Increasing pesticide diversity impairs soil microbial functions

Significance

Understanding the impacts of pesticide diversity on soil microbial functions remains limited, particularly under different nitrogen fertilizer management practices. Our study demonstrated that higher pesticide diversity can significantly and adversely affect soil microbial functions, thereby intensifying microbial consumption of soil resources. This finding suggests that pesticide diversity acts as an environmental stressor that drives substantial changes in the genomic traits of microbial communities and nutrient cycling in agricultural soils. However, these adverse effects of pesticide diversity can be mitigated by nitrogen fertilizer input. These insights enhance our understanding of the role of pesticide diversity in biogeochemical cycling and soil health, supporting the development of effective nitrogen management strategies in agriculture.

Abstract

Pesticide application is essential for stabilizing agricultural production. However, the effects of increasing pesticide diversity on soil microbial functions remain unclear, particularly under varying nitrogen (N) fertilizer management practices. In this study, we investigated the stochasticity of soil microbes and multitrophic networks through amplicon sequencing, assessed soil community functions related to carbon (C), N, phosphorus (P), and sulfur (S) cycling, and characterized the dominant bacterial life history strategies via metagenomics along a gradient of increasing pesticide diversity under two N addition levels. Our findings show that higher pesticide diversity enriches the abundance of bacterial specialists and opportunists capable of degrading or resisting pesticides, reducing the proportion of bacterial generalists in the absence of N addition. These shifts can complicate multitrophic microbial networks. Under increased pesticide diversity, selective pressure may drive bacteria to streamline their average genome size to conserve energy while enhancing C, N, P, and S metabolic capacities, thus accelerating soil nutrient loss. In comparison, N addition was found to reduce bacterial niche differentiation at higher pesticide diversity, mitigating the impacts of network complexity and functional traits associated with pesticide diversity, ultimately alleviating soil nutrient loss. Our results reveal the contrasting impacts of pesticide diversity on microbial functions under different N input scenarios and emphasize that strategic N fertilizer management can mitigate the ecological effects of pesticide use in agricultural systems.

Pesticides are essential for protecting crops from diseases and pests, making them a crucial component of modern agriculture (1–3). However, pesticides, as bioactive and toxic substances, also create environmental stress for microbes in agricultural soils, with 30 to 50% of applied pesticides not reaching their intended targets and instead settling on the soil surface (3–5). Most research has focused on the effects of a single pesticide type or quantity on soil microbes (6–9). For example, soil microbial carbon (C) function genes (cbhl and chiA) were enriched as pesticide application doses (7). The use of metalaxyl-M (a fungicide) may promote metalaxyl-M resistance within the fungal community (7, 10) and shift the microbial community composition toward fungi, which can potentially impact soil organic matter formation (7, 11). However, real-world applications often involve combinations of multiple pesticides with diverse targets. Studies indicate that over 51% of agricultural topsoils in the European Union contained more than five pesticides (12), and over 58% of managed grassland soils contained two or more pesticide residues (13). Therefore, pesticide diversity (i.e., the number of pesticides used) poses an environmental stress that requires urgent investigation (6, 14, 15). To date, no quantitative analysis has assessed the impact of pesticide diversity on microbial communities in agricultural soils.

Understanding community assembly mechanisms is essential to linking community structure and interactions with the functions of soil microbial communities (16–19). Specifically, the relative contributions of deterministic and stochastic processes to community assembly can be influenced by environmental stress (16, 19–22). Environmental stress typically favors more deterministic microbial community assembly, promoting microbial community differentiation (23, 24) and enhancing microbial co-occurrence due to complementary microbial traits (25, 26); however, microbial community assembly has also been found to be more stochastic under stressful conditions (27–31). Evidence suggests that pesticide-induced environmental stress could alter microbial competitive interactions (3, 32) and reduce microbial network complexity (5, 33, 34). Studies also consistently report adverse pesticide effects on soil nitrogen (N) cycling (3, 35, 36). These findings apply to single pesticide types or quantities, not multiple pesticide stress. Furthermore, N is the primary limiting nutrient for photosynthesis and crop growth, compared to phosphorus (P), potassium (K), and sulfur (S) (37, 38). Managing N is central to agricultural nutrient practices for achieving high crop yield (38, 39). N addition can increase soil nutrient availability and the soil C/N ratio, potentially exerting environmental selection for well-adapted taxa or promoting competitive dominance, leading to niche convergence and reducing niche differentiation, thereby decreasing microbial co-occurrence (25, 40, 41). N addition has also been shown to increase the abundance of functional genes linked to N cycling (e.g., amoA, nifH) (42–44) and labile C degradation (e.g., chitinase) but decrease genes associated with recalcitrant C degradation (e.g., amyA, cda), potentially reducing the processing of persistent soil C molecules (43, 45). Varying N addition levels frequently coincide with pesticide application in agriculture. Studies suggest that N addition can influence pesticide effectiveness and alter pesticide impacts on soil microbes (46–48). However, it remains unclear to what extent pesticide diversity affects soil microbiome assembly and interactions or whether these changes threaten essential soil functions, such as elemental cycling, under various N addition levels in agricultural soils.

