When cells are starved of oxygen, they usually die. This is why numerous anti-cancer treatments aim to prevent the growth of blood vessels in tumours, thereby cutting off their oxygen supply. But, in the latest addition to a long list of lessons we are learning about the insidious nature of cancer, it seems that tumours may in fact turn a lack of oxygen to their advantage. Pennacchietti and colleagues1, writing in Cancer Cell, have uncovered a molecular pathway that is switched on by low oxygen levels, and which could cause cancers to become more aggressive and invade surrounding tissues.

Normal tissues receive a constant supply of oxygen from oxygenated haemoglobin molecules, carried by a continuous flow of blood. When such tissues are subjected to oxygen starvation (hypoxia) — because of a reduction either in blood flow or in the oxygen content of the blood — the eventual result is cell death. Depending on their individual metabolic requirements, some tissues can survive a hypoxic state for longer than others; for example, brain tissue can survive if hypoxia is limited to only a few minutes. To ensure survival, tissues react in two ways: they switch into a protective mode by using a specific set of hypoxia-sensing proteins called hypoxia-inducible factors, or HIFs; and they produce 'angiogenesis' proteins that will attract new blood vessels as a way of restoring local blood flow.

Hypoxia has found a place on both sides of the war on cancer. On the one hand, the idea that depriving a primary tumour of essential nutrients and oxygen will stop its growth and spread has won wide appeal. When tumours reach a certain size they outgrow their blood supply, and various studies have linked the resulting hypoxia, via HIFs, to the production of vascular endothelial cell growth factor and consequent stimulation of blood-vessel growth (Figs 1, 2). Following these findings, advocates of anti-angiogenesis strategies have looked for ways to block this molecular cascade by targeting the constituent signalling molecules or the responsive blood-vessel cells (endothelial cells)2,3. On the other hand, several clinical studies4 have shown that the presence of hypoxic regions within tumours correlates with poor prognosis and an increased risk of spread to other parts of the body (metastasis), irrespective of treatments used. These findings have cast a different light on tumour starvation as a therapeutic strategy.

Figure 1: Short of oxygen.
figure 1

The left half of this section of a human tumour had a normal oxygen supply; the right half did not, and is dying. The cells in the right half are shrivelled and fragmented.

Figure 2: How tumours cope with — and can benefit from — a lack of oxygen.
figure 2

a, At this stage, the growing tumour is well supplied with oxygen and nutrients by its own blood supply. b, The tumour has now outgrown its blood supply and cells at the front line are becoming hypoxic. However, this need not always mean cell death, as is shown on the right. Hypoxia-inducible factors (HIFs) sense the low oxygen levels and stimulate the production of vascular endothelial growth factor (VEGF), a protein that attracts new blood vessels. As shown by Pennacchietti et al.1, HIFs also lead to higher levels of c-Met protein. This protein, on binding hepatocyte growth factor (HGF, which can be produced by nearby stromal cells), can result in increased cell motility, invasion and metastasis. These findings raise concerns about the idea of treating cancer by cutting off its blood supply.

Several biological mechanisms have been proposed to explain the observed correlation between hypoxia and accelerated cancer progression. But although there is indirect evidence for these mechanisms, discrete causal links between hypoxia and tumour aggressiveness have been elusive. The report by Pennacchietti et al.1 finally makes a firm molecular connection.

Pennacchietti et al. looked at the c-Met protein, which is a receptor for hepatocyte growth factor (HGF), and is found on the cell surface. In a nutshell, they show that when the HIF proteins sense hypoxia, they trigger an increase in the levels of c-Met. The authors first found that hypoxia led to higher levels of c-Met messenger RNA, indicating protein synthesis, and of c-Met protein in cultured tumour cells. They also identified regions of the c-Met gene that contain typical sequences for binding HIF, and showed that these sequences could be used to drive the expression of a different gene (a so-called reporter gene) in response to hypoxia. Next, the authors turned to grafted tumours in animals and to specimens of human tumours, and found that c-Met and one type of HIF protein were expressed in the same cells. The expression of HIFs is usually taken to indicate hypoxia. In the same samples, the expression of the endothelial-cell marker CD31 and c-Met was spatially distinct. This implies that regions of a tumour that are close to blood vessels (and so probably have normal oxygen levels) express less c-Met.

It is known5 that, when HGF binds to c-Met, it switches on the receptor's intrinsic enzymatic activity and causes it to phosphorylate itself. This addition of phosphate groups stabilizes c-Met's catalytic activity and allows intracellular signalling proteins, such as Gab1, to latch on and also be phosphorylated. Pennacchietti et al. now show that, in cultured cells, hypoxia leads to enhanced HGF-induced phosphorylation of c-Met and Gab1. It also results in greater HGF-stimulated cell motility, invasion and branching. But hypoxia-enhanced, HGF-stimulated cell branching is lost when c-Met is repressed.

So, hypoxia leads to increased c-Met levels in tumours, and to increased activity of this signalling pathway (Fig. 2). But why should that influence the aggressiveness of cancer? HGF, the ligand for c-Met, normally stimulates growth, migration and shape changes in a range of cells, including epithelial, endothelial, blood, neural and skin cells, as well as hepatocytes (liver cells)6. These diverse effects are crucial to development, organ growth and tissue regeneration7,8,9,10,11. But overactive HGF signalling has also been linked to many different cancers11. For instance, inherited activating mutations in the gene encoding c-Met are associated with renal papillary carcinomas in humans12,13. This is probably partly because activation of the c-Met pathway by HGF stimulates cell division. But it is also known to cause the dissociation of one cell from another, increased cell mobility, and greater production of protease enzymes that degrade the matrix in which cells are embedded — all features that increase the ability of cells to migrate through the extracellular matrix in vitro, and which correlate with tumour metastasis in vivo11. So the newly discovered connection between hypoxia and c-Met may explain why hypoxia often correlates with a poor prognosis for cancer patients.

This connection also sheds new light on a type of renal-cell carcinoma14. c-Met is reportedly overexpressed in these carcinomas, which are associated with inherited inactivating mutations in the von Hippel–Lindau (VHL) gene. The VHL protein usually targets HIF proteins for degradation when oxygen levels are normal. So, when VHL is mutated, HIF can no longer be degraded, presumably leading to increased c-Met expression. In effect, the loss of VHL mimics hypoxia, and this can contribute to tumour aggressiveness.

The findings of Pennacchietti and colleagues1 raise concerns about treatments that aim to cut off a tumour's blood supply. Although the resulting hypoxia might starve the main mass of the tumour, it could also stimulate the motility, invasion and metastasis of tumour cells at the front line — the tumour–host interface. In other words, tumour cells that survive the hypoxia might be selected for increased aggressiveness. But perhaps there is a silver lining to this cloud, if anti-angiogenesis strategies could be combined with drugs that target the proteins needed for motility and invasion — which now include HGF and c-Met. Such an approach would circumvent the insidious ability of tumour cells to co-opt, and be sustained by, the hypoxia-induced programme of events.