Why some tumours trigger neovascularisation and others don’t: the story thus far

In 1927, Otto Warburg described what would be called the “Warburg effect,” in which tumour cells exhibited characteristic changes in metabolism, particularly the use of glycolysis rather than oxidative phosphorylation, despite the presence of adequate amounts of oxygen [1, 2]. Warburg believed that this process was the actual cause of neoplastic transformation [3].

Tumour development is now known to be driven by genetic damage. However, mutations in some metabolic enzymes, such as succinate dehydrogenase (SDH) and fumaratehydratase (FH), both parts of the tricarboxylic acid (TCA) cycle, can drive neoplastic transformation, and intermediate products of metabolism can also promote neoplastic progression [4].

Proliferating cancer cells have a high energy requirement to maintain homeostatic cellular processes. The shift in energy production to aerobic glycolysis, while allowing for more rapid production of adenosine triphosphate (ATP), yields far less energy than oxidative phosphorylation: there are two net molecules of ATP per glucose molecule in glycolysis versus 36 molecules of ATP via oxidative phosphorylation [5].

The reasons for the glycolytic energy dependence of proliferating tumour cells are still being debated. Initially, it was believed that the mitochondria in tumours were intrinsically defective. However, it was determined that tumour mitochondria are actually functional, retaining the capacity for oxidative phosphorylation and consuming oxygen at similar rates to normal tissues [6], although it should be appreciated that a degree of variability in mitochondrial activities exists across different neoplasms. Alternatively, high rates of glycolysis might be co-selected with factors that promote the expression of hypoxia-related genes (such as those required for angiogenesis) as an oxygen-independent energy source. Finally, increased intermediate products of glycolysis can easily be shunted into the biosynthetic pathways required for serine and nucleotide synthesis [7].

According to Folkman’s original theory [8], the onset of hypoxia in tumour triggers angiogenesis, which in turn is essential for supplying neoplastic cells with nutrients and oxygen and evacuating metabolic waste and carbon dioxide. The best understood hypoxia signaling mechanism is the stabilization and post-transcription activation of the hypoxia-inducible factor (HIF) proteins, which lead to the activation of many different pathways, including the vascular endothelial growth factor (VEGF) pathway. The VEGF pathway prompts and supports neoangiogenesis and glycolysis. Hypoxia-inducible pathway activation also has other effects, which include reducing the activity of mammalian target of rapamycin (mTOR), which in turn can reignite autophagy and promote survival under stress [9].

HIF is a heterodimer of an alpha subunit that is unstable in normoxia and a constitutively present and stable beta subunit. The hypoxia activation of HIF causes the heterodimer to bind to DNA at specific locations, called hypoxic response elements (HREs), eliciting the transcriptional up-regulation of genes required to respond appropriately to hypoxia [9].

In addition to triggering the VEGF pathway, the ubiquitously expressed HIF1 isoform promotes the transcription of glucose transporter 1 (GLUT1), which activates glucose transport inside the cell, lactate dehydrogenase-A (LDH-A), which is involved in the glycolytic pathway, erythropoietin (EPO), which enhances erythropoiesis, and nitric oxide synthase (NOS), which promotes angiogenesis and vasodilatation [9].

HIF1 also prevents the entry of pyruvate into the TCA cycle by inducing the expression of pyruvate dehydrogenase kinase 1 (PDK1), thus altering the expressed isoform of cytochrome c and inhibiting mitochondrial biogenesis. This process causes reduced levels of oxygen consumption and a shift away from oxidative phosphorylation. Interestingly, HIF1 can also be activated under normoxic conditions by a variety of oncogenic pathways, such as phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha (PIK3CA), and by mutations in von Hippel-Lindau tumour suppressor (VHL), SDH, and FH [10].

In the classic angiogenic pathway, VEGF binds to VEGF receptor 2 (VEGFR2) on endothelial cells, increasing the expression of the Notch ligand Delta-like 4 (DLL4) on the same cells. DLL4 then binds to its receptor Notch on the adjacent endothelium. Further expression of VEGFR2 and VEGFR1, as well as a smaller amount of VEGFR3, then follows, leading to triggering/amplification of the downstream phospholipase C familyγ (PLCγ)–protein kinase C (PKC)–Raf kinase–MAP kinase-ERK kinase (MEK)–mitogen-activated protein kinase (MAPK) pathway, concomitantly prompting cell proliferation and cell survival throughout the phosphoinositide 3-kinase (PI3 K)/protein kinase B (AKT) pathway [11].

The switch to glycolysis in neoplasia was, according to Warburg, irreversible [3], yet a more complex picture has emerged over the last decade. There have been observed instances in which oxidative phosphorylation predominates during neoplastic transformation [12]. This variation between OxPhos and glycolysisin cancer cells has been increasingly linked to specific disturbances in cell signaling pathways [13].

Additionally, tumours of the same genetic lineage can develop different metabolic adaptations depending on the host tissue from which they arise, suggesting that the stromal environment might play a crucial role in shaping the metabolic profile [14]. The different molecular mechanisms being postulated to explain this variability of the Warburg effect include the following: inhibition of pyruvate dehydrogenase (PDH) by PDK1, reduction of mitochondrial biogenesis and inhibition of oxidative phosphorylation, both are caused by P53 inactivation and mutations [15].