Community aggregated traits (CATs), calculated from metagenomic sequencing data, are frequently used to characterize changes in the organismal strategies that dominate bacterial communities in situ (49). Previous research has confirmed that CATs effectively evaluate theoretical life history strategy schemes for microbial communities in their natural environments (49–53). In these studies, the dominant strategies in bacterial communities were used to explain the responses of key bacterial groups, which carry essential functional genes, to soil pH, C/N ratio (environmental stress), and annual precipitation (global climate change) (49, 50). According to genome streamlining theory, environmental stress tends to select for gene loss, as smaller genomes offer adaptive advantages over larger ones (49, 54). Streamlining average genomes at the microbial community level may enhance taxonomic diversity and the abundance of functional genes related to C, N, P, and S cycling, potentially accelerating biogeochemical cycles, as observed when soil pH shifts from acidic to neutral (51). However, the impact of changes in microbial life history strategies on nutrient cycling in agricultural soils under environmental stress from pesticide diversity with varying N addition levels has not yet been explored.

In this study, we examined the assembly processes, community interactions, and functionality of soil microbial communities across a gradient (0 to 4) of pesticide diversity under two N addition levels (N control and N addition: 0 and 100 kg N ha⁻¹ y⁻¹) in Northeastern China. Additionally, we used a dataset of soil metagenomic sequences from major biomes to quantify the key trait dimensions of soil bacterial communities and their responses to pesticide diversity, to elucidate underlying mechanisms under different N addition levels. We hypothesized that 1) increasing pesticide diversity would reduce soil microbial stochasticity and promote microbial community differentiation, thereby complicating multitrophic networks without N addition; 2) increasing pesticide diversity would streamline bacterial average genome size and enhance the abundance of specific functional genes related to C, N, P, and S cycling, thus intensifying microbial processing of soil resources; and 3) N addition would reduce soil microbial community differentiation due to nutrient inputs, mitigating the effects of pesticide diversity on microbial interactions and functions, ultimately alleviating microbial processing of soil resources. Our study emphasizes the ecological impact of pesticide type diversity, which has often been overlooked, rather than pesticide concentration alone. This research enhances our understanding of microbial adaptation to pesticide diversity through the lens of multitrophic interactions and bacterial life history strategies. These insights into the combination of nutrient N management and pesticide pollution mitigation will strongly support sustainable agricultural development.

Results

Effects of Pesticide Diversity on Soil Properties.

In the treatments without N addition (N control), increased pesticide diversity caused a significant reduction in several soil property indexes (Fig. 1A and SI Appendix, Figs. S1–S3). For example, pesticide application led to a notable decrease in soil total C (TC) and total N (TN) content [regression coefficients (β): −0.027 and −0.020, P < 0.01) in the no-N-added treatments, as shown by linear mixed-effects models (Fig. 1A). Electrical conductivity (EC) and NO₃⁻-N contents were also strongly negatively associated with pesticide diversity (P < 0.05; SI Appendix, Fig. S3). Overall, pesticide diversity was negatively correlated with soil resource-related properties (SRP) in the no-N-added treatments (SI Appendix, Fig. S3). In contrast, the effects on soil property indices were attenuated in the N-added treatments (Fig. 1 A and B and SI Appendix, Fig. S3).

Fig. 1.

Effects of pesticide diversity on soil properties, microbial diversity, and network topological parameters. (A) Effects of pesticide diversity on various soil properties, microbial diversity, and network topological parameters based on linear mixed-effects models (n = 16). We selected 13 soil property indexes, 13 microbial diversity indexes, and 22 network topological parameters. Detailed information on methods is available in SI Appendix, Supplementary Methods (6). APL, average path length; Average CC, average clustering coefficient. (B−D) Relationships between the overall soil properties, biodiversity, network complexity, and pesticide diversity in treatments with no N added (N control) and N addition treatments (n = 16). The significance of each equation was determined using an F-test (***, **, and * indicate P < 0.001, < 0.01, and < 0.05, respectively). The equations, adjusted R², and P values from the regressions are shown near the regression lines. The error bands shaded in surrounding the regression lines indicate the 95% CI. Biotic Shannon, biotic richness, and biotic evenness represent the Shannon, richness, and evenness of multiple trophic level communities (including bacteria, fungi, and protists).

Response of Microbial Diversity and Community Composition to Pesticide Diversity.

Bacterial richness and the richness of multitrophic communities were significantly (P < 0.05) positively associated with pesticide diversity in the no-N-added treatments, while fungal Shannon and Pielou indices showed significantly (P < 0.05) negative correlations with pesticide diversity in the N-added treatments (SI Appendix, Figs. S4 and S5). These findings aligned with the results from linear mixed-effects models (Fig. 1C). Compared to microbial diversity, community composition was less affected by pesticide diversity in both N-added and no-N-added treatments (SI Appendix, Fig. S6 A and B).

