

Recent evidence from early clinical trials raises concerns that overlapping toxicities from antiangiogenic combinations may limit feasibility in the long term due to cumulative toxicity. It will be important to determine whether sequencing these agents can provide similar benefit while sparing the toxicities from the combinations. This is important because, unlike conventional cytotoxics, antiangiogenic treatments may require chronic dosing, and therefore traditional DLT endpoints within the first cycle of treatment will underestimate the severity of adverse events from these agents together. Nevertheless, interest in pursuing alternative doses and schedules of a combined antiangiogenic approach remains, largely due to the encouraging efficacy signals in terms of increased tumor response, especially in traditionally chemoresistant tumors and in patients who have been heavily pretreated.

As more antiangiogenic agents that target pathways other than VEGF enter clinical trials, the opportunities for testing novel antiangiogenic combinations will continue to expand. For new drugs that target the VEGF pathway itself, this could increase the potential for achieving successful combinations if they show a different safety profile. However, the list of available agents (see Table 1) is already long, and a key goal now is to gain an improved understanding of the mechanisms that underlie clinical resistance to VEGFR inhibitors. This should lead to the further development of rational combinations, such as VEGFR/C-MET combinations.

A key area under investigation in this arena is the development of relevant predictive biomarkers and pharmacodynamic endpoints. This would permit appropriate patient selection and advance the concept of treating to a biologically effective dose rather than the MTD and thus accelerate the therapeutic development of better-tolerated combinations and schedules for future use in clinical practice.

S.B. Kaye is a consultant/advisory board member of Roche. No other potential conflicts of interest were disclosed.

Conception and design: V.M. Garcia, B. Basu, L.R. Molife, S.B. Kaye

Development of methodology: V.M. Garcia, L.R. Molife

Analysis and interpretation of data: B. Basu, L.R. Molife, S.B. Kaye

Writing, review, and/or revision of the manuscript: V.M. Garcia, B. Basu, L.R. Molife, S.B. Kaye

The Drug Development Unit of the Royal Marsden NHS Foundation Trust and the Institute of Cancer Research is supported in part by a program grant from Cancer Research UK. Support was also provided by the Experimental Cancer Medicine Centre (to the Institute of Cancer Research) and the National Institute for Health Research Biomedical Research Centre (jointly to the Royal Marsden NHS Foundation Trust and the Institute of Cancer Research). V.M. Garcia was supported by a grant from the Fundacion Para la Investigacion del Hospital Universitario La Paz (REX-09).