Antimicrobial resistance (AMR) poses a global threat to human, public, and animal health and predicting the evolution, persistence and transmission of AMR has been a mainstay challenge. Shotgun metagenomic sequencing helps overcome this by enabling characterization of resistance genes within all bacterial taxa, including those uncultivatable using standard laboratory methods. In this way, shotgun sequencing provides a more comprehensive assessment of the AMR ‘potential’ within samples, i.e., the “resistome”. The data-dense metagenomic assessment of the resistome, however, encumbers a clear understanding of how and where resistance likely develops and spreads. Recently, it has been proposed that risk stratification of all AMR determinants of a sample are most informative if their profile is considered in light of the mobilizable genomic milieu flanking or encompassing microbial resistomes. This includes mobility factors such as plasmids, integrative conjugative elements (ICE), insertional sequences, transposons, and other mobile genetic elements (MGEs). As these components are responsible for horizontal gene transfer (HGT) and potentially the uptake of resistance by pathogenic species, investigators are turning their attention to studying that portion of the resistome closely associated with HGT elements, i.e., the “mobilome”. As metagenomics is poised to be at the forefront of characterizing resistance mobility potential, there is currently an inexorable need to perform resistome-mobilome colocalization analyses. In this study, we explored currently available colocalization approaches with a focus on alignment-based methods as well as co-occurrence analysis using assembly-based techniques, using default or common literature-supported parameters. We analyzed a clinical (human) and an agricultural (cattle) publicly available fecal metagenomic dataset, obtained from trials employing antimicrobials in individuals sampled over time. For human and cattle datasets, AMR and MGE input data for colocalization analysis starkly differed in gene class richness, depending on alignment and assembly techniques used. Ordination revealed that tulathromycin use in cattle was associated with a shift in ICE and plasmid composition relative to untreated animals, the resistome was not significantly impacted (ANOSIM P >0.05) during the 11-day monitoring period. Contrarily, in the human dataset, while ordination of resistome and ICE, plasmid, and prophage components of the mobilome showed a shift shortly after the administration of antimicrobials (ANOSIM P <0.01), the composition rebounded to pre-treatment levels. However, Procrustes biplot analysis of superimposed AMR and MGE ordinations indicates significant compositional correlation between resistomes and mobilomes in the face of antimicrobial exposure, for both datasets. Co-occurring AMR and MGE genes identified with Bayesian networks representing > 70% bootstraps accounted for 19% of edges of the human network and 2% of edges of the cattle network. Conversely, using the Mobility Index (MI) at the level of the metagenomic sample (defined as proportion of all present AMR-containing contigs with flanking MGE sequences) colocalizations identified from de novo assembly indicates that AMR-MGE co-occurrence increases shortly after exposure to antibiotics within the human metagenome (up to 75%), however >40 days after peak antimicrobial exposure, such contigs were rare (~2 %). For the cattle metagenome, MI was not altered by antimicrobial exposure, ranging 0.5–4.0%. Our results highlight that alignment-based and assembly based techniques used in colocalization will yield often contradictory and incomplete conclusions about resistance mobility, and that current bioinformatic approaches are limited by technical and computational challenges that prevent reliable colocalization analysis. We conclude by discussing development of laboratory, sequencing, and computational methods that may be useful in contextualizing that portion of the resistome most likely to be mobilizable, and therefore, enhance relevance of metagenomic resistome analysis in clinical, regulatory, and commercial applications.
BLAST parsing scripts for colocalization analysis using metagenomic short-read assemblies
Sponsorship: National Institute of Allergy and Infectious Diseases (NIAID) of the U.S. National Institutes of Health (NIH), project no. 1R01AI141810-01