Salmonella bacteria are best known as a causes of food poisoning and typhoid fever. Every year, they sicken millions of people. But those in Jeff Hasty’s lab at the University of California, San Diego, are different. They’ve been neutered and modified so that rather than causing gastro-catastrophes, they kill tumors.

Hasty’s team has engineered the microbes to produce a variety of anti-cancer drugs, and to self-destruct when they reach a certain density. In their death throes, the bacteria release their toxic payloads to kill the tumor cells around them. “It’s like a kamikaze mission,” says Hasty. But not all the bacteria die. About 10 percent of them survive and can re-seed the population, triggering many more rounds of self-destruction and drug delivery.

These engineered microbes have only been tested in laboratory cells and mice. Needless to say, there are many unresolved questions about safety and efficiency, and this preliminary approach is a long way from being a legitimate cancer treatment.

But that’s almost not the point. What Hasty has created is a probiotic on a timer—or rather, with a snooze button. He’s proved that it’s possible to engineer bacteria to produce and release drugs at regular intervals, to the beat of a genetic metronome that he can set. That’s useful not just for cancer, but for diseases like diabetes and high blood pressure that require regular doses of medicine. “It’s going to open up opportunities more general than cancer,” says Hasty. It’s part of a growing number of attempts to turn the bacteria that share our lives—the microbiome—into medical assistants.

Scientists and doctors have been trying to use bacteria to fight cancer for more than a century. In 1891, the surgeon William Coley infected cancer patients with Streptoccocus, thinking that the microbes would trigger an immune response that would also destroy the cancer cells. Coley treated over a thousand people in this way, and although some recovered, others didn’t. Amid mixed results, the approach fell into neglect.

It resurfaced when doctors started realizing that tumors were not sterile, as had long been thought, but often contains multitudes of microbes. The oxygen-hating bacteria that thrive in our gut can readily grow in the interior of tumors, which are similarly low in oxygen and sheltered from patrolling immune cells. The presence of such microbes was attractive to scientists because one of the big challenges in cancer medicine is actually getting drugs into tumours in the first place. If tumour-seeking microbes could be engineered to make those drugs, they could be the perfect Trojan horses.

First, you need to pick the right microbe. Salmonella is a good choice. It can survive without oxygen and readily accumulates in tumors. It’s also closely related to Escherichia coli, the favoued bacterium of the modern biologist. Which means that all the techniques that geneticists have developed for engineering E. coli can be used to modify Salmonella too. “We can quickly leverage an enormous toolbox,” says Hasty.

But, wait—bacteria are living things. They grow. They multiply. They trigger immune responses. Other scientists have tested engineered strains of drug-delivering Salmonella, with poor results. “We killed our mice,” says Seigfried Weiss from the Helmholtz Centre for Infection Research. “They had such a high load of bacteria that they probably suffered a toxic shock.” So if you want to use Salmonella to shuttle drugs into tumors, you need some way of controlling their numbers.

That’s what Hasty’s team, led by student Omar Din, developed. They started with a de-fanged strain of Salmonella that doesn’t cause disease, and gave it the ability to make an anti-tumor drug. They also added genes that bacteria-killing viruses use to destroy their victims, and cleverly wired these up to genes that Salmonella and other bacteria use to sense each other.

The result is a medicine-making microbe with built-in population control. Once the population hits a certain size, 90 percent of the cells rupture synchronously, releasing their cargo. The survivors can then repopulate, and the whole cycle beings again, like clockwork.

Of course, “you’re still putting in a live Salmonella strain,” says Beth McCormick from the University of Massachusetts Medical School. They might be neutered but they’re still bacteria, and they could still trigger inflammation and other immune problems. “But these synchronized circuits allow a persistent but low state of infection, which might overcome the issue with immune responses. It’s a great preliminary step.”

The self-destruct approach doesn’t just stop the bacteria from over-growing. It also delivers a big pulse of a drug directly to the site of a tumor. And it makes it easy to program the bacteria to release many kinds of drugs. Hasty’s team tested three: a non-specific toxin that bores into mammalian cells, a substance that stimulates the immune system, and one that causes cancer cells to commit suicide.

The team’s colleagues, led by Sangeeta Bhatia at MIT, then tested these microbes in mice with incurable tumors that had spread to their livers. On their own, the bacteria did no better than a standard chemotherapy drug. But when the team used both together, the tumors shrank by a third and the mice lived 50 percent longer. That makes sense, says Hasty. “Salmonella can grow in the oxygen-poor region inside the tumor, where chemotherapy isn’t as effective. The bacteria get behind enemy lines and work from the inside out while the chemotherapy works at the growing rim of the tumour.”

That’s promising, but there are many kinks to iron out. If the bacteria were ever to be tested in humans, they’d need more safety features to stop them from growing, spreading, or mutating unpredictably. They’d need to be more effective, too; the team’s mice initially benefited from their microbial injections but their tumors started growing again after three weeks. “There are many things that need to be worked out before we think about testing these things in humans,” says Hasty. “I want to show some restraint.”

He’s right to do so. There’s a long history of medical techniques that seemed exciting in mice but that faceplanted in people. For example, in the late 1990s, David Bermudes from California State University also developed Salmonella strains that could shuttle drugs into tumours. These fared well in mice and monkeys, but when they were eventually tested in human clinical trials, they didn’t work.

“We could show safety but there was no measurable anti-tumor activity,” says Bermudes. “In mice, the bacteria targeted a wide range of tumors 100 percent of the time. In the human study, they targeted tumors just a third of the time and only at the highest dose. We all know mice don’t predict tumor therapies, or we’d have cured cancer already.”

Still, Bermudes thinks the approach has merit, as do many other scientists. Some are trying to program Salmonella to release toxins after invading cancer cells directly. Others are modifying the bacterium to release substances that either remove the shields that cloak cancer cells from the immune system, or that activate immune cells directly.

“One of the problems in using bacteria against tumors was the lack of acceptance by oncologists,” says Weiss. “Very few groups worked on it. But slowly, more are getting interested. New ideas will fertilize the field, and I’m sure that eventually, this will bring bacterial therapy into the clinics.”