Resistance and Escape From Antiangiogenesis Therapy: Clinical Implications and Future Strategies

Selection of clonal populations with upregulated alternative and compensatory pathways.

Various data raise the question of whether anti-VEGF therapy selects for clonal populations that have upregulated alternative and compensatory proangiogenic signaling cascades. In pancreatic cancer models, treatment with a VEGF receptor 2 (VEGFR-2) –inhibiting antibody resulted in 10 to 14 days of tumor stasis followed by tumor regrowth, with apparent acceleration of tumor vascularity and increased expression of fibroblast growth factor 1 (FGF-1), FGF-2, Ephrin-A1 (Eph-A1), Eph-A2, and angiopoietin-1.⁷ In tumors whose angiogenesis should otherwise be suppressed by ectopic expression of angiogenesis inhibitors, compensatory upregulation of VEGF, platelet-derived growth factor (PDGF), and FGF-2 is observed.¹⁷ In human studies, predictable upregulation of FGF-2, VEGF, and placental growth factor (PlGF) has been detected in response to antiangiogenic therapy.¹⁸ In sunitinib-treated mice, VEGF, granulocyte colony-stimulating factor, stromal cell–derived factor 1-alpha (SDF-1α), stem-cell factor, and osteopontin have all been found to be increased.¹⁸

FGFs are important promoters of angiogenic and mitogenic activity.¹⁹ FGF-1 and FGF-2 induce angiogenesis, and preclinical models suggest that FGF binding protein releases FGF-2 from the extracellular matrix, allowing it access to receptors.^20,21 Cross talk between VEGF and FGF may stimulate angiogenesis synergistically, with variable effects on vessel size and function.¹⁹ In a phase II trial of FOLFIRI+B (folinic acid, fluorouracil, irinotecan, and bevacizumab) in colorectal cancer, significant increases in FGF-2 as well as hepatocyte growth factor, PlGF, SDF-1, and macrophage chemoattractant protein-3 were observed in plasma samples.²² In patients with glioblastoma, treatment with cediranib, a pan-VEGFR inhibitor, resulted in decreased tumor edema by dynamic magnetic resonance imaging; subsequent increases in FGF-2 and SDF-1 levels correlated with progression.²³ Murine models of pancreatic and renal-cell carcinoma further support that upregulation of FGF-2 is associated with tumor progression in the face of VEGFR-2 blockade, and upregulation of FGF-2, angiopoietin-2, and PDGF-A has been implicated in bypass of antiangiogenic signaling.¹⁷

Recognizing the complexity of angiogenesis, other signaling systems are potentially implicated. In tumor models without hypoxia inducible factor 1-alpha (HIF-1α), interleukin-8 (IL-8) expression has been shown to support angiogenesis.²⁴ In a phase II trial of FOLFIRI+B in colorectal cancer, increased baseline IL-8 correlated with decreased PFS.²² Increased insulin-like growth factor 1 has been associated with increased prostaglandin E2 expression, which correlates with increased VEGF expression and enhanced angiogenesis.²⁵ VEGF precursor mRNA undergoes alternative splicing that generates both pro- and antiangiogenic forms, and variant splicing has been hypothesized to explain both initial and acquired resistance to antiangiogenic therapy.²⁶ Studies of the murine double minute (MDM2) oncogene have demonstrated a p53-independent, hypoxia-driven translocation to the cytoplasm, with binding and stabilization of VEGF mRNA, increasing translation, VEGF protein production, cell survival, and angiogenesis.²⁷ Epidermal growth factor receptor (EGFR) –mediated activation of the phosphoinositide 3–kinase/Akt/mammalian target of rapamycin pathway can also upregulate HIF-1α and VEGF, and there have been separate pathways described by which VEGF production via phosphoinositide 3–kinase is independent of HIF.²⁸ One report suggested that blockade with anti-EGFR antibody resulted in selection of tumor cell subpopulations with increased angiogenic potential.²⁹ These studies form the basis of dual targeting of EGFR and VEGFR with drugs such as vandetanib.³⁰

Selection of clones with greater invasive capacity in nonangiogenic environment.

