Metabolite co-variation networks reveal keystone functions and an emergent pathogen state in the human urobiome.

Microbial communities within the human urobiome are intricate ecosystems that undergo dynamic changes influenced by metabolic interactions, including cross-feeding, competition, and cooperation. These interactions, rather than taxonomic diversity or individual metabolic potential, drive the collective behavior of these communities.

In postmenopausal women, recurrent urinary tract infections (rUTIs) are linked to persistent infection dynamics involving multiple bacterial species. The ability of resident microbial communities to thwart pathogen establishment, termed colonization resistance, is increasingly recognized as a function of the metabolic interactions within the urobiome itself, not solely attributable to a single species.

However, traditional methods such as taxonomic profiling and classical differential abundance analysis fall short in fully characterizing these interactions. To unravel this complexity, we introduced PhenoRewire, a network-based framework that evaluates how metabolite co-variation is altered between distinct biological states using untargeted metabolomics data.

We employed this framework using induced pluripotent stem cell (iPSC) urothelial organoid-derived barriers co-cultured with synthetic urobiome communities, serving as a model system for urobiome-pathogen dynamics relevant to rUTIs. In the first approach, we co-cultured clinically isolated uropathogens Escherichia coli and Enterococcus faecalis with a three-member urobiome community comprising Lactobacillus gasseri, Lactobacillus crispatus, and Gardnerella vaginalis.

Our findings revealed that E. coli induced a metabolic reorganization, while E. faecalis disproportionately amplified this reorganization. PhenoRewire disentangled the six-fold increase in the metabolic network, mediated by E. faecalis, as a metabolic facilitator, unveiling an emergent urobiome-pathogen co-variation architecture that was not present in either community individually (1,781 vs 227 edges).

Furthermore, in a six-member urobiome single-strain dropout experiment, the absence of the sole Actinomycete Winkia anitrata resulted in significant network collapse, with a drastic reduction in Louvain modularity from 0.707 to 0.038. This finding identified W. anitrata as the sole non-redundant keystone of the community. Our results underscore the utility of untargeted metabolomics co-variation network analysis in conjunction with a urothelial host model to elucidate community dynamics.

This framework provides a template that can be applied to other complex microbial communities where ecological behavior remains an open question.