Introduction

The growth and metastasis of tumors depends, in part, on their ability to induce the growth of new blood vessels (Folkman, 1971, 1990). It is widely accepted that most tumors and metastases originate as small avascular structures, which must induce the development of new vessels in order to grow beyond a few millimeters in size (Folkman, 1971, 1990). An initial phase of avascular growth certainly seems to be a feature of tumor cells seeded into avascular structures such as the cornea, the anterior eye chamber or an artificial tumor window chamber, as well as of tumor cells implanted into virtual spaces, such those implanted subcutaneously or into the peritoneum (Folkman, 1990). It also seems likely to be characteristic of early tumors arising in epithelial structures that are separated by a basement membrane from underlying vasculature (Folkman, 1990). However, other evidence suggests that many tumors may not initially grow in an avascular fashion, particularly when they are growing within the confines of a vascularized tissue (Wesseling et al., 1994; Holmgren et al., 1995; Pezzella et al., 1997). In such settings, tumor cells often seem to coopt existing blood vessels. The interplay between such initial coopting of existing vessels and subsequent tumor-induced angiogenesis has not been extensively examined.

Tumor-induced angiogenesis is thought to depend on the production of pro-angiogenic growth factors by the tumor cells, which overcome other forces that tend to keep existing vessels quiescent and stable (Hanahan and Folkman, 1996). This process involves a diverse array of molecules that includes both those that regulate the maintenance and destruction of the perivascular milieu (which includes both extracellular matrix and perivascular cells) as well as those which stimulate endothelial cell division and migration.

Perhaps the best characterized of the pro-angiogenic agents is vascular endothelial growth factor (VEGF), which is relatively unique among growth factors in terms of its specificity for the vascular endothelium (Ferrara, 1999). VEGF acts rather specifically on vascular endothelium because the receptors for VEGF are largely restricted to these cells; VEGF acts to initiate proliferation and the sprouting of processes, and can promote the formation of new vessel-like structures. Since VEGF expression is induced in tumor cells by hypoxia (Shweiki et al., 1995), VEGF production is thought to reflect a poorly vascularized tumor, and the production of VEGF by tumor cells is thought to contribute to onset of tumor-associated angiogenesis. Consistent with this notion, anti-VEGF approaches slow the growth of many tumors for example, (Kim et al., 1993; Millauer et al., 1994, 1996; Asano et al., 1995; Warren et al., 1995; Goldman et al., 1998). The role of VEGF in tumor-associated angiogenesis apparently recapitulates its actions during normal vascular development. Gene knockout studies in mice have proven that VEGF and one of its receptors, Flk1/KDR, are absolutely required for the early stages of vascular development in the embryo (Shalaby et al., 1995; Carmeliet et al., 1996; Ferrara et al., 1996). Mice lacking VEGF exhibit gross abnormalities in endothelial cell differentiation, proliferation and vessel formation (Figure 1a).

Figure 1
figure 1

Model for the co-ordinated and complementary roles of VEGF and the angiopoietins during vascular development and remodeling, normally and during tumorigenesis. (a) In the embryo, VEGF is required for differentiation and proliferation of endothelial cells, and for tubule formation. After VEGF promotes the formation of immature vessels, Angiopoietin-1, acting through the Tie2 receptor, promotes the remodeling, maturation and stabilization of blood vessels, which apparently involves optimizing interactions with surrounding support cells and extracellular matrix. (b) By adulthood, Tie2 is constitutively engaged with Angiopoietin-1, and vessels remain in a stable state. When Angiopoietin-2 expression is upregulated, for instance, in the normal ovary or by tumor cells, the interaction between Tie2 and Angiopoietin-1 is disrupted, and the vessel is destabilized. Once endothelial cells are separated from perivascular cells and extracellular matrix, they appear particularly plastic and vulnerable. In the presence of VEGF, angiogenesis is promoted; however, lack of stimulatory factors results in the regression of destabilized vessels

