The increasing AMR caused by the overuse of antibiotics has emerged as a significant threat to global health. A precise and timely report of ARB is critical for reducing empirical antibiotic prescription, which is vital for reducing clinical AMR. Unfortunately, the unculturable or unknown isolation conditions make about 90% of bacteria that could not be cultured in vitro [16, 17]. In our study, bacteria could be isolated in only 16.17% of infectious samples, AMR information obtained from isolated bacteria could not meet the requirement that can guide antibiotic use. Compared with the species isolated by culture, metagenomics had significantly higher effectiveness in detecting bacterial species, especially those in the environment. Up to 1573 bacterial species were found in the hospital sewage. Proteobacteria and Bacteroidetes were the main phyla identified, comprising a variety of gram-negative bacteria such as Escherichia, Salmonella, Vibrio, and other pathogenic genera. Among the high abundance bacterial species found in sewage, E. coli, K. pneumonia, P. aeruginosa, Acidovorax_sp._KKS102, P. mirabilis, and S. aureus were also the main bacteria isolated in hospital, which are eventually discharged into sewage. Thus, bacterial species in wastewater can reflect the total number of clinical bacteria.
The relationships between antibiotic use and AMR involve the hospital population, pathogen genotypes, intensity of infection control, and transmission dynamics [18–20]. When analyzing the effect of antibiotic consumption on AMR, antibiotic surveillance based on sales data describes an incomplete story. Building a nationwide surveillance program to quantify antibiotic prescriptions from all sources, i.e., pharmacies, clinics, hospitals, is quite challenging to achieve . In our study, the dosage of antibiotics was calculated based on the actual consumption of hospitalized patients. Cephalosporin, β-lactase inhibitors, and Fluoroquinolones were the most frequently used antibiotics. Their high average monthly dose prescription results in higher levels of ARBs, including E. coli, P. aeruginosa, and P. mirabilis resistant to the corresponding antibiotic. The amount of antibiotics affects the distribution of ARBs with or without delay effect. A greater abundance of Proteobacteria identified in the hospital environment may be the result of antimicrobial use in the hospital [7, 22]. Consistent with the previous study, our results also indicated the close correlations of the quantities of ARBs with those of corresponding antibiotics . The reason for the correlation failed to find in low abundance bacteria, may due to the number of cases is too small to meet the statistical sample requirements. The difficulty in obtaining the samples from the vast population, the complexity of pathogens, behavioral and antibiotic response patterns, made antibiotic use guiding by bacterial sensitivity tests in a hospital almost impossible in the present period. In monitoring the occurrence of AMR, a significant advantage of hospital sewage is that such samples can be easily obtained and analyzed without ethical concerns compared to samples directly collected from humans .
The induction and development of ARGs could be related to the concentration effects of antibiotics on the bacterial groups harboring these genes . For tracking the potential hosts of ARGs, the co-occurrence patterns between ARG subtypes and microbial taxa were explored by correlation analysis [25, 26]. In our study, more than 40 mobile ARGs were all from Proteobacteria, and many novel pairwise of ARG-bacteria or ARG-ARG examples observed from the network. E-coli, the most abundant bacterium isolated from our clinical samples, was highly resistant to Quinolones, Penicillin, Beta-lactamase inhibitor, Cephalosporin, Tetracycline, and Aminoglycosides. In the network, E-coli was co-occurrence with multiple ARGs, including AAC(6')-30/ AAC(6')Ib', acrD, acrF, baeR, CRP, cmlA5, emrA, marA, patA, pmrF, floR, and pp.flo, indicating that multi-resistance of E-coli may be induced by these ARGs. The intrinsic ARGs are merely antibiotic determinants conferring resistance phenotype. Thus, understanding the intrinsic resistome will, therefore, contribute to the prediction of the emergence and evolution of antibiotic resistance in the future [27, 28].
