Skip to main content
Advertisement
  • Loading metrics

Advancing antimicrobial resistance monitoring in surface waters with metagenomic and quasimetagenomic methods

  • Andrea Ottesen ,

    Roles Conceptualization, Data curation, Investigation, Methodology, Project administration, Supervision, Validation, Visualization, Writing – original draft

    andrea.ottesen@gmail.com, andrea.ottesen@fda.hhs.gov, aottesen@umd.edu

    Affiliations Center for Veterinary Medicine, Food and Drug Administration, Laurel, Maryland, United States of America, National Antimicrobial Resistance Monitoring System (NARMS), Laurel, Maryland, United States of America

  • Brandon Kocurek,

    Roles Formal analysis, Investigation, Methodology, Project administration, Software, Validation

    Affiliations Center for Veterinary Medicine, Food and Drug Administration, Laurel, Maryland, United States of America, National Antimicrobial Resistance Monitoring System (NARMS), Laurel, Maryland, United States of America

  • Padmini Ramachandran,

    Roles Formal analysis, Methodology, Software

    Affiliation Center for Food Safety and Applied Nutrition, Food and Drug Administration, College Park, Maryland, United States of America

  • Elizabeth Reed,

    Roles Formal analysis, Visualization

    Affiliation Center for Food Safety and Applied Nutrition, Food and Drug Administration, College Park, Maryland, United States of America

  • Seth Commichaux,

    Roles Formal analysis

    Affiliation Center for Food Safety and Applied Nutrition, Food and Drug Administration, College Park, Maryland, United States of America

  • Gunnar Engelbach,

    Roles Formal analysis, Software

    Affiliation Center for Food Safety and Applied Nutrition, Food and Drug Administration, College Park, Maryland, United States of America

  • Mark Mammel,

    Roles Formal analysis

    Affiliation Center for Food Safety and Applied Nutrition, Food and Drug Administration, College Park, Maryland, United States of America

  • Sanchez Saint Fleurant,

    Roles Investigation

    Affiliations Center for Veterinary Medicine, Food and Drug Administration, Laurel, Maryland, United States of America, National Antimicrobial Resistance Monitoring System (NARMS), Laurel, Maryland, United States of America

  • Shaohua Zhao,

    Roles Supervision, Writing – review & editing

    Affiliations Center for Veterinary Medicine, Food and Drug Administration, Laurel, Maryland, United States of America, National Antimicrobial Resistance Monitoring System (NARMS), Laurel, Maryland, United States of America

  • Claudine Kabera,

    Roles Conceptualization, Project administration

    Affiliations Center for Veterinary Medicine, Food and Drug Administration, Laurel, Maryland, United States of America, National Antimicrobial Resistance Monitoring System (NARMS), Laurel, Maryland, United States of America

  • Amy Merrill,

    Roles Investigation, Writing – review & editing

    Affiliations Center for Veterinary Medicine, Food and Drug Administration, Laurel, Maryland, United States of America, National Antimicrobial Resistance Monitoring System (NARMS), Laurel, Maryland, United States of America

  • Nathalie Bonin,

    Roles Formal analysis

    Affiliation Department of Computer Information Science and Engineering, University of Florida, Gainesville, Florida, United States of America

  • Hannah Worley,

    Roles Formal analysis

    Affiliation Department of Veterinary Population Medicine, University of Minnesota, Falcon Heights, Minnesota, United States of America

  • Noelle Noyes,

    Roles Formal analysis

    Affiliation Department of Veterinary Population Medicine, University of Minnesota, Falcon Heights, Minnesota, United States of America

  • Christina Boucher,

    Roles Formal analysis

    Affiliation Department of Computer Information Science and Engineering, University of Florida, Gainesville, Florida, United States of America

  • Patrick McDermott,

    Roles Supervision, Writing – review & editing

    Affiliations Center for Veterinary Medicine, Food and Drug Administration, Laurel, Maryland, United States of America, National Antimicrobial Resistance Monitoring System (NARMS), Laurel, Maryland, United States of America

  •  [ ... ],
  • Errol Strain

    Roles Supervision, Writing – review & editing

    Affiliations Center for Veterinary Medicine, Food and Drug Administration, Laurel, Maryland, United States of America, National Antimicrobial Resistance Monitoring System (NARMS), Laurel, Maryland, United States of America

  • [ view all ]
  • [ view less ]

Abstract

The National Antimicrobial Resistance Monitoring System (NARMS) has monitored antimicrobial resistance (AMR) associated with pathogens of humans and animals since 1996. In alignment with One Health strategic planning, NARMS is currently exploring the inclusion of surface waters as an environmental modality for monitoring AMR. From a One Health perspective, surface waters function as key environmental integrators between humans, animals, agriculture, and the environment. Surface waters however, due to their dilute nature present a unique challenge for monitoring critically important antimicrobial resistance. Selective enrichments from water paired with genomic sequencing effectively describe AMR for single genomes but do not provide data to describe a broader environmental resistome. Metagenomic data effectively describe a broad range of AMR from certain matrices however, depth of coverage is usually insufficient to describe clinically significant AMR from aquatic matrices. Thus, the coupling of biological enrichments of surface water with shotgun NGS sequencing has been shown to greatly enhance the capacity to report an expansive profile of clinically significant antimicrobial resistance genes. Here we demonstrate, using water samples from distinct sites (a creek in close proximity to a hospital and a reservoir used for recreation and municipal water), that the AMR portfolio provided by enriched (quasimetagenomic) data is capable of describing almost 30% of NARMS surveillance targets contrasted to only 1% by metagenomic data. Additionally, the quasimetagenomic data supported reporting of statistically significant (P< 0.05) differential abundance of specific AMR genes between sites. A single time-point for two sites is a small pilot, but the robust results describing critically important AMR determinants from each water source, provide proof of concept that quasimetagenomics can be applied to aquatic AMR surveillance efforts for local, national, and global monitoring.

