Culturing microbiome therapeutics with big data

The human microbiome was first sequenced in the early 2010s, giving rise to the first ideas that therapies targeting the microbiome could be developed. However, determining the direct role of the human microbiome in health and disease was difficult due to small data set sizes, confounding environmental factors, and technical limitations. Early clinical trials failed, dampening initial enthusiasm, and it was clear that more detail was needed on the biological links between microbes and disease. In recent years, these links have been enriched with huge data sets and important biological information. A recent study used large-scale metagenomic profiling of gut microbiomes and matched host health data from approximately 34,000 participants in the US and UK to define and develop a ranking of bacterial species associated with markers of human health, establishing links between diet, gut microbiota and health outcomes.¹. Another study integrated 168,464 publicly available 16S ribosomal RNA gene amplicon sequencing samples from 68 countries to create the Human Microbiome Compendium.². Advances in multi-omics approaches and computational pipelines have defined the healthy human microbiome, highlighted the correlation between gut, vaginal, skin, and oral microbiomes and human health, and identified predictive, diagnostic, and prognostic markers for various diseases.^3.4.

However, increasingly large human datasets and multi-omics analyzes have shown substantial inter-individual variability in microbiome composition, which slows the clinical translation of disease markers. Additionally, the complexity of microbial interactions makes it difficult to predict outcomes, particularly for clinical trials. Microbes have been shown to possess unique and evolving antimicrobial resistance mechanisms when confronted with antibiotics. Additionally, antibiotic use leads to dysbiosis of the gut, skin, and oral microbiome – a significant disruption of the balance between the microbiota and loss of beneficial microbes.^5,6,7. Providing evidence that microbiome modification is associated with a positive and predicted outcome and learning how antimicrobial resistance arises requires a high-level understanding of biological mechanisms.

In late 2022 and early 2023, the U.S. Food and Drug Administration approved the first microbiome-based treatments, both intended to treat recurrences. Clostridioides difficile infections⁸. C. difficile the infection mainly occurs after intensive and prolonged use of antibiotics and leads to gastrointestinal disorders. However, direct causal evidence for the activity of these drugs was, at that time, unknown, limiting their use more widely. Recently, experiments showed that one of the drugs caused an increase in the number of Firmicutes species, which produced metabolites to directly inhibit the growth of C. difficile and other opportunistic pathogens^9.10.

Although these early approvals did not result in much follow-up, researchers are now providing evidence for the direct functional effect of metabolic output from microbial communities, and insights into host-microbe interactions are poised to lead to breakthroughs. There is now evidence linking microbiome dysbiosis to disease onset and/or severity, and companies are using this data to develop treatments to treat and prevent disease.

For example, microbiome dysbiosis has been associated with atopic diseases (atopic dermatitis, food allergy, and allergic asthma). These conditions ultimately lead to immune system dysfunction that triggers IgE antibody production and inflammatory responses. Given the recent wealth of data on the gut microbiome of infants, Siolta Therapeutics is combining this data with bioinformatics tools to isolate specific symbiotic organisms from healthy infants that can be manufactured as drugs to treat dysbiosis. Their most promising candidate prevents atopic disease in infants as a preventative and durable oral live biotherapeutic. Even a year after taking the drug, it reduced the risk of atopic dermatitis by 64% and the risk of food allergy by 77% in a child population. It remains to be seen whether this durability is due to long-lasting engraftment of beneficial microbes or adaptive immune modulation.

Metabolites such as short-chain fatty acids (SCFAs) and bile acids produced by the microbiota can enter the circulation and affect human health¹¹. Community-scale metabolic models have now been used to predict individual-specific SCFA production profiles and can help design combinatorial prebiotic, probiotic, and dietary interventions in silico.¹². However, it remains to be demonstrated whether directly administering these metabolites to treat diseases is as effective as enriching the microbes that produce them, particularly in the complex ecosystem of the gut.

Microbes producing SCFA butyrate have been shown to promote a healthy intestinal lining and inhibit inflammatory immune cells in the gut.¹³. Maat Pharma combined a proprietary cryoprotectant for butyrate-producing microbes with a pool of donor-derived microbes from stool to restore the intestinal ecosystem in acute graft-versus-host gastrointestinal disease. The severity of this disease is associated with a decrease in Gram-positive anaerobic bacteria and a reduction in butyrate production.¹⁴. Batch-to-batch variability, traditionally a problem when using single-donor stool treatments, is reduced by pooling donor-derived microbes, ensuring consistency¹⁵. The company is also developing drugs to improve response to immune checkpoint inhibitor (ICI) therapies in solid tumors, based on previous studies showing that fecal microbial transplants from healthy donors or ICI responders directly improve responses to melanoma treatment.^16,17,18. By using a co-culture system that maintains the full functional diversity of the microbiome, they are better equipped to demonstrate the mechanistic details of these host-microbe interactions.

Other small molecules produced by microbes can be identified through computational and experimental approaches and used as therapeutic targets. One such example is curli, a cell surface amyloid protein produced by certain bacteria. CsgA, a major subunit of curli fibers, promotes α-synuclein pathology in the gut and brain in mice overexpressing human amyloid α-synuclein¹⁹. Vertero Therapeutics is developing an oral small molecule targeting CsgA in Parkinson’s disease patients who are positive for its expression to slow disease progression.

Non-gut microbiome datasets are also growing in size and number, with several companies working on therapies targeting the skin, oral and vaginal microbiome. Freya Biosciences correlates dysbiosis of the vaginal microbiome with changes in immune profiles, particularly increased activation of the immune response.²⁰. Increased immune responses could potentially lead to negative reproductive health consequences, such as premature birth or infertility. The company has created a cell bank for beneficial strains of Lactobacilli donor-derived communities, which can be grown on a large scale by fermentation and inserted into capsules for vaginal transplantation. Trials for the treatment of women with asymptomatic vaginal dysbiosis are ongoing, in addition to a clinical trial in women undergoing in vitro fertilization to evaluate how the vaginal microbiome is associated with pregnancy outcomes after frozen embryo transfer.

With all these new attempts to block their activity, microbes will continue to evolve their defense systems. Companies are using data sets to identify and design antimicrobial peptides or small molecules with distinct modes of action to continually combat the emergence of antimicrobial resistance. Phare Bio uses generative AI to design and develop small molecule antibiotics with novel modes of action, staying ahead of the evolution of microbes. The company expects it to produce up to 15 new small molecules by 2030. As part of a streamlined path to commercialization, once an antibiotic has passed preclinical development, large pharmaceutical companies partner with Phare Bio to move the product into clinical trials.

Researchers have been collecting microbiome data for over 15 years and now there are tools to process and use it effectively. Additionally, combining this data with AI and synthetic engineering technologies will continue to facilitate the generation of therapeutic products. Direct engineering of human microbes for therapeutic purposes is in its infancy, but it has been accomplished²¹and more will come as biological pathways continue to be explored.