Spatial Patterns,driving Mechanisms,and Prediction Of Soil Antibiotic Resistance Genes At The Global Scale E

Antibiotics,secondary metabolites secreted by microorganisms,are able to inhibit the growth of other microorganisms.Antibiotics are widely used for clinical treatment,desease prevention,and livestock growth promotion,with great socioeconomic and health benefits.However,the overuse and abuse of antibiotics have resulted in serious antibiotic resistance problems.For example,pathogens with antibiotic resistance genes（ARGs）that allow antibiotics lose their efficacy have posed ever-growing threats to human and animal health.Soils are essential reservoirs for ARGs and serve as an important habitat for many pathogens that were associated with clinical infection and plant disease outbreak.One serious issue regarding edaphic antibiotic resistance is the growing evidence that ARGs have transferred from soils to anthropogenic,animal,and plant settings.Previous studies concerning soil antibiotic resistance have mainly focused on regional and local scales,which could not provide comprehensive understandings of global patterns of ARGs and their interactions with ecological factors.This knowledge gap is restricting taking global actions to address serious challenges from antibiotic resistance.Here,we attempt to leverage metagenomics samples to profile antibiotic resistome across the world.we then disentangled the response of soil ARGs to environmental covariates and their microbial mechanisms,on the basis of 169 layers of climatic,anthropogenic,physiochemical,and land use features.Lastly,we trained four machine learning algorithms and selected the best one to generate a high-resolution map of the normalized abundance of soil ARGs across the globe.Our work would improve the understandings of antibiotic resistance on a large scale,provding scientific basis to control soil antibiotic resistance.The major findings are as follows.Our soil metagenomic data resulted in a total of 23 ARG types and 558 ARG subtypes,of which genes encoding resistance to multidrug was the dominant type.Normalized abundance of ARGs ranged from 29.34 to 250.02 ppm,with the average of 121.20±38.44 ppm.ARG composition discrepancy was found in varied habitats and latitudes.Normalized abundance of ARGs in agricultural soils was significantly higher than that in other soils,and ARG abundance also showed an untrend with absolute latitudes.Our global observations annotated 9 mobile gene element（MGE）types and 157 MGE subtypes,of which transposase was the type with the highest detection frequency and normalized abundance.Normalized abundance of MGEs has the mean of 55.26 ppm,ranging from 7.31 to 391.05 ppm.Normalized abundance of MGEs and ARGs are positively correlated,which suggested horizonal gene transfer would be an important controller to regulate soil ARGs.The annotated microbiome represented 4 kingdoms,69 phyla,134 classes,265 orders,618 families,2,397 genera,and 15,071 species,of which 3,150 species carried ARGs or MGEs.Gut microbes and pathogens belong to Enterobacterales and Pseudomonadales were dominant microbes carrying ARGs or MGEs.The most important factors to regulate ARGs were anthropogenic activities,where the normalized abundance of ARGs increased with human development index,horse density,anthropogenic biomes,proportion of nitrogen in product pigs,and chlorpyrifos used in corn.Climatic varibales,that ranked the second important drivers to influence ARGs,were significantly correlated with ARG composition based on Procrustes analysis.Increased annual mean temperature was associated with decreasing normalized abundance of all ARGs,genes encoding resistance to multidrug and fosmidomycin.Soil ARGs are also driven by soil properties,where higher normalized abundance of ARGs allowed to lower nitrogen,organic carbon,and organic phosphorus.Structural equation model revealed that anthropogenic activities largely controlled ARG abundance via their microbial hosts（pathogens and gut microbes）.The impact of climate on soil ARGs was mediated via soil properties,land use features,and microbial factors.Similarly,the impacts of soil properties and land use on ARGs were largely indirect and mediated via microbial factors.We trained support vector machine,k-nearest neighbor,gradient boosting decision tree,and random forest to predict ARG abundance through feature selection and hyperparameter tuning.Random forest was the optimal algorithm whose R~2,RMSE,and MAE reached 0.47,20.94,and 16.23,when feature count,mtry,ntree,and nodesize were set as 71,6,1200,and 9.We used random forest to predict the normalized abundance of soil ARGs in 2015 at the global scale,which revealed the hotspots of ARGs were largely located in highly populated regions,such as Western Europe,East Asia,South Asia,and Eastern United States.These hotspots resulted largely from anthopogenic activities that input selective agents,pathogens,and gut microbes to rasie the normalized abundance of ARGs.Our global map also illustrated a high normalized abundance of ARGs in North Europe,Southeast America,and New Zealand,which was mainly attributed by abundant nutriments,high organic carbon content,low temperature,plentiful precipitation,as well as developed husbandry.