The Company We Keep

In 2016 Ed Yong’s I Contain Multitudes argued that every animal is a crowd. Ten years on, which parts of the microbiome story held up, which were hype, and how much of the hype had the book already warned against?

The title comes from Walt Whitman: “Do I contradict myself? / Very well then I contradict myself, / (I am large, I contain multitudes.)”[1] Ed Yong took it literally. His book I Contain Multitudes: The Microbes Within Us and a Grander View of Life, published in 2016,[2] is about the bacteria, fungi and other microbes that live on and inside every animal, and about the claim that an animal is less a single organism than a community: a body plus the microbes it keeps.

I came to the subject through the book. Its stories are memorable: a squid that lights its underside with bacteria, wasps that coat their young in antibiotic-making microbes, babies seeded with their mother’s bacteria at birth, transplants of microbes that save frogs[3] and cure people. Stephen Curry, reviewing it in The Guardian, noted that Yong kept “dosing the text with warnings and caveats”, and guessed that “in five or ten year’s time a new edition is required”.[4] It has now been ten years. This essay is not that new edition, but it asks the question a new edition would have to answer: of the claims that made microbiome science famous around 2016, which held up, which were revised, and which turned out to be hype? Reading the book again with that question in mind gives a second answer, about the book itself. On most of the points where the field ran ahead of its evidence, Yong had already said so.

The count

Not ten to one

For decades, articles and textbooks repeated that bacteria in the human body outnumber human cells ten to one, so that we are only one-tenth human. Yong called the ratio “a wild guess, based on a back-of-the-envelope calculation that became unfortunately enshrined as fact”.[2] In 2016, the year the book came out, Ron Sender, Shai Fuchs and Ron Milo traced the figure to its source. Every reference they followed led back to a single back-of-the-envelope estimate in a short 1972 paper by Thomas Luckey,[5] which multiplied 10¹¹ bacteria per gram of gut content by a litre of gut.[6] Nobody had redone the sum.

They redid it. Almost all our bacteria live in the colon; the mouth holds less than 1% of the colon’s number and the stomach almost none. With better figures for the volume of the colon and the density of bacteria in it, they estimated 38 trillion bacteria in a 70 kg man, against 30 trillion human cells: a ratio of about 1.3 to 1, not 10 to 1. The human count is dominated by red blood cells, 84% of all human cells, small cells with no nucleus. The bacteria weigh about 0.2 kg, around 0.3% of the body. And since each trip to the toilet removes a quarter to a third of the colon’s content, the ratio swings every day, after a third of the bacteria are gone, about 25 trillion are left, fewer than the human cells.[6] The book already used the new count: “around 30 trillion human cells and 39 trillion microbial ones – a roughly even split”, and added that it “does not really matter: by any reckoning, we contain multitudes.”[2] Two small slips survive in it. The text says the microbes “weigh just a few pounds”, against the 0.2 kg of the recount, and a note still gives 100 trillion microbes and a Milky Way of “between 100 million and 400 million stars”, where astronomers count hundreds of billions.

Fig. 1 — The 10:1 ratio, traced to a 1972 estimate, and the 2016 recount for a 70 kg man. Sources: Luckey 1972; Sender, Fuchs & Milo 2016.

Yong was right that the correction matters less than it seems. Thirty-eight trillion is still a crowd, and the case for taking microbes seriously never rested on the ratio. The genes make the point better: a catalogue built from the faeces of 124 Europeans in 2010 found 3.3 million microbial genes, about 150 times the human set, with each person carrying at least 160 common bacterial species.[7] Whatever the count of cells, most of the chemistry that goes on in a gut is done by genes that are not ours.

The partners

Animals built with microbes

The strongest part of Yong’s book was never about people. It was about how deeply other animals depend on microbes, and that part has not needed revising, because the evidence was solid before he wrote.

The Hawaiian bobtail squid, the book’s first great character, carries in its underside light organs full of the luminous bacterium Vibrio fischeri. At night their glow “supposedly matches the moonlight that wells down from above, cancelling out the squid’s silhouette”: “This animal casts no shadow.”[2] A newly hatched squid has none; it collects them from seawater and, through a series of steps that Margaret McFall-Ngai and Spencer Nyholm called “the winnowing”, lets in this one species out of the millions of bacteria in the water.[8] The bacteria, in turn, “control the development” of the organ they live in, as Curry put it.[4]

Aphids carry a bacterium called Buchnera inside special cells. Its genome is only a seventh of the size of its relative E. coli; it has kept the genes for making the amino acids the aphid needs and lost those it can get from its host, and the partnership is “so obligate that neither can reproduce independently”.[9] In some cicadas the bacterial partner Hodgkinia has split into several lineages that only work together, some of them in separate cells, a complex with almost ten times the genome of its single-genome ancestor.[10] The bacterium Wolbachia, which manipulates the reproduction of its hosts, is estimated to infect about 40% of arthropod species.[11]

