Combined pesticide pollution enhances the dissemination of the phage-encoded antibiotic resistome in the soil under nitrogen deposition

Significance

Phage-mediated dissemination of antibiotic resistance genes (ARGs) is a key driver of environmental antibiotic resistance and is increasingly influenced by global change and human activities. However, the mechanisms through which soil phages facilitate ARG spread under multiple stresses remain poorly understood. This study shows that nitrogen deposition intensifies pesticide-induced soil stress, promoting the activation and transfer of phage-encoded ARGs. Strengthened phage–host interactions accelerate ARG mobilization, which in turn enhances bacterial resistance and provides adaptive advantages under environmental pressures. Our findings uncover an overlooked pathway through which global change accelerates ARG mobilization via phage dynamics, offering valuable insights for tackling the global antibiotic resistance crisis amid environmental change.

Abstract

Phage-mediated dissemination of antibiotic resistance genes (ARGs) intensifies health threat in the environment. Increasing amounts of pesticides are entering the soil ecosystem, yet their potential influence on phage-mediated ARG spread, particularly under conditions of global change, remains poorly understood. In this study, we performed a long-term field experiment simulating pesticide contamination under nitrogen deposition and examined the role of soil phages in ARG spread and host adaptation using metagenomic and viromic sequencing. Combined pesticide markedly elevated the abundance of phage-encoded ARGs under nitrogen deposition. By enhancing phage–host interactions and increasing the co-occurrence of auxiliary metabolic genes with ARGs, phages may further facilitate the transfer of ARGs to bacterial hosts, conferring hosts a competitive edge in intensified microbial competition driven by combined pesticide exposure under nitrogen deposition. The phage-driven mechanism was supported by in vitro cultivation experiments, demonstrating that phages harboring ARGs, shaped by long-term combined pesticide exposure under nitrogen deposition, can infect bacterial hosts and confer resistance. Collectively, our findings underscore the pivotal role of phages in ARG mobilization under environmental stressors, reinforcing the importance of accounting for phage activity in ARG risk assessments under global change.

Antibiotic resistance genes (ARGs) have emerged as one of the major global environmental health risk factor (1, 2). Phages in soil environments impacted by anthropogenic activities have been identified as overlooked reservoirs of ARGs (3). Phages play a critical role in soil ecosystems due to their abundance, rich genetic diversity, and intricate interactions with their host (4, 5). They can transfer ARGs to host bacteria through phage-mediated transduction (6). Specifically, transduction relies on phage-mediated DNA transfer (7); the ability of phages to carry bacterial DNA fragments, including ARGs, may be pivotal in ARG dissemination (8). Phage-mediated horizontal gene transfer (HGT) is essential in bacterial evolution and environmental adaptation (7). Lysogenic phages can facilitate adaptive evolution in hosts by expanding the pool of functional genes available through lysogenic conversion (e.g., ARGs) (9). Phages can even spread ARGs heterologously, posing an added challenge to managing antibiotic resistance. This phage-mediated ARG transfer enables bacteria to survive antibiotic exposure, promoting antibiotic-resistant strains and complicating clinical treatments (10). Recently, the finding that phages contribute more to the lysis of antibiotic-resistant bacteria than to ARG transduction has sparked debate over their dual roles in antibiotic resistance (11). Due to the complexity of soil ecosystems, soil phages remain poorly understood, and their involvement in ARG dynamics warrants further investigation.

Pesticides are essential for boosting crop yields and controlling pests in modern agriculture (12, 13). However, their global use is expected to increase with population growth (14), and poor management results in persistent residues that threaten biodiversity (15) and elevate the dissemination of ARGs and environmental risk (16). Evidence suggests that over half of agricultural topsoil in the European Union is contaminated with mixed pesticides (17). Such combinations are more likely to induce environmental stress than single pesticides. For example, increased pesticide diversity can impair soil microbial function (18). Pesticides exert selective pressure on bacterial communities and promote HGT, thereby accelerating the emergence and spread of ARGs (19). While the effects of pesticides on ARG propagation have been studied, most research has focused on bacteria as the primary hosts and vectors of ARGs. However, recent studies have identified phages as a key reservoir, highlighting their crucial role in ARGs dissemination (3). In polluted environments, phages play an increasingly important role in ARG spread through host cell lysis and transduction, releasing host DNA (e.g., plasmids and ARGs) (20, 21). Reports suggest that phages can undergo lysis and promote ARG release under oxidative stress induced by nonantibiotic pollutants (e.g., nanosilver) (20). Nevertheless, the mechanisms by which pollutants influence phage-mediated ARG spread remain unclear, and further research is needed to investigate the impact of pesticide pollution on phage-mediated ARG dissemination, especially from combined pesticide exposure.

Nitrogen deposition has become a key driver of global change, significantly impacting the stability and function of soil ecosystems. While it enhances soil fertility and favors dominant bacteria (18), excessive input can lead to soil acidification, nutrient imbalances, and biodiversity loss (22–24). Moreover, nitrogen addition can increase the abundance of ARGs in soil (25) and alter phage diversity (26). The intensity of soil stress caused by nitrogen deposition rises with its rate and duration (27). Nitrogen deposition also affects the environmental behavior of pollutants entering the soil. Its synergistic interaction with soil Cd and Cu may enhance heavy metal uptake by crops, threatening food safety (28). It may also accelerate dissolved organic carbon production, increasing heavy metal mobility (29). Excessive nitrogen can reduce key soil enzyme activity, acidify soil, and inhibit oligotrophic degraders, thereby impeding pollutant degradation (30, 31). Environmental stress is expected to intensify, as nitrogen deposition often co-occurs with other adverse factors (e.g., climate warming and pesticides). Although atmospheric nitrogen deposition and pesticide pollution may jointly drive environmental change, their combined effects remain largely understudied, especially on soil phages and ARGs.

