The study was conducted in the dominant distribution range of Pinus armandii plantations in Bijie City (26°21’N-27°46’N, 103°36’E-106°43’E), Guizhou Province, China. This region is characterized by a subtropical humid monsoon climate, with altitudes ranging from 457 to 2910.3 m, average annual precipitation between 849 and 1,399 mm, and average temperatures ranging from 10°C to 15°C.

We investigated mono-specific P. armandii plantations at three successional stages (young, middle-aged, mature) with homogeneous geomorphic conditions in August 2021 during peak vegetation growth. A nested sampling design was implemented with three replicate 20 × 20 m plots per stand age. Plots were intentionally spaced ≥150 m apart to minimize spatial autocorrelation and ensure independence (Mariotte et al., 1997). Plot selection adhered to national forest resource survey protocols for age classification (GB/T 26424–2010). Within each plot, all trees with DBH > 3 cm (diameter at breast height, 1.3 m) were measured for dendrometric parameters (height, DBH, crown width). Understory vegetation was quantified through five randomly positioned shrub (2 × 2 m) and herbaceous (1 × 1 m) subplots per main plot, recording species composition and structural parameters (density, coverage, vertical stratification).

Rhizosphere and bulk soils were systematically collected using compartment-specific techniques with explicit replication at multiple spatial scales to ensure representativeness and analytical robustness.

Soil pH was determined potentiometrically in 1:2.5 (w/v) soil-water suspensions using a calibrated pH meter (InsMark^™ IS126, Shanghai, China). Total carbon (TC) quantification employed high-temperature combustion with an Elementar TOC analyzer (Hanau, Germany). Soil organic carbon (SOC) was quantified via dichromate oxidation-external heating (Walkley-Black modified method). Nitrogen fractions were analyzed through: (1) total nitrogen (TN) by micro-Kjeldahl digestion, (2) available nitrogen (AN) via alkaline diffusion. Potassium speciation included: (1) total potassium (TK) by atomic absorption flame photometry, (2) available potassium (AK) through ammonium acetate extraction. Phosphorus fractions were determined calorimetrically using molybdenum antimony anti-colorimetric method for both total (TP) and available phosphorus (AP). All analyses followed standardized procedures (Bao, 2000).

Total DNA was extracted from both bulk soil and rhizosphere samples using the E.Z.N.A^™ Mag-Bind Soil DNA Kit (OMEGA) according to the manufacturer’s instructions. Hypervariable regions were amplified with proofreading polymerase: bacterial 16S rRNA V3-V4 using 515F/806R primers and fungal ITS1-5F with ITS5-1737F/ITS2-2043R. PCR reactions (25 μL) contained: 2 μL template DNA, 5X buffer, 2.5 mM dNTPs, 0.25 μL HiFi polymerase, and 1 μL primers. Thermal cycling conditions:

Purified amplicons (Qiagen Gel Extraction Kit) were sequenced on Illumina NovaSeq (2 × 250 bp) at Novogene (Beijing).

Raw sequences were demultiplexed in QIIME2 (2020.06) and processed through DADA2 (v1.8) for quality filtering, paired-end merging, and ASV clustering. Taxonomic assignment used SILVA v132 (bacteria) and UNITE v8.0 (fungi) databases with 97% similarity thresholds. Rarefaction to 1,067,590 bacterial and 1,095,406 fungal sequences ensured uniform analysis. Data are deposited in NCBI SRA (PRJNA1063200, PRJNA1132892).

To elucidate microbial interaction patterns across the chronosequence and soil compartments, bacterial and fungal co-occurrence networks were generated through Sparse Correlations for Compositional data (SparCC). Spearman’s rank correlation analysis of bacterial and fungal ASVs abundance tables yielded correlation coefficients (R) and significance values (P) (Barberán et al., 2012). Network topology was defined using adjacency matrices derived from correlation coefficients, with edge significance thresholds established through random matrix theory (RMT). Final network visualizations were rendered using Gephi v0.9.2.

