SELECT 2.0: Refined and open access SELection Endpoints in Communities of bacTeria (SELECT) method to determine concentrations of antibiotics that may select for antimicrobial resistance in the environment

Antimicrobial resistance (AMR) is a significant and growing threat to human, plant and animal health, the global economy, and food security. The One Health approach to AMR recognises the role of the environment in the evolution, emergence, and dissemination of AMR. In part, this is due to anthropogenic pollution that releases AMR organisms alongside cocktails of compounds that may select for AMR in situ, which then pose an exposure risk to humans and animals. This has spurred growing interest from cross-sectoral stakeholders in environmental risk assessment (ERA) of antibiotics, with regards to their selective potential. Many different experimental and modelling approaches have been used to determine the lowest concentration of an antibiotic that may select for AMR. Debates continue regarding which individual approach, if any, may be best for determining concentrations of antibiotics that may select for AMR, for ERA purposes. This paper contributes to this ongoing discourse by refining and using a previously published method SELECT (SELection Endpoints in Communities of bacTeria) to rapidly generate predicted no effect concentrations for resistance (PNECRs) for 32 antibiotics on the premise that reduction in growth of complex community of bacteria correlates with selection for AMR resistance genes. The database of PNECRs of antibiotics presented here is the largest generated using a single experimental, empirical approach that will aid future efforts towards creating a standardised test. PNECR data were used to conduct ERAs using measured environmental concentrations of antibiotics to rank antibiotics by potential selection risk in different environments. The experimental approach and statistical code have been made open access, with online tutorials available to facilitate other laboratories using the SELECT 2.0 method. Finally, we discuss the limitations of this approach and how these could be addressed in future studies.