To examine whether and how nitrogen deposition affects the spread and risk of pesticides on ARGs in soils via viral–host interactions, we employed metagenomic and viromic sequencing alongside microbial culture experiments. Our aim was to assess the combined effects of nitrogen deposition as a global change factor and pesticide pollution on phage–host interactions, community dynamics, and phage-mediated ARG transfer contributing to host resistance. To this end, we conducted a 3-y field experiment investigating the effects of two commonly used pesticides (chlorpyrifos: BW; a mixture of azoxystrobin and propiconazole: AB) and two nitrogen application levels (N control: 0 kg N ha⁻¹ y⁻¹; N addition: 100 kg N ha⁻¹ y⁻¹), simulating pesticide pollution and nitrogen deposition. We hypothesize that nitrogen deposition promotes phage-encoded ARGs in pesticide-contaminated soils and further elevates ARG risk in bacterial hosts through phage-mediated transfer. To verify this, we established an experiment using lysogenic phages isolated from environments exposed to long-term combined pesticide application and nitrogen deposition, infecting Escherichia coli to evaluate phage impacts on host resistance phenotypes. Aligned with the One Health concept, our study focuses on soil phages and aims to enhance understanding of the potential risk of phage-mediated ARG dissemination to human health. This research offers insights into the critical role of phages in ARG dissemination influenced by combined pesticides under global change factors.

Results

Effects of Combined Pesticide Pollution on Phage-Encoded ARGs under Nitrogen Deposition.

Soil samples were obtained from a long-term field control experiment examining the interaction between pesticides and nitrogen deposition. For each sample, five soil cores (10 cm in diameter) were randomly collected within a quadrat, and soil from the top 0 to 20 cm was combined and homogenized. DNA from the soil samples was subjected to viromic sequencing. Candidate viral sequences were screened following the method of Xia et al. (32). After viral sequence identification, ARGs within viral operational taxonomic units (vOTUs) were detected by aligning protein sequences against the Structured ARG (SARG) database using BLASTX, with an e-value ≤1e⁻⁵ and sequence identity ≤60% (33). A total of 55 ARG subtypes across 16 classes were annotated in 128 vOTUs from all soil samples. The five dominant ARG classes were macrolide–lincosamide–streptogramin (MLS), novobiocin, trimethoprim, bacitracin, and tetracycline, accounting for 21.94%, 15.75%, 15.69%, 13.06%, and 10.77% of the total ARG abundance, respectively. The NABBW treatment, which combines AB and BW composites under nitrogen deposition, increased ARG abundance to 3.81-fold of the control (CK) (ANOVA, P < 0.05; Fig. 1A). Additionally, Nonmetric multidimensional scaling (NMDS) analysis revealed significant differences in the structure of ARG-carrying vOTUs under pesticide pollution and nitrogen deposition (ANOSIM, R² = 0.3398, P = 0.026; SI Appendix, Fig. S1). Similarly, pesticide application under nitrogen deposition significantly influenced the mobile genetic elements (MGEs) annotated in the virome (ANOVA, P < 0.05; Fig. 1B). In total, 75 MGE-carrying vOTUs were identified, including transposase, istB, insertion_element_IS91, and integrase. Transposase was the most prevalent MGE type, accounting for 80.34% of the total MGE abundance. The NABBW treatment increased MGE abundance 5.40-fold compared to the CK (ANOVA, P < 0.05; Fig. 1B).

Fig. 1.

Phage-associated ARG composition in soil. (A) The abundance of ARGs detected in soil vOTUs (ANOVA, Duncan’s multiple comparison test). (B) The abundance of MGEs detected in soil vOTUs (ANOVA, Duncan’s multiple comparison test). (C) The abundance of ARGs across different risk levels (Q1, Q2, Q3, Q4) was assessed. Risk index levels were classified into four categories: Q1 (Top 25%), Q2 (25 to 50%), Q3 (50 to 75%), and Q4 (Bottom 25%) (ANOVA, Duncan’s multiple comparison test). High-risk ARGs (Q1, Q2) in soil were defined according to host pathogenicity, gene mobility, and human environmental enrichment. (D) ARG composition colored by ARG type. The outer and inner circles represent ARG types and subtypes, respectively. Note: Different letters indicate significant differences in treatment groups.

High-risk ARGs are defined by high mobility, pathogenic hosts, and human accessibility (HA), posing significant threats to public health (34). In total, 41 high-risk ARGs were identified (Dataset S1), accounting for 25.35% of the total ARG abundance. The abundance of high-risk ARGs under the NABBW treatment was 3.69-fold higher than that of the CK (t test, P < 0.05; Fig. 1C). Nitrogen deposition further enhanced the abundance of high-risk ARGs in soils exposed to pesticide pollution. Pesticide application, particularly combined pesticides, significantly increased the abundance of high-risk ARGs, including MLS resistance genes (macB) (t test, P < 0.05), multidrug resistance genes (efrB and msbA), tetracycline resistance gene (tet(A)), and polymyxin resistance gene (ugd) under nitrogen deposition (Fig. 1D). Moreover, a significant linear correlation was observed between ARG abundance and the number of vOTUs, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways, and CAZy annotations (P < 0.05; SI Appendix, Fig. S2).

Effects of Combined Pesticide Pollution on Macro- and Microdiversity and Interactions of Phage and Bacteria under Nitrogen Deposition.

The phages in this study formed 605 viral clusters (VCs) with existing phage datasets, soil phages shared 4 VCs with human gut, 29 VCs with forest soil, 53 VCs with freeze-thawing phages, and 5 VCs with deep-sea sediment phages, showing similarity to phages from forest soil and freeze-thawing samples. Among these, soil phages shared 16,288 viral contigs with freeze-thawing phages and 16,119 viral contigs with forest phages (Fig. 2A, SI Appendix, Fig. S3, and Dataset S2). Additionally, 50.30% of vOTUs were annotated, revealing Caudoviricetes as the dominant class across all samples (28.83 to 61.89%), followed by Malgrandaviricetes and Arfiviricetes (Fig. 2B). Under the influence of nitrogen deposition, combined pesticide application significantly increased the richness (richness index) (t test, P < 0.01) and diversity (Shannon index) (t test, P < 0.05) of soil phages. The richness and diversity in the NABBW treatment were 5.51-fold and 1.56-fold higher than that of the nitrogen addition treatment (NCK) (Fig. 2C). Moreover, pesticide pollution altered phage lifestyles, particularly under nitrogen deposition. Compared to pesticide pollution treatments, nitrogen deposition significantly exacerbated the impact of pesticide pollution on phage lifestyles, with the average proportion of lysogenic phage decreased from 70.73% in all pesticide pollution treatments without nitrogen deposition (including AB, BW, and ABBW) to 61.33% in all nitrogen deposition combined with pesticide pollution treatments (including NAB, NBW, and NABBW) (t test, P < 0.05; Fig. 2D and SI Appendix, Fig. S4).

