The past two decades have seen intensive research into the human microbiome and its contributions to human health and disease, but there have been relatively few victories in terms of parlaying those findings into therapies. As of mid-2024, only two microbiome-based therapies have gained US Food & Drug Administration approval, both targeting recurrent bacterial infections of the gut.

Empress founding team (left to right): John Mendlein, executive partner, Flagship Pioneering; Sabrina Yang, chief innovation officer and co-founder, Empress Therapeutics and principal, Flagship Pioneering; Doug Cole, general partner, Flagship Pioneering; Jason Park, CEO and co-founder, Empress Therapeutics and operating partner, Flagship Pioneering. Credit: Empress Therapeutics

But there is still good reason to believe that buried treasure is waiting within the microbial communities that dwell within the human body. “I’m an immunologist by training, and they’re fundamentally foreign — so why would we allow all these cells to inhabit our guts, our skins, in our mouth?” asks Murray McKinnon, CSO at Empress Therapeutics. The answer, he says, is because of our powerful symbiotic relationship with these communities, which have been strongly shaped by a process of ‘co-evolution’ with their human hosts. “They generate chemistry that has a beneficial impact on our immune system, that has a beneficial impact on our metabolism.”

Empress is using this underlying principle of co-evolution to guide the discovery of drugs based on biologically active small molecules that are naturally produced by human resident microbiota. Co-founded by Jason Park, Sabrina Yang and colleagues at Flagship Pioneering in 2020, the company exited stealth in June of 2023 with an initial investment of $50 million in funding from Flagship.

Park, who is now the company’s CEO, cites his past experience in the small-molecule drug development world as part of his motivation for launching Empress. He notes that this category of agents has remained a bastion of the pharmaceutical world because of their long track record of versatility, flexible formulation and capacity for readily making their way into the interiors of cells. “The major issue that we saw was that of uncertainty,” says Park, noting that the small-molecule drug development process remains highly vulnerable to failures from unexpected toxicity, poor bioavailability and other factors.

Turning to the microbiome could take some of that uncertainty out of the equation. “We have a real chance to rethink how we do small-molecule drugs,” says Park. The evolutionary forces that have shaped humanity’s symbiotic relationship with resident microbes would be expected to favor production of compounds that are nontoxic, stable and biologically active within the human body. And judging from microbiome research to date, these compounds have the innate potential to directly modulate physiological processes associated with conditions ranging from autoimmunity to cancer to neurodegenerative disease.

The Chemilogics discovery platform developed by Empress begins with an in-depth metagenomic survey of the microbial populations from many patients with a particular condition as well as cohorts of healthy controls. These sequencing data are then fed into a machine learning-powered algorithm, which attempts to differentiate these two groups on the basis of specific ‘biosynthetic gene clusters’ (BGCs) — tightly-linked assembly lines of genes that work together serially to manufacture a particular biomolecular product. “It’s not just about the chemistries, but the causal association between those chemistries and a clinical phenotype,” says McKinnon. This analysis is facilitated by the inclusion of patient transcriptomic data, which can reveal disease-related shifts in gene expression that correlate with the presence or absence of a candidate BGC of interest. Ideally, this will also uncover target genes or pathways that are modulated by the product of that BGC.

These newly identified BGCs are then evaluated by expressing them in genetically reprogrammed model bacteria species like Escherichia coli. This gives the Empress team the opportunity to thoroughly characterize the chemical outputs of these clusters. Potential drug candidates identified at this stage are then subject to various biological assays to assess their therapeutic efficacy and characterize their mechanism of action.

Princeton University biochemist Mohammad Seyedsayamdost, whose lab focuses on the discovery of useful microbially derived compounds, is happy to see a startup looking into such naturally derived products as a starting point for drug discovery. “I think it’s a renewed field in some ways, because we’ve just realized how vast the number of biosynthetic gene clusters is and how little we know about them,” he says. Identifying BGCs from the metagenomic data should be a relatively straightforward endeavor, according to Seyedsayamdost, but he cautions that experimentally characterizing such clusters can be much more challenging. “You can’t always stick the genes in E. coli and expect it to make these compounds …. In many cases, I think you need to go back to the original host,” he says. This could create considerable complexity, as many microbial species remain difficult to culture and manipulate.

MacKinnon acknowledges that “the biology is the rate-limiting factor” and that the analysis and development process gets considerably more difficult once the initial wave of algorithmic analysis is complete. But the company benefits from a scientific advisory board (SAB) that includes domain experts in microbiome research and metagenomics, synthetic biology and metabolomics, who have been engaged with the development and refinement of the Chemilogics platform since the company’s inception. “Every one of our SAB members — all six of them — have been working with us for over five years,” says Park.

Empress is still at a preclinical stage, but the team is exploring a range of indications including immunological and inflammatory disorders as well as oncology and metabolic conditions. Park says that their first few years of work with the platform have been fruitful, yielding roughly 15 promising drug leads with in vivo activity against multiple different categories of targets. He hopes to move their first program into clinical testing in the next year or two.

Importantly, this initial round of leads also seems to be living up to the expectation that microbiome-sourced compounds might bypass many of the adverse effects that can come into play with fully synthetic drugs. “We’ve undergone some toxicology studies for lead molecules and been pleasantly surprised to see these molecules are well tolerated,” says McKinnon, adding that they’ve generally had to do far less chemical optimization and refinement than would be typical in a small-molecule drug discovery campaign.

Empress’s approach also represents a substantial departure from most drug discovery work in the microbiome field, which has largely focused on treating patients using defined consortia of therapeutically beneficial microbes. McKinnon notes that there are numerous challenges associated with that therapeutic strategy, including the difficulties of manufacturing — again, many potentially clinically useful microbes are challenging to cultivate — as well as the regulatory issues involved with developing an optimized live microorganism-based therapeutic. In contrast, Empress is focused on exploring the inner workings of these microbial communities to find better starting points for launching otherwise conventional small-molecule drug programs.

That, however, doesn’t mean the company’s focus ends at the boundaries of the microbiome. And Park is optimistic that, as the Empress team gains more insights into microbial biosynthetic diversity, they will be able to start exploring new frontiers in terms of engineering novel genetically encoded chemical synthesis pathways. “It doesn’t have to be about commensal microbes,” he says. “Anywhere there’s biosynthetic DNA, we could apply the same concepts.”