Global health risk assessment of antibiotic resistance in agricultural soils

Global health risk assessment of antibiotic resistance in agricultural soils | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Global health risk assessment of antibiotic resistance in agricultural soils Yanzheng Gao, Zekai Li, Sean Fettrow, Thomas Borch, Miao Han, and 6 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7635655/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Antibiotic resistance is posing a major threat to public health, yet there is little effort to quantitatively assess risks associated with antibiotic resistance genes (ARGs), particularly in agricultural soils. We therefore propose a risk assessment framework by integrating metagenomic profiling, quantitative health risk assessments, and machine learning to evaluate distribution and health risks of ARGs in a diverse set of global agricultural soil samples. Based on 985 selected metagenomic samples from diverse agricultural systems, we identified 1745 subtypes from 30 major ARG families, revealing patterns between major agricultural settings. Approximately 1% of global agricultural areas were classified as high-risk, primarily concentrated in regions with intensive farming practices and high antibiotic usage. Our framework enables the identification of risk hotspots which seem to be driven by socioeconomic, climatic, and land use factors. These findings facilitate a methodological advancement for predicting ARG risk through mechanism-driven models, rather than descriptive abundance metrics. The proposed framework will support targeted soil management strategies to mitigate antibiotic resistance propagation in agroecosystems. Earth and environmental sciences/Biogeochemistry Earth and environmental sciences/Environmental sciences Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction The transmission of antibiotic resistance genes (ARGs) leads to the development of antibiotic-resistant bacteria (ARB), which has emerged as one of the greatest global health challenges of the 21 st century 1,2 . Soils act as a natural reservoir for both ARGs and ARB to emerge, evolve and disseminate 3 , with agricultural soils being of particular concern. This is because agricultural practices (e.g. application of manure and wastewater) continuously introduce antibiotics (which exert sustained selective pressure for ARG proliferation), ARGs and ARB into soils 4–7 . Edible leafy plant tissue and cereal grains can contain ARG residues 8–12 , sometimes harboring high-risk resistant pathogens 13,14 . As a primary source of ARGs found in agricultural products, soils facilitate the spread of ARGs along the food chain 15,16 . Consequently, the consumption of fresh produce, particularly raw vegetables, is a direct route of human exposure to soil microorganisms and ARGs, posing a substantial threat to human health 17 . Yet, the health risks and future trends of ARGs in global agricultural systems remains largely unknown. Although advances in qPCR and metagenomic profiling have delineated ARG distribution patterns in agricultural soils at the regional scale 6,18–21 , there remain essential knowledge gaps at the global scale. Sampling areas are currently focused on specific agroecosystems in similar pedoclimatic zones 18,22–24 . Moreover, risk assessments predominantly rely on abundance metrics, neglecting the synergistic risks from MGE (mobile genetic element) and pathogen tripartite interactions 20,25,26 . Furthermore, conventional linear models fail to capture nonlinear interactions among climatic, edaphic, and agronomic drivers 27 . The above limitations have resulted in incomplete understanding of global distribution patterns and future health risks associated with ARGs. A comprehensive research strategy is urgently required to address these gaps and to construct a comprehensive health risk assessment of ARGs in global agricultural systems. Here, we analyze 985 global agricultural soil metagenomes spanning 7 different crop type systems across 6 continents. Resistance profiles are generated to elucidate the occurrence and distribution patterns of ARGs in agricultural soils worldwide using bioinformatics. Specifically, we develop a risk assessment framework based on the bioinformatics results using a random forest machine learning algorithm. Our framework encompasses global soils and analyzes the contribution of driving factors such as climate change, fertilization intensity, and soil pH to ARG-related health risks. Furthermore, we predict future trends of high-risk areas for 2030 under the Shared Socioeconomic Pathways (SSP) scenarios, which are global development trajectories widely used in climate change research 28 . This work provides the first global risk stratification map of ARGs in agricultural soils, identifies priority areas for control, and provides theoretical support for developing effective strategies to mitigate ARGs and improve the health and development of agricultural systems. Results ARG types and subtypes in global agricultural soils The 985 metagenomic samples were distributed across representative countries (e.g., the United States, China, Australia, France, and Brazil) and 6 continents including Asia, Africa, Europe, South America, North America, and Oceania (Fig. 1a and Supplementary Table 1). The soil samples were divided into various soil crop types, including 141 fruit-growing samples, 35 livestock and poultry samples, 462 major crops (maize, rice, and wheat) samples, and 154 vegetable-growing samples. For the 31 samples without specified land use type information in the NCBI database, we determined their land use types according to the latitude and longitude of the samples via a global gridded agricultural production map 29 . Overall, 1745 ARG subtypes from 30 ARG families are identified (Supplementary Fig. 1, Supplementary Table 2, and Table 3). The detection of only a few ARG types, including bacitracin, macrolide-lincosamide-streptogram (MLS), multidrug, polymyxin, and tetracycline, encompasses most of the abundance, while the vast majority of ARG types (1425 out of 1745) makes up considerably less of the abundance (<10%) (Supplementary Table 2). At the more refined subtype level, bacA , MexB , MexF , MuxB and MuxC makes up the majority of the abundance (Supplementary Fig. 2, Supplementary Table 3). Statistical tests revealed significant differences in total ARG abundance (RPKM; Reads per kilobase per million) between soil land use types (Fig. 1b). The abundance in livestock and poultry farm soils is greatest (3226.32 RPKM), with a mean value approximately 20 times higher than the other soil types. This was followed, in descending order, by maize-growing soil (155.96 RPKM), fruit-growing soil (140.65 RPKM), vegetable-growing soil (118.42 RPKM), others soil (113.40 RPKM), wheat-growing soil (108.74 RPKM), and rice-growing soil (93.87 RPKM), which showed the lowest ARG abundance. There are also notable differences in ARG types composition between soils, with the primary difference being between cropland soil types (i.e., fruit, maize, rice etc.) and livestock and poultry soils (Fig. 1c). Bacitracin, polymyxin and rifamycin are widely detected in most cropland soils, but are infrequent in livestock and poultry farm soils. Conversely, aminoglycoside, multidrug, MLS and tetracycline have relatively high abundances in the livestock and poultry soils compared to cropland soil types. (Fig. 1c, Supplementary Table 2). Regarding the distribution of ARG subtypes across soil types, livestock and poultry soils and maize-growing soils displayed distinct compositional profiles compared to other five soil categories (Fig. 1d). These two soil types were characterized by a higher abundance of sulfonamide and MLS-resistant ARG subtypes, but relatively lower proportions of novobiocin and rifamycin resistance subtypes. In contrast, the other five soil types consistently contained over 20% bacitracin-resistant ARGs (specifically the bacA ) and approximately 10% novobiocin-resistant ARGs (specifically the novA ) (Fig. 1d). For ARG subtypes, bacA have the greatest abundance, followed by nov A, sul1 and floR, all of which exceed 8000 RPKM (Fig. 1e). For a vast majority of detected ARG subtypes (1487 out of 1745), abundance is comparatively low (i.e., <100 RPKM) (Supplementary Table 2). Antibiotic inactivation and antibiotic efflux are the main resistance mechanisms found in all agricultural soil types (Supplementary Fig. 3, Supplementary Table 4). Furthermore, more than half of the ARGs (1128 out of 1745) have developed resistance to a single drug class, while 617 ARGs harbor resistance to 2~16 drug classes (Supplementary Fig. 4, Supplementary Table 4). Profile s of soil MGEs and ARGs host bacteria Mobile genetic elements (MGEs) and their host bacteria play a crucial role in promoting the prevalence and persistence of ARGs in soils. Through quantitative annotation, 41 MGE types and 217 MGE subtypes are identified across all soil samples, with total abundance value ranging from 0.0021 to 1726.10 RPKM (Supplementary Fig. 5 and Supplementary Table 5). Among these, transposase and IS91 account for 63.7% and 20.0% of the total detected MGE subtypes, respectively (Supplementary Fig. 5). Statistical tests revealed significant variations in MGE abundance across different soil land use types. Notably, livestock and poultry soils demonstrate the highest MGE abundance, which is significantly greater than that observed in other cropland soils—a pattern consistent with previously reported ARG distributions (Fig. 2a). Regarding the compositional profile of MGEs, transposase is the predominant type across all soil categories, exceeding 50% prevalence, followed by IS91, which consistently represented over 19%. A notable exception is observed in fruit-growing soil, where plasmid content reached 17%, significantly higher than in other soil types. In contrast, ist elements are nearly undetectable in livestock and poultry soils (Fig. 2b). The Pearson correlation analysis revealed strong positive correlation between the total abundance of ARGs and several major MGE types, including transposase ( r =0.81, p <0.001), integrase ( r =0.73, p <0.001), plasmid ( r =0.70, p <0.001), IS91 ( r =0.78, p <0.0001), and ist ( r =0.43, p <0.001), as well as the total abundance of MGEs ( r =0.83, p <0.0001) (Fig. 2c). These results indicate a consistent micro-scale co-occurrence of ARGs with specific MGEs within individual soil samples. At the macro-scale across different soil types, the abundance trends of major MGEs mirrored those of ARGs. Specifically, transposase, integrase, and IS91 show significant higher abundances in livestock and poultry farm soils compared to other soil types (Supplementary Fig. 6), which is highly consistent with the spatial distribution pattern of ARGs described in the previous section. Furthermore, Procrustes analysis reveals a significant concordance between the community structures of ARG and MGE subtypes (M 2 =0.3871, P < 0.001, Fig. 2d), indicating coordinated variation in their profiles across samples. This strong correlation implies that the compositional patterns of ARG subtypes are intimately associated with those of MGE subtypes, likely resulting from shared environmental drivers or direct interactions such as HGT. Bacteria identified as potential hosts of ARGs in this study belong to 75 distinct phyla (Supplementary Table 6). Pseudomonadota and Actinomycetota are the most dominant groups, representing 43.6% and 42.8% of the total microbial population, respectively (Supplementary Fig. 7, Supplementary Table 6). Other phyla with relatively high abundances (≥1.8%) include Myxococcota (2.6%), Planctomycetota (1.9%), Bacillota (1.8%) and Bacteroidota (1.8%). The Shannon Index of microbial diversity varied significantly across soil types (Fig. 2e, Supplementary Fig. 9). Specifically, rice-growing soil exhibited the highest microbial diversity (mean value: 5.51), followed closely by maize-growing soil (5.50), whereas livestock and poultry farm soil showed the lowest diversity (4.97). To further identify potential hosts of ARGs and MGEs at phylum level, a Spearman correlation-based co-occurrence network was constructed (Supplementary Fig. 8). The network included 32 phyla, 200 ARG subtypes and 76 MGE subtypes. Phyla such as Bacteroidota , Bdellovibrionota and Spirochaetota exhibit positive correlations with more than 40 ARG subtypes, suggesting their roles as potential hosts for the associated ARGs and MGEs. We further provide an ARG risk framework based on classifications by the World Health Organization (WHO)and following a previous study 27 (Supplementary Table 7). High-risk ARGs (Rank I/II) represent 9–22% of the total abundance (Fig. 3a) and 7-9% of the total richness (Fig. 3b) across the soil types. When all soil types are combined, high-risk ARGs (Rank I/II) represent 17.4% of the total abundance (Fig. 3c) and 7.5% of the total richness (Fig. 3d). Furthermore, a bioinformatics software package previously used to assess ARG risk 27 was used to calculate the risk index (RI) in each sample. The RIs of all metagenome samples were then classified into 6 risk levels (Fig. 3e) using the K-means clustering method. Samples in Ranks 6 and 5 were classified as high-risk and represent 6% of the total ARGs. The standardized risk values for ARGs in Ranks 6 and 5 ranged from 0.12 to 1 and from 0.0045 to 0.11, respectively. The standardized risk values for ARGs classified as Ranks 1 to 4 ranged from 0 to 0.0028. Global ARG risk assessment model in agricultural soils We further used a machine learning model to visualize the potential health risks of ARGs in agricultural soils at the global scale. We applied this model to a global dataset, mapping ARG risks at a resolution of 0.25° (Fig. 4a). Most of the obtained ARG high-risk areas (Rank 5/6) are in the Northern Hemisphere, mainly in the North China Plain and North India (Fig. 4c, d), both of which are major global agricultural production areas. This map highlights the distribution of risks and threats of ARGs in agricultural soils worldwide, providing critical insights for ARG detection and control. The model demonstrated