Metagenomic analysis reveals differential effects of sewage treatment on the microbiome and antibiotic resistome in Bengaluru, India

Antimicrobial resistance (AMR) is a complex global health challenge involving the transfer of bacteria and genes between humans, animals, and the environment. The resistance is mediated by antibiotic resistance genes (ARGs), also known as ‘the resistome’, which circulate among microbiomes representing different sectors of the One Health Concept. Antibiotics are the main preventive agents for bacterial infectious diseases in human and veterinary medicine due to their high efficacy and safety. However, the extent and scale of AMR are largely shaped by selection pressure resulting from antibiotic usage. The environment plays a significant role in the ecology, evolution, and transmission of AMR among pathogenic and non-pathogenic bacteria, and transmission occurs in multiple directions. For instance, AMR developed in animal reservoirs can spill over to humans through food, water, or other environmental routes, or through reverse zoonotic transmission of AMR (Larsson & Flach 2022). Therefore, it is necessary to develop an understanding of the major pathways and drivers of AMR in the environment.

Environmental (sewage) surveillance (ES) is an important part of the One Health approach. Urban sewage, particularly in low-resource settings, is a major source for the dissemination of ARGs in various environments (Schlüter et al. 2007; Bouki et al. 2013; Gatica & Cytryn 2013; Rizzo et al. 2013). ES is an effective tool for collecting reliable data to assess the spatiotemporal patterns of ARGs diversity at the community level. Many antibiotics are currently used to treat infections, as well as ARGs acquired by human pathogens originating from the environment (Martínez 2008). Large quantities of antibiotics are discharged into municipal wastewater through incomplete metabolism in humans or the disposal of unused antibiotics (Nagulapally et al. 2009). Sewage treatment plants (STPs), like gut microbiomes, provide environments with abundant ecological opportunities and niche availability, which promote antibiotic persistence and adaptation among community members. Eukaryotes such as protozoa, fungi, parasites, and algae play vital roles in enhancing the efficiency of STP processes (Siles & Michán 2020). STPs are critical environments where high microbial density and diverse ARGs are exposed to selective agents such as antibiotics, disinfectants, and heavy metals. STPs act as melting pots where selective agents facilitate horizontal gene transfer (HGT) through mobile genetic elements (MGEs), resulting in the emergence of antibiotic-resistant bacteria (Gillings & Stokes 2012; von Wintersdorff et al. 2016). Wastewater treatment typically involves various mechanical, biological, physical, and chemical processes that can potentially impact the fate of antibiotics, antibiotic-resistant bacteria, and ARGs as well as the spread of resistance in the environment, thereby increasing the risk of transmission back to humans (Rizzo et al. 2013).

Climate change and health are closely linked to urban wastewater. Water security is a pressing issue in India. Water scarcity and reduced availability of agricultural water have spurred increased interest in the use of recycled irrigation water. Currently, only 28% of wastewater in India is treated and reused (Bassi et al. 2023). Wastewater is a valuable source of nutrients (nitrogen, phosphorus, potassium). Therefore, reusing treated wastewater for irrigation can have economic and environmental benefits. However, there is limited tracking of treated wastewater parameters such as pH (5 − 5.9), total suspended solids (TSS: 20 mg/l), nitrogen (10 mg/l), phosphorus (1 mg/l), biological oxygen demand (BOD: 10 mg/l), chemical oxygen demand (COD: 50 mg/l), and fecal coliform levels (less than 100 MPN/100 ml) and regular monitoring usually does not include a complete profiling of pathogens (NJS 2017). It is crucial to identify and address any water quality issues associated with the use of recycled water including the possible presence of bacterial, viral, and protozoan pathogens. Furthermore, there is a concern that the continuous addition of antibiotics through treated wastewater for irrigation poses a significant risk to soil and plant quality and health. This includes antibiotic uptake and accumulation by crops as well as the potential selection of antibiotic resistance from consuming contaminated vegetables (LaPara et al. 2011). Even low concentrations of antibiotics in plants could promote the selection of ARGs in bacterial ecosystems and contribute to antibiotic resistance in the environment (Azanu et al. 2016).