Fig. 2.

Changes of soil phages with pesticide pollution under nitrogen deposition. (A) The Gene-sharing network associates grassland phages (our study, red nodes) in this study to viral genome datasets, including human gut (yellow nodes), forest soil (light blue nodes), freeze-thawing soil (dark blue nodes), and deep-sea sediment (green nodes). (B) Community compositions (family level) of phages. (C) Richness index and Shannon index of phages (t test; *P < 0.05, **P < 0.01). (D) Proportions of lytic phages and lysogenic phages (t test; *P < 0.05).

Intrapopulation variations (i.e., microdiversity) can improve ecological resilience and provide insights into population- and gene-level selective pressures (35–38). Examining shifts in viral microdiversity (via nucleotide diversity, π value) can reveal species’ stress responses in harsh or disturbance-prone environments. Compared with the untreated control group, the average nucleotide diversity (π-value) of phages declined significantly under pesticide stress (BW and ABBW; t test, P < 0.05, Fig. 3A) and all NP treatments (NAB, NBW, and NABBW; t test, P < 0.05, Fig. 3A). Under nitrogen deposition, pesticide stress further reduced the average nucleotide diversity of phages. Relative to the BW treatment, NBW showed a highly significant decrease in nucleotide diversity (t test, P < 0.0001, Fig. 3A). In nitrogen deposition treatments, nucleotide diversity related to repair and recombination functions increased, including genes associated with replication, recombination, and repair (wilcoxon-test, P < 0.05); posttranslational modification and protein turnover chaperones; and coenzyme transport and metabolism (Fig. 3B and SI Appendix, Fig. S5). Growth rate is central to understanding microbial interactions and community dynamics (39). Genome size is an important microbial trait that reflects the adaptability of an organism’s metabolic strategy to complex environments (40). Bacterial genome size and growth rates can help us to explore adaptive strategies and functional trade-offs of bacteria in different environments. Measurements of bacterial genome size and growth rate can help reveal adaptive strategies and functional trade-offs across environments. Compared with pesticide pollution alone, the average minimal doubling time of bacteria tended to increase when pesticide pollution occurred under nitrogen deposition, indicating reduced growth rates. Specifically, the average minimal doubling time in the NBW treatment was significantly higher than the BW treatment (t test, P < 0.05; Fig. 3C). Nitrogen deposition also affected bacterial genome size under pesticide exposure. Compared with the NCK treatment, genome size was significantly reduced under all NP treatments (NAB, NBW, and NABBW, t test, P < 0.05; Fig. 3D). Similarly, compared with the BW treatment, genome size in the NBW treatment was significantly lower (t test, P < 0.05; Fig. 3D). Compared to pesticide pollution alone, nitrogen deposition further increased the complexity of microbial networks in pesticide-polluted soils—evidenced by a higher number of nodes and edges, greater connectance, and increased average degree—and intensified microbial competition, as indicated by stronger negative correlations (SI Appendix, Fig. S6).

Fig. 3.

Genome-level changes of phages and bacteria. (A) Average nucleotide diversity of phages (t test; *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001). (B) Nucleotide diversity of functional genes annotated in eggNOG database of phages (wilcoxon-test; *P < 0.05). Note: The meaning of the letters on the x-axis is provided in SI Appendix, Fig. S5. (C) Average bacterial doubling time (t test; *P < 0.05). (D) Bacterial genome size (t test; *P < 0.05, **P < 0.01).

Phage–Host Interaction Exacerbated the Risk of ARGs with Pesticide Pollution under Nitrogen Deposition.

In total, phage hosts were identified across seven bacterial phyla, with Acidobacteriota, Verrucomicrobiota, Firmicutes, and Actinobacteriota serving as the primary hosts. Compared to pesticide pollution alone, pesticide pollution under nitrogen deposition altered the host composition, increasing the relative abundance of Verrucomicrobiota and decreasing the relative abundance of Acidobacteriota (Fig. 4A). Across all soil samples, 506 phage–host pairs were identified. The NCK treatment markedly reduced the number of phage–host pairs to only 57.09% of the CK. Interestingly, pesticide exposure under nitrogen deposition led to a 117.57% increase in phage–host associations and 251.48% increase in abundance of phages carried by the host, compared to pesticide pollution alone (P < 0.05, Fig. 4B and SI Appendix, Figs. S7 and S8). Additionally, antiphage defense systems were predominantly concentrated within these main host phyla, with Acidobacteria exhibiting the highest diversity of defense mechanisms (Fig. 4 C and D). Compared with pesticides alone, phage–Acidobacteria pairings decreased when exposed to pesticides under nitrogen deposition (Fig. 4B). In total, we identified 23 defense types, 32 subtypes, and 187 defense system genes across the seven bacterial phyla. These defense systems corresponded to seven known mechanisms, with “degrading nucleic acids,” “toxin–antitoxin,” and the “CRISPR-Cas system” being the three most prevalent. Among them, restriction modification system (RM system) was the most abundant defense system in all samples (Fig. 4D).

Fig. 4.