These are not curiosities. They show that a partner can become part of the body: the line between an animal and its microbes is a matter of degree. And the same microbe can help or harm. One of the book’s central ideas follows from that. Yong describes the immune system as an “immunostat”, a dial that “stabilises our relationships with our microbes”: it “manages the benign trillions that live with us, while thwarting invasions by an infectious minority”.[2]

The training

Microbes that teach the guards

That idea links this essay to The Second Lock, which described how the immune system is restrained by regulatory T cells, the cells defined by the gene FOXP3 that keep attacks on the body in check.[a] Some of the strongest evidence that microbes shape the immune system concerns exactly those cells.

Mice raised in sterile isolators, with no microbes at all, have stunted immune systems. In 2005 Sarkis Mazmanian and colleagues showed that a single molecule, a sugar polymer made by the common gut bacterium Bacteroides fragilis, corrected several of these defects; a mutant bacterium without it did not.[12] In 2011 Koji Atarashi, Kenya Honda and colleagues found that regulatory T cells were most abundant in the lining of the colon, and that spore-forming bacteria of the Clostridium group drove their accumulation. Mice given these bacteria early in life resisted colitis and allergic antibody responses as adults.[13] Two years later the same group isolated 17 strains from a healthy person’s faeces that, given to germ-free mice, expanded regulatory T cells and reduced colitis and allergic diarrhoea.[14]

The human version of the story is older and less tidy. In 1989 David Strachan, studying some 17,000 British children, found that the more older siblings they had, the less likely they were to get hay fever, and proposed that allergy might be “prevented by infection in early childhood, transmitted by unhygienic contact with older siblings”. His title, “Hay fever, hygiene, and household size”, gave the idea its name.[15][2] The “hygiene hypothesis” has been reshaped many times since, and Yong points out that its name misleads: later versions put the weight on “old friends”, the microbes of mud, animals and other people, lost to smaller families, cities and distance from livestock, more than to personal cleanliness. The most striking modern test is a comparison of two farming communities in the United States. The Amish and the Hutterites share ancestry and much of their way of life, but the Amish farm in the traditional way and the Hutterites use industrial methods. Among 60 children studied, asthma was 4 times and allergic sensitisation 6 times less common among the Amish, whose house dust held 6.8 times as much bacterial endotoxin. Dust from Amish homes protected mice from experimental asthma; dust from Hutterite homes did not, and the protection needed the innate immune system’s signalling.[16][b]

That is strong, but it is a comparison of two communities, and the protective ingredient has not been isolated. Ten years on, the claim that microbes train the immune system has held up firmly in mice and is very likely true in people; what is still missing is a way to use it, a product or a practice proven to prevent allergy or autoimmune disease.

The cures

Where microbes worked as medicine

The most solid success is the least glamorous: curing chronic diarrhoea with someone else’s faeces. Clostridioides difficile infection often follows a course of antibiotics, which clears away the bacteria that normally keep it out, and it returns again and again in many patients. In a Dutch trial published in 2013, 13 of 16 patients given an infusion of donor faeces were cured by the first infusion, against 4 of 13 given the antibiotic vancomycin alone and 3 of 13 given vancomycin with a bowel wash; the trial was stopped after an interim analysis.[17] The book reported the same trial as 94% cured against 27%, counting a second infusion and pooling the two vancomycin groups.[2] A capsule of purified bacterial spores taken from donor faeces, tested against placebo in 182 patients, cut the rate of recurrence from 40% to 12%.[18] The US Food and Drug Administration approved a faecal microbiota product, Rebyota, on 30 November 2022,[19] and that capsule, Vowst, on 26 April 2023, as the first such product taken by mouth.[20] The next step is to leave the faeces out altogether. VE303, a mixture of eight cultured strains of Clostridia, cut recurrence to 13.8% against 45.5% on placebo at the higher of two doses, in a trial of 79 patients; a larger trial is still needed.[21]

The book judged faecal transplants for inflammatory bowel disease less convincing, with success rates “lower and more inconsistent”.[2] For ulcerative colitis the evidence has improved since. Pooling six double-blind trials with 324 patients, transplants roughly quadrupled the odds of remission confirmed both by symptoms and by endoscopy, with side effects no more common than on placebo. The trials were short, and whether the benefit lasts is unknown.[22]

