NEW YORK – In a recently launched observational study, San Diego-based biotech startup Persephone Biosciences is hoping to explore how cancer patients' gut microbiome composition influences their immune system and their ability to respond to treatments.
In the ARGONAUT study, Persephone hopes to enroll roughly 4,000 patients with various solid cancers, making it possibly the largest prospective gut microbiome trial of its kind. The aim within the trial is to not only use microbiome composition to predict cancer patients' outcomes to treatment, but ultimately, to develop microbiome-modulating therapies that can improve patients' responses to immunotherapy drugs.
The study is rooted in the notion that the gut, and the hundreds of species of microbes within it, play a meaningful role in modulating cancer patients' immune system response to cancer. While past studies have interrogated the link between gut microbiome composition, immune function, and response to immunotherapy drugs, most studies have been small or preclinical, and have yielded incomplete findings.
"Up to this point, [what we know] is mostly correlative," Persephone CEO Stephanie Culler said in an interview. "We know, for example, that the microbiome in certain patients correlates to a disease or response to a certain treatment, but no one has really unlocked the mechanism behind how the microbiome in a person impacts such responses."
Persephone originated in 2017 as part of a Johnson & Johnson incubator, called JLABS. The company proceeded to raise seed funding from Y Combinator, and in the years since its launch, Persephone's scientists have developed an artificial intelligence-based platform that analyzes DNA, RNA, metabolite, and protein data from multi-omics analyses of stool and blood samples, and identifies biomarkers that may predict responses to cancer treatment.
This machine-learning platform, which has been trained on stool and blood samples from healthy individuals and cancer patients, will help Persephone unlock key findings in the ARGONAUT study. Specifically, the platform will be used to parse multi-omics data gleaned from the 4,000 patients enrolled in the trial with non-small cell lung, triple-negative breast, colorectal, or pancreatic cancer.
These patients may simultaneously be enrolled in a therapeutic drug study running separately from ARGONAUT, and over the course of six months, investigators will collect two blood samples and two stool samples from each patient before and after they begin treatment. In the NSCLC cohort, patients must be on an immune checkpoint inhibitor that they have not previously received, and across the other three cohorts patients can receive an immune checkpoint inhibitor or any other type of treatment that they have not received before including chemotherapy.
According to Sandip Patel of the University of California, San Diego, one of the site investigators for the ARGONAUT study and a clinical advisor to Persephone, the choice to enroll patients across these four specific tumor types was based on preclinical data and insights from earlier pilot studies.
"A lot of the data we've seen with the microbiome relates to how it modulates either immune response in terms of bacterial species that could foster an environment that lets the body best attack cancer or how the specific bacterial species in the microbiome can actually digest and break down the chemotherapy before it's able to get into the tumor," he said.
Patients enrolled in ARGONAUT will be monitored for up to two years, allowing the researchers to incorporate longer-term outcomes into their investigation of the microbiome's role. The primary endpoint of the study is to deduce whether microbiome composition can predict patients' progression-free survival, and secondarily, researchers want to determine the same in terms of overall survival and identify correlations between the microbiome composition and immune markers found in the blood. A third study aim is to build a library of patient samples for further research.
Metagenomic WGS, multiomics analysis
To fully characterize patients' microbiome composition and how it correlates to immune markers, ARGONAUT's investigators will perform what Culler called a "suite of omics analyses" on the collected samples. Metagenomic whole-genome sequencing will be conducted on the stool samples, allowing researchers to characterize patients' microbiome compositions beyond the species level, interrogating the individual bacteria's strains and substrates.
Patel offered a classic bacteria example, Escherichia coli, to emphasize the importance of this deeper, sub-species assessment using metagenomic WGS. "All of us have E. coli in our guts. It's actually the most common gut bacteria," he said. "But it's certain strains of E. coli that make you sick [and] calling them all E. Coli is only going to give you the street level view, so to speak. If you really want to nail down the specific address and the unique features of a bacteria, you need to go in depth."
Beyond the metagenomic WGS, the researchers will drill down to the functions of these species using untargeted metabolomics analyses to detect over a thousand metabolites. Using computational models to analyze this data, the researchers will map the pathways that are producing these metabolites and identify the microbes with the most important functions, Culler explained. From there, researchers will go on to analyze these microbes on the single-cell level.
