Most studies of the causes of resistance to cancer treatment focus on the tumour itself. However, some resistance mechanisms might involve alterations in the host rather than the cancer. A particularly conspicuous gap in our knowledge concerns the possibility that dietary factors influence the outcome of some cancer treatments. This has been widely assumed not to be the case, but writing in Nature, Hopkins et al.1 show that cancer drugs that inhibit the signalling protein PI3K are considerably more effective in mice if the animals are on a specific diet. The authors provide a plausible mechanism for why this is so.
A person with cancer might wonder whether their diet could affect their prognosis. A wide range of dietary recommendations are available, both on the Internet and from physicians and dieticians. Such advice is often conflicting. For example, a patient might read that extreme dietary calorie restriction helps to ‘starve’ a tumour in a clinically useful manner, but might also come across information suggesting that the opposite approach of maximizing calorie intake is beneficial, to avoid cancer-associated weight loss linked to later stages of the disease. Clinical data to support either of these approaches are not compelling. Physicians lack high-quality data on which to base dietary advice for people undergoing cancer treatment.
Hopkins and colleagues now provide evidence from mouse experiments that a diet that keeps levels of the hormone insulin low improves the effectiveness of cancer drugs that inhibit PI3K. There is great interest in trying to inhibit PI3K signalling in cancer cells, because mutations that cause excessive activation of this pathway are common in many kinds of cancer2. The pharmaceutical industry has invested heavily in developing drugs that inhibit the PI3K signalling pathway, but most clinical trials of these agents have revealed only modest benefits.
The authors offer a fresh perspective on the effect of PI3K inhibition by taking into account the fact that these inhibitors not only target cancer cells, but also act on tissues that regulate blood glucose levels. In glucose regulation, insulin is secreted by the β-cells of the pancreas when blood glucose levels rise. Insulin binding to receptors on its target cells activates the PI3K signalling pathway in liver, muscle and fat, causing changes in glucose production and uptake that reduce the glucose concentration in the bloodstream.
Cancer cells commonly also express insulin receptors, and, as in normal cells, signalling through the insulin receptor activates the PI3K signalling pathway. However, in cancer cells, pathway activation causes an increase in cell proliferation and a reduction in cell death, rather than affecting blood-glucose regulation. This is in keeping with the observation that insulin-like hormones and the PI3K signalling pathway are both ancient in evolutionary terms3,4, and their roles in stimulating cellular nutrient use and proliferation pre-date their function of blood-glucose regulation.
Hopkins and colleagues’ results confirm previous reports5–7 that PI3K inhibitors raise blood glucose by blocking signalling downstream of the insulin receptor in tissues involved in blood-glucose regulation. The authors go on to show that this increase in blood glucose causes a substantial elevation in insulin levels in the bloodstream. This rise in insulin probably results from pancreatic β-cells responding to high blood glucose concentrations by secreting extra insulin in an attempt to restore normal glucose levels. The authors make the key finding that, in cancer cells that express insulin receptors, this rise in insulin is sufficient to increase signalling downstream of the insulin receptor to activate PI3K and overcome the action of the PI3K inhibitor on the pathway (Fig. 1). This lessens the drug’s therapeutic effect and enables cancer cells to proliferate despite drug treatment.
Furthermore, Hopkins et al. show that combining a PI3K inhibitor with a pharmaceutical or dietary intervention that lowers blood glucose reduces the drug-induced elevation in insulin, and that this increases the effectiveness of PI3K inhibitors in slowing cancer growth, compared with the effect of the drug in animals that do not receive a glucose-lowering intervention. The most effective intervention tested was a type of low-carbohydrate, high-fat diet termed a ketogenic diet, which improved the action of the PI3K inhibitor to a greater extent than was observed for the drug metformin, which reduces glucose output from the liver, or for the drug canagliflozin, which causes glucose loss in the urine. Notably, the authors found that a ketogenic diet in the absence of the inhibitor drug did not curb cancer growth in mice. Consistent with this, early-stage clinical trials have not shown that this diet alone improves cancer survival8.
Hopkins and colleagues’ findings are important because they identify a mechanism of resistance to cancer therapy that is based on a hormonal response of the host, rather than on alterations in cancer cells. The key element of this response is a rise in insulin levels, but whether other hormones involved in regulating glucose metabolism also contribute to the effect is not known. Previous research9 indicates that high insulin levels associated with obesity can increase the risk of cancer or worsen prognosis. The work by Hopkins et al. reveals an additional connection between insulin and cancer by illuminating a way in which this hormone can influence the usefulness of a cancer drug.
Certain PI3K-inhibitor drugs block only the δ version of the protein, which is found in blood cancers, but not in common solid tumours or in the organs that regulate blood glucose. These inhibitors presumably do not significantly elevate blood insulin levels, so the mechanism of resistance described by Hopkins et al. does not occur. If this is the case, it might help to explain why this particular type of PI3K inhibitor was the first to show sufficient activity in clinical trials to be approved for clinical use.
Hopkins et al. provide a firm basis for investigating dietary or pharmaceutical approaches to reduce insulin signalling in cancers to enhance the effectiveness of PI3K inhibitors. The authors’ data indicate that previous attempts10 to predict whether this class of drug would be effective by examining tumour characteristics would have overlooked the part played by host insulin levels.
Substantial patient-to-patient variability in the response to anticancer drugs often occurs even among patients whose tumours show similar genetic abnormalities. Along with another paper11, published last month, which demonstrates that a diet rich in the amino acid histidine can increase the effectiveness of the anticancer drug methotrexate, Hopkins and colleagues’ study supports the idea that differences in diet could contribute to variability in response to cancer treatments. Excitingly and unexpectedly, we now have a rationale for clinical research to determine whether the efficiency of certain cancer drugs might be improved by pairing them with specific diets.