Effects of copper and cadmium stress on resistance genes in farmland soil

Abstract

Background

Current farmland soils are increasingly affected by contamination from antibiotic resistance genes (ARGs) and heavy metals, with the evolution of microbial antibiotic resistance potentially influenced by metal stress. The distribution effects of ARGs in farmland soils under cadmium (Cd) and copper (Cu) combined pollution remain unclear.

Results

Cu and Cd exerted significant main effects on bacitracin, fosfomycin, sulfonamide, trimethoprim–β-lactam, polymyxin, tetracycline, trimethoprim, and vancomycin resistance genes. Interactions between Cd/Cu and fosmidomycin, Macrolide-Lincosamide Streptogramin (MLS), multidrug resistance, sulfonamide, and vancomycin abundance are significant. Co-occurrence network analysis identifies Pseudomonadota, Actinobacteriota, and Acidobacteriota as primary potential host groups for ARGs. Structural equation modeling analysis further demonstrates that Cd exhibits more pronounced effects on ARGs abundance compared to Cu. In contrast to some previous studies, bacterial alpha-diversity and mobile genetic elements (MGEs) did not act as primary drivers of ARGs variation, whereas metal resistance genes (MRGs) showed a strong positive association with ARGs abundance.

Conclusion

In our study, MGEs did not act as primary drivers of ARGs variation but were positively associated with ARGs abundance through their linkage with MRGs, suggesting that metal exposure may increase the mobility potential of resistance determinants. These findings hold substantial implications for assessing potential ecological risks in soil metal pollution and provide new insights into the co-selection mechanisms between Cd and Cu in shaping ARGs dynamics.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12866-026-05267-2.

Introduction

In recent years, the overuse of antibiotics has primarily induced the development of antibiotic resistance genes (ARGs) in animals, accelerating their spread among bacteria in the environment. Currently, ARGs have emerged as a new category of environmental pollutants. Their transmission and proliferation across various media may pose greater ecological risks than antibiotics themselves. Research indicates that ARGs are widely distributed in multiple environments including soil, water bodies, and air [1–3]. Ultimately, antibiotics and ARGs from diverse sources accumulate in farmland soils, making agricultural land a major environmental sink. Meanwhile, heavy metals also exist in soil. Due to their toxic and non-degradable characteristics, heavy metals are considered as one of the main soil pollutants [4]. Research evidence suggests that heavy metals can exert selection pressure not only on heavy metal resistance genes (MRGs), but also on ARGs [5, 6]. Heavy metals are often difficult to degrade, and a variety of metals exist in complex forms in the soil and persist at high concentrations [7]. Therefore, the spread of antibiotic resistance genes in agricultural soils contaminated with heavy metals is a major concern.

Heavy metal-induced soil pollution usually involves multiple metals. Heavy metals coexist and interact in the environment, forming complex mixed contamination scenarios [8]. Among them, Cu and Cd are common heavy metal pollutants that enter farmland ecosystems through mining, smelting, sewage irrigation, and fertilizer application, posing serious threats to soil health and crop safety [9]. Previous surveys have reported Cu concentrations in agricultural soils typically ranging from 20 to 50 mg kg⁻¹, with substantially higher levels occurring in areas affected by long-term fertilization or mining activities. Cd concentrations in agricultural soils are generally below 0.5 mg kg⁻¹ under background conditions but may exceed 1–5 mg kg⁻¹ in contaminated regions. Regulatory thresholds for these metals have been established by national and international agencies, such as the Chinese Soil Environmental Quality Standard (GB 15618–2018) and comparable international guidelines, to define risk screening and intervention values for agricultural soils [10, 11]. In addition, heavy metal pollution not only directly damages soil ecological functions, but also may induce microorganisms and plants to produce MRGs, which can exacerbate the environmental transmission risk of resistance genes through the synergistic selection mechanism with ARGs [12]. For example, studies have shown that Cu and Cd stress can significantly increase the abundance of multidrug-resistant genes in soil [13]. Microorganisms like Pseudomonadota, which may carry multiple resistance genes simultaneously, serve as important vectors for horizontal transfer of resistance genes [14, 15]. Previous research has demonstrated that heavy metals can significantly promote ARGs transfer in pure bacterial strains [16, 17], Wright et al. suggested that the increase in ARGs abundance may be related to the release of heavy metals from waste [18]. Research indicates that multiple mechanisms are involved in the co-selection of antibiotic and heavy metal resistance [19]. One such mechanism is co-resistance, where genes with specific resistance phenotypes coexist within common genetic elements like transposons, plasmids, and integrases [20]. Di Cesare discovered that some ARGs, MRGs, and integrases in wastewater treatment plants show high correlation, demonstrating co-selection between MRGs and ARGs [21].

