Antimicrobial resistance bacteria and genes detected in hospital sewage provide valuable information in predicting clinical antimicrobial resistance

Extensive use of antibiotics is significantly associated with development of antibiotic-resistant (AR) bacteria. However, their causal relationships have not been adequately investigated, especially in human population and hospitals. Our aims were to understand clinical AR through revealing co-occurrence patterns between antibiotic-resistant bacteria and genes (ARB and ARGs), and their association with antibiotic use, and to consider impact of ARB and ARGs on environmental and human health. Antibiotic usage was calculated based on the actual consumption in our target hospital. ARB was identified by culture. In isolates collected from hospital sewage, bacterial-specific DNA sequences and ARGs were determined using metagenomics. Our data revealed that the use of culture-based single-indicator-strain approaches only captured ARB in 16.17% of the infectious samples. On the other hand, 1573 bacterial species and 885 types of ARGs were detected in the sewage. Furthermore, hospital use of antibiotics influenced the resistance profiles, but the strength varied among bacteria. From our metagenomics analyses, ARGs for aminoglycosides were the most common, followed by sulfonamide, tetracycline, phenicol, macrolides, and quinolones, comprising 82.6% of all ARGs. Association analyses indicated that 519 pairs of ARGs were significantly correlated with ARB species (r > 0.8). The co-occurrence patterns of bacteria-ARGs mirrored the AR in the clinic. In conclusion, our systematic investigation further emphasized that antibiotic usage in hospital significantly influenced the abundance and types of ARB and ARGs in dose- and time-dependent manners which, in turn, mirrored clinical AR. In addition, our data provide novel information on development of certain ARB with multiple antibiotic resistance. These ARB and ARGs from sewage can also be disseminated into the environment and communities to create health problems. Therefore, it would be helpful to use such data to develop improved predictive risk model of AR, to enhance effective use of antibiotics, and to reduce environmental pollution.