Unraveling the activity of phage-carrying antibiotic resistance genes in constructed wetlands.

Antimicrobial resistance (AMR) is a global public health challenge, and risk assessments based solely on gene abundance often underestimate the immediacy of resistance dissemination. This study presented a carrier-centric framework integrating metagenomic and metatranscriptomic profiling with deep learning-based identification of mobile genetic elements, applied to a full-scale constructed wetland (CW). CW overall reduced ARG burdens, with genomic abundance in plants, sediments, and water decreasing by 98.5%, 80.9%, and 88.8%, respectively. However, transcriptional activity showed an opposite trend, with sediments exhibiting the highest ARG expression, highlighting their pivotal role in the persistence and dissemination of resistance. In sediments, phage-mediated expression increased sharply from 4.0% to 92.5%, exceeding plasmid-associated levels by ~276-fold, revealing a low-abundance but high-activity residual risk pattern. Furthermore, 16 of the 310 recovered nonredundant MAGs were identified as phage hosts, 11 of which were potentially pathogenic, antibiotic-resistant bacteria (PARB) and were more active in sediments than in water or plants. These findings indicate that transduction within high-density, biofilm-associated niches constitutes a key terminal risk source. In addition, sediment acts as a high-risk reservoir where redox and ionic gradients, together with residual lomefloxacin and other antibiotics, enhance phage infectious activity and the accumulation of ARGs. Through cross-compartment transmission along the sediment-water interface, these phage-associated and PARB populations continuously seed the overlying water. It is recommended that ARG risk assessment shift from static abundance to an activity-aware, carrier- and host-resolved approach, prioritizing sediment-targeted transcript monitoring and phage transduction early warning to support risk mitigation in CW.