Linking environmental antimicrobial resistance to human acquisition: evidence, knowledge gaps and a mechanistic framework

The contribution of environmental reservoirs to antimicrobial resistance (AMR) acquisition in humans remains poorly quantified, despite extensive evidence that resistant bacteria and resistance genes circulate through wastewater, surface waters, food and other environmental matrices. The central difficulty is no longer to demonstrate that these reservoirs are connected, but to quantify the direction and relative contribution of the pathways linking environmental AMR to human acquisition. Here, we synthesize the main processes that need to be represented at the interface between the environment and humans and propose a mechanistic framework for Escherichia coli producing extended spectrum beta lactamases (ESBL Ec), linking human, water and food compartments. The framework explicitly represents contamination of the environment by humans and multiple routes of human exposure to environmental reservoirs, including drinking water, recreational water exposure and food contamination mediated by irrigation. We apply the model to two contrasting illustrative settings, France and Madagascar, and use scenarios involving antibiotic use, environmental contamination and climate to explore how changes in specific drivers propagate across compartments. The simulations show responses that depend strongly on the compartment, while sensitivity analyses identify wastewater treatment efficiency and parameters describing transmission, exposure and environmental selection at specific interfaces as influential components of the system. The framework is used to explore how different transmission pathways and environmental drivers can shape AMR dynamics across interconnected human and environmental compartments. By comparing contrasting scenarios and identifying the parameters that most strongly influence model behaviour, we highlight both potential leverage points and critical gaps in current empirical knowledge. More broadly, this work illustrates how decomposing One Health AMR into tractable reservoir interfaces can help structure future data collection and progressively improve our ability to quantify transmission between reservoirs.