This study investigates the effects of high and low concentrations of Cd and Cu as independent and combined stressors in agricultural soils. To represent prevalent antibiotic and metal resistance groups in soil, microbial community structure was characterized using 16 S rRNA gene amplicon sequencing, while ARGs, MRGs, and MGEs were profiled using shotgun metagenomic sequencing. The research aims to achieve three objectives: (i) compare the effects of single and combined Cd and Cu contamination on the abundance and composition of ARGs in agricultural soils; (ii) assess the responses of microbial community structure and diversity to different metal stress scenarios; and (iii) evaluate the relative contributions of metal concentrations, soil properties, microbial communities, and MRGs to the observed patterns of ARGs distribution.

Materials and methods

Collection and processing of soil samples

For the soil cultivation experiment, soil samples were collected from Xingshi Town, Longli County, in the northwest of Qiannan Buyi and Miao Autonomous Prefecture, central Guizhou Province (longitude: 106°48′36″; latitude: 26°30′11″), According to the Chinese Soil Taxonomy, the studied soil was classified as Loess soil, which corresponds to inceptisols in the USDA Soil Taxonomy and Cambisols in the World Reference Base for Soil Resources (WRB, FAO). After air-drying and sieving through a 2-millimeter nylon sieve, 5 kg of soil was placed into each circular flowerpot measuring 133 cm² × 15 cm. The soil was homogenized before undergoing metal contamination and immersion experiments. Cd and Cu were added in the form of Cd(NO3)2·4H2O and Cu(NO3)2·3H2O (analytical purity ≥ 99.0%), with treatment concentrations as follows (Values indicate Cu-Cd concentrations, mg kg⁻¹ dry soil): no heavy metals (CK); single Cu (Cu100; Cu200) labeled as T1, T2; single Cd (Cd 2.4; Cd 4.8) labeled as T3, T4; Cd-Cu mixture (Cd 2.4 + Cu 100; Cd 2.4 + Cu 200) labeled as T5, T6; Cd 4.8 + Cu 100; Cd 4.8 + Cu200) labeled as T7, T8. A total of 9 treatment groups were established, with exposure levels approximating those found in agricultural soils contaminated by certain heavy metals [22].Each group was treated in triplicate and passivated for two weeks [23], Soils were watered every two days during the passivation period to maintain moisture.

Soil samples for heavy metal determination and DNA extraction were collected only after the two-week equilibration period following Cd and Cu amendment. Baseline samples prior to metal amendment were not collected, as the study focused on comparative responses among controlled metal treatments under identical soil background conditions.

Determination of soil physical and chemical properties and heavy metal concentrations

1. Determination of physical and chemical properties: the SOM content was determined via the potassium dichromate volumetric method [24]. Soil pH was measured using a glass electrode by the potentiometric method in a soil–water suspension (1:2.5, w/v). Cu and Cd determination: To measure total Cd and Cu concentrations in soil, soil samples were air-dried at room temperature, ground, and sieved through a 0.149 mm nylon mesh. The digestion was performed using a pressure digestion vessel (HNO3/HClO4/HF = 3/3/2) with the digestion solution analyzed by ICP-MS [25].

2. 16 S rRNA gene sequencing: Using primers 799 F and 1193R to amplify the 16 S rRNA genes (V5–V7) from soil bacteria and plant endophytes. The amplification products were sequenced using dual-end 2 × 250 bp sequencing on the Illumina NovaSeq 6000 platform at Shanghai Ling’en Biotechnology Co., Ltd. (Shanghai, China) using a SP kit with 500 cycles. PCR amplification was performed under the following conditions: predenaturing at 95 °C for 3 min, followed by denaturation at 95 °C for 30 s, annealing at 55 °C for 30 s, and extension at 72 °C for 45 s. Finally, the procedure was stably extended at 72 °C for 10 min and maintained at 10 °C until the end of the reaction. DNA purity and integrity were examined. The PCR products were detected via 2% agarose gel electrophoresis, and the PCR products were quantified via the QuantiFluor™-ST Blue Fluorescence Quantification System (Promega, USA).

