Identification of novel toxins associated with the extracellular contractile injection system using machine learning.

Secretion systems play a crucial role in microbe-microbe or host-microbe interactions. Among these systems, the extracellular contractile injection system (eCIS) is a unique bacterial and archaeal extracellular secretion system that injects protein toxins into target organisms. However, the specific proteins that eCISs inject into target cells and their functions remain largely unknown. Here, we developed a machine learning classifier to identify eCIS-associated toxins (EATs). The classifier combines genetic and biochemical features to identify EATs. We also developed a score for the eCIS N-terminal signal peptide to predict EAT loading. Using the classifier we classified 2,194 genes from 950 genomes as putative EATs. We validated four new EATs, EAT14-17, showing toxicity in bacterial and eukaryotic cells, and identified residues of their respective active sites that are critical for toxicity. Finally, we show that EAT14 inhibits mitogenic signaling in human cells. Our study provides insights into the diversity and functions of EATs and demonstrates machine learning capability of identifying novel toxins. The toxins can be employed in various applications dependently or independently of eCIS. Synopsis: Toxins delivered by the extracellular contractile injection system (eCIS) of 950 microbial genomes were identified using XGBoost classifier algorithm, four of which were experimentally confirmed and functionally characterized. eCIS is a poorly studied microbial toxin delivery system encoded by many environmental microbes. The first algorithm for predicting eCIS-associated toxins (EATs) was developed using genomic and biochemical features, and was used to identify ~2,200 EATs from 950 genomes. Features characterizing the N terminal signal peptide of EATs were identified and combined into a signal peptide score. Four new EATs genes (EAT14-17) were experimentally validated as being toxins, and critical EAT residues within the predicted catalytic sites were identified. EAT14 inhibits SRE-dependent mitogenic signaling in HEK293T cells. Toxins delivered by the extracellular contractile injection system (eCIS) of 950 microbial genomes were identified using XGBoost classifier algorithm, four of which were experimentally confirmed and functionally characterized. [ABSTRACT FROM AUTHOR]