Metagenomic Insight into The Global Dissemination of The Antibiotic Resistome

Abstract

The global crisis in antimicrobial resistance continues to grow. Estimating the risks of antibiotic resistance transmission across habitats is hindered by the lack of data on mobility and habitat-specificity. Metagenomic samples of 6092 are analyzed to delineate the unique core resistomes from human feces and seven other habitats. This is found that most resistance genes (≈85%) are transmitted between external habitats and human feces. This suggests that human feces are broadly representative of the global resistome and are potentially a hub for accumulating and disseminating resistance genes. The analysis found that resistance genes with ancient horizontal gene transfer (HGT) events have a higher efficiency of transfer across habitats, suggesting that HGT may be the main driver for forming unique but partly shared resistomes in all habitats. Importantly, the human fecal resistome is historically different and influenced by HGT and age. The most important routes of cross-transmission of resistance are from the atmosphere, buildings, and animals to humans. These habitats should receive more attention for future prevention of antimicrobial resistance. The study will disentangle transmission routes of resistance genes between humans and other habitats in a One Health framework and can identify strategies for controlling the ongoing dissemination and antibiotic resistance.

Keywords: antibiotic resistome; horizontal gene transfer; metadata; metagenome; one health.

Figure 1

Historical variation and driving factor of human fecal core resistomes. A) We collected metagenomic data from palaeofeces (n = 20) and modern‐human feces (n = 3018), and found that the prevalence of modern‐human fecal core ARGs was extremely low in palaeofeces. B) and C) The dominant ARGs in palaeofeces (>1 abundance (reads per kilobase per million mapped reads, RPKM) and >60% frequency), and the associated classification. A&A, aminoglycoside and aminocoumarin antibiotic; N&A, nucleoside antibiotic and acridine dye. D) Machine learning with the random‐forest algorithm was used to determine the importance of optimal factors (VIFs < 5) driving the core resistomes. Age was the most important factor to drive the core resistomes in modern human feces. GNI, gross national income; VIFs: variance inflation factors.

Figure 2

Variation of the core resistome across all habitats. A) We collected 3562 metagenomic datasets from the air (413), aquatic (471), buildings (508), invertebrate (404), plant (346), terrestrial (493), vertebrate, (419) and human feces (508). B) Principal coordinate analysis with Bray‐Curtis dissimilarity and Adonis analysis showed that the pattern of resistomes from different habitats is clearly separated from human feces. Red and blue circles indicate human feces and other habitats, respectively. C) The classification of the core resistome from different external habitats. D) Sharded and unique core ARGs between human feces and other habitat samples.

Figure 3

Shared pattern of resistomes across human feces and the other habitats. A) Network of resistomes shared between human feces and the external habitats (air, terrestrial habitats, aquatic habitats, buildings, invertebrates, vertebrates and plants). A total of 708 ARGs were shared among all habitats, and human feces had no unique ARGs. B) The abundance (RPKM) of shared ARGs varied considerably across habitats. C) Fast expectation‐maximization for microbial source tracking (FEAST) estimating the source contribution of human fecal resistome to the different habitats. These shared ARGs mainly conferred beta‐lactams resistance; most of the ARGs (≈85%) in other habitats were sourced from human fecal resistome. D) The total abundance of shared ARGs in human feces and the other habitats. E) The antibiotic resistance risk of samples from various habitats. Different letters represent significant differences between habitats (Kruskal‐Wallis test, adjust p < 0.05).

Figure 4

Habitat filtering for the ARG hosts. A) Phylogenetic taxonomic network of ARG hosts in human feces. A total of 720 ARG hosts (species level) were identified based on the 177134 metagenome‐assembled genomes (MAGs) of the human fecal microbiome collected from 33 countries across six continents compiled by Almeida et al⁴⁵. The color and size of the circles indicate the taxonomic level and ARG numbers (per species), respectively, and the width of the lines indicates the richness of the taxonomic levels. Clostridia, Gammaproteobacteria, Bacilli, and Bacteroidetes were the main ARG hosts in human feces. B) Principal coordinate analysis with Bray‐Curtis dissimilarity showed the variation pattern of ARG hosts between human feces and other habitats. Adonis analysis indicates the ARG hosts from each habitat were significantly separated from human feces (“*”, adjust p < 0.05). C) The relative abundance of the pathogenic ARG hosts from human feces in the other habitats. Based on the database of Pathogen Host Interactions and the list of antibiotic‐resistant “priority pathogens” (World Health Organization), we defined the pathogenic ARG hosts as R1 (Priority 1: CRITICAL), R2 (Priority 2: HIGH), R2 (Priority 3: MEDIUM), and Rn (other human pathogens). Different letters represent significant differences between the relative abundance (Kruskal‐Wallis test, adjust p < 0.05).

Figure 5

Pattern of dissemination of resistomes across human feces and the other habitats. A) Based on the list of 720 ARG hosts from human feces, we collected 10 274 bacterial genomes from the various habitats based on the database of the National Center for Biotechnology Information to construct the ARG exchange network with the global shared resistome (n = 400). We used Blast hits with 100% similarity and lengths >500 bp. B) We obtained 5555932 horizontal gene transfers (HGTs), including 2, 088998 genomic pairs, and 191/414 transferable ARGs. C) The proportions of HGT efficiency between and within species in the various habitats. D) The rate of detection of ARGs between and within species in the various habitats. E) We collected 2090 E. coli genomes to calculate the efficiency of transfer within species across the external habitats and human feces (See “Method”). The cross‐transmission routes between human feces and air, invertebrates and plants had higher HGT efficiencies within E. coli strains than within human feces; the efficiency of transfer within E. coli was higher in invertebrates and plants than the other external habitats. F) The number of single nucleotide polymorphisms (SNPs) for each transferable ARG in different routes of transmission. The average number of SNPs for transferable ARGs in genomic pairs from cross‐transmission routes between human feces and air, invertebrates and plants was lower than for the other habitats.

Figure 6

Overview of this study. We collected three datasets including metagenomics (n = 8061), metagenome‐assembled genomes (MAGs, n = 177134) and bacterial genomes (n = 10274) from paleofeces, modern‐human feces, buildings, air, aquatic, and terrestrial, plants, vertebrates and invertebrates. First, we determined the historical and habitat variations of human fecal resistome, identified core resistome from various habitats and quantified the ARGs risk of different habitats by using a metagenomic dataset. Second, we identified the ARG host (n = 720) in human feces by using MAG datasets and evaluated their risk in various habitats. Finally, we calculated the HGT efficiency of ARGs (n = 400) across various habitats to determine the transmission risk by using the bacterial genomes. The ongoing process of acquiring genes by HGT has overcome and altered the ecological and phylogenetic barriers to such horizontal transfers, probably by the co‐selection of promiscuous mobile genetic elements that carry resistance genes as cargo.