Disease-specific loss of microbial cross-feeding interactions in the human gut

The human gut contains hundreds of microbial species forming a complex and interdependent metabolic network. Over half of the metabolites consumed by gut microbes are by-products of microbial metabolism with the waste of one species serving as nutrients for others. Species interdependence can render microorganisms vulnerable to local extinction if a partner is lost unless alternative species are available to fill that niche. In this context, having functionally redundant species with the ability to produce or consume the same nutrients is beneficial for the host. While it is generally accepted that high functional redundancy is a characteristic of resilient human gut microbiomes, the human health impacts of redundancy in metabolic interactions remain largely uncharacterized. Restoring the diversity of cross-feeding microbial partners represents a logical but still largely unexplored rubric to fight a wide range of diseases linked with an unbalanced gut microbiome.

Mechanistic models that simulate microbial metabolism in silico hold the promise to fill our knowledge gap on microbial metabolic interactions. Genome-scale metabolic models (GEMs) are based on increasingly comprehensive databases linking genes to biochemical and physiological processes. These models have been used to estimate metabolic exchanges between pairs of bacterial species for over a decade. Developments in automating the reconstruction of GEMs and the availability of manually-curated GEMs for thousands of gut microorganisms have paved the way to build metabolic models for complex microbial communities. Methodological advances now allow modelling interactions between multiple species, and a recently developed workflow by Zorrilla and colleagues now allows reconstructing metabolic models directly from large-scale metagenome datasets. Studies using community-wide metabolic models have found dozens to hundreds of significantly different metabolic exchanges in the gut microbiome associated with type 2 diabetes and in inflammatory bowel disease when compared to healthy controls. A method to rank these metabolic interactions according to an ecology-based framework provides the opportunity to generate targeted hypotheses underlying mechanistic links between the gut microbiome and diseases.

Here, we introduce a metabolite exchange scoring system derived from metagenome-scale metabolic models, designed to identify the potential microbial cross-feeding interactions most affected in disease. We apply our conceptual framework to an integrated dataset of 1661 publicly available stool metagenomes, encompassing 15 countries and 11 disease phenotypes.

Our framework identified both known and novel microbiome-disease associations, including a link between colorectal cancer and the microbial metabolism of ethanol, a connection between rheumatoid arthritis with microbially-derived ribosyl nicotinamide, and links between Crohn’s disease and specific bacteria that metabolise hydrogen sulfide. The scoring system can help quantify and identify context-dependent disruptions of microbial interactions, which may be targets for microbiome-based medicines.