Evidence suggests that when VEGFR/PDGF receptor (PDGFR) signaling is blocked with sunitinib (preclinical pancreatic and glioblastoma models) in addition to anticipated antitumor effects, a tumor phenotype is observed with heightened invasive and metastatic potential. In an orthotopic mouse model of glioblastoma multiforme, it was observed that blockade of VEGF with a neutralizing antibody resulted in a tumor phenotype that continued to grow, albeit more slowly, with persistent invasive capacity.³¹ These models remain unsubstantiated and controversial, but they raise potential concerns regarding unintended consequences of blockade of these pathways, and mechanisms remain under exploration.^32,33

Aside from the upregulated mitogenic pathways discussed previously, the hypoxic, acidic, high–interstitial fluid pressure features of the tumor microenvironment seem to favor tumor clones with more aggressive behavior.³⁴ It has been argued that this microenvironment fuels nonproductive angiogenesis that fails to relieve the hypoxic stress, perpetuating a self-reinforcing loop of pathologic angiogenesis.³⁵ Hypoxic conditions created by VEGF pathway inhibitors correlate with upregulation of the MET oncogene, promoting invasive behavior.^36,37 MET overexpression has been observed in variety of solid tumors including advanced ovarian cancer.³⁶ Cabozantinib (XL-184) is an oral, potent inhibitor of MET and VEGFR-2, and a phase II trial (100 mg once daily orally over 12 weeks) in advanced, progressive epithelial ovarian cancer showed a promising clinical response rate (overall, 24%).³⁸

Selection of clones resistant to hypoxia.

Pancreatic ductal adenocarcinoma is characteristically hypovascular and in clinical trials has shown no treatment response to bevacizumab.^37,39 This tumor type seems to have underlying resistance to hypoxia, indicating a high probability that other cell populations could evolve this trait. Approximately 75% of these tumors carry inactivating mutations of the p53 gene, which may contribute to survival in hypoxic conditions.^40,41 Additionally, 75% of pancreatic cancer cell lines in one study constitutively expressed HIF-1α, correlating with resistance to apoptosis induced by hypoxia and/or glucose deprivation. Constitutive HIF-1α expression correlates with increased expression of glucose transporter 1 and Aldolase-A, both of which are associated with anaerobic metabolism.⁴² These data suggest that preferential utilization of particular metabolic pathways may favor certain cell populations when VEGF-targeted therapy imposes a hypoxic environment. The combination of p53 loss supporting cell survival and the utilization of metabolic pathways that additionally favor survival in a hypoxic environment would offer a simple mechanism by which clonal populations may thrive in the hypoxic environment.

Compensation for VEGF blockade and resistance to hypoxia can be mediated by the microenvironment. Hypoxic conditions promote recruitment of bone marrow–derived cells that include vascular progenitors (eg, endothelial and pericyte progenitors) and vascular modulators.^43,44 Vascular modulators are a class of cells including tumor-associated macrophages, immature monocytes, VEGFR-1 hemangiocytes, and CD11b-myeloid cells.^45,46 HIF-1α has been implicated in the recruitment of CD45 myeloid cells with various subpopulations expressing Tie-2, VEGFR-1, CD11b, mature F4/80 tumor-associated macrophages, and endothelial and pericyte progenitors.^46,47 This environment favors selection of tumor clones protected by CD11b+Gr1+ myeloid cells that express proangiogenic factors such as Bv8, which has been shown to be partially responsible for angiogenesis promotion during VEGF blockade.^48,49 These cells facilitate progression to frank carcinoma and may represent an important alternative therapeutic target.⁵⁰

Selection and support of normalized vasculature, supporting tumor cell growth and function.