Full size image

Recently, a second family of growth factors specific for the vascular endothelium has been identified, with members of this family termed the Angiopoietins (Davis et al., 1996; Suri et al., 1996; Maisonpierre et al., 1997). Similar to VEGF, the specificity of the Angiopoietins for the vascular endothelium results from the restricted distribution of the Angiopoietin receptors (known as the Tie receptors) to these cells. Tie1 and Tie2 are receptor tyrosine kinases just as are the receptors for VEGF. All known Angiopoietins bind to Tie2, but it is still unclear as to whether they utilize the closely related receptor Tie1. The actions of the Angiopoietins seem to be quite different from those of VEGF. In fact, the Angiopoietins seem to act in complementary and coordinated fashion with VEGF, playing a later role in vascular development. Thus, in mouse embryos lacking either Angiopoietin-1 or Tie2, the early stages of VEGF-dependent vascular development seem to occur rather normally, resulting in the formation of a primitive vasculature (Dumont et al., 1994; Sato et al., 1995; Suri et al., 1996). However, remodeling and stabilization of this primitive vasculature is severely perturbed (Figure 1a), leading to embryonic lethality. Angiopoietin-1 is thought to play a key role in mediating interactions between endothelial cells and surrounding support cells such as smooth muscle cells (Folkman and D'Amore, 1996; Suri et al., 1996). Ultrastructural examination of vessels in mice lacking Angiopoietin-1 showed that endothelial cells failed to interact and adhere properly to underlying support cells; in the absence of such optimized cellular interactions, vessels apparently fail to undergo normal remodeling events and may also be at risk for subsequent regression. Angiopoietin-1 seems to be constitutively expressed in the adult, where it may continue to play a required stabilizing role for existing vessels. Transgenic overexpression of Angiopoietin-1 leads to striking hypervascularization, more impressive than that reported for VEGF, presumably by promoting vascular remodeling events and perhaps by decreasing normal vascular pruning (Suri et al., 1998).

Shortly after the discovery of Angiopoietin-1, the cloning of Angiopoietin-2 was described (Maisonpierre et al., 1997). Remarkably, though Angiopoietin-2 bound to the Tie2 receptor as did Angiopoietin-1, it could not activate it in endothelial cells. Instead, Angiopoietin-2 provided the first example of a naturally occurring antagonist for a vertebrate receptor tyrosine kinase, suggesting that the ability to turn off the Tie2 receptor was just as important as the ability to activate it. Consistent with the notion that Angiopoietin-2 acts as a natural antagonist for the Angiopoietin-1/Tie2 interaction, transgenic overexpression of Angiopoietin-2 during embryogenesis (Maisonpierre et al., 1997) leads to a lethal phenotype reminiscent of that seen in embryos lacking either Angiopoietin-1 or Tie2, with severe disruptions in vascular development (Figure 1a).

Examination of Angiopoietin expression patterns in vivo suggested a very interesting role for Angiopoietin-2 at sites of vascular remodeling in an otherwise stable adult vasculature (Maisonpierre et al., 1997). While Angiopoietin-1 is widely expressed in normal adult tissues (where Tie2 is constitutively phosphorylated (Wong et al., 1997)), consistent with it playing a continuously required stabilization role, Angiopoietin-2 is typically expressed at sites of vascular remodeling in the adult, notably in the female reproductive tract (Maisonpierre et al., 1997). Detailed localization of Angiopoietin-2 in the ovary by in situ hybridization revealed that in regions of active vascular remodeling it was either expressed together with VEGF at sites of vessel sprouting and ingrowth (e.g. developing corpus luteum), or in the absence of VEGF at sites of frank vessel regression (e.g. atretic follicles). These expression patterns led to the proposal of a model in which Angiopoietin-2 plays a facilitative role at sites of vascular remodeling in the adult by blocking a constitutive stabilizing action of Angiopoietin-1 (Figure 1b). Further, it was suggested that such destabilization by Angiopoietin-2 in the presence of high VEGF levels primes the vessels to mount a robust angiogenic response reminiscent to that of early embryonic vessels prior to maturation (Figure 1b). However, such destabilization by Angiopoietin-2 in the absence of VEGF is instead proposed to lead to frank vessel regression (Figure 1b).