3. Analysis of the microbial community in soil: Quality control of the raw sequenced sequences was performed via fastp software (https://github.com/OpenGene/fastp, version 0.20.0) Splicing was performed with FLASH software (http://www.cbcb.umd.edu/software/flash, version 1.2.7) [26], the optimized data was obtained after quality control and splicing. The DADA2 plug-in in the QIIME2 workflow was used to denoise the optimized sequences after quality control and splicing [27]. The sequences obtained after DADA2 denoising are often referred to as amplicon sequence variants (ASVs). To minimize the impact of sequencing depth on subsequent alpha diversity and beta diversity data analysis, the number of sequences in all the samples was adjusted to 20,000. After leveling, the average sequence coverage (Good’s coverage) of each sample was still 99.09%. On the basis of the Sliva 16 S rRNA gene database (v 138), species taxonomic analysis of ASVs was performed via the Naive Bayes classifier in QIIME 2. Sequencing was performed at Shanghai Ling’en Biotechnology Co., Ltd., and microbial community analysis was performed on the corresponding cloud platform (http://www.cloud.biomicroclass.com/CloudPlatform/home).

4. Metagenomic analysis: Microbial DNA was extracted from soil samples using the E.Z.N.A.^® soil DNA Kit (Omega Bio-tek, Norcross, GA, U.S.) according to manufacturer’s protocols. Metagenomic shotgun sequencing libraries were constructed and sequenced at Shanghai Biozeron Biological Technology Co. Ltd. Briefly, for each sample, 1 µg of genomic DNA was sheared by Covaris S220 Focused-ultrasonicator (Woburn, MA USA) and sequencing libraries were prepared with a fragment length of approximately 450 bp. All samples were sequenced using next generation sequencing platform (MGI-T7 or Illumina) (Shanghai BIOZERON Biotech. Co., Ltd) with PE150 mode according to the standard protocols. Raw sequence reads underwent quality trimming using Trimmomatic(http://www.usadellab.org/cms/uploads/supplementary/Trimmomatic) to remove adaptor contaminants and low quality reads. Reads through quality control were then mapping against humangenome (version: hg19) by BWA mem algorithm (parameters: -M -k 32 -t 16, http://bio-bwa.sourceforge.net/bwa.shtml). The reads removing host-genome contaminations and low-quality data were called as clean reads and used for the further analysis.

5. Annotation and quantification of resistance genes: The ARGS-OAP v2.0 [28]and SARG (the structured ARG reference database) database [29] were used to annotate and quantify ARGs using clean reads. The BacMet (Biocide and Metal Resistance Gene Database) database, MRGs [30] and Mobile Genetic Elements database [31] were also identified and quantified. The criteria for gene annotation included a similarity threshold of more than 90% and matching length of more than 25 amino acids between the reads and reference sequence. The resistance gene annotation data were used to quantify the gene abundance (ARG copy per 16 S rRNA, cell, ppm) in each sample, categorised by type and subtype. In addition, we also calculated the average abundance of each gene within the group.

Data analysis

Data processing was performed using Microsoft Excel 2010. Alpha diversity indices, including Observed Species, Chao1, ACE, Shannon, Simpson, Pielou’s evenness (Pielou-J), and Faith’s phylogenetic diversity (Pd-Faith), were calculated in R using the vegan package. Observed Species, Chao1, and ACE were used as richness estimators, whereas Shannon and Simpson indices quantified community diversity. Community evenness was assessed using Pielou-J. Python-based two-factor nonparametric tests were applied to evaluate changes in ARGs and MRGs abundances under different heavy metal treatments [32]. PCoA based on Bray–Curtis dissimilarity was used to visualize compositional differences, and PERMANOVA was applied to test their statistical significance. PERMANOVA based on Bray–Curtis dissimilarities with 999 permutations. Structural equation modeling (SEM) and random forest analyses were conducted using the plspm package to evaluate interactions among multiple influencing factors [33]. Microbial co-occurrence networks were constructed using the R package WGCNA through the CNSknowall platform (https://cnsknowall.com), a comprehensive web service for biomedical data analysis and visualization.

