A few years ago, biologist Ze’ev Ronai discovered that the mouse strain he’d been working with for the last decade was particularly resistant to melanoma. He’d been studying the mice—which lack a functional copy of the protein-regulating enzyme RNF5—with colleagues at Sanford Burnham Prebys Medical Discovery Institute in La Jolla, CA, to learn more about the cellular stress response pathways of which RNF5 is a part. When, during research on those stress pathways’ involvement in cancer progression, the scientists inoculated the RNF5 knockouts with melanoma cells, they were surprised to find that tumors grew more slowly, and the mice showed heightened antitumor immune responses compared with wild-type animals.

Excited by the implications for immunotherapy research, Ronai’s group planned to demonstrate that the knockout mice’s immune cells, which are produced by bone marrow, were sufficient to elicit the same cancer-resistant phenotype when transferred into wild-type animals. But the experiment didn’t work—a bone marrow transfer didn’t slow tumor growth. “To us,” Ronai says, “this indicated that there was something else beyond the immune system component that elicited the phenotype.”

Scouring the melanoma literature for inspiration, the team came across a 2015 study by the University of Chicago’s Thomas Gajewski and colleagues. That research showed that mice with melanoma responded better to immunotherapy treatment—specifically, to anti-PD-L1 immune checkpoint therapy, which works by enhancing antitumor responses—if their guts contained high concentrations of a bacterial genus called Bifidobacterium. It piqued Ronai’s interest, and “led us to examine whether the differences we saw here could be linked to the microbiota,” he says.

Sure enough, administering antibiotics to Ronai’s RNF5-knockout mice wiped out their special antitumor response. The same thing happened when the researchers housed the knockout and wild-type mice together: the knockout mice, which ended up exchanging gut microbiota with their wild-type cage mates within a few weeks, no longer showed an enhanced resistance to melanoma. “Clearly, the microbiota change led to loss of antitumor immunity, and consequently tumor growth inhibition,” Ronai says. “From there, the road was fairly straightforward.”

Expanding the team to include microbiologists, computer scientists, and immunologists, Ronai and his colleagues identified 11 microbial strains that were enriched in the guts of RNF5-knockout mice and whose abundances negatively correlated with tumor size. Feeding those 11 strains to antibiotic-treated wild-type mice boosted the recipient animals’ immune responses to tumor cells and restricted tumor growth. The group also found that mice, and indeed human patients in three independently studied cohorts totaling nearly 100 people, were more responsive to anti-PD-1 or other immune checkpoint treatment if their tumors expressed lower levels of proteins that, like RNF5, were involved in cellular stress responses.

In a paper published earlier this year, the researchers argue that the findings suggest that crosstalk between cellular stress responses and the gut microbiome could help shape antitumor immunity. Although the mechanisms involved are unclear, Ronai notes that particular metabolites or antigens presented by members of that microbiome could somehow help raise the immune system’s sensitivity to foreign-looking cells, such as those that compose tumors.

“It’s a very interesting paper,” says Juan Cubillos-Ruiz, a cancer immunologist at Weill Cornell Medical College, who was not involved in the work. For one thing, the results suggest that the expression of proteins involved in these cellular stress responses “could be used as a biomarker of whether a patient will or will not respond to immunotherapy in the future,” he notes. “It’s the first study demonstrating that that’s the case.” However, the finding that microbiome composition is altered in the mouse model is “a little bit complicated to interpret,” he says—not least because the RNF5 protein was knocked out in all of the mouse model’s tissues, so “you don’t know specifically in which cell types the protein is playing a role.”

The results come at an interesting time for research into the gut microbiome’s role in antitumor immunity. Shortly before Ronai’s group submitted its paper last year, Science published a trio of studies—one by Gajewski’s group, one from Jennifer Wargo and colleagues at MD Anderson Cancer Center, and one from Laurence Zitvogel and colleagues at INSERM in France—that all found a connection between the success of immunotherapy and the composition of a human patient’s gut microbiome. Each paper also highlighted a handful of bacterial strains or phyla that appeared to be linked to better responses, although the specific taxa varied between studies.

Attempts to apply these findings to improve immunotherapy outcomes are already underway. A team led by Hassane Zarour at the University of Pittsburgh, for example, is recruiting patients with stage III or IV melanoma for a trial of immunotherapy drug pembrolizumab combined with fecal matter transplants from other patients who have already responded well to the drug. “It’s an ambitious trial,” says Zarour, who is collaborating with Merck and Massachusetts-based Evelo Biosciences for the project. “Of four patients so far, we have one stable partial response,” he notes, adding that more will be treated over the coming months.

Meanwhile, MD Anderson researchers have teamed up with Massachusetts-based Seres Therapeutics to run a separate clinical trial, also with late-stage melanoma patients, that will combine anti-PD-1 immunotherapy with an oral dose of certain bacterial strains and/or a fecal matter transplant from melanoma patients who responded well to treatment. “The premise behind this is that you can actually change the microbiome and make patients respond better,” says Wargo.

Nevertheless, several aspects of the microbiome-immunotherapy link remain unclear. For starters, “there’s not complete overlap of the different taxa that are response-associated between the published cohorts, which is vexing and confusing to some,” Wargo says. “We still don’t know what is the ideal, ‘rockstar’ microbiome.”

And with mechanism-focused studies such as the one by Ronai’s group still relatively few and far between, researchers don’t understand exactly how the microbiome might be boosting antitumor immunity, Zarour tells The Scientist. A start would be to pay more attention to function than to taxonomy when studying the links between microbiome composition and cancer, he adds. Instead of focusing on a list of potentially beneficial commensal species in mice or humans, researchers should perhaps be asking, “What pathways are activated?”

—Catherine Offord