robust predictive performance, as evidenced by the receiver operating characteristic (ROC) curve and confusion matrix (Supplementary Fig. 10 and 11). Variable importance analysis identified “Climate, Soil pH, and Moisture” as the most important factor affecting ARG risk in our model, consistent with previous reports (Fig. 4b) 23,28 . Precipitation and soil texture are also identified as influential yet secondary factors. Significant correlations were observed between these predictors and ARG risk levels (Supplementary Fig. 12). In particular, “Climate, Soil pH, and Moisture” and “Temperature” showed strong positive correlations with ARG risk across a broad range of values. In contrast, factors such as “Agricultural Production and Fertilizer Use” and “Carbon and Nitrogen Content” display nonlinear or threshold-dependent relationships with risk. We further predicted the spatial distribution of ARG risks under future scenarios using a modeling approach that incorporated projections of key drivers, including future antibiotic consumption, socioeconomic development, and climate change—with increased soil temperature serving as a primary climatic effector. Projections were made under three Shared Socioeconomic Pathways (SSPs), namely SSP1-2.6 (sustainability), SSP3-7.0 (regional rivalry), SSP5-8.5 (fossil-fueled development) 28 (Fig. 5a-c), representing divergent trajectories of global change. Under all scenarios, a substantial expansion of high-risk areas was predicted (Fig. 5d). These shifts in risk levels are closely associated with spatial patterns of temperature change, socioeconomic intensification, and land-use modification (Fig. 5e-f), consistent with previous findings 29 . In particular, future antibiotic use was identified as a major input variable contributing to risk increases in the model. Overall, these results highlight pronounced regional differences in ARG risk and underscore the influence of future environmental and anthropogenic factors on risk dynamics in agricultural soils. Further validation through experimental and monitoring data remains necessary. Discussion Livestock and poultry farms have increased ARG risk A total of 985 global soil samples were analyzed to characterize ARG profiles in agricultural soils (Fig. 1a) and our findings reveal significant variations in ARG distribution across different agricultural soil types. Specifically, both livestock and poultry soils harbor significantly greater ARG abundance compared to other agricultural soil types (Fig. 1b). The predominant ARG types in livestock and poultry farm soil are aminoglycoside, MLS, and tetracycline-related ARGs (Fig. 1c), suggesting that ARGs in these soils exhibit higher activity and potential health risks. This elevated ARG activity is likely due to the prevalent use of antibiotics in these locations, since livestock and poultry feces are major sources of ARGs 30 , and contribute to the substantial influx of antibiotics and ARGs into the soil environment. In contrast, significantly lower ARG abundances were found in fruit-growing soil, major crops (maize, rice, and wheat) soil, and vegetable-growing soil. This is likely due to fewer contamination sources 31 , as well as the mitigating effect of plants 32,33 . Across all soil types, bac A was the most abundant and widely detected gene (100% prevalence), which is commensurate with findings from preivous studies 34,35 .This suggests that bacA could serve as a broad indicator for ARGs across different environments. Furthermore, Bac A has demonstrated resistance to degradation in compost and was associated with bacitracin, indicating the the necessity for stricter regulation of bacitracin use 34 . Risks associated with ARG specific types and subtypes The transmission and potential health risks of ARGs is closely associated with MGEs (mobile genetic elements) through horizontal gene transfer (HGT). By analyzing the correlation between ARG abundance and MGE abundance, we confirmed that ARG abundance had a positive correlation with MGE abundance (Fig. 2b). Importantly, our comparative analysis revealed significant enrichments of transposases, integrases, and IS91 in livestock and poultry farm soil compared to other croplands, suggesting high HGT frequency mediated by these MGEs. This phenomenon may be attributed to the intensive antibiotic selection pressure of this environment. This soil type is a primary source for antibiotic application, which contributes to a different ARG profile compared to other croplands. Furthermore, the Shannon index variability in livestock and poultry soil is notably complex, indicating the complexity of antibiotic contamination in these soils (Fig. 2c). The risk level of ARGs is not always directly correlated with their abundance in the environment; even low-abundance ARGs can pose a threat to human health. Assessing the health risks associated with ARGs also requires assessing their ability to colonize and proliferate in the human body, as well as their adaptability and virulence 36 . A comprehensive risk assessment of ARGs also involves complex biological processes (such as HGT and host pathogenicity), which, unlike other environmental pollutants, cannot be easily quantified using clear quantitative analysis 37 . We used a quantitative assessment of ARG risk levels, which contained three critical nodes of ARG transmission from the environment to humans: i) enrichment of ARGs in human environments; ii) ARG mobility; and iii) host pathogenicity 38 . This framework simultaneously accounts for ARG types and abundance, the significant role of MGEs in the HGT of ARGs, as well as the pathogenicity of antibiotic-resistant bacteria. According to our model annotation results, 5.99% of the ARGs detected in our agricultural soil dataset are classified as high risk (Rank 5/6) (Fig. 3e), which is lower than that reported in hospital wastewater (23.76%) 39 and global groundwater (10.76%) 40 , but higher than that in freshwater aquaculture ponds (5.25%) 41 . Global agricultural soil ARG risk assessment framework We constructed a global risk assessment framework based on a machine learning model and mapped the associated health risks of ARGs. Our model indicates that ~0.90% of the data points represent high-risk ARG regions for the global agricultural system. These regions are predominantly located in the North China Plain and North India (Fig. 4c-d). In 2021, these two agricultural regions ranked 4th, and 2nd globally in terms of agricultural area, covering 1.09×10 8 , and 1.54×10 8 hectares, respectively (https://data.worldbank.org/). Therefore, these hotspots of elevated ARG risk poses unparalleled concern that necessitates global attention. Our model revealed that “Climate, Soil pH and Moisture” and “Precipitation” are significantly correlated with the distribution patterns of ARG health risks (Fig. 4b). Climate has previously been reported as an important factor affecting global distribution of ARGs 25,42–45 . We illustrate the impact of climate change on potential changes to soil ARG health risks (Fig. 4b).Additionally, pH is considered a key factor in predictive models and plays a crucial role in shaping soil microbial communities, thereby influencing the composition of ARGs 46,47 . Precipitation is considered a critical influencer of soil microbiota at a global scale 41,44,48 . Our scenario-based predictions reveal that ARG health risks will escalate synergistically with rising socioeconomic and land use pressures (Fig. 5d-f). In our model, “socioeconomic and land use indicators” primarily refers to changes in GDP and population. As scenarios such as SSP126, SSP370, and SSP585 evolve, economic activity and population growth could lead to increased urbanization and heightened demands on agricultural activities, which might contribute to the excessive use of antibiotics 49 . The region with the highest-ranked ARGs (rank 6) was mainly distributed in China. In ResistanceMap (https://resistancemap.onehealthtrust.org/Animals.php), China has the highest antimicrobial consumption in the world, reaching 319 mg per population correction unit (PCU), significantly higher than South Korea’s 188 mg per PCU and Spain’s 182 mg per PCU, this could explain the reason of high-risk areas. Global livestock antibiotic consumption is expected to increase by 52% by 2030, since China and India emerging as the fastest-growing contributors (13% and 18% increases, respectively) (https://resistancemap.onehealthtrust.org/Animals.php). Our results indicate that if sustainable development pathways are implemented (Fig. 5d), ARG risks could be reduced by 47% globally. In addition, ARG health risks are expected to escalate synergistically with rising global temperatures 50 . This phenomenon is primarily mediated by two mechanisms. The first mechanism is the thermodynamic enhancement of HGT. Elevated temperatures can increase plasmid conjugation rates 51,52 . And the second mechanism is pathogen proliferation. Pathogenic bacteria play a significant role in the dissemination of ARGs. These bacteria exhibit increased activity with increasing temperature 53 . Conclusion In summary, we have investigated the global distribution of ARGs in different agricultural soil environments, identifying a total of 1745 ARG subtypes of 30 types. Several ARGs were consistently detected across soil types, highlighting their potential roles in the global soil ARG profile. Through a quantitative risk assessment framework built from a machine learning model, the highest-risk regions have been identified, particularly in the North China Plain and India. Predictive assessments revealed that high economic activity and population densities may impact future ARG risk, while climate and increased temperature was also an important variable affecting ARG risk. We might reduce ARG risks by following sustainable development pathways while mitigating the impacts of climate change. Our research provides a global risk assessment framework and predictive analysis which may guide future ARG risk research. Methods Metagenome dataset collection From September 10, 2022, to February 10, 2024, we conducted a search in the NCBI SRA database (https://ncbi.nlm.nih.gov/sra/) via the keywords “agricultural soil”. The following screening criteria were applied to standardize the subsequent analysis steps: (1) exclusion of experimental treatment groups and rhizosphere soil samples; (2) selection of sequencing samples from platforms labeled “paired”, “genome”, and “illumina” while filtering out data generated from single layouts, exomes, and other sequencing platforms; and (3) downloading raw SRA data for reanalysis. These criteria helped mitigate potential uncertainties in experimental results arising from varied sequencing standards and sample types. Ultimately, we obtained 985 metagenomic samples and compiled their library information, size, land use classification, geographical coordinates (latitude and longitude), continent, country of origin, and DOI, as listed in Supplementary Table 1. ARGs, MGEs, taxonomic annotation, and abundance calculations The raw data from the metagenomic samples were subject to quality assessment and filtered via fastp (v0.23.3; https://github.com/OpenGene/fastp). The required length was 50, and the qualified quality phred was 20. The ARGs in the samples were subsequently annotated via ARG-OAP (v3.2.2; https://github.com/xinehc/ARGs_OAP) with default parameters. For the annotation of MGEs, we utilized a specialized database (https://github.com/KatariinaParnanen/MobileGeneticElementDatabase) containing 278 gene name annotations and over 2000 unique sequences, employing the same default parameters as those used for ARG annotation. Relative abundance comparisons among different samples were conducted using reads per kilobase per million mapped reads (RPKM) as units. To identify the hosts of ARGs and MGEs within the metagenome, we employed kraken2 (v2.1.2; https://github.com/DerrickWood/kraken2) with default parameters. Species abundance statistics based on kraken2 annotations were performed using bracken. Quantification of the risk of ARGs to humans We assessed the risk associated with ARGs in the metagenome via Arg_ranker (v3.5; https://github.com/caozhichongchong/arg_ranker). The health risk index (RI) calculation method is as follows: The rank percentage, ranking risk code, and total abundance were calculated with the default parameters of Arg_ranker. This approach was used to quantify the health risk posed by ARGs in agricultural soils based on the RI. Machine learning and global mapping To predict the global health risks of ARGs in agricultural soil, we prepared a total of 65 gridded covariates, encompassing climate, soil, agricultural, and socioeconomic factors. The climate data included 19 bioclimatic variables were downloaded from WorldClim (https://www.worldclim.org/). Soil information and characteristics were downloaded from SoilGrids (https://www.soilgrids.org/). And other variables, including the number of livestock, human population, GDP and global antibiotic usage, were downloaded from databases such as the Food and Agriculture Organization (FAO; https://www.fao.org/livestock-systems/en/), Socioeconomic Data and Applications Center (SEDAC; https://sedac.ciesin.columbia.edu/), and ResistanceMap (https://resistancemap.onehealthtrust.org/index.php). Specific covariate information is presented in Supplementary Table 9. For tabular data, we used ArcGIS (v10.8) to allocate all values based on administrative boundaries. For raster layers with different resolutions, we used the nearest neighbor method to resample all raster data to a resolution of 0.25° (approximately 27.75 km at the equator). Our resampled covariate layers had latitude and longitude values ranging from -180° to 180° and from -90° to 90°, respectively. For those maps that did not fall within the latitude or longitude range, we redefined the latitude and longitude value range via the “Mosaic to New Raster” method to ensure that all covariate layers had the same range. To enhance the interpretability of variables, we conducted a factor analysis. Specifically, we started with a Principal Component Analysis (PCA) to compute the