Results and discussion

Effects of copper and cadmium treatment on microbial communities

Alpha diversity is used to represent microbial community diversity, and through single-sample diversity analysis (Alpha diversity), the abundance and diversity of microbial communities within a sample can be reflected [34]. As shown in Fig. 1, similar to earlier studies, soil bacterial species abundance was adversely affected by heavy metal pollution [35]. Specifically, Observed Species, Chao1, and ACE were used to describe community richness (i.e., the number of taxa), which is particularly sensitive to environmental filtering by heavy metals. In our results, these richness indices showed clearer and more consistent responses to Cu/Cd and Cd–Cu combined stress than diversity or evenness indices. In this study, soil bacterial richness showed a gradual upward trend in the Observed Species index as Cu concentration increased. This may be attributed to the dual role of Cu as both an essential micronutrient and a toxicant at elevated concentrations. As Cu is an essential nutrient at low concentrations and a crucial component of many enzymes, it may promote enzyme synthesis and increase Observed Species at moderate to low concentrations [7]. As Cd concentration increased, the Observed Species index (OSI) showed a significant decline. The combined Cd-Cu treatment significantly reduced the soil bacterial OSI. Notably, the decrease in bacterial alpha-diversity caused by the Cd-Cu treatment was substantially greater than the combined effect of single treatments with Cd and Cu (Fig. 1A) (p < 0.05). This suggests a non-additive response of bacterial richness under combined metal exposure [7]. In contrast, Shannon diversity, which integrates both richness and evenness, exhibited no significant variation across most treatments (Fig. 1D). This suggests that dominant bacterial taxa remained relatively stable under metal stress, while reductions were mainly associated with rare or metal-sensitive lineages. Consistent with this interpretation, Pielou’s evenness indices showed limited changes among treatments (Fig. 1E). Pielou’s evenness (Pielou-J) was used to evaluate whether changes in diversity were driven by shifts in dominance structure. The absence of significant changes in evenness suggests that heavy metal stress did not strongly alter species dominance patterns, reinforcing the conclusion drawn from Shannon indices [36]. This indicates that heavy metal exposure did not substantially alter dominance structure within the bacterial community [37]. One possible explanation is that when Cu pollution occurs, the development of microorganisms is hindered and the diversity of microorganisms is reduced, while the high level of Cu may enhance the resistance of various bacteria to Cu [38]. Pollution-induced community tolerance (PICT) is a phenomenon characterized by changes in microbial responses to pollution, resulting in increased tolerance to contaminants. Faith’s phylogenetic diversity (Pd-Faith) also declined significantly across treatments (p < 0.001). This indicates a pronounced loss of phylogenetic breadth (Fig. 1F). The inclusion of Pd-Faith allowed us to assess whether heavy metal stress disproportionately eliminated phylogenetically distinct lineages, providing additional ecological context to the observed richness loss [39]. Overall, these results demonstrate that Cd and Cd–Cu stress reduced bacterial richness without inducing major shifts in community evenness or dominance patterns. Therefore, richness loss rather than community reorganization appears to be the primary microbial response to heavy metal contamination in this system.

Fig. 1.

Alpha diversity indices of the bacterial community of agricultural soil subjected to pollution by Cu and Cd. A Observed Species; B ACE; C Chao 1; D Shannon; E Pielou_J; F Pd-Faith

Microbial communities require complex defense mechanisms to withstand various environmental pressures [39]. In order to further understand the mechanism of microbial adaptation to the outside world, we conducted VPA analysis and used PCoA algorithm to study the similarity or difference of microbial community composition in samples, and used PERMANOVA to test and identify. Then we selected the top30 species for cluster analysis, and show the percentage of the top ten categories. As shown in Fig. 2A, the sample exhibited the highest percentage of Pseudomonadota at the phylum level, followed by Acidobacteriota and Actinobacteriota. Notably, over 70% of the abundance was composed of Pseudomonadota and Acidobacteriota. This finding aligns with previous reports indicating that bacterial communities in metal-contaminated areas are predominantly Pseudomonadota-based, VPA was used to determine the contributions of Cd, Cu, Ph, and SOM to microbial community differences (Fig. 2B). Cd and Cu were the primary contributors to microbial changes at concentrations of 0.05 and 0.017, respectively. However, their contribution rates to microbial changes remained relatively low under the heavy metal concentrations tested in this study (soil physicochemical properties are detailed in Supplementary Material Table 1). Therefore, Cu/Cd stress contributes to variation in bacterial community composition rather than directly altering total bacterial abundance. To investigate abundance changes of microorganisms under Cu-Cd stress. Under Cu-Cd treatment, Nitrospirota, Thermodesulfobacteriota, Deinococcola, Chloroflexota, and Cyanobacteria were clustered together, indicating similar abundance patterns across samples. This may be because they are either chemolithoautotrophs or phototrophs (e.g., Cyanobacteria, Chloroflexota), or are involved in nitrogen/sulfur cycles (Nitrospirota, Thermodesulfobacteriota), while Deinococcola is renowned for its stress resistance. These groups likely exhibit similar responses to heavy metal stress. Subsequently, Fusobacteria, SAR324 clade (Marine group B), Myxococcota, Planctomycetota, Bacillota, Abdilibacteriota, and Pseudomonadota were clustered together, suggesting comparable stress response patterns. Notably, Pseudomonadota typically contains many heavy metal-tolerant bacteria, while Bacillota also includes some highly tolerant species. When samples treated with CK were grouped with those treated with T1 and T2, the effects of Cu concentration treatments on microbial communities showed consistency with the control group, and no significant differences between different treatments. This result is similar to previous studies where bacterial community diversity in Cu-contaminated soil was found to be comparable to that in unpolluted soil [40].The overlapping clusters of Cd treatment and Cu-Cd stress showed that the effects of Cd on microbial communities in soil concentration treatment were inconsistent with that in blank, and there were obvious differences. This functional-based sample clustering pattern is similar to the microbial composition, indicating that there is a sharing or horizontal transfer of functional genes between bacterial genera. PERMANOVA indicated that Cd explained the largest proportion of variation in microbial community composition (R² = 0.38). Although Cd explained the largest proportion of variation, this effect did not reach statistical significance, indicating a trend rather than a definitive driver (Fig. 2D).

