Tumorigenesis and the angiogenic switch

REVIEWS TUMORIGENESIS AND THE ANGIOGENIC SWITCH Gabriele Bergers* and Laura E. Benjamin‡ It has become evident that we cannot understand tumour growth without considering components of the stromal microenvironment, such as the vasculature. At the same time, the tumour phenotype determines the nature of the tumour vasculature. Much research is now devoted to determining the impact of angiogenesis on tumour development and progression, and the reciprocal influences of tumour products on the microvasculature. A more detailed understanding of the complex parameters that govern the interactions between the tumour and vascular compartments will help to improve anti-angiogenic strategies — not only for cancer treatment, but also for preventing recurrence. ANGIOGENESIS

*University of California San Francisco, Department of Neurological Surgery, Brain Tumor Research Center and Comprehensive Cancer Center, HSE 722, 513 Parnassus Avenue, San Francisco, California 94143-0520, USA. ‡ Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, Massachusetts 02215, USA. Correspondence to G.B. e-mail: [email protected] doi:10.1038/nrc1093

It is easy to forget that cancer is not a single disease, but many diseases. In the past 50 years, the complexity of cancer has been rendered more tangible by a large body of work that has identified common principles of pathogenesis. Recently, several models of multistep tumorigenesis, via discrete genetic changes, have been proposed to explain the transformation of a normal cell into a cancer cell. These include oncogene activation, loss of telomerase and induction of aneuploidy as critical initiating events1–3. Some of the debate in the tumour biology field over alternate models might be due to tumour heterogeneity — whether the tumour initiates in an epithelial cell or in another cell type — or to the model organism used for experimentation. For carcinomas, the stepwise progression towards malignancy has been classically defined in colon cancer4. In this and many other models, accumulated spontaneous genetic changes occur, and cell populations that have acquired growth and invasion advantages are expanded. However, in addition to the genetic and epigenetic changes that occur during transformation, another discrete step is required to allow tumour propagation and progression — the induction of a tumour vasculature, termed the ‘angiogenic switch’5,6. Like normal tissues, tumours require an adequate supply of oxygen, metabolites and an effective way to remove waste

products7. These requirements vary, however, among tumour types, and change over the course of tumour progression8. But gaining access to the host vascular system and the generation of a tumour blood supply are rate-limiting steps in tumour progression. A classical model of the regulation of tumour angiogenesis that has held up nicely in practice is illustrated by a scale laden with anti-angiogenic molecules on one side and pro-angiogenic molecules on the other (FIG. 1). Induction of the angiogenic switch depends on how heavily that balance tips in favour of pro-angiogenesis. Pro-angiogenic gene expression is increased by physiological stimuli, such as hypoxia, which results from increased tissue mass9, and also by oncogene activation or tumour-suppressor mutation (recently reviewed in REF. 10). The angiogenic switch can occur at different stages of the tumour-progression pathway, depending on the tumour type and the environment. It has been shown that dormant lesions, and, in some instances, premalignant lesions, also initiate neovascularization, which allows them to progress6,11. The fact that tumours are dependent on blood supply has inspired many researchers to search for anti-angiogenic molecules, and to design antiangiogenic strategies for cancer treatment. But what is the importance of the angiogenic switch in the tumour-progression pathway, and how is it regulated?

NATURE REVIEWS | C ANCER

VOLUME 3 | JUNE 2003 | 4 0 1

© 2002 Nature Publishing Group

REVIEWS

PERIVASCULAR CELLS

Mural cells that surround the endothelium. These cells are related to the vascular smoothmuscle lineage and share many markers with these cells. ENDOTHELIAL PRECURSOR CELLS

Undifferentiated cells in the adult bone marrow that can travel through the blood to sites of ongoing angiogenesis, and differentiate into mature endothelial cells. They have recently been implicated in tumour angiogenesis. These cells are identified by co-expression of haematopoietic stem-cell markers (CD34 and AC133), and vascular endothelial-cell markers (VEGFR2 and TIE2).

Summary • Tumour heterogeneity leads to heterogeneity in the tumour vasculature. Just as there are multiple phenotypes for any given tumour type, so can there be multiple phenotypes of the tumour angiogenic process. • The onset of angiogenesis, or the ‘angiogenic switch’, is a discrete step that can occur at any stage of tumour progression. It depends on the type of tumour and its microenvironment. • Tumour angiogenesis differs significantly from physiological angiogenesis. Differences include aberrant vascular structure, altered endothelial-cell–pericyte interactions, abnormal blood flow, increased permeability and delayed maturation. • The abnormal features of the tumour vasculature are believed to result from the disproportionate expression of angiogenic cytokines and inhibitors. Expression of these varies from tumour to tumour. • Tumour hypoxia complicates the angiogenic response, depending on the status of p53, which can regulate key angiogenic cytokines and inhibitors. • The angiogenic activity of a tumour does not necessarily correlate with tumour aggressiveness. Nonetheless, it can be a prognostic factor for certain tumour types. • Anti-angiogenic agents can be used not only for the treatment of cancer, but also for the prevention of cancer recurrence or metastasis. Given the heterogeneity of tumour and blood-vessel growth, a multidrug approach that targets various factors might be more successful than monotherapy in restraining cancer growth.

Tumour versus physiological angiogenesis

During embryonic vasculogenesis, blood vessels are formed de novo, from endothelial-cell precursors (angioblasts) that assemble into a primary capillary plexus. This primitive network then differentiates, and new blood vessels sprout and branch from pre-existing capillaries — the process of angiogenesis12. The vasculature is usually quiescent in the adult, and endothelial cells are among the longest-lived cells outside the nervous system. In fact, physiological endothelial-cell turnover is reportedly measured in

Inhibitors: Thrombospondin-1 The statins: Angiostatin Endostatin Canstatin Tumstatin

Activators VEGFs FGFs PDGFB EGF LPA

Figure 1 | The angiogenic balance. Angiogenesis is orchestrated by a variety of activators and inhibitors — only a few of which are listed above. Activators of endothelial-cell proliferation and migration are mainly receptor tyrosine kinase ligands12, such as vascular endothelial growth factor (VEGF), fibroblast growth factors (FGFs), platelet-derived growth factor (PDGF) and epidermal growth factor (EGF), but can also be of very different origin, such as lysophosphatic acid (LPA)89. EGF upregulates VEGF, FGF and interleukin-8, whereas LPA upregulates VEGF levels. The first described angiogenic inhibitor was thrombospondin-1, which modulates endothelial-cell proliferation and motility90. Remarkably, many inhibitory molecules, such as ‘statins’, are derived from larger proteins that have no effect on angiogenesis. Among those that are listed are angiostatin36 (a fragment of plasminogen that binds ATP synthase and annexin II), as well as endostatin37, tumstatin91 and canstatin92 (fragments of collagens that bind to integrins) (see the review article by R. Kalluri on page 420 of this issue). In general, the levels of activators and inhibitors dictate whether an endothelial cell will be in a quiescent or an angiogenic state. It is believed that changes in the angiogenic balance mediate the angiogenic switch.

