In February 2021, a pair of remarkable clinical trials was published1,2. Each one involved people with melanoma. Some participants had been successfully treated using drugs called immune checkpoint inhibitors, which help the body to destroy cancer cells. The others had failed to respond to the treatment.
In both studies, researchers collected stool samples from the people who had benefited from the immunotherapy. They then implanted these samples — and the gut bacteria that they contained — into the people who had not responded to the drugs, and administered the checkpoint inhibitors again. The hope was that this faecal microbiota transplant (FMT) would transfer the ability to respond to this potentially life-saving treatment.
Experiments in mice had already suggested that differences in the composition of people’s intestinal microbiomes might account for much of the notorious variability in responses to checkpoint inhibitors3. Mice implanted with gut microbes from people who had responded positively to immunotherapy tended to respond well, too. But when transplants came from non-responders, the drugs were ineffectual.
Unconventional as the clinical trials were, they attracted many volunteers, says Hassane Zarour, an oncologist at the University of Pittsburgh, Pennsylvania, who led one of them. “People are really eager either to try something else when they have no other option, or to help others with something which is easy to give.”
In Zarour’s study, 6 of the 15 people who had received FMT benefited from the same form of immunotherapy that had previously done nothing for them1. In the second trial, led by researchers at the Sheba Medical Center in Tel HaShomer, Israel, three of the ten participants became responders after receiving a transplant2. “The data were very impressive and very convincing, even if very preliminary,” says Zarour.
Last year, Bertrand Routy, an oncologist at the University of Montreal, Canada, co-led a study involving 20 people with melanoma4. After receiving FMT, 13 responded to checkpoint inhibitors, with 4 entering complete remission. However, this time, the implanted gut microbes came from healthy volunteers, rather than people who had already responded well to immunotherapy — expanding the potential pool of stool donors enormously.
Preliminary results from a trial in which FMT was administered before first-line immunotherapy with two checkpoint inhibitors also indicate success — around three-quarters of people responded to the drugs, compared with a historical average of 58% when FMT is not used5. “It was associated with really strong activity,” says University of Montreal oncologist Arielle Elkrief. The group will now start a large randomized study in people with melanoma.
But many oncologists want to take this idea further — so much so that FMT could be rendered obsolete almost as quickly as it emerged. “FMT is a stepping stone,” says Jennifer Wargo, a cancer researcher at the MD Anderson Cancer Centre in Houston, Texas. She and others instead want to create standardized packages of known bacteria to modify the gut microbiome in a more controlled fashion. Other researchers intend to abandon live bacteria entirely and develop drugs that target the molecular mechanisms by which the microbiome affects immunotherapy responses.
FMT has provided a proof of concept that changing the gut microbiome of someone with cancer can save their life, and has set the benchmark against which these prospective treatments will be judged.
Microbial effects
Cancer immunotherapies allow the immune system to recognize malignant cells as foreign objects and use T cells to destroy them. Checkpoint inhibitors bind to and nullify the checkpoint proteins that certain cancers manipulate to suppress T-cell activity. Under the right circumstances, this frees these immune cells to attack and eliminate cancers.
Clinical trials in the 2000s showed that checkpoint inhibitors were effective against some previously untreatable tumours. However, responses to these drugs vary dramatically. The chances of success are highest with melanoma: the likelihood of the treatment working amounts roughly to a coin toss. In lung cancer, it’s closer to the odds of rolling a six on a die. “There’s a difference between patients in terms of the magnitude of endogenous immune responses,” says Thomas Gajewski, an oncologist at the University of Chicago in Illinois.
To improve the odds, researchers needed to understand what shapes the immune responses of people with cancer. Numerous factors influence this — not least how immunogenic the tumour itself is. But it had also long been understood that the ecosystem of trillions of bacteria living in people’s guts affects immune function.
In 2015, Gajewski’s group demonstrated that mice with melanoma that had distinct intestinal microbiomes would respond to treatment with checkpoint inhibitors differently. One bacterial genus — Bifidobacterium — seemed to be key to their ability to respond6. At the same time, a team at the Gustave Roussy Cancer Campus in Paris showed that introducing Bacteroidales bacteria into mice with no intestinal microbes could get them to respond7.
Three years later, Gajewski and his colleagues showed that the guts of people who responded to immunotherapy contained different bacterial species from the guts of those who didn’t, and that transplanting the participants’ microbiomes into mice generally made the rodents respond to treatment in the same way as the donor8. These findings were published alongside two similar papers — one from the Parisian group3, and another from Wargo’s team9.
