A fertilization experiment with long-term monitoring was set up in 2014 within a Phyllostachys edulis (Moso bamboo) forest stand in Fujian Province's Daiyun Mountain National Nature Reserve (25°38′7″–25°43′40″N, 118°5′22″–118°20′15″E), located in Dehua County. The site’s soil is classified as Ultisols under the USDA Soil Taxonomy system. The subtropical monsoon climate of the region features 15.6–19.5°C average yearly temperatures and receives 1700–2000 mm precipitation annually (Zeng et al. 2024). The experiment comprised three N treatments delivered as ammonium nitrate: control (CT; ambient N deposition), low-N addition (LN, 20 kg N ha^− 1 yr^− 1), and high-N addition (HN; 80 kg N ha^− 1 yr^− 1). The experiment followed a completely randomized design with nine plots (each 3 m × 10 m), including three replicates per treatment, separated by 2-m buffer zones. Annual applications of 20 L ammonium nitrate solutions (prepared with Milli-Q water) were administered during the March-September growth period.

Soil sampling was conducted annually in July from 2017–2019. From each plot, five randomly selected soil cores (0–15 cm depth) were combined to create homogenized composite samples. After manually removing gravel, plant roots, and other debris, each composite sample was subdivided into three portions for subsequent analyses: 1) chemical analysis after 2 mm sieving of air-dried samples; 2) DOM characterization from 4°C-stored portions; and 3) microbial community analysis via DNA extraction and high-throughput sequencing of − 80°C-stored portions.

Statistical analysis employing SPSS 27 assessed N treatment effects via repeated-measures ANOVA and Duncan's multiple comparisons (p < 0.05), examining soil chemical properties, DOM traits, and microbial communities. Microbial α-diversity (Shannon index) and β-diversity (Bray-Curtis dissimilarity) were analyzed with the 'vegan' package based on ASV tables. Partitioning of β-diversity into three constituent elements was accomplished using the beta.div.comp method from the 'adespatial' package (Baselga 2010; Shen et al. 2020). Microbial life-history strategies (i.e., K- and r-strategists) were inferred using two complementary approaches. First, taxa were classified at the phylum level (Table S3). Second, bacterial life-history strategies were identified based on copy numbers of the ribosomal RNA operon (rrn) linked to each ASV, following methods outlined in prior research (Chen et al. 2022; Gao et al. 2025). Bacterial populations exhibiting lower ribosomal RNA operon (rrn) copy numbers are generally indicative of K-selected dominance (Dai et al. 2022; Roller et al. 2016). Associations between DOM traits and microbial community composition were examined using the “ggcor” package through partial Mantel tests (9,999 permutations). Network-based approaches enabled exploration of possible relationships between bacterial DOM components and amplicon sequence variants (ASVs). Bacterial and fungal ASVs were filtered based on relative abundance thresholds (> 0.05%) to streamline downstream analyses. Spearman correlation analysis of DOM-ASVs relationships was performed employing the 'psych' package. According to Li et al. (2019), microbial association networks were constructed based on statistically significant correlations (p < 0.05) with coefficient thresholds set at |r| >0.8, with |r| >0.60 considered biologically relevant in ecological studies. The co-occurrence network was visualized in Gephi 0.10.1, where nodes corresponded to either microbial ASVs or DOM components, connected by edges denoting statistically correlations. The 'igraph' package was employed to analyze key network topological parameters of DOM-microbe associations, quantifying node and edge counts, average degree, average path length, and others (Csardi and Nepusz 2006). These network parameters served as indicators of structural complexity, consistent with established ecological network analysis frameworks (Li et al. 2024). The 'rfPermute' package was employed to build random forest models for identifying crucial drivers of DOM-microbe network complexity. To assess monotonic association magnitude and direction, Spearman's rank correlation was applied to statistically significant predictors (p < 0.05). All packages are run in R 4.1.1. The final polishing of all data figures was carried out in Adobe Illustrator 2025.

N addition over three consecutive years significantly increased soil NO₃⁻-N content, particularly under high-N treatment in 2017 and 2019 (p < 0.05). In contrast, N addition produced no significant alterations in pH, SOC, total phosphorus (TP), total nitrogen (TN), or NH₄⁺-N concentrations (Table S1). DOC content decreased significantly with N addition, with no interaction with time (p < 0.05; Fig. 1a). Neither DON content nor DOC:DON ratio showed significant responses to N input (p > 0.05; Fig. 1b, c), although the three-year mean DOC:DON ratio declined significantly over time under high-N addition (p < 0.05). Among DOM optical characteristics, only specific UV absorbance at 254 nm (SUV₂₅₄) responded significantly to N addition, showing increased three-year means under high-N input (p < 0.05; Table S2). Other indices, including SUV₂₆₀, HIX, FI, BIX, and the HIX:FI ratio, were not significantly affected (p > 0.05). Parallel factor (PARAFAC) analysis identified three fluorescent components: C1 (humic-like substances; Ex = 75/320 nm, Em = 465 nm), C2 (fulvic-like substances; Ex = 245 nm, Em = 425 nm), and C3 (aromatic protein-like substances; Ex = 225 nm, Em = 332 nm) (Fig. 1g–i). While neither C1 nor C3 proportions were significantly affected by N addition (p > 0.05; Fig. 1d, f), a marked increase in the three-year mean proportion of C2 was observed specifically under low-N conditions (p < 0.05; Fig. 1e).

