Co-occurrence of resistance genes to antibiotics, biocides and metals reveals novel insights into their co-selection potential

Abstract

Background: Antibacterial biocides and metals can co-select for antibiotic resistance when bacteria harbour resistance or tolerance genes towards both types of compounds. Despite numerous case studies, systematic and quantitative data on co-occurrence of such genes on plasmids and chromosomes is lacking, as is knowledge on environments and bacterial taxa that tend to carry resistance genes to such compounds. This effectively prevents identification of risk scenarios. Therefore, we aimed to identify general patterns for which biocide/metal resistance genes (BMRGs) and antibiotic resistance genes (ARGs) that tend to occur together. We also aimed to quantify co-occurrence of resistance genes in different environments and taxa, and investigate to what extent plasmids carrying both types of genes are conjugative and/or are carrying toxin-antitoxin systems.

Results: Co-occurrence patterns of resistance genes were derived from publicly available, fully sequenced bacterial genomes (n = 2522) and plasmids (n = 4582). The only BMRGs commonly co-occurring with ARGs on plasmids were mercury resistance genes and the qacE∆1 gene that provides low-level resistance to quaternary ammonium compounds. Novel connections between cadmium/zinc and macrolide/aminoglycoside resistance genes were also uncovered. Several clinically important bacterial taxa were particularly prone to carry both BMRGs and ARGs. Bacteria carrying BMRGs more often carried ARGs compared to bacteria without (p < 0.0001). BMRGs were found in 86 % of bacterial genomes, and co-occurred with ARGs in 17 % of the cases. In contrast, co-occurrences of BMRGs and ARGs were rare on plasmids from all external environments (<0.7 %) but more common on those of human and domestic animal origin (5 % and 7 %, respectively). Finally, plasmids with both BMRGs and ARGs were more likely to be conjugative (p < 0.0001) and carry toxin-antitoxin systems (p < 0.0001) than plasmids without resistance genes.

Conclusions: This is the first large-scale identification of compounds, taxa and environments of particular concern for co-selection of resistance against antibiotics, biocides and metals. Genetic co-occurrences suggest that plasmids provide limited opportunities for biocides and metals to promote horizontal transfer of antibiotic resistance through co-selection, whereas ample possibilities exist for indirect selection via chromosomal BMRGs. Taken together, the derived patterns improve our understanding of co-selection potential between biocides, metals and antibiotics, and thereby provide guidance for risk-reducing actions.

Fig. 1

Overview of resistance information in genomes and plasmids, and relative proportions of ARG-carrying plasmids and genomes. Resistance gene profiles of a 2522 genomes irrespective of location on chromosomes or plasmids, b 1926 plasmids harboured by the 2522 genomes and c 4582 plasmids, respectively. d Relative proportions of ARG-carrying plasmids with or without BMRGs on the same plasmid, and e Relative proportions of ARG-carrying genomes with or without BMRGs in the same genomes, either on a plasmid or chromosome

Fig. 2

Overview of resistance information of 4582 plasmids found across bacterial genera. Only the 25 genera hosting the highest number of resistance plasmids are shown. Each individual section of a bar represents the proportion that carries different types of resistance genes. The green section in each bar indicates the number of plasmids that have both BMRGs and ARGs present on the same plasmid, and thus potential for co-selection

Fig. 3

Overview of resistance information from genomes and plasmids from different environments. Resistance information from (a) 2522 completely sequenced bacterial genomes and (b) 1926 plasmids harboured by those genomes from different environments. Each individual section of a bar represents the proportion that carries different types of resistance genes. The green section of a bar represents the proportions of genomes (a) or plasmids (b) that have both ARGs and BMRGs present in the same strain (a) or on the same plasmid (b), thus representing the potential for co-selection

Fig. 4

Relative abundance of biocide/metal resistance genes across environments. a Chromosomal and b Plasmid-borne biocide/metal resistance genes per genome

Fig. 5

Resistance gene characteristics of 4582 plasmids in relation to their size. The individual sections of each bar represent the number of plasmids that carry different resistance gene types. The green section of each bar represents the number of plasmids with both ARGs and BMRGs on the same plasmid, and thus the potential for co-selection

Fig. 6

Distribution of plasmids with or without co-selection and mobility potential relative to their size. a Distribution of plasmids carrying both ARGs and BMRGs. b Distribution of plasmids carrying either ARGs or BMRGs or no resistance genes, thus lacking potential for co-selection of resistance based on co-occurrences. c Distribution of conjugative plasmids. d Distribution of non-conjugative plasmids (i.e. non-transmissible or mobilizable)

Fig. 7

Overview of mobility potential and toxin-antitoxin systems on plasmids with different combinations of resistance genes. a Mobility potential (i.e. presence or absence of conjugations systems) in plasmids with different combinations of resistance genes. b Toxin-antitoxin (TA) systems in plasmids with different combinations of resistance genes

Fig. 8

Co-occurrence network of resistance genes on plasmids. The network was built based on the observed co-occurrence pattern of BMRGs with ARGs on 4582 plasmids. The network was filtered such that only connections between BMRGs and ARGs, and between integrases/transposases and resistance genes were kept if the connected genes occurred together on at least 10 plasmids. No connections are shown between ARGs to better emphasize the co-selection potentials. The thickness of each connection (edge) between two resistance genes (nodes) is proportional to the number of times the two resistance genes co-occurred on the same plasmids

Fig. 9

Co-occurrence network of resistance genes in genomes. The network was built based on the observed co-occurrence pattern of BMRGs with ARGs in 2522 completely sequenced bacterial genomes (2666 chromosomes and 1926 plasmids). For BMRGs, only a subset of resistance genes towards metals (i.e. mercury, arsenic, cadmium, copper, zinc, bismuth, antimony and silver) was used in order to simplify the network and make trends visible. The network was filtered such that only connections between BMRGs and ARGs, and between integrases/transposases and resistance genes were kept if the connected genes occurred together in at least ten genomes (i.e. strains) irrespective of their location (chromosomes and/or plasmids). For clarity and illustration purposes, the connections between BMRGs are not shown. No connections are shown between ARGs to better emphasize the co-selection potentials. The thickness of each connection (an edge) between two resistance genes (nodes) is proportional to the number of times the two resistance genes co-occurred in the same genomes. ‘ARG + BMRG’ refers to the genes that confer resistance/tolerance to both antibiotics and biocides/metals