Evidence Showing That Tumors Can Grow Without Angiogenesis and Can Switch Between Angiogenic and Nonangiogenic Phenotypes

Although sorafenib is approved for the treatment of hepatocellular carcinoma (HCC), the ability of this anti-angiogenic tyrosine kinase inhibitor to extend survival in HCC patients is limited because of acquired drug resistance, whose mechanism is poorly understood. In this issue of the Journal, Kuczynski and colleagues (1) report a mechanism of acquired resistance to sorafenib using an orthotopic HCC xenograft model. The authors tell us how, although these HCC are initially angiogenic and respond to sorafenib, resistance occurs because the neoplastic cells switch to a non-angiogenic phenotype and actively co-opt the normal liver vasculature instead of inducing angiogenesis. These results have important implications for understanding this particular form of resistance to anti-angiogenic drugs and the biology of the tumors that can develop such a resistance.

The non-angiogenic growth of tumors has already been hypothesized to be one of the possible reasons for intrinsic and/or acquired cancer resistance to anti-angiogenic treatment (2). In this paper, it is now proven that an angiogenic malignancy can, following treatment with a drug inhibiting sprouting angiogenesis, switch to a non-angiogenic phenotype in order to keep growing.

This finding further supports the notion that neoplasia is not necessarily angiogenesis dependent as some tumors can be completely non-angiogenic while many more can present with a mixture of both angiogenic and non-angiogenic regions (3). Furthermore, it had been already described that the angiogenic status of a tumor is not an absolutely fixed characteristic but can change: eg, primary angiogenic breast carcinomas can relapse as non-angiogenic lung metastases (4) while non-angiogenic primary lung cancers can progress with angiogenic brain secondary (5). Now, the findings of Kuczynksi and coworkers show that a tumor can switch from angiogenic to non-angiogenic growth in response to treatment with an anti-angiogenic drug. Taken together, these observations demonstrate that the angiogenic status of a tumor is not an absolutely fixed characteristic, but can change.

This is in sharp contrast to the concept, first enunciated in 1971 by Judah Folkman (6) and subsequently included among the hallmarks of cancer (7), that tumor growth is always angiogenesis-dependent as “any increase in tumor population must be preceded by an increase in new capillaries converging on the tumor” (8).

Non-angiogenic tumors were first described in histopathology studies of primary non–small cell lung carcinomas (NSCLCs) and carcinoma metastases, from a variety of primaries, in the lung as these tumors fill the air spaces without destroying the parenchyma and co-opt the pre-existing alveolar vessels (9). Non-angiogenic human neoplasms were subsequently reported to occur as brain tumors (10), liver metastases growing by replacing the haepatocytes (11,12) or by colonizing the hepatic sinusoids (13) and lymph node metastases (14,15). Mouse models have now also been established for non-angiogenic malignant growths in brain (16,17) and lung (18).

The demonstration that some tumors do not induce angiogenesis is achieved on the grounds that in these lesions the only vessels present have the distribution and the phenotype of normal vessels and the overall architecture of the organ is preserved (10–12,19,20). This evidence now includes also a recent intravital microscopy study in an animal model (15). A second type of tumor growing without inducing sprouting angiogenesis and without co-opting vessels has also been discovered. This one exploits vascular mimicry, in which the neoplastic cells themselves assemble into channels that act like vessels (21).

The angiogenic NSCLC can also be told apart from non-angiogenic ones by gene transcription analyses. An investigation of differentially transcribed genes in these tumors suggested that metabolic reprogramming, with a switch toward oxidative phosphorylation under the control of selected heat shock proteins, oncogenes, and tumors suppressor genes, could be one of the key issues (22,23).

Another noteworthy piece of information is that the transcription and expression of genes associated with hypoxia and angiogenesis seems to be comparable in both angiogenic and non-angiogenic tumors (15,22,24). Therefore, the sensing of hypoxia and expression of VEGF protein by cancer cells do not invariably lead to new vessel formation.

Only a few studies so far have started to address the mechanisms of vessel co-option. Observations in animal models of brain malignancies indicate that vessel co-option is an active process that involves selected cell adhesion molecules and protection from apoptosis for the cell that is successfully co-opting a vessel (25–28).

All these results have therefore established the study of non-angiogenic tumors as a new field in cancer biology, but so far we have only just scratched the surface. So what now are the most pressing questions we face?

One is: what is the biology underlying the non-angiogenic phenotype, and what are the mechanisms that allow a cancer cell to switch between angiogenic and non-angiogenic phenotypes?

Second, how do the cancer cells interact with and co-opt the pre-existing vessels? Data published so far suggests that this is an active process (26,29). Moreover, how does this process change in different organs?

Third, does vessel co-option also facilitate resistance to other classes of anti-angiogenic drugs, such as the VEGF-neutralizing antibody bevacizumab?

Fourth, given that vessel co-option occurs in many human cancers, including some of the most prevalent (eg, malignancies from breast, colon, rectum, and lung), is vessel co-option also a mechanism of resistance to anti-angiogenic therapy in humans and not just in animal models?

Last, but not least, can vessel co-option be inhibited with drugs? The data of Kuczynski and colleagues shows that the response to sorafenib in HCC might be more durable if sorafenib were to be combined with a drug that targets vessel co-option. However, there are currently no drugs designed to target vessel co-option in humans. In our opinion, this represents a major deficiency in the current portfolio of oncology drugs and needs to be addressed urgently.

Hopefully, this new field of cancer biology will lead to novel therapeutic interventions designed according to the relationship observed between neoplastic cells and vessels in tumor lesions, in the knowledge that tumors can also grow without inducing angiogenesis.