correlation matrix and extract eigenvalues. These eigenvalues helped us determine the contribution rate of each principal component, as well as the cumulative contribution rate. We selected the principal components with a cumulative contribution rate of at least 80%, resulting in 10 retained variables. Subsequently, a Varimax rotation was applied to improve the interpretability of the factor loadings. The 10 principal components and their explanations are presented in Supplementary Table 10. A random forest combined with 10-fold cross validation was used for machine learning. The random forest algorithm is an ensemble learning algorithm that combines the modelling capabilities of decision trees with the advantages of ensemble learning. To enhance the model’s generalization ability, 10-fold cross-validation was used. Relying on the factor loading results, we regenerated the data for machine learning. Hence, the new dataset was randomly divided into 10 equally sized subsets with 9 subsets used for training and 1 subset for testing in each model fitting process. This procedure was repeated 10 times until each subset was used as a testing set. Additionally, the result of this model was determined by averaging the 10 fitting data results, and the fitting ability of the model was evaluated via confusion matrixes and ROC curves aligned with the utility of mean imputation to fill the missing values. In addition, global data from 65 covariate layers were also regenerated based on the factor loading results and excluded regions lacking data, which made our model more accurate. Afterwards, the 78301 remaining data points were available for prediction, followed by visualizing the anticipated health risks caused by ARGs existence in agricultural soils worldwide via ArcGIS (v10.8) at a resolution of 0.25°. To predict the future health risks of ARGs in global agricultural soils, we collected antibiotic usage data for 2030 from ResistanceMap (https://resistancemap.onehealthtrust.org/index.php), grid-based population and GDP data from two studies corresponding to 2030 53–55 , and climate data from BCC-CSM2-MR, which is a version of the Coupled Model Intercomparison Project Phase 6 (CMIP6) model. CMIP6 is a climate comparison program organized by the World Climate Research Programme (WCRP), and the associated products provide important insights for predicting future climate change and climate impacts. Subsequently, we have chosen three prediction scenarios for future development (SSP126, SSP370 and SSP585) operating as follows: SSP126 represents sustainable development pathways with low climate change challenges, SSP370 is regionally differentiated and high climate change challenges, finally SSP585 that describes traditional development pathways dominated by fossil fuels. For each future scenario, we obtained 19 bioclimatic variables. The results were visualized via ArcGIS (v10.8) at a resolution of 0.25°. Statistical analysis and visualization Data preprocessing and analysis was accomplished using Microsoft Excel (v16.0). In R (v4.3.3), ggplot2 was used to plot geographic location maps of metagenomic samples, Vegan package was implemented for Procrustes analysis and plotting further relationships between ARGs and MGEs, but the psych package was applied for factor analysis. OriginLab (v2024b) was used to plot the abundance and richness of ARGs, as well as the compositions of and relationships between ARGs and MGEs. Gephi (v0.10.1) was used to construct co-occurrence networks of ARGs, MGEs, and microorganisms. Machine learning was performed using Jupyter Notebook (v6.5.4), and all generated maps were visualized with ArcGIS (v10.8). Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Declarations Data availability All of the data supporting this research are available in the main text and the Supplementary Information, and original data can be obtained from the corresponding authors upon reasonable request. Acknowledgements C.Q. and Y.G. acknowledge the support from the Joint Funds of the National Key Research and Development Program of China (2023YFE0110800, 2023YFC3708103) and the National Natural Science Foundation of China (42477419, 42107221, 22161132011, U22A20590). Author contributions All authors contributed to the work presented in this paper. Y.G. conceived the study. Z.L. completed the majority of the experiments and wrote the original paper. M.H. helped to conduct the theoretical calculations. S.F., T.B., A.K., X.H., J.W., and H.W. revised the manuscript. C.Q., J.G., and Y.G. proposed and discussed the concepts of this study and made extensive revisions to this work. Competing interests The authors declare no competing interests. References Antimicrobial resistance: global report on surveillance. https://www.who.int/publications/i/item/9789241564748. Okeke, I. N. et al . The scope of the antimicrobial resistance challenge. The Lancet 403 , 2426–2438 (2024). Goh, Y. X. et al . Evidence of horizontal gene transfer and environmental selection impacting antibiotic resistance evolution in soil-dwelling Listeria. Nat. Commun. 15 , 10034 (2024). Udikovic-Kolic, N., Wichmann, F., Broderick, N. A. & Handelsman, J. Bloom of resident antibiotic-resistant bacteria in soil following manure fertilization. Proc. Natl. Acad. Sci. U. S. A. 111 , 15202–15207 (2014). Kuppusamy, S. et al . Veterinary antibiotics (VAs) contamination as a global agro-ecological issue: A critical view. Agric. Ecosyst. Environ. 257 , 47–59 (2018). Chen, C. et al . Occurrence of antibiotics and antibiotic resistances in soils from wastewater irrigation areas in Beijing and Tianjin, China. Environ. Pollut. 193 , 94–101 (2014). Chen, C. et al . Effect of antibiotic use and composting on antibiotic resistance gene abundance and resistome risks of soils receiving manure-derived amendments. Environ. Int. 128 , 233–243 (2019). Campos, J. et al . Microbiological quality of ready-to-eat salads: An underestimated vehicle of bacteria and clinically relevant antibiotic resistance genes. Int. J. Food Microbiol. 166 , 464–470 (2013). Guo, Y. et al . Diversity and abundance of antibiotic resistance genes in rhizosphere soil and endophytes of leafy vegetables: Focusing on the effect of the vegetable species. J. Hazard. Mater. 415 , 125595 (2021). Zhao, C. X., Su, X. X., Xu, M. R., An, X. L. & Su, J. Q. Uncovering the diversity and contents of gene cassettes in class 1 integrons from the endophytes of raw vegetables. Ecotox. Environ. Safe. 247 , 114282 (2022). Zhou, S. Y. D. et al . Prevalence of antibiotic resistome in ready-to-eat salad. Front. Public Health 8 , 92 (2020). Yu, Y. et al . Plants select antibiotic resistome in rhizosphere in early stage. Sci. Total Environ. 858 , 159847 (2023). Zekar, F. M., Granier, S. A., Touati, A. & Millemann, Y. Occurrence of third-generation cephalosporins-resistant klebsiella pneumoniae in fresh fruits and vegetables purchased at markets in Algeria. Microb. Drug Resist. 26 , 353–359 (2020). Nkhebenyane, S. J., Khasapane, N. G., Lekota, K. E., Thekisoe, O. & Ramatla, T. Insight into the prevalence of extended-spectrum β-Lactamase-producing enterobacteriaceae in vegetables: A systematic review and meta-analysis. Foods 13 , 3961 (2024). Gao, F. Z. et al . Untreated swine wastes changed antibiotic resistance and microbial community in the soils and impacted abundances of antibiotic resistance genes in the vegetables. Sci. Total Environ. 741 , 140482 (2020). Marti, R. et al . Impact of manure fertilization on the abundance of antibiotic-resistant bacteria and frequency of detection of antibiotic resistance genes in soil and on vegetables at harvest. Appl. Environ. Microbiol. 79 , 5701–5709 (2013). Cordovez, V., Dini Andreote, F., Carrion, V. J. & Raaijmakers, J. M. Ecology and evolution of plant microbiomes. Annu. Rev. Microbiol. 73 , 69-88 (2019). Zhou, Y., Niu, L., Zhu, S., Lu, H. & Liu, W. Occurrence, abundance, and distribution of sulfonamide and tetracycline resistance genes in agricultural soils across China. Sci. Total Environ. 599 , 1977–1983 (2017). Nogrado, K., Unno, T., Hur, H. G. & Lee, J. H. Tetracycline-resistant bacteria and ribosomal protection protein genes in soils from selected agricultural fields and livestock farms. Appl. Biol. Chem. 64 , 42 (2021). Wu, J. et al . Antibiotics and antibiotic resistance genes in agricultural soils: A systematic analysis. Crit. Rev. Environ. Sci. Technol. 53 , 847–864 (2023). Wang, B. et al . Tackling soil ARG-carrying pathogens with global-scale metagenomics. Adv. Sci. 10 , 10 (2023). Salam, L. B. Metagenomic insights into the microbial community structure and resistomes of a tropical agricultural soil persistently inundated with pesticide and animal manure use. Folia Microbiol. 67 , 707–719 (2022). Cadena, M. et al . Tetracycline and sulfonamide antibiotic resistance genes in soils from nebraska organic farming operations. Front. Microbiol. 9 , 1283 (2018). Sun, J. et al . Antibiotic resistance genes (ARGs) in agricultural soils from the Yangtze River Delta, China. Sci. Total Environ. 740 , 140001 (2020). Zheng, D. et al . Global biogeography and projection of soil antibiotic resistance genes. Sci. Adv. 8 , eabq8015 (2022). Fang, H. et al . Dissemination of antibiotic resistance genes and human pathogenic bacteria from a pig feedlot to the surrounding stream and agricultural soils. J. Hazard. Mater. 357 , 53–62 (2018). Ghannam, R. B. & Techtmann, S. M. Machine learning applications in microbial ecology, human microbiome studies, and environmental monitoring. Comp. Struct. Biotechnol. J. 19 , 1092–1107 (2021). Riahi, K. et al . The shared socioeconomic pathways and their energy, land use, and greenhouse gas emissions implications: An overview. Global Environ. Chang. 42 , 153–168 (2017). Yu, Q. et al . A cultivated planet in 2010-Part 2: The global gridded agricultural-production maps. Earth Syst. Sci. Data 12 , 3545–3572 (2020). Duan, M. et al . Factors that affect the occurrence and distribution of antibiotic resistance genes in soils from livestock and poultry farms. Ecotox. Environ. Safe. 180 , 114–122 (2019). Wang, J., Zhang, Q., Chu, H., Shi, Y. & Wang, Q. Distribution and co-occurrence patterns of antibiotic resistance genes in black soils in Northeast China. J. Environ. Manage. 319 , 115640 (2022). Li, S. et al . Plant diversity reduces the risk of antibiotic resistance genes in agroecosystems. Adv. Sci. 12 , e2410990 (2025). Shen, Y. et al . Dominant microbiome iteration and antibiotic resistance genes propagation way dictate the antibiotic resistance genes contamination degree in soil-plant system. J. Clean Prod. 464 , 142786 (2024). Yue, Z. et al . Antibiotic degradation dominates the removal of antibiotic resistance genes during composting. Bioresour. Technol. 344 , 126229 (2022). Gao, Q. et al . Diverse and abundant antibiotic resistance genes from mariculture sites of China’s coastline. Sci. Total Environ. 630 , 117–125 (2018). Manaia, C. M. Assessing the risk of antibiotic resistance transmission from the environment to humans: Non-direct proportionality between abundance and risk. Trends Microbiol. 25 , 173–181 (2017). Sanderson, H., Johnson, D. J., Wilson, C. J., Brain, R. A. & Solomon, K. R. Probabilistic hazard assessment of environmentally occurring pharmaceuticals toxicity to fish, daphnids and algae by ECOSAR screening. Toxicol. Lett. 144 , 383–395 (2003). Zhang, A. N. et al . An omics-based framework for assessing the health risk of antimicrobial resistance genes. Nat. Commun. 12 , 4765 (2021). Shuai, X. et al . Ranking the risk of antibiotic resistance genes by metagenomic and multifactorial analysis in hospital wastewater systems. J. Hazard. Mater. 468 , 133790 (2024). Liu, C. et al . Meta-analysis addressing the characterization and risk identification of antibiotics and antibiotic resistance genes in global groundwater. Sci. Total Environ. 860 , 160513 (2023). Wang, C., Liu, X., Yang, Y. & Wang, Z. Antibiotic and antibiotic resistance genes in freshwater aquaculture ponds in China: A meta-analysis and assessment. J. Clean Prod. 329 , 129719 (2021). Hendriksen, R. S. et al . Global monitoring of antimicrobial resistance based on metagenomics analyses of urban sewage. Nat. Commun. 10 , 1124 (2019). Gao, M., Zhang, Q., Lei, C., Lu, T. & Qian, H. Atmospheric antibiotic resistome driven by air pollutants. Sci. Total Environ. 902 , 165942 (2023). Zheng, D. et al . Metagenomics highlights the impact of climate and human activities on antibiotic resistance genes in China’s estuaries. Environ. Pollut. 301 , 119015 (2022). Pepi, M. & Focardi, S. Antibiotic-resistant bacteria in aquaculture and climate change: A challenge for health in the mediterranean area. Int. J. Environ. Res. Public Health 18 , 5723 (2021). Liu, Z. et al . Deterministic effect of pH on shaping soil resistome revealed by metagenomic analysis. Environ. Sci. Technol. 57 , 985–996 (2023). Xu, Y. et al . Unraveling the determinants of antibiotic resistance evolution in farmland under fertilizations. J. Hazard. Mater. 474 , 134802 (2024). Tecon, R. & Or, D. Biophysical processes supporting the diversity of microbial life in soil. Fems Microbiol. Rev. 41 , 599–623 (2017). Zhang, L. et al . The antibiotic resistance and risk heterogeneity between urban and rural rivers in a pharmaceutical industry dominated city in China: The importance of social-economic factors. Sci. Total Environ. 852 , 158530 (2022). MacFadden, D. R., McGough, S. F., Fisman, D., Santillana, M. & Brownstein, J. S. Antibiotic resistance increases with local temperature. Nat. Clim. Chang. 8 , 510–514 (2018). Headd, B. & Bradford, S. A. Physicochemical factors that favor conjugation of an antibiotic resistant plasmid in non-growing bacterial cultures in the absence and presence of antibiotics. Front. Microbiol. 9 , 2122 (2018). Baomo, L., Lili, S., Moran, R. A., van Schaik, W. & Chao, Z. Temperature-regulated IncX3 plasmid characteristics and the role of plasmid-encoded H-NS in thermoregulation. Front. Microbiol. 12 , 765492 (2022). Baker Austin, C. et al . Emerging Vibrio risk at high latitudes in response to ocean warming. Nat. Clim. Chang. 