To quantify assembly processes, we applied null model-based phylogenetic β-diversity analysis using the picante R package. The β-Nearest Taxon Index (β-NTI) values were derived from 999 randomizations of the observed ASV table and phylogenetic tree. |β-NTI| > 2 indicated deterministic processes (heterogeneous selection for β-NTI > 2; homogeneous selection for β-NTI < −2), while |β-NTI| < 2 indicated stochastic processes (Stegen et al., 2012). For stochastic processes, the contribution of dispersal limitation (RCBray > 0.95), homogenizing dispersal (RCBray < −0.95), and ecological drift (|RCBray| < 0.95) was further quantified using the RCBray value (Stegen et al., 2013).

Alpha diversity indices (Chao1 and Shannon) were calculated using the sequencing company’s analysis platform. Functional classifications of microbial communities were predicted using PICRUSt for bacteria and FUNGuild for fungi (Langille et al., 2013). Parametric comparisons were conducted through one-way ANOVA with post-hoc LSD tests for multi-group comparisons, supplemented by independent t-tests for pairwise chronosequence and soil compartment analyses. Microbial community structure was ordinated via principal coordinate analysis (PCoA) with Bray–Curtis dissimilarity matrices. Mantel tests (MT) quantified correlations between soil physicochemical parameters and alpha diversity metrics. Taxon-environment relationships were elucidated through redundancy analysis (RDA) implemented in Canoco 5.0. All statistical procedures were executed in SPSS 23 (IBM Corp.) and R v3.5.3, with significance thresholds defined at p < 0.05.

Alpha diversity (Shannon and Chao1 indices) varied significantly across stand age and soil niches (Figure 1). In middle-aged and mature forests, rhizosphere soil exhibited lower bacterial and fungal alpha diversity than bulk soil. However, in young forests, rhizosphere soil showed lower bacterial but higher fungal alpha diversity. Overall, significant differences in alpha diversity were observed across the chronosequence for both soil compartments (p < 0.05), generally showing an initial decline followed by an increase with increasing stand age, except for fungal Chao1 diversity in bulk soil.

Beta diversity analysis using Principal Coordinates Analysis (PCoA) revealed distinct bacterial community structures across stand age in both soil compartments (Figure 2A). Bulk soil showed more pronounced changes (p < 0.001; R² = 0.66) than rhizosphere soil (p < 0.001; R² = 0.59). Fungal community structure varied spatially; bulk soil communities separated along the second principal coordinate, while rhizosphere communities separated along the first (Figure 2B). PCoA based on Bray–Curtis similarity confirmed significant differences in fungal community composition across stand age in both bulk (p < 0.001, R² = 0.67) and rhizosphere (p < 0.001, R² = 0.64) soils.

Taxonomic analysis at the phylum level revealed consistent dominant bacterial phyla (Proteobacteria, Acidobacteria, Actinobacteria, and Chloroflexi) across soil compartments, accounting for approximately 85% of the total relative abundance (Figure 3A). However, their relative abundances varied significantly across stand age and soil niches (Figures 4A,B; p < 0.05). In bulk soil, Proteobacteria decreased while Chloroflexi increased in middle-aged forests. Conversely, Acidobacteria and Actinobacteria showed a gradual increase across the chronosequence. In rhizosphere soils, Acidobacteria and Chloroflexi peaked in the middle-aged forest, while Proteobacteria and Actinobacteria displayed a declining trend. While Proteobacteria were more abundant in bulk soil and Chloroflexi in rhizosphere soil, these differences were not statistically significant (p > 0.05).

Similarly, the dominant fungal phyla (Ascomycota, Basidiomycota, and Mortierellomycota) comprised approximately 97% of the total relative abundance (Figure 3B). Their relative abundances varied across stand age and soil niches (Figures 4C,D), although significant differences between rhizosphere and bulk soil were not observed (p > 0.05). Basidiomycota abundance was lowest in middle-aged rhizosphere soil, while Ascomycota peaked at this stage. The opposite trend was observed in bulk soil. Mortierellomycota abundance increased consistently across the chronosequence.

