Interactions between plant pathogens and root-associated microbes play an important role in determining disease outcomes. While several studies have suggested that steering these interactions may improve plant health, such approaches have remained challenging in practice. Because of low iron availability in most soils, competition for iron via secreted siderophore molecules might influence microbial interaction outcomes.Here we tested if bacterial interactions mediated by iron-scavenging siderophores can be used to predict the disease suppressiveness of microbial consortia against soil-borne Ralstonia solanacearum bacterial pathogen in tomato rhizosphere. Iron availability significantly affected the interactions within inoculated consortia and between the consortia and the pathogen. We observed contrasting effects of siderophores and other non-siderophore metabolites on the pathogen growth, while the siderophore effects were relatively much stronger. Specifically, disease incidence was reduced in vivo when the inoculated consortia produced siderophores that the pathogen could not use for its own growth. Employing siderophore-mediated interactions to engineer functionally robust microbial inoculants show promise in protecting plants from soil-borne pathogens.

All data were collected from indoor and greenhouse pot experiments, please check the corresponding article for details.

General linear mixed models were used to examine the siderophore and non-siderophore metabolite effects on the pathogen and consortia member growth in mono, pairwise and consortia co-cultures. In these analyses, we conducted separate models to explore consortia community richness, composition and strain identity effects (presence and absence of certain strain) on strain interactions. Differences in disease incidence was explained by three quantitative factors that were measured in vitro: consortia siderophore production, consortia siderophore-mediated effect on the pathogen growth and siderophore-mediated interactions between inoculated consortia members. All variables were fitted as continuous variables and one separate model was used for each variable. All analyses were performed using R 3.3.1 program (www.r-project.org).