The Ohashi Medical Center in Toho University is a general hospital opened in 1973. Prior to relocation, the hospital consisted of four buildings: West building (BW), Administration building (BA), Central building (BC), and East building (BE) and had a total of 430 beds, and accepted an average of 850 outpatients per day (Fig. 1). In the BW, there were intensive care wards (ICU, cardiovascular surgery, cardiovascular internal medicine, respiratory, and internal medicine), but no outpatient departments. In the BA, there were several outpatient departments, but the toilets were for only healthcare workers. In the BC, patients with digestive diseases, urinary diseases, and bone fracture were mainly accommodated, and the number of beds was the largest among the old hospital buildings. The BE included the highest number of outpatient departments with no intensive care wards.
The hospital relocated to the new building (BN) on June 20th, 2018. In the BN, all outpatient departments and wards were integrated into one building. The sewage from patients was not mixed with other wastewater sources (laboratory and general wastewater) in the wastewater systems of the old and new hospital (Fig. 1).
The average number of hospitalized patients in May was 246, and the number of hospitalized patients in BC, BE, and BW building started to decrease in June. Consequently, 66 patients were transferred to BN on June 16 (Fig. 2). After the new hospital was opened on June 20, we started outpatient examinations and accepted a new hospitalized patient, and the number of inpatients gradually increased and exceeded 250 on July 2.
Metagenomic analysis of tank flora in each building
To elucidate the differential microbial flora in the hospital sewage tank (tank flora), metagenome DNA-seq analysis of sewage samples was conducted. The dominant bacteria in the sewage were classified according to the metagenomic data （Supplemental Data Set S1), wherein the most common genus was Aeromonas, followed by Citrobacter and Bacteroidetes (Fig. 3a). In addition, the proportion of genera varied depending on the building; the largest genus comprised of Citrobacter in STA, Aeromonas in STC, Bacteroides in STE, Acinetobacter and Citrobacter in STW, and variable genera in STN. Notably, less diversity was observed in STA (Fig. 3b) tanks because the toilet users in BA were limited to staff members, and the amount of excretion was also small.
In our hospital's sewage tank, bacteria originating from human gut flora were expected to be predominant as only stool and urine excreted from the toilet are stored. Composition of bacterial flora in hospital sewage has been reported to comprise of human gut flora Bacteroides, Faecalibacterium, Bifidobacterium, Blautia, Roseburia, and Ruminococcus, in addition to Klebsiella, Aeromonas, Enterobacter, Prevotella, and Comamonas . In the gut flora of healthy Japanese adults, the anaerobes Bifidobacterium, Blautia, Bacteroides, Faecalibacterium, Eubacterium, Ruminococcus, and Collinsella are predominant . However, hospital sewage is mainly influenced by the patients’ gut flora, which is different from healthy individuals. Thus, flora in each tank at the old hospital (STA, STW, STC, and STE) exhibited different bacterial compositions depending on the characteristics of the patients present in each building (Fig. 1).
