Overcoming challenges in metagenomic AMR surveillance with nanopore sequencing: a case study on fluoroquinolone resistance

Abstract

Introduction: Antimicrobial resistance is an alarming public health problem, and comprehensive surveillance across environments is required to reduce its impact. Phenotypic testing and whole-genome sequencing of isolates are efficient, but culture-free approaches like metagenomic sequencing potentially allow for broader investigation of resistance gene occurrence, evolution and spread. However, technical challenges such as difficulties in associating antimicrobial resistance genes with their bacterial hosts and the collapse of strain-level variation during metagenome assembly, hinder its implementation.

Methods: To illustrate how these challenges can be overcome, we applied Oxford Nanopore Technologies long-read metagenomic sequencing and novel bioinformatic methods to a case study focused on fluoroquinolone resistance in chicken fecal samples.

Results: We demonstrate plasmid-host linking based on detecting common DNA methylation signatures. Additionally, we use new bioinformatic approaches for strain haplotyping, enabling phylogenomic comparison and uncovering fluoroquinolone resistance determining point mutations in metagenomic datasets.

Discussion: We leverage long-read sequencing, including DNA methylation profiling and strain-level haplotyping, to identify antimicrobial resistance gene hosts, link plasmids to their bacterial carriers, and detect resistance-associated point mutations. Although some limitations remain, our work demonstrates how these improvements in metagenomic sequencing can enhance antimicrobial resistance surveillance.

Keywords: DNA methylation; antimicrobial resistance (AMR); metagenomic sequencing; nanopore sequencing; plasmid host prediction; strain-resolved metagenomics.

Figure 1

Overview of metagenomic antimicrobial esistance (AMR) investigation. We sampled chicken feces from farms, added spike-in control bacteria, and used Oxford Nanopore Technologies sequencing with methylation calling to generate long reads with methylation base tags. Antimicrobial resistance genes (ARGs) were then identified using a read-based approach (A), and through metagenomic assembly and binning (assembly-based approach) (B). For the read-based approach, ARG-carrying reads were also retrieved from their taxonomic alignment for host linking. For the assembly-based approach, methylation motifs were used for bin improvement and plasmid-host linking (C). Strain haplotyping then allowed for phylogenomic investigation and detection of strain-level AMR, including point mutations associated with AMR, which we did for fluoroquinolone resistance (D).

Figure 2

(A, B) Resistome profile in terms of logarithmically scaled depth of coverage per antimicrobial resistance gene (ARG) for samples A and B, respectively, with fluoroquinolone (FQ) ARGs indicated in bold on the x-axis. (C, D) ARG-host links as determined via ARG detection in reads and read taxonomic alignment. FQ ARGs and their hosts are highlighted with bold nodes, bold text, and highlighted edges. In (C), both I. halotolerans and A. halotolerans are indicated as carrying an ARG. In (D), A. halotolerans is reported as ARG host as well. MLS-B: macrolide, lincosamide, streptogramin-B.

Figure 3

Contig methylation profiles in sample A, visualized in hierarchically clustered heatmaps (A, C) and UMAP plots (B, D). Colors next to the rows (representing contigs) in (A) and (C) indicate the metagenomic bins as determined by the Nanomotif-refined binning method, while columns represent methylation motifs. Cell colors represent methylation percentage. In the UMAP plots (B, D), contigs are shaped according to their genomad classification, and colored by bin [cfr (A, C)]. In all plots, motifs were only included if they were at least 50% methylated in at least a single contig. Bins with completeness <10% are excluded, and for each bin the largest 10 contigs are shown. The upper plots (A, B) focus on the unbinned plasmids (contig 847, 1,029, and 640, gray) and the bins with the most similar methylation profiles to these plasmid: E. coli (light red) and K. pneumoniae (green). The qnrS1-carrying plasmid [contig 847 in (A) and indicated in (B)] most closely resembles the E. coli genome in terms of methylation [contig 501 in (A), indicated in (B)]. Contig 866 (light red) is binned as E. coli, but clusters more closely with K. pneumoniae contigs in (A, B). The lower plots (C, D) show the overall methylation profiles of the metagenomic assembly. A legend shows al bins by color. Some bins (e.g., contig_bins_binned.tsv_23_sub) were >10% complete, but could not be classified to species or genus level using gtdb-tk.

Figure 4

SNP-based phylogeny of E. coli ST10 isolates obtained from multiple farms and the sample A E. coli MAG. As reference sequence, a hybrid assembly of Farm A Isolate 3 was used, from which plasmids were removed. All Farm A isolates (red branch) cluster closely together, contain a similar antimicrobial resistance gene (ARG) profile and carry an IncFII, MOBP plasmid with the qnrS1 gene. All ARGs in the farm A isolates were carried on plasmids. The ILMN: Illumina short reads, ONT: Oxford Nanopore Technologies long reads. Branch length and three scale represent average number of substitutions per site. The numbers on the branches represent bootstrap values.