402

years in tissues that do not require ongoing angiogenesis. The few adult tissues that do require ongoing angiogenesis include female reproductive organs, organs that are undergoing physiological growth11 or injured tissue. The point at which these ‘normal’ processes differ from pathological angiogenesis is in the tightly regulated balance of pro- and anti-angiogenic signals. During normal physiological angiogenesis, new vessels rapidly mature and become stable. By contrast, tumours — described as “wounds that never heal”13 — have lost the appropriate balances between positive and negative controls. One characteristic feature of tumour blood vessels is that they fail to become quiescent, enabling the constant growth of new tumour blood vessels. Consequently, the tumour vasculature develops unique characteristics and becomes quite distinct from the normal blood supply system. Tumour blood vessels are architecturally different from their normal counterparts — they are irregularly shaped, dilated, tortuous and can have dead ends. They are not organized into definitive venules, arterioles and capillaries like their normal counterparts, but rather share chaotic features of all of them. The vascular network that forms in tumours is often leaky and haemorrhagic, partly due to the overproduction of vascular endothelial growth factor (VEGF; also known as vascular permeability factor, VPF). PERIVASCULAR CELLS, which are usually in close contact with the endothelium, often become more loosely associated or less abundant14,15. Tumour vessels have also been reported to have cancer cells integrated into the vessel wall16,17, and some tumours rely heavily on vasculogenesis, recruiting ENDOTHELIAL PRECURSOR CELLS from the bone marrow18. Blood flows irregularly in tumour vessels, moving more slowly and sometimes even oscillating. This leads to dysfunctional capillaries. Tumours can be quite heterogeneous in their vascular patterns, and are able to overproduce their capillary networks. In normal tissues, by contrast, vessel density is dynamically controlled by

| JUNE 2003 | VOLUME 3

www.nature.com/reviews/cancer


REVIEWS

a Dormant

b Perivascular detachment

c Onset of angiogenic sprouting

and vessel dilation

d Continuous sprouting;

e Tumour vasculature

new vessel formation and maturation; recruitment of perivascular cells

Normal cell

Blood vessel with pericyte

Cancer cell

Apoptosing, necrotic cell

Dividing cell

Figure 2 | The classical angiogenic switch. The angiogenic switch is a discrete step in tumour development that can occur at different stages in the tumour-progression pathway, depending on the nature of the tumour and its microenvironment. Most tumours start growing as avascular nodules (dormant) (a) until they reach a steady-state level of proliferating and apoptosing cells. The initiation of angiogenesis, or the ‘angiogenic switch’, has to occur to ensure exponential tumour growth. The switch begins with perivascular detachment and vessel dilation (b), followed by angiogenic sprouting (c), new vessel formation and maturation, and the recruitment of perivascular cells (d). Blood-vessel formation will continue as long as the tumour grows, and the blood vessels specifically feed hypoxic and necrotic areas of the tumour to provide it with essential nutrients and oxygen (e).

the metabolic needs of nutrients and oxygen. So, the structural and functional abnormalities in tumour vessels reflect the pathological nature of their induction. Although we do not fully understand the molecular controls of all of these abnormalities, we can surmise that they are the result of the imbalanced expression and function of angiogenic factors. PERICYTES

How are tumour blood vessels formed?

Cells that are related to vascular smooth muscle. These cells are adjacent to and surrounding the endothelium, share a common basement membrane with the endothelium, and have gapjunction connections with the endothelial cells. Whether these cells are multipotent, with the ability to differentiate into either vascular smooth muscle or even endothelium, is still unclear.

The vasculogenic process that is used by the early embryo has been adapted for use in adults, under certain situations, and by tumours. In this setting, endothelial precursor cells can be mobilized from the bone marrow and transported through the bloodstream to become incorporated into the walls of growing blood vessels19. Factors that stimulate angiogenesis, such as VEGFA, placental growth factor (PlGF) and angiopoietin-1 (ANG1), have been shown to stimulate this process20,21.

The percentage of endothelial-precursor-cell incorporation is generally low and depends on the nature of the tumour, supporting the concept that most tumour neovascularization seems to occur via angiogenesis. However, in some model systems, tumours are reliant on endothelial-precursor-cell mobilization18 (FIG. 2). High levels of VEGFA expression alone are capable of initiating angiogenesis in a quiescent vasculature, and the steps and morphological changes that follow from the early formation of ‘mother vessels’ through maturation have been characterized22. The earliest stages are defined by vasodilation and an increased vascular permeability of pre-existing capillaries, or post-capillary venules, in response to VEGF. This allows extravasation of plasma proteins, which lay down a provisional matrix on which activated endothelial cells migrate. This is accompanied by the loosening of PERICYTE covering, which is thought to be a function of ANG2 (REF. 23) — a ligand of TIE2. TIE2 is a tyrosine kinase receptor that is selectively expressed in endothelial cells24.


VOLUME 3 | JUNE 2003 | 4 0 3


REVIEWS

a

b

c

d

e ?

Figure 3 | New blood-vessel formation. Blood vessels arise from pre-existing capillaries or postcapillary venules in tumours (a). b | First, pericytes (green) detach and blood vessels dilate before the basement membrane and extracellular matrix is degraded. c | This allows endothelial cells (red) to migrate into the perivascular space towards angiogenic stimuli produced by the tumour cells or host cells. d | Endothelial cells proliferate, loosely following each other, and are presumably guided by pericytes. e | Behind the migration columns, endothelial cells adhere to each other and create a lumen, which is accompanied by basement-membrane formation and pericyte attachment. Finally, blood-vessel sprouts will fuse with other sprouts to build new circulatory systems. Little is known about this fusion mechanism.

MURAL CELLS

Cells of the blood-vessel wall.

404

The vascular basement membrane and extracellular matrix are locally degraded to allow underlying endothelial cells to migrate into the perivascular space towards chemotactic angiogenic stimuli. The endothelial cells then multiply, loosely following each other into the perivascular space and forming a migration column. There is a debate as to whether these sprouts are guided by immature MURAL CELLS in tumours15, similar to the guidance provided by other mesenchymal cells during normal development (such as astrocytes in the developing retina25). During normal angiogenesis, migration columns lead to a differentiation zone, where endothelial cells change shape and adhere to each other to form a lumen. Continued proliferation within the vascular wall allows for enlargement of the blood-vessel diameter. Perivascular cells are attracted, and a vascular basal lamina is produced around the newly formed blood vessels. In normal vasculature, pericyte associations reduce endothelial-cell proliferation and decrease their dependence on host-tissue production of VEGFA26,27. Conversely, the ‘improper’ or decreased vessel association with pericytes in tumours might partially explain the abnormal vessel diameters and sensitivity to VEGFA inhibition14,28. Inhibition or loss of VEGF activity induces apoptosis of endothelial cells due to its function as a survival factor. Recent literature indicates that other angiogenic cytokines, such as ANG1 (REF. 29) or PlGF30, can also provide survival signals, and could be equally capable of rescuing immature blood vessels from VEGFA loss. This redundancy in function will be an important consideration in planning VEGF-specific anti-angiogenic therapies to induce regression of tumour blood vessels.Very little is

known about the final step in the angiogenic process, in which vascular sprouts fuse with other sprouts to form loops and circulate the blood into the newly vascularized area (FIG. 2e). Tumours express various pro-angiogenic factors (FIG. 2). Although FGF (fibroblast growth factor) was the first to be discovered, VEGFA is perhaps the most ubiquitous. As well as VEGFA, other VEGF-family members are produced by human tumours, including VEGFC and VEGFD, which both bind to VEGF receptor-2 (VEGFR2) and activate many of the same pathways as VEGFA31. Another VEGF-family member, PlGF, has a unique function in cancer and other pathological settings, although it is not essential for normal development32. Notably, PlGF does not bind VEGFR2, but rather VEGFR1. ANG molecules are also involved in tumour angiogenesis24, although they seem to function after the induction of angiogenesis, vessel assembly or mural-cell regulation. One reason that VEGFA expression is so widespread might be due to its induction under ischaemic conditions. Ischaemia occurs during organ growth, when the increased tissue mass outgrows the ability of the existing vasculature to provide sufficient nutrients. Tumours contain elevated VEGF levels at the rim of necrotic and hypoxic tissue. This is also where new blood-vessel sprouting occurs. In addition to sensing glucose and other nutrients provided by the circulation, most cells are sensitive to hypoxia. Oxygen and nutrient requirements vary among tumours, depending on their stage and the microenvironment they are growing in. Commonly, oxygen consumption is reduced in tumour cells, in comparison to the normal cells. Tumours are able to adapt their metabolism to survive under conditions of reduced oxygen availability by increasing glycolysis to maintain ATP production33. Nevertheless, every tissue — including tumorigenic tissue — is dependent on adequate oxygen delivery. If oxygen is not sufficiently supplied, a battery of genes are induced by hypoxiainducible transcription factors (HIFs) that allow them to overcome hypoxic conditions34. Angiogenesis and tumour progression?

