Establishing a Link between Oncogenes and Tumor Angiogenesis

The progressive growth and expansion of solid tumors beyond microscopic sizes of about 1–2 mm in diameter requires the formation of new blood vessels—a process known as tumor angiogenesis. Two major types of event are thought to be involved in the ability of tumors to switch on the angiogenic phenotype: a gain-of-function event in which various stimulators of angiogenesis are induced, or up-regulated, in tumor cells, and a loss-of-function event involving down-regulation of one or more endogenous inhibitors of angiogenesis (1). Activation of oncogenes and inactivation of tumor suppressor genes are generally associated with gain- and loss-of-function events, respectively (2). Hence, these types of genetic alterations could be one of the main ways in which the process of tumor angiogenesis is switched on and sustained. This has potentially important therapeutic implications for the use of drugs designed to target and inhibit the overex-pressed proto-oncogenes or mutant oncogenes or their encoded oncoproteins. In short, such drugs may function in vivo, at least in part, as inhibitors of tumor angiogenesis—an anti-tumor effect that would be missed in tissue culture assays used for drug development and activity.

Positive and Negative Regulators of Tumor Angiogenesis

Folkman first put forward the hypothesis that the growth of solid tumors is “angiogenesis dependent” (3,4), which is the basis for anti-cancer treatment strategies aiming to develop drugs that selectively target newly formed, immature blood vessels in tumors while leaving normal mature vessels elsewhere in the body unharmed (3,4). This hypothesis has stood the test of time, at least in preclinical models of tumor growth and therapy (5,6). There is now great interest in academic research laboratories and in the biotech/pharmaceutical industry in the development of drugs that inhibit or enhance angiogenesis. There are a number of reasons for this interest. (1) The possibility exists that resistance to certain angiogenesis inhibitors may not develop in tumor cells exposed to such drugs, even over prolonged periods of time (7–9). (2) A growing number of molecular targets on “activated” endothelial cells have been identified that are associated with newly formed blood vessel capillaries, such as acutely up-regulated receptor tyrosine kinases (10–12), integrins (13), and adhesion molecules (14,15). (3) A number of potentially powerful endogenous protein inhibitors of angiogenesis that can cause tumor regressions (16) and are usually internal fragments of higher molecular weight proteins have been discovered (16). (4) There is a growing sense of urgency about the need to devise new and innovative anti-cancer strategies or drugs to replace or supplement the ones that have been used with such limited success for the past 50 years.

When Folkman first put forward his seminal hypothesis more than 25 years ago, he initially envisioned a scenario in which developing tumor masses remained microscopic and dormant as long as they were incompetent to induce angiogenesis (3,4). He suggested that termination of this dormant phenotype was brought about by the induction and release of soluble and diffus-able growth factor, which he called tumor angiogenic factor (TAF) (3,4). Release of this hypothetical molecule into the extracellular environment would set in motion the various chains of events associated with the formation of new blood vessel capillaries sprouting from the mature and pre-existing host vasculature located in the vicinity of a TAF-producing tumor mass. These events include localized proteolysis of the basement membranes surrounding mature blood vessels, migration of endothelial cells through the newly created breach, endothelial cell proliferation, the formation of rudimentary tubes or vascular sprouts, and the joining of such newly formed vessels to form a vascular network (1,6). It took another 15 years before a molecule that could be equated with TAF was identified. It turned out to be basic fibroblast growth factor (bFGF), also known as FGF-2 (17,18). Since then, at least a dozen more growth factors have been identified as having potential tumor angiogenesis promoting activity (6). These include growth factors also known to have mitogenic autocrine activity for tumor cells, such as transforming growth factor alpha (TGF-α), and a number of tumor- or stromal cell-derived paracrine growth factors (for endothelial cells), including vascular endothelial cell growth factor VEGF, which is also known as vascular permeability factor (VPF); scatter factor (SF), also known as hepatocyte growth factor (HGF); transforming growth factor beta (TGF-β); angio-poietin 1 and 2; platelet-derived growth factor (PDGF); and interleukin-8 (IL-8), among others (6). In addition to these growth factors, whose expression can be induced or up-regulated in tumor cells, there is a family of angiogenesis inhibitory molecules, some of which are now known to be strongly down-regulated in tumor cells. Thrombospondin-1 is perhaps the best example of such an inhibitor (6). Bouck and colleagues first reported evidence that the wild-type p53 gene is a positive regulator of throm-bospondin expression; inactivation of p53 by mutational or deletion events can result in loss of thrombospondin expression (19). Indeed, this led Bouck to propose that a major consequence of inactivation of wild-type suppressor genes would be the facilitation of tumor angiogenesis, primarily by eliminating or reducing the expression of endogenous inhibitors of angiogenesis (20). This would contribute (indirectly) to the tumor-promoting function of such genetic alterations, in addition to their direct effects on enhancing unrestricted cell proliferation and cell survival (6,19).

