In his 45 years as a physician, Arthur Frankel has witnessed a striking evolution in available cancer treatments, and in patient outcomes. Not so long ago, for instance, “metastatic melanoma was truly a horrible disease,” says the University of South Alabama physician-researcher. In the US, more than 10,000 people each year were diagnosed with the cancer, which starts in the skin and has typically carried little chance of survival once it spreads to the lymph nodes and internal organs. With the recent approvals of several checkpoint inhibitors and another class of drugs known as BRAF inhibitors, however, Frankel says he saw “dramatic responses” in some of his patients: their tumors would disappear, and the patients would go into years-long remissions—but not all of them. The response rate of melanoma patients to one common combination immunotherapy, Nivolumab plus Ipilimumab, is less than 60 percent.


There was already some support for the idea. Taking a course of antibiotics shortly before an immunotherapy had been linked with poor patient response, for example, and studies in animal models supported a causal relationship. In 2013, Giorgio Trinchieri and Romina Goldszmid of the National Cancer Institute (NCI) and colleagues reported that germ-free mice or animals treated with antibiotics before being given either a cocktail of immunotherapies or platinum-based chemotherapy drugs were less able to fend off tumors.And two years later, two separate research groups published studies that documented different rates of response to immunotherapies in mice with melanoma2 and other cancers3 depending on their microbiota.

More than a dozen types of cancers for which immunotherapies on the market or in clinical trials can drive tumors into remission—but they work for fewer than half of patients.

Beginning in 2016, Frankel, then at the University of Texas (UT) Southwestern Medical Center, teamed up with UT Southwestern colleague Andrew Koh and other oncologists to collect stool samples from more than three dozen metastatic melanoma patients who were about to receive one or more checkpoint inhibitor (CPI) immunotherapies, which unleash the immune system against cancer. As they had suspected, the patients who responded to the treatments differed in their microbiomes from nonresponders.4

Frankel and Koh’s study was the first to document how microbiome composition is linked with immunotherapy outcome in humans. That finding joined a growing body of research that points to the influence of resident microbes, particularly those in the gut, on people’s response to medical therapies. Most studies in this area have focused on cancer treatment, but results suggest that drugs for Parkinson’s, high cholesterol, and many other conditions can be similarly affected by a patient’s microbiome. While some scientists caution that a far better understanding of these interactions is needed before they can be harnessed clinically to improve drug responses, trials are already underway to test probiotic formulations or fecal transplants aimed at improving cancer immunotherapy outcomes.

“Will [microbiome manipulations] solve everyone’s problem? No, I don’t think so,” Frankel says. “But a probiotic or prebiotic, whether in a pill form or a powder, is a lot easier for patients than more-aggressive interventions.”

Cancer and the gut microbiome

Early last year, a trifecta of human studies published in Science found new evidence of the connection between cancer drug efficacy and the resident bacteria of the body. In two of the studies, researchers collected fecal samples from patients with melanoma before they received an anti-PD-1 checkpoint inhibitor. These drugs block a T cell receptor called PD-1 that cancer cells can use to fend off attack from the immune system. Both studies identified certain bacterial species that were present in greater numbers in patients who responded positively to the drug, and microbes transferred from those patients into germ-free mice enhanced the immunotherapy’s effects on the rodents’ tumors.

Despite the similarities between the two studies, the characteristics of the microbiome that they found to correlate with treatment response differed. At the University of Chicago, immunotherapy researcher Thomas Gajewski and colleagues observed greater quantities of Bifidobacterium longum, Collinsella aerofaciens, and Enterococcus faecium in patients for whom immunotherapy was successful,5 whereas the research team led by oncology researcher Jennifer Wargo of MD Anderson Cancer Center found that patients with greater species diversity and a profusion of bacteria in the Ruminococcaceae family tended to do better on treatment than those with different microbiome compositions.6 The results also differed from those identified by Frankel and his colleagues at UT Southwestern the year before; they had found that higher numbers of Faecalibacterium prausnitzii, Bacteroides thetaiotaomicron, and Holdemania filiformis were associated with a response to a combination of the anti-PD-1 CPI nivolumab and a different type of CPI called ipilimumab among patients with metastatic melanoma, whereas patients who got better when given the anti-PD-1 agent pembrolizumab had microbiomes enriched with Dorea formicogenerans. And the third 2018 Science study, led by Laurence Zitvogel and Guido Kroemer, both of INSERM, identified yet another strain of gut bacteria, Akkermansia muciniphila, linked with positive outcomes among people with lung, renal, or urothelial carcinoma who underwent anti-PD-1 immunotherapy.7

Microbial Effects on Drug Metabolism

Gut bacteria harbor enzymes and pump out other molecules that can influence how medications are activated or broken down. One example is the Parkinson’s drug levodopa (L-dopa), for which studies have suggested these interactions help explain differences in efficacy among individuals.

