• Machine learning and mass spectrometry clarify sulfonamide transmembrane transport. • Molecular dynamics simulations reveal passive diffusion capacity of sulfonamides. • Data-driven approach decouples active-passive transport roles in sulfonamide uptake. • LrABCG35 was identified as a potential sulfonamide transporter in plants. Emerging organic contaminants, including pharmaceuticals, are increasingly threatening global ecological systems. Phytoremediation is a sustainable strategy for remove pharmaceuticals, and its efficiency depends on the transporters capable of handling pollutants. However, the transporter identification is constrained by the high cost of developing mutant libraries that require a well-characterized plant genome, and the trial-and-error inefficiencies arising from unguided transporter screening without prior assessment of dominant transmembrane pathways. Herein, we developed a machine learning-molecular dynamics convergence framework that mechanistically decouples sulfonamide transport pathways in plants. Machine learning successfully predicted the overall transmembrane ability of five sulfonamides by training a Random Forest model on the transmembrane capabilities of 1366 pharmaceuticals. Molecular dynamics simulations quantified the passive diffusion capabilities of five sulfonamides. The discrepancies between overall transport and passive diffusion suggested the significance of active transport in sulfonamide uptake by plants. Transcriptome analysis, molecular docking, and heterologous expression identified LrABCG35 as a transporter for five sulfonamide antibiotics, which exhibits potential to enhance plant-mediated pollutant transport. Our approach eliminates futile engineering trials and accelerates the development of precision phytoremediation for polluted water bodies by pre-screening contaminants for transporter dependency, paving the way for more effective environmental remediation strategies.
Zhao et al. (Sun,) studied this question.