Versatile wastewater monitoring of pathogens and antimicrobial resistance enabled by metatranscriptomics and long-read metagenomics

Abstract

Widespread interest in the development of population-wide pathogen and antimicrobial resistance (AMR) monitoring has revealed wastewater's microbial footprint as a marker of public health. Near-source wastewater remains a difficult sample type for microbiome analyses but represents a closer link to human health than the downstream products of its treatment. Few studies integrate methods for non-targeted monitoring applications, and critically, current methods cannot connect AMR genes to species, nor resolve full genomes. We address these challenges by developing a pipeline that enables untargeted metagenomics, metatranscriptomics, and novel long-read metagenomics (LRG). We achieve untargeted pathogen detection, limited by highly abundant resident species, while retaining microbial information with near-source sampling. Furthermore, LRG identifies antibiotic resistance gene-containing microbes and enables assembly of culture-independent genomes with previously unreported AMR genes. We establish an integrated approach to broadly monitor pathogens in wastewater, while demonstrating the importance of LRG to illuminate microbial AMR at the species level.

Figure 1

Community-level analysis reveals distinct microbial diversity patterns in near-source and wastewater treatment plant samples. A) Automated wastewater processing workflow. 24-hour flow-weighted wastewater composites collected from near-source sampling (NSS, n = 285) at campus manholes or sewer cleanouts, as well as at influent pumping stations of wastewater treatment plants (WWTPs n = 286; Point Loma and Encina facilities). Samples were processed in automation-compatible microplates with stepwise increases in sample density, followed by total nucleic acid extraction for parallel metagenomic and metatranscriptomic library preparation from a single sample. B) Alpha diversity comparisons. Boxplots display distribution of alpha diversity metrics of uniquely Observed Features and Faith’s Phylogenetic Diversity from metagenomic reads (Mann-Whitney-Wilcoxon two-sided test with Benjamini-Hochberg correction, **** denotes P ≤ 1× 10⁻⁴). C) Beta diversity and community structure. Principal coordinate analysis (PCoA) of unweighted UniFrac distances illustrate differences in metagenomic community composition between NSS samples from the UC San Diego campus compared to WWTP samples from San Diego County (PERMANOVA P = 0.001). WWTP microbial communities further segregated by county catchment area, while NSS communities differed between residential and non-residential buildings. Panel A is created in BioRender. Din, O. (2025) https://BioRender.com/.

Figure 2

Pathogen detection in wastewater short-read metagenomics is shaped by background microbial abundance. A) Background prevalence of representative pathogens. Relative abundance distributions for E. coli, C. difficile, and C. auris across NSS and WWTP samples, with detection prevalence indicated for each taxon. B) Spike-in experimental design. Total nucleic acids from the three pathogens was spiked into extracted nucleic acids from NSS and WWTP samples (and pure water controls) at concentrations ranging from 10⁻¹ to 10⁵ genome copies, followed by metagenomic library preparation and sequencing. C–D) Detection responses to spike-ins. Genome coverage (%) and relative abundance as a function of spike-in concentration for each pathogen in NSS (i) and WWTP (ii) backgrounds. E) Conceptual model of detection sensitivity. Illustration of how pathogen background abundance influences the ability of metagenomics based read-mapping and coverage approaches to enable detection from complex wastewater samples. Panels B and C are created in BioRender. Din, O. (2025) https://BioRender.com/.

Figure 3

Metatranscriptomics enables detection of pathogens and AMR expression A) Detection response of metatranscriptomic reads to pathogen spike-ins. Relative abundance (log10) as a function of spike-in concentration for each pathogen in NSS and WWTP backgrounds. B) Comparison of ARG detection across wastewater sample sources. Boxplots (left) display distribution of percentages of ARG mapped reads across NSS and WWTP samples (P = 2.848 × 10⁻², U = 3.683 × 10⁴). Countplots (right) of unique ARGs and their distribution among reference databases. Genes observed from metatranscriptomic reads aligning to distinct alleles, represented by either CARD or Resistomes & Variants reference DB, are displayed as a reference DB overlap (orange). C-D) Description of ARG types in wastewater metatranscriptome samples, split by reference database. Top barplots depict prevalence of ARG types among samples. Middle boxplots display log10 transformation of relative abundance of genes by ARG type. Bottom countplots (log10) tally number of unique genes of each ARG type.

Figure 4

Long-read metagenomic reads enables species-level AMR gene context A) Connecting ARG-genes to species-level context with long-reads. Schematic of the analysis approach to delineate antibiotic resistance gene (ARG) information in species context using Argo. B) Microbial genera across Campus samplers. Taxa barplot showing the relative abundance at the genus level across near-source wastewater samplers. Colored taxa represent the top 10 genera across all samplers (see legend), while all other genera are in gray. C) Distribution of microbial species across ARG types. Taxa barplot showing the relative abundance of microbial species across identified ARG types. Species in the legend represent the top 20 species in terms of abundance across all ARG types. Above each bar, the number of classified species (black text) and unclassified species (blue text) are labeled. D) Top ARGcontaining species. Barplot showing the top 15 assigned species taxonomies in terms of number of unique ARG genes. Panel A is created in BioRender. Din, O. (2025) https://BioRender.com/.

Figure 5

Long-read metagenomics produces full genomes of wastewater microbes A) Complete and high-quality MAGs. Plot showing the contamination and completeness of all the complete MAGs (MAGs, n=42) generated from long-read metagenomic assemblies. Each point represents one MAG, all less than 20 contigs. The color of each point indicates the number of contigs for each MAG (as shown in legend). Color gradient follows green to magenta for lower to higher numbers of contigs per MAG. B) Single-contig, complete MAGs. Barplot showing the length of the single-contig genomes (cMAGs) with identified taxa at the species, genus (shown as the genus name followed by sp.), or family level (shown in parentheses). Note that some taxa have multiple assemblies, such as Cloacibacterium caeni, denoted by a number for identification (1, 2, 3, etc..). C) Long-read MAGs as sources of unclassified, ARG-containing reads. Unclassified ARG reads identified in Fig. 4 mapped against the long-read de novo generated MAGs (inset). Plot shows the numbers of reads mapped to each MAG, colored circles, from its corresponding taxonomic classification represented on the x-axis. The color for each circle conveys the unique number of ARGs identified in the unclassified reads mapped to the MAG. D) C. caeni comparison with closely related species. Phylogenetic tree of marker genes rooted by midpoint of our newly assembled and publicly available genomes from Cloacibacterium. Tips are annotated by GTDB-Tk species identification and country of origin. Inset for panel C is created in BioRender. Din, O. (2025) https://BioRender.com/.