Since June 5th, only construction company and hospital staff visited the BN, and the STN was composed of a few genera such as Citrobacter, Aeromonas, Escherichia, and Klebsiella. Thereafter, Arcobacter, Citrobacter, Aeromonas, Bacteroides, and Acinetobacter comprised the main constituents, and the variety of genera increased with time. In the STN, the composition of bacteria was partially consistent with that in the STC, STE, and STW. It appears that all bacterial species detected in the old sewage tanks (STA, STC, STE and STW) were integrated into STN. For instance, Aeromonas originated from STC and Citrobacter and Acinetobacter originated from STW. In the STN, Comamonas and Arcobacter were significantly detected after the transfer of inpatients. Comamonas is a genus of aerobic proteobacteria, generally considered as environmental bacteria with less pathogenicity. It is reported that one of the Comamonas species, C. testosterone, aerobically degrades testosterone when cultured in testosterone-containing media . In the STN, the enrichment of Comamonas in sewage suggests the possibility of its role in decontamination of medical compounds. Arcobacter spp. are gram-negative rod-shaped bacteria of the family Campylobacteraceae. A. butzleri, A. cryaerophilus, and A. skirrowii have been implicated in severe acute gastrointestinal infections and bacteremia after exposure to shellfish and/or ingestion of contaminated water, or inadequately cooked meat/shellfish [International Commission on Microbial Specification for Foods (ICMSF, 2002)]. In the present study, there were no instances of infectious enteritis due to Arcobacter in the hospital; however, the causative bacteria of infectious enteritis are not easily identifiable, except in limited conditions. Arcobacter is detected in WWTP in several countries [68-71], and the presence of antimicrobial resistant Arcobacter has also been documented [70, 72]. E. coli and Klebsiella were not abundant, but were consistently detected, in all sewage tanks of buildings. It is unclear whether the difference in the bacterial composition of each tank reflects characteristics of each building (department, diseases, drugs administered). Unlike the sewage system in the old hospital (STA, STC, STE, and STW), the new hospital has a single sewage tank (STN). Thus, the bacterial composition of STN looks similar to that of STW, suggesting that the flora in STW was most influenced by STN.
Similarity and diversity of bacterial population amongst tanks was analyzed using principal co-ordinates analysis (PCoA) based on bacterial genus level (Fig. 4). A PCoA used a total of 25 sewage samples of each tank excluding STA, and showed that the STN and STW groups were closely plotted, while the STE and STC groups were largely separated. The BW and BN contain rooms where seriously ill patients are treated, such as intensive care unit (ICU), coronary care unit (CCU), and respiratory disease wards (Fig. 1). According to PCoA analysis, the distribution of flora in STN after the transfer of patients was similar to that of STW, indicated by the presence of Aeromonas, Citrobacter, and Comamonas (Fig. 4). The distribution of tank flora can be strongly influenced by the wards for patients who were prescribed strong doses of antibiotics and other therapeutic agents. Furthermore, tank flora can be instantly affected by excrement of hospital patient/staff/visitor because the bacterial composition of STN was comparable to that of STW within one month after the relocation.
In this study, the old hospital sewage tanks (STA, STC, STE, and STW) were separately installed in each building; thus, a marked difference was observed among tanks. In contrast, it was difficult to identify the relationship between the flora and specific departments in the new hospital sewage tank, STN, because of a single tank in the building. For the purpose of monitoring department-specific ARB, it may be beneficial to install department-specific sewage tanks.
Analysis of β-lactamase genes in flora of tanks
The metagenome next generation sequence (NGS) reads corresponding to β-lactamase genes were identified in original hospital sewage samples (Fig. 5a) and EPOs from each tank were selected on CHROMagar ESBL (Fig. 5b). In the original hospital sewage samples, various β-lactamase genes such as blaGES, blaOXA, and blaCMY were detected in flora of STC, STW, and STN, while almost no β-lactamase genes were detected in flora of STA and STE. This may be due to the limited sensitivity of metagenomic detection in this study. The blaIMP gene was detected from STC and STW samples, and the blaCTX-M gene was detected from STW and STN samples (Fig. 5a). CHROMagar ESBL selection facilitated the detection of EPOs as β-lactamase genes were detected in STA and STE using this technique. Moreover, blaIMP and blaCTX-M genes were present in flora of all sewage tanks (Fig. 5b).
Genome comparison between sewage and clinical isolates
Although metagenomic analysis showed the number of each bacterial species and ARG, it is impossible to deduce which bacterial strain carries a particular ARG. Thus, whole genome sequencing was performed for 78 EPO isolates from STW0522 and STN0717, and 20 EPO clinical isolates (May 8, 2018 to July 17, 2018) for comparison of genomic sequences of EPOs from sewage tanks and clinical sources. In the STW0522 sample, bacteria such as E. coli, Klebsiella, Enterobacter, Citrobacter, and Achromobacter were detected, and in the STN0717 sample, E. coli, Klebsiella, Enterobacter, and Citrobacter were detected (Table 1). Clinical isolates included only E. coli and the Klebsiella spp., wherein E. coli displayed sequence variations compared to sewage isolates (Table 2).
