Citation:

Hill, R., M. Jahne, S. Keely, N. Brinkman, R. Haugland, S. Leibowitz, E. Wheaton, J. Garland, AND R. Martin. Modeling and predicting the occurrences of antibiotic resistance genes in US rivers and streams. Annual Meeting of the Society for Freshwater Science, Salt Lake City, Utah, May 19 - 23, 2019.

Impact/Purpose:

Antimicrobial resistance (AMR) of pathogens is a critical human health threat. Human activity on landscapes enhances AMR spread, and rivers often have detectable concentrations of antibiotic-resistance genes (ARGs). We developed models to predict the distributions of four ARGs (sul1, blaTEM, tetW, and KPC), a mobile genetic element (intI1) that facilitates the spread of ARGs, and two pathogen indicators (entercocci and Escherichia coli) in US rivers and streams with data from the 2013-2014 National Rivers and Streams Assessment and 50 predictor variables from the US EPA’s StreamCat dataset. We applied the models back to the StreamCat dataset to predict probability of target gene occurrences at 1.1 million streams segments across the conterminous US. Most studies of AMR development and spread have been based on opportunistic sampling (e.g., upstream and downstream of local wastewater treatment plants) and are limited in their spatial extent or range of environmental conditions over which they were conducted. The current study is a spatially extensive analysis of genes associated with AMR across the conterminous US and provides an unprecedented view of this threat. We showed that ARGs are highly responsive to human-related watershed stressors. Their occurrence may be moderated by improving watershed integrity or maintaining flow from groundwater sources. Some ARGs are highly predictable and can be mapped with a reasonable degree of certainty. The current study also includes maps that could prove critical for focusing field surveillance and mitigation of emerging AMR threats. In addition, the five-year cycle of the NRSA will allow monitoring of the spread of AMR and early warning detection of new threats. This research supports the development of robust national maps of stream conditions, which is of interest to the Monitoring Branch within the Office of Water.

Description:

Antimicrobial resistance (AMR) of pathogens is a critical threat to human and animal health. Human activities enhance the spread of AMR and AMR can be detected in streams by PCR targeting antibiotic resistance genes (ARGs). Understanding anthropogenic drivers of this spread is critical to develop effective monitoring and mitigation. Here, we identified potential ecological drivers to model and predict the probability of occurrence of four ARGs (sul1, tetW, blaTEM, and KPC), a gene associated with AMR mobilization (intI1), and two potential pathogenic indicators (E. coli and enterococci). We paired gene occurrences from ~2,000 stream samples across the US (USEPA NRSA) with several watershed attributes, including urbanization and agriculture (StreamCat dataset). We used random forests to model gene occurrences in response to watershed land use intensity. The models correctly predicted ARGs at 70%-82% of sites but varied in their ability to balance type I and II errors. We then predicted ARG occurrences at 1.1 million stream reaches. The resulting maps reflect the relative importance of urbanization, agriculture, or other landscape stressors on the occurrence of each gene and can indicate where focused surveillance and mitigation of AMR may be needed.

Record Details: