Costs of antibiotic resistance genes depend on host strain and environment and can influence community composition

Antibiotic resistance genes (ARGs) benefit host bacteria in environments containing corresponding antibiotics, but it is less clear how they are maintained in environments where antibiotic selection is weak or sporadic. In particular, few studies have measured the effect of ARGs on host fitness in the absence of direct selection or determined if any costs are fixed or depend on the host strain, perhaps marking some ARG-host combinations as reservoirs that can maintain ARGs in the absence of antibiotic selection. We quantified the fitness effects of six ARGs in 11 diverse Escherichia spp. strains. Three ARGs (blaTEM-116, cat, and dfrA5, encoding resistance to β-lactams, chloramphenicol, and trimethoprim, respectively) imposed an overall cost but all ARGs had an effect in at least one host strain, reflecting a significant strain interaction effect. A simulation predicts these interactions cause the success of ARGs to depend on available host strains, and, to a lesser extent, for successful host strains to depend on the ARGs present in a community. These results indicate the importance of considering ARG effects over different host strains, especially the potential of reservoir strains that allow resistance to persist in the absence of direct selection, in efforts to understand resistance dynamics.

Methods
To measure the fitness effect of each ARG, we performed a series of competition assays. The fitness effect of an ARG was estimated as the fitness difference between cells carrying an ARG plasmid (WARG) and cells carrying the control pmFP plasmid vector (WVec). The fitness of ARG and vector plasmid-carrying cells was estimated indirectly by competing each against reference cells carrying a GFP expressing plasmid (pUA66-PrpsLGFP).

To measure the copy number of the pmFP::blaTEM116* plasmid we used qPCR. Primers were designed to amplify the chromosomal dxs gene, in order to determine the absolute Ct value of the bacterial chromosome (Ct_ch) and the pUA66 plasmid encoded aph(3’)-IIa gene, to determine the Ct value of the pmFP::blaTEM116* plasmid (Ct_p). The relative copy number of the plasmid to the chromosome (ΔCt) was calculated as 2^-(Ct_p - Ct-ch). The ΔCt of different bacterial strains was then normalized to the lab strain, REL606, to obtain the ΔΔCt value.