Antibiotic resistance gene classes show no detectable shift despite extensive microcystin-associated river microbiome reorganization

Antibiotic resistance gene (ARG) signals can reflect increasing environmental resistance or changes in the microorganisms carrying those genes. Cyanobacterial blooms make this distinction especially important because they can reorganize entire freshwater microbial communities. We combined repeated metagenomic sampling at ten sites over eleven rounds along an 85 km Nakdong River reach with taxonomic, functional, and genome-resolved analyses and site-aware machine learning. Sampling round explained 60.9% of microbial-community variation, compared with 5.0% for site, indicating that temporal turnover dominated persistent spatial structuring. After adjusting for site, season, and water temperature, high-microcystin conditions were associated with 3,241 microbial features. None of the 18 tested ARG drug classes showed a corresponding association. The resistance-candidate inventory sums candidate annotations in representative metagenome-assembled genomes (MAGs) assigned to origin samples; drug-class profiles partition the same annotations rather than measuring sample-wide ARG abundance. Across 301 representative MAGs, candidate counts were strongly positively associated with genome size, indicating that the inventory partly reflected the microbial genomic content represented. In exploratory models, predicted resistance-candidate inventories differed between community states. High-microcystin conditions therefore coincided with extensive microbial reorganization without a detectable parallel ARG-class response. Microcystin measurements should be paired with direct ARG measurements and community profiles to track toxin hazard, resistance determinants, and microbial turnover.