Normalization of Tumor Vasculature: An Emerging Concept in Antiangiogenic Therapy

Solid tumors account for more than 85% of cancer mortality. Because cancer cells in these tumors require access to blood vessels for growth and metastasis, inhibiting vessel formation offers hope for reducing the mortality and morbidity from these tumors (1). When administered as single agents, antiangiogenic drugs have produced modest objective responses in clinical trials (2, 3), but overall they have not yielded long-term survival benefits (4). In contrast, when given in combination with chemotherapy, bevacizumab, an antibody targeted against the potent angiogenic molecule vascular endothelial growth factor (VEGF), produced an unprecedented increase in survival (5 months) in colorectal cancer patients (5).

These clinical data support the earlier predictions of Teicher (6), who postulated that combined administration of antiangiogenic and cytotoxic (chemo- and radiation) therapies would yield maximal benefit because such combinations would destroy two separate compartments of tumors—cancer cells and endothelial cells. Cytotoxic agents would kill cancer cells directly, and antiangiogenic agents would kill cancer cells indirectly by depriving them of nutrients. Emerging data suggest that chemotherapy and radiation therapy may also have antiangiogenic effects, directly damaging or killing tumor endothelial cells, bone marrow–derived cells (e.g., endothelial progenitor cells), and/or circulating endothelial cells, and thus enhancing the indirect killing of cancer cells (7). Furthermore, cancer cells may express receptors for angiogenic growth factors (e.g., VEGFR1 or VEGFR2), and thus antiangiogenic drugs (e.g., antibody to VEGF) could lead to the direct killing of cancer cells by interfering with survival pathways and/or enhancing sensitivity to other treatments (7). All of these mechanisms imply that an antiangiogenic agent would always augment the response to radiation or chemotherapy.

This is paradoxical, however. One would expect that destroying the vasculature would severely compromise the delivery of oxygen and therapeutics to the solid tumor, producing hypoxia that would render many chemotherapeutics, as well as radiation, less effective. Indeed, some studies show that antiangiogenic therapy can compromise the delivery of drugs to tumors (8), as well as antagonize the outcome of radiation therapy (9, 10). How can we reconcile these diametrically opposed outcomes of combination therapy with the positive preclinical and clinical data? The answer to this question is critical for optimizing the efficacy of combined antiangiogenic and cytotoxic therapy.

To resolve this paradox, I have hypothesized that the judicious application of antiangiogenic agents can “normalize” the abnormal tumor vasculature, resulting in more efficient delivery of drugs and oxygen to the targeted cancer cells (Fig. 1A) (11). The increased penetration of drugs throughout the tumor would enhance the outcome of chemotherapy, and the ensuing increased level of oxygen would enhance the efficacy of radiation therapy and many chemotherapeutic agents. Here, I review our current understanding of vascular normalization, discuss recent preclinical and clinical data supporting this counterintuitive hypothesis, and point out potential research avenues for further exploration.

• Download high-res image
• Open in new tab
• Download Powerpoint

Why Normalize the Tumor Vasculature?

To obtain nutrients for their growth and to metastasize to distant organs, cancer cells co-opt host vessels, sprout new vessels from existing ones (angiogenesis), and/or recruit endothelial cells from the bone marrow (postnatal vasculogenesis) (12). The resulting vasculature is structurally and functionally abnormal (Table 1) (13). Blood vessels are leaky, tortuous, dilated, and saccular and have a haphazard pattern of interconnection (Fig. 1). The endothelial cells lining these vessels have aberrant morphology, pericytes (cells that provide support for the endothelial cells) are loosely attached or absent, and the basement membrane is often abnormal—unusually thick at times, entirely absent at others.

These structural abnormalities contribute to spatial and temporal heterogeneity in tumor blood flow. In addition, solid pressure generated by proliferating cancer cells compresses intratumor blood and lymphatic vessels, which further impairs not only the blood flow but also the lymphatic flow (14). Collectively these vascular abnormalities lead to an abnormal tumor microenvironment characterized by interstitial hypertension (elevated hydrostatic pressure outside the blood vessels), hypoxia, and acidosis.