See full infographic: WEB | PDF


Researchers found that some gut bacteria produce an enzyme called tyrosine decarboxylase that can convert L-dopa into dopamine as the drug passes through the small intestine, before it can reach the brain. Testing the stool of patients with Parkinson’s, the team discovered that the abundance of the bacterial gene for tyrosine decarboxylase correlated with a need for a higher dose of L-dopa to control their symptoms (Nat Commun, 10:310, 2019). Another team identified a small-molecule inhibitor that appears to block the enzyme’s action in mice (Science, 364:eaau6323, 2019).


After crossing the blood-brain barrier, L-dopa is converted to dopamine by neurons’ own enzymes to treat the symptoms of Parkinson’s disease. Because there is no transporter protein for dopamine, it can’t cross the blood-brain barrier itself, so L-dopa that’s converted to dopamine prematurely in the intestine can’t reach the brain.

To find out whether differences in the analysis techniques could explain this tangle of results, two University of Florida researchers reanalyzed the data from the three Science studies. Their work turned up results similar to those of the original publications, and their paper, published earlier this year, concludes that differences in the analyses are not the explanation.8 The authors of that reanalysis suggest that the different taxa may be performing similar functions that boost CPIs’ efficacy, a hypothesis that would need to be tested by looking at what genes the bacteria are expressing.

Trinchieri, who was not involved in any of the three studies or the reanalysis, has another idea about what might explain the discrepancies. He points to a phenomenon reported in 2018: computational models that used microbiome composition to predict which people from certain districts in China had metabolic disease failed when applied to people from nearby districts. This suggests, Trinchieri says, that people in the same region have broadly similar gut microbiomes, which influence whether particular microbial species are associated with a given health effect, and this may explain why studies conducted in different places yielded diverse signals.

“The microbiome is almost like an infectious consortium of bacteria” that can be exchanged among people in the same region, he explains. “The major determinant of the composition of the microbiome [appears to be] where people live.”

Gut bugs, systemic effects

Gut microbes are known to modulate the immune system, and this interaction may go a long way in explaining the microbiome’s influence on cancer immunotherapy.

Several possibilities have been proposed for how this works, says Koh. For example, antigens from gut bacteria could be similar to those of a tumor, training the immune system to fight the cancer. (See “Do Commensal Microbes Stoke the Fire of Autoimmunity?The Scientist, June 2019.) What Koh considers most likely is that commensal microbes can activate the immune system. “Some bacteria . . . may be more immunogenic than others,” he says. “They kind of are that first step to priming or pumping up the immune system, and then [immune cells] can go do a better job fighting against cancer.”

While some scientists caution that a far better understanding of drug-microbiome interactions is needed, trials are already underway to test probiotic formulations or fecal transplants aimed at improving cancer immunotherapy outcomes.

Indeed, mouse studies have shown that microbiome composition can affect the activity of some types of immune cells, such as memory T cells and myeloid cells. In 2007, Chrystal Paulos, now at the Medical University of South Carolina, Nicholas Restifo of the NCI, and colleagues reported a curious phenomenon: that tumors in mice treated with CD8+ T cells programmed to attack the cancer were more likely to shrink if the animals were first subjected to total-body ir­radiation. The researchers found that the radiation damaged the lining of the gut, freeing commensal microbes, which lodged themselves in other parts of the body and churned out immune-stimulating lipopolysaccharides. If the researchers gave the mice a broad-spectrum antibiotic to deplete their gut microbiomes throughout the treatment, the animals developed fewer activated dendritic cells, which work to enhance T cell function. Giving antibiotics to nonirradiated mice had no effect on the efficacy of the T cell treatment. This suggests, Paulos says, that the microbes that escaped the gut mediated the immune system’s activity.9

In their 2013 study, Trinchieri, Goldszmid, and colleagues homed in on the effects of gut microbes on the tumor microenvironment itself. Among their findings was that hours after treatment with a type of immunotherapy known as CpG-oligodeoxynucleotides—short, synthetic segments of DNA that can help stimulate the immune system to attack cancer cells—antibiotic-treated mice had lower numbers of various types of immune cells that produced tumor necrosis factor, a molecule required for CpG-oligodeoxynucleotides to work. Such findings, along with those of numerous studies identifying crosstalk between gut microbes and various immune cells, point to intimate links between the microbiome and immunity, and by extension, the body’s response to immunotherapy. This suggests that targeting gut bacteria in conjunction with immunotherapy could boost the success rate of the drugs.