3 , 73–77 (2013). Wang, X., Meng, X. & Long, Y. Projecting 1 km-grid population distributions from 2020 to 2100 globally under shared socioeconomic pathways. Sci. Data 9 , 563 (2022). Wang, T. & Sun, F. Global gridded GDP data set consistent with the shared socioeconomic pathways. Sci. Data 9 , 221 (2022). Additional Declarations There is NO Competing Interest. Supplementary Files SupplementaryTable1Imformationofthesamplescollectedinthisstudy.xlsx Dataset 1 SupplementaryTable2ARGstypesdetectedinthisstudy.xlsx Dataset 2 SupplementaryTable3ARGssubtypesdetectedinthisstudy.xlsx Dataset 3 SupplementaryTable4ARGresistancemechanismsanddrugclass.xlsx Dataset 4 SupplementaryTable5MGEsdetectedinthisstudy.xlsx Dataset 5 SupplementaryTable6MicroorganisminPhylumleveldetectedinthisstudy.xlsx Dataset 6 SupplementaryTable7HighriskARGfamiliesidentifiedaccordingtotheWorldHealthOrganizationWHO.xlsx Dataset 7 SupplementaryTable8Gridsearchresultsofrandomforest.xlsx Dataset 8 SupplementaryTable9Specificcovariateinformationcollectedinthisstudy.xlsx Dataset 9 SupplementaryTable10The10principalcomponentswithacumulativecontributionrateofatleast80Xandtheirinterpretations.xlsx Dataset 10 SupportingInformation.docx Supporting Information Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7635655","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":527726089,"identity":"94810aa4-af16-4eb7-97d1-66df19daecfc","order_by":0,"name":"Yanzheng Gao","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABCElEQVRIiWNgGAWjYBCDBCBmfMDAA+YYEK2F2YBkLWwSUA5+LfLuh49J/NxRl8cv3X6tukDGLrGBvXmbBEPNHZxaDM+kpUn2nmErlpxzpuz2DJ7kxAaeY2USDMee4dbSkGMmwdvGk7jhRk7abR4e5sQGCaAIY8Nh3Fr635hJ/m2TSNwP1FLMw1Of2CD/Br8WeaCZ0rxtBokbJNKPMfPwHAbawoNfi4HEs2Rr2baExBk3cpileXiOG7fxpBVbJBzDY0t/8sGbb9vqEvtnpD/8zNtTLdvPfnjjjQ81eGw5wMACjQ4eAwbGHmDsgNgJODUAbWlgYP4AYbI/YGD4gUfpKBgFo2AUjFgAAJg2UqdzBN/tAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0002-3814-3555","institution":"Nanjing Agricultural University","correspondingAuthor":true,"prefix":"","firstName":"Yanzheng","middleName":"","lastName":"Gao","suffix":""},{"id":527726090,"identity":"92a57995-ada4-49d7-8b56-5e756ca8cacc","order_by":1,"name":"Zekai Li","email":"","orcid":"","institution":"Nanjing Agricultural University","correspondingAuthor":false,"prefix":"","firstName":"Zekai","middleName":"","lastName":"Li","suffix":""},{"id":527726091,"identity":"73a6ea29-c67c-4478-b953-7946ded80e3c","order_by":2,"name":"Sean Fettrow","email":"","orcid":"","institution":"Colorado State University","correspondingAuthor":false,"prefix":"","firstName":"Sean","middleName":"","lastName":"Fettrow","suffix":""},{"id":527726092,"identity":"fcc53223-fcf6-47b3-9536-c08643124930","order_by":3,"name":"Thomas Borch","email":"","orcid":"","institution":"Colorado State University","correspondingAuthor":false,"prefix":"","firstName":"Thomas","middleName":"","lastName":"Borch","suffix":""},{"id":527726093,"identity":"c8cca1d4-3ace-41aa-a6c5-f0e945cb76d8","order_by":4,"name":"Miao Han","email":"","orcid":"","institution":"College of Resources and Environmental Sciences, Nanjing Agricultural University","correspondingAuthor":false,"prefix":"","firstName":"Miao","middleName":"","lastName":"Han","suffix":""},{"id":527726094,"identity":"b901815c-828f-429b-8ce6-384e36357993","order_by":5,"name":"Antti Karkman","email":"","orcid":"https://orcid.org/0000-0003-0983-3319","institution":"University of Helsinki","correspondingAuthor":false,"prefix":"","firstName":"Antti","middleName":"","lastName":"Karkman","suffix":""},{"id":527726095,"identity":"b1e798e3-5fcb-4f1e-af3b-2a9cb50c60c4","order_by":6,"name":"Xiaojie Hu","email":"","orcid":"","institution":"College of Resources and Environmental Sciences, Nanjing Agricultural University","correspondingAuthor":false,"prefix":"","firstName":"Xiaojie","middleName":"","lastName":"Hu","suffix":""},{"id":527726096,"identity":"1a855ff8-685c-4b70-895f-894392d519a8","order_by":7,"name":"Jian Wang","email":"","orcid":"","institution":"Nanjing Agricultural University","correspondingAuthor":false,"prefix":"","firstName":"Jian","middleName":"","lastName":"Wang","suffix":""},{"id":527726097,"identity":"3f10c483-62a5-4f15-aed1-24c9cdb95308","order_by":8,"name":"Hefei Wang","email":"","orcid":"","institution":"Nanjing Agricultural University","correspondingAuthor":false,"prefix":"","firstName":"Hefei","middleName":"","lastName":"Wang","suffix":""},{"id":527726098,"identity":"0da3d350-8262-4ed6-b298-6e5822b1948e","order_by":9,"name":"Chao Qin","email":"","orcid":"","institution":"Nanjing Agricultural University","correspondingAuthor":false,"prefix":"","firstName":"Chao","middleName":"","lastName":"Qin","suffix":""},{"id":527726099,"identity":"aceddaa7-69c8-4381-b252-9df3d179ffd9","order_by":10,"name":"Jianhua Guo","email":"","orcid":"https://orcid.org/0000-0002-4732-9175","institution":"The University of Queensland","correspondingAuthor":false,"prefix":"","firstName":"Jianhua","middleName":"","lastName":"Guo","suffix":""}],"badges":[],"createdAt":"2025-09-17 04:45:37","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7635655/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7635655/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":93366586,"identity":"e32c85b1-469b-4cdd-a8de-c8357a20b7b5","added_by":"auto","created_at":"2025-10-13 05:08:58","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":327894,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eGlobal distribution and ARG diversity based on agricultural soil metagenomic data.\u003c/strong\u003e \u003cstrong\u003ea\u003c/strong\u003e Geographic information and land use types associated with the metagenomic data collected in this study (n=985). Point size represents the number of samples, and different colors represent different land use types. \u003cstrong\u003eb\u003c/strong\u003e Total ARG abundance across different land use types. Squares represent mean values of total ARG abundance for different soil types. Hollow lines represent median values of total ARG abundances for different soil types (using Kruskal‒Wallis test). \u003cstrong\u003ec\u003c/strong\u003e Composition of ARG types across various soil types. \u003cstrong\u003ed \u003c/strong\u003eComposition of ARG subtypes across various soil types. \u003cstrong\u003ee\u003c/strong\u003e The 20 most prevalent ARG types and subtypes. Identical colors indicate belonging to the same ARG type.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7635655/v1/512b0764d9be935ada538ed5.png"},{"id":93366747,"identity":"a29dd5bf-8d10-4634-999d-fd0d16a8d1d4","added_by":"auto","created_at":"2025-10-13 05:16:58","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":310456,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDistribution of microbial hosts for MGEs and ARGs in agricultural soil environments.\u003c/strong\u003e \u003cstrong\u003ea\u003c/strong\u003e Total MGE abundance across different land use types. Squares represent mean values of total MGE abundance for different soil types. Hollow lines represent median values of total MGE abundances for different soil types (using Kruskal‒Wallis test). \u003cstrong\u003eb \u003c/strong\u003eComposition of MGE types across various soil types.\u003cstrong\u003e c \u003c/strong\u003ePearson’s correlation analysis of the total MGE abundance, transposase, integrase, plasmid, IS91, ist and total abundance of ARGs. \u003cstrong\u003ed\u003c/strong\u003e Procrustes analysis demonstrated a significant correlation between all ARG subtypes and MGE subtypes (n=985). The length of the lines connecting the two points quantifies the differences between ARG subtypes (blue) and MGE subtypes (green) in the same sample. \u003cstrong\u003ee\u003c/strong\u003e Microbial diversity assessed via the Shannon index across different habitats. Hollow lines represent the median values of the Shannon index of microbial diversity for different soil types (Using Kruskal‒Wallis test; pairwise comparison plots are shown in Supplementary Fig. 9).\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7635655/v1/264e797837036d114fb9be02.png"},{"id":93366589,"identity":"0ac6e028-7e21-4698-96f9-ef13c2e1303a","added_by":"auto","created_at":"2025-10-13 05:08:58","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":149670,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eIdentification of core high-risk ARGs in soil metagenome samples. \u003c/strong\u003eQuantification of known high-risk ARG families was identified by the World Health Organization (WHO) in our metagenome samples. \u003cstrong\u003ea \u003c/strong\u003eThe abundance of high-risk (Rank I/II) ARGs across different habitat subtypes. \u003cstrong\u003eb\u003c/strong\u003e The richness of high-risk (Rank I/II) ARGs across different habitat subtypes. \u003cstrong\u003ec\u003c/strong\u003e The proportion of high-risk (Rank I/II) ARGs in all metagenome samples expressed in terms of abundance. \u003cstrong\u003ed\u003c/strong\u003e The proportion of high-risk (Rank I/II) ARGs in all metagenome samples expressed in terms of richness. \u003cstrong\u003ee\u003c/strong\u003e The health risks of ARGs in the samples and the risk levels of the samples after k-means clustering. The numerical values of health risk are normalized to the [0, 1] interval, with each vertical line representing one sample. The sample risk levels are categorized into six levels.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-7635655/v1/45cecb06b821d5d858cd8000.png"},{"id":93366749,"identity":"334551d0-149e-482b-8bf9-f380f033beb9","added_by":"auto","created_at":"2025-10-13 05:16:58","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":205396,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eGlobal mapping of the health risk of ARGs in agricultural soils.\u003c/strong\u003e \u003cstrong\u003ea\u003c/strong\u003e Map of ARGs risk in agricultural soils with prediction results from machine learning, drawn via ArcGIS at a 0.25°×0.25° resolution. \u003cstrong\u003eb\u003c/strong\u003e Feature permutation importance of the variable layers obtained from the random forest model. \u003cstrong\u003ec \u003c/strong\u003eHealth risks of ARGs in agricultural soils in China. \u003cstrong\u003ed\u003c/strong\u003e Health risks of ARGs in agricultural soils in India.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-7635655/v1/190105ea5f3b40be04f4f80b.png"},{"id":93366595,"identity":"087bf884-770a-4cba-8064-cf5fc2447803","added_by":"auto","created_at":"2025-10-13 05:08:58","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":184504,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePrediction of health risks of ARGs in agricultural soils worldwide in different scenarios (SSP126/370/585).\u003c/strong\u003e Risklevels of ARGs in global agricultural soils under three different scenarios: \u003cstrong\u003ea\u003c/strong\u003eSSP126, \u003cstrong\u003eb\u003c/strong\u003e SSP370, and \u003cstrong\u003ec\u003c/strong\u003e SSP585 in 2030. Relationships between high-risk ARGs areas and influencing factors under three scenarios: \u003cstrong\u003ed\u003c/strong\u003e high-risk ARGs areas, \u003cstrong\u003ee\u003c/strong\u003e temperature, \u003cstrong\u003ef\u003c/strong\u003esocioeconomic and land use indicators.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-7635655/v1/532c494e98d8c331eb85a41c.png"},{"id":97895335,"identity":"5531a7d6-3ee5-4c97-9018-73d8532f05e7","added_by":"auto","created_at":"2025-12-10 15:34:01","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1989015,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7635655/v1/c5db9469-e9f0-4b4c-b998-d29de3aaa522.pdf"},{"id":93366588,"identity":"b3e79b34-f220-424b-ab93-a145245bfd3b","added_by":"auto","created_at":"2025-10-13 05:08:58","extension":"xlsx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":106886,"visible":true,"origin":"","legend":"Dataset 1","description":"","filename":"SupplementaryTable1Imformationofthesamplescollectedinthisstudy.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7635655/v1/02184023205b9aba498103c7.xlsx"},{"id":93367217,"identity":"42e43838-a2ce-45df-b244-6638a7f12d66","added_by":"auto","created_at":"2025-10-13 05:24:58","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":117095,"visible":true,"origin":"","legend":"Dataset 2","description":"","filename":"SupplementaryTable2ARGstypesdetectedinthisstudy.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7635655/v1/c59834255570377fb899d4c5.xlsx"},{"id":93366592,"identity":"71c4508e-bf5a-44b6-b4f2-02998181cbda","added_by":"auto","created_at":"2025-10-13 05:08:58","extension":"xlsx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":133973,"visible":true,"origin":"","legend":"Dataset 3","description":"","filename":"SupplementaryTable3ARGssubtypesdetectedinthisstudy.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7635655/v1/d8ea1e0ce93df3008a10d782.xlsx"},{"id":93366593,"identity":"93a83556-a09d-4de8-b62f-eddcf704f025","added_by":"auto","created_at":"2025-10-13 05:08:58","extension":"xlsx","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":60526,"visible":true,"origin":"","legend":"Dataset 4","description":"","filename":"SupplementaryTable4ARGresistancemechanismsanddrugclass.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7635655/v1/cbe1ffb49e47a4a9e0f86bda.xlsx"},{"id":93366601,"identity":"5d05f2a2-82da-434f-91f2-75d08b994d53","added_by":"auto","created_at":"2025-10-13 05:08:59","extension":"xlsx","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":189930,"visible":true,"origin":"","legend":"Dataset 5","description":"","filename":"SupplementaryTable5MGEsdetectedinthisstudy.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7635655/v1/18c01e4f048e02d54acd1174.xlsx"},{"id":93366590,"identity":"e8d53403-07bf-4a66-bcdd-4f90f563fa05","added_by":"auto","created_at":"2025-10-13 05:08:58","extension":"xlsx","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":14158,"visible":true,"origin":"","legend":"Dataset 6","description":"","filename":"SupplementaryTable6MicroorganisminPhylumleveldetectedinthisstudy.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7635655/v1/93dcdc517c25a6c0b67e868d.xlsx"},{"id":93366750,"identity":"ca155965-37b7-4819-97c5-efbb737fb92a","added_by":"auto","created_at":"2025-10-13 05:16:59","extension":"xlsx","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":10782,"visible":true,"origin":"","legend":"Dataset 7","description":"","filename":"SupplementaryTable7HighriskARGfamiliesidentifiedaccordingtotheWorldHealthOrganizationWHO.