The concept that angiogenesis is required for the expansion of a tumour mass raises the question as to when during tumour progression neovascularization occurs — is it simply a necessity to overcome size limitations, or is it coupled to a specific stage in a tumour’s life cycle? There is increasing evidence from a variety of human tumours and mouse models of multistage cancers that angiogenesis can be switched on at different stages of tumour progression, depending on the nature of the tumour and the microenvironment. Historically, tumour-associated angiogenesis was thought to go through two phases, which are separated by the ‘angiogenic switch’ (FIG. 3). The first is defined as an avascular phase, which corresponds to small and occult lesions of not more than 1–2 mm in diameter. These lesions stay dormant by reaching a steady state between proliferation and apoptosis. Dormant tumours have been discovered during autopsies of




REVIEWS individuals who died of causes other than cancer35. This supports the notion that only a very small subset of dormant tumours enter the second phase — the vascular phase — in which exponential tumour growth ensues. This same principle applies to tumour metastasis. Dormant metastases are a clinically difficult problem, as they can sometimes become activated by removal of the primary tumour10. Given the requirement of angiogenesis for growth and progression of dormant lesions, it would be assumed that inhibition of the angiogenic switch could prevent progression of tumours and their metastases. Preventative anti-angiogenic strategies could be especially useful in patients who are at high risk for developing cancer or recurrence of cancer, or development of metastases. A few experimental studies in animals, as well as in clinical trials, have already shown promising results. For example, angiostatin and endostatin reduced the formation of metastases in the Lewis lung cancer mouse model36,37. Linomide, which inhibits angiogenesis induction by various pro-angiogenic factors (such as VEGF, FGF and tumour necrosis factor-α (TNF-α)), was able to prevent the development of N-methylnitrosurea-induced prostate and mammary carcinomas in rats by 50% (REF. 38). Celecoxib, an inhibitor of the inducible prostaglandin G/H synthase COX2, was able to reduce the number of colorectal polyps in patients with familial adenomatous polyposis (FAP). Patients with FAP are at high risk of developing colorectal neoplasias39. COX2 inhibitors are anti-inflammatory agents with anti-angiogenic activities that downregulate the expression of the pro-angiogenic factors FGF and VEGF40. These drugs have also been considered for the prevention of prostate and cervical adenocarcinomas, which both express high levels of COX2 — clinical trials are already underway 41–43. The classical characterization of tumour angiogenesis has been demonstrated in various oncogene-driven mouse tumour models. These include RipTag mice44,45, which develop pancreatic-islet carcinomas, and K14HPV16 mice, which develop squamous-cell carcinomas of the skin and cervix46–48. In these mice, neovascularization begins early in the dysplastic stages and is a prerequisite for tumour formation. Tumours develop through temporally and histologically distinct stages — normal cells progress to hyperplasia, angiogenic dysplasia and, eventually, to highly vascularized invasive tumours that are similar to many human cancers. It is important to note that neither the oncogene SV40 Tag, which is used in the creation of the RipTag mice, nor the oncogenes E6 and E7, which are used in the creation of the K14-HPV16 mice, were sufficient to induce angiogenesis. Additional changes that occurred along the progression pathway enabled neovascularization. SV40 Tag, E6 and E7 bind and inactivate the tumour-suppressor gene products p53 and RB. In the pancreatic-islet cancer model, VEGF was shown to be required for the angiogenic switch to occur45,49. Although VEGF and its receptors are constitutively expressed during the tumour-progression pathway, inhibition of VEGF activity blocked the angiogenic switch and restrained

tumour growth. The matrix metalloproteinase (MMP)-9 was also found to be a component of the angiogenic switch. By releasing sequestered VEGF, this proteinase makes VEGF available for interaction with its receptors45. Similarly, mammary-tumour-prone mice (MMTVneu) that lacked the endogenous angiogenesis inhibitor thrombospondin-1 (TSP1) developed larger, more-vascularized tumours than the parental strain due to higher levels of MMP9 production. This was shown to increase the association of VEGF with its receptor VEGFR2 (REF. 50). TSP1 overexpression in MMTV-neu mice, conversely, resulted in delayed tumour growth and even hindered tumour development by about 20%, supporting the hypothesis that angiogenesis can be a prerequisite for tumour formation50. Similar results were observed in studies of patients with carcinogenesis of breast, cervix and prostate, whose premalignant carcinoma in situ lesions become vasoproliferative11. In breast cancer, overall vascular density correlates to poor cancer prognosis51, and TSP1 expression is inversely correlated with malignant progression of mammary and lung carcinomas, as well as melanoma52,53. In contrast to the multistep model described above, in which the angiogenic switch is distinct and a prerequisite to tumour development and progression, expression of deregulated MYC, in conjunction with anti-apoptotic factors such as BCL2 or BCL-XL, resulted in the formation of angiogenic and invasive pancreatic-islet tumours that did not require additional mutations to progress through premalignant stages54. Whether this model represents an accelerated version of normal cancer progression, or simply a bypass of the premalignant stages, is still unclear. Human tumours that bypass premalignant stages have not been described. The importance of the angiogenic switch was also recently demonstrated by Weinberg and colleagues55. They showed that expression of the SV40 early region, TERT and activated RAS were sufficient to transform primary epithelial cells in vitro. However, the ability to grow in vivo depended on the level of HRAS expression — cells that expressed low levels of RAS were dormant and non-angiogenic, whereas cells that expressed high levels of RAS developed into full-blown tumours. What is the explanation for this difference? Whereas VEGFA levels increased only modestly (1.4-fold) in tumours that expressed high levels of RAS, the TSP1 levels increased markedly (8-fold) in these cells. The authors were able to induce tumour formation from cells that expressed low levels of RAS by simply overexpressing VEGFA. This confirms that the block in tumour progression was due to a failure to activate the angiogenic switch. It also provides further support for the model of an angiogenic switch that is determined by the ultimate balance in pro- and anti-angiogenic factors, and that oncogene expression can influence this balance. Angiogenesis differs among tumour types

Despite the fact that angiogenesis begins at different stages in all of these examples, neovascularization is usually a prerequisite for tumour progression. There are, however, some examples of tumours that do not


VOLUME 3 | JUNE 2003 | 4 0 5


REVIEWS

a Tumour cells grow along blood vessels

b Increased tumour growth leads

c Angiogenic sprouting is initiated

to hypoxia and necrosis

Figure 4 | Blood vessel co-option precedes angiogenesis in astrocytoma progression. Astrocytomas first acquire their blood supply by co-opting existing normal brain blood vessels without the necessity to initiate angiogenesis. They instead grow along blood vessels, without a tumour capsule, eliciting an invasive character (a) . When grade III astrocytomas progress into glioblastomas (GBM or grade IV astrocytoma), they become hypoxic and necrotic — partly due to vessel regression and increased tumour-cell proliferation (b). These conditions, in turn, induce formation of new blood vessels (angiogenic sprouting) (c) that supply the tumour with the necessary metabolites. In fact, glioblastomas are partly defined by the appearance of proliferating endothelial cells and a high blood-vessel density that distinguishes grade IV tumours from the lower-grade astrocytomas.