Anti-VEGF therapy results in vascular normalization that remains of unclear therapeutic significance.³⁵ Abnormal tumor vasculature functions poorly, reflected by chaotic, stagnant, even reversed flow.^34,51 The hypoxic, acidic microenvironment promotes protease-mediated matrix remodeling, anchorage-independent growth, and resistance to apoptosis despite increasing genetic instability.^52,53 Inhibition of VEGF reduces vessel size and tortuosity; the remaining vessels have greater pericyte coverage, and the basement membrane is normalized.⁵⁴ Some studies have posited a therapeutic drug-delivery advantage (within a limited therapeutic window) gained by changes in vascular permeability.³⁴ Other studies have suggested that normalization and maturation may represent mechanisms of therapeutic escape. In one study, myeloid cell–driven angiogenesis was selectively ablated in a mouse model of breast cancer, resulting in reduced VEGF, reduced vascular density, increased maturation, and concomitant maturation/normalization; this normalization correlated with increased tumor growth and progression.⁵⁵

The oxygen-sensing prolyl hydroxylase domain (PHD) proteins may also play a role in vascular normalization and maturation and contribute to more aggressive tumor behavior after antiangiogenic therapy. PHD-2 is an oxygen-sensing enzyme that hydroxylates HIF when sufficient oxygen is available, targeting the protein for degradation.⁵⁶ Under hypoxic conditions resulting from anti-VEGF treatment, a compensatory release of proangiogenic cytokines generates vessels characterized by irregular borders, hypermobile cells, loosely attached layers, and denuded areas.^34,57 In PHD-2 heterozygous mice, endothelial cells upregulate VEGFR-1 and VE-cadherin to stabilize vessels, reduce leakage, and improve vessel perfusion, allowing normalization of neovasculature; homozygous deficiency does not allow this function.⁵⁷ In some non–small-cell lung cancers, PHDs are present and upregulated, suggesting a possible vascular normalization function associated with overall tumor growth.⁵⁸

The hyperactivated angiogenic state characteristic of some tumor microenvironments results in a dense, nonproductive vascular network that fails to support tumor growth. Recent data suggest that the Notch ligand Delta-like ligand-4 (DLL4), which is normally induced by VEGF, is actually a negative-feedback regulator of vascular sprouting and branching. In mice lacking DLL4, or in circumstances where it is inhibited, there is excessive and nonproductive angiogenesis (blind-end budding).⁵⁹ Despite the excess of angiogenesis, tumor growth is decreased, even in tumors resistant to anti-VEGF therapy. The current hypothesis is that in the absence of DLL4, there is inadequate maturation, and therefore inadequate function, of the vessels.⁵⁹ It has been shown that tumor resistance to bevacizumab can be induced by transfection of DLL4; blockade of Notch signaling reverses this resistance. Mechanisms implicated include increased stromal VEGFR-1, decreased vascular VEGFR-2, diffusely reduced VEGFR-3, and increased signaling through FGF-2/FGF receptor and Ephrin-B4/Ephrin-B2. Cell lines transfected with DLL4 were also resistant to a VEGFR-targeted multikinase inhibitor.⁶⁰

Among vascular support structures, pericytes provide survival factors and temper the proliferation rate of endothelial cells.^61–63 Pericytes are positioned around endothelial cell junctions, provide physical support, and release low levels of endothelial survival factors such as VEGF; for new vascular branches to form, pericytes must detach.⁶⁴ Within tumors, pericyte coverage is variable, but it is less extensive than seen in normal tissue, and cells take on an abnormal shape and express markers characteristic of immature, less contractile mural cells.^64–66 Evidence suggests that tumor cell shedding into the circulation is inversely proportional to pericyte coverage.⁶⁷ Chronic VEGF expression in the tumor microenvironment seems to interfere with PDGF-B signaling, interfering with vascular smooth muscle cell function and smooth muscle recruitment.⁶⁸ VEGF blockade increases the signaling of angiopoietin-1, resulting in improved endothelial cell function and pericyte recruitment.⁶⁹