In India, obtaining representative data on AMR for healthy human and animal populations is challenging. Until 2010, India was the largest consumer of antibiotics (Gelband et al. 2015), with antibiotics easily accessible without a prescription (Laxminarayan & Chaudhury 2016). The rise in AMR among bacterial pathogens is primarily attributed to the acquisition of ARGs or the accumulation of spontaneous mutations in the bacterial genome. Six bacterial pathogens, namely Escherichia coli, Staphylococcus aureus, Klebsiella pneumoniae, Streptococcus pneumoniae, Acinetobacter baumannii, and Pseudomonas aeruginosa (ESKAPE), are the leading cause of death worldwide.

In the Indian context, studies have demonstrated the presence of diverse ARGs in wastewater treatment plants, river basins, or estuaries. These include ARGs coding for resistance against tetracyclines, beta-lactams, aminoglycosides, and macrolides. For example, Saxena et al.’s (2020) study on emerging contaminants and antibiotic resistance in STPs, focused on the variability in ARG profiles influenced by treatment processes and environmental factors. Hazra & Durso (2022) evaluated the performance efficiency of conventional treatment plants and constructed wetlands toward reducing antibiotic resistance, highlighting different ARG removal efficiencies. Similarly, Samson et al. (2023) explored the spatiotemporal variation of the microbiome and resistome repertoire along the Ganges River, providing detailed analyses of ARG distribution patterns. However, these studies have been cross-sectional with limited spatiotemporal resolution and lack associated data on microbial communities. To address this gap, we designed a comprehensive longitudinal study in the urban sewage network of Bengaluru city to understand the overall microbiological water quality and the role of sewage digestion processes, which potentially affect the prevalence and abundance of AMR genes in the environment.

Bengaluru (12.9716° N, 77.5946° E, Karnataka, India) is the third largest city (∼11 million inhabitants) in India with an efficient sewage network that processes ∼1,142.5 million liters per day (m³/d) of wastewater. Each STP follows a water treatment technology depending on the quality of raw sewage to make the treated water reusable (https://bwssb.karnataka.gov.in/info-1/About+BWSSB/en). In Bengaluru, the predominant sewage treatment method is the activated sludge process (ASP). This process utilizes aeration along with microbial flocs or granules to effectively remove carbon (C), nitrogen (N), phosphorus (P), various micropollutants (such as toxins, pesticides, hormones, and pharmaceuticals), as well as pathogens (van Loosdrecht & Brdjanovic 2014), biochemical oxygen demand (BOD), fecal coliforms, and turbidity. Removal of these components is essential for maintaining the environmental and safety standards of treated wastewater. There are two other variations of ASP used: extended aeration, sequential batch reactor (SBR) and upflow anaerobic sludge blanket reactor (UASB). Extended aeration is better suited for treating low sewage loads and is often used in apartment complexes and residential areas. SBR, on the other hand, can remove nutrients like nitrogen from sewage, allowing treated water to be directly released into lakes. Bengaluru has the largest water footprint compared with any other Indian city and is the second-largest city after Mexico where treated wastewater is used not only for agriculture but also for recharging groundwater in drought-prone regions outside the city. For example, in the Kolar district of Karnataka, one of the worst-affected districts in terms of drought and climate change, treated wastewater pumped from Bengaluru has helped improve groundwater quality and ensure the availability of water regardless of weather and climate patterns. While it is essential to minimize the public health risks before using treated wastewater, it is important to understand the effectiveness of treatment plants in removing harmful parasites and whether there is spatial and temporal segregation in AMR and related microbial diversity. To explore this, we used amplicon-based sequencing of the 16S ribosomal RNA gene (16S rRNA gene) and an AMR panel targeting 478 ARGs to characterize the spatiotemporal variations in the antibiotic resistome and corresponding bacterial community. We are specifically interested in understanding the following:

• (1) How does bacterial diversity change from the inlet to the outlet in STPs based on season and processing methods?

