Antimicrobial resistome and mobilome in the urban river affected by combined sewer overflows and wastewater treatment effluent

Antimicrobial resistance (AMR) is among the top-10 global public health threats and strains the modern healthcare system and global economy (Thangaraju & Venkatesan 2019). Approximately 214,000 newborn babies died in 2015 due to pathogenic infections with AMR (Balasegaram 2021). Prolonged illness due to failing last-resort antimicrobials and lack of new antimicrobials could cause millions of deaths in the coming decades (de Kraker et al. 2016; Holmes et al. 2016; Aarestrup & Woolhouse 2020). AMR, an emerging public health hazard, is projected to cause up to 10 million deaths by 2050 if no effective regulations are adopted (O'Neill 2016; Luong et al. 2020; Pascucci et al. 2021). Aquatic environments are major reservoirs of antimicrobial resistant bacteria (ARB) and antimicrobial resistance genes (ARGs) (Hatosy & Martiny 2015; Cuadrat et al. 2020). Recently, the World Health Organization (WHO) has emphasized the necessity of controlling AMR in environmental waters for better sanitation and safe water (WHO 2020). ARGs have been broadly identified in different environments, such as wastewater (Jang et al. 2018; Raza et al. 2021), drinking water (Mattioli et al. 2017; Su et al. 2018), livestock farms and aquaculture (He et al. 2014; Jo et al. 2021), groundwater, surface water reservoirs (Lu et al. 2015; Zainab et al. 2020), and pristine environments (Van Goethem et al. 2018).

Municipal wastewater treatment plants (WWTPs) are important for preventing the spread of AMR from anthropogenic sources into the aquatic environment. Urban wastewater contains diverse ARB and ARGs, with some remaining in municipal wastewater treatment effluent (Pazda et al. 2019; Honda et al. 2020a; Kumar et al. 2020). The ARGs persisting after treatment are eventually released into receiving waters that are used as a drinking water source or for recreational purposes (Guo et al. 2020). Cell stresses, including subinhibitory concentrations of antibiotics in a WWTP, reportedly exert selection pressure on microbial communities in wastewater and activated sludge (Chait et al. 2016; Singer et al. 2016; Sulfikar et al. 2018). Furthermore, mobile genetic elements (MGEs) are engaged in horizontal gene transfer (HGT) of ARGs among microbial communities (Gupta et al. 2018). Although WWTPs decrease the absolute abundance of ARGs to a certain extent (Ben et al. 2017; Hultman et al. 2018), WWTP effluent is one of the major sources of AMR in various aquatic environments, including urban rivers (Osińska et al. 2020; Reddy et al. 2022). Urban stormwater is also an AMR source, which greatly affects aquatic environments. Untreated sewage is discharged as combined sewer overflows (CSOs) into nearby surface water bodies when the quantity of combined sewage exceeds the capacity of sewers and WWTPs due to heavy precipitation. Reportedly, 20–25% of urban sewage is discharged annually as untreated CSOs (USEPA 2004; Honda et al. 2020b). Because CSOs contain a much heavier ARB concentration than WWTP effluent, their annual ARB loading was estimated as 3.7-log larger than that of WWTP effluent (Honda et al. 2020b). Among aquatic environments, rivers have a high capacity for ARG transportation especially following heavy precipitation (Xu et al. 2015; Qiao et al. 2018; Singh et al. 2019; Zhang et al. 2021). Such weather conditions not only result in high AMR loading by CSOs in urban rivers but also cause transportation of the AMR farther downstream.

Untreated wastewater and WWTP effluent have different ARG and MGE profiles, as revealed through recent metagenomic approaches (Guo et al. 2017; Honda et al. 2023). When urban rivers are subjected to WWTP effluents and CSOs, the antimicrobial resistome and mobilome in the river water change dynamically depending on weather conditions. Specifically, the change in the MGE composition in CSOs and WWTP effluents may lead to differences in the likelihood of HGT in the urban river environment. However, the impacts of CSOs and WWTP effluents on the antimicrobial resistome and mobilome in urban rivers have not been sufficiently investigated. This study aimed to elucidate the dynamics of the antimicrobial resistome and mobilome in an urban river associated with CSOs and WWTP effluents. The composition of ARGs and MGEs in the river water was determined through shotgun metagenomic sequencing and compared under dry weather conditions, where WWTP effluents was dominant, and rainy weather during the CSO events. Impacts of CSOs on the microbial community and their correlation with ARGs were also clarified. The findings of this study are expected to contribute to efficiently monitoring and controlling AMR and HGT potential in urban rivers.