Network analysis revealed significant differences in microbial co-occurrence patterns between rhizosphere and bulk soils across all stand ages (Figures 5A,C; Supplementary Table S1). Bacterial networks in rhizosphere soil were larger and more connected than those in bulk soil, with more nodes, a higher proportion of positive edges and greater average connectivity. However, rhizosphere networks had lower diameter, modularity, and average path length. No network hubs were detected in either soil type, although 10 and 12 module hubs were observed in rhizosphere and bulk soil networks, respectively. Keystone taxa were identified in both soil types, with Proteobacteria and Actinobacteria consistently prominent. The top seven keystone species in rhizosphere soil networks included Proteobacteria (26.67%), Actinobacteria (17.33%), Bacteroidota (12%), Firmicutes (10.67%), Chloroflexi (9.33%), Acidobacteria (6.67%), and Verrucomicrobiota (5.33%). In contrast, the bulk soil networks were dominated by Proteobacteria (25.68%), Actinobacteria (17.57%), Firmicutes (12.16%), Chloroflexi (9.46%), Bacteroidota (9.46%), Acidobacteria (6.67%), and Myxococcus (4.05%).

Fungal networks in rhizosphere soil showed a larger proportion of positive edges, higher average connectivity, average cluster coefficient, density, and modularity compared to bulk soil networks (Figures 5B,D; Supplementary Table S1). No network hubs were identified in either soil type, but a greater number of module hubs was observed in bulk soil (15) compared to rhizosphere soil (11). The top five keystone species in bulk soil networks included Ascomycota (73.68%), Basidiomycota (22.37%), Chytridiomycota (1.32%), Mucoromycota (1.32%), and Olpidiomycota (1.32%). In rhizosphere networks, the dominant taxa were Ascomycota (75.71%), Basidiomycota (20%), Chytridiomycota (1.43%), Glomeromycota (1.43%), and Olpidiomycota (1.43%). Overall, bacterial networks exhibited greater complexity than fungal networks, as indicated by multiple topological properties (Figures 5E,F; Supplementary Table S2).

Null model analysis revealed distinct stand age-dependent variations in the relative contributions of stochastic and deterministic processes to microbial community assembly (Figure 6). Bacterial communities were predominantly structured by deterministic processes across all stand ages. Notably, stochasticity exerted a stronger influence on bacterial composition in rhizosphere soils than in bulk soils (Figures 6A,C). Conversely, fungal community assembly was primarily driven by stochastic mechanisms. A critical transition was observed in rhizosphere fungal communities, where the dominant assembly process shifted from stochastic to deterministic during forest development (Figures 6B,D).

PICRUSt analysis of bacterial communities revealed 44 pathways, predominantly associated with environmental information processing, genetic information processing, and metabolic pathways (Figures 7A,B). While no significant differences in the relative abundance of bacterial functional categories were observed between rhizosphere and bulk soils, 35 and 28 pathways showed substantial changes (p < 0.05) across stand age for bacterial communities in bulk and rhizosphere soils, respectively.

FUNGuild analysis of fungal communities revealed three trophic modes (pathotroph, symbiotroph, and saprotroph), with rhizosphere soil exhibiting higher relative abundances of pathotrophic functions (Figures 7C,D). Significant changes in specific fungal functional groups were observed across stand age in both soil compartments (p < 0.05). Specifically, in bulk soils, the relative abundance of plant pathogens, fungal parasites, lichen parasites, leaf saprotrophs, and dung saprotrophs significantly changed (p < 0.05) with stand age. In rhizosphere soils, significant changes (p < 0.05) were observed in fungal parasites, arbuscular mycorrhizal fungi, leaf saprotrophs, and dung saprotrophs.