Revving up the Immune System

In addition to releasing products that directly act on drugs in the body, resident microbes can affect drugs’ action through the immune system, which is particularly relevant for patients’ responses to immunotherapy. While little is known about the mechanisms by which certain bacterial species seem to boost a patient’s chance of success with these immune-modulating treatments, hints of a causal role have emerged from studies that compared responses in tumor-ridden mice with normal microbiomes to those in rodents whose gut microbes had been depleted by antibiotics.


In one study, gut microbes appeared to prime murine immune cells to secrete tumor necrosis factor (TNF) when the animals were treated with CpG-oligonucleotide immunotherapy—short, synthetic segments of DNA that can help stimulate the immune system to attack cancer cells. TNF in turn induced tumor necrosis (Science, 342:967–70, 2013).


In another study, mice were treated with tumor-specific CD8+ T cells from other mice—a similar procedure to CAR T cell therapies—and gut microbes promoted dendritic cell maturation via the transmembrane protein toll-like receptor 4 (TLR4). The mature dendritic cells in turn activated the CD8+ T cells, inducing them to kill tumor cells (J Clin Invest, 117:2197–204, 2007).

See full infographic: WEB | PDF

The march toward translation

With so many unanswered questions about which microbes affect which treatments and through what mechanisms, Koh thinks it’s premature to move microbiome manipulations into the clinic to augment cancer therapies. “It’s really important to try to figure out why the bugs work. . . . I’m just a little wary about jumping too quickly to therapeutic trials in humans,” he says. “If you get a bad result on one of those trials . . . [it] sets the field back.”

Yet the draw of moving toward commercialization sooner than later is clear. “My patients kept yelling at me that I was taking their stool, and I wasn’t giving them any answers,” says Frankel. After he finally told one patient a bacterial species—F. prausnitzii—appeared to correlate with better outcomes from the therapy the patient was receiving, the patient found a product sold online that is composed of spores from five Bacillus species; Bacillus spores have been found to increase the abundance of F. prausnitzii in humans. Frankel is now planning a clinical trial sponsored by the product’s maker, Microbiome Labs, to test the effects of supplementing CPI therapy with capsules of the probiotic.


Another researcher with an eye toward clinical intervention is MD Anderson’s Wargo. Her findings on how the microbiome affects the efficacy of anti-PD-1 therapy for melanoma in some patients were licensed to Cambridge, Massachusetts–based biotech company Seres Therapeutics, which has designed a consortium of live bacteria to be taken orally before a CPI is given. A Seres-sponsored trial began recruiting patients with melanoma at multiple sites early this year. Meanwhile, Vedanta Biosciences and Bristol-Myers Squibb (BMS) have announced they will team up to conduct a clinical trial of a bacterial concoction devised by Vedanta in conjunction with BMS’s anti-PD-1 therapy nivolumab. And a small trial at the Lawson Health Research Institute in Ontario, Canada, is currently recruiting patients to test the safety of fecal transplants from healthy donors for melanoma patients about to begin nivolumab or pembrolizumab.

Some patients aren’t waiting on the results of such trials. In a recent, not-yet-published study, MD Anderson’s Jennifer McQuade and colleagues asked 113 patients about to undergo immunotherapy about their diets and analyzed bacterial species in their stool. More than 40 percent of the subjects reported they were taking over-the-counter probiotics. McQuade worries this may be counterproductive, however, as people taking probiotics tended to have less-diverse gut microbiomes than those who don’t. On the other hand, those patients whose self-reported diets were rich in fiber were more likely to have high numbers of types of Ruminococcaceae and Faecalibacterium bacteria—both of which were associated with better anti-PD-1 outcomes in the 2018 Science study from MD Anderson.