• (2) Is there spatial and temporal heterogeneity in ARG diversity?

• (3) Is there a shift in differential abundance across bacteria and ARGs from the inlet to the outlet of the treatment?

• (4) What type of co-occurrence networks exist in the microbial community and ARGs between the inlet and the outlet?

• (5) How stable is the microcosm in the inlet and outlet community (equalization analysis)?

Samples were thawed at 4 °C and processed in a Biosafety Level-2 laboratory. First, each sample was prefiltered using a sterile muslin cloth to remove large debris. After the debris was removed, the sample was filtered using 0.22 μm membranes (MF-Millipore^® Membrane Filter). DNA was isolated from the filter paper using the Qiagen DNeasy PowerWater DNA Isolation Kit. A negative control was included with each extraction batch. DNA samples with a nucleic acid ratio of 260/280, greater than 1.7 nm and a concentration greater than 4.5 ng/μL were considered for sequencing and stored at −20 °C.

Universal bacterial primers, the forward primer 338F and reverse primer 802R by Klindworth et al. (2013), tagged with an Illumina adaptor sequence, were used to amplify the V3–V4 hypervariable region of the 16S rDNA gene through polymerase chain reaction (PCR). The PCR reaction mix (25 μL) contained 12.5 μL of KAPA HiFi HotStart ReadyMix (Roche), 1.0 μL (5 μM) of the forward and reverse primer, 2.0 μL of template DNA (5 ng/μL), and 8.5 μL of Ambion nuclease-free water (Invitrogen). Each PCR plate included a negative control to detect potential contamination. The PCR conditions included an initial denaturation step at 95 °C for 3 min, followed by 25 cycles of denaturation at 95 °C for 30 s, annealing at 55 °C for 30 s, and a final extension at 72 °C for 30 s. Amplification success was confirmed through gel electrophoresis, using a 2% agarose gel. PCR products were purified using bead purification with Agencourt AMPure XP (Beckman Coulter, USA). Samples were normalized, then multiplexed with the Nextera XT Index kit (96 indexes) (Illumina) and sequenced on an Illumina MiSeq flowcell using a V3 sequencing chemistry kit (2 × 350) at the BLiSC (Bengaluru Life Science Cluster) NGS facility.

Raw reads were processed in R (version 3.4.3) (R Development Core Team 2008) using a modified version of the Divisive Amplicon Denoising Algorithm 2 (DADA2) pipeline (Callahan et al. 2016) and online tutorials v1.6 and workflow for big data v1.4 (benjjneb.github.io/dada2/tutorial.html). Briefly, the filterAndTrim function was used to remove the low-quality reads. Error rates were calculated for each amplicon dataset. The identical reads were dereplicated, and a sequence variant inference algorithm was applied to each dataset. Tables of ASV sequences per sample within each run were then combined, and chimera detection using all pooled samples was performed (see Supplementary Table S1 for the number of reads retained across each step). Taxonomy assignment from Kingdom to Genus was performed using the RDP classifier and MiDAS 4 (Dueholm et al. 2022) reference database formatted for DADA2 (benjjneb.github.io/dada2/training.html), using the assignTaxonomy function. All ASV tables produced by the pipelines were converted into phyloseq objects using the phyloseq package (McMurdie & Holmes 2013).