Mantel tests and redundancy analysis (RDA) demonstrated significant associations between soil physicochemical parameters and microbial community composition (Figures 8, 9; Table 1). In bulk soil systems, bacterial Shannon diversity exhibited significant associations with total potassium (TK) and stoichiometric ratios (C/N, C/P) (Figure 8A), whereas fungal Chao1 diversity demonstrated broader nutrient sensitivity, correlating with total carbon (TC), organic carbon (SOC), nitrogen species (TN, AN), phosphorus (TP), calcium (TCa), and pH (Figure 8B). In rhizosphere soil, bacterial alpha diversity was significantly linked to TK and N:P (Figure 8C), while fungal Shannon diversity was significantly correlated with pH (Figure 8D).

Ordination analyses revealed differential explanatory power between soil compartments. For bacterial communities, the first two RDA axes explained 90.86% (bulk soil) and 86.86% (rhizosphere) of compositional variation (Figures 9A,B). Fungal communities demonstrated significantly greater axis explanatory power at 99.84% (bulk) and 99.92% (rhizosphere) (Figures 9C,D). Bacterial communities were predominantly structured by pH and stoichiometric ratios (C/N, C/P, N/P) (Figures 9A,B; Table 1), while fungal communities were more strongly affected by TN, TP, and AN (Figures 9C,D; Table 1). Notably, compartment-specific modulated these relationships: rhizobacterial composition were governed by nutrient availability (AN, AK) and elemental ratios (N/P, C/N, C/P), while bulk soil bacteria responded to pH and total nutrient pools (TC, TN) (Figures 9A,B; Table 1). Fungal biogeography exhibited similar patterns, with rhizosphere communities influenced by AN, TN, and C:P ratios, contrasting with bulk soil variations governed by TN, AN, and SOC (Figures 9C,D; Table 1).

Long-term reforestation significantly altered soil microbial alpha diversity and community composition across the chronosequence. Consistent with expectations, P. armandii plantation exhibited significant changes in soil microbial alpha diversity, aligning with observations in temperate and subtropical forests (Li et al., 2020; Yan et al., 2020). Bacterial and fungal Shannon diversity exhibited a distinct pattern: initially decreasing, then increasing with stand age. This dynamic primarily stems from shifting resource availability, changes in root exudation, and successional niche dynamics driven by plant community development (Wang et al., 2023). In young forests, harsh, variable conditions coupled with highly fluctuating resources favor r-selected microbial strategists (high reproductive rates, rapid growth, broad niches) (Liu W. et al., 2020). As plant biomass, root density, and competition increase during mid-succession, resource limitation intensifies: heightened plant and microbial nutrient uptake depletes pools, and shifts in root exudate composition and litter chemistry selectively favor fewer microbial taxa (Kasel and Bennett, 2007; Malchair and Carnol, 2009). Concurrently, successional niche dynamics occur, where early-successional specialists are replaced but stable, K-selected communities are not yet established, resulting in the observed diversity minimum (Zhang et al., 2016). Ultimately, mature forests provide stable conditions with predictable, diverse resource inputs (e.g., litterfall, complex exudates). This promotes K-selected strategists (high competitiveness, specialization) and allows for resource partitioning among microbes across a broader resource spectrum, leading to higher alpha diversity than mid-succession (Cui et al., 2018; Wang et al., 2023).

Crucially, rhizosphere and bulk soil exhibited divergent microbial diversity patterns. Rhizosphere soil generally had lower alpha diversity than bulk soil, except for fungal Chao1 diversity in young forests, which aligns with previous studies in grasslands and agricultural soils (Marilley and Aragno, 1999; Ai et al., 2012). This is likely attributed to strong host plant selection via root exudates (Liu L. et al., 2020), which create a distinct physicochemical microhabitat favoring specific microbial populations. This rhizosphere filtering, driven by plant-derived signals and organic compounds, shapes a community distinct from the bulk soil reservoir. Microhabitat heterogeneity is thus fundamentally higher in the rhizosphere. However, the persistent similarity between compartments underscores that the rhizosphere microbiome is shaped by both recruitment from the bulk soil reservoir and local plant-microbe feedbacks within the rhizosphere environment (Bram et al., 2017; Veach et al., 2019).