For the river water and WWTP effluent, 10 L of the sample was collected and first concentrated using a hollow-fiber ultrafiltration (UF) membrane unit (APS-25SA, Asahi Kasei) (Wang et al. 2022). Briefly, the membrane unit was pretreated by circulating 200 mL of 5% fetal bovine serum solution prior to sample filtration. The microbe concentrate retained in the membrane unit was recovered using an elution buffer containing 0.1% (v/v) Tween 80, 100 mg/L sodium polyphosphate, and 10 ppm (v/v) antifoam A. Since the river water at the CSO event was too turbid to filtrate the whole 10-L sample, 5 L of NTR-1, 1.5 L of NTR-2, 750 mL of NTR-3, and 3 L of NTR-4 site were concentrated through the UF unit to the final volume of 135–150 mL. The concentrate was centrifuged at 10,000 × g for 30 min for further concentration. DNA was extracted from the centrifuged pellets using a FastDNA Spin Kit for Soil (MPBiomedicals, USA). For the WWTPs influents, 50 mL of the sample was centrifuged at 10,000 ×g for 30 min, and the pellet was then subjected to DNA extraction via the aforementioned method. The DNA concentration of each extract was measured using a spectrophotometer (μCuvete Biophotometer, Eppendrof, Hamburg, Germany).

The microbial community was analyzed by targeting the V3-V4 hypervariable region of the 16S rRNA gene. The sequence library was prepared through a two-step tailed polymerase chain reaction (PCR) using ExTaq HS (Takara-bio, Japan). The primer sequences and PCR amplification conditions are listed in Supplementary Table S2. The prepared library was purified using a QuantiFluor dsDNA kit with a Synergy H1 system (Bio Tek), and library quality was checked using a Fragment Analyzer with a dsDNA 915 Reagent kit (Agilent Technologies, USA). The paired-end reads (2 × 300 bp) were acquired using the MiSeq System (Illumina, USA) with MiSeq Reagent Kit v3 (Illumina, USA). The raw reads were trimmed and filtered using Fastx toolkit version 0.0.14 and sickle version 1.33. The clean reads were merged using FLASH version 1.2.11 to obtain the paired-end library. Taxonomy analysis was conducted by Quantitative Insights into Microbial Ecology 2 (QIIME2) version 2021.4. After chimeric sequences were removed using a dada2 plugin, the library was classified into operational taxonomic units (OTUs) under the Greengene reference library (ver. 13_8) with 97% identity. The sequence data were deposited in the DRA database of DNA Data Bank of Japan (DDBJ) as accession numbers of SAMD00561001-00561012 (Supplementary Table S3).

The microbial population in a sample was quantified as the concentration of the 16S rRNA gene using a real-time PCR assay with a universal primer set of 341f (5′-CCTACGGGNGGCWGCAG-3′) and 805R (5′-GACTACHVGGGTATCTAATCC-3′) (Herlemann et al. 2011). For the qPCR assay, a 20-μL PCR reaction mixture of each sample was prepared using Brilliant III Ultra-Fast SYBR Green qPCR Master Mix Kit (Agilent Technologies, USA). Amplification was achieved using a real-time PCR system (Mx3000P, Agilent Technologies, USA) under the following PCR conditions: initial denaturation for 3 min at 95 °C followed by 40 cycles of denaturation for 30 s at 95 °C, annealing for 45 s at 55 °C, and extension for 1 min at 72 °C.