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7635655/v1/2c5f4a9b5c1e2950d69652ab.xlsx"},{"id":93366594,"identity":"022829b6-328f-455a-ba16-35a1895cd838","added_by":"auto","created_at":"2025-10-13 05:08:58","extension":"xlsx","order_by":8,"title":"","display":"","copyAsset":false,"role":"supplement","size":25854,"visible":true,"origin":"","legend":"Dataset 8","description":"","filename":"SupplementaryTable8Gridsearchresultsofrandomforest.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7635655/v1/c8f216dfe708cedc261cafc1.xlsx"},{"id":93366600,"identity":"b196f41d-63af-4bd1-86ca-0991518cebee","added_by":"auto","created_at":"2025-10-13 05:08:59","extension":"xlsx","order_by":9,"title":"","display":"","copyAsset":false,"role":"supplement","size":18110,"visible":true,"origin":"","legend":"Dataset 9","description":"","filename":"SupplementaryTable9Specificcovariateinformationcollectedinthisstudy.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7635655/v1/f6ed86670e11ec8e41cbc732.xlsx"},{"id":93366599,"identity":"df729a1a-96cc-4299-8ee5-ccc9c8bcb653","added_by":"auto","created_at":"2025-10-13 05:08:59","extension":"xlsx","order_by":10,"title":"","display":"","copyAsset":false,"role":"supplement","size":20798,"visible":true,"origin":"","legend":"Dataset 10","description":"","filename":"SupplementaryTable10The10principalcomponentswithacumulativecontributionrateofatleast80Xandtheirinterpretations.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7635655/v1/05db25ab64f673fe5c6d1cdb.xlsx"},{"id":93366596,"identity":"dbd9bc1d-f343-4d14-875c-73593c6fa825","added_by":"auto","created_at":"2025-10-13 05:08:58","extension":"docx","order_by":11,"title":"","display":"","copyAsset":false,"role":"supplement","size":10891950,"visible":true,"origin":"","legend":"Supporting Information","description":"","filename":"SupportingInformation.docx","url":"https://assets-eu.researchsquare.com/files/rs-7635655/v1/8cd636afe55d6db6170003b7.docx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Global health risk assessment of antibiotic resistance in agricultural soils","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe transmission of antibiotic resistance genes (ARGs) leads to the development of antibiotic-resistant bacteria (ARB), which has emerged as one of the greatest global health challenges of the 21\u003csup\u003est\u003c/sup\u003e century\u003csup\u003e1,2\u003c/sup\u003e. Soils act as a natural reservoir for both ARGs and ARB to emerge, evolve and disseminate\u003csup\u003e3\u003c/sup\u003e, with agricultural soils being of particular concern. This is because agricultural practices (e.g. application of manure and wastewater) continuously introduce antibiotics (which exert sustained selective pressure for ARG proliferation), ARGs and ARB into soils\u003csup\u003e4–7\u003c/sup\u003e. Edible leafy plant tissue and cereal grains can contain ARG residues\u003csup\u003e8–12\u003c/sup\u003e, sometimes harboring high-risk resistant pathogens\u003csup\u003e13,14\u003c/sup\u003e. As a primary source of ARGs found in agricultural products, soils facilitate the spread of ARGs along the food chain\u003csup\u003e15,16\u003c/sup\u003e. Consequently, the consumption of fresh produce, particularly raw vegetables, is a direct route of human exposure to soil microorganisms and ARGs, posing a substantial threat to human health\u003csup\u003e17\u003c/sup\u003e. Yet, the health risks and future trends of ARGs in global agricultural systems remains largely unknown.\u003c/p\u003e\n\u003cp\u003eAlthough advances in qPCR and metagenomic profiling have delineated ARG distribution patterns in agricultural soils at the regional scale\u003csup\u003e6,18–21\u003c/sup\u003e, there remain essential knowledge gaps at the global scale. Sampling areas are currently focused on specific agroecosystems in similar pedoclimatic zones\u003csup\u003e18,22–24\u003c/sup\u003e. Moreover, risk assessments predominantly rely on abundance metrics, neglecting the synergistic risks from MGE (mobile genetic element) and pathogen tripartite interactions\u003csup\u003e20,25,26\u003c/sup\u003e. Furthermore, conventional linear models fail to capture nonlinear interactions among climatic, edaphic, and agronomic drivers\u003csup\u003e27\u003c/sup\u003e. The above limitations have resulted in incomplete understanding of global distribution patterns and future health risks associated with ARGs. A comprehensive research strategy is urgently required to address these gaps and to construct a comprehensive health risk assessment of ARGs in global agricultural systems.\u003c/p\u003e\n\u003cp\u003eHere, we analyze 985 global agricultural soil metagenomes spanning 7 different crop type systems across 6 continents. Resistance profiles are generated to elucidate the occurrence and distribution patterns of ARGs in agricultural soils worldwide using bioinformatics. Specifically, we develop a risk assessment framework based on the bioinformatics results using a random forest machine learning algorithm. Our framework encompasses global soils and analyzes the contribution of driving factors such as climate change, fertilization intensity, and soil pH to ARG-related health risks. Furthermore, we predict future trends of high-risk areas for 2030 under the Shared Socioeconomic Pathways (SSP) scenarios, which are global development trajectories widely used in climate change research\u003csup\u003e28\u003c/sup\u003e. This work provides the first global risk stratification map of ARGs in agricultural soils, identifies priority areas for control, and provides theoretical support for developing effective strategies to mitigate ARGs and improve the health and development of agricultural systems.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003eARG types and subtypes in global agricultural soils\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe 985 metagenomic samples were distributed across representative countries (e.g., the United States, China, Australia, France, and Brazil) and 6 continents including Asia, Africa, Europe, South America, North America, and Oceania (Fig. 1a and Supplementary Table 1). The soil samples were divided into various soil crop types, including 141 fruit-growing samples, 35 livestock and poultry samples, 462 major crops (maize, rice, and wheat) samples, and 154 vegetable-growing samples. For the 31 samples without specified land use type information in the NCBI database, we determined their land use types according to the latitude and longitude of the samples via a global gridded agricultural production map\u003csup\u003e29\u003c/sup\u003e. Overall, 1745 ARG subtypes from 30 ARG families are identified (Supplementary Fig. 1, Supplementary Table 2, and Table 3). The detection of only a few ARG types, including bacitracin, macrolide-lincosamide-streptogram (MLS), multidrug, polymyxin, and tetracycline, encompasses most of the abundance, while the vast majority of ARG types (1425 out of 1745) makes up considerably less of the abundance (\u0026lt;10%) (Supplementary Table 2). At the more refined subtype level, \u003cem\u003ebacA\u003c/em\u003e, \u003cem\u003eMexB\u003c/em\u003e, \u003cem\u003eMexF\u003c/em\u003e, \u003cem\u003eMuxB\u003c/em\u003e and \u003cem\u003eMuxC\u003c/em\u003e makes up the majority of the abundance (Supplementary Fig. 2, Supplementary Table 3). Statistical tests revealed significant differences in total ARG abundance (RPKM; Reads per kilobase per million) between soil land use types (Fig. 1b). The abundance in livestock and poultry farm soils is greatest (3226.32 RPKM), with a mean value approximately 20 times higher than the other soil types. This was followed, in descending order, by maize-growing soil (155.96 RPKM), fruit-growing soil (140.65 RPKM), vegetable-growing soil (118.42 RPKM), others soil (113.40 RPKM), wheat-growing soil (108.74 RPKM), and rice-growing soil (93.87 RPKM), which showed the lowest ARG abundance.\u003c/p\u003e\n\u003cp\u003eThere are also notable differences in ARG types composition between soils, with the primary difference being between cropland soil types (i.e., fruit, maize, rice etc.) and livestock and poultry soils (Fig. 1c). Bacitracin, polymyxin and rifamycin are widely detected in most cropland soils, but are infrequent in livestock and poultry farm soils. Conversely, aminoglycoside, multidrug, MLS and tetracycline have relatively high abundances in the livestock and poultry soils compared to cropland soil types. (Fig. 1c, Supplementary Table 2). Regarding the distribution of ARG subtypes across soil types, livestock and poultry soils and maize-growing soils displayed distinct compositional profiles compared to other five soil categories (Fig. 1d). These two soil types were characterized by a higher abundance of sulfonamide and MLS-resistant ARG subtypes, but relatively lower proportions of novobiocin and rifamycin resistance subtypes. In contrast, the other five soil types consistently contained over 20% bacitracin-resistant ARGs (specifically the \u003cem\u003ebacA\u003c/em\u003e) and approximately 10% novobiocin-resistant ARGs (specifically the \u003cem\u003enovA\u003c/em\u003e) (Fig. 1d).\u003c/p\u003e\n\u003cp\u003eFor ARG subtypes, \u003cem\u003ebacA\u003c/em\u003e have the greatest abundance, followed by \u003cem\u003enov\u003c/em\u003eA, \u003cem\u003esul1\u003c/em\u003e and \u003cem\u003efloR,\u0026nbsp;\u003c/em\u003eall of which exceed 8000 RPKM (Fig. 1e). For a vast majority of detected ARG subtypes (1487 out of 1745), abundance is comparatively low (i.e., \u0026lt;100 RPKM) (Supplementary Table 2). Antibiotic inactivation and antibiotic efflux are the main resistance mechanisms found in all agricultural soil types (Supplementary Fig. 3, Supplementary Table 4). Furthermore, more than half of the ARGs (1128 out of 1745) have developed resistance to a single drug class, while 617 ARGs harbor resistance to 2~16 drug classes (Supplementary Fig. 4, Supplementary Table 4).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eProfile\u003c/strong\u003e\u003cstrong\u003es of\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003esoil MGEs and ARGs host bacteria\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMobile genetic elements (MGEs) and their host bacteria\u0026nbsp;play a crucial role in promoting the prevalence and persistence of ARGs in soils. Through quantitative annotation, 41 MGE types and 217 MGE subtypes are identified across all soil samples, with total abundance value ranging from 0.0021 to 1726.10 RPKM (Supplementary Fig. 5 and Supplementary Table 5). Among these, transposase and IS91 account for 63.7% and 20.0% of the total detected MGE subtypes, respectively (Supplementary Fig. 5). Statistical tests revealed significant variations in MGE abundance across different soil land use types. Notably, livestock and poultry soils demonstrate the highest MGE abundance, which is significantly greater than that observed in other cropland soils—a pattern consistent with previously reported ARG distributions (Fig. 2a). Regarding the compositional profile of MGEs, transposase is the predominant type across all soil categories, exceeding 50% prevalence, followed by IS91, which consistently represented over 19%. A notable exception is observed in fruit-growing soil, where plasmid content reached 17%, significantly higher than in other soil types. In contrast, \u003cem\u003eist\u003c/em\u003e elements are nearly undetectable in livestock and poultry soils (Fig. 2b).\u003c/p\u003e\n\u003cp\u003eThe Pearson correlation analysis revealed strong positive correlation between the total abundance of ARGs and several major MGE types, including transposase (\u003cem\u003er\u003c/em\u003e=0.81, \u003cem\u003ep\u003c/em\u003e\u0026lt;0.001), integrase (\u003cem\u003er\u003c/em\u003e=0.73, \u003cem\u003ep\u003c/em\u003e\u0026lt;0.001), plasmid (\u003cem\u003er\u003c/em\u003e=0.70, \u003cem\u003ep\u003c/em\u003e\u0026lt;0.001), IS91 (\u003cem\u003er\u003c/em\u003e=0.78, \u003cem\u003ep\u003c/em\u003e\u0026lt;0.0001), and \u003cem\u003eist\u003c/em\u003e (\u003cem\u003er\u003c/em\u003e=0.43, \u003cem\u003ep\u003c/em\u003e\u0026lt;0.001), as well as the total abundance of MGEs (\u003cem\u003er\u003c/em\u003e=0.83, \u003cem\u003ep\u003c/em\u003e\u0026lt;0.0001) (Fig.\u0026nbsp;2c). These results indicate a consistent micro-scale co-occurrence of ARGs with specific MGEs within individual soil samples. At the macro-scale across different soil types, the abundance trends of major MGEs mirrored those of ARGs. Specifically, transposase, integrase, and IS91 show significant higher abundances in livestock and poultry farm soils compared to other soil types (Supplementary Fig. 6), which is highly consistent with the spatial distribution pattern of ARGs described in the previous section. Furthermore, Procrustes analysis reveals a significant concordance between the community structures of ARG and MGE subtypes (M\u003csup\u003e2\u003c/sup\u003e=0.3871, P \u0026lt; 0.001, Fig. 2d), indicating coordinated variation in their profiles across samples. This strong correlation implies that the compositional patterns of ARG subtypes are intimately associated with those of MGE subtypes, likely resulting from shared environmental drivers or direct interactions such as HGT.