Analysis of phage–host interactions to pesticide pollution under nitrogen deposition. (A) Relative abundance of phages carried by the host (B) Predictive analysis of phage–host interactions. (Left) Different bacterial host MAGs at the family level. (Middle) The number of phage–host pairings in different samples. (Right) the vOTUs that can match to host MAGs. The bar chart below counts the number of pairings in each sample, and the classification of the different hosts in the bar is consistent with the information of the hosts on the Left. (C) The phylogenetic tree of the phage hosts. The first layer shows MAGs colored by taxonomic assignment at the phylum level. The second layer shows the phages linked to the host MAG. The third layer shows the relative abundance of each MAG. Circles indicate the number of vOTUs with predicted hosts in each phylum. (D) The antiphage defense systems in bacteria, including different defense types (vertical ordinate) and defense mechanisms.

A total of 553 viral auxiliary metabolic genes (AMGs) encoding various functional genes were identified through VIBRANT and DRAMv analyses, 215 of which were associated with metabolic functions. These AMG-carrying vOTUs tended to exhibit a lysogenic lifestyle (SI Appendix, Fig. S9). Compared to pesticide pollution alone, pesticide pollution under nitrogen deposition significantly reduced both the number (t test, P < 0.05; Fig. 5A and SI Appendix, Fig. S10) of phage-encoded AMGs related to carbohydrate metabolism, amino acid metabolism, and nucleotide metabolism. A similar pattern was observed for AMG abundance, with significant reductions under combined pesticide pollution. Compared with the AB and BW composite treatment (ABBW), the NABBW treatment showed a significant decline in AMG abundance (t test, P < 0.05; SI Appendix, Fig. S11). Putative AMGs linked to carbohydrate metabolism accounted for 52.09% of all metabolism-related AMGs. Specifically, 78 AMGs belonging to glycosyltransferase family 1 (GT1) and 8 to glycosyltransferase family 2 (GT2) were identified. Notably, the abundance of GT1 AMGs in the NABBW treatment significantly declined to 14.84% of the ABBW treatment and to 11.58% of the CK (Fig. 5B). Furthermore, we identified vOTUs capable of encoding both ARGs and AMGs. For instance, AMGs (ABC-2. P) and ARGs [tetA(41)] related to Adenosine Triphosphate-binding cassette (ABC) transporters co-occurred within the same vOTU (SI Appendix, Table S1). Strikingly, when compared with pesticide pollution alone, we observed that the abundance of vOTUs co-occurring with ARGs and AMGs markedly increased in pesticide pollutions with nitrogen deposition, especially in treatments with combined pesticides (t test, P < 0.05; Fig. 5C and SI Appendix, Fig. S12). Among these, half of the AMGs were related to the RmlD substrate-binding domain, which is linked to bacterial and pathogen virulence, while one AMG encoded an ABC-2 type transport system permease protein (Fig. 5D). Structural predictions using Phyre2 showed no less than 99.6% and 99.4% confidence for the predicted structures of phage-encoded AMGs and ARGs, respectively. These high-confidence predictions suggest that the AMGs and ARGs can encode functional proteins capable of exerting biological activity (Fig. 5D and SI Appendix, Fig. S13).

Fig. 5.

Phage-encoded AMGs with pesticide under nitrogen deposition. (A) The number of AMGs associated with metabolism (t test; *P < 0.05). (B) AMGs abundance of carbohydrate metabolism-related processes in viral communities based on KEGG, CAZY, and Pfam database. (C) The abundance of vOTUs that co-occurrence with ARGs and AMGs. Significance: Asterisks of different colors indicate statistically significant differences relative to the treatment group represented by the corresponding color (t test; *P < 0.05, **P < 0.01). (D) Genomic context of five phages encoding AMGs and ARGs. (E) Protein structure of phages encoding ARGs. Note: Different letters indicate significant differences among treatment groups. Tertiary structural homology and protein structure predictions of AMGs were generated from their amino acid sequences using Phyre2 (v2.0).

Phage-Mediated Evolution of Host Antibiotic Resistance.

To further investigate whether combined pesticides under nitrogen deposition enhance the potential of phages to mediate ARG transfer and increase host resistance, we conducted experiments evaluating the resistance phenotype of E. coli in response to phage exposure. Screen from the results of predicting the viral lifestyle of phages encoding ARG to calculate the lytic/lysogenic proportion. Our results revealed that the majority of phages encoding ARGs were lysogenic, accounting for 91.60 to 99.53% (ANOVA, P < 0.05, Fig. 6A). These lysogenic phages stimulated and promoted the growth and proliferation of E. coli, with the number of E. coli in the NABBW treatment reaching 2.19-fold that of the CK treatment (ANOVA, P < 0.05; Fig. 6B). Similarly, the abundance of host progrowth genes carried by lysogenic phages, which can enhance host fitness, was substantially higher in the NABBW treatment than in the CK group (SI Appendix, Fig. S14 and Dataset S3). At a concentration of 16 μg mL⁻¹ tetracycline, phage-encoded ARGs significantly enhanced E. coli resistance compared to the CK treatment. Under lysogenic phage coculture, the number of E. coli in the NABBW treatment was 25.07-fold that of the CK treatment (ANOVA, P < 0.05, Fig. 6C). Combined pesticide application with nitrogen deposition also reduced the inhibitory effect of tetracycline on E. coli, with inhibition rates of 99.32% in the CK treatment and 92.49% in the NABBW treatment (ANOVA, P < 0.05, Fig. 6D). Similarly, at 0.5 μg mL⁻¹ ciprofloxacin, phage-encoded ARGs significantly increased E. coli resistance compared to the CK treatment. The E. coli population in the NABBW treatment was 17.54-fold of the CK treatment under lysogenic phage coculture (ANOVA, P < 0.05, Fig. 6E), with inhibition rates of 92.60% in the CK treatment and 40.69% in the NABBW treatment (ANOVA, P < 0.05, Fig. 6F).

Fig. 6.