Shotgun metagenomic sequencing was performed using a DNBSEQ-G400RS High-throughput Sequencing System. After a DNA extract was fragmented using MGIEasy FS DNA Library Prep Set according to the instruction manual, the library was prepared using an MGIEasy DNA Adapter-96 (Plate) Kit and a dsDNA HS Assay Kit (Thermo Fisher Scientific). After the library was validated using a Fragment Analyzer with a dsDNA 915 Reagent Kit (Advanced Analytical Technologies), circular DNA from the library was created using an MGIEasy Circularization Kit. DNA nanoballs (DNBs) were prepared using the DNBSEQ-G400RS High-throughput Sequencing Set. The paired-end sequence of 2 × 200 bp was acquired from the DNBs using DNBSEQ-G400. The sequences were filtered and trimmed using the Enveomics collection pipeline (Rodriguez-R & Konstantinidis 2016). After quality trimming, each end of the fragment sequence was directly subjected (without matching paired ends) to ARG identification using BLASTn with an E-value cutoff of 1 × 10⁻⁵ against the Comprehensive Antibiotic Resistance Database (CARD) ver. 3.0.7 (https://card.mcmaster.ca/) (Jia et al. 2016). The BLAST output was aggregated to the raw ARG profile, which is the list of ARGs with a read count in each sample. To remove the bias of gene length on read counts, the read count of each ARG was normalized as reads per kilobase (RPK) based on the subject sequence length (slen) in the CARD database. The ARG composition of each sample was prepared as the relative proportion of each ARG, which was calculated as the RPK ratio of each ARG to the sum of RPK of all ARGs in the sample. The RPK counts of gene variants, which are indicated with suffix of ‘-’ & number in the CARD database (e.g. OXA-3, OXA-5, etc.) were aggregated into RPK counts of the main gene (e.g. OXA). The read count of the 16S rRNA gene was recorded using Parallel-META 3 (Jing et al. 2017). The read count of the 16S rRNA gene was also normalized as RPK according to the 1,541 bp of 16S rRNA gene length of E. coli (Brosius et al. 1978). The total ARG abundance of each sample was calculated as the ratio of the sum of RPK of all ARGs to the RPK of the 16S rRNA gene. The shotgun metagenomic sequence data were deposited in DRA database of DDBJ as accession numbers SAMD00561434-00561443 (Supplementary Table S3).

For the mobile genetic elements (MGEs), merged pair-end reads were identified against two databases, ACLAME (ver. 0.4) and MGE, using BLASTn with an E-value cutoff of 1 × 10⁻¹⁰. The ACLAME database contains different gene types, including plasmid, transposons, and prophage (http://aclame.ulb.ac.be/), whereas the MGE database comprises 2,706 non-redundant sequences belonging to more than 270 gene types, including transposase, plasmid, integrase, insertion elements, and qacEdelta (Pärnänen et al. 2018) (https://github.com/KatariinaParnanen/MobileGeneticElementDatabase). The BLAST output of each sample was aggregated in the same manner as the ARG profile in the above section. The read counts of the MGEs were normalized as RPK based on the subject sequence length (slen) of the MGE in the databases. The MGE composition of each sample was prepared as the relative proportion of each MGE to the sum of RPK of all MGEs in a sample. The total abundance of MGE in each sample was calculated as the ratio of the sum of RPK of all MGEs to the RPK of the 16S rRNA gene.

The antimicrobial resistome and microbial community of each sample were compared via principal component analysis (PCA) using R version 4.0.0. The PCA for the ARG composition was performed using a relative proportion of ARGs with scaling to reflect the relative changes in each component. The PCA for the microbial community was performed using a relative abundance of genus-level OTUs with scaling. Pearson's correlation analysis was performed between the relative proportion of ARGs and relative abundance of class-level OTUs.

The antimicrobial resistome, which is defined as the relative abundance of ARG composition, in the river water during dry weather and the CSO event was compared through hierarchic cluster analysis and PCA. According to the hierarchic cluster analysis, ARG compositions in the river water during dry weather and the CSO event were distinct (Figure 3(b)). River water in the dry weather and WWTP effluent clustered closely, suggesting that the WWTP effluent affected the river under dry weather. In contrast, the river water during the CSO event exhibited a similar ARG composition as the WWTP influent, revealing the contribution of untreated wastewater. Moreover, the ARG composition in the river water during the CSO event was more uniform from the upstream (NTR-1) to the downstream (NTR-4) than during dry weather. PCA revealed the variation in ARGs during dry weather and the CSO event in the urban river.