Increased pesticide diversity significantly enhanced network complexity indices (e.g., network complexity represented by the first axis of principal coordinate analysis based on 12 network parameters, connectance, average degree, average clustering coefficient, centralization degree, and centralization closeness) in the no-N-added treatments, as shown by linear mixed-effects models (β = 0.142−0.349, P < 0.05; Fig. 1A). Linear regression analyses further revealed positive correlations between pesticide diversity and both links and modularity (P < 0.05; Fig. 2 A and B and SI Appendix, Fig. S7). In contrast, no significant relationships between network complexity indices and pesticide diversity were observed in the N-added treatments (P > 0.05; Figs. 1A and 2 A and B and SI Appendix, Fig. S7). Additionally, N addition significantly (P < 0.001) increased link constancy and strengthened the positive association between total cohesion, positive cohesion, and network complexity parameters (e.g., network complexity, links, average degree) compared to the no-N-added treatments (SI Appendix, Fig. S8 A–C).

Fig. 2.

Links of microbial network topological parameters with pesticide diversity in N control and N addition treatments. (A) Visualization of the microbial networks for each pesticide diversity level in treatments with no N added (N control) and N addition treatments. The networks within black, blue, and purple circles represent three types of pesticide combination, that is, None, Ⅰ, and Ⅱ, respectively. None indicates no application of pesticides, Ⅰ indicates the application of a single type of pesticide (fungicide or insecticide), and Ⅱ indicates the combined application of fungicide and insecticide. The presence of letters “O,” “C,” “P,” and “M” in the treatment label indicates the use of Omethoate, Chlorpyrifos, Azoxystrobin + Propiconazol, and Mefenoxam + Mancozeb, respectively, and the presence of letter “N” in the treatment label indicates the addition of N fertilizer. “Control” indicates treatment without the application of pesticide and N fertilizer. Details of treatments and pesticide application are described in Dataset S1. The nodes are colored according to microbial taxonomy (bluish violet: bacteria, orange: fungi; green: protists), and node size is weighted according to the edge number connected with others. Detailed information on empirical network topology attributes is listed in Dataset S7. (B) Relationships between network topological parameters (including the number of links and nodes, connectance, average degree, average path length, average clustering coefficient, modularity, and centralization degree) and pesticide diversity in N control and N addition treatments (n = 16). The equations, adjusted R², and P values from the regressions are shown near the regression lines. The error bands shaded in surrounding the regression lines indicate the 95% CI.

Network Motifs as Affected by Pesticide Diversity and Their Impacts on Network Parameters.

Among the seven theoretical three-way associations (i.e., motifs), only five were detected in our data. Among these, transitive competition and facilitation (trancomfac) and transitive facilitation (tranfac) were the most common motifs, with frequencies ranging from 33.4 to 38.8% and 17.4 to 28.3%, respectively (Fig. 3 A and B). Unlike the motifs traffic, transitive competition (trancom), and trancomfac, the overrepresentation values of two network motifs—cycle facilitation (cycfac) and facilitation-mediated competition (facmcom)—were higher than zero (Z > 0), indicating that they were statistically more frequent than expected by chance (P < 0.05; Fig. 3C). In the no-N-added treatments, the cycfac motif generally decreased along the pesticide diversity gradient, while the facmcom motif showed a significant positive correlation with pesticide diversity (P < 0.05; Fig. 3D). In contrast, these strong associations between cycfac, facmcom, and pesticide diversity were not observed in the N addition treatments (P > 0.05; Fig. 3D).

Fig. 3.

Network motifs within the soil micro–food web among different treatments. (A) All possibilities of network motifs among triads of soil microbial taxa. Green dots represent soil microbial genus. Magenta and sky blue links indicate positive and negative associations, whereas gray dashed links indicate neutral associations. (B) Bar plot showing the composition of network motifs among triads of soil taxa detected in 32 networks. (C) Heatmap showing the overrepresentation of network motifs. The importance of network motifs was assessed by the z-score of motifs relative to chance. Dark green (Z > 0, P < 0.05) and brown (Z < 0, P < 0.05) blocks denote motifs that significantly occurred more and less than expected by chance, respectively, while gray blocks (P ≥ 0.05) denote motifs that did not occur differently relative to chance. (D) Relationships between diverse network motifs and pesticide diversity in treatments with no N added (N control) and N addition treatments (n = 16). The significance of each equation was determined using an F-test (***, **, and * indicate P < 0.001, < 0.01, and < 0.05, respectively). The equations, adjusted R², and P values from the regressions are shown near the regression lines. The error bands shaded in surrounding the regression lines indicate the 95% CI. The presence of letters “O,” “C,” “P,” and “M” in the treatment label indicates the use of Omethoate, Chlorpyrifos, Azoxystrobin + Propiconazol, and Mefenoxam + Mancozeb, respectively, and the presence of letter “N” in the treatment label indicates the addition of N fertilizer. “Control” indicates treatment without the application of pesticide and N fertilizer.

Motifs significantly influenced the composition of network parameters, including complexity and stability (SI Appendix, Fig. S9 A–F). Most network complexity parameters showed significant correlation with cycfac (negative) and facmcom (positive) in both N-added and no-N-added treatments (P < 0.05; SI Appendix, Fig. S9G). Additionally, N addition intensified the positive associations between trancomfac and complexity parameters (P < 0.05; SI Appendix, Fig. S9G).

Microbial Assembly Processes, Niche Width, and Specialization as Affected by Pesticide Diversity and Their Impacts on Network Motifs.

