In this study, to examine the soil resistome we profiled 36 metagenomes and 41 high-quality MAGs by referring to the SARG database along with a recommended pipeline . The comprehensive information of ARGs was investigated with both the antibiotic classes and functional genes, and thus compositional characteristics of background ARG in natural soils and potential microbial hosts were revealed. ARG profiles generated from the shotgun data and MAGs were mutually supportive. While the short-read-based approach provided more sensitive detections of ARGs, de novo assembled genome analysis would better reflect the co-occurrence and relative abundance of potentially functional ARGs as well as the identity of their microbial host , although low abundant ARGs might be lost during the assembly process. It is indeed the case in our results that ARGs for nine classes of antibiotics were both frequently detected in MAGs and short reads (Fig. 6 and Addition file 1: Figure S3), meanwhile genes that were low detected in short reads have vanished from MAGs (i.e. genes for kasugamycin, puromycin and carbomycin). However, with regards to the relative abundance of ARGs, results of short reads and MAGs show some contradictions. In MAGs genes like bcrA (confer resistance to bacitracin), macB (extrude macrolide and peptide), and vanSR (regulon for vancomycin resistance) were more abundant than the short-read-popular multidrug resistance genes like mdtBC and mexF. (Addition file 2: Table S11). For environmental samples, such contradictions due to gene assembly seem to occur from time to time, as the mapped sequences of ARGs are only a small fraction of all reads in a sample. In a recent review, Coelho et al. summarized 13174 publicly available metagenomes and found MAGs generally capture a proportion less than 5% of all genes in a soil sample . In addition, the ability of binning method to correctly recover ARGs on mobile genetic elements was systemically problemed by the variable copy number and sequence composition of the gene elements . Therefore, to have a consistent perspective, we mainly discuss the results obtained from short reads in this study.
As we mentioned in the introduction, it is unsurprising that ARGs are ubiquitous in the soil , since many antibiotics are natural secondary metabolites of soil microbes and were produced as chemical arms in the competition for survival . ARGs serve as a genetic blueprint for fortifications and is necessary for the survival of microbes. However, what are the physiological and ecological nature behind the predominance of multidrug resistance genes (or more broadly, efflux pump genes), in the soil resistome is still worth exploring. In this study genes for efflux pump mechanisms show dominancy in soil resistome in terms of either total abundance, average relative abundance or detection frequency (Fig. 2 and Additional file 2: Table S4-S9). The dominancy could reflect their higher fitness in natural soils.
As Larson et al. concluded the concentration of a single antibiotic in the environment is generally insufficient to cause a selective pressure, and even if the antibiotic is accidentally exposed due to human factors, it will degrade within a short period . Therefore, the real situation is supposed to be the coexistence of multiple naturally secreted antimicrobials at low concentrations, where multidrug efflux pumps were reasonably selected by microbes due to their functional versatility . For instance, MdtABC-TolC was the efflux pump with the highest average abundance detected in this study (Fig. 3), it is associated with the efflux of antibiotics such as β-lactams and novobiocin as well as with the exclusion of copper and zinc and the formation of biofilms . Similarly, MexEF-OprN could pump out aromatic hydrocarbons and population-sensing signaling molecules; MacAB-TolC extrudes lipopolysaccharides and peptides as non-antibiotic resistant functions . These functions are closely related to interfacial colonization and nutrient acquisition and are themselves a compensatory mechanism for the cost of resistance expression of microbes in the soil environment. Furthermore, consistent with our observation, previous studies also reported the prevalence of multidrug resistance genes in natural and anthropogenically affected soil [15, 26]. In a study of the Savannah River Site , a high abundance of multidrug resistance genes were found in all 24 acidic (i.e. pH in 3.98–4.38) and heavy metal polluted soils. If in that situation it can still be explained by heavy metal co-selection at contaminated sites, the prevalence of multidrug resistance genes in 26 rainforest and pasture soil samples could only be attributed to the natural selection of soil microbes .