The Connection of Oncogenes to Tumor Angiogenesis

A survey of some of the growth factors up-regulated in tumor cells immediately suggests a possible role for oncogenes in tumor angiogenesis. For example, both TFG- α and TGF- β, which are proangiogenic molecules in vivo (6), are known to be induced or up-regulated in mutant ras transformed cells (21,22). The same is true for bFGF (23). Two other considerations led us to speculate in 1995 that a major, unappreciated function of oncogenes is to contribute to the angiogenic phenotype in tumors. First, it is known that the growth fraction of many solid tumors is actually quite low (24,25), in some cases surprisingly so, as in rapidly expanding metastatic prostate cancer deposits growing in the bone (26). The low fractions of solid tumors is thought to be a major source of the intrinsic resistance characteristics of such tumors to cytotoxic chemotherapeutic drugs, which generally target rapidly dividing cells (27). This presents an interesting paradox, given the emphasis that has been placed on aberrant cell cycle regulation (“unrestricted cell proliferation”) as the predominant functional effect of most oncogenes on the transformed phenotype. For example, because metastatic prostatic cancers have tumor cell growth fractions in the range of only 2% (26), it is difficult to accept the idea that oncogenes that contribute to prostate cancer do so primarily, if at all, through their effects on directly promoting aberrant cell proliferation. In this regard, it is worth pointing out that when tumor cells are grown in monolayer cell culture, it is not uncommon to encounter growth fractions in the range of 75%, which is far greater than that observed in vivo (27). However, growth in vitro as three-dimensional multicellular spheroids can lead to a very significant reduction in the growth fraction (27), despite the presence of numerous inactivated tumor suppressor genes and mutant oncogenes in the cultured tumor cells. Perhaps the predominant use of monolayer cell culture systems to study cancer biology in vitro has resulted in a somewhat distorted view of the relative importance and contribution of uncontrolled cell-cycle proliferation to the overall growth and expansion of solid tumors in vivo (27). If so, the following question emerges: how do oncogenes contribute to the ability of tumors to grow indefinitely in addition to, or instead of, promoting aberrant cell proliferation? Could an indirect mechanism of growth promotion be involved — namely, induction of or contribution to angio-genesis?

A second paradoxical observation that caught our attention concerns the therapeutic effects of a new class of anti-cancer drugs generally known as signal transduction inhibitors. The best known examples of such drugs include small-molecular-weight Ras farnesyltransferase inhibitors (FTIs) (28,29) and monoclonal neutralizing antibodies to overexpressed cell-surface receptor tyrosine kinases, such as the EGF receptor or erbB2/HER-2/neu (30,31). In general, such drugs are considered to be cytostatic rather than cytotoxic agents because they inhibit growth of target tumor cells in (monolayer) cell culture in the absence of any significant killing (28,29,32). Moreover, the cytostatic effects are often modest —in the range of 30–60% inhibition at the highest drug concentrations (28,29,32). Consequently, a reasonable prediction would be that such drugs possess only modest anti-tumor effects in vivo when tested on established tumors, i.e., overt regressions of tumor mass would not be a feature of treatment with such drugs. Rather, tumors would be kept from expanding by an induced state of dormant growth. Surprisingly, this is not necessarily the case (32–34). For example, regression of established experimental tumors in mice has been observed, in some cases, by using Ras FTIs on certain transgenic “oncomouse” strains (33). Moreover, the extent and rapidity of tumor regressions can match or even exceed maximum tolerated doses of conventional cytotoxic anticancer drugs such as adriamycin (33).