Wastewater samples were sequenced for ARGs using the AmpliSeq Illumina AMR panel (Supplementary Table S3). The PCR conditions were adapted from Illumina's AmpliSeq for Illumina On-Demand, Custom, and Community Panels reference guide. Briefly, the total reaction volume was 20 μL, consisting of 12.5 μL of KAPA 2G Fast Multiplex Mix (Roche). Two PCRs were set up simultaneously as the panel offers two pools (Pool 1: 408 amplicons, Pool 2: 407 amplicons). The primer volume for individual runs was 5.0 μL each, template DNA was 1.0 μL (5 ng/μL), and nuclease-free water (Invitrogen) was added. The ARGs PCR conditions used included an initial denaturation step at 99 °C for 2 min, followed by 20 cycles of denaturation at 99 °C for 15 s, annealing at 60 °C for 4 min, and a hold at 4 °C. A negative control was included in each PCR. Samples showing positive amplification for ARGs were considered for sequencing. Amplicon libraries for the Illumina platform were prepared using the AmpliSeq Library PLUS for Illumina kit (Cat. No. 20019102). Library quality was assessed using a 2100 Bioanalyzer with a DNA high sensitivity assay kit (Agilent CA, USA). Sequencing was carried out using an Illumina MiSeq system with a MiSeq reagent kit v2 500 cycles (2 × 150) paired-end chemistry.

To identify ARGs, we analyzed raw reads using the CLC Genomics Workbench 21.0.5 Microbial Genomics Module (CLC MGM) and the Find Resistance with ShortBRED (FRSB) tool (similar to ShortBRED) (Kaminski et al. 2015). Within the FRSB tool (CLC Microbial Genomics Module), we used DIAMOND v0.9.31 to match queried sequences against the QMI-AR Peptide Marker database (released 2019-11). We retained all default settings in the FRSB tool except for the ‘more sensitive search’ parameter, which allowed us to run DIAMOND in its highest sensitivity mode. We cross-checked the identified AMR genes against entries in the Comprehensive Antibiotic Resistance Database (CARD) (Alcock et al. 2020). It is important to note that our methodology can only detect ARGs that have been annotated in CARD. Therefore, some novel types of ARGs present in the samples may be missed since the analysis is based on a similarity search.

All analyses were conducted using R software v.3.4.3. Rarefied operational taxonomic unit (OTU) tables were generated by performing multiple rarefactions on bacterial reads using the phyloseq package. The OTU tables were rarefied to the sample with the lowest number of sequences, with a threshold of >10,000 sequences (any samples with less than 10,000 sequences were excluded from the analyses before the rarefaction step; see Supplementary Figure S1). Rarefaction curves were generated using the rarefy_even_depth function in the phyloseq package, to equalize sequencing depth among samples (see Additional File 1).

The unweighted UniFrac distances between samples were calculated using the rarefied OTU abundance tables in the phyloseq package. We assessed alpha diversity (Observed richness, Shannon diversity), conducted beta diversity analyses using principal coordinate analysis (PCoA) and performed permutational multivariate analysis of variance (PERMANOVA) using the adonis2 function. Additionally, non-parametric multidimensional scaling (NMDS) and ANOSIM tests were used to assess the differences in communities based on inlet versus outlet, season, and STP processing (Supplementary Table S2). All statistical analyses associated with sample grouping were performed using the vegan package (Oksanen 2017). Relative abundance plots were constructed using the ggplot2 package (Wickham 2016).

To identify potential interactions between ARGs and bacteria, a gene co-occurrence network was created. The correlation matrix between the ARG pairs in the inlet and outlet samples was calculated using a pairwise Spearman's correlation rank test (Spearman 1904). The values in the matrix indicated the co-occurrence correlation of the ARGs, with positive values indicating a positive correlation and negative values indicating a negative interaction. These matrix values were input into Gephi 0.10.1 (Bastian et al. 2009) to plot the gene network using built-in functions. The resulting network included interactions between ARGs, ARG and species, and species using the Force Atlas layout format. To create a clear network, only statistically significant values with a P-value < 0.01 and a correlation coefficient >0.6 for inlet samples and >0.5 for outlet samples were plotted in the constructed network for an easier interpretation (Barberán et al. 2012; Mandakovic et al. 2018; Yasir 2020). Additionally, the node size was scaled based on the abundance of ARGs in the data.