Microbial community composition varied significantly across stand ages (Figures 3, 4), driven primarily by shifts in the abundance of key bacterial and fungal phyla. These compositional shifts reflect functional adaptations associated with stand ages. Changes in bacterial composition were largely attributed to fluctuations in Proteobacteria, Acidobacteria, and Actinobacteria. The high abundance of Proteobacteria in young forests is consistent with their broad ecological niche and nitrogen fixation capabilities (Kim et al., 2021; Rojas et al., 2016), supporting early forest growth. In contrast, the increased abundance of Acidobacteria and Actinobacteria in middle-aged and mature forests aligns with Acidobacteria’s tolerance to nutrient-poor soils and Actinobacteria’s role in lignin decomposition (Kirby, 2005), facilitating conifer forest development under changing conditions. The decline in Proteobacteria abundance in middle-aged and mature forests may be linked to reduced nitrogen availability, favoring non-nitrogen-fixing bacteria (Zheng et al., 2020). Fungal community shifts were primarily driven by changes in the relative abundance of Ascomycota and Basidiomycota. A progressive enrichment of Basidiomycota and reduction in Ascomycota across the chronosequence indicates a transition in decomposition strategy: from preferential utilization of labile carbon compounds by Ascomycota towards the breakdown of recalcitrant complex polymers (e.g., lignin) dominated by Basidiomycota (Lodato et al., 2021). This functional shift has significant ecosystem implications: it enhances the formation of stable soil organic carbon pools through the accumulation of recalcitrant residues and microbial necromass, contributing to long-term carbon sequestration. Concurrently, the slower decomposition of complex compounds may regulate nutrient cycling rates, influencing nitrogen and phosphorus mineralization and availability within the ecosystem.

Principal Coordinates Analysis (PCoA) distinctly separated microbial communities into three groups corresponding to forest age classes (Figure 4), reinforcing the strong taxonomic dependence observed across the P. armandii chronosequence. Despite these significant temporal shifts in community structure, core bacterial and fungal phyla maintained relative stability throughout forest development. Furthermore, although rhizosphere filtering consistently differentiated the composition between rhizosphere and bulk soil compartments (Ning et al., 2022; Luo et al., 2021; Baudoin et al., 2003), we observed no significant temporal differences in the compositional divergence between these compartments across age classes. This finding further underscores the dual importance of bulk soil reservoirs and rhizosphere recruitment processes in shaping the microbial communities.

Soil microbial communities, like plant communities, exhibit complex interaction networks encompassing both mutualistic (e.g., symbiosis) and antagonistic (e.g., competition, predation) relationships (de Menezes et al., 2017). While co-occurrence network analysis does not directly reveal in situ interactions, it remains a valuable tool for elucidating microbial coexistence patterns in environmental samples (Barberán et al., 2012) and exploring links between ecosystem complexity and stability (Zhu et al., 2020). Our analysis revealed distinct co-occurrence network properties in rhizosphere versus bulk soil, indicating that differing niche conditions driven by plant roots sculpt unique bacterial and fungal interaction patterns. This differentiation is primarily attributed to plant root exudates, which alter nutrient availability, suppress pathogens, and recruit specific microbes (Chaparro et al., 2014; Zhao et al., 2021), demonstrating strong host-driven selection and metabolic influences. Furthermore, both forest successional stage and microhabitat significantly modulated rhizosphere network structure, highlighting the synergistic effects of plant development and environmental filtering.