Gajewski says the human-to-mouse transplant experiments suggest that, in around half of the people who don’t respond to checkpoint inhibitors, the gut microbiome is to blame. “If you took care of the microbiome, you could double the response rate.”
This alluring possibility could see gut-microbiome screening become a routine precursor to immunotherapy — if it is proved correct. But developing reliable treatments is a challenge. “The more we advance, the more complex we realize this ecosystem is,” says Routy.
Inconsistent medicine
Wargo sees ways in which these findings could quickly change clinical practice. In 2021, her group showed that people with high-fibre diets responded better to checkpoint inhibitors than did those with low-fibre diets10. Her colleagues at MD Anderson are working on trials to formally assess the effectiveness of high-fibre diets, and Wargo thinks that increasing fibre intake could soon become standard advice for people taking checkpoint inhibitors.
Several studies have also linked antibiotics — which can severely deplete the gut microbiome — to poorer immunotherapy outcomes11. Reducing their use in people undergoing immunotherapy could help, Wargo says. When antibiotics can’t be avoided, restorative action might be needed. “If someone’s seen broad-spectrum antibiotics, they may need something as drastic as a whole faecal transplant to really be able to respond,” she says.
The encouraging FMT trial results from Zarour, Routy, Elkrief and others have intensified efforts to develop ways to directly transform the gut microbiomes of people with cancer. The poster child for FMT is its use for treating recurrent Clostridium difficile infections. Transplants from healthy donors are an effective remedy for this debilitating intestinal disease, and the procedure has since 2013 had regulatory approval in the United States.
However, practical challenges — such as difficulties recruiting donors, demanding screening requirements and unease among physicians — have limited clinical uptake. Physicians also worry that human stool is an inconsistent medicine. “You don’t know if a donor has the right bacteria, and you don’t know which bacteria the recipient needs,” says Gajewski.
Notably, two companies last year gained approval from the US Food and Drug Administration for specific bacterial consortia to treat recurrent C. difficile infections. Defined bundles of microbes should be more reliable and replicable treatments, and can negate certain FMT-related safety concerns, such as the potential for pathogen transfer.
These developments are inspiring many researchers trying to create microbiome-targeted interventions to augment cancer immunotherapy.
“We’ve read the C. diff. book,” says Nadim Ajami, a microbiome researcher who is working with Wargo at MD Anderson to develop bacterial consortia for immunotherapy. “Getting down to a more scalable and reproducible drug is really important.”
However, initial efforts to create bacterial consortia have failed. Ajami says most companies involved have either moved into other areas or folded altogether.
One problem is that the studies that found differences between the microbiomes of responders and non-responders in small cohorts failed to converge on a set of gut microbes that robustly predict immunotherapy responses across study sites. So, in 2022, gastroenterologist Laura Bolte at the University of Groningen in the Netherlands co-led a meta-analysis that combined data from several sources to examine the microbiomes and clinical outcomes of 165 people12. She hoped that the large sample size would bring clarity, but “it’s not that easy”, she says. “It’s more complex than just species differing between responders and non-responders.”
Bolte and her colleague Johannes Björk, a microbiome researcher at the University of Groningen, think it might be necessary to go beyond the species level and also catalogue the subspecies and strains of bacteria present. This hypothesis is supported by a small study published in March that found this finer analysis improved predictions of checkpoint-inhibitor responses in people with some cancers13. Another possibility is that researchers should focus on the functions of intestinal bacteria. For example, two microbial species might have the same effects if they both release the same metabolite.
In contrast to Bolte’s work, a 2022 meta-analysis by Zarour and his colleagues was able to link some bacterial species to positive responses to checkpoint inhibitors. However, it uncovered a much stronger signature in non-responders, Zarour says — raising the possibility that removing unhelpful microbes might be more important than delivering ‘good’ ones14.
Researchers hope that interventional studies will reveal what works in practice. For example, Routy stresses that a central part of ongoing FMT clinical trials is carefully monitoring microbial changes to pin down which bacteria might be the most effective components of a standardized microbial therapy. “We’re going to be really understanding how these interventions are working,” he says. “We can then go on to develop the next generation of microbiome therapeutics.”
Other groups are using human-to-mouse FMT to characterize effective bacteria. Gajewski and his colleagues are mindful of the fact that they’ll probably have to find a way to deal with problematic gut microbiomes, and have recreated the microbiomes of non-responding people in mice in an effort to discover bacteria that can “compete with what’s there already and flip the system, so the mice become responders”.