require neovascularization. For example, astrocytomas first acquire their blood supply by co-opting existing normal brain blood vessels without initiating angiogenesis, making them a ‘non-angiogenic tumour’ (FIG. 4). They instead grow along blood vessels, without a tumour capsule, eliciting an invasive character, and can enlarge as much as some angiogenic tumours56. However, when grade III astrocytomas progress to grade IV astrocytomas/glioblastomas (glioblastoma multiforme, or GBM), they become hypoxic and necrotic — partly due to vessel regression caused by ANG2 upregulation57 and increased tumour-cell proliferation. Although it seems to be a paradox that blood vessels first regress as tumours progress, the hypoxic conditions eventually induce the formation of new blood vessels. The most prominent hypoxia-induced angiogenic factors are VEGF and VEGFR2, which, in conjunction with ANG2, promote vascular remodelling and sprouting23,56. In fact, glioblastomas are partly defined by the appearance of proliferating endothelial cells and a high blood-vessel density, which distinguishes grade IV tumours from the lower-grade astrocytomas. Little is known about the genetic or epigenetic changes that occur during GBM progression that cause the reduction in vessel density. Much research has been done, however, to describe the process of hypoxiainduced angiogenesis. Glial tumour cells are among the most sensitive to hypoxia, and produce very high levels of VEGFA under hypoxic conditions58. Grade I pilocytic brain tumours are highly angiogenic, showing characteristics of endothelial-cell proliferation and haemorrhages, but are defined as slow-growing cancers that neither progress nor metastasize 59. On the other hand, there are aggressive tumours, such as chondrosarcomas, that show a very low vessel density 60. An even more puzzling observation is that some tumours that originate in the lung, brain, colon or other organs have a lower vessel density than their normal parental tissue61.

406

So, how can angiogenesis be a rate-limiting step if there seems to be no correlation with tumour progression? The idea that vessel density is a definite indicator of angiogenesis dependence and aggressiveness of a tumour is a misconception that is reviewed in detail by Lynn Hlatky and Judah Folkman 8. Tumours generally require less oxygen and metabolites than the organ they are growing in, which explains the discrepancies in vessel densities between tumour and normal tissue. Also, the fact that tumours have not activated angiogenesis, or have a low vessel density, does not mean that they don’t require an adequate supply of oxygen and nutrients. The best examples, again, are astrocytomas, which are very oxygen sensitive, but the vascular-rich brain parenchyma allows the tumour cells to grow along blood vessels. A drop in vessel density, however, immediately generates tumour necrosis, which, in turn, initiates hypoxiainduced angiogenesis to restore oxygen to the tumour. As soon as these tumours are re-vascularized, they become extremely aggressive — patients who are diagnosed with glioblastomas have a median survival of only 1 year. In summary, microvessel density is only a useful prognostic marker for some cancer types. Overcoming angiogenic dependency?

In contrast to tumour cells that are, by nature, genetically unstable and heterogeneous, endothelial cells are normal diploid cells that do not acquire mutations, and therefore should not become resistant to therapy. This rationale stimulated researchers to dream of a pantumour therapy focused on blood-vessel ablation. This dream, however, seems to be naïve. Both animal and preliminary human clinical trials revealed that different tumours respond differently to anti-angiogenic therapy. Do these differences in response depend on the type or amount of angiogenic molecules that are expressed, or on other intrinsic tumour-cell characteristics, such as oncogene expression or resistance to hypoxia? Does anti-angiogenic therapy simply select for tumour cells




REVIEWS

Table 1 | p53-regulated angiogenic molecules Target gene

Target-gene function

Genes upregulated by p53 Thrombospondin-1 (TSP1)93,94

Endogenous angiogenesis inhibitor

Brain-specific angiogenesis inhibitor 1 (BAI1)95

Brain-specific endogenous angiogenesis inhibitor

Matrix metalloproteinase 2 (MMP2)96

Promotes endogenous antiangiogenic peptide formation from basement-membrane proteins

Eph receptor A2 (EphA2)97–99

EphA2 induces apoptosis; soluble EphA2 receptors are anti-angiogenic

Genes downregulated by p53 Matrix metalloproteinase 1 (MMP1)100,101 Promotes angiogenesis in bone and activates latent MMP2 Cyclooxygenase-2 (Cox2)102

Promotes angiogenesis

Vascular endothelial growth factor A (VEGFA)103

Potent pro-angiogenic and permeability factor

Hypoxia inducible factor 1 alpha (HIF-1α)77

Induces angiogenesis by activation of VEGFA transcription under hypoxic conditions

that can grow at very low oxygen levels, without initiating neoangiogenesis or that use alternative pathways to circumvent neovascularization? Initial findings indicate that these effects might indeed happen in some tumour types. Anti-VEGF therapy in a rat model of glioblastoma resulted in decreased tumour vessel density and increased apoptosis. However, treatment also led to the increased co-option of existing blood vessels, circumventing the necessity of the tumour to initiate angiogenesis62. Examples of emerging resistance to anti-angiogenesis therapy include studies that showed that glioblastomas become resistant to TSP1 therapy 63, renal-cell carcinomas become insenstitive to thalidomide64 and tumours that lose p53 expression become less responsive to combinations of low-dose cytotoxic and anti-angiogenic agents65. How does ‘resistance’ occur? In addition to the hypoxia that often naturally accompanies tumour growth, anti-angiogenesis treatments can also contribute to hypoxia — an expected result of the loss of the oxygen-transporting blood vessels66. Unfortunately, hypoxia and high levels of VEGFA production provide some protection from cytotoxic drugs by inducing expression of anti-apoptotic genes67,68 and of multidrug-resistance-associated proteins69. So, there is a complexity of opposing factors that are involved in the hypoxia response. In a normal tissue, hypoxia leads to induction of VEGF RNA stability, transcription and translation58,70,71. It also reduces the levels of the endogenous anti-angiogenesis factor TSP1 (REF. 72), and under conditions of severe hypoxia, p53-mediated angiogenesis is induced73. It is important to note that more than 50% of human cancers contain mutations in p53, and that these mutations can affect the angiogenic balance. Kerbel and colleagues recently reported that cancer cells with p53 mutations are selected for their ability to survive under hypoxic conditions65. As well as their increased resistance to apoptosis, cells with p53 defects

