NEW YORK – Fecal microbiota transplants (FMT) could convert a molecularly defined subset of patients with advanced melanoma from immunotherapy non-responders to responders, according to a paper published last week in Science.
Prior preclinical studies — mostly in mice — have suggested that altering the microbiome composition might convert immunotherapy non-responders into responders. Now, researchers from the University of Pittsburgh Medical Center and the National Institutes of Health have conducted what they say is the first clinical trial to show the same in humans. Specifically, the researchers conducted a Phase II clinical trial in which they administered FMTs to 15 patients with advanced melanoma who had not responded to treatment with pembrolizumab (Merck's Keytruda). The transplants, on the other hand, were derived from seven patients who had responded quite well to pembrolizumab, achieving a median progression-free survival of 56 months.
In the trial, the 15 patients received FMTs with additional pembrolizumab, followed by pembrolizumab every three weeks until disease progression or intolerable toxicity. Six patients, or 40 percent, either responded to the treatment or experienced durable stable disease. Though a minority of the cohort responded, this was still a marked improvement over the fewer than 10 percent of patients who, after failing to respond initially to anti-PD-1 therapy, benefit with additional treatment. Additionally, the FMT plus pembrolizumab retreatment was well-tolerated, with adverse events remaining mostly low-grade.
The proof-of-principle study, while small in terms of patient numbers, could have significant implications, according to one of the authors, both in terms of demonstrating the feasibility of the approach and, importantly, determining which patients refractory to checkpoint inhibitors could be candidates for microbiome modulation.
"We do not think that fecal microbiota transplant, or even microbiome-based therapies more generally, will be the key for all patients who do not respond to anti-PD-1 [drugs]," UPMC's Hassane Zarour, one of the authors of the Science study, said in an interview. "It's a subset of patients for whom the microbiome could play a key role. … It's people who we know should respond but don't respond because they don't have the right microbiome."
Defining these two elements — what sets apart patients who "should respond" to checkpoint inhibitors and what defines the "right" microbiome — could ultimately enable patient selection for FMT and immunotherapy, Zarour said. "The goal is to be able to predict ahead of time if a patient may be a candidate," he explained.
According to Zarour, the initial way that the researchers wanted to go about this study was to select stool donors based on the presence of a favorable microbiome gene signature detected in their stool. However, while there are certain species of microbiota associated with improved immune system function, there is no consensus in the published literature regarding one overarching "favorable" signature.
"The reason why we could not be more granular and select donors based on the gene signature is that… there is a major discrepancy between [previous] studies," Zarour said, explaining why donors were selected on the basis of their immunotherapy responses alone. Going forward, Zarour said that he and his colleagues are working on analyzing greater numbers of stool samples collected from pembrolizumab responders to better define the microbiome gene signature of the ideal donor stool sample.
For now, he noted that metagenomic analysis of the seven donors' samples showed their composition was fairly similar, with a high abundance of specific species such as phyla Firmicutes, including Lachnospiraceae and Ruminococcaceae families, that have been shown in prior studies to correlate with immunotherapy response.
With the goal of someday selecting recipients for FMT, on the other hand, Zarour and colleagues had to perform analyses on far more data sets than just the taxa of the microbiota. "The correlation of microbes and outcome is a complex question, because there are so many confounding variables," he said, noting there are many more factors at play when a patient's cancer resists checkpoint inhibition than the specific species of microbiota in their guts.
Among these factors are the unique makeup of patients' immune systems. To better home in on how these elements might play a role in treatment response, the researchers performed multiparameter flow cytometry and single-cell RNA sequencing analyses of peripheral blood mononuclear cells and tumor samples. These analyses were performed both before and after the FMT-pembrolizumab treatment.
In the peripheral blood, the researchers found that patients who ended up responding to pembrolizumab after the transplants had lower percentages of naïve CD8-positive T cells and higher percentages of differentiated T cells. In the tumor samples, for which the researchers used 10X Genomics' Chromium single-cell RNA sequencing analysis, they observed a higher frequency of myeloid cells and CD4-positive regulatory T cells among the CD45-positive cells in non-responders compared with responders. These findings suggested a myeloid gene signature associated with tumor progression in non-responders.
Additionally, the researchers performed multiomics analysis of patients' serum samples, including serum metabolomics and lipidomics. Among the differences noted between non-responders and responders based on these analyses was that the responders down-regulated circulating cytokines and chemokines associated with resistance to PD-1 inhibitors, including IL-8, an immunosuppressive cytokine.
The circulating IL-8 was particularly interesting to Zarour. "IL-8, in the case of microbiome-based therapy, may be a major driver of resistance," he said.
Finally, to confirm that the microbiome was actually playing a role in modulating these signatures, and in turn, affecting patients' treatment response, the researchers used a bioinformatics approach called transkingdom analysis.
"It was a small cohort of patients, but one [for which] we had tons of data," Zarour said, listing microbiome, flow cytometry, metabolite, and lipidomics among the various data types generated on patients. "We used [transkingdom analysis] to connect all of the data together to see how the data talk to each other … and say how likely it is that one set of data is responsible for the other sets of data."
Through this bioinformatics approach, the microbiome emerged as a major driver of all of the other variables. "Putting it all together, this bioinformatics analysis seems to suggest that … the good microbes changes operate with good clinical outcome, while the bad microbes were linked to biologic changes that were linked to non-response," Zarour said.
The depth of analysis conducted in this study to parse the molecular and biological differences between responders and non-responders to FMT-pembrolizumab treatment was possible in part due to the small number of patients enrolled. That said, Zarour explained that the findings could ultimately help shed light on a straightforward, scalable signature to sort potential responders versus non-responders.
"FMT shifted microbiome composition toward taxa favoring anti-PD-1 efficacy to induce clinical responses to anti-PD-1 in PD-1 refractory melanoma patients who had an immunological ability to respond to the treatment but exhibited an unfavorable microbiota composition," concluded Zarour and colleagues in their Science paper.
That "immunological ability to respond to the treatment" together with the unfavorable microbiome composition will be the key to zeroing in on the group of patients likely to benefit from this treatment approach.
Going forward, Zarour and colleagues plan on conducting studies in larger patient populations to better define this signature. Such analyses are now possible since increasingly researchers are routinely collecting stool samples from patients treated with immunotherapy, the same way that cancer centers might collect blood or tumor samples.
"Major centers are doing that now, but it's not something that has been routinely done for many years," Zarour said of stool sample collection.
That said, a number of biotech companies have entered the scene in recent years with the goal of developing and commercializing microbiome-modulating therapies, and several of these companies, including San Diego-based biotech startup Persephone, are ramping up stool sample collection to accomplish this. Another startup, Bio-Me, is working on developing a qPCR-based microbiome profiling tool to help physicians predict immunotherapy response.
"A number of small companies are very active and already have launched [microbiome-based] trials," Zarour said.
Importantly, both academic researchers like Zarour's team and companies like Persephone acknowledge that the end goal, when it comes to microbiome modulation, will not be fecal transplants but rather an oral medication that can be taken like a probiotic. Such an intervention will only be a reality once the signature for identifying who can benefit from receiving such a pill, and the ideal consortium of microbes to include within it, become crystal clear. In the meantime, FMT and immunotherapy-based clinical trials, which Zarour and colleagues have demonstrated to be feasible, are a step in this direction.