Using cross-species co-expression to predict metabolic interactions in microbiomes.

Abstract

In microbial ecosystems, metabolic interactions are key determinants of species relative abundance and activity. Given the immense number of possible interactions in microbial communities, their experimental characterization is best guided by testable hypotheses generated through computational predictions. However, widely adopted software tools - such as those utilizing microbial co-occurrence - typically fail to highlight the pathways underlying these interactions. Bridging this gap will require methods that utilize microbial activity data to infer putative target pathways for experimental validation. In this study, we explored a novel approach by applying cross-species co-expression to predict interactions from microbial co-culture RNA-sequencing data. Specifically, we investigated the extent to which co-expression between genes and pathways of different bacterial species can predict competition, cross-feeding, and specialized metabolic interactions. Our analysis of the Mucin and Diet-based Minimal Microbiome (MDb-MM) data yielded results consistent with previous findings and demonstrated the methods potential to identify pathways that are subject to resource competition. Our analysis of the Hitchhikers of the Rhizosphere (THOR) data showed links between related specialized functions, for instance, between antibiotic and multidrug efflux system expression. Additionally, siderophore co-expression and further evidence suggested that increased siderophore production of the Pseudomonas koreensis koreenceine BGC deletion-mutant drives siderophore production in the other community members. In summary, our findings confirm the feasibility of using cross-species co-expression to predict pathways potentially involved in microbe-microbe interactions. We anticipate that the approach will also facilitate the discovery of novel gene functions through their association with other species metabolic pathways, for example, those involved in antibiotic response. ImportanceAn improved mechanistic understanding of microbial interactions can guide targeted interventions or inform the rational design of microbial communities to optimize them for applications such as pathogen control, food fermentation, and various biochemical processes. Existing methodologies for inferring the mechanisms behind microbial interactions often rely on complex model-building and are therefore sensitive to the introduction of biases from the incorporated existing knowledge and model-building assumptions. We highlight the microbial interaction prediction potential of cross-species co-expression analysis, which contrasts with these methods by its data-driven nature. We describe the utility of cross-species co-expression for various types of interactions and thereby inform future studies on use-cases of the approach and the opportunities and pitfalls that can be expected in its application.