The apparently coordinated actions of VEGF and the Angiopoietins during normal embryonic vascular development, coupled with the suggestion of a key destabilizing role for Angiopoietin-2 during normal adult vascular remodeling, demanded that the potential actions of the Angiopoietins be examined during pathological adult angiogenesis, such as occurs in tumors. We recently reported on the results of such an examination (Holash et al., 1999; Zagzag et al., 1999), and these findings are reviewed in detail below.

Rat gliomas initially co-opt host vessels which then regress and only belatedly co-opted vessels undergo angiogenesis to support tumor growth

We initially performed an in-depth temporal analysis of tumor-vessel interactions in the rat C6 glioma model (Nagano et al., 1993). This model involves stereotaxically implanting a relatively small number of cells in a small volume (thus preventing the formation of a large cavity at the implantation site) into the striatum. As early as 1 week after implantation, we observed small, well-vascularized tumors (Holash et al., 1999). Similar to what had previously been described by others (Nagano et al., 1993), we found that rather than growing as an avascular mass, the tumor cells apparently had coopted the vessels of the surrounding brain. Interestingly, a morphologically obvious angiogenic response was not apparent in these tumors. Instead, the center of the tumor became progressively hypovascular as the tumor grew. Although in early stages of tumor development it appeared that the hypovascularity was a consequence of the tumor growth in the absence of accompanying angiogenesis, at later stages it became obvious that vessels within the center of the tumor had regressed. It was tempting to speculate that this vessel regression represented a host defense mechanism that somehow sensed inappropriate co-opting and perturbation of existing vessels by the tumor, and that it led to dramatic death of tumor cells surrounding the regressed vessels (Holash et al., 1999). Despite this process, however, tumor was ultimately rescued by a robust angiogenic response occurring at the margins of the tumor, where the tumor cells continued to invade host brain. The changes in vascularity which occur during the growth of gliomas are diagrammed in Figure 2.

Figure 2
figure 2

Changes in tumor vasculature during growth. After co-opting host vessels, tumors (gray) initially grow as well vascularized masses. In relatively small tumors, as represented in (a), vessels continue to express Tie receptors at levels similar to those in vessels in host brain. A striking up-regulation of Angiopoietin-2 message becomes apparent in tumor vessels whereas expression of VEGF in surrounding tumor cells is minimal. As tumor growth progresses, many of the central tumor vessels regress (b), resulting in massive tumor cell death and necrosis (stippled region). Surviving tumor cells form cuffs around the few remaining internal vessels. Tumor cells at the periphery of the cuffs, as well as those at the margin of the tumor and the host brain, express very high levels of VEGF. Coincidentally, vessels at the interface between tumor and host brain remain strongly Angiopoietin-2 positive and express high levels of Tie receptors. This results in a region of robust angiogenesis

Full size image

We were fascinated by the apparent regression of vessels within the center of the gliomas, and with the possibility that this vessel regression preceded and actually led to loss of surrounding glioma cells. To pursue this possibility, we proceeded to label sections with a TUNEL stain to identify cells undergoing apoptosis. Strikingly, TUNEL labeling revealed that endothelial cell apoptosis occurs prior to the widespread loss of tumor cells, and thus precedes the development of the necrotic tumor core (Holash et al., 1999). To further investigate the mechanisms which were responsible for endothelial cell death, we doubled labeled sections with both an endothelial cell marker and antibodies to smooth muscle cell actin (SMA). Although SMA only labels a limited number of vascular profiles within the brain, it proved to be a very useful marker. In normal brain, when apparent, SMA-positive cells closely invest the endothelial cells forming the walls of blood vessels. However, as tumor growth progresses, vascular profiles in which the SMA positive cells appear to have detached from the endothelial wall become apparent, suggesting that the signal which maintains the normally stable relationship between the endothelial cell and perivascular cell has been disrupted (Holash et al., 1999).