McQuade, who was not involved in that Science study, is now planning a follow-up experiment that will involve feeding patients undergoing immunotherapy either a conventional healthy diet or the same diet plus an extra 30 grams per day of fiber for 12 weeks, to see whether fiber will change microbiome composition and treatment outcomes.

“It might be that a single-strain probiotic is not the answer, but that a consortia, kind of a mini-environment, might have different effects,” says McQuade. Perhaps one day, she says, “we can . . . provide people with specific bacteria based on their individual microbiome profiles, but we’re not there yet.”

Intratumor Bacteria and Drug Resistance

It turns out it’s not just gut microbes that can influence the success of cancer treatment—bacteria within tumors themselves can drive resistance to anticancer drugs. In a 2017 study, a collaboration led by Ravid Straussman of the Weizmann Institute of Science in Israel stumbled on an intriguing hint of this while studying the influence of connective tissue cells known as stromal cells on drug response. They found that skin fibroblasts from one patient upped resistance to the chemotherapy gemcitabine when cocultured with cancer cell lines (Science, 357:1156–60).

It turned out that a bacterium called Mycoplasma hyorhinis had infected the stromal cells and was converting gemcitabine to an inactive form, protecting the cancer cells. To find out how widespread that phenomenon might be, the researchers tested 27 other strains of bacteria and found that half—many of them in the Gammaproteobacteria class—shared Mycoplasma’s ability to chew up gemcitabine. The drug is commonly used to treat pancreatic ductal adenocarcinoma (PDAC), so the research team collected 113 human PDAC tumor samples and found three-quarters contained bacterial DNA, mostly from Gammaproteobacteria species. When bacteria from fresh PDAC samples were cultured with cancer cell lines, the cells became resistant to gemcitabine.

Straussman says it’s too soon to tell how important bacteria living inside tumor cells are to treatment outcomes. “There’s not a huge amount of bacteria in these tumors,” he says. Moreover, the mechanism his study identified is only one of many ways tumors can resist therapies. “It’s not very clear at this point what will be the effect of inhibiting this specific mechanism,” he says—although he suspects that, as with gut microbes, intratumor bacteria may turn out to have wide-ranging effects on cancer biology. “I do think that there is potential here.”

Drug metabolism and the microbiome

In addition to the microbiome’s ties to cancer immunotherapy, scientists have begun to link our resident microbes with the efficacy of drugs for a wide range of conditions. Rather than being mediated by the immune system, these effects have been traced to the actions of bacterial products on the drugs themselves. In 2009, for example, researchers at Imperial College London correlated urine levels of a bacterial product, p-cresol sulfate, with the rate at which people metabolize the painkiller acetaminophen.10 The scientists theorized that this is because p-cresol effectively competes with acetaminophen in a reaction where either molecule accepts the addition of a sulfonate group. While the effect of sulfonation on acetaminophen’s activity was not clear, the authors suggested that other medications that are sulfonated after ingestion might be similarly affected.

In other cases, researchers have gathered evidence that certain bacterial species may enhance medications’ efficacy. For example, in 2011, Duke University metabolomics researcher Rima Kaddurah-Daouk and colleagues reported that patients’ responsiveness to a cholesterol-lowering statin drug correlated with blood levels of three bile acids produced by gut bacteria.11 Bile acids and the statin share transporter proteins in the liver and intestine, note the study authors, who suggest that such competition could help explain the association. Similarly, a 2014 study at South Korea’s Kyung Hee University looked at mice treated with the cholesterol-lowering drug lovastatin. Animals that were also treated with antibiotics had lower levels of the statin’s active β-hydroxy acid metabolite, they found.12 Kaddurah-Daouk thinks gut bacteria will turn out to have an effect on responses to other drug classes, too, via their metabolites.


In recent years, metabolomics research has helped bring the microbiome’s enormous influence on human physiology into focus, Kaddurah-Daouk says. For example, metabolic processes in our resident bacteria produce more circulating molecules than our own cells. “It was daunting to realize that the majority of the chemicals in blood . . . are products of the bacterial metabolism,” she says. “You and I are not all that different genetically . . . the biggest differences between us [are] probably a result of the diet and the gut bacteria.”

Another such metabolic relationship emerged earlier this year, when Sahar El Aidy of the University of Groningen in the Netherlands and colleagues reported that patients with Parkinson’s disease who have higher levels of a bacteria-produced enzyme called tyrosine decarboxylase require higher doses of the drug levodopa to control their symptoms.13 El Aidy explains that her group, which focuses on the relationship of gut bacteria to neurotransmitters, was studying how tyrosine decarboxylase converts one amino acid to another, and realized the chemical change it catalyzes is the same one that converts levodopa to dopamine.

