In the early experiments it looked like the virus called VP882 was doing something that should be impossible for a thing that is not a bacterium, and not technically even alive: intercepting molecular messages exchanged by its host bacteria, and reading them to determine the best time to annihilate the whole bacterial colony. “As scientists, this is just unimaginable to us,” says Bonnie Bassler, a molecular biologist at Princeton University. “We were delighted and skeptical at the same time. It was almost too good to be true.”

Not only did it turn out to be true for VP882; Bassler learned there is a family of bacteria-infecting viruses (a subgroup of a kind called bacteriophages, or just “phages”) that eavesdrop on their hosts’ routine molecular communications with other bacteria. That means VP882’s kill trigger could be easily manipulated to target any bacteria, Bassler says—opening the possibility that the virus could be engineered into an ideal killing machine for dangerous pathogens.

In 2009 scientists in Taiwan first found VP882 in a bacterium related to cholera. The bacterium “probably made someone sick,” Bassler says, and in investigating the illness, “researchers came across the virus. They sequenced it and dropped it in a [DNA] repository.” When Bassler’s graduate student Justin Silpe came across the virus in that database, he had been looking for a gene that encodes a receptor for a particular molecule called DPO. Cholera and related bacteria utilize this molecule for something called “quorum sensing”—a kind of chemical “speech” that bacteria use to figure out how many of their own kind are nearby.

When a host detects a high concentration of quorum-sensing molecules in its environment, it is as if the bacteria are hearing a high volume of chatter in a room. “This tells them they have enough neighbors around to do collective behavior”—such as causing disease—Bassler says. She and Silpe had hoped to find the DPO receptor in other kinds of bacteria than cholera. Instead, it showed up in VP882. “Quorum sensing is supposed to be about bacteria. But here it was on a viral genome,” she notes. “We were like, ‘What is that virus doing?’”

When they conducted experiments with VP882, Silpe noticed that when it was in a cholera colony without DPO, the phage and the bacteria coexisted peacefully. But “we clearly noticed that where we added DPO, the cells died,” he says. “That was a clear indication the phage was killing them.” To confirm the quorum-sensing signal was indeed the kill trigger, Silpe mutated the DPO receptor on VP882—and showed the mutated phages did not kill.

This told the duo the virus tunes into the quorum-sensing information and uses it to figure out how many potential victims are around. If the virus senses enough DPO, it means a buffet is nearby; that signals VP882 to start destroying its host bacterium and sending copies of itself careening into the colony. Otherwise VP882 lies quietly and waits for the bacteria to reproduce. “It’s a wonderful and insidious strategy,” Bassler says. “It allows the phage to optimally find its next victim.” The team published their results in Cell Thursday.

This finding about VP882 is “very cool,” says Mark Mimee, a biologist at Massachusetts Institute of Technology who was not involved in the study. “It evokes this phage spy that is really listening to its bacteria and choosing the opportune time to strike.” He says one of the most intriguing parts of the discovery was how Bassler and Silpe could genetically engineer VP882. Most phages live inside a host bacterium’s nucleoid (a large tangle of DNA comprising its genome), but VP882 behaves like a plasmid—a simple ring of DNA that floats freely in the cell.

That is a huge advantage when it comes to engineering the virus itself. Plasmids are relatively simple to modify, and researchers can force any bacteria to accept any plasmid; they have been doing so since plasmids were discovered decades ago. Because VP882 behaves the same way, Bassler and Silpe were able to engineer its kill switch to turn on when it senses signals unique to Escherichia coli or salmonella or any other bacteria they might want to eliminate.

This could solve certain problems in phage therapy—the use of phages to treat infections—Mimee says. For one thing, most phages attack only one kind of bacteria. That makes phage therapy somewhat less attractive in a lot of situations. “If you have a lung infection, you might not be able to diagnose what bacteria [are] responsible in time and choose the right phage,” he notes. “To get around that, people use cocktails of different phages. But manufacturing cocktails and adhering to drug regulations is too expensive.” Phage therapy is used to treat infections in some countries—but not in the U.S., where each individual phage must go through rigorous and costly drug trials before authorities can approve it. Because VP882 can be engineered to kill any kind of bacteria, it might be more attractive to researchers developing phage therapy for the clinic, Mimee says: “A single recombinant phage—yeah, that would be really interesting.”

Turning VP882 into a highly controlled bacteria assassin is “our dream,” Bassler says. Silpe “can engineer the phage to kill on demand. It has really attractive features.” But there’s no telling if VP882 would be as good of a killer in the real world as it is in the petri dish, she adds. For one thing, VP882’s only known natural prey are cholera and its relatives—so it is unclear if the phage would infect other types of bacteria without a scientist forcing the infection or using more extensive engineering.

Still, cholera infects millions of people every year. “If you want to kill cholera in industrial wastewater or in the gut and not perturb the [beneficial bacteria in the] microbiome,” Bassler says, “that’s really interesting to us.”