Introduction

From a One Health perspective, surface waters function as key environmental integrators. They receive human, agricultural, and wildlife input and provide that same water for human, agricultural, and wildlife sustenance. Industrial and agricultural chemicals, metals, food additives, antibiotics, and even non-antibiotic drugs, have all been shown to play influential roles in the spread of AMR [1, 2]. Understanding the presence of pathogens and antimicrobial resistance (AMR) determinants in surface waters helps to inform risks across a wide range of applications and provides an integrative approach to public health. Recognizing the significant health impact that environmental water has on humans, animals, and the environment [35] the National Antimicrobial Resistance Monitoring System (NARMS) is investigating surface waters as a potential environmental modality for One Health AMR monitoring. This strategy requires methodological approaches capable of reporting AMR from an ecosystem as complex and dilute as water. Currently, NARMS monitoring efforts use standard in vitro antimicrobial susceptibility testing (AST) to generate minimum inhibitory concentrations (MICs) and whole genome sequencing (WGS) to predict resistant phenotypes directly from nucleotide data [6, 7]. These approaches rely on preliminary selective enrichments that produce high resolution data characterizing AMR phenotypes of pathovars. Metagenomic data is useful for providing ‘big pictures’ of certain environmental, human, and animal microbiomes; however, for the pathogens under active surveillance by NARMS, metagenomic data usually doesn’t provide sufficient depth of coverage to describe AMR phenotypes–especially for water. Quasimetagenomic data (QMGS), due to its inclusion of an enrichment step, provides coverage of critically important resistance determinants with a broader throughput than culture independent (CI) metagenomics or WGS.

Critically important antimicrobials (CIA) are ranked by the World Health Organization (WHO) according to their importance in human medicine in efforts to develop risk management for control of AMR in humans and animals. CIAs from WHO comprise more than a dozen different classes of antibiotics [8]. Currently NARMS monitors a subset of genes conferring resistance to CIAs in Escherichia, Salmonella, Campylobacter and Enterococcus primarily isolated from human, food-producing animals, raw retail meats, and feed environments. NARMS seafood monitoring also tracks resistance in Aeromonas and Vibrio species. For NARMS, critically important determinants for monitoring efforts include genes conferring resistance to aminoglycosides, quinolones, ß-lactams, colistin, macrolides and ketolides, oxazolidinones, penicillins (Table 1) [9].

thumbnail
Table 1. Genes encoding resistance to critically important antimicrobial agents.

https://doi.org/10.1371/journal.pwat.0000067.t001

For NARMS’s expanded One Health mission, new approaches that are capable of detecting and describing critically significant resistance determinants will be needed. In the present study dead end ultrafiltration [10] was used to collect 50L volumes from each of two sites: 1) an urban creek in close proximity to a hospital (Sligo), and 2) a recreational reservoir which is the source of drinking water for Prince George’s county, Maryland (Patuxent). DNA extracted from water was evaluated using both metagenomic and quasimetagenomic data to assess the ability of each data type to describe critically important NARMS resistance determinants.

Quasimetagenomics, which uses shotgun sequencing of enriched microbiomes at strategic temporal increments during pathogen recovery protocols, has previously proven useful for expedited source tracking in outbreak investigations [11], identification of multi-serovar diversity associated with samples linked to outbreaks, and identification of co-competitors, co-enrichers, and recovery biases in state of the art FDA microbiological culturing methods [12, 13]. Here we demonstrate that the approach also greatly enhances capacity for resistome reporting and description of critically important resistance determinants, and taxa important to global AMR morbidity and mortality.

Results

Resistome composition of the Sligo and Patuxent water by CI and QMGS approaches

The difference between the CI and QMGS AMR profiles was always statistically significant, with QMGS demonstrating far more sensitivity for detection of AMR. Out of 132 AMR gene/class comparisons, there were only 9 times that CI was not zero. This is demonstrated using results from four bioinformatic pipelines (and corresponding databases) for Sligo and Patuxent water in Figs 1 and 2. Fig 1 shows annotations of ß-lactams by AMRPlusPlus (RPKM) [14], AMRFinder Plus, CARD [15], and COSMOSID (www.cosmosid.com) for both CI and QMGS samples. No ß-lactam annotations (using default parameters and normalization) were reported by any of the pipelines for CI water samples except by the kmer based COSMOSID pipeline. (Kmer based approaches have well demonstrated sensitivity, but without dedicated curation and validation, specificity can be unreliable). Reporting for almost every class of drug, biocide, metal, and multi-compound resistance was vastly expanded by QMGS data (Figs 1 and 2). With CI data alone, it was not possible to evaluate differences between the two sites for trimethoprims, sulfonamides, phenicols, oxazolidinones, nucleosides, metronidazoles, lipopeptides, glycopeptides, fosfomycins, ß-lactams, bacitracins, and many other groups due to lack of coverage in at least one set of CI samples and often both.

thumbnail
Fig 1. Resistome Composition: ß-lactams by different annotation pipelines.

Four different analytical approaches, AMRPlusPlus 2.0 [14], AMRFinder Plus [16], CARD [15] and COSMOSID [17], consistently reported substantially more ß-lactam antimicrobial resistance genes in QMGS data compared to CI data. a) AMRPlusPlus, b) CARD, c) AMRFinder, and d) COSMOS ID.

https://doi.org/10.1371/journal.pwat.0000067.g001

thumbnail
Fig 2. Resistome composition: Sligo and Patuxent water by CI and QMGS approaches.