have been shown to downregulate the expression of TSP1 and upregulate the expression of VEGFA74. This is supported by the finding that wild-type p53 inhibits VEGF transcription75,76, and that p53 promotes MDM2-mediated ubiquitylation and proteasomal degradation of HIF-1α — the main transcriptional activator of VEGF under hypoxic conditions77. However, other reports have shown that VEGF expression is not under p53 control78 (TABLE 1). A potential explanation for these conflicting reports was recently provided by Laderoute et al. who showed that, in normal cells, hypoxia increased the VEGFA and decreased the TSP1 expression levels — regardless of p53 status. But, when these cells were transformed by E1A and RAS, TSP1 expression was downregulated, regardless of the presence of hypoxia, whereas the VEGFA response remained normal72. If the classical model of a balance between pro- and anti-angiogenic factors as the final determinant of angiogenic phenotype is considered, then an oncogene that inhibits the anti-angiogenic side of the scale would lead to a net increase in neovascularization. Additional data to support the instrumental role of p53 in regulating tumour angiogenesis comes from two independent studies in which the introduction of wild-type p53 both induced dormancy and inhibited metastasis by means of alteration in the TSP1:VEGFA ratio79,80. These examples strengthen the proposition that p53 mutations allow tumours to overcome anti-angiogenic therapy. This, however, cannot be described as ‘classical’ drug resistance, because combinations of angiogenic inhibitors should be able to overcome p53-mediated resistance. So, does anti-angiogenic therapy in p53 wild-type cells enable other ‘escape’ mechanisms, such as increased reliance on vessel co-option or dependence on alternate angiogenic mechanisms? This is an important unanswered question. What has become apparent is that, in many experimental settings, two drugs are better than one. For example, co-administration of angiogenic inhibitors, such as TNP470 (Takeda-Abbott Pharmaceuticals), DC101 (ImClone Systems; VEGF receptor inhibitor), SU5416 (Sugen Inc., VEGF receptor inhibitor) or batimastat (broadspectrum MMP inhibitor from British Biotech Pharmaceuticals), with low doses of a cytotoxic drug (which has an anti-angiogenic effect, known as ‘metrononomic’ therapy), results in disease stabilization, or even regression. This combination has been shown to be more effective and less toxic than highdose chemotherapy 81–83. Antibodies that block the VEGF receptors VEGFR1 and VEGFR2 (ImClone Systems) more effectively block tumour growth than either one does alone18. What has also become apparent from studying the mouse model of pancreatic-islet carcinogenesis is the fact that the efficacy of angiogenesis inhibitors depends on the tumour stage84 (TABLE 2). Endostatin, MMP and VEGF inhibitors are most successful in treating earlystage disease — both in targeting angiogenic switching in dysplastic lesions and in blocking the expansive


VOLUME 3 | JUNE 2003 | 4 0 7


REVIEWS

Table 2 | Preventing pancreatic cancer with angiogenic inhibitors Angiogenesis inhibitor Target

Efficacy in tumour stage Pre

Sm

Lg

Endostatin

Integrin α5β310

++

++

+/–

Angiostatin

ATP synthase104 Angiopoietin105

+/–

++

+/–

Endostatin + Angiostatin

See main text

++

n.d.

++

TNP470/AGM1470

EC proliferation106 SMC migration85

–

++

++

Batimastat/BB-94

MMP1, 2, 3, 7, 8, 9107

++

++

+/–

SU5416

VEGF receptors108

+++

++

+/–

SU6668

PDGF receptor, VEGF receptor, ++ FGF receptor109

++

++

SU5416 + SU6668

See main text

+++

+++

++++

See REFS 45,84,86. EC, epithelial cell; FGF, fibroblast growth factor; Lg, large tumour; MMP, matrix metalloproteinase; n.d., not determined; PDGF, platelet-derived growth factor; Pre, pre-malignant; Sm, small tumour; SMC, smooth-muscle cell; VEGF, vascular endothelial growth factor.

growth of small solid tumours. These drugs do not, however, stabilize or even cause regression of bulky end-stage tumours. On the other hand, inhibitors such as TNP-470/AGM-1470, which block endothelial-cell proliferation and smooth-muscle-cell migration85, are effective at reducing the mass of bulky end-stage tumours, but have no effect on preventing progression during early-stage disease. These differential responses indicate that there might be qualitative differences in the angiogenic vasculature in early and late stages, or in the regulatory mechanisms that control induction of angiogenesis and persistence of the tumour vasculature. Indeed, treatment of islet tumours with the receptor tyrosine kinase inhibitors SU6668 or Glivec resulted in detachment of pericytes of established tumour vessels, which was partly due to inhibition of platelet-derived growth factor (PDGF) receptor (PDGFR) signalling. Alternatively, inhibition of VEGF receptor signalling by SU5416 blocked neovascularization without any signs of pericyte detachment. Interestingly, combination of SU5416 (which blocks VEGFR) and SU6668 (which blocks PDGFR, and to a lesser extent FGFR and VEGFR) are able to block the angiogenic switch, as well as to induce regression of small and large tumours in the pancreatic-islet model, having improved synergistic effects at all stages, which, in every case, is better than either single agent alone86. Combination of a VEGF receptor inhibitor with another distinctive kinase inhibitor Glivec (which only overlaps with SU6668 in targeting PDGFR activity), was also able to induce regression of late-stage tumours, causing disruption and regression of the tumour vasculature, as observed with the two SUGEN compounds. These synergistic inhibitory effects are achieved by targeting different signalling circuits, not only in endothelial cells but also in perivascular cells. These studies also highlight the potential significance of PDGFR signalling in tumour-associated pericytes, thereby implicating this cell type as a functionally important component of the tumour vasculature and a new target for anti-angiogenic therapy 86. The synergistic

408

effects of PDGF and VEGFA inhibition can be partially explained by studies in developing retina and prostate cancer, which showed that blood vessels with few pericytes were more sensitive to VEGFA inhibition14,27. So, how can researchers determine which combination of angiogenesis inhibitors will be most effective in the clinic? Cline et al. have performed gene-expression profiling studies on endothelial cells following different anti-angiogenic treatments87. This clever approach allowed them to accurately predict which anti-angiogenic agents would act synergistically in inhibiting tumour growth in animal models. Studies that involve combination therapy, gene profiling and the use of more realistic animal models will improve clinical-trial design and help to identify drugs that are most effective in treating and preventing human cancers. What lessons should we take to the lab?

Angiogenesis is one of many discrete steps in tumour progression. Angiogenesis itself is a complex, multistep process that follows stage- and tissue-specific regulations. Although much effort has been put into the identification of the still-growing number of endogenous angiogenic inhibitors and activators, much less is known about how these factors work together to form vessels at any given stage of tumour development. For example, what does it mean if a tumour expresses a certain number of angiogenesis activators and inhibitors, with regard to its growth requirements? How do these factors interact to form tumour vessels, and how does this differ from the formation of a normal blood network? What happens when tumours co-opt blood vessels? Do those vessels undergo changes and have signs of activated endothelium that would respond to an antiangiogenic therapy? These are only a few questions that need to be addressed to have a better understanding of the angiogenic switch. Another important aspect of angiogenesis research is to use an appropriate experimental setting to study the efficacy of anti-angiogenesis molecules in vivo. It has always been a challenge to extrapolate animal data into the clinical setting, which has been partly attributed to differences in mouse and human metabolism. Another unmentioned complication is that traditional xenotransplant models involve cultured tumour cells that are inoculated into different sites — most frequently subcutaneously, where tumour cells assemble into nodules and grow. Human cancers arise de novo, originating out of once-normal cells in natural tissue microenvironments. So, spontaneous tumour models, in which normal cells become malignant within their natural microenvironments via a multistep pathway, are more likely to recapitulate the human situation. Based on the multistep progression pathway, these models can be used to study the impact of angiogenesis inhibitors on blocking the angiogenic switch in early premalignant lesions or reducing small or bulky latestage tumours. So far, these models have been of predictive value. For example, the RipTag model confirmed recent clinical trials showing that broad-spectrum




REVIEWS MMP inhibitors are not effective in treating end-stage disease, whereas combinatorial strategies involving lowdose chemotherapy have been (TABLE 2). Therefore, there is reason to be optimistic that genetically engineered mouse models of organ-specific carcinogenesis will indeed provide new insight into the potential efficacy of targeted anti-angiogenic therapies for different kinds of cancer. What lessons should we take to the clinic?

The first anti-angiogenesis clinical trials in cancer patients have not lived up to their high expectations to significantly reduce tumour burden and prolong life, as observed in various pre-clinical studies. On the other hand, it is not necessarily surprising that the aggressive, therapy-resistant tumours of end-stage patients do not regress when confronted with a single anti-angiogenic agent such as endostatin, or a factor that targets VEGF signalling.As discussed, tumours at this stage have already activated various pathways that allow them to easily override the angiogenic restrictions of one inhibitor.