Impaired blood supply and interstitial hypertension interfere with the delivery of therapeutics to solid tumors. Hypoxia renders tumor cells resistant to both radiation and several cytotoxic drugs. Independent of these effects, hypoxia also induces genetic instability and selects for more malignant cells with increased metastatic potential (15). Hypoxia and low pH also compromise the cytotoxic functions of immune cells that infiltrate a tumor. Unfortunately, cancer cells are able to survive in this abnormal microenvironment. In essence, the abnormal vasculature of tumors and the resulting abnormal microenvironment together pose a formidable barrier to the delivery and efficacy of cancer therapy. This suggests that if we knew how to correct the structure and function of tumor vessels, we would have a chance to normalize the tumor microenvironment and ultimately to improve cancer treatment. The fortified tumor vasculature may also inhibit the shedding of cancer cells into the circulation—a prerequisite for metastasis.

In the past, higher doses of drugs and hyperbaric oxygenation have been used to increase the tumor concentrations of drugs and oxygen, respectively. These strategies have not shown much success in the clinic, however. One reason for this failure is that tumor vessels have large holes in their walls (16). As stated earlier, this leakiness leads to interstitial hypertension as well as spatially and temporally nonuniform blood flow. If the delivery system is flawed, it does not matter how much material is pumped into it. The drugs and oxygen will become concentrated in regions that already have enough and will still not reach the inaccessible regions (17). However, if we fix the delivery system, more cells are likely to encounter an effective concentration of drugs and oxygen. This is the rationale for developing therapies that normalize the tumor vasculature. These therapies do not merely increase the total uptake of drugs and oxygen but also distribute these molecules to a larger fraction of the tumor cells by fixing the delivery system.

How Should We Normalize the Tumor Vasculature?

In normal tissues, the collective action of angiogenic stimulators (e.g., VEGF) is counterbalanced by the collective action of angiogenic inhibitors such as thrombospondin-1 (Fig. 1D). This balance tips in favor of the stimulators in both pathological and physiological angiogenesis (18). However, in pathological angiogenesis, the imbalance persists. Therefore, restoring the balance may render the tumor vasculature close to normal. On the other hand, tipping this balance in favor of inhibitors may lead to vascular regression and, ultimately, to tumor regression.

If we had antiangiogenic agents that completely destroyed tumor vessels without harming normal vessels, we would not need to add cytotoxic therapy. Unfortunately, such agents are not currently available. It is conceivable that increased doses of currently available antiangiogenic agents could produce complete tumor regression, but such doses are likely to adversely affect the vasculature of normal tissues, including the cardiovascular, endocrine, and nervous systems (12). Indeed, antiangiogenic therapy with agents such as bevacizumab is associated with an increased risk of arterial thromboembolic events (19), and such adverse effects could be more pronounced with increased doses. Furthermore, excessive vascular regression may be counter-productive because it compromises the delivery of drugs and oxygen (Fig. 1). Indeed, suboptimal doses or scheduling of antiangiogenic agents might lower tumor oxygenation and drug delivery and, thus, antagonize rather than augment the response to radiotherapy or chemotherapy (8–10). This need for a delicate balance between normalization and excessive vascular regression emphasizes the requirement for careful selection of the dose and administration schedule for antiangiogenic agents.

Can Blocking VEGF Signaling Normalize Tumor Vessels?

Of all the known angiogenic molecules, VEGF (also referred to as VEGF-A) appears the most critical (12, 20, 21). VEGF promotes the survival and proliferation of endothelial cells, increases the display of adhesion molecules on these cells, and increases vascular permeability. During mouse embryonic development, the exquisite regulation of VEGF expression sets in motion a chain of events that leads to the development of a mature vasculature from primordial cells (18). Deletion of a single allele of VEGF results in embryonic lethality. So, too, does overexpression of VEGF. In adults, ectopic overexpression of VEGF results in a highly abnormal vasculature (22). Collectively, these results indicate that the normal vasculature requires precise spatial and temporal control of VEGF levels.