Detection of AMR genes in CI and QMGS data are shown for a) multidrug resistance b) ß-lactams c) multimetal resistance and d) multibiocide resistance. The AMRPlusPlus pipeline with Megares database was used for all annotations [14].

https://doi.org/10.1371/journal.pwat.0000067.g002

Resistome composition: Identification of NARMS critically important antimicrobial resistance

A key objective for this work was to evaluate metagenomic and quasimetagenomic data for their potential contribution to detection of resistance genes important to the NARMS monitoring program. From the list of NARMS ‘critically important resistance gene targets’ (Table 1), more than 30% were identified across the two water sources using the QMGS approach compared to only 1% identified using metagenomics. Critically important gene targets were observed at 100% identity across 100% coverage using the Comprehensive Antibiotic Resistance Database (CARD) for Sligo and Patuxent QMGS data; blaTEM-207, blaTEM-15, blaSHV-30, blaSHV-12, qnrS2, qnrB9, qnrB7, qnrB6, qnrB2, qnrB19, qnrB1, oqxB, oqxA, msrE, mphE, mphA, mefC, blaFOX-5, ereA, blaCMY-43, blaCMY-4, blaCMY-30, blaCMT-2, acrB. With the exception of ereA, observed in CI Sligo water samples, all genes of importance to NARMS monitoring were observed only in QMGS samples (Fig 3).

thumbnail
Fig 3. NARMS critically important genes.

From the list of NARMS critically important antimicrobial (CIA) resistance genes, using 100% identity across 100% of gene length annotated using the CARD database [15], almost 30% (29.4) were identified across the two water sources using the QMGS approach compared to 1% identified in CI samples.

https://doi.org/10.1371/journal.pwat.0000067.g003

Resistome composition: Quinolone resistance genes in Sligo and Patuxent water

Fluoroquinolone resistance genes and variants were reported by the AMRPlusPlus pipeline (Megares 2.0 database) and additional confirmation analytics were conducted ‘by hand’ comparing resistance determining alleles to genes identified in Sligo and Patuxent water to ensure validated annotation of fluoroquinolone resistance determinants [18]. From Patuxent water, the resistance determining allele of gyrB S464Y (1) was confirmed while all others were identified as wildtype alleles. In Sligo water there were more fluoroquinolone resistance determining alleles confirmed; including gyrB E466D (1), S464Y (1), gyrA S83N(2), S83T, and parC S80R, S80I. Additional flouroquinolone genes reported by the AMRFinderPlus pipeline included; qnrD, qnrD1, and qnrS2.

Resistome Composition: ß-lactams

No ß-lactams were detected in CI samples at all, except for a few by the kmer based COSMOSID pipeline. ß-lactam reporting from QMGS samples however, provided sufficient data to describe statistically significant differential abundance of ß-lactam determinants (Fig 4). Fig 4 shows statistically significant (p< 0.05) in ß-lactam resistance genes associated with Sligo and Patuxent water.

thumbnail
Fig 4. Statistically significant differential abundance of ß-lactam resistance determinants in Sligo and Patuxent water by QMGS data.

Counts of ß-lactam resistance genes with statistically significant differential abundance (p< 0.05) (binomial exact test with expected probability of 0.5) between Patuxent (grey) and Sligo (black) are shown here. A key for the genes shown here is reproduced from the Megares database here: OXA: The OXA genes encode Ambler class D beta-lactamases. cphA: The cphA gene encodes an Ambler class B beta-lactamase. CARB: The CARB genes encode Ambler class A beta-lactamases. CMY: The CMY genes encode Ambler class C beta-lactamases. CTX: The CTX genes encode Ambler class A beta-lactamases. MOX: The MOX genes encode Ambler class C beta-lactamases. FOX: The FOX genes encode Ambler class C beta-lactamase. AMPC: The AMPC group of genes confer resistance to betalactams through the mechanism, Class C betalactamases. PBP2: Penicillin bindings proteins are the target of beta-lactam antibiotics. Multiple types of PBPs exist, and it has been demonstrated that mutations in each can confer resistance to beta-lactam antibiotics. CEPH: The CEPH group of genes confer resistance to betalactams through the mechanism, Class C betalactamases. CEPS: The CepS gene encodes an Ambler class C beta-lactamase. PBP4B: Penicillin bindings proteins are the target of beta-lactam antibiotics. Multiple types of PBPs exist, and it has been demonstrated that mutations in each can confer resistance to beta-lactam antibiotics.

https://doi.org/10.1371/journal.pwat.0000067.g004

Resistome composition: Plasmids

Using Platon, a pipeline for identification and characterization of bacterial plasmid contigs in short-read draft assemblies by exploiting protein sequence-based replicon distribution scores [12], we were able to annotate plasmids in both Sligo and Patuxent water using QMGS data (Fig 5). IncFIIK, IncQ2, Col(pHAD28) were only observed in Sligo while others were observed in both Sligo and Patuxent. InC, Inc, and IncI plasmids often carry AMR genes especially in Enterobacteriaceae. IncHI and IncHII’s plasmid types are also frequently associated with AMR genes [19] and observed IncI plasmid (Gamma1AP005147) shares genomic similarity to the pESI megaplasmid found in Salmonella serovar Infantis.

thumbnail
Fig 5. Plasmids in Sligo and Patuxent water.

Composition of plasmids that could be identified in QMGS data by Platon from Sligo and Patuxent water (none could be annotated from CI data).

https://doi.org/10.1371/journal.pwat.0000067.g005

While the vast majority of antimicrobial resistance genes were identified in QMGS data, there was a small number of genes for which the converse was true; i.e., more information was provided by CI assessment than by QMGS. Almost half of the AMR genes observed exclusively in CI data (by AMRPlusPlus) were Mycobacterium tuberculosis drug specific, although each annotation had the addendum that further confirmation would be needed to confirm resistance-determining single nucleotide polymorphisms (SNPs). SNP validation is currently in development for Mycobacterium tuberculosis resistance determinants but the observation that Mycobacterium spp. were almost exclusively observed in culture independent data is important to environmental Mycobacterium surveillance efforts. The length of time needed to culture a Mycobacterium isolate has been described as 6 days or more [20] so it is not surprising that a 24H timepoint in a broth designed primarily for enrichment of Enterobacteriales would not provide the right conditions to observe Mycobacterium populations.

