Geographics and bacterial networks differently shape the acquired and latent global sewage resistomes

Abstract

Antimicrobial resistance genes (ARGs) have rapidly emerged and spread globally, but the pathways driving their spread remain poorly understood. We analyzed 1240 sewage samples from 351 cities across 111 countries, comparing ARGs known to be mobilized with those identified through functional metagenomics (FG). FG ARGs showed stronger associations with bacterial taxa than the acquired ARGs. Network analyses further confirmed this and showed potential for source attribution of both known and novel ARGs. The FG resistome was more evenly dispersed globally, whereas the acquired resistome followed distinct geographical patterns. City-wise distance-decay analyses revealed that the FG ARGs showed significant decay within countries but not across regions or globally. In contrast, acquired ARGs showed decay at both national and regional scales. At the variant level, both ARG groups had significant national and regional distance-decay effects, but only FG ARGs at a global scale. Additionally, we observed stronger distance effects in Sub-Saharan Africa and East Asia compared to North America. Our findings suggest that differential selection and niche competition, rather than dispersal, shape the global resistome patterns. A limited number of bacterial taxa may act as reservoirs of latent FG ARGs, highlighting the need of targeted surveillance to mitigate future resistance threats.

Fig. 1. Abundance and beta-diversity of ARGs across geographical regions.

a The country-wise ALR abundances of acquired ARGs and the FG ARGs. b, c PCA biplots of resistance genes (98% homology grouping), in which the PCA loadings were calculated from CLR values. Each marker represents a sewage sample and is colored by the world region for b acquired and c FG ARGs.

Fig. 2. Distance-Decay analyses of read abundances for city resistomes and bacteriomes.

Distance-Decay relationships for the bacteriome and resistome city communities (n = 60,030 pairwise comparisons) for the abundance of a Acquired ARGs, b FG ARGs, and c bacterial genera. The x-axis shows the pairwise city distances in kilometres (km), and the y-axis shows similarities. The dashed line represents the fit across all spatial scales and individual fits in solid: cities within the same country, cities within the same region, and cities that are in different regions. Model parameters and adjusted R² values are listed for each model in its corresponding plot.

Fig. 3. UMAP and distance-decay analysis of resistomes.

UMAP clustering of shared variants among the cities for a acquired ARGs and b FG ARGs. Only cities with more than 100 non-singleton alleles were retained, and Hellinger transformed and clustered with the UMAP algorithm. Each marker represents a city, colored by region and sized by the number of unique variants in that city. City labels were optimized using the ggrepel package to minimize overlap. Distance-decay relationships for assembled city resistomes across different spatial scales for c acquired (n = 4656 pairwise comparisons) and d the FG variants (n = 16,471 pairwise comparisons). The x-axis shows the pairwise city distances in kilometers (km), and the y-axis shows resistome similarities. The dashed line represents the fit across all spatial scales and individual fits in solid: cities within the same country, cities within the same region, and cities that are in different regions (between regions). Model parameters and adjusted R² values are listed for each model in its corresponding plot.

Fig. 4. The core of the correlation network of the abundances of bacteria and ARGs in the wastewater samples.

Each node represents either a species or an ARG, where an edge is the positive (blue) or negative (red) correlation between two nodes. The nodes are shaped by the database source (mOTU, acquired, or FG) and sized by the mean CLR value across samples. a Network graph showing the distribution of species and ARG, where a node is colored by either mOTU family or antimicrobial resistance class. b Network graph showing the communities detected. Nodes are here colored by whether they are an ARG or if the bacterial species is related to the human gut. The edge colors of the nodes highlight which community they were part of. c The count of the type of nodes in each community.