VEGF is overexpressed in the majority of solid tumors. Thus, if one were to judiciously down-regulate VEGF signaling in tumors, then the vasculature might revert back to a more “normal” state. Indeed, blockade of VEGF signaling passively prunes the immature and leaky vessels of transplanted tumors in mice and actively remodels the remaining vasculature so that it more closely resembles the normal vasculature (Fig. 1). This “normalized” vasculature is characterized by less leaky, less dilated, and less tortuous vessels with a more normal basement membrane and greater coverage by pericytes (Fig. 1C). These morphological changes are accompanied by functional changes—decreased interstitial fluid pressure, increased tumor oxygenation, and improved penetration of drugs in these tumors (Table 1) (6, 16, 23–30).

What About Human Tumors?

Thousands of patients worldwide have received anti-VEGF therapy. The effect of VEGF blockade on human tumors was recently studied in rectal carcinoma patients receiving an antibody to VEGF, bevacizumab, together with radiation and chemotherapy (31). The results in patients mirrored those seen in transplanted tumors in mice: Two weeks after a single injection of bevacizumab alone, the global (mean) blood flow of tumors, as measured by contrast-enhanced computed tomography (CT), decreased by 30 to 50% in six consecutive patients. Tumor microvascular density, vascular volume, and interstitial fluid pressure were also found to be reduced. Surprisingly, however, there was no concurrent decrease in the uptake of radioactive tracers in tumors, which suggests that vessels in the residual “normalized” tumor vasculature were more efficient in delivering these agents to tumor parenchyma than they were prior to bevacizumab treatment.

Similar reductions in blood flow, as measured by magnetic resonance imaging (MRI), had been noted previously in patients treated daily with small-molecule inhibitors of VEGFR tyrosine kinase activity (PTK787 and SU6668) (32, 33). Interestingly, however, positron emission tomography (PET) analysis of patients treated with endostatin, an endogenous inhibitor of angiogenesis, revealed a biphasic response—that is, an increase in tumor blood flow at lower doses and a decrease at intermediate doses (34).

As we interpret these human data obtained from MRI, PET, and CT, two key limitations of these imaging methods must be kept in mind. First, because most of these methods yield a parameter that depends on both blood flow and permeability, the blood flow and permeability cannot be calculated unambiguously. Second, tumor blood flow is highly heterogeneous. It is not the total blood flow, but the distribution of blood flow, that determines the distribution of a drug or oxygen in tumors. Therefore, the global (total) blood flow, as estimated by the currently available resolution of MRI, CT, or PET, does not inform us about the degree of spatial heterogeneity in vascular normalization or drug distribution (17). Thus, improved imaging techniques, which can measure the spatial and temporal changes in blood flow and other physiological parameters with higher resolution, are needed to definitively establish the effects of antiangiogenic treatment on vascular function in human tumors growing at different sites.

Is There an Optimal Time or Drug Dose for Normalization?

Optimal scheduling of antiangiogenic therapy with chemotherapy and/or radiation therapy requires knowledge of the time window during which the vessels initially become normalized, as well as an understanding of how long they remain in that state. Recent studies, in which human tumors growing in mice were treated with an antibody to VEGF receptor-2, have identified such a “normalization window,” that is, a period during which the addition of radiation therapy yields the best therapeutic outcome (Fig. 2) (29). This window was short-lived (about 6 days) and was characterized by an increase in tumor oxygenation, which enhances radiation therapy by increasing the concentration of reactive oxygen species created by the radiation. During the normalization window, but not before or after it, VEGFR2 blockade was found to increase pericyte coverage of vessels in a human brain tumor grown in mice. Vessel normalization was accompanied by upregulation of Angiopoietin 1 and activation of matrix metalloproteinases (MMPs). The prevailing hypothesis is that VEGF blockade passively prunes nascent vessels that are not covered with pericytes. In contrast, this study found that pericyte coverage increased prior to vascular pruning (29). Improved understanding of the molecular mechanisms of vessel normalization may suggest new strategies for extending the normalization window to provide ample time for cytotoxic therapy.