According to PCA scores, PC1 distinguished the river water during the CSO event and WWTP influent from the river water in dry weather and WWTP effluent (Figure 3(c)). The river water under dry weather and WWTP effluents were characterized by negative PC1 scores. Among the top-30 ARGs with highly negative PC1 loading, 22 exhibited multidrug antimicrobial resistance to more than three drug classes, including clinically important drugs such as macrolide, fluoroquinolone, and aminoglycoside. Most of the multiple ARGs belonged to antibiotic efflux, which included mex and mux genes and target protection, which included tlrC, oleB, and carA (Supplementary Table S4). Therefore, these multidrug ARGs were relatively abundant in the river water during the dry weather. Meanwhile, the river water during the CSO event and the WWTP influents were characterized by positive PC1 scores, mostly corresponding to ARGs with single resistance to fluoroquinolone (qnrS and qnrVC), aminoglycoside (mdtN, marA, and baeR), peptide (mcr genes), and β-lactams including cephalosporin, and cephamycin (VEB, cepA, FOX, and CfxA) (Supplementary Table S5). These antimicrobials are popularly used for clinical purposes in Japan (AMR Clinical Reference Center 2021) and are reported to be abundant in WWTP influent (Honda et al. 2023). Among the top-30 ARGs with highly positive PC1 loadings, 8 were characterized by antibiotic inactivation and 17 ARGs to efflux (Supplementary Table S5). This finding is consistent with a previous study, which reported that antibiotic efflux and antibiotic inactivation were the dominant resistant mechanisms during CSO events (Zhang et al. 2022a).

In contrast, microbial group C (shown in Figure 5(a)) had high correlations with ARGs to cephalosporin, macrolide, quinolone, and tetracycline. Group C mostly contained anaerobic and commensal bacteria, which are often abundant in the gut microbiome, including classes of Bacteroidia and Flavobacteriia of Bacteroidetes; Bacilli, Clostridia, and Erysipelotrichi classes in Firmicutes; and Epsilonproteobacteria class in Proteobacteria. Interestingly, microbial group C had an opposite correlation tendency to ARGs from groups A and B. Group C exhibited a high correlation with most ARGs to macrolide (mefB, ermF, ermG, erm(35), and erm(49)), quinolone (qnrS and qnrVC), β-lactams (ACI, AER, CblA, CepA, CfxA, FOX, and MOX), tetracycline (tetM, tetO, tetS, tetQ, tetT, tet36, and tet44), and aminoglycoside (aadA4, aadS, aad(6), and ant) (Figure 5(a)). Hence, group C classes were abundant during the CSO event and possibly harbored ARGs to β-lactams, macrolide, quinolone, tetracycline, and some of aminoglycoside. A previous study also reported a higher correlation of ARGs to macrolide, quinolone, and tetracycline with commensal bacteria and anaerobic bacteria in WWTP influent (Honda et al. 2023). Furthermore, the abundance of bacteria from sewage and stormwater reportedly exhibited a strong positive correlation with abundance of ARGs to fluoroquinolone (Carney et al. 2019). Because CSO abundantly contains potential pathogens of fecal origin, these results demonstrated that the CSO event poses a risk of spreading such pathogens that are resistant to clinically important drugs with the highest priority into the aquatic environment.