\u003c/p\u003e\n\u003cp\u003eBacteria identified as potential hosts of ARGs in this study belong to 75 distinct phyla (Supplementary Table 6). \u003cem\u003ePseudomonadota\u003c/em\u003e and \u003cem\u003eActinomycetota\u003c/em\u003e are the most dominant groups, representing 43.6% and 42.8% of the total microbial population, respectively (Supplementary Fig. 7, Supplementary Table 6). Other phyla with relatively high abundances (≥1.8%) include \u003cem\u003eMyxococcota\u0026nbsp;\u003c/em\u003e(2.6%), \u003cem\u003ePlanctomycetota\u0026nbsp;\u003c/em\u003e(1.9%), \u003cem\u003eBacillota\u0026nbsp;\u003c/em\u003e(1.8%) and \u003cem\u003eBacteroidota\u0026nbsp;\u003c/em\u003e(1.8%). The Shannon Index of microbial diversity varied significantly across soil types (Fig. 2e, Supplementary Fig. 9). Specifically, rice-growing soil exhibited the highest microbial diversity (mean value: 5.51), followed closely by maize-growing soil (5.50), whereas livestock and poultry farm soil showed the lowest diversity (4.97). To further identify potential hosts of ARGs and MGEs at phylum level, a Spearman correlation-based co-occurrence network was constructed (Supplementary Fig. 8). The network included 32 phyla, 200 ARG subtypes and 76 MGE subtypes. Phyla such as \u003cem\u003eBacteroidota\u003c/em\u003e, \u003cem\u003eBdellovibrionota\u003c/em\u003e and \u003cem\u003eSpirochaetota\u003c/em\u003e exhibit positive correlations with more than 40 ARG subtypes, suggesting their roles as potential hosts for the associated ARGs and MGEs.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWe further provide an ARG risk framework based on classifications by the World Health Organization (WHO)and following a previous study\u003csup\u003e27\u003c/sup\u003e (Supplementary Table 7). High-risk ARGs (Rank I/II) represent 9–22% of the total abundance (Fig. 3a) and 7-9% of the total richness (Fig. 3b) across the soil types. When all soil types are combined, high-risk ARGs (Rank I/II) represent 17.4% of the total abundance (Fig. 3c) and 7.5% of the total richness (Fig. 3d). Furthermore, a bioinformatics software package previously used to assess ARG risk\u003csup\u003e27\u003c/sup\u003e was used to calculate the risk index (RI) in each sample. The RIs of all metagenome samples were then classified into 6 risk levels (Fig. 3e) using the K-means clustering method.\u0026nbsp;Samples in Ranks 6 and 5 were classified as high-risk and represent 6% of the total ARGs. The standardized risk values for ARGs in Ranks 6 and 5 ranged from 0.12 to 1 and from 0.0045 to 0.11, respectively. The standardized risk values for ARGs classified as Ranks 1 to 4 ranged from 0 to 0.0028.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGlobal ARG risk assessment model in agricultural soils\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe further used a machine learning model to visualize the potential health risks of ARGs in agricultural soils at the global scale. We applied this model to a global dataset, mapping ARG risks at a resolution of 0.25° (Fig. 4a). Most of the obtained ARG high-risk areas (Rank 5/6) are in the Northern Hemisphere, mainly in the North China Plain and North India (Fig. 4c, d), both of which are major global agricultural production areas. This map highlights the distribution of risks and threats of ARGs in agricultural soils worldwide, providing critical insights for ARG detection and control. The model demonstrated robust predictive performance, as evidenced by the receiver operating characteristic (ROC) curve and confusion matrix (Supplementary Fig. 10 and 11).\u003c/p\u003e\n\u003cp\u003eVariable importance analysis identified “Climate, Soil pH, and Moisture” as the most important factor affecting ARG risk in our model, consistent with previous reports (Fig. 4b)\u003csup\u003e23,28\u003c/sup\u003e. Precipitation and soil texture are also identified as influential yet secondary factors.\u0026nbsp;Significant correlations were observed between these predictors and ARG risk levels (Supplementary Fig. 12). In particular, “Climate, Soil pH, and Moisture” and “Temperature” showed strong positive correlations with ARG risk across a broad range of values. In contrast, factors such as “Agricultural Production and Fertilizer Use” and “Carbon and Nitrogen Content” display nonlinear or threshold-dependent relationships with risk.\u003c/p\u003e\n\u003cp\u003eWe further predicted the spatial distribution of ARG risks under future scenarios using a modeling approach that incorporated projections of key drivers, including future antibiotic consumption, socioeconomic development, and climate change—with increased soil temperature serving as a primary climatic effector. Projections were made under three Shared Socioeconomic Pathways (SSPs), namely SSP1-2.6 (sustainability), SSP3-7.0 (regional rivalry), SSP5-8.5 (fossil-fueled development)\u003csup\u003e28\u003c/sup\u003e (Fig. 5a-c), representing divergent trajectories of global change. Under all scenarios, a substantial expansion of high-risk areas was predicted (Fig. 5d). These shifts in risk levels are closely associated with spatial patterns of temperature change, socioeconomic intensification, and land-use modification (Fig. 5e-f), consistent with previous findings\u003csup\u003e29\u003c/sup\u003e. In particular, future antibiotic use was identified as a major input variable contributing to risk increases in the model. Overall, these results highlight pronounced regional differences in ARG risk and underscore the influence of future environmental and anthropogenic factors on risk dynamics in agricultural soils. Further validation through experimental and monitoring data remains necessary.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDiscussion\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLivestock and poultry farms have increased ARG risk\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA total of 985 global soil samples were analyzed to characterize ARG profiles in agricultural soils (Fig. 1a) and our findings reveal significant variations in ARG distribution across different agricultural soil types. Specifically, both livestock and poultry soils harbor significantly greater ARG abundance compared to other agricultural soil types (Fig. 1b). The predominant ARG types in livestock and poultry farm soil are aminoglycoside, MLS, and tetracycline-related ARGs (Fig. 1c), suggesting that ARGs in these soils exhibit higher activity and potential health risks. This elevated ARG activity is likely due to the prevalent use of antibiotics in these locations, since livestock and poultry feces are major sources of ARGs\u003csup\u003e30\u003c/sup\u003e, and contribute to the substantial influx of antibiotics and ARGs into the soil environment.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn contrast, significantly lower ARG abundances were found in fruit-growing soil, major crops (maize, rice, and wheat) soil, and vegetable-growing soil. This is likely due to fewer contamination sources\u003csup\u003e31\u003c/sup\u003e, as well as the mitigating effect of plants\u003csup\u003e32,33\u003c/sup\u003e. Across all soil types, \u003cem\u003ebac\u003c/em\u003eA was the most abundant and widely detected gene (100% prevalence), which is commensurate with findings from preivous studies\u003csup\u003e34,35\u003c/sup\u003e.This suggests that \u003cem\u003ebacA\u003c/em\u003e could serve as a broad indicator for ARGs across different environments. Furthermore, \u003cem\u003eBac\u003c/em\u003eA has demonstrated resistance to degradation in compost and was associated with bacitracin, indicating the the necessity for stricter regulation of bacitracin use\u003csup\u003e34\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRisks associated with ARG specific types and subtypes\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe transmission and potential health risks of ARGs is closely associated with MGEs (mobile genetic elements) through horizontal gene transfer (HGT). By analyzing the correlation between ARG abundance and MGE abundance, we confirmed that ARG abundance had a positive correlation with MGE abundance (Fig. 2b).\u0026nbsp;Importantly, our comparative analysis revealed significant enrichments of transposases, integrases, and IS91 in livestock and poultry farm soil compared to other croplands, suggesting high HGT frequency mediated by these MGEs. This phenomenon may be attributed to the intensive antibiotic selection pressure of this environment. This soil type is a primary source for antibiotic application, which contributes to a different ARG profile compared to other croplands. Furthermore, the Shannon index variability in livestock and poultry soil is notably complex, indicating the complexity of antibiotic contamination in these soils (Fig. 2c).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe risk level of ARGs is not always directly correlated with their abundance in the environment; even low-abundance ARGs can pose a threat to human health. Assessing the health risks associated with ARGs also requires assessing their ability to colonize and proliferate in the human body, as well as their adaptability and virulence\u0026nbsp;\u003csup\u003e36\u003c/sup\u003e. A comprehensive risk assessment of ARGs also involves complex biological processes (such as HGT and host pathogenicity), which, unlike other environmental pollutants, cannot be easily quantified using clear quantitative analysis\u003csup\u003e37\u003c/sup\u003e. We used a quantitative assessment of ARG risk levels, which contained three critical nodes of ARG transmission from the environment to humans: i) enrichment of ARGs in human environments; ii) ARG mobility; and iii) host pathogenicity\u003csup\u003e38\u003c/sup\u003e. This framework simultaneously accounts for ARG types and abundance, the significant role of MGEs in the HGT of ARGs, as well as the pathogenicity of antibiotic-resistant bacteria. According to our model annotation results,\u0026nbsp;5.99% of the ARGs detected in our agricultural soil dataset are classified as high risk (Rank 5/6) (Fig. 3e), which is lower than that reported in hospital wastewater (23.76%)\u003csup\u003e39\u003c/sup\u003e and global groundwater (10.76%)\u003csup\u003e40\u003c/sup\u003e, but higher than that in freshwater aquaculture ponds (5.25%)\u003csup\u003e41\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGlobal agricultural soil ARG risk assessment framework\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe constructed a global risk assessment framework based on a machine learning model and mapped the associated health risks of ARGs. Our model indicates that ~0.90% of the data points represent high-risk ARG regions for the global agricultural system. These regions are predominantly located in the North China Plain and North India (Fig. 4c-d). In 2021, these two agricultural regions ranked 4th, and 2nd globally in terms of agricultural area, covering 1.09×10\u003csup\u003e8\u003c/sup\u003e, and 1.54×10\u003csup\u003e8\u003c/sup\u003e hectares, respectively (https://data.worldbank.org/). Therefore, these hotspots of elevated ARG risk poses unparalleled concern that necessitates global attention.\u003c/p\u003e\n\u003cp\u003eOur model revealed that “Climate, Soil pH and Moisture” and “Precipitation” are significantly correlated with the distribution patterns of ARG health risks (Fig. 4b). Climate has previously been reported as an important factor affecting global distribution of ARGs\u003csup\u003e25,42–45\u003c/sup\u003e. We illustrate the impact of climate change on potential changes to soil ARG health risks (Fig. 4b).Additionally, pH is considered a key factor in predictive models and plays a crucial role in shaping soil microbial communities, thereby influencing the composition of ARGs\u003csup\u003e46,47\u003c/sup\u003e.\u0026nbsp;Precipitation is considered a critical influencer of soil microbiota at a global scale\u003csup\u003e41,44,48\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eOur scenario-based predictions reveal that ARG health risks will escalate synergistically with rising socioeconomic and land use pressures (Fig. 5d-f). In our model, “socioeconomic and land use indicators” primarily refers to changes in GDP and population. As scenarios such as SSP126, SSP370, and SSP585 evolve, economic activity and population growth could lead to increased urbanization and heightened demands on agricultural activities, which might contribute to the excessive use of antibiotics\u003csup\u003e49\u003c/sup\u003e. The region with the highest-ranked ARGs (rank 6) was mainly distributed in China. In ResistanceMap (https://resistancemap.onehealthtrust.org/Animals.php), China has the highest antimicrobial consumption in the world, reaching 319 mg per population correction unit (PCU), significantly higher than South Korea’s 188 mg per PCU and Spain’s 182 mg per PCU, this could explain the reason of high-risk areas. Global livestock antibiotic consumption is expected to increase by 52% by 2030, since China and India emerging as the fastest-growing contributors (13% and 18% increases, respectively) (https://resistancemap.onehealthtrust.org/Animals.php). Our results indicate that if sustainable development pathways are implemented (Fig. 5d), ARG risks could be reduced by 47% globally. In addition, ARG health risks are expected to escalate synergistically with rising global temperatures\u003csup\u003e50\u003c/sup\u003e.\u0026nbsp;This phenomenon is primarily mediated by two mechanisms. The first mechanism is the thermodynamic enhancement of HGT. Elevated temperatures can increase plasmid conjugation rates\u003csup\u003e51,52\u003c/sup\u003e. And the second mechanism is pathogen proliferation. Pathogenic bacteria play a significant role in the dissemination of ARGs. These bacteria exhibit increased activity with increasing temperature\u003csup\u003e53\u003c/sup\u003e.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn summary, we have investigated the global distribution of ARGs in different agricultural soil environments, identifying a total of 1745 ARG subtypes of 30 types. Several ARGs were consistently detected across soil types, highlighting their potential roles in the global soil ARG profile. Through a quantitative risk assessment framework built from a machine learning model, the highest-risk regions have been identified, particularly in the North China Plain and India. Predictive assessments revealed that high economic activity and population densities may impact future ARG risk, while climate and increased temperature was also an important variable affecting ARG risk. We might reduce ARG risks by following sustainable development pathways while mitigating the impacts of climate change. Our research provides a global risk assessment framework and predictive analysis which may guide future ARG risk research.