Our study provides the first atlas showing the global distribution of environmental ARGs in soils, through conducting a global field survey of 1012 sites across 35 countries from all continents. We pinpointed the major global hotspots of soil ARGs and identified global drivers for diversity and abundance of the dominant soil ARGs. This study represents an important advancement in our understanding of the ecology, biogeography, and potential changes in soilborne ARGs in a changing world, which is integral to increase our capacity to address future health crises driven by antibiotic resistant infections.
Our study shows that, in general, soils support relatively low ARG proportions, with only a few soils supporting relatively high proportions of topsoil ARGs (Fig. 2A). On the contrary, according to our histogram, intermediate levels of ARG richness are common in soils worldwide (Fig. 2A). Multidrug resistance genes, and efflux pump machineries, are the most dominant ARG types and antibiotic resistance mechanisms of action, respectively, found in soils across the globe (Figs. 3A and 4). These ARGs are especially effective against a broad spectrum of antibiotics because they allow bacteria to pump antibiotic peptides out of their cells [31]. Efflux pumps, like oprJ and oprD revealed in this study, are evolutionarily ancient and a very common antibiotic resistance mechanism in pristine ecosystems [31, 32]. Multidrug efflux pumps are commonly intrinsically encoded by chromosome and exhibit different functions with physiological and ecological significances that go beyond their activity as antibiotic resistance elements. In fact, multidrug efflux pumps have a wide range of substrate, and their original function was not, in most cases, to resist to antibiotics [32]. Their physiological roles also involve regulating intracellular pH, transporting quorum sensing molecules, and enhancing bacterial pathogenicity. Our results are in agreement with previous local and experimental work highlighting the dominance of this type of ARGs in soils [31, 33] and support the many studies that have demonstrated the ubiquity of ARGs in terrestrial ecosystems with contrasting level of anthropogenic disturbance, from pristine to croplands [34,35,36].
We further show that only 14 ARGs can be considered dominant and ubiquitous in soils worldwide, including the beta-lactamase gene fox5 and the multidrug resistance genes oprJ, oprD, and acrA-05. These dominant ARGs were present in all biomes (Supplementary Table 2). The only exception was oleC, which was found in all biomes except in cold shrublands. The potential public health risks associated with these ubiquitous ARGs should be interpreted with caution, as these ubiquitous soil ARGs are commonly involved in basic processes in bacterial physiology and should be regarded as a potential risk only if they are captured by transferable genetic elements [37]. However, when these genes are subject to a high antibiotic load (e.g., in farms or in natural ecosystems where antibiotic concentrations are locally high; 44), they are more likely to become relevant for resistance development. Indeed, the spread of fecal matter across the globe via animal waste, sewage effluents, and birds transporting microbes from urban habitats has contributed to the ARG dissemination in farmland soils and estuaries [38,39,40]. Even so, it should be noted that there are extremely stringent bottlenecks for the transfer of ARGs from soil bacterial hosts to human pathogens [41], especially in natural habitats that are rarely colonized by human pathogens. The abundant and ubiquitous ARGs identified in our study are clear targets to further investigate the potential global contribution of soils in increasing the resistance of microbial pathogens to several antibiotics. Taken together, these results provide global insights into the most common types of ARGs found in soils globally.