Fig. 2.

Changes in soil bacterial community composition and structure under Cu and Cd contamination. A Percentage of dominant bacterial phylum; (B) Variation partitioning analysis (VPA) showing the contributions of Cu, Cd, pH, and SOM; (C) Clustering tree heat map; (D) Principal coordinates analysis (PCoA) of bacterial community composition and identified by PERMANOVA

Effect of copper and cadmium treatment on resistance genes

The SARG (Structured ARG Reference Database) database has annotated 293 ARGs across 22 types, with vancomycin, bacitracin, phosphomycin, rifamycin, tetracycline, and multidrug-resistant genes showing higher abundance than others. Significant differences in total resistance gene abundance were observed between treatment groups, with the T3 treatment demonstrating the highest resistance gene abundance. All metal treatments significantly exceeded the control group (Kruskal-Wallis test, p < 0.05) (Fig. 3A). This indicates significant compositional differences in ARGs detected in soil samples. The increased abundance of ARGs may be attributed to elevated metal concentrations, suggesting strong selective pressure that enhances bacterial resistance to metals. PERMANOVA explained the largest proportion of variation of ARGs (R² = 0.62, p < 0.01), while Cu exerted a weaker yet significant main effect. Moreover, a significant Cu × Cd interaction revealed non-additive influences on ARGs assemblage structure (Fig. 3C).

Fig. 3.

Overview of ARGs and MRGs abundance and composition across treatments: (A) ARGs; (B) MRGs; The histogram above the heat map shows the sum of each column of variables, and the lollipop map on the left side of the heat map shows the sum of each row. C and (D) are the overall distribution patterns of ARGs and MRGs differences between samples, respectively. Based on the Bray-Curtis distance, the PERMANOVA results of resistance gene samples under heavy metal stress

The BacMet2 database annotation identified 262 resistance genes associated with 18 metallic and metalloid elements. Resistance genes related to Ag, Al, As, Au, Cd, Co, Cr, Cu, Fe, Hg, Ni, Pb, Se, Te, V, W, and Zn were particularly abundant, accounting for 69.4% of the total MRGs. Notably, resistance genes involving Zn, Fe, Al, As, and Cr showed significant increases (p < 0.05) (Fig. 3B). Furthermore, given the crucial roles of Fe and Al in cellular functions, the abundance levels of MRGs associated with these elements remained exceptionally high [41]. Furthermore, we observed that treatments with Cu and Cd did not significantly alter the related resistance genes, but instead caused significant impacts on resistance genes associated with Zn, Fe, Al, As, Cr, and other elements. This indicates significant differences in the composition characteristics of MRGs found in soil samples (p < 0.05), as microbial resistance to heavy metals is not dependent on a single gene but is achieved through multi-level network coordination involving efflux pumps, metal-binding proteins, antioxidant systems, and other mechanisms [42]. Cu and Cd exposure did not primarily increase the abundance of their corresponding metal resistance genes. Instead, significant increases were observed in resistance genes associated with other metals, particularly Fe and Al. PERMANOVA explained the largest proportion of variation of MRGs (R² = 0.61, p < 0.01), while Cu exerted a weaker yet significant main effect. Moreover, a significant Cu × Cd interaction revealed non-additive influences on MRGs assemblage structure (Fig. 3D).