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003eMetagenome dataset collection\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFrom September 10, 2022, to February 10, 2024, we conducted a search in the NCBI SRA database (https://ncbi.nlm.nih.gov/sra/) via the keywords \u0026ldquo;agricultural soil\u0026rdquo;. The following screening criteria were applied to standardize the subsequent analysis steps: (1) exclusion of experimental treatment groups and rhizosphere soil samples; (2) selection of sequencing samples from platforms labeled \u0026ldquo;paired\u0026rdquo;, \u0026ldquo;genome\u0026rdquo;, and \u0026ldquo;illumina\u0026rdquo; while filtering out data generated from single layouts, exomes, and other sequencing platforms; and (3) downloading raw SRA data for reanalysis. These criteria helped mitigate potential uncertainties in experimental results arising from varied sequencing standards and sample types. Ultimately, we obtained 985 metagenomic samples and compiled their library information, size, land use classification, geographical coordinates (latitude and longitude), continent, country of origin, and DOI, as listed in Supplementary Table 1.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eARGs, MGEs, taxonomic annotation, and abundance calculations\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe raw data from the metagenomic samples were subject to quality assessment and filtered via fastp (v0.23.3; https://github.com/OpenGene/fastp). The required length was 50, and the qualified quality phred was 20. The ARGs in the samples were subsequently annotated via ARG-OAP (v3.2.2; https://github.com/xinehc/ARGs_OAP) with default parameters. For the annotation of MGEs, we utilized a specialized database (https://github.com/KatariinaParnanen/MobileGeneticElementDatabase) containing 278 gene name annotations and over 2000 unique sequences, employing the same default parameters as those used for ARG annotation. Relative abundance comparisons among different samples were conducted using reads per kilobase per million mapped reads (RPKM) as units. To identify the hosts of ARGs and MGEs within the metagenome, we employed kraken2 (v2.1.2; https://github.com/DerrickWood/kraken2) with default parameters. Species abundance statistics based on kraken2 annotations were performed using bracken.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eQuantification of the risk of ARGs to humans\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe assessed the risk associated with ARGs in the metagenome via Arg_ranker (v3.5; https://github.com/caozhichongchong/arg_ranker). The health risk index (RI) calculation method is as follows:\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003cimg src=\"data:image/png;base64,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\" width=\"746\" height=\"91\"\u003e\u003c/p\u003e\n\u003cp\u003eThe rank percentage, ranking risk code, and total abundance were calculated with the default parameters of Arg_ranker. This approach was used to quantify the health risk posed by ARGs in agricultural soils based on the RI.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMachine learning and global mapping\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo predict the global health risks of ARGs in agricultural soil, we prepared a total of 65 gridded covariates, encompassing climate, soil, agricultural, and socioeconomic factors. The climate data included 19 bioclimatic variables were downloaded from WorldClim (https://www.worldclim.org/). Soil information and characteristics were downloaded from SoilGrids (https://www.soilgrids.org/). And other variables, including the number of livestock, human population, GDP and global antibiotic usage, were downloaded from databases such as the Food and Agriculture Organization (FAO; https://www.fao.org/livestock-systems/en/), Socioeconomic Data and Applications Center (SEDAC; https://sedac.ciesin.columbia.edu/), and ResistanceMap (https://resistancemap.onehealthtrust.org/index.php). Specific covariate information is presented in Supplementary Table 9. For tabular data, we used ArcGIS (v10.8) to allocate all values based on administrative boundaries. For raster layers with different resolutions, we used the nearest neighbor method to resample all raster data to a resolution of 0.25\u0026deg; (approximately 27.75 km at the equator). Our resampled covariate layers had latitude and longitude values ranging from -180\u0026deg; to 180\u0026deg; and from -90\u0026deg; to 90\u0026deg;, respectively. For those maps that did not fall within the latitude or longitude range, we redefined the latitude and longitude value range via the \u0026ldquo;Mosaic to New Raster\u0026rdquo; method to ensure that all covariate layers had the same range.\u003c/p\u003e\n\u003cp\u003eTo enhance the interpretability of variables, we conducted a factor analysis. Specifically, we started with a Principal Component Analysis (PCA) to compute the correlation matrix and extract eigenvalues. These eigenvalues helped us determine the contribution rate of each principal component, as well as the cumulative contribution rate. We selected the principal components with a cumulative contribution rate of at least 80%, resulting in 10 retained variables. Subsequently, a Varimax rotation was applied to improve the interpretability of the factor loadings. The 10 principal components and their explanations are presented in Supplementary Table 10.\u003c/p\u003e\n\u003cp\u003eA random forest combined with 10-fold cross validation was used for machine learning. The random forest algorithm is an ensemble learning algorithm that combines the modelling capabilities of decision trees with the advantages of ensemble learning. To enhance the model\u0026rsquo;s generalization ability, 10-fold cross-validation was used. Relying on the factor loading results, we regenerated the data for machine learning. Hence, the new dataset was randomly divided into 10 equally sized subsets with 9 subsets used for training and 1 subset for testing in each model fitting process. This procedure was repeated 10 times until each subset was used as a testing set. Additionally, the result of this model was determined by averaging the 10 fitting data results, and the fitting ability of the model was evaluated via confusion matrixes and ROC curves aligned with the utility of mean imputation to fill the missing values.\u003c/p\u003e\n\u003cp\u003eIn addition, global data from 65 covariate layers were also regenerated based on the factor loading results and excluded regions lacking data, which made our model more accurate. Afterwards, the 78301 remaining data points were available for prediction, followed by visualizing the anticipated health risks caused by ARGs existence in agricultural soils worldwide via ArcGIS (v10.8) at a resolution of 0.25\u0026deg;. To predict the future health risks of ARGs in global agricultural soils, we collected antibiotic usage data for 2030 from ResistanceMap (https://resistancemap.onehealthtrust.org/index.php), grid-based population and GDP data from two studies corresponding to 2030\u003csup\u003e53\u0026ndash;55\u003c/sup\u003e,\u0026nbsp;and climate data from BCC-CSM2-MR, which is a version of the Coupled Model Intercomparison Project Phase 6 (CMIP6) model. CMIP6 is a climate comparison program organized by the World Climate Research Programme (WCRP), and the associated products provide important insights for predicting future climate change and climate impacts. Subsequently, we have chosen three prediction scenarios for future development (SSP126, SSP370 and SSP585) operating as follows: SSP126 represents sustainable development pathways with low climate change challenges, SSP370 is regionally differentiated and high climate change challenges, finally SSP585 that describes traditional development pathways dominated by fossil fuels. For each future scenario, we obtained 19 bioclimatic variables. The results were visualized via ArcGIS (v10.8) at a resolution of 0.25\u0026deg;.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistical analysis and visualization\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eData preprocessing and analysis was accomplished using Microsoft Excel (v16.0). In R (v4.3.3), ggplot2 was used to plot geographic location maps of metagenomic samples, Vegan package was implemented for Procrustes analysis and plotting further relationships between ARGs and MGEs, but the psych package was applied for factor analysis. OriginLab (v2024b) was used to plot the abundance and richness of ARGs, as well as the compositions of and relationships between ARGs and MGEs. Gephi (v0.10.1) was used to construct co-occurrence networks of ARGs, MGEs, and microorganisms. Machine learning was performed using Jupyter Notebook (v6.5.4), and all generated maps were visualized with ArcGIS (v10.8).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eReporting summary\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFurther information on research design is available in the Nature Portfolio Reporting Summary linked to this article.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll of the data supporting this research are available in the main text and the Supplementary Information, and original data can be obtained from the corresponding authors upon reasonable request.\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eC.Q. and Y.G. acknowledge the support from the Joint Funds of the National Key Research and Development Program of China (2023YFE0110800, 2023YFC3708103) and the National Natural Science Foundation of China (42477419, 42107221, 22161132011, U22A20590).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll authors contributed to the work presented in this paper. Y.G. conceived the study. Z.L. completed the majority of the experiments and wrote the original paper. M.H. helped to conduct the theoretical calculations. S.F., T.B., A.K., X.H., J.W., and H.W. revised the manuscript. C.Q., J.G., and Y.G. proposed and discussed the concepts of this study and made extensive revisions to this work.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAntimicrobial resistance: global report on surveillance. https://www.who.int/publications/i/item/9789241564748.\u003c/li\u003e\n\u003cli\u003eOkeke, I. N. et al\u003cem\u003e.\u003c/em\u003e The scope of the antimicrobial resistance challenge. \u003cem\u003eThe Lancet\u003c/em\u003e \u003cstrong\u003e403\u003c/strong\u003e, 2426\u0026ndash;2438 (2024).\u003c/li\u003e\n\u003cli\u003eGoh, Y. X. et al\u003cem\u003e.\u003c/em\u003e Evidence of horizontal gene transfer and environmental selection impacting antibiotic resistance evolution in soil-dwelling Listeria. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 10034 (2024).\u003c/li\u003e\n\u003cli\u003eUdikovic-Kolic, N., Wichmann, F., Broderick, N. A. \u0026amp; Handelsman, J. Bloom of resident antibiotic-resistant bacteria in soil following manure fertilization. \u003cem\u003eProc. Natl. Acad. Sci. U. S. A.\u003c/em\u003e \u003cstrong\u003e111\u003c/strong\u003e, 15202\u0026ndash;15207 (2014).\u003c/li\u003e\n\u003cli\u003eKuppusamy, S. et al\u003cem\u003e.\u003c/em\u003e Veterinary antibiotics (VAs) contamination as a global agro-ecological issue: A critical view. \u003cem\u003eAgric. Ecosyst. Environ.\u003c/em\u003e \u003cstrong\u003e257\u003c/strong\u003e, 47\u0026ndash;59 (2018).\u003c/li\u003e\n\u003cli\u003eChen, C. et al\u003cem\u003e.\u003c/em\u003e Occurrence of antibiotics and antibiotic resistances in soils from wastewater irrigation areas in Beijing and Tianjin, China. \u003cem\u003eEnviron. Pollut.\u003c/em\u003e \u003cstrong\u003e193\u003c/strong\u003e, 94\u0026ndash;101 (2014).\u003c/li\u003e\n\u003cli\u003eChen, C. et al\u003cem\u003e.\u003c/em\u003e Effect of antibiotic use and composting on antibiotic resistance gene abundance and resistome risks of soils receiving manure-derived amendments. \u003cem\u003eEnviron. Int.\u003c/em\u003e \u003cstrong\u003e128\u003c/strong\u003e, 233\u0026ndash;243 (2019).\u003c/li\u003e\n\u003cli\u003eCampos, J. et al\u003cem\u003e.\u003c/em\u003e Microbiological quality of ready-to-eat salads: An underestimated vehicle of bacteria and clinically relevant antibiotic resistance genes. \u003cem\u003eInt. J. Food Microbiol.\u003c/em\u003e \u003cstrong\u003e166\u003c/strong\u003e, 464\u0026ndash;470 (2013).\u003c/li\u003e\n\u003cli\u003eGuo, Y. et al\u003cem\u003e.\u003c/em\u003e Diversity and abundance of antibiotic resistance genes in rhizosphere soil and endophytes of leafy vegetables: Focusing on the effect of the vegetable species. \u003cem\u003eJ. Hazard. Mater.\u003c/em\u003e \u003cstrong\u003e415\u003c/strong\u003e, 125595 (2021).\u003c/li\u003e\n\u003cli\u003eZhao, C. X., Su, X. X., Xu, M. R., An, X. L. \u0026amp; Su, J. Q. Uncovering the diversity and contents of gene cassettes in class 1 integrons from the endophytes of raw vegetables. \u003cem\u003eEcotox. Environ. Safe.\u003c/em\u003e \u003cstrong\u003e247\u003c/strong\u003e, 114282 (2022).\u003c/li\u003e\n\u003cli\u003eZhou, S. Y. D. et al\u003cem\u003e.\u003c/em\u003e Prevalence of antibiotic resistome in ready-to-eat salad. \u003cem\u003eFront. Public Health\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 92 (2020).\u003c/li\u003e\n\u003cli\u003eYu, Y. et al\u003cem\u003e.\u003c/em\u003e Plants select antibiotic resistome in rhizosphere in early stage. \u003cem\u003eSci. Total Environ.\u003c/em\u003e \u003cstrong\u003e858\u003c/strong\u003e, 159847 (2023).\u003c/li\u003e\n\u003cli\u003eZekar, F. M., Granier, S. A., Touati, A. \u0026amp; Millemann, Y. Occurrence of third-generation cephalosporins-resistant klebsiella pneumoniae in fresh fruits and vegetables purchased at markets in Algeria. \u003cem\u003eMicrob. Drug Resist.\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 353\u0026ndash;359 (2020).\u003c/li\u003e\n\u003cli\u003eNkhebenyane, S. J., Khasapane, N. G., Lekota, K. E., Thekisoe, O. \u0026amp; Ramatla, T. Insight into the prevalence of extended-spectrum \u0026beta;-Lactamase-producing enterobacteriaceae in vegetables: A systematic review and meta-analysis. \u003cem\u003eFoods\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 3961 (2024).\u003c/li\u003e\n\u003cli\u003eGao, F. Z. et al\u003cem\u003e.