We first examined the alpha diversity measures across inlet and outlet samples by season. While the taxa diversity (observed: F = 43.56, P < 0.001; Shannon diversity: F = 38.25, P < 0.001) was higher in outlet samples than in inlet samples (Figure 3(a) and 3(b)). However, there was no change in diversity across seasons except in October to December probably due to low sample size. To determine whether the overall microbial composition of inlet and outlet samples was different, we examined β-diversity measures by season and sample types (inlet and outlet). In PCoA plots, these groups were not clustered together by season, indicating that there were no differences (PERMANOVA: R² = 0.04, P > 0.37; Figure 3(c)). Similar results were observed in NMDS plots (ANOSIM: R = 0.02, P > 0.24; Figure 3(d)).

In total, 36.9 million reads were obtained using Illumina MiSeq, which corresponds to an average of 38,4858.12 million reads per sample with an average length of 151 bp. We found 154 ARGs and 147 ARGs, respectively, in the inlet and outlet samples (Figure 1 and Supplementary Table S4). Out of these 273 types of ARGs, 54 genes were dominant with a relative abundance of more than 1% belonging to 8 dominant antibiotic classes (resistance gene types): tetracycline (teto, tetM, tet(G), tet(A), tet (M), tetx), beta-lactams (blaOXA-5, blaGES, blaLCR-1, blaOXA, blaOXA-256, blaCARB-5, blaOXA-31, blaCMY, blaVEB, blaDIM-1, blaTEM, blaADC, blaCTX-M, blaNDM, blaPER-4), MLSB (ErmF), aminoglycosides (aadA23, aadA6/aadA10, ANT(3″)-Ia, aadS, APH(3′)-Ia, ANT(6), ANT(6)-Ib, APH(3′)-IIIa, APH(6)-Id), macrolides (mphF, ereA, mphE, mel, mef(B), mphA), nucleoside antibiotic (SAT-2), diaminopyrimidine antibiotic (DHFR_1, dfrG, dfrA13), fluroquinolones (QnrVC1), MGE (lsaE, lnuC), MLSB (ErmF, ErmT, ErmB, ErmQ), and phenols (cmx, CAT). ARGs with a relative abundance of less 1% were categorized as ‘other’ (Supplementary Table S5).

Differentially abundant species analysis for bacterial communities showed a significant difference between the inlet and outlet samples (Supplementary Table S6). Among ESKAPE pathogens, the differential analysis identified Escherichia–Shigella (FC_log2 2.42), Klebsiella (FC_log2 2.20) including Enterobacter (FC_log2 2.58), Megasphaera indica (FC_log2 4.49), Catenibacterium mitsuokai (FC_log2 −2.37), Lactobacillus ruminis (FC_log2 −2.37), Megasphaera indica (FC_log2 4.49), Streptococcus henryi (FC_log2 4.59), Streptococcus parasuis (FC_log2 2.77), and among others, showed a significant increase from inlet to outlet.

In general, there were more complex and dense correlations between ARG–ARG, ARG–bacteria, and bacteria–bacteria in inlet samples. For example, ARG–ARG pairs, such as bla_TEM (Pseudomonas aeruginosa) and bla_NDM (Klebsiella pneumoniae), as well as ARG–bacteria pairs, such as CAT (Campylobacter coli) and bla_CMY-1 (Klebsiella pneumoniae), showed strong positive correlations. Additionally, there was an exclusive positive correlation between dfr and aadA6/aadA10 (Pseudomonas aeruginosa). These ARGs showed no interactions with bacterial phyla. MLSB genes like ErmQ and ErmF (Streptomyces spp.) showed a negative association with bla_OXA-31 (Pseudomonas aeruginosa), mphA (Escherichia coli), and mef(B) (Escherichia coli). Among ARG–bacterial phyla interactions, Campylobacterota and Desulfobacterota showed the highest positive networks with various ARGs. However, Firmicutes, Actinobacteriota, and Proteobacteria exhibited many negative correlations.