• Download high-res image
• Open in new tab
• Download Powerpoint

The dose of antiangiogenic agents also determines the efficacy of combination therapy. Although it is tempting to increase the dose of antiangiogenic agents or to use a more potent angiogenic blocker, as one would for chemotherapeutic agents, doing so might lead to normal-tissue toxicity and compromise the tumor vessels to the point that drug delivery is impaired. Indeed, renal cell carcinoma patients on a high dose of bevacizumab (10 mg per kg of body weight every 2 weeks) were more likely to develop hypertension and proteinuria than those on a low dose, although the sample size was too small for comparison of the rates of serious adverse events (2). Even the low dose of bevacizumab (5 mg/kg) given in combination with chemotherapy has contributed to an increased risk of cardiovascular problems, including death, in some cancer patients (19). Although no dose comparison has yet been made in large clinical trials, it is conceivable that such serious adverse events may increase with higher doses. In studies of mice, more potent blockers of VEGF signaling have induced regression of normal tracheal and thyroid vessels (35).

Do We Need an Antiangiogenic Cocktail for Normalization?

The constellation of angiogenic molecules expressed in a tumor increases with malignant progression, rendering certain tumors less dependent on VEGF. For example, early stages of breast tumors may require only VEGF for angiogenesis, whereas at later stages, angiogenesis in these tumors may be driven by additional factors, including fibroblast growth factor 1 (FGF-1), FGF-2, transforming growth factor–β (TGF-β), platelet-derived endothelial cell growth factor (PD-ECGF), and placental growth factor (PlGF) (36). Thus, a late-stage breast tumor may escape anti-VEGF treatment by exploiting alternative angiogenic factors to generate its neovasculature. This may help explain why the addition of bevacizumab to chemotherapy did not prolong the survival of breast cancer patients in a recent phase III trial (4). Optimal cancer treatment may require the targeting of multiple angiogenic pathways, and the challenge for the oncologist will be to formulate cocktails of antiangiogenic agents specifically tailored to the angiogenic profile of individual tumors.

Small-molecule inhibitors that target multiple kinases involved in tumor angiogenesis (e.g., VEGFR2, PDFGRβ) are currently in clinical trials (7). Whether these kinase inhibitors cover the necessary spectrum of angiogenic pathways to normalize the tumor vasculature is not known. If they do not, it will be important to conduct clinical trials that combine antiangiogenic agents produced by competing pharmaceutical companies before each agent is approved by the Food and Drug Administration (FDA).

An alternative approach is to develop agents that mimic such antiangiogenic cocktails by targeting upstream pathways that regulate the production of angiogenic molecules. Such drugs already exist and have been approved by the FDA. For example, one recent study has shown that an antibody to HER2, trastuzumab (also known as Herceptin), acts on multiple angiogenic pathways in HER2-overexpressing human breast cancer xenografts (37). Herceptin lowers the expression of several proangiogenic molecules while increasing expression of the antiangiogenic molecule thrombospondin-1. Interestingly, although Herceptin lowered the expression of VEGF in tumor cells, the host cells within the tumor stroma produced compensatory VEGF; thus, additional anti-VEGF treatment could improve the efficacy of Herceptin. On the basis of these preclinical results, a clinical trial has been initiated in which Herceptin is combined with bevacizumab for treatment of HER2-positive breast cancer (38).