Warburg raised two important issues: first, how tumour cells are supplied with glucose; and second, how they are supplied with oxygen [1]. Folkman’s work addressed the latter question with the hypothesis that tumour growth is strictly angiogenesis-dependent [16]. The work undertaken to test this hypothesis led to the inclusion of “angiogenesis” as one of the hallmarks of cancer [8].

Although there is strong evidence that angiogenesis frequently occurs in cancer, we also now know that this event does not always occur. Indeed, some tumours, called “non-angiogenic tumours,” can grow without triggering new vessel formation by co-opting preexisting vessels [17, 18].

Non-angiogenic growth was first identified by histology in primary and metastatic lung carcinomas because neoplastic cells filled the alveolar spaces, co-opting the pre-existing capillary network and exhibiting a characteristic “chicken-wire” appearance [17]. A gene expression signature for non-angiogenic non-small cell lung cancer (NSCLC) was published in 2005 [19]. Surprisingly, rather than the classic angiogenesis-related genes, the differentially expressed genes were involved in mitochondrial metabolism, transcription, protein synthesis, and the cell cycle. Lack of differential mRNA expression between tumour phenotypes was noted for genes classically associated with hypoxia and angiogenesis. This result suggested that the response to hypoxia does not necessarily trigger neovascularisation, as would be observed in angiogenic tumours, but could actually be dependent on the genetic background of neoplastic cells, and in some instances, it could lead to metabolic reprogramming [19]. We therefore postulated that the degree to which a tumour will rely on angiogenic or non-angiogenic growth could be associated with a variety of events, including hypoxia, pseudo-hypoxia, and metabolic re-programming.

In the first part of the present study, we investigated whether there were truly no differences in the expression of hypoxia-and angiogenesis-related proteins between angiogenic and non-angiogenic tumours, as suggested by mRNA profiling. We also investigated the degree of these proteins expression and the expression of some mitochondrial biogenesis proteins via immunohistochemistry. Notably high cytoplasmic P53 expression in non-angiogenic tumours, compared to angiogenic tumours, was found after completing the first part of this study. We therefore performed a second investigation, in which we examined and sequenced the p53 gene in these tumours.

As initially suggested by our mRNA profiling work [19] and also confirmed by the immunohistochemical data presented here, we failed to reveal significant differences between angiogenic and non-angiogenic NSCLCs as far as the expression of proteins associated with the classic hypoxia/angiogenesis pathway is concerned. Because neovascularisation is found in some tumours but not in others, we suggest that activation of the classical angiogenic pathways is necessary, but not sufficient, to induce the sprouting of new vessels in cancer.

The higher levels of expression of proteins associated with extracellular matrix, cell adhesion, and inflammation were in agreement with the observed presence of tumour-associated stroma and chronic inflammation in many angiogenic tumours [18, 22]. mRNA profiling of these tumours has equally shown that stromal remodeling, cell adhesion, and inflammation are enhanced in angiogenesis [19]. Both mRNA and immunohistochemical data suggest a crucial role in angiogenic cancers for FBJ murine osteosarcoma viral oncogene homolog (FOS), a protein involved in cell proliferation, remodeling, and inflammation. The question remains of whether tissue remodeling is a consequence or a cause of the triggering of angiogenesis. Because non-angiogenic tumours usually preserve the pre-existing architecture, we hypothesize that the triggering of tissue destruction could be a secondary step necessary for the activation of angiogenesis. Nonetheless, this process would be far from a general rule because we have observed, in a subset of NSCLC, that angiogenesis can occur in the absence of tissue destruction [23]. Clearly, other mechanisms must exist.

If classical angiogenesis pathways are similarly active in both angiogenic and non-angiogenic tumours, how does their biology differ? Our previous transcriptional profiling work demonstrated higher levels of mRNA coding for molecules associated with oxidative phosphorylation and mitochondrial biogenesis in non-angiogenic tumours [19]. In the present study, we did not examine components of the oxidative phosphorylation pathway, although we did investigate the expression of some proteins involved in mitochondrial functions. Our results showed that proteins related to mitochondria and to cell biogenesis-promoting processes were more highly expressed in non-angiogenic tumours than in angiogenic tumours. This feature was consistent with higher mitochondrial regulatory activity and a possible metabolic switch in non-angiogenic tumours. Again, whether this switch is a cause or an effect of HIF activation remains unclear.

Interestingly, our data suggested increased involvement of response to hypoxia and to HIF regulation by oxygen in non-angiogenic tumours. We hence speculate that, in these tumours, although the HIF pathway failed to induce new vessel formation, it could well be involved in metabolic reprogramming.

Finally, we noted markedly higher levels of cytoplasmic P53 expression in non-angiogenic tumours than in angiogenic tumours via immunohistochemistry. In a pilot study, we sequenced p53 in a limited number of angiogenic and non-angiogenic NSCLCs. Non-angiogenic tumours had a higher incidence of mutations, which were all missense mutations, whereas angiogenic tumours had an amalgam of frame shifts and missense and nonsense mutations. Because p53 also affects mitochondrial respiration [24], it will be necessary to investigate further how the observed p53 mutations could functionally affect its ability to regulate respiration and/or angiogenesis.

On the basis of our observations collected so far, from the mRNA profiling, immunohistochemical and histopathologic data, we conclude that all tumours, angiogenic and non-angiogenic, experience hypoxia/HIF and VEGF pathway activation. However, angiogenesis does not always ensue. Based on these findings, we suggest that in non-angiogenic tumours, HIF pathway activation could be the driving force toward metabolic reprogramming.