The human activities using antibiotics lead to significant amplification of the original ARGs in social clinical settings . Underexposure to antibiotics, higher incidences of co-occurrence ARG subtypes of the same type may generally reflect selective pressure exerted by the same antibiotics in sewage, as our results showed, antibiotic consumption of a few months before influenced the abundance of ARGs in sewage later, with the peak of ARGs lagging that of antibiotics use for 2–6 months. Previous studies showed that Cephalosporin consumption of 7–10 months before had a significant influence on resistance, while consumption of Penicillin / Beta-lactamase inhibitor produced resistance without delay . The possible explanation for this phenomenon may be that different antibiotics have different selective capabilities , and the abundance of ARGs is driven by complex interplays of different environmental, antibiotic types, and anthropogenic variables, rather than being related to any single factor. That clinical antibiotic usage affects the abundance and distribution of ARGs in sewage was marked in Aminoglycosides, β-lactamase inhibitor, Macrolides, Penicillin, and the enriched occurrence of these ARGs was closely associated with the antibiotic’s usage in the target hospital.
The correlation between antibiotics and ARGs was not observed in tetracycline and the sulfonamide. A previous study revealed that ARGs of tetracycline and sulfonamide were the top three widely transferred ARGs . Despite deficient consumption at present, a high abundance of ARGs to tetracycline were account for 10.49% of total ARGs. Consistent with the results of ARGs, high resistance to tetracycline was found in 80% E.coli, in 100% P. mirabilis, and in 75% K. pneumonia of clinical isolates. A possible explanation may be a potential source of ARGs in the environment was caused by the continuous prophylactic use of antibiotics as a for animals . Another reason is that antibiotics use and their corresponding ARGs do not always appear simultaneously .
The inter-types co-occurrence of ARGs, such as beta-lactam and aminoglycoside resistance by methicillin resistant S. aureus  and prevalent co-resistance to Aminoglycosides, Tetracycline, and Quinolones by beta-lactamase (ESBL) producers , are more likely driven by a co-resistance mechanism, i.e., presence of different resistance determinants on the same genetic element . Our correlation network revealed bacteria bearing multiple resistance genes, as well as cross-resistance between antibiotics. The most abundance ARG, sul1, usually found in the conserved region of class1 integrons, is capable of transferring between bacteria in different environments. Other high abundance ARGs, like the AAC.6...Ib8, AAC(6')-30/ AAC(6')Ib', patA, mexF, aadA3, and QnrS2, located in the hub of a network, may own high potential risks for horizontal gene transfer between different bacteria in the sewage. ARGs, including aadA1, aac(3)-II and aph(6)-I ( aminoglycoside ), tetA and tetG (tetracycline ), sul1 and sul2 (sulfonamide ), catA and catB (chloramphenicol ), were carried by the same genome . When the genes that cause resistance phenotypes are located together on the same genetic element (MGEs), such as a plasmid, transposon, or integron, co-resistance occurs [38, 39]. This information may provide some clues regarding multi-drug resistance ARBs happened. The correlations between ARGs and bacterial taxa in the network could provide some insights into the potential mechanism of human pathogens prone to the acquisition of ARGs . Based on the ARGs located in the hub of a network, and we could track the possible routes of ARGs transmission and dissemination among ARB. The interactions between the number of resistant bacteria and antibiotics may have contributed to a significant increase of some ARGs, which confer resistance to the most frequently used antibiotics in the hospitals [41, 42]. Therefore, the co-occurrence patterns between bacteria and the ARGs could help to propose a novel approach for risk assessment of ARGs acquisition by HBPs in human-impacted environments , and the future level of resistance, esp. short- to medium-term, could be forecasted based on antibiotic usage and incidence of bacteria , though it remains unclear for the diversity, distribution, and fate of ARGs in urban water systems .
Hospital discharges are essential reservoirs of ARB and ARGs [44, 45], which promote the potential spread of AMR to the environment [46, 47]. More attention should be paid to assess ARGs in hospital sewage, which could provide a great solution to clinical AMR monitoring.