Investigators on Wednesday said there could be a clue to explain why dietary restriction succeeds or fails in braking the growth of cancer.
Scientists spotted a century ago the link between calorie intake and the progression of certain kinds of tumour.
But how this works at the molecular level has been unclear. In addition, tumours may remain puzzlingly immune to fasting among some individuals but not among others, which is partly why the practice has not been incorporated into cancer treatment.
Nada Kalaany and David Sabatini of the Massachusetts Institute of Technology (MIT) looked at the impact of food curbs on six types of human cancers that had been grafted onto mice which had been genetically modified to lack immune resistance.
The mice were then split into two groups, with one batch able to tuck into a standard rodent diet for two to three weeks. The others received daily meals that amounted to a 40 percent reduction in calories compared to their counterparts, which led to a fall of between 20 and 30 percent in body weight.
The researchers conclude that tumours' sensitivity to fasting depends on a key pathway -- activation of an enzyme called phosphatidylinositol-3-kinase, also called PI3K, which plays an important role in the proliferation of cancer cells.
If the PI3K pathway is closed, tumours respond to dietary restriction, and vice versa.
The study is important because it helps drug designers who are working hard on prototype treatments that mimic the effects of dietary curbs on cancer.
As several cancer types carry mutations that could cause the PI3K mechanism to open up, the paper suggests that the approach will not work in certain cases.
The paper is published by the British-based journal Nature.
Previous research on lab rodents suggests that a calorie decrease of 10 to 50 percent leads to a reduction in the incidence of certain cancers and brakes the speed at which they spread.
One theory is that the change in caloric intake beefs up cell repair mechanisms against destructive agents called free radicals.