On the collected blood samples, the researchers will perform deep immune profiling using CyTOF single-cell mass cytometry techniques, and will look at the chemokines and cytokines, as well as the metabolites, ultimately allowing them to link the microbiome data from the stool samples to the immune data from the blood samples. The combined analyses will be processed by the machine-learning platform, which Culler explained will be helpful for identifying those microbes or signatures of microbes that could be used as predictive biomarkers.
Companion diagnostics, microbiome therapeutics
Culler is ultimately hoping that the large multi-omics datasets collected within ARGONAUT will facilitate the development of companion diagnostics and microbiome therapeutics.
For example, if the data derived from patients' stool and blood samples are predictive of their responses to specific immunotherapy drugs, then Culler said that could lead to the development of a non-invasive companion diagnostic for informing which patients are likely to benefit from these treatments. "The vision is that, prior to going on treatment, a patient could provide a stool sample for analysis, which would help predict which treatment they should take," she said.
Persephone will likely have to partner with drug sponsors to advance such companion diagnostics, and these tests will also have to be affordable and scalable.
In an October presentation during the European Society for Medical Oncology's virtual meeting, Jennifer Wargo, who leads MD Anderson Cancer Center's Program for Innovative Microbiome and Translational Research, shared research showing that "the gut microbiome could be used as a biomarker for immune checkpoint blockade," but also acknowledged that the ideal assay is not yet known. "The assay we use needs to be rapid, accurate, and cost-effective," she said in her presentation.
Based on these same considerations, Patel explained that the suite of in-depth omics analyses performed on stool samples in ARGONAUT would not be the methods used in a broadly marketed companion diagnostic. A more cost-effective and scalable option, he suggested, would be a companion test that relies on 16s rRNA sequencing, a well-adopted laboratory method for identifying bacterial strains.
The second application for the data, which both Culler and Patel highlighted as an ultimate goal for the young company, is to develop therapeutics that can actually modulate patients' microbiome composition such that it enhances the efficacy of their treatment.
The microbiome therapy, which Culler said would likely be taken alongside a given treatment, such as a checkpoint inhibitor, would consist of engineered microbes designed to have the same set of key functions that the researchers identified as beneficial for treatment outcomes. Persephone has already presented data from a mouse-model study to the US Food and Drug Administration and demonstrated as a proof of concept that a consortia of bacteria improved immunotherapy responses in mice.
"While that worked in a mouse, we really believe that we can boil [the therapy] down to a key set of functions … and with a few metabolic functions in-hand [that are missing from some patients], we can engineer that into a single microbe and have that be our therapy," Culler said. The researchers have not yet determined whether this microbiome therapy would be given to all patients treated with a given agent, or just to those whose microbiomes lack key functions.
"This is actually something we're trying to understand from the study," Patel said. "Even if someone already has a 'good' microbiome signature, maybe we can make it even better. These are questions that we will have to learn as we go."
Persephone's goal is to design an oral therapy. This would depart from most microbiome-modulating treatments developed in the past involving fecal transplants, which are a one-time procedure and tend to carry risks.
"This would likely be an orally administered encapsulation of the good bacteria within a capsule that is able to get past the stomach," Patel said. "And that capsule would protect those bacteria as they get into the distal bowel, so that they can colonize the bowel and create the virtuous immune microenvironment that we're going for."
Equitable research, diversity benchmarks
The human microbiome differs significantly depending on patients' ethnicity, sex, and geographic location. As such, it is critical that microbiome-based clinical trials enroll a diverse population if the resulting findings that lead to drugs and diagnostics are to be useful for a broad population. For this reason, and because Persephone's mission is "equitable precision medicine," the ARGONAUT trial has gone a step beyond many clinical trials — including those sponsored by pharma giants with far greater resources —and set diversity benchmarks for patient enrollment.
While it is not an official requirement for the study to proceed, the researchers have committed to making every effort to ensure that the ARGONAUT cohort will be evenly divided between male and female participants, and that 20 percent of them will be African American, 10 percent will be Hispanic, and 10 percent will be Asian.
"What we've recognized is that, in cancer datasets, there are tremendous disparities," Culler said. "And [the studies] really don't reflect those populations with the highest cancer risks. Even in our own initial proof-of-concept dataset, we noticed it was heavily biased toward individuals of European descent. They don't accurately reflect the patient population that we want to treat."
To accomplish the goal of profiling a diverse population in the ARGONAUT study, Persephone has already reached out to minority oncologists and patient advocacy groups. "We've been going about it in a very grassroots way," she said. "We're obviously a very small organization, but the power of having these conversations to build trust is exceptionally important."