The modified stochasticity ratio (MST) of bacterial, fungal, and protistan communities across all 32 treatments ranged from 0.593 to 0.801, 0.585 to 0.823, and 0.543 to 0.787, respectively (SI Appendix, Fig. S10A). In the no-N-added treatments, bacterial MST decreased as pesticide diversity increased, while N addition inhibited the effect of pesticide diversity on bacterial MST (SI Appendix, Figs. S10C and S12A). Similarly, bacterial niche width was strongly negatively associated with pesticide diversity in the no-N-added treatments (P < 0.01), a pattern that did not occur in the N-added treatments (SI Appendix, Figs. S11B and S12A). N addition significantly increased bacterial niche width compared to no-N-added treatments (P < 0.01; SI Appendix, Fig. S11C). Across all treatments, the proportions of generalists were 41.8 to 58.5%, 27.2 to 47.9%, and 29.7 to 47.7%, while the proportions of specialists were 2.7 to 11.1%, 0.5 to 31.6%, and 0.0 to 12.9% in bacterial, fungal, and protistan communities, respectively (SI Appendix, Fig. S13). In the no-N-added treatments, the proportions of bacterial generalists and neutrals were significantly negatively associated with pesticide diversity (P < 0.001), while specialists showed a positive association (P = 0.014, 0.088) (SI Appendix, Figs. S12A and S14 A–C). Compared to no-N-added treatments, N addition mitigated the strong associations between bacterial generalists and neutrals with pesticide diversity and resulted in a positive effect of pesticide diversity on bacterial specialists (P < 0.05; SI Appendix, Figs. S12A and S14 A–C).

The motif facmcom was significantly negatively associated with bacterial MST, niche width, and generalists in the no-N-added treatments, a trend also observed in the N-added treatments (P < 0.05; SI Appendix, Fig. S12B). Additionally, N addition strengthened the positive associations between cycfac and tranfac motifs with bacterial niche width and generalists, as well as between facmcom and bacterial neutrals and specialists (P < 0.05; SI Appendix, Fig. S12B).

Changes in the Functional Traits along with the Pesticide Diversity Increase.

In the no-N-added treatments, 40 genes associated with C, N, P, and S metabolisms showed significant associations with pesticide diversity, which was substantially higher than the 10 related genes in the N-added treatments (P < 0.05; Fig. 4 A and B). Among the C cycling genes, those involved in aerobic respiration (coxA, coxB, coxD) and anaerobic fixation (korB) significantly increased, while a gene related to methanogenesis (fwdA) significantly decreased as pesticide diversity increased in the no-N-added treatments (P < 0.05; Fig. 4C). Compared to the no-N-added treatments, N addition altered the types of genes influenced by pesticide diversity. For instance, genes involved in aerobic fixation (rbcS) and methane oxidation (pmoA, pmoB, pmoC, mmoY) significantly increased with higher pesticide diversity in the N-added treatments (P < 0.05; Fig. 4C). Regarding N cycling, increased pesticide diversity strongly (P < 0.05) enhanced the abundance of genes for nitric oxide reduction (norC), nitrate reduction (narG, narH), and glutamate dehydrogenase (gdha), while reducing the abundance of genes for ammonia oxidation (amoA, amoB) and dissimilatory N reduction (napA) in no-N-added treatments. N addition, however, mitigated the strong associations between these functional genes and pesticide diversity (P > 0.05; Fig. 4D). This mitigating effect of N addition was also evident in P and S cycling (Fig. 4 E and F). For example, genes for acid and alkaline phosphatase (ACP5, phoA, phoD), polysulfide reduction (cysJ), sulfate dissimilatory reduction (aprA), and sulfate assimilatory reduction (cysC, cysD) were significantly affected (P < 0.05) by increasing pesticide diversity in the no-N-added treatments but not in the N-added treatments (Fig. 4 E and F).

Fig. 4.

Responses of functional genes to pesticide diversity. Association between environmental variables and specific genes involved in carbon, nitrogen, phosphorus, and sulfur metabolisms drawn from functional pathways in (A) N control and (B) N addition treatments (n = 48). Node size is weighted according to the edge number connected with others. Differences in the relative abundance of functional genes related to (C) carbon cycling, (D) nitrogen cycling, (E) phosphorus cycling, (F) sulfur metabolisms (n = 48).

Selection of Bacterial Life History Strategies and Their Response to Pesticide Diversity.

Nearly 75% of the total variation in metagenomic CATs was captured in the two dimensions of the multitable coinertia analysis (MCOA), with MCOA1 and MCOA2 accounting for 65.9% and 10.2% of the metagenomic trait variation, respectively (Fig. 5). Average genome size, functional gene diversity, and functional trait stability (including C, N, P, and S metabolisms) were significantly negatively correlated with MCOA1 (P < 0.05, 0.01). In contrast, certain complex metabolic functions (e.g., C and N metabolisms, bacterial and fungal biomass degradation enzymes) and resource acquisition strategies (e.g., uptake system, exoenzymes, motility, and chemotaxis), along with microbial diversity indices (Shannon, richness, evenness, and multitrophic community diversity), were strongly positively associated with MCOA1 (P < 0.05, 0.01) (Fig. 5 and SI Appendix, Figs. S15 A–D and S16A). In the positive direction of MOCA2, genes related to membrane synthesis and repair, sporulation, and osmolyte production decreased significantly (P < 0.01; Fig. 5 and SI Appendix, Fig. S15 A–D).