PDGF/PDGFR signaling itself is implicated in rescue and escape from VEGF blockade.⁷⁰ The PDGF family provides mitogenic signaling necessary for pericyte recruitment and maturation. Immature vessels with poor pericyte investment seem vulnerable to anti-VEGF treatment, and post-treatment (surviving) vessels seem to have relatively high pericyte coverage.^71,72 In one study, PDGF-BB was expressed by endothelial and tumor cells, and PDGFRβ was expressed in pericyte-like cells; PDGF-BB increased the migration and VEGF production of these pericyte-like cells, and these noted functions could be blocked by PDGFRβ inhibitors.⁷³ AX102, a highly specific inhibitor of PDGF-B signaling, was highly effective in combination with bevacizumab in ovarian cancer models.⁷⁴ Dual targeting of endothelial cells and pericytes has been considered to hold potential as an antivascular therapeutic approach in ovarian carcinoma, and agents such as pazopanib, sunitinib, sorafenib, and BIBF-1120 are being tested in clinical trials.

The role of PlGF in tumor angiogenesis remains controversial and poorly understood. PlGF overexpression in tumors in vivo has been show to correlate with decreased tumor growth, and it has also been shown to correlate with normalization of tumor vasculature, possibly through heterodimerization with VEGF, reducing VEGF potency.^75,76 PlGF-null mice demonstrate reduced response to VEGF. However, PlGF is chemotactic for endothelial cells and monocytes in vitro; it may be involved with mobilization of bone marrow–derived cells, and it increases the response of cultured endothelial cells to VEGF-induced survival, proliferation, and migration.^77,78 Aflibercept (VEGF Trap) is a protein that contains the VEGF-binding regions of VEGFR-1 and VEGFR-2 fused to the Fc portion of human IgG1. It acts as a high-affinity soluble VEGFR decoy receptor and therefore inhibits the activity of both VEGF and PlGF.⁷⁹ Two randomized phase II studies showed that even in heavily pretreated patients, single-agent aflibercept could induce tumor response and delay progression.^80,81 A recent combined phase I/II trial of docetaxel plus aflibercept in patients with recurrent ovarian, primary peritoneal, or fallopian tube cancer resulted in a 54% response rate, and 24% of patients had stable disease.⁸²

Ang1 and Ang2 exist in balance to promote organization and maturation of neovasculature. Overexpression of Ang1 compared with Ang2 results in dense hypervascularization with large vessels, and excess Ang2 binding to Tie-2 results in destabilized and leaky blood vessels.⁸³ Perivascular Tie-2–expressing monocytes are also proangiogenic.³⁵ Coordinated Ang1 and Ang2 signaling through Tie-2 receptors is thought to be an alternative pathway to VEGF through which neovascularization can occur, and coordinated upregulation may represent an additional mode of anti-VEGF–therapy bypass.

Selection of subpopulations capable of co-opting normal/existing vasculature.

Grade 2 and 3 astrocytomas have been observed to develop neovasculature without an identified proliferative endothelial component.^84,85 Vascular acquisition in this case seems to be via perivascular tumor invasion with incorporation of normal vasculature into the tumor structure. As an apparent corollary, three human studies reported patients with multifocal recurrence while receiving bevacizumab therapy.^37,86,87 These findings, in conjunction with data suggesting that fibroblasts can draw existing vessels toward and into a fresh wound, point toward cell populations that have the phenotypic capacity to home toward existing vasculature and to incorporate it rather than relying on a cytokine-driven neovascular response. These clonal populations would be favored in an environment exposed to antiangiogenic therapy.

Endothelial cell genetic instability and resistance to therapy.

Contrary to previous dogma stating that endothelial cells within tumors are genetically stable and therefore less likely to evolve resistance to therapy, recent work has identified cytogenetic abnormalities in human tumor-associated endothelial cells.^88,89 Although they have not been validated, these studies raise the possibility that high rates of endothelial cell mutation within tumors could contribute to therapeutic resistance to a greater degree than previously surmised.^88,89 What remains to be elucidated is why tumor-based endothelial cells have significant cytogenetic instability and whether this high rate of mutation would confer the capacity to select for unique subpopulations of endothelial cells that would develop therapeutic resistance over (relatively) short periods of time.⁹⁰