How can such an unexpected discrepancy be accounted for? One possible explanation is that the agents may be found to be cytotoxic when tested against tumor cells grown in a solid tumor (multicellular) context, rather than in monolayer cell cultures, which is the usual way tumor cells are grown for drug-testing studies in cell culture. Indeed, there is evidence that ras oncogenes can function as potent survival factors by suppressing the massive levels of apoptosis of epithelial cells observed when such cells are grown nonphysiologically as multicellular spheroids (35). This probably explains why Ras FTIs can induce apoptosis in ras-transformed cells grown anchorage independently, but not in monolayer cell culture (36), and as it now turns out, in cells grown in vivo, as established solid tumors (37).

However, an alternative or additional mechanism of cell killing by such signal transduction inhibitors could involve inhibition of tumor an-giogenesis, as we first suggested in 1995 (38) and as schematically outlined in Figure 1. If it is supposed that activation of an oncogene such as ras leads to a marked induction in tumors cells of a paracrine-acting angiogenesis growth factor such as VEGF/VPF, treatment of such cells with a drug such as a Ras FTI could result in down-regulation or suppression of that tumor cell-derived angiogenesis factor. This in turn could endow the drug with potential anti-angiogenic properties, which could then lead to an increase in tumor cell apoptosis, because blocked angiogenesis is often associated with an increase in the levels of apo-ptotic cells detected in tumors (39,40). Moreover, even modest (2- to 3-fold) reductions in VEGF/VPF expression can be associated with profound degrees of tumor growth inhibition (41) as well as endothelial cell apoptosis (42,43), leading to blood vessel destruction (42,44). The latter could lead to the former (42).

Fig. 1

Outline of how oncogenes and signal transduction inhibitor drugs may contribute to induction and inhibition of tumor angiogenesis, respectively.

Oncogene (e.g., ras) activation can lead to induction of VEGF/VPF, a potent mediator of angiogenesis. VEGF/VPF cannot function as an autocrine growth factor for tumor cells as tumor cells generally lack receptors for VEGF/VPF. In contrast, activated endothelial cells can express high levels of VEGF/VPF, perhaps because of the inductive effect of VEGF/VPF itself. Treatment of the VEGF/VPF-positive tumor cells with a signal transduction inhibitor e.g., a Ras farnesyltransferase inhibitor (FTI), can lead to, among many other changes, a reduction in VEGF/VPF expression. This could in turn lead to a suppressed in vivo angiogenic response. Oncogene activation could also lead to expression of growth factors having both autocrine and paracrine/angiogenesis-promoting functions (such as TGF- α) and/or down-regulation of angiogenesis inhibitory molecules, such as thrombospondin (see text).

In an effort to test these hypotheses, we initiated a series of experiments designed to determine first, whether mutant ras oncogenes, and subsequently, other oncogenes, could act as inducers of VEGF/VPF gene and protein expression. Assuming such an association could be uncovered, we could then test the effects of various signal transduction anti-tumor inhibitory drugs for their effects on VEGF/VPF expression both in vitro and in vivo against appropriate target tumor cells. Our first experiments utilized IEC-18 cells, which is a spontaneously immortalized cell line of rat intestinal epithelial origin totally incapable of forming tumors in nude mice (38). A number of clonal ras-transformed sublines were obtained by transfection of IEC-18 cells with a mutant human H-ras oncogene (38), all of which were found to be highly tumorigenic in nude mice, e.g., the IEC-ras3, IEC-ras4, and IEC-ras7 cell lines (38). As shown in Figure 2, the parental IEC-18 cell line was essentially negative for VEGF/VPF gene expression, as assessed by Northern blotting experiments (38). In marked and obvious contrast, the ras-transformed sublines were all strongly VEGF/VPF positive. This included a clone in which mutant H-ras was put under the control of heavy metal inducible promoter —so long as the cells were exposed to zinc and cadmium (38), as shown in Figure 2. Similarly, IEC-18 cells transfected with a tetracyclin-regulated ras gene expression construct were VEGF/VPF mRNA positive in the absence of tetracyclin and negative in its presence (unpublished observations). At the time our results were published, Grugel and colleagues reported a virtually identical pattern of results using v-ras (and v-raf) transformed NIH 3T3 fibroblasts (45). Since then, a number of confirmatory reports have appeared showing an association between ras expression, which is often, but not always, mediated by mutant ras oncogene transfection, and induction or up-regulation of VEGF/VPF gene expression in mouse, rat, hamster, and human cells of variable origin (46–54). This is summarized in Table 1. In virtually all of these experiments, the increase in mRNA expression was matched by a commensurate increase in protein expression.