The other clear success is not in humans at all. Aedes aegypti mosquitoes infected with the wMel strain of Wolbachia transmit dengue poorly, and the bacteria spread through wild populations by manipulating the mosquitoes’ reproduction. In a randomised trial in Yogyakarta, Indonesia, 12 areas of the city received releases of these mosquitoes and 12 did not. Among patients with fever at local clinics, 2.3% had dengue in the treated areas against 9.4% in the others: a protective efficacy of 77%. Hospitalisation for dengue fell by 86%.[23] Releases of mosquitoes cannot be hidden, so the trial was not blinded as a drug trial would be, although the staff who tested patients did not know where they lived; for three years it was the only completed randomised trial of the method.[24] In 2026 a second one, in Singapore, used Wolbachia the other way round, releasing infected males whose matings with wild females produce no offspring, to shrink the mosquito population. Among people tested for dengue in the release areas 6% were positive, against 21% elsewhere: a protective efficacy of 71 to 72%.[25] The book ends with the first releases of these mosquitoes in Cairns, Australia, in 2011, and with Scott O’Neill’s own caution: he could “declare victory only when his mosquitoes repress dengue in the countries where it’s most prevalent”, which was why the work was expanding to Indonesia and elsewhere.[2] The Yogyakarta trial was that test, and it passed.

Fig. 2 — Four results from randomised trials, as the share of patients with the bad outcome in each arm. Dengue: share of fever patients at clinics whose illness was dengue, and share hospitalised with dengue (intention to treat). Sources: van Nood et al. 2013; Feuerstadt et al. 2022; Utarini et al. 2021.

Both successes share a feature. They work where a disease follows a clear ecological failure: an empty gut that a pathogen colonises, or an insect that carries a virus. They replace a missing community, or add a competitor in one place. Yong saw this in 2016. “C-diff is the exception that proves the rule,” he wrote: antibiotics leave a wasteland in which a donor’s microbes “find few competitors”, and “if you wanted to design a disease that could be easily treated by FMT, you’d create something like a C-diff infection”.[2] Repairing a whole microbiome to treat a complex disease is a much harder thing, and it is in the complex diseases that the story ran ahead of the evidence.

The mice

What mice can and cannot show

The experiment that made the field famous was simple. Take germ-free mice, give them the gut microbes of a fat or a lean donor, and see what happens. In 2006 Peter Turnbaugh, Jeffrey Gordon and colleagues found that mice colonised with the microbes of genetically obese mice gained more body fat than mice given the microbes of lean ones.[26] In 2013 the same group did it with human twins in which one was obese and the other lean: mice given the obese twin’s microbes gained more fat, and mice housed with “lean” cage mates were protected, depending on their diet.[27]

These are good experiments, and the book described them with their limits. “Even the sincerely proposed links are mostly correlations,” Yong wrote of the long list of diseases tied to the microbiome; the mouse transfers went further, but “even if experiments show that microbes can cause diseases in mice and other animals, we still don’t know if they actually do so in people.” He also reported that one of the field’s most famous results, a higher ratio of Firmicutes to Bacteroidetes in obese people, had turned out to be “a mirage” when the studies were pooled.[2] The trouble is what came after. In 2020 Jens Walter, Anissa Armet, Brett Finlay and Fergus Shanahan reviewed every study they could find that transferred human gut microbes into germ-free rodents. Of 38 published studies, 36 reported that the disease of the human donor, whatever it was, had been transferred to the animals. They called this rate “implausible” and concluded that it “overstates the role of the gut microbiome in human disease”.[28] A success rate like that usually means that failures go unpublished, or that experiments are built in ways that favour the expected result.

Fig. 3 — Published studies that transferred human gut microbes into germ-free rodents, and how many reported that the donor’s disease came with them. Source: Walter et al. 2020.

When the obesity idea was tested in people, it did not show. The book reported a small Dutch trial in which 18 obese volunteers received either their own gut microbes or a lean donor’s: those given lean microbes became more sensitive to insulin, “but they didn’t lose any weight”.[2] In a later placebo-controlled trial at a US medical centre, 24 adults with obesity were given weekly capsules of faecal microbes from lean donors, or placebo, for six weeks. The donors’ bacteria took hold in most recipients and stayed for at least 12 weeks, but there was no clinically significant change in insulin sensitivity, body fat or the other metabolic measures.[29] It was a small trial, and its authors say so; it does not prove that gut microbes play no part in obesity. It shows that the mouse result does not translate simply. Pooled analyses point the same way: across nine randomised trials with 303 people, transplants from lean donors brought small short-term improvements in blood sugar, insulin and HDL cholesterol, and no difference in weight.[30]

The link between gut and brain is where caution is most needed. Large studies find real correlations: in the Flemish Gut Flora Project, with 1,054 people, two groups of bacteria were consistently depleted in people with depression, even after allowing for antidepressants.[31] But correlation runs in both directions, since mood changes diet and diet changes microbes. Autism is the clearest case. The book told how Sarkis Mazmanian’s team gave B. fragilis to mice with autism-like behaviours and reversed most of them, and how mice given the gut microbes of autistic children developed similar behaviours; it also quoted the sceptic Emily Willingham, that “mice don’t have autism”.[2] The best-designed test in people since, with 247 participants, found “negligible direct associations” between the diagnosis and the gut microbiome; the differences followed from the restricted diets that often go with autism, and the authors cautioned “against claims that the microbiome has a driving role”.[32] Yong’s warnings were in the book; much of what was written about the microbiome after it did not take them.