A further finding from this study was that Herceptin “normalized” the vasculature of human breast cancer xenografts (37). Whereas vessels in the control antibody-treated tumors were dilated and leaky, those in the Herceptin-treated tumors had diameters and vascular permeability closer to those of normal vessels. Thus, Herceptin and other drugs that target upstream mutant receptors [e.g., cetuximab, an antibody to HER1 (39), or kinase inhibitors such as gefitinib or imatinib (40)] act as mimics of antiangiogenic cocktails, that is, these drugs improve their own delivery as well as that of other therapeutics given in combination. This improvement in delivery and alleviation of hypoxia presumably contributes to their efficacy.

These findings point to an urgent need to investigate whether other molecularly targeted agents also mimic antiangiogenic cocktails. This concept may not only change our view of how these therapeutics actually work in vivo but will also provide us with the knowledge of how to use them for improving the delivery of other therapeutic agents given in combination. Furthermore, once we discover the set of angiogenic pathways affected by each drug, we can develop algorithms to combine them for maximal efficacy in a manner tailored to the angiogenic profile of each tumor. Further support for the use of these agents as antiangiogenic cocktails comes from emerging evidence that most oncogene and tumor-suppressor gene pathways are implicated in angiogenesis, either directly or indirectly (41–43).

Is Tumor Growth Accelerated During Vascular Normalization?

One would expect that the improved delivery of oxygen and nutrients during vascular normalization would enhance tumor growth. However, both preclinical and clinical studies to date show that, despite normalization, tumor growth is not accelerated during antiangiogenic monotherapy. There are several possible explanations for this apparent paradox.

1) It is important to remember that vascular normalization occurs in the context of antiangiogenic treatment and that the main effect of this treatment is a reduction in the number of blood vessels (vessel density), which should lead to tumor regression. Moreover, tumors are highly heterogeneous; not all regions are equally vascularized, some tumor vessels are more mature than others, and the balance of pro- and antiangiogenic molecules differs from region to region and from one moment to the next. Hence, it may be that the effects of vessel normalization in some regions of the tumor are swamped by simultaneous vessel regression in other regions. In addition, the inability of tumors to grow new vessels during antiangiogenic therapy limits the ability of this transient increase in vascular efficiency to expand the tumor mass. If it were easy to achieve complete tumor regression with antiangiogenic monotherapy, vascular normalization would be of marginal importance, because it is expected to affectonlyasubsetofcellsandtodosoonlytemporarily. Unfortunately, some tumor cells are able to survive antiangiogenic monotherapy, and these cells must be targeted with combined therapy.

2) The transient normalization of tumor vessels produces a temporary increase in oxygen and nutrient delivery to the cancer cells that surround these “normalized” vessels. This might be expected to enhance the proliferation of these cells and hence to accelerate tumor growth. However, this intuitive notion is not supported by published data. For example, Gullino found no correlation between tumor growth rate in vivo and blood flow rate, vascular volume, or use of oxygen or glucose [reviewed in (44)]. Even if the proliferation rate of cancer cells around normalized vessels was increased, this may well enhance therapeutic efficacy, because rapidly proliferating cells are more sensitive to radiation and to many cytotoxic drugs.

3) It is widely assumed that hypoxia leads to the death of cells. Therefore, alleviation of hypoxia during transient normalization of tumor vasculature should accelerate tumor growth. However, a growing body of evidence indicates that hypoxia may in fact promote cancer progression (45, 46). These two competing effects of antiangiogenic therapy may cancel each other.

4) Finally, in some tumors, cancer cells depend on the same angiogenic growth factors (e.g., VEGF) for their survival as do the endothelial cells. In these tumors, antiangiogenic agents may kill both cancer cells and endothelial cells and will likely induce tumor regression—similar to hormone withdrawal from a hormone-dependent tumor (47)—despite vessel normalization.

For all these reasons, any acceleration in tumor growth during transient normalization is presumably masked by indirect and direct killing of cancer cells by antiangiogenic agents. Thus, it is not surprising that tumor regression is slow and/or modest after antiangiogenic monotherapy despite a significant decrease in microvascular density (31).