Combined pesticides under nitrogen deposition promoted the increase of resistance phenotype to E. coli. (A) Proportions of lytic and lysogenic phages that encoding ARGs. The specific workflow was as follows: i) viral contig identification; ii) viral lifestyle prediction; iii) annotation of ARGs within vOTUs; and iv) screening and calculating viral lifestyle annotations encoding ARGs. (B) The number of E. coli cocultured with lysogenic phages. (C) The number of E. coli exposed to tetracycline. (D) The inhibition of E. coli exposed to tetracycline. (E) The number of E. coli exposed to ciprofloxacin. (F) The inhibition of E. coli exposed to ciprofloxacin. Note: Different letters indicate significant differences in treatment groups.

Discussion

Our results indicated that combined pesticide pollution significantly increased the abundance of phage-encoded ARGs and MGEs, whereas no significant difference was observed under single pesticide pollution. Two possible explanations may account for this. First, the strong lipophilicity of chlorpyrifos may lead to prolonged coadsorption of the compound pesticides (azoxystrobin + propiconazole) onto soil particle surfaces. This could damage cell membranes and exacerbate oxidative stress, thereby promoting phage attachment to bacterial DNA and free ARGs (42, 43). Second, the quinone structure of azoxystrobin and the organophosphorus component of chlorpyrifos may induce or interfere with phage genomes, leading to replication errors or gene recombination events (44, 45). Furthermore, we found that nitrogen deposition significantly enhanced the abundance of phage-encoded ARGs and MGEs, suggesting that nitrogen deposition acts as an additional selective pressure, potentially intensifying the effects of combined pesticide pollution on phages. It is likely that nitrogen deposition leads to soil acidification (SI Appendix, Fig. S15), which may increase pesticide solubility and mobility (46). Additionally, nitrogen input may suppress oligotrophic bacteria and reduce the activity of key enzymes such as catalase, thereby hindering pesticide degradation (30, 31). In our study, the increase in phage-encoded high-risk ARG abundance was paralleled by a rise in phage-encoded MGEs. The elevated presence of MGEs, such as transposases, facilitates the “capture” and “release” of ARGs in a dynamic process, promoting the spread of high-risk phage-encoded ARGs among different bacterial hosts. Among these high-risk ARGs, macB, tet(A), efrB, and msbA show high clinical accessibility, human pathogenicity (HP), and mobility potential in the environment (Dataset S4). Recent studies have reported a global rise in macrolide resistance in Mycoplasma pneumoniae. Notably, macB, a membrane protein in the ABC transporter family, mediates the extrusion of macrolide antibiotics and is believed to play a key role in efflux pump–mediated resistance in M. pneumoniae (47). Tetracyclines are also widely used in clinical settings. For example, resistance to tigecycline in highly pathogenic Salmonella is predominantly conferred by tet(A) (48). Compared with natural environments, phages with broad host ranges are more likely to infect multiple hosts under anthropogenically impacted stressful conditions (41). Such phages tend to carry a greater abundance of ARGs with higher transcriptional activity, accelerating the dissemination of high-risk ARGs within potentially pathogenic bacteria (3). This, in turn, heightens the challenge of managing infectious diseases and poses greater risks to public health.

We found that pesticide pollution altered the composition of soil phages, increasing both the proportion of lytic phages and the alpha diversity of phages under nitrogen deposition. A likely explanation is that the combined stresses of pesticides and nitrogen deposition trigger the host’s SOS response, leading to shifts in bacterial communities that, in turn, drive changes in phage composition and promote the transition of temperate phages into the lytic cycle (10). As pesticide pressure increases under nitrogen deposition, phage evolution and genetic variation may be affected by reduced microdiversity, leading phages to selectively retain specific genes (e.g., those involved in replication, repair, and genome reorganization). At the genomic level, the retention of genes that reduce or eliminate excess metabolic burden may represent a survival strategy for host microbes under environmental stress (49). In our study, the observed reduction in bacterial growth rate and genome size suggests that smaller genomes may facilitate a rapid response to environmental changes. This likely results from the removal of redundant genes (e.g., nonessential metabolic pathways), which simplifies gene expression regulation and supports the retention of core stress-resistance genes (40, 50). Soil bacteria with smaller genomes have also been reported to play crucial roles in adaptation to arid environments (49). In line with the Black Queen Hypothesis, this genome streamlining strategy reduces metabolic burden (51), allowing bacteria to better withstand pesticide exposure under nitrogen deposition.

We observed that phage infection was enhanced under pesticide pollution combined with nitrogen deposition, as evidenced by strengthened phage–host interactions, increased abundance of phages carried by the host, and a reduction in bacterial antiphage defense systems. The decline in such defense systems facilitates greater susceptibility to viral infection (32). For example, the RM system, the most abundant defense system across all samples, prevents phage replication by cleaving nonbacterial DNA with restriction endonucleases, suggesting that its reduction favors viral infection (52). This decrease in antiphage defenses further supports the observed enhancement of phage–host interactions under pesticide exposure with nitrogen deposition. Additionally, the increased presence of genes related to replication, repair, and genome reorganization in phages further supports the notion of intensified phage–host interactions under these conditions. This is likely because phages may undergo mutation or recombination to integrate new genes, enhancing their ability to infect hosts and adapt to environmental stress (53). The enhanced phage–host interactions may increase the transfer of genes carried by phages. There are three main reasons. First, intensified phage infection and damaged to host cell membranes in response to the combined pesticides may facilitate more efficient injection of phage DNA. Second, phage infection can activate the host’s DNA repair system (e.g., RecA), promoting homologous recombination between phage-encoded ARGs and the host genome (54). Third, as phage infection increases, the integration efficiency of phage genes also improves (55), elevating the likelihood of phage-encoded ARG transfer to the host. In summary, our results indicate that combined environmental stress intensifies phage–host interactions, increasing the potential for phage-mediated gene transduction.