Trials of treatments are the better test, and they have begun. Pooling 20 randomised trials in people with diagnosed depression or anxiety, probiotics gave a large average reduction in depressive symptoms, but the trials were small, 1,401 patients in all, and their results varied widely.[33] In autism, eight trials with 318 children and young people found a modest average improvement in behaviour.[34] The largest controlled study of faecal transplants in autistic children that I found had 30 children, was not blinded, and saw no difference at eight weeks in its main clinician-scored measure, although parents’ ratings improved; its authors call it exploratory.[35] None of this shows yet that changing the microbiome treats a disorder of the brain. It is no longer only mice and correlations.

Pills and plates

Probiotics, diet and the shape of a life

If the microbiome matters, the obvious product is a pill of good bacteria. The book was blunt about them. Even the strongest products hold a few hundred billion bacteria, against a gut that “already holds at least a hundredfold more”, and most are species chosen because they are easy to grow, not because they live in adult guts. Yong summed up the Cochrane reviews: probiotics shorten infectious diarrhoea, reduce the diarrhoea that follows antibiotics and save premature babies from necrotising enterocolitis. “And there ends the list. Compared to the hype, it’s modest.”[2] Later work made the case against them for healthy adults stronger. In 2018 Eran Elinav, Eran Segal and colleagues examined the gut lining itself, by endoscopy, not just stool. In mice the resident microbes resisted an 11-strain probiotic; in people, whether the bacteria took hold depended on the person, the part of the gut and the strain, and the effect was “transient” and individual.[36] In a companion study, people who took probiotics after a course of antibiotics recovered their own microbiome more slowly and less completely than those who took nothing, while a transplant of their own stored faeces restored it within days.[37]

The exception has held, but it has become contested. In very premature babies, the 2023 Cochrane review pooled 60 trials with 11,156 infants: probiotics may roughly halve necrotising enterocolitis, a bowel disease that often kills, on low-certainty evidence, and probably reduce deaths slightly; about 33 babies have to be treated to prevent one case.[38] The FDA warned in 2023 against their routine use, citing safety and the quality of products sold as supplements;[39] paediatric societies in Europe disagreed, arguing for “a more nuanced and supportive approach”.[40] The disagreement is about product quality and regulation as much as about the biology.

Diet is more promising, and less marketable. In a 17-week randomised study with 18 people in each arm, a diet rich in fermented foods steadily increased the diversity of the gut microbiome and lowered markers of inflammation; a diet rich in fibre did not raise diversity, although it changed what the microbes could digest.[41] The study was small, it had no control group, comparing the two diets with each other, and its main measure of immune response did not change in either arm: the fall in inflammatory markers was a secondary finding. The fermented-food group went from under one serving a day to about six.[41] In mice carrying human gut microbes, a diet low in fibre caused losses that could be reversed within one generation, but over several generations species went extinct and did not return when fibre came back; they had to be reintroduced.[42] The Hadza, hunter-gatherers in Tanzania, have a microbiome that cycles with the seasons, and the species that come and go with their seasons are the same ones that distinguish traditional from industrialised populations.[43]

What shapes a person’s microbiome? The Human Microbiome Project found in 2012 that even healthy people “differ remarkably” in the microbes of their gut, skin and vagina, while the metabolic pathways those microbes carry were stable from person to person.[44] In 1,046 healthy people of several ancestries, host genetics played only a minor role, while people who shared a household had similar microbes even when unrelated; diet, drugs and body measurements were associated with over 20% of the differences between people.[45] In more than 2,000 Belgians and Dutch, medicines explained the largest share of the variation, and the way a person was born left no trace in the adult microbiome.[46] That bears on one of the book’s worries. Babies born by caesarean section get their first microbes from their mother’s skin and the hospital rather than her vagina, and Yong reported the suggestion that this might explain why they are more likely to develop allergies, asthma, coeliac disease and obesity, while noting that “it’s not clear what these differences mean in the long term”.[2] A study of 596 British babies confirmed that caesarean birth changes the first weeks, with fewer of the mother’s Bacteroides and more hospital bacteria,[47] but by adulthood the difference is not visible. Whether those weeks matter for later health is still unknown.