1. 2.

3. 4.

5. 6. 7.

8.

9.

10.

11.

12. 13.

14.

15.

16. 17.

18.

Hahn, W. C. et al. Rules for making human tumor cells. N. Engl. J. Med. 347, 1593–603 (2002). Duesberg, P. et al. How aneuploidy may cause cancer and genetic instability. Anticancer Res. 19, 4887–4906 (1999). Pelengaris, S. et al. c-MYC: more than just a matter of life and death. Nature Rev. Cancer 2, 764–776 (2002). Vogelstein, B. et al. Genetic alterations during colorectaltumor development. N. Engl. J. Med. 319, 525–532 (1988). Hanahan, D. et al. The hallmarks of cancer. Cell 100, 57–70 (2000). Folkman, J. in Cancer Medicine (Holland, M.D. et al.) 2546 (BC Decker, Hamilton, Ontario, 2000). Papetti, M. et al. Mechanisms of normal and tumor-derived angiogenesis. Am. J. Physiol. Cell Physiol. 282, C497–C970 (2002). Hlatky, L. et al. Clinical application of antiangiogenic therapy: microvessel density, what it does and doesn’t tell us. J. Natl Cancer Inst. 94, 883–893 (2002). This important review discusses several misconceptions in the angiogenesis field. It specifically focuses on microvessel density and the information that can and cannot be gained from it. Dor, Y. et al. Vascular endothelial growth factor and vascular adjustments to perturbations in oxygen homeostasis. Am. J. Physiol. Cell Physiol. 280, C1367–C1374 (2001). Kerbel, R. et al. Clinical translation of angiogenesis inhibitors. Nature Rev. Cancer 2, 727–739 (2002). This is a recent and detailed review about antiangiogenic therapy. Hanahan, D. et al. Patterns and emerging mechanisms of the angiogenic switch during tumorigenesis. Cell 86, 353–364 (1996). Carmeliet, P. Mechanisms of angiogenesis and arteriogenesis. Nature Med. 6, 389–395 (2000). Dvorak, H. F. Tumors: wounds that do not heal. Similarities between tumor stroma generation and wound healing. N. Engl. J. Med. 315, 1650–1659 (1986). Benjamin, L. E. et al. Selective ablation of immature blood vessels in established human tumors follows vascular endothelial growth factor withdrawal. J. Clin. Invest. 103, 159–165 (1999). Morikawa, S. et al. Abnormalities in pericytes on blood vessels and endothelial sprouts in tumors. Am. J. Pathol. 160, 985–1000 (2002). Folberg, R. et al. Vasculogenic mimicry and tumor angiogenesis. Am. J. Pathol. 156, 361–381 (2000). McDonald, D. M. et al. Vasculogenic mimicry: how convincing, how novel, and how significant? Am. J. Pathol. 156, 383–388 (2000). Lyden, D. et al. Impaired recruitment of bone-marrowderived endothelial and hematopoietic precursor cells blocks tumor angiogenesis and growth. Nature Med. 7, 1194–1201 (2001).

Another promising anti-angiogenic approach that should be further explored is the ability of angiogenesis inhibitors not only to reduce tumour growth, but also block the progression of dormant lesions into aggressive cancers or metastases in high-risk cancer patients. The most successful approaches are likely to involve combinatorial strategies to target cancer cells themselves, along with the stroma (that is, endothelial, perivascular and inflammatory cells). Clinical trials, however, have been hampered by intellectual property issues that prevent combinatorial testing of agents that are produced by different drug companies. Fortunately, some dual-action inhibitors have emerged that will allow us to test this approach, such as drugs that inhibit both PDGF and VEGF receptors88. These include SU11248 from Sugen, PTK787 from Novartis and AG013736 from Pfizer — all of which are in early clinical development. We anticipate and hope that inhibitors such as these — used alone or in combination with other classes of anticancer drug — will be more effective in restraining cancer growth.

19. Rafii, S. et al. Efficient mobilization and recruitment of marrow-derived endothelial and hematopoietic stem cells by adenoviral vectors expressing angiogenic factors. Gene Ther. 9, 631–641 (2002). 20. Hattori, K. et al. Vascular endothelial growth factor and angiopoietin-1 stimulate postnatal hematopoiesis by recruitment of vasculogenic and hematopoietic stem cells. J. Exp. Med. 193, 1005–1014 (2001). 21. Hattori, K. et al. Placental growth factor reconstitutes hematopoiesis by recruiting VEGFR1(+) stem cells from bone-marrow microenvironment. Nature Med. 8, 841–849 (2002). 22. Pettersson, A. et al. Heterogeneity of the angiogenic response induced in different normal adult tissues by vascular permeability factor/vascular endothelial growth factor. Lab. Invest. 80, 99–115 (2000). 23. Holash, J. et al. Vessel cooption, regression, and growth in tumors mediated by angiopoietins and VEGF. Science 284, 1994–1998 (1999). The is the first report showing that tumours do not always start as avascular lesions, but can also initially grow by co-opting existing blood vessels. The coopted host vasculature does not immediately undergo angiogenesis to support the tumour — it first regresses before hypoxia-induced neovascularization is initiated. 24. Yancopoulos, G. D. et al. Vascular-specific growth factors and blood vessel formation. Nature 407, 242–248 (2000). 25. Stone, J. et al. Development of retinal vasculature is mediated by hypoxia-induced vascular endothelial growth factor (VEGF) expression by neuroglia. J. Neurosci. 15, 4738–4747 (1995). 26. Hirschi, K. K. et al. Pericytes in the microvasculature. Cardiovasc. Res. 32, 687 (1996). 27. Benjamin, L. E. et al. A plasticity window for blood vessel remodelling is defined by pericyte coverage of the preformed endothelial network and is regulated by PDGF-B and VEGF. Development 125, 1591–1598 (1998). 28. Benjamin, L. E. et al. Conditional switching of vascular endothelial growth factor (VEGF) expression in tumors: induction of endothelial cell shedding and regression of hemangioblastoma-like vessels by VEGF withdrawal. Proc. Natl Acad. Sci. USA 94, 8761–8766 (1997). 29. Papapetropoulos, A. et al. Direct actions of angiopoietin-1 on human endothelium: evidence for network stabilization, cell survival, and interaction with other angiogenic growth factors. Lab. Invest. 79, 213–223 (1999). 30. Adini, A. et al. Placental growth factor is a survival factor for tumor endothelial cells and macrophages. Cancer Res. 62, 2749–2752 (2002). 31. Ferrara, N. VEGF and the quest for tumour angiogenesis factors. Nature Rev. Cancer 2, 795–803 (2002). 32. Carmeliet, P. et al. Synergism between vascular endothelial growth factor and placental growth factor contributes to


33.

34. 35.

36.

37. 38.

39.

40. 41.

42. 43.

44.

45.

46.

47.

48.

49.

