Compartment-specific host association and mobility shape ARG risk in aquaculture systems

Antimicrobial resistance in aquaculture threatens environmental and public health, but the risk of ARGs cannot be inferred from abundance alone; host context and mobility potential are essential. Here, we investigated how ecological compartments shape ARG host background, mobility, and risk in aquaculture systems. We analyzed 437 metagenomes from water and sediment in freshwater and marine aquaculture across China using resistome profiling, host assignment, genetic localization, ARG-MGE co-occurrence, a four-tier risk framework, and machine learning. We detected 1413 nonredundant ARG subtypes (28 classes). Water had higher ARG diversity, stronger associations with opportunistic pathogens, and stronger mobility-related signals than sediment. High-risk ARGs were concentrated in water: Rank I ARGs were exclusive to water, and water-specific Rank II ARGs accounted for 7.2% (freshwater) and 6.9% (marine) of total ARG diversity, versus 4.2% (freshwater sediment) and 2.9% (marine sediment). The LightGBM model identified salinity, temperature, and pH as key mobility predictors. Together, these results show that ARG risk in aquaculture is jointly shaped by the ecological compartment, host association, and mobility potential, with water acting as the principal high-risk interface. This risk-oriented analytical framework provides a transferable basis for prioritizing surveillance and intervention in aquaculture environments.