Organic fertilization co-selects genetically linked antibiotic and metal(loid) resistance genes in global soil microbiome

Antibiotic resistance genes (ARGs) and metal(loid) resistance genes (MRGs) pose a significant threat to human and animal health. Agricultural soils, particularly those receiving organic fertilizers, are hotspots for ARG dissemination due to the selective pressures exerted by antibiotics and metal(loid)s introduced through manure application. While correlations between ARG and MRG abundances have been observed in organically fertilized soils, evidence for direct genetic linkage remains scarce. This study addresses this gap by analyzing a large dataset of global agricultural soil metagenomes to determine if organic fertilization co-selects genetically linked ARGs and MRGs, and investigates the geographical distribution of these co-resistant genes. The increasing use of antibiotics in livestock and the subsequent presence of these compounds in manure applied to fields creates a selective pressure favoring the survival and proliferation of bacteria carrying ARGs. Similarly, the use of metal(loid)s in livestock feed for growth promotion introduces these elements into the soil, selecting for bacteria possessing MRGs. The potential for co-selection arises from shared mechanisms of resistance, genetic linkage on mobile genetic elements (MGEs), and the potential for horizontal gene transfer (HGT). Previous studies have shown correlations between ARG and MRG abundances, but this study uses a more comprehensive approach to investigate genetic linkage within complex soil microbial communities.

Prior research has established a link between organic fertilization and increased abundance of both ARGs and MRGs in agricultural soils. Studies have demonstrated the presence of various antibiotics (e.g., tetracyclines, sulfonamides, quinolones, macrolides) and metal(loid)s (e.g., copper, zinc, arsenic, chromium) in livestock manure, contributing to the selective pressure in fertilized soils. The long-term application of manure has been shown to increase metal(loid) concentrations in soil. Previous work has also indicated a co-occurrence of ARGs and MRGs in mining-impacted environments with high metal(loid) levels, suggesting a co-selective pressure. While some studies based on bacterial isolate genomes have shown genetic linkages between ARGs and MRGs, the extent of this linkage in complex soil microbial communities remains unclear. This study leverages the increasing availability of global agricultural soil metagenomic datasets to comprehensively assess ARG-MRG associations and their potential risks.

The study utilized 511 agricultural soil metagenomes collected from 17 countries worldwide. Samples were categorized as organically fertilized (OF) or non-organically fertilized (NOF) based on available information or a trained random forest classification model (F1 score 0.97). Metagenomic data were processed to obtain contigs, ORFs, and metagenome-assembled genomes (MAGs). ARGs, MRGs, mobile genetic elements (MGEs), and virulence factor genes (VFGs) were identified using BLAST searches against relevant databases. ARG-MRG-carrying contigs (AMCCs) indicating genetic linkage were analyzed. Twelve agricultural soil samples were further analyzed using metagenomic and metatranscriptomic sequencing to assess ARG-MRG co-expression under different arsenic stress levels. A partial least squares-structural equation model (PLS-SEM) was used to identify factors influencing ARG-MRG coexistence, and machine learning algorithms were applied to predict the global distribution and risk of antibiotic and metal(loid) co-resistant bacteria (AMRB). Various statistical analyses, including two-sided Wilcoxon tests, Spearman correlations, and linear regressions, were employed to analyze the data.

The study revealed significantly higher abundances and diversity of ARGs, risk ARGs, and MRGs in organically fertilized soils compared to non-organically fertilized soils. Organically fertilized soils showed significantly higher abundances of AMCCs (0.44 vs. 0.11 copies per cell), a higher fraction of AMCCs carrying MGEs (4.59% vs. 0.86%), and a threefold increase in the number of AMCC types (63 vs. 22). Metatranscriptomic analysis demonstrated higher abundance and transcriptional activity of specific AMCCs under higher arsenic stress, confirming co-selection and co-regulation of ARGs and MRGs. The PLS-SEM model indicated that fertilizer type, soil properties, climate, and socioeconomic factors all influenced ARG-MRG coexistence, with fertilizer type indirectly affecting coexistence through soil properties and ARG/MRG abundances. Approximately 15% of high-quality MAGs carried AMCCs, with many representing potentially novel species and displaying antibiotic and metal(loid) co-resistance. Machine learning predictions generated a global map indicating higher AMRB risk in Central North America, Eastern Europe, Western Asia, and Northeast China.

The findings confirm the hypothesis that organic fertilization co-selects genetically linked ARGs and MRGs in the global soil microbiome. The increased abundance of AMCCs in organically fertilized soils, coupled with the metatranscriptomic data showing co-expression under stress, strongly supports this. The significant contribution of fertilizer type, soil properties, climate, and socioeconomic factors to ARG-MRG coexistence highlights the complex interplay of factors driving the spread of co-resistant genes. The identification of numerous potentially novel AMRB species underscores the need for further research into the ecology and evolution of these co-resistant organisms. The global risk map provides crucial insights into areas with heightened potential for the dissemination of these genes, facilitating targeted interventions.

This study provides compelling evidence for the co-selection of genetically linked ARGs and MRGs in agricultural soils through organic fertilization. The increased abundance and transcriptional activity of AMCCs under metal(loid) stress highlight the co-selection process. The global risk map identifies high-risk regions, guiding future research and mitigation strategies. Further research should focus on characterizing the novel AMRB identified, investigating the specific mechanisms of co-resistance, and developing effective strategies to reduce the spread of co-resistant genes in agricultural settings.

The study relied on publicly available metagenomic datasets, potentially introducing biases due to variations in sampling methods, sequencing depth, and data processing across different studies. The random forest model for classifying fertilizer types, while exhibiting high accuracy, might still contain some misclassifications. The global risk map is based on predictive modeling and may not perfectly reflect the actual risk in all locations. Future studies could benefit from standardized sampling protocols and more detailed characterization of soil properties.