Fig. 5.

The life history strategies of soil bacterial communities. Two-dimensional trait space from an MCOA describing trait linkages across soil bacterial communities. Dots indicate the positions of the 96 samples used in this study along these two dimensions. In the trait lists, letters in brackets represent how CSR (competitor, stress tolerant, ruderal) and YAS (high yield, resource acquisition, stress tolerance) strategies have been related with these traits in previous theoretical works (Dataset S2).

Soil dissolved organic C (DOC) content and pesticide diversity were the main predictors for MCOA1, while soil TC, TN, and related soil properties were the primary predictors for MCOA2, according to random forest analyses (SI Appendix, Fig. S17 A and B). MCOA1 values (i.e., values of CATs on MCOA1) increased with rising pesticide diversity when soil DOC content was below 90 mg kg⁻¹. However, at DOC levels exceeding 90 mg kg⁻¹, higher MCOA1 values were observed at pesticide diversity levels of 1 and 2 (SI Appendix, Figs. S17C and S18A). Correlation analyses showed that most life history traits were strongly positively correlated with increasing pesticide diversity in the no-N-added treatments and with increasing soil DOC content in the N-added treatments (SI Appendix, Fig. S19). Additionally, MCOA2 values increased with higher TC and TN contents and with the ratio of TC and TN (SI Appendix, Figs. S17D and S18B).

Discussion

Effects of Pesticide Diversity on Microbial Assembly Processes, Niche Width, Specialization, and Potential Interactions.

In the treatments with no added N, the bacterial community exhibited decreasing stochasticity with increasing pesticide diversity. This effect was likely due to the stress that higher pesticide diversity imposed on microbial community members, resulting in environmental filtering (20). In contrast to previously reported findings, where habitat generalists with broader niche widths were shown to have greater environmental tolerance and a survival advantage in high-stress environments (55), we observed that the proportion of bacterial generalists decreased with increased pesticide diversity. This reduction may be because higher pesticide diversity enriched the abundance of bacterial neutrals (i.e., opportunists) and specialists capable of degrading or resisting pesticides (56, 57). In general, generalists are considered less dependent on the presence of other species for growth, indicating that interactions among species are not essential for their survival (54, 58, 59). Compared to generalists, specialists and opportunists with narrower niche widths are expected to rely more heavily on external nutrients or metabolites produced by other microorganisms (54, 58). The increase in the proportion of bacterial specialists and opportunists suggests a narrowing of niche widths, which subsequently strengthens microbial interactions as pesticide diversity increases. This finding aligns with our observation that the complexity of multitrophic level networks increased with higher pesticide diversity. Collectively, these findings support our initial hypothesis.

In the present study, two network motifs with facilitative (positive) associations (cycfac and facmcom), which occurred more frequently than expected across networks, contributed to the complexity of the multitrophic level network. This suggests that direct cross-feeding facilitation or mutualism (e.g., cycfac) and indirect cooperation (e.g., facmcom) within phylotype triads with diverse optimal niches are the dominant ecological strategies forming the fundamental structure of multitrophic level ecological networks in grassland soils (60–62). These triads may also indicate resistance to predation when various taxa co-occur, as described in a previous experimental study (63). Our findings align with a recent global soil study highlighting the predominant role of positive associations (60). Collaborative loops (i.e., cycfac) can create dependency, meaning species face challenges when cooperative partners are lost due to environmental stress (e.g., increased pesticide diversity) (64, 65). This observation corresponds to our finding that motif cycfac decreased with higher pesticide diversity in the no- N-added treatments. The collapse of collaborative loops at higher pesticide diversity provided coexistence opportunities for pairs of microbial species with complementary traits (25, 26), potentially increasing the complexity of multitrophic level networks. These pairs of microbial species with complementary traits may have stronger adaptation to high pesticide diversity than other species (62), which led to an increase in the facmcom motif (i.e., one species is excluded from favorable assemblages of two species) under higher pesticide diversity.

Effects of Pesticide Diversity on the Functional Traits and Soil Nutrient Turnover.

Our MCOA results demonstrate links between microbial life history strategies and pesticide diversity in treatments without N addition. Under high pesticide diversity, bacteria with smaller genome sizes dominated the “metabolic capacity” and “nutrient recycling” dimensions, suggesting that increased pesticide diversity led bacteria to streamline their genomes while enhancing their metabolic activity. Previous studies have shown that maintaining a larger genome entails high costs (maintenance, replication, and regulation) (54), so selection under high pesticide diversity likely favored genome reduction. According to the Black Queen Hypothesis, generalists, as common-good producers or “helpers”, tend to have larger genomes, possibly with expansion (54, 58). Our finding that the proportion of bacterial generalists decreased with rising pesticide diversity supports this selection. Bacterial specialists became more competitive, and their specific metabolic genes likely contributed to overall metabolic capacity (49, 51). Community-level genome streamlining could reduce functional redundancy (51), potentially compromising the stability of functional traits related to C, N, P, and S metabolism. Notably, taxonomic diversity (biodiversity) increased when average genome size and functional trait stability decreased. This effect may stem from bacterial specialists investing energy in degrading and utilizing pesticides, promoting the coexistence of other species across multiple trophic levels and thereby increasing taxonomic diversity (51, 54).