Personal variation is real. In an 800-person study, people’s blood sugar responses to identical meals varied widely, and an algorithm that used, among other things, their gut microbes predicted those responses; in a small blinded trial, diets built on its predictions lowered blood sugar after meals.[48]

Cancer

Microbes and the brakes

One of the most surprising findings since the book connects the microbiome to The Brakes Released, the essay on cancer drugs that release the immune system’s brakes.[c] In January 2018 three groups reported in the same issue of Science that the gut microbes of patients predicted their response to anti-PD-1 drugs. In France, antibiotics reduced the benefit of the drugs, and the bacterium Akkermansia muciniphila was associated with response.[49] In Houston, among melanoma patients, responders had more diverse gut microbes.[50] In Chicago, different species were more abundant in responders.[51] In each case, mice given the faeces of responders controlled tumours better.

The next step was to test the idea in patients whose melanoma had stopped responding to the drugs. In two small early-stage trials published in 2021, faecal transplants from patients who had responded, followed by the drug again, produced responses in 3 of 10 patients in Israel[52] and clinical benefit in 6 of 15 in Pittsburgh.[53] Those are patients for whom the drugs had failed, so the results are notable, but neither trial had a control group.

Randomised evidence arrived in 2026. In the TACITO trial, 45 patients with advanced kidney cancer starting standard immunotherapy received, double-blind, either a transplant from a donor who had responded completely to such drugs or a placebo transplant. After a year, 70% of the transplant group and 41% of the placebo group were free of progression, a difference that just missed the trial’s threshold for significance; the median time before the cancer worsened was 24 months against 9, which did reach it.[54] In a single-arm trial published alongside it, patients given a capsule transplant before first-line immunotherapy responded at high rates: 16 of 20 with lung cancer and 15 of 20 with melanoma. What marked out the responders was not which donor bacteria took hold, but which of their own species disappeared.[55]

So this has moved from association to a treatment effect measured against a placebo, in one trial of 45 patients that missed its own main target. The bacteria credited with the effect still differ from study to study, and no microbial treatment for cancer has been approved. It is the most promising medical frontier of the microbiome, and it is still early.

Ten years on

What is known and what is hoped

Put together, the scorecard is mixed in a way that says something about the field (Figure 4), and something about the book. What held up is ecology: animals built with microbial partners, immune systems trained by them, and treatments that restore a missing community or add a competitor in one specific place. What failed or remains unproven is the story told on top of it, in which a disrupted microbiome lies behind obesity, depression, autism and much else, and a pill or a diet can set it right. The last five years have not overturned that summary, but they have narrowed it: where microbial treatments are now being tested properly, in ulcerative colitis, in cancer, in depression, the early results are better than the sceptical reading of 2021 would have predicted, and still short of proof.

Fig. 4 — Eight claims of the microbiome story around 2016 and the state of the evidence in 2026, as judged in this essay. Sources: as in the text.

The trap was the mouse. A germ-free mouse is a powerful tool for showing that microbes can do something, and a poor guide to whether they do it in people living ordinary lives. Many of the most quoted findings came from transferring human microbes into such mice, and the rate at which they “worked” should have been a warning in itself. The fair summary is not that the microbiome was hype. It is that the microbiome is real, important and mostly ecological, and that ecology resists simple products. The book holds up better than the field around it. Its stories about other animals were solid, its count was already right, and its caveats on obesity, autism, probiotics and faecal transplants have mostly been borne out. Where it now reads as too hopeful is in its tone more than its claims: the promise that we might “control these multitudes for our own benefit” has been kept, so far, mainly for C. difficile and dengue.[2]

What changes, then, about being one person? Less than the ten-to-one figure suggested, and more than the corrected count implies. The cells are roughly even, but most of the genes at work in a gut are microbial, the microbes are needed to build a normal immune system, and the boundary between self and partner is, in other animals, sometimes impossible to draw. Yong’s “grander view of life” was right in its main claim: an animal is a crowd that has learned to live together. What the last ten years added is a sense of how hard it is to change that crowd on purpose, and how often the attempts work only where the crowd was missing in the first place.

Go deeper

Ideas this essay uses without stopping to explain them, and where this blog does. The letters in the text point here; the links go straight to the section.

[a] Regulatory T cells

The immune cells that restrain attacks on the body, and the gene FOXP3 that defines them. The Second Lock, “Terms used across the series”.

[b] Innate immune system

The fast, inherited first line of defence that senses molecules typical of microbes. The Second Lock, “Terms used across the series”.

[c] Checkpoint drugs

Antibodies such as anti-PD-1 that release a brake on T cells so they attack a tumour. The Second Lock, “Terms used across the series”.

