Culture-enriched sequencing reveals river confluence sediments as high-risk reservoirs of viable multidrug-resistant...

Antimicrobial resistance (AMR) in aquatic ecosystems is an escalating One Health concern. However, viable antibiotic-resistant bacteria (ARB)-particularly pathogenic strains-and their mobility remain poorly characterized at hydrological interfaces such as river confluences. Here, we integrated culture enrichment, high-throughput 16S rRNA gene sequencing, isolate phenotyping and whole-genome analysis to profile ARB and antibiotic-resistant bacterial pathogens (ARBPs) in sediments from the Fenhe River-Yellow River confluence. Non-selective enrichment reduced community complexity yet uniquely recovered dozens of rare taxa absent from direct sequencing. Antibiotic enrichment induced pronounced, drug-specific ARB shifts; antibiotic type explained more variance (19.3%) than hydrological region (11.2%). Pathogen signals were strongly amplified by enrichment, and ARBP communities retained significant regional clustering. Notably, the confluence hydrodynamic region (CHR) consistently exhibited the highest ARBP richness. Of the 121 recovered isolates, 94.2% were phenotypically resistant and 73.5% were multidrug-resistant; 89.3% matched in situ ASVs, bridging community profiles and cultivable strains. We recovered seven high-risk multidrug-resistant pathogens (belonging to Pseudomonas, Acinetobacter, Aeromonas) as viable isolates, even though they were rare or undetected by direct sequencing. Whole-genome sequencing revealed 658 virulence factors and 312 antibiotic-resistance genes (ARGs). Clinically relevant determinants (e.g., AAC(6')-Iaa, OXA-917, OprN) were embedded within mobile genetic elements, including transposons, plasmid-like contigs, and integrative and conjugative elements (ICEs). The edeine acetyltransferase gene edeQ showed 100% nucleotide identity to alleles from clinical sources, indicating overlap between environmental and clinical resistomes. Collectively, our findings highlight river confluences as priority surveillance nodes and demonstrate that culture-enriched sequencing more effectively quantifies viable AMR hazards than sequencing alone.