Targeted metagenomics to track and characterize the antimicrobial resistome (RESCAP1.0)

Abstract

[Background]: Metagenomics has recently been used to analyze the ensemble of antimicrobial resistance genes in a particular environment (resistome). All metagenomic open and closed formats exhibit low sensitivity and lack of specificity to uncover the size and diversity of minor fractions of total metagenomes. To achieve better sensitivity and specificity in resistome mining, we developed a targeted sequence capture panel (ResCap v.01) based on SeqCapEZ technology (NimblegeneRoche), designed to enrich sequences from genes encoding resistance to antibiotics, biocides and heavy metals. [Material & Methods]: ResCap is composed by a panel of approximately 79,000 non-redundant genes (80 Mb) including 7,963 well-known antibiotic resistance genes, 30,740 biocide & heavy metal resistance genes, and 2,517 relaxases, signature proteins of plasmid conjugation. Design of the probes was based on a non-redundant database (DB), constructed by mixing curated DBs for antibiotic resistance (CARD, ARG-ANNOT and RED-DB). To improve the detection of “novel” AbR genes, a Hidden Markov Model (HMM) from each AbR family was constructed for detection of homologous sequences by using hammer3 software against Uniref100 DB. Sequences coding for resistance to heavy metals and antiseptics were recruited from BacMet DB, which includes both experimentally confirmed and predicted antibacterial biocide- and metal-resistance genes. Relaxases were retrieved from the ConjDB repository. ResCap workflow consisted on i) whole-metagenome shotgun library construction, ii) hybridization, and iii) capture. All steps were performed according Nimblegene standard protocols for Illumina platforms. To evaluate ResCap efficiency, samples were sequenced before and after capture. ResCap was validated by analysing fecal samples from 9 humans and 8 swine. Robustness of the platform was tested by comparative analysis of two technical replicates of two samples. Bioinformatic analysis was performed by mapping sequenced reads against well-known gene DBs (ARG-ANNOT,BacMet Experimental and ConjDB) using Bowtie2 with the output option of all hits (-a). All statistical operations were performed using R and homemade Perl scripts. [Results]: An average of 1.9x107 reads were obtained from the ResCap pool (9,2-32 106). The average on-target hits was 0,11% (0,07%-0.18%) for pre-captured samples and 30% (20%-41%) for post-captured sequences. The enhanced sensitivity of ResCap represents a gain of 279 fold (170-480 fold). The diversity of genes encoding antimicrobial resistance and relaxases was 627 (436-937) in pre-capture samples and 1367 (1071-1616) in samples captured with ResCap (Figure 1). Reproducibility was addressed by the correlation between replicates, showing R factors of 0.81 and 0.96, respectively. [Conclusion]: ResCap v0.1 substantially enhances the sensitivity and specificity of metagenomics shotgun sequencing for analyzing the resistome. Moreover, the ability to accurately detect low abundant or rare populations and its robustness, make ResCap a promising tool for a wide number of research and diagnostic applications and suitable for being used in both longitudinal and cross-sectional studies The platform constitutes one of the first examples of targeted metagenomics for analysing bacterial populations.