Angiopoietin-2 initially marks co-opted glioma vessels for regression, but subsequently acts with VEGF to promote tumor angiogenesis

We and others have shown that although Angiopoietin-1 cannot induce endothelial cell proliferation in vitro, it is an effective endothelial cell survival factor in vitro (Holash et al., 1999; Kwak et al., 1999; Papapetropoulos et al., 1999). It has also been proposed that Angiopoietin-1 is involved in stabilizing vessels through recruitment of perivascular cells and matrix in vivo (Suri et al., 1996). We wondered whether Angiopoietin-2, the natural antagonist for Angiopoietin-1, might be playing a role in gliomas develolpment since we were observing both endothelial cell death and disrupted perivascular interactions in tumor vessels. To pursue this possibility, we used in situ hybridization analyses to localize the expression of VEGF, the Tie receptors and the Angiopoietins during glioma growth (Holash et al., 1999). Interestingly, Angiopoietin-2 was strikingly induced in the co-opted vessels of relatively small 2 week gliomas, while the other genes were not notably up-regulated at this early stage (Figure 2). Remarkably, Angiopoietin-2 was expressed by endothelial cells themselves; such autocrine expression would seem to be the most efficacious way of delivering an antagonist to the receptor it was designed to block. Early induction of Angiopoietin-2 in blood vessels precedes vessel regression, and it appears that Angiopoietin-2 may be the earliest marker of glioma associated vessels yet described. Consequently it seems that Angiopoietin-2 might indeed be the mediator of a host defense system designed to cause regression of initially co-opted tumor vessels.

Prominent Angiopoietin-2 expression continued to mark vessels throughout later stages of tumor growth, and at these later stages, Tie receptor expression was also upregulated in tumor associated vessels. Perhaps most importantly, dramatic VEGF induction also was noted in late-stage gliomas that had already undergone massive vessel regression. VEGF induction was not dramatic in early, well-vascularized tumors, consistent with the notion that the VEGF induction reflected the hypoxic state of the remaining tumor cells following vessel regression. During normal vascular remodeling it has been noted that Angiopoietin-2 expression in the absence of VEGF expression leads to vessel regression, whereas expression of Angiopoietin-2 in the presence of VEGF seems to prime vessels and facilitate responsiveness to VEGF (Figure 1b and Maisonpierre et al., 1997). Consistent with this notion, late expression of Angiopoietin-2 in the presence of tumor-derived VEGF was associated with robust angiogenesis, apparently allowing tumor rescue and further invasion of tumor into normal tissue. These findings are consistent with the idea that, in gliomas, Angiopoietin-2 and VEGF reprise the roles they play during normal vascular remodeling to regulate a previously underappreciated balance between vessel regression and vessel growth. Thus, Angiopoietin-2 expression in the absence of VEGF expression apparently leads to glioma vessel regression, while Angiopoietin-2 expression together with VEGF expression leads to robust glioma-associated angiogenesis (Figure 2).

Throughout glioma development, Angiopoietin-1 expression is not particularly striking. Consistent with observations that C6 cells express low levels of Angiopoietin-1 in culture (unpublished observations and Enholm et al., 1997) Angiopoietin-1 expression was noted diffusely in small C6 gliomas. Lack of further Angiopoietin-1 elevation in later gliomas contrasted with the specific elevation of VEGF in hypoxic regions of the tumor, consistent with in vitro studies showing that, unlike VEGF, Angiopoietin-1 is not upregulated by hypoxia (Enholm et al., 1997). Perhaps the lack of angiopoietin–1 in tumors may help to contribute to the tenuous nature of tumor vessels.