To identify the factors influencing the interactions between DOM components and microbial communities under three consecutive years of N addition, we assessed network topological parameters (e.g., node and edge numbers) using random forest models (Fig. 6a, b). The complexity of DOM-bacteria networks was primarily predicted by C1 proportion, rrn copy numbers, and the relative abundance of Actinobacteria. In contrast, NO₃⁻-N concentration, DOC:DON ratio, and MBC were the dominant factors shaping DOM-fungi network complexity. Subsequent linear regression analyses confirmed significant associations between these predictors and network complexity (p < 0.05). Specifically, DOM-bacteria network complexity significantly correlated positively with C1 proportion (R² = 0.14, p < 0.05; Fig. 6c) and rrn copy numbers (R² = 0.30, p < 0.05; Fig. 6d). In contrast to a negative correlation with DOC:DON ratio (R² = 0.10, p < 0.05; Fig. 6f), a positive correlation was observed between DOM-fungi network complexity and NO₃⁻-N (R² = 0.42, p < 0.001; Fig. 6e) as well as MBC (R² = 0.20, p < 0.05; Fig. 6g).

Contrary to well-documented patterns of DOC accumulation following N input in most ecosystems (Ngaba Junior et al. 2022; Tian et al. 2022; Xu et al. 2021), our results demonstrated significant decreases in soil DOC content following three consecutive years of N addition (Fig. 1a). This finding aligns with a recent meta-analysis showing asymmetric DOC responses to N addition between humid and non-humid subtropical forests, where N addition decreasing DOC content in humid regions while increasing it in non-humid ones (Ren et al. 2025). Our study site, located in a relatively N-rich humid subtropical forest, supports the hypothesis proposed by Ren et al. (2025). This reduction in DOC likely stems from two key processes. First, in N-rich systems, N addition often reduces or stabilizes C inputs by suppressing net primary productivity, while simultaneously lowering C outputs through inhibition of soil respiration, jointly resulting in reduced DOC availability. Second, elevated microbial activity (evidenced by increased MBC, Fig. S1a) may enhance microbial uptake and utilization of DOC, further depleting its pool. Despite substantial DOC content changes, N addition had limited effects on DOM quality. Solely the low-N addition increased the proportion of component C2 (fulvic acid-like substances; Fig. 1e, h). This increase probably results from fulvic acids' enhanced mobility and reactivity, which are driven by their oxygenated functional groups, low molecular weight, and high solubility (Hayes 1983; Xue et al. 2024). Moderate N input may also facilitate the activation of soil humus, thereby influencing organic C transformation and stability. Together, these findings provide novel perspectives on soil C dynamics in humid subtropical forests affected by N deposition. Future research incorporating long-term experiments is warranted to elucidate the temporal trajectories and broader implications for soil C cycling.

Bacterial communities exhibited greater sensitivity to N addition than fungal communities, as evidenced by significant shifts in taxonomic composition and life-history strategies (Fig. 2a and 3). This differential sensitivity probably stems from the higher N requirements and metabolic adaptability of bacteria in subtropical soils, enabling quicker adaptation to nutrient perturbation (Cao et al. 2023). Contrary to the documented stimulation of copiotrophic bacterial growth under N enrichment (Liu et al. 2020; Wang et al. 2024), our findings revealed a transition toward oligotrophic predominance. This shift was characterized by declining relative abundances of r-selected taxa (e.g., Proteobacteria, Actinobacteria) and reduced rrn copy numbers (Fig. 2a and 3g). This apparent discrepancy may be explained by the high baseline N availability in the study region (Liu et al. 2024b; Zhang et al. 2025), where additional N input decreased soil C:N ratio (e.g., lower DOC: DON ratio), potentially restructuring competitive dynamics among bacterial taxa. Under these conditions, r-strategists reliant on labile C sources likely faced C limitation, whereas K-strategists may have obtained a competitive advantage. Despite these compositional changes, neither bacterial nor fungal α- or β-diversity showed significant responses to N addition during the three-year study (Fig. 2b, c). The overall patterns of β-diversity were influenced by both species turnover and differences in species richness (Fig. 2d), consistent with findings from other N deposition researchs (Shen et al. 2020; Yuan et al. 2025).