A pairwise single nucleotide variation (SNV) analysis of the core genome was conducted for all E. coli strains (Fig. 6). The E. coli sequence types (STs) included ST393 (n = 16) , ST38 (n = 8), ST131 (n = 5) , ST1011 (n = 4) , ST12 (n = 1), ST73 (n = 1), ST9586 (n = 1), and ST224 (n = 1). E. coli ST12, ST73, ST131, and ST1011 were detected exclusively in clinical isolates (Table 2). Clinical EPO isolates (THO-008 and -019 from same patient; Patient-No. 8) comprised of ST393 harboring blaCTX-M-27 and there was no difference in SNVs between isolates (Fig. 6), indicating that the clone was identified from the same patient. In addition, these clinical isolates also showed no SNVs with sewage isolates (14 isolates; STN0717-1 to -11, 14, 15, and 19), suggesting that these sewage isolates may have originated from the same patient (Patient-No. 8) (Supplemental Data Set S2). Monitoring of ARB/ARGs in hospital sewage may enable detection of latent carriers or nosocomial infections.
Three ST38 clinical isolates (THO-002 identified from Patient-No. 2, THO-007 and -020 identified from Patient-No. 7) harboring blaCTX-M-14 were identified within a three-week duration, and further SNV analysis revealed marked 21-110 SNVs in the core-genome. This result suggested that all three isolates are closely related; however, there was no strong evidence of a nosocomial outbreak associated with this single clone. It is reported that molecular evolution of E. coli genome is possible with less than 5 SNVs within a 60-day duration . In contrast, ST38 sewage isolates (STN0717-12, -13, -17, -18, and -21) harboring blaCTX-M-55 exhibited strict clonality with 3 SNVs, and showed ≥123 SNVs with clinical isolates (THO-002, -007, and -020) (Fig. 6). ST131 is an E. coli isolate responsible for a worldwide pandemic and carries a broad range of pathogenicity and antimicrobial resistance-associated genes, including a variety of β-lactamase genes (CTX-M family, TEM, SHV, and CMY β-lactamases) on a transferable plasmid [74-77]. Five ST131 isolates (THO-001, -003, -005, -015, and -016) harboring various CTX-M genes (blaCTX-M-15, CTX-M-27, CTX-M-44) were identified from individual patients and 32−84 SNVs were observed between the isolates, suggesting that the isolates were closely related but not identical (Fig. 6).
Thus far, the healthy carriage rate of EPOs has been rising worldwide. In Japan, detection rate of EPO was reported at 12.2% in healthy adult volunteers  and at 15.6% in healthy food handlers . Additionally, it is reported that 92.9% of EPOs were blaCTX-M gene positive . The CTX-M genes (blaCTX-M-14, blaCTX-M-27, blaCTX-M-15 and blaCTX-M-2 in descending order in size) are mainly identified in Japan and its reported sequence is similar to that observed in EPOs in the hospital sewage and clinical isolates in this study (Table 3). Prominent global trends in CTX-M epidemiology indicate that the reduced occurrence of blaCTX-M-2 and emergence of blaCTX-M-27 , suggesting that healthy carriers could be at risk for nosocomial infections by EPOs.
In the sewage samples, isolates possessed various β-lactamase genes; blaSHV-12, blaCTX-M-14, and blaCTX-M-62, in Enterobacter kobei; blaSHV-12 in Enterobacter cloacae; blaCTX-M-9 in Enterobacter homaechei; blaIMP-11 in Pseudomonas monteilii, and blaSHV-12 in Raoultella planticola (Supplemental Data Set S2). Many of the potential EPOs harboring ARGs in the sewage tanks were different from the clinical isolates. It is not clear whether these EPOs were excreted by healthy carriers or were transformed by acquiring the ARGs in the sewage tank.