Observations in rat glioma model relate to findings in human glioblastomas

The patterns of angiopoietin and VEGF expression we observed in experimental rat gliomas are correlative to what occurs in a progressing glioblastoma multiforme in humans (Stratmann et al., 1998; Holash et al., 1999; Zagzag et al., 1999). While Angiopoietin-2 is not expressed by normal human brain vessels, it is dramatically induced in the vessels of human glioblastomas. Even relatively normal-appearing glioblastoma vessels with only slight hyperplastic changes show upregulated levels of Angiopoietin-2. As vessels progress through various stages of hyperplasia, and even as they show early sclerotic changes, they continue to express high levels of Angiopoietin-2. Coincident with expression of Angiopoietin-2, vascular profiles can be characterized by abnormal deposition of extracellular matrix and abnormal associations with perivascular cells. Vessels with early hyperplastic changes are associated with fragmented perivascular cell staining. As hyperplasia progresses, deposition of matrix and recruitment of perivascular cells occurs suggesting that vessels are trying to establish appropriate associations with stabilizing factors.

Thus, examination of human glioblastoma revealed that the rat C6 model was quite predictive. Angiopoietin-2 was not detected in normal human brain vessels but was dramatically induced in co-opted tumor vessels, apparently preceding vessel regression that was once again noted to occur in association with a disruption in endothelial cell/smooth muscle interactions and endothelial cell apoptosis; diffuse Angiopoietin-1 expression in the human tumors also resembled that seen in the rat model.

Findings in gliomas are true for other tumors

Implantation of the RBA rat mammary adenocarcinoma into the rat brain revealed a similar progression of tumor growth suggesting that the above findings are generalizable to other types of tumors. (Holash et al., 1999). Instead of growing avascularly, the implanted RBA cells rapidly associated with and migrated along cerebral blood vessels in a manner even more striking than that observed in gliomas. Consistent with the well-vascularized state of these early tumors, there was minimal upregulation of VEGF in early RBA tumors. However, the coopted vessels once again displayed striking and specific upregulation of Angiopoietin-2, which was not detected in the vessels of neighboring normal brain. Furthermore, preliminary analysis of later RBA tumors indicated that Angiopoietin-2 expression is associated with a pattern of vascular regression and angiogenesis which resembles that evident in glioblastomas.

Examination of a model of tumor metastasis, in which mouse lung is colonized by intravenously injected Lewis Lung Carcinoma cells, once again yielded similar results (Holash et al., 1999). Tiny tumor metastases as well as moderately-sized tumor nodules closely associated with pulmonary vessels, and these vessels displayed dramatic induction of Angiopoietin-2 expression. Progressively larger tumor nodules appeared to be characterized by vessel regression as well as new-onset angiogenesis, co-incident with high levels of expression of Angiopoietin-2 and VEGF.

Conclusions and Discussion

Our analyses in several different tumor models suggest a modification of the prevailing view that most malignancies and metastases originate as avascular masses that only belatedly induce angiogenic support. Instead, we suggest that many tumors rapidly co-opt existing host vessels to form an initially well-vascularized tumor mass. Perhaps as part of a host defense mechanism, there is widespread regression of these initially co-opted vessels, leading to a secondarily avascular tumor and massive tumor loss; unfortunately for the host, the remaining tumor is ultimately rescued by robust angiogenesis at its rim (Figure 2). The expression patterns of VEGF and the natural Tie2 receptor antagonist, Angiopoietin-2, strongly implicate them in the above processes. Angiopoietin-2 is strikingly induced in co-opted vessels, prior to VEGF induction in the adjacent tumor cells, providing perhaps the earliest marker of tumor vasculature. The intense autocrine expression of Angiopoietin-2 by endothelial cells in tumor-associated vessels appears to counter a paracrine stabilization/survival signal provided by low level constitutive expression of Angiopoietin-1 in normal tissues. This apparently marks the co-opted vessels for regression by an apoptotic mechanism that seems to involve disrupted interactions between endothelial cells and the surrounding extracellular matrix and supporting cells. Subsequently, VEGF upregulation coincident with Angiopoietin-2 expression at the tumor periphery is associated with robust angiogenesis. This late expression of tumor-derived VEGF may nullify the regression signal provided by Angiopoietin-2; alternatively, the effect of this VEGF may actually be facilitated when vessels are destabilized by Angiopoietin-2. Interestingly, newly formed tumor vessels are often tenuous, poorly differentiated and undergo regressive changes even as blood vessel proliferation continues. The failure of many solid tumors to form a well differentiated and stable vasculature may indeed be attributable to the fact that newly formed tumor vessels continue to overexpress Angiopoietin-2. Thus, a persistent blockade of Tie2 signaling may prevent vessel differentiation and maturation, contributing to the generally tenuous and leaky quality of tumor vessels.

