Trung Phan’s science experiment began with a dare. His boss, Princeton University physicist Robert Austin, challenged him to design a maze that Austin couldn’t solve.
To be sure, the challenge was just a thought experiment—Phan wasn’t about to actually start planting Versailles-scale hedges and throw his boss in the middle of it. But Phan, Austin’s graduate student, took the assignment to heart. He gave Austin a few easy puzzles to start, to learn Austin’s maze-solving strategy. “When he hit a dead end, he just traced his path back, which is a very traditional way to solve a maze,” says Phan. “So my idea was: How about a maze with no dead ends?”
In Phan’s final blueprint, wrong paths merge into other wrong paths, with the goal of throwing even the most patient navigator into an infinite loop of despair. “Inside that maze, you don’t know where you are,” says Austin. “I don’t know how long it would take me to solve that maze from the inside, because you can end up going in circles.”
But Austin wasn’t actually the intended player of this game, and designing the maze was only the first step in answering a bigger question about how organisms solve problems. In fact, the lab’s true maze runners are bacteria, which Austin and Phan are studying to learn about the microbes’ collaborative abilities. Phan came up with the idea of a maze test “to see how smart these bacteria really are,” says Austin.
Curiously, bacteria—single-celled organisms that are among the simplest living things—are well known for working together, creating problem-solving units that are more than the sum of their parts. For example, to protect themselves from your immune system, the bacteria in your mouth will unionize to form a film on your teeth known as dental plaque. Myxococcus, a type of bacteria that lives in soil, forms thread-like networks between microbes so that they can hunt prey in a pack. Many bacteria, including E.coli, are also capable of communicating among themselves to determine if microbes nearby are their own species or an enemy, by exchanging certain chemicals in a process known as “quorum sensing.”
In Phan’s case, he wanted to see if the bacteria could navigate his labyrinth. So in the next phase of the research, a colleague etched Phan’s winding pathways on a small silicon chip, and the researchers trapped around 10 E.coli bacteria in the center. Then they flooded the chip with the bacteria’s favorite food, a broth that “smells like chicken soup,” according to Phan, and then observed them through a microscope.
In a new paper accepted to Physical Review X, his team showed that bacteria succeeded at the task as they ate—and reproduced—their way around the maze. (By the end of the experiment, the 10 bacteria had become more than a million.) As they cleared paths of food, the E.coli tended to move toward unexplored, broth-rich areas, which ultimately helped them evacuate the maze. It took about 10 hours for about 1 percent of the multiple generations of bacteria to collectively solve the puzzle. That may not sound fast, but it’s five times faster than if the organisms had just been swimming around randomly, says Phan.
In addition to running that maze, Phan confined the bacteria to the center of a different puzzle, a tree-shaped trap resembling the fractal-like structure inside human lungs that had no exits. The motive of this experiment was to study how the bacteria would behave when facing a dead end. They found that the bacteria would quickly get trapped in the smallest branches of the fractal, but then, unexpectedly, they would build up in clumps and collectively launch themselves in waves out of the dead ends. The wave behavior seemed to arise from inter-bacterial communication, with the microbes responding to chemicals emitted by their compatriots. “The bacteria definitely work together,” says Phan.
It’s “not surprising” that the bacteria were able to navigate Phan’s puzzles, given the intricate natural landscapes in which E.coli are known to thrive, says microbiologist James Berleman of Saint Mary’s College, who was not involved in the work. “It’s worth pointing out that our small intestine, which E.coli can reside in, is certainly a more complex environment,” he says.
Still, Phan’s maze may be one of the most sophisticated human-made settings that anyone has watched bacteria navigate. “I haven’t seen anything quite like this,” says Berleman. “The fractal and maze structure that they use are really pretty complicated.”
Researchers often use mazes to study animal, behavior because they can imitate the complexity of nature but are easy to control in the lab, says ecologist Inon Scharf of Tel Aviv University in Israel, who studies insect behavior. The maze, in a way, serves as a metaphor for an organism’s life. At its core, any organism’s existence involves a series of forks in the road—ones that lead to survival or to death. A maze just treats those forks literally.
One larger goal behind the Princeton lab’s experiments is to better understand the motion of bacteria in various environments, which could help elucidate how microbes mutate to develop antibiotic resistance, says Austin. The maze provides a framework to study how bacteria move. He and Phan were surprised at how fast the bacteria were able to traverse the maze and the fractal, and they think their experiment could point to a previously unknown communication mechanism between bacteria, beyond chemical sensing.
For example, Austin and Phan have noticed that the bacteria leave a mystery residue on the surface of the maze. “We don’t know what it is,” Austin wrote in an email to WIRED. “We do know it is extremely difficult to remove.” They have only managed to clean it off by completely removing the surface of the maze with strong acid and high heat. They hypothesize that the bacteria leave this residue as clues for subsequent microbes, known among math researchers as a “Hansel and Gretel” mode of solving a maze.
Berleman, however, is skeptical of these claims. Austin and Phan draw their experimental conclusions from comparing the performance of two strains of E.coli, one strain capable of chemical communication, and one that is incapable. But the two strains of E.coli have other differences, which makes it difficult to determine how the bacteria solved the maze, says Berleman. The communicative strain’s advantage over the other could still be due to factors other than unknown communication abilities, such as a more advanced turning capability.
Regardless of the bacteria’s escape mechanism, the experiment conjures questions about bacteria sophistication. “They definitely possess an incredible ability to solve problems, to find food and escape structures,” Phan says. “Whether that really means intelligence, I will say I do not know.”
Biologists tend to avoid the word “intelligence,” because nobody agrees on what it means, says Scharf. He thinks that people often misinterpret the word, thinking it signifies humanlike abilities. In the context of a scientific experiment, intelligence is relative, depending on the skill being tested. “There are some tests where pigeons do better than humans,” says Scharf.
Scharf prefers to describe his studies in terms of measurable quantities, like the time it takes to solve a maze, instead of an abstract concept like intelligence. “It’s always better to use terms that are more specific,” he says. “I always make it clear what I did, what I measured.”
Nobody’s arguing that the kind of smarts the bacteria exhibited in running a maze are humanlike: The two species are just too different. “E.coli, in terms of metabolism, is way more complex than us,” Berleman says. “It can make all 20 amino acids. We can’t. It’s a different complexity than our complexity.” Unlike a human going through a corn maze, the microbes are continually reproducing as they solve the puzzle. And they work together in a way that millions of humans could never do. But still, they seem pretty, well … “Their behavior is quite clever, if we are allowed to use that word,” says Phan.