Our results further revealed a substantial increase in the abundance of phages co-occurring with both ARGs and AMGs by pesticide pollution under nitrogen deposition, especially for combined pesticides. This co-occurrence may enhance phage-encoded ARG dissemination through three primary mechanisms. First, it may facilitate ARG spread by increasing the diffusion range of phages and the likelihood of contact with host bacteria (56). Second, since AMGs and ARGs were located in the same region of the phage genome, their co-occurrence allows for simultaneous transfer to host bacteria via phage-mediated transduction. Finally, these AMGs may promote host survival and competitiveness, indirectly facilitating the dissemination of phage-encoded ARGs. This could be attributed to the ability of AMGs to optimize host metabolism or stress responses, thereby prolonging host survival or enhancing competitiveness, which in turn supports greater phage replication and the spread of their associated genes (41). For example, in our study, AMGs carried by phages were associated with microbial virulence and ABC transporter proteins. These AMGs were found to encode structurally intact proteins capable of performing functional roles. ABC transporters, a superfamily of integral membrane proteins, facilitate small molecule transport, enhance virulence, and contribute to multidrug resistance by enabling bacteria to acquire extracellular DNA fragments containing ARG (57). These findings suggest that phage-encoded AMGs may enhance microbial stress resistance and host adaptation. However, the co-occurrence of ARGs with AMGs also increases the likelihood of their simultaneous transfer.

In our study, the major ARG-encoding phages were lysogenic. These lysogenic phages exhibit a higher frequency of HGT than lytic phages and are more likely to transfer genes into their host cells (58). The increased growth of host bacteria following cocultivation with lysogenic phages under combined pesticide exposure and nitrogen deposition suggests that lysogenic phages may enhance host metabolism and environmental adaptation by transferring AMGs that promote host fitness and resistance. Phages encoding AMGs can participate in the host’s nutritional metabolism and may include genes that directly enhance host fitness (59). These AMGs were primarily associated with nucleic acid metabolism, nucleotide metabolism, amino acid metabolism, energy metabolism, protein synthesis, and amino acid recycling. For example, genes related to DNA (cytosine-5)-methyltransferase may influence cell division by regulating DNA replication and gene expression (60); genes encoding small or large subunit ribosomal proteins contribute to ribosome assembly, essential for protein synthesis (61) and genes associated with glucose-6-phosphate 1-dehydrogenase contribute to the production of NADPH and nucleotide precursors (62) (SI Appendix, Fig. S14 and Dataset S3). Lysogenic phages can carry AMGs that favor better host metabolism, growth, and fitness (63). In extreme environments, for instance, phage–bacteria interactions can shift from parasitism to protective mutualism, highlighting the important role of lysogenic phages in bacterial adaptation to environmental stress (64). Furthermore, as bacterial competition was intensified by pesticide pollution under nitrogen deposition, phages encoding ARGs were able to rapidly reinfect other hosts and confer resistance, significantly enhancing the competitive advantage of their bacterial hosts. This occurs because, unlike vertical gene transfer, phage-mediated transduction can transfer ARGs within hours, allowing bacteria to rapidly acquire resistance and bypass the slower process of natural selection (65). This clearly demonstrates the competitive advantage conferred by phage-mediated gene transduction. These phages with specific genes (e.g., ARGs and AMGs), can directly influence the evolution and adaptation of their hosts (66, 67). For instance, the Erysipelothrix phage can acquire mel and tetM genes from its immediate host and transmit ARGs to other bacteria across genera (e.g., Bacillus coagulans) (68). Phage genomes also harbor numerous AMGs that confer microbial resistance to various stresses, including arsM and efflux pump genes linked to heavy metal detoxification and antibiotic resistance (69, 70). Similar findings were reported in paddy soils, where phage-mediated HGT of arsM helped microorganisms adapt to arsenic stress (71). Our in vitro incubation experiments further supported this hypothesis, providing strong evidence that lysogenic phages isolated from long-term combined pesticides under nitrogen deposition improved the phenotypic resistance of E. coli to tetracycline and ciprofloxacin, compared with phages from control soils. These findings highlighted the critical role of phage-mediated gene transfer in driving host antibiotic resistance, particularly in the context of global environmental change.

In conclusion, our study provides important insights into the critical role of phages in the spread of ARGs under global change. Specifically, nitrogen deposition significantly increased the abundance of phage-encoded high-risk ARGs in pesticide-polluted soils. This primarily facilitated the spread of phage-encoded ARGs to host bacteria by intensifying phage–host interactions and increasing the co-occurrence of ARGs and AMGs. Such phage-mediated ARG transfer may elevate the potential risk of ARG exposure to humans through the food chain, water cycle, and direct environmental contact, indicating that phages could act as a potential conduit for ARGs to enter the human body from the environment. Comparing the soil phages in this study with those from typical and extensively studied habitats holds significant ecological relevance, as it enables the identification of shared and unique phages. Unique phages expand viral databases and uncover novel viral resources, whereas shared phages provide insights of potential universal significance, particularly in response patterns of prevalent phages. Such cross-environment comparisons reveal the universality and specificity of phages in biogeochemical cycles and improve our understanding of their host interactions, environmental adaptations, human health connections, and ecological roles. Our soil viral community shared phages with other habitats (e.g., forest soils and freeze-thawing soils), suggesting a potential risk of phage-mediated ARG dissemination in other environments. These findings underscore the importance of investigating the combined effects of soil pollution and global change factors, as well as their interactions. Our results also highlight the significant role of phage-mediated ARG dissemination in bacterial resistance, emphasizing the need to account for phage contributions in ARG risk assessments. Based on the ‘one health’ concept, the development of environmental protection and agricultural management strategies should prioritize phage–host interactions and evolutionary processes in habitats affected by interacting global change drivers. This study demonstrates that inevitable nitrogen deposition intensifies the adverse effects of pesticide pollution on soil ecosystems, underscoring the urgent need for stricter pesticide industry standards. It is equally important to establish long-term monitoring networks and integrate predictive modeling approaches to assess the combined impacts of pesticide pollution, nitrogen deposition, and other global changes on the future dynamics of ARG networks.

Materials and Methods

Site Selection and Sample Setup.