The taxonomic composition, biodiversity patterns, and life-history strategies of bacterial and fungal communities all showed strong associations with DOM characteristics. Notably, bacterial community composition exhibited stronger correlations with DOM traits than did fungal community composition (Fig. 4), supporting our first hypothesis. Bacterial taxa exhibited significantly stronger associations with DOM components relative to fungal taxa (Fig. 5). These differences likely arise from contrasting metabolic strategies. Dominant bacterial phyla (e.g., Proteobacteria, Actinobacteria) preferentially metabolize labile C sources (e.g., simple lipids, carbohydrates, and proteins) (Liu et al. 2019) and thrive under nutrient-rich conditions (Bergmann et al. 2011; Pascault et al. 2013). Although their relative abundance declined, the increased prevalence of other bacterial taxa may have compensated through alternative metabolic pathways, thereby preserving the community’s functional capacity for DOM utilization. Conversely, dominant fungal taxa (e.g., Ascomycota, Basidiomycota) specialize in decomposing recalcitrant DOM compounds with complex structures, slow degradation rates, and specific substrate requirements (Zhang et al. 2022). As a result, bacterial communities formed more intricate and dynamic interaction networks with DOM than fungal communities. These results offer essential insights into the distinct patterns of DOM-microbe interactions and the fundamental ecological strategies differentiating bacterial and fungal responses to N enrichment.

DOM-bacteria and DOM-fungi interactions exhibited significantly opposite responses under N addition, with DOM-bacteria interactions demonstrating a pronounced attenuation (Fig. 5), thereby supporting hypothesis 2. Specifically, under low-N addition, DOM-bacteria interactions were notably attenuated (Fig. 5c), primarily due to shifts in DOM traits and bacterial life-history strategies. Humic-like components (C1) and bacterial rrn copy numbers emerged as primary determinants of DOM-bacteria network complexity, demonstrating positive correlations with DOM-bacteria association (Fig. 6a, c, d). Low-N treatment did not significantly alter the proportion of humic-like substances, leading to reduced energy availability and less frequent interactions between DOM and bacterial communities. Concurrently, N-induced shifts in microbial metabolic strategies likely influenced bacterial C sources utilization (Yang et al. 2024; Zhang et al. 2024). Specifically, the observed transition from copiotrophic to oligotrophic strategies following N addition may decrease bacterial metabolic activity, thereby further weakening DOM-bacteria interactions. Additionally, the increase in negative links within the DOM-bacteria network structure under N addition (Fig. 5a, c, e) indicated intensified competition or antagonism between microbial taxa and resources, further reducing positive associations.

In contrast, high-N addition significantly enhanced interactions between DOM components and fungal communities (Fig. 5f), likely due to the increased susceptibility of fungal communities to variations in available energy and nutrients. Elevated NO₃⁻-N concentrations and improved C substrate quality (i.e., decreased DOC:DON ratio; Table S1, Fig. 1c) under high-N input augmented energy supply, facilitating fungal decomposition and DOM assimilation. Notably, fungal taxa such as Ascomycota and Basidiomycota possess extensive extracellular enzyme systems capable of efficiently degrading complex organic matter. Under conditions of reduced C:N ratio, the metabolic potential of these taxa is activated (Baldrian 2009; Lindahl and Tunlid 2015). Furthermore, high-N addition increased MBC (Fig. S1a), reflecting enhanced microbial metabolic activity. Compared to bacterial communities, fungal communities responded more strongly to changes in resource quality, consequently developing stronger interactions with DOM components (Treseder 2008). This pattern aligns with fungi's ecological strategy of preferentially utilizing complex organic matter in nutrient- and energy-rich environments (Wang and Kuzyakov 2024).

Crucially, alterations in DOM-microbe interactions showed no significant correlation with microbial community diversity. Within the EDTiA framework, this implies that variation in energy input and the functional traits of both DOM and microbes, rather than microbial diversity, primarily govern DOM-microbe interactions under N addition. While microbial community diversity contributes to functional potential, the pronounced shifts in resource availability induced by N addition may concentrate metabolic activity among a limited number of functionally dominant taxa, thereby diminishing the role of diversity in shaping overall functional output (Hu et al. 2022b). Moreover, increased functional redundancy within microbial communities under N addition, where different taxa perform overlapping functions, may further obscure the influence of diversity on DOM-microbe associations. Additionally, we need to acknowledge that we did not incorporate DOM molecular diversity into our discussion. Future research will focus on the impact of DOM molecular diversity on DOM-microbe interactions to more comprehensively reveal the underlying mechanisms. Collectively, when interpreted through the EDTiA framework, these findings highlight distinct response patterns of bacterial and fungal communities to DOM components under varying N addition regimes underscoring the pivotal role of energy supply and DOM/microbial traits in mediating DOM-microbe interactions.