On method and tools

This essay was researched and written with Claude Opus 5.5 (Anthropic): human specification, editorial direction and critical review; machine research, drafting and figure generation. The subject came from Luis, who was reading Ed Yong’s I Contain Multitudes; quotations from the book were checked against the US e-book edition. Figures and quotations from research come from the abstracts of the papers cited, each checked against its PubMed record, except for three read in full: Sender, Fuchs and Milo, for the cell counts, the 1972 estimate and the effect of a trip to the toilet; Utarini and colleagues and Wastyk and colleagues, whose limitations are reported here from their own texts. The papers published since 2021 were found by searching PubMed in eight areas where the evidence seemed most likely to have moved, which is not a systematic review; the papers cited from before then were chosen as landmarks of the field. Strachan’s 1989 paper has no abstract and I did not read it: its finding and the quotation are given as the book reports them. The Dutch obesity trial with 18 volunteers is also cited from the book, not from the paper. There is no model in this essay. scripts/multitudes.py draws the four figures from the published numbers, which it writes to docs/multitudes-results.json. The verdicts in Figure 4 are my judgement of the evidence cited, not a systematic review. The cover is drawn by the same script, one mark per hundred billion cells in the proportions of the 2016 count.

Authored by: Luis Matos Ferreira — Physicist, Developer, Writer

Related essays on this blog

1. The Second Lock — the immune system’s brakes, regulatory T cells and autoimmune disease.
2. Why So Many Locks — why the immune system checks and rechecks before it attacks.
3. Other Locks — could the immune system be designed better? Bacteria, plants, lampreys and engineered cells.
4. The Brakes Released — the same brakes released against cancer.
5. The Complete Parts List — why knowing every part of a living thing does not explain it.