To investigate the long-term effects of pesticides and nitrogen on soil, we selected a grassland steppe with minimal human disturbance and low pesticide residue levels near the Erguna Forest-Steppe Ecotone Ecosystem Research Station (50°10′N, 119°22′E, Northeast China) for a field experiment. The area experiences cold winters and hot summers, with average annual precipitation ranging from 200 to 450 mm, mostly occurring in summer. The soil type is black calcareous soil, and the vegetation is primarily composed of tufted grasses and conifers. This midtemperate, semiarid continental grassland climate zone is ecologically unique and functionally important, playing a vital role in maintaining regional ecological balance and promoting sustainable development. As global food security remains a pressing issue, increased agricultural production often parallels rising pesticide exposure. Simultaneously, global reactive nitrogen emissions are increasing, and excessive nitrogen deposition poses significant threats to ecosystems (72). Drawing on concentrations from numerous long-term nitrogen deposition experiments, 100 kg N ha⁻¹ y⁻¹ is considered an appropriate level with reliable simulation results (73, 74). To simulate nitrogen deposition, we applied slow-release urea, establishing two treatments: an N control (0 kg N ha⁻¹ y⁻¹) and a nitrogen addition treatment (100 kg N ha⁻¹ y⁻¹).

In addition, we selected two commercial pesticides widely used in agriculture (75, 76): azoxystrobin + propiconazol (AB) and chlorpyrifos (BW). Chlorpyrifos (C₉H₁₁Cl₃NO₃PS), a typical organophosphate pesticide targeting subterranean insects, can induce cell death by releasing proapoptotic protein (44). Azoxystrobin (C₂₂H₁₇N₃O₅) and propiconazol (C₁₅H₁₇Cl₂N₃O₂), which target aboveground fungi, affect microbial growth by inhibiting mitochondrial energy synthesis and disrupting cell membrane integrity, respectively (18). Our experiment included eight treatments combining nitrogen fertilizer and pesticide application, with three replicates (i.e., three plots) per treatment. Each quadrat measured 4 m² (2 m × 2 m), with adjacent quadrats within a plot separated by 1 m and adjacent plots separated by 2 m. Pesticide exposure and concentrations were based on application rates commonly used by local farmers and the methodology of Ni et al. (18). The eight treatments were labeled as “CK,” P (“AB,” “BW,” “ABBW”), “NCK,” and NP (“NAB,” “NBW,” “NABBW”). Here, AB, BW, and ABBW refer to azoxystrobin + propiconazole, chlorpyrifos, and their combined application, respectively; ‘N’ represents nitrogen fertilizer addition. CK refers to the control treatment with no pesticide or nitrogen application, while NCK refers to nitrogen addition only. Soil samples were collected after 3 y of exposure. From each quadrat, five soil cores (10 cm diameter) were randomly collected from the top 0 to 20 cm and thoroughly mixed. Detailed information is provided in Dataset S5.

Amplicon Sequencing and Metagenome Data.

DNA was extracted from grassland soil samples using the FastDNA Spin Kit for Soil (MP Biomedicals, USA). High-throughput sequencing of 16S rRNA amplicons was performed using the 515F/806R primers. DNA extraction, sequencing, trimming, and mapping followed the procedures described in previous studies (50). Microbial groups (bacteria) at the phylum level were used for microbial network analysis based on Spearman correlation (≥0.6) and | P | < 0.05. Detailed methods are provided in SI Appendix, Text S1.

Metagenomic sequencing was performed on DNA extracted from the soil samples. Libraries were prepared using the NEB Next Ultra™ DNA Library Prep Kit and sequenced on the NovaSeq 6000 platform (Illumina, San Diego, CA). Raw reads from the metagenomes were trimmed using fastp (v0.22.08) (77) and contigs were independently assembled using MEGAHIT (v1.2.9) (78). Configuration grouping was conducted with MetaBAT2 (v2.12) using default parameters (79). The completeness and contamination of metagenome-assembled genomes (MAGs) were evaluated using CheckM (v1.2.0), retaining MAGs with completeness ≥50%, contamination <10%, and strain heterogeneity <10%) (80), and taxonomically classified using GTDB-Tk (v2.1.1) (81). Minimal doubling time was estimated by analyzing codon usage bias patterns in each MAG containing more than 10 ribosomal proteins, using gRodon v.1 (39). In addition, the average genome size per bacterial contig was assessed using the MicrobeCensus pipeline (82).

Viromic Sequencing and Analysis.

Total DNA was extracted from grassland soil viromes using the MiniBEST Viral RNA/DNA Extraction Kit 5.0. High-quality DNA samples were used to construct viral community sequencing libraries with the NEB Next Ultra DNA Library Prep Kit for Illumina, and 150 bp bipartite sequencing reads were generated using the Illumina platform. Subsequent assembly was identical to that of the metagenome.

Candidate phage sequences were identified using VIBRANT (v1.2.1) (83), DeepVirFinder (v1.1) (84), and VirSorter2 (v2.0) (85). vOTUs were clustered using MUMmer, based on a 95% average nucleotide identity and 85% sequence length threshold (86). The quality and completeness of viral sequences were assessed using CheckV (v1.0.1) (87). Read counts were normalized to reads per kilobase per million mapped (RPKM) for vOTU abundance estimation using bwa (v0.7.12) (88). Phage coding potential was analyzed using DRAMV (89) (AMGs scored 1 to 3, AMG flags -M and -F) in combination with VIBRANT (83). Tertiary structural homology and protein structure predictions of AMGs were generated based on their amino acid sequences using Phyre2 (v2.0) (http://www.sbg.bio.ic.ac.uk/phyre2/html/page.cgi?id=index). Phage taxonomic assignment and vOTU annotation were conducted using geNomad (v.1.5.1) with “end-to-end” and other default parameters (90). Viral shared networks constructed by vConTACT3 were visualized using Cytoscape (v3.10.0) (91). Functional annotation of related genes was performed using DIAMOND against the KEGG (92), CAZy (93), and eggNOG databases (94). Population microdiversity metrics, including nucleotide diversity (π) of vOTUs, were calculated using inStrain (v1.5.3) (95). ARGs within vOTUs were annotated by aligning sequences against the SARG (v3.2) database using BLASTp, with thresholds of e-value ≤10⁻⁵, query coverage ≥80%, and amino acid identity ≥60%. ARG risk in soil was assessed following Zhang et al. (96), classifying ARGs into six risk levels: high-risk (Q1, Q2), low-risk (Q3, Q4), zero, and unassessed. The overall health risk of each ARG was assessed using four calculated metrics: HA, mobility (MO), HP, and clinical availability (CA). The risk index (RI) is calculated as

$$\mathrm{RI}=\mathrm{HA}\times \mathrm{MO}\times \mathrm{HP}\times \mathrm{CA}.$$

