Identification and Risk Assessment of Antibiotics and Their Transformation Products in a Large-Scale River Using Suspect and Nontarget Screening and Machine Learning

Abstract

The occurrence of antibiotics in aquatic environments has been well-known, and more research is needed about their transformation products (TPs) and associated environmental risks, especially antimicrobial resistance, in large-scale rivers. Here, we developed a highly comprehensive target, suspect and nontarget screening machine-learning workflow based on high-resolution mass spectrometry to identify unknown antibiotic TPs in a large-scale river. We identified 46 antibiotics and 144 TPs with a confidence level of 1 to 3. Parent antibiotics were dominated by sulfonamides (26.1%), while TPs were dominated by macrolides (34.0%), with main transformation pathways of oxidation and hydrolysis. More TPs were found than the number of parent antibiotics. Fourteen, 18, 97, and 36 TPs had greater persistence, bioaccumulation, mobility and toxicity, respectively. One PBT compound was identified, and no PMT compounds were observed. Fifteen (10 antibiotics and 5 TPs) and 12 (3 antibiotics and 9 TPs) compounds were assessed as high ecological risk and high antimicrobial resistance risk, respectively. The presence of TPs leads to higher total risk. Thus, TPs should be included in future monitoring and risk assessment.