A microbe that normally lives in the soil has been genetically engineered to destroy cancerous tumours and leave healthy tissue untouched, a study shows. Scientists found the microbe, a bacterium called Clostridium novyi, can invade and kill the multiplying cells of large, solid tumours, eventually causing the cancer to shrink and disappear.

Cancer specialists say further work is needed before human clinical trials can begin but the preliminary results are highly encouraging.

A research team led by Bert Vogelstein of the Johns Hopkins School of Medicine in Baltimore, Maryland, used the bacteria with conventional anti-cancer drugs in a treatment they called Cobalt, combination bacteriolytic therapy.

"The results ... show Cobalt can result in rapid and dramatic regressions of experimental tumours in mice," the researchers say in the Proceedings of the National Academy of Science.

"Even relatively large tumours could be treated successfully with Cobalt, although tumours of the size used in our experiments do not generally respond well to chemotherapeutic agents," say the researchers.

Dr Vogelstein's team tested 26 strains of bacteria for their capacity to destroy tumours but found that C. novyi "appeared particularly promising".

Normally, the bacterium produces a highly lethal toxin but the scientists engineered the microbe's DNA to remove the toxin gene. They then injected spores of the bacterium into mice with large tumours.

Within 24 hours the tumours began to break up and shrink. The scientists said the effect was "significant and prolonged". The tumours decomposed and turned to blackened scars, leaving surrounding tissues unharmed. They disappeared after two weeks of treatment.

Solid tumours are difficult to treat with conventional therapies because their densely packed cells form a natural barrier which is difficult to penetrate. Soil bacteria are perfectly adapted to living in an oxygen-starved environment and the scientists believed they would be perfectly able to live inside a tumour. However, the researchers admit they do not fully understand how the treatment works. "It is clear that many questions remain," they say. "For example, the basis for the potent tumour cell killing in the vicinity of the germinating bacteria is not understood."

Other strains of bacteria able to germinate and grow inside tumours did not exhibit this ability to destroy cancerous cells and the Clostridium's killing ability was evidently not due to its toxin gene.

Some of the experimental mice died soon after being "cured" of their tumours, possibly as a result of the toxic by-products released by the rapid destruction of cancerous cells.

One of problems the scientists still have to address is how to avoid the side-effects of rapid tumour destruction before clinical trials on human volunteers can begin.

"We hope that this research will add a new dimension to cancer treatment, but realise that the way tumours respond to treatment in mice can be different than in humans," said Professor Kenneth Kinzler, a member of the research team.