We then used structural equation modeling (SEM; Supplementary Fig. 2) to generate a system-level understanding of the most important ecological factors controlling the distribution of topsoil ARGs across the globe (Fig. 5; Supplementary Fig. 2 and Supplementary Tables 3–5). Our results suggest that the proportion of soil MGEs (see “Materials and methods”) is, by far, the most important factor associated (positively) with the proportion of soil ARGs (Fig. 5; bootstrap P = 0.001; Supplementary Tables 3–4). Moreover, our SEM shows for the first time that the direct relationship between the proportion of soilborne MGEs and that of ARGs is far more important than the effects of other essential environmental factors such as location, climate, vegetation, and soil properties (Fig. 5A; Supplementary Tables 3–4). For example, the direct relationship between the proportion of soil MGEs and that of ARGs is three times more important than that associated with mean annual temperature and between 18 and 230 times more important than that linked to soil C:N ratio and soil C, respectively (Fig. 5A; Supplementary Tables 3–4). Similarly, the richness of soil MGEs was also the most important factor associated with the richness of ARGs (Fig. 5B), which is consistent with previous findings from a regional-scale study of Chinese forest ecosystems [12]. Here, we further show some novel indirect positive associations between ARG richness with both soil pH and plant cover via increasing richness of MGEs (Supplementary Table 5). Additional analyses showed that the proportion of soil MGEs had the strongest correlation with the relative abundance of multiple ARG types and mechanisms of action and with the abundance of the most dominant individual ARGs (Fig. 6). MGEs (including plasmids, integrons, and transposons) are common in soils [12], can transfer genetic information from one species or replicon to another, and allow ARGs to efficiently disperse across different organisms [14]. They can also potentially facilitate the transfer of important ARGs from soil microorganisms to clinically important human pathogens [42]. The importance of MGEs was maintained even when removing from our analyses the subset of locations belonging to croplands (Supplementary Fig. 3). Our findings are therefore important because they provide evidence of the potential capacity of soils to contribute to the rapid spread of genes associated with the resistance to medically relevant antibiotics via horizontal gene transfer mediated by MGEs. This knowledge further contributes to better understanding the rapidly increasing amount of information on soil ARGs (e.g., via RefSoil+) [13]. It is also important to note that the strong correlations of ARGs and MGEs implicate the genetic potential of ARGs transfer, but the frequency of horizontal gene transfer in soil is generally low, as revealed by the evidence that the composition of soil resistome is correlated with the taxonomic composition of bacteria [37].
Tundra ecosystems (nine locations in Antarctica, Chile, and Iceland), boreal forests (Northern Hemisphere high latitude forests), other cold forests, and shrublands had the greatest proportion of soil ARGs globally (Fig. 2B; see Supplementary Table 2 to a biome classification). Soils in these ecosystems have, on average, between two- and nine times higher proportions of ARGs than soils from other ecosystems. Remarkably, Antarctica, which was represented by only three sites near each other, included the soils with the highest proportions of ARGs (Fig. 2C). These results agree with those from local studies observing an accumulation of ARGs in Arctic ecosystems [19, 43]. Our SEM provided further evidence of direct and significant negative associations between mean annual temperature and temperature seasonality with the proportion of soil ARGs. We also found a direct and significant positive relationship between the proportion of soil ARGs and precipitation seasonality, an environmental condition shared by cold deserts such as those from Antarctica and many temperate shrublands [44, 45]. These relationships were still found when we focused on samples from natural ecosystems and removed those from croplands (Supplementary Fig. 3). Together, our results show that a greater proportion of soil ARGs is found in extreme environments and highlight potential co-evolutive mechanisms aiming to provide resistance genes and adapt to harsh environments. Microbial antibiotics and extreme cold temperatures are known to cause similar types of damage in cellular components [46]. Consequently, soil microbes might use ARGs to withstand both types of stressors, and this may explain the patterns of ARG proportion observed in cold ecosystems [11, 46]. Interestingly, although we also found a positive correlation between soil C:N ratio and the proportion of soil ARGs (Fig. 6), as previously reported by studies based on shotgun sequencing [11], this positive association vanished when we considered environmental factors such as location, climate, vegetation, and soil properties in our SEM. The combination of these factors has not been previously considered as predictors of soil ARG abundance and diversity at a global scale. As previously reported [11], we also found a correlation between mean annual precipitation and the proportion of soil ARGs (Fig. 6). However, this association was only evident when croplands were removed from our analyses (Supplementary Fig. 3). Cropland ecosystems are often irrigated, and this could have masked the importance of precipitation for ARGs when all data were analyzed together.