Based on the prevalence of multidrug efflux pump genes, the deterministic shaping role of soil pH in the resistome was highlighted. In terms of the abundance of ARGs, the total abundance was found to be highest when the soil pH was the lowest (Figure S3 in Additional file1) and showed a linearly decreasing trend with the increase of pH (Fig. 2D). Moreover, all the changes in total ARG abundance seem to be caused by the changes in multidrug resistance genes (Fig. 2E and Additional file1: S6, similar slopes in regression). Especially for the genes related to inner membrane permease, i.e. mdtBC, they were hyper dominant in groupings of acidic soils (Fig. 3) but almost decreased to none in groupings with soil pH higher than neutral, reflecting a dependence on the low pH (i.e. strong proton activity). MdtB and MdtC have been proved to be the PMF driven multidrug transporter since both of them had five charged and polar amino acid residues which conserved with AcrB and were essential for the proton transportation . Therefore, there is a biochemical basis for the selection of low soil pH for MdtBC. In addition, based on the calculation of the diversity index of ARGs, we found the evenness of ARGs linearly increased with pH, although the richness of ARGs is almost constant. This indicates a shift from the predominance of a few types of ARGs to the coexistence of multiple types, accompanied by an increase in soil pH.
In general, the pH shaping rule we found is consistent with the observation of Bahram et al. in a recent study where global topsoil samples (7560 subsamples in 189 sites from 12 ecosystems) were investigated and a negative correlation between ARGs abundance and soil pH was found . However, the deterministic role of pH in shaping soil resistome is more evident in our study (see the VPA result in Additional file1: S5), which could due to the differences in pH of the observed soil. In particular, their pH values ranged from 2.5 to 7.5 and were mainly concentrated in acidic conditions, while our pH values are evenly distributed between 4.37 and 9.69 (Additional file1: S2). Bahram et al. demonstrated that higher fungal biomass and stronger fungal-bacterial antagonism in acidic soils would lead to a higher abundance of ARGs, this should be a reliable nature. Fungi are well-known natural producers of many antimicrobials, not just natural β-lactam antibiotics, but also ergot alkaloids, fungal polyketides and kinds of antimicrobial peptides . Therefore, in low pH soils where fungal antimicrobials are present in a large amount, PMF-driven multidrug efflux pumps (such as MdtBC and MexF in our case) would be carried by bacterial hosts to counteract multiple antimicrobials stresses is intuitively correct, due to their versatility and the easier access to protons. When soil pH rises it is another situation, where fungal antimicrobials are reduced and access to protons was relatively limited, thus the abundance of ARGs, especially for multidrug genes, gradually decreased. Then the abundance increasing in non-multidrug resistance genes will be foregrounded, such as the ARGs for vancomycin and tetracycline in this study (Figure S3 in Additional file1: S3).
To further explore the relationship between microbes and ARGs in soil, a strict correlation analysis was performed in this study, where indirect correlations were deduced from the network (Fig. 5). And we found ARGs and correlated bacterial phyla have separated into an acid module and an alkali module. In the alkali module, the unique pH-positive-correlated phylum Actinobacteria was also positively associated with the tetracycline and vancomycin genes, this is consistent with the “producer hypothesis” [13, 14]. That microbes in the Actinobacteria phylum (i.e. Streptomyces spp.) are natural producers of tetracycline and vancomycin, and thus are the origins of tetracycline and vancomycin resistance genes. While no direct correlation between pH and ARGs was indicated the changes in tetracycline and vancomycin resistance genes were primarily driven by the natural selection of the Actinobacteria phylum. A similar conclusion was also given by Qian et al., who demonstrated environmental factors select bacterial populations and hence enrich the ARGs they carry . However, in the acid module of our network, environmental factors directly connected with ARGs and their counts were generally higher than the positive correlations between ARGs and bacterial phyla, thus a stronger linkage was found between ARGs and environmental factors. To further confirm, we used SEM to examine relationships among environmental factors, microbes, and ARGs (Additional file1: S9), consistent with the result of correlation analysis, the causality between ARGs and environmental factors was higher weighted than it between ARGs and bacterial phyla. Combining these results, we assumed that in acidic soils ARGs corresponding to high fitness functions may be directly selected by environmental factors rather than through a particular bacterial population, indeed the selected ARGs might be carried by more than one bacterial population. For instance, mdtBC and mexF genes were carried by more than half MAGs and have no specific taxonomic affiliation (Figure S10.1 in Additional file1: S10).
Until here a panorama of pH-driven changes in soil resistome becomes clear. Unlike the neutral assembly of bacterial phylum composition (Additional file1: S7), soil pH variation has a deterministic effect on shaping resistome. The specific manifestations are 1) pH decrease (until 4.5): the total abundance of ARGs increased to meet the putative stronger fungal-bacterial antagonism, multidrug efflux pumps are preferred due to their versatility and stronger fitness in low pH soils and thus were selected; and 2) pH increase (until 9.5): decreased strength of inter-kingdom competition, thus the lower total abundance of ARGs, lower abundance of multidrug genes, when Actinobacteria enriched the abundance of corresponding ARGs like for vancomycin and tetracycline increased.