Microbiome composition: Bacteria from Sligo and Patuxent by CI & QMGS approaches

Diverse taxonomic compositions were observed between the CI Sligo and Patuxent water microbiota; Burkholderiales, Comamonadaceae, Aeromonadales, and Caulobacterales were abundant in Sligo water but not in the Patuxent samples. Actinobacteria, Alphaproteobacterium, and Chroococcales were abundant in Patuxent water and not in Sligo (CI data shown in Fig 6A and 6B). Sligo and Patuxent QMGS taxonomy was much more homogenous (QMGS data shown in Fig 6C and 6D). Enterobacteriales were present in less than 1% of CI samples but comprised between 15% and 20% of QMGS water enrichments. After enrichment, samples are biased toward recovery of taxa that thrive in the nutrient conditions of the media, in this case BPW, comprised of peptone, sodium chloride, disodium phosphate, mono-potassium phosphate and distilled water [21]. While enrichments obscure certain taxonomic and functional features of the true microbiome, they also provide robust data to monitor critically important AMR from medically relevant taxa.

thumbnail
Fig 6. Microbiome composition: Bacterial taxa from Sligo and Patuxent CI and QMGS.

Enterobacteriales were present in less than 1% of CI samples (a & b) but comprised between 15% and 20% of QMGS water enrichments (c & d). Shaded regions are highlighted if they correspond to taxa with >1% abundance in a sample. If the abundance is less than 1 percent, there is no area shading but a node may still have a color when observed at less than 1 percent.

https://doi.org/10.1371/journal.pwat.0000067.g006

The Sligo and Patuxent data were examined for key NARMS monitoring targets; Campylobacter, Salmonella, Escherichia coli, Enterococcus, and NARMS seafood targets; Aeromonas and Vibrio (Figs 7 and 8). Data were also examined for presence of ‘ESKAPE’ nosocomial pathogens, known for carriage of multidrug resistance and virulence. These are Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, and Enterobacter spp., and for pathogens with the highest number of global deaths attributable to AMR; Escherichia coli, Staphylococcus aureus, Klebsiella pneumoniae, Streptococcus pneumoniae, Acinetobacter baumannii, and Pseudomonas aeruginosa [22]. ESKAPE bacteria and bacteria attributable to the highest numbers of AMR associated deaths, were combined to create a list of 8 important bacterial monitoring targets: Staphylococcus aureus, Klebsiella pneumonia, Acinetobacter baumannii, Pseudomonas aeruginosa, Escherichia coli, Streptococcus pneumoniae, Enterococcus faecium and Enterobacter spp. (Fig 9). Figs 79 show target species in the first panel and all observed species from those same genera in the second panel. This comprehensive ecological view of all species in the genus of a surveillance target may prove useful for identification of emerging pathogens with critically important resistance.

thumbnail
Fig 7. Species important to NARMS Monitoring.

Panel A shows NARMS targets; Campylobacter, Salmonella, Enterobacter and Escherichia genera and panel b show all related species.

https://doi.org/10.1371/journal.pwat.0000067.g007

thumbnail
Fig 8. Panel a shows the relative abundance of observed NARMS seafood pathogens Vibrio and Aeromonas and panel b shows all observed species.

https://doi.org/10.1371/journal.pwat.0000067.g008

thumbnail
Fig 9. ESKAPE8 pathogens.

The ‘ESKAPE8’ list used here is comprised of the 6 ESKAPE pathogens; Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, and Enterobacter spp., and two additional species Escherichia coli, Streptococcus pneumoniae, which play a key role in AMR associated mortality (see Box 1. Definitions). In panel a) genus and species of pathogens under active surveillance are shown and in panel b) all observed species from these genera are shown.

https://doi.org/10.1371/journal.pwat.0000067.g009

Box 1. Definitions

ESKAPE8 The ESKAPE8 list used here is comprised of the 6 ESKAPE pathogens; Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, and Enterobacter spp. merged with the only two not represented in that list from the top 6 species associated with global AMR mortality ie: Escherichia coli, Staphylococcus aureus, Klebsiella pneumoniae, Streptococcus pneumoniae, Acinetobacter baumannii, and Pseudomonas aeruginosa. Thus, ‘ESKAPE8’ = Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, Enterobacter spp., Escherichia coli and Streptococcus pneumoniae.

Bacterial taxa of importance to veterinary monitoring efforts

Important veterinary pathogens were also detected from Sligo and Patuxent water samples. The Veterinary Laboratory Investigation and Response Network (Vet-LIRN) regularly monitors E. coli and Salmonella (NARMS targets, Fig 7) and the Staphylococcus intermedius group (SIG). The SIG, comprised of Staphylococcus pseudointermidius, Staphylococcus intermedius, and Staphylococcus delphini, is a collective of emerging importance [23]. The European Antimicrobial Resistance Surveillance network in Veterinary medicine (EARS-Vet) monitors 11 bacterial targets; Escherichia coli, Klebsiella pneumoniae, Mannheimia haemolytica, Pasteurella multocida, Actinobacillus pleuropneumoniae, Staphylococcus aureus, Staphylococcus pseudintermedius, Staphylococcus hyicus, Streptococcus uberis, Streptococcus dysgalactias and Streptococcus suis, across six animal species (cattle, swine, chickens (broilers and laying hens) turkeys, cats, and dogs) [24]. Genera and species from EARS-Vet surveillance that were observed in Sligo and Patuxent water are shown in Fig 10. As previously indicated, most pathogen species under active surveillance require general and selective enrichment protocols before they can be recovered, sequenced, and described. Certain genera, however, were more robustly reported by culture independent data. Mycobacterium, Neisseria, Mycoplasma, and Borrelia for example, were predominantly observed in CI data from Sligo and Patuxent water compared to QMGS data (Fig 11).

thumbnail
Fig 10. EARS-veterinary surveillance targets.

