Anthropogenic land use changes have been recognized with significant effects on the abundance and diversity of antibiotic resistance genes (ARGs) in soil, but their impacts on ARGs with potential health risk remained poorly understood. In this study, paired metagenomes and viromes were obtained from soils (Anthrosols and Nitisols) with different land uses including urban parks, road verge, forests, vegetable and paddy in a subtropical city, Xiamen, and soils (Anthrosols) with various long-term fertilization treatments in Dezhou located in temperate region, respectively, to explore the influence of anthropogenic activity on soil resistome. The diversity and abundance of antibiotic resistance genes (ARGs) were profiled, and the risk associated factors of ARGs, i.e., genetic location, host, and co-existence with virulence factors (VFs), were systematically investigated at reads and contigs level. We observed that agricultural areas significantly enriched human-related ARGs and viruses, and positively related with clinical ARGs. Most of the ARG-carrying contigs were chromosomes (～85 %), while, human-related ARGs presented a higher odds ratio to locate on plasmids. Soil VFs exhibited land use pattern and distinct distribution between chromosome and plasmids, but less mobile than ARGs. Analysis of 131,014 soil viral genomes indicated that they barely encoded ARGs, nevertheless, transduction of VLPs was implicated in the spread of ARGs. The results can be mutually verified in Xiamen and Dezhou datasets. Overall, the agricultural soils with dry-farming are hotspots for the clinical ARGs, and the transmission of clinical ARGs between human dominated environments and soil is primarily mediated by plasmids, rather than bacterial chromosomes, and the transduction of human-gut related viruses could participate the process. These results highlight the importance of tracking the fate of clinical ARGs for better evaluating the impacts of human activities on soil resistome.

Fig. 1. (A). ARG abundance in soils with different types of land uses. (B). The boxplot showed ARGs abundance among different types of land uses. The median and quartiles were shown. The difference between types of land use was tested using the Wilcox.test, ** and * represents p < 0.01 and p < 0.05, respectively. (C). The diversity of ARG subtypes. (D). The linear model indicated the agricultural (paddy and vegetable plot) (adjusted R2 = 0.68, p < 0.01) and non-agricultural area (park, road verge and forest) (p > 0.05) presented different correlation between ARG subtypes diversity and abundance (copy per 16S rRNA gene).