A neural network model predicts the consequences of crosstalk in bacterial quorum sensing

Many bacteria use chemically similar molecules to communicate, resulting in crosstalk between bacterial signaling networks. Such crosstalk can have unexpected consequences for decision making in heterogeneous communities of cells. Here we examine crosstalk within a community composed of five strains of Bacillus subtilis, with each strain producing a variant of the quorum sensing peptide ComX. Co-cultures of these strains create in a mixture of ComX variants, resulting in variable levels of gene expression in each strain. To predict gene regulation in communities producing multiple signals, we implement a neural network model. Experimental quantification of crosstalk between pairs of strains parametrized the model, enabling the accurate prediction of activity within the full five-strain network. Interactions weights between the five signaling networks were both positive and negative, of variable magnitude, and asymmetric. Signal crosstalk within the five-strain community results in multiple community-level quorum sensing states, each with a unique combination of quorum sensing activation among the five strains. The community-level signaling state was influenced by the ratio of strains as well as dynamics of community composition.