Table 1 Mutant ras oncogene induction or up-regulation of VEGF/VPF

Fig. 2

Mutant ras oncogene regulates expression of VEGF mRNA.

(A) Northern blot of VEGF mRNA is shown of human HT 1080 human fibrosarcoma cells (as a positive control), nontumorigenic rat intestinal epithelial cell line-18 (IEC-18 cells), and two clones of IEC-18, Ras-7 and Ras-3, which were obtained by transfection of IEC-18 cells with a mutant human H-ras oncogene. These latter two lines are tumorigenic, and unlike IEC-18 cells, they express abundant VEGF transcripts. MT-ras is a clone of IEC-18 which expresses its transfected mutant H-ras oncogene only when exposed to heavy metals such as zinc and cadmium, as the gene is under the control of a metallothionein promotor. VEGF mRNA are not expressed in this clone unless they are exposed to these metals. (B) Northern blot for VEGF mRNA in human DLD-1 and HCT-116 human colorectal carcinoma cells, each of which contains a single mutant K-ras allele. DKS-8 and Hkh-2 are sublines obtained from DLD-1 and HCT-116, respectively, in which the mutant K-ras allele has been disrupted by gene targeting (55). The DLD-1 and HCT-116 cell lines are tumorigenic in nude mice, whereas DKS-8 and Hkh-2 are not. The suppression of VEGF mRNA is matched by a down-regulation in VEGF protein released by the cells into the conditioned medium. The results are taken from Rak et al. (38).

An alternative and complimentary method to demonstrate a cause-and-effect relationship between oncogenic ras mutations is to examine VEGF/VPF expression in VEGF/VPF-positive human colorectal carcinomas that carry a single mutant K-ras allele, and sublines of such tumors in which the dominantly acting mutant K-ras allele is disrupted by gene-targeting methods (38). Such sublines have been obtained by Shira-sawa et al. from the highly tumorigenic HCT-116 and DLD-1 human colorectal carcinomas (55). It is remarkable that the sublines containing a disrupted K-ras allele were found to be nontumori-genic in nude mice (38,44,55) despite the retention of the numerous other genetic alterations normally associated with, and presumed to be causative of, advanced colon cancer. Could this profound loss in tumorigenicity be related to a marked suppression of VEGF/VPF expression in the mutant K-ras knockout sublines? We have speculated that the answer to this question is an affirmative one, based on several findings (44). First, the knockout sublines were found to express an approximately 4-fold reduction in VEGF/VPF mRNA (see Fig. 2) and protein levels (44). Second, a similar reduction in VEGF/VPF expression mediated by transfection of HCT-116 or DLD-1 cells with a VEGF₁₂₁ antisense cDNA expression construct resulted in the derivation of several clones from each parent line which were suppressed 2- to 4-fold for VEGF/VPF protein expression (44). All of these clones were profoundly suppressed in their tumor-forming ability in nude mice, grow normally in cell culture. Third, when VEGF/VPF-deficient mutant K-ras knockout sublines were used as recipients for a VEGF/VPF₁₂₁ “sense” transfection procedure, a number of the VEGF/VPF-expressing variants showed a weak but detectable increase in tumor-forming ability in nude mice but no growth advantage in tissue culture (44). Taken together, these results suggest that ras oncogene-induced VEGF/VPF expression is necessary, but clearly not sufficient, for aggressive tumorigenic growth in vivo. This conclusion would appear to make intuitive sense because knocking out the mutant ras allele would lead to suppression of numerous and different pro-cell transformation events involving, for example, growth, invasion, and survival, in addition to angiogenesis. Restoring some degree of VEGF/VPF expression would not affect these other vital, transformed cell-associated phenotypes. Indeed, even the degree of angiogenesis competence that is restored by a VEGF₁₂₁ transfection method might be an underestimate, since other VEGF/VPF isoforms (e.g., VEGF₁₆₅) and additional pro-angiogenic growth factors (e.g., bFGF, IL-8, etc.) suppressed in the knockout sublines would not have their levels restored by transfection of a sense VEGF/VPF₁₂₁ cDNA expression construct (44).

