(26g) A Split-FAST Based Biosensor for In Vivo Screening of Protein Soluble Expression and Stability

Understanding protein solubility and structural stability is pivotal in protein design, elucidating the pathogenesis of certain human diseases, and optimizing industrial biocatalysis applications. These factors determine the functionality of proteins under physiological conditions. Establishing an accurate and high-throughput method to correlate protein sequence with solubility/stability has garnered significant interest. While fluorescent protein-based biosensing methods have been developed for this purpose, they often encounter challenges related to protein co-translational folding, and sequence modifications which can alter their original properties, thereby limiting their applicability. In this study, we present a universal in vivo protein solubility/stability biosensing method by integrating the Protein of Interest (POI) into the Split-FAST (Fluorescence-Activating and Absorption-Shifting Tag) system. This engineered system effectively separates intracellular protein synthesis and detection phases, enhancing accuracy in measuring soluble protein content. Validation was performed using mutation libraries of RbTA (Rhodobacter sp.-derived ω-transaminase), Fluc (Firefly luciferase), hAβ (human amyloid-β-peptide), and human α-synuclein. In all validation cases, our method achieved an correlation coefficient R-square value of 0.85 or higher when compared with solubility levels obtained from SDS-PAGE. Notably, for RbTA and Fluc, our approach enabled in vivo characterization of both protein solubility and thermostability, which was unattainable with traditional split-GFP biosensors. Leveraging mutation library design and high-throughput screening, we identified more than ten robust RbTA biocatalysts with industrial applications in chiral amine synthesis, among which the best performed mutant exhibited a 17.9-fold increase in heat inactivation half-life and a 11.2℃ increase in T50. Furthermore, in conjunction with dSort-seq (deep-learning–assisted Sort-Seq) approach, we utilized this system to investigate the mechanistic impact of the N-terminal sequence of nattokinase on its intracellular expression and solubility.