In contrast, outlet samples showed stronger and positive correlations between ARG–ARG pairs and bacterial assemblages, with very few interactions between ARG–bacteria pairs.

Studying the microflora of wastewater is of great importance to public health and ecology. The microbial diversity in wastewater is constantly changing and highly dynamic, influenced by various factors. Our large-scale urban sewage sampling in Bengaluru revealed that there is a shift in microbial diversity and the antibiotic resistome from the inlet to the outlet of a treatment plant. This study represents the most comprehensive longitudinal research on Bengaluru wastewater to date, significantly enhancing our understanding of the potential variations in the risk of exposure to bacterial pathogens and ARGs originating from diverse sewage processing methods used in urban STPs. Our data confirms that conventional STPs effectively eliminate a significant number of bacterial cells, along with their associated resistance genes.

We found significant bacterial diversity in the inlet and outlet wastewaters. At the phylum level, Firmicutes (40%), Actinobacteriota (18%), Proteobacteria (18%), and Patescibacteria (12%) were the most dominant in the inlet samples. These major bacterial phyla are found in the gut microbiota of healthy human beings (Arumugam et al. 2011) and are dominant in the gut microbiome of Indian communities (Das et al. 2018). Firmicutes showed a downward shift in the outlet and were replaced by Proteobacteria and Patescibacteria. The environmental bacteria genera Acinetobacter, Aeromonas, and Pseudomonas were present in both the inlet and outlet samples, contributing between 0.74 and 0.01%. These three genera have shown strains that become multi-resistant and should receive special attention. For example, Zhang et al. (2009) showed an increase in antibiotic-resistant Acinetobacter spp. in a wastewater treatment plant. At the phylum level, there were two distinct clusters based on the inlet and outlet specific bacterial communities, which showed no significant differences by season or processing methods. It is possible that abiotic parameters such as oxygen concentration, as well as competition among different bacterial species with different metabolic characteristics, could be driving these compositional differences in the bacterial community between the inlet and outlet.

We found a similar downward shift from the inlet to the outlet at species and genus level. We recorded seven pathogens – Escherichia coli, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Streptococcus pneumoniae, Pseudomonas aeruginosa, and Enterobacter spp. (ESKAPE) – in inlet samples, albeit in small proportions, which showed a further decrease in the outlet samples. This matches findings from studies in other parts of the world, which show that sewage treatment is not able to remove the total burden of microbial diversity but minimizes the effect of pathogenic bacteria (Zhang et al. 2018; Ju et al. 2019). Catenibacterium mitsuokai was dominant and has been positively associated with obesity-related insulin resistance. Lactobacillus ruminis, a commensal motile lactic acid bacterium living in the intestinal tract of humans and animals, contributed to the top 10 bacterial species. Streptococcus parasuis (S. parasuis), a close relative of Streptococcus suis and a potential opportunistic zoonotic pathogen, was also recorded. Megasphaera indica, an obligate anaerobic bacterium isolated from human feces and the genus Subdoligranulum, which was associated with autoantibody development, were present. The genus Arcobacter has been considered an emergent enteropathogen and a potential zoonotic agent (Snelling et al. 2006; Houf 2010). We found Arcobacter midas_s_2255, a potential pathogen commonly abundant in the influent wastewater with high relative abundance in both the inlet and outlet sewage water. Kristensen et al. (2020) reported similar results from Danish wastewater treatment plants and attributed the main reason for the high abundance of this genus since Arcobacter cells do not flocculate and attach well to the activated sludge flocs, thus remaining in high concentration in the water phase.