Sources

1. W. Whitman, “Song of Myself”, section 51, in Leaves of Grass, poets.org.
2. E. Yong, I Contain Multitudes: The Microbes Within Us and a Grander View of Life (The Bodley Head, 2016; US edition, HarperCollins, 2016), prologue and chapters 3, 4, 5, 9 and 10.
3. R. N. Harris, R. M. Brucker, J. B. Walke et al., “Skin microbes on frogs prevent morbidity and mortality caused by a lethal skin fungus”, The ISME Journal 3, 818–824 (2009), doi.org.
4. S. Curry, “I Contain Multitudes by Ed Yong – full of life’s little surprises”, The Guardian, 25 August 2016, theguardian.com.
5. T. D. Luckey, “Introduction to intestinal microecology”, The American Journal of Clinical Nutrition 25, 1292–1294 (1972), doi.org.
6. R. Sender, S. Fuchs & R. Milo, “Revised Estimates for the Number of Human and Bacteria Cells in the Body”, PLoS Biology 14, e1002533 (2016), doi.org.
7. J. Qin, R. Li, J. Raes et al., “A human gut microbial gene catalogue established by metagenomic sequencing”, Nature 464, 59–65 (2010), doi.org.
8. S. V. Nyholm & M. J. McFall-Ngai, “The winnowing: establishing the squid-vibrio symbiosis”, Nature Reviews Microbiology 2, 632–642 (2004), doi.org.
9. S. Shigenobu, H. Watanabe, M. Hattori et al., “Genome sequence of the endocellular bacterial symbiont of aphids Buchnera sp. APS”, Nature 407, 81–86 (2000), doi.org.
10. M. A. Campbell, J. T. Van Leuven, R. C. Meister et al., “Genome expansion via lineage splitting and genome reduction in the cicada endosymbiont Hodgkinia”, Proceedings of the National Academy of Sciences of the United States of America 112, 10192–10199 (2015), doi.org.
11. R. Zug & P. Hammerstein, “Still a host of hosts for Wolbachia: analysis of recent data suggests that 40% of terrestrial arthropod species are infected”, PloS One 7, e38544 (2012), doi.org.
12. S. K. Mazmanian, C. H. Liu, A. O. Tzianabos et al., “An immunomodulatory molecule of symbiotic bacteria directs maturation of the host immune system”, Cell 122, 107–118 (2005), doi.org.
13. K. Atarashi, T. Tanoue, T. Shima et al., “Induction of colonic regulatory T cells by indigenous Clostridium species”, Science (New York, N.Y.) 331, 337–341 (2011), doi.org.
14. K. Atarashi, T. Tanoue, K. Oshima et al., “Treg induction by a rationally selected mixture of Clostridia strains from the human microbiota”, Nature 500, 232–236 (2013), doi.org.
15. D. P. Strachan, “Hay fever, hygiene, and household size”, BMJ (Clinical Research Ed.) 299, 1259–1260 (1989), doi.org.
16. M. M. Stein, C. L. Hrusch, J. Gozdz et al., “Innate Immunity and Asthma Risk in Amish and Hutterite Farm Children”, The New England Journal of Medicine 375, 411–421 (2016), doi.org.
17. E. van Nood, A. Vrieze, M. Nieuwdorp et al., “Duodenal infusion of donor feces for recurrent Clostridium difficile”, The New England Journal of Medicine 368, 407–415 (2013), doi.org.
18. P. Feuerstadt, T. J. Louie, B. Lashner et al., “SER-109, an Oral Microbiome Therapy for Recurrent Clostridioides difficile Infection”, The New England Journal of Medicine 386, 220–229 (2022), doi.org.
19. US Food and Drug Administration, “Rebyota”, product page with the approval letter of 30 November 2022, fda.gov.
20. US Food and Drug Administration, “FDA Approves First Orally Administered Fecal Microbiota Product for the Prevention of Recurrence of Clostridioides difficile Infection”, news release, 26 April 2023, fda.gov.
21. T. Louie, Y. Golan, S. Khanna et al., “VE303, a Defined Bacterial Consortium, for Prevention of Recurrent Clostridioides difficile Infection: A Randomized Clinical Trial”, JAMA 329, 1356–1366 (2023), doi.org.
22. N. El Hage Chehade, S. Ghoneim, S. Shah et al., “Efficacy of Fecal Microbiota Transplantation in the Treatment of Active Ulcerative Colitis: A Systematic Review and Meta-Analysis of Double-Blind Randomized Controlled Trials”, Inflammatory Bowel Diseases 29, 808–817 (2023), doi.org.
23. A. Utarini, C. Indriani, R. A. Ahmad et al., “Efficacy of Wolbachia-Infected Mosquito Deployments for the Control of Dengue”, The New England Journal of Medicine 384, 2177–2186 (2021), doi.org.
24. T. Fox, Y. Sguassero, M. Chaplin et al., “Wolbachia-carrying Aedes mosquitoes for preventing dengue infection”, The Cochrane Database of Systematic Reviews 4, CD015636 (2024), doi.org.
25. J. T. Lim, C. S. Chong, C. C. Chang et al., “Dengue Suppression by Male Wolbachia-Infected Mosquitoes”, The New England Journal of Medicine 394, 1175–1183 (2026), doi.org.
26. P. J. Turnbaugh, R. E. Ley, M. A. Mahowald et al., “An obesity-associated gut microbiome with increased capacity for energy harvest”, Nature 444, 1027–1031 (2006), doi.org.
27. V. K. Ridaura, J. J. Faith, F. E. Rey et al., “Gut microbiota from twins discordant for obesity modulate metabolism in mice”, Science (New York, N.Y.) 341, 1241214 (2013), doi.org.
28. J. Walter, A. M. Armet, B. B. Finlay et al., “Establishing or Exaggerating Causality for the Gut Microbiome: Lessons from Human Microbiota-Associated Rodents”, Cell 180, 221–232 (2020), doi.org.
29. E. W. Yu, L. Gao, P. Stastka et al., “Fecal microbiota transplantation for the improvement of metabolism in obesity: The FMT-TRIM double-blind placebo-controlled pilot trial”, PLoS Medicine 17, e1003051 (2020), doi.org.