Pesticide diversity negatively correlated with soil TC and TN contents, and SRP, consistent with life history strategies suggesting that high pesticide diversity enhances soil microbial “metabolic capacity” and “nutrient recycling.” This trend was linked to the stimulation of cycling genes involved in aerobic respiration (coxA, coxB, coxC) and glutamate dehydrogenase (gdha) in the no-N-added treatments. Our results emphasize accelerated soil nutrients (particularly recalcitrant substrates) consumption under high pesticide diversity in treatments without N addition, supporting our second hypothesis. Additionally, denitrification-related genes (narG, narH, and norC) increased with rising pesticide diversity in no-N-added treatments, potentially reducing NO₃⁻-N accumulation. Denitrification is a key source of the greenhouse gas N₂O emitted from soils (66, 67). Our findings thus indicate a greater potential for N₂O production in the no-N-added treatments under higher pesticide diversity. Overall, these results suggest that high pesticide diversity may enhance soil resource processing by microorganisms, potentially reducing long-term C and N storage and increasing ecological risks, such as greenhouse gas emissions, in no-N-added treatments. Given that the four pesticide types studied have minimal impact on nontarget organisms (68–72), using highly toxic, broad-spectrum pesticides could substantially impair soil microbial functions. As the use of such pesticides increases, the negative effects of pesticide diversity on soil microbial functions could intensify. Unlike grassland soils, frequent agronomic practices (e.g., tillage) in farmland can disrupt soil aggregate structure and expose organic matter (73), which may facilitate soil organic matter utilization and decomposition by microorganisms (74, 75). Therefore, higher soil nutrient consumption may be observed in farmland soils under elevated pesticide diversity. A comprehensive assessment of pesticides’ effects on soil microbial functions in agricultural production should account for pesticide diversity rather than concentration alone.

N Addition Alleviates the Effects of Pesticide Diversity on the Functional Traits and Soil Nutrient Turnover.

Our results indicated that N addition significantly increased the content of inorganic N (NH₄⁺-N and NO₃⁻-N), consequently improving soil available N levels compared to treatments without N addition. The bacterial niche width in treatments with N addition was also significantly greater than in the no-N-added treatments (P < 0.01), suggesting that enhanced soil nutrients can effectively reduce niche differentiation within the bacterial community. Microbes with broader niche widths may be more competitive in utilizing soil nutrient resources to withstand environmental stress (76), consistent with our finding that bacterial niche width did not significantly decrease under higher pesticide diversity in N addition treatments. We observed that microbial networks were more stable (i.e., had higher link constancy) in N addition treatments than in no-N-added treatments (P < 0.001). Higher link constancy reflects microbial interactions with increased environmental tolerance (20, 25, 77, 78). As a result, network complexity in N addition treatments did not respond to pesticide diversity as it did in the no-N-added treatments. In the N-added treatments, DOC rather than pesticide diversity was the main driver influencing bacterial life history strategies. One possible explanation is that in the presence of ample N, soil microorganisms utilized available organic C to degrade and assimilate low C/N ratio pesticides (79, 80). Our study showed that N addition specifically promoted N cycling genes—such as those for ammonification (ureC), nitrification (hao), and nitrate assimilation reduction (nasA)—while having less effect on C cycling genes, indicating a higher potential for N turnover. This partially aligns with previous studies showing that N addition significantly affects both C and N cycling genes (42, 43), potentially because N addition meets microbial N demands more than C demands in nutrient-limited grassland soils (43, 52). Unlike in the no N treatments, soil nutrients (e.g., TC, TN, NO₃⁻-N, NH₄⁺-N, SRP) did not decline, or declined more slowly, with increased pesticide diversity in the N addition treatments, suggesting that N addition mitigated soil resource depletion under high pesticide diversity. This mitigation may be due to N addition enhancing plant growth and increasing photosynthetic allocation to roots (81). Associated mycorrhizal fungi facilitate soil macroaggregate formation through fine root and hyphal enmeshment of microaggregates (82), which may restrict soil nutrient consumption under high pesticide diversity stress. This observation aligns with our finding that fewer genes involved in C, N, P, and S cycling were affected by pesticide diversity in N addition treatments than in no-N-added treatments. Collectively, these results support our third hypothesis, indicating that N inputs can mitigate the adverse effects of pesticide diversity on soil functions.

In contrast to grassland soils, which are often limited by P, farmland soils receive P inputs from synthetic fertilizers and manure (83, 84). Consequently, genes involved in P metabolism may exhibit less sensitivity to pesticide diversity stress in farmland soils than in grasslands, given the higher available P supply in farmlands. Considering significant differences between these land-use types (e.g., nutrient status, plant types), further studies are needed to assess the ecological effects of high pesticide diversity in farmland soils or in scenarios where other nutrients (e.g., P) are supplemented. Compared to slow-release urea and neutral N fertilizers (e.g., KNO₃), acidic N fertilizers [e.g., NH₄NO₃ and (NH₄)₂SO₄] exert a stronger acidifying effect on soils (85, 86). Acidification can shift bacterial functions by excluding certain intolerant bacterial groups (87, 88), though the impact of this shift under pesticide diversity stress remains understudied and warrants further investigation. In conclusion, our findings highlight the potential of nutrient management to mitigate pesticide diversity stress, thereby supporting soil health and the sustainable development of agriculture.