The specific calculation formula is as follows:

$${\mathrm{RI}}_{\mathrm{sample}}={\sum }_{\mathit{i} =1}^{\mathrm{n}}{\mathrm{Aundance}}_{\mathit{i}}\times {\mathrm{RI}}_{\mathit{i}}.$$

Among these variables, “Abundance” and “RI” denote the abundance and risk value of ARG “i” within a sample, respectively, where “n” represents the total number of ARGs detected.

RI values are classified into four categories: Q1 (top 25%), Q2 (25 to 50%), Q3 (50 to 75%), and Q4 (bottom 25%). MGEs of vOTUs were annotated using the SARG database with BLASTp, applying stricter thresholds (e-value ≤10⁻¹⁰, query coverage ≥80%, and amino acid identity ≥60%).

Phage–Host Interaction Analysis.

Three predictive computational methods were employed to accurately determine phage hosts. The first method involved identifying CRISPR spacer sequences within bacterial MAGs using the CRISPR-match identification tool. These spacer sequences were then aligned to viral genomes using BLASTn, with a threshold of 95% identity and ≤2 single nucleotide polymorphisms. The second method used tRNAscan-SE to predict tRNA genes in phage contigs under default settings. Predicted tRNA sequences were compared with bacterial sequences using BLASTn, and hits with at least 95% sequence identity were retained to infer phage–host associations (97). The third method involved direct comparison of bacterial and viral genome features using BLASTn, with filtering criteria of bitscore ≥50, e-value ≤10⁻³, identity ≥70%, and alignment length ≥2,500 bp). Phage–host matches were determined by integrating the results of all three methods. Additionally, DefenseFinder was used to identify antiphage defense systems in bacteria, with defense mechanisms classified according to previously established categories (98). All predictions regarding viral lifestyle followed the identification methods described in ref. 32. Viral lifestyles were predicted using DeePhage (v1.0) and classified as lysogenic (cutoff ≥0.6) or lytic (cutoff ≤0.4) (99). For cases where the DeePhage probability score was inconclusive (0.4 to 0.6), lysogenic marker proteins in phage sequences were identified using “hmmscan” in HMMER (v3.1b2) with an e-value threshold of 10⁻⁵ (100). Any phage containing a lysogenic marker protein was classified as lysogenic. The specific calculation formula is as follows:

$$\mathrm{Lysogenic} \mathrm{proportion}=\frac{{\mathrm{Abundance}}_{\mathrm{lysogenic} \mathrm{pahges}}}{{\mathrm{Abundance}}_{\mathrm{total} \mathrm{phages}}},$$

$$\mathrm{Lytic} \mathrm{proportion}=\frac{{\mathrm{Abundance}}_{\mathrm{lytic} \mathrm{pahges}}}{{\mathrm{Abundance}}_{\mathrm{total} \mathrm{phages}}}.$$

Antibiotic Resistance of Lysogenic Phages to E. coli.

To verify whether lysogenic phages enhance the resistance phenotypes of soil microbes, we selected tetracyclines (tetracycline) and quinolones (ciprofloxacin), two widely used and distinct classes of antibiotics, for comparative analysis (101). Tetracyclines are a commonly used class of natural broad-spectrum antibiotics (102, 103), whereas quinolones are synthetic broad-spectrum antibiotics. For more comprehensive validation, we screened for tetracycline resistance genes detected in the virome that ranked among the top five most abundant phage-encoded ARG classes. Conversely, quinolone (ciprofloxacin) resistance genes were not detected in the virome. Phages were first isolated and extracted by mixing fresh soil samples (CK and NABBW) with M9 medium, followed by centrifugation. The supernatant was filtered and concentrated using a TFF system (Merck Cogent μScale, USA; 0.22 μm and 100 kDa cutoffs). The resulting 0.22 μm cell suspension was retained as the bacterial community. Mitomycin-C was then added to induce lysogenic phages, which were subsequently collected using an Amicon Ultra filter (a control group was prepared without added phages). We selected the nonpathogenic strain E. coli DH5α for cultivation due to its high transformation efficiency, plasmid stability, and ability to yield high-quality DNA. E. coli preserved in glycerol at −80 °C was revived and inoculated into freshly sterilized LB liquid medium, followed by incubation at 37 °C with shaking at 220 rpm for 12 h. Three treatments, control, lysogenic phage, and inactivated lysogenic phage, were established for incubation experiments with E. coli. Each treatment was incubated at 25 °C for 2 d in dark conditions. All tests were conducted in triplicate. The resistance phenotype was evaluated by plate counting. Detailed procedures are provided in SI Appendix, Text S2.

Statistical Analysis.

Data analysis was conducted using R (v4.0.3), Excel 2019, and IBM SPSS Statistics 26. One-way ANOVA and Student’s t test were used to assess the significance of differences. NMDS ordination was performed in R using the “vegan” package. The alpha diversity index of the virome was calculated using the “diversity” function in the vegan package in R (v4.0.3). Data visualization and plotting were conducted using Origin 2021, Adobe Illustrator 2024, GraphPad Prism 9.5, and Chiplot (https://www.chiplot.online/).

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Dataset S05 (XLSX)

Data, Materials, and Software Availability

Metagenomic and viromic sequencing data generated in this study have been deposited in the National Centre for Biotechnology Information (NCBI) Sequence Read Archive under accession number PRJNA1212801 (104). The 16S rRNA gene sequencing reads are available under BioProject PRJNA1158439 (105). All other data are included in the article and/or supporting information.