Boreal, cold, temperate, and tropical forests supported the highest richness of ARGs in soils (Fig. 2B). Remarkably, on average, boreal and cold forests supported a 64% higher ARG richness than all the other ecosystems (Fig. 2B). Our analyses further demonstrate the contribution of forest biomes to the diversity of topsoil ARGs globally (Figs. 5B and D and 6). Croplands did not show significantly different levels of ARG richness compared with most other biomes. However, we would like to stress that most croplands in the present dataset are from Asia, and that future studies would need to better address the global impact of agriculture on soil ARGs. Vegetation structure was also an important predictor of the community composition of individual ARGs (Fig. 6). Our findings indicate, therefore, that any land use promoting reforestation or deforestation [47] may have important consequences for the global management of soil ARGs. Our SEM analyses further identified a direct and significant positive association between soil ARG richness and temperature seasonality (Fig. 5B and D). These results were consistent when removing the subset of locations belonging to croplands (Supplementary Fig. 3). However, after removing croplands, we found that many vegetation impacts on ARG richness were driven via changes in soil pH (the major driver of soil bacterial diversity [11]), and that precipitation seasonality also positively influenced the richness of ARGs in exclusively non-cropland ecosystems. Finally, we found a positive and significant correlation between the fungal-to-bacterial ratio and the richness and proportion of soil ARGs at those sites where this comparison was possible (r > 0.299; P < 0.005; n = 87). Many fungal species are known to produce antibiotics [19], supporting the positive association with the richness of ARGs across soils reported here and elsewhere [11]. We also considered the possibility that human influence, as measured with the Global Human Influence Index [48] (see “Materials and methods”), could influence our results. However, we could not find any significant correlation between ARG richness and this index of human influence (P > 0.05; n = 1012).
We generated the first atlas for the current distribution of the richness and proportions of soil ARGs, a necessary step to identify topsoil ARG hotspots and to predict sources of potential resistance to antibiotics associated with soil ARG reservoirs. Our atlas suggests that soils from cold and boreal forests in North America and Asia support intermediate to high proportions of ARGs (Fig. 7A), with similar proportions being found in highly seasonal arid regions across the globe (Fig. 7A; Supplementary Fig. 4A). Our findings further suggest that soils from high latitudinal regions of North America and Asia, as well as tropical and subtropical regions in South America, Africa, and Asia, are the most important hotspots for the richness of soil ARGs (Fig. 7B). Many of these locations correspond to forest environments and regions with high-temperature seasonality (Fig. 7B; Supplementary Fig. 4B). Soils from Asia had some of the highest ARG richness, matching regions with the highest current and forecasted human casualties associated with antibiotic resistance, and with some of the largest rates of antibiotic applications for animal production on Earth [39]. Moreover, soils from highly populated areas in Australia showed relatively low proportion and richness of soil ARGs, matching areas with the lowest human casualties associated with antibiotic resistance [49]. These areas are also some of the most (e.g., China) and least (e.g., Australia) populated regions on Earth. Of course, soil ARGs are not necessarily directly implicated in these casualties. However, they potentially support a reservoir of multiple ARG types and defense mechanisms that can be acquired by human pathogens, increasing their virulence or incidence in areas already severely affected by antibiotic resistance.
Our study opens the door to better understanding the global distribution of soilborne ARGs in terrestrial ecosystems. In this respect, we targeted genes associated directly or indirectly with soil ARGs. Even so, we would like to highlight that some of the selected genes are essential to microbial physiology and not always related to drug resistance, especially when considered alone and not accounting for the presence of multiple genes simultaneously. For instance, tetR is a gene regulator, which may be out of context without another tetracycline resistance mechanism such as tetA or an efflux pump. Similarly, vanY and vanT are equally nonspecific and out of context when not considering other van genes from the same cluster simultaneously. This said, we aimed to provide a holistic view of soil ARGs by simultaneously considering 285 genes directly or indirectly associated with ARGs offering a valuable baseline for future work about the soil resistome.