Notably, rhizosphere networks exhibited significantly higher topological complexity (e.g., positive edge proportion, clustering coefficient, average connectivity, and density) than bulk soil networks (Figure 6; Supplementary Table S1). This increased complexity, potentially reflecting greater species interactions (e.g., commensalism, syntrophy, mutualism) and niche overlap (Yuan et al., 2021), may support enhanced community stability and resilience against environmental fluctuations. Three mechanisms likely underpin this complexity: (1) microenvironmental modifications induced by roots (e.g., hydrologic shifts) (Wang et al., 2018); (2) enhanced diversity of potential interactions, including cascading effects (Shi et al., 2016); and (3) exudate-mediated stimulation of microbial exchanges via carbon substrates (e.g., organic acids, sugars, amino acids) (Philippot et al., 2013).

Central to understanding the ecological implications of these networks are keystone taxa, identified based on their topological roles (e.g., connectors, module hubs), which are hypothesized to disproportionately influence network stability and key ecosystem functions such as soil organic matter (SOM) metabolism and nitrogen (N) cycling (Xun et al., 2021). Our analysis identified keystone bacterial taxa primarily from Proteobacteria, Actinobacteria, Ascomycota, and Mortierellomycota, and fungal keystones from Firmicutes, Chloroflexi, and Bacteroidota, serving as critical connectors and module hubs. The ecological dominance of these keystones stems from physiological adaptations (e.g., high-affinity transporters in bacteria) (Ward et al., 2009) and functional versatility crucial for ecosystem processes (e.g., lignocellulose degradation and mycoparasitism in fungi) (Wang et al., 2023). Environmental selection in forests, favoring taxa adept at utilizing complex organic polymers abundant in detritus (e.g., chitin, phospholipids) (Chen et al., 2019), further shapes their prominence. Crucially, the differences in keystone species composition and function between rhizosphere and bulk soil reflect the interplay of vegetation development and soil properties, directly linking these network hubs to spatial variations in nutrient cycling potential. In summary, during plantation development, the versatile metabolisms of these keystone species likely contribute significantly to shaping community structure and mediating interspecies interactions that underpin ecosystem functions (Fan et al., 2018).

However, it is essential to acknowledge the limitations inherent in co-occurrence network inference from sequencing data. These networks represent statistical associations (co-occurrence/co-exclusion patterns), not confirmed biological interactions. Factors like shared environmental preferences or dispersal limitations can generate patterns indistinguishable from direct interactions. Therefore, the inferred interactions (mutualism, competition, etc.) and the direct functional roles of “keystone” taxa based solely on topology remain hypotheses requiring further validation (e.g., through targeted experiments, cultivation, or functional genomics). Our interpretations regarding stability, resilience, and specific functional contributions (e.g., to nutrient cycling rates) are thus cautious extrapolations based on network theory and identified taxa, avoiding claims of direct mechanistic proof.

Deciphering the mechanisms governing microbial community assembly remains a central challenge in microbial ecology (Zhou and Ning, 2017). In our study, deterministic processes were found to drive bacterial community assembly during the development of P. armandii plantations in both rhizosphere and bulk soils, aligning with previous research (Xu et al., 2022; Zhao et al., 2018). This deterministic dominance in bacteria can be attributed to several key drivers: (1) Their rapid generation times (Sun et al., 2017) and higher transmission rates (Schmidt et al., 2014) enable swift responses to environmental gradients such as shifts in soil pH, moisture, or nutrient availability (Rinnan et al., 2007; Ren et al., 2021). (2) Crucially, biotic interactions, particularly with the host plant, exert strong selective pressures. This is especially pronounced in the rhizosphere, where deterministic processes were significantly stronger than in bulk soil. The primary driver of this rhizosphere effect is the concentrated flux of diverse root exudates (e.g., carbon substrates, nitrogenous compounds, flavonoids, salicylic acid, phytoalexins) (Zhu et al., 2016; Vieira et al., 2020). These exudates create a distinct chemical environment gradient that selectively enriches microbial taxa phylogenetically or functionally adapted to utilize these resources, often enhancing nutrient cycling or promoting plant growth (Zhang et al., 2019). In contrast, the attenuated chemical gradient and weaker plant-driven biotic interactions in bulk soil result in comparatively weaker deterministic filtering.