The European Antimicrobial Resistance Surveillance network in Veterinary medicine (EARS-Vet) monitors 11 bacterial targets; Escherichia coli, Klebsiella pneumoniae, Mannheimia haemolytica, Pasteurella multocida, Actinobacillus pleuropneumoniae, Staphylococcus aureus, Staphylococcus pseudintermedius, Staphylococcus hyicus, Streptococcus uberis, Streptococcus dysgalactias and Streptococcus suis, across six animal species (cattle swine, chickens (broilers and laying hens) turkeys, cats and dogs). Incidence of these taxa is shown for Sligo and Patuxent water.

https://doi.org/10.1371/journal.pwat.0000067.g010

thumbnail
Fig 11. Incidence of Borrelia, Mycobacterium, Mycoplasma and Neisseria in Sligo and Patuxent water is primarily associated with CI data.

As previously described for Mycobacterium, it is important for monitoring efforts to understand when CI data will be more useful than QMGS data and vice versa. All the species shown here were primarily observed in CI data.

https://doi.org/10.1371/journal.pwat.0000067.g011

Discussion

Water plays perhaps the most important role in states of health and disease in humans and other animals, yet it is one of the most complex matrices to monitor due to its dilute nature. Metagenomics is often a valuable way to monitor a wide breadth of AMR from complex environmental samples but for certain matrices such as surface waters, there is simply not enough sequence coverage of AMR genes to describe critically important resistant determinants in water. Work by Gweon et al. used 200 million CI reads to describe AMR in pig caeca, effluent, and stream sediment. For pig caeca, the highest number of hits to AMR genes reached about 55,000, for effluent, about 22,000, but for stream sediment only 22 hits were observed [25]. This is consistent with numbers observed by metagenomic methods in the water samples examined here (stream and reservoir) with 70 to 150 million reads. CI data was insufficient for the monitoring aims of the NARMS program, but QMGS data met the challenge of reporting clinically important antimicrobial resistance. While QMGS produced robust data, there will inevitably be biases and limitations associated with this approach that will be important to evaluate in future work. The use of established AMR multiplex PCR panels side by side QMGS data will inform on just how much is missed by a QMGS approach and perhaps also, how much is gained. While PCR approaches may be more sensitive for a wide range of genes, they will never provide data beyond the panel. QMGS may have less sensitivity for certain genes, but it will also support discovery and provide complementary environmental genomic data to describe plasmids and co-occurring taxa to better understand community dynamics.

Quasimetagenomic approaches have an established history of expediting WGS based source tracking and plasmid identification, as well as describing multi-serovar diversity associated with outbreaks [11, 12, 26]. Here we demonstrate that QMGS data also provides robust utility for AMR reporting from surface waters. The QMGS data presented here represents water incubated for 24 H at 37° in Buffered Peptone Water (BPW). There are many opportunities for advanced, precision QMGS targeting of important species and consortia of species by adjusting nutrients, oxygen tension, time, temperature, and addition of antibiotics. Even in enriched (QMGS) samples for Sligo and Patuxent, with an average of 70 million reads per replicate, the coverage of taxa responsible for the most AMR deaths worldwide (Escherichia coli, Staphylococcus aureus, Klebsiella pneumoniae, Streptococcus pneumoniae, Acinetobacter baumannii, and Pseudomonas aeruginosa [22]) comprised less than 6% of the total data. This, in contrast to less than 1% coverage in CI data, is a vast improvement but still leaves opportunity for further optimization. Bioinformatic methods will also require optimization, validation, and synchronization for harmonized reporting of taxonomy and resistance determinants.

Important work has described AMR in water and effluent [35, 2737], but lack of a consensus set of tools to harmonize methodologies and analyses across multiple groups’ efforts from water collection and laboratory processing to bioinformatic analysis and reporting, leaves us without a ready, common source of information to support local, national, and global AMR surveillance of key environmental integrators like surface waters. Metagenomic approaches provided valuable description of species such as Mycobacterium, Mycoplasma and Neisseria (as shown in Fig 11), but for monitoring of NARMS critically important gene targets, we recommend that metagenomics be coupled with quasimetagenomics.

The Centers for Disease Control and Prevention (CDC) estimate that millions of people in the United States are impacted by waterborne diseases every year [38]. The WHO reports that 144 million people rely on untreated surface water [39], with projections that by 2025, more than half of the world’s population will live in water stressed areas [39]. Surface waters have the potential to serve as sentinels for global One Health monitoring of antimicrobial resistance. Murray et al. estimate there were 1.27 million deaths attributable to antimicrobial resistance in 2019 [22]. That number is close to the global number of HIV (680,000) and malaria (627,000) deaths combined [4042]. Understanding the flow of antimicrobial resistance through ecosystems is a fundamental objective for One Health research and a new priority for the National Antimicrobial Resistance Monitoring System.

As laboratory and informatic methods are harmonized in collaborative endeavors spanning microbiology, molecular biology, chemistry, hydrology, epidemiology, and interoperable metadata ontology, it will be possible to develop global resources to describe phenotype evolution, plasmid dynamics and an improved understanding of the flow of AMR through human, animal and environmental ecosystems. Our findings suggest that metagenomic and quasimetagenomic data used together, provide a valuable framework to support a new era of One Health AMR monitoring and identification of emerging resistance.