Conclusion

Overall, our study provides a perspective on the effects of pesticides on soil microbial functions under different N fertilizer management practices. The results indicate that increasing pesticide diversity can have a strongly adverse impact on soil microbial functions, intensifying microbial consumption of soil resources. This finding suggests that pesticide diversity acts as an environmental stressor, driving substantial changes in the genomic traits of microbial communities and altering nutrient cycling in agricultural soils. However, these adverse impacts of pesticide diversity can be mitigated by N fertilizer input. The insights gained here enhance our understanding of the role of pesticide diversity in biogeochemical cycling and soil environmental health. Our study further suggests that effective N fertilizer management can help mitigate the negative effects of pesticide use, promoting the sustainable development of agriculture.

Materials and Methods

Site Description and Experimental Schemes.

To avoid the influence of pesticide residues commonly present in most arable soils, steppe grassland with minimal pesticide pollution was chosen for the field experiment. The study site was located in a steppe grassland near the Erguna Forest-Steppe Ecotone Ecosystem Research Station (50°10′N, 119°22′E, Northeast China), an area free from human interference, beginning in 2020. The experimental site experiences a mid-temperate, semiarid continental grassland climate with an average annual temperature of −2.4 °C and average annual precipitation of 360 mm. The soil type is Chernozem, and the dominant vegetation includes Leymus chinensis and Stipa baicalensis. However, with potential agricultural expansion in the future, this region may also face pesticide exposure. Urea is the most widely used N fertilizer in global agricultural production (89), and slow-release urea is particularly recommended as it provides a gradual N supply to crops, reducing N loss and mitigating greenhouse gas emissions (90–92). The recommended N application rate for the primary crops in Northeast China—wheat, maize, and soybean is approximately 100 kg N ha⁻¹ (93–96). Accordingly, N was applied as slow-release urea at two levels: a control with no N addition (N control, 0 kg N ha⁻¹ y⁻¹) and a treatment with N addition (N addition, 100 kg N ha⁻¹ y⁻¹).

Four types of commercial pesticides, widely used in agricultural and natural ecosystems globally (68–72), were selected for this study. The active ingredients in these pesticides were omethoate (O), chlorpyrifos (C), azoxystrobin + propiconazole (P), and mefenoxam + mancozeb (M). Combined applications of these pesticides are common due to their varied target pests and pathogens. Pesticides were applied at two levels: a control with no pesticide application and a conventional application rate. Omethoate (C₅H₁₂NO₄PS) and chlorpyrifos (C₉H₁₁C_l3NO₃PS) target aboveground and underground insects, respectively (71, 72). Both are organophosphorus pesticides that inhibit acetylcholinesterase activity in insects, leading to nervous system dysfunction and death (71, 72). The compound pesticides azoxystrobin + propiconazole, and mefenoxam + mancozeb serve as aboveground (foliar) and underground (soil) fungicides, respectively (69, 70). Azoxystrobin (C₂₂H₁₇N₃O₅) and propiconazol (C₁₅H₁₇Cl₂N₃O₂) restrict microbial growth by inhibiting mitochondrial respiration and interfering with cell membrane synthesis and function. Mefenoxam (C₁₅H₂₁NO₄) and mancozeb ([C₄H₆MnN₂S₄]_xZn_y, x: y = 1: 0.091) exert biological inhibition by disrupting ribosomal RNA1 and protein synthesis in fungal mycelium and by damaging cell walls and membranes through the actions of zinc and manganese ions, respectively.

In summary, the study included 32 treatments arranged in a completely randomized block design with 10 replicates (i.e., 10 plots), totaling 320 quadrats. The treatments included five factors: N fertilizer (N) and four types of commercial pesticides. The 32 treatments were labeled as “Control,” “N,” “O,” “C,” “P,” “M,” “NO,” “NC,” “NP,” “NM,” “OC,” “OP,” “OM,” “CP,” “CM,” “PM,” “NOC,” “NOP,” “NOM,” “NCP,” “NCM,” “NPM,” “OCP,” “OCM,” “OPM,” “CPM,” “NOCP,” “NOCM,” “NOPM,” “NCPM,” “OCPM,” and “NOCPM.” In each label, “O,” “C,” “P,” and “M” represent omethoate, chlorpyrifos, azoxystrobin + propiconazol, and mefenoxam + mancozeb, respectively, while “N” denotes the addition of N fertilizer. “Control” indicates a treatment with no pesticide or N fertilizer application. Pesticide diversity within each treatment was defined by the number of commercial pesticides used, ranging from 0 to 4. Each quadrat measured 4 m² (2 m × 2 m), with 1 m separating adjacent quadrats within a plot and 2 m separating adjacent plots. Pesticides were applied close to the ground on calm, sunny days to minimize cross-contamination. Detailed treatment and application information can be found in Dataset S1.