30. B. Qiu, J. Liang & C. Li, “Effects of fecal microbiota transplantation in metabolic syndrome: A meta-analysis of randomized controlled trials”, PloS One 18, e0288718 (2023), doi.org.
31. M. Valles-Colomer, G. Falony, Y. Darzi et al., “The neuroactive potential of the human gut microbiota in quality of life and depression”, Nature Microbiology 4, 623–632 (2019), doi.org.
32. C. X. Yap, A. K. Henders, G. A. Alvares et al., “Autism-related dietary preferences mediate autism-gut microbiome associations”, Cell 184, 5916–5931.e17 (2021), doi.org.
33. A. Asad, M. Kirk, S. Zhu et al., “Effects of Prebiotics and Probiotics on Symptoms of Depression and Anxiety in Clinically Diagnosed Samples: Systematic Review and Meta-analysis of Randomized Controlled Trials”, Nutrition Reviews 83, e1504–e1520 (2025), doi.org.
34. S. Soleimanpour, M. Abavisani, A. Khoshrou et al., “Probiotics for autism spectrum disorder: An updated systematic review and meta-analysis of effects on symptoms”, Journal of Psychiatric Research 179, 92–104 (2024), doi.org.
35. D. Varnas, A. Kunevičius, A. Burokas et al., “Fecal Microbiota Transplantation for Autism Spectrum Disorder in Children: Results from a Prospective Open-Label Controlled Observational Study”, Medicina (Kaunas, Lithuania) 62 (2025), doi.org.
36. N. Zmora, G. Zilberman-Schapira, J. Suez et al., “Personalized Gut Mucosal Colonization Resistance to Empiric Probiotics Is Associated with Unique Host and Microbiome Features”, Cell 174, 1388–1405.e21 (2018), doi.org.
37. J. Suez, N. Zmora, G. Zilberman-Schapira et al., “Post-Antibiotic Gut Mucosal Microbiome Reconstitution Is Impaired by Probiotics and Improved by Autologous FMT”, Cell 174, 1406–1423.e16 (2018), doi.org.
38. S. Sharif, N. Meader, S. J. Oddie et al., “Probiotics to prevent necrotising enterocolitis in very preterm or very low birth weight infants”, The Cochrane Database of Systematic Reviews 7, CD005496 (2023), doi.org.
39. S. Viswanathan & K. S. Gautham, “Navigating the United States FDA advisory: Probiotics in Preterm Infants”, Journal of Perinatology 46, 874–877 (2026), doi.org.
40. C. H. P. van den Akker, N. D. Embleton, A. Lapillonne et al., “Reevaluating the FDA's warning against the use of probiotics in preterm neonates: A societal statement by ESPGHAN and EFCNI”, Journal of Pediatric Gastroenterology and Nutrition 78, 1403–1408 (2024), doi.org.
41. H. C. Wastyk, G. K. Fragiadakis, D. Perelman et al., “Gut-microbiota-targeted diets modulate human immune status”, Cell 184, 4137–4153.e14 (2021), doi.org.
42. E. D. Sonnenburg, S. A. Smits, M. Tikhonov et al., “Diet-induced extinctions in the gut microbiota compound over generations”, Nature 529, 212–215 (2016), doi.org.
43. S. A. Smits, J. Leach, E. D. Sonnenburg et al., “Seasonal cycling in the gut microbiome of the Hadza hunter-gatherers of Tanzania”, Science (New York, N.Y.) 357, 802–806 (2017), doi.org.
44. Human Microbiome Project Consortium, “Structure, function and diversity of the healthy human microbiome”, Nature 486, 207–214 (2012), doi.org.
45. D. Rothschild, O. Weissbrod, E. Barkan et al., “Environment dominates over host genetics in shaping human gut microbiota”, Nature 555, 210–215 (2018), doi.org.
46. G. Falony, M. Joossens, S. Vieira-Silva et al., “Population-level analysis of gut microbiome variation”, Science (New York, N.Y.) 352, 560–564 (2016), doi.org.
47. Y. Shao, S. C. Forster, E. Tsaliki et al., “Stunted microbiota and opportunistic pathogen colonization in caesarean-section birth”, Nature 574, 117–121 (2019), doi.org.
48. D. Zeevi, T. Korem, N. Zmora et al., “Personalized Nutrition by Prediction of Glycemic Responses”, Cell 163, 1079–1094 (2015), doi.org.
49. B. Routy, E. Le Chatelier, L. Derosa et al., “Gut microbiome influences efficacy of PD-1-based immunotherapy against epithelial tumors”, Science (New York, N.Y.) 359, 91–97 (2018), doi.org.
50. V. Gopalakrishnan, C. N. Spencer, L. Nezi et al., “Gut microbiome modulates response to anti-PD-1 immunotherapy in melanoma patients”, Science (New York, N.Y.) 359, 97–103 (2018), doi.org.
51. V. Matson, J. Fessler, R. Bao et al., “The commensal microbiome is associated with anti-PD-1 efficacy in metastatic melanoma patients”, Science (New York, N.Y.) 359, 104–108 (2018), doi.org.
52. E. N. Baruch, I. Youngster, G. Ben-Betzalel et al., “Fecal microbiota transplant promotes response in immunotherapy-refractory melanoma patients”, Science (New York, N.Y.) 371, 602–609 (2021), doi.org.
53. D. Davar, A. K. Dzutsev, J. A. McCulloch et al., “Fecal microbiota transplant overcomes resistance to anti-PD-1 therapy in melanoma patients”, Science (New York, N.Y.) 371, 595–602 (2021), doi.org.
54. S. Porcari, C. Ciccarese, V. Heidrich et al., “Fecal microbiota transplantation plus pembrolizumab and axitinib in metastatic renal cell carcinoma: the randomized phase 2 TACITO trial”, Nature Medicine 32, 1316–1324 (2026), doi.org.
55. S. Duttagupta, M. Messaoudene, S. Hunter et al., “Fecal microbiota transplantation plus immunotherapy in non-small cell lung cancer and melanoma: the phase 2 FMT-LUMINate trial”, Nature Medicine 32, 1337–1350 (2026), doi.org.