In 2009, Danny Cahill won the eighth season of The Biggest Loser, a reality TV show in which contestants compete to lose the most weight. Over the program’s seven months, Cahill’s weight dropped from 430 pounds to just 191. But since then, he has regained 100. The same is true for most of the show’s contestants, several of whom are now heavier than they were before they took part.

Their story is all too common. Even when people successfully manage to lose weight, in the majority of cases, the vanished pounds return within a year—and often with reinforcements. For many people, weight loss isn’t just hard, it’s Sisyphean.

No one really understands the reasons behind this “weight cycling”, this so-called “yo-yo effect”. It seems to happen no matter your starting weight, or how much exercise you do. As my colleague Julie Beck noted earlier this year, the speed at which people lose weight might be important—but even that’s controversial. “There’s a lot of speculation but very little knowledge,” says Eran Elinav from the Weizmann Institute of Science in Israel.

Now, by studying mice, Elinav and his colleague Eran Segal have shown that the yo-yo effect might be at least partly driven by the microbiome—the huge community of bacteria and other microbes that share our bodies. These microbes help us to digest our food and extract energy from it. And if we change what we eat, our microbiome changes too, in terms of which species are abundant, and how they process our food. Some of these changes happen very quickly but, as Elinav and Segal discovered, others do not.

Their student Cristoph Thaiss duplicated the yo-yo effect in mice. He fed them high-fat chow for a month, and they gained weight and became obese. He switched them to a diet of normal chow, and they lost weight. He put them back on the high-fat chow, and this time, they put on even more weight than before. Just like humans, their successful bouts of dieting predisposed them to accelerated weight gain.

When Thaiss examined their bodies, he couldn’t find an obvious explanation. After dieting, the obese rodents returned to their original baseline weight, and everything else about their bodies normalized—their cholesterol, blood sugar, insulin, body fat, appetites, physical activity, and metabolic rates.

But their microbiomes did not. When they first gained weight, several species of microbes in their guts became more common, others became rarer, and the overall community became less diverse. But as the mice lost weight, those microbial changes were slow to revert.

Through two experiments, Thaiss showed that these lagging microbes explain why the mice were prone to putting on even more weight after their successful diets. First, he transferred those microbiomes into mice that had been raised in sterile conditions and had no microbes of their own. When the recipients then ate fatty food, they put on more weight than usual.

Second, Thaiss treated the dieting rodents with high doses of antibiotics to kill off their existing gut microbes and reset their microbiomes. When these animals went back to high-fat food, they put on weight, but to a less exaggerated degree than before. The antibiotics had reduced the yo-yo effect. “This is by no means a recommendation to take antibiotics while dieting,” says Elinav. “That would have many side effects, and I don’t think it would be helpful. It was just a scientific way of showing that the microbes are doing something.”

Not all the mice reacted in the same way. As in humans, some yo-yoed more strongly than others, and Thaiss showed that a computer algorithm could learn to predict the size of these rebounds. By analyzing the rodents’ microbiomes, it accurately predicted the extent of their weight gain, with a score of 0.72 on a scale from 0 to 1. And tellingly, it based its predictions by looking at the levels of 189 microbe species. “There’s no one silver bullet microbe,” says Elinav. It’s the whole community that affects our health—a common theme that runs throughout microbiome studies.

To summarize: when rodents lose weight, their microbiome lags. During that window, if the mice break their diets and eat high-fat food again, they put on weight faster than before. The altered microbiome doesn’t make the mice fat on its own. Rather, it makes them especially vulnerable to the effects of a poor diet.

Maybe this is a way for animals to withstand long periods of hunger and famine. “The microbiome could act as a buffer to preserve energy until the next feeding opportunity comes,” says Elinav. But if famine gives way too suddenly to feast, “the same mechanisms that protect us might act against us and increase our tendency to accumulate fat.”

“The study is likely to be a milestone towards understanding the complex mechanisms that lead to weight cycling,” says Abdul Dulloo from the University of Fribourg. Still, he notes that it doesn’t fully capture the complexity of the yo-yo effect. For example, Thaiss’s mice initially became obese after eating fatty chow, but the yo-yo effect can strike lean people who eat normal healthy diets, like athletes or military personnel who lose weight after intense training. And “when you feed mice, you give them a standard diet that isn’t generalizable to humans,” adds Krista Casazza from the University of Alabama at Birmingham.

True enough, which is why the Israeli team are now studying human volunteers to see if they are affected by the same microbial lag—and to what extent. For example, after a 4-week cycle of weight gain and dieting, Thaiss’s mice took 21 weeks for their microbial communities to normalize. If that span of time scales to human lifespans, it would mean that people are vulnerable to the yo-yo effect for years after dieting. But if it stays the same across species, it would mean a more manageable window of several months. “Until you measure it in humans, we just don’t know how long the window of susceptibility is,” says Elinav.

“We know from our human clinical trials that there seems to be a point of no return in weight loss, beyond which subjects can keep their lost weight off,” says Liping Zhao from Shanghai Jiao Tong University. “That point seems to be personalized, and we’re now inspired to more precisely define it for each individual [in our studies].”

Elinav also hopes to find ways of manipulating the microbiome to prevent the yo-yo effect. His team managed to slightly reduce it by giving their dieting mice a faecal transplant—filling them with stool from healthy individuals to reset their microbiomes. In humans, “no one has attempted a trial of faecal transplants after successful dieting,” says Elinav. “But these transplants are a very aggressive and non-specific approach.” He’s after something a bit more precise.

Again, the mice provide some important clues. When they ate fatty food, their gut microbes started enthusiastically breaking down flavonoids—nutrients that normally increase the burning of fat. Fewer flavonoids meant more weight. And when the team fed their dieting mice with flavonoid supplements, they forestalled the subsequent rebound.

So, is the answer to eat flavonoid-rich foods after dieting? Again, the results provide hope for the future rather than actionable tips for the present. “Everything we’ve discovered was in mice,” Elinav emphasizes. “It may be flavonoids in humans too, or it may be something else. The devil is in the details. By understanding those detail, we’ll be able to intervene in a more accurate way, which is something we haven’t been able to do.”