Urban sewage resistomes partially reflect clinical resistomes

Antimicrobial resistance (AMR) poses a major global public health threat, and ongoing surveillance of antimicrobial resistance genes (ARGs) is critical to mitigate current and future risks. Sewage-based ARG surveillance is gaining traction, but insight into how it compares to surveillance by clinical bacterial isolates is limited, especially when it comes to ARG mutational variants. We compared ARGs identified in clinical bacterial isolates (n = 2,989) with those detected in sewage metagenomes (n = 468) across 33 countries. ARG variant detection data from clinical isolates and sewage metagenomes shared some regional patterns in detection, but many ARG variants were detected exclusively in either sewage metagenomes or clinical isolates. We found that across all samples, only 69% of ARG clusters detected in clinical isolates were also detected via read mapping in sewage. Some ARGs highly prevalent in clinical isolates were not detected in sewage. Among clinically widespread ARGs, prevalence varied across bacterial species and clinical isolate types depending on whether the ARGs were also detected in sewage. This could indicate that sewage surveillance is better suited for detection of clinically relevant ARGs prevalent in certain bacterial species and infection sites than others. Spearman correlation between ARG abundance in sewage and the proportion of clinical isolates from the same country with detection was 0.28 overall, with stronger correlations for certain ARGs. The results demonstrate that sewage ARG profiles correlate, to some extent, to the clinical AMR landscape, but do not capture the full spectrum of clinically relevant ARGs at currently realistic sequencing depths.IMPORTANCEAntimicrobial resistance (AMR) is a major public health threat. Surveillance of AMR is important and can be conducted via the detection of antimicrobial resistance genes (ARGs). Sewage can be used as a medium for surveillance as an alternative to analyzing individual bacterial isolates from health clinics. We compared detection in large global data collections of sewage metagenomes and clinical isolates. We found that while there were significant positive correlations between findings in sewage and clinical isolates, some widespread clinical ARGs were not detectable in sewage. This should be considered if establishing sewage surveillance systems.