Gajewski predicts that some degree of personalization will be required, and that various bacterial consortia might need to be developed to meet individuals’ needs. He hopes to begin clinical trials of tailored bacterial consortia in the next two years.
In Houston, Ajami says that, every few months, he welcomes to MD Anderson a woman who is currently in remission from colon cancer following checkpoint-inhibitor treatment three years ago. She has already donated stool for use in an ongoing FMT clinical trial, but is now providing samples to Ajami’s laboratory, too. The scientists use techniques designed to maintain the complex bacterial ecosystems that thrive in the gut’s airless conditions. Within 15 minutes of a bowel movement, her gut bacteria are cultured in anaerobic chambers. Later, these are introduced into mouse models of cancer, and their impact on immunotherapy outcomes is assessed. Ajami hopes to begin a human trial of microbial consortia next year.
Drugs not bugs
Some researchers, however, are unconvinced that administering microbes — even in carefully calculated mixtures — is the best strategy. “I’m personally hesitant about using live bacteria,” says Francesca Gazzaniga, who studies the physiological impacts of the microbiome at Harvard Medical School in Boston, Massachusetts. She worries that newly introduced bacteria could either die or lose their antitumour properties inside people with problematic microbiomes.
Instead, Gazzaniga favours bypassing bacteria altogether. She hopes to uncover the mechanisms by which the gut microbiome influences cancer immunity, and design conventional drugs that use the same mechanisms to promote positive immunotherapy outcomes. “Let’s just completely circumvent the need to have the right microbiome,” she says.
However, the exact pathways that these drugs would need to target, Gajewski says, remain “a little bit of a black box”. Several potential mechanisms exist, he explains — including direct interactions between bacteria and gut immune cells, interactions between bacterial metabolites and gut immune cells and bacterial metabolites entering the circulation and acting systemically.
In 2015, Gazzaniga, then an immunology postdoc at Harvard, decided to tackle this problem with fellow postdoc and cancer specialist Joon Seok Park. They first looked for differences in immune-gene expression between mice whose intestines had been colonized with healthy human microbiota and responded to checkpoint inhibitors, and mice whose microbiomes had been obliterated with antibiotics and did not respond.
The researchers found that, in certain immune cells in draining lymph nodes of antibiotic-treated mice, levels of a little-studied protein called PD-L2 were markedly elevated15. This protein is a cousin of PD-L1, a checkpoint-activating protein made by certain cancer cells, whose T-cell-suppressing actions are the target of most currently used checkpoint inhibitors. What’s more, the T cells of mice that did not respond to checkpoint inhibitors sported elevated levels of an inhibitory receptor activated by PD-L2.
Gazzaniga then asked whether supplementing a typical checkpoint inhibitor with antibodies that blocked PD-L2 made these normally insensitive mice respond to the immunotherapy. It was a success: together, the two drugs shrank tumours, even in mice with no gut microbes. She also showed that mice colonized with gut bacteria from some of Wargo’s non-responding trial participants responded to checkpoint inhibitors when given PD-L2-targeting antibodies. The findings suggest that some people with cancer could benefit from combining PD-L2-targeting drugs with existing immunotherapy options. “It has the potential to work in a lot of cases,” Gazzaniga says.
“There are going to be many mechanisms by which the microbiome can impact antitumour immunity, and this was one of them,” she adds. “We need to keep looking for others.”
Uncertain future
Antibodies to PD-L2 and defined mixtures of live bacteria are likely to be more scalable and have a more consistent composition than FMT, and their developers are confident that they will succeed. If so, FMT might become nothing more than a footnote in the development of better therapies.
Presented with this view of the future, Routy and Elkrief both agree that using FMT for C. difficile infections has revealed many challenges. “There are a lot of logistical hurdles associated with such a programme,” Elkrief says. Most biotechnology companies are uninterested in FMT, she adds, and would prefer to develop more targeted treatments.
But Routy thinks that if trials of FMT for cancer keep yielding positive data, scientists and physicians will find ways to put it to clinical use. Whatever the excitement associated with other approaches, he says, so far FMT is the only strategy with robust clinical evidence supporting its ability to improve immunotherapy outcomes.
“We have now patients that are still alive, cancer free, four years after the first trial,” he says. Ultimately, this is the metric that matters most.