“[W]e sought to test whether the same bacterial enzyme can also convert levodopa to dopamine and it did,” she writes in an email to The Scientist. Dopamine can’t cross the blood-brain barrier, whereas levodopa can, so El Aidy thinks patients with many tyrosine decarboxylase-producing bugs must take more of the drug for sufficient amounts to make it to the central nervous system. There, it provides dopamine that the brain can no longer produce itself, easing motor symptoms of Parkinson’s. Last month, another group reported that they’d identified a compound that inhibits tyrosine decarboxylase’s activity when given to mice—suggesting, the researchers write, that combination therapies could be developed that both deliver levodopa and prevent it from being activated prematurely.14

Other researchers have found microbiome effects on the metabolism of drugs ranging from the anti-epileptic zonisamide to insulin and the hormone calcitonin, which is used to treat high calcium in the blood and some diseases of the bone. Indeed, researchers recently reported that two-thirds of the 276 different drugs in which they incubated 76 human gut bacterial species were in some way modified by the microbes.15

Given the microbiome’s wide-ranging effects, greater knowledge of the interplay between resident microbes and drugs has the potential to transform the practice of medicine, and will be combined with other types of information, such as genetics, to make treatments more individualized and effective, says Kaddurah-Daouk. “The future looks very different from where we are today.”


  1. N. Iida, “Commensal bacteria control cancer response to therapy by modulating the tumor microenvironment,” Science, 342:967–70, 2013.
  2. A. Sivan et al., “Commensal Bifidobacterium promotes antitumor immunity and facilitates anti–PD-L1 efficacy,” Science, 350:1084–89, 2015.
  3. M. Vétizou et al., “Anticancer immunotherapy by CTLA-4 blockade relies on the gut microbiota,” Science, 350:1079-84, 2015.
  4. A.E. Frankel et al., “Metagenomic shotgun sequencing and unbiased metabolomic profiling identify specific human gut microbiota and metabolites associated with immune checkpoint therapy efficacy in melanoma patients,” Neoplasia, 19:848–55, 2017.
  5. V. Matson et al., “The commensal microbiome is associated with anti–PD-1 efficacy in metastatic melanoma patients,” Science, 359:104–8, 2018.
  6. V. Gopalakrishnan et al., “Gut microbiome modulates response to anti–PD-1 immunotherapy in melanoma patients,” Science, 359:97–103, 2018.
  7. B. Routy et al., “Gut microbiome influences efficacy of PD-1–based immunotherapy against epithelial tumors,” Science, 359:91–97, 2018.
  8. R.Z. Gharaibeh, C. Jobin, “Microbiota and cancer immunotherapy: in search of microbial signals,” Gut, 68:385–88, 2019.
  9. C.M. Paulos et al., “Microbial translocation augments the function of adoptively transferred self/tumor-specific CD8+ T cells via TLR4 signaling,” J Clin Invest, 117:2197–204, 2007.
  10. T.A. Clayton et al., “Pharmacometabonomic identification of a significant host-microbiome metabolic interaction affecting human drug metabolism,” PNAS, 106:14728–33, 2009.
  11. R. Kaddurah-Daouk et al., “Enteric microbiome metabolites correlate with response to simvastatin treatment,” PLOS ONE, 6:e25482, 2011.
  12. D.-H. Yoo et al., “Gut microbiota-mediated drug interactions between lovastatin and antibiotics,” Drug Metab Dispos, 42:1508–13, 2014.
  13. S.P. van Kessel et al., “Gut bacterial tyrosine decarboxylases restrict levels of levodopa in the treatment of Parkinson’s disease,”  Nat Commun, 10:310, 2019.
  14. V.M. Rekdal et al., “Discovery and inhibition of an interspecies gut bacterial pathway for Levodopa metabolism,” Science, 364:eaau6323, 2019.
  15. M. Zimmermann et al., “Mapping human microbiome drug metabolism by gut bacteria and their genes,” Nature, doi:10.1038/s41586-019-1291-3, 2019.

Interested in reading more?

On Target July Issue The Scientist

Become a Member of

Receive full access to digital editions of The Scientist, as well as TS Digest, feature stories, more than 35 years of archives, and much more!