Harry Sokol on Gut Microbiome and Health

A curious thing happens when a scientific field gets popular before it gets settled: the public starts acting on answers the scientists themselves do not have yet. The gut microbiome, the trillions of bacteria living in the human intestine, is a case study. There are tests, supplements, and diet plans built around the idea of a "healthy microbiome." And yet Professor Harry Sokol, a gastroenterologist at Saint Antoine Hospital and Sorbonne University in Paris who runs one of France's main fecal transplant centers, says plainly that no one can define what that actually means (05:31). The field, he explains, is only twenty to twenty-five years deep into serious study, and many of the basic functions of gut bacteria are still unknown. Without knowing what the bacteria normally do, there is no way to say what "normal" looks like.

That gap matters because it shapes how easily a finding gets oversold. Sokol's broader point is that most of the popular claims about the microbiome rest on correlation, not causation, and correlation is a trap. He offers a simple example: catch a stomach virus, and a microbiome sample taken during the illness will look disturbed. That disturbance is not a cause of the disease, it is a result of it, a pattern researchers call reverse causality (08:34). To tell the difference between a bystander and a driver, Sokol's own research follows a specific sequence: take the microbiome from sick patients and transplant it into germ-free mice, animals raised with no microorganisms of their own, and see whether the disease follows the bacteria (09:25). Even that is not proof, since mice are not humans. The real test is intervention: change a person's microbiome, through a fecal transplant or an antibiotic, and see if the disease itself changes.

Fecal microbiota transplantation, known as FMT, is the clearest case of doing this well, and it also shows how unevenly it works across diseases. For recurrent Clostridioides difficile infection, a bacterial overgrowth that typically follows a course of antibiotics, FMT succeeds in 80 to 90 percent of cases (13:41). Sokol does several hundred of these transplants a year at his center.

"Fecal microbiota transplantation is, I have to say, almost magical in this disease." — Harry Sokol [12:42]

Contrast that with ulcerative colitis, a form of inflammatory bowel disease. FMT there induces remission in only 30 to 35 percent of patients, and the effect usually fades over time, probably because the transplanted microbiome does not stay put (14:05). For irritable bowel syndrome, a common condition involving abdominal pain and irregular bowel habits with no visible lesions on standard tests, the published FMT trials have been largely negative, and Sokol says there is currently no good rationale for using the procedure there at all (15:23). The pattern across these three conditions tells its own story: the more the microbiome is the actual engine of a disease, the better a microbiome-based treatment works, and the more other factors are involved, the less a single intervention can do.

What Diet Can and Cannot Do

Diet is the lever people reach for most often, and Sokol is careful about what it can realistically deliver. He points to the Mediterranean diet, built around diverse fruits and vegetables for fiber, less ultra-processed food, less red meat, and some fermented foods, as the best-evidenced dietary pattern for microbiome health (18:30). But he frames its value narrowly: it is mainly a tool for prevention in people who are already healthy, not a treatment for people who already have a disease (18:03). That framing comes from a deeper complaint he has about how the field measures success.

"As a physician, to be honest, I don't care about the microbiome diversity. I care about the health of my patient." — Harry Sokol [17:11]

There is one narrower case where diet is starting to look like real medicine rather than general wellness advice. A study from Kevin Whelan's research group in London tested removing emulsifiers, common additives in ultra-processed foods, from the diets of patients with Crohn's disease, a form of inflammatory bowel disease. The unpublished data so far show a lower inflammatory signal in patients who cut emulsifiers (22:14). It will not replace drug treatment, Sokol says, but it is a plausible add-on for a specific, already-diagnosed population, which is a different claim than telling a healthy person to avoid emulsifiers for general wellness.

The diet conversation connects to something even less familiar: gut bacteria quietly altering how medications work. Sokol describes three documented mechanisms. Certain bacteria can metabolize metformin, the most widely prescribed drug for type 2 diabetes, effectively breaking it down before it can act, which may explain why some patients do not respond to the drug (24:58). In Parkinson's disease, the standard treatment, levodopa, is designed to cross into the brain and convert into dopamine there. A Boston research team found that some patients carry gut bacteria that convert levodopa into dopamine too early, in the gut, where it cannot cross the blood-brain barrier, and some resulting byproducts may even be toxic (26:19). And in a chemotherapy used for pancreatic cancer, the liver deactivates the drug and sends the inactive form into the gut through bile, where certain bacteria can reactivate it.

"The bacteria re-toxify the molecule. So the molecule get toxic again, and this is responsible for very strong diarrhea, which is the main side effect of this chemotherapy." — Harry Sokol [27:15]

Sokol is careful to note these findings are compelling but not fully proven as causal in humans. The next step, already underway for metformin, is stratifying patients by how much of the drug-degrading bacteria they carry, and eventually testing whether killing that bacterium, perhaps with a targeted virus called a phage, restores the drug's effect (29:20).

Turning a Bacterium Into a Drug

Sokol's own research has moved from description to product. Years ago, his lab noticed that a bacterium called Faecalibacterium prausnitzii was depleted in patients with Crohn's disease, and that lower levels tracked with a higher risk of flares. This bacterium is unusual: it is present in nearly every healthy person on Earth and typically makes up about 5 percent of a healthy gut microbiome (31:24). After showing the bacterium had protective effects in mouse models of colitis, Sokol co-founded a company, Exelium, to develop it as what is called a live biotherapeutic product, essentially a bacterium manufactured and tested like a pharmaceutical drug. A phase one safety trial in eight Crohn's disease patients showed no safety issues and some signal of benefit, and a randomized phase two trial is now underway (32:48).

The same bacterium turned out to matter somewhere else entirely: cancer immunotherapy. Researcher Laurence Zitvogel's team in France showed that giving antibiotics to mice or patients receiving immune checkpoint inhibitors, a class of cancer drugs that unleash the immune system against tumors rather than poisoning cells directly, sharply reduces how well the drugs work (34:06). Faecalibacterium prausnitzii emerged as one of the bacteria most associated with a good response. Sokol's team is now running four clinical trials testing the bacterium alongside checkpoint inhibitors, in gastric, liver, and lung cancer in France, and kidney cancer in the United States (35:09).

What ties these projects together is Sokol's insistence that microbiome products be held to the same standard as any other drug, tested through phases one, two, and three, with clinical outcomes like survival or remission as the endpoint, not microbiome diversity scores. That standard is exactly what is missing from most of the probiotic industry, and exactly what is missing from consumer microbiome testing. Sokol points to a study published in the journal Gut in which the same stool sample was sent to five different microbiome testing companies and came back with five different results and different recommendations (40:19). Combine that with the unresolved question of what a "normal" microbiome even is, and the conclusion is blunt.

"I mean, there is no good reason to recommend doing this test for now." — Harry Sokol [42:38]

Asked for a single piece of daily advice outside the realm of diet and supplements, Sokol did not reach for another microbiome intervention. He said to find something that brings genuine pleasure, in his case listening to music, because happier people tend to be healthier people (43:55). It is a fitting close to a conversation built around resisting shortcuts: the body's bacteria are not a dial to be optimized by a test result, and neither, it turns out, is a life.