Inferring antibiotic resistance selection in the environment can be confounded by correlations between resistance genes and unrelated functional traits

Antimicrobial resistance (AMR) is a silent pandemic that is coupled with other crises such as climate change in the polycrisis humanity is facing. One of the key questions is whether antibiotic resistance genes (ARGs) are selected for at the low antibiotic concentrations typical for most environments. Many studies have observed changes in the relative abundance of ARGs from one environmental compartment to the next, e.g. from wastewater treatment plant influent to effluent. Fewer studies have directly tested for selection by incubating environmental samples in mesocosms or laboratory models at different concentrations of antibiotics to infer minimal selective concentrations. We developed a mathematical model to demonstrate that these studies can be confounded by shifts in the microbial community composition that occur when a microbiome is transported from one environmental compartment to another or when incubated under different conditions. Such community shifts will confound tests of selection when there is an association between carriage of ARGs and other functional traits. As an example, we show that there is a phylum-dependent association between the number of ARGs and the number of ribosomal RNA genes, which are both higher in fast growing, copiotrophic bacteria. We then show that specific growth rate or nutrient concentration upshifts increased the proportion of copiotrophs in the community and thus the relative abundance of ARGs. This result generalizes to community shifts for other reasons if there is some association between ARGs and ecological niches. Therefore, most studies of selection for ARGs in the environment are confounded. Solutions are proposed.