More recently, we have found that other genetic alterations that effectively function as dominant oncogenes, e.g., overexpression of the EGF and erbB2/neu/Her2 receptor tyrosine kinases, are also associated with induction or up-regulation of VEGF/VPF mRNA and/or protein expression (56). This is consistent with the previous results of some other studies showing, for example, that TGF- α or IGF-1 can induce VEGF/VPF expression in vitro (57,58). It is also now known that other classes of oncogenes, e.g., genes that encode transcription factors such as c-fos (46) or protein translational initiation factors such as ElF-4e (59), can function as potent inducers of VEGF/VPF expression both in vitro and in vivo (46). Thus a generic function of many oncogenes may be to promote tumor growth and survival indirectly through an angio-genesis-dependent mechanism, as well as directly through their effects on enhancing intrinsic tumor cell proliferation and survival. This suggests that various signal transduction inhibitors being developed as anti-cancer drugs may function in vivo as de facto anti-angiogenic agents.

Signal Transduction Inhibitors as Anti-Tumor Agents: Do They Inhibit Tumor Angiogenesis In Vivo?

This is a difficult question to answer in a definitive manner, but the evidence obtained thus far would appear to indicate that one possible effect of administering a variety of signal transduction inhibitors to tumor-bearing mice would be to suppress VEGF/VPF expression and in all probability, a number of other pro-angiogenic growth factors as well, thereby endowing such drugs with the potential to block or suppress tumor growth made of tumor cells by inhibiting the angiogenesis competence of the treated tumors.

Our first attempt at analyzing this question involved an examination of the effects of Ras FT1 called L-739, 749 (38) on VEGF/VPF expression using VEGF/VPF cultured ras-transformed IEC-18 cells as a target population in vitro (38). The results showed that one effect of drug treatment of such cells in vitro was down-regulation of VEGF/VPF expression (38). We have not yet determined whether a similar effect of the drug would be observed in vivo in drug-treated tumor-bearing mice.

However, such an in vivo effect has been observed with a different class of anti-tumor signal transduction inhibitory agent, namely, monoclonal neutralizing antibodies to the human EGF receptor (56). We had found that human A431 squamous carcinoma cells, which overexpress the EGF receptor, displayed reduced (up to 2-fold) levels of VEGF/VPF mRNA and protein in vitro after treatment in culture with varying concentrations of the EGF receptor neutralizing antibody, which is known as C225 (32). We therefore attempted a similar experiment with A431 squamous carcinoma cells grown as subcutaneous xenografts in nude mice (56). For these experiments, nude mice were injected with A431 cells and the tumors allowed to grow for approximately 3–4 weeks. One group of mice was injected four times with C225 intraperitoneally (one injection every 2 days) and the tumors were removed shortly after the last injection. They were assessed for their relative expression of VEGF/VPF protein by immunostaining, in comparison to the control mice. The results showed a rather striking down-regulation of VEGF/VPF expression in the A431 tumors removed from mice that had been given the C225 antibody (56). The level of reduction could not be quantitated with accuracy but appear to exceed 3-fold (56). As discussed above, this level of suppression, when induced by an antisense method, can have profound suppressive consequences on tumor growth in vivo in the absence of any anti-tumor effect in cell culture (41,44).