To further elucidate the changes in ARGs and MRGs under Cu-Cd stress, the Scheierer-Ray-Hare test was employed to determine how different heavy metal combinations affect the relative abundance of ARGs (Table 1) or MRGs (Table 2). Among ARGs, Bacitracin, fosfomycin, sulfonamide, and Trimethoprim showed significant effects from Cu exposure. Beta-lactams, polymyxins, tetracycline, Trimethoprim, and vancomycin were notably impacted by Cd. The interaction between Cu and Cd exhibited a synergistic effect on fosmidomycin, MLS, multidrug resistance, sulfonamide, and vancomycin. This demonstrates that different heavy metals induce distinct patterns of antibiotic resistance gene accumulation. Compared with single-metal exposure, combined Cu–Cd contamination resulted in a significantly greater accumulation of resistance genes, indicating a synergistic effect of multi-metal stress on resistance gene profiles.

Table 1.

Scheirer-Ray-Hare test for ARGs (2-factor nonparametric ANOVA)

ARGs	Cu	Cd	Cu×Cd
Aminoglycoside	0.552	0.947	0.109
Bacitracin	0.006***	0.654	0.452
Betalactam	0.938	0.023*	0.616
Chloramphenicol	0.511	0.894	0.186
Fosfomycin	0.001***	0.898	0.479
Fosmidomycin	0.138	0.496	0.046*
Kasugamycin	0.759	0.272	0.097
MLS	0.821	0.094	0.021*
Multidrug	0.177	0.513	0.026*
Polymyxin	0.365	0.004***	0.948
Puromycin	0.421	0.180	0.509
Quinolone	0.436	0.526	0.092
Rifamycin	0.588	0.271	0.038*
Sulfonamide	0.030*	0.373	0.045*
Tetracenomycin	0.598	0.467	0.594
Tetracycline	0.495	0.039*	0.462
Trimethoprim	0.019*	0.028*	0.908
Unclassified	0.953	0.097	0.238
Vancomycin	0.142	0.047*	0.022*
SUM	4	5	5

Cu main effect of Cu, Cd main effect of Cd, Cd×Cu interaction effect between Cu-Cd, SUM sum of significant differences (p < 0.05)

Note: Asterisks denote significance levels among treatments: * significant, ** highly significant

Table 2.

Scheirer-Ray-Hare test for MRGs genes (2-factor nonparametric ANOVA)

MRGs	Cu	Cd	Cu×Cd
Al	0.266	0.016*	0.163
As	0.953	0.213	0.002**
Au	0.275	0.276	0.029*
Cd	0.247	0.250	0.590
Co	0.791	< 0.001***	0.938
Cr	0.704	0.618	0.015*
Cu	0.750	0.199	0.835
Fe	0.972	0.002**	0.247
Hg	0.250	0.580	0.307
Ni	0.588	0.849	0.118
Pb	0.689	0.011*	0.864
Se	0.046*	0.097	0.686
Te	0.808	0.279	0.187
V	0.964	0.329	0.018*
W	0.380	0.821	0.330
Zn	0.937	0.393	0.003**
Multi_metals	0.195	0.546	0.093
SUM	1	4	5

Cu main effect of Cu, Cd main effect of Cd, Cd×Cu interaction effect between Cu-Cd, SUM sum of significant differences (p < 0.05)

Note: Asterisks denote significance levels among treatments: * significant, ** highly significant

In Table 2, Se, Al, Co, Fe, and Pb were affected by Cu and Cd respectively, while As, Au, Cr, V, and Zn showed significant combined impacts from both metals. This difference fundamentally stems from the interaction between heavy metal geochemical behaviors (such as ionic radius and oxidation states) and biomolecular mechanisms (including transporter specificity and detoxification pathways). Future research should integrate in-situ mineral characterization (e.g., Mossbauer spectroscopy) with multi-omics analysis to further elucidate gene regulatory networks under complex heavy metal contamination. This approach will provide a scientific basis for developing targeted strategies to mitigate resistance gene pollution in such environments.