Perspective

The approval of the first antiangiogenic agent for clinical use in patients with colorectal carcinoma has taught us many lessons, the most important of which is that these agents must be used in combination with agents that target cancer cells to have an appreciable impact on patient survival. Increasing the dose of antiangiogenic agent may harm normal tissues and destroy too much of the tumor vasculature, leading to hypoxia and poor drug delivery in the tumor and to toxicity in normal tissues. However, optimal doses and schedules of these reagents tailored to the angiogenic profile of tumors can normalize tumor vasculature and microenvironment without harming normal tissue.

At least three major challenges must be met before therapies based on this vascular normalization model can be successfully translated to the clinic. The first challenge is to determine which other direct or indirect antiangiogenic therapies lead to vascular normalization. In principle, any therapy that restores the balance between pro- and antiangiogenic molecules should induce normalization. Indeed, withdrawing hormones from a hormone-dependent tumor lowers VEGF levels and leads to vascular normalization (47). Recently, metronomic therapy—a drug delivery method in which low doses of chemotherapeutic agents are given at frequent intervals—has also been shown to increase the expression of thrombospondin-1, which is a potent endogenous angiogenesis inhibitor (48). Conceivably, this therapy might also induce normalization and improve oxygenation and drug penetration into tumors. Whether various synthetic kinase inhibitors (e.g., Novartis PTK787, Bayer 43-9006, Pfizer SU11248, and AstraZeneca AZD2171), endogenous inhibitors (e.g., angiostatin, endostatin, and tumstatin), antivasocrine agents [i.e., razoxane (49)], conventional chemotherapeutic agents [e.g., taxol (50)], and vascular targeting agents (51–54) do the same remains to be seen. Some of these agents may be effective because they target both stromal and cancer cells. To date, most clinical trials are designed primarily to measure changes in the size of the tumor and may therefore not shed light on changes in the vascular biology of tumors. Clinical studies, such as the rectal carcinoma study described earlier (31), and other ongoing translational clinical trials should help bridge the gaps in this aspect of our knowledge.

The second challenge is to identify suitable surrogate markers of changes in the structure and function of the tumor vasculature and to develop imaging technology that will help to identify the timing of the normalization window during antiangiogenesis therapy. Measurement of blood-vessel density requires tissue biopsy and provides little information on vessel function. Although imaging techniques are expensive and far from optimal, they can provide serial measures of vascular permeability, vascular volume, blood perfusion, and uptake of some drugs and can therefore be used to monitor the window of normalization in patients. The number of circulating mature endothelial cells and their less differentiated progenitors does decrease after VEGF blockade, both in animals and in patients, but whether this decline coincides with the normalization window is not known (31). During the course of therapy, serial blood measurements of molecules involved in vessel maturation have the potential to identify surrogate markers. PET with 18-fluoromisonidzole and MRI can provide some indication of tumor oxygenation (55, 56) and might be useful for tracking the normalization window. Finally, the measurement of the interstitial fluid pressure is minimally invasive, inexpensive, and easy to implement for anatomically accessible tumors. Hence, this parameter could be used in the interim as a useful indicator of vessel function until novel noninvasive methods are developed.

The third challenge is to fill gaps in our understanding of the molecular and cellular mechanisms of the vascular normalization process (29). With rapid advances in genomic and proteomic technology and access to tumor tissues during the course of therapy, we can begin to monitor tumor response to antiangiogenic therapies at the molecular level.

Addressing all of these challenges not only will benefit patients with cancer but also may benefit patients with other diseases (18, 57). For example, a low dose of anti-VEGF aptamer was recently shown to improve the vision of patients with macular degeneration, whereas a higher dose was ineffective (58). These principles may also be useful for regenerative medicine and tissue engineering, in which the goal is to create and maintain a functionally normal vasculature (59).