Materials and methods

Water collection

Dead end ultrafiltration (DEUF) was used to collect 50 L of water from Sligo and Patuxent water sources [28]. DEUF was done using a Hemodialyzer Rexeed 25S filters (AsahiKasei, Chiyoda, Tokyo, Japan) and a Geopump peristaltic pump (Geotech, Denver, CO). Cells and particles are caught in the hollow fiber membranes within the ultrafilter filter, while filtrate passes through. Ultrafilter membrane separation collects particles between nano and micro (pore size range of 0.001–0.05 μm) and all manner of larger organisms and debris. The size range captured by ultrafiltration is ideal for examination of viruses, bacteria, fungi, and protists in water. After collection, filters are capped, bagged and stored at 4°C, until backflushing.

Laboratory processing

Backflushing.

First step for laboratory processing is the ‘backflushing’ of the filter.

Backflush concentrate was used for culture independent (CI) metagenomics by direct DNA extraction of water filters passed through 0.2 micron filters. Nucleic acid extraction was conducted directly on the filters as part of the Qiagen DNeasy PowerWater DNA extraction protocol (Qiagen, Germantown, MD, United States) according to the manufacturer’s specifications (Qiagen PowerWater Kit Handbook).

Enrichment

Enrichment for quasimetagenomes was achieved by adding BPW at a 1 to 1 ratio (25 ml to 25 ml) (4 replicates of backflushed filtrate from each 50L ultrafilter collections with incubation at 37° for 24 H). After 24 H of incubation at 37°, 2 ml aliquots were removed, centrifuged and DNA was extracted from the pellet using the Zymo High Molecular Weight DNA Extraction kit according to the manufacturers specifications (Zymo Quick-DNA Magbead Handbook).

Library preparation and sequencing

DNA libraries from both CI and QMGS samples was prepared using the Illumina DNA Library Prep according to the manufacturers specifications (Illumina). https://www.protocols.io/edit/illumina-dna-prep-sop-bzstp6en.

Sequencing was performed on a NextSeq 500 according to the manufacturer’s specifications. All sequencing runs were performed in paired end mode with 2 x 150 cycles using the NextSeq 500/550 v2.5 High Output Kit (150 Cycles). Libraries were diluted to 1.8 pM according to the manufacturer’s specifications (NextSeq Denature and Dilute Libraries Guide).

Bioinformatic analyses

Files were demultiplexed (bcl to fastq) and screened/trimmed using Trimmomatic [43].

Four replicates of each treatment with reads per sample spanning 20 million to 150 million reads were used for further downstream analyses. Fastqs were run on the AMR++ pipeline [7] with the Megares database v2 using the CFSAN High Performance Cluster (HPC) with default parameters. All fluoroquinolones requiring SNP confirmation were verified by identifying the SNPs conferring resistance in Tablet [44] using the AMR++ output. For annotation by AMR FinderPlus, fastq files were aligned against the AMRFinder Plus database using SAUTE [45] on the CFSAN HPC. https://github.com/ncbi/amr/wiki/Methods. Blast was used with the CARD [15] database to annotate sequence data according to default parameters. Reads were also evaluated using the COSMOS ID analytical pipeline (AMR database update July 2021 https://www.cosmosid.com). Counts and abundances from AMR annotation outputs were ‘normalized’ using scripts to assess ‘reads per kilobase of transcript’ (RPKM) to normalize gene reporting between different sites by accommodating for variation in number of sequencing reads per sample and gene length. Total reads in the sample were divided by 1,000,000 “per million” scaling factor to normalize for sequencing depth and provide ‘reads per million’ (RPM). RPM values were then divided by length of each gene in kilobases to report ‘RPKM. https://github.com/SethCommichaux/AMRplusplus

Annotation of metagenomic outputs of antimicrobial resistance

Each pipeline used to examine Sligo and Patuxent data generated annotations by different algorithmic approaches and slightly different databases and output styles. AMRFinderPlus uses Hidden Markov Models (HMMs) with customized algorithms to identify AMR genes, point mutations, stress responses, and virulence genes. AMRFinderPlus can use both protein and nucleotide sequences. Outputs include gene length and contig position information for users to further their own evaluations of the annotations [16]. The pipeline was primarily designed for use with genomes. AMRPlusPlus was designed for use with large datasets of short read metagenomic data. It handles terabyte sized data fast and accurately for count-based data. The latest update in AMRPlusPlus’s associated database, MEGARes 2.0 incorporates published resistance sequences (~8,000 hand curated) for antimicrobial drugs, metal, and biocide resistance determinants [14]. CosmosID uses kmers which are excellent for detection (sensitivity) but sometimes lacking in specificity. While Cosmos’s approach and database is proprietary, there is very clear overlap with AMRFinderPlus annotation which uses the NCBI databases. The Comprehensive Antibiotic Resistance Database (CARD) (https://card.mcmaster.ca) provides highly curated reference sequences of both nucleotides and proteins with a highly structured ontology to support analysis and prediction [15]. Plasmids were annotated using ‘Platon’ according to the default parameters [46] (https://github.com/oschwengers/platon).

Taxonomy

All taxonomic annotation was accomplished using an in-house FDA kmer pipeline developed at CFSAN, FDA, hand curated since 2014 to address pathogens monitored by FDA, available on GalaxyTRAKR http://galaxytrakr.org.

Data reporting and sharing

Reporting.

Pipeline annotation outputs were visualized using R and Graphlan [47]. Replicates of water samples were merged for certain visualizations to simplify reporting.

Data sharing.

All sequences have been deposited in NARMS Water: NARMS Water Metagenomes in the BioProject ID: PRJNA794347 http://www.ncbi.nlm.nih.gov/bioproject/794347.

Acknowledgments

We would like to acknowledge and profoundly thank the architects and scientists that support the CFSAN High Performance Cluster.