Although our findings appear rather consistent in several rodent tumor models as well as in one human cancer, the relevance of our findings may well depend on the tumor in question. For example, when tumor cells are introduced into a virtual or potential space, as is the case when tumors are deliberately implanted into the vitreous or subcutaneously, tumor cells do initially grow as space-filling avascular masses, which only later recruit vessel ingrowth. Similarly, many epithelial tumors, such as those deriving in the skin which are initially separated from their blood supply in the underlying dermis by a well-organized basement membrane, may have to undergo significant growth before they recruit vascular support. However, when tumors arise in or metastasize to a solid tissue, they may initially associate with and grow preferentially along existing vascular channels.

A great deal of evidence indicates that VEGF plays an important role in tumor angiogenesis, though our findings are the first that put this role in a context that includes the Angiopoietins as well. In tumors, Angiopoietin-2 and VEGF apparently reprise the complementary and co-ordinated roles they play during vascular remodeling in normal tissues, acting to regulate the previously under-appreciated balance between vasculature regression and growth.

Our findings clearly bolster the case for anti-VEGF therapies in cancer. Other studies have suggested that VEGF may not only play a role in inducing angiogenesis, but in promoting the survival of fragile new vessels that have not been stabilized by mature interactions with extracellular matrix and supporting cells, such as smooth muscle. This has been most carefully documented during vascularization of the retina (Benjamin et al., 1998) and more recently in tumors (Benjamin et al., 1999). Thus, anti-VEGF therapies may be considered not only for blocking angiogenesis, but perhaps also for their ability to cause the regression of extant tumor vessels that are immature and tenuous in nature.

Angiopoietin-2 appears to be the earliest marker of blood vessels that have been perturbed by invading tumor cells. As such, Angiopoietin-2 may prove to be useful in the imaging of very small tumors and metastases, and may even be useful in schemes designed to specifically target chemotoxic therapy to tumor vasculature. Anti-Angiopoietin therapies employing soluble Tie2 receptors also have recently been reported to be efficacious in animal models of cancer (Lin et al., 1997, 1998). However, it is somewhat difficult to envision how these soluble receptors might work, as they have the potential to inhibit both Angiopoietin-1 and Angiopoietin-2. One possibility is that soluble Tie2 receptor therapy potentiates the blockade of endogenous Tie2 receptors by endogenous Angiopoietin-2. In addition, immature tumor vessels may be particularly susceptible to blockade of an Angiopoietin-1 mediated stabilization or survival signal, explaining the lack of toxicity of such agents to the vasculature in general. Such a mechanism may also be relevant to other anti-angiogenic therapies. Thus, understanding the mechanisms underlying vessel regression is of great importance, not only in the context of anti-VEGF or anti-Angiopoietin therapies, but for other anti-angiogenic therapies as well. It will be of particular interest to see whether such therapies differentially affect regression of new as opposed to mature, existing vessels, and whether they do this by altering expression of either VEGF or the Angiopoietins.