Soil Sampling and Measurement.

Soil samples were collected in September 2022, after 3 y of treatment exposure. In each quadrat, a composite soil sample was created from five randomly collected soil cores (10 cm in diameter) taken from the top 0 to 20 cm after removing stones and plant debris. All samples were stored at −80 °C prior to DNA extraction. Subsamples were air-dried and sieved to analyze soil properties, including pH, EC, soil organic matter (SOM), TC, TN, total P (TP), DOC, ammonium N (NH₄⁺-N), nitrate N (NO₃⁻-N), and available P (AP). These properties were individually standardized using the equation (97):

$${\mathit{SP}}_{\mathit{stand}}=\frac{\mathit{SP}-{\mathit{SP}}_{\mathit{min}}}{{\mathit{SP}}_{\mathit{max}}-{\mathit{SP}}_{\mathit{min}}},$$ [1]

Where SP and SP_stand are the original and standardized values, respectively, and SP_min and SP_max are the minimum and maximum values across all samples. SRP were calculated as the average of standardized values for specific properties, including soil organic C (SOC), TN, TP, the ratio of ammonium and nitrate N to TN, and the ratio of AP to TP, representing soil resource availability (97). Overall soil properties were calculated as the average of all standardized soil properties, including pH, EC, SOM, TC, TN, TP, DOC, NH₄⁺-N, NO₃⁻-N, AP, and the ratio of TC to TN (TC/TN). Further details on detection methods are available in SI Appendix. Soil property data are displayed in SI Appendix, Figs. S1–S3.

Amplicon Sequencing and Metagenomic Data.

The extracted DNA was sent to Guangdong Magigene Biotechnology Co., Ltd. (Guangdong, China) for high-throughput sequencing of the 16S and 18S rRNA amplicons, using the 515F/806R and Euk1391f/EukBr primers, respectively. DNA extraction, sequencing, trimming, and mapping procedures followed methods outlined in previous studies (49, 98), with further details available in SI Appendix, Supplementary Methods (2).

Microbial Community Assembly Processes.

The MST was used to evaluate the relative influence of stochastic processes in community assembly (58, 99). The MST index ranges from 0 to 1, with 0.5 serving as the threshold to determine whether community assembly is more deterministic (<0.5) or more stochastic (>0.5) (58, 99). Levins’ niche width was calculated to assess species taxonomy and dispersal limitation patterns and their effects on microbial communities across treatments (76). Additionally, we categorized generalist, neutral (i.e., opportunist), and specialist groups based on prior studies (58, 100). The MST, niche width, and ecotypes for all treatments across bacterial, fungal, and protistan communities were computed using the “NST,” “spaa,” and “EcolUtils” packages, respectively.

Network Construction and Structure Characterization.

Co-occurrence networks, constructed at the genus level, were used to examine associations among bacterial, fungal, and protistan communities across all 32 treatments. Networks were developed in R (version 4.2.3) using the “Hmisc” and “igraph” packages with Spearman correlation (|r| ≥ 0.60; Benjamini–Hochberg adjusted P < 0.05) (97, 98). Genera present in at least 20% of samples were included to ensure network reliability (101). Gephi software (version 0.10.1) was used to visualize networks with the Fruchterman–Reingold layout. Network motifs, composed of triads of microbial genera, are considered the fundamental building blocks of complex networks (60, 61). The frequency of these motifs was calculated to provide insight into the structure of soil microbial networks (60). Additional details on network parameters are available in SI Appendix, Supplementary Methods (3).

Bacterial Community Aggregated Trait Calculation.

In this study, four CAT databases (life history traits, eggNOG categories, CAZy types, and KEGG pathways) were created according to prescribed protocol (49). The life history traits database included 26 traits, such as chaperones, carbohydrate-degrading enzymes, antibiotic resistance, chemotaxis, sigma factor, osmolytes, and others (Dataset S2). The average genome size and rRNA gene copy number were calculated using MicrobeCensus (102) and the method described in ref. 103. Data for the four CAT databases are provided in Datasets S3−S6.

Statistical Analysis.

To assess the effects of pesticide diversity on soil properties, microbial diversity, and network parameters, linear mixed-effects models were applied within each of the two N input levels (with and without N addition) using the R package “lme4” (104). The regression coefficients (β) for each model were extracted to represent effect sizes across treatments. To identify the multivariate axes that best explain variations in soil bacterial metagenomic CATs, an MCOA was conducted using the “ade4” package in R (version 4.2.3), based on the four databases (life history traits, eggNOG, CAZy, KEGG) (49). Random forest models were then used to identify potential environmental drivers impacting the two MCOA coordinates (105). Further details on analytical methods are provided in SI Appendix, Supplementary Methods (6).

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Dataset S05 (XLSX)

Dataset S06 (XLSX)

Dataset S07 (XLSX)

Data, Materials, and Software Availability

The 16S and 18S rRNA gene sequencing reads generated in this study have been deposited in the National Center for Biotechnology Information (NCBI) under accession number PRJNA1158439 (106). All other data are included in the manuscript and/or supporting information.