References

  1. 1. Wang Y, Lu J, Zhang S, Li J, Mao L, Yuan Z, et al. Non-antibiotic pharmaceuticals promote the transmission of multidrug resistance plasmids through intra- and intergenera conjugation. ISME J. 2021;15(9):2493–508; pmid:33692486
  2. 2. Maier L, Pruteanu M, Kuhn M, Zeller G, Telzerow A, Anderson EE, et al. Extensive impact of non-antibiotic drugs on human gut bacteria. Nature. 2018;555(7698):623–8; pmid:29555994
  3. 3. Sanganyado E, Gwenzi W. Antibiotic resistance in drinking water systems: Occurrence, removal, and human health risks. Science of The Total Environment. 2019;669:785–97; pmid:30897437
  4. 4. Wang Q, Yang Z. Industrial water pollution, water environment treatment, and health risks in China. Environmental Pollution. 2016;218:358–65; pmid:27443951
  5. 5. Zhang X-X, Zhang T, Fang HHP. Antibiotic resistance genes in water environment. Applied Microbiology and Biotechnology. 2009;82(3):397–414; pmid:19130050
  6. 6. Nguyen M, Long SW, McDermott PF, Olsen RJ, Olson R, Stevens RL, et al. Using Machine Learning To Predict Antimicrobial MICs and Associated Genomic Features for Nontyphoidal Salmonella. J Clin Microbiol. 2019;57(2); pmid:30333126
  7. 7. Tyson GH, Sabo JL, Rice-Trujillo C, Hernandez J, McDermott PF. Whole-genome sequencing based characterization of antimicrobial resistance in Enterococcus. Pathogens and Disease. 2018;76(2); pmid:29617860
  8. 8. Collignon PC, Conly JM, Andremont A, McEwen SA, Aidara-Kane A, for the World Health Organization Advisory Group BMoISoAR, et al. World Health Organization Ranking of Antimicrobials According to Their Importance in Human Medicine: A Critical Step for Developing Risk Management Strategies to Control Antimicrobial Resistance From Food Animal Production. Clinical Infectious Diseases. 2016;63(8):1087–93; pmid:27439526
  9. 9. NARMS. Evaluating the Safety of Antimicrobial New Animal Drugs with Regard to Their Microbiological Effects on Bacteria of Human Health Concern. Guidance for Industry #152. 2003;https://www.fda.gov/media/69949/download.
  10. 10. Smith CM, Hill VR. Dead-End Hollow-Fiber Ultrafiltration for Recovery of Diverse Microbes from Water. Appl Environ Microbiol. 2009;75(16):5284–9; pmid:19561183
  11. 11. Ottesen A, Ramachandran P, Chen Y, Brown E, Reed E, Strain E. Quasimetagenomic source tracking of Listeria monocytogenes from naturally contaminated ice cream. BMC Infectious Diseases. 2020;20(1):1–6. pmid:31996135
  12. 12. Ottesen A, Ramachandran P, Reed E, White JR, Hasan N, Subramanian P, et al. Enrichment dynamics of Listeria monocytogenes and the associated microbiome from naturally contaminated ice cream linked to a listeriosis outbreak. BMC Microbiology. 2016;16(1):275. pmid:27852235
  13. 13. Pettengill JB, McAvoy E, White JR, Allard M, Brown E, Ottesen A. Using metagenomic analyses to estimate the consequences of enrichment bias for pathogen detection. BMC Research Notes. 2012;5(1):378. pmid:22839680
  14. 14. Doster E, Lakin SM, Dean CJ, Wolfe C, Young JG, Boucher C, et al. MEGARes 2.0: a database for classification of antimicrobial drug, biocide and metal resistance determinants in metagenomic sequence data. Nucleic Acids Research. 2019;48(D1):D561–D9; pmid:31722416
  15. 15. Alcock BP, Raphenya AR, Lau TTY, Tsang KK, Bouchard M, Edalatmand A, et al. CARD 2020: antibiotic resistome surveillance with the comprehensive antibiotic resistance database. Nucleic acids research. 2020;48(D1):D517–D25; pmid:31665441
  16. 16. Feldgarden M, Brover V, Gonzalez-Escalona N, Frye JG, Haendiges J, Haft DH, et al. AMRFinderPlus and the Reference Gene Catalog facilitate examination of the genomic links among antimicrobial resistance, stress response, and virulence. Scientific Reports. 2021;11(1):12728; pmid:34135355
  17. 17. Yan Q, Wi YM, Thoendel MJ, Raval YS, Greenwood-Quaintance KE, Abdel MP, et al. Evaluation of the CosmosID Bioinformatics Platform for Prosthetic Joint-Associated Sonicate Fluid Shotgun Metagenomic Data Analysis. Journal of clinical microbiology. 2019;57(2):e01182–18; pmid:30429253
  18. 18. Slizovskiy IB, Mukherjee K, Dean CJ, Boucher C, Noyes NR. Mobilization of Antibiotic Resistance: Are Current Approaches for Colocalizing Resistomes and Mobilomes Useful? Frontiers in Microbiology. 2020;11(1376); pmid:32695079
  19. 19. Loftie-Eaton W, Rawlings DE. Comparative biology of two natural variants of the IncQ-2 family plasmids, pRAS3.1 and pRAS3.2. J Bacteriol. 2009;191(20):6436–46; pmid:19684126
  20. 20. Drancourt M, Carrieri P, Gévaudan MJ, Raoult D. Blood agar and Mycobacterium tuberculosis: the end of a dogma. Journal of clinical microbiology. 2003;41(4):1710–1; pmid:12682165
  21. 21. FDA. Buffered Peptone Water (BPW). Bacteriological Analytical Manual. 2005;8th Edition, Revision A.
  22. 22. Murray CJL, Ikuta KS, Sharara F, Swetschinski L, Robles Aguilar G, Gray A, et al. Global burden of bacterial antimicrobial resistance in 2019: a systematic analysis. The Lancet. 2022; pmid:35065702
