Prediction of secreted uncharacterized protein structures from Beauveria bassiana ARSEF 2860 unravels novel toxins-like families

Abstract

Insecticides are toxic substances used to control a wide variety of agricultural insect pests. Most of these are chemicals in nature, and their increasing residues in soil, water, and fruits contribute to environmental pollution, chronic human illnesses, and the emergence of insecticide resistance phenomenon. In the context of a green environment, bioinsecticide metabolites, including proteins, are a safe alternative that mostly has selective toxicity to insects. Thus, this study aimed to predict and identify new toxin-like families through uncharacterized secreted proteins from one of the most potent entomopathogenic fungi, Beauveria bassiana ARSEF 2860, which was selected as a model. In this work, a total of 2483 amino acid sequences of uncharacterized proteins (Ups) were retrieved from the RefSeq database. Among these, 365 UPs were identified as secreted proteins using the SignalP web server. We implemented the integration of well-designed bioinformatic tools to characterize and anticipate their homologous similarities at the sequence (InterPro) and structural (AlphaFold2) levels. The structural function annotation of these proteins was predicted using DeepFRI. With 269 successfully predicted folds, we identified new putative families with pathogenesis functions related to toxins like Janus-faced atracotoxins (insecticidal spider toxin), Cry toxins (commercial insecticide from Bacillus thuringiensis), ARTs-like toxins, and other insecticidal toxins. Furthermore, some proteins that are not homologous to any known experimental data were functionally predicted as cation metal ion binding (Zn, Na, and Co) with potential toxicity. Collectively, computational structural genomics can be used to study host–pathogen interactions and predict novel families.