Conversely, fungal community assembly exhibited stronger stochasticity, implying weaker environmental filtering overall and highlighting the importance of other drivers. This aligns with growing recognition of neutral processes in mycobiome organization (Wang et al., 2016). The enhanced stochasticity in fungi likely stems from inherent biological traits imposing constraints: (1) Limited dispersal capacity, associated with larger propagule sizes (Chen W. et al., 2020; Chen W. M. et al., 2020), particularly for specialists like symbionts or biotrophs dependent on spatially constrained hosts or resources (Wardle, 2006; Wang et al., 2022), reduces their ability to track environmental gradients efficiently. (2) Multicellular growth strategies enhance substrate exploitation but inherently limit mobility, further amplifying dispersal limitation and rapid response to environmental gradients across larger spatial scales (Heaton et al., 2020). These observations support theoretical frameworks linking microbial traits to assembly, such as the “size plasticity” (Farjalla et al., 2012) and “size-dispersal tradeoff” hypotheses (Shurin et al., 2009). Consequently, dispersal limitation and drift (random birth/death events) become more critical drivers than fine-scale environmental filtering or intense biotic selection for fungal communities. The specificity of many fungal interactions (e.g., host-pathogen or mycorrhizal symbioses) can also create patchy resource distributions, further amplifying the role of stochastic processes like dispersal limitation in assembly.

This study confirms that the structure of soil microbial communities is intricately regulated by multiple environmental factors, and this regulation varies depending on microbial groups (bacteria vs. fungi) and their specific soil microhabitats (rhizosphere vs. bulk soil). Bacterial community composition was primarily governed by soil pH (Chen W. et al., 2020; Chen W. M. et al., 2020) and key elemental stoichiometric ratios (C/N, C/P, N/P) (Delgado-Baquerizo et al., 2019; Cao et al., 2010; Dini-Andreote et al., 2014), supporting the prevailing view that pH acts as a critical filter for bacterial diversity and that stoichiometric ratios impose fundamental constraints on bacterial metabolism and community structure. Within the rhizosphere microhabitat shaped by root activity, bacterial composition was mainly driven by available nutrients (AN, AK) and elemental stoichiometric ratios (N/P, C/N, C/P), reflecting the selective effects of root exudates in enhancing nutrient availability and altering substrate stoichiometry. In contrast, bacterial communities in the bulk soil responded more strongly to factors reflecting fundamental soil conditions and total nutrient pools, such as pH, TC, and TN.

Conversely, fungal community composition exhibited a stronger response to total nutrient pools (TN, TP) and AN, likely related to their diverse roles in nutrient cycling (e.g., saprotrophic, symbiotic fungi) and acquisition strategies. The specific driving factors also differed by microhabitat. Rhizosphere fungal communities were primarily associated with AN, TN, and the C:P ratio, underscoring the importance of nitrogen (both total and available forms) and carbon-phosphorus balance in the root-influenced zone (Li D. D. et al., 2018; Li J. et al., 2018; Wang et al., 2019). In contrast, variation in bulk soil fungal communities was mainly driven by TN, AN, and SOC in combination, indicating that their structure is simultaneously influenced by nitrogen status and organic carbon availability (a key energy source for saprotrophic fungi) (Li D. D. et al., 2018; Li J. et al., 2018; Wang et al., 2019).

Collectively, these findings deepen our understanding of the mechanisms shaping soil microbial biogeographical patterns. They highlight the necessity of considering spatial heterogeneity (particularly the rhizosphere effect) and the specificity of microbial functional groups when studying the relationships between microbial communities and their environment.