  23. 23. Lynch SA, Helbig KJ. The Complex Diseases of Staphylococcus pseudintermedius in Canines: Where to Next? Vet Sci. 2021;8(1):11; pmid:33477504
  24. 24. Mader R, EU-JAMRAI , Bourély C, Amat J-P, Broens EM, Busani L, et al. Defining the scope of the European Antimicrobial Resistance Surveillance network in Veterinary medicine (EARS-Vet): a bottom-up and One Health approach. Journal of Antimicrobial Chemotherapy. 2022; pmid:35022739
  25. 25. Gweon HS, Shaw LP, Swann J, De Maio N, AbuOun M, Niehus R, et al. The impact of sequencing depth on the inferred taxonomic composition and AMR gene content of metagenomic samples. Environmental Microbiome. 2019;14(1):7; pmid:33902704
  26. 26. Commichaux S, Javkar K, Ramachandran P, Nagarajan N, Bertrand D, Chen Y, et al. Optimizing source tracking of Listeria monocytogenes with quasimetagenomics and integrated long and short read sequencing. 2020.
  27. 27. Białasek M, Miłobędzka A. Revealing antimicrobial resistance in stormwater with MinION. Chemosphere. 2020;258:127392; pmid:32947654
  28. 28. Cagle R, Ramachandran P, Reed E, Commichaux S, Mammel MK, Lacher DW, et al. Microbiota of the hickey run tributary of the Anacostia river. Microbiol Resour Announc. 2019;8(12):e00123–19. pmid:30938701
  29. 29. Che Y, Xia Y, Liu L, Li A-D, Yang Y, Zhang T. Mobile antibiotic resistome in wastewater treatment plants revealed by Nanopore metagenomic sequencing. Microbiome. 2019;7(1):44; pmid:30898140
  30. 30. Danko D, Bezdan D, Afshin EE, Ahsanuddin S, Bhattacharya C, Butler DJ, et al. A global metagenomic map of urban microbiomes and antimicrobial resistance. Cell. 2021;184(13):3376–93 e17; pmid:34043940
  31. 31. Guo X, Tang N, Lei H, Fang Q, Liu L, Zhou Q, et al. Metagenomic Analysis of Antibiotic Resistance Genes in Untreated Wastewater From Three Different Hospitals. Frontiers in Microbiology. 2021;12; pmid:34504480
  32. 32. Haenni M, Dagot C, Chesneau O, Bibbal D, Labanowski J, Vialette M, et al. Environmental contamination in a high-income country (France) by antibiotics, antibiotic-resistant bacteria, and antibiotic resistance genes: Status and possible causes. Environment International. 2022;159:107047; pmid:34923370
  33. 33. Hendriksen RS, Munk P, Njage P, van Bunnik B, McNally L, Lukjancenko O, et al. Global monitoring of antimicrobial resistance based on metagenomics analyses of urban sewage. Nature Communications. 2019;10(1):1124; pmid:30850636
  34. 34. Markkanen MA, Haukka K, Pärnänen KMM, Dougnon VT, Bonkoungou IJO, Garba Z, et al. Metagenomic analysis of antimicrobial resistance genes in wastewaters in Benin and Burkina Faso indicates a serious health risk from untreated hospital wastewaters in low-income countries. medRxiv. 2022:2021.10.19.21265183;
  35. 35. Niegowska M, Sanseverino I, Navarro A, Lettieri T. Knowledge gaps in the assessment of antimicrobial resistance in surface waters. FEMS Microbiol Ecol. 2021;97(11); pmid:34625810
  36. 36. Rowe W, Verner-Jeffreys DW, Baker-Austin C, Ryan JJ, Maskell DJ, Pearce GP. Comparative metagenomics reveals a diverse range of antimicrobial resistance genes in effluents entering a river catchment. Water Science and Technology. 2015;73(7):1541–9; pmid:27054725
  37. 37. Yin X, Deng Y, Ma L, Wang Y, Chan LYL, Zhang T. Exploration of the antibiotic resistome in a wastewater treatment plant by a nine-year longitudinal metagenomic study. Environment International. 2019;133:105270; pmid:31683155
  38. 38. CDC. Waterborne Disease in the United States. Waterborne Disease & Outbreak Surveillance Reporting. 2014.
  39. 39. Organization WH. Drinking Water. Fact sheets.
  40. 40. Laxminarayan R. The overlooked pandemic of antimicrobial resistance. The Lancet. 2022;399(10325):606–7; pmid:35065701
  41. 41. WHO. HIV/AIDS. 2021.
  42. 42. WHO. Malaria. 2021.
  43. 43. Bolger AM, Lohse M, Usadel B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics. 2014;30(15):2114–20; pmid:24695404
  44. 44. Milne I, Bayer M, Stephen G, Cardle L, Marshall D. Tablet: Visualizing Next-Generation Sequence Assemblies and Mappings. In: Edwards D, editor. Plant Bioinformatics: Methods and Protocols. New York, NY: Springer New York; 2016. p. 253–68.
  45. 45. Souvorov A, Agarwala R. SAUTE: sequence assembly using target enrichment. BMC Bioinformatics. 2021;22(1):375; pmid:34289805
  46. 46. Schwengers O, Barth P, Falgenhauer L, Hain T, Chakraborty T, Goesmann A. Platon: identification and characterization of bacterial plasmid contigs in short-read draft assemblies exploiting protein sequence-based replicon distribution scores. Microb Genom. 2020;6(10):mgen000398; pmid:32579097
  47. 47. Asnicar F, Weingart G, Tickle TL, Huttenhower C, Segata N. Compact graphical representation of phylogenetic data and metadata with GraPhlAn. PeerJ. 2015;3:e1029–e; pmid:26157614