An equalization analysis using alpha diversity indices revealed a significant increase in the ARG/16S ratio suggesting strong competition between ARG diversity and 16S, which could be detrimental for HGT in a dynamic microbial community in both the inlets and outlets. This further suggests that the microbial equilibrium of the bacterial and fungal communities is regulated by species richness and diversity (Shi et al. 2021).

Bengaluru sewage harbored a wide range of ARGs including clinically relevant CTX-M, qnr, sul-1, and NDM-1 which confer resistance to six major classes of antibiotics. Higher richness of ARGs was detected in the sewage inlets compared with the outlets, and there was significant variation in the clustering of ARGs by STPs but not by sewage processing methods. This is in contrast to studies conducted on a continental scale in Europe (Cacace et al. 2019) and China (Su et al. 2017), where the absolute abundance of ARGs did not exhibit biogeographical patterns in STP effluents. Our data showed spatial structuring of ARGs by STPs, suggesting that there might be environmental factors such as temperature, catchment area size, or number of hospitals that could influence the patterns at a city scale. Hospitals are major contributors of ARGs in wastewater (Hutinel et al. 2022). It is possible that hospitals in the catchment areas have an impact on the effluents of STPs, which requires further investigation. Among the nine antibiotic classes examined, aminoglycosides and beta-lactams were the most dominant. Resistance genes found in many clinical isolates tend to spread beyond the boundaries of hospital catchments. For example, there was an increased prevalence of NDM-1 in river sediment following a pilgrimage in the upper Ganges (Ahammad et al. 2014). Aminoglycosides are used against bacteria that are already resistant to beta-lactams and fluoroquinolones, which further suggests the presence of aminoglycoside-resistant genes in multidrug-resistant (MDR) bacteria, and pan-drug resistant bacteria, particularly in the hospital environment (Ahmed et al. 2021).

The Indian healthcare system lacks strict regulations on over-the-counter access to antibiotics (Laxminarayan & Chaudhury 2016). Furthermore, there are no regulatory provisions for the use of antimicrobials in livestock, including pigs and poultry raised for domestic consumption. Therefore, it is difficult to determine the overall rates of antibiotic prescription or how antibiotic usage varies by season. We observed a significant high in ARG abundance from April to June, which coincides with the opening of schools and an increase in public movement, as well as a decrease in remote working. This led to a huge surge in SARS-CoV-2 infections, as reflected by an increasing viral load in wastewater in June 2022 in Bengaluru city (Lamba et al. 2023). Overall, there was a decrease in ARG abundance from July to October, which could be attributed to a decrease in SARS-CoV-2 infections, as observed through a decreasing viral load in wastewater. It is important to understand if the high abundance of ARGs could be used as a proxy for a seasonal increase in antibiotic usage in the city. Lu et al. (2018) demonstrated a positive correlation between the concentration of tetracyclines and the abundance of total tetracycline-resistant genes in environmental samples. However, there are contrasting findings regarding the significant (P > 0.05) correlation between the absolute concentrations of selected genes and antibiotics. Liyanage et al. (2021) found that resistance genes were detected even in the absence of tetracycline and penicillin in water. It is important to highlight that the type of environmental samples and sampling design, such as a longitudinal study across the sewage network in an urban environment or fecal samples, may capture these dynamics better than an exposed river basin or coastal environments where the prevalence of ARGs is caused by the persistence of these genes even in the absence of selection pressure (Kim et al. 2012). Nonetheless, our data indicate that fluctuating abundance of ARGs could be driven by selective pressure from either microbial communities or antibiotics (Franje et al. 2010).