50.

angiogenesis and plasma extravasation in pathological conditions. Nature Med. 7, 575–583 (2001). Semenza, G. L. HIF-1 and tumor progression: pathophysiology and therapeutics. Trends Mol. Med. 8, S62–S67 (2002). Maxwell, P. H. et al. Activation of the HIF pathway in cancer. Curr. Opin. Genet. Dev. 11, 293–299 (2001). Black, W. C. et al. Advances in diagnostic imaging and overestimations of disease prevalence and the benefits of therapy. N. Engl. J. Med. 328, 1237–1243 (1993). O’Reilly, M. S. Angiostatin: an endogenous inhibitor of angiogenesis and of tumor growth. EXS. 79, 273–294 (1997). O’Reilly, M. S. et al. Endostatin: an endogenous inhibitor of angiogenesis and tumor growth. Cell 88, 277–285 (1997). Joseph, I. B. et al. Antiangiogenic treatment with linomide as chemoprevention for prostate, seminal vesicle, and breast carcinogenesis in rodents. Cancer Res. 56, 3404–3408 (1996). Subbaramaiah, K. et al. Cyclooxygenase 2: a molecular target for cancer prevention and treatment. Trends Pharmacol. Sci. 24, 96–102 (2003). Iniguez, M. A. et al. Cyclooxygenase-2: a therapeutic target in angiogenesis. Trends Mol. Med. 9, 73–78 (2003). Lieberman, R. Chemoprevention of prostate cancer: current status and future directions. Cancer Metastasis Rev. 21, 297–309 (2002). Hussain, T. et al. Cyclooxygenase-2 and prostate carcinogenesis. Cancer Lett. 191, 125–135 (2003). Chen, Y. J. et al. High cyclooxygenase-2 expression in cervical adenocarcinomas. Gynecol. Oncol. 88, 379–385 (2003). Folkman, J. et al. Induction of angiogenesis during the transition from hyperplasia to neoplasia. Nature 339, 58 (1989). Bergers, G. et al. Matrix metalloproteinase-9 triggers the angiogenic switch during carcinogenesis. Nature Cell Biol. 2, 737 (2000). Coussens, L. M. et al. Inflammatory mast cells up-regulate angiogenesis during squamous epithelial carcinogenesis. Genes Dev. 13, 1382–1397 (1999). Bergers, G. et al. Angiogenesis and apoptosis are cellular parameters of neoplastic progression in transgenic mouse models of tumorigenesis. Int. J. Dev. Biol. 42, 995 (1998). Elson, D. A. et al. Sensitivity of the cervical transformation zone to estrogen-induced squamous carcinogenesis. Cancer Res. 60, 1267–1275 (2000). Inoue, M. et al. VEGF-A has a critical, nonredundant role in angiogenic switching and pancreatic beta cell carcinogenesis. Cancer Cell 1, 193–202 (2002). Rodriguez-Manzaneque, J. C. et al. Thrombospondin-1 suppresses spontaneous tumor growth and inhibits activation of matrix metalloproteinase-9 and mobilization of vascular endothelial growth factor. Proc. Natl Acad. Sci. USA 98, 12485–12490 (2001).

VOLUME 3 | JUNE 2003 | 4 0 9


REVIEWS 51. Weidner, N. Tumoural vascularity as a prognostic factor in cancer patients: the evidence continues to grow. J. Pathol. 184, 119–122 (1998). 52. Zabrenetzky, V. et al. Expression of the extracellular matrix molecule thrombospondin inversely correlates with malignant progression in melanoma, lung and breast carcinoma cell lines. Int. J. Cancer 59, 191–195 (1994). 53. Weinstat-Saslow, D. L. et al. Transfection of thrombospondin 1 complementary DNA into a human breast carcinoma cell line reduces primary tumor growth, metastatic potential, and angiogenesis. Cancer Res. 54, 6504–6511 (1994). 54. Pelengaris, S. et al. Suppression of Myc-induced apoptosis in beta cells exposes multiple oncogenic properties of Myc and triggers carcinogenic progression. Cell 109, 321–334 (2002). 55. Watnick, R. S. et al. Ras modulates Myc activity to repress thrombospondin-1 expression and increase tumor angiogenesis. Cancer Cell 3, 219–231 (2003). This study strengthens the proposition that the angiogenic switch is a crucial step in tumour progression and that is determined by the ultimate balance in pro- and anti-angiogenic factors. 56. Vajkoczy, P. et al. Microtumor growth initiates angiogenic sprouting with simultaneous expression of VEGF, VEGF receptor-2, and angiopoietin-2. J. Clin. Invest. 109, 777–785 (2002). 57. Holash, J. et al. New model of tumor angiogenesis: dynamic balance between vessel regression and growth mediated by angiopoietins and VEGF. Oncogene 18, 5356–5362 (1999). 58. Shweiki, D. et al. Vascular endothelial growth factor induced by hypoxia may mediate hypoxia-initiated angiogenesis. Nature 359, 843–845 (1992). 59. Tomlinson, F. H. et al. The significance of atypia and histologic malignancy in pilocytic astrocytoma of the cerebellum: a clinicopathologic and flow cytometric study. J. Child Neurol. 9, 301–310 (1994). 60. Brem, S. et al. Tumor angiogenesis: a quantitative method for histologic grading. J. Natl Cancer Inst. 48, 347–356 (1972). 61. Eberhard, A. et al. Heterogeneity of angiogenesis and blood vessel maturation in human tumors: implications for antiangiogenic tumor therapies. Cancer Res. 60, 1388–1393 (2000). 62. Rubenstein, J. L. et al. Anti-VEGF antibody treatment of glioblastoma prolongs survival but results in increased vascular cooption. Neoplasia 2, 306–314 (2000). 63. Filleur, S. et al. In vivo mechanisms by which tumors producing thrombospondin 1 bypass its inhibitory effects. Genes Dev. 15, 1373–1382 (2001). 64. Douglas, M. L. et al. Renal cell carcinoma may adapt to and overcome anti-angiogenic intervention with thalidomide. BJU Int. 89, 591–595 (2002). 65. Yu, J. L. et al. Effect of p53 status on tumor response to antiangiogenic therapy. Science 295, 1526–1528 (2002). 66. Gee, M. S. et al. Hypoxia-mediated apoptosis from angiogenesis inhibition underlies tumor control by recombinant interleukin 12. Cancer Res. 59, 4882–4889 (1999). 67. Tran, J. et al. A role for survivin in chemoresistance of endothelial cells mediated by VEGF. Proc. Natl Acad. Sci. USA 99, 4349–4354 (2002). 68. Wang, J. H. et al. Hypoxia upregulates Bcl-2 expression and suppresses interferon-gamma induced antiangiogenic activity in human tumor derived endothelial cells. Cancer 94, 2745–2755 (2002). 69. Zhang, X. S. et al. Multiple drug resistance phenotype of human endothelial cells induced by vascular endothelial growth factor 165. Acta Pharmacol. Sin. 22, 731–735 (2001). 70. Forsythe, J. A. et al. Activation of vascular endothelial growth factor gene transcription by hypoxia-inducible factor 1. Mol. Cell. Biol. 16, 4604–4613 (1996). 71. Stein, I. et al. Stabilization of vascular endothelial growth factor mRNA by hypoxia and hypoglycemia and coregulation with other ischemia-induced genes. Mol. Cell. Biol. 15, 5363–5368 (1995).