Effects of microbial resistance genes

Bacteria are one of the driving forces in microbial environment activity and the main carrier of resistance genes in the environment [43], A network analysis was conducted to investigate the correlation between resistance genes and bacterial communities (Phylum-level) under Cu-Cd stress. The results revealed a strong positive correlation between resistance genes and potential bacterial hosts within the microbial community. As shown in Fig. 4A, the network comprises 375 nodes and 1,167 edges. Pseudomonadota, Actinobacteriota, and Acidobacteriota showed the highest co-occurrence frequency with ARGs, serving as primary carriers of ARGs. These findings align with other research outcomes [43, 44]. Most of the bacteria resistant to antibiotics belong to the phylum Pseudomonadota and Actinobacteriota. In the collinear network of bacteria and MRGs in (Fig. 4B), there are 373 nodes and 117 edges, among which Pseudomonadota phylum and Actinobacteriota have the highest co-occurrence frequency with MRGs, which are the main carriers of MRGs. This finding is consistent with the results reported by Li et al. [45]. The results are consistent, They discovered that the Pseudomonadota, Actinobacteriota are the primary metal-resistant bacteria under heavy metal stress. The main carriers of resistance genes are the Pseudomonadota and Actinobacteriota families, as these groups possess the genetic capacity to produce multiple antibiotics and carry various ARGs and MRGs [21], Therefore, in the metagenomic study of soil samples, actinomycetes and Pseudomonadota are one of the most frequently predicted sources of resistance genes.

Fig. 4.

A-B Network analysis of associations between resistance genes and bacterial communities (phylum-level). Connections indicate strong correlations (Pearson correlation coefficient r ≥ 0.6; significant p < 0.01). Red lines represent positive correlations, while gray lines indicate negative correlations

Any statistical inference regarding ARGs/MRGs–host relationships should be interpreted with caution. As false positives cannot be completely eliminated and a small number of connections may be coincidental. To further examine the relationship between Pseudomonadota, Actinobacteriota composition and resistance gene profiles under Cu–Cd stress, principal coordinate analysis (PCoA) was first performed independently on the Pseudomonadota, Actinobacteriota data and on ARGs/MRGs abundance matrices using Bray–Curtis dissimilarity. PCoA served as a dimensionality-reduction and visualization tool, allowing comparison of overall compositional patterns among samples.

Specifically, the ordination of Pseudomonadota, Actinobacteriota showed a strong and significant alignment with ARGs composition (M² = 0.83, p = 0.018), indicating that variations in ARG profiles closely tracked changes in Pseudomonadota, Actinobacteriota community structure. A similarly significant relationship was observed between microbial communities and MRGs, with an M² value of 0.82 (p = 0.025). In both cases, the relatively short Procrustes residual vectors suggested a high degree of structural correspondence between microbial taxa and resistance gene assemblages.

These results demonstrate that both ARGs and MRGs are non-randomly structured in relation to the Pseudomonadota, Actinobacteriota community and support the view that shifts in resistance gene composition are tightly associated with microbial community dynamics under Cu and Cd stress (Fig. 5).

Fig. 5.

Procrustes analysis linking ARGs, MRGs and bacterial assemblages. A ARGs–bacterial community ordination match; B MRGs–bacterial community ordination match

Effects of different factors on ARGs

To determine the primary drivers of ARGs distribution in soil bacterial communities Cd and Cu concentrations, soil characteristics, microbial community diversity, MGEs and MRGs, we conducted structural equation modeling (SEM) analysis (Fig. 6A) and random forest analysis to verify the contributions of different factors to ARGs.

Fig. 6.

A The structural equation model (SEM) demonstrates the effects of Cd, Cu, soil properties, microbial diversity, MGEs, and MRGs on ARGs composition changes in soil microbiota. The SEM-based standardized direct and indirect effects model shows excellent data fit: 4 degrees of freedom, non-significant probability level (p = 0.928), strong goodness-of-fit index (GFI = 0.992), and low root mean square error (RMSEA < 0.001). Positive and negative correlations are indicated by black and orange arrows respectively. R² values represent explained variance. B Schematic diagram of the random forest algorithm for analyzing the impact of different factors on ARGs