Thus, it would seem reasonable to postulate that long-term therapy of tumors with an agent such as C225 could lead to an indirect mechanism of suppression of tumor growth as a result of blocked angiogenesis. This may be true of many drugs designed to inhibit the expression of different classes of mutant or overexpressed oncoproteins. If so, it could partially explain why such drugs appear to be more potent in vivo than one would anticipate from their behavior as antitumor drugs on cells grown in monolayer cell culture where generally only modest and non-cytotoxic effects are observed. In this respect, it would be of interest to determine whether the putative anti-angiogenic effects of such agents can be separated from their direct anti-proliferative effects on tumor cells. Are the drug concentrations and scheduling that are optimal for inhibiting (directly) tumor cell growth (or survival) the same as those for inhibiting angiogenesis?

Oncogenes (and Tumor Suppressor Genes) and Angiogenesis Inhibitors

The emphasis in this discussion has been on the idea that oncogenes can contribute to tumor angiogenesis primarily by virtue of their stimulatory effects on the expression of pro-angiogenic growth factors such as VEGF/VPF. However, as discussed earlier, the angiogenic switch is also affected by the loss of angiogenesis inhibitors such as thrombospondin-1 as a result of inactiva tion of the p53 gene (19). It is conceivable that oncogenes could contribute to the angiogenic switch by causing a similar down-regulation of various angiogenesis inhibitors. Indeed, several groups have reported that the levels of throm-bospondin in ras oncogene-transformed fibroblasts can be strongly suppressed (60,61). Likewise, inactivation of the von Hippel Lindau suppressor gene can lead to a marked induction or up-regulation of VEGF/VPF gene and protein expression (62,63). Hence it is becoming clear that the ways in which mutant or deleted tumor suppressor genes and oncogenes can influence angiogenesis is not only by loss of angiogenesis inhibitors and induction of angiogenesis stimulators, respectively.

Interaction of Oncogenes with Physiologic Regulators of Tumor Angiogenesis

One important and potent mediator of VEGF/VPF expression both in vitro and in vivo is reduced oxygen concentrations, i.e., hypoxia (64,65). This effect of hypoxia is mediated both by a transcriptional effect and an increase in mRNA stability (66); the latter seems to be the more important (66,67). This has led to the view that physiologic stresses such as hypoxia in solid tumors may be the major inducing influence of angiogenesis rather than genetic changes per se (68). It is becoming increasingly evident, however, that a combination of genetic and epigentic (i.e., hypoxia) changes can function together in a synergistic manner to boost VEGF/VPF expression in tumor cells (47,52,56). This effect, as observed in mutant neu oncogene-transformed NIH 3T3 fibroblasts, can be quite dramatic (56). The signalling pathways that are involved in this interaction are now being analyzed (47,50). For example, activation of PI3 kinase, but not MAP kinases, has been implicated in combined oncogene-hypoxia induction of VEGF/VPF in ras-transformed fibroblasts (50).

Summary

We have tried to stress that mutant oncogenes or overexpressed, nonmutated proto-oncogenes, in addition to their direct affect on promoting aberrant tumor cell proliferation (and survival), may possess a crucial indirect means of stimulating tumor cell growth through regulation of angiogenesis. This effect would never be observed in tissue culture studies of oncogene function using pure cultures of tumor cells, which probably helps explain why the pro-angiogenic function of oncogenes has not been appreciated until only relatively recently. Indeed, the very first indication of a possible contributory role of oncogenes, such as ras and myc, to tumor angiogenesis was first reported by Thompson et al. in 1989, who used reconstituted organ cultures of the mouse prostate gland for their studies (69). This potentially important contribution of oncogenes to tumor growth and development may prove to have an impact on how various signal transduction inhibitors that are now in early phase clinical trials, e.g., monoclonal neutralizing antibodies to the human EGF receptor (70), function in vivo as anti-tumor agents.