One of the striking findings was the high prevalence of tetracycline in urban sewage. India accounts for about 3% of the global consumption of antimicrobials in food animals (Van Boeckel et al. 2015). Poultry is one of the most widespread food industries in India. A large variety of antimicrobials is used to raise poultry, and tetracycline class of antibiotics is especially used in animal feed to prevent diseases caused by Mycoplasma, Pasteurella multocida, and E. coli. The growth enhancement properties of antibiotics such as chlortetracycline and oxytetracycline also contribute to their increased commercial use in animal feeds, particularly for broiler chickens (Mehdi et al. 2018). TetA and tetB are the commonly found tetracycline resistance genes in livestock associated Enterobacteriaceae (Bryan et al. 2004). There is reported tetracycline resistance in poultry even without the administration of this antibiotic (Bogaard & Stobberingh 2000).

Our goal is to quantify the differential effects of water treatment mechanisms on ARG diversity and abundance in the outlet of four STPs in Bengaluru where treated water is used for groundwater recharging and agriculture purposes (Bassi et al. 2023). Using a combination of metagenomic approaches, our preliminary findings based on 273 ARGs screened in wastewater showed 54 dominant genes with a downward shift in drug class from inlet to outlet samples, except for aminoglycosides, beta-lactams, MLSB and tetracycline.

According to the United Nations statistics, it has been estimated that up to 90% of perishable vegetables consumed in cities globally are provided by 200 million city dwellers engaged in urban farming. The significant role of urban agriculture in contributing to the proper and sufficient availability of food and poverty alleviation has been recognized and promoted (Martellozzo et al. 2014). Therefore, as part of environmental surveillance, it is important to know the diversity of the microbial community and its role in the digestion process. This knowledge can help improve the performance of STPs and the total abundance of ARGs in environmental samples (von Wintersdorff et al. 2016). Bengaluru has the second-largest treated water system used for agriculture in the world. Further research is needed to understand the dynamics that occur downstream of the STPs and to improve water treatment mechanisms in four key STPs that supply treated water for agriculture and groundwater recharging in peri-urban areas. Extending this study to Bengaluru peri-urban areas (e.g., Kollar, Chikkaballapura, Tumkuru, and Kollar) and profiling ARGs, bacterial, and fungal diversity to evaluate the health impact of treated water in the areas while working with the sewage board in developing a policy and strategy for water quality issues associated with the use of recycled water (e.g., possible persistence of bacterial, viral, and protozoan pathogens) will be crucial for mitigating the impact of wastewater on the health of the ecosystem.

This study does not represent the complete metagenome of the STP. We used an AMR panel designed to screen only select ARGs. It is possible that the diversity of genes in the wastewater is larger than we sequenced. Our data will serve as a baseline study that could aid in quantifying genes from both culturable and non-culturable taxa. We have a very low representation of genes that confer resistance to last resort antibiotics. Nonetheless, this was an improvement over PCR-based approaches which may have greater sensitivity to low-abundance ARGs due to targeted amplification but are limited in the ARGs they can detect.

Going forward, shotgun metagenomics can help alleviate some of the drawbacks of amplicon-based sequencing. The inherent connections between microbial ecology and environmental biotechnology (of which wastewater treatment is an important aspect) have recently been emphasized. A longitudinal study that combines antibiotic levels in wastewater and ARGs could help us understand the impact of treated wastewater on the microbial community.

We sincerely thank the Bengaluru Water Supply and Sewerage Board for providing access to the sewershed sites. We are grateful to NGS facility at the National Centre for Biological Sciences for facilitating the sequencing of wastewater samples. D.M. is supported by funding from the Rockefeller Foundation (grant 2021 HTH 018). F.I. and A.N. are supported by TIGS.

This research was financially supported by the Tata Trusts to Tata Institute for Genetics and Society (TIGS).

F.I. and V.S. conceptualized and designed the study; A.N. processed samples for molecular and CLC analysis; D.M. helped with statistical analysis; V.S. facilitated sewage sample collection and understanding of sanitation network; and F.I. led the sewage surveillance, analyzed the data, and wrote the manuscript. All authors approved the manuscript.

All relevant data are available from https://www.ncbi.nlm.nih.gov/sra/PRJNA1114410.

The authors declare there is no conflict.