In the target river, microbial groups A and B, which inhabit soil and activated sludge, were abundant in the river water during dry weather and in the WWTP effluents, whereas microbial group C, which inhabit the gut, were abundant in the river water during the CSO event and the WWTP influents (Supplementary Figure S2(a)). The high abundance of anaerobic bacteria that are abundant in gut microbiome suggests that the main origin of microbes in the target river was CSO including untreated wastewater. The target river also received surface runoff water; however, the impact of surface runoff water on antimicrobial resistome in the target river was likely limited compared to the CSO. The correlation pattern between abundant ARGs and microbial groups implied that a change in bacterial community could be one of the drivers of changes in the antimicrobial resistome in the targeted river. During the CSO event, the relative abundance of phylum Firmicutes and Bacteroidetes increased from dry weather, whereas Proteobacteria abundance decreased (Figure 5(b)). Specifically, Firmicutes was found only in the river water during the CSO event and in the WWTP influent. The abundant Firmcutes classes of Bacilli, Clostridia, and Erysipelotrichi; Tenericutes class of Mollicutes; and Lentisphaerae class of Lentisphaeria belonged to microbial group C, which had high correlations with ARGs to β-lactams, macrolide, quinolones, and some tetracycline and aminoglycoside. It has been reported that Firmicutes tends to increase by 10–35% during CSO events in a recreational beach (Jang et al. 2021). These results suggest that changes in the antibiotic resistome in the river are directly linked with the microbial community under the effects of dry weather or a CSO event. A genus-level comparison via PCA indicated a similar microbial community in the river water during the CSO event as in the WWTP influent, which were mostly characterized by genera belonging to Bacilli, Clostridia, and Erysipelotrichi classes (Supplementary Figure S2(b) and Table S6). Importantly, potentially human pathogenic Streptococcus and Clostridium were detected abundantly in the river water during the CSO event (Supplementary Table S6). Streptococcus and Clostridium could be host to clinically important ARGs because they exhibited a high correlation with ARGs to macrolide, quinolone, and β-lactams. A previous study reported that the increased abundance of sewage-related ARGs and microorganisms during the CSO event caused an increase in the diversity and abundance of ARGs at a recreational beach (Jang et al. 2021). A higher abundance of Proteobacteria (56–72%) and Cyanobacteria (3.0–14%) was observed during dry weather than the CSO event. A genus-level comparison via PCA indicated a similar microbial community in the river water in dry weather as in the WWTP effluent, which was mostly composed of aerobic bacteria of Alphaproteobacteria, Betaproteobacteria, Planctomycetes, Rhizobiales, and Sphingomonadales classes (Supplementary Figure S2(b) and Table S7). Most of these classes belonged to microbial groups A and B, which had high correlations with ARGs to multidrug by efflux, sulfonamide, and some β-lactams and aminoglycoside. These results suggested that the antimicrobial resistome of the target river in dry weather was mainly associated with bacteria in the WWTP effluent, which hosted ARGs to multidrug and sulfonamide. Importantly, AMR to aminoglycoside was abundantly present under both dry weather and the CSO event, even though it originated from different ARG species hosted by different microbial groups. Consequently, a specific link between the microbial community and ARGs observed in the target river was present, which may reflect the reserve conditions in the source of AMR in the target river. However, it was unable to distinguish intrinsic ARGs from mobile ones in the present study. It is also important to know whether the ARG is on chromosome or on MGEs in order to understand the potential of HGT in the target river.

Referring to the MGE database, transposase exhibited the highest proportion both during dry weather (78–82%) and the CSO event (85–88%) (Figure 6(c)). Transposases are involved in transposition, by which ARGs can transfer from chromosome to plasmids and vice versa (Partridge et al. 2018). The high abundance of transposase suggested that transposition is quite common in river water both during dry weather and at the CSO events. On the other hand, the insertion sequence (IS) was more abundant during dry weather and the WWTP effluents (12–19%) than during the CSO event and the WWTP influents (8.2–8.9%). Previous studies also demonstrated a high relative abundance of transposase (more than 50%) followed by IS and integrase in river water (Song et al. 2022; Zhang et al. 2022b). Transposase is associated with the intracellular mobility of ARGs, whereas integrase is associated with intercellular mobility of ARGs. Meanwhile, transposase may have indirect association with HGT because transposition of ARGs into plasmids possibly enhances intercellular mobility by HGT. Transposase has been observed to mediate ARG mobility in aquatic environments (Knapp et al. 2008). Raza et al. (2021) reported that tnpA and IS91 were major MGEs carrying ARGs in WWTP effluent. At the MGE-gene level, tnpA, IS91, and intI1 were found to be dominant (Figure 6(d)). The most abundant MGE was tnpA followed by IS91 and intI1. Specifically, tnpA and IS91 were remarkably high during both dry weather and the CSO event. These MGEs were found along ARGs in various wastewaters and in the river during the CSO event (Zhang et al. 2020, 2022a; Zhou et al. 2020).