\u003c/em\u003e Untreated swine wastes changed antibiotic resistance and microbial community in the soils and impacted abundances of antibiotic resistance genes in the vegetables. \u003cem\u003eSci. Total Environ.\u003c/em\u003e \u003cstrong\u003e741\u003c/strong\u003e, 140482 (2020).\u003c/li\u003e\n\u003cli\u003eMarti, R. et al\u003cem\u003e.\u003c/em\u003e Impact of manure fertilization on the abundance of antibiotic-resistant bacteria and frequency of detection of antibiotic resistance genes in soil and on vegetables at harvest. \u003cem\u003eAppl. Environ. Microbiol.\u003c/em\u003e \u003cstrong\u003e79\u003c/strong\u003e, 5701\u0026ndash;5709 (2013).\u003c/li\u003e\n\u003cli\u003eCordovez, V., Dini Andreote, F., Carrion, V. J. \u0026amp; Raaijmakers, J. M. Ecology and evolution of plant microbiomes.\u003cem\u003e Annu. Rev. Microbiol.\u003c/em\u003e \u003cstrong\u003e73\u003c/strong\u003e, 69-88 (2019).\u003c/li\u003e\n\u003cli\u003eZhou, Y., Niu, L., Zhu, S., Lu, H. \u0026amp; Liu, W. Occurrence, abundance, and distribution of sulfonamide and tetracycline resistance genes in agricultural soils across China. \u003cem\u003eSci. Total Environ.\u003c/em\u003e \u003cstrong\u003e599\u003c/strong\u003e, 1977\u0026ndash;1983 (2017).\u003c/li\u003e\n\u003cli\u003eNogrado, K., Unno, T., Hur, H. G. \u0026amp; Lee, J. H. Tetracycline-resistant bacteria and ribosomal protection protein genes in soils from selected agricultural fields and livestock farms. \u003cem\u003eAppl. Biol. Chem.\u003c/em\u003e \u003cstrong\u003e64\u003c/strong\u003e, 42 (2021).\u003c/li\u003e\n\u003cli\u003eWu, J. et al\u003cem\u003e.\u003c/em\u003e Antibiotics and antibiotic resistance genes in agricultural soils: A systematic analysis. \u003cem\u003eCrit. Rev. Environ. Sci. Technol.\u003c/em\u003e \u003cstrong\u003e53\u003c/strong\u003e, 847\u0026ndash;864 (2023).\u003c/li\u003e\n\u003cli\u003eWang, B. et al\u003cem\u003e.\u003c/em\u003e Tackling soil ARG-carrying pathogens with global-scale metagenomics. \u003cem\u003eAdv. Sci.\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 10 (2023).\u003c/li\u003e\n\u003cli\u003eSalam, L. B. Metagenomic insights into the microbial community structure and resistomes of a tropical agricultural soil persistently inundated with pesticide and animal manure use. \u003cem\u003eFolia Microbiol.\u003c/em\u003e \u003cstrong\u003e67\u003c/strong\u003e, 707\u0026ndash;719 (2022).\u003c/li\u003e\n\u003cli\u003eCadena, M. et al\u003cem\u003e.\u003c/em\u003e Tetracycline and sulfonamide antibiotic resistance genes in soils from nebraska organic farming operations. \u003cem\u003eFront. Microbiol.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 1283 (2018).\u003c/li\u003e\n\u003cli\u003eSun, J. et al\u003cem\u003e.\u003c/em\u003e Antibiotic resistance genes (ARGs) in agricultural soils from the Yangtze River Delta, China. \u003cem\u003eSci. Total Environ.\u003c/em\u003e \u003cstrong\u003e740\u003c/strong\u003e, 140001 (2020).\u003c/li\u003e\n\u003cli\u003eZheng, D. et al\u003cem\u003e.\u003c/em\u003e Global biogeography and projection of soil antibiotic resistance genes. \u003cem\u003eSci. Adv.\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, eabq8015 (2022).\u003c/li\u003e\n\u003cli\u003eFang, H. et al\u003cem\u003e.\u003c/em\u003e Dissemination of antibiotic resistance genes and human pathogenic bacteria from a pig feedlot to the surrounding stream and agricultural soils. \u003cem\u003eJ. Hazard. Mater.\u003c/em\u003e \u003cstrong\u003e357\u003c/strong\u003e, 53\u0026ndash;62 (2018).\u003c/li\u003e\n\u003cli\u003eGhannam, R. B. \u0026amp; Techtmann, S. M. Machine learning applications in microbial ecology, human microbiome studies, and environmental monitoring. \u003cem\u003eComp. Struct. Biotechnol. J.\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 1092\u0026ndash;1107 (2021).\u003c/li\u003e\n\u003cli\u003eRiahi, K. et al\u003cem\u003e.\u003c/em\u003e The shared socioeconomic pathways and their energy, land use, and greenhouse gas emissions implications: An overview. \u003cem\u003eGlobal Environ. Chang.\u003c/em\u003e \u003cstrong\u003e42\u003c/strong\u003e, 153\u0026ndash;168 (2017).\u003c/li\u003e\n\u003cli\u003eYu, Q. et al\u003cem\u003e.\u003c/em\u003e A cultivated planet in 2010-Part 2: The global gridded agricultural-production maps. \u003cem\u003eEarth Syst. Sci. Data\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 3545\u0026ndash;3572 (2020).\u003c/li\u003e\n\u003cli\u003eDuan, M. et al\u003cem\u003e.\u003c/em\u003e Factors that affect the occurrence and distribution of antibiotic resistance genes in soils from livestock and poultry farms. \u003cem\u003eEcotox. Environ. Safe. \u003c/em\u003e\u003cstrong\u003e180\u003c/strong\u003e, 114\u0026ndash;122 (2019).\u003c/li\u003e\n\u003cli\u003eWang, J., Zhang, Q., Chu, H., Shi, Y. \u0026amp; Wang, Q. Distribution and co-occurrence patterns of antibiotic resistance genes in black soils in Northeast China. \u003cem\u003eJ. Environ. Manage.\u003c/em\u003e \u003cstrong\u003e319\u003c/strong\u003e, 115640 (2022).\u003c/li\u003e\n\u003cli\u003eLi, S. et al\u003cem\u003e.\u003c/em\u003e Plant diversity reduces the risk of antibiotic resistance genes in agroecosystems. \u003cem\u003eAdv. Sci.\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, e2410990 (2025).\u003c/li\u003e\n\u003cli\u003eShen, Y. et al\u003cem\u003e.\u003c/em\u003e Dominant microbiome iteration and antibiotic resistance genes propagation way dictate the antibiotic resistance genes contamination degree in soil-plant system. \u003cem\u003eJ. Clean Prod.\u003c/em\u003e \u003cstrong\u003e464\u003c/strong\u003e, 142786 (2024).\u003c/li\u003e\n\u003cli\u003eYue, Z. et al\u003cem\u003e.\u003c/em\u003e Antibiotic degradation dominates the removal of antibiotic resistance genes during composting. \u003cem\u003eBioresour. Technol.\u003c/em\u003e \u003cstrong\u003e344\u003c/strong\u003e, 126229 (2022).\u003c/li\u003e\n\u003cli\u003eGao, Q. et al\u003cem\u003e.\u003c/em\u003e Diverse and abundant antibiotic resistance genes from mariculture sites of China\u0026rsquo;s coastline. \u003cem\u003eSci. Total Environ.\u003c/em\u003e \u003cstrong\u003e630\u003c/strong\u003e, 117\u0026ndash;125 (2018).\u003c/li\u003e\n\u003cli\u003eManaia, C. M. Assessing the risk of antibiotic resistance transmission from the environment to humans: Non-direct proportionality between abundance and risk. \u003cem\u003eTrends Microbiol.\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 173\u0026ndash;181 (2017).\u003c/li\u003e\n\u003cli\u003eSanderson, H., Johnson, D. J., Wilson, C. J., Brain, R. A. \u0026amp; Solomon, K. R. Probabilistic hazard assessment of environmentally occurring pharmaceuticals toxicity to fish, daphnids and algae by ECOSAR screening. \u003cem\u003eToxicol. Lett.\u003c/em\u003e \u003cstrong\u003e144\u003c/strong\u003e, 383\u0026ndash;395 (2003).\u003c/li\u003e\n\u003cli\u003eZhang, A. N. et al\u003cem\u003e.\u003c/em\u003e An omics-based framework for assessing the health risk of antimicrobial resistance genes. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 4765 (2021).\u003c/li\u003e\n\u003cli\u003eShuai, X. et al\u003cem\u003e.\u003c/em\u003e Ranking the risk of antibiotic resistance genes by metagenomic and multifactorial analysis in hospital wastewater systems. \u003cem\u003eJ. Hazard. Mater.\u003c/em\u003e \u003cstrong\u003e468\u003c/strong\u003e, 133790 (2024).\u003c/li\u003e\n\u003cli\u003eLiu, C. et al\u003cem\u003e.\u003c/em\u003e Meta-analysis addressing the characterization and risk identification of antibiotics and antibiotic resistance genes in global groundwater. \u003cem\u003eSci. Total Environ.\u003c/em\u003e \u003cstrong\u003e860\u003c/strong\u003e, 160513 (2023).\u003c/li\u003e\n\u003cli\u003eWang, C., Liu, X., Yang, Y. \u0026amp; Wang, Z. Antibiotic and antibiotic resistance genes in freshwater aquaculture ponds in China: A meta-analysis and assessment. \u003cem\u003eJ. Clean Prod.\u003c/em\u003e \u003cstrong\u003e329\u003c/strong\u003e, 129719 (2021).\u003c/li\u003e\n\u003cli\u003eHendriksen, R. S. et al\u003cem\u003e.\u003c/em\u003e Global monitoring of antimicrobial resistance based on metagenomics analyses of urban sewage. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 1124 (2019).\u003c/li\u003e\n\u003cli\u003eGao, M., Zhang, Q., Lei, C., Lu, T. \u0026amp; Qian, H. Atmospheric antibiotic resistome driven by air pollutants. \u003cem\u003eSci. Total Environ.\u003c/em\u003e \u003cstrong\u003e902\u003c/strong\u003e, 165942 (2023).\u003c/li\u003e\n\u003cli\u003eZheng, D. et al\u003cem\u003e.\u003c/em\u003e Metagenomics highlights the impact of climate and human activities on antibiotic resistance genes in China\u0026rsquo;s estuaries. \u003cem\u003eEnviron. Pollut.\u003c/em\u003e \u003cstrong\u003e301\u003c/strong\u003e, 119015 (2022).\u003c/li\u003e\n\u003cli\u003ePepi, M. \u0026amp; Focardi, S. Antibiotic-resistant bacteria in aquaculture and climate change: A challenge for health in the mediterranean area. \u003cem\u003eInt. J. Environ. Res. Public Health\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 5723 (2021).\u003c/li\u003e\n\u003cli\u003eLiu, Z. et al\u003cem\u003e.\u003c/em\u003e Deterministic effect of pH on shaping soil resistome revealed by metagenomic analysis. \u003cem\u003eEnviron. Sci. Technol.\u003c/em\u003e \u003cstrong\u003e57\u003c/strong\u003e, 985\u0026ndash;996 (2023).\u003c/li\u003e\n\u003cli\u003eXu, Y. et al\u003cem\u003e.\u003c/em\u003e Unraveling the determinants of antibiotic resistance evolution in farmland under fertilizations. \u003cem\u003eJ. Hazard. Mater.\u003c/em\u003e \u003cstrong\u003e474\u003c/strong\u003e, 134802 (2024).\u003c/li\u003e\n\u003cli\u003eTecon, R. \u0026amp; Or, D. Biophysical processes supporting the diversity of microbial life in soil. \u003cem\u003eFems Microbiol. Rev.\u003c/em\u003e \u003cstrong\u003e41\u003c/strong\u003e, 599\u0026ndash;623 (2017).\u003c/li\u003e\n\u003cli\u003eZhang, L. et al\u003cem\u003e.\u003c/em\u003e The antibiotic resistance and risk heterogeneity between urban and rural rivers in a pharmaceutical industry dominated city in China: The importance of social-economic factors. \u003cem\u003eSci. Total Environ.\u003c/em\u003e \u003cstrong\u003e852\u003c/strong\u003e, 158530 (2022).\u003c/li\u003e\n\u003cli\u003eMacFadden, D. R., McGough, S. F., Fisman, D., Santillana, M. \u0026amp; Brownstein, J. S. Antibiotic resistance increases with local temperature. \u003cem\u003eNat. Clim. Chang.\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 510\u0026ndash;514 (2018).\u003c/li\u003e\n\u003cli\u003eHeadd, B. \u0026amp; Bradford, S. A. Physicochemical factors that favor conjugation of an antibiotic resistant plasmid in non-growing bacterial cultures in the absence and presence of antibiotics. \u003cem\u003eFront. Microbiol.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 2122 (2018).\u003c/li\u003e\n\u003cli\u003eBaomo, L., Lili, S., Moran, R. A., van Schaik, W. \u0026amp; Chao, Z. Temperature-regulated IncX3 plasmid characteristics and the role of plasmid-encoded H-NS in thermoregulation. \u003cem\u003eFront. Microbiol.\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 765492 (2022).\u003c/li\u003e\n\u003cli\u003eBaker Austin, C. et al\u003cem\u003e.\u003c/em\u003e Emerging Vibrio risk at high latitudes in response to ocean warming. \u003cem\u003eNat. Clim. Chang.\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, 73\u0026ndash;77 (2013).\u003c/li\u003e\n\u003cli\u003eWang, X., Meng, X. \u0026amp; Long, Y. Projecting 1 km-grid population distributions from 2020 to 2100 globally under shared socioeconomic pathways. \u003cem\u003eSci. Data\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 563 (2022).\u003c/li\u003e\n\u003cli\u003eWang, T. \u0026amp; Sun, F. Global gridded GDP data set consistent with the shared socioeconomic pathways. \u003cem\u003eSci. Data\u003c/em\u003e\u003cstrong\u003e9\u003c/strong\u003e, 221 (2022).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-7635655/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7635655/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Antibiotic resistance is posing a major threat to public health, yet there is little effort to quantitatively assess risks associated with antibiotic resistance genes (ARGs), particularly in agricultural soils. We therefore propose a risk assessment framework by integrating metagenomic profiling, quantitative health risk assessments, and machine learning to evaluate distribution and health risks of ARGs in a diverse set of global agricultural soil samples. Based on 985 selected metagenomic samples from diverse agricultural systems, we identified 1745 subtypes from 30 major ARG families, revealing patterns between major agricultural settings. Approximately 1% of global agricultural areas were classified as high-risk, primarily concentrated in regions with intensive farming practices and high antibiotic usage. Our framework enables the identification of risk hotspots which seem to be driven by socioeconomic, climatic, and land use factors. These findings facilitate a methodological advancement for predicting ARG risk through mechanism-driven models, rather than descriptive abundance metrics. The proposed framework will support targeted soil management strategies to mitigate antibiotic resistance propagation in agroecosystems.","manuscriptTitle":"Global health risk assessment of antibiotic resistance in agricultural soils","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-13 05:08:53","doi":"10.21203/rs.3.rs-7635655/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"fe23426c-f035-45ad-8895-95916197c0dd","owner":[],"postedDate":"October 13th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":56097541,"name":"Earth and environmental sciences/Biogeochemistry"},{"id":56097542,"name":"Earth and environmental sciences/Environmental sciences"}],"tags":[],"updatedAt":"2025-12-08T18:11:09+00:00","versionOfRecord":[],"versionCreatedAt":"2025-10-13 05:08:53","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7635655","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7635655","identity":"rs-7635655","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}