Among the 156 CHROMagar ESBL-positive strains from hospital sewage tanks (STW0522 and STN0717) (Supplemental Data Set S2), carbapenemase gene (blaIMP-11) was identified only in Pseudomonas monteilli, but not in E. coli and Klebsiella (Table 3). The blaIMP-11 gene was originally found in clinical isolates of E. cloacae , K. pneumoniae, E. coli , P. aeruginosa , and Acinetobacter species  in Japan. P. monteilli is a Gram-negative bacteria of the P. putida group and occurs in the soil, garbage, and drains . P. monteilli is less pathogenic to humans, but was isolated from clinical samples [49, 86-88]and hospital environment [85, 89]. P. monteilii may play a role as an Metallo-β-lactamase (MBL) reservoir and transfer of MBL genes to other species such as Enterobacteriaceae may be a cause of concern, especially in hospital sewage tanks [49, 87, 89, 90].
Concentration of residual antimicrobial agents in hospital sewage tanks
There is a possibility that sewage tanks provide an environment conducive to the development of novel drug resistant bacteria. Sewage tanks are reservoirs of ARGs and transfer of genes between bacteria may occur under the selective pressure of antimicrobial agents and antiseptics. Indeed, concentration of chemical contaminants in the tank (Table 4) showed that the most predominant antimicrobial agent was levofloxacin (32,500 ng/l) and clarithromycin (13500 ng/l), which may potentially exhibit inhibitory action in the sewage tank environment, although their concentrations were below MIC breakpoints.
In hospital wastewater, ciprofloxacin concentration was reported to be 22,000 - 179,000 ng/L in Germany , 19,110 - 428,00 ng/L in Vietnam, and clarithromycin was reported to be up to 960 ng/L in Portugal  and 760 - 72,800 ng/l in Singapore . The present study detected an antibiotic concentration similar to the above reports. Although β-lactam antibiotics were not measured in this study, they are known to be almost undetectable in environmental samples [93, 94]. This is because unstable lactam rings increase susceptibility of β-lactam antibiotics to hydrolysis immediately after excretion [94, 95]. Hospital sewage tanks containing a large amount of residual antibiotics  may change the composition of tank flora and further promote the development of AMR by high selective pressure on bacteria [20, 97, 98]. Antibiotics exert selective pressure in favor of the growth resistant bacteria, even at very low concentrations [99-101]. Moreover, there are some reports that concentration of antibiotics in hospital sewage are often higher than the reported no-effect or minimum concentrations for resistance selection, suggesting that the selection of ARB can occur in this environment [97, 102]. We presume that selective pressure of antibiotics exists in our hospital sewage tanks; however, this will be verified in future studies.
Horizontal gene transfer is a cause of great concern as it is one of the most important mechanisms of the spread of antibiotic resistance in the environment [103, 104]. It is suggested that nutrient-rich environments such as wastewater offer optimal conditions for horizontal gene transfer, involving the transconjugation of AMR plasmids encoding ARGs  and transformation of non-pathogenic bacteria into reservoirs of ARGs. It is known that a microbial gut flora composed of a spectrum of bacteria functions as a reservoir for ARGs and horizontal plasmid transfer between bacteria is common . This is plausible as sewage tanks consists of an accumulation of excrement and acquisition of resistance may occur frequently.
In Sweden, carbapenemase-producing Klebsiella was also detected from clinical specimens, in Svartån River and the rivers downstream of it . In Japan, KPC-2 producing Klebsiella [16, 17] and NDM-5-coproducing Escherichia coli  were detected from the effluent of urban wastewater treatment plants. Effective actions should be taken including advanced wastewater treatment processes such as ozone and UV treatment[47, 106, 107], and ultrafiltration  to accelerate the removal of ARB in WWTP. However, even the above methods do not ensure a complete removal of resistant bacteria, therefore, treatment processes may be introduced prior to the release of hospital sewage into the main sewage to reduce ARB and residual drug concentrations. In Japan, there are a few reports of contamination of sewage with ARB and presence of ARGs in microbial flora of hospital sewage tanks [109, 110]. Nevertheless, this study is the first comprehensive description of AMR in a hospital setting using metagenomic and whole genome analysis.