410

72. Laderoute, K. R. et al. Opposing effects of hypoxia on expression of the angiogenic inhibitor thrombospondin 1 and the angiogenic inducer vascular endothelial growth factor. Clin. Cancer Res. 6, 2941–2950 (2000). 73. Royds, J. A. et al. Response of tumour cells to hypoxia: role of p53 and NFkB. Mol. Pathol. 51, 55–61 (1998). 74. Kwak, C. et al. Thrombospondin-1, vascular endothelial growth factor expression and their relationship with p53 status in prostate cancer and benign prostatic hyperplasia. BJU Int. 89, 303–309 (2002). 75. Mukhopadhyay, D. et al. Wild-type p53 and v-Src exert opposing influences on human vascular endothelial growth factor gene expression. Cancer Res. 55, 6161–6165 (1995). 76. Zhang, L. et al. Wild-type p53 suppresses angiogenesis in human leiomyosarcoma and synovial sarcoma by transcriptional suppression of vascular endothelial growth factor expression. Cancer Res. 60, 3655–3661 (2000). 77. Ravi, R. et al. Regulation of tumor angiogenesis by p53induced degradation of hypoxia-inducible factor 1alpha. Genes Dev. 14, 34–44 (2000). 78. Agani, F. et al. p53 does not repress hypoxia-induced transcription of the vascular endothelial growth factor gene. Cancer Res. 57, 4474–4477 (1997). 79. Holmgren, L. et al. p53 induces angiogenesis-restricted dormancy in a mouse fibrosarcoma. Oncogene 17, 819–824 (1998). 80. Gautam, A. et al. Aerosol delivery of PEI-p53 complexes inhibits B16-F10 lung metastases through regulation of angiogenesis. Cancer Gene Ther. 9, 28–36 (2002). 81. Klement, G. et al. Continuous low-dose therapy with vinblastine and VEGF receptor-2 antibody induces sustained tumor regression without overt toxicity. J. Clin. Invest. 105, R15–R24 (2000). 82. Browder, T. et al. Antiangiogenic scheduling of chemotherapy improves efficacy against experimental drug-resistant cancer. Cancer Res. 60, 1878–1886 (2000). 83. Bergers, G. et al. Combining anti-angiogenic agents with metronomic chemotherapy enhances efficacy against latestage pancreatic islet carcinomas in mice. CSH Symp. Quant. Biol. 67, 293–300 (2002). 84. Bergers, G. et al. Effects of angiogenesis inhibitors on multistage carcinogenesis in mice. Science 284, 808–812 (1999). 85. Shigematsu, K. et al. Anti-angiogenic drug AGM1470 suppresses smooth muscle cell migration induced by endothelial PDGF. Eur. J. Vasc. Endovasc. Surg. 14, 290–298 (1997). 86. Bergers, G. et al. Benefits of targeting both pericytes and endothelial cells in the tumor vasculature. J. Clin. Invest. 111, 1287–1295 (2003). 87. Cline, E. I. et al. Prediction of in vivo synergistic activity of antiangiogenic compounds by gene expression profiling. Cancer Res. 62, 7143–7148 (2002). 88. Dancey, J. et al. Issues and progress with protein kinase inhibitors for cancer treatment. Nature Rev. Drug Discov. 2, 296–313 (2003). 89. Hu, Y. L. et al. Lysophosphatidic acid induction of vascular endothelial growth factor expression in human ovarian cancer cells. J. Natl Cancer Inst. 93, 762–768 (2001). 90. Volpert, O. V. et al. The modulation of thrombospondin and other naturally occurring inhibitors of angiogenesis during tumor progression. Breast Cancer Res. Treat. 36, 119–126 (1995). 91. Maeshima, Y. et al. Identification of the anti-angiogenic site within vascular basement membrane-derived tumstatin. J. Biol. Chem. 276, 15240–15248 (2001). 92. Kamphaus, G. D. et al. Canstatin, a novel matrix-derived inhibitor of angiogenesis and tumor growth. J. Biol. Chem. 275, 1209–1215 (2000). 93. Dameron, K. M. et al. Control of angiogenesis in fibroblasts by p53 regulation of thrombospondin-1. Science 265, 1582–1584 (1994). 94. Lawler, J. et al. Thrombospondin-1 gene expression affects survival and tumor spectrum of p53-deficient mice. Am. J. Pathol. 159, 1949–1956 (2001).


95. Nishimori, H. et al. A novel brain-specific p53-target gene, BAI1, containing thrombospondin type 1 repeats inhibits experimental angiogenesis. Oncogene 15, 2145–2150 (1997). 96. Bian, J. et al. Transcriptional activation by p53 of the human type IV collagenase (gelatinase A or matrix metalloproteinase 2) promoter. Mol. Cell. Biol. 17, 6330–6338 (1997). 97. Brantley, D. M. et al. Soluble Eph A receptors inhibit tumor angiogenesis and progression in vivo. Oncogene 21, 7011–7026 (2002). 98. Dohn, M. et al. Receptor tyrosine kinase EphA2 is regulated by p53-family proteins and induces apoptosis. Oncogene 20, 6503–6515 (2001). 99. Cheng, N. et al. Blockade of EphA receptor tyrosine kinase activation inhibits vascular endothelial cell growth factorinduced angiogenesis. Mol. Cancer Res. 1, 2–11 (2002). 100. Sun, Y. et al. Human metalloproteinase-1 (collagenase-1) is a tumor suppressor protein p53 target gene. Ann. NY Acad. Sci. 878, 638–641 (1999). 101. Zhou, Z. et al. Impaired endochondral ossification and angiogenesis in mice deficient in membrane-type matrix metalloproteinase I. Proc. Natl Acad. Sci. USA 97, 4052–4057 (2000). 102. Subbaramaiah, K. et al. Inhibition of cyclooxygenase-2 gene expression by p53. J. Biol. Chem. 274, 10911–10915 (1999). 103. Pal, S. et al. Central role of p53 on regulation of vascular permeability factor/vascular endothelial growth factor (VPF/VEGF) expression in mammary carcinoma. Cancer Res. 61, 6952–6957 (2001). 104. Moser, T. Angiostatin binds ATP synthase on the surface of human endothelial cells. Proc. Natl Acad. Sci. USA 96, 2811–2816 (1999). 105. Troyanovsky, B. et al. Angiomotin: an angiostatin binding protein that regulates endothelial cell migration and tube formation. J. Cell. Biol. 152, 1247–1254 (2001). 106. Yeh, R. J. et al. The antiangiogenic agent TNP470 requires p53 and p21CIP/WAF for endothelial cell growth arrest. Proc. Natl Acad. Sci. USA 97, 12782–12787 (2000). 107. Sternlicht, M. D. et al. Matrix proteinases as emerging targets in anticancer therapy: status and prospects. ETT 4, 609–633 (2000). 108. Fong, T. A. et al. SU5416 is a potent and selective inhibitor of the vascular endothelial growth factor receptor (Flk–1/KDR) that inhibits tyrosine kinase catalysis, tumor vascularization, and growth of multiple tumor types. Cancer Res. 59, 99 (1999). 109. Laird, A. D. et al. SU6668 is a potent antiangiogenic and antitumor agent that induces regression of established tumors. Cancer Res. 60, 4152–4160 (2000).

Acknowledgements We thank T. Tihan (University of California San Francisco) for advising us on brain-tumour pathology.

Online links DATABASES The following terms in this article are linked online to: Cancer.gov: http://www.cancer.gov/cancer_information/ astrocytoma | breast cancer | cervical cancer | colon cancer | glioblastoma | lung cancer | melanoma | prostate cancer | renalcell carcinoma GenBank: http://www.ncbi.nih.gov/Genbank/ E1A | E6 | E7 LocusLink: http://www.ncbi.nih.gov/LocusLink/ ANG1 | ANG2 | BCL2 | BCL-XL | endostatin | FGF | HIF-1α | MDM2 | MMP9 | MYC | p53 | PDGF | PDGFR | PlGF | RAS | RB | TERT | TIE2 | TNF-α | TSP1 | VEGFA | VEGFC | VEGFD | VEGFR1 | VEGFR2 OMIM: http://www.ncbi.nlm.nih.gov/Omim/ familial adenomatous polyposis FURTHER INFORMATION National Cancer Institute — Understanding angiogenesis: http://press2.nci.nih.gov/sciencebehind/angiogenesis/angio01.htm Access to this interactive links box is free online.