Cu exerts direct and significant effects on ARGs, with its direct impact being more pronounced than indirect effects. Cu can influence the abundance of related ARGs through cross-resistance mechanisms. Although Cu itself is an essential trace element, high concentrations exhibit potent antimicrobial properties. Moreover, Cu concentration significantly affects soil properties, which indirectly impacts soil microbial diversity and consequently influences resistance gene abundance through soil characteristics. However, this antimicrobial pressure may indirectly promote the selection, enrichment, and dissemination of ARGs [46]. Moreover, the indirect impact of Cd on ARGs outweighs its direct effect. Soil MRGs may influence the distribution of ARGs in soil. Cd pollution drives the enrichment of mobile elements carrying MRGs through selective pressure, leading to passive accumulation and dissemination of co-localized ARGs. Moreover, heavy metal stress promotes horizontal gene transfer, accelerating the spread of MRGs (along with ARGs), findings consistent with previous studies [47]. Cd concentration has a significant effect on soil microorganisms and MRGs, which can indirectly affect the abundance of ARGs in soil through soil microorganisms and MRGs. Within the SEM framework, Cd showed a stronger association with ARGs abundance in soil than Cu. This may be because Cd is more toxic than Cu, which is a micronutrient and does not have any beneficial function in this regard, but is completely a toxic metal [48], In this study, treatment with different concentrations of Cu and Cd exhibited distinct effects on ARGs. Specifically, both low and high concentrations of Cd increased the abundance of ARGs. Conversely, high concentrations of Cu showed antagonistic effects on microbial diversity when co-exposed with Cd (Fig. 1A), suggesting potential competition for binding sites between high-concentration copper and cadmium. These findings indicate that heavy metal exposure may influence the distribution and abundance of ARGs through both direct and indirect mechanisms. Most studies indicate that microbial communities play a crucial role in promoting ARGs in soil. As microbes serve as the primary carriers of ARGs, they are recognized as one of the main driving forces behind their evolution [49], However, this study did not find microbial community diversity to be the main driver of variation in target ARGs, which is consistent with some findings from a previous study [50], In this study, MRGs in soil bacterial communities significantly influenced ARGs. Previous studies have shown strong correlations between Cu and zinc resistance genes, pcoA, and czcD with multidrug-resistant strains. It has been widely reported that the mer operon in mercury-resistant strains is frequently genetically linked to multiple antibiotic resistance genes, often co-located on mobile genetic elements such as Tn21-family transposons, which facilitate the co-selection and horizontal transfer of metal and antibiotic resistance traits [51], Transposons as potential vehicles for ARGs–MRGs co-selection. Metal stress may promote the dissemination of ARGs not only through host community shifts but also via mobile genetic elements. In particular, transposon-associated functions (e.g., transposases and resolvases) and class 1 integron integrase (intI1) have frequently been reported to co-localize with metal resistance operons (e.g., mer) and ARGs on composite MGEs such as Tn21-family transposons, facilitating co-selection and horizontal transfer [52]. In our dataset, the abundance patterns of ARGs were positively associated with MGE-related features, suggesting that metal exposure may increase the mobility potential of resistance determinants.

Conclusion

This study demonstrates that copper (Cu) and cadmium (Cd) contamination, particularly under combined exposure, significantly reshapes soil bacterial communities and resistance gene profiles in agricultural soils. Combined Cd–Cu treatment caused a stronger reduction in bacterial richness than single-metal treatments, indicating a non-additive response of microbial communities to mixed-metal stress. Multivariate analyses showed that Cd exerted a stronger influence on community compositional variation than Cu, although this effect did not reach statistical significance.

Resistome analyses revealed that ARGs enrichment was more closely associated with MRGs than with changes in microbial community structure alone. Structural equation modeling further supported a stronger association between Cd and ARGs abundance than Cu, primarily through indirect pathways mediated by MRGs, rather than direct effects of microbial diversity.

Overall, these findings highlight the importance of metal-driven co-selection processes in shaping soil resistomes under multi-metal contamination. This study provides insight into the ecological risks posed by combined heavy metal pollution and underscores the need to consider metal mixtures when assessing the dissemination potential of ARGs in agricultural soils. These findings provide a foundation for future studies integrating functional assays and gene transfer analyses to more explicitly evaluate ARGs spread under heavy metal stress.

Authors’ contributions

Ren He: Performed the most of the experiments, Formal analysis, Writing – original draft. Ke Liu: Conceptualization, Funding acquisition, Writing – review & editing, Supervision. Fan yang: Visualization. Shiqing Feng: Methodology. Lili Zhang: Performed the part of the experiments.

Funding

The authors would like to acknowledge the support of the National Natural Science Foundation of China (No 21866008).

Data availability

The sequencing data generated in this study have been deposited in the NCBI Sequence Read Archive (SRA) under accession number [PRJNA1435093], https://www.ncbi.nlm.nih.gov/sra/PRJNA1435093 .

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.