Overall, the higher prophage and transposase abundance during the CSO event suggested that CSO had a greater impact on the transduction potential in the river microbiome than dry weather. Phages are reportedly the potential carriers of ARGs in an aquatic environment (Jebri et al. 2021). Hence, the dissemination of ARGs by HGT could be further enhanced after the CSO event because of the increased transduction potential. Meanwhile, IS and plasmids were more abundant in dry weather, mainly due to the effect of WWTP effluents. IS is the smallest MGE that can move within a genome or horizontally as part of plasmids or phages (Siguier et al. 2006; Vandecraen et al. 2017). IS is not only involved in the transfer of ARGs but also modulates the expression of ARGs by providing an active promoter (Vrancianu et al. 2020). Reportedly, IS can increase the expression of efflux pumps and the acquisition of multidrug resistance (Olliver et al. 2005). Zhang et al. (2022b) reported that IS and plasmid were the primary MGEs shaping the ARG profile, and that IS was found to be associated with the transfer of multidrug resistance in wastewater treatment processes. In the present study, multidrug ARGs exhibited a higher abundance during dry weather than the CSO event, where IS and plasmid were also found at high abundance. Therefore, IS may also play an important role in acquiring multidrug resistance in the river microbiome.

The major source of the target urban river during dry weather was effluents from the two WWTPs. Hence, the antimicrobial resistome, mobilome, and microbial community during dry weather were highly affected by those in the WWTP effluents. During dry weather, the WWTP effluent was the source of ARGs to multidrug and sulfonamide, which were likely hosted by aerobic bacteria inhabiting the soil, sediment, and activated sludge. However, the antimicrobial resistome and mobilome differed slightly during dry weather among the sampling sites from the upstream to the downstream, suggesting the effect of the nearby water source. For example, ARG abundance at NTR-4 was much lower than that at NTR-2 and NTR-3. ARG abundance was not significantly different between NTR-2 and NTR-3, which were both located downstream of WWTP I. However, ARG abundance at NTR-4 decreased by 67% after the river received WWTP II effluent. A possible reason for the reduction in ARG abundance was dilution with the WWTP II effluent. Dilution alone is expected to reduce the ARG abundance at NTR-4 by only 44%. Hence, another factor contributed to the ARG reduction at this location. Presumably, the ARGs were degraded by residual chlorine in the WWTP II effluent, which was disinfected through chlorination before discharge. Hence, ARGs from upstream were also degraded at NTR-4. Meanwhile, WWTP I effluent was disinfected through ozonation; therefore, no significant reduction was observed at NTR-2 and NTR-3. MGE abundance was also likely reduced at NTR-4 for the same reason (Figure 6(a)).

During the CSO event, the ARG composition was rather uniform compared to that during dry weather (Figure 3(c)). However, ARG abundance increased gradually from upstream to downstream. The antimicrobial resistome, mobilome, and microbial community at the CSO event were highly affected by the untreated wastewater, which was expected to be similar to the WWTP influents. Interestingly, ARG abundances during the CSO event at downstream sites NTR-2 and NTR-4 were even higher than those of the WWTP influents. This could be due to the dispersion of ARGs from sediments that were whirled up by the high flow of the CSO. Zhang et al. (2022a) also reported that sediments flushed out with CSOs had a significant effect on the antibiotic resistome. Most of the ARGs detected in the river water during CSO were common with those in the WWTP influents. However, 28 ARGs in the river water during the CSO event were not common with those from other samples (Figure 2(b)). These ARGs, including resistance to macrolide, β-lactam, and aminoglycoside, may have originated from the whirled sediments in the sewer or river. In this study, the river water during the CSO event was found to mainly contain ARGs to clinically important drugs, including macrolide, quinolone, and β-lactams, which were likely to be hosted by anaerobic bacteria in the gut microbiome. Jang et al. (2021) reported that CSO may bring an increased concentration of sewage-related ARGs to a recreational beach. More importantly, the river water at the CSO event was found to have higher transduction potential because of the high prophage abundance. During the CSO event, a high load of the microbes with a higher ARG abundance and high transduction potential were transported farther than that during dry weather due to the high flow rate of the river. A part of the transported microbes originating from the CSO would later settle in river sediments. Consequently, ARGs from the CSO are expected to widely spread and be reserved in the river sediment, where ARGs may be further transferred to other microbes in the sediment.

Developing an AMR monitoring framework for various aquatic environments is essential to mitigating the spread of AMR in the environmental dimensions. Liguori et al. (2022) proposed several genetical indicators for monitoring AMR in various water environments: intI1, blaCTX-M, bla(NDM-1), bla(OXA), bla(TEM), bla(kpc), bla(vim), bla(shv), sul1, tet(A), tetW, tetM, vanA, mecA, ermB, qnrS, and aac(6)-ib. Among them, intI1 or a combination of sul1 and tetM were the most applicable to evaluating the ARG abundance in the target river. In the present study, sul1 exhibited a higher correlation with ARGs on multidrug and sulfonamide (Supplementary Figure S3). In contrast, tetM exhibited an inverse correlation pattern to sul1, and highly correlated with ARGs on quinolone, β-lactams, aminoglycoside, tetracycline, and macrolide (Supplementary Figure S3). This indicated that a combination of sul1 with tetM could represent the distribution of ARGs to most drug classes in the urban river. In the present study, the abundance of the tetM gene was significantly high during the CSO event, where β-lactams, quinolone, and macrolide drug resistance were abundant. Meanwhile, the abundance of sul1 in dry weather was significantly high, during which multidrug and sulfonamide drug resistance was abundant. tetW exhibited a similar correlation pattern with drug classes as tetM; however, its abundance was much lower than that of tetM. Hence, a combination of sul1 and tetM gene can not only represent the abundance of most drug classes but can also distinguish AMR from different sources. Among β-lactam genes, OXA and CTX-M were observed to exhibit moderate correlations with most of bacterial classes in groups A and C (Figure 5(a)). However, TEM represented an abundance of some specific classes in group C. The ermB and qnrS genes also tended to represent the abundance of macrolide and fluoroquinolone resistance in some specific bacterial classes. These genes may not be applicable to monitoring AMR when the target water body contains AMR from multiple sources. Among the AMR indicators suggested by Liguori et al. (2022), vanA, NDM, bla(kpc), bla(vim), bla(shv), and mecA were not detected at all in the present study.

The antimicrobial resistome and the mobilome in an urban river was elucidated during dry weather and a CSO event via shotgun metagenomic sequencing. In the target river, antimicrobial resistome during dry weather was mainly composed of ARGs to multidrug by efflux and to sulfonamide, which were likely hosted by aerobic bacteria originating from the soil, sediment, and activated sludge. During the CSO event, the antimicrobial resistome in the river water shifted toward a higher abundance of ARGs to clinically important drug classes, including macrolide, fluoroquinolone, and β-lactams, which were probably hosted by the anaerobic bacteria abundant in untreated wastewater. Regarding mobilome, plasmid and transposase were the most abundant MGEs during both dry weather and the CSO event. However, an increase in prophage and transposase during the CSO event suggested an enhancement in HGT via transduction, which could have caused the further spread of ARGs in the river microbiome after the CSO event. The antimicrobial resistome during the CSO event was more uniform from upstream to downstream than during dry weather. Consequently, ARB to clinically important drugs with a high potential for transduction could be transported farther downstream during the CSO event. As monitoring indicators of AMR in the target river, intI1 and tnpA genes showed high correlation with the total ARG abundance. Moreover, a combination of sul1 and tetM genes was useful for representing the abundance of ARG originating from different sources of bacteria. Although only a single CSO event was observed in the present study, these results demonstrated that CSO is an important source of AMR and MGEs in an urban river. A reduction in CSO is necessary to mitigate the spread of AMR in an urban river and downstream aquatic environment.

This study was supported by JSPS KAKENHI fund (Grant Nos. 19H02272, 21KK0073, 21H03617, 18KK0114), JST MIRAI Program (Grant No. JPMJMI18DC), and ‘Asia-Pacific Researcher Network on Environmental Dimensions of Antimicrobial Resistance (END-AMR-Asia)’ funded by Kurita Water and Environmental Foundation (Grant No. 22T007). Additional support was provided by MEXT scholarship for Doctorate Program.

M.A.S.: Conceptualization, Methodology, Formal analysis, Investigation, Data Curation, Writing – Original Draft; T.V.H., Y.S., H.W., B.Z.: Investigation, Data Curation; N.M.: Methodology, Software, Resources, Writing – Review & Editing; T.W.: Supervision, Funding acquisition; M.I., H.T.: Supervision, Resources, Funding acquisition, Writing – Review & Editing; R.H.: Conceptualization, Methodology, Supervision, Project administration, Funding acquisition, Writing – Review & Editing.

All sequence data are available from the DRA database of DNA Data Bank of Japan (DDBJ